Abstract:
Developing parallel software for unstructured problems continues to be a difficult undertaking, particularly for distributed memory machines. Framework and library support are limited for non-standard applications and developers are often forced to code from scratch. This is particularly true for complex, unstructured applications. In this paper, we show that this needn't always be the case. We describe a set of simple primitives which can be combined to provide solutions to a variety of unstructured parallel computing problems. Specifically, we show how a small set of tools can yield efficient parallel algorithms for particle modeling, crash simulations and transferring data between two independent grids in multiphysics simulations. The use of such tools allows the application developer to program at a higher level without sacrificing performance.