Hybrid progress over last 18 months
- Richard and I have almost entirely been pulled onto the main
UKESM1 project
- Richard has got a 3D coupling branch onto trunk (untested with
hybrid, as no hybrid model since UM10.4).
- Richard has created drivers for the hybrid model
- Ran hybrid model for 5 years
- We had some out-of-memory problems with UM10.4 version.
- Investigation showed that there seemed to be a memory leak in
microphysics
- This is common to all jobs (incl GA + StratTrop & UM10.2 hybrid)
- Hybrid is more likely to suffer this problem because of extra
memory for
- Full chemistry
- Coupling fields
- Hoping that problem is less bad at UM10.9. Otherwise, use more
cores