[gmx-users] errors depending on (lack of) solvent

[Erik Lindahl]

>
> So, isn't there a way to run parallel simulations of a single molecule
> on 2 or more machines?
>
Yes, it's possible if you don't constrain all bonds. It's pretty common 
in the literature to constrain only bonds with hydrogens for this very 
reason.

We might implement a parallel constraint solver, but I doubt it will be 
useful apart from testing stuff since it will involve extra 
communication (reducing performance).

Cheers,

Erik