CGAL users discussion list

[erickee <>]

Update:  I have not yet resolved this issue.  Furthermore, I have found that
multiplying my homogenous points by a large number causes problems.

If I multiply by 1000, I frequently have overflow issues, even though I am
using ring type long, which I thought would have a sufficient range to work
with my typically small homogeneous points (e.g. [1000 1000 1000]).   

On the other hand, if I multiply by only 10, I don't have overflow issues,
but I am throwing away a lot of precision.  For example, a homogenous
half-plane [1.0  1.04  1.0] has a 46º orientation, but round(10*[1.0 1.04
1.0]) has a 45º orientation. 

I could use a ring type like CGAL::Gmpz, but I haven't figured out how to
convert homogenous Gmpz back to a homogeneous double after my computation is
done.  (I need doubles for post-processing in matlab). 

I would greatly appreciate any advice on this topic.

--
View this message in context: 
http://cgal-discuss.949826.n4.nabble.com/Converting-doubles-to-Filtered-extended-homogeneous-tp4332165p4353379.html
Sent from the cgal-discuss mailing list archive at Nabble.com.