SHF: Small: Virtual Probe: A Statistically Optimal Framework for Affordable Monitoring and Tuning of Large-Scale Digital Integrated Circuits

[ID: 0915912]标题：虚拟探针：大规模数字集成电路可负担监测和调谐的统计最优框架[j]: Xin Li和Rob Rutenbar，卡内基梅隆大学摘要片上可变性监测和后硅调谐已经成为一种联合策略，以对抗纳米级工艺变化的有害影响，以保持集成电路（ic）的积极缩放。然而，这些新技术的设计开销（例如，模具面积，功耗等）是一个日益严重的问题，因为设备继续缩小，关键工艺波动的相对幅度继续增长。该项目建议开发一种称为虚拟探针（VP）的新型统计框架，以最大限度地减少可变性监测和硅后调谐的开销。VP从尽可能小的测量数据集准确地预测全芯片的空间变化，从而实现最低成本/最高精度的硅测试、表征和调谐，随着IC技术进一步进入纳米级体系。该项目旨在创建一个从根本上改进的平台，用于大规模数字集成电路的片上统计监测和调谐；预计在从消费电子到航空航天控制器的广泛应用中，先进ic的性能将提高5-10倍。此外，所提出的数学框架也适用于许多其他科学和工程问题，因此，为研究和理解这些问题提供了新的途径。最后，鉴于其在多个研究领域的广泛覆盖，拟议的项目激发了统计学家、计算机科学家和电路设计师之间的密切合作，从而为跨学科创新创造了巨大的机会。该项目的跨学科性质也为在多个科学和工程领域培养下一代美国研究人员提供了极好的机会。

ID: 0915912Title: Virtual Probe: A Statistically Optimal Framework for Affordable Monitoring and Tuning of Large-Scale Digital Integrated CircuitsPI: Xin Li and Rob Rutenbar, Carnegie Mellon UniversityAbstractOn-chip variability monitoring and post-silicon tuning have emerged as a joint-strategy to combat the deleterious effects of nanoscale process variations, to maintain the aggressive scaling of integrated circuits (ICs). However, the design overhead (e.g., die area, power consumption, etc.) of these new techniques is a growing problem as devices continue to shrink, and the relative magnitude of critical process fluctuations continues to grow. This project proposes to develop a novel statistical framework called virtual probe (VP) to minimize the overhead of variability monitoring and post-silicon tuning. VP accurately predicts full-chip spatial variation from the smallest possible set of measurement data, thereby enabling lowest-cost / highest-accuracy silicon testing, characterization and tuning as IC technologies move further into the nanoscale regime.The proposed project aims to create a radically improved platform for on-chip statistical monitoring and tuning of large-scale digital ICs; it is expected to yield 5-10 times performance improvement for advanced ICs in a broad range of applications, from consumer electronics to aerospace controllers. In addition, the proposed mathematical framework is applicable to many other scientific and engineering problems and, hence, offers a new avenue to study and understand these. Finally, given its broad coverage over multiple research areas, the proposed project motivates close collaboration among statistician, computer scientists and circuit designers, thereby creating enormous opportunities for interdisciplinary innovations. The interdisciplinary nature of this project also offers an excellent opportunity to train the next generation of U.S. researchers in multiple science and engineering domains.