[lammps-users] (no subject)

Hello Group,
I was wondering if anyone has any experience running LAMMPS on a Mac cluster. I have devised a simple test case with flexible water, where I conducted the simulations on our Mac cluster as well as our Opteron cluster. In this case, I get the exact same energies after a 1000 timesteps.

However, if I devise a new test case where I use SHAKE bond constraints, I get different energies that are significantly off …

For the opteron cluster, using the Portland Group (PG) compiler, the total energy of the system is -4973.8062.
For the mac cluster, using GCC and XLC compilers, the total energy of the system is -4973.8497 for both compilers.

Does anyone have any insight into why this may be happening?

I think I replied to this, but the SourceForge mail lists seem to have
been messed up for several days ...

I don't think an energy difference of a few parts on 10^6 is significant
on two different machine, after 1000 steps. So I would attribute
it to round-off.