[MITgcm-devel] sx8 latest testing

[Martin Losch]

Hi Jean-Michel,

finally, I have a "good" test on the SX8 (sometimes I seriously doubt my
scripting capabilities): All experiments compile, run (except for two
fizhi and adjustment.cs-32x32x1.nlfs, the latter is expected, the former
also) and testreport actually realizes that runs were successful.

Now I can focus on iblade ...

Martin

Jean-Michel Campin wrote:
> Hi Martin,
> 
> On Mon, Jan 18, 2010 at 12:06:00PM +0100, Martin Losch wrote:
>> Hi Jean-Michel,
>>
>> I don't know why the job on iblade did not start on the weekend, probably a problem with cron, I started on "by hand" to if anything works.
>>
>> About the SX8: I think I found a solution to the many fails: I reversed the order in which I "cat" "NORMAL END" into run.log and the copying of STDOUT.0000 to output.txt, so that the cp is the last command in the job script. That should do it.
>>
>> You new -norun flag does not work for on the SX8, unfortunately. I do have make on the compute nodes, but that make thinks it needs to relink and regenerate the executable, which spoils everything ... I don't know why I always have to deal with broken installations.
> 
> OK, it could be that the compute node has a different feeling about time,
> and then it's hard to make it change its mind. But what matter is that 
> you managed to make it work the other way.
> 
> Thanks,
> Jean-Michel
> 
>> Martin
>>
>>
>>
>>
>> On Jan 16, 2010, at 11:15 PM, Jean-Michel Campin wrote:
>>
>>> Hi Martin,
>>>
>>> I think we can live with a single output file for diagnostics,
>>> even with binary (MDS) output. Just to remind that the list of
>>> fields are stored in the meta files and can be loaded in matlab
>>> with rdmds.m, so should be OK.
>>>
>>> By the way, I've just check-in the -nr option for testreport,
>>> but still depend on whether or not you can run the command "make"
>>> on compute node when there is nothing to build.
>>>
>>> Thanks,
>>> Jean-Michel
>>>
>>> On Sat, Jan 16, 2010 at 10:55:19PM +0100, Martin Losch wrote:
>>>> Hi,
>>>> back to SX8: I'll change data.diagnostics in offline_exf_seaice:
>>>> about lab_sea: I can change the data.diagnostics, too, but then, the MDS-output in the other three sub-experiments is awkard (although I have no problem with that).
>>>> Or maybe I can supress the standard output for the first experiment by
>>>> dumpFreq = 0.,
>>>> tavefreq = 0.,
>>>> dumpInitAndLast = .false.
>>>>
>>>> What's your opinion?
>>>>
>>>> Martin
>>>>
>>>> On Jan 16, 2010, at 9:13 PM, Jean-Michel Campin wrote:
>>>>
>>>>> Hi Martin,
>>>>>
>>>>> Sorry, I miss the 2nd message you sent the other day (on Jan 12).
>>>>>
>>>>> Regarding point 1:
>>>>> I agree, it would be better to change data.diagnostics to much fewer 
>>>>> files, go for it. 
>>>>> Not sure about turning off mnc for standard lab_sea (since it's the only 
>>>>> forward test with full seaice and MNC). 
>>>>> If you try also to change data.diagnostics in the same way, would this
>>>>> work ?
>>>>
>>>> _______________________________________________
>>>> MITgcm-devel mailing list
>>>> MITgcm-devel at mitgcm.org
>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

-- 
Martin Losch
Alfred Wegener Institute for Polar and Marine Research
Postfach 120161, 27515 Bremerhaven, Germany;
Tel./Fax: ++49(0471)4831-1872/1797