what


These two MPI based projects demonstrate two different approaches to carrying out heavy computational Molecular Dynamics projects on NANCO. Both provide speedup data for simulations of realistic sizes and times.

The first is a model of nanotube vibrations from a project co-supervised by Yuval Yaish, is not memory bound, at least for now. It is mainly complicated by the need to run the same code for many different parameters and by an enormous load of fourier transfom analysis. We may call this a bookeeping bound project.

  • Polina Pine (email) has written a code that enables simultaneous submission of multiple jobs with outputs that are arranged in a way that minimises bookeeping needs. (See image on left of a vibrating nanotube.)

    The second project is severely memory bound and also somewhat time bound It is the start of a project on HCP crystal melting which is being co-supervised by Emil Polturak and also a test of NANCO's capabilities. The sample sizes needed are way beyond those that can be modeled on a few procesors, hence the speedup is calculated for small samples on a few processors and for larger ones on more processors.

  • Pavel Bavli (email) has achieved super linear speedup between 4 and 8 processors. Although the exhaustive speedup statistics are for smaller samples, Pavel achieved Molecular Dynamics simulation of one million particles and could go larger too. (See image on right of only 256,000 atoms.)

    HOME Projects Publications