Re: [Adonthell-devel] Drawing order

[Enrico Kochon]

Hi,

why dont you just adjust the drawing order according to the viewpoint?
Beginning from the last to the first: this is the world-famous
"z-buffer" algo.
However, I don't know much about the adonthell-rendering pipeline. So
maybe this is not appropriate.

Cheers,
Enrico


Am Samstag, den 25.10.2008, 20:08 +0200 schrieb Kai Sterker:
> Wish I could report completion of that part, but in fact I didn't get
> much done since my last commit.
> 
> Besides, what I had done started looking like spaghetti code where it
> should have been a simple and elegant algorithm. So I just drew a
> little scene (see attached image) and tried figuring out a more
> algorithmic solution.
> 
> The image itself shows the scene seen from the side, with an
> (invisible) scary eye looking at it from the top right.
> 
> To figure out the drawing order, two rays are projected through a
> shapes min and max corner points. If those rays do not pass through
> any other shape, we can render it and remove it from the queue.
> 
> For that scene, that's all there is to it. For shapes that do not
> overlap, it is obviously not important in which order they are
> rendered (hence they share the same level of drawing order).
> 
> A more formal algorithm would look as follows:
> 
> while drawing_queue not empty:
>   for each shape in drawing_queue:
>      do ray test against every other shape
>      if ray test negative: // rays didn't pass through any other shape
>        draw shape
>        remove shape from drawing_queue
> 
> That means that we must make several passes over the queue, until it
> is empty. Not sure if we could do faster ... we ought to, but lets
> start simple :-).
> 
> Remains the question how the ray test works. Obviously, by projecting
> the object shadow on the y-axis, we can check if two objects overlap,
> but it doesn't tell us which object is in the path of the rays.
> Figuring out boils down to a ray - AABB intersection test in 2D. From
> what I've seen, one could turn the ray(s) into an AABB too and check
> both boxes for overlap. Again, there could be something more simple,
> but for now lets try that ...
> 
> But enough talk, on to the implementation. I'll let you know how it goes :-)
> 
> Kai
> _______________________________________________
> Adonthell-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/adonthell-devel