基于混合混沌大爆炸算法的三维片上网络低功耗映射

Low-power Mapping Method for Three-dimensional Network on Chip Based on Hybrid Chaotic Big Bang-big Crunch

三维片上网络(3D NoC)被认为是提高多核处理系统性能的一种方式。对于3D NoC的设计,如何将给定应用特征图(APCG)上的IP核适当地分配到3D NoC架构中是IP核映射的关键问题。一种优秀的映射算法及一次合理的映射可以大幅改善片上网络的通信功耗、发热、延时等指标。大爆炸算法(BB-BC)是一种新型的元启发式群体智能优化算法;混合混沌大爆炸(HCBB-BC)算法是在大爆炸算法基础上进行改进的一种算法,它具有参数简单、收敛速度快等优点。文中提出将混合混沌大爆炸算法用于解决三维片上网络映射问题,这是首次用大爆炸算法的相关算法来解决3D NoC映射问题。仿真实验结果证明,与现有的3D NoC映射算法相比,所提方法可以用更少的迭代次数和时间来找到更好的解决方案,同时有效地降低3D NoC的映射功耗。在经典任务图映射条件下,混合混沌大爆炸算法与遗传算法(GA)相比,收敛速度提高了36.73%,与粒子群算法(PSO)相比,收敛速度提高了22.45%;同时,混合混沌大爆炸算法的平均功耗比遗传算法的平均功耗的最大值低5.75%,并且比粒子群算法的平均功耗的最大值低3.90%。在随机任务图映射条件下,混合混沌大爆炸算法仍然能够保持稳定的功耗优化效率和更快的收敛速度。

Three-dimensional network on chip(3D NoC)is envisioned as a way to achieve high performance of multi-processor systems.For the design of 3D NoC,how to properly allocate IP cores on a given application feature map(APCG)to the 3D NoC architecture is the key problem of IP core mapping.An excellent mapping algorithm and a reasonable mapping can greatly improve the power consumption,heating,delay and other indicators of network-on-chip.The big bang-big crunch(BB-BC)algorithm is a new type of meta-heuristic swarm intelligence optimization algorithm.The hybrid chaotic big bang-big crunch(HCBB-BC)is a modified algorithm based on the big bang-big crunch(BB-BC)algorithm,which has simple parameters and fast convergence speed.In this paper,the hybrid chaotic big bang-big crunch(HCBB-BC)was proposed to solve the problem of 3D NoC mapping.This is the first time that big bang-big crunch(BB-BC)algorithm is used to solve the 3D NoC mapping problem.Simulations are conducted to prove that the proposed method requirs less number of iterations and time to find a better solution and can reduce energy consumed compared with existing NoC mapping algorithms.Under the condition of the classical task graphs,compared with the genetic algorithm(GA)algorithm,the convergence speed of the hybrid chaotic big bang-big crunch(HCBB-BC)is increased by 36.73%,and the 22.45%improvement is witnessed compared with the particle swarm optimization(PSO)algorithm.The average power consumption of the hybrid chaotic big bang-big crunch(HCBB-BC)mapping is lower than that of GA with the maximum as 5.75%,and lower than that of PSO with the maximum as 3.90%.Under the condition of the random tasks,the hybrid chaotic big bang-big crunch(HCBB-BC)algorithm can still maintain stable optimization efficiency and higher convergence speed.