一、线程池
大多数的网络服务器，包括Web服务器都具有一个特点，就是单位时间内必须处理数目巨大的连接请求，但是处理时间却是比较短的。在传统的多线程服务器模型中是这样实现的：一旦有个请求到达，就创建一个新的线程，由该线程执行任务，任务执行完毕之后，线程就退出。这就是"即时创建，即时销毁"的策略。尽管与创建进程相比，创建线程的时间已经大大的缩短，但是如果提交给线程的任务是执行时间较短，而且执行次数非常频繁，那么服务器就将处于一个不停的创建线程和销毁线程的状态。这笔开销是不可忽略的，尤其是线程执行的时间非常非常短的情况。
线程池就是为了解决上述问题的，它的实现原理是这样的：在应用程序启动之后，就马上创建一定数量的线程，放入空闲的队列中。这些线程都是处于阻塞状态，这些线程只占一点内存，不占用CPU。当任务到来后，线程池将选择一个空闲的线程，将任务传入此线程中运行。当所有的线程都处在处理任务的时候，线程池将自动创建一定的数量的新线程，用于处理更多的任务。执行任务完成之后线程并不退出，而是继续在线程池中等待下一次任务。当大部分线程处于阻塞状态时，线程池将自动销毁一部分的线程，回收系统资源。
下面是一个简单线程池的实现，这个线程池的代码是我参考网上的一个例子实现的，由于找不到出处了，就没办法注明参考自哪里了。它的方案是这样的：程序启动之前，初始化线程池，启动线程池中的线程，由于还没有任务到来，线程池中的所有线程都处在阻塞状态，当一有任务到达就从线程池中取出一个空闲线程处理，如果所有的线程都处于工作状态，就添加到队列，进行排队。如果队列中的任务个数大于队列的所能容纳的最大数量，那就不能添加任务到队列中，只能等待队列不满才能添加任务到队列中。
主要由两个文件组成一个threadpool.h头文件和一个threadpool.c源文件组成。源码中已有重要的注释，就不加以分析了。
threadpool.h文件：
struct job
{
void* (*callback_function)(void *arg); //线程回调函数
void *arg; //回调函数参数
struct job *next;
};
struct threadpool
{
int thread_num; //线程池中开启线程的个数
int queue_max_num; //队列中最大job的个数
struct job *head; //指向job的头指针
struct job *tail; //指向job的尾指针
pthread_t *pthreads; //线程池中所有线程的pthread_t
pthread_mutex_t mutex; //互斥信号量
pthread_cond_t queue_empty; //队列为空的条件变量
pthread_cond_t queue_not_empty; //队列不为空的条件变量
pthread_cond_t queue_not_full; //队列不为满的条件变量
int queue_cur_num; //队列当前的job个数
int queue_close; //队列是否已经关闭
int pool_close; //线程池是否已经关闭
};
//================================================================================================
//函数名： threadpool_init
//函数描述： 初始化线程池
//输入： [in] thread_num 线程池开启的线程个数
// [in] queue_max_num 队列的最大job个数
//输出： 无
//返回： 成功：线程池地址 失败：NULL
//================================================================================================
struct threadpool* threadpool_init(int thread_num, int queue_max_num);
//================================================================================================
//函数名： threadpool_add_job
//函数描述： 向线程池中添加任务
//输入： [in] pool 线程池地址
// [in] callback_function 回调函数
// [in] arg 回调函数参数
//输出： 无
//返回： 成功：0 失败：-1
//================================================================================================
int threadpool_add_job(struct threadpool *pool, void* (*callback_function)(void *arg), void *arg);
//================================================================================================
//函数名： threadpool_destroy
//函数描述： 销毁线程池
//输入： [in] pool 线程池地址
//输出： 无
//返回： 成功：0 失败：-1
//================================================================================================
int threadpool_destroy(struct threadpool *pool);
//================================================================================================
//函数名： threadpool_function
//函数描述： 线程池中线程函数
//输入： [in] arg 线程池地址
//输出： 无
//返回： 无
//================================================================================================
void* threadpool_function(void* arg);
#include "threadpool.h"
struct threadpool* threadpool_init(int thread_num, int queue_max_num)
{
struct threadpool *pool = NULL;
do
{
pool = malloc(sizeof(struct threadpool));
if (NULL == pool)
{
printf("failed to malloc threadpool!\n");
break;
}
pool->thread_num = thread_num;
pool->queue_max_num = queue_max_num;
pool->queue_cur_num = 0;
pool->head = NULL;
pool->tail = NULL;
if (pthread_mutex_init(&(pool->mutex), NULL))
{
printf("failed to init mutex!\n");
break;
}
if (pthread_cond_init(&(pool->queue_empty), NULL))
{
printf("failed to init queue_empty!\n");
break;
}
if (pthread_cond_init(&(pool->queue_not_empty), NULL))
{
printf("failed to init queue_not_empty!\n");
break;
}
if (pthread_cond_init(&(pool->queue_not_full), NULL))
{
printf("failed to init queue_not_full!\n");
break;
}
pool->pthreads = malloc(sizeof(pthread_t) * thread_num);
if (NULL == pool->pthreads)
{
printf("failed to malloc pthreads!\n");
break;
}
pool->queue_close = 0;
pool->pool_close = 0;
int i;
for (i = 0; i < pool->thread_num; ++i)
{
pthread_create(&(pool->pthreads[i]), NULL, threadpool_function, (void *)pool);
}
return pool;
} while (0);
return NULL;
}
int threadpool_add_job(struct threadpool* pool, void* (*callback_function)(void *arg), void *arg)
{
assert(pool != NULL);
assert(callback_function != NULL);
assert(arg != NULL);
pthread_mutex_lock(&(pool->mutex));
while ((pool->queue_cur_num == pool->queue_max_num) && !(pool->queue_close || pool->pool_close))
{
pthread_cond_wait(&(pool->queue_not_full), &(pool->mutex)); //队列满的时候就等待
}
if (pool->queue_close || pool->pool_close) //队列关闭或者线程池关闭就退出
{
pthread_mutex_unlock(&(pool->mutex));
return -1;
}
struct job *pjob =(struct job*) malloc(sizeof(struct job));
if (NULL == pjob)
{
pthread_mutex_unlock(&(pool->mutex));
return -1;
}
pjob->callback_function = callback_function;
pjob->arg = arg;
pjob->next = NULL;
if (pool->head == NULL)
{
pool->head = pool->tail = pjob;
pthread_cond_broadcast(&(pool->queue_not_empty)); //队列空的时候，有任务来时就通知线程池中的线程：队列非空
}
else
{
pool->tail->next = pjob;
pool->tail = pjob;
}
pool->queue_cur_num++;
pthread_mutex_unlock(&(pool->mutex));
return 0;
}
void* threadpool_function(void* arg)
{
struct threadpool *pool = (struct threadpool*)arg;
struct job *pjob = NULL;
while (1) //死循环
{
pthread_mutex_lock(&(pool->mutex));
while ((pool->queue_cur_num == 0) && !pool->pool_close) //队列为空时，就等待队列非空
{
pthread_cond_wait(&(pool->queue_not_empty), &(pool->mutex));
}
if (pool->pool_close) //线程池关闭，线程就退出
{
pthread_mutex_unlock(&(pool->mutex));
pthread_exit(NULL);
}
pool->queue_cur_num--;
pjob = pool->head;
if (pool->queue_cur_num == 0)
{
pool->head = pool->tail = NULL;
}
else
{
pool->head = pjob->next;
}
if (pool->queue_cur_num == 0)
{
pthread_cond_signal(&(pool->queue_empty)); //队列为空，就可以通知threadpool_destroy函数，销毁线程函数
}
if (pool->queue_cur_num == pool->queue_max_num - 1)
{
pthread_cond_broadcast(&(pool->queue_not_full)); //队列非满，就可以通知threadpool_add_job函数，添加新任务
}
pthread_mutex_unlock(&(pool->mutex));
(*(pjob->callback_function))(pjob->arg); //线程真正要做的工作，回调函数的调用
free(pjob);
pjob = NULL;
}
}
int threadpool_destroy(struct threadpool *pool)
{
assert(pool != NULL);
pthread_mutex_lock(&(pool->mutex));
if (pool->queue_close || pool->pool_close) //线程池已经退出了，就直接返回
{
pthread_mutex_unlock(&(pool->mutex));
return -1;
}
pool->queue_close = 1; //置队列关闭标志
while (pool->queue_cur_num != 0)
{
pthread_cond_wait(&(pool->queue_empty), &(pool->mutex)); //等待队列为空
}
pool->pool_close = 1; //置线程池关闭标志
pthread_mutex_unlock(&(pool->mutex));
pthread_cond_broadcast(&(pool->queue_not_empty)); //唤醒线程池中正在阻塞的线程
pthread_cond_broadcast(&(pool->queue_not_full)); //唤醒添加任务的threadpool_add_job函数
int i;
for (i = 0; i < pool->thread_num; ++i)
{
pthread_join(pool->pthreads[i], NULL); //等待线程池的所有线程执行完毕
}
pthread_mutex_destroy(&(pool->mutex)); //清理资源
pthread_cond_destroy(&(pool->queue_empty));
pthread_cond_destroy(&(pool->queue_not_empty));
pthread_cond_destroy(&(pool->queue_not_full));
free(pool->pthreads);
struct job *p;
while (pool->head != NULL)
{
p = pool->head;
pool->head = p->next;
free(p);
}
free(pool);
return 0;
}
#include "threadpool.h"
void* work(void* arg)
{
char *p = (char*) arg;
printf("threadpool callback fuction : %s.\n", p);
sleep(1);
}
int main(void)
{
struct threadpool *pool = threadpool_init(10, 20);
threadpool_add_job(pool, work, "1");
threadpool_add_job(pool, work, "2");
threadpool_add_job(pool, work, "3");
threadpool_add_job(pool, work, "4");
threadpool_add_job(pool, work, "5");
threadpool_add_job(pool, work, "6");
threadpool_add_job(pool, work, "7");
threadpool_add_job(pool, work, "8");
threadpool_add_job(pool, work, "9");
threadpool_add_job(pool, work, "10");
threadpool_add_job(pool, work, "11");
threadpool_add_job(pool, work, "12");
threadpool_add_job(pool, work, "13");
threadpool_add_job(pool, work, "14");
threadpool_add_job(pool, work, "15");
threadpool_add_job(pool, work, "16");
threadpool_add_job(pool, work, "17");
threadpool_add_job(pool, work, "18");
threadpool_add_job(pool, work, "19");
threadpool_add_job(pool, work, "20");
threadpool_add_job(pool, work, "21");
threadpool_add_job(pool, work, "22");
threadpool_add_job(pool, work, "23");
threadpool_add_job(pool, work, "24");
threadpool_add_job(pool, work, "25");
threadpool_add_job(pool, work, "26");
threadpool_add_job(pool, work, "27");
threadpool_add_job(pool, work, "28");
threadpool_add_job(pool, work, "29");
threadpool_add_job(pool, work, "30");
threadpool_add_job(pool, work, "31");
threadpool_add_job(pool, work, "32");
threadpool_add_job(pool, work, "33");
threadpool_add_job(pool, work, "34");
threadpool_add_job(pool, work, "35");
threadpool_add_job(pool, work, "36");
threadpool_add_job(pool, work, "37");
threadpool_add_job(pool, work, "38");
threadpool_add_job(pool, work, "39");
threadpool_add_job(pool, work, "40");
sleep(5);
threadpool_destroy(pool);
return 0;
}上面的文章介绍了线程池的原理及意义。
下面，介绍的这个线程池与上面提到的那个线程池有一部分相似的地方。
主要区别为：
1、线程池中的每个线程都有自己的互斥量和条件变量，而不是线程池共享一个。
2、线程池中的线程在程序结束时，等待线程池中线程停止的机制不同。
该程序主要由两个文件构成，分别为ThreadPool.h和ThreadPool.cpp文件。
ThreadPool.h文件：
#define MAXT_IN_POOL 200
#define BUSY_THRESHOlD 0.5
#define MANAGE_INTREVAL 2
class ThreadPool;
typedef void (*dispatch_fn)(void*);
//线程函数参数
typedef struct tagThread
{
pthread_t thread_id; //线程ID
pthread_mutex_t thread_mutex; //信号量
pthread_cond_t thread_cond; //条件变量
dispatch_fn do_job; //调用的函数，任务
void* args; //函数参数
ThreadPool *parent; //线程池指针
}_thread;
//线程池
class ThreadPool
{
public:
//================================================================================================
//函数名： ThreadPool
//函数描述： 构造函数
//输入： [in] max_threads_in_pool 线程池最大线程数
//输入： [in] min_threads_in_pool 线程池最小问题数
//输出： 无
//返回： 无
//================================================================================================
ThreadPool(unsigned int max_threads_in_pool, unsigned int min_threads_in_pool = 2);
~ThreadPool();
//================================================================================================
//函数名： dispatch_threadpool
//函数描述： 将任务加入线程池，由线程池进行分发
//输入： [in] dispatch_me 调用的函数地址
//输入： [in] dispatch_me 函数参数
//输出： 无
//返回： 无
//================================================================================================
void dispatch_threadpool(dispatch_fn dispatch_me, void* dispatch_me);
private:
pthread_mutex_t tp_mutex; //信号量
pthread_cond_t tp_idle; //线程池中线程有空闲线程的条件变量
pthread_cond_t tp_full; //线程池中线程为满的条件变量
pthread_cond_t tp_empty; //线程池中线程为空的条件变量
int tp_min; //线程池的最小线程数
int tp_max; //线程池的最大线程数
int tp_avail; //线程池中空闲的线程数
int tp_total; //线程池中已创建的线程数
_thread** tp_list; //指向线程池中所有空闲线程的参数的指针
bool tp_stop; //线程池是否已停止
//================================================================================================
//函数名： add_avail
//函数描述： 加入空闲线程
//输入： [in] avail 线程的参数
//输出： 无
//返回： 成功：true，失败：false
//================================================================================================
bool add_avail(_thread* avail);
//================================================================================================
//函数名： work_thread
//函数描述： 线程函数
//输入： [in] args 参数
//输出： 无
//返回： 无
//================================================================================================
static void* work_thread(void* args);
//================================================================================================
//函数名： add_thread
//函数描述： 添加一个线程
//输入： [in] dispatch_me 函数指针
//输入： [in] args 函数参数
//输出： 无
//返回： 无
//================================================================================================
bool add_thread(dispatch_fn dispatch_me, void* args);
//================================================================================================
//函数名： syn_all
//函数描述： 等待线程池中所有线程空闲
//输入： 无
//输出： 无
//返回： 无
//================================================================================================
void syn_all();
};
ThreadPool::ThreadPool(unsigned int max_threads_in_pool, unsigned int min_threads_in_pool)
{
pthread_t manage_id;
if (min_threads_in_pool <= 0 || max_threads_in_pool < 0 || min_threads_in_pool > max_threads_in_pool || max_threads_in_pool > MAXT_IN_POOL)
{
return ;
}
tp_avail = 0; //初始化线程池
tp_total = 0;
tp_min = min_threads_in_pool;
tp_max = max_threads_in_pool;
tp_stop = false;
tp_list = (_thread * *) malloc(sizeof(void *) * max_threads_in_pool);
if (NULL == tp_list)
{
return;
}
memset(tp_list, 0, sizeof(void *) * max_threads_in_pool);
pthread_mutex_init(&tp_mutex, NULL);
pthread_cond_init(&tp_idle, NULL);
pthread_cond_init(&tp_full, NULL);
pthread_cond_init(&tp_empty, NULL);
}
bool ThreadPool::add_avail(_thread* avail)
{
bool ret = false;
pthread_mutex_lock(&tp_mutex);
if (tp_avail < tp_max)
{
tp_list[tp_avail] = avail;
tp_avail++;
pthread_cond_signal(&tp_idle); //线程池中有线程为空闲
if (tp_avail >= tp_total)
{
pthread_cond_signal(&tp_full); //线程池中所有线程都为为空闲
}
ret = true;
}
pthread_mutex_unlock(&tp_mutex);
return ret;
}
void* ThreadPool::work_thread(void* args)
{
_thread* thread = (_thread*) args;
ThreadPool *pool = thread->parent;
while (pool->tp_stop == false)
{
thread->do_job(thread->args);
pthread_mutex_lock(&thread->thread_mutex); //执行完任务之后，添加到空闲线程队列中
if (pool->add_avail(thread))
{
pthread_cond_wait(&thread->thread_cond, &thread->thread_mutex);
pthread_mutex_unlock(&thread->thread_mutex);
}
else
{
pthread_mutex_unlock(&thread->thread_mutex);
pthread_mutex_destroy(&thread->thread_mutex);
pthread_cond_destroy(&thread->thread_cond);
free(thread);
break;
}
}
pthread_mutex_lock(&pool->tp_mutex);
pool->tp_total--;
if (pool->tp_total <= 0)
{
pthread_cond_signal(&pool->tp_empty);
}
pthread_mutex_unlock(&pool->tp_mutex);
return NULL;
}
bool ThreadPool::add_thread(dispatch_fn dispatch_me, void* args) //添加一个线程
{
_thread* thread = NULL;
pthread_attr_t attr;
thread = (_thread *) malloc(sizeof(_thread));
if (NULL == thread)
{
return false;
}
pthread_mutex_init(&thread->thread_mutex, NULL);
pthread_cond_init(&thread->thread_cond, NULL);
thread->do_job = dispatch_me;
thread->args = args;
thread->parent = this;
pthread_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
if (pthread_create(&thread->thread_id, &attr, work_thread, (void *) thread) != 0)
{
pthread_mutex_destroy(&thread->thread_mutex);
pthread_cond_destroy(&thread->thread_cond);
pthread_attr_destroy(&attr);
free(thread);
return false;
}
tp_total++;
return true;
}
void ThreadPool::dispatch_threadpool(dispatch_fn dispatch_me, void* args)
{
_thread* thread = NULL;
pthread_mutex_lock(&tp_mutex);
if (tp_avail <= 0 && tp_total >= tp_max) //无可用线程，而且线程数已达最大值，等待空闲线程
{
pthread_cond_wait(&tp_idle, &tp_mutex);
}
if (tp_avail <= 0) //无可用线程，而且线程数未达最大值，添加线程
{
if (!add_thread(dispatch_me, args))
{
return;
}
}
else //有可用线程
{
tp_avail--;
thread = tp_list[tp_avail];
tp_list[tp_avail] = NULL;
thread->do_job = dispatch_me;
thread->args = args;
pthread_mutex_lock(&thread->thread_mutex);
pthread_cond_signal(&thread->thread_cond);
pthread_mutex_unlock(&thread->thread_mutex);
}
pthread_mutex_unlock(&tp_mutex);
}
void ThreadPool::syn_all()
{
if (tp_avail < tp_total) //等待线程池中所有线程都为空闲状态
{
pthread_cond_wait(&tp_full, &tp_mutex);
}
tp_stop = true;
int i = 0;
for (i = 0; i < tp_avail; i++) //唤醒线程池中所有线程
{
_thread *thread = tp_list[i];
pthread_mutex_lock(&thread->thread_mutex);
pthread_cond_signal(&thread->thread_cond);
pthread_mutex_unlock(&thread->thread_mutex);
}
if (tp_total > 0)
{
pthread_cond_wait(&tp_empty, &tp_mutex); //等待线程池中所有线程都结束
}
}
ThreadPool::~ThreadPool()
{
sleep(MANAGE_INTREVAL);
pthread_mutex_lock(&tp_mutex);
syn_all(); //等待线程池为空
int i = 0;
for (i = 0; i < tp_total; i++) //资源释放
{
free(tp_list[i]);
tp_list[i] = NULL;
}
pthread_mutex_unlock(&tp_mutex);
pthread_mutex_destroy(&tp_mutex);
pthread_cond_destroy(&tp_idle);
pthread_cond_destroy(&tp_full);
pthread_cond_destroy(&tp_empty);
free(tp_list);
}
737
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
