High Performance Estuarine and Coastal Environmental Modeling: Part II

摘要

As ecosystem management and restoration becomes increasingly more important so does the researcher's reliance on high performance techniques to solve the long-term, fine resolution, dynamically coupled problems of today and tomorrow. Two such techniques, parallelism via shared memory constructs applied to a legacy code and parallelism via message passing applied to a new code, are discussed herein. First, a CH3D-based integrated modeling system (IMS) which includes the major components of a 3D coupled, hydrodynamic, wave, flushing, sediment, water quality and light models has been made parallel using a shared memory approach. Using an Indian River Lagoon grid system (477 x 43 x 4), the integrated hydrodynamics, sediment and water quality models achieved a parallel speedup of 3 when using four processors of an SGI Origin 2000. Second, a new 2D hydrodynamic, wetting and drying model (PEM) has been developed specifically for parallel application. The explicit and semi-implicit versions of the model achieved speedups of 10.5 and 18.6, respectively, when executed on a 20 processor custom built Beowulf Cluster.