[Paraview] Integrating new volume renderer into VTK/Paraview

[Steven Callahan]

We are working on integrating the Hardware-Assisted Visibility Sorting 
(HAVS) volume renderer into VTK and ParaView.  We have a prototype 
version that performs at about twice the speed of  Projected Tetrahedra 
with about the same image quality as ZSweep, incorporated into 
ParaView.  (ZSweep has much better image quality than PT.)

We would like to know what steps we need to take to get this added to 
VTK and ParaView.

Our implementation is a new volume mapper 
vtkHAVSUnstructuredGridVolumeMapper that can be used in place of any of 
the other unstructured grid volume mappers.  The actual code consists of 
the mapper class along with several shader files which are written in 
ARB assembly and work with ATI and NVidia video cards.  The code also 
supports a new high quality, dynamic Level-of-Detail approach by passing 
in a desired frame rate...

For integration with VTK, we see the following issue: Our code requires 
the most recent OpenGL extension header files.  VTK currently uses 
headers that are slightly out of date.  The newest headers are always 
available at the SGI website 
(http://oss.sgi.com/projects/ogl-sample/registry/).

We have integrated our code into ParaView.   For screenshots comparing 
PT with HAVS, see http://www.sci.utah.edu/~stevec/images.

For complete integration with ParaView, we see the following issues:
1.  We do not understand ParaView's LOD selection.  Thus, we have not 
incorporated our LOD technique inside ParaView.  We would appreciate 
getting help from someone that knows the details of how this works.
2.  Although our code works on both ATI and NVidia, ParaView does not 
seem to use hardware-acceleration on machines with ATI cards.  Even with 
the precompiled Windows binaries from the web, the PT algorithm is 
extremely slow, indicating that hardware rendering is not being used.  
Is this an intrinsic limitation of ParaView?

Any help or advice on getting this code into VTK and ParaView would be 
helpful.  Thanks

Steve Callahan and Claudio Silva

PS. We also have an efficient software volume renderer based on the same 
algorithm that we would like to incorporate in the near future.