E.T. wrote:

> Now, if i don't sound too bothering, i would like to ask 'why' on those two big
> NO's ( about GPGPU and about AA on the GPU ), if that's not a huge explanation
> of course.

The idea is actually not so far-fetched. Although the traditional
interfaces for scanline rendering (typically via OpenGL or DirectX)
are not helpful for raytracing, the processing power which is idly
adding some extra heat to your system for most application just
begs to be tapped in a more direct way.

This process is still in its infancy, yet a range of NVIDIA cards
is already programmable in C (and not just for Windows), see

   http://www.nvidia.com/object/cuda_home.html

Of course, the architecture is very different from normal CPUs.
What they are really good at is massively parallel calculation
intensive stuff with few dependencies on global shared memory.
Doesn't sound too bad for offloading some raytracing duty.

But of course it would be a lot of work and it's still too
hardware specific. Wait a few years and it might just be a
further supported target platform for POV-Ray ;)

Post a reply to this message