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:20181221T160727Z
LOCATION:D166
DTSTART;TZID=America/Chicago:20181112T110000
DTEND;TZID=America/Chicago:20181112T113000
UID:submissions.supercomputing.org_SC18_sess173_ws_espm103@linklings.com
SUMMARY:Design of Data Management for Multi-SPMD Workflow Programming Mode
 l
DESCRIPTION:Workshop\nAccelerators, Exascale, Parallel Programming Languag
 es, Libraries, and Models, Workshop Reg Pass\n\nDesign of Data Management 
 for Multi-SPMD Workflow Programming Model\n\nDufaud, Tsuji, Sato\n\nAs bot
 h the complexity of algorithms and architecture increase, development of s
 cientific software becomes a challenge. In order to exploit future archite
 cture, we consider a Multi-SPMD workflow programing model. Then, data tran
 sfer between tasks during computation highly depends on the architecture a
 nd middleware used. In this study, we design an adaptive system for data m
 anagement in a parallel programming environment which can express two leve
 l of parallelism. We show how the consideration of multiple strategies bas
 ed on I/O and direct message passing can improve performances and fault to
 lerance in the YML-XMP environment. On a real application with a sufficien
 tly large amount of local data, speedup of 1.36 for a mixed strategy to 1.
 73 for a direct message passing method are obtained compared to our origin
 al design.
URL:https://sc18.supercomputing.org/presentation/?id=ws_espm103&sess=sess1
 73
END:VEVENT
END:VCALENDAR