Usage of Discrete Particle Swarm Optimization Method for the Searching of Optimal Tile Size

Abstract

The paper considers actual methods of the software optimizations. Main attention is directed to the tiling method and parallelization. Showing the opportunities of the tiling method authors accent that tile size of the tiling method has critical influence of the optimization level. Main part of the paper dedicated to the searching of the optimal tile sizes for the better execution time of the test program. As a conclusion authors demonstrate results of the reducing execution time by applying of the particle swarm optimization method for the searching of the optimal tile sizes. The obtained results reveal that particle swarm optimization method can be used for the optimal tile size search. In the same time, parameters of the particle swarm optimization method should be picked up carefully for the faster method search.