The Effect of Speculative Computation on Combinatorial Optimization Problems

In recent years, multicore or many-core processors have gained significant attention as they enable computation with a large degree of parallelism on desktop computers. However, conventional parallel processing methods often cannot easily achieve parallel effects due to various factors. In this study, we evaluated the effect of applying MultiStartbased speculative parallel computation to combinatorial optimization problems. Using probability theory, we performed theoretical verification to determine whether speculative computation is more effective than conventional parallel computation methods. In addition, we conducted experiments and compared the result with those of conventional parallel processing. In this paper, we report the results of the theoretical verification and experiments, and we show that speculative computation is more effective than conventional parallel processing.