Derek Gaston wrote:
> Ok - I went ahead and committed that so that I can work on the next  
> step... which is allowing you to attach a Preconditioner to the linear  
> solver.  But now we need some discussion on that.
> 
> First, do we want to support the old style of specifying a  
> preconditioner type for a LinearSolver?

Supporting old APIs is always good.  If you want to stick a deprecated() 
warning in there now and plan to get rid of it in the future, though, 
that's fine.

In the near term, I'd like to have a factory method that takes a 
PreconditionerType and generates a Preconditioner.  Then, when you call 
set_preconditioner_type it uses that factory to build a new 
Preconditioner, or when you call set_preconditioner it calls 
set_preconditioner_type(pre.type()) to keep in sync.

Naturally, your status as "guy writing the code" gives you veto power 
over "guy pontificating about how the code *should* be written".

> Along these same lines... if we keep the preconditioner type setting  
> capability should the linear solvers build Preconditioners and set  
> their appropriate type?  Or should we continue to do everything the  
> way we currently do unless you specifically attach a Preconditioner  
> object to the linear solver?

I'd say initialize the LinearSolver with whatever Preconditioner 
corresponds to the default ILU_PRECOND type; these things should be 
lightweight enough (until they're initialized with a matrix) that that's 
convenient and cheap enough.
---
Roy

------------------------------------------------------------------------------
Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
software. With Adobe AIR, Ajax developers can use existing skills and code to
build responsive, highly engaging applications that combine the power of local
resources and data with the reach of the web. Download the Adobe AIR SDK and
Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to