Parallel strategies for crash and impact simulations

We describe a general strategy we have found effective for parallelizing so lid mechanics simulations. Such simulations often have several computationa lly intensive parts, including finite element integration, detection of mat erial contacts, and particle interaction if smoothed particle hydrodynamics is used to model highly deforming materials. The need to balance all of th ese computations simultaneously is a difficult challenge that has kept many commercial and government codes from being used effectively on parallel su percomputers with hundreds or thousands of processors. Our strategy is to l oad-balance each of the significant computations independently with whateve r balancing technique is most appropriate. The chief benefit is that each c omputation can be scalably parallelized. The drawback is the data exchange between processors and extra coding that must be written to maintain multip le decompositions in a single code. We discuss these trade-offs and give pe rformance results showing this strategy has led to a parallel implementatio n of a widely used solid mechanics code that can now be run efficiently on thousands of processors of the Pentium-based Sandia/Intel TFLOPS machine. W e illustrate with several examples the kinds of high-resolution, million-el ement models that can now be simulated routinely. We also look to the futur e and discuss what possibilities this new capability promises, as well as t he new set of challenges it poses in material models, computational techniq ues, and computing infrastructure. (C) 2000 Elsevier Science S.A. All right s reserved.