Jean-christophe Penalva wrote:
>
> Hello,
>
> my question is simple, is there somewhere a document about the
> source (directory, description, link between files). We'd like to modify
> some parts of the source for 2 reasons.
>
> 1 : we'd like to see if it's possible to use Volume rendering with
> hardware option on us SGI Onyx2. By now we are not sure that this is
> possible without completly recode the application... so we're looking
> for explanations or descriptions of the structure of the source.
You will not need to change any of the source, just write a
new image module. We've done several variations on this concept
over the past 6 years. Long ago, I wrote a CAVE module using
EVL's cave libraries. The core of the module simply descended
the DX data structure and made OpenGL calls when it encountered
triangles, quads, lines etc. It was not as full-featured as the image
module (no support for screen objects), but it could do the essentials.
I also had a student write an Image module that supported stereo hardware
and better mouse interface (interactive pan). That code was, unfortunately
lost a few years ago when the student left.
In addition to writing modules, you have the option of writing stand-alone
code that starts the dxexec and can load and run dx visual programs (DXLINK).
This is very similar in concept to MatLab's Engine API. I had a student
recently create a stand-alone program running in our VR environment that could
invoke
a DX program and display the graphics in our CAVE in real time. He had to
play fast and loose with DXLink, but as a proof of concept it worked well!
Ability to hook into the image window at a much lower level has now
been added in the form of SuperviseWindow/SuperviseState modules.
We are only just begining to explore this option. Another
functionality that has hardly been exploited is Image's picking.
Good place to start is the Programmer's guide. It's a bit old now,
but not much has changed at the module coding level.
---- while I'm thinking about Image
Here is my wish list for new image module features:
1. better mapping of camera and object transformations to mouse movement
(solved to some extent by SuperViseWindow etc.)
2. ability to use depthcuing and clipping planes interactively with mouse
(interactive subwindow showing side view of perspective cone with
draggable
front and back clip/depth planes)
3. updated OpenGL functionality: textures, volviz, reflection maps etc
4. better integrated picking. Click-highlite-drag-rotate parts of objects
in real time. Perhaps even basic building functionality with output
tab for modified structures
5. simpler, better integrated hardware stereo
---- perhaps I'll have my next student start a project like this
>
> 2 : SGI has developped a library (MPU) for use of their cave (and
> others periph stereo ...). The purppose is to keep the "same"
> application (that genereate OpenGL) and use it independantly with 1
> pipe, 2 pipes .... the pipe is the periph hardware on Onyx2 to use for
> generating images.
You must be referring to the ability of new Infinite Reality hardware to
devote multiple engines to one display. The highest graphics benchmarks
they cite are actually for a system in which every IR they can cram into
a box are crunching on one picture. It should be perfectly possible to
do this at the module level.
I should also mention that we have a current project underway to
create a new CAVE module using the open source VR Juggler program
(vrjuggler.vrac.iastate.ed). This environment may support many of the features
you want.
>
> What do you think of these ideas ? Is it possible for someone who
> know OpenGL to recode the parts of OpenDX ?
>
> Thank you for your answers, please cc to my email adress.
>
> ##################################################
> Centre Informatique National de l'Enseignement Superieur
> CINES Montpellier France
> 33 4 67 14 14 14
> Fax 33 4 67 52 37 63
> http://www.cines.fr
> mailto:[EMAIL PROTECTED]
