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:20240626T180035Z
LOCATION:3003\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240625T110000
DTEND;TZID=America/Los_Angeles:20240625T111500
UID:dac_DAC 2024_sess120_RESEARCH016@linklings.com
SUMMARY:GDR-HGNN: A Heterogeneous Graph Neural Networks Accelerator Fronte
 nd with Graph Decoupling and Recoupling
DESCRIPTION:Research Manuscript\n\nRunzhen Xue (State Key Lab of Processor
 s, Institute of Computing Technology, Chinese Academy of Sciences); Mingyu
  Yan (Institute of Computing Technology, Chinese Academy of Sciences); Den
 gke Han and Yihan Teng (State Key Lab of Processors, Institute of Computin
 g Technology, Chinese Academy of Sciences); and Zhimin Tang, Xiaochun Ye, 
 and Dongrui Fan (Institute of Computing Technology, Chinese Academy of Sci
 ences)\n\nHeterogeneous Graph Neural Networks (HGNNs) have broadened the a
 pplicability of graph representation learning to heterogeneous graphs. How
 ever, the irregular memory access pattern of HGNNs leads to the buffer thr
 ashing issue in HGNN accelerators.\n\nIn this work, we identify an opportu
 nity to address buffer thrashing in HGNN acceleration through an analysis 
 of the topology of heterogeneous graphs. To harvest this opportunity, we p
 ropose a graph restructuring method and map it into a hardware frontend na
 med GDR-HGNN. \nGDR-HGNN dynamically restructures the graph on the fly to 
 enhance data locality for HGNN accelerators.\nExperimental results demonst
 rate that, with the assistance of GDR-HGNN, a leading HGNN accelerator ach
 ieves an average speedup of 14.6$\times$ and 1.78$\times$ compared to the 
 state-of-the-art software framework running on A100 GPU and itself, respec
 tively.\n\nTopic: AI, Design\n\nKeyword: AI/ML Architecture Design\n\nSess
 ion Chairs: Andrey Ayupov (Google) and Subhendu ROY (Cadence Design System
 s, Inc.)
END:VEVENT
END:VCALENDAR