矩阵相乘Cannon并行算法在工作站机群上的实现

Implementation of Matrix Multiple Cannon Parallel Algorithm based on Ouster of Workstations

矩阵相乘Cannon并行算法是一个基于分布式存储多处理机模型的并行数值算法,本文研究了它在工作站机群上的实现。在满足负载平衡和减少网络间数据传输的条件下,主要探讨了子任务在工作站上的优化分配策略,最后给出了在pvm并行编程环境下的具体实现方法。

Matrix multiple Cannon parallel algorithm is a kind of parallel algorithm based on the distributed memory and multiprocessor model. This paper sets out to an analysis of its implementation on the cluster of workstations. The purpose of this paper is to make a research into the optimized allocation policy of the sub-jobs on the workstations on the condition of load balance and a minimum of data transmission among workstations. Finally,a practical implementation under the PVM parallel programming environment is provided.