多核多线程并行求解线性方程组  

Parallel solving linear equations using multi-core computing and multiple threads

作　　者：冯佩[1] 钟诚[1] 韦伟[1] 

机构地区：[1]广西大学计算机与电子信息学院,广西南宁530004

出　　处：《合肥工业大学学报（自然科学版）》2011年第2期237-240,250,共5页Journal of Hefei University of Technology：Natural Science

基　　金：高校优秀人才资助计划项目(RC2007004);广西研究生教育创新计划资助项目(105931003036);广西大学拔尖创新团队建设计划资助项目

摘　　要：线性方程组求解在科学与工程计算领域具有广泛的应用。文章依据多核计算机共享二级缓存和私有一级缓存的容量,采取将线性方程组的增广矩阵按行划分并合理地分布存储到各级缓存中,各个处理核以多线程方式并行计算矩阵行的方法,给出了一种在多核计算机上实现的线程级并行求解n阶线性方程组的算法。实验结果表明,与原Gauss-Seidel并行算法相比,文中所提出的算法具有较好的加速比和可扩展性。Solving linear equations is common in the filed of scientific and engineering computation.According to the size of shared L2Cache and private L1Cache on the multi-core computers,a thread-level parallel algorithm for solving n linear equations on multi-core computers is designed.The algorithm divides the augmented matrix of equations into several rows which are distributed and stored in the caches on the all processing cores.All the processing cores compute the matrix rows in parallel by using multi-threading technology.The experimental results on a multi-core computer show that,compared with the Gauss-Seidel parallel algorithm,the presented thread-level parallel algorithm for solving n linear equations has good speedup ration and scalability.

关 键 词：线性方程组求解 Gauss-Seidel算法 多核计算 线程级并行 多级缓存 

分 类 号：TP338.6[自动化与计算机技术—计算机系统结构]