Parallel simulation of a large-scale aerospace system in a multicomputer environment

This work describes the application of newly developed parallel processing techniques that exploit the irregularly-structured functional parallelism present within a test-bed aerospace system (i.e., a Space Shuttle main rocket engine simulation) utilizing a reduced complexity multicomputing architecture with an arbitrary topology. The methodologies are applicable to real time parallel simulation of large-scale dynamic systems and include deterministic model optimization and decomposition techniques, automated translation mechanisms, and a newly developed task allocation heuristic. The effectiveness of these techniques are illustrated with performance being measured empirically for a wide range of topological configurations using a partially reconfigurable network of Transputers.