[Aces-support] opt files

[Matthew Mazloff]

OK...sorry for the spam...I will try to be less jumpy to conclusions in 
the future

The mpich environment is not what causes my model to crash...it appears 
my model crashes occasionly and the problem is that it is not reading 
correctly from a linked data folder (linked to scratch-1)....this seems 
to be a somewhat random occurence (except for the 4am prob which is very 
predictable). 

Matt




Matthew Mazloff wrote:

> Hello,
>
> I have figured out what was causing my problem.  Even though I have:
>
> if ( -f /etc/profile.d/modules.csh ) then
>    source /etc/profile.d/modules.csh
> endif
> #
> module add mpich/intel
>
> in my runscript, if I don't use
>
> module add mpich/intel
>
> at least once sometime before I run my script, my model will crash.
>
> Don't know if this is worthy of the support list....but someone might 
> be interested
>
> Thanks,
> Matt
>
>
>
> Matthew Mazloff wrote:
>
>> Hello everyone,
>>
>> I don't understand it, but I just recompiled the model with
>>
>> linux_ia32_ifort+mpi_aces
>>
>> and it is running fine...
>>
>> Sorry about that, I must have done something wrong before...
>>
>> Matt
>>
>>
>> Patrick Heimbach wrote:
>>
>>> Hi Matt,
>>>
>>> re. 1), order doesn't matter.
>>>
>>> I just ran the global_ocean.90x40x15
>>> on 2 processors under 'module load mpich/intel'
>>> and it ran fine.
>>> Have a look under
>>> /home/heimbach/ecco/MITgcm/verification/global_ocean.90x40x15/build/
>>> for the compile and output
>>> and
>>> /home/heimbach/ecco/MITgcm/jobs/
>>> for the job script (which is basically yours).
>>>
>>> For compile, I did
>>> ../../../tools/genmake2  -of 
>>> ../../../tools/build_options/linux_ia32_ifort+mpi_aces -mpi -mods 
>>> ../code
>>>
>>> Maybe a different problem?
>>> For that, a more detailed bug report would be needed
>>> (which file does it try to read when crashing?)
>>>
>>> -Patrick
>>>
>>>
>>>
>>> On Thursday 16 December 2004 14:08, Matthew Mazloff wrote:
>>>  > Hi,
>>>  >
>>>  > I def. used genmake2 -mpi ....
>>>  >
>>>  > 2 things I may have done to mess things up
>>>  >
>>>  > 1) I may have added the  -mpi  in diff. order...would it matter 
>>> if I put
>>>  > the -mpi aftr the -of=...I am not sure if I did this, but it is 
>>> possible
>>>  >
>>>  > 2) I noticed a few extra lines at the end of the
>>>  > 'linux_ia32_ifort+mpi_aces' file that I may have accidentally put
>>>  > there...I know I shouldn't have messed with this file...it was a
>>>  > mistake...would these extra 'returns' have messed things up.  I did
>>>  > check the CVS repository and everything else, besides these extra 
>>> lines
>>>  > is the exact same.  However, I know Fortran is sensitive enough that
>>>  > these lines may have cause the model to crash...if so I am very 
>>> sorry
>>>  > for making people search for an answer...my mistake
>>>  >
>>>  > If 1 or 2 would not have made the model crash, then I really have no
>>>  > idea why it crashed...everything else I did was identical...
>>>  >
>>>  > Matt
>>>  >
>>>  > Patrick Heimbach wrote:
>>>  > >Hi Ed,
>>>  > >
>>>  > >I saw an "inconsistency" as to when we explicitly specify
>>>  > >-DALLOW_USE_MPI -DALWAYS_USE_MPI
>>>  > >in the optfiles and when we don't.
>>>  > >If '+mpi' is included in the optfile name people tend to assume,
>>>  > >rightly so, that -mpi isn't needed as explicit genmake2 option.
>>>  > >
>>>  > >Matt may have run into this.
>>>  > >the ...+mpi_itrda optfile does have the -DALLOW_USE_MPI 
>>> -DALWAYS_USE_MPI
>>>  > >the ...+mpi_aces doesn't.
>>>  > >
>>>  > >Matt, could you check whether you did 'genmake2 -mpi ...' ?
>>>  > >
>>>  > >-Patrick
>>>  > >
>>>  > >On Wednesday 15 December 2004 14:53, Matthew Mazloff wrote:
>>>  > >  > Hello,
>>>  > >  >
>>>  > >  > I wanted to bring to attention, the fact that when I 
>>> compiled the
>>>  > >  > MITGCM with the optfile
>>>  > >  >  linux_ia32_ifort+mpi_aces
>>>  > >  > the model crashed.  By changing to the optfile
>>>  > >  > linux_ia32_ifc+mpi_itrda
>>>  > >  >  everything ran smoothly.  For both I used the  mpich/intel
>>>  > >  > environment.
>>>  > >  >
>>>  > >  > Thanks,
>>>  > >  > Matt
>>>  > >  >
>>>  > >  >
>>>  > >  >
>>>  > >  > 
>>> <http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/tools/build_options/li
>>>  > >  >nux_i a32_ifc%2Bmpi_itrda>
>>>  >
>>>  > _______________________________________________
>>>  > Aces-support mailing list
>>>  > Aces-support at acesgrid.org
>>>  > http://acesgrid.org/mailman/listinfo/aces-support
>>>
>>>  
>>>
>>
>> _______________________________________________
>> Aces-support mailing list
>> Aces-support at acesgrid.org
>> http://acesgrid.org/mailman/listinfo/aces-support
>>
>
> _______________________________________________
> Aces-support mailing list
> Aces-support at acesgrid.org
> http://acesgrid.org/mailman/listinfo/aces-support
>