面向片上网络的多播吞吐率和能量模型
DOI:
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

基金项目:

国家自然科学基金资助项目(60873212、60970036、61003075);国家“核高基”重大专项“超高性能CPU新型架构研究”资助项目(2011ZX01028-001-001)


Throughput and Energy Model of Multicast Communication in Networks on Chip
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    片上网络逐渐成为片上众核中非常有前景的互连方式。基于目录的cache一致性协议的维护需要片上互连网络高效的支持多播。在借鉴单播的网络吞吐率模型基础上,建立了面向多播的网络吞吐率模型和体系结构级的能量模型。相对于传统的多播路由算法,负载平衡维序路由算法(BDoR)和最小路径维序路由算法(MPDoR)通过平衡X和Y两个方向上的网络负载,从而能够获得较大的网络吞吐率。基于功能模拟器的模拟结果显示:在4×4的Mesh网络中,随机广播通信下,MPDoR算法的网络吞吐率是XY多播路由算法的1.6倍;网络输出加速比为2时,网络才能够进入饱和状态;MPDoR算法下网络消耗能量与OPT算法相当。

    Abstract:

    Networks-on-Chip (NoC) has become a promising interconnect solution between processor cores and cache banks in many-core-processor. Routers in the network should support multicast efficiently, due to the fact that the invalidation requests in the directory-based cache coherent protocols often cause multicasting packets. Throughput and energy model of multicast communication proposed in this paper can evaluate the network ideal throughput and energy under the different multicasting routing algorithms. Network can get higher throughput under Balanced Dimension order Routing algorithm (BDoR) and Minimal Path Dimension order Routing algorithm (MPDoR) proposed in the paper than the conventional multicast routing algorithms, because they can balance the link load between X direction and Y direction. Simulation results show that the throughput under MPDoR is 1.6 times than that under the XY multicasting routing algorithm under random broadcast traffic in the 4×4 mesh network; the link will not get saturated if the output speedup of the network is less than two; and the energy consumption under MPDoR is almost the same as that under OPT alogithm.

    参考文献
    相似文献
    引证文献
引用本文

齐树波,蒋江,李晋文,等.面向片上网络的多播吞吐率和能量模型. Throughput and Energy Model of Multicast Communication in Networks on Chip[J].国防科技大学学报,2011,33(6):36-41.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2011-06-20
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2012-09-12
  • 出版日期:
文章二维码