Well, it is nice to see Carmack is still passionate about what he does.Anyone that says that is basically an idiot, not appreciating the problems.
More a "Why would you want to"?Was that a "what are" or "why would you use" for voxels?
So why not model with Catmull-Rom splines?Well there's voxels and then there's voxels. The voxel engines used in old PC games were just using height maps.
I'm not familiar with the term "epipolar" but I'm taking a wild guess that what you are describing is (IIRC**) White's algorithm; the sort of thing one programmed on a circa 1981 computer to draw functions such as y=sin(sqrt(x^2+z^2)) with asterisks.If you don't have a 3D card at your disposal height maps (and displacement maps) can be much more efficiently rendered than their corresponding meshes through iterative raycasting along the epipolar lines, which was what those old PC games were doing (although generally they restricted the viewpoint such that the epipolar lines were vertical).
I'm not familiar with the term "epipolar" but I'm taking a wild guess that what you are describing is (IIRC**) White's algorithm;
It's an image based rendering term. I was indeed referring to the method from the paper Mate Kovacs linked, although the link to epipolar lines was made in a later paper "Efficient displacement mapping by image warping". A better explanation of the theory in my opinion, lousy job on the literature search thoughI'm not familiar with the term "epipolar"
Sigh... yet another paper on the pile to read.It's an image based rendering term. I was indeed referring to the method from the paper Mate Kovacs linked, although the link to epipolar lines was made in a later paper "Efficient displacement mapping by image warping". A better explanation of the theory in my opinion, lousy job on the literature search though
Surprised you haven't read them yet ... I've spammed these quite a few times over the years.
The link actually isn't clear whether it is 16 or 8. It could be 8 cores since that is what they used for the demonstration. Still doubling cores is getting cheaper than doubling frequency so...