Journal of Computer Science and Technology


Path-based Multicast Routing for Network-on-Chip of the Neuromorphic Processor

Zi-Yang Kang (康子扬), Shi-Ming Li (李石明), Shi-Ying Wang (王世英), Lian-Hua Qu (曲连华), Rui Gong (龚锐), Wei Shi (石伟), Wei-Xia Xu (徐炜遐), and Lei Wang (王蕾), Member, CCF, ACM   

  1. College of Computer Science and Technology, National University of Defense Technology, Changsha 410000, China

Network-on-Chip (NoC) is widely adopted in neuromorphic processors to support communication between neurons in spiking neural networks (SNNs). However, SNNs generate enormous spiking packets due to the one-to-many traffic pattern. The spiking packets may cause communication pressure on NoC. We propose a path-based multicast routing to alleviate the pressure. Firstly, all destination nodes of each source node on NoC are divided into several clusters. Secondly, multicast paths in the clusters are created based on Hamiltonian path algorithm. The proposed routing can reduce the length of path and balance the communication load of each router. Lastly, we also design a lightweight microarchitecture of NoC, which involves a customized multicast packet and routing function. We use six datasets to verify the proposed multicast routing. Compared with the unicast routing, the running time of path-based multicast routing achieves 5.1x speedup, the number of hops and the maximum number of transmission latency of path-based multicast routing is reduced by 68.9% and 77.4%, respectively. The maximum length of path is reduced by 68.3% and 67.2% compared with Dualpath (DP) or Multi-path (MP) multicast routing, respectively. Therefore, the proposed multicast routing has improved performance in terms of average latency and throughput compared with DP or MP multicast routing.


1. 研究背景:(context)
类脑处理器中广泛采用片上网络(NoC)来支持脉冲神经网络 (SNN) 中神经元之间的通信。然而,由于SNN一对多的流量模式会产生大量的脉冲数据包,从而对 NoC 造成通信压力。
2. 研究目的:(Objective)
3. 方法:(Method)
首先,将每个源节点的目的节点根据其在NoC上的坐标划分为若干簇。其次,采用基于哈密顿路算法来创建这些簇中的多播路径。这种方式可以减少多播路径的长度,平衡每个路由器的通信负载。最后,我们设计并实现了一个轻量级的 NoC 结构,包括一个特定的多播数据包和路由计算组件,以支持多播路由。
4. 结果:(Result & Findings)
我们使用六个数据集来验证所提出的基于路径的多播。 与单播路由、双路径 (DP) 和多路径 (MP) 组播路由相比,所提出的组播在运行时间、总跳数、传输延迟和吞吐量方面提高了 NoC 性能。
5. 总结:(Conclusions)
我们提出了一种考虑 NoC 上神经元分布的基于路径的多播路由。它可以减少多条路径的长度,平衡每个路由器的通信负载。我们应用六个 SNN 数据集来验证所提出的基于路径的多播。与单播协议相比,运行时间实现了5.1倍的加速,总跳数和最大传输延迟分别降低了68.9%和77.4%。实验表明,组播路径的最大长度分别减少了68.3%和67.2%。与双路径 (DP) 或多路径 (MP) 组播路由相比,所提出的组播路由在平均延迟和吞吐量方面具有改进的性能。与 DP 和 MP 相比,平均延迟分别大幅降低了 83.5% 和 67.5%。与 DP 和 MP 路由相比,吞吐量分别显着提高了 2.8 倍和 1.6 倍。 这些路径存储在与本地核心对应的神经元的 SRAM 中。 SRAM 的开销取决于许多因素,例如映射算法、SNN 模型的类型。在今后的工作中,我们将继续完成设计。

Key words: Neuromorphic processors, Spiking Neural Networks, Network-on-Chip, path-based multicast


No related articles found!
Full text



[1] Li Wei;. A Structural Operational Semantics for an Edison Like Language(2)[J]. , 1986, 1(2): 42 -53 .
[2] Li Wanxue;. Almost Optimal Dynamic 2-3 Trees[J]. , 1986, 1(2): 60 -71 .
[3] Feng Yulin;. Recursive Implementation of VLSI Circuits[J]. , 1986, 1(2): 72 -82 .
[4] C.Y.Chung; H.R.Hwa;. A Chinese Information Processing System[J]. , 1986, 1(2): 15 -24 .
[5] Wu Enhua;. A Graphics System Distributed across a Local Area Network[J]. , 1986, 1(3): 53 -64 .
[6] Zhang Cui; Zhao Qinping; Xu Jiafu;. Kernel Language KLND[J]. , 1986, 1(3): 65 -79 .
[7] Wang Jianchao; Wei Daozheng;. An Effective Test Generation Algorithm for Combinational Circuits[J]. , 1986, 1(4): 1 -16 .
[8] Huang Heyan;. A Parallel Implementation Model of HPARLOG[J]. , 1986, 1(4): 27 -38 .
[9] Min Yinghua; Han Zhide;. A Built-in Test Pattern Generator[J]. , 1986, 1(4): 62 -74 .
[10] Huang Xuedong; Cai Lianhong; Fang Ditang; Chi Bianjin; Zhou Li; Jiang Li;. A Computer System for Chinese Character Speech Input[J]. , 1986, 1(4): 75 -83 .

ISSN 1000-9000(Print)

CN 11-2296/TP

Editorial Board
Author Guidelines
Journal of Computer Science and Technology
Institute of Computing Technology, Chinese Academy of Sciences
P.O. Box 2704, Beijing 100190 P.R. China
  Copyright ©2015 JCST, All Rights Reserved