Does Darkplaces/Nexuiz support Levels of Detail models for the player models, are there any plans to support this in the near future, or would plans be made if player models with LoDs were contributed under the GPL?

I have recently started work on player models for Nexuiz, and it would be most helpful if I knew that I should shoot for high and low budget models with LoD in mind or more like medium budget models if I can't count on any LoD support from the Nexuiz engine.

I don't know if LoD is supported (I haven't heard of such a thing at least), but if it isn't try not to get above current poly count anyway. As more and more stuff is being made things starts to get intensive for the computers. 3000 is in my opinion the roof of how much poly's to have. But I suppose a more involved person in this, like a dev, could come fourth and state his opinion.

Some1 recently said to me that LoD is possible with DarkPlaces, there are 2 LoD supported. The bad thing is that they aren't used :/

Using this tecnique surely will improve the performance of the game.

DP model LOD is not automatic - the models will need to be manually duplicated and reduced in detail for each LOD. This will reduce the number of triangles to draw, but since the models aren't that complex to begin with, it probably won't yield sufficient return to justify the work.

LoD support exists for map details (as used on the map "yourewinner", and for player models it is supported but commented out due to the models not supporting it, but it taking noticable server CPU.

Enable it by compiling with the option -DALLOW_VARIABLE_LOD and naming models like

pyria.zym
pyria_1.zym
pyria_2.zym

It is not good yet though, e.g. it allows using pyria_1.zym as normal player model. In case we EVER get usable LODded player models, this will get fixed.

divVerent wrote:In case we EVER get usable LODded player models, this will get fixed.

Excellent.

ai wrote:I don't know if LoD is supported (I haven't heard of such a thing at least), but if it isn't try not to get above current poly count anyway. As more and more stuff is being made things starts to get intensive for the computers. 3000 is in my opinion the roof of how much poly's to have. But I suppose a more involved person in this, like a dev, could come fourth and state his opinion.

This might sound like a dumb question, but 3000 whats? There's a big difference between quad count and tri count; there's quite alot I could do with 3000 or (now being able to count on LoD support) 5000 quads.

So that is my next question:

Does Darkplaces/Nexuiz support (untriangulated) Quad meshes for the player models, are there any plans to support this in the near future, or would plans be made to support this if player models with Quad meshes were contributed under the GPL?

Poly's = Tris.
Games count in tris, so when one says 3000 polygons it's 3000 tris.

And yes, you can have quads on your mesh and it works just fine in Nexuiz.

ai wrote:Poly's = Tris.
Games count in tris, so when one says 3000 polygons it's 3000 tris.

I have heard that 1 quad is a good deal less demanding than two tris in a modern game engine.

And yes, you can have quads on your mesh and it works just fine in Nexuiz.

That's great then. I was getting the impression only tris were supported since all of the current character models appear to be only tri meshes and with the old meshes there wouldn't be much incentive to update any quake 1 mesh limitations like support for tri meshes exclusively.

Flying Steel wrote:

ai wrote:Poly's = Tris.
Games count in tris, so when one says 3000 polygons it's 3000 tris.

I have heard that 1 quad is a good deal less demanding than two tris in a modern game engine.

I have no idea about that, don't know that much about game engines. Would be nice to find out though.

Depends on whether quad is tessellated into 2 triangles (by the gpu), or not (manipulated directly). OpenGL has specific command to draw rectangles (glRect), which could/should be optimized more than drawing two GL_TRIANGLES or one GL_QUADS, but not all quads are rects.

You could experiment with creating very high poly model from quads and another from triangles. (quad_count = triangle_count). We could test this on different GPU's, set of drivers (OpenGL implementations).

Alien wrote:Depends on whether quad is tessellated into 2 triangles (by the gpu), or not (manipulated directly). OpenGL has specific command to draw rectangles (glRect), which could/should be optimized more than drawing two GL_TRIANGLES or one GL_QUADS, but not all quads are rects.

Which just leaves the question of how does Nexuiz treate a quad, as a quad or a couple of tris?

You could experiment with creating very high poly model from quads and another from triangles. (quad_count = triangle_count). We could test this on different GPU's, set of drivers (OpenGL implementations).

That's a good idea, I think I'll do that when I know how to get models into and working in Nexuiz.

I doubt darkplaces use software tessellation of quads. There would be no obvious reason, but either lh or div knows better.