The Development of an E cient Direct

This paper describes initial research into the development of an eeective parallel implementation of the direct simulation Monte Carlo (DSMC) method. DSMC is a widely established particle simulation method for the computation of rareeed gas ows, and in many cases is the only eeective means of ow prediction. The method is however computationally expensive in terms of time and memory usage, and further applicability is hampered by current serial computer power. Massively parallel architectures provide a means of overcoming this diiculty, not only through their sheer power but also because the DSMC method is inherently parallel. A review of the DSMC method is presented along with justiication for parallelisa-tion. The parallel implementation is then discussed with some speciic information for the AP1000. Results are presented showing some encouraging speedups. However the need for an eeective load balancing algorithm is highlighted by these, and an adaptive domain decomposition algorithm is presented for this purpose.