[gmx-users] A few general questions on Gromacs
Stéphane Téletchéa
steletch at jouy.inra.fr
Tue Jun 26 15:56:07 CEST 2007
Corina Stratan a écrit :
> Hi,
>
> I would like to thank you and David for the answers. By "multiple
> programs", I meant batch scripts that contain other
> preprocessing/postprocessing programs besides mdrun (like grompp,
> pdb2gmx,...) - so this would be a case in which the processing is done
> in multiple steps. I asked this because we're especially interested in
> multi-step processes (workflows); and in the Gromacs tutorials most of
> the examples are like this (some other programs are invoked before and
> after mdrun).
>
> Thanks again!
> Corina
>
Your mileage may depend on your cpu availability, for instance, you can
do in one batch job either:
- grompp+mdrun
- grompp+mdrun(em1), grompp+mdrun(em2), grompp+mdrun(pr1), ...
Ideally, one would do (in order to share ressources:
- 1 batch file containing all the "setup" steps (those are quite "rapid"
usually), namely energy minimisation and positional restraints
- this batch job launches the next step (after some verification if you
need to) automatically, thus *chaining* them
- 1 batch file for the "production" (the one you'll use for your
analyses/statistics), which can itself be splitted.
That would be:
batch-setup --
|
->batch-production1 (--
|
->batch-production2...n).
The more you chain, the more you'll need to validate your steps :-)
For the benchmark in itself, you can easily generate a *big* box of
water but you'll not really benchmark the system even with such a big
system if your interconnection is not fast enough (i.e. you have fast
network *and* low latency, gigabit for instance is not very efficient
for a good scaling).
Cheers,
Stéphane
--
Stéphane Téletchéa, PhD. http://www.steletch.org
Unité Mathématique Informatique et Génome http://migale.jouy.inra.fr/mig
INRA, Domaine de Vilvert Tél : (33) 134 652 891
78352 Jouy-en-Josas cedex, France Fax : (33) 134 652 901
More information about the gromacs.org_gmx-users
mailing list