Minutes 24 July 2006

Present: Shane, Karl, Peter, Poul, Travis, Greg.

An audio recording is available.

Libmesh

• Want to contact developers.
  • Karl wants to talk about mesh parallelization
  • Want to talk about a test framework. Has it been thought about?
    • Shane doesn't see this as important as getting something working.
  • libmesh has already been used for reaction-diffusion. Could try to collaborate with these developers.
• Problems to work on
  • Shane suggests duplicating existing cm examples.
    • Start with one example.
• Test framework
  • Not much point developing unit tests for our code (using libmesh) if there are no unit tests in libmesh.
    • testing framework/examples?
    • visualisation
  • Easy enough to add complete problems using libmesh to our example tree.
  • Discussion over whether to have a separate tree for libmesh or to merge these examples in with similar problem types.
    • Peter prefers libmesh examples alongside cm examples.
    • Karl thought it was clearer to users to have separate trees.
• Discussion over what type of code would be contributed to libmesh and what code we would keep separate but would just use the libmesh library.
  • based on level: e.g. basisfuncs in libmesh, problems separate.
  • also issues of when code is suitable (stable enough) to submit.

Greg: Interactive or batch style?

• Shane can't see a good argument for an interactive interface.
  • C++ code is similar to com file code.
• Poul: visualization requires interactive interface.
  • Shane: export from C++ code at appropriate stage, then use separate visualization program.

Peter will chase up code associated with Ed Vigmond.

How do we get PhD students using libmesh?