[petsc-dev] Failure converting MPIBAIJ -> MPIAIJ

[Barry Smith]

On Sep 19, 2011, at 10:12 AM, Lisandro Dalcin wrote:

> On 19 September 2011 11:06, Hong Zhang <hzhang at mcs.anl.gov> wrote:
>> Lisandro :
>> I cannot reproduce this error using a C code (ex160.c attached).
>> 
> 
> You have to explicitly ask for MATMPIBAIJ, change the MatSetType like below:
> 
>  ierr = MatSetType(A,MATMPIBAIJ);CHKERRQ(ierr);
> 
> With that change, you C code fails exactly the same as my Python code.
> 
> I knot this makes little sense (I mean, creating MPIBAIJ on a single
> process), and perhaps fixing this issue is not worth it,

   It is worth fixing and must be fixed. If MPIBAIJ matrices don't work in the trivial case one cannot expect them to work in the parallel case.

   Thanks for bringing this up.

    Barry

   
> but I wanted
> to let you know about the error. I think making MPIXXX types work in a
> single process is an extra sanity check for the code.
> 
>> 
>> Why MatConvert_MPIBAIJ_MPIAIJ()  is being called for sequential matrix?
>> 
> 
> As I said, this is because I'm forcibly creating MPIBAIJ on a single process.
> 
> 
> -- 
> Lisandro Dalcin
> ---------------
> CIMEC (INTEC/CONICET-UNL)
> Predio CONICET-Santa Fe
> Colectora RN 168 Km 472, Paraje El Pozo
> 3000 Santa Fe, Argentina
> Tel: +54-342-4511594 (ext 1011)
> Tel/Fax: +54-342-4511169