I was playing with this today and I tought it is worth sharing… Im plannig to extend it with bilinear interpolation if I get some time tommorow so it could fill a four-sided polygon. I ment to do it as shading algorythm for a 3D engine, so if you will have any suggestions for optimization, they are very welcome 
It keeps on about 20 milliseconds per line on this 1GHz machine, thats a lot 