Designação de tarefas em aplicações de multiprocessadores de processamento digital de sinal utilizando algoritmos genéticos
Abstract
This work consists in the development of genetic algorithms for the Task-to-Processo Assignment Problem in multiprocessor applications. Specifically, the objective is to find the task-to-processor assignment that minimizes the total delay in a particular multiprocessor digital signal processing architecture. We present a description of our algorithm implementations and the results found with a set of 117 randomly generated and real-life instances. The algorithms performance is compared with the results provided by a competitive dynamic list heuristic and a multiple start search algorithm. The results indicate lower delays in more than 68% of the instances, at a higher computational cost.