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:20260202T201258Z
LOCATION:Second Floor Atrium
DTSTART;TZID=America/Chicago:20251121T080000
DTEND;TZID=America/Chicago:20251121T120000
UID:submissions.supercomputing.org_SC25_sess620_post176@linklings.com
SUMMARY:Learning To Select Scheduling Algorithms in OpenMP
DESCRIPTION:Jonas H. Müller Korndörfer (University of Bern, University of 
 Basel); Ali Mohammed and Ahmed Eleliemy (HPE HPC/AI EMEA Lab); Quentin Gui
 lloteau (Inria); and Reto Krummenacher and Florina Ciorba (University of B
 asel)\n\nScientific and data science applications demand increasing comput
 ational performance, requiring effective scheduling and load balancing on 
 high performance computing (HPC) systems. While OpenMP libraries such as L
 B4OMP provide several scheduling algorithms, selecting the best one for a 
 given application-system pair remains an open challenge. This work address
 es the scheduling algorithm selection problem by investigating automated a
 pproaches that can adapt to diverse workloads and architectures.\n\nWe pro
 pose and evaluate two automated selection strategies: expert- and reinforc
 ement learning-based (RL). We use six applications and three systems to co
 nduct the performance evaluation, revealing trade-offs between exploration
  overhead and optimality of selection of the methods. We further demonstra
 te that combining expert knowledge with RL improves overall performance.\n
 \nWith the poster we will present the methodology, results, and insights o
 f expert- versus RL-based approaches. We highlight implications for future
  heterogeneous and multi-level systems and advertising the open source lib
 rary (LB4OMP) where the methods were implemented.\n\nTag: Research & ACM S
 RC Posters\n\nRegistration Category: Technical Program Reg Pass\n\nSession
  Chairs: Kento Sato (RIKEN Center for Computational Science (R-CCS)); Anja
  Gerbes (Georg-August-Universität Göttingen); and Chris Schlipalius (Pawse
 y Supercomputing Research Centre; Commonwealth Scientific and Industrial R
 esearch Organisation (CSIRO), Australia)\n\n
END:VEVENT
END:VCALENDAR