I am writing a little game using OpenGL. I want to speed it up doing some culling. At the moment I am using OCTREE subdivision stuff (based on cone3doctree-stuff) doing frustum culling. But now I heard that matrix transformations will speed up my prog. I started to read about them and I can''t imagine that this would be a speed-up. Imagine:
2000 Triangles in a scene. I got my matrix and now I render it. For translating the scene I have to translate each vertex of my triangles with my matrix. - That would be great, cause I could then store the translated scene in an array and cull some more stuff (like terrain behind high hills or something like that). But that would actually slow down my prog, wouldn''t it?
Actually, if my memory serves me correctly, the OpenGL Red Book (OpenGL Programming Guide) specifically states that using the pre-made translation/rotation functions are, in general, a much better idea than a ''home-made'' way of doing it. Just something to think about.