[lammps-users] optimizing lammps

LAMMPS LAMMPS Mailing List Mirror
1

I am currently trying to optimize LAMMPS on our cluster and stumbled upon
this talk:

http://www.arl.hpc.mil/events/Clusters2006/speaker.php?id=59

from James Fischer. I was wondering if these optimizations will find their
way into the core soon.

Here is the program listing:

http://www.arl.hpc.mil/events/Clusters2006/program.html

Naveen

2

I'm familiar with Jim Fischer and his collaborator's work. I don't think
it's complete yet, but I've advised them what I would need eventually
to add it to the main distribution. One of their ideas (for pair hybrid)
is already in the latest LAMMPS with the accompanying speed-up
for hybrid systems. The other work is for EAM and pair-CHARMM
but it's less clear to me how much speed-up it will really offer.

Steve

3

Hi,
     I have a bunch of data files, some of them have 2 bond types, some of
them have only one. I want to use the same lammps input script for all of
them, but only to change this line everytime:
     read_data MYDATAFILE

    However, it turned out bond coefficient caused problem. In the input
script, I put:
      bond_coeff 1 1000 1
      bond_coeff 2 2000 1

    Because some data file have only one bond type, lammps stopped with error
saying that
   ERROR: Incorrect args for bond coefficients.

   So that I had to check bond type in each data file and also change
bond_coeff in my input script accordingly. It was more complicated. I wonder
would it be nicer if lammps simply ignore unused setting rather than comlain
about it and terminate?
    Thanks

jiwu

4

So that I had to check bond type in each data file and also change
bond_coeff in my input script accordingly. It was more complicated. I >wonder
would it be nicer if lammps simply ignore unused setting rather than >comlain
about it and terminate?

You can use "*" for the bond type if you want it to cover an (unknown)
range of bond types. But I don't think it is safe to just ignore
an incorrect bond type specification, else the user won't know
when he screwed up. You might be better off
making multiple input scripts for these kinds of diffferernt problems
and using one master script (with jump commands) to call
them all.

Steve