Massively Parallel Computing

Our density functional code( Dacapo ) is fully parallelized.  Parallelization can be done over either the k-points used for sampling the surface Brillouin zone or the plane-wave used in the calculations.  The code can be used with either the Message Passing Interface or the Parallel Virtual Machine message passing libraries.

Our principle computational workhorse is a set of 3 Beowulf clusters with Compaq Alpha and AMD/Intel processors.  These facilities, under the control of the Portable Batch System queueing software, allows us to perform calculations for unit cells of up to 30 metal or semiconductor atoms within reasonable time frames.   

In addition to our in-house computational resources, we make extensive use of National Energy Research Scientific Computing Center , National Partnership for Advanced Computational Infrastructure and ERDC resources.  These government systems, sponsored by DOE, NSF and DOD respectively are comprised of clusters of IBM SP-2 and SP-3 workstations.