Research On Shuflfed Frog Leaping Algorithm And Its Application To Noc Low-Power Mapping

With the continuous development of the semiconductor technology, the existingSoC(System on Chip) has corresponding bottlenecks in different ways. NoC(Networkon Chip) can solve these problems fundamentally, thus it has gradually become a hotspot in this research area. Based on nanotechnology, the scale of NoC circuit is veryhuge. Thus, under corresponding design constraints, reducing the power consumptionof by reasonable mapping algorithms is a key problem.Firstly, the fundamental theory of is studied in this paper and the commonmapping algorithms are analyzed. Secondly, in order to eliminate the shortcomings oforiginal shuffled frog leaping algorithm (SFLA) such as non-uniform initial population,slow convergence rate and low precision, a new improved chaos annealing shuffledfrog leaping algorithm(CASFLA) is proposed. Utilizing its ergodicity, the idea ofchaos is used here to obtain the initial population of new algorithm, which makes thedistribution of initial individuals more even and ergodic. After local search the worseindividuals are optimized by chaos optimization algorithm. By this approach the worseindividuals can be updated effectively in earlier iterations and have a probability toescape from local optimum in later iterations, thus accelerating the convergence speedand avoiding premature convergence. At the same time the better individuals areoptimized by the strategy of multi-neighborhood annealing, thus improving the searchprecision. Finally, the corresponding encoding rules and local search strategy aredesigned based on the idea of adjustment sequence, which makes suitablefor low-power mapping problem.At last, algorithm, PSO algorithm, GA algorithm andalgorithm are applied to solve VOPD problem respectively. The experimental resultsshow that compared with the above three algorithms the average communication valueobtained by algorithm is lower by8.832%,8.834%and14.833%, thus caneffectively reduce the power consumption of.In the follow-up research and work, it is necessary to continue the study ofmapping based on diverse topology and multiple objectives, thus enhancing thealgorithm’s practicability.