CGAL users discussion list

[Andreas Fabri <>]

Thomas Zangl - Home wrote:
>  
> Am Thu, 03 Jul 2008 15:21:43 +0200, schrieb "Andreas Fabri" <>:
>
> Hi!
>
> Another question which arise during testing:
>
> in ftp://ftp-sop.inria.fr/geometrica/pion/publis/lazy-kernel.pdf you
> make a comparison between different kernels. Now, I ask myself:
>
> if I use:
> typedef CGAL::Simple_cartesian<CGAL::Gmpq> K;
> typedef CGAL::Gmpq dbl;
>
> all predicates and operations are done exactly? Like if the
> filter of the Filtered_Kernel always returns "NO_IDEA"? 
>
> The reason is, if I use the plain Exact_pred_exact_constr. kernel, my
> test app does not produce any degenerate cases. If I use the kernel
> typedef'd above, I get a lot of them. It seems, computation accuracy is
> far worse with the Simple_cartesian<CGAL::Gmpq> in my case.

Can you give a minimalistic code example.

andreas

> Any idea why this can happen? How can I clarify that all computations
> are done using exact arithmetic like described in your paper? 
>
> TIA,