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:20220812T074335Z
LOCATION:Nairobi Room
DTSTART;TZID=Europe/Stockholm:20220629T113000
DTEND;TZID=Europe/Stockholm:20220629T120000
UID:submissions.pasc-conference.org_PASC22_sess161_msa242@linklings.com
SUMMARY:Development of Multi-GPU Solvers for Nonlinear Multi-Physics with 
 Julia
DESCRIPTION:Minisymposium\n\nDevelopment of Multi-GPU Solvers for Nonlinea
 r Multi-Physics with Julia\n\nOmlin, Räss, Utkin, Narayanan, Senden\n\nWe 
 present an efficient approach for the development of massively scalable hi
 gh performance multi-GPU solvers for nonlinear multi-physics with Julia. P
 owerful costless abstractions, multiple dispatch and metaprogramming enabl
 e to write a single code that is usable for both productive protoyping on 
 a single CPU and for large scale production runs on GPU or CPU supercomput
 ers. High performance stencil computations are expressable with math-close
  notation in hardware-agnostic compute kernels for which launch parameters
  can be automatically derived from the kernel arguments. Halo updates need
 ed for distributed memory parallelization require only a simple function c
 all, are fully overlapable with computations and achive performance close 
 to hardware limit. We demonstrate the wide applicability of our approach b
 y reporting on several multi-GPU solvers for geosciences as, e.g., 3-D sol
 vers for poro-visco-elastic twophase flow and for reactive porosity waves.
  As reference, the latter solvers were ported from MPI+CUDA C to Julia and
  achieve 90% and 98% of the performance of the original solvers, respectiv
 ely, and a nearly ideal parallel efficiency on thousands of NVIDIA Tesla P
 100 GPUs at the Swiss National Supercomputing Centre. We further show that
  the approach is applicable beyond geosciences, showcasing a computational
  cognitive neuroscience application modelling visual target selection usin
 g HPC.\n\nDomain: Climate, Weather and Earth Sciences, Computer Science an
 d Applied Mathematics, Physics
END:VEVENT
END:VCALENDAR