CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

Sebastian Kapfer wrote:
> Dear Sebastien!
>
> On Wed, Oct 20, 2010 at 11:27:26AM +0200, Sebastien Loriot (GeometryFactory) 
> wrote:
> > I can see you have a 64bits machine. Mine is 32.
> > I try on a 64bits machine and still you code ran without
> > assertion failure.
> >
> > Did you used cmake to build CGAL?
>
> Yes.
>
> > can you try using cmake to generate the makefile to compile
> > your example (you can use the script cgal_create_cmake_script to
> > generate a CMakeLists.txt)
>
> Using CMake has nothing to do with it.

I think it does.
The config flag of gcc should have been set when configuring CGAL.
Can you give me the content of CGAL_CXX_FLAGS in CMakeCache.txt
where you configure CGAL.


> In the meantime, I tried on a more modern machine. Ubuntu Lucid, GCC
> 4.4, AMD64, Boost 1.43, CGAL 3.7.  Same problem.
>
> But your mention of your 32 bit machine tripped me off.  Of course, on
> AMD64, GCC uses the SSE unit to implement floating point, which has
> 64-bit floats all the way through, and doesn't use 80-bit floats for
> registers.
>
> > sk@noether(~/spielwiese/fast_thicken/MinimalExample)>
> >  g++ -O3
> > -frounding-math -mfpmath=387 -lCGAL -odistance_browsing
> > distance_browsing.cpp 
> > sk@noether(~/spielwiese/fast_thicken/MinimalExample)>
> > ./distance_browsing 
> > sk@noether(~/spielwiese/fast_thicken/MinimalExample)> g++ -O3
> > -frounding-math -mfpmath=sse -lCGAL -odistance_browsing
> > distance_browsing.cpp 
> > sk@noether(~/spielwiese/fast_thicken/MinimalExample)>
> > ./distance_browsing 
> > terminate called after throwing an instance of
> > 'CGAL::Assertion_exception'
> >   what():  CGAL ERROR: assertion violation!
> >   Expr: new_rd >= copy_rd
> >   File:
> >   
> > /opt/cgal/3.7_boost1.40/include/CGAL/Orthogonal_incremental_neighbor_search.h
> >   Line: 267
> >   Aborted
>
> (see http://gcc.gnu.org/onlinedocs/gcc/i386-and-x86_002d64-Options.html)
>
> So it seems CGAL works when using 80-bit floats, but not with true
> 64-bit floats?