How to parallelize a loop?

Question

How to parallelize a loop?

I have about a trace. code

for (;;)
{
   do
  {
   //тут нужно расспараллелить
  }
}

I would like to parallelize what is in the loop. But the trouble is that if I write #pragma omp parallel where I need to parallelize, it will create and delete threads as many times as the do is executed. And I would like to say outwardly that we need to create two threads that will execute this code, while they do not need to be deleted and created 100500 times.

0

openmp

Author: van9petryk, 2016-05-31

Source

2 answers

score 1 · Answer 1

Omp_set_dynamic + omp_set_num_threads will allow you to prevent changing the number of threads and set this number, respectively.

1

Author: Владимир Мартьянов, 2016-05-31 16:24:29

score 0 · Answer 2

Threads will be created when entering #pragma omp parallel. Next, they will be assigned iterations, the threads will continue to live without being destroyed. So it will be possible to parallelize only for.