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:20250626T234542Z
LOCATION:B306
DTSTART;TZID=America/New_York:20241117T094500
DTEND;TZID=America/New_York:20241117T100000
UID:submissions.supercomputing.org_SC24_sess734_ws_pawatm101@linklings.com
SUMMARY:Resource Adaptivity at Task-Level
DESCRIPTION:Jonas Posner (University of Kassel, Germany)\n\nTraditional st
 atic resource allocation in supercomputers (jobs retain a fixed set of res
 ources) leads to inefficiencies. Resource adaptivity (jobs can change reso
 urces at runtime) significantly increases supercomputer efficiency.\n\nThi
 s work exploits Asynchronous Many-Task (AMT) programming, which is particu
 larly well suited to adaptivity, thanks to its transparent resource manage
 ment. The AMT runtime system dynamically assigns user-defined small tasks 
 to workers to achieve load balancing and adapt to resource changes.\n\nCon
 tributions of this work include techniques for malleability and evolving c
 apabilities, allowing programs to dynamically change resources without int
 errupting computation. Heuristics for automatic load detection determine w
 hen to start or terminate processes, which is particularly beneficial for 
 unpredictable workloads. Practicality is demonstrated by adapting the GLB 
 library. A generic communication interface enables interaction between pro
 grams and resource managers. Evaluations with a prototype resource manager
  show significant improvements in batch makespan, node utilization, and jo
 b turnaround time for both malleable and evolving jobs.\n\nTag: Heterogene
 ous Computing, Parallel Programming Methods, Models, Languages and Environ
 ments, PAW-Abstract, Task Parallelism\n\nRegistration Category: Workshop R
 eg Pass\n\nSession Chairs: Engin Kayraklioglu (Hewlett Packard Enterprise 
 (HPE)); Daniele Lezzi (Barcelona Supercomputing Center (BSC)); Karla Vanes
 sa Morris Wright (Sandia National Laboratories); Irene Moulitsas (Cranfiel
 d University); Elliott Slaughter (SLAC National Accelerator Laboratory); a
 nd Kenjiro Taura (The University of Tokyo, Japan)\n\n
END:VEVENT
END:VCALENDAR
