I have added some new features to my software renderer and also fixed a bug.

There was a bug which caused the clipping rectangle not to be obeyed when the clipping rectangle was different than the viewport. This is now fixed and no pixels will be written outside the specified clipping rectangle.

I've added a per triangle callback function which can be used to compute the per triangle mipmap level. The example also has been updated to show the function signature.

With the renderer it is now possible to do interlaced rendering. For example you can specify that you only want every second scanline to be rendered. I tried this with my demo on the GP2X and it gave a performance increase of 11-68% (avg. ~50%). So even though I only render half as many pixels it is not twice as fast as one might think. There is the triangle processing overhead (probably the biggest factor in the logo scene where only a 11% performance increase was gained) and then there is the cache. That the performance doesn't double might very well be cache related.