Solving large sparse systems of linear equations on GPU
This thesis focuses on solving large sparse linear systems which formulate problems, defined on a two dimensional grid, described by elliptic partial differential equations, the general Laplace-Beltrami equation and its subsets, Poisson and Laplace equations. The iterative methods for solving sparse linear systems as well as applications from the field of image processing, where such systems are solved, are presented. The thesis also provides an explanation of basic principles of GPGPU programming in CUDA framework. The CUDA implementation of the solver used for interactive image segmentation is described and the results are compared with the output from MATLAB.
Dean's award for outstanding master thesis