CGAL users discussion list

[torchi borchi <>]

i followed your suggestion and tried this:

typedef CGAL::Gmpq RT;
typedef CGAL::Simple_cartesian<RT> K;

but it still crashes. is Gmpq using reference counting?
i need exact constructions.


> -----Ursprüngliche Nachricht-----
> Von: "Sylvain Pion" 
> <>
> Gesendet: 24.06.09 13:46:34
> An: 
> 
> Betreff: Re: [cgal-discuss] CGAL+GMP and multithreading crash


> 
>  wrote:
> > hi,
> > 
> > i am using CGAL with exact arithmetic withGMP/MPFR.
> > 
> > I am running MSVC2005. To install CGAL i used CGAL-3.4-Setup.exe.
> > I compile with Multithreaded DLL (debug) runtime libs.
> > 
> > My app binds agains (for debug version) among other libs: 
> > 
> > gmp-vc80-mt-gd.lib 
> > mpfr-vc80-mt-gd.lib 
> > CGAL_Core-vc80-mt-gd.lib 
> > CGAL-vc80-mt-gd.lib 
> > 
> > To speed up computation i changed my algorithm to use more than one 
> > thread.
> > 
> > i found out, that copying a Polygon_2 as an input for the threads is 
> > causing 
> > my app to crash. it looks like:
> > Polygon_2 a;
> > 
> > // ... init a ...
> > 
> > Polygon_2 b_thread1;
> > Polygon_2 b_thread2;
> > Polygon_2 b_thread3;
> > Polygon_2 b_thread4;
> > b_thread1 = a;
> > b_thread2 = a;
> > b_thread3 = a;
> > b_thread4 = a;
> > 
> > // ... start 4 threads each using b_threadX
> > 
> > // CRASH
> > 
> > when i do a workaround like this to copy "a":
> > 
> > b_trheadX.push_back( Point_2( toDouble( a[0].x() ), toDouble( a[0].y() ) 
> > ) );
> > 
> > there is no crash.
> > 
> > It looks like assigining:
> > b_threadX = a;
> > does not realy do a seperate copy in memory and when using multiple 
> > threads
> > there happen evil things.
> > 
> > is there a clean way to do a copy of a Polygon_2  without that 
> > workaround? i
> > use exact numbers and toDouble() is rounding.
> > 
> > may it be a problem with CGAL-3.4-Setup.exe? maybe they are compiled not 
> > for
> > multithreading? any idea?
> 
> Which kernel are you using ?
> 
> Cartesian_kernel and Exact_predicates_exact_constructions_kernel both
> use a reference counting mechanism which is not thread-safe.
> In comparison, Simple_cartesian and 
> Exact_predicates_inexact_constructions_kernel
> should both be fine.
> 
> -- 
> Sylvain Pion
> INRIA Sophia-Antipolis
> Geometrica Project-Team
> CGAL, http://cgal.org/
> -- 
> You are currently subscribed to cgal-discuss.
> To unsubscribe or access the archives, go to
> https://lists-sop.inria.fr/wws/info/cgal-discuss
> 


______________________________________________________
GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://movieflat.web.de