Hi, I should work fine. I'm running some rough tests now. It will be difficult to come up with good tests though as the different optimisation classes (grid search, standard optimisation, clustered and non-clustered), different models, and different systems (multi-core vs. a real cluster) will all result in different scaling efficiencies. Feel free to play around with it and see how it performs on your cluster. For some of your aims, it may prove to not be of any benefit. Regards, Edward On 11 September 2013 18:39, Troels Emtekær Linnet <tlinnet@xxxxxxxxx> wrote:
Hi Edward. I just tried it out, and I got it to work. :-) I added these sections to the wiki: http://wiki.nmr-relax.com/Installation_linux#mpi4py http://wiki.nmr-relax.com/Installation_linux#Script_for_running_the_disp_branch_of_relax_with_maximum_number_of_processors_available Section on mpi4py http://wiki.nmr-relax.com/Mpi4py Should some speed test be provided here? Looking forward to try it ! Troels Emtekær Linnet 2013/9/11 Edward d'Auvergne <edward.dauvergne@xxxxxxxxx>:That looks really interesting!:)Is there a special code which need to be specified to run with multiprocesser. http://www.nmr-relax.com/manual/Usage_multi_processor.htmlThe code will run on the uni-processor or the multi-processor. It's the same code.Would you have to start relax with this mpirun, and analyse a script? mpirun -np 257 /usr/local/bin/relax -multi=`mpi4py' -tee log dauvergne_protocol.pyNo, you can run relax as you normally would, but replace the 'relax' at the start with: mpirun -np N+1 /usr/local/bin/relax -multi=`mpi4py' where N is the number of slaves you have. See the mpirun documentation for details - this is not part of relax. This code runs in the GUI, the script UI and the prompt UI, i.e. everywhere! Regards, Edward