多核计算机上并行计算的实现与分析  

Implementation and Analyses of Parallel Computation on Multi-core Computers

作　　者：王顺绪[1] 

机构地区：[1]淮海工学院理学院,江苏连云港222005

出　　处：《淮海工学院学报（自然科学版）》2009年第3期30-33,共4页Journal of Huaihai Institute of Technology:Natural Sciences Edition

基　　金：淮海工学院科研经费资助项目(KQ08031);淮海工学院教学改革立项课题(5509002)

摘　　要：研究多核计算机上OpenMP编程模式和MPI消息传递机制编程模式的运行效能,并在双核和四核计算机上,分别使用传统算法和并行算法计算了和式sum from i=1 to n i和n×n矩阵的乘积以及矩阵的Cholesky分解,数值试验结果表明,传统的串行程序只能利用二分之一的双核CPU资源和四分之一的四核CPU资源,而并行程序在双核和四核计算机上的并行效率很高,并且OpenMP编程模式的运行效率比MPI编程模式的运行效率稍高。Parallel programming in Fortran using OpenMP and MPI on multi-core computers are studied. The serial and concurrent experiments of matrix-matrix multiplication, summation of 1＋2＋…＋n and Cholesky factorization are implemented on 2-core and 4-core computers respectively. Experiment results show that the serial program can only use one core of multi-core computers. Meanwhile, the parallel efficiency of the program using OpenMP and MPI is very high, and the efficiency of the program using OpenMP is a little higher than that of the program using MPI.

关 键 词：多核编程 OPENMP MPI 

分 类 号：TP338[自动化与计算机技术—计算机系统结构] O246[自动化与计算机技术—计算机科学与技术]