BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160557Z
LOCATION:Track 7
DTSTART;TZID=America/New_York:20201113T160000
DTEND;TZID=America/New_York:20201113T163000
UID:submissions.supercomputing.org_SC20_sess226_ws_ross107@linklings.com
SUMMARY:Locality-Aware Scheduling for Scalable Heterogeneous Environments
DESCRIPTION:Workshop\n\nLocality-Aware Scheduling for Scalable Heterogeneo
 us Environments\n\nKamatar, Friese, Gioiosa\n\nHeterogeneous computing pro
 mises to boost performance of scientific applications by allowing massivel
 y parallel execution of computational tasks. Manually managing extremely h
 eterogeneous, multi-device systems is complicated, however, and may result
  in sub-optimal performance. Specifically, data management is an extremely
  challenging problem on multi-device systems. In this work, we introduce t
 wo locality-aware schedulers for the Minos Computing Library (MCL), an asy
 nchronous, task-based programming model and runtime for extremely heteroge
 neous systems. The first scheduler implements a pure locality-aware algori
 thm to maximize data reuse, though it might incur "hot-spots" that limit s
 ystem utilization. The second scheduler mitigates this drawback by dynamic
 ally targeting between locality-awareness and system utilization based on 
 the current workload and available computing devices.\n\nOur results show 
 that locality-awareness greatly benefits applications that exhibit data re
 use, providing up to 6.9x and 7.9x over the original MCL scheduler and equ
 ivalent OpenCL implementations, respectively. Moreover, our schedulers int
 roduce negligible overhead compared with the original MCL scheduler and ac
 hieve similar performance for applications that do not benefit from data l
 ocality.\n\nTag: System Software and Runtime Systems\n\nRegistration Categ
 ory: Workshop Reg Pass
END:VEVENT
END:VCALENDAR