What I've done

  • Coupling code to and from MEDUSA (with Richard Hill)
  • Restartability for UKESM model (with Richard Hill)
    • NRUN + NRUN = NRUN + CRUN
    • NRUN + CRUN = LRUN, e.g. 2 * 10 day cycle = 1 * 20 day cycle
    • One thread = Two threads
    • Lots and lots of bug fixes to get this to work
      • Especially MEDUSA
  • I've built a lot of the proto-type UKESMs
    • UKESM N96 ORCA1 with MEDUSA
    • UKESM AMIP
    • N216-ORCA025, N96-ORCA025, N48-ORCA1
  • Improved performance of UKESM
    • I added OpenMP into most of chemistry (~10% faster)
      • Chemistry solver
      • Photolysis
      • Emissions
    • I pushed for
      • Remove climate meaning (~20% faster)
      • Huge pages, improvement to I/O server settings, MPICH settings (~10% faster)