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:20190613T174500
DTEND;TZID=Europe/Stockholm:20190613T181500
UID:submissions.pasc-conference.org_PASC19_sess168_msa189@linklings.com
SUMMARY:Accelerating Cloud Physics and Atmospheric Models using GPUs, KNLs
  and FPGAs
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Clima
 te and Weather, Physics\n\nAccelerating Cloud Physics and Atmospheric Mode
 ls using GPUs, KNLs and FPGAs\n\nBrown\n\nThe Met Office NERC Cloud model 
 (MONC) is an atmospheric model used to study clouds and turbulent flows. T
 his is coupled with the CASIM microphysics model which provides the capabi
 lity to investigate interactions at the millimetre scale and study the for
 mation and development of moisture. Crucially, MONC and CASIM are both hea
 vily computationally intensive, and in this talk I will explore the role t
 hat accelerators might provide for improving the capability of these 
 code. Using OpenACC for GPU acceleration, the approach adopted for MONC, w
 here we were able to isolate specific kernels, differed from the approach 
 for CASIM where we were forced to offload the entirety of this microphysic
 s model due to the tightly coupled nature of the code. Whilst there were s
 everal caveats and lessons learnt, experiments on Piz Daint demonstrated a
  benefit to using GPUs for these models. We will also consider the role ot
 her HPC technologies, namely KNL and FPGAs, might have to play in accelera
 ting cloud physics codes at exascale. For the later, as part of the EXCELL
 ERAT CoE, we have ported aspects of the MONC model onto FPGAs and will det
 ail the challenges, along with performance characteristics, contrasted aga
 inst using CPUs, GPUs and KNL.
END:VEVENT
END:VCALENDAR