[GSoC 2011] Nextgenify (as in 'go nextgen')

[so0os]

Since GSoC ideas started poppin up, maybe I add my two cents.

My idea is to commit a bundle of changes that will update Ogre with latest graphics rendering tech support, so it can be used to produce AAA looking games and visualisations. I may be missing a few facts, but that's what comes to my mind so far:

- Finish Direct3D 11 RenderSystem
- Provide an easy deferred shading solution, that can be enabled as easily as changing a scheme for a viewport. It's not that bad, but I guess providing DS support straight of the bat is essential for this task.

So.... is the idea worth anything?

[DavlexDesign]

G'day so0s,
Wait for the next incarnation of this demo, I think you might be pleasantly surprised, and hopefully some of this stuff could end up in the mainstream and help the guys get just that (NextGen).

Alex

[spacegaier]

Wait for the next incarnation of this demo

Which demo? Am I missing something?

[DavlexDesign]

G'day spacegaier,

Which demo? Am I missing something?

The OGRE ROAM planet engine demo

Alex

[so0os]

Cool But what about the idea? Is it even a good one? Also, isn't it too much?

[Deathgunner]

With a completed directx11 implementation, Solid wireframe could be implemented

link: http://developer.download.nvidia.com/wh ... eframe.pdf

[mkultra333]

Moving toward DX11 sounds like a very good idea. Essential, even. While DX9 might still be big, I don't think programmers will be able to get away with it for that much longer. With so much emphasis being given to hardware tesselation, it becomes an obvious Ogre blindspot. (Ogre doesn't support hardware tesselation atm, does it?)

Although that's probably too big for a GSoC project. I don't know, depends how much leverage you get from the existing DX work.

Deferred shading I'm not so excited about, because you can already set it up yourself. I use my own version of it and think it's better when you customize it to your exact needs. Not that it would hurt to have it simplified, but it wouldn't seem like the engine was being pushed forward that much.

[spacegaier]

Well, Assaf is already working on some DX11 parts: http://www.ogre3d.org/forums/viewtopic. ... ad#p418676

Those things are on his plate:
1. multi-GPU support for GL and D3d11.
2. Tessellation shaders.
3. Texture array support.
4. Samples compatibility with D3D11.

But I guess parts of it or other things in that area could be perfect GSoC projects (if the size matches the time).

[DavlexDesign]

Cool But what about the idea? Is it even a good one? Also, isn't it too much?

Of course it's a good idea, differed rendering also, as I think it is the way to go for inside environments, if you want believable lighting, then you need it I think.
As for the rest of the stuff, well, anything that improves efficiency and performance has to be a good thing, don't you think.

Alex

[madmarx]

Provide an easy deferred shading solution, that can be enabled as easily as changing a scheme for a viewport.

Concerning Deferred shading (I will write 'DS'), there was already a soc on the subject, and it went correctly and the student is now member of the team (Noman).

I think that the real world problem for DS is to create all the different shaders (ex: if you have a material with 1normal map, 1 env map, 1 lightmap on coordtex 2, 1 diffuse, 1 projected shadow, then you need another one if you have only 1 map that differs, or if there is transparency etc...). And after the shaders, you need to create the pipeline for the content (<=> exporters that use the new pipeline). Each DS project has different needs (motion blur? specular? hardware skinning? refraction?), so writing the shaders and exporters become the limiting factor IMHO.

IMO, the RTSS system is a good thing to correct this (even if I never used it myself).

[Noman]

I think the two ideas are both good, but each one of them is big enough to be a SoC project.
Finishing a rendersystem is a lot of work, especially since many 'minor' features need to be also tested.
As for deferred shading, I agree that the correct thing to do is to adapt the RTSS to work in the deferred pipeline. This is an interesting project, but is also big enough to be independent.

[al2950]

I agree with Noman, I think adapting the RTSS to work in a deferred pipeline is a great idea, infact I have been looking at it myself. However it is quite a large task and dont have much time to play !

[fzh]

Having been using Ogre and in the forum (lurking - as could be seen from post counts...) for a little over 4 years, from the perspective of an user, D3D 11 support, especially hardware tessellation support is something I would really love to see included in Ogre. In fact, I wanted a free open source D3D 11 3D engine so bad it's the cause for my first attempt at a 3D engine directly from d3d api, which didn't go too well considering how much work it takes to have an engine as good as Ogre. Will be participating in this year's GSoC as well, hope to have a chance to work on this great engine.

[jonim8or]

Maybe implementing a Deferred version of PSSM would be part of the NextGenification...
Deferred PSSM: I thought it might be interesting to render the frustra of PSSM shadow cameras as render shapes, similar to spot lights. In this way the shadow texture can be re-used