[gmx-users] Larger simulation times because of many [ pairs ]?

[Mark Abraham]

Hi,

Without having checked the code, adding [pairs] forces mdrun to compute
that non-bonded interaction. So you should expect the cost of computing
those interactions to go up. If that dominates the cost of the simulation
(e.g. you have no other non-bonded interactions) then your observations are
consistent. The implementation of [pairs] is not optimized for this
application.

Mark

On Fri, Feb 13, 2015 at 11:13 PM, Fernando Yrazu <fyrazu at rice.edu> wrote:

> Hi,
>
> I am bulding a C-alpha coarse grained model of a protein, and I would like
> to use a modified Lennard-Jones potential for every possible non-bonded
> contact.  For that purpose, I have built a list of all the possible pairs
> under the [ pairs ] heading of the topol.top file.
>
> Before, I used this strategy for a model with about 300 non bonded
> parameters and it worked fine. Now that I am using about 6400 non bonded
> parameters (all of them listed under the [ pairs ] heading), simulations
> take about 10 times longer to complete. There is still only one table for
> all of the contacts.
>
> Given that I am only changing the factor by which I multiply the geometric
> part of the potential for each pair (meaning, a single table is read), I
> wonder whether this large increase in computation time is expected. The
> Gromacs 4.6.7 manual (section 3.4 - page 15), when referring to energy
> monitor groups, states that:
>
> "All non-bonded interactions between pairs of energy-monitor groups can be
> excluded (see sec. 7.3). Pairs of particles from excluded pairs of
> energy-monitor groups are not put into the pair list. This can result in a
> significant speedup for simulations where interactions within or between
> parts of the system are not required.”
>
> Therefore, my question is whether adding a large number of pairs to the
> list (even when using a single potential table) can result in such a
> slowdown of the simulations.
>
> Thank you,
>
> Fernando Yrazu
> PhD student
> Rice University
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>