[MITgcm-support] IBM

Fri May 7 11:14:20 EDT 2004

Well, with xlf,  there is a runtime option:
via the environment variable XLFRTEOPTS you can set the name list 
behavior
either
setenv XLFRTEOPTS namelist=old
or
setenv XLFRTEOPTS namelist=new (default)
but in both cases the & as the namelist terminator does not seem to be 
allowed. But as Patrick pointed out the program complains but the 
execution is not affected. The problems I am having stem from something 
different, that must be related to MPI or exchanges in general, because 
I see ridiculous values in Eta (-536 to 8) and W after the first time 
step. It must be something really stupid. Patrick just pointed out to me
usingMPI=.true. in eedata, but that didn't do it, what about the other 
settings in eedata (taken from the verification experiments 
global_ocean.cs32x15)? I don't know anything about that.

Martin

PS. Maybe the attached figure helps (SSH after 1 timestep, plotted 
range is +/-1m, actual range is -580 to 8m)

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ssh_ts1.eps.gz
Type: application/x-gzip
Size: 34951 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20040507/446f0835/attachment.gz>
-------------- next part --------------

On Friday, May 7, 2004, at 04:34 PM, Chris Hill wrote:

> Hmmm, not sure what you mean - we'll take a closer look when we get 
> some IBM
> access back. Let us know what you think needs to be done (if anything) 
> to
> tidy this up.
>
> Thanks,
>
> Chris
>
>> -----Original Message-----
>> From: mitgcm-support-bounces at mitgcm.org
>> [mailto:mitgcm-support-bounces at mitgcm.org] On Behalf Of Martin Losch
>> Sent: Friday, May 07, 2004 10:29 AM
>> To: mitgcm-support at mitgcm.org
>> Subject: Re: [MITgcm-support] IBM
>>
>> Yes, but not for xlf for AIX, there you can only switch between "old"
>> and "new" -style namelists, where "old" means something very
>> IBM-specific.
>>
>> M.
>>
>> On Friday, May 7, 2004, at 04:17 PM, Chris Hill wrote:
>>
>>> Martin/James,
>>>
>>>  There is an options setting these days that does that
>> automatically
>>> for some platforms.
>>>  See tools/build_options/linux_ia32_lf95. The CPP option
>>> -DNML_TERMINATOR makes the switch.
>>>
>>>  BTW - we had been using NCAR for our IBM testing, but they
>> have been
>>> out of service for the last month due to security problems at NCAR.
>>> Ed, Patrick and I will talk about getting some time at BU to do
>>> testing.
>>>
>>> Chris
>>>
>>>> -----Original Message-----
>>>> From: mitgcm-support-bounces at mitgcm.org
>>>> [mailto:mitgcm-support-bounces at mitgcm.org] On Behalf Of James B.
>>>> Girton
>>>> Sent: Friday, May 07, 2004 10:08 AM
>>>> To: mitgcm-support at mitgcm.org
>>>> Subject: Re: [MITgcm-support] IBM
>>>>
>>>> Hi Martin,
>>>>
>>>> I've been using an IBM at Boston University and remember
>> having some
>>>> issues like this at the start. One thing I had to do was
>> change all
>>>> of the namelist termination characters in data, data.obcs,
>> etc., from
>>>> "&"
>>>> to "/".
>>>>
>>>> -James
>>>>
>>>>
>>>> Martin Losch wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> I am supposed to do some test runs on our "IBM Regatta" computer,
>>>>> which is a test computer for a different much bigger IBM system.
>>>>> uname -a gives
>>>>> AIX edvir1 1 5 0010117A4C00
>>>>> and the avaible compiler is xlf, so I guess I should use
>> the sp3 or
>>>>> sp4 build_options file. With sp4, I get this:
>>>>> 1. cc not found (Argh! Apparently there's a gcc and xlc but
>>>> no cc, so
>>>>> there should be a way to include CC='gcc' into the
>> options file) I
>>>>> solved that by modifying the makefile to include CC=gcc
>> or aliasing
>>>>> gcc to cc 2. Code compiles fine, but when I run it I get:
>>>>>
>>>>>> 1525-090 The NAMELIST READ statement cannot be completed
>>>> because an
>>>>>> incorrect character was encountered in a NAMELIST group
>>>> name or item
>>>>>> name.  The program will recover by discontinuing further
>>>> processing
>>>>>> of the READ statement.
>>>>>
>>>>> as often as a namelist is read and then the model
>> explodes after 4
>>>>> timesteps
>>>>>
>>>>> Any experience with this? Would you like me to run
>>>> testscript with the
>>>>> -addr option?
>>>>>
>>>>> Martin
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> MITgcm-support mailing list
>>>>> MITgcm-support at mitgcm.org
>>>>> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
>>>>>
>>>>
>>>> --
>>>> -----------------------------------------------------------
>>>> James B. Girton, Postdoctoral Investigator Department of Physical
>>>> Oceanography, MS 29 Woods Hole Oceanographic Institution,
>> Woods Hole,
>>>> MA 02543 Clark 214B, Phone:
>>>> 508-289-2632, FAX: 508-457-2163 girton at whoi.edu,
>>>> http://ohm.apl.washington.edu/~girton
>>>> -----------------------------------------------------------
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> MITgcm-support mailing list
>>>> MITgcm-support at mitgcm.org
>>>> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
>>>>
>>>
>>> _______________________________________________
>>> MITgcm-support mailing list
>>> MITgcm-support at mitgcm.org
>>> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
>>
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
>>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support