Minutes 22 January 2007

Present: Travis, Shane, Karl, Jagir, Peter.

Examples on hpc/AIX.

• Shane will talk to Tiong about loadleveller 'pools'.

BoomerAMG from Hypre through PETSc

• Travis thinking about the best way to initialize PETSc solver in LibMesh.
• Travis comparing performance optimized and debug PETSc versions.
  • Not looking like a big difference.

Penalty boundary conditions.

• When Karl changed his 2D diffusion example to 3D, some solution values became negative even though no initial values were negative. This was resolved by reducing the penalty parameter from 10^20 to 10^15.

LibMesh CellML integration

• Jagir has written a proposal for the design and programming model. Jagir introduced the concepts of application-developer-designed:
  • "factory" for handling the set of cell models.
  • "marshaler" for handling communication with the cell models.

Next meeting 2007-01-23 (tomorrow) 11am to look at cellml/solver interaction.