We've implemented Eric Chan's hybrid algorithm for accurate shadows, and we're now taking another step. Our shadow volumes algorithm involves extruding silhouette edges in the Geometry Shader.
The question I want to ask is this:
Suppose in the Geometry Shader, we were able to check whether a silhouette edge is already in shadow (using, perhaps, the shadow map). Is it O.K to not extrude this silhouette edge? i.e. simply neglect the shadow planes that will be created when this edge is extruded? We're only dealing with watertight meshes, if that is relevant.
We expect some performance gain if we can reject some edges inside the geometry shader, as that will reduce the fragment shader's load.
Thanks for your help!
Shadow volume hybrid algorithm
1 reply to this topic