[gmx-users] Segmentation fault

[Mark Abraham]

Marcelo Müller wrote:
> Hello,
> 
> I'm getting some problems about segmentation fault. I'm doing a protein 
> simulation during 20 ns on a water cubic box with a total of 23000 
> atoms.  I'm using the TIP3P water model at 300 K. Interestingly, when I 
> use the Berendsen barostat for pressure coupling, I've got a 
> segmentation fault about 1 ns of trajectory. When I don't use it the 
> simulation extends untill 10 ns, but the fault also occurs. Could it be 
> some problem with memory limitation of the computer? I'm using a cluster 
> to run the simulations, and I'm trying to check the memory capacity of 
> the node that I've used.
> Besides, the message that I've seen is something like that:
> Step #number of the step and time     mdrun -v -deffnm run   
> Segmentation fault

Check out the .log file for the real error message, rather than stderr. 
I expect you will see LINCS warnings, and that your system hasn't been 
prepared/minimized/equilibrated carefully enough, which is causing 
problems down the road. See 
http://wiki.gromacs.org/index.php/Errors#LINCS_warnings and 
http://wiki.gromacs.org/index.php/Steps_to_Perform_a_Simulation

There are many other more likely sources of problems than a memory 
limitation.

Mark