Study of Parallel Software Development with HPF and MPI for Composite Process Modeling Simulations

摘要

High performance and parallel computers have greatly increased the feasibility of performing large-scale analysis for engineering applications. However, it does take some effort to effectively use these high performance computing (HPC) resources. Appropriate parallel algorithms must be developed, tested, and optimized. The investment in time and resources required to accomplish this task is most often substantial. Therefore, developments should take place in a framework that is reasonably portable with minimal code rewrite. In this light, we look at two portable parallel programming methodologies: High Performance Fortran (HPF) and the Message Passing Interface (MPI). These are discussed in the context of simulations for composite manufacturing processes. We describe the mathematical modeling and related implementation approaches in the two methods, highlight some strengths and weaknesses, and provide some preliminary comparisons. We also briefly present one of the current research applications in the composites process modeling arena.