Re: [AMBER-Developers] netcdf restart files?

From: case <>
Date: Thu, 6 Jan 2011 08:14:30 -0500

On Wed, Jan 05, 2011, B. Lachele Foley wrote:

> Has there been discussion of making it possible to read and write
> netcdf-formatted restart files? If not, can I start that discussion?
> If so, is it planned or already available?

a. We already have an "unformatted" restart file option, which might be
fine since the data inside a restart file is very simple, and might not
need the complexity of a netcdf wrapper. I couple of revisions ago, this
option was disabled, but that was probably a bad decision. I have made
repeated requests for developers to turn it back on and try things out,
since I don't think it is broken, just disabled. So far, there have not
been any takers, which may reflect how much influence I really have....

b. There are advantages to going to a more elaborate netcdf format, since
libraries for reading/writing these files are now routinely included in
sander, ptraj and NAB. (We were reluctant to do this earlier, since netcdf
did not compile cleanly everywhere.) If someone wants to work on this, that
is fine as well.

I think this is an important thing to do, since it would eliminate (to some
extent) the need to set iwrap in sander/pmemd and remove the dreaded ***
problem. It would also remove the small amount of noise that occurs on
restarts due to truncation of precision in current (formatted) restart files.

The downside (and the reason we turned off binary restarts in the first place)
is that increases the complexity of everything: having two different restart
formats means that every program that interfaces with these needs to know
about both options. This includes not only things we control (ptraj, mm-pbsa,
NAB....) but programs written by others (VMD, Chimera, ....).

But I still think it is worth doing.


AMBER-Developers mailing list
Received on Thu Jan 06 2011 - 05:30:02 PST
Custom Search