Faster voronoi

i m looking for ways to produce faster voronoi … any suggestions how ??? FAcet dome is fast but doesnt behave well on random point cloud. here is what i m trying to do … why i want a faster voronoi

If you Google the topic, you will find a lot of information on this, including algorithms and code. You may need to implement this yourself.

Facet dome uses exactly the same algorithm as Voronoi3D. But of course the purpose of the component is different so you won’t be able to substitute one for the other.

The 3D voronoi stuff in GH doesn’t use Nurbs surface or breps internally, it has a custom made intersector and solid boolean slicer which is very fast because it assumes that each shape it is working with is:

  • Closed
  • Convex
  • Has only planar faces
  • Each face has only polyline edges
  • The edges of each face are convex

In order to make Voronoi3D faster you have two options, both of which probably require you to write your own code:

  1. Make it multi-threaded
  2. Reduce the number of intersections needed per cell.

The algorithm in Grasshopper for 3D voronoi cells does not use a 3D delaunay dual to figure out which intersections need to be performed, this is where some speed gains can be made.

2 Likes

Thank u David.that helped.

hello
I hope not to be late.
Attached is a component
In which you will find a simple script for Voronoï with parallel function
I hope this will help you
.Parallel Voronoï_1.gh (13.3 KB)

sincerely Hugo

Hugo , i can’t download ur file . can u plz upload it again ??
Thank u

Parallel Voronoï_2.gh (13.8 KB)
sorry I did not check the script.
here’s a script that works properly

kind regards

I don’t understand … not much of difference here or at times py code is slower than native component .

yes i have see that but i send u another definition using the same component parallel but with another GH fonction and u going to see the diferenceutilisation de tous les coeurs.gh (13.3 KB)

1 Like

Further Recomputing gives good results but this didn’t work with Voronoi.