BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250822T115805Z
LOCATION:Campussaal - Plenary Room
DTSTART;TZID=Europe/Stockholm:20250617T103000
DTEND;TZID=Europe/Stockholm:20250617T110000
UID:submissions.pasc-conference.org_PASC25_sess150_pos107@linklings.com
SUMMARY:P41 - SYCL and Block-Structured Grids: Performance Impact on Simul
 ations of Complex Costal Ocean Domains
DESCRIPTION:Jonathan Schmalfuß (University of Bayreuth), Daniel Zint (New 
 York University), Sara Faghih-Naini (ECMWF), Julian Stahl (Friedrich-Alexa
 nder-Universität Erlangen-Nürnberg), Markus Büttner (University of Bayreut
 h), Roberto Grosso (Friedrich-Alexander-Universität Erlangen-Nürnberg), an
 d Vadym Aizinger (University of Bayreuth)\n\nDeveloping the next generatio
 n of climate modelling tools to increase throughput and ensure performance
  portability is crucial. The choice of an underlying grid for ocean modell
 ing, an important climate compartment, is difficult. The almost fractal-li
 ke boundaries of ocean domains and quickly changing bathymetry often make 
 unstructured triangular meshes the preferred choice. Their construction is
  simple and adaptability high. An alternative is Block-Structured Grids (B
 SG), an unstructured collection of blocks, each containing a topologically
  structured mesh.  We present the performance impact of utilizing BSG on d
 iverse hardware employing SYCL and the current state of BSG generation. Co
 mputation on unstructured grids is associated with suboptimal performance 
 due to irregular memory access patterns, whereas structured grids enable n
 ear-optimal efficiency. Our shallow water equations solver UTBEST exploits
  the regular memory access pattern on a per-block basis provided by the BS
 Gs to enable performance gains. This impact is studied with respect to the
  block size and the influence of unstructured blocks. As SYCL allows progr
 amming for heterogeneous parallel computing in C++ on CPU, GPUs (and FPGAS
 ), two major SYCL implementations (oneAPI, AdaptiveCpp) with multiple back
 ends CUDA, OpenMP, OpenCL are evaluated.\n\nSession Chair: David Moxey (Ki
 ng's College London)\n\n
END:VEVENT
END:VCALENDAR