[vtk-developers] Rendering of vtkPolyData broken on Nvidia systems

[Marcus D. Hanwell]

On Fri, Jun 22, 2012 at 7:34 PM, Sean McBride <sean at rogue-research.com> wrote:
> On Fri, 22 Jun 2012 12:30:36 -0400, Cory Quammen said:
>
>>Sigh. Nothing like removing useful features from a widely used API for
>>no good (in my mind) reason.
>
> Indeed.  However, looked at in another way, it is perhaps also an indication that VTK is not keeping up with the evolution of OpenGL.
>
If you search for display lists and nVidia you will find results such as,

http://www.nvidia.com/object/linux-display-ia32-295.20-driver.html

This shows bug fixes for Fermi based GPUs where display lists could
result in "missing rendering" in some cases. While there are certainly
advantages to using VBOs, I don't see nVidia dropping support for
display lists any time soon. If there are rendering bugs in the
polydata mapper we can work to track them down, but in my tests I was
not able to reproduce the rendering issue - it could be driver version
and/or GPU specific.

In terms of keeping up with the evolution of OpenGL, I am very eager
to add in functionality to support VBOs and other new GL features. I
think there is some exciting work in related projects that we would
like to see appear in VTK in the near future. I know we have been
doing some work in VES for example with VBOs on mobile GPUs (GL ES
2.0), and we have also been working on a GL 2.1 desktop rendering
pipeline with no fixed function calls (very similar to ES 2.0) for
efficient rendering of large chemical structures.

There is a lot more to do before these approaches can handle
everything VTK currently supports.

Thanks,

Marcus