I have read this tutorial: http://www.opengl-tutorial.org/intermediate-tutorials/tutorial-16-shadow-mapping/
I wonder, is this the current best technique for generating shadows in an arbitrary 3D scene? Is it the way current modern games implements directional light shadows or are they using some other better way of calculating shadows on the fly?
What I mean is if you know your geometry so good that you know exactly every impact a shadow can have on an objects colors, it must be better to use that information instead of rendering a Shadow map?
For example, if we have a 3D grid world with all perfect cubes at perfect offset (think minecraft) and the directional light if exactly 45 degree in all dimensions, that leave us with a situation where a cube closer to the light will also cast shadow on the entire cubes at position offset -1, -1, -1 from that cube, and so on. The shadows always go diagonal and put other equally sized cubes sides in 1 of three states: all shadow, all light or 50% in shadow.
If we know our geometry so good, then using that information must be much faster than rendering shadow maps right? Even when most of such calculation is done on the CPU instead of the GPU.