本文已被:浏览 979次 下载 465次
投稿时间:2007-03-20
投稿时间:2007-03-20
中文摘要: 在并行计算中,矩阵与矩阵相乘的Cache利用率最高,为了提高并行计算的速度,在进行向量与向量的运算时,可以先把向量补成矩阵,参加运算后,再把相应的部分分离出来.其中的关键问题是将原来的向量下标复位,以确保计算结果与原来的向量运算结果相同.
Abstract:In the parallel computing,the cache using ratio(CUR) of the multiplying of matrixes is the highest.To increase the parallel computational speed,when multiplying vector and vector,the vector can be transferred to matrix dealing.And when finished,it is restored to the former.The most important process is to restore the matrix's subscript to guarantee that the correct result is restored.
keywords: parallel computing communicator root process
文章编号:20070322 中图分类号: 文献标志码:
基金项目:
引用文本:
刘辉,欧广宇.并行处理中的矩阵复位的算法实现[J].上海电力大学学报,2007,23(3):281-283.
LIU Hui,OU Guang-yu.The Algorithm of Matrix's Subscript Restoring in Parallel Computing[J].Journal of Shanghai University of Electric Power,2007,23(3):281-283.
刘辉,欧广宇.并行处理中的矩阵复位的算法实现[J].上海电力大学学报,2007,23(3):281-283.
LIU Hui,OU Guang-yu.The Algorithm of Matrix's Subscript Restoring in Parallel Computing[J].Journal of Shanghai University of Electric Power,2007,23(3):281-283.