about MatNullSpace

[Barry Smith]

On Sep 1, 2008, at 4:36 PM, Lisandro Dalcin wrote:

> 1) Given that we can attach null spaces to KSP, Do you believe there
> are good use cases for attaching them to Mat?

No.

That said, MatMFFDAddNullSpace() does exist. This is because due to
round-off in the differencing the null space may be introduced where it
doesn't belong.
>
>
> 2) Given that currently the A and B operators of KSP actually live
> inside the PC, perhaps the KSP null space should also live inside the
> PC, ant PC should be in charge of removing the null space. What do you
> think about this?

    When the null space is removed is really a property of the KSP  
method
not the PC. You can't just say "everytime you apply the PC you remove  
the
null space". It might be possible to force the null space down to the  
PC but
I don't think it is the right place for it.

    Barry

>
>
>
>
> -- 
> Lisandro Dalcín
> ---------------
> Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
> Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
> Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
> PTLC - Güemes 3450, (3000) Santa Fe, Argentina
> Tel/Fax: +54-(0)342-451.1594
>