Dear all,
My system includes some water(2226 water molecules(spc/e)),a membrane(4000 carbon atoms,no charge) consists of amorphous carbon and a graphite wall using for compressing water.After running several steps,it gives an error: Shake atoms missing.It is difficult for me to understand this problem because the shake output seems no problem.Another question is that it takes 14404s CPU(16-core parallel) time to run 6500 steps for my system;Why my MD speed so slow?How can I optimize my parameters in Lammps to speed up my calcation?
Dear all,
My system includes some water(2226 water molecules(spc/e)),a
membrane(4000 carbon atoms,no charge) consists of amorphous carbon and
a graphite wall using for compressing water.After running several
steps,it gives an error: Shake atoms missing.It is difficult for me to
understand this problem because the shake output seems no
but using neigh_modify delay 10 is very aggressive.
problem.Another question is that it takes 14404s CPU(16-core parallel)
time to run 6500 steps for my system;Why my MD speed so slow?How can I
optimize my parameters in Lammps to speed up my calcation?
hard to say without knowing more about your system,
e.g. the dimensions of your box, but there are two
likely reasons: load imbalance and using ewald instead
of pppm. also, with the processors 2 2 2 line, you
can only run with 8 MPI tasks. with proper updates
to the neighbor lists, you may be able to increase
the time step to 2fs.
cheers,
axel.
The error is b/c one or more atoms moved too far.
Re: speed - you are running Ewald which can
be very slow, esp for big systems. Try PPPM.
And try running on 1 proc for a short run and look
at the timing breakdown. That will tell you where
it is spending time and help you judge whether it is
faster on 16 procs,
Steve