摘要
文中提出一个IPBPS(InterconnectedProcessor BasedParallelSorting)并行分类算法,运行在由独立处理器构成的计算机网络上,以解决网络分布式数据库的分类计算问题。基于并行算法应与并行计算的拓扑结构相匹配的思想,设计了一种旨在减小处理器间通信开销的网络结构。在此并行计算环境中,每个处理器执行同样的程序,计算负载均匀分布在每个处理器中,具有较高的加速比。同时,这种基本的处理器互联结构可灵活扩展,且随着网络的扩大,算法的并行加速比更高。
A parallel sorting scheme called IPBPS(Interconnected ProcessorBased Parallel Sorting) based on a computer network connected by individual processors was presented,with the aim to solve the timeconsuming sorting problem in distributed database applications. Based on the idea of mapping the parallel algorithm with the topological structure of parallel computing, a network configuration with minimized interprocessor communication cost is designed. Under this parallel computing environment, every processor in the network executes the same program, and computing load is evenly distributed among the processors in the network, high parallelism can be achieved. In addition, the basic 10 processor configuration can be expanded flexibly into larger networks. As the network size grows, the parallel speedup of IPBPS algorithm becomes even high.
出处
《计算机应用》
CSCD
北大核心
2003年第7期9-12,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目 (6 9976 0 2 0 )
