摘要
海洋数值预报技术的发展与高性能计算密切相关。为提高OVALS海洋资料同化系统的时效性,本文实现了OVALS系统的并行化。在温盐资料同化模块并行化过程中,本文提出了层优先处理器划分算法,并研究了基于该算法的并行I/O、全局通讯等实现方法;在高度计资料同化模块并行化过程中,设计实现了基于预处理的非规则区域分解算法,较好地实现了OVALS并行计算负载平衡。数值实验表明,OVALS并行系统在36并行规模下取得了17.45的并行加速比。
High performance computing is closely related to the development of numerical oceanic prediction. In order to enhance the in-time property of the OVALS, this paper focuses on the parallelism of the OVALS system. In the process of parallelizing the ASSIMI part, the Layer-First-Processor-Allocation algorithm is proposed and the parallel I/O and global communication operations are implemented based on the LFPA algorithm. In the process of parallelizing the ASSssh part, an irregular domain decomposition algorithm based on the pre-condition is designed and implemented, which makes the computation load balancing of the parallel OVALS system. The numerical experiment shows that the parallel OVALS system gains a 17. 45 speedup when running with 36 processes.
出处
《计算机工程与科学》
CSCD
北大核心
2010年第1期113-116,共4页
Computer Engineering & Science
关键词
OVALS海洋资料同化
并行算法
处理器划分
非规则区域分解
OVALS
ocean data assimilation
parallel algorithm
processor allocation
irregular domain decomposition
