[gmx-developers] Re: Shall we ditch gro and g96 files?

[Brian Helfrich]

> Date: Tue, 1 Apr 2008 14:34:38 +0200
> From: Erik Lindahl <lindahl at cbr.su.se>
> Subject: Re: [gmx-developers] Shall we ditch gro and g96 files?
> ...
>
> Mathias/John, do you or anybody else have any experience from using
> HDF5 for development? Have there been different library versions that
> you need to install, or do packages usually include their own copy of
> the library?
>

Hi, we have an experimental version of GMX 3.3.2 that writes HDF5 trajectory
files and it works quite well. In fact, we're bundling it (in source and
binary form) with our (free, open-source) molecular cad tool which ships the
end of this month. Our goal is to make our implementation very robust, then
submit it for consideration to become part of the official GMX code.
Something like a --enable-hdf5 configure flag would turn it on in the build.

Some notes:
- We've built our GMX+HDF5 on Mac (10.3.9 thru 10.5.2), Windows XP, and
several Linuxes without a problem. We haven't tried Solaris, or AIX, etc.
- HDF5 supports parallel I/O very well with one small caveat: you can't turn
on I/O stream compression - it isn't that big a deal
- Using HDF5 to write trajectories allows us to view them in real time with
a trajectory player. This is nice for us since we're minimizing large, novel
structures and can quickly tell how a minimization is progressing and abort
it if it's going the wrong way.

In addition to the trajectory data, we also include the input parameters,
some input files, simulation results measurements, and some notes in text,
all in the HDF5 data store. The purpose for this is so that the HDF5 file
becomes a compact simulation results package that we can send to each other
for viewing or repeating. This scheme does extend the official GMX file
format implementation somewhat, and does affect more of the code than a pure
file format implementation would, but it's experimental and can be scaled
back to a pure GMX file format.

Brian Helfrich
www.nanorex.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20080401/95e6a3de/attachment.html>