[Wien] "Segmentation fault" in lapw2 program

[Laurence Marks]

Increasing NMATMAX is probably not going to be the right approach, and
I doubt that it is a memory problem in lapw2. It is hard to know
exactly where the problem is, and it will be useful to try and find
out exactly where it is happening, and unfortunately sometimes the
scripts that Wien2k uses hide the problem. First, check case.dayfile
to see if it has more information. If still nothing you have to go
around how Wien2k runs shell scripts and find out what is causing the
code to crash.

The command "x lapw2" does two things:

a) Runs "x lapw2 -fermi" to work out the occupancies. Test this by
hand and see if it fails. If it  does run it at an even lower level as
"lapw2 lapw2.def". If this crashes with luck it will show you where
and you can go and look at the code. It might be necessary to use -g
in compilation. If nothing, go to the next step

b) The command runs a series of commands (on different nodes if you
are running in parallel) which you can replicate by something like
"lapw2 lapw2_1.def" -- look at what files you have. Again, there is a
good chance this will tell you what is going on.

You can get some extra information about what commands are being run
by setting the debug flag in lapw2para. (General note: it might be
good to have the debug flag reflect the commands being run.)

On 6/7/07, cesar <cesar at unizar.es> wrote:
> To whom may help me,
>
>
> Quite recently, I ve been found the following error in lapw2 program when I
> run the SCF.
> ------------------------------------------------------------------
>     start       (Thu Jun  7 11:00:07 CEST 2007) with lapw0 (20/20 to go)
>
>     cycle 1     (Thu Jun  7 11:00:07 CEST 2007)         (20/20 to go)
>
> >   lapw0       (11:00:07) 12.646u 0.241s 0:12.88 100.0%        0+0k 0+0io
> 0pf+0w
> >   lapw1       (11:00:20) 14.442u 0.387s 0:14.85 99.7% 0+0k 0+0io 0pf+0w
> >   lapw2       (11:00:35) Segmentation fault
> 0.797u 0.053s 0:00.85 98.8%     0+0k 0+0io 0pf+0w
>
> >   stop error
> ------------------------------------------------------------------
> (I m using a reduced number of k-points to check the SCF convergence, and
> the -traceback compiling option is enabled but optimization flag are still
> enabled)
> ------------------------------------------------------------------
>
> This fault seems to be related with the increment of the number of atoms in
> the unit cell because on decreasing this number the fault disappears (of
> course for other system).
>
> I ve change the compiling dimension NMATMAX from 10000 to 20000 or higher to
> increase the matrixsize of the basis function (the memory should be enough,
> 6 Gb), but lapw2 fails again in the first SCF cycle.
>
> Can this problem be associated to the memory limitation of em64t systems
> above 2Gb? (mentioned in "www.ghfecher.de/Fecher_CompileIntel.pdf", section
> 5.1)
>
> Is it possible to fix this problem with a dynamic linking of some
> libraries?, or Any idea to fix this problem?
>
> Sincerely,
> Cesar de la Fuente
> cesar at unizar.es
>
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> PS:
> -I ve tried to calculate using TEMP option instead TETRA, but nothing.
> -I ve not installed MPI and SCALAPACK libraries, only the usual Intel and
> scalapack_wien2k libraries.
> -I m using ifc 9.0, mkl 8.0 and Linux (RHE4-ud2 for WS). My system is a Dual
> Xeon Pentium with 6Gb of (real) memory.
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


-- 
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
Web: www.numis.northwestern.edu
EMM2007 http://ns.crys.ras.ru/EMMM07/
Commission on Electron Diffraction of IUCR
www.numis.northwestern.edu/IUCR_CED