The primary implications of parallel processing on h-p adaptive finite element methods for electromagnetic analysis are investigated. Aside from the conventional benefits and costs associated with the parallelization of the essentially non-adaptive finite element modules, significant fundamental advantages that are unique to the adaptive process itself are explored. First, the overall speedup potential of local error estimator evaluation and h-p discretization refinement is superior to that of finite element solver execution in parallel environments, and therefore justifies the use of more complex and computationally intensive adaption control strategies. Second, the availability of parallel processing motivates the comparative assessment of different discretization strategies at each h-p refinement step to help guide the evolution of the adaption. Practical results representing a range of parallel configurations are computed to illustrate the concepts.
展开▼
