Question

我有以下简单程序

#include <iostream>
#include <omp>

int main() {

    std::cout << "max threads: " << omp_get_max_threads() << "\n";

    #pragma parallel num_threads(4)
    {
        int tid = omp_get_thread_num();
        std::cout << "Hello from " << tid << " of " << omp_get_num_threads() << "\n";

        #pragma omp for
        for (int i = 0; i < 5; i++) {
            std::cout << "(" << tid << ", " << i << ")\n";
        }
    }
}

我正在使用clang++ -fopenmp=libomp main.cpp进行编译。我能够编译和运行以这种方式编译的其他OpenMP程序。

我希望num_threads(4)能够使并行区域跨越4个线程运行。相反，我会遇到以下输出：

max threads: 4
Hello from 0 of 1
(0, 0)
(0, 1)
(0, 2)
(0, 3)
(0, 4)

为什么并行区域没有跨越4个线程运行？

Answer 1

您已将omp从parallel pragma中删除。

#pragma omp parallel num_threads(4)

num_threads子句不设置线程数

1 个答案: