AttenPIM: Accelerating LLM Attention with Dual-mode GEMV in Processing-in-Memory

Large Language Models (LLMs) have demonstrated unprecedented generative performance across a wide range of applications. While recent heterogeneous architectures attempt to address the memory-bound bottleneck from attention computations by processing-in-memory (PIM) offloading, they overlook two cri...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:2025 62nd ACM/IEEE Design Automation Conference (DAC) s. 1 - 7
Hlavní autori: Chen, Liyan, Lyu, Dongxu, Li, Zhenyu, Jiang, Jianfei, Wang, Qin, Mao, Zhigang, Jing, Naifeng
Médium: Konferenčný príspevok..
Jazyk:English
Vydavateľské údaje: IEEE 22.06.2025
Predmet:
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Abstract Large Language Models (LLMs) have demonstrated unprecedented generative performance across a wide range of applications. While recent heterogeneous architectures attempt to address the memory-bound bottleneck from attention computations by processing-in-memory (PIM) offloading, they overlook two critical characteristics of attention GEMVs that distinguish them from traditional PIM scenarios: (1) dynamic matrix dimensions that scale with token length, and (2) distinct GEMV patterns between score computation (Q \times K_{t}) and context computation (S \times V). Existing PIM designs, employing either uniform or transposed computing modes, suffer from inefficiencies in newly generated element preparation or distinct GEMV execution. To address these limitations, we propose AttenPIM, a software-hardware co-design for efficient PIM-based attention acceleration. For bank-level execution, we propose dual-mode computing modes tailored for score and context computations with PIM-oriented data layouts and execution flows for KV storage, supported by a low-cost configurable per-bank PIM unit (PU). For system-level execution, we leverage token-level and head-level concurrency to ensure workload balance and maximize bank PU parallelism. Furthermore, dynamic allocation and kernel fusion methods are proposed to further minimize memory overhead. Experimental results demonstrate that AttenPIM achieves 1.13 \times-5.26 \times speedup and reduces energy consumption by 17 %-49 % compared to two state-of-the-art PIM baselines.
AbstractList Large Language Models (LLMs) have demonstrated unprecedented generative performance across a wide range of applications. While recent heterogeneous architectures attempt to address the memory-bound bottleneck from attention computations by processing-in-memory (PIM) offloading, they overlook two critical characteristics of attention GEMVs that distinguish them from traditional PIM scenarios: (1) dynamic matrix dimensions that scale with token length, and (2) distinct GEMV patterns between score computation (Q \times K_{t}) and context computation (S \times V). Existing PIM designs, employing either uniform or transposed computing modes, suffer from inefficiencies in newly generated element preparation or distinct GEMV execution. To address these limitations, we propose AttenPIM, a software-hardware co-design for efficient PIM-based attention acceleration. For bank-level execution, we propose dual-mode computing modes tailored for score and context computations with PIM-oriented data layouts and execution flows for KV storage, supported by a low-cost configurable per-bank PIM unit (PU). For system-level execution, we leverage token-level and head-level concurrency to ensure workload balance and maximize bank PU parallelism. Furthermore, dynamic allocation and kernel fusion methods are proposed to further minimize memory overhead. Experimental results demonstrate that AttenPIM achieves 1.13 \times-5.26 \times speedup and reduces energy consumption by 17 %-49 % compared to two state-of-the-art PIM baselines.
Author Li, Zhenyu
Jiang, Jianfei
Chen, Liyan
Mao, Zhigang
Jing, Naifeng
Lyu, Dongxu
Wang, Qin
Author_xml – sequence: 1
  givenname: Liyan
  surname: Chen
  fullname: Chen, Liyan
  email: liyan.chen@sjtu.edu.cn
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
– sequence: 2
  givenname: Dongxu
  surname: Lyu
  fullname: Lyu, Dongxu
  email: sjtuj@sjtu.edu.cn
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
– sequence: 3
  givenname: Zhenyu
  surname: Li
  fullname: Li, Zhenyu
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
– sequence: 4
  givenname: Jianfei
  surname: Jiang
  fullname: Jiang, Jianfei
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
– sequence: 5
  givenname: Qin
  surname: Wang
  fullname: Wang, Qin
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
– sequence: 6
  givenname: Zhigang
  surname: Mao
  fullname: Mao, Zhigang
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
– sequence: 7
  givenname: Naifeng
  surname: Jing
  fullname: Jing, Naifeng
  organization: Shanghai Jiao Tong University,Department of Micro/Nano Electronics,Shanghai,China
BookMark eNo1j99KwzAYxSO4C517A5G8QGaSr2kb70o356DFXkxvR9J800CbShuRvb3FPzfnwPkdDpxrchmGgITcCb4Wguv7TVGmkCd6LblUcyQAJPALstKZzgGE4sCT_Io0RYwYmn39QIu2xQ5HE314o1VV0x8U_RDol4_vdPNpOtYPDuluW79SH2gzDi1O09xnPrAa-2E835DFyXQTrv58SV4et4fyiVXPu31ZVMyITEc2i1OoWpMpx5XT3HKZCatSC86mCEYlaJVMEY0Dm3PTanCSJ0KJkzS5gCW5_d31iHj8GH1vxvPx_yh8A3hpS_w
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/DAC63849.2025.11133230
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP) 1998-present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
EISBN 9798331503048
EndPage 7
ExternalDocumentID 11133230
Genre orig-research
GrantInformation_xml – fundername: National Natural Science Foundation of China
  funderid: 10.13039/501100001809
GroupedDBID 6IE
6IH
CBEJK
RIE
RIO
ID FETCH-LOGICAL-a179t-179d5e5ca75d05d90b0271b56b3db6e3a54eb526eead3b80ac93d204151f2a813
IEDL.DBID RIE
IngestDate Wed Oct 01 07:05:15 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-a179t-179d5e5ca75d05d90b0271b56b3db6e3a54eb526eead3b80ac93d204151f2a813
PageCount 7
ParticipantIDs ieee_primary_11133230
PublicationCentury 2000
PublicationDate 2025-June-22
PublicationDateYYYYMMDD 2025-06-22
PublicationDate_xml – month: 06
  year: 2025
  text: 2025-June-22
  day: 22
PublicationDecade 2020
PublicationTitle 2025 62nd ACM/IEEE Design Automation Conference (DAC)
PublicationTitleAbbrev DAC
PublicationYear 2025
Publisher IEEE
Publisher_xml – name: IEEE
Score 2.295538
Snippet Large Language Models (LLMs) have demonstrated unprecedented generative performance across a wide range of applications. While recent heterogeneous...
SourceID ieee
SourceType Publisher
StartPage 1
SubjectTerms Computer architecture
Concurrent computing
Design automation
Dynamic scheduling
Energy consumption
Kernel
Large language models
Layout
Parallel processing
Resource management
Title AttenPIM: Accelerating LLM Attention with Dual-mode GEMV in Processing-in-Memory
URI https://ieeexplore.ieee.org/document/11133230
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LSwMxEA62ePCkYsU3OXhNu5vdbDbeSh8qdMseVHorecxKQbalbgX_vZntVvHgwVtIAoHJhC-TzDcfIbfGprEA45jlYFishWJK-5in8L4cmQIhtlYtmcjpNJ3NVN6Q1WsuDADUyWfQxWb9l--WdoNPZT2URY_8nblFWlLKLVmrYf2GgeoN-wPvTTHST7jo7ib_kk2pUWN8-M_1jkjnh39H829kOSZ7UJ6QvF_5-23-mN3RvrUeLXDvylc6mWS0HkITU3xXpcONfmOocUPvR9kLXZS04QP4-WxRsgzTaz875Hk8eho8sEYPgWl_bCqsJOoECKulcIFwKjA-pgyNSEzkTAKRFjEYwRPw3hGZNNBWRY4jBz8suE7D6JS0y2UJZ4RKrjUIJ5XHqjgJpOZGa1FwEwAoH8Kckw6aY77alryY7yxx8Uf_JTlAo2MOFedXpF2tN3BN9u1HtXhf39Qb9QXahpSU
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LSwMxEA5aBT2pWPFtDl7T7mY33Y230oct7pY9VOmt5DErBdlK3Qr-ezPbVfHgwVtIAoGZCV8mmS8fIbfaxKEAbZnhoFmohGRSuZwnd7Ec6BwhtlItSaLJJJ7NZFaT1SsuDABUxWfQwmb1lm-XZo1XZW2URQ_cmXmb7Igw5P6GrlXzfn1PtvvdnounEAkoXLS-pv8STqlwY3jwzxUPSfOHgUezb2w5IltQHJOsW7oTbjZO72jXGIcX6L3imSZJSqshNDLFm1XaX6sXhio39H6QPtFFQWtGgJvPFgVLscD2o0keh4Npb8RqRQSm3MYp8S9RK0AYFQnrCSs97bJKX4uODqzuQKBECFrwDrj4CHTsKSMDy5GF7-dcxX5wQhrFsoBTQiOuFAgbSYdWYceLFNdKiZxrD0C6JOaMNNEc89fNpxfzL0uc_9F_Q_ZG0zSZJ-PJwwXZRwdgRRXnl6RRrtZwRXbNe7l4W11XTvsExt2X2w
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=proceeding&rft.title=2025+62nd+ACM%2FIEEE+Design+Automation+Conference+%28DAC%29&rft.atitle=AttenPIM%3A+Accelerating+LLM+Attention+with+Dual-mode+GEMV+in+Processing-in-Memory&rft.au=Chen%2C+Liyan&rft.au=Lyu%2C+Dongxu&rft.au=Li%2C+Zhenyu&rft.au=Jiang%2C+Jianfei&rft.date=2025-06-22&rft.pub=IEEE&rft.spage=1&rft.epage=7&rft_id=info:doi/10.1109%2FDAC63849.2025.11133230&rft.externalDocID=11133230