CGAL users discussion list

[Marc Glisse <>]

On Thu, 8 Oct 2015, Scherrer Daniel wrote:

> If I understand right, this performance issue has also something to do 
> with the fact that our data points are lying on regular grid positions 
> and often there are many possibilities to triangulate and CGAL is 
> searching for the exact best using the exact number type, which in our 
> case would not be necessary as just one of best the possibilities could 
> be chosen.

It isn't looking for "the best", it needs to know for sure that they are 
equivalent before picking an arbitrary one. Otherwise the algorithm may 
crash, loop forever, etc.

> Can switching to exact number type be disabled somehow?

It isn't really meant to be, and bad things could happen. But something 
like this (undocumented and likely to break at some point):

struct K : 
CGAL::Kernel_d_interface<CGAL::Cartesian_wrap<CGAL::Cartesian_base_d<double, 
CGAL::Dimension_tag<6>>, K>> {};

might work (you may need to add a couple constructors as in Epick_d.h). It 
means all computations are done with double, no filtering, no exact type.


--
Marc Glisse