[CIG-SEISMO] SPECFEM3D_GLOBE unstable in fluid

[Marcin Zielinski]

Dear Dimitri, others,

I think removing -qsave is not really a fix. It is merely a way around 
and it will work only when the stacksize on the given machine is set by 
default to unlimited, just like on our Power6 machine here in Amsterdam.
Obviously, what you can do to reassure that, is make a little script (bash):
$> cat setstacksize
ulimit -s unlimited
$*

and then run mesher/solver with:
./setstacksize bin/xmeshfem3D

Should work and should reassure us that the stack size is always set to 
unlimited during the runs. Then the qsave can go away.

More over, to fix this problem permanently, it is more important, I 
think, to put deallocate() for each allocate() in the code, which exists 
in vast amounts in specfem3D.F90(?), for example.

best regards and good luck.

On 01/25/2012 10:43 PM, Dimitri Komatitsch wrote:
>
> Dear Chad,
>
> Thank you for your message. Indeed as you say we discussed this point a
> few months ago, and I thought the problem had been fixed. I remember
> that somebody had a look (maybe Carl, Daniel and/or Yang? I cc them).
>
> If anybody has worked on this already and fixed the problem, please let
> me know (and I will check why the bug has reappeared). Otherwise in a
> few days I will have a look at the SVN logs to see what went wrong (the
> code used to be 100% OK; this problem appeared a few months ago for some
> reason).
>
> Are you using a Cray or IBM machine? if so, it could just be the
> "-qsave" flag that needs to be removed from "flags.guess" before running
> the "configure" script. I was about to add this to the users manual
> (thanks to Marcin Zielinski, who found and fixed this compiler option
> problem; I cc him).
>
> Thanks,
> I will keep you informed,
>
> Dimitri.
>
> On 01/25/2012 07:41 PM, Chad Hogan wrote:
>> Hello,
>>
>> I've been having some trouble with SPECFEM3D_GLOBE, as all of my runs
>> terminate with
>> "forward simulation became unstable and blew up in the fluid". I have
>> seen previous messages on this list suggesting that there may be code
>> issues.
>>
>> To remove the possibility that it is a bug in my own changes to the
>> code, I checked out the original code from SVN
>>
>> $> svn checkout
>> http://geodynamics.org/svn/cig/seismo/3D/SPECFEM3D_GLOBE/trunk
>> SPECFEM3D_GLOBE
>>
>> -----
>> Checked out external at revision 19417.
>>
>> Checked out revision 19416.
>> -----
>>
>> For my testing I am using "small_benchmark_run_to_test_a_new_machine"
>>
>> Meshing completes successfully, but the simulation becomes unstable in
>> the fluid.
>>
>> The meshing and simulation were run with the unmodified
>>
>> Par_file_this_example (md5sum: 8d385f2b18e35e0ecd3acb0c0d8dd612 )
>> CMTSOLUTION_this_example (md5sum: 4d57b5e215d5e2f39988e1031000cc53 )
>> STATIONS_this_example (md5sum: ffc137053fe5e15008c4a749c4eb50f3 )
>>
>> and the unmodified checked-out code as above.
>>
>> I would very much appreciate any suggestions or advice on how I can
>> address this problem.
>>
>> Thank you very much,
>>
>> Chad Hogan
>> cmhogan at ucalgary.ca
>> _______________________________________________
>> CIG-SEISMO mailing list
>> CIG-SEISMO at geodynamics.org
>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>

-- 
Marcin Zielinski
Supercomputing group, OSD
SARA
Science Park 140           mob +31 (0)65 123 5602
1098 XG Amsterdam          tel +31 (0)20 888 4074
www.sara.nl                fax +31 (0)20 668 3167
marcin.zielinski at sara.nl