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:20241118T160700
DTEND;TZID=America/New_York:20241118T161300
UID:submissions.supercomputing.org_SC24_sess751_ws_p3hpc119@linklings.com
SUMMARY:Exploring SYCL for batched kernels with memory allocations
DESCRIPTION:Aymeric Millan, Thomas Padioleau, and Julien Bigot (Maison de 
 la Simulation, Atomic Energy and Alternative Energies Commission (CEA))\n\
 nBatched parallelism with local allocations is an extremely common pattern
  in HPC, appearing in multi-dimensional FFTs, neural networks processing, 
 or split computation of numerical operators.\nIts efficient support is esp
 ecially complex on GPU where memory per work-item is limited and dynamic m
 emory allocations are challenging.\n\nThis study investigates whether the 
 native abstractions of SYCL can support performance portability for this p
 attern. We implement versions of a batched semi-Lagrangian advection kerne
 l using each parallel construct of SYCL. We evaluate them in terms of main
 tainability, performance portability and memory footprint on CPUs and GPUs
  (AMD, Intel, NVIDIA), with two distinct SYCL implementations (AdaptiveCpp
  and DPC++). Our results demonstrate that no single parallel construct of 
 SYCL emerges as best solution and that a construct offering a higher level
  of abstraction would be required to support this common pattern.\n\nTag: 
 Performance Optimization, Programming Frameworks and System Software\n\nRe
 gistration Category: Workshop Reg Pass\n\nSession Chairs: CJ Newburn (NVID
 IA Corporation), Scott J. Parker (Argonne National Laboratory (ANL)), John
  Pennycook (Intel Corporation), and Kenneth Weiss (Lawrence Livermore Nati
 onal Laboratory (LLNL))\n\n
END:VEVENT
END:VCALENDAR