> o Will make your suggested mods and try them on the demos and then our model. 
>  Thanks very much, James, for the insight into FiPy's innards.
> 
> Regards,
> +jtg+

Also, on another quick note: the modifications for eliminating preconditioners 
and matching residual norms that I've detailed are kind of hacky as they have 
you modifying internal FiPy code. With Trilinos, you can currently instantiate 
a solver from the top level with the parameter ``precon=None'' passed to 
disable the preconditioner. However, when I began working on this problem, 
doing the same thing using PySparse yielded no change, as the preconditioner 
was hardwired in for PySparse solvers. That's why I had you modifying source 
directly.

In light of all this, I've written up a quick patch that'll let you specify one 
(or None) of two preconditioners (JacobiPreconditioner() or 
SsorPreconditioner()) for PySparse solvers. Selection works in exactly the same 
way that it does for Trilinos; simply pass in ``precon=Foo'' when you 
instantiate a PySparse solver. These changes will be committed to branch 
momentarily.

NB: You are still going to need to modify internal FiPy code to match the 
residual norms.

James

Reply via email to