这个C++多线程程序有什么问题？

dkqlctbz 于 2024-01-09 发布在其他

关注(0)|答案(1)|浏览(156)

我使用std::thread编写了一个多线程C++程序，如下所示：

#include <iostream>
#include <thread>
#include <mutex>
#include <condition_variable>
using namespace std;
const int threadNum = 4;
mutex mt[threadNum];
condition_variable cv[threadNum];
thread threadList[threadNum];
bool threadWork[threadNum];
void work(int id) {
    while (true) {
        unique_lock<mutex> lck(mt[id]);
        cv[id].wait(lck, [&]() { return threadWork[id]; }); // wait for incoming tasks
        
        // do something
        
        threadWork[id] = false;
        cv[id].notify_all();
    }
}
int main() {
    for (int i = 0; i < threadNum; i ++) {
        threadWork[i] = false;
        threadList[i] = thread(work, i);
    }
    while (true) {
        for (int i = 0; i < threadNum; i ++) {
            // allocate tasks for each threads
            threadWork[i] = true;
            cv[i].notify_all();
        }
        for (int i = 0; i < threadNum; i ++) {
            // wait until all tasks finish
            unique_lock<mutex> lck(mt[i]);
            cv[i].wait(lck, [&]() { return !threadWork[i]; });
            cout << "Finish thread " << i << endl;
        }
        // do something
    }
    for (int i = 0; i < threadNum; i ++)
        threadList[i].join();
    return 0;
}

字符串
在程序的迭代过程中，主线程和子线程交替执行，主线程首先为多个子线程分配任务，所有线程完成任务后，主线程将汇总信息并进行其他计算。
但是，在执行过程中，程序会随机地在几次迭代后崩溃或陷入死锁。我不知道为什么会发生这种情况。这是什么问题？
我的环境：ubuntu 22.04，g++ 11.4

c++

来源：https://stackoverflow.com/questions/77732314/what-is-wrong-with-this-c-multi-threaded-program

1条答案

按热度按时间

kyxcudwk1#

下面代码中的注解：

//...
mutex mt[threadNum];           // Mutexes are for synchronizing 
                               // between threads.  Having 1 mutex
                               // per thread is not proper the way
                               // to guard access to data.
                               // try having 1 mutex per block of 
                               // data that you want guarded instead.
condition_variable cv[threadNum];  // same applies for the condition variables
                                   // you want to indicate that the DATA 
                                   // is ready.
// ... 
// The rest of the code will need a bit of reordering.

字符串
对于你的实验，你应该尝试一些更简单的方法，比如说，从一个受保护的数据块开始。下面是一个例子。
请注意，在下面的代码中，工作线程等待一个唯一的条件变量，告诉它们需要做一些事情。
我还添加了一个简单的机制，一旦工作完成，就可以优雅地从程序中退出。这和其他部分一样重要，注意原子<>的使用，它保证所有核心都会在值发生变化时看到。

#include <condition_variable>
#include <iostream>
#include <mutex>
#include <thread>
#include <vector>
#include <atomic>
std::mutex my_data_mutex;
std::condition_variable my_data_cv;
std::vector<int> my_data;
constexpr size_t threadNum = 4;
std::atomic<bool> halt = false;
void work(int id) {
    for(;;)
    {
        std::unique_lock<std::mutex> _(my_data_mutex);
        my_data_cv.wait(_);  // wait for incoming tasks
        if (halt)
            return;
        // do something
        my_data.push_back(my_data.back() + id);
    }
}
int main() {
    std::vector<std::thread> tasks;
    for (int i = 0; i < threadNum; i++) {
        tasks.emplace_back(work, i);
    }
    while(!halt)
    {
        std::unique_lock<std::mutex> _(my_data_mutex);
        my_data.push_back(0);             // do some silly job
                                          // note tht the worker threads
                                          // would seg fault when 
                                          // the array is empty (try it!)
                                          // thank god, they're waiting, for now.
        if (my_data.size() > 1'000'000)   // exit when job is done.
            halt = true;                  // tell all that the jobn is done
        my_data_cv.notify_all();          // tell all threads to do their silly part
    }
    for (auto& t : tasks)
        t.join();
   return 0;
}

型

展开查看全部

赞(0）回复(0）举报 2024-01-09

我来回答

这个C++多线程程序有什么问题？

1条答案

相关问题

热门标签

最新问答