片上网络二维和三维结构的通信性能分析

钱悦1，鲁中海2，窦强1，窦文华1

doi:10.3969/j.issn.1007130X.2011.

计算机工程与科学 >

2011 , Vol. 33 >Issue 3: 34 - 40

DOI: https://doi.org/10.3969/j.issn.1007130X.2011.

论文

片上网络二维和三维结构的通信性能分析

展开

(1.国防科学技术大学计算机学院，湖南长沙 410073；
2.瑞典皇家理工学院电子、计算机与软件系统系，瑞典斯德哥尔摩 SE16440)

钱悦(1980),男,江苏江都人，博士，助理研究员，CCF会员（E200009531G），研究方向为计算机体系结构。鲁中海(1968),男，湖北石首人，博士，教授，研究方向为计算机体系结构。窦强(1973),男,山西平定人，博士，研究员，研究方向为计算机体系结构。窦文华(1946),男，山西平定人，教授，研究方向为计算机体系结构和计算机网络。

收稿日期: 2009-10-20

修回日期: 2010-02-25

网络出版日期: 2011-03-25

基金资助

国家自然科学基金资助项目(60633050)

收起

Communication Performance Analysis of the NoCs in 2D and 3D Architectures

Expand

（1.School of Computer Science,National University of Defense Technology,Changsha 410073,China;
2.Department of Electronics,Computer and Software Systems,
Royal Institute of Technology,Stockholm SE16440,Sweden）

Received date: 2009-10-20

Revised date: 2010-02-25

Online published: 2011-03-25

Fold

摘要

芯片集成技术的迅猛发展，使得片上网络从二维向三维扩展成为可能。研究表明三维片上网络因拓扑维度的增加而缩短了通信距离，极大地提升了网络的平均通信性能。本文对比分析了kary2mesh网络及其对应的三维网络在最差情形下的通信性能，得出了以下结论：三维网络的平均通信性能虽然更优，但受垂直信道影响其最差情形下的通信性能可能劣于其对应的二维网络。本文的分析基于网络演算理论，该理论广泛应用于计算信息流穿越各种网络元素的延迟上界。

关键词： 三维片上网络; 网络演算; 延迟上界; 性能分析

本文引用格式

钱悦1，鲁中海2，窦强1，窦文华1 . 片上网络二维和三维结构的通信性能分析[J]. 计算机工程与科学, 2011 , 33(3) : 34 -40 . DOI: 10.3969/j.issn.1007130X.2011.

Abstract

Advanced integration technologies enable the construction of NetworkonChip (NoC) from two dimensions to three dimensions. Studies have shown that 3D NoCs can improve the average communication performance because of the possibility of using the additional dimension to shorten the communication distance. In this paper, we present a comparative analysis on the worstcase communication performance in the regular kary2mesh networks and their 3D forms. We show that, though 3D networks achieve better average latency, this may not be the case for the worstcase performance mainly due to the constraints on vertical channels. Our analysis is based on network calculus, which allows to calculate the theoretical delay bounds for constrained flows traversing network elements.

Key words： 3D NoC;network calculus;delay bound;performance analysis

参考文献

［1］Pavlidis V,Friedman E. 3D Topologies for NetworksonChip［J］. IEEE Transactions on Very Large Scale Integration Systems, 2007,15(10):10811090.
［2］Loi,Mitra S, Lee T H, Fujita S, et al. A LowOverhead Fault Tolerance Scheme for TSVBased 3D Network on Chip Links［C］∥Proc of the 2008 IEEE/ACM Int’l Conf on ComputerAided Design, 2008:598602.
［3］Weerasekera R, Zheng LR, Pamunuwa D, et al. Extending SystemsonChip to the Third Dimension: Performance, Cost and Technological Tradeoffs［C］∥Proc of the 2007 IEEE/ACM Int’l Conf on ComputerAided Design, 2007:212219.
［4］Feero B,Pande P. NetworksonChip in a Three Dimensional Environment: A Performance Evaluation［J］. IEEE Transactions on Computers, 2008, 58(1):3245.
［5］Qian Y, Lu Z, Dou W. Analysis of Communication Delay Bounds for Network on Chips［C］∥Proc of the 14th Asia and South Pacific Design Automation Conf, 2009:712.
［6］Cruz R L. A Calculus for Network Delay, Part I: Network Elements in Isolation; Part II: Network Analysis［J］. IEEE Transactions on Information Theory, 1991, 37(1):114141.
［7］Chang CS. Performance Guarantees in Communication Networks［M］. Berlin:SpringerVerlag, 2000.
［8］Le Boudec JY,Thiran P. Network Calculus: A Theory of Deterministic Queuing Systems for the Internet［M］. Berlin:SpringerVerlag, 2004.
［9］Stiliadis D,Varma A. LatencyRate Servers: A General Model for Analysis of Traffic Scheduling Algorithms［J］. IEEE/ACM Transactions on Networking, 1998, 6(5):611624.
［10］Lenzini L, Martorini L, Mingozzi E, et al. Tight EndtoEnd PerFlow Delay Bounds in FIFO Multiplexing SinkTree Networks［J］. Performance Evaluation, 2006, 63(910):956987.
［11］OMNeT++ Discete Event Simualtion System［EB/OL］.［20081015］.http://www.omnetpp.org/.

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献