摘要
研究了DFP算法在消息传递型MIMD并行机上的并行实现问题,首先讨论并行一维寻优问题,然后利用矩阵和向量分解,并充分考虑到Transputer并行机的特点,给出了一个负载平衡度高,通讯量小的并行化方案。理论分析和实际算例表明,当问题的规模充分大时,其并行效率趋于1。
The parallel implementation of DFP algorithm on message passing type MIMD computer systems is discussed, and a parallel scheme of DFP algorithm with high load balance and small communication volume is proposed. The theoretical analysis and practical examples of computation indicate that, as the scale of the problem to be solved is large enough, the efficiency of the parallel DFP algorithm approaches to unity.
出处
《南京航空航天大学学报》
CAS
CSCD
1995年第2期266-269,共4页
Journal of Nanjing University of Aeronautics & Astronautics
关键词
最优化算法
数据传输
通信
并行算法
optimization algorithms
information transmission
MIMD
communication