Re: [ublas] proper use of lu_factorize

From: Paul C. Leopardi (leopardi_at_[hidden])
Date: 2005-02-23 18:01:58

Shawn,
I just checked GluCat 0.1.8, and in matrix_multi_imp.h it has

[snip]
      const matrix_t& AT = ublas::trans(rhs.m_matrix);
      matrix_t LU = AT;
      typedef ublas::permutation_matrix<matrix_index_t> permutation_t;
      permutation_t pvector(AT.size1());
      if (! ublas::lu_factorize(LU, pvector))
[snip]

Does this help?
Best regards, Paul Leopardi

On Thursday 24 February 2005 07:56, Shawn D. Pautz wrote:
> Hello,
>
> What is the proper use of the pivoting version of lu_factorize?  In
> particular, are there requirements on the initial contents of the
> pivoting vector 'pm'?  I get an assertion thrown unless I initialize
> each element of the vector to be equal to its index.  I had thought that
> I only needed to make sure it had the right size.  An example would be
> very helpful.
>
> This occurs in Boost 1.32.0 and also in the head version in the Boost
> repository.