On Tue, 3 Aug 2010, rochan wrote:

> I have been trying to move to the direction of vtk + Paraview. Libmesh
> says that the vtk io is 'is untested, experimental, or likely to see future
> API changes'. Anyway there seem to be some issues.

Some major issues, it looks like.  Comments in vtk_io.C include
"single processor implementation" and "now this is known to write
nonsense on AMR meshes".  I'm still happy Wout started this class; 
serial unrefined grids are better than nothing at all and at least the
parallelism problem will be easy to fix when someone has time.  For
now I don't have time to fix VTKIO for general cases, but I'll at
least rephrase those comments in "libmesh_assert()" form where they'll
be a little more helpful at runtime in the future.

And it looks like we need something better than this to replace GMV as
our default visualization format.  Ben uses Tecplot, but that's out
since it's commercial.  Derek uses ExodusII, which Paraview also
reads... but IIRC that currently doesn't work with hybrid meshes.  I
guess it could be usable for now if you stick to one element type per
mesh.  And we have figured out a design to get around the "only one
element type per block" limitation in the ExodusII format, so
implementing hybrid mesh support could be done in the short term if we
do need it.
---
Roy

------------------------------------------------------------------------------
This SF.net email is sponsored by 

Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev 
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to