Naive methods and parallelizations? about meshing.jl HOT 1 OPEN

By simple squares, do you mean a voxel-like approach? That should be possible by modifying one of the algorithms here. If you mean quads, we have naive surface nets. Though it may not be as easy to extend this one to a voxel approach. My guess is marching cubes is likely the easiest to modify by modifying the lookup tables cases and removing interpolation. Now if you mean an axis-aligned bounding box, that is something different entirely and I'm not sure how I would proceed on that.

A hacky approach might be to set the eps parameter really large in Marching Cubes or Marching Tetrahedra to get voxel-like appearance. This will increase the tolerance to snap the interpolated value to the voxel corner. Though it will go through the same code paths so you won't notice any performance difference.

I've done some experimentation with multithreading here: #52 and here #67
One thing of note is that the current algorithms do not scale well with multiple threads using pre-computed array inputs. The avant garde in multithreaded isosurface extraction is Flying Edges. I tried an attempt at this in 2D in JuliaGeometry/Contour.jl#51.

from meshing.jl.