What technique are you referring to here?
Both actually. I guess my point was that bandwidth on either shadow technique is low if its just used for the characters. But, add in world geometry and it becomes measurable.
nAo said:
With shadow maps you could use some super simplified geometry for the stadium and draw it once per shadow map with minimal overdraw -> big win.
We kind of do that now with shadow volumes. Our stadium shadow volumes for example are tiny, < 300 verticies for an entire stadium which at 16 bit precision means we need < 2k to get realtime stadium shadows. With < 300 verts, extrusion is also cheap.
Fafalada said:
Actually there's one situation where I have been recently thinking stencil could still be a win, but it's pretty game specific, and it requires mixing of both approaches.
We actually do use both. For stadiums, shadow volumes seemed to make sense for a bunch of reasons. The memory footprint is tiny, performance is very fast. Plus, all the stadium structure is far from where the shadows ultimately get cast (on ground and players), so we can do a uniform blur on the shadow and it will look correct since the shadow casting structure is effectively at infinity.
In cases where shadow casting geometry is in contact with shadow receiving geometry, there would be issues since you'd expect a sharp shadow where they meet, and generally getting fuzzier as they got further apart. But that situation doesn't arise with stadium geometry so a uniform blur looks ok. Plus, the simplicity of shadow volumes made them very desirable for a sports game that must be released every year like clockwork or face massive fines!
Anyways, that was our line of thinking.
