FuSe-TG: Domain-Specific 3D ReRAM-based Processing-in-Memory Accelerators for Streaming Time Series Applications

FuSe-TG：用于流时间序列应用的特定领域的基于 3D ReRAM 的内存处理加速器

• 批准号: 2235398
• 负责人: Philip Brisk
• 金额: $ 40万
• 依托单位: University of California-Riverside
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-06-15 至 2025-05-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2235398&HistoricalAwards=false
• 关键词: FuSe; TG; Domain; Specific; 3D

Crossbar Array-based Processing-in-Memory (CBA-PIM) technology can be integrated in 3D with domain-specific heterogeneous compute-engines to achieve unprecedented improvements in application performance and power consumption. This FuSe-Teaming Grant will establish a collaboration between researchers at UC Riverside (UCR), Arizona State University, the Georgia Institute of Technology, and Washington State University. A pilot research project will define domain-specific CBA-PIM accelerators for streaming time series data analytics algorithms, integrate state-of-the-art techniques to achieve fault tolerance and reliability, and advance the state-of-the-art in monolithic 3D layout tools to ensure that the proposed devices can be physically realized. An Education and Workforce Development Plan will provide a comprehensive learning experience for graduate and undergraduate student trainees through which they will learn to design, layout, characterize, and program crossbar arrays, through which they will be exposed to a variety of open-source Electronic Design Automation (EDA) tools and crossbar simulators. It will also provide trainees with Diversity, Equity, and Inclusion (DEI) competencies that will prepare them for employment working with future semiconductor technologies.Moore’s Law has characterized miniaturization in semiconductor microelectronic systems for decades. With the slowing and inevitable end of Moore’s Law on the horizon, advances in general-purpose computing technology can no longer sustain semiconductor industry growth. Subsequent advances are projected to occur through application-specific specialization that synergistically exploits future semiconductor technologies. By exposing trainees to these challenges and empowering them to explore potential solutions, the proposed activities will contribute to the development of the skilled technical workforce. The project leads plan to recruit trainees from underrepresented backgrounds to participate in the project; UCR is a Hispanic Serving Institution (OPEID-00131600). Research and education-related materials will be disseminated via a public-facing website dedicated to the project, and research outcomes will be integrated into a graduate course taught at UCR. This material will be also shared with the other project participants and with the general public.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Crossbar基于阵列的内存处理（CBA-PIM）技术可以在3D中与特定领域的异构计算引擎集成，从而在应用性能和功耗方面实现前所未有的改进。这项FuSe团队资助将在加州大学滨江分校（UCR）、亚利桑那州立大学、格鲁吉亚理工学院和华盛顿州立大学的研究人员之间建立合作。一个试点研究项目将为流时间序列数据分析算法定义特定领域的CBA-PIM加速器，集成最先进的技术以实现容错和可靠性，并推进最先进的单片3D布局工具，以确保所提出的设备可以物理实现。教育和劳动力发展计划将为研究生和本科生学员提供全面的学习体验，通过这些学习，他们将学习设计、布局、表征和编程交叉开关阵列，并通过这些学习接触各种开源电子设计自动化（EDA）工具和交叉开关模拟器。它还将为学员提供多样性，公平性和包容性（DEI）的能力，这将使他们为未来半导体技术的就业做好准备。摩尔定律几十年来一直是半导体微电子系统小型化的特征。随着摩尔定律的缓慢和不可避免的终结，通用计算技术的进步不再能维持半导体行业的增长。后续的进展预计将通过特定应用的专业化来协同利用未来的半导体技术。通过让受训人员接触这些挑战，并使他们有能力探索潜在的解决办法，拟议的活动将有助于培养熟练的技术劳动力。该项目导致计划招募来自代表性不足的背景的学员参加该项目; UCR是一个西班牙裔服务机构（OPEID-00131600）。研究和教育相关的材料将通过一个面向公众的网站专门用于该项目的传播，研究成果将被纳入在加州大学河滨分校教授的研究生课程。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。