On Wed, Jun 17, 2015 at 12:28 PM, Gerald Monard <
Gerald.Monard.univ-lorraine.fr> wrote:
> Hi Daniel,
>
> A good point is raised here.
>
> I've made a small test, but from experience it's the same for bigger
> systems: I run a molecular dynamics and saved the same number of steps
> (ntwx) with ioutfm = 0 or ioutfm = 1, then I compressed the files. The
> results are the following:
> -rw-r--r-- 1 gmonard users 6777381 Jun 17 19:52 tmp.crd-ioutfm0
> -rw-r--r-- 1 gmonard users 2527101 Jun 17 19:53 tmp.crd-ioutfm0.gz
> -rw-r--r-- 1 gmonard users 3348408 Jun 17 19:51 tmp.crd-ioutfm1
> -rw-r--r-- 1 gmonard users 3004141 Jun 17 19:53 tmp.crd-ioutfm1.gz
>
> This means that indeed using ioutfm=1 gives a lower file size than
> ioutfm=0, but the compression of the ASCII file yields the smallest file
> size... which can be directly used by trajin.
>
> Does this mean that, if size matters, it's better to use ioutfm=0?
> What's the advantage, except size, of using the NetCDF format?
>
You've already gotten lots of reasons to use NetCDF instead of mdcrd (even
gzipped). All I'll do is explain why gzipped mdcrd files are smaller: the
coordinates are stored in MUCH lower precision (%8.3F format), compared to
full single precision in NetCDF trajectories. NetCDF trajectories contain
more information, so it's only natural that they're less compressible. (I
had noticed this exact same thing before and I worked out this explanation
back then.)
All the best,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 17 2015 - 21:30:02 PDT