[MITgcm-support] ifort/g77 issue

[Patrick Heimbach]

Hi Tom,

seems a bit tricky.
My understanding is, it's not just a recent checkpoint,
it also occurs with older checkpoints.
So what seems new is your specific config.(?).
I would try to gradually simplify the setup to see
up to when the problem persists. In particular:
* is it advection-scheme dependent (try simplest, which is 2)
* does it depend on topography (a numerical rounding that in
   one setup may have a partial cell, in the other a full cell),
   i.e. what happens if you change the partial cell parameters
* which extra packages are you using (kpp, gmredi, seaice,...)?

For ifort, maybe try using compiler option '-check all -fpe0 -traceback'
to trap non-initialized variables or floating point exceptions.

-Patrick



On Jan 22, 2008, at 11:13 AM, Thomas Haine wrote:

> Hi Martin, OLx/y=3 works fine in g77, but the ifort code dies very
> quickly. See attached output files (no mpi, Olx,y=3). The same also
> happens with no forcing.
>
> The output shows some diffs in Ro_surf, but they look small. Perhaps
> something is different in the grid/mask setup, however.
>
> Any ideas about what I should try next?
>
> Thanks, Tom.
>
>
> On Mon, 2008-01-21 at 09:13 +0100, Martin Losch wrote:
>> Hi Tom,
>> is it possible, that your configuration/parameters require Olx/y = 4,
>> but we don't test for it (e.g. with advection scheme 7, that's the
>> case as far as I know)?
>> Maybe there is some initialization issue, that is handled differently
>> by ifort and g77.
>>
>> Martin
>> On 18 Jan 2008, at 19:00, Thomas Haine wrote:
>>
>>> Hi Folks,
>>>
>>> I have an ifort/g77 issue on dolphin (x86_64, linux
>>> 2.6.9-55.0.6.ELsmp)
>>> using checkpoint59m (and an old checkpoint too):
>>>
>>> Building with linux_amd64_ifort I get theta blowing up in 3-10
>>> steps (in
>>> 1 step it rises to O(10^3)). The exact details are not reproducible.
>>>
>>> Building with linux_amd64_g77 gives no problem.
>>>
>>> I get similar problems with ifort+mpi, where increasing OLx,OLY from
>>> 3->4 seems to stabilise the problem (not fully explored).
>>>
>>> I have: gcc version 3.4.6 20060404 (Red Hat 3.4.6-8), ifort Version
>>> 10.0. My pkg/data configuration is very simple.
>>>
>>> Any ideas?
>>>
>>> Cheers, Tom.
>>> --  
>>> --------------------------------------------
>>>  Thomas W. N. Haine, Professor,
>>>  Department of Earth & Planetary Sciences,
>>>  329 Olin Hall, The Johns Hopkins University,
>>>  Baltimore, MD 21218, USA.
>>>  Tel : 410 516 7048,   Fax : 410 516 7933
>>>  Thomas.Haine at jhu.edu
>>>  http://www.jhu.edu/~eps/faculty/haine
>>> --------------------------------------------
>>> _______________________________________________
>>> MITgcm-support mailing list
>>> MITgcm-support at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>
>>
> --  
> --------------------------------------------
>  Thomas W. N. Haine, Professor,
>  Department of Earth & Planetary Sciences,
>  329 Olin Hall, The Johns Hopkins University,
>  Baltimore, MD 21218, USA.
>  Tel : 410 516 7048,   Fax : 410 516 7933
>  Thomas.Haine at jhu.edu
>  http://www.jhu.edu/~eps/faculty/haine
> -------------------------------------------- 
> <output_g77.txt><output_ifort.txt>____________________________________ 
> ___________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support

---
Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS 54-1518 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach