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:20241120T082410Z
LOCATION:HG F 30 Audi Max
DTSTART;TZID=Europe/Stockholm:20240604T094300
DTEND;TZID=Europe/Stockholm:20240604T094300
UID:submissions.pasc-conference.org_PASC24_sess158_pos125@linklings.com
SUMMARY:P24 - Fully Spectral Dynamo Simulations for Heterogeneous Computin
 g
DESCRIPTION:Poster\n\nGiacomo Gastiglioni, Philippe Marti, and Dmitrii Tol
 machev (ETH Zurich); Daniel Ganellari (ETH Zurich / CSCS); and Andrew Jack
 son (ETH Zurich)\n\nOur CFD framework QuICC, based on a fully spectral met
 hod, has been successfully used for various dynamo simulations in spherica
 l and Cartesian geometries.\nIt runs efficiently on a few thousands of cor
 es using a 2D data distribution based on a distributed memory paradigm (MP
 I).\nIn order to better harness the computing power of current and upcomin
 g HPC systems, which are increasingly based on heterogeneous nodes built f
 rom multi-core processors and accelerators (GPU), we present our work on r
 efactoring the framework to introduce a hybrid distributed and shared memo
 ry parallelization (MPI + X).\nA critical part of this strategy is the ref
 actorization of the nonlinear transform step.\nThe nonlinear transform ste
 p is refactored to be described and manipulated in ''qir'' a bare-bone ins
 truction-oriented language.\nThough passes, the tree is pruned and optimiz
 ed, temporary resources and communication grouping is handled programmatic
 ally.\nA visitor pattern is used both for the instantiation and the dispat
 ch of the correct back-ends at runtime.\nThe refactored tree enables QuICC
  to run full simulations on hybrid machines efficiently.\nThe implementati
 on of ''qir'' and a performance comparison for different back-ends will be
  presented.\n\nSession Chair: Iva Kavcic (Met Office)
END:VEVENT
END:VCALENDAR