For those of you with ACM access and who havent seen them yet, most of the Siggraph papers are now online here. One of the coolest papers IMO is the shadow silhouette maps paper (although curiously that one is only on the web, not on the ACM site).
Please excuse my ignorance on how all this technology works, but could the shadow silhouette map technique be used on the shadow buffer technique used by Splinter Cell? The shadow jaggies in the Xbox version does get on my nerve sometimes. It would be cool if they could use this method to update the shadows in the next Splinter Cell game.
Well it uses one 60 instruction and one 46 instruction fragment shader, and it has to calculate silhouettes ... probably not on the X-Box.
I find the idea interesting mostly in theory, because you can think up a lot of variants.
For instance you could combine silhouettes (which can also be calculated on the graphics card) with shadow maps to generate shadow volumes, using the shadow map as a non-conservative HSR mechanism to make the inherent overdraw of normal shadow volumes go away, instead of generating a silhouette map (this is an adaption of McCool's shadow volume reconstruction method). Probably hard to do efficiently on present hardware though.
