基于高性能微机群集的并行计算环境的设计与实现

Design and Implementation of Parallel Computing Environment Based on High Performance PCs Cluster

基于工作站或高性能微机群组的并行计算 ,是并行计算领域研究的热点 .常见的并行环境有 MPI( MessagePassing Interface)、PVM( Parallel Virtual Machine)等 ,但它们编程复杂 ,并行程序的可移植性差 .该文介绍了一种有效的并行计算模型 BSP,它的并行编程容易 ,性能可预测 ,程序的可移植性好 .最后 ,讨论了它在高性能微机群集上的程序设计和实现 .实验结果说明这种并行计算环境具有良好的性能 ,可扩展性好 .关键词 :并行计算 ;微机群集 ;BSP模型基于工作站或高性能微机群组的并行计算 ,是并行计算领域研究的热点 .常见的并行环境有 MPI( MessagePassing Interface)、PVM( Parallel Virtual Machine)等 ,但它们编程复杂 ,并行程序的可移植性差 .该文介绍了一种有效的并行计算模型 BSP,它的并行编程容易 ,性能可预测 ,程序的可移植性好 .最后 ,讨论了它在高性能微机群集上的程序设计和实现 .实验结果说明这种并行计算环境具有良好的性能 。

The research of parallelism based on workstations cluster or high performance PCs cluster, is a hot area in parallel computing. Both MPI and PVM are common parallel environments now. But they are too complicate for programming and poor in portability. This paper recommends a kind of effective parallel computing model BSP, which provides good programming environments and portability. Then its design and implementation on high performance PCs clusters are discussed. The experimental results show that this parallel computing environment exhibits good performance and scalability.