Shattered Func_Groups

Posted: Sun Feb 07, 2010 8:15 am Post subject: Re: Shattered Func_Groups

If you think about it, splitting a brush face like that has got nothing to do with the VIS process. VIS doesn't care about tris (polygons) anyway as it creates its own leaf nodes, which do not follow at all how you cut your own brushes (at least I don't think it cares). However, since the rendering engine is based on triangles, you can see how it handles the floor and the various wall shapes when you look at the picture showing the "problem".

Posted: Sun Feb 07, 2010 3:44 pm Post subject: Re: Shattered Func_Groups

I couldn't think of how it would help VIS, either. But how do I get it to stop splitting the faces like this? I'm not sure I've ever had this problem before...maybe I just didn't notice it? It creates a whole lot of unnecessary tris that should be eliminated if possible, so hopefully there is a way to fix it.

Posted: Mon Feb 08, 2010 2:01 am Post subject: Re: Shattered Func_Groups

Heh your right. Sorry for confusing you all. :D

Posted: Mon Feb 08, 2010 7:33 am Post subject: Re: Shattered Func_Groups

The floor is always going to be split in a number of triangles if the lower portion of the wall touching it isn't just a single horizontal brush but multiple vertical ones. That's the t-junction fixing in practice as q3map2 will only allow a triangle side to touch one other triange side. So, many brushes lining the edge of a longer brush will split the longer brush in as many triagnles.

However, how to prevent that starry splitting scheme is another matter. I'm sure it's actually a result of a particular condition. Whether it really has any extra tris, though, is an entirely different matter. I suppose you could manually split the floor to two parts and see how that would affect the t-junction splitting. However, no matter what you do, as long as the walls have multiple vertical elements touching the floor, you are going to get multiple splits as well.

Posted: Mon Feb 08, 2010 12:23 pm Post subject: Re: Shattered Func_Groups

Posted: Mon Feb 08, 2010 6:39 pm Post subject: Re: Shattered Func_Groups

I suppose something like in my attached shot is the scene we mostly are used to seeing, but if you begin to count the triangles, there isn't that much difference, I think. I don't really feel like trying to hunt for the condition for getting the star shape, but I'm sure there's a specific one.

Posted: Mon Feb 08, 2010 10:27 pm Post subject: Re: Shattered Func_Groups

So there's no way to get around this? No special strategy to get the same/similar look without the extra tris?

Posted: Tue Feb 09, 2010 7:31 am Post subject: Re: Shattered Func_Groups

Like Pande posted:

Posted: Tue Feb 09, 2010 3:33 pm Post subject: Re: Shattered Func_Groups

Oh, okay. When you said it didn't have anything to do with t-junctions, I thought that negated Pande's q3map_notjunc advice, as well.

Posted: Tue Feb 09, 2010 4:53 pm Post subject: Re: Shattered Func_Groups

Who said? I was saying from the beginning it's connected to t-junctions, even before Pande's post. T-junctions have got nothing to do with VIS, though. VIS is about lessening the burden on the engine by making sure only the visible parts of the map are drawn, t-junction fixing is about avoiding sparklies and other adverse effects caused by adjoining edges not meeting perfectly. The Quake engine is based on a singe triange edge meeting another single triangle enge. Under those conditions it can ensure the textures meet perfectly: Nothing is visible from behind or no pixel is drawn on top of another.

Posted: Thu Feb 18, 2010 5:36 pm Post subject: Re: Shattered Func_Groups

Whoops, missed the entire second page. Bad me.

*post removed*