i thought about a way to simulate real-time shadows in a very cheap way, if stencil shadow volumes or shadow mapping are to expensive (or such accurate shadow simulations aren’t needed for the sort of game) or the system too slow. let’s say we extrude vertices the way it’s done for stencil shadow volumes. could we get the world’s vertices which are “in” this volume? and then set their color to black while the vertices outside the volue remain white? should should simulate a shadowed region. very vague - would of course only make sense if the world has a high polycount, a terrain e.g., but it should be very fast to compute.
would this work or is a crappy idea?

If the world has a high poly count, then the work to know if a polygon is lit or not will be very long and you will loose what you’ll win. I also tried some stuffs like limiting the size of the extruded silhouette but that really didn’t help.

However you can find out some experiments like merging shadow volume and shadow maps. I haven’t tested that yet however (since my shadow map isn’t still good enough )

Well, now, I can remember that there are some studies about limiting the polygons to be taken in account when doing the shadow volume. I was able to find that quiete easily on google, so you should find it as easy as me. Try ‘fast shadow volume’ or so. But this only helps for static geometry I guess, which is less and less important as time goes on.

Note that those technics have their counterparts ! but in the examples I’ve seen they can accelerate the rendering about twice or more.

Hope that could help a bit.

Originally posted by Vexator:
i thought about a way to simulate real-time shadows in a very cheap way, if stencil shadow volumes or shadow mapping are to expensive (or such accurate shadow simulations aren’t needed for the sort of game) or the system too slow. let’s say we extrude vertices the way it’s done for stencil shadow volumes. could we get the world’s vertices which are “in” this volume? and then set their color to black while the vertices outside the volue remain white? should should simulate a shadowed region. very vague - would of course only make sense if the world has a high polycount, a terrain e.g., but it should be very fast to compute.
would this work or is a crappy idea?

ild say that scores about 9 on the crappy scale
A/ it wont look good (even with a terrain unless u heaps of verts onscreen)
B/ meshes will need to be pretty evenly tesselated
C/ how do u find if a vert is in a region fast? + u wanna do this with millions of verts a second

The quality would be poor, but it might work for example with character verts in the world volume thanks to vertex density but not for world verts. The main problem though is really the cost. Calculating whether something is in or out of a shadow volume would require a complex beam tree insertion. Stencil works fast because it boils this comparrison down to a simple stencil raster operation. Your technique in eliminating that stage probably becomes more expensive and cumbersome for less quality.