摘要
提出了采用双核处理器的共享存储多处理机(SMP)作为计算节点时,多粒度并行计算集群的结构。研究了此类系统的并行计算粒度和优化方法,描述了该集群MPI+OpenMP的多粒度并行计算平台构建方法。利用此平台,实现了求解现行方程组的SOR迭代算法,通过数值测试,表明此类集群具有良好的计算性能。
The multi-granularity parallel computing cluster＇s structure is introduced, when the computing node is Symmetric Multiprocessor （SMP） with the dual core CPUs. The granularity of parallel computing and optimization are researched. The details of constructing platform of hybrid MPI ＋ OpenMP parallel programming are provided. SOR （Successive Over Relaxation）＇s parallel algorithm for solving linear systems of equations is implemented based on this platform. Numerical results show that the cluster is effective.
出处
《电力学报》
2007年第4期436-438,共3页
Journal of Electric Power
基金
重庆市科委资助项目(CST2005BB0061)的资助
