"I'm performing slicing operations on a three-dimensional libfive::Tree. The fastest method I can think of is to call Contours::render layer by layer and use OpenMP for parallelization, but I still find it slow. Because I think you need to build the tree independently for each layer. This image is my code.

In many cases, the layers are highly similar with only slight differences. Is it possible to refer to the situation of layer n when building the tree for layer n+1? Or build trees for multiple layers, such as 16 or 32 layers, simultaneously? Would this be faster?"