摘要
在过程控制、图像处理等应用领域中需要用到大量的矩阵乘法操作,并且矩阵乘法计算性能是系统性能的关键因素。本文根据矩阵相乘的特点,提出了带状划分的矩阵相乘的并行算法。同时着重分析和推导了当并行机的处理器个数小于和远小于矩阵规模的情况下实现快速的矩阵乘法。
Many application areas, such as process control and image processing, make extensive use of matrix multiplication operations, and the computational performance of this operation is critical for the whole system. A Striped Partitioning parallel algorithm for matrix multiplication is proposed according to the characteristic of matrix multiplication. At the sametime a systolic method for matrix multiplication is resulted by the analysis,when the processors number of the parallel computers less than the scale of Matrix or far less than the scale of Matrix,how to realizing fast matrix multiplication.
出处
《武汉科技学院学报》
2008年第11期4-6,共3页
Journal of Wuhan Institute of Science and Technology
关键词
并行计算
带状划分
矩阵乘法
parallel algorithm
striped partitioning matrix multiplicatio
