From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Thu Oct 07 2010 - 10:07:53 CDT
On Thu, 2010-10-07 at 13:40 +0200, Bjoern Olausson wrote:
bjoern,
first off. an answer to your previous question on the dcd format.
the size of a dcd file is determined by its header, which is a
constant for all files, and then a block with coordinates and
(optionally) the unit cell dimensions and angles that scales
linearly with the number of time steps stored.
[...]
> I know that the results might diverge since job is running on 288 cores and
> this raised a question:
>
> How large must the difference be to be of relevance?
>
> I mean in what case would it be noticeable that a 1ns slice diverges slightly?
after a nanosecond, you will see significant differences.
MD is a chaotic process. as soon as you have the tiniest
difference - and the combination of floating point math
and (non-deterministic) load balancing in NAMD will guarantee
that - the divergence will grow exponentially.
> Is there some kind of test to check if all the runs are sane?
> Maybe a energy peak, jump in motion, ... while proceeding from the last
> coordinate set from the recalculated trajectory to the first coordinate set of
> the following correct trajectory?
for the reasons from above, there is not much point in redoing
the missing parts (unless for making nicer animations).
if you are looking at properties that are just averaging over
the whole data set, you can analyze the pieces that you have
just as well. if you are looking for dynamical properties, where
you need consecutive frames, then you cannot "make up" the
missing pieces. you better bite the bullet and either only
look at pieces, or pick the longest consecutive segment that
you (still) have and continue from there, and regard the
rest as very thorough equilibration.
cheers,
axel.
> [] Trajectory (dcd-file)
> - Coordinate set in the correct trajectories
> * Coordinate set in the recalculated trajectory
>
> [-----][*****][-----]
>
> Anything to check for while proceeding from the last * to the first -?
> How to determine the threshold for "Too large difference to be consecutive"
>
> Cheers
> Bjoern
>
-- Dr. Axel Kohlmeyer akohlmey_at_gmail.com http://sites.google.com/site/akohlmey/ Institute for Computational Molecular Science Temple University, Philadelphia PA, USA.
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:34 CST