Scalability of LS-DYNA on SGI Systems

In parallel computation, the scalability of the application software is critical. It is especially important when large number of processors are used. In this paper, we present the scalability results of LS-DYNA on the SGI multiprocessor computer systems. Furthermore, since MPP- DYNA is a domain decomposition based software, data partitioning algorithms play an important role in the scalability of the code. We will show in this paper that for some car crash models, special data partitioning techniques may improve the scalability significantly.