Re: [Algorithms] Triangle strips still useful?
Brought to you by:
vexxed72
From: Tomas A. @ Z. <ta...@ze...> - 2006-10-12 16:44:35
|
Another way to think about the triangle strips lists, etc. It is just a way to represent a patch object. Where the patch is already been texelated for you. Then it will be worth adding a CULL cone-sphere. Cone-sphere I define it as a sphere that can become a cone such you will have a normal and a parametric value that will go from [1,-1] along the normal which will describe the uni-sphere extent. Such if you do the dot product between the camera and the normal you can reject base on the parametric value. The hardware could use this cone-sphere to skip a hold batch of triangles. But there is a good chance that triangles are going to just become a specialize primitive as pixel shader becomes better at giving good parallax and the geometry shader may be able to improve the silhouette. The hardware guys just need to add a new texture format that will allow for real time raytracing of it. Similar to the http://www.ati.com/developer/gdc/2006/GDC06-Tatarchuk-Parallax_Occlusion_Map ping.pdf. It will be a good first step towards a more robust raytracer. But I guess I am changing the subject here. Tomas |