[MOAB-dev] moab, tpetra::map

Nico Schlömer nico.schloemer at gmail.com
Mon Nov 2 07:42:42 CST 2015


I trust the last line is supposed to be
```
rval = mb->tag_get_data(gid, verts, &globalids[0]);
```
Anyhow, good to know! That would have cost me time for sure.

Cheers,
Nico

On Mon, Nov 2, 2015 at 2:37 PM Grindeanu, Iulian R. <iulian at mcs.anl.gov>
wrote:

>
> ------------------------------
> *From:* moab-dev-bounces at mcs.anl.gov [moab-dev-bounces at mcs.anl.gov] on
> behalf of Nico Schlömer [nico.schloemer at gmail.com]
> *Sent:* Monday, November 02, 2015 4:30 AM
> *To:* moab-dev at mcs.anl.gov
> *Subject:* [MOAB-dev] moab, tpetra::map
>
> I would liket to create a Tpetra::Map [1] from a MOAB parallel partition.
> To this end, I need to following:
>
>  * The notion of global ID of a node. I trust EntityHandle is what I'm
> looking for? EntityID appears to be the index a lot of times, aka, a local
> ID.
>
> No, the EntityHandle refers only to a local partition
> In general, we keep a Global ID of a node in a tag called exactly like
> that
> "GLOBAL_ID"
>
> When you just load a file, in different partitions, EntityHandle for
> vertices starts at 1 on each partition
> So they "overlap", it can't be used as global id;
>
> to get the global id, you need to do something like this, on each partition
> Range verts;
> rval = mb->get_entities_by_dimensions(0,0, verts);
> std::vector<int> globalids;
> globalids.resize(verts.size());
> Tag gid;
> rval = mb->tag_get_handle("GLOBAL_ID", gid);
>
> rval = mb->tag_get_data(gid, verts, &gid[0]);
>
>
>  * Translate a Range into a std::vector. Seems easy given Range's
> operator[].
>  * Get owned nodes. Get owned plus shared nodes.
>     I'm still not quite familiar on how to best pick out entity sets from
> a pcomm. I see there's `get_pstatus_entities()`, but I don't see a
> PSTATUS_OWNED.
>
> Cheers,
> Nico
>
> [1]
> https://trilinos.org/docs/dev/packages/tpetra/doc/html/classTpetra_1_1Map.html
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20151102/3a30921d/attachment.html>


More information about the moab-dev mailing list