MPI并行计算性能的研究  

Research on the performance of MPI parallel computing

作　　者：李久楷[1][2] 朱俊[1] 宁交贤[1] Jiu-Kai Jiao-Xian 

机构地区：[1]四川大学物理科学与技术学院,成都,610064 [2]四川大学原子与分子物理研究所,成都610065

出　　处：《四川大学学报（自然科学版）》2009年第6期1659-1662,共4页Journal of Sichuan University(Natural Science Edition)

摘　　要：探索了一种基于Windows系统平台的、用于实现高性能计算的MPI并行环境.采用MPI最新版本MPICH2-1.0.6作为并行计算的支撑环境,通过编制的三个具有代表性的MPI并行计算程序,并在以100M bps交换式局域网作为互连的机群上和具有双核处理器的PC机上分别进行了并行效率的实际测试,得到了预期结果,并做了相应分析.A MPI (Message Passing Interface) parallel environment for high performance computing based on Windows platform is presented in article.Its parallel computing is controlled and supported by MPICH2 software,a latest version of MPI implementation.And a high-speed 100Mb/s Ethernet network interconnects all PCs.Tests using three parallel computing programs measure the parallel efficiency of the systems which are the PC cluster and Dual Core Processor PCs.At last,The results which are expected are analyzed and compared.

关 键 词：MPI并行计算 高性能计算 PARALLEL COMPUTING high performance COMPUTING PARALLEL efficiency Windows系统 交换式局域网 MPICH2 双核处理器 PC cluster Interface 支撑环境 实际测试 并行效率 并行环境 software expected Ethernet version results 

分 类 号：N55[自然科学总论]