【c++】OpenMP自动并行

标题

自动并行相关技术

文档教程OpenMP 推荐！！！下面的几篇博客为此教程的翻译整理版本。
官网：OpenMP应用程序编程接口

入门博客：

  1. 最简单的并行计算——OpenMP的使用
  2. OpenMP入门教程（一）
  3. OpenMP入门教程（二）
  4. OpenMP入门教程（三）
  5. 通过 GCC 学习 OpenMP 框架
  6. OpenMP编程入门之一
  7. OpenMP中几个容易混淆的函数（线程数量/线程ID/线程最大数）以及并行区域线程数量的确定
  8. 并行编程-OpenMP（必看!）
  9. 嵌套并行操作（必看!）

omp例子

#include <iostream>
#include <omp.h>
#include <thread>

void report_num_threads(int level)
{
  // #pragma omp single
  {
      printf("Level %d: number of threads in the team: %d\n",
                level, omp_get_num_threads());

    // 获取当前线程的 ID
    std::thread::id this_id = std::this_thread::get_id();
    // 输出线程 ID
    std::cout << " Current thread ID: " << this_id << std::endl;
  }
 }

int main() {
    const int n = 32;
    int sum = 0;
    // int data[n];
    int a1 = 0;
    // int data2[n];
    int a2 = 0;
    // int data3[n];
    int a3 = 0;
    // int data4[n];
    int a4 = 0;
    // int data5[n];
    int a5 = 0;

    bool is_nested = false;
    is_nested = omp_get_nested();
    std::cout<<"是否支持并行嵌套：" << is_nested << "\n";
    omp_set_nested(1);//设置支持嵌套并行, 嵌套则会产生 n*m个线程(线程id不同)
    is_nested = omp_get_nested();
    std::cout<<"是否支持并行嵌套：" << is_nested << "\n";

    omp_set_num_threads(4); // 这个设置一个最大线程数，其他地方不指定线程数，然后开启嵌套，不会增加线程数，但是可以嵌套并行，然后指定的for里面会按照指定的来，没有指定的全部按照这个默认值

    double start_time = omp_get_wtime();

    for (int j = 0; j < 1; j++) {
      #pragma omp parallel for // num_threads(4)
      for (int i = 0; i < 1; ++i) {
          std::cout << "Thread " << omp_get_thread_num() << ": "  << std::endl;
         int sum = 0;
         report_num_threads(1);
         #pragma omp parallel for //num_threads(3)
         for (int k = 0; k < 10; ++k) {
            sum = sum + 1;
            report_num_threads(2);
         }
         report_num_threads(3);
         std::cout << "Thread " << omp_get_thread_num()
                   << " sum=" << sum 
                   << std::endl;
      }
    }
    report_num_threads(4);
    double end_time = omp_get_wtime();
    std::cout << "Time taken: " << end_time - start_time << " seconds" << std::endl;

    return 0;
}

// cmd: g++ -fopenmp -o sum_openmp openmp_test.cpp && ./sum_openmp

/*
output:
ID: 0, Max threads: 52, Num threads: 1 
ID: 0, Max threads: 5, Num threads: 1 
ID: 0, Max threads: 5, Num threads: 5 
ID: 4, Max threads: 5, Num threads: 5 
ID: 2, Max threads: 5, Num threads: 5 
ID: 1, Max threads: 5, Num threads: 5 
ID: 3, Max threads: 5, Num threads: 5 
ID: 0, Max threads: 5, Num threads: 1 
ID: 0, Max threads: 6, Num threads: 1 


ID: 0, Max threads: 6, Num threads: 6 
ID: 3, Max threads: 6, Num threads: 6 
ID: 1, Max threads: 6, Num threads: 6 
ID: 4, Max threads: 6, Num threads: 6 
ID: 2, Max threads: 6, Num threads: 6 
*/

嵌套并行操作

嵌套并行操作
例子：

#include <iostream>
#include <omp.h>

void report_num_threads(int level)
{
  // #pragma omp single
  {
      printf("Level %d: number of threads in the team: %d\n",
                level, omp_get_num_threads());
  }
 }

int main() {
    const int n = 32;
    int sum = 0;
    // int data[n];
    int a1 = 0;
    // int data2[n];
    int a2 = 0;
    // int data3[n];
    int a3 = 0;
    // int data4[n];
    int a4 = 0;
    // int data5[n];
    int a5 = 0;

    bool is_nested = false;
    is_nested = omp_get_nested();
    std::cout<<"是否支持并行嵌套：" << is_nested << "\n";
    omp_set_nested(1);//设置支持嵌套并行
    is_nested = omp_get_nested();
    std::cout<<"是否支持并行嵌套：" << is_nested << "\n";

    omp_set_num_threads(4);

    double start_time = omp_get_wtime();

    for (int j = 0; j < 1; j++) {
      #pragma omp parallel for num_threads(4)
      for (int i = 0; i < 2; ++i) {
          std::cout << "Thread " << omp_get_thread_num() << ": "  << std::endl;
         int sum = 0;
         report_num_threads(1);
         #pragma omp parallel for num_threads(16)
         for (int k = 0; k < 3; ++k) {
            sum = sum + 1;
            report_num_threads(2);
         }
         report_num_threads(3);
         std::cout << "Thread " << omp_get_thread_num()
                   << " sum=" << sum 
                   << std::endl;
      }
    }
    report_num_threads(4);
    double end_time = omp_get_wtime();
    std::cout << "Time taken: " << end_time - start_time << " seconds" << std::endl;

    return 0;
}

// cmd: g++ -fopenmp -o sum_openmp openmp_test.cpp && ./sum_openmp

/*
是否支持并行嵌套：0
是否支持并行嵌套：1
Thread 0: Thread 
1: 
Level 1: number of threads in the team: 4
Level 1: number of threads in the team: 4
Level 2: number of threads in the team: 16
Level 2: number of threads in the team: 16
Level 2: number of threads in the team: 16
Level 2: number of threads in the team: 16
Level 2: number of threads in the team: 16
Level 2: number of threads in the team: 16
Level 3: number of threads in the team: 4
Thread 1 sum=3
Level 3: number of threads in the team: 4
Thread 0 sum=3
Level 4: number of threads in the team: 1
Time taken: 0.000629196 seconds*/