Ask HN: Is 4-dimensional depth buffering feasible with modern APIs?

A recent post (https://news.ycombinator.com/item?id=31750981) shows an implementation of 2-Dimensional Voronoi diagrams, by rendering a number of cones, collapsing along the z-axis, and using the GPU's z-culling as a de facto optimiser (it even bakes in free inter-cell anti-aliasing).

(This also allows weighted voronoi to be implemented simply by changing the gradient of the cones.)

I'm treating this as an analogue to 2-dimensional light-cones with a depth axis treated as the dimension of time. The z-axis collapse leaves everything within the domain of the first cone to intersect any given [x, y] 2-Dimensional location.

If we get depth culling in 4-dimensions, we can take a space [x, y, z, Ω] and treat our Ω dimension as the dimension of time (as we did for z in 2-dimensions).

And that would give us 3D voronois (with a lot of skipped stuff about volumetric rendering).

We can alter the gradients of the 4D cones, then we have weighted 3D voronois.

I think that this could be used for an interesting voxel experiment.