From: Martin Kronbichler Date: Wed, 22 Apr 2009 12:47:08 +0000 (+0000) Subject: Mention the change in the boost DoF renumbering scheme. On the other hand, we can... X-Git-Tag: v8.0.0~7790 X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=ad78626fd345df409981719f01d44c8eb5514efd;p=dealii.git Mention the change in the boost DoF renumbering scheme. On the other hand, we can now also use constraints with boost. git-svn-id: https://svn.dealii.org/trunk@18689 0785d39b-7218-0410-832d-ea1e28bc413d --- diff --git a/deal.II/deal.II/source/dofs/dof_renumbering.cc b/deal.II/deal.II/source/dofs/dof_renumbering.cc index 4435682ef7..25db55869d 100644 --- a/deal.II/deal.II/source/dofs/dof_renumbering.cc +++ b/deal.II/deal.II/source/dofs/dof_renumbering.cc @@ -173,7 +173,6 @@ namespace DoFRenumbering types::Graph &graph, types::property_map::type &graph_degree) { - Assert (use_constraints == false, ExcNotImplemented()); { // create intermediate sparsity pattern // (faster than directly submitting diff --git a/deal.II/doc/news/changes.h b/deal.II/doc/news/changes.h index d9ac8f816e..c31235da72 100644 --- a/deal.II/doc/news/changes.h +++ b/deal.II/doc/news/changes.h @@ -30,6 +30,16 @@ inconvenience this causes.

    +
  1. +

    + Changed: The implementation of the reordering algorithms in the boost + namespace created multiple edges between two nodes. The new implementation + submits not more than one edge, but that affects the ordering of the + degrees of freedom and deviates from previous behavior. +
    + (Martin Kronbichler 2009/04/22) +

    +
  2. Changed: The files implementing the class ConstraintMatrix are now @@ -617,12 +627,22 @@ inconvenience this causes.

    lac

      +
    1. +

      + Updated: The SparseMatrix::precondition_SSOR function was implemented + inefficiently. The new implementation uses about half the instruction + count and reduces execution times on typical applications by ten to + thirty percent. +
      + (Martin Kronbichler 2009/04/21) +

      +
    2. Updated: The SparseILU::vmult kernel has been re-written to use similar data structures as SparseMatrix::vmult, which reduces the count of - operations by one third and the execution times on typical problems by - ten to twenty precent. + operations by one third and the execution times on typical applications + by ten to twenty percent.
      (Martin Kronbichler 2009/04/16)