基于Julia语言的并行计算方法初探.pdfVIP

基于Julia 语言的并行计算方法初探 1,2 1+ 1 1 巩庆奎 , 张常有 , 张先轶 , 张云泉 1( 中国科学院软件研究所 并行软件与计算科学实验室,北京 100190) 2(石家庄铁道大学 信息科学技术学院,石家庄 河北,050043) Primary investigation into Parallel Computing in Julia Language 1,2 1+ 1 1 GONG Qing-kui , ZHANG Chang-you , ZHANG Xian-yi , ZHANG Yun-quan 1(State Key Laboratory of Computing Science, the Chinese Academy of Sciences, Beijing 100190, China) 2(School of Information Science and Technology, Shijiazhuang Tiedao University, Hebei 050043, China) + Corresponding author: Phn: +86- 10 Fax: +86- 10 E-mail: changyou@ Abstract : Julia language is a free developing scripting language under MIT under the MIT license. Its goal is to ease the difficulty of parallel programming. Based on the language mechanisms of Julia, we construct a use case of computing the average running-time between every two bus stops. And then, we examples the Julia programming framework and the code refining steps. Julia language supports both multi-cores/CPUs parallel programming modes. To full use all the computing resources, we developed some experiments on new policies about how to higher the computing performance. Experiments show that managing processors in parallel computing model consume working time, but with the increasing of problem size, this impact can be gradually ignored, and we gained nearly linear speedup. Key words : Julia language; computing scale; auto-adapting; performance analysis 摘 要: Julia 语言是一种在MIT 许可证下免费的开发中脚本语言（beta 0.2.0 ），目标降低并行程序的编程难 度。本文将基于Julia 现有语法机制，逐步深入Julia 语法特性，结合公交线路的平均走行时间统计案例，研 究Julia 并行编程框架和程序逐步精化的方法。Julia 程序支持本地多核心/ 多CPU 并行计算，为充分发挥实验 平台的计算潜能，本文尝试了提高 Julia 程序计算性能的策略。对案例程序的实验分析表明，Julia 并行程序 在管理计算核心方面耗费一定的工作时间，但随着问题规模的增大，其影响将渐可忽略，获得了接近线性的 加速比。 关键词: Julia 语言；脚本；并行程序；程序精化；性能分析 中图法分类号: 文献标识码: A Julia 是一种正在开发中的动态高性能的高级编程语言，适合科学和数值计算，性能可与传统静