Topic Options
#23380 - 02/09/10 06:07 AM ERROR IN FREHP
sgd Offline
Forum Member

Registered: 02/23/04
Posts: 55
I am trying to run vibran module, using a serial version of charmm running on one processor of a cluster. I have tested two different systems having 1440 atoms and 200 atoms respectively. The smaller system runs perfectly whereas for the bigger system I am getting the following error message just after invoking of the diag command:

VEHEAP> Expanding heap size by 18677760 words.

***** LEVEL -4 WARNING FROM <LINKHP> *****
***** OVERLAP OR LOGIC ERROR IN FREHP.
******************************************
BOMLEV ( -2) IS REACHED - TERMINATING. WRNLEV IS 5



Here is the script in brief:
------
read rtf
read parameters
read psf
read coor

delete waters and ions

bomlev -2

minimization

set nmod ?natom
calc nmod = @nmod * 3

vibran nmod @nmod
diag ENTROPY TEMP 300
end
--------

I set bomlev -2 to avoid ther termination due to the waring of "NAT3 for second derivatives is >3600".

Surprisingly, both system run perfectly on a desktop PC and I have used the same charmm script/input. Is it something to do with the charmm executable itself (e.g. installation) or with the machine?

Top
#23381 - 02/09/10 06:34 AM Re: ERROR IN FREHP [Re: sgd]
lennart Online   content

Forum Member

Registered: 09/25/03
Posts: 4741
Loc: ~ 59N, 15E
Should not be a problem for CHARMM. Older versions may have issues with memory management (heap) on new 64-bit machines, but it is hard to tell without more details, eg concerning CHARMM version, installation options and the platform you are using.
_________________________
Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden

Top
#23395 - 02/09/10 11:41 PM Re: ERROR IN FREHP [Re: lennart]
sgd Offline
Forum Member

Registered: 02/23/04
Posts: 55
Charmm version 34b2

The platform is AMD O 8380 2.54 GHz
32 cpu cores per node, total 32 nodes with infiniband interconnection.

Linux version: centOS 6.2

I will provide the installation options once I hear from our supercomputing centre who install all applications.

A weird thing is that out of many attempts, a very few times the calculations were successfully completed, although the same script gave problem in the next run and so on.

The following command I am using (as per the instruction for the cluster users):
mpirun -machinefile $PBS_NODEFILE -np 1 charmm-s.exe < input > output

--------
Shubhra Ghosh Dastidar
Singapore


Edited by sgd (02/09/10 11:43 PM)

Top

Moderator:  John Legato, lennart