Towards efficient GPGPU Cellular Automata model implementation using persistent active cells