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:20230712T175212Z
LOCATION:3010\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20230711T141000
DTEND;TZID=America/Los_Angeles:20230711T142500
UID:dac_DAC 2023_sess156_RESEARCH761@linklings.com
SUMMARY:A Convolution Neural Network Accelerator Design with Weight Mappin
 g and Pipeline Optimization
DESCRIPTION:Research Manuscript\n\nLixia Han, Peng Huang, Zheng Zhou, Yiya
 ng Chen, Xiaoyan Liu, and Jinfeng Kang (Peking University)\n\nIn-memory co
 mputing (IMC) convolution neural network (CNN) accelerators overcome von N
 eumann's bottleneck by in-situ computing. While the improvement of energy 
 efficiency and throughput suffers from challenges from the access and move
 ment of input and output data. A CNN accelerator design with overall syste
 m integration insight is desired. In this work, we propose a CNN accelerat
 or design named MDopt that fully considers buffer access energy and latenc
 y. It includes a novel CNN mapping method and a pipeline optimization sche
 me. Experiments show that MDopt can reduce 49% and 22% energy consumption 
 and improve 7.0х and 12.6х throughput for ResNet18 and VGG19, respectively
 .\n\nTopic: Design\n\nKeyword: In-memory and Near-memory Computing Archite
 ctures, Applications and Systems\n\nSession Chairs: Laura Fick (Mythic) an
 d Biresh Kumar Joardar (Duke University)
END:VEVENT
END:VCALENDAR