From: David Hardy (dhardy_at_ks.uiuc.edu)
Date: Wed Apr 08 2009 - 09:33:40 CDT
I've got a suggestion.
Run a simulation that produces a saved binary restart file every 100
steps, then do post-processing in which you start a NAMD simulation
that runs for just 2 steps from each of the binary restart files to
produce a DCD file, then join the resulting DCD files together. You
might have to adjust this recipe because I recall that DCD files
don't store the zeroth frame.
On Apr 8, 2009, at 8:23 AM, Axel Kohlmeyer wrote:
> On Wed, 2009-04-08 at 12:44 +0300, Mert Gür wrote:
>> Writing every 1 step gives me an huge DCD which is hard to handle
>> for short simulation times.
>> How about I perform 3 simulations at the same time.
> won't work. particularly, if you run the individual runs
> as parallel jobs.
>> I have a previous simulation and I will use its restart files so that
>> my initial system conditions are specified.
>> 1- First simulation I run with a DCD frequency of every 100th step
>> 2-For the second simulation, I first run my simulation for 1 step and
>> then use the final restart files for my second simulation and get the
>> DCD frequency every 100th step
>> 3--For the third simulation I first run my simulation for 2 step and
>> then use the final restart files for my third simulation and then get
>> the DCD frequency every 100th step.
>> So I need these simulations to be identically the same so that I will
>> get the consequent coordinates as I desire. In other words the 100.
>> step of my first simulation should be the 99. of the second and the
>> 98. of the third.
>> As much as my understanding goes that should be possible. In case it
>> is what do I have to be carefull about?
> please update your understanding. it has been discussed on this list
> many times that subsequent (parallel) runs are sampling the same
> ensemble, but are not guaranteed to be binary identical. due to
> inherently divergent nature of coupled partial differential equations,
> that you (only) solve numerically, trajectories will always diverge
> again, if you would explain _why_ you need this strange output
> frequency, perhaps somebody can suggest an alternate solution.
> as it stands, you have no choice but to modify the NAMD source
> code and introduce your own flag for your purposes.
>> Would it be enough if I set only the following parameters to 1?
>> nonbondedFreq 1
>> fullElectFrequency 1
>> stepspercycle 1
> this would be _very_ wasteful as you would not only run the same
> simulation 3 times, but also run each of them highly inefficiently.
> and yet it would still not sufficient. if you want this output so
> badly, you better get your preferred text editor and compiler ready
> and attack the sources.
>> On Tue, Apr 7, 2009 at 2:24 PM, Mert Gür <gurmert_at_gmail.com> wrote:
>> Hi everyone!
>> I want to write the coordinates of 3 subsequent time frames
>> into my DCD file.
>> For example
>> Lets say I have a DCD writing frequency of every 100th step.
>> Hence I am writing the coordinates for the 100.time step and
>> then for the 200. timestep......
>> But what I need to do is to write my coordinates for the
>> 101, 102. timesteps and then for the 200., 201, 202. time
>> and then again for the 300., 301. ,302, timesteps and so
>> Can I do that?
>> Any suggestion or even alternative ways to do that would
>> be of
>> big help to me.
>> Center for Computational Biology and Bioinformatics
>> Koc University
>> Sariyer 34450, Istanbul, Turkey
> Axel Kohlmeyer akohlmey_at_cmm.chem.upenn.edu http://
> Center for Molecular Modeling -- University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA
> tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
> If you make something idiot-proof, the universe creates a better
-- David J. Hardy, Ph.D. Theoretical and Computational Biophysics Beckman Institute, University of Illinois dhardy_at_ks.uiuc.edu http://www.ks.uiuc.edu/~dhardy/
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:52:34 CST