[Pw_forum] Trying to find reasonable parallelization parameters

J. J. Ramsey jjr19 at uakron.edu
Thu Jan 22 18:01:09 CET 2009


The cluster I'm running on uses Infiniband for MPI communications, and
the queueing software by default limits me to 1280 processor cores for a single
job, 1.8 GB per core, and a wall time of 96 hours. As warned about in the user guide, I'm running into problems with running out of memory when using pw.x, so I need to find appropriate values for npool, ntg, and ndiag. (I don't have an NEB calculation, so I don't thing that the # of images is an issue.) I'm a bit unsure about sane values of these, since as far as I can tell, my simulation isn't a "very large" job but isn't exactly small, either: 213 atoms, 727 electrons, and I'm using 4x4x1 k-point sampling. (It's a slab calculation.)

How many cores, roughly, is reasonable for a simulation like this, and what's a reasonable way to divide it up into pools, ortho groups, etc.?



      



More information about the users mailing list