TVSM

[Satanka_Patka]

Hi

I'm also implemented TVSM  into my project, but I have 2 question.

Must I create a low poly model for all my high poly models?

What is mean exactly tvsm advancedscene  scenenminimum and scenemaximum ?

I think the occluder bounding box.

My problem is that, I have a big ground in my game, so when I use TVSM,  sometimes the ground is disappear.

[GD]

Must I create a low poly model for all my high poly models?

No, not needed. It's optional to speed up occlusion culling. In occ. cull. biggest objects on screen get rendered to test which objects are hidden. So it's good to have occluder low poly (to be rendered fast).

What is mean exactly tvsm advancedscene  scenenminimum and scenemaximum ?

It's the bounding box of your world. Same as TVPhysics.SetWorldSize. TVSM and TVPhysics work the same way, they need to know the maximum world size to slice it.
This "feature" in new versions will probably be deprecated. New space partioning algorithms will be added!

My problem is that, I have a big ground in my game, so when I use TVSM,  sometimes the ground is disappear.

One of the reasons can be the above question; you set too small world size?
Hope I was of help.

[Satanka_Patka]

Thanks GD! 

Your answers were very helpful for me.

I have a mapeditor and I set the physic wordsoze boundingbox in it.
When I use the tvsm advancescene I set the wordsize to same size as physic wordsize, but then   I get a big black screen, because everything is hidden.

 

[DavidAWinter]

Could you use this code on only part of a scene?

I'm reworking my stadium models and the fans in the seats need to be dynamic. My current system to get 70,000 + fans in the seats is to draw many fans on a single texture (whole section rows) and then just render mesh strips that have all the fans.

You can get a sence of what I'm talking about in this screen shot;
http://www.maximum-football.com/2/artwork/ss11.jpg

There are about 60,000 fans represented in this stadium.

I'd like to separate each fan into it's own mini-mesh. This would allow me more options as far as fan animation and textures goes. It also means the fans can properly face towards or away from the camera.

Obviously doing a for loop 60,000 times is slow, but I'm wondering if your DLL would help with that.

thanks
David

[GD]

Sure it's possible to put any part of the scene in TVSM, and the other part cull/render manually. You have nothing to lose if you try since it's only couple of more lines of code.
But I would recommend you to rather use minimesh - split your crowd in 2 dozen of minimeshes, I'm sure it will be fast.

Announcement:
I havent programmed this for a long long time amd seeing it now, I'm not satisfied with the current state and coding of TVSM. So I announce core rewrite with multiple spatial techniques, extended features and more generic library so any engine and language can use it. During this winter I hope. We'll see... Maybe not. Probably yes. Dunno, I have to consult with my boss Raine.

[petrus]

Can we not also check if a pixel is occluded by another pixel. Kinda something like shadow mapping from the camera's point of view. If a pixel is behind another pixel then, in the shader, simply return the color as float4(0.0f,0.0f,0.0f,0.0f) don't even worry about calculating specularity and normals. Thus instruction counts on the GPU decrease dramatically.
I don't know if anybody tried this... I should actually look it up. Else I take patent thanx 

[Raine]

I think that's common in deferred rendering

[Zaknafein]

Yeah, in fact I don't think this can be done without doing deferred rendering.

The idea is to make a first pass that does as little computation as possible, but does the Z-ordering and outputs only the stuff that the camera sees.
Then other passes (or another pass) is made with this data to calculate the lighting as a post-process, fullscreen shader, and it's done for each pixel that the camera sees instead of for each object that's in the frustum, thus removing the possibility of overdraw.

The main drawback(s) is that you have two or more passes, and end up pushing a lot of pixels after all... and since you're working with RenderSurfaces most of the time, it's hard to get anti-aliasing. Among other things. It's pretty complicated to do right.

[GD]

petrus, i'm already using gpu pixel occlusion, but as Zaknafien said it requires one more render pass that renders low poly occluders.

[JohnB]

Glad to hear you might get back to this over the winter GD!
We're looking forward to seeing your progress.

John B.


PS - Raine will probably be too busy to notice you're working on it ;-)