当前位置: X-MOL 学术IEEE Trans. Emerg. Top. Comput. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Self-Adjusting Monitor for Measuring Aging Rate and Advancement
IEEE Transactions on Emerging Topics in Computing ( IF 5.1 ) Pub Date : 2020-07-01 , DOI: 10.1109/tetc.2017.2771441
Somayeh Sadeghi-Kohan , Mehdi Kamal , Zainalabedin Navabi

Time-variant age information of different parts of a system can be used for system-level performance improvement through high-level task scheduling, thus extending the life-time of the system. Progressive age information should provide the age state that the system is in, and the rate that it is being aged at. In this paper, we propose a structure that monitors certain paths of a circuit and detects its gradual age growth, and provides the aging rate and aging state of the circuit. The proposed monitors are placed on a selected set of nodes that represent a timing bottleneck of the system. These monitors sample expected data on these nodes, and compare them with the expected values. The timing of sampling changes as the circuit ages and its delay increases. The timing of sampling will provide a measure of aging advancement of a circuit. To assess the efficacy of the proposed method and compare it with other state-of-the-art aging monitors, we use them on selected nodes of the execution unit of different processors, as well as some circuits from ITC99 benchmarks. The results reveal that the precision of our proposed method is between 0.12 (ns) to 0.401 (ns). Its Area and power overhead are negligible and are about 2.13 and 0.69 percent respectively.

中文翻译:

用于测量老化率和先进性的自调整监视器

系统不同部分的时变年龄信息可以通过高层任务调度用于系统级性能提升,从而延长系统的生命周期。渐进式年龄信息应该提供系统所处的年龄状态,以及它正在老化的速度。在本文中,我们提出了一种结构,可以监视电路的某些路径并检测其逐渐老化的增长,并提供电路的老化率和老化状态。建议的监视器放置在代表系统时序瓶颈的一组选定节点上。这些监视器对这些节点上的预期数据进行采样,并将它们与预期值进行比较。采样时间随着电路老化及其延迟的增加而变化。采样时间将提供电路老化进展的度量。为了评估所提出方法的有效性并将其与其他最先进的老化监视器进行比较,我们在不同处理器的执行单元的选定节点以及 ITC99 基准测试中的一些电路上使用它们。结果表明,我们提出的方法的精度在 0.12 (ns) 到 0.401 (ns) 之间。它的面积和功率开销可以忽略不计,分别约为 2.13% 和 0.69%。
更新日期:2020-07-01
down
wechat
bug