History for Minutes 1 May 2006

changed:
-
Present: Shane, Peter (until 11:30), Andre, Karl, Poul, Travis.

Apologies: Greg.

"An audio recording is available":ftp://ftp.bioeng.auckland.ac.nz:/cmiss/opencmiss/cmiss-meeting-010506.ogg

* MPI job progress

  *  Has now been made ready to be advertised (as soon as Peter put a word in).

* Software Resources

  * Jon Pearce has used a few codes that could be useful.

  * Andre: Merryn's been using gmesh.

  * Peter: mpich-G2 is mpich extended by Globus Toolkit for grids.

  * Shane: If we choose an existing FEM framework it will come with a field
  framework and will likely not work well with our fieldml.  How well we can
  implement the structures we want is something we need to consider.

    * Poul: interface between fieldml and existing framework could be used.

    * Shane was hoping that fieldml defined data structures.

  * Karl: should we exclude GPL as cannot distribute complete binaries without
    putting our source under GPL?

    * No objections.

    * Mention for now but don't spend too much time investigating,
      unless it looks like everything we need.

  * Peter: vpac in Melbourne have parallelization support libraries the might be
    useful.

  * Travis: are we looking at features or intending to use the software itself?

    * Shane: hoping to use software itself.

  * Travis: Next PetSc release is meant to provide more support for unstructured
    meshes.

  * Peter: Tinsley-Oden in University of Texas at Austin may have mentioned a
    computed-variables-like approach.

  * Travis: Metis can separate things into subdomains but doesn't tell you the
    neighbours.

* Parallel machine arriving in 3 weeks (but will be sometime longer before its
  ready to be running):

  * 8 nodes of 2 processors.

  * 32 GB per node.

  * 2 TB SAN.

  * Infiniband.

  * 1.9 !GHz POWER5.

* Should we be in contact with Computer Science?

* Travis:  Can we have a goal for something to run when parallel system ready?

  * Karl: Do some evaluation of potential framework.

  * Poison equation as a start.

  * Simple mesh but represented in a general unstructured format.

  * Shane: should it look like a ventricle?

    * Karl/Travis: much easier to mesh generate a cube for example.

Meeting finished 11:44.