Stochastic Cellular Automata (CA) form a subclass of asynchronous cellular automata models of nonlinear spatial dynamics. Functioning of a stochastic CA is represented by a
random sequence of simple actions (given as local operators) in randomly chosen cells of a discrete space. Behavioral properties of stochastic CA evolutions exhibit tendency to stable state, stable oscillations, autowaves or chaos, which shows a wide range of their application. In the report most attention is made on computational characteristics osf stochastic CA and their dependencies on the mode of local operator application, which may be modified by means of introducing partial synchronization. These modifications are important for achieving acceptable values of parallelization of large scale CA-models.
By means of performing a number of simulation “reaction-diffusion” processes on supercomputers parallelization efficiency estimations have been obtained for different modifications of stochastic mode of opration. It is found out that block-synchronous transformation results in superefficiency due to reduced amount of random generator calls, as well as reduces of “modN” operations when periodic boundary conditions are used.