[MITgcm-support] namelist; end of file reached without finding group

Jean-Michel Campin jmc at ocean.mit.edu
Mon Jul 13 16:30:10 EDT 2009


Hi Matt,

I am going to check-in your modifs, with a small difference:
TARGET_BGL or TARGET_CRAYXT would need to be specified either as a cpp flag 
 (e.g. in our optfile, by adding -DTARGET_BGL to DEFINES list), 
or by adding "#defined TARGET_BGL" in our CPP_EEOPTIONS.h (and not 
CPP_OPTIONS.h).

I am not going to change "eesupp/src/nml_filter.F" since I don't
find any call to this subroutine. Patrick, may be we should remove it ?

Also, will not be able to really test this, so if you could give it a try.

Thanks,
Jean-Michel

On Mon, Jul 13, 2009 at 09:30:57AM -0700, Matthew Mazloff wrote:
> Yeah, I think this is a poorly behaved mpi barrier issue -- I think I  
> wrote about this earlier
>
> when you have many processors some will over-write scratch before others 
> have finished using it
>
> until it gets fixed by someone, I suggest having each processor write  
> its own scratch -- so you can put the following routines in your code  
> folder and make sure you
> #define TARGET_BGL
> in your CPP_OPTIONS.h
>
> -Matt
>
> if you want try googling for this -- my last email may give more info
>




>
>
>
> On Jul 13, 2009, at 9:09 AM, Michael Schaferkotter (Contractor) wrote:
>
>> greetings all;
>>
>> i updated my production version of MITgcm a few days ago.
>>
>> when i now rerun an experiment i get the following:
>>
>> C--   Input files
>> PGFIO-F-228/namelist read/unit=11/end of file reached without finding 
>> group.
>> File name = scratch1    formatted, sequential access   record = 1
>> In source file ini_parms.f, at line number 1991
>> PGFIO-F-228/namelist read/unit=11/end of file reached without finding 
>> group.
>> File name = scratch1    formatted, sequential access   record = 1
>> In source file ini_parms.f, at line number 1991
>> [NID 1372]Apid 1023803: initiated application termination
>> PGFIO-F-228/namelist read/unit=11/end of file reached without finding 
>> group.
>> PGFIO-F-228/namelist read/unit=11/end of file reached without finding 
>> group.
>> File name = scratch1    formatted, sequential access   record = 1
>>
>>
>>   1988       WRITE(msgBuf,'(A)') 'S/R INI_PARMS ; starts to read  
>> PARM01'
>>   1989       CALL PRINT_MESSAGE( msgBuf, standardMessageUnit,
>>   1990      &                    SQUEEZE_RIGHT, myThid )
>>   1991       READ(UNIT=iUnit,NML=PARM01) !,IOSTAT=errIO)
>>   1992       IF ( errIO .LT. 0 ) THEN
>>   1993        WRITE(msgBuf,'(A)')
>>   1994      &  'S/R INI_PARMS'
>>
>>
>> this is an mpi compiled executable on an xt5.
>>
>> there was a similar issue in the past, in which a list member  
>> suggested a fix involving editting
>>
>> eeset_parms.F and open_copy_data_file.F. the problem was with the  
>> scratch files for the namelists.
>>
>> in my code directory, i still have:
>>
>> eeset_parms.F and open_copy_data_file.F
>>
>> any ideas on this?
>>
>> michael
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>

> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list