Dear All,
I am trying to learn the effect of cutoff length used in the simulation. I am using
pair_style lj/charmm/coul/long/opt 10.0 11.5 and
pair_style lj/charmm/coul/long/opt 9.0 10.0 with PPPM.
Simulation with lower cutoff (9Angs) runs fine but one with slightly larger cutoff crashes with any error. System is ~5000 atoms.
Any input to figure out the source if error will be helpful.
Thanks
Hemant
Dear All,
I am trying to learn the effect of cutoff length used in the
simulation. I am using
pair_style lj/charmm/coul/long/opt 10.0 11.5 and
pair_style lj/charmm/coul/long/opt 9.0 10.0 with PPPM.
Simulation with lower cutoff (9Angs) runs fine but one with slightly larger
cutoff crashes with any error. System is ~5000 atoms.
Any input to figure out the source if error will be helpful.
impossible to say from this little information.
what version of LAMMPS?
running on what OS? serial/parallel?
how do you run it?
can you reproduce the same behavior with the "rhodo" benchmark example
(if yes, what are your modifications?)
or the "peptide" example?
axel.
Thank you for quick reply.
I am using Lammps version (5 May 2012) with 32 processor compiled with intel12 compiler and mvapich.
I haven’t tried to reproduce the error with benchmark system, I will try and will update you.
However I have other systems with larger cutoff but this particular system somehow larger cutoff leads instability. I am unable to figure out which setup parameter should I change to complete the run.
Thanks
Hemant
Thank you for quick reply.
I am using Lammps version (5 May 2012) with 32 processor compiled with
intel12 compiler and mvapich.
I haven't tried to reproduce the error with benchmark system, I will try and
will update you.
However I have other systems with larger cutoff but this particular system
somehow larger cutoff leads instability. I am unable to figure out which
setup parameter should I change to complete the run.
there may be a gazillion possible reasons.
the fact that a job completes doesn't mean that it is running correctly.
the other option to check is to try running with a more recent version
of the code. while new features occasionally add a bug to the code,
newer versions have less bugs.
axel.
Thank you for quick reply.
I am using Lammps version (5 May 2012) with 32 processor compiled with
intel12 compiler and mvapich.
I haven't tried to reproduce the error with benchmark system, I will try and
will update you.However I have other systems with larger cutoff but this particular system
somehow larger cutoff leads instability. I am unable to figure out which
setup parameter should I change to complete the run.there may be a gazillion possible reasons.
the fact that a job completes doesn't mean that it is running correctly.the other option to check is to try running with a more recent version
of the code. while new features occasionally add a bug to the code,
newer versions have less bugs.
1) strongly agree. Please get the newest version, (or at least a much
more recent version) of LAMMPS. LAMMPS gets modified all the time.
My old lammps input scripts from 2012-5 no longer run on the latest
versions of LAMMPS.
2) Perhaps this does not need to be said, but have you tried watching
a movie of your system to see what is going on? (Write to dump file
often. I attached some generic instructions for visualizing using
VMD.)
Simulation with lower cutoff (9Angs) runs fine but
one with slightly larger cutoff crashes with any error.
3) Are you sure there was no error message? I've noticed that some
LAMMPS error messages do not get saved to the log.lammps file, but get
printed to stderr or stdout instead. If you are running these
simulations on a 32-core cluster with a queuing system, then these
messages may be lost (or redirected to another file). Also, if your
system only has 5000 atoms, perhaps you can run the simulation and
reproduce the crash using a desktop computer. (possibly not, but
worth trying). I hate queueing systems.
Forgive me if these suggestions are not relevant or if you have tried
them already.
Andrew
README_visualize.txt (2.85 KB)