[MITgcm-devel] do_the_model_io

[Jean-Michel Campin]

Hi, 

I propose to change the code and to call "do_the_model_io"
also at the beginning (to dump the initial state) in "the_main_loop", 
where "write state" is called.
Then, I think the "call monitor" can also be put in "do_the_model_io".
(for instance at the beginning).

The motivation is for pkg Output and diagnostic:
It's simpler (easy to modify) and clearer when all pkg-Out-diag
are called from the same place, with only 1 (main) call for each. 
This is the case for almost all packages, that call their 
diagnostic-Output part in "do_the_model_io". And this is fine.

But before the 1rst iter (initial state), things are less clear: 
presently pkgs do their Output (if needed) within the 
initialization (package_ini_varia) but for the main state variable, 
it's done in "the_main_loop". It's not very clear but it works.
 
Things becomes tricky for ${pkg}_monitor :
I can put "call ice_monitor" in "ice_diag" and call it in
ice_ini_vars.F (in package_ini_varia) for the 1rst iter
and again in do_the_model_io. But then the 1rst ice_monitor output
will be written before the 1rst main monitor, whereas all the 
others will appears just after the main monitor output.

I don't know exactly what is done for gchem_monitor at the
beginning, but should have similar problems.

adding "call do_the_model_io" in "the_main_loop" will
fix it. And within S/R do_the_model_io, could always put 
"if myIter.ne.nIter0" in front of S/R call(s) that do not need
to be called at the beginning.

Any comments are welcome.

Jean-Michel