I have uploaded a new version of the software renderer and also updated the example provided.

I have done some refactoring and created a base class for the rasterizer. Now the GeometryProcessor class (formerly Renderer) is not tied to a fixed Rasterizer implementation any more. Now one instantiates a rasterizer (currently RasterizerSubdivAffine is provided) and tells the GeometryProcessor to use this. The rasterizer can also be changed at runtime. This would for example make it possible to use a rasterizer that uses an s-buffer internaly (like Quake1) for level geometry and another rasterizer for models.

I have in mind to create a rasterizer which delegates work to the second CPU so that I can use both CPUs for rasterizing in parallel.