CGAL users discussion list

[Yann GAVET <>]

Dear all, dear Sebastien,

I am requesting your help on this tentative parallell version of the 
lloyd_optimize_mesh_2.

I made a clone on github, started my own branch, pushed a few 
modifications and I am ready to start the pull request. I am not really 
comfortable with this process, thus please accept again my apologies in 
advance.

2 files are impacted: Lloyd_move_2.h and Mesh_global_optimizer_2.h

An example is provided for testing.


The pull request reference is parallel lloyd_optimize_mesh_2 #5674

Thank you very much for your time

best regards

Yann


Le 17/04/2021 à 10:50, Yann GAVET ( via cgal-discuss 
Mailing List) a écrit :
> Thank you Sebastien,
>
>
> I will do this pull request, probably next month.
>
> I am not used to this process, accept my apologies for these naive 
> questions: Do you request a fork before submitting a PR ? Is there a 
> page describing this process with cgal ?
>
> best regards
>
> yann
>
>
> Le 17/04/2021 à 07:12, Sebastien Loriot ( via 
> cgal-discuss Mailing List) a écrit :
> > Except if you changed the default traits to use some caching for 
> > circumcenters, this particular location should not be problematic
> > as the container is local.
> >
> > There are some flags set in faces so it might be related to that.
> > However without seeing the modifications you did, it is hard to
> > tell you what can go wrong. If you want you can open a pull-request
> > on github and I'm sure developer will be happy to help making work.
> >
> > Best regards,
> >
> > Sebastien.
> >
> > On 4/16/21 5:46 PM, Yann GAVET ( via cgal-discuss 
> > Mailing List) wrote:
> > > Dear all,
> > >
> > >
> > > I am trying to modify the function lloyd_optimize_mesh_2 in order to 
> > > have a parallel computation of all movements. As a reminder, each 
> > > vertex of the triangulation should move to the geometrical center of 
> > > its cell (defined by the voronoi diagram).
> > >
> > > When I simulate the threaded behavior (using a serie of different 
> > > function calls), I have to problem and it seems to work. However, if 
> > > I use actual threads, I get an error that occurs at 
> > > Constrained_voronoi_diagram_2.h:360 (this line is 
> > > polygon.push_back(m_cdt.circumcenter(face));)
> > >
> > > This error is malloc(): unaligned tcache chunk detected
> > >
> > >
> > > Thus, my question is for cgal experts: do you think this method 
> > > could be parallel ? is there any static variable that could not be 
> > > used by multiple threads ?
> > >
> > > Thank you very much for your time
> > >
> > > Yann