[gmx-developers] How to treat energy jumps at the cut-off?

[Berk Hess]

Hi,

We are soon releasing a 4.6 pre-release with a new cut-off scheme.
This uses exact, buffered, cut-off's:
http://www.gromacs.org/Documentation/Cut-off_schemes
In our current git code there is a jump in the energy at the cut-off,
both for LJ and PME. For checking of energy conservation it would be better
to (also) calculate energies with the LJ and PME real-space potential 
shifted
to zero. But we could also choose to only report the shifted energies.
The issue is that shifting the potential adds a (small) cost and complicates
the kernels a little. If we optionally want to shift it, we need to come 
up with
names for the non-shifted and shifted potentials (and they shouldn't be
confused with the "old" shifted interactions, which shift the force).

Do you have an opinion/suggestion?

Cheers,

Berk