GPU usage for solving Data Intensive applications

Abstract

 Development of parallel algorithms for solving the problems of the class Data Intensive is important, because the number of exchanges with the RAM compared to the computational load of these tasks is sufficiently large, and the computational power of processors is increasing faster than the speed of the RAM. Proposed two parallel algorithms for solving problems of breadth-first search and the search of the shortest paths in undirected graphs. Also this article describes the various optimizations to work effectively with the memory of the graphics accelerator.