Minutes 5 December 2005

Present: Greg, Travis, Andre, Shane, Carey.

Apologies: Peter, Martyn, Karl, Chris.

• Contributions. What existing projects and research may give ideas, indicate directions or directly contribute to this project?

Thanks to Karl, Travis and Greg for working on the dependencies list. We seem to be struggling with what we should be trying to decide. Some discussion on what is it that needs to be done first?

FEM Data structures is agreed to be task 1.

Some discussion of HDF, but agreed we are working at a higher level than that initially, similarly for PETSCI.

• What role are dynamic languages going to play?

One view is to do the whole thing in python then reimplement bits that this turns out to not be suitable for (possibly Karl's view but absent today).

Another is for the glue or only for the scripting.

Difficult to discuss without Karl today.

Looks like the first task, FEM data structures, will need to be implemented in a compiled language.

We now need to formally record the things we want our data structures to satisfy. We we data structures we are initially talking about the structures that appear to be presented by the API - there could be other things going on under the API for MPI etc.

Language for the data structure module. Much discussion of the pros and cons, choice between fortran 95/2003 and C++. Does this storage module need to be able to manipulate arrays if so fortran is good if not maybe less reason for fortran. C++ offers templating and gets the compiler to do a lot of work for you.

Seems if we can find somebody adept at doing this part Fortran 95 we would accept this done in Fortran. Although there is some sceptism that

• somebody could be found to do it
• it will actually perform as claimed in relation it its ability to be called from c++ & python.

We would like the interfacing by fortran95 with other languages to be proven to work as early as possible.

Resourcing: So now we need a job description for a numerical computational programmer, needs to mention both fortran and C++. Shane to follow this up with Peter now.