FABM 1.0 released

[Jorn Bruggeman]

Hi all,

 

We are happy to announce the release of FABM 1.0. This a major release that includes changes in nearly every part of the code. It foremost improves consistency and clarity for developers, but it also benefits users as it offers modest improvements in performance.

 

FABM 1.0 is not fully backward compatible: hydrodynamic models that couple to FABM will need an update. Biogeochemical model code will typically work as-is, unless it uses conventions from before 2015. Details at https://github.com/fabm-model/fabm/wiki/FABM-1.0.

 

As usual, key information can be found on the wiki (http://fabm.net/wiki). A few things are worth noting:

• It will take some time before all models that couple to FABM are updated. The 0d driver, pyfabm and GOTM (but not GOTM-lake) are already FABM 1.0 compatible.
• The wiki homepage now shows hydrodynamic/biogeochemical models coupled to FABM, as well as FABM users. If you want to see your models or institute mentioned there, let me know.
• FABM releases can now be cited by DOI: https://doi.org/10.5281/zenodo.3774497

 

Any improvements in performance will be highly specific to the models, the platform and the compiler used. Here an impression of the changes seen for a complex model, ERSEM, with FABM’s host emulator compiled with ifort:

 

host    change in runtime (%)

0d      -29 %

brom    -18 %

fvcom   -25 %

getm    -15 %

glm     -18 %

gotm    -23 %

hycom   -2 %

mom4p0d -51 %

mom4p1  -18 %

mom5    -17 %

nemo    -11 %

python  -31 %

roms    -16 %

schism  -10 %

 

NB these are changes in compute time for the source calculations only. In practice you will see less of a change, since source terms are only one part of a model simulation (transport terms often dominate runtime).

 

A few biogeochemical models included with FABM have not yet been updated for the 1.0 release. We expect this to happen over the coming weeks. After this is complete, we will release a minor, backward-compatible update, likely 1.1. If you have any biogeochemical models or couplers (fabm_driver.h) you would like to see included in this next release, please let me know.

 

If you run into issues with the 1.0 release, please post these to one of the mailing lists.

 

Best,

 

Jorn and Karsten