[gmx-users] multiple time step algorithm

[Cristina GRECO]

Dear Berk Hess,

I am Cristina Greco.  When I first wondered why there were no multiple or adaptive timestep methods implemented I suspected that the reasons had to do with compatibility with other options.I was just interested because I want to set up a simulation to study problems connected with folding and unfolding of polymers; thus I am trying to get informed about the possible strategies .

Thank to you and all other Gromacs developers 

cristina greco

> 
> That it not what I meant.
> To properly support multiple time stepping (or many other features) it should 
> work correctly
> with all other options, such as parallelization, constraints, free-energy 
> calculations, pressure scaling, etc.
> Usually user contributions do not fulfill most of these criteria.
> So am I not talking about the amount of additional code, but more that 
> someone has to make sure
> that everything works with everything, which could mean additional code in many files
> that currently assume that everything is updated simultaneously.
> 
> Another option would be to have something like multiple 
> time stepping only work for certain combinations.
> But we would not want such code in the core of Gromacs.
> Even if a user would provide such code, there should be checks that stop the 
> code for combinations
> that are not supported.
> 
> Berk.
>