[Pw_forum] compile error on Xeon with intel fortran

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Sat Apr 5 15:27:03 CEST 2008


On Sat, 5 Apr 2008, Min Seung Kyu wrote:

MSK> Dear users and developers,

dear min,

MSK> I compiled espresso3.2.3 on Xeon machine with Intel fortran 10.1.012 and
MSK> 
MSK> MKL 10.0.1.014 version. Whenever I run pw.x, however, I get segementation

what linux version and which release? which type of xeon cpu
(check out /proc/cpuinfo)? are you running 32-bit mode or x86_64?
 
MSK> fault message at the beginning.

check out the usual recommendations with intel:
- increase stack size limit. 
- disable MKL threads with setting the environment 
variable OMP_NUM_THREADS to 1. (as of version 10
MKL always tries to run multiple threads on multi
cpu/core machines, regardless whether you also
are running MPI or just don't want it)

- try compiling with reduced optimization.

MSK>       Program PWSCF     v.3.2.3  starts ...
MSK>       Today is  4Apr2008 at 21:35:35
MSK> 
MSK>       Ultrasoft (Vanderbilt) Pseudopotentials
MSK> 
MSK>       Current dimensions of program pwscf are:
MSK> 
MSK>       ntypx = 10   npk = 40000  lmax =  3
MSK>       nchix =  6  ndmx =  2000  nbrx = 14  nqfx =  8
MSK> forrtl: severe (174): SIGSEGV, segmentation fault occurred
MSK> Image              PC                Routine            Line        Source
MSK> pw.x               0000000000DF55B7  Unknown               Unknown  Unknown
MSK> libpthread.so.0    0000003A0550C4F0  Unknown               Unknown  Unknown
MSK> pw.x               0000000000DF5593  Unknown               Unknown  Unknown
MSK> libpthread.so.0    0000003A0550C4F0  Unknown               Unknown  Unknown
MSK> 
MSK> My maks.sys file contains
MSK> 

MSK> BLAS_LIBS      = -L/opt/intel/mkl/10.0.1.014/lib/em64t -lmkl_em64t -lmkl_core -lguide -lpthread

where are the rest of the settings? they matter as much as 
the BLAS lib. what makes you think that MKL is the culprit?

MSK> 
MSK> Please let me know what I should do.

the most important thing you should do is to report problems 
in a way, that somebody can reproduce it. you provide some
information, but it is not complete...

cheers,
  axel.

MSK> Sincerely, Min.
MSK> 
MSK> 
MSK> 
MSK> 
MSK> _______________________________________________
MSK> Pw_forum mailing list
MSK> Pw_forum at pwscf.org
MSK> http://www.democritos.it/mailman/listinfo/pw_forum
MSK> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.


More information about the Pw_forum mailing list