-
集成电路技术的进步使得传统的基于总线结构的片上通信方式面临着功耗、性能、时延和可靠性等诸多方面的问题,已经逐渐不能满足片上多模块间通信的需要。为了克服总线结构的不足,人们将通信网络的思想运用到芯片通信结构的设计上,提出了片上网络(NoC)[1]。为了解决时钟树设计的难题,降低时钟偏斜,NoC通常使用本地同步全局异步(GALS)的互连方法。在这种互连结构中,系统被分为多个功能模块,每一个功能模块都有自己的工作时钟[2]。
NoC的测试可以分为3个主要方面:资源核的测试[3]、路由的测试[4-6]以及路由间互连链路[7-9]的测试,本文关注NoC互连线的测试问题。在当前先进半导体生产工艺下,集成电路片上延时的主要来源已经从门延迟变为互连线延迟。工艺误差以及串扰等问题又使得互连线上的延迟存在较大不确定性。GALS结构对延迟匹配的要求较高,很容易因为线路的延迟引起电路故障[10]。延迟故障通常只在电路工作在正常工作频率时表现出来,若测试时电路的工作频率低于正常工作频率,延迟故障不会被发现。全速测试(at-speed test)面临着测试矢量生成以及测试应用等多方面的问题[11],内建自测试(BIST)很好地解决了这些问题[12-13]。文献[14-15]研究了由串扰和噪声引起的互连线故障,并都提出了相应的测试硬件电路。文献[16]提出了另一种BIST解决方案,并分析了测试结果。但所有这些都是针对同步电路的测试方法,不能被应用到异步电路的测试中,本文提出了一种针对片上跨时钟域互连线的全速测试方法。
-
硬件实现采用Digilent Nexys2 FPGA开发板,该开发搭载一块Xilinx Spartan3E(XC3S500E) FPGA芯片,图 13为开发板上实现的测试电路的结构图。其中,Sender为数据发送端BIST电路;Receiver为数据接收端BIST电路,故障信息输出直接驱动8位LED,通过LED的亮灭指示相应的互连导线上是否存在延迟故障;tpgen_r和tpgen_s分别为Sender和Receiver的ATPG模块;delayModule为数据链路延迟调制模块,通过它设置数据链路的延迟;Test_db完成test_en信号输入的去抖动,test_en信号由开发板上的按键开关产生;seg7为7段数码管的驱动电路,它根据Sender和Receiver的信号输出判断测试电路的当前状态,并将其显示在4位7段数码管上,当测试完成时,数码管上显示done字样;ClkModule为时钟生成模块,它由两个数字时钟管理单元(digital clock management,DCM)组成,对板上50 MHz晶振产生的时钟信号进行倍频与分频,产生测试电路的工作时钟。
完成了电路的实现后,可以利用Xilinx ISE提供的时序分析工具提取出dataout和data_en信号上的延迟,如表 1所示,判断电路是否存在延迟故障,以及所存在的延迟故障类型。通过与FPGA上的运行结果相比较,可知测试电路的运行结果是否正确。
表 1 各链路上延迟数值
From To 延迟/ns dataout_2 tdata_buf_2 8.56 dataout_5 tdata_buf_5 8.06 dataout_7 tdata_buf_7 7.81 data_en data_en_r 6.44 dataout_6 tdata_buf_6 5.04 dataout_3 tdata_buf_3 4.89 dataout_1 tdata_buf_1 4.55 dataout_0 tdata_buf_0 4.30 dataout_4 tdata_buf_4 3.19 将设计实现后的配置文件下载到FPGA中运行,所得结果如图 14所示。LED灯右起第2位、第5位、第7位亮,其他位不亮,表明第2位、第5位、第7位存在延迟故障,与表 1中数据完全相符。
BIST测试需要在系统中添加额外的硬件结构,会占用芯片上有限的资源,所以,需要严格控制BIST电路的资源占用。表 2为本文设计的BIST电路在FPGA中实现时所占用触发器以及LUT的数量。从表中数据可以看出,只需要98个触发器与152个4输入LUT就可以实现本论文中所提出的的测试电路,片上逻辑资源占用较少。
表 2 BIST电路逻辑资源消耗
已用 可用 利用率/% 触发器 98 9 312 1 LUT 152 9 312 1
Research of Interconnection Delay Fault Detection of Network-on-Chip
-
摘要: 基于GALS结构的NoC节点间通常拥有较长的互连线,并且采用异步方式进行通信,对延迟匹配的要求较高。该文提出了一种内建自测试方法,完成跨时钟域互连链路的延迟测试问题。针对该方法完成了相应的测试电路以及测试矢量生成模块的设计与仿真,并在FPGA中实现该电路以验证测试电路的功能和性能。仿真与硬件验证结果都表明,所设计的测试电路以及ATPG模块能够实现NoC互连线延迟故障诊断的功能;该文的延迟故障诊断方法能够快速准确地发现互连线上存在的延迟故障。Abstract: There are quite long interconnection lines between the nodes of a network-on-chip (NoC) based on the globally asynchronous locally synchronous (GALS) structure. It is difficult to match the delay requirement for the communication with the asynchronous ways. In this paper, we propose a build-in self-test (BIST) method to solve the problem of interconnection delay fault across different clock domains. The test circuit and the module of automatic test pattern generation (ATPG) are designed and simulated. The circuit with FPGA is realized to verify the function and performance of the test circuit. The results of simulation and hardware verification indicate that test circuit and the module of ATPG can carry out the function of interconnection delay fault diagnosis of NoC, and the proposed method of interconnection delay fault diagnosis can detect the delay fault existed in the interconnection line rapidly and accurately.
-
表 1 各链路上延迟数值
From To 延迟/ns dataout_2 tdata_buf_2 8.56 dataout_5 tdata_buf_5 8.06 dataout_7 tdata_buf_7 7.81 data_en data_en_r 6.44 dataout_6 tdata_buf_6 5.04 dataout_3 tdata_buf_3 4.89 dataout_1 tdata_buf_1 4.55 dataout_0 tdata_buf_0 4.30 dataout_4 tdata_buf_4 3.19 表 2 BIST电路逻辑资源消耗
已用 可用 利用率/% 触发器 98 9 312 1 LUT 152 9 312 1 -
[1] LAI Ming-che, GAO Lei, XIAO Nong, et al. An accurate and highly-efficient performance evaluation approach based on queuing model for on-chip network[J]. Sicience China Information Sciences, 2013, 56(7):1-20. https://www.researchgate.net/publication/257686935_An_accurate_and_highly-efficient_performance_evaluation_approach_based_on_queuing_model_for_on-chip_network [2] KHETADE V E, LIMAYE S S. Novel data dependent pausible clocking scheme with pll calibration for GALS NOC[C]//Recent Advances in Intelligent Computational Systems (RAICS).[S.l.]:IEEE, 2011:813-818. [3] NAEEM A. Scalability analysis of memory consistency models in NoC-based distributed shared memory SoCs[J]. Computer-Aided Design of Integrated Circuits and Systems, 2013, 32(5):760-773. doi: 10.1109/TCAD.2012.2235914 [4] FU Fang-fa, MA Jian-xin, WU Zi-xu, et al. Low-cost router microarchitecture based on dimension-switch for fault-tolerance in 2D-mesh NoC[C]//Optoelectronics and Microelectronics Technology (AISOMT).[S.l.]:IEEE, 2011:313-319. [5] GALVEZ J J, RUIZ P M. Efficient rate allocation, routing and channel assignment in wireless mesh networks supporting dynamic traffic flows[J]. Ad Hoc Networks, 2013, 11(6):1765-1781. doi: 10.1016/j.adhoc.2013.04.002 [6] LOTFI-KAMRAN P, RAHMANI A M, DANESHTALAB M, et al. EDXY-A low cost congestion-aware routing algorithm for network-on-chips[J]. Journal of Systems Architecture, 2010, 56(7):256-264. doi: 10.1016/j.sysarc.2010.05.002 [7] 姜书艳, 罗刚, 吕小龙, 等. 片上网络互连串扰故障模型的研究及改进[J]. 电子测量技术, 2012, 35(4):123-127. http://www.cnki.com.cn/Article/CJFDTOTAL-DZCL201204032.htm JIANG Shu-yan, LUO Gang, LÜ Xiao-long, et al. Improvement and verification of interconnection crosstalk fault model of network-on-chip[J]. Electronic Measurement Technoloty, 2012, 35(4):123-127. http://www.cnki.com.cn/Article/CJFDTOTAL-DZCL201204032.htm [8] JIANG Shu-yan, LUO Gang, LIANG Hao, et al. A test method of interconnection online detection of NoC based on 2D torus topology[C]//2011 International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2011. Sydney, Australia:[s.n.], 2011:183-187. [9] 姜书艳, 罗刚, 吕小龙, 等. 90 nm和65 nm工艺下片上网络互连串扰故障模型分析[J]. 电子测量与仪器学报, 2012, 26(3):267-272. http://www.cnki.com.cn/Article/CJFDTOTAL-DZIY201203016.htm JIANG Shu-yan, LUO Gang, LÜ Xiao-long, et al. Analysis of interconnection crosstalk fault model of network-on-chip undertechnique of 90 nm and 65 nm[J]. Jounal of Electronic Measurement and Instrument, 2012, 26(3):267-272. http://www.cnki.com.cn/Article/CJFDTOTAL-DZIY201203016.htm [10] JIANG Shu-yan, LUO Gang, CHEN Su, et al. Study of synchronization test methods of NoC at multi-clock domains[J]. COMPEL-the International Jounal for Computation and Mathematics in Electrical and Electronic Engineering, 2013, 32(2):504-515. doi: 10.1108/03321641311296909 [11] CHANG Kai-hui, CHOU Hong-zu, MARKOV I L. RTL analysis and modifications for improving at-speed test[C]//Proceedings Design, Automation and Test in Europe, DATE. Dresden, Germany:IEEE:400-405. [12] CASELLI N, STRANO A, LUDOVICI D, et al. Cooperative built-in self-testing and self-diagnosis of NOC bisynchronous channels[C]//Proceedings-IEEE 6th International Symposium on Embedded Multicore SoCs, MCSoC 2012. Washington, DC, USA:IEEE Computer Society, 2012:159-166. [13] KUIKEN O J, ZHANG Xiao, KERKHOFF H G. Built-in self-diagnostics for a NoC-based reconfigurable IC for dependable beamforming applications[C]//IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems. Washington, DC, USA:IEEE Computer Society, 2008:45-53. [14] ATTARHA A, NOURANI M. Testing interconnects for noise and skew in gigahertz SoC[C]//Proc of ITC.[S.l.]:IEEE, 2001:305-314. [15] SU Chau-chin, CHEN Yue-tsang, HUANG Mu-jeng, et al. All digital built-in delay and crosstalk measurement for on-chip buses[C]//Proc of DATE Conf. Paris, France:IEEE Computer Society, 2000:527-531. [16] JUTMAN A. At-speed on-chip diagnosis of board-level interconnect faults[C]//Proc of 9th European Test Symposium. Washington, DC, USA:IEEE Computer Society, 2004:2-7. [17] CUVIELLO M, DEY S, Bai Xiao-liang, et al. Fault modeling and simulation for crosstalk in system-on-chip interconnects[C]//1999 IEEE/ACM International Conference on Computer-Aided Design. San Jose, CA, USA:IEEE, 1999:297-303.