使用boost asio的线程池

Question

我正在尝试使用boost :: asio创建一个有限的线程池类.但是我被困在某一点可以帮助我.

唯一的问题是我应该减少反击的地方？

代码无法按预期工作.

问题是我不知道我的线程何时完成执行以及我将如何知道它已经返回池中

#include <boost/asio.hpp>
#include <iostream>
#include <boost/thread/thread.hpp>
#include <boost/bind.hpp>
#include <boost/thread/mutex.hpp>
#include <stack>

using namespace std;
using namespace boost;

class ThreadPool
{
    static int count;
    int NoOfThread;
    thread_group grp;
    mutex mutex_;
    asio::io_service io_service;
    int counter;
    stack<thread*> thStk ;

public:
    ThreadPool(int num)
    {   
        NoOfThread = num;
        counter = 0;
        mutex::scoped_lock lock(mutex_);

        if(count == 0)
            count++;
        else
            return;

        for(int i=0 ; i<num ; ++i)
        {
            thStk.push(grp.create_thread(boost::bind(&asio::io_service::run, &io_service)));
        }
    }
    ~ThreadPool()
    {
        io_service.stop();
        grp.join_all();
    }

    thread* getThread()
    {
        if(counter > NoOfThread)
        {
            cout<<"run out of threads \n";
            return NULL;
        }

        counter++;
        thread* ptr = thStk.top();
        thStk.pop();
        return ptr;
    }
};
int ThreadPool::count = 0;


struct callable
{
    void operator()()
    {
        cout<<"some task for thread \n";
    }
};

int main( int argc, char * argv[] )
{

    callable x;
    ThreadPool pool(10);
    thread* p = pool.getThread();
    cout<<p->get_id();

    //how i can assign some function to thread pointer ?
    //how i can return thread pointer after work done so i can add 
//it back to stack?


    return 0;
}

Answer 1

简而言之,您需要使用另一个函数来包装用户提供的任务:

• 调用用户函数或可调用对象.
• 锁定互斥锁并减少计数器.

我可能无法理解此线程池的所有要求.因此,为清楚起见,这里有一个明确的列表,我认为是什么要求:

• 池管理线程的生命周期.用户不应该删除驻留在池中的线​​程.
• 用户可以以非侵入方式将任务分配给池.
• 分配任务时,如果池中的所有线程当前正在运行其他任务,则该任务将被丢弃.

在我提供实现之前,我想强调一些关键点:

• 一旦启动了一个线程,它将一直运行直到完成,取消或终止.线程正在执行的函数无法重新分配.为了允许单个线程在其生命周期内执行多个函数,线程将希望使用将从队列读取的函数启动,例如io_service::run(),并且可调用类型被发布到事件队列中,例如来自io_service::post().
• io_service::run()返回,如果没有挂起的工作io_service,io_service则停止,或者从线程正在运行的处理程序抛出异常.为了防止io_serivce::run()在没有未完成的工作时返回,io_service::work可以使用该类.
• 定义任务的类型要求(即任务的类型必须可以通过object()语法调用)而不是要求类型(即任务必须继承process),为用户提供更大的灵活性.它允许用户提供任务作为函数指针或提供nullary的类型operator().

实施使用boost::asio:

#include <boost/asio.hpp>
#include <boost/thread.hpp>

class thread_pool
{
private:
  boost::asio::io_service io_service_;
  boost::asio::io_service::work work_;
  boost::thread_group threads_;
  std::size_t available_;
  boost::mutex mutex_;
public:

  /// @brief Constructor.
  thread_pool( std::size_t pool_size )
    : work_( io_service_ ),
      available_( pool_size )
  {
    for ( std::size_t i = 0; i < pool_size; ++i )
    {
      threads_.create_thread( boost::bind( &boost::asio::io_service::run,
                                           &io_service_ ) );
    }
  }

  /// @brief Destructor.
  ~thread_pool()
  {
    // Force all threads to return from io_service::run().
    io_service_.stop();

    // Suppress all exceptions.
    try
    {
      threads_.join_all();
    }
    catch ( const std::exception& ) {}
  }

  /// @brief Adds a task to the thread pool if a thread is currently available.
  template < typename Task >
  void run_task( Task task )
  {
    boost::unique_lock< boost::mutex > lock( mutex_ );

    // If no threads are available, then return.
    if ( 0 == available_ ) return;

    // Decrement count, indicating thread is no longer available.
    --available_;

    // Post a wrapped task into the queue.
    io_service_.post( boost::bind( &thread_pool::wrap_task, this,
                                   boost::function< void() >( task ) ) );
  }

private:
  /// @brief Wrap a task so that the available count can be increased once
  ///        the user provided task has completed.
  void wrap_task( boost::function< void() > task )
  {
    // Run the user supplied task.
    try
    {
      task();
    }
    // Suppress all exceptions.
    catch ( const std::exception& ) {}

    // Task has finished, so increment count of available threads.
    boost::unique_lock< boost::mutex > lock( mutex_ );
    ++available_;
  }
};

关于实施的一些评论:

• 需要围绕用户的任务进行异常处理.如果用户的函数或可调用对象抛出一个异常的类型是不boost::thread_interrupted,那么std::terminate()被调用.这是Boost.Thread 在线程函数行为中的异常的结果.同样值得阅读Boost.Asio对处理程序抛出异常的影响.
• 如果用户提供taskvia boost::bind,则嵌套boost::bind将无法编译.需要以下选项之一:
  • 不支持task创建boost::bind.
  • 元编程进行编译时分支基于是否将用户的类型,如果结果boost::bind,这样boost::protect可以使用,boost::protect适当对某些函数对象只有函数.
  • 使用其他类型task间接传递对象.我选择以boost::function丢失确切类型为代价来提高可读性. boost::tuple虽然可读性稍差,但也可用于保留确切的类型,如Boost.Asio的序列化示例中所示.

应用程序代码现在可以使用thread_pool非侵入式:

void work() {};

struct worker
{
  void operator()() {};
};

void more_work( int ) {};

int main()
{ 
  thread_pool pool( 2 );
  pool.run_task( work );                        // Function pointer.
  pool.run_task( worker() );                    // Callable object.
  pool.run_task( boost::bind( more_work, 5 ) ); // Callable object.
}

thread_pool可以在没有Boost.Asio 的情况下创建,对于维护者来说可能稍微容易一些,因为他们不再需要了解Boost.Asio行为,例如何时io_service::run()返回,以及io_service::work对象是什么:

#include <queue>
#include <boost/bind.hpp>
#include <boost/thread.hpp>

class thread_pool
{
private:
  std::queue< boost::function< void() > > tasks_;
  boost::thread_group threads_;
  std::size_t available_;
  boost::mutex mutex_;
  boost::condition_variable condition_;
  bool running_;
public:

  /// @brief Constructor.
  thread_pool( std::size_t pool_size )
    : available_( pool_size ),
      running_( true )
  {
    for ( std::size_t i = 0; i < pool_size; ++i )
    {
      threads_.create_thread( boost::bind( &thread_pool::pool_main, this ) ) ;
    }
  }

  /// @brief Destructor.
  ~thread_pool()
  {
    // Set running flag to false then notify all threads.
    {
      boost::unique_lock< boost::mutex > lock( mutex_ );
      running_ = false;
      condition_.notify_all();
    }

    try
    {
      threads_.join_all();
    }
    // Suppress all exceptions.
    catch ( const std::exception& ) {}
  }

  /// @brief Add task to the thread pool if a thread is currently available.
  template < typename Task >
  void run_task( Task task )
  {
    boost::unique_lock< boost::mutex > lock( mutex_ );

    // If no threads are available, then return.
    if ( 0 == available_ ) return;

    // Decrement count, indicating thread is no longer available.
    --available_;

    // Set task and signal condition variable so that a worker thread will
    // wake up andl use the task.
    tasks_.push( boost::function< void() >( task ) );
    condition_.notify_one();
  }

private:
  /// @brief Entry point for pool threads.
  void pool_main()
  {
    while( running_ )
    {
      // Wait on condition variable while the task is empty and the pool is
      // still running.
      boost::unique_lock< boost::mutex > lock( mutex_ );
      while ( tasks_.empty() && running_ )
      {
        condition_.wait( lock );
      }
      // If pool is no longer running, break out.
      if ( !running_ ) break;

      // Copy task locally and remove from the queue.  This is done within
      // its own scope so that the task object is destructed immediately
      // after running the task.  This is useful in the event that the
      // function contains shared_ptr arguments bound via bind.
      {
        boost::function< void() > task = tasks_.front();
        tasks_.pop();

        lock.unlock();

        // Run the task.
        try
        {
          task();
        }
        // Suppress all exceptions.
        catch ( const std::exception& ) {}
      }

      // Task has finished, so increment count of available threads.
      lock.lock();
      ++available_;
    } // while running_
  }
};