vasp 5.3.3 -Segmentation fault (core dumped) when trying Handson1 exercises

Message

stana · #1 Post by **stana** » Thu Jun 13, 2013 5:20 pm

I am trying to get started with VASP. After installation I ran the "Test Suite for VASP" by Peter Larrson and "Quick suite" and "Geoopt suit"e gave all pass. In "Production suite" only PbO2 passed. Rest gave Error due to Segmentation fault.
Then I got working with the "Workshop Lectures". From Handson1 1_1_O_atom, when running gives

POSCAR found : 1 types and 1 ions
LDA part: xc-table for Pade appr. of Perdew
POSCAR, INCAR and KPOINTS ok, starting setup
WARNING: small aliasing (wrap around) errors must be expected
FFT: planning ...
WAVECAR not read
WARNING: random wavefunctions but no delay for mixing, default for NELMDL
entering main loop
N E dE d eps ncg rms rms(c)
Segmentation fault (core dumped)

Same happens when trying to run 1_2_O_atomspin, 1_4_Odimer, 1_5_CO, 1_6_, 1_7_, 1_8_, 1_9_

1_3_O_atomspinlowsym, 1_10_, 2_1_, 2_2_ seem to work fine

Might that be because I did something wrong when installing VASP?

Kind regards!
markus

#2 Post by **admin** » Fri Jun 14, 2013 10:56 am

Have you checked "segmentation fault" issues?
E.g. proper setting in .bashrc file?

stana · #3 Post by **stana** » Fri Jun 14, 2013 12:20 pm

Thank you for the reply!
The only thing I changed in .bashrc is adding
"source /opt/intel/composerxe/bin/compilervars.sh intel64"
as I run a 64 bit system. I tried sourcing .bashrc again, just to be sure, but the error still occurs. I do not think that my ram is too small (8GB). Unfortunately I can't seem to find a core file. Could it be the installation of vasp.*.lib?
<span class='smallblacktext'>[ Edited Fri Jun 14 2013, 12:34PM ]</span>

alex · #4 Post by **alex** » Mon Jun 17, 2013 8:38 am

Hmm, difficult. Some questions:
- Did you u(n)limit your resources?
- Does it run on a single core?
- Are you sure you are running a 64 bit Linux?
- Did the compilation work properly?
- Have you tried a different compiler version?
- Is your MPI-Library used by different people (in other words: proven to be working)
... (to be continued)

Cheers,

alex

stana · #5 Post by **stana** » Mon Jun 17, 2013 9:58 am

- I am not aware of having done anything to u(n)limit my resources
- I tried to set it up to work parallel and it appears to use several cores according to the system monitor
- I am running Release 12.04 (precise) 64-bit, Kernel Linux 3.5.0-34-generic
- the compilation gave several warnings. compiling the lib only worked properly when using gfortran instead of ifort.
- I only tried ifort composer_xe_2013.4.183. Should I try an older version?
- No, unfortunately my MPI_library is not tested. Is there a relatively easy way to do that?

Thank you very much for your time!
markus

alex · #6 Post by **alex** » Mon Jun 17, 2013 4:31 pm

Last one first: check for sth. like
make test
or similar in you mpi installation directory. Mostly it'll _not_ test more than one node. You would have to go for some 'hello world' example from the web. But after your description I don't think this is your problem.

Maybe it's related to the mixing of compilers gfortran and ifort. Please give the error messages of compiling the lib with ifort here.

Cheers,

alex

#7 Post by **admin** » Thu Jun 27, 2013 12:00 pm

please check what get if you give
ulimit -s
(it should be unlimited)
if your account has hard-set stack size limits, please add the following line to your .bashrc:
ulimit -s unlimited

stana · #8 Post by **stana** » Mon Jul 01, 2013 5:47 pm

Thank you very much for your response. "ulimit -s" gave 8192. So "ulimit -s unlimited" solved my problem.
Kind regards!
markus