Re: [ublas] A x = b; binding for large sparse matrix A

From: Gunter Winkler (guwi17_at_[hidden])
Date: 2007-12-05 17:43:15

Am Montag, 3. Dezember 2007 15:14 schrieb Per Abrahamsen:
> Gunter Winkler <guwi17_at_[hidden]> writes:
> > The AFAIK simplest way is to try
> >
> > http://guwi17.de/ublas/ublas_csparse.hpp
>
> Do you have a small example preferable of how to use this to solve a
> linear equation system?

// works on 32bit and 64bit systems
compressed_matrix<
  double,
  basic_row_major<unsigned int, int>,
  0,
  unbounded_array<unsigned int>,
  ublas::unbounded_array<double>
>  a_compressed_matrix;

// declare variable (T=double, size_type=unsigned int)
CS::cxsparse_type_traits<T, size_type>::lu_type     cs_lu;

// decompose
cs_lu = CS::cs_ul_decompose( a_compressed_matrix, 1.0 );

// solve
ublas::vector<double> temp = <rhs>;
CS::cs_ul_solve( cs_lu, temp );
<solution> = temp;

// free
CS::cs_lu_free(cs_lu);

HTH
Gunter