CSR: Medium: Collaborative Research: Scale-Out Near-Data Acceleration of Machine Learning

CSR：媒介：协作研究：机器学习的横向扩展近数据加速

• 批准号: 1705047
• 负责人: Nam Sung Kim
• 金额: $ 60万
• 依托单位: University of Illinois at Urbana-Champaign
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-10-01 至 2022-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1705047&HistoricalAwards=false
• 关键词: CSR; Medium; Collaborative; Research; Scale

A growing number of commercial and enterprise systems increasingly rely on machine learning algorithms. This shift is, on the one hand, due to the breakthroughs in machine learning algorithms that extract insights from massive amounts of data. Therefore, such systems need to process ever-increasing amounts of data, demanding higher memory bandwidth and capacity. However, the bandwidth between processors and off-chip memory has not increased due to various stringent physical constraints. Besides, data transfers between the processors and the off-chip memory consume orders of magnitude more energy than on-chip computation due to the disparity between interconnection and transistor scaling.Exploiting recent 3D-stacking technology, the researcher community has explored near-data processing architectures that place processors and memory on the same chip. However, it is unclear whether or not such processing-in-memory (PIM) attempts will be successful for commodity computing systems due to the high cost of 3D-stacking technology and demanded change in existing processor, memory and/or applications. Faced with these challenges, the PIs are to investigate near-data processing platforms that do not require any change in processor, memory and applications, exploiting deep insights on commodity memory subsystems and network software stack. The success of this project will produce inexpensive but powerful near-data processing platforms that can directly run existing machine learning applications without any modification.

越来越多的商业和企业系统越来越依赖机器学习算法。这种转变一方面是由于机器学习算法的突破，这些算法从大量数据中提取见解。因此，这样的系统需要处理不断增加的数据量，要求更高的存储器带宽和容量。然而，由于各种严格的物理约束，处理器和片外存储器之间的带宽并没有增加。此外，由于互连和晶体管缩放之间的差异，处理器和片外存储器之间的数据传输消耗的能量比片上计算多几个数量级。利用最新的3D堆叠技术，研究人员已经探索了将处理器和存储器放置在同一芯片上的近数据处理架构。然而，由于3D堆叠技术的高成本以及现有处理器、存储器和/或应用程序中的所需改变，不清楚这种存储器中处理（PIM）尝试对于商用计算系统是否会成功。面对这些挑战，PI将研究不需要对处理器、内存和应用程序进行任何更改的近数据处理平台，并利用对商品内存子系统和网络软件堆栈的深刻见解。该项目的成功将产生廉价但功能强大的近数据处理平台，可以直接运行现有的机器学习应用程序而无需任何修改。

项目成果

SiMul: An Algorithm-Driven Approximate Multiplier Design for Machine Learning

• DOI: 10.1109/mm.2018.043191125
• 发表时间: 2018-07
• 期刊: IEEE Micro
• 影响因子: 3.6
• 作者: Zhenhong Liu;A. Yazdanbakhsh;Taejoon Park;H. Esmaeilzadeh;N. Kim

BabelFish: Fusing Address Translations for Containers

• DOI: 10.1109/isca45697.2020.00049
• 发表时间: 2020-05
• 期刊: 2020 ACM/IEEE 47th Annual International Symposium on Computer Architecture (ISCA)
• 作者: Dimitrios Skarlatos;Umur Darbaz;Bhargava Gopireddy;N. Kim;J. Torrellas

Rethinking DRAM's page mode with STT-MRAM

使用 STT-MRAM 重新思考 DRAM 的页面模式

• DOI: 10.1109/tc.2022.3207131
• 发表时间: 2022
• 期刊: IEEE Transactions on Computers
• 影响因子: 3.7
• 作者: Oh, Byoungchan;Abeyratne, Nilmini;Kim, Nam Sung;Ahn, Jeongseob;Dreslinski, Ronald G.;Mudge, Trevor
• 通讯作者: Mudge, Trevor

Simulating PCI-Express Interconnect for Future System Exploration

模拟 PCI-Express 互连以进行未来系统探索

• DOI: 10.1109/iiswc.2018.8573496
• 发表时间: 2018
• 期刊: 2018 IEEE International Symposium on Workload Characterization (IISWC)
• 作者: Mohammad Alian;K. Srinivasan;N. Kim
• 通讯作者: N. Kim

Don’t Forget the I/O When Allocating Your LLC

• DOI: 10.1109/isca52012.2021.00018
• 发表时间: 2021-06
• 期刊: 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA)
• 作者: Yifan Yuan;Mohammad Alian;Yipeng Wang;Ren Wang;I. Kurakin;Charlie Tai;N. Kim