Progress in parallelizing XOOPIC

XOOPIC (Object Oriented Particle in Cell code for X11-based Unix workstations) is presently a serial 2d 3v particle-in-cell plasma simulation. The present effort focuses on using parallel and distributed processing to optimize the simulation for large problems. The benefits include increased capacity for memory intensive problems, and improved performance for processor-intensive problems. The MPI library enables the parallel version to be easily ported to massively parallel, SMP, and distributed computers. The philosophy employed here is to spatially decompose the system into computational regions separated by virtual boundaries, objects which contain the local data and algorithms to perform the local field solve and particle communication between regions. This implementation will reduce the changes required in the rest of the program by parallelization. Specific implementation details such as the hiding of communication latency behind local computation will also be discussed. The initial implementation includes manual, partitioning in one spatial coordinate, electromagnetic models, diagnostics by computational region, and effective transmission of both fields and particles across virtual boundaries. This version was able to perform greater than 600,000 particle-pushes-per-second using 8 200MHz UltraSPARC CPU's. In this work the authors extend parallel XOOPIC to have 2-d partitioning, automated partitioning, and global diagnostics