摘要
§1.前言 并行计算是近十几年来随着并行计算机发展而发展起来的一门新兴学科,特别是对于多指令流多数据流(MIMD)的并行计算机,由于它是由多台普通计算机甚至是向量计算机相互以一定方式联结起来的新型计算机系统。
In this paper the parallel direct method on multiprocessors with local memoryis considered for finite element structure analysis. In §2 parallel computing of anelement stiffness matrix and an assembly of the global stiffness matrix is descrbed.In §3 two direct algorithms are given for banded linear systems with variable bandwidth in which the coefficients are stored by row-wise interleaving storage or column-wise interleaving storage and data exchange bctween processors by pipelinetransmition. In §4 an upper-bound of the number of processors related the halfbandwidth of the matrix and corresponding optimal parallel efficiency are givenFinally, a numerical example on hypercube architecture computer ipsc-2 is provided.
出处
《计算数学》
CSCD
北大核心
1991年第4期433-438,共6页
Mathematica Numerica Sinica
基金
国家自然科学基金
