[Nexus] Release of Version 2.0.0 (including HDF5 support)
Mark Koennecke
Mark.Koennecke at psi.ch
Thu Sep 19 07:03:59 BST 2002
Brian,
On Mon, 16 Sep 2002, Ray Osborn wrote:
> On 9/16/02 12:33 PM, "Brian Tieman" <tieman at aps.anl.gov> wrote:
>
> > Are there any driving reasons to use HDF5 over HDF4? Does it provide better
> > performance? Better compression? Easier to handle files? Increased
> > compatability? Etc...
> >
In addition to what Ray already mentioned: I have been forced to use
HDF-5 at two of our SINQ instruments because I hit HDF-4 limititations.
The first is a four circle diffractometer with PSD, operated like a
protein crystallography rotation instrument. With this HDF-4 permitted
me to store only 49 frames in one file because HDF-4's limit of 64000
objects. I did not have that many datasets but apparently almost
everything consists of many objects in HDF-4. HDF-5 solved this
limitation.
The second case is a reflectometer with a PSD operated in TOF mode. I
get datasets of up to ~200MB here. In order to keep memory
requirements low and keep some kind of reponsiveness of my instrument
control program I read this in chunks of 16MB from the histogram
memory and write it to file. HDF-5 now allows to use
chunked datasets (or dataset slabs) together with compression which I
obviously want on those datasets. HDF-4 only supports compression on
datasets which are written in one go. This is not working yet because
of a bug in HDF-5 but the NCSA people promised to provide a fixed
version in autumn.
This also means that HDF-5 and the new NeXus-API is stable enough to
use it in a production environment.
Regards,
Mark Koennecke
More information about the NeXus
mailing list