PIMCOMP: A Universal Compilation Framework for Crossbar-based PIM DNN Accelerators

Crossbar-based PIM DNN accelerators can provide massively parallel in-situ operations. A specifically designed compiler is important to achieve high performance for a wide variety of DNN workloads. However, some key compilation issues such as parallelism considerations, weight replication selection,...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2023 60th ACM/IEEE Design Automation Conference (DAC) S. 1 - 6
Hauptverfasser: Sun, Xiaotian, Wang, Xinyu, Li, Wanqian, Wang, Lei, Han, Yinhe, Chen, Xiaoming
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 09.07.2023
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Abstract Crossbar-based PIM DNN accelerators can provide massively parallel in-situ operations. A specifically designed compiler is important to achieve high performance for a wide variety of DNN workloads. However, some key compilation issues such as parallelism considerations, weight replication selection, and array mapping methods have not been solved. In this work, we propose PIMCOMP - a universal compilation framework for NVM crossbar-based PIM DNN accelerators. PIMCOMP is built on an abstract PIM accelerator architecture, which is compatible with the widely used Crossbar/IMA/Tile/Chip hierarchy. On this basis, we propose four general compilation stages for crossbar-based PIM accelerators: node partitioning, weight replicating, core mapping, and dataflow scheduling. We design two compilation modes with different inter-layer pipeline granularities to support high-throughput and low-latency application scenarios, respectively. Our experimental results show that PIMCMOP yields improvements of 1.6× and 2.4× in throughput and latency, respectively, relative to PUMA.
AbstractList Crossbar-based PIM DNN accelerators can provide massively parallel in-situ operations. A specifically designed compiler is important to achieve high performance for a wide variety of DNN workloads. However, some key compilation issues such as parallelism considerations, weight replication selection, and array mapping methods have not been solved. In this work, we propose PIMCOMP - a universal compilation framework for NVM crossbar-based PIM DNN accelerators. PIMCOMP is built on an abstract PIM accelerator architecture, which is compatible with the widely used Crossbar/IMA/Tile/Chip hierarchy. On this basis, we propose four general compilation stages for crossbar-based PIM accelerators: node partitioning, weight replicating, core mapping, and dataflow scheduling. We design two compilation modes with different inter-layer pipeline granularities to support high-throughput and low-latency application scenarios, respectively. Our experimental results show that PIMCMOP yields improvements of 1.6× and 2.4× in throughput and latency, respectively, relative to PUMA.
Author Wang, Lei
Han, Yinhe
Li, Wanqian
Sun, Xiaotian
Wang, Xinyu
Chen, Xiaoming
Author_xml – sequence: 1
  givenname: Xiaotian
  surname: Sun
  fullname: Sun, Xiaotian
  organization: University of Chinese Academy of Sciences,Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
– sequence: 2
  givenname: Xinyu
  surname: Wang
  fullname: Wang, Xinyu
  organization: University of Chinese Academy of Sciences,Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
– sequence: 3
  givenname: Wanqian
  surname: Li
  fullname: Li, Wanqian
  organization: University of Chinese Academy of Sciences,Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
– sequence: 4
  givenname: Lei
  surname: Wang
  fullname: Wang, Lei
  organization: University of Chinese Academy of Sciences,Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
– sequence: 5
  givenname: Yinhe
  surname: Han
  fullname: Han, Yinhe
  organization: University of Chinese Academy of Sciences,Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
– sequence: 6
  givenname: Xiaoming
  surname: Chen
  fullname: Chen, Xiaoming
  email: chenxiaoming@ict.ac.cn
  organization: University of Chinese Academy of Sciences,Center for Intelligent Computing Systems, Institute of Computing Technology, Chinese Academy of Sciences
BookMark eNo1j91KwzAYQCMoqLNvIJIX6MzPlybxrnROB_tD3PX42n6DYteMpCi-vQP16lydA-eWXQ5hIMYepJhKKfzjrKxM4ZWfKqH0VAoF1it3wTJvvdNGaKXByWuWpdTVohDGgSjghr1tF6tqs9o-8ZLvhu6TYsKeV-F46nocuzDwecQjfYX4wQ8h8iqGlGqMeY2JWn62-Wy95mXTUE8RxxDTHbs6YJ8o--OE7ebP79Vrvty8LKpymaPyYswNAIFsa3SeDFkE6ZRWAMI00JAB1foGLWKNxkirbWt07c6GsyjaRoGesPvfbkdE-1Psjhi_9__r-gfalE_w
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/DAC56929.2023.10247928
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 9798350323481
EndPage 6
ExternalDocumentID 10247928
Genre orig-research
GrantInformation_xml – fundername: National Natural Science Foundation of China
  funderid: 10.13039/501100001809
GroupedDBID 6IE
6IH
ACM
ALMA_UNASSIGNED_HOLDINGS
CBEJK
RIE
RIO
ID FETCH-LOGICAL-a290t-544e41dba89e5e7a4182324405c4ce542d9ca7aaba551737d53b8e4187a0dc243
IEDL.DBID RIE
ISICitedReferencesCount 3
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001073487300235&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
IngestDate Wed Aug 27 02:50:59 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-a290t-544e41dba89e5e7a4182324405c4ce542d9ca7aaba551737d53b8e4187a0dc243
PageCount 6
ParticipantIDs ieee_primary_10247928
PublicationCentury 2000
PublicationDate 2023-July-9
PublicationDateYYYYMMDD 2023-07-09
PublicationDate_xml – month: 07
  year: 2023
  text: 2023-July-9
  day: 09
PublicationDecade 2020
PublicationTitle 2023 60th ACM/IEEE Design Automation Conference (DAC)
PublicationTitleAbbrev DAC
PublicationYear 2023
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssib060584064
Score 2.312683
Snippet Crossbar-based PIM DNN accelerators can provide massively parallel in-situ operations. A specifically designed compiler is important to achieve high...
SourceID ieee
SourceType Publisher
StartPage 1
SubjectTerms Accelerator architectures
compilation framework
deep neural network
Design automation
Low latency communication
NVM
Parallel processing
PIM accelerator
Pipelines
Power demand
Throughput
Title PIMCOMP: A Universal Compilation Framework for Crossbar-based PIM DNN Accelerators
URI https://ieeexplore.ieee.org/document/10247928
WOSCitedRecordID wos001073487300235&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NSwMxEA22ePCkYkWtSg5eU3ez2c3GW9laFOy6iEJvZZLMQkFa6Ye_3yTdVTx48BYCQ2CS8GYy72UIuVEuJ5BSCxZbZZjgvGaguGVJprM8AysRw-_6T7Is8-lUVY1YPWhhEDGQz3Dgh6GWb5dm65_K3A3nQiqed0hHSrkTa7WHx5f3HDiJRgUcR-p2NCzSzMH_wLcIH7TGv9qoBBQZH_5z_SPS-9Hj0eobaY7JHi5OyEv1OCmeJ9UdHdKGXgHv1N_v-Y7fRsct8Yq6yJQWHg81rJgHLkudNR2VJR0a45AnFNvXPfI2vn8tHljTIYEBV9GGpUKgiK2GXGGKEoTLFlyE5IIwIwymgrsNAAmgwQVGMpE2TXTuLHIJkTVcJKeku1gu8IzQCLM6TWyaR5r7nE9DJGsEqONYoojgnPS8Q2Yfu08wZq0vLv6Y75MD7_bAbFWXpLtZbfGK7JvPzXy9ug5b9wVyg5iO
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NSwMxEA1aBT2pWPHbHLym7mazm423srW02K6LVOitzCazUJBW-uHvN0m7igcP3pbAQJjZ8N5k5mUIuVc2J5CyFCw0SjPBecVAccOipEzSBIxE9K_rD2Sep-OxKrZida-FQUTffIYt9-lr-Wau1-6qzJ5wLqTi6S7Zi4Xg4UauVf8-rsBn4UlsdcBhoB467SxOLAFouSHhrdr81yAVjyPdo3_u4Jg0fxR5tPjGmhOyg7NT8lr0h9nLsHikbbptsIB36k74dNPhRrt16xW13JRmDhFLWDAHXYZaa9rJc9rW2mKPL7cvm-St-zTKemw7I4EBV8GKWT-gCE0JqcIYJQibL1iOZGmYFhpjwW0IQAKUYKmRjKSJozK1FqmEwGguojPSmM1neE5ogEkVRyZOg5K7rK-EQFYIUIWhRBHABWk6h0w-Ns9gTGpfXP6xfkcOeqPhYDLo589X5NCFwPe5qmvSWC3WeEP29edqulzc-jB-AZSIm9U
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=2023+60th+ACM%2FIEEE+Design+Automation+Conference+%28DAC%29&rft.atitle=PIMCOMP%3A+A+Universal+Compilation+Framework+for+Crossbar-based+PIM+DNN+Accelerators&rft.au=Sun%2C+Xiaotian&rft.au=Wang%2C+Xinyu&rft.au=Li%2C+Wanqian&rft.au=Wang%2C+Lei&rft.date=2023-07-09&rft.pub=IEEE&rft.spage=1&rft.epage=6&rft_id=info:doi/10.1109%2FDAC56929.2023.10247928&rft.externalDocID=10247928