addressing for texture mapping with non-power-of-two textures

OpenGL
1

Hi all,

I have a question concerning the way non-power-of-two textures are implemented in OpenGL. More precisely, I would like to know how the address of a texel is obtained in a fast and efficient way (or so I believe).

This is an easy thing to do for power-of-two textures, using bitshift and mask, as described in this article: http://theglide.wordpress.com/2007/01/08/oldskool-tiled-texture-mapping/

However, what about non-power-of-two textures? The ideas described in the article do not apply anymore. Does anybody know more tricks to quickly compute the address of a texel in such a case?

Thanks in advance

Alex

2

Does anybody know more tricks to quickly compute the address of a texel in such a case?
I guess the trick is specialized hardware. With enough transistors, you can do almost everything in a single clock tick :wink: