[MITgcm-devel] Re: it's probably my faults again, but ...

[Dimitris Menemenlis]

Martin, I am transferring your e-mail to devel list as Chris or others may have 
comments.  What makes you think that it is the sea-ice model that causes 
trouble?  I have used the s1500t_17x51/SIZE.h_500 configuration in the past 
successfully but have not done so in a very long time.  What is special about 
this configuration is that there are holes in the domain, i.e., no computations 
take place over some of the land.

 >>> MY QUESTION TO THE DEVEL LIST IS WHETHER ANYONE ELSE HAS USED
 >>> DOMAINS WITH HOLES RECENTLY?

I don't think that this part of the exch2 package is tested on regular basis, so 
it may be broken.  I am in process of rearranging the description of 
experiments, etc., as per your suggestions and there is a small test with holes 
on the 32x6x32 domain that I plan to test and get back.  D.


> ... just to make sure that I am not trying something stupid: After having
> made one day of integration on 216 CPUs (and actually picking up and running
> a second day with 216 CPUs), I have tried using a higher number of CPUs
> (which is more effective on the machine that I am running on), that is the
> SIZE.h_500 in the s1500t directory. So I replaced SIZE.h with SIZE.h_500 and
> w2_ee2setup.F and W2_EXCH_TOPOLOGY.h and recompiled. Then I tried to restart
> from the same pickup, from which I have already successfully started with
> 216CPUs. The model starts (after waiting 4days in the queue) and seems to
> pickup fine, at least the model part.
> 
> [ now it's definitely my fault, that I send this email prematurely, sorry for
> that ]
> 
> I am attaching the stdout, and you can see that something is wrong with the
> seaice model and then the first timestep is already very wrong. My eedata is
> correct this time. And I am using the MULTICATEGORY seaice (all the way, it's
> also in the pickup files), but that shouldn't make a difference, should it?
> 
> So my question is really: Do you regularily use this configuration at all or
> have I made one of my famous mistakes again?
> 
> Martin