Reax segmentation fault

LAMMPS LAMMPS Mailing List Mirror
1

Hi

I compiled lammps with reax option, when I run my code I get the errors below, the manual say this : reax_defs.h contains dimensions of statically-allocated arrays created by the ReaxFF library. The size of these arrays must be set small enough to avoid exceeding the available machine memory, and large enough to fit the actual data generated by ReaxFF. If you change the values in reax_defs.h, you must first rebuild the library and then rebuild LAMMPS. Is this the reason for the errors?

Setting up run …
[s6-1:27667] *** Process received signal ***
[s6-1:27667] Signal: Segmentation fault (11)
[s6-1:27667] Signal code: Address not mapped (1)
[s6-1:27667] Failing at address: (nil)
[s15-2:00706] *** Process received signal ***
[s15-2:00706] Signal: Segmentation fault (11)
[s15-2:00706] Signal code: Address not mapped (1)
[s15-2:00706] Failing at address: (nil)
[s5-3:10892] *** Process received signal ***
[s5-3:10892] Signal: Segmentation fault (11)
[s5-3:10892] Signal code: Address not mapped (1)
[s5-3:10892] Failing at address: (nil)
[s6-1:27668] *** Process received signal ***
[s6-1:27668] Signal: Segmentation fault (11)
[s6-1:27668] Signal code: Address not mapped (1)
[s6-1:27668] Failing at address: (nil)
[s5-10:01445] *** Process received signal ***

Thank you

2

Possibly, but I think the error message for
that is usually more informative. So I don't
know what error you are seeing - there is not
enough information.

Steve

3

Possibly, but I think the error message for
that is usually more informative. So I don't

indeed, it is more likely that either the compilation,
your machine setup, or your input is messed up.

can you run the example inputs shipped with LAMMPS
with your executable?

what platform/os/compiler are you running on?
have you tried getting a stack trace?
( http://lammps.sandia.gov/threads/msg19060.html )

cheers,
    axel.

4

Hi

I changed atom_style from atomic to charge and now the simulation runs without problems

Ulises

5

Charge is clearly required for ReaxFF, since
it sets the charge on each atom. LAMMPS
should check if charge isn't defined and throw
an error. Were you using pair_style reax or
reax/c ??

Steve

6

I was using pair_style reax and I did not get the charge error just a lot of segmentation fault messages, by the way I am using LAMMPS 29 May 2010

Ulises