dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On Jun 16 2008, Anders Logg wrote:

> On Sat, Jun 14, 2008 at 05:31:56PM -0400, Hatef Monajemi wrote:
> > Hello,
> >
> > I am trying to solve a problem with different parameters; I am having a
> > loop and each time I am getting a solution vector (say Function u1) and
> > I need to save each vector as a column of an arbitrary matrix in dolfin
> > (say Matrix S). I would appreciate if you let me know how it is
> > possible. I tried the following scheme:
> >
> > Loop over i:
> >
> > Matrix S;
> > d=u1.vector();
> >
> > for(j)
> > {  
> >   S(i,j)=d(j);
> > {
> >
> > But it is not going to work! any help on this matter is appreciated.
> >
> >
> > Hatef:)
>
> Accessing columns in our sparse matrices is very inefficient. You
> could use a dense matrix, but I suggest storing each solution to a
> separate file as a Function:
>
>  file << u1;

Columns can be accessed efficiently using a uBlasMatrix with an underlying 
uBLAS column-major matrix, e.g.

 uBlasMatrix<ublas_sparse_matrix_cmajor> A;

What won't then work well is zeroing rows to apply boundary conditions.

Garth

>