Optimizing exact computation of Betweenness Centrality for CUDA