[gmx-users] time estimation script
steletch at jouy.inra.fr
Tue Feb 27 11:53:17 CET 2007
Martin Höfling a écrit :
> Am Dienstag, 27. Februar 2007 schrieb David van der Spoel:
>> but it says at the bottom of the md.log file how many ns/day you will get.
> ah ok, I know what you mean: Setting up a short run ~200 steps or so and then
> parsing md.log file. That's a good idea. My solution would have been killing
> mdrun, and parsing the last intact output line and then recalculate ns/day
> Thx for this hint.
This is what i've done with the script i've attached, but as a security,
i would also recommend verifying the real time taken.
For instance, if you use a queuing system (PBS, SGE, ...) then you could
look at the 'Wallclock' time. Most of the time the reported performance
by gromacs (amber does the same) is over-estimated since it does not
take into account some setup steps (but in your real life, you need
them). The difference is not very important but from experience there is
attention to port to it :-) (if there is a big discrepancy, you'll need
to find why)
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