PARALLEL PERFORMANCE OF LS-DYNA ON THE NEXT GENERATION OF COMPUTER SYSTEMS

LS-DYNA is available on a wide variety of computer platforms, ranging from commonplace personal computers to sophisticated vector processors to high-performance scalable servers. This presentation will review historical LS-DYNA performance and look forward to what can be expected in the next generation of computer platforms. The capability of available computer platforms always has a bearing on what type of simulations LS-DYNA end users can run. In the past, departmental machines such as DEC VAX servers provided many computing cycles to users. Porting of the code to Cray vector machines made overnight full-car crash simulations practical. Models grew larger but could still be run overnight as machine speeds increased and SMP parallel versions became available. At the same time, UNIX workstations increased in capability to the point where many sophisticated simulations could be run on desktop machines. Most recently, significant advances in LS-DYNA and computer hardware have taken place in two different areas: scalable servers, and machines based on low-cost processors. The advances are being driven by the growing acceptance of the MPI version of LS-DYNA, the level of performance available from low-cost commodity processors, and the level of performance available from high-performance processors available in scalable systems. Current performance examples and trends in microprocessor improvement can be used to give users guidance on what levels of LS-DYNA performance will be attainable in the next few years.