CGAL users discussion list

[Peter Hachenberger <>]

Hi Fred,

> I managed to have something working with Cartesian<Gmpq> under Linux but it 
> is
> dead slow ( I am doing solid intersections inside a 3D rendering loop ). 
> Would
> it be faster with Homogenous<Gmpz> ? Cartesian<Gmpz> doesn't work because
> to_double() returns only the integral part of numbers.

Yes, Homogeneous<Gmpz> is much faster. Also try Homogeneous<mpz_class>.
But be careful; with a homogeneous kernel your input files must have
integer coordinates. These problems will hopefully be gone in the next
CGAL Version. Then floating point filtering will be effective with
Nef_polyhedron_3. This means that the fastest Kernel will be a Cartesian
kernel and it will be much faster. 

Do you do a single intersection per loop cycle?

Peter