Change in the project technical specifications
As i was in the process of making a realtime raytracer based on traditional geometry (triangles) and kd-trees, i suddenly had some interrest in sparse voxel octrees. As a result i am restarting a new raytracer for this project. This makes sense as sparse octrees have the faculty of containing the geometry in the acceleration structure itself, and the traversal is also much cheaper as the one of a kd-tree and doesn't require any stack. Furthemore, it is possible to handle streaming and LOD in a very natural way with this structure. The only drawback may be the size of the dataset, which i dont expect to be very nice anyway for this project. :)