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:20260422T143140Z
LOCATION:B306
DTSTART;TZID=America/New_York:20241117T115000
DTEND;TZID=America/New_York:20241117T121000
UID:submissions.supercomputing.org_SC24_sess734_ws_pawatm107@linklings.com
SUMMARY:Mitigating synchronization bottlenecks in high-performance actor-m
 odel-based software
DESCRIPTION:Kyle Klenk and Mohammad Mahdi Moayeri (University of Saskatche
 wan, Canada); Junwei Guo and Martyn P. Clark (University of Calgary, Canad
 a); and Raymond J. Spiteri (University of Saskatchewan, Canada)\n\nBulk sy
 nchronous programming (in distributed-memory systems) and the\n  fork-join
  pattern (in shared-memory systems) are often used for\n  problems where i
 ndependent processes must periodically\n  synchronize. Frequent synchroniz
 ation can greatly undermine the\n  performance of software designed to sol
 ve such problems. We use \n  the actor model of concurrent computing to ba
 lance the load of \n  hundreds of thousands of short-lived tasks, and miti
 gate \n  synchronization bottlenecks by buffering communication via actor\
 n  batching. The actor model is becoming increasingly popular in scientifi
 c and\n  high-performance computing because it can handle heterogeneous ta
 sks\n  and computing environments with enhanced programming flexibility an
 d\n  ease relative to conventional paradigms like MPI. For a hydrologic \n
   simulation of continental North America with over 500,000 elements, \n  
 the proposed buffering approach is approximately 4 times faster than \n  n
 o buffering, outperforms MPI on single and multiple nodes, and \n  remains
  competitive with OpenMP on a single node and MPI+OpenMP on \n  multiple n
 odes.\n\nTag: Heterogeneous Computing, Parallel Programming Methods, Model
 s, Languages and Environments, PAW-Full, PAW-Abstract, Task Parallelism\n\
 nRegistration Category: Workshop Reg Pass\n\nSession Chairs: Engin Kayrakl
 ioglu (Hewlett Packard Enterprise (HPE)), Daniele Lezzi (Barcelona Superco
 mputing Center (BSC)), Karla Vanessa Morris Wright (Sandia National Labora
 tories), Irene Moulitsas (Cranfield University), Elliott Slaughter (SLAC N
 ational Accelerator Laboratory), and Kenjiro Taura (The University of Toky
 o)\n\n
END:VEVENT
END:VCALENDAR