openmp 并行中的 std::lock_guard

Question

#include <iostream>
#include <omp.h>
#include <mutex>
using namespace std;

int main()
{
    mutex l;

#pragma omp parallel for
    for(int i=0;i<24;i++){
        int thread_num = omp_get_thread_num();
        std::lock_guard<std::mutex> lg(l);
        cout<<"i: "<<i<<" thread_num:"<<thread_num<<endl;
    }

    return 0;
}

Hi, I can't understand the output of such code. Looks like that mutex is not unlocked after each loop. But lock_guard is defined inside loop, isn't it?

i: 0 thread_num:0
i: 1 thread_num:0
i: 2 thread_num:0
i: 3 thread_num:0
i: 4 thread_num:0
i: 5 thread_num:0
i: 12 thread_num:2
i: 13 thread_num:2
i: 14 thread_num:2
i: 15 thread_num:2
i: 16 thread_num:2
i: 17 thread_num:2
i: 6 thread_num:1
i: 7 thread_num:1
i: 8 thread_num:1
i: 9 thread_num:1
i: 10 thread_num:1
i: 11 thread_num:1
i: 18 thread_num:3
i: 19 thread_num:3
i: 20 thread_num:3
i: 21 thread_num:3
i: 22 thread_num:3
i: 23 thread_num:3

My system is linux with g++ 9.1.0

Answer 1

Google is your friend. You can find documentation of std::lock_guard at https://en.cppreference.com/w/cpp/thread/lock_guard

If you read it it explains how this works.