[Pw_forum] segmentation faults in pwcond.x

[Derek Stewart]

Hi everyone, 

I am trying to do some calculations with pwcond to determine the complex 
band structure for a fairly large unit cell structure (22 atoms).  I am 
using QE version 4.0.4 compiled with the intel compilers (ifort & icc v10.1, 
mkl 10.0.4) in parallel with lam-mpi 7.1.4.  I have no problem running the 
scf calculation. 

However, my pwcond calculation crashes after printing out the atomic 
positions and the following lines: 

        k(    1) = (   0.0000000   0.0000000), wk =   1.0000000 

 ngper, shell number =          951         256
 ngper, n2d =          951         212
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image              PC        Routine            Line        Source
Stack trace terminated abnormally.
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image              PC        Routine            Line        Source 

The run was done in parallel over 10 processors.  I have set the stack as 
unlimited. 

Looking back through the pwscf archive, it appears that other people have 
run into segmenation faults at a similar point in their calculation.  It was 
mentioned that pwcond performs a diagonilization over a non-symmetric matrix 
that occasionally crashes.  Has this issue been resolved?  Also, are there 
problems with larger systems for pwcond and total RAM required? 

All my parameters for the pwcond input file are similar to those used in 
example12. 

Any suggestions would be greatly appreciated! 

Thanks, 

Derek 


################################
Derek Stewart, Ph. D.
Scientific Computation Associate
http://www.people.cornell.edu/pages/das248/
250 Duffield Hall
Cornell Nanoscale Facility (CNF)
Ithaca, NY 14853
stewart (at) cnf.cornell.edu
(607) 255-2856