GPU Sparse Matrix Multiplication with CUDA
College of Arts and Sciences, Department of Computer Science Research, Florida State University
Florida State University, 2013
@article{rose2013gpu,
title={GPU Sparse Matrix Multiplication with CUDA},
author={Rose, Sean},
year={2013}
}
Matrix multiplication is a commonly-used mathematical operation that has many practical applications. It is used to solve a number of problems in a wide variety of fields including science, engineering, and computer science. Given two matrices, A and B, and a resultant matrix C. The concept of density is used to describe the number of nonzero elements in a matrix relative to the total number of elements. For an NxM matrix with Z nonzero elements, the density is defined as Z=(NxM). A sparse matrix is one which has a low density. Sparse matrices can be stored in special formats to eliminate the need for the zero elements to be stored. The storage format and potentially large matrix size presents a challenge when designing an efficient sparse matrix multiplication algorithm.
May 9, 2013 by hgpu