[Mitgcm-support] Re: [Fwd: calendar]

[mitgcm-support at dev.mitgcm.org]

Benny,

Benny Cheng wrote:
> 
> Nothing wrong with CAL_SET itself.
> You see, I was using EXTERNAL_FIELD_LOAD, instead of the EXF package,
> so the calendar macros are commented out. BUT... some residual
> calendar routines are still present, so they crash the code.
> BECAUSE.... cal_set is not set!

Ok, good to know.
I may try to properly disentangle these two packages for a 
future version (right now cal is only used in conjunction with exf).

> Anyways, I switch over the EXF package, that solves the problem.
> 
> Any answers to my question about the checkpoint file as global file ?

As to my knowledge, the old checkpoint.F routine
which you probably have at JPL (checkpoint18 or so?)
has been replaced (instead of complemented) by a new routine
which reads/writes pickups for each variable individually.

A solution for you would be to
- either replace the c43 checkpoint.F routine by your old one
  which reads/writes a single file (not recommended), or
- to actually add your routine as an additional option
  (e.g. create checkpoint_single.F     -> your old routine
               checkpoint_multiple.F   ->   current c43 routine
  create corresponding runtime flags 
  (to be added to PARAMS.h, ini_parms.F, set_defaults.F:
   useCheckpointSingle, useCheckpointMultiple)
  and communicate the changes back to us so we can
  add them to a later checkpoint (recommended).

Patrick

> - Benny
> 
> Patrick Heimbach wrote:
> >
> > Benny,
> >
> > what was the problem with cal_set?
> >
> > Patrick
> >
> > Benny Cheng wrote:
> > >
> > > Thanks. That was a good clue. Now it works after cal_set is set!.
> > >
> > > Another question, can the checkpoint files (called pickup ) be a
> > > single global file ? I see no support for this... must it be
> > > a separate file for each individual processor under MPI ?
> > >
> > > - Benny