The paper addresses two closely related problems, namely themapping and scheduling tasks of a parallel program in parallelcomputers. To develop new efficient algorithms to solve theaforementioned problems we propose a theoretical framework based on amodel of N-person games which serves as a model of interaction of agentsin a multi-agent system. We implement the multi-agent system with use ofevolutionary techniques and apply it to develop parallel and distributedalgorithms of dynamic mapping and scheduling
展开▼