Re: [ublas] Problems with stability and ublas:.vector assignment!

From: Vardan Akopian (vakopian_at_[hidden])
Date: 2007-03-08 14:46:45

Gunter,

Looking at the line where the error is produced, it looks like
prod(sparse, dense) is passed to vector_assign (it is going through
the constructor, btw) with a sparse tag. Is this normal? Shouldn't
prod(sparse, dense) generally be considered as dense?

-Vardan

On 3/8/07, Gunter Winkler <guwi17_at_[hidden]> wrote:
> On Thursday 08 March 2007 10:33, Preben Hagh Strunge Holm wrote:
> > It's nice to see that the matrix is of size:
> > size1_ = 66, size2_ = 72
> >
> > but I don't know if it's nice to see that index1_data_ and index2_data_
> > doesn't match these sizes. _index1_data_ is of size 67 and index2_data_
> > of size 1056... Does this matter?
>
> the maximal size of index2_data_ is size1_*size2_ so the number is ok. You can
> check all these values via
> a.nnz(), a.filled1() (should be 1+size1), a.filled2() (should be ==a.nnz())
> and a.nnz_capacity() should be the size of index2_data_ and value_data_
>
> about the assignement problem:
> can you try to explicitly use the copy constructor syntax?
>
> vector dg( prod( ... ) );
>
> (Hoever this should be equivalent for the compiler)
>
> and try to split the creation and the assignement
>
> vector dg( prod_size );
> noalias(dg) = prod(...);
>
>
> mfg
> GUnter
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>