LS-DYNA on Advanced SGI Architectures

LS-DYNA’s implicit solver integration with explicit software allows large time steps transient dynamics as well as linear statics and normal modes analysis. Until recently,this capability could only be run on large Shared Memory Parallel (SMP) systems,where the application had access to large memory address space of the model. Distributed Memory Parallel (DMP) implementation of LS-DYNA’s implicit solver now allows the factorization of smaller mass and stiffness matrices of the decomposed problem domain by corresponding tasks in less memory. Performance enhancement through SMP processing is moreover also available in the recently introduced ‘hybrid’ mode. This paper demonstrates how advanced SGI computer systems, ranging from SMP servers addressing large memory space through multi-node clusters can be used to architect and accelerate solutions to meet complex analysis requirements.