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 104
DTSTART;TZID=Europe/Stockholm:20260701T150000
DTEND;TZID=Europe/Stockholm:20260701T153000
UID:submissions.pasc-conference.org_PASC26_sess163_msa210@linklings.com
SUMMARY:Performance, Precision and Accuracy: Mixed-precision modelling for
  weather and climate applications at the MetOffice
DESCRIPTION:Christopher Michael Maynard (Met Office)\n\nNumerical calculat
 ions using floating point numbers suffer from accumulated round-off error 
 and moreover, they are not associative. This is problematic for  weather a
 nd climate applications as they rely on global sums in PDE solvers and the
  numerical answer can then depend on the parallel decomposition. Typically
 , 64-bit arithmetic is used in these applications to mitigate the again th
 e lack of robust error propagation in such complex numerical algorithms.\n
 Recently, the performance gains available in reduced precision calculation
 s, primarily driven by ML applications on GPUs, have have provoked interes
 t in reducing the reliance on 64-bit arithmetic. The Met Office operationa
 l model, the Unified Model (UM) uses a a mixed-precision, 32-bit solver ru
 nning on CPUs and the performance and accuracy is described. Moreover, the
  Met office is developing a new atmospheric model, LFRic, which will repla
 ce the UM in the next few years, which  has been developed explicitly as a
  mixed-precision code. The model and flexible precision framework which al
 lows different model components to run in different precision, will be dis
 cussed and performance results shown for 32-bit arithmetic on CPUs. Finall
 y, some preliminary, numerical experiments at 16-bit arithmetic for GPU ar
 chitectures will be discussed before concluding will some remarks on futur
 e directions.\n\nDomain: Climate, Weather, and Earth Sciences\n\nSession C
 hairs: Balthasar Reuter (ECMWF) and Claudia Frauen (DKRZ)\n\n
END:VEVENT
END:VCALENDAR
