[Trilinos-Users] EpetraExt::MatrixMatrix::Multiply

[Williams, Alan B]

The Multiply function needs a pre-processing step to first determine the structure of the result matrix, and then create/fill the result matrix.

I'll look into correcting this.

Alan


> -----Original Message-----
> From: trilinos-users-bounces at software.sandia.gov
> [mailto:trilinos-users-bounces at software.sandia.gov] On Behalf
> Of Kurt Stokbro
> Sent: Wednesday, February 20, 2008 7:40 AM
> To: trilinos-users at software.sandia.gov
> Subject: [Trilinos-Users] EpetraExt::MatrixMatrix::Multiply
>
> We have been using EpetraExt::MatrixMatrix::Multiply, and
> discovered that it uses enourmously amounts of memory, for
> 1000x1000 matrices, it may use more than 100 times of the
> memory of one of the matrices to perform the product. In our
> case this resulted in core dumps.
>
> The behaviour seems to be related with that FillComplete is
> first called when the function returns, thus all terms in the
> matrix product is stored temporarily.
>
> Maybe the behaviour of the distributed
> EpetraExt::MatrixMatrix::Multiply should be changed.
>
> We will use our own version.
>
>
> Kurt Stokbro
> Professor,
> Department of Computer Science
> University of Copenhagen
> Universitetsparken 1
> DK-2100 Copenhagen, Denmark
> cell: 51320610
>
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users
>