Discovery of parallel scheduling algorithms in cellular automata - based systems

Abstract

In this paper we report new results concerning developing parallel multiprocessor scheduling algorithms working in cellular automata (CAs) -based scheduler. We consider the simplest case when a multiprocessor system is limited to two processors, but we not imply any limitations on a size and parameters of parallel programs. We use an approach called a selected neighborhood to design a nonlinear structure of CAs for a given program graph. We propose to use coevolutionary genetic algorithm (GA) to discover rules of parallel CAs, suitable for solving the scheduling problem. We compare sequential and parallel scheduling algorithms discovered in the context of CAs -based scheduling system.