摘要
处理单元映射是NoC设计的重要环节,其实施结果对NoC通信功耗及QoS约束的满足均起着决定性的作用。本文选取的NoC拓扑结构为星型子网的片上网路结构SSBM NoC(Star-Subnet-Based-Mesh NoC)。为保证系统实时性和低能耗,文中建立了延时约束下功耗最小的NoC映射模型,提出了一种基于改进粒子群算法的优化方法。针对NoC映射的特点,在粒子群算法中引入了交换子和交换序算子,确保了映射的有效性。实验结果表明,文中算法在满足延时约束下,降低NoC通讯功耗20%以上,证明该算法优化效果显著。
The core mapping is the important part of NoC design.It plays a decisive role in the communication energy and QoS constraints.This paper selects Star-Subnet-Based-Mesh NoC（SSBM NoC） topologic.For ensuring the system real time property and low power consumption,this paper models NoC mapping building which considering delay-constrained and the total communication energy.A core mapping method based on Particle Swarm Optimization（PSO）is presented.In the consideration of NoC mapping property,we induct the operator of recon and recon sequence into PSO,which ensure the mapping results valid.The experimental results show that this method can adapt the NoC to satisfy the time-constrained and reduce the total communication energy more than 20%,proofing that this method optimizes well.
出处
《国外电子测量技术》
2010年第4期28-31,共4页
Foreign Electronic Measurement Technology
关键词
星型子网
映射
功耗
粒子群算法
交换子和交换序
Star-Subnet-Based-Mesh NoC（SSBM NoC）
mapping
energy
particle swarm optimization
recon and recon sequence
