PVM下矩阵相乘并行算法的研究与实现

在许多实际计算机科学、数学、工程等问题的计算中，经常遇到一些大型的高阶矩阵的有关计算，尤其是两矩阵相乘最为常见，当矩阵阶数较高时，通常的计算过程需要占用较多的工作单元和较大的计算机内存，计算效率受到影响。文中研究了一种矩阵要乘的并行算法，并根据其时间复杂度，进行了一定的改进，并在PVM环境下实现了该改进算法。此算法减少了所需要的处理要的数量和在一台处理机上分配的子任务数。通过对时间复杂度的分析可知，此改进算法减少了进程间的选择性传输所带来的过重通信开销，提高了程序的运行效率。

Implementation of Matrix Multiple Parallel Algorithm on PVM

In many of domain,such as computer science,mathematics,the project,often meet some large-scale high-order matrix calculations,especially two matrix multiple.When matrix steps numbers are relatively high,the common computational process needs to take up more working cells and relatively big memory of computer,and computational efficiency is influenced.This paper reseaches a parallel algorithm of matrix multiple,and analyses its time complexity.According to its time complexity,carried on some improvement,and realized this algorithm under the enviroment of PVM.This algorithm reduces the quantities of processor needed and task assigned at a processor.According to the analysis to the complexity of time,this algorithm reduces the overweight communication expenses brought in alternative transmission among the process,and have improved the operational efficiency of the procedure.