Version 8 (modified by rblod, 14 years ago) (diff) |
---|
Question for the Developper committee
Last edited Timestamp?
Strategic questions
- transition from a vector optimised code with reasonable mpp performances to a mpp optimised code with reasonable vector performances. Agreement ?
- switch to dynamical allocation Agreement ?
- vvl case as the default option : introduction of a key_linssh for linear free surface (=constant volume) Agreement ?
- embedded sea ice default. Levitating case as namelist option or suppression of levitating case Agreement ?
- Off Line priority ?
- Separation between the initialisation and the time-stepping phase: call of the xxx_yyy_init routine at the initialisation phase, not as today at the first call of step. Needed for the adjoint
- How do we manage open boundaries condition : BDY and OBC, suppression of OBC?, merge of the two?
Open questions
- How to manage sea-ice in mpp : introduce another domain decomposition for sea-ice? How is it done in CICE?
- surf module en stand Alone? both to have stand alone bulk and stand alone sea-ice model.
- unequal i-j partition?
- time splitting at high latitudes to save ~20-30% of a global high resolution configuration?
- space averaged output for very high resolution configurations?
- Update the Equation of State with a faster and accurate one ? Priority?
- Non-boussinesq version of the code ? Need, priority??
- Inclusion of academic test cases (Kelvin, soliton ...)? Need, priority??
- Inclusion of timing for mpp? Need, priority??
- Introduce nstock as a list rather than a frequency ? Need, priority??
- Introduce MPI_TYPES from PRACE? Need, priority??