[Mitgcm-support] Re: exf

[mitgcm-support at dev.mitgcm.org]

Hi Patrick,
Another issue that we've found, connected with exf, is that the calendar
package runs out of seconds after about 70 years. The problem is that
2^31 seconds is about 68 years, and cal uses integers.  This is not
really long enough for many of our runs.

Our current way round this is with '-i8' but that's not really
satisfactory, especially for IA32 processors...

Is it possible to use exf without cal ? 

John


Patrick Heimbach wrote:
> 
> Hi exf users,
> 
> I am thinking of making a major simplification to the
> exf package which should considerably facilitate the
> inclusion of new fields and features (such as
> constant or cyclic forcing option for each field).
> Right now, every field has its own read routine
> (exf_set_hflux.F, etc.) which is very cumbersome.
> 
> I think that this is historically due to a
> misunderstanding of the problems with 'field swapping'
> (i.e. only read new forcing field when necessary,
> e.g. every 24 hours) in the adjoint.
> I think the reasons why there were problems with the
> adjoint were not understood at the time the package
> was first conceived, and a 'play safe' solution was adopted.
> 
> If there are no objections, I'd like to give it a try
> and see whether the simplification works.
> Please let me know. It's not going to happen immediately,
> but maybe in June.
> 
> Cheers
> -Patrick
>