## GPU-picking * :-1: takes time to create (measure) * we now calc snapped points for every 1e6 pointset update, that is slower than kdtree * :-1: test how much time it takes to re-render the framebuffer * :-1: picking is pretty slow O(N), reading from GPU * :-1: obscures points within a range * :-1: no mechanism of providing selection, so it happens O(N) ## KD-tree * :-1: takes time to initialize * :-1: point updates force re-initializing the tree, which mb slow for 1e6 points (point diffs? quad-tree?) * :+1: O(logN) picking time * :+1: selects all points within a range * :+1: fast selection (O(logN) for poly, box
GPU-picking
KD-tree