librelist archives

« back to archive

p4est struct API

p4est struct API

From:
Carsten Burstedde
Date:
2014-10-26 @ 15:41
Hi all,

I'm collecting ideas about a hopefully final p4est version with respect
to the member variables of the major struct data types.  I'll start:

 1. Finalize what belongs into mesh_t (optional by mesh_new_ext)
 1. Remove tree->quadrants_per_level
 1. Add p4est->local_maxlevel
 1. Deprecate p4est_nodes but keep it around for reference
 1. Without nodes we could do away with P4EST_QMAXLEVEL and the nodes
    functionality in p4est_bits.  This is a rather involved issue
    though.

Comments?  More?  :)

Carsten

Re: [p4est] p4est struct API

From:
Alex Fikl
Date:
2014-10-27 @ 07:08
On Sun, Oct 26, 2014 at 5:41 PM, Carsten Burstedde
<burstedde@ins.uni-bonn.de> wrote:
> Hi all,

Hello,
> I'm collecting ideas about a hopefully final p4est version with respect
> to the member variables of the major struct data types.  I'll start:
>
>  1. Finalize what belongs into mesh_t (optional by mesh_new_ext)
>  1. Remove tree->quadrants_per_level
>  1. Add p4est->local_maxlevel

What about also adding p4est->local_minlevel? For completeness, if nothing else.

>  1. Deprecate p4est_nodes but keep it around for reference
>  1. Without nodes we could do away with P4EST_QMAXLEVEL and the nodes
>     functionality in p4est_bits.  This is a rather involved issue
>     though.

p4est_nodes_t is used in p4est_vtk_* to construct the connectivity
when scale is 1.0. If you deprecate it, is there another way to get a
unique numbering of nodes? Maybe with p4est_lnodes_t?

Regards,
Alex

> Comments?  More?  :)
>
> Carsten

Re: [p4est] p4est struct API

From:
Carsten Burstedde
Date:
2014-10-28 @ 09:54
> > I'm collecting ideas about a hopefully final p4est version with respect
> > to the member variables of the major struct data types.  I'll start:
> >
> >  1. Finalize what belongs into mesh_t (optional by mesh_new_ext)
> >  1. Remove tree->quadrants_per_level
> >  1. Add p4est->local_maxlevel
> 
> What about also adding p4est->local_minlevel? For completeness, if nothing else.
And tree->minlevel then.  The hardest part would be to implement this in
balance.  Toby, how complicated would this be?

> >  1. Deprecate p4est_nodes but keep it around for reference
> >  1. Without nodes we could do away with P4EST_QMAXLEVEL and the nodes
> >     functionality in p4est_bits.  This is a rather involved issue
> >     though.
> 
> p4est_nodes_t is used in p4est_vtk_* to construct the connectivity
> when scale is 1.0. If you deprecate it, is there another way to get a
> unique numbering of nodes? Maybe with p4est_lnodes_t?

I'm thinking of doing this via mesh_new_ext, disabling all other mesh
fields except the (yet to be written) local vertex numbers/coordinates.
Would anybody want to volunteer for this one?

Carsten