SHF:SMALL:Collaborative Research: Exploring Nonvolatility of Emerging Memory Technologies for Architecture Design

SHF:SMALL:合作研究：探索新兴内存技术的非易失性以用于架构设计

• 批准号: 1816833
• 负责人: Yuan Xie
• 金额: $ 24.9万
• 依托单位: University of California-Santa Barbara
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-08-01 至 2021-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1816833&HistoricalAwards=false
• 关键词: SHF; SMALL; Collaborative; Research; Exploring

In modern computers, by combining the speed of traditional cache technology, the density of traditional main memory technology, and the non-volatility of flash memory, a new class of emerging byte-addressable nonvolatile memories (NVMs) have great potential to be used as the universal memories of the future. Such memory types include technologies such as phase-change memory, spin-transfer-torque magnetoresistive memory, and resistive memory. As these emerging memory technologies mature, it is important for computer architects to understand their pros and cons in a comprehensive manner in order to improve the performance, power, and reliability of future computer systems incorporating these systems which will be used in various application domains. Yet, most of previous research on NVM architecture is focused only on the performance, power, and density benefits and how to overcome challenges, such as write overhead and wearout issues. The non-volatility characteristic of NVM technologies is not fully explored. Therefore, this project examines how to exploit the non-volatility characteristic that distinguishes the emerging NVM technologies from traditional memory technologies, and investigate new memory architecture design with novel applications.The goal of this project is to advance the memory architecture design of various types of computer systems with a full exploration of the non-volatility characteristic of NVM technologies across architecture, system, and application levels. To this end, the project explores the design space of various types of computer systems, ranging from severs to embedded systems. In particular, the project identifies and addresses design issues in nonvolatile cache architecture, re-architects main memory structure to leverage the non-volatility characteristic to improve system performance and energy consumption, supports persistent memory systems in various use cases with emerging NVM technologies, and studies near-data-computing techniques applied for these NVM technologies. The successful outcome of this research is expected to provide the design guidelines for enabling both large capacity and fast-bandwidth nonvolatile memory/storage, which are beyond the present state-of-the-art. Consequently, the research will spawn new applications involving the computation on the exascale of data, e.g., data mining, machine learning, visual or auditory sensory data recognition, bio-informatics, etc.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.

在现代计算机中，通过结合传统的缓存技术的速度，传统的主要记忆技术的密度以及闪存的非挥发性，一类新的新出现的字节可调性的非挥发性记忆（NVM）具有巨大的潜力，可以用作未来的普遍记忆。这样的内存类型包括诸如相变内存，旋转转移磁力磁盘的内存和电阻内存等技术。随着这些新兴内存技术的成熟，对于计算机架构师来说，重要的是要以全面的方式理解其优点和缺点，以提高结合这些系统将在各种应用程序领域中使用的未来计算机系统的性能，功率和可靠性。然而，先前关于NVM体系结构的大多数研究仅集中在性能，力量和密度收益上，以及如何克服挑战，例如写开销和磨损问题。 NVM技术的非挥发性特征没有得到充分探索。因此，该项目探讨了如何利用非挥发性特征，该特征将新兴的NVM技术与传统的记忆技术区分开来，并使用新颖的应用来研究新的内存架构设计。该项目的目的是推进各种计算机系统的内存架构设计，并全面探索NVM技术范围内的NVM技术层面和系统级别，系统和系统，系统，系统，系统，系统，系统，系统，系统，系统和系统。为此，该项目探讨了各种类型的计算机系统的设计空间，范围从裂缝到嵌入式系统。 特别是，该项目识别并解决了非易失性高速缓存架构中的设计问题，重新构造主要记忆结构以利用非挥发性特征来提高系统性能和能源消耗，在各种新兴NVM技术中支持持续的内存系统，并为这些NVM技术提供了近乎DATA的技术。预计这项研究的成功结果将提供设计指南，以实现大容量和快速宽度的非易失性内存/存储，这超出了当前的最新时间。因此，这项研究将产生涉及数据计算的新应用程序，例如，数据挖掘，机器学习，视觉或听觉感觉数据识别，生物信息学等。该奖项反映了NSF的法定任务，并被认为是通过基金会的知识功能和广泛的影响来评估CRITERIA的智力功能和广泛影响。

项目成果

MEDAL: Scalable DIMM based Near Data Processing Accelerator for DNA Seeding Algorithm

• DOI: 10.1145/3352460.3358329
• 发表时间: 2019-10
• 期刊: Proceedings of the 52nd Annual IEEE/ACM International Symposium on Microarchitecture
• 作者: Wenqin Huangfu;Xueqi Li;Shuangchen Li;Xing Hu;P. Gu;Yuan Xie

A Survey of Accelerator Architectures for Deep Neural Networks

• DOI: 10.1016/j.eng.2020.01.007
• 发表时间: 2020-03-01
• 期刊: ENGINEERING
• 影响因子: 12.8
• 作者: Chen, Yiran;Xie, Yuan;Tang, Tianqi
• 通讯作者: Tang, Tianqi

Rescuing RRAM-Based Computing From Static and Dynamic Faults

• DOI: 10.1109/tcad.2020.3037316
• 发表时间: 2021-10-01
• 期刊: IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS
• 影响因子: 2.9
• 作者: Lin, Jilan;Wen, Cheng-Da;Xie, Yuan
• 通讯作者: Xie, Yuan

Improving Streaming Graph Processing Performance using Input Knowledge

• DOI: 10.1145/3466752.3480096
• 发表时间: 2021-10
• 期刊: MICRO-54: 54th Annual IEEE/ACM International Symposium on Microarchitecture
• 作者: Abanti Basak;Zheng Qu;Jilan Lin;Alaa R. Alameldeen;Zeshan A. Chishti;Yufei Ding;Yuan Xie

Overcoming the Memory Hierarchy Inefficiencies in Graph Processing Applications

克服图形处理应用程序中内存层次结构的低效率

• DOI: 10.1109/iccad51958.2021.9643434
• 发表时间: 2021
• 期刊: 2021 IEEE/ACM International Conference On Computer Aided Design (ICCAD
• 作者: Lin, Jilan;Li, Shuangchen;Ding, Yufei;Xie, Yuan
• 通讯作者: Xie, Yuan