序列化部分工作 - boost::asio

Posted 2023-02-22

【中文标题】序列化部分工作 - boost::asio

【英文标题】：Serializing part of work - boost::asio

【发布时间】：2011-04-21 11:33:01

【问题描述】：

    void wt(shared_ptr<io_service> io_servicee)

    io_servicee->run();


void calculateAndReturn(int myNumber, shared_ptr<vector<int> > vectorOfResults)

    /*
     * Here will be a lot of pararel computing (first part)...
     */

    int result = myNumber;

    /* This part (second part) below I want to be executed sequentially depending on the number of thread.
     * Thread which has myNumber == 1 should be first, second should be thread with
     * myNumber == 2, third with myNumber == 3 etc.
     *
     */
    mt.lock();
    vectorOfResults->push_back(result);
    mt.unlock();


int main()

    shared_ptr< io_service > io_servicee(new io_service);
    shared_ptr< io_service::work > work(new io_service::work( *io_servicee ));
    shared_ptr<vector<int> > vectorOfSums(new vector<int>);
    thread_group worker_threads;


    for( int x = 0; x < 4; ++x )
    
            worker_threads.create_thread(bind( &wt, io_servicee ));
    

    for( int x = 0; x < 4; ++x )
    

        io_servicee->post(bind( &calculateAndReturn, x, vectorOfSums));
    
    work.reset();
    worker_threads.join_all();

    for ( int i = 0; i < vectorOfSums->size(); i++)
    
        cout << (*vectorOfSums)[i] << endl;

    

我想序列化部分calculateAndReturn 函数。它应该是这样的：

First part of thread 1 ------->  \ 
                                  \
First part of thread 2 -------> -- Second part of thread 1 --> second part of thread2...
                                  /
     ....                        /
                                /
First part of thread n ------->/

我不想在 main 函数中使用 strand，而是在 calculateAndReturn 的第二部分之前使用。有什么优雅的解决方案吗？

最好的问候

【参考方案1】：

如果您在单个进程中执行此操作，我建议使用OpenMP。典型例子：

#include <omp.h>
#define N 1000
#define CHUNKSIZE  100

main ()  

    int i, chunk;
    float a[N], b[N], c[N];

    /* Some initializations */
    for (i=0; i < N; i++)
      a[i] = b[i] = i * 1.0;
    chunk = CHUNKSIZE;

    #pragma omp parallel for shared(a,b,c,chunk) private(i) schedule(static,chunk)
    for (i=0; i < n; i++)
        c[i] = a[i] + b[i];

在 gnu 上：安装 libgomp，构建

g++ -fopenmp -lgomp source.cpp

【讨论】：

我没听说过这个库，看起来很有用，你的代码更简单。但在这种情况下，我想使用 boost 和 asio 库，因为我想编写简单的 tcp 服务器。还是谢谢！