[gmx-developers] LINCS + L-BFGS

[Berk Hess]

On 11/25/12 22:08 , David van der Spoel wrote:
> Is this combination supposed to work?
>
> It does not for me, independent of lincs-order (8, 12, 100) the bond 
> lengths vary randomly during a minimization.
>
It seems that L-BFGS doesn't do any constraints. But apparently there's 
also no check for this.
It doesn't call do_em_step. Implementing this should be straightforward, 
at least without
domain decomposition.

Cheers,

Berk