Everyone is correct. You (currently) can't have multiple element types with the same subdomain id. You also can't write two different meshes to the same file. If the mesh changes you must start writing a new file.
The multiple element types with the same subdomain id restriction is a hard restriction from Exodus. exodus actually has this assumption built in... It's core to the way Exodus lays out storage. That doesn't mean that we couldn't detect this case and make up some subdomain ids to write out... But that would give you significant other issues when trying to do things like restart. In practice... How often are element types mixed in the same place in the mesh? I don't know of any mesh generators that do this. Sure... You might have a transition from quads to tris when trying to capture geometry... But those transitions are well contained... And there is nothing stopping you from treating both subdomains exactly the same during the simulation (we could easily do the same in ex10). I do agree that this is a slight bummer... But in practice it doesn't hurt. Derek Sent from my iPhone On Mar 15, 2010, at 10:15 AM, Cody Permann <[email protected]> wrote: > I'm pretty sure you're right John regarding the subdomain ID issue, > I looked at that awhile back and thought it would be a problem but > it turns out that we never run meshes of mixed element types so we > haven't hit that issue. The other real issue here though is that > the exodus format does not support a changing mesh in the same file > (to the best of my knowledge). One must write a new mesh file each > time the mesh changes and deal with that on the visualization side. > > Cody > > On Mar 15, 2010, at 10:03 AM, John Peterson wrote: > >> On Sun, Mar 14, 2010 at 3:57 PM, Nestor M. Solalinde >> <[email protected]> wrote: >>> Dear libmesh developers: >>> >>> I'm having trouble visualizing example 10 results. I can't use GMV >>> because >>> it doesn't have an open licence anymore. Up to now, I've been >>> saving the >>> results of my (transient) simulation using an ExodusII_IO object >>> to store >>> each timestep, but it doesn't seem to work with example 10 (mesh >>> refinements). I tried also saving just one timestep with >>> >>> ExodusII_IO (mesh).write_equation_systems >>> ("filename.ex2",equation_systems); >>> >>> getting the following error, >>> >>> Num elem block: 1 >>> Assertion `i < this->n_nodes()' failed. >>> [0] /opt/libmesh/libmesh_3697/include/geom/elem.h, line 1183, >>> compiled Mar >>> 14 2010 at 18:02:13 >>> terminate called after throwing an instance of 'libMesh::LogicError' >>> what(): Error in libMesh internal logic >>> make: *** [run] Aborted >>> >>> Any suggestion? thanks a lot, >> >> Derek can correct me if I'm wrong, but it looks like the Exodus >> writer >> is expecting all elements with a given subdomain ID to be of the same >> geometric type -- which isn't the case for ex10, the mesh consists of >> quads and triangles all with the same subdomain ID. >> >> In order to change over to Exodus for our default writer, we'll need >> to fix up ExodusII_IO_Helper::write_elements() to not make this >> assumption -- the subdomain IDs are completely user-defined, we can't >> assume they divide the elements into subsets of different geometric >> types. >> >> -- >> John >> >> --- >> --- >> --- >> --------------------------------------------------------------------- >> Download Intel® Parallel Studio Eval >> Try the new software tools for yourself. Speed compiling, find bugs >> proactively, and fine-tune applications for parallel performance. >> See why Intel Parallel Studio got high marks during beta. >> http://p.sf.net/sfu/intel-sw-dev >> _______________________________________________ >> Libmesh-users mailing list >> [email protected] >> https://lists.sourceforge.net/lists/listinfo/libmesh-users > > > --- > --- > --- > --------------------------------------------------------------------- > Download Intel® Parallel Studio Eval > Try the new software tools for yourself. Speed compiling, find bugs > proactively, and fine-tune applications for parallel performance. > See why Intel Parallel Studio got high marks during beta. > http://p.sf.net/sfu/intel-sw-dev > _______________________________________________ > Libmesh-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/libmesh-users ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
