performance - open omp slower than without it

Question

I'm new in "parallelization". I'm trying to parallelize an easy code to get a better performance, but I'm having very bad runtimes. here is the code:

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

using namespace std;
double omp_get_wtime(void);

int main(){
const double starttime = omp_get_wtime();
    #pragma omp parallel num_threads(10)
            {
            cout<<"sbam"<<endl;
            }
const double endtime = omp_get_wtime();
    cout << "work took time: " << endtime-starttime << " s"<<endl;
    cout << "work took time: " << (endtime-starttime)*1000000.00 << " mus"<<endl;
    return 0;
}

what am I doing wrong?? thank you!!

Answer 1

控制台是独占资源，当 10 个线程想要同时使用它时，需要锁定！

那个补充。锁定会增加开销，这解释了您所看到的时间

---

顺便说一句，磁盘访问或更确切地说是任何专有/有限资源也是如此