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:20260202T201806Z
LOCATION:265
DTSTART;TZID=America/Chicago:20251116T163000
DTEND;TZID=America/Chicago:20251116T170000
UID:submissions.supercomputing.org_SC25_sess226_ws_ross105@linklings.com
SUMMARY:Reproducible Performance Evaluation of OpenMP and SYCL Workloads u
 nder Noise Injection
DESCRIPTION:Christoffer Persson, Mathias Pretot, Minyu Cui, and Miquel Per
 icas (Chalmers University of Technology and University of Gothenburg)\n\nP
 erformance instability caused by unpredictable system noise remains a pers
 istent challenge in high-performance and parallel computing. This work pre
 sents a reproducible methodology to characterize this variability through 
 noise injection, tested using workloads implemented in OpenMP and SYCL to 
 compare their performance resilience under noisy conditions. We design a n
 oise injector that captures real system traces and replays the deltas as c
 ontrolled noises. Using this approach, we evaluate multiple mitigation eff
 orts, that is, thread pinning, housekeeping core isolation, and simultaneo
 us multithreading (SMT) toggling, under both default and noise-injected ex
 ecutions.  Experiments with two benchmarks (N-body, Babelstream) and one m
 ini-application (MiniFE) across two processor platforms show that while Op
 enMP consistently achieves higher raw performance, SYCL tends to exhibit g
 reater resilience in noisy environments. Mitigation effectiveness varies w
 ith workload characteristics, system configuration, and noise intensity, w
 ith housekeeping core isolation offering the clearest benefits, particular
 ly in high-noise scenarios.\n\nRecording: Livestreamed, Recorded\n\nRegist
 ration Category: Technical Program Reg Pass, Workshop Reg Pass\n\nSession 
 Chairs: Kurt B. Ferreira (Sandia National Laboratories, University of New 
 Mexico); Balazs Gerofi (Intel Corporation, RIKEN Center for Computational 
 Science (R-CCS)); Torsten Hoefler (ETH Zürich, Microsoft Corporation); and
  John Lange (Oak Ridge National Laboratory (ORNL), University of Pittsburg
 h)\n\n
END:VEVENT
END:VCALENDAR