摘要
针对基于PVM的微机网络并行计算环境下，处理机的运算速度较快而处理机间的通信相对较慢的实际情况，给出了一种网上并行求解线性方程组的Guass—Seidel迭代算法。该算法将方程组的增广矩阵按行卷帘方式分布存储在各处理机中，循环传送每一次的迭代向量以减少处理间的通信次数，同时，采用计算与通信部分重叠技术，提高并行算法的效率。并用1—12台桌面PC机联成的局域网，在PVM3．4 on Windowsi2000，VC6．0并行计算平台上编程对该算法进行了数值试验，试验结果表明，该算法较传统的基于列扫描法的Guass—Seidel并行迭代算法优越。
With the consideration of the speciality of network parallel computing environment based on PVM system that the speed of the operation of processors is high while that of the communication among processors is relatively low, a parallel Guass - Seidel iterative algorithm for solving linear equations on LAN is presented in this paper, in which the data are distributed among local memories of all procesthrough row toms - wrap mapping, and the time of communication is decreased by loop - conveying the iterative vector. Meanwhile, this algorithm adopts the technology of overlapping computation and communication to improve the efficiency of the parallel algorithm. The results of numerical experiments, performed on LAN made up of 1-12 personal computers which are installed Windows2000 operation system, VC6.0 and PVM3.4, show that the algorithm presented in this paper is superior to the GuassSeidel parallel iterative algorithm based on column-scanning.
出处
《贵州师范大学学报（自然科学版）》
CAS
2006年第1期76-80,共5页
Journal of Guizhou Normal University：Natural Sciences
