EBE-PCG算法在有限元并行计算中的应用研究  

Research on Application of EBE-PCG Algorithm in Parallel Computing of FEM

作　　者：陈荣征[1] 李代平[1] 黄健[1] 秦昭晖[1] 

机构地区：[1]广东工业大学计算机学院,广东广州510006

出　　处：《计算机技术与发展》2008年第3期232-235,共4页Computer Technology and Development

基　　金：广西自然科学基金资助项目(桂科自0229009)

摘　　要：目前,在研究有限元并行计算时,讨论并行算法理论和并行算法设计与分析的居多,研究并行算法的实现并解决实际问题较少。在Beowulf集群环境下,采用EBE策略设计出基于PVM平台的EBE-PCG算法,并通过一个电法勘探的典型工程算例对该算法在有限元计算中的性能进行了测试。实验结果表明,该算法加速比和并行效率均较为理想;在处理同等规模的问题时,同CG算法、PCG算法相比,具有并行度更高,耗时更少等优点。Currently, research on the parallel computing of finite element method, discussing parallel algorithm theory and parallel algorithm design and analysis are in the majority. However,on how to achieve and solve practical problems using the algorithm is less. In the Beovatlf cluster environment, using EBE strategy to design the F.BE- PCG algorithm based on PVM platform,and puts forward a typical use case in electrical surveying to test the performance of the algorithm in parallel computing of FEM. Experimental results show that the speedup and the parallel efficiency are both excellent. Compared with CA; algorithm and PCG algorithm in dealing with the .mrne scale problems, the EBE - PCG algorithm has the merits of high degree of parallelima and less running - time.

关 键 词：有限元法 BEOWULF集群 EBE-PCG 并行计算 并行虚拟机 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]