三维片上网络映射算法研究综述

Survey on Mapping Algorithm of Three-dimensional Network on Chip

三维片上网络(Three-dimensional Network on Chip,3D No C)以其更短的全局互连、更高的封装密度、更小的体积等诸多优势,已成为国内外工业界和学术界一个重要的研究领域.对于3D No C的研究,映射是一个关键问题,对系统的功耗、延迟等性能均有很大的影响.本文介绍了3D No C映射的基本问题;归纳了3D No C映射算法的分类方法,将其分为启发式映射算法和非启发式映射算法两大类,其中,启发式映射算法又分为基于遗传算法、基于粒子群算法、基于模拟退火算法以及基于蚁群算法的映射算法4种类型,非启发式算法具有多样性;另外,对现有的3D No C映射算法进行了分类综述,分析比较了各种映射算法的特点;最后,探讨了3D No C映射的研究方向.

Three-dimensionalnetworkonchip （3D NoC） with its shorter global interconnections, higher packing density, smaller vol- ume, and many other advantages, has drawn more and more attention from both industry and academia. Mapping is a key problem on 3D NoC research and has a great impact on power consumption, latency and many other system performances. In this paper, some essential issues on 3D NoCmapping are introduced and many 3D NoC mapping algorithms are studied and classified into heuristic mapping algorithms and non-heuristic mapping algorithms, in which heuristic mapping algorithms include mapping algorithms based on genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm and ant colony algorithm. Existing mapping algorithms for 3D NoC are then summarized with the characteristics of each algorithm outlined. The possible hot research areas in fu- ture on3D NoCmapping algorithms arealso discussed in the last part.