Scalable
VORPAL has been designed as a parallel code, but can also be run in
serial. Its powerful domain decomposition algorithms, checkpoint
capabilities, and the use of standard data formats enable VORPAL to
execute complex plasma simulations on thousands of processors.
VORPAL is designed to run as both a serial code for single- processor
workstations and as a parallel code for systems that support MPI. Domain
decomposition techniques employed in the code enable static and dynamic
load balancing. Scaling on several thousands of processors has been
achieved for VORPAL simulations (see graphic). The use of parallel I/O
enables high performance even if gigabytes of data are dumped by
thousands of processors. Checkpoint and restarting capabilities enable
the continuation of long running simulations even if they are limited by
computing center usage policies.
Powerful
VORPAL runs on typical laptop and workstation hardware and supports
Linux (32-bit, 64-bit), IBM SP running AIX, Mac OS X (Intel), and
Windows (serial version only). For support of other
platforms, please contact Tech-X Corporation. Researchers working with
different hardware and operating systems can share input files, greatly
facilitating rapid collaboration. The data dump/restore files are also
binary compatible across hardware using the HDF5 file format. Hence,
the results of a long simulation on a dedicated high-performance Linux
cluster, for example, can be copied to a laptop running Linux, Microsoft
Windows, or Mac OS X for further study.