并行计算在胶州湾水质预报系统中的应用研究

发布时间：2018-03-04 07:13

本文选题：水质预报　切入点：并行计算　出处：《河北工程大学》2012年硕士论文　论文类型：学位论文

【摘要】：海洋水质预报的核心任务是快速准确地预测未来海水环境状况。随着业务海洋学的发展,水质预报系统所考虑的过程越趋完善,计算量也越来越大。基于单处理器的串行预报程序已渐不能胜任海洋水质预报的任务。为此,将串行程序改为并行程序,在高性能集群计算机系统上运行,已是大势所趋。本文利用消息传递接口(Message Passing Interface,简称MPI)技术对胶州湾水质预报系统中串行的生态模式进行了并行化处理,主要工作如下： 1.阐述了胶州湾水质预报系统结构,主要针对水质模式部分的设计模型以及代码进行了详细的分析,在此基础上,使用并行模式子区域划分的方法。 2.分别采用聚合通信技术和非阻塞通信技术对原串行程序进行了并行化处理,并编制相应代码。从代码实现的角度来看,聚合通信方式编程简单易懂,但不能达到通信与计算时间同步；而非阻塞通信方式能够最大限度的达到通信与计算相同步,但代码编写相对要复杂。 3.对系统的并行化程序进行了效能测试与评估。采用聚合通信方式,系统的运行时间比串行程序的运行时间可缩短29%,且并行效率最大可达55%；采用非阻塞通信方式,系统的的运行时间可缩短35%,且并行效率最大可达88%。所以,非阻塞通信方式在对系统的并行化改进中是很有优势的。胶州湾水质预报系统采用非阻塞通信方式对其进行并行化处理,大大的缩短了运行时间,有效的改善了水质预报系统的预报能力。
[Abstract]:The core task of marine water quality prediction is to predict the future marine environment quickly and accurately. With the development of operational oceanography, the process of water quality forecasting system is becoming more and more perfect. The serial prediction program based on single processor is no longer competent for ocean water quality prediction. Therefore, it is the trend to change the serial program into a parallel program and run on a high performance cluster computer system. In this paper, the serial ecological model in Jiaozhou Bay water quality prediction system is parallelized with message Passing Interface (MPI) technology. The main work is as follows:. 1. The structure of water quality prediction system in Jiaozhou Bay is expounded. The design model and code of water quality model are analyzed in detail. On the basis of this, the method of sub-region partition of parallel model is used. 2. The parallel processing of the original serial program is carried out by using the aggregate communication technology and the non-blocking communication technology respectively, and the corresponding codes are compiled. From the point of view of the code realization, the programming of the aggregate communication mode is simple and easy to understand. But it can not achieve the synchronization of communication and computing time, and the non-blocking communication mode can maximize the synchronization of communication and calculation, but the code writing is relatively complex. 3. The efficiency test and evaluation of the parallelization program of the system are carried out. The running time of the system can be shortened by 29% than that of the serial program, and the maximum parallel efficiency can reach 55%, and the non-blocking communication mode is adopted. The running time of the system can be shortened by 35%, and the parallel efficiency can be up to 88%. Therefore, the non-blocking communication mode has the advantage in the improvement of the system parallelization. The water quality prediction system of Jiaozhou Bay adopts non-blocking communication to process it in parallel, which greatly shortens the running time and effectively improves the forecasting ability of water quality forecasting system.
【学位授予单位】：河北工程大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：X834;TP338.6

【参考文献】