Design and implementation of GPU-based matrix chain multiplication using C++AMP

Matrix Chain Multiplication is one of the optimization problems widely used in graph algorithms, signal processing and network industry. The Matrix Chain multiplication is the process of multiplying number of nxn matrices cumulatively, where n is the size of the matrix. In this paper, the focus of the work is to accelerate the computation time for matrix chain multiplication with the help of high performance C++ Accelerated Massive Parallelism (AMP) code for Graphic Processing Units (GPUs). The proposed work is implemented in C++ AMP and tested with matrix sizes varying from 100 to 1000 for chain multiplication of three and four matrices. The experimental results show that, the speed up achieved through GPU implementation is 97.27% when compared with Sequential implementation of the matrix chain multiplication.