[gmx-users] Implicit solvent simulation

[steletch at jouy.inra.fr]

Selon Erik Lindahl <lindahl at cbr.su.se>:

> Hi,
>
> On Jul 18, 2007, at 5:08 PM, David van der Spoel wrote:
>
> > The version used by Alan Mark has never made it into gromacs.
> > Rather, Erik Lindahl started the implementation of this stuff but
> > did never finish it completely, due to other priorities I presume.
> > We should however take out the variables to avoid confusion (unless
> > someone provides a full implementation).
>
> Actually, technically the GB implementation is already in Gromacs
> (even assembly kernels), and quite correct, it's just that you don't
> have the born radii yet :-)
>
> The last couple of years we've mainly done membrane stuff where it's
> not useful for my students, and thus not a priority to implement.
> However, recently Per Larsson has finished C implementations of the
> Still radius calculations that I've started optimizing.
>
> The big remaining task is the SA part of GB/SA, i.e. the surface
> calculation, which also needs to be optimized and preferrably
> parallelized (harder) to be useful.
>
>
> Fan Hao actually offered us the code he wrote with Alan, but the
> challengefor us isn't to write a trivial Still GB implementation, but
> to make it perform well. There's little point in having a GB
> implementation that is slower than the fast explicit-solvent
> simulations in Gromacs!
>
>

Happy to hear that, if you need some beta-tester for validation and performance
tests, i'm your man :-)

Cheers,

Stéphane Téletchéa