Statpt segmentation fault for systems > 500 atoms

[mgt16]

All,
I have been performing geometry optimizations of nanoclusters of more than 500 atoms. Statpt is being run by the jobex script. The jobex script is failing after the ridft energy is converged during the geometry optimization step and from what I can see the statpt might be running out of memory.

The GEO_OPT_FAILED file contains only a warning to look at the statpt.tmpout file - but the statpt.tmpout file ends abruptly with no uncommon error:

Code: [Select]

  norm of actual CARTESIAN gradient:  2.99848E+00

  ENERGY =  -23769.9570481000 a.u.; # of cycle =    1


 keyword $hessapprox missing in file <control>


  USING DIAGONAL INITIAL HESSIAN
  Diagonal elements set to:    0.50000

I found this not very useful so I looked to the jobex output file.

The jobex output file only contains this:

Code: [Select]

/ihome/crc/install/turbomole/7.02/TURBOMOLE/scripts/jobex: line 528: 149335 Segmentation fault      /ihome/crc/install/turbomole/7.02/TURBOMOLE/bin/em64t-unknown-linux-gnu_smp/statpt

I have run several similar systems before on our cluster - but have recently received this error for two important systems.
I would appreciate any help!
Thanks,
Michael.

[uwe]

Hello,

one usually gets 'Segmentation fault' messages if the (user or system) memory limits are too low, or if you ask for more memory than your machine really has.

Since the statpt module  is not memory-consuming, I assume that you have hit a user limit (especially the stack size limit).

To check and resolve this, please see:

http://www.turbo-forum.com/index.php/topic,23.0.html  -- I get "Segmentation Fault", SIGSEGV or "Memory fault"

If that does not help, just contact the Turbomole support team to get help (send an email to turbomole at cosmologic.de).

Regards,

Uwe

[mgt16]

This solved my issue right away!
Thanks!
Michael.