Representation of intervals: Will there be a better way to represent the coverage intervals of the children? center and radius? other?

I did a test calculating the membership of the interval without using the bit mask array, but it is noticeably slower

It is possible to take advantage of the coherence in the "rays"? only varies x when each line is drawn, I have no proof but the most expensive operation seems to be the advance in the octree

One way to see the algorithm is to think that it is raytracing but the intersection is calculated with the intervals of the children, with the operation AND. The isometric perspective produces the symmetry of the form between the root and the children, and the permanence of the visiting order. It occurs to me how to try to produce this mechanism when the children have different sizes, that is, from a real perspective, but I cannot think of how to change the order of visit. The next experiment will be this test, but the order of visit will be wrong in some cases.

In a paper by Donald Meagher he talks about using an octree that combines several other to build a scene, it doesn't occur to me how this can be done, but, if it were possible, the zbuffer could be unnecessary since this algorithm always returns the closest voxel first.

This image is a previous version with ilumination

See the glitches in the voxel draw, the ilimunation is dynamic, build with the zbuffer