call function in openmp parallelized loop

Question

I am new to OpenMP and I have a loop that I want to parallelize with OpenMP. Inside the parallel loop a subroutine is called. Here my code:

#pragma omp parallel for shared(dir,utilitiespath,frequency,solvent,Method,dispersion)  private(beginning,file_cp2k,file_geo,file_energy)
for(i=1;i<=n_conformers;i=i+1)
{
    time(&beginning);
    file_cp2k="conformer"+ QString::number(i) +".inp";
    file_geo= "conformer" +  QString::number(i) + "_geoMM.sdf";
    file_energy=dir+ "conformer" + QString::number(i)  + "_enerSE";
    cout<<"loop "<<i<<" time "<<beginning<< endl;
    int n_atom=Makecp2kOptInput(file_geo, file_cp2k, dir,utilitiespath, frequency, solvent, Method,dispersion);
}

I get the following error:

* Error in `./ChemAliveMolOpt': double free or corruption (fasttop): 0x00000000018d7c00 * Aborted (core dumped)

If I remove the call to the subroutine there is no problem.

How should I proceed to parallelize the loop properly?

Thanks

Make your subroutine call sequential. Also, calling time won't work as you're expecting it to, probably. — erip
– erip, Commented Feb 8, 2016 at 12:20
@erip Well this is the subroutine that is long and I would like to have several instance of it running — Laetis
– Laetis, Commented Feb 8, 2016 at 14:29
Seeing as you're listing the behavior of every variable, you should also make i private and n_conformers shared. However, without us knowing what the function Makecp2kOptInput does and how it interacts with these variables, we can't help you further. — NoseKnowsAll
– NoseKnowsAll, Commented Feb 8, 2016 at 15:51

Sven Nilsson · Accepted Answer · 2016-02-08 15:55:35Z

0

All code that the OpenMP-enabled loop executes must be thread safe. Your subroutine is probably not thread safe. Thread safety essentially means that you synchronize the access to all global/shared objects.

This is not an easy topic, so I recommend you study thread safety and mutex-based synchronization in order to solve your problem.

answered Feb 8, 2016 at 15:55

Sven Nilsson

1,90912 silver badges16 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

call function in openmp parallelized loop

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related