http://www.spider3d.com/img/onebigfig.jpg
I found this representation a long time ago. It looks like it would be extremly fast in theory.
I don’t understand what happens when you compare the light source z depth with the view z depth. Also, how would you align the light z depth with the object in the rendered view?
He, the old double Z-Buffer method! Mainly used in software engines, can be done in hardware too. But slow and generate pixellised shadows - even with this bluring.
actually this is the way, how the SGIX_shadow extension works. GF3 and GF4(*) supports this extension in hardware.
take a look at the “hw_shadowmaps_simple” demo from the NVIDIA OpenGL SDK how to use it…
(*) at least some of them. i got a little bit confused about the features of all this GF4-versions out there…