On Fri, Aug 12, 2011 at 2:22 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
> > I dealt with this problem a couple of days ago in mdgx and had to fix
> > the
> > zeros by intercepting them as I populated the list of the scaling
> > factors
> > actually implemented in the calculation. I make a list of "adjustment"
> > factors for 1-4 scaling terms, which are basically 1 - 1.0/T where T is
> > the 1-4 term read in from the topology. When T is zero I make T=1.0,
> > so
> > that the adjustment term comes out to be zero and the overall effect is
> > that the 1-4 interaction is unchanged. So, as far as I'm concerned I
> > have
> > dealt with this issue. But, it would seem simpler to designate "1.0"
> > in
> > the topology where a 1-4 bonded term is not meant to be scaled.
>
> But what's the issue with just dividing by zero? Since the actual element
> of
> the array should never be used. Unless you calculate all the dihedral terms
> and then multiply any duplicates by zero.
>
Not all languages allow this. Fortran does, certainly, but C doesn't (from
the sound of Dave's issue). I knew that (at least the Fortran compilers
I've tested out) allow division by 0 in double precision, and just say it's
infinity. Python raises a ZeroDivisionError exception.
Perhaps we should just make these dihedral terms very very small? Something
like 1E-6 or so, so that it still causes the dihedral energy to explode, but
doesn't cause issues in other languages (since I agree it's nice to know
when those terms are being applied when they shouldn't be).
All the best,
Jason
Either way I think having the scaling term for all impropers and duplicate
> dihedrals set to 0 is the best approach since then it will catch any
> implementation that might be double counting 1-4s by mistake.
>
> All the best
> Ross
>
> /\
> \/
> |\oss Walker
>
> ---------------------------------------------------------
> | Assistant Research Professor |
> | San Diego Supercomputer Center |
> | Adjunct Assistant Professor |
> | Dept. of Chemistry and Biochemistry |
> | University of California San Diego |
> | NVIDIA Fellow |
> | http://www.rosswalker.co.uk | http://www.wmd-lab.org/ |
> | Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
> ---------------------------------------------------------
>
> Note: Electronic Mail is not secure, has no guarantee of delivery, may not
> be read every day, and should not be used for urgent or sensitive issues.
>
>
>
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Aug 12 2011 - 11:30:05 PDT
