[gmx-developers] Generalized Born - Crashes

Erik Lindahl lindahl at cbr.su.se
Sun Apr 12 17:33:40 CEST 2009


On Apr 12, 2009, at 1:17 PM, David van der Spoel wrote:
>
> Why is there a need for a special neighbor list? Why not use the  
> normal one (possibly without exclusions)

Because doing that too early without extensive is a guaranteed way to  
introduce stupid bugs. Just a few examples:

1. It is a pain to debug GB when the particles are in three separate  
neigborlists (LJ-only, Coul-only, both)
2. If you use twin-range cutoffs you will get problems when LJ-only  
atoms are not included in the long-range neighborlist, but they should  
still be included for calculating born radii.
3. There is no "normal neighborlist without exclusions" - the  
neighborlist does not include excluded atoms. You would end up writing  
two copies of every single routine you need, one processing  
neighborlists, and one processing entries from the exclusion list.
4. Long-term, there is no specific reason why rgb should be the same  
as the interaction cutoff


Cheers,

Erik
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20090412/37fac28b/attachment.html>


More information about the gromacs.org_gmx-developers mailing list