基于COW系统的并行求和算法研究  

Research on Parallel Summary Algorithm Based on COW

作　　者：洪雄[1] 彭雷[1] 戴光明[1] 

机构地区：[1]中国地质大学计算机科学与技术系,湖北武汉430074

出　　处：《微机发展》2005年第5期111-113,共3页Microcomputer Development

摘　　要：针对并行求和问题,在Linux环境下组建了一个COW系统,并介绍了在COW系统上的并行求和算法的设计和实现,分析了影响算法性能的因素并提出改进方法,最后给出了并行求和算法在改进前后的实验数据。实验结果表明,当n=2000000时,用改进的算法在两台联想启天4000上测得并行求和效率为80%,而算法改进前在相同的条件下测得并行效率为66.7%,并行效率明显提高。COW is built based on Linux to solve the parallel summary.The design and implementation of the parallel summary algorithm is introduced in this paper.It also analyses the complications influenced performance of the algorithm and advances an improving method.Finally it also shows the result of the algorithm tested with two computers (LEGEND QiTian 4000).When n is up to 2 000 000,can get the parallel efficiency 66.7% before improving algorithm,and at the same condition,can get the parallel efficiency 80% after improving algorithm.

关 键 词：工作站集群 并行求和 并行效率 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]