Performance Opimization of a Parallelized Three-Dimensional Monte-Carlo Ion Implantation Simulator

摘要

We present a parallelization method based on MPI (Message Passing Interface) for a Monte-Carlo simulator for the simulation of ion implantation into three-dimensional structures. By this method the simulation domain is geometrically distributed among several CPUs which exchange simulation data during the simulation. We optimize the performance gain by identifying bottlenecks of this strategy when it is applied to arbitrary shaped simulation domains consisting of various materials, which requires the application of varying physical models within the simulation domain and which makes it impossible to determine a reasonable domain distribution before starting the simulation. Due to a feedback procedure between the parallelization strategy and the simulation by online performance measurements, we obtain an almost linear performance gain on a cluster of workstations with just slightly varying processor loads. Besides an increase in the performance the parallelization method achieves a distribution of the memory requirement, which allows also the use of small workstations for three-dimensional simulations.