This assigment implements hierarchical radiosity with adaptive refinement. This is similar to homework #2, but solves the system of equations using the Gauss-Seidel method rather than estimating the solution with the Monte Carlo method.

The mesh is developed quite nicely, but there seems to be some problem with the links... for instance, many child nodes are the same color. Actually, they are different numerically, but out in the third or fourth decimal place, so it is not terribly apparent. I would have also expected the indirect lighting to be stronger. The ceiling and box sides are very dark for some reason. I guess this is a case where the mesh looks better than the final picture!

The following picture was generated after 9 iterations. 16 samples were taken to estimate the point to polygon form factor and another 16 samples were taken to calculate the visiblility term (for each element). Total running time on an UltraSparc was 19.68 seconds, of which 0.33 seconds were spent in the solving of the matrix (gather/sweep).

So, without further ado...