Load Balancing in PIC Parallel Implementation

摘要

Heterogeneous computing opens up new challenges and opportunities in fields such as parallel processing, design of algorithms for applications, partitioning and mapping of parallel tasks, interconnection network technology and the design of heterogeneous programming environment. Lots of load balancing algorithm have been proposed and experimented with in the past years for homogeneous parallel and distributed systems. In this paper, an efficient parallel processing methodology which is based upon the particle-in-cell (PIC) code technique for plasma simulations is described. The methodology can be applied to solve fully three-dimensional problems in plasma physics which occur in fusion machines and space. In this paper, the methodology is shown to effectively partition and allocate a sizable plasma dynamics problem (i.e. the nonlinear evolution of lower hybrid waves in an auroral plasma) onto multiple processing nodes of a cluster of heterogeneous workstations using the Parallel Virtual Machine (PVM).