Hmmm, highly reliable restart writing must be something added to sander
since amber 6, and I just have not picked up on it. I can certainly improve
the robustness of this for the next release, at least doing whatever sander
does. I presume that this is a really really rare event that someone gets
clobbered by this, though? I would expect that the "vulnerability window"
is pretty small unless you push the rate of restart file writing way up;
pmemd intentionally backs off the default restart write frequency as the
processor count goes up because it is generally inefficient and unnecessary
to be writing this out all the time (so say you only write once every 5000
steps, and lose the last 4999 steps in a crash - big deal, basically).
Anyway, I'll take it under consideration.
Regards - Bob Duke
----- Original Message -----
From: "Austin B. Yongye" <ayongye.chem.uga.edu>
To: <amber.scripps.edu>
Sent: Saturday, May 12, 2007 2:21 PM
Subject: AMBER: PMEMD and restart files
> Dear amber users,
> I noticed that PMEMD writes restart files during an MD simulation, then
> deletes their contents before writing new restart coordinates again. This
> is an issue when a simulation is abruptly terminated when there are no
> coordinates in the restart file. In an instance whereby the coordinates of
> water molecules are not written in the trajectory file, it becomes really
> problematic because the restart file can no longer be generated from the
> last frame of the simulation.
> I was wondering if/how the pmemd code could be modified so that the
> restart file is written as in the sander code.
>
> Austin B. Yongye
> Complex Carbohydrate Research Center,
> University of Georgia,
> 315 Riverbend Road,
> Athens, GA 30602.
> Phone: 706 542 0263
>
> "...then, we learn better in a free spirit of curiosity, than under fear
> and compulsion..." -St. Augustine of Hippo
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun May 13 2007 - 06:07:51 PDT
