II-NEW: Infrastructure Acquisition for Research and Education in Statistical Design

设计和分析集成系统(微处理器、SoC、ASIC、存储器、MEMS等)由于制造过程和操作环境中都存在极端的变化，因此越来越困难。因此，在设计过程中简单地执行最坏情况分析不再可行。相反，这些变化导致的设计分布必须高效而准确地建模。日益扩大的一体化规模加剧了这一需求。随着更大的片上存储器，越来越多的不同计算和通信模块的使用，以及越来越多的异构性，这些系统必须有良好的特征，并从统计学上设计达到甚至超过六西格玛。卡内基梅隆大学的硅系统实施中心(CSSI)专注于统计设计和分析的所有方面，从系统级架构到模拟/数字电路，再到半导体制造中发现的复杂性的物理和建模。利用在NSF计算研究基础设施(CRI)计划的支持下获得的共享计算基础设施，CSSI正在开发各种统计设计和分析方法，以支持继续使用按比例扩展的CMOS技术。CSSI的教师在与集成系统设计和制造相关的不同领域领导着70多个不同的研究项目，并指导20多个涵盖这些相同领域的不同的研究生/高级课程。通过支持国家科学基金研究和创新方案而获得的计算机基础设施，大大增强了开发和验证统计设计新方法以及培训下一代综合系统研究人员和从业人员的能力。

Design and analysis of integrated systems (microprocessors, SoCs, ASICs, memories, MEMS, etc.) is increasingly difficult due to the extreme variations that exist both in the fabrication process and the operating environment. As a result, it is no longer viable to simply perform worst-case analysis during design. Instead, the distribution of designs resulting from these variations must be efficiently and accurately modeled. Compounding this need is the increasing scale of integration. With larger on-chip memories, growing use of diverse computing and communication blocks, and increasing heterogeneity, it is imperative that these systems be well characterized and "statistically" designed up to and even beyond "six sigma". The Center for Silicon System Implementation (CSSI) at Carnegie Mellon is focused on all aspects of statistical design and analysis that spans from system-level architectures to analog/digital circuits, to the physics and modeling of the complexities found in semiconductor manufacturing. Using the shared compute infrastructure acquired with support from the NSF Computing Research Infrastructure (CRI) program, CSSI is developing various statistical design and analysis methodologies that enable continued use of scaled-CMOS technology. CSSI faculty lead over 70 different research projects in various areas related to integrated system design and manufacturing, and instruct over 20 different graduate/senior-level courses that cover these same areas. The ability to develop and validate new methodologies for statistical design, and train the next-generation of integrated systems researchers and practitioners is being significantly enhanced by the computer infrastructure acquired through support of the NSF CRI program.