Mixed MPI-CUDA implementation of full PB equation