>hi jeff
>
>i have successfully imported netCDF files with DX, up to 8.9meg file,
>however my 16meg file still gives me trouble.  is there something i
>should be doing with setting the memory.  i setenv DXMEMORY 128 before i
>run DX, and i also setenv DXSHMEM 0.  what does DXSHMEM do, btw?
>
>do you know of any file size limitations in other aspects of DX that
>might be haunting me here, ie why can it import an 8 meg file, but not a
>16 meg file?  this makes me think that DX has file size limitations, but
>i can't find any reference to this in the onlince documentation...
>
>thanks
>rob
>


There are no  file size limitations per se (other than OS limits), but
everything that DX creates must fit in available memory: Import is not
passive: it is creating (a) DX object(s) as it parses the input file. Jeff
suggested some possible reasons that relate to the structure of your
object. For example, a client gave me a 1.4M vertex grid, irregular
topology, thus every single vertex had to be loaded as a 3-vector float
value. In addition, the topology (connectivity) of the vertices had to be
loaded: since they were cubic, that took 8 integers for each cube element.
There were several data values associated with each vertex: so add in 1
float per scalar data value per vertex; some were vectorial, so add in 3
floats per data vector value per vertex. Coloring it added 3 more floats
per vertex, as did normals for shading.

Then the big gotcha! That was 1/12th of a cylindrical symmetric object, so
he wanted me to reproduce the above critter under rotation 12 times and
knit the pieces together. Yikes! It just fit in a 1Gb memory machine.

Another example: a client gave me an AVIRIS remote sensed image file,
nominally 140Mb, but structurally, 224 layers of about 600x500 (I forget
the details) images of grayscale (byte). DX thinks images are grids, so
would have tried to create a 224x600x500 volume. No way, jose'! (not in my
memory, you don't). So I used a prototype (now dead and gone) scientific
database program to preselect smaller volumes or slices before feeding them
into DX. Moral: DX is NOT a database program, so you may very well have to
preprocess your data differently than it currently is to VISUALIZE it which
is what DX is for.

So, without any idea of your netcdf file's contents, two ideas: if you are
describing some godawful huge grid, you may be out of luck on your current
hardware. If you are instead importing a series file, in which each time
step could theoretically fit in memory one (or more, but not all) at a
time, you need to cut it up in advance. That's a shame since IF you could
import it into DX, you COULD then output a series file for working on a
smaller memory machine. But since you can't read it all in, you can't use
any functionality of DX (yet).

Chris Pelkie
Vice President/Scientific Visualization Producer
Conceptual Reality Presentations, Inc.
30 West Meadow Drive
Ithaca, NY 14850
[EMAIL PROTECTED]
(607) 257-8335 or (607) 254-8794

Reply via email to