[ase-users] problem with jacapo: symmetry=True

[Marcin Dulak]

Hi,

John Kitchin wrote:
>
>     John Kitchin wrote:
>
>         I guess this is a problem in the Fortran code.
>
>         1.  does the serial executable have a problem with the
>         symmetry=True flag?
>
>     yes, it freezes before the first SCF step.
>     I have just tested that there is no energy jump with
>     https://svn.fysik.dtu.dk/projects/ase/trunk/ase/calculators/dacapo.py
>     (using equivalent parameters to jacapo and usesymm=True).
>     Still even with dacapo.py the job freezes when run on 6 cores.
>
>     Marcin
>
>
>
> I am a little confused by your reply. The first line "yes, it freezes 
> before the first SCF step" suggests there is a Fortran problem in both 
> serial and parallel (they are almost the same codes in most places).
it's confusing, but this is what happens.
As Lars reported the behavior depends on machine and number of cores job 
is running on.
I made some more tests with the following results (with symmetry=True):
- serial job freezes for both jacapo and dacapo on our opteron nodes, on 
our xeon nodes serial job runs and dacapo does not show the energy jump 
- jacapo does.
The only difference between the two interfaces in the txt output are 
different atomic positions for 3 atoms (!)
- on 6 cores job behaves as the one on 1 core,
- on 8 cores job shows the energy jump for jacapo on both opteron and 
xeon, and does not show the jump for dacapo for both opteron and xeon
Can you check if you see these problems?
jacapo and dacapo inputs attached.
I think also we should start adding some jacapo tests to 
https://trac.fysik.dtu.dk/projects/ase/browser/trunk/ase/test

Marcin
>
> The second line "I have just tested that there is no energy jump with 
> https://svn.fysik.dtu.dk/projects/ase/trunk/ase/calculators/dacapo.py 
> (using equivalent parameters to jacapo and usesymm=True)." sounds like 
> the serial  job ran fine and finished with symmetry=True using dacapo.py
>
> and the third line says that in parallel on 6 cores it freezes with 
> dacapo.py
>
> the python modules don't do anything but setup an nc file that is read 
> by a fortran code that runs the calculations. you could easily see the 
> difference between a jacapo created ncfile and dacapo.py created 
> ncfile with ncdump. but as long as the UseSymmetry variable has the 
> same value, it is a fortran problem that cannot be fixed from python.
>
> I guess that someone may have to do some fortran debugging to figure 
> out precisely where the freezing occurs.
>
> j
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ase-users mailing list
> ase-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/ase-users

-- 
***********************************
 
Marcin Dulak
Technical University of Denmark
Department of Physics
Building 307, Room 229
DK-2800 Kongens Lyngby
Denmark
Tel.: (+45) 4525 3157
Fax.: (+45) 4593 2399
email: Marcin.Dulak at fysik.dtu.dk

***********************************

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Pt111.py
Type: text/x-python
Size: 1299 bytes
Desc: not available
URL: <http://listserv.fysik.dtu.dk/pipermail/ase-users/attachments/20101029/bbdb35a9/attachment.py>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Pt111_dacapo.py
Type: text/x-python
Size: 1383 bytes
Desc: not available
URL: <http://listserv.fysik.dtu.dk/pipermail/ase-users/attachments/20101029/bbdb35a9/attachment-0001.py>