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:20260421T090513Z
LOCATION:Bldg. 6 - Room 103
DTSTART;TZID=Europe/Stockholm:20260701T100000
DTEND;TZID=Europe/Stockholm:20260701T103000
UID:submissions.pasc-conference.org_PASC26_sess153_msa171@linklings.com
SUMMARY:Co-Designing Performance Portability for ICON: An OpenACC to Kokko
 s Journey
DESCRIPTION:Pradipta Samanta (German Climate Computing Centre)\n\nModern h
 eterogeneous HPC systems, which include multi-core CPUs and GPU accelerato
 rs from various vendors, require portability strategies that go beyond sin
 gle-architecture optimization. This is particularly challenging for large 
 production models like ICON, as it requires a careful balancing act betwee
 n respecting decades of Fortran heritage and adopting modern portability a
 bstractions. We tackle this by taking an incremental approach, migrating I
 CON's physics and mathematics components from Fortran/OpenACC to C++/Kokko
 s without disrupting operational workflows.\n\nA key requirement for this 
 work is a dedicated build system infrastructure that integrates C++/Kokkos
  into ICON's Fortran-based ecosystem, forming the foundation for current a
 nd future component migrations. Building on this, we demonstrate the incre
 mental rewriting strategy through two representative components. The math 
 library in ICON, iconmath, which covers interpolation, divergence, rotatio
 n, gradient, and Laplacian operators, represents the migration of core mat
 hematical routines. The graupel microphysics scheme from the AES physics p
 ackage represents the migration of a physical parameterization. In both ca
 ses, Kokkos provides a unified programming model that supports CPUs and GP
 U backends, including CUDA and HIP, providing measurable performance porta
 bility across different architectures. Together, these serve as a replicab
 le co-design template for the wider Earth-system modeling community.\n\nDo
 main: Climate, Weather, and Earth Sciences, Computational Methods and Appl
 ied Mathematics\n\nSession Chairs: Georgiana Mania (DKRZ) and Xavier Lapil
 lonne (MeteoSwiss)\n\n
END:VEVENT
END:VCALENDAR