[petsc-users] How to get cell number from node number?

[barry]

Thanks!

On 11/30/18 9:28 PM, Matthew Knepley wrote:
> On Fri, Nov 30, 2018 at 8:19 AM barry via petsc-users 
> <petsc-users at mcs.anl.gov <mailto:petsc-users at mcs.anl.gov>> wrote:
>
>     Sorry for the imprecisely description.
>
>     I have an unstructured grid (DMPLEX) with triangle mesh.
>
>     cell number (2 dim)={0};    line number (1 dim)={1, 2, 3};    node
>     number (0 dim)={4, 5, 6}
>
> So, you have a 2D interpolated simplex mesh in DMPlex, which means it 
> has vertices, edges, and cells.
> If you want the vertices on a given cell, you would get the closure 
> and then filter out any points which are
> not vertices. For example,
>
> PetscInt *closure = NULL;
> PetscInt  clSize, cl;
>
> ierr = DMPlexGetDepthStratum(dm, 0, &vStart, &vEnd);CHKERRQ(ierr);
> ierr = DMPlexGetTransitiveClosure(dm, cell, PETSC_TRUE, &clSize, 
> &closure);CHKERRQ(ierr);
> for (cl = 0; cl < clSize*2; cl += 2) {
>   const PetscInt point = closure[cl];
>
>   if (point >= vStart && point < vEnd) {
>     <this is a vertex>
>   }
> }
> ierr = DMPlexRestoreTransitiveClosure(dm, cell, PETSC_TRUE, &clSize, 
> &closure);CHKERRQ(ierr);
>
> If you do not use edges, then its much easier. Just call 
> DMPlexGetCone(). If you do use edges, but also
> get the vertices all the time, you can make an index for this query, 
> which will speed it up greatly at the
> cost of some memory.
>
>   Thanks,
>
>      Matt
>
>     Thank you,
>
>     Barry
>
>     On 11/30/18 8:33 PM, Stefano Zampini wrote:
>>
>>
>>     Il giorno Ven 30 Nov 2018, 15:05 Tsung-Hsing Chen via petsc-users
>>     <petsc-users at mcs.anl.gov <mailto:petsc-users at mcs.anl.gov>> ha
>>     scritto:
>>
>>         Hi,
>>         Is there any function that can get cell number from the given
>>         node number exist already?
>>
>>
>>     Cell number of what? DMPLEX? And what is "node number"?
>>
>>
>>         Thank you,
>>         Barry
>>
>
>
> -- 
> What most experimenters take for granted before they begin their 
> experiments is infinitely more interesting than any results to which 
> their experiments lead.
> -- Norbert Wiener
>
> https://www.cse.buffalo.edu/~knepley/ 
> <http://www.cse.buffalo.edu/~knepley/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20181130/ed1aca99/attachment.html>