BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211207T055408Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211114T103100
DTEND;TZID=America/Chicago:20211114T110000
UID:submissions.supercomputing.org_SC21_sess426_ws_mchpc103@linklings.com
SUMMARY:FreeLunch: Compression-based GPU Memory Management for Convolution
 al Neural Networks
DESCRIPTION:Workshop\n\nFreeLunch: Compression-based GPU Memory Management
  for Convolutional Neural Networks\n\nPatel, Liu, Guan\n\nRecently, there 
 is a trend to develop deeper and wider Convolutional Neural Networks (CNNs
 ) to improve task accuracy. Due to this reason, the GPU memory quickly bec
 omes the performance bottleneck since its capacity cannot keep up with the
  increase of the memory requirement of CNN models. Existing solutions expl
 oit techniques such as swapping and recomputation to accommodate the short
 age of memory. However, they suffer from performance degradations due to e
 ither the limited CPU-GPU bandwidth or the significant recomputation cost.
   This paper proposes a compression-based technique called FreeLunch that 
 actively compresses the intermediate data to reduce the memory footprint o
 f large CNN models. Based on our evaluation, FreeLunch has up to 35% less 
 memory consumption and up to 70% better throughput than swapping and recom
 putation.\n\nTag: Online Only, Architectures, Memory Systems, Parallel Pro
 gramming Languages and Models, System Software and Runtime Systems\n\nRegi
 stration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR