A Hybrid MPI-OpenMP Implementation of an Implicit Finite-Element Code on Parallel Architectures

摘要: SummaryThe hybrid MPI-OpenMP model is a natural parallel programming paradigm for emerging architectures that are based on symmetric multiprocessor (SMP) clusters. This paper presents implementation adapted an implicit finite-element code developed groundwater transport simulations. The original was parallelized distributed memory using MPI (Message Passing Interface) domain decomposition strategy. OpenMP directives were then added to the (a straightforward loop-level implementation) use multiple threads within each process. To improve performance, several loop modifications adopted. performance results compared four modern architectures. show most of cases tested, pure approach outperforms model. exceptions this observation mainly due limitation in library one A general conclusion ...