This directory has a C++ code multiple.cpp which illustrates how a driver code can create multiple instantiations of LAMMPS to run independent simulations simultaneously, on a set of processors allocated by MPI.

Once you have built LAMMPS as a library (see examples/COUPLE/README), you can then build the driver codes with compile lines like these, which include paths to the LAMMPS library interface, MPI (an installed MPICH in this case), and FFTW (assuming you built LAMMPS as a library with its PPPM solver).

g++ -I/home/sjplimp/lammps/src -c multiple.cpp g++ -L/home/sjplimp/lammps/src multiple.o \

-llammps -lfftw -lmpich -lmpl -lpthread -o multiple

If you then run this command:

% mpirun -np 8 multiple 4 in.chain 1.0 0.5

you will create 4 instances of LAMMPS, each running on 2 of the allocated 8 processors. The final two arguments are the base temperature and a delta temperature added by each instance. The provided input script in.chain uses those temperatures, set as a variable "t" when each instance of LAMMPS is created.

After a few seconds you should get this output to the screen:

Instance 1, final temp = 1.00469 Instance 2, final temp = 1.50395 Instance 3, final temp = 1.99873 Instance 4, final temp = 2.50862

indicating the runs equilibrated to 4 temperatures separated by 0.5 (in LJ units). There should also be 4 screen files (screen.N) and 4 log files (log.lammps.N) from the individual LAMMPS runs.