Minutes 21 November 2005

Present: Andre, Carey, Greg, Karl, Martyn, Peter, Shane, Travis

** Refinement of initial development program **

Shane has begun a development plan, initially contains things considered in last weeks meeting.

Discussion on whether we are doing FEM and/or FV first. We do want both long term. Consensus is to start with FEM, but not to comprimise our ability to FV later too. Unstructured multigrid is more suited to FV for example, and hosted mesh based generation requirements are different to free form mesh generation.

The outcome should include a statement on scalability. Discussion on what is the development target number of processors, 64 proc like we have or say 1000 like we can get access to. The code for these two scales is possibly different at a low level. Getting a common API is probably possible and is the real task. There is also a choice between doing shared and distributed memory. The feeling seems to be to write for a low number of procs and get scaling to 64 procs using distributed memory, using MPI. Then somebody like IBM could help us take that to 10000. This needs further consideration.

** Development target. What enviroment (OS's), interaction method and users are we targeting? **

AIX and Linux. Linux for local development and AIX for running jobs.

** Scripted or a single program **

Scripted. Defered granularity discussion for later. Shortlisted Python and PERL as the languages of choice.

** Everybody please add to the References by next week **