Uma estratégia baseada na metaheurística VNS para encontrar efetivas sequências de otimizações

Abstract

The Optimization Selection Problem aims to find the best optimizations to use in a specific source code. The exhaustive search is an impracticable approach to solve this problem because are length of the search space. This article aims to mitigate the Optimization Selection Problem with an application of the Variable Neighborhood Search metaheuristic. The achieved results by the proposed algorithm, for the SPEC CPU2006 benchmark, indicate that it has potential to find good solutions and outperforms the algorithm proposed by Pan e Eigenmann, Combined Elimination.