Work to come

The next year

• UKESM-hybrid: to a finish
• UKESM-fast: increase top speed from about 5 model years/day to ~20 model years a day, by
  • Removing the full chemistry, i.e. use UKESM-CN
  • Increasing timestep: 20 mins -> 30 mins/45 mins/1 hr
  • Reducing vertical levels: 85 -> 38 (same 38 as used in HadGEM2-ES)
  • Increasing radiation & aerosol timesteps: 1 hr -> 3 hrs
  • Lots more ...
• Support for UKESM (along with the rest of the group)

Next 2-3 years

• Probably optimisations for UKESM model
  • OpenMP for MEDUSA (necessary for ORCA025 at CMIP7)
  • Work stealing for UKCA (pass work from heavily loaded PEs to lightly loaded PEs)
  • Run the full chemistry concurrently with dynamical core by running on its own threads