Development and performance of a PVM based parallel geometric modeler for MEMS

This paper describes a successful approach to improving the robustness and speed of Sandia National Laboratory’s 3-D MEMS geometry modeler through a combination of mask subdivision and code parallelization. Symptoms of the robustness problems experienced with the original modeler that suggested the need for the subdivision strategy pursued here are explained. The basic elements of the subdivision approach and the subsequent parallel implementation are described. Results and timings for a number of different modeling problems are presented and discussed in order to illustrate the effectiveness of the current approach.