Hi, i was making some tests of shadow map implementation in my engine, and i saw that the depth resolution was too low ( depth value not the texture resolution ), causing some glitchs on self shadowing, i then tried to do the shadow texture with float to float3 codification, and so after that codification the quality of the shadow have increased a lot, the glitchs disapeared. My question is, why no one use this? Every book that i've readed used to put the depth value in one component or in all components of the depth texture. Anybody here already done this too?

Haven't done shadow mapping for quite a while now, but if I'm not mistaken you usually use a texture with only one 24 bit channel or something similar.

Thanks people!
I will try it with a floating-point texture instead of the actual texture, and will be posting the results soon.