BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180002Z
LOCATION:3002\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240626T103000
DTEND;TZID=America/Los_Angeles:20240626T120000
UID:dac_DAC 2024_sess116@linklings.com
SUMMARY:Memories Have a Mind of Their Own
DESCRIPTION:Research Manuscript\n\nComputing-in-memory (CIM) continues to 
 advance energy efficiency for deep learning with breakthroughs in architec
 ture, circuitry, and devices. The first paper introduces a tri-gear hetero
 geneous digital CIM for flexible data reuse in Diffusion models. The secon
 d paper unveils a fine-grained digital CIM with hessian trace-based quanti
 zation and approximate computing. The third paper presents a hybrid-domain
  SRAM CIM macro, harmonizing accuracy and energy efficiency through digita
 l-analog computing synergy. The next two papers harness RRAM and IGZO in C
 IM applications. The final paper presents a drop-in-replacement design all
 owing fine-grained interleaving of processing-in-memory accesses and host 
 accesses.\n\nFDCA: Fine-grained Digital-CIM based CNN Accelerator with Hyb
 rid Quantization and Weight-Stationary Dataflow\n\nDigital-Compute-In-Memo
 ry (DCIM) has demonstrated significant energy and area efficiency in convo
 lutional neural network (CNN) accelerators, particularly for high precisio
 n applications. However, to mitigate parasitic effects on word and bit lin
 es, most DCIMs employ fine-grained multiply-accumulate ...\n\n\nBo Liu, Qi
 ngwen Wei, Yang Zhang, Xingyu Xu, Zihan Zou, Xinxiang Huang, Xin Si, and H
 ao Cai (Southeast University)\n---------------------\nAddition is Most You
  Need: Efficient Floating-Point SRAM Compute-in-Memory by Harnessing Manti
 ssa Addition\n\nThe compute-in-memory (CIM) paradigm holds great promise t
 o efficiently accelerate machine learning workloads. Among memory devices,
  static random-access memory (SRAM) stands out as a practical choice due t
 o its exceptional reliability in the digital domain and balanced performan
 ce. Recently, there ...\n\n\nWeidong Cao (George Washington University), J
 ian Gao (Northeastern University), Xin Xin (University of Central Florida)
 , and Xuan Zhang (Northeastern University)\n---------------------\nRWriC: 
 A Dynamic Writing Scheme for Variation Compensation for RRAM-based In-Memo
 ry Computing\n\nRRAM-based compute-in-memory (CIM) suffers from programmin
 g variation issues, specifically device-to-device variation (DDV) and cycl
 e-to-cycle variation (CCV), which can have a detrimental impact on inferen
 ce accuracy. To address these variation issues, we propose RWriC, a dynami
 c Writing scheme for...\n\n\nYucong Huang, Jingyu He, Tim Cheng, and Chi Y
 ing Tsui (Hong Kong University of Science and Technology (HKUST)) and Terr
 y Ye (Southern University of Science and Technology)\n--------------------
 -\nIG-CRM: Area/Energy-Efficient IGZO-Based Circuits and Architecture Desi
 gn for Reconfigurable CIM/CAM Applications\n\nArtificial intelligence is e
 volving with various algorithms such as deep neural network (DNN), Transfo
 rmer, recommendation system (RecSys) and graph convolutional network (GCN)
 . Correspondingly, multiply-accumulate (MAC) and content search are two ma
 in operations, which can be efficiently executed o...\n\n\nZeyu Guo, Jinsh
 an Yue, Shengzhe Yan, Zhuoyu Dai, Xiangqu Fu, Zhaori Cong, Zening Niu, Ke 
 Hu, Lihua Xu, Jiawei Wang, Lingfei Wang, Guanhua Yang, Di Geng, and Ling L
 i (Institute of Microelectronics, Chinese Academy of Sciences)\n----------
 -----------\nHAIL-DIMM: Host Access Interleaved with Near-Data Processing 
 on DIMM-based Memory System\n\nNear-data processing (NDP), a solution to r
 educe data movement overhead between host and memory, should not interfere
  with host access to ensure system fairness. We propose a cost-effective a
 nd energy-efficient LRDIMM-based NDP architecture (HAIL-DIMM) that can sea
 mlessly interleave NDP and regular ...\n\n\nMinkyu Lee, Sang-Seol Lee, Kyu
 ngho Kim, Eunchong Lee, and Sung-Joon Jang (Korea Electronics Technology I
 nstitute)\n---------------------\nAIG-CIM: A Scalable Chiplet Module with 
 Tri-Gear Heterogeneous Compute-in-Memory for Diffusion Acceleration\n\nThe
  emergence of Diffusion models has gained significant attention in the fie
 ld of Artificial Intelligence Generated Content. While Diffusion demonstra
 tes impressive image generation capability, it faces hardware deployment c
 hallenges due to its unique model architecture and computation requirement
 ....\n\n\nYiqi Jing, Meng Wu, Jiaqi Zhou, Yiyang Sun, Yufei Ma, Ru Huang, 
 Tianyu Jia, and Le Ye (Peking University)\n\nTopic: Design\n\nKeyword: In-
 memory and Near-memory Computing Architectures, Applications and Systems\n
 \nSession Chairs: Yannan Nellie Wu (Google) and Win-San Vince Khwa (TSMC)
END:VEVENT
END:VCALENDAR