On Tue, 2012-02-07 at 11:25 +0100, Gabriele Sclauzero wrote: > The OpenMP parallelization is useful only if the code stops > complaining that there are not enough plane waves. In my experience, > before that number of processors is reached, the MPI parallelization > gives much better performances than the OpenMP one, probably because > there are still several parts of the code that have not been "threaded" so > far.
I guess you are referring to PWscf. I have been working with the Car-Parrinello code instead, which shows good OpenMP scalability. In any case, things are (slowly) improving for OpenMP parallelization in all codes. P. -- Paolo Giannozzi, IOM-Democritos and University of Udine, Italy
