Design and Analysis of Parallel Matrix Multiplication
Ei Ei Myint, Lai Lai Win Kyi · 2014
Problem in science and engineering are often large, complex and highly irregular. Many type of problem can be converted into matrix form. Multiplication of large matrices requires a lot of time as its complexity is O(n3). Because most current applications require higher computational through puts with minimum time, parallel algorithm is developed. If matrix multiplication is done in the efficient way, many problems can be solved easily and efficiently. As a result, this paper tends to implement parallel matrix multiplication in Java threads. The application distributes the products of rows and columns on different servers. One server and four clients are run to find the product of matrix multiplication. The server distributes the determine blocks of rows and columns on the registered clients. The clients return their product blocks to a server, which calculate the final product of matrix multiplication. In this paper, a theoretical analysis for the performance and evaluation of the parallel matrix multiplication algorithms is carried out. And an experimental analysis is performed to support the theoretical analysis results by calculating 100×100 to 500×500 matrices.