CGAL users discussion list

[Gueroult Renaud <>]

Thanks a lot for your help.
Apparently I had a mistunderstanding of the settings. Now it works fine 
and fast.
Thanks again.

Renaud

Efi Fogel a écrit :
> I've tried it myself on a Windows machine (AMD Athlon 64x2 Core 
> Processor 3800+ 2.01GHz, 1GB RAM). I did change the storage type of 
> time_task1 to double. When compiled without optimization I got the 
> following:
>
>    Elpased: 3
>    Elpased: 3
>    Elpased: 2
>    Elpased: 3
>    Elpased: 2
>
> after I adding "-DNDEBUG /Ox /O2" to the compilation command line, I got:
>
>    Elpased: 0
>    Elpased: 1
>    Elpased: 0
>    Elpased: 1
>    Elpased: 0
>
> The entire command lines were:
>
>    "cl.exe" -TP -GR -EHsc -Zm900 -nologo -MD -DCGAL_NO_AUTOLINK
>    '-Ic:/CGAL/CGAL-3.3-I/include' '-Ic:/boost/include/boost-1_33_1'    
> -DNDEBUG -Ox -O2 -c bso.cpp
>
>    "cl.exe"  -Febso bso.obj  -link '-LIBPATH:c:/CGAL/CGAL-3.3-I/lib'
>    -nologo CGAL.lib CGALPDB.lib mpfr.lib gmp.lib psapi.lib
>
> You will still have to sort out the GMP issue, but at least these 
> results should give you confidence that there are no fundamental 
> problem on Windows.
>
> Gueroult Renaud wrote:
> > Andreas Fabri a écrit :
> > > Gueroult Renaud wrote:
> > > > Yes the VC2005 project options are release, O2 and I had the NDEBUG 
> > > > flag on for the time given previously.
> > > > If I change this flag to DNDEBUG it takes about 18 seconds to make 
> > > > 1000 intersections in release under windows with O2.
> > > >
> > > > Thanks for your help. I'm at your disposal if you need anything else.
> > >
> > >
> > > Did you define CGAL_USE_GMP ?  Otherwise
> > > the exact predicates exact constructions kernel
> > > might use another multi precision integer
> > > number type under the hood.
> > >
> > >
> > > Do you use the same GMP for VC8 and cygwin?
> > > The precompiled version that comes with the
> > > windows installer doesn't use processor specific
> > > optimisations, that is no assembler routines.
> > >
> > > andreas
> > I was not using the CGAL_USE_GMP preprocessor definition, I tried it 
> > but unfortuantely it doesn't improve the computation time.
> >
> > On the other hand, I can't really tell if I'm using the same GMP for 
> > VC and cygwin. The one I use with VC8 is the precompiled one which 
> > come directly with the windows installer (I didn't do anything on 
> > that side). With cygwin, I installed CGAL as explained in the manual, 
> > so I don't really now how does it work, all I know is that I was not 
> > using the flag neither.
> > Do you think there's something to do on windows with GMP?
> > > > Renaud
> > > >
> > > > Andreas Fabri a écrit :
> > > > > Did you -O2  -DNDEBUG
> > > > >
> > > > > andreas
> > > > >
> > > > >
> > > > > Gueroult Renaud wrote:
> > > > > > Hi CGAL users and developpers,
> > > > > >
> > > > > > I've got a speed issue running some polygons operations. Using the
> > > > > > CGAL-3.3 in release with CGAL_NO_ASSERTIONS and 
> > > > > > CGAL_NO_PRECONDITIONS, I
> > > > > > get an approximate computation time of 15 seconds for 1000 
> > > > > > intersections
> > > > > > of really simple polygons on  a windows OS under VS2005. I tried it
> > > > > > using cygwin (liux environment for windows) with the same 
> > > > > > computer and
> > > > > > the same operation takes approximately 4 sec. Unfortunately I'd 
> > > > > > like to
> > > > > > keep developping under windows and VS2005 because I'm using others
> > > > > > library I don't want to implement with cygwin.
> > > > > > If someone would like to see the options of the VS2005 project, 
> > > > > > tell me
> > > > > > and I would be please to send  it to him.
> > > > > > I tried with various kernels, it certainly has an impact on the
> > > > > > computation time, but that is definitely not the explanation for 
> > > > > > this
> > > > > > huge difference.
> > > > > >
> > > > > > Best regards,
> > > > > >
> > > > > > Renaud
> > > > > >
> > > > > >
> > > > > > Below is the code sample:
> > > > > >
> > > > > > /#include "stdafx.h"
> > > > > > #include <CGAL/Exact_predicates_exact_constructions_kernel.h>
> > > > > > #include <CGAL/Cartesian.h>
> > > > > > #include <CGAL/Polygon_2.h>
> > > > > > #include <CGAL/Polygon_with_holes_2.h>
> > > > > > #include <CGAL/Boolean_set_operations_2.h>
> > > > > > #include <iostream>
> > > > > > #include "time.h"
> > > > > > #include "bso_rational_nt.h"
> > > > > >
> > > > > > typedef CGAL::Cartesian<double> K;
> > > > > > typedef CGAL::Exact_predicates_exact_constructions_kernel  Kernel;
> > > > > > typedef Kernel::Point_2 Point_2;
> > > > > > typedef CGAL::Polygon_2<Kernel> Polygon_2;
> > > > > > typedef CGAL::Polygon_with_holes_2<Kernel> Polygon_with_holes_2;
> > > > > > typedef std::list<Polygon_with_holes_2>        Pwh_list_2;
> > > > > > using std::cout; using std::endl;
> > > > > >
> > > > > >
> > > > > > int main()
> > > > > > {
> > > > > >    Polygon_2 P3,P4;
> > > > > >    P3.push_back(Point_2(0,0));
> > > > > >    P3.push_back(Point_2(1,0));
> > > > > >    P3.push_back(Point_2(1,1));
> > > > > >    P3.push_back(Point_2(0,1));
> > > > > >    P4.push_back(Point_2(0.5,0.5));
> > > > > >    P4.push_back(Point_2(1.5,0.5));
> > > > > >    P4.push_back(Point_2(1.5,1.5));
> > > > > >    P4.push_back(Point_2(0.5,1.5));
> > > > > >
> > > > > >    Polygon_with_holes_2 UnionR;
> > > > > >    Pwh_list_2 intR;
> > > > > >
> > > > > >    int i,j;
> > > > > >    time_t Start_t, End_t;
> > > > > >    int time_task1;
> > > > > >    FILE *out = fopen("D:\\code\\TimeTestFromScratch\\Test.txt","w");
> > > > > >    for (i=0;i<5;i++)
> > > > > >    {
> > > > > >        Start_t = time(NULL);
> > > > > >        for (j=0;j<1000;j++)
> > > > > >        {    //UnionR.clear();
> > > > > >            //CGAL::join(P3,P4,UnionR);
> > > > > >            intR.clear();
> > > > > >            CGAL::intersection(P3,P4,std::back_inserter(intR));
> > > > > >        }
> > > > > >        End_t = time(NULL);
> > > > > >        time_task1 = difftime(End_t, Start_t);
> > > > > >        fprintf(out,"Elpased : %d\n",time_task1);
> > > > > >    }
> > > > > >
> > > > > >    return 1;
> > > > > > }/


--
------------------------------------------------------
/Renaud Gueroult/

 
<mailto:>

/Peter Heppel Associates/
Design-Engineering-Mathematics
www.peterheppel.com <http://www.peterheppel.com>