libMesh: A C++ Finite Element Library

I have a question about the initialization of a petsc_matrix.

The init call is

  init (const unsigned int m,
          const unsigned int n,
          const unsigned int m_l,
          const unsigned int n_l,
          const std::vector<unsigned int>& n_nz,
          const std::vector<unsigned int>& n_oz);

where n_nz is the number of non-zeros per row. Hence n_nz.size() = m_l 
on each processor, no?

However there's an assert in petsc_matrix.C, line 110:

   libmesh_assert (n_nz.size() == n_l);

What am I missing here?

Thanks!

Jens

Re: [Libmesh-users] petsc_matrix From: Roy Stogner <roystgnr@ic...> - 2012-08-15 21:38 On Wed, 15 Aug 2012, Jens Lohne Eftang wrote: > init (const unsigned int m, > const unsigned int n, > const unsigned int m_l, > const unsigned int n_l, > const std::vector<unsigned int>& n_nz, > const std::vector<unsigned int>& n_oz); > > where n_nz is the number of non-zeros per row. Hence n_nz.size() = m_l > on each processor, no? > > However there's an assert in petsc_matrix.C, line 110: > > libmesh_assert (n_nz.size() == n_l); > > What am I missing here? It looks to me like you caught a bug. Most of our apps only test symmetric matrices. Looks like David Knezevic added this code, though; I'll Cc: him before fixing it. --- Roy

Re: [Libmesh-users] petsc_matrix From: David Knezevic <dknezevic@se...> - 2012-08-15 21:53 On 08/15/2012 05:38 PM, Roy Stogner wrote: > > On Wed, 15 Aug 2012, Jens Lohne Eftang wrote: > >> init (const unsigned int m, >> const unsigned int n, >> const unsigned int m_l, >> const unsigned int n_l, >> const std::vector<unsigned int>& n_nz, >> const std::vector<unsigned int>& n_oz); >> >> where n_nz is the number of non-zeros per row. Hence n_nz.size() = m_l >> on each processor, no? >> >> However there's an assert in petsc_matrix.C, line 110: >> >> libmesh_assert (n_nz.size() == n_l); >> >> What am I missing here? > > It looks to me like you caught a bug. Most of our apps only test > symmetric matrices. Looks like David Knezevic added this code, > though; I'll Cc: him before fixing it. I agree, looks like a bug. Apologies! David

9 messages have been excluded from this view by a project administrator.