基于可压缩格子Boltzmann方法的高可扩展并行算法研究  

Research on High Scalable Parallel Algorithm Based on Compressible Lattice Boltzmann Methods

作　　者：何冰[1] 封卫兵[1] 张武[1] 武频[1] 白文[2] 李立[2] 

机构地区：[1]上海大学计算机工程与科学学院,上海200072 [2]中国航空计算技术研究所,陕西西安710069

出　　处：《航空计算技术》2008年第5期65-68,72,共5页Aeronautical Computing Technique

基　　金：航空科学基金(20061431)

摘　　要：Lattice Boltzmann Methods(LBM)是近年来发展的求解流体问题的计算新方法,该方法具有编程相对简单,并行计算效率高的特点,但是现有的D2Q9 LBM模型只能计算速度在0.3马赫下的不可压缩流体,提出的新的LBM模型可以处理速度0.7马赫以下的流体问题,并且具有较好的数值稳定性,对计算程序并行性能深入研究的基础上,提出了基于cache的性能优化,经过程序性能测试证明该方法具有较好并行计算效率,并具有很好的可扩展性。Lattice Boltzmann Method（LBM） is developed recently to solve complex fluid dynamics problems. The advantages of LBM have been addressed as simplicity of programming, and high parallel efficiency. The traditional lattice Boltzmann model has a constraint of small Maeh number（ thevelocity of fluid must be less than O. 3 mach）. This paper presents a novel lattice Boltzmann method to simulate compressible flows. The new model is capable to deal with the velocity of flows reach to 0.7 Mach. The numerical experiments show that our algorithm is more numerical stable than traditional LBM methods. It proposes cache- based performance optimization of the computing program after deep research on the parallel capability. The parallel performance of this algorithm on workstation clusters demonstrates high parallel computing efficiency and good scalability.

关 键 词：并行计算 格子波尔兹曼方法 空腔流 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP391.77[自动化与计算机技术—计算机科学与技术]