改进混合蛙跳优化的片上网络低功耗映射算法

Improved shuffled frog-leaping algorithm for low-power network-on-chip mapping

随着片上网络集成度的提高,低功耗映射逐渐成为研究的热点.文中利用调整序思想,将混合蛙跳算法应用到片上网络低功耗映射问题中;同时对混合蛙跳算法进行改进,引入自适应混沌方法优化较差个体,采用多邻域退火策略优化较优个体.实验结果表明,相比于传统混合蛙跳算法和模拟退火混合蛙跳算法,改进混合蛙跳算法具有更优的搜索能力,从而能进一步降低通信功耗.

With the density of network-on-chip（NoC） integration getting higher, the low-power mapping has become a hotspot research. A novel shuffled frog-leaping algorithm（SFLA） is used for solving the NoC mapping problem based on the concept of adjustment sequence. To further enhance searching capability, the proposed SFLA is improved with the adaptive chaos tactic and strategy of multi-neighborhood annealing. Experimental results shows that the improved SFLA has the advantage over the SFLA and annealing SFLA of better optimizing performance, thus reducing the communication power further.