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:20190719T085744Z
LOCATION:HG D 1.2
DTSTART;TZID=Europe/Stockholm:20190613T144500
DTEND;TZID=Europe/Stockholm:20190613T151500
UID:submissions.pasc-conference.org_PASC19_sess148_msa259@linklings.com
SUMMARY:dawn: A Toolchain for Weather and Climate DSLs Applied to COSMO Dy
 namical Core
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Clima
 te and Weather\n\ndawn: A Toolchain for Weather and Climate DSLs Applied t
 o COSMO Dynamical Core\n\nWicky, Osuna, Fuhrer\n\nWe present a compiler-to
 olchain for domain-specific languages (DSLs) in weather and climate. The l
 arge diversity of modern computing architectures and their respective prog
 ramming models is imposing a great challenge in scientific applications th
 at need to run efficiently on them. DSLs are developed for individual appl
 ications which leads to little reuse between these tools and the performan
 ce optimizations. Our toolchain aims at supporting high-level, descriptive
  DSLs for a wide range of weather and climate models, allowing different c
 ommunities to use it from their own front-ends. A set of domain specific c
 ode analysis and optimizations are performed in the toolchain, which allow
 s the DSL front-ends for different models to exploit the efficient code ge
 neration multiple the architectures. This approach can separate concerns b
 etween model developers and HPC experts. Combining the modularity of the t
 oolchain, the light weight front-end and a more advance code analays leads
  to a significant decrease in maintenance which allows for higher producti
 vity in module development. We implemented the dynamical core of COSMO and
  are able to see significant performance improvement on NVIDIA GPUs over e
 xpert tuned production code in both code size and performance. Productivit
 y of the high-level DSL as well as performance for weather application is 
 evaluated.
END:VEVENT
END:VCALENDAR