Michael,
> From: "Plonski, Michael" <mplonski@xxxxxxx>
> To: "'John Caron'" <caron@xxxxxxxxxxxxxxxx>
> Cc: <mplonski@xxxxxxx>
> Sent: Saturday, September 15, 2001 11:30 AM
> Subject: RE: netcdf and multiarray directories in Netcdf ver 2 .vs. Ver 1
The above message contained the following:
> > I haven't tried it yet because I am about to give up with VisAD.
> > I thought if I got it to read netCDF better (read pieces of large
> > files), it would be more useful to me. I have some home grown netcdf
> > files that are typically in the several hundred Mb range with X,y
> > dimensions in the order of 2000 x 1500 and then some meteorological data
> > with bands and other with levels. Even running VisAD with something
> > like -mx800m, it still just croaks on things. I was trying to modify
> > the ingest so that I could just read a few netcdf variables at a time
> > using your version 2 netcdf. I wrote up some code that just lets me
> > extract one variable at a time from the file, without using the visad
> > "plain". Plain() reads the whole files and just croaks on my data
> > files at the start. The problem is there seems to be some memory leaks
> > in VIsad and even after you get the data in, memory still seems to go up
> > way too high and things really run slow. I am surprised how slow visad
> > runs on a Sun Ultra 60 - it seems to do much better on a Pentium 4.
You can reduce memory-usage by telling the netCDF-import
module to maximize the number of VisAD FileFlatField-s
used to represent the data. This is done by setting
the Java property "visad.data.netcdf.in.Strategy" to
"visad.data.netcdf.in.MaxFileFieldStrategy". For example:
$ java
-Dvisad.data.netcdf.in.Strategy=visad.data.netcdf.in.MaxFileFieldStrategy ...
Additionally, a version of VisAD which should be released soon will have
the ability to import an arbitray subset of the parameters in a local
netCDF file.
Regards,
Steve Emmerson <http://www.unidata.ucar.edu>
