OMP_PLACES 未在此范围内声明

Question

我正在寻找在我的代码中使用 OMP_PLACES 和 OMP_PROC_BIND=close 的正确方法。当我运行这个时，我得到的错误是：

error: ‘OMP_PLACES’ was not declared in this scope


#include <iostream>
#include <random>
#include <fstream>
#include <iomanip>
#include <cfloat>
#include <omp.h>
#include <stdlib.h>
using namespace std; 

int main()
{
    const int N = 100;
    unsigned seed = 0;
    double sum = 0.0;
    double avg;
    srand(seed);
    default_random_engine g(seed);
    uniform_real_distribution<double> distribution(0.0f, nextafter(1.0f, DBL_MAX));
    #pragma omp parallel 
    {
        OMP_PLACES="cores(4)";
        OMP_PROC_BIND=close;
    
     #pragma omp for reduction (+: sum) 
     for (int i = 0; i < N; i++) {

        double x = distribution(g);
        sum = sum + x;
    }

    avg = sum / static_cast<double> (N);
    
     
    }

    

}

Answer 1

必须在 shell 环境中设置 OMP_PLACES 和 OMP_PROC_BIND 变量。所以，如果你使用 bash，那么你可以这样做：

$ export OMP_PLACES="cores(4)"
$ export OMP_PROC_BIND=close
$ ./your_application

另一种方法是在您的代码中使用 proc_bind() 子句：

// header omitted

int main()
{
    const int N = 100;
    unsigned seed = 0;
    double sum = 0.0;
    double avg;
    srand(seed);
    default_random_engine g(seed);
    uniform_real_distribution<double> distribution(0.0f, nextafter(1.0f, DBL_MAX));
    #pragma omp parallel proc_bind(close)
    {
        #pragma omp for reduction (+: sum) 
        for (int i = 0; i < N; i++) {
            double x = distribution(g);
            sum = sum + x;
        }
       // remainder of code to follow
    }
  ...
}

你仍然需要

$ export OMP_PLACES="cores(4)"

指定您希望代码运行所在的位置。

PS：您可能还想查看有关随机数生成器的 C++ 线程安全性的 SO 页面：C++11 Thread safety of Random number generators

OMP_PLACES 未在此范围内声明

OMP_PLACES was not declared in this scope

multithreading

openmp

multiprocessing