> > First of all I would like to congratulate both the developers and the
> > users for making a rather impressive application framework. It is in many
> > ways significantly better than anything I have seen to date for scientific
> > applications.
> >
Yes, ROOT team has deserved to receive not only bug reports. Here is
another modest contribution.
A week ago I used ROOT to make the slides for a presentation where quality
of the slides was quite an issue. Thanks Rene, Fons, Valerij and Nenad!
Great job! ... Bug report will follow ...
> >
> > 3) Have there been any benchmarks relating non-application framework
> > programs to ROOT ported ones? On the ROOT web site there are several
> > tables, but none seem to provide an idea of the performance hit
> > endured by programs that use ROOT. (Please do not tell me that ROOT
> > makes special purpose software faster-- the only time this happens is
> > when the original program is shoddily written).
> >
Here is one of the user comparisons. Not on the performance though...
For many years HEP community has been
using a FORTRAN-written histogramming package called HBOOK which among
many other features provided tools to store ntuples/histograms on disk
in an architecture-independent format (it should be mentioned that
Rene was one the authors of HBOOK...). Having converted one of my
HBOOK ntuple files into ROOT format I obtained a compression factor
of about 12 (!):
-rw-r--r-- 1 murat cdfupg_d 10788864 May 22 11:51 mbr_proc_001.3.hrz
-rw-r--r-- 1 murat cdfupg_d 889792 May 26 20:25 mbr_proc_001.3.root
Some details on this comparison. Default settings for all the parameters
affecting disk space usage have been used for both HBOOK and ROOT.
The HBOOK file contained ntuples with 1- and 2-dimensional arrays,
significant part of the ntuples is in fact empty and ROOT compression
procedure took care of it. But the result is impressive nevertheless.
Pasha.