基于MPI和Taurus高性能计算系统的Jacobi并行迭代算法  

Jacobi parallel iteration algorithms based on MPI and Taurus high performance computing system

作　　者：张海龙[1,2,3] 张萌 王杰[1] 冶鑫晨 王万琼[1] 朱艳 ZHANG Hai-long;ZHANG Meng;WANG Jie;YE Xin-chen;WANG Wan-qiong;ZHU Yan(Xinjiang Astronomical Observatory,Chinese Academy of Sciences,Urumqi 830011,China;University of Chinese Academy of Sciences,Beijing 100049,China;Key Laboratory of Radio Astronomy,Chinese Academy of Sciences,Nanjing 210008,China)

机构地区：[1]中国科学院新疆天文台,乌鲁木齐830011 [2]中国科学院大学,北京100049 [3]中国科学院射电天文重点实验室,南京210008

出　　处：《吉林大学学报（工学版）》2019年第2期606-613,共8页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(11873082;U1531125;11803080;11503075);"973"国家重点基础研究发展计划项目(2015CB857100);中国科学院青年创新促进会项目;中国科学院天文台站设备更新及重大仪器设备运行专项经费支持项目

摘　　要：针对Jacobi迭代的海量计算问题,设计了大规模并行计算算法。通过非阻塞通信函数替代阻塞通信函数、采用虚拟进程拓扑方式改进数据的区块划分,并利用高性能集群系统多计算节点协同处理对Jacobi并行迭代进行了尝试。实现了基于MPI的C语言串行与并行算法,利用Taurus HPC分别对串行、并行,单节点、多节点并行算法进行了系统测试。测试结果表明,进程间数据通信效率是影响并行程序性能的重要因素;跨多节点执行对于海量计算任务可显著提高计算速度;合理的数据区块划分有利于处理器的任务调度,可有效提高Jacobi并行迭代算法的执行效率。Large-scale parallel computing algorithms have been designed for Jacobi iterative computing problems.Systematic improved the algorithm efficiency of Jacobi iteration by using a non-blocking communication function against blocking one,the topology method of virtual process,and the high performance computing system.The MPI-based C language serial and parallel algorithms were implemented,and the Taurus HPC was used to test the serial,parallel,single-node and multi-node parallel algorithms respectively.Experimental results shows that data communication efficiency between processes was an important factor of parallel program performance;across multiple computing nodes for compute-intensive tasks could significantly speed up the calculation;reasonable data block division was helpful for the task scheduling of the processors,which could greatly improve the parallel execution efficiency of Jacobi iteration algorithms.

关 键 词：计算机应用 MPI程序 JACOBI迭代 并行算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]