Adaptation of stochastic search methods realized on a homogeneous computational array under uncertainty (Q1069261)

The development of computer systems (multiprocessor systems) is considered. The effectiveness of realization of the Monte Carlo method on a homogeneous computational array with a common memory field is studied. The computational array has m processor elements (PE) and v control devices (CD) each of which can control all the PE. The PE that are controlled by the same CD form a branch. A priori the optimal allocation of the PE to the branches is not known. During the search the branches can accumulate information about the problem to be solved and then reallocate the PE in such a way that the time of search for a solution is minimized. Four algorithms are presented to reallocate the PE. These algorithms are based on a theorem a proof of which is given in the appendix. A numerical example is given to find optimal allocation for 16 PE to 4 branches at the j-th iteration for given numbers of trials.