AMBER Archive (2005)

Subject: RE: Re: AMBER: a trivial question

From: David LeBard (david.lebard_at_asu.edu)
Date: Fri Aug 05 2005 - 11:32:47 CDT


Dear Linchen,

I am not sure if this will help your problem, but this is my experience
nonetheless:

We are have two clusters, one is a cluster of Compaq Alphas and one is a
cluser of 64-bit AMD opterons, and both run Amber 8. For some reason,
the Alpha cluster will update the mdcrd file automatically at the
expected interval, but the opterons are different. For some reason, the
opterons must fill a memory buffer first, and once this buffer is full
of data (like that made for the the mdcrd file) only then will it write
to the file. This means that there is some lag or delay to writing to
the mdcrd file. Also this unfortunately means that if some power outage
or computer failure occurs, then the mdcrd file is always missing the
most recent data (that mdinfo will report because its opened and closed
every time). The problem is not isolated to Amber either, as the same
functionality can be seen with DL_POLY and many other programs.

To find what you indeed have in your trajectory file, you can always use
ptraj's trajin/trajout function. At the end of the ptraj run it outputs
the number of saved configurations to the screen so you no longer have
to guess.

Hope this helps,

David LeBard
Dept of Chemistry and Biochemistry
Arizona State University

On Fri, 2005-08-05 at 08:26, Linchen Gong wrote:
> Dear Carlos:
> I think nstlim/ntwx is a approximate estimate. For my question I found that the first 100000 step breeds 1000 frames, however the subsequent 100000 step only produces 999 frames.Further more, things become more strange when computer crash occurs, e.g a process was crashed at 11ps, restarted from 10ps and crashed again at 111.4ps, 0.002ps per step ntwx=200, ntpr=100. The whole process brings 252 frames, but I think it ought to be more. It seems that amber doesn't write mdcrd file every ntwx steps, instead, it delays a little.
> This question is quite trivial, but I want to know the answer to avoid some potential analysis error.
> Thanks a lot!
> ----- Original Message -----
> From:Carlos Simmerling <carlos_at_ilion.bio.sunysb.edu>
> To:amber_at_scripps.edu
> Subject:Re: AMBER: a trivial question
> Date:Thu, 4 Aug 2005 19:28:36 +0800
> > Linchen Gong wrote:
> >
> > >Dear amber users:
> > > I have a trivial question that I do not understand. If I run a 100000 step MD simulation with a writing frequency to mdcrd file of 100, then howmany frames will the mdcrd contains?
> > > If I run another MD simulation with the restrt file of the former one. the total number of step is also 100000 and mdcrd writing frequency is still 100, howmany frames will be there in the mdcrd file and why?
> > > Since my computer shut down automaticly sometimes, I want to know the answer.
> > > Thanks a lot
> > >
> > >
> > >
> > >
> > divide nstlim by ntwx.
> >
> >
> >
> > -----------------------------------------------------------------------
> > The AMBER Mail Reflector
> > To post, send mail to amber_at_scripps.edu
> > To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
> >
> >
>
> ______________________________________
> Linchen Gong
> basic science class
> math and physics oriented
> Department of physics
> Tsinghua University
> ===================================================================
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu