Volume rendering, more maps supported.

[wyszolmirski]

Hi,

I though that it's good to start this thread. To start with, is there any chance to see current to-do list / road map for RT ?

And here comes the wishlist. At the moment I would love to see:

1. Volume rendering. That's would make me most happy at the moment.

2. More procedural shading features including some random texture / colors. Also improving things that don't work now as well when used in weird combination like plugging composite to composite freaks RT out . Most issues are with bump, that basically works nicely with only one bitmap (sometimes it doesn't for some reason).

3. I would also love if RT would get more priority, for example if there is some new feature in Adv, it gets implemented to RT CUDA as well.

One final thought: Once the cards get more powerful and get stocked with more memory I don't see a reason to render on CPU. Just a waste of money and time .

Anyway good job so far.

Best,

Tomasz

[vlado]

We don't usually give details about roadmaps; they almost never seem to be followed anyways

With that said, we actually did do some experiments with volume rendering; it worked, but not in a satisfactory way so I left it to go back to it at a later time. We also had a branch with bump enabled for procedural textures, but that needed more work as well. Eventually we will get back to it.

Better support for shading nodes is certainly coming and is something that we are always looking into. Keep in mind that we don't have just the 3ds Max shading nodes to look into - we must also support the various Maya shaders; at some point also the various Modo shaders and so on. At some point I really want to get OSL working on the GPU so that these things can be done in OSL, rather than us manually changing the GPU code every single time. OSL will also give 3rd party developers the chance to write textures compatible with both the CPU and the GPU renderers.

We also have two potentially major refactors of the GPU code that we need to do. One is related to the fact that AMD basically told us that our current GPU code is more complicated than what their GPUs and drivers can handle and we need to split it into several smaller GPU programs. I'm somewhat sceptical this will change anything, but we have to at least give it a try, even if this means that we have to rearchitect the whole thing. The other change is related to the ability to evaluate multiple light sources per shading hit. Right now the GPU renderer uses an extreme version of the probabilistic lights. This works well for scenes with a relatively small number of lights, but is very inefficient for more complicated scenes and something that we need to resolve.

Best regards,
Vlado

[wyszolmirski]

Thanks you for very detailed response.

I'm looking forward to what new builds will bring and hope that keeping engine compatible with Nvidia and AMD cards won't slow development too much. Keep up the good work.

Best regards,

Tomasz

[super gnu]

"AMD basically told us that our current GPU code is more complicated than what their GPUs and drivers can handle and we need to split it into several smaller GPU programs."

sounds like a far more sensible option would be for AMD to sort it out! It doesnt sound ideal for their dreams of opencl dominance if their cards cant handle complex code as well as the opposition.

presumably their next generation of cards will be better in this regard, unless they have really dropped the ball on gpu compute.

[Alan Iglesias]

Wow...what you have to do to create a little competition around here! I wish AMD would just get it's you-know-what together with GPU rendering and give nVidea a run for it's money!

Sheesh...

-Alan

[glacierise]

+1 for volumeGrid support! That would be awesome

[Moriah]

any ETA on Volumegrid on GPU? Maybe you'll show us something at GTC? :P

[vlado]

We needed to get other things in order before we could add the volume code (which we already coded it a while back) but we'll get there in a few months.

Best regards,
Vlado