BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:How not to write a convection parameterisation code? - Mike Whital
 l - UK Meteorological Office
DTSTART:20231123T130000Z
DTEND:20231123T140000Z
UID:TALK204715@talks.cam.ac.uk
CONTACT:Jack Atkinson
DESCRIPTION:Convection parameterisations are a crucial element of global a
 tmospheric models.  They simulate the vertical transport of heat\, moistur
 e and momentum by convective clouds\, and associated rainfall.  The majori
 ty of Tropical rainfall is associated with these clouds\, which are too sm
 all-scale to explicitly resolve on the model’s grid and so need to be pa
 rameterised.\n\nMost global atmosphere models use a so-called “mass-flux
 ” form of convection parameterisation\, which consists of a diagnostic v
 ertical integral to compute the properties of the clouds and the amount of
  heat / moisture entrained / detrained at each height.  The calculations i
 n a given vertical column of model grid-points are completely independent 
 of those in other neighbouring columns\, so it would be simplest to write 
 the code so that it only considers a single column at a time.  However\, s
 ince we are performing a vertical integral the calculations at a given hei
 ght within each column depend on the results from those calculations at th
 e level below\, so the scheme must be structured in a vertically sequentia
 l manner considering a single height-level at a time.\n\nConsidering only 
 a single column and a single height-level at a time amounts to computing o
 nly a single grid-point at a time.  On CPU architectures\, this is extreme
 ly inefficient\; far greater computation speeds are obtained by doing many
  identical calculations simultaneously\, via vectorisation.  Another chall
 enge / opportunity is the sparsity of the required calculations\, since co
 nvective clouds only occupy a small fraction of the atmosphere’s volume.
 \n\nIn this talk I discuss routes to exploiting both vectorisation and sha
 red memory parallelisation\, and how to make efficient use of memory given
  the sparsity\, in the comorph convection parameterisation fortran code cu
 rrently under development at the Met Office.\n\nHowever\, current and futu
 re changes in software and HPC architectures (such as GPUs) may radically 
 change the optimal code structure.  Is there any way to adapt our convecti
 on code to these changes without completely rewriting it\, or write it in 
 a “future proof” way?\n\nHybrid Zoom details will be emailed to the RS
 E mailing list\; if you are not on the list\, please contact the organiser
 s.
LOCATION:JJ Thomson Seminar Room\, Maxwell Centre
END:VEVENT
END:VCALENDAR