Parallel Regions(contd.)

• Using replicated execution

  • make each thread work on different data
  • Do work based on thread number in team
    • Use library routine: iam = omp_get_thread_num()

• Number of threads specified by user

  • Use library routine: call omp_set_num_threads(128)
    • Applies to next parallel region
  • Environment variable setenv OMP_NUM_THREADS 128

Implementation Default < Environment variable < Runtime Library call

Previous slide

Next slide

Back to first slide

View graphic version