制片人/消费者实施 - 反馈意向

时间:2010-05-24 21:11:05

标签: c++ boost multithreading producer-consumer

我准备在几个星期内接受采访,并且我希望能够提供一些线索,以及我在学校学到的简单的生产者/消费者问题。

没有做过一段时间,所以我很好奇你们对此有何看法?我应该添加什么来使它成为一个更好的例子等。感谢您的反馈! :)

//////////////////////////////////////////////////////////////////////////
boost::mutex bufferMutex;
deque<int> buffer;
const int maxBufferSize = 5;
//////////////////////////////////////////////////////////////////////////

bool AddToBuffer(int i)
{
    if (buffer.size() < maxBufferSize)
    {
        buffer.push_back(i);
        return true;
    }
    else
    {       
        return false;
    }
}

bool GetFromBuffer(int& toReturn)
{
    if (buffer.size() == 0)
    {
        return false;
    }
    else
    {
        toReturn = buffer[buffer.size()-1];
        buffer.pop_back();
        return true;
    }
}

struct Producer 
{
    int ID;
    void operator()()
    {
        while (true)
        {
            boost::mutex::scoped_lock lock(bufferMutex);
            int num = dice();
            bool result = AddToBuffer(num);
            lock.unlock();
            //safe area done
            if (result)
            {
                cout << "Producer " << this->ID << " Added " << num << endl;
            }
            else
            {
                cout << "!!Buffer was Full!!" << endl;
            }
            //Added
            //Now wait
            boost::xtime xt;
            xtime_get( &xt, boost::TIME_UTC);
            xt.nsec += 1000000 + 100000 * (rand() % 1000);
            boost::thread::sleep(xt);
        }
    }
};

struct Consumer 
{
    int ID;
    void operator()()
    {
        while (true)
        {
            int returnedInt = 0;
            boost::mutex::scoped_lock lock(bufferMutex);
            bool result = GetFromBuffer(returnedInt);
            lock.unlock();
            //safe area done
            if (result)
            {
                cout << "\tConsumer " << this->ID << " Took Out " << returnedInt << endl;
            }
            else
            {
                cout << "!!Buffer was Empty!!" << endl;
            }
            //Added
            //Now wait
            boost::xtime xt;
            xtime_get( &xt, boost::TIME_UTC);
            xt.nsec += 1000000 + 100000 * (rand() % 1000);
            boost::thread::sleep(xt);
        }
    }
};



void main()
{
    Producer p, p2;
    Consumer c, c2;

    p.ID = 1;
    p2.ID = 2;

    c.ID = 1;
    c2.ID = 2;

    boost::thread thread1(boost::ref(p));
    boost::thread thread2(boost::ref(c));
    boost::thread thread3(boost::ref(p2));
    boost::thread thread4(boost::ref(c2));

    int x;
    cin >> x;
}

1 个答案:

答案 0 :(得分:3)

如果您已经在AddToBuffer和GetFromBuffer之类的调用中包装缓冲区对象,那么将锁定置于包装函数中会更有意义。另外,你正在明确调用unlock,这完全违背了scoped_lock的目的; scoped_lock使用资源获取初始化(RAII)来获取和释放锁。更好的用法是将临界区放在一个块中,以便由于锁超出范围而释放互斥锁,而不是由于显式调用解锁函数,因为范围不太脆​​弱。例如:

// Code that doesn't need locking
{
    boost::mutex::scoped_lock lck(bufferMutex); // Lock is acquired here
    // Code that needs to be synchronized
} // Lock is automatically released here without explicit call to unlock()
// More code that doesn't need locking