CAREER: A Self-Tuning Cache Architecture for Multi-Core Systems

由于从单核技术到多核技术的范式转换导致设计复杂性的增加，设计人员需要先进的优化技术来在不断缩短的上市时间内满足设计限制(例如，功率、能源、性能、面积)。自主自调整体系结构和方法使系统能够动态地适应输入激励，以满足设计限制，而只需很少或根本不需要设计者的工作。然而，自调优系统的设计非常困难，需要定制算法、灵活的可重构体系结构、非介入式自我评估技术等。本研究探索并设计了一种用于节能嵌入式系统的自调优高速缓存层次结构、算法和方法，并设计了新颖、轻量级和高效的可重构体系结构、算法和方法。这项研究将提供关于核心相互依赖如何使其他系统组件(例如，通信拓扑和协议、电压调节、任务到核心映射等)复杂化和揭示新的优化机会和适用于其他系统组件的自调整方法的基础知识。这些成果将催生未来的研究努力，加强学术课程，并通过更快速地发展最先进的技术来影响社会，通过减少设计时间负担，使应用程序设计人员能够更多地专注于增强功能，而不是系统优化。一个主要由代表性不足的少数民族和妇女组成的研究小组将承担该项目的努力。此外，该项目将包括对从成功的女工程师那里收集的现实生活经验进行深入评估，以提供切实的动机，并将通过学术出版物、高中生暑期研究体验计划以及为K-12学生、教师和家长举办的研讨会进行传播。

Due to the increase in design complexity as a result of the paradigm shift from single-core to multi-core technology, designers require advanced optimization techniques to meet design constraints (e.g. power, energy, performance, area) within an ever shrinking time-to-market. Autonomous self-tuning architectures and methodologies enable a system to dynamically adapt to input stimuli to meet design constraints with little or no designer effort. However, self-tuning systems are exceedingly difficult to design, requiring custom algorithms, flexible reconfigurable architectures, non-intrusive self-evaluation techniques, etc. This research explores and architects novel, lightweight, and efficient reconfigurable architectures, algorithms, and methodologies in the context of a self-tuning cache hierarchy for energy efficient embedded systems. This research will provide fundamental knowledge of how core inter-dependencies both complicate and reveal new optimization opportunities and self-tuning methodologies applicable to other system components (e.g. communication topologies and protocols, voltage scaling, task-to-core mapping, etc.). The outcomes will spawn future research endeavors and enhance academic curricula as well as impact society through more rapidly advancing state-of-the-art technology made possible by reducing design time burdens, enabling application designers to focus more heavily on enhanced functionality rather than system optimization. A research group composed largely of underrepresented minorities and women will undertake the project's endeavors. In addition, the project will include an in-depth evaluation of real-life experiences gathered from successful women engineers to provide tangible motivation, which will be disseminated through scholarly publications, summer research experience programs for high school students, as well as seminars for K-12 students, teachers, and parents.