I'm curious about how did people come up with this guess. Did they assume that each primitive might be uniformly distributed in the scene and might has similar size?
outousan
Is N the number of voxels or primitives?
buzz
Well if the number of voxels should be approximately equal to the number of primitives, N could be the number of either.
Parker
Always surprising how simple some of the most effective heuristics end up. I'm with Snack, coming up with these things seems deceptively difficult.
sitongc
Why the intersection cost is this formula?
supernova
Still not quite understand the intersection cost function
yee
Assume there will be m primitive in one voxel, the volume of a cube equals to mN. So the length of each side of the cube equals to cube root mN. Remember that when we do line rasterization, the computational cost is o(side length). So here the intersection cost is o(cube root N).
I'm curious about how did people come up with this guess. Did they assume that each primitive might be uniformly distributed in the scene and might has similar size?
Is N the number of voxels or primitives?
Well if the number of voxels should be approximately equal to the number of primitives, N could be the number of either.
Always surprising how simple some of the most effective heuristics end up. I'm with Snack, coming up with these things seems deceptively difficult.
Why the intersection cost is this formula?
Still not quite understand the intersection cost function
Assume there will be m primitive in one voxel, the volume of a cube equals to mN. So the length of each side of the cube equals to cube root mN. Remember that when we do line rasterization, the computational cost is o(side length). So here the intersection cost is o(cube root N).