基于二维网孔互连网络的并行矩阵乘算法

Parallel Algorithms for Matrix Multiplication on the 2-D Mesh Interconnection Networks

矩阵乘法是数值分析领域中一种十分常用的基本运算,被广泛应用于模式识别、图像和信号处理。由于矩阵运算具有局部性、一致性的特点,特别适合用二维网孔并行计算机来实现。文章讨论了基于二维网孔互连网络的矩阵乘并行算法的实现,首先给出了一种正方网孔处理机阵列的并行算法,然后将其推广到长方网孔处理机阵列中。最后通过在LSMPP计算机的应用,证明算法是可行的、有效的。

Matrix multiplication is one of the most important basic operations in numerical analysis, which is widely used in pattern recognition, image and signal processing. Having the advantages of locality and uniformity, matrix operation is especially suitable for 2-D mesh interconnection network to process. In this paper, we discuss the realization of parallel algorithm for matrix multiplication based on 2-D mesh interconnection network. Firstly, we propose a parallel algorithm on the square mesh array, then develop it on the rectangle mesh array. Through the application on LS MPP, the proposed algorithms are approved to be applicable and efficient.