Algoritmus A* Search implementovaný pomocí technologie CUDA

Abstract

This thesis presents a speed comparison of the well know A* Search algorithm implemented on processor and graphics card. The goal is to find the ideal thread division where each thread searches through a common graph created by a triangulation. The work compares several ways of connecting randomly generated points on a plane together and shows why it is best to use Delaunay triangulation. The results are evaluated by graphical representation generated by OpenCV libraries. It is presumed that the final programs for computing algorithms, either on the processor or CUDA, are very similar. Every search only computes the A* algorithm and does not take into account any other auxiliary procedures.