Performance of the NAS Benchmarks on a Cluster of SMP PCs Using a Parallelization of the MPI Programs with OpenMP

The availability of multiprocessors and high performance networks offer the opportunity to build CLUMPs (Cluster of Multiprocessors) and use them as parallel computing platforms. The main distinctive feature of the CLUMP architecture over the usual parallel computers is its hybrid memory model (message passing between the nodes and shared memory inside the nodes). To be largely used, the CLUMPs must be able to execute the existing programs with few modifications. We investigate the performance of a programming approach based on the MPI for inter-multiprocessor communications and OpenMP standards for intra-multiprocessor exchanges. The approach consists in the intra-node parallelization of the MPI programs with an OpenMP directive based parallel compiler. The paper details the approach in the context of the biprocessor PC CLUMPs and presents a performance evaluation for the NAS parallel benchmarks.