View in EDS

ReCIM: A SRAM‐Based Digital–Analogue Hybrid CIM Reformer Accelerator Macro.

Saved in:
Bibliographic Details
Title: ReCIM: A SRAM‐Based Digital–Analogue Hybrid CIM Reformer Accelerator Macro.
Authors: Liu, Yu1,2 (AUTHOR), Li, Hao1 (AUTHOR), He, Xin1 (AUTHOR), Wu, Xiulong1,2 (AUTHOR), Li, Xin1,2 (AUTHOR), Peng, Chunyu1,2 (AUTHOR), Lu, Wenjuan1,2 (AUTHOR), Lin, Zhiting1,2 (AUTHOR) ztlin@ahu.edu.cn
Source: Electronics Letters (Wiley-Blackwell). Jan2025, Vol. 61 Issue 1, p1-5. 5p.
Subjects: Static random access memory, Hybrid computers (Computer architecture), Computer architecture, Random access memory, Mechanical efficiency
Abstract: Reformer reduces redundant self‐attention computations via hash bucketing. In this study, we introduce a SRAM‐based digital‐analogue hybrid reformer computing‐in‐memory (ReCIM) accelerator macro. This macro presents an absolute maximum value addressing circuit which facilitates the hash bucketing process and enables the utilisation of strongly‐correlated (S‐C) vectors for attention mechanism computations, thereby improving computational efficiency and saving memory space. Additionally, we introduce a reusable weight array which is suitable for matrix operations across various processes of self‐attention, minimising unnecessary area overhead and enhancing device reusability. The proposed 4 Kb ReCIM macro was analysed using 28‐nm CMOS technology. Simulation results demonstrate that the macro achieves a frequency of 500 MHz at a supply voltage of 0.9 V. During the hash bucketing process, energy efficiency reaches 9.74 TOPS/W. [ABSTRACT FROM AUTHOR]
Copyright of Electronics Letters (Wiley-Blackwell) is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
FullText Text:
  Availability: 0
Header DbId: egs
DbLabel: Engineering Source
An: 190527146
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: ReCIM: A SRAM‐Based Digital–Analogue Hybrid CIM Reformer Accelerator Macro.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Liu%2C+Yu%22">Liu, Yu</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Li%2C+Hao%22">Li, Hao</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22He%2C+Xin%22">He, Xin</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Wu%2C+Xiulong%22">Wu, Xiulong</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Li%2C+Xin%22">Li, Xin</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Peng%2C+Chunyu%22">Peng, Chunyu</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Lu%2C+Wenjuan%22">Lu, Wenjuan</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Lin%2C+Zhiting%22">Lin, Zhiting</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<i> ztlin@ahu.edu.cn</i>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Electronics+Letters+%28Wiley-Blackwell%29%22">Electronics Letters (Wiley-Blackwell)</searchLink>. Jan2025, Vol. 61 Issue 1, p1-5. 5p.
– Name: Subject
  Label: Subjects
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Static+random+access+memory%22">Static random access memory</searchLink><br /><searchLink fieldCode="DE" term="%22Hybrid+computers+%28Computer+architecture%29%22">Hybrid computers (Computer architecture)</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+architecture%22">Computer architecture</searchLink><br /><searchLink fieldCode="DE" term="%22Random+access+memory%22">Random access memory</searchLink><br /><searchLink fieldCode="DE" term="%22Mechanical+efficiency%22">Mechanical efficiency</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Reformer reduces redundant self‐attention computations via hash bucketing. In this study, we introduce a SRAM‐based digital‐analogue hybrid reformer computing‐in‐memory (ReCIM) accelerator macro. This macro presents an absolute maximum value addressing circuit which facilitates the hash bucketing process and enables the utilisation of strongly‐correlated (S‐C) vectors for attention mechanism computations, thereby improving computational efficiency and saving memory space. Additionally, we introduce a reusable weight array which is suitable for matrix operations across various processes of self‐attention, minimising unnecessary area overhead and enhancing device reusability. The proposed 4 Kb ReCIM macro was analysed using 28‐nm CMOS technology. Simulation results demonstrate that the macro achieves a frequency of 500 MHz at a supply voltage of 0.9 V. During the hash bucketing process, energy efficiency reaches 9.74 TOPS/W. [ABSTRACT FROM AUTHOR]
– Name: AbstractSuppliedCopyright
  Label:
  Group: Ab
  Data: <i>Copyright of Electronics Letters (Wiley-Blackwell) is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.)
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=190527146
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1049/ell2.70389
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 5
        StartPage: 1
    Subjects:
      – SubjectFull: Static random access memory
        Type: general
      – SubjectFull: Hybrid computers (Computer architecture)
        Type: general
      – SubjectFull: Computer architecture
        Type: general
      – SubjectFull: Random access memory
        Type: general
      – SubjectFull: Mechanical efficiency
        Type: general
    Titles:
      – TitleFull: ReCIM: A SRAM‐Based Digital–Analogue Hybrid CIM Reformer Accelerator Macro.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Liu, Yu
      – PersonEntity:
          Name:
            NameFull: Li, Hao
      – PersonEntity:
          Name:
            NameFull: He, Xin
      – PersonEntity:
          Name:
            NameFull: Wu, Xiulong
      – PersonEntity:
          Name:
            NameFull: Li, Xin
      – PersonEntity:
          Name:
            NameFull: Peng, Chunyu
      – PersonEntity:
          Name:
            NameFull: Lu, Wenjuan
      – PersonEntity:
          Name:
            NameFull: Lin, Zhiting
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Text: Jan2025
              Type: published
              Y: 2025
          Identifiers:
            – Type: issn-print
              Value: 00135194
          Numbering:
            – Type: volume
              Value: 61
            – Type: issue
              Value: 1
          Titles:
            – TitleFull: Electronics Letters (Wiley-Blackwell)
              Type: main
ResultId 1