Yes that should definitely be interesting, please get back to us with your experiences!
For 2d Delaunay you use the determinant of a 4x4 or 3x3 matrix for the predicate. For convex hulls in 3d, you use literally the same determinant. I'm not saying that a general convex hull code in n+1 dims will be as fast as Delaunay in n dims, but if that convex hull code can be parametrized with a generator type that represents a point on the quadratic surface where the Delaunay points are embedded, I think that the differences will be small. On Sunday, 25 May 2014 14:51:54 UTC+2, Ariel Keselman wrote: > > when using hulls to calculate Delaunay then you calculate everything in > d+1 dimensions, which for 2D has a 4X slower predicate. Also predicates are > mostly precalculated (already implemented in the updated gist above), so > their repeated calculation in the "swap" method should be fast. > > Anyway, let me finish a robust 2D Delaunay implementation in Julia and do > some benchmarks against [CGAL](https://www.cgal.org/), [QHull]( > http://www.qhull.org/) and [Triangle]( > http://www.cs.cmu.edu/~quake/triangle.html) > > Should be interesting :) > > > >
