Font Size: a A A

Researches On Particle Swarm Optimization And Its Applicatons In Low Energy Mapping Of NoC

Posted on:2015-12-07Degree:MasterType:Thesis
Country:ChinaCandidate:Q J YeFull Text:PDF
GTID:2308330464966748Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the high intergration of the system on chip(So C) based on bus, as a new communication framework, Network on chip(Noc) was born with a strage of globally asynchronous locally synchronous, it solve a series of problem brought by So C, such as, reliability, power consumption, scalability and so on. Currently, some optimizaton algorithm, such as, colony algorithm, genetic algorithm and so on, are utilized to minimize power consumption, and achieve good results, but they are still have some limitations, such as, complex operations, easy to fall into local optimum, lower power consumption et al. Therefore, in the condition of ensuring the speed and accuracy of the algorithm, the design of No C is always difficult and important for persuing the lower power and finding more efficient algorithm.For the low power optimization problems of No C in this paper, the low power mapping algorithm of No C based on the improved particle swarm algorithm(MPSO) was proposed according to the existing network communication model and the mapping objective function. The improvement of PSO are: good point set based on chaos disturbed is designed to initialize particle populations, which makes the particles distribute uniform and can travel the whole space, it also improves the quality of the initialized particles; the overall adaptive weight strategy is proposed, the dimensions of the weight are reduced due to the structure of good point set being irrelevant to its space dimensions, and the weight of particles are adaptively adjusted according to its global optimal position and historical optimal position, thus the efficiency of the algorithm is improved; the local search strategy of electromagnetism-like mechanism algorithm is introduced to make a local fine search around the optimal particle, it can avoid missing some particles around the optimal particle, the global and local are searched at the same time, so that the search efficiency and accuracy of the algorithm are improved; Cauchy mutation based on fuzzy theory is designed, the particles concentrated in α-cut set mutate with Cauchy in a probability way, it can increase the diversity of the population and ease the problem of easily falling into the local optimum of the PSO. The improved PSO algorithm is applied to solve the discrete problems by encoding mechanism,so MPSO algorithm is able to quickly and accurately determine the optimal mapping scheme of NoC low power mapping.Experiment results show that MPSO algorithm has lower communication traffic and power consumption, power can respectively save 18.23%, 14.81%, 9.19%,compared with the existing genetic algorithm, simulated annealing algorithm and Electromagnetism-like mapping algorithm.The mapping algorithm for low power of No C is researched in this paper, and the PSO algorithm is improved and then to be applied to solve the optimal mapping scheme, for the follow-up work, the modified PSO algorithm may be used for other performance indicators of No C optimization, or the study of multi-objective optimization mapping algorithm also can research.
Keywords/Search Tags:Network-on-Chip, low power, particle swarm algorithm, good point set, Cauchy mutation
PDF Full Text Request
Related items