多群粒子输运问题在多核集群系统上的混合并行计算  

Hybrid Parallel Computation of Multi-Group Particle Transport Equations on Multi-Core Cluster Systems

作　　者：迟利华[1] 刘杰[1] 龚春叶[1] 徐涵[1] 蒋杰[1] 胡庆丰[1] 

机构地区：[1]国防科技大学计算机学院,湖南长沙410073

出　　处：《计算机工程与科学》2009年第11期94-97,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60673150;60603061);国家863计划资助项目(2008AA01Z137)

摘　　要：本文分析了非结构网格多群粒子输运Sn方程求解的并行性,拟合多核机群系统的特点,设计了MPI/OpenMP混合程序,针对空间网格点采用区域分解划分,计算结点间基于消息传递MPI编程,每个MPI计算进程在计算过程中碰到关于能群的计算,就生成多个OpenMP线程,计算结点内针对能群进行多线程并行计算。数值测试结果表明,非结构网格上的粒子输运问题的混合并行计算能较好地匹配多核机群系统的硬件结构,具有良好的可扩展性,可以扩展到1024个CPU核。The parallel performance of solving the multi-group particle transport equations on the unstructure meshes is analyzed. Adapting to the characteristics of multi-core cluster systems, this paper desgins a MPI/OpenMP hybrid parallel code. For the meshes, the space domain decomposition is adopted, and MPI between the computations of multi-core CPU nodes is used. When each MPI process begin to compute the variables of the energy groups, several OpenMP threads will be forked, and the threads start to compute simultaneously in the same mutli-core CPU node. Using the MPI/OpenMP hybrid parallel code, we solve a 2D mutli-group particle transport equation on a cluster with mutli-core CPU nodes, and the results show that the code has good scalability and can be scaled to 1024 CPU cores.

关 键 词：MPI/OpenMP混合并行计算 粒子输运 Sn方法 多核机群系统 

分 类 号：TP301[自动化与计算机技术—计算机系统结构] O571.51[自动化与计算机技术—计算机科学与技术]