dolfin team mailing list archive

[Ola Skavhaug <skavhaug@xxxxxxxxx>]

Garth N. Wells skrev den 08/07-2008 følgende:
>
>
> Ola Skavhaug wrote:
>> Anders Logg skrev den 08/07-2008 følgende:
>>> On Tue, Jul 08, 2008 at 09:35:15AM +0200, Ola Skavhaug wrote:
>>>> Could we add a default constructor to Parameter, e.g., letting it default to
>>>> int or similar? This will make it possible to wrap it with SWIG. The current
>>>> state of SWIG will declare variables like this (depending on where
>>>> Parameter lives in the argument list):
>>>>
>>>> Parameter arg3;
>>>>
>>>> This fails due to the lack of default constructor for Parameter. I don't know
>>>> how to get around this silly SWIG behaviour.
>>>>
>>>> Ola
>>> We could add a default constructor for an int parameter named
>>> "undefined" with value 0 or similar. Parameter is never used directly
>>> by a user anyway (since it will be cast automatically to the
>>> underlying type). 
>>
>> I don't see how naming the parameter "undefined" would help, since 0 should be
>> a valid interger option. Except from that, I follow.
>>
>> I have made some improvements to the SLEPc eigenvalue solver, making it
>> possible to set tolerance and max iter from dolfin. Another thing we should
>> consider is making it possible to specify the problem type, EPS_HEP, EPS_NHEP,
>> EPS_GHEP, or EPS_GNHEP (or so).
>>
>
> The SLEPc interface is very primitive, so improvements are welcome.
>
> Ola, I guess this means that you're using it?

Yes, I'm trying to solve very big matrix eigenvalue problems. My problems are
non-symmetric and very sparse, and I'm only interested in the largest real
eignvalue to find out if the system
u' = Au
is stable or not. It seems like this is a quite difficult problem, at least
the current way dolfin communicates and integrates with SLEPc.

Ola 

> Garth
>
>> Ola
>>  
>>> -- Anders
>>
>>
>> _______________________________________________
>> DOLFIN-dev mailing list
>> DOLFIN-dev@xxxxxxxxxx
>> http://www.fenics.org/mailman/listinfo/dolfin-dev
>
>