摘要
本文讨论了摹矩阵连乘积M_1M_2…M_q的并行计算问题,其中M_是n_(-1)×n_矩阵,证明了如果n_=min{N_0,…,n_q},则从右至左计算M_1…M_,从左至右计算M_(+1)…M_q,再将二者摹乘的计算方案是最优的。最优方案的并行计算量为 n_(sum from =1 to q-1 n_-n_+min(n_0,n_1}
In this paper, the problem of the efficient parallel computation of modulo-matrix chain products M_1M_2…M_q is considered. It points out that the policy is optimal if n_s= min {n_i} and the optimal cost is o<f<q n_(n_i-n_+min{n_o,n_q}).
出处
《武汉大学学报（自然科学版）》
CSCD
1989年第1期9-13,共5页
Journal of Wuhan University(Natural Science Edition)
关键词
摹矩阵
并行计算
嘉量原理
modulo-matrix, parallel computation, jar-matrix principle, outer product algorithm。
