Yes, sorry I forgot to attach the files. I tried sending both the input and data file after my first email however the data file is rather large and is awaiting moderation. In the meantime here is simply the input file in case that is of any use.
Many thanks, Jeff.
in.dumb (415 Bytes)
Yes, sorry I forgot to attach the files. I tried sending both the input and
data file after my first email however the data file is rather large and is
awaiting moderation. In the meantime here is simply the input file in case
that is of any use.
it looks as if your bond definitions are off somehow,
but there could also be a bug in LAMMPS. this is
touching a code path that was changed not so long ago.
i presume, you are using the latest version.
before attaching larger text files, consider compressing them.
cheers,
axel.
Try to ramp up the value of "neighbor"
Try to ramp up the value of "neighbor"
This is likely the problem. The error is b/c one
processor cannot find the ghost atom that one
of its own atoms is bonded to. Your cutoff
is 3.2, the neigh skin is 0.3, so if you have a bond
that is more than 3.5 in length, it will fail
in parallel. Either that's a bad model (too long a bond),
or you need to use the communicate cutoff command to extend
the range of the ghost atoms each proc owns.
Steve
I have done some testing of a small system and so far have only been able to deduce one thing. The problem disappears when I double the simulation box dimensions.
The initial configuration is generated such that all atomic positions are within the unit cell and such that the intra atomic distance in a molecule is that of the equilibrium bond distance, so I was under the impression your suggestion might not be the case. (I also tested the model for very short bond lengths and still have the same issue).
I attach the input script and data file. In my test I change the xhi,yhi,zhi from 10 to 20 in order to get the script to run with mpi. The simulation in fact starts running in mpi from values of 19.99 upward.
Many thanks, Jeff.
in.dumb (403 Bytes)
data.dumb (15.2 KB)
jeff,
this input deck works for me with no modifications using a current
version of lammps and any number of processors between 1 and 32.
you are likely a victim of a bug that has since been fixed.
please upgrade to the latest version.
axel.