Personal tools
You are here: Home → openCMISS → Wiki → Minutes 20 November 2006
Navigation
Log in


Forgot your password?
New user?
 
Views
Actions
Technical committee meetings >>

Minutes 20 November 2006

last edited 2 years ago by karl

Present: Jagir, Karl, Shane, Peter.

AIX

  • Shane built PETSc on hpc.
  • MPI headers on AIX seem different to mpich so some tweaks were required in libMesh build.
  • Parallel environment doesn't seem to be working at the moment: seems to need rsh.

CellML

  • Integrator
    • cvodes (from Sundials) maintains a lot of state information.
      • Expensive in memory
      • Not sure how well swapping the history information in and out of a single solver will work.
    • How complex an integrator do we need?
      • Complex integrators work well on single cell models but if we need a small time step for diffusion, is there any point in having a complex integrator?
      • Start with integrators that don't need their own history information.

Material base vectors from angles

  • Karl has written some code for calculating these but it is not working yet.


comments:

cell model integrator --andre, Mon, 20 Nov 2006 21:12:17 +1300 reply
The LSODA implementation from the current cm could be a good option for use in integrating cellular models. Its the common one to use currently for stiff models that do still require a decent integrator even when using a small diffusion time step. And it is already built as a separate library, although possibly some of the interface code will need to be taken from cm. It may also be worth checking with Alan Garny, who possibly uses CVODE with multi-cell simulations in COR.

cell model integrator --karl, Thu, 23 Nov 2006 16:33:43 +1300 reply
Thanks, Andre. Alan Garny mentioned that CVODE provides a Newton integrator which doesn't use much memory (or not as much as others) when used with the right linear solver.