基于MPI和OpenMP的排序算法并行优化研究  

The Performance Analysis of Sorting Algorithms based on MPI and OpenMP

作　　者：王帅[1] 喻歆[1] 何嘉[1] 

机构地区：[1]成都信息工程大学计算机学院,四川成都610225

出　　处：《成都信息工程大学学报》2016年第3期277-284,共8页Journal of Chengdu University of Information Technology

基　　金：四川省科技计划资助项目(2012GZ0111)

摘　　要：排序是计算机程序设计中的一项重要操作,其性能好坏决定整个程序性能的优劣。针对常见的快速排序、冒泡排序、归并排序、计数排序和选择排序这5种排序算法,分别用MPI(message passing Interface)和OpenMP(open multi-processing)并行化编程环境对其进行并行程序优化,研究分析MPI和OpenMP并行优化时的优缺点,并对比不同并行化技术下的加速比和开销等性能,为更高效的排序算法的并行程序设计奠定基础。Sorting is an important operation of designing a computer program. Its performance usually determines the quality of the whole program. This paper carries out parallel optimization of five common sorting algorithms,including quick sort,bubble sort,merge sort,counting sort and select sort,based on the MPI（ Message Passing Interface） and Open MP（ Open Multi-Processing） parallel programming environment. The pros and cons using MPI and Open MP are analyzed,and the performances including speedup and overhead are compared. The purpose of this paper is to lay the foundation for designing more efficient parallel sorting programs.

关 键 词：计算机应用 高性能计算 排序 并行编程 加速比 MPI OPENMP 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构]