/* ML: A Multilevel Preconditioner Package Copyright (2002) Sandia Corporation Under terms of Contract DE-AC04-94AL85000, there is a non-exclusive license for use of this work by or on behalf of the U.S. Government. This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with this library; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA Questions? Contact Michael A. Heroux (maherou@sandia.gov) */ ML -- History of visible changes ================================ This file reports (some of) the visible changes to ML. Please submit ML bugs at software.sandia.gov/buzgilla, or report bugs to jhu@sandia.gov. ------------------------------------------------------------------------------ Jan 2005 * The first cut of the MLAPI is now available. MLAPI is the MATLAB-like parallel programming module of ML. Right now only a limited subset of ML functionalities is available. * Improved the MultiLevelPreconditioner class (now derived from Epetra_Operator and no longer from Epetra_RowMatrix). Updated the filtering/GGB algorithm, which now supports only Anasazi and the non-symmetric cycle. Some never-used methods have been deleted (like scaling). * Improved compilation on several machines; now ML *should* compile with no warnings on Linux using `-g -ansi -pedantic -Wall -Wno-format-extra-args'. ------------------------------------------------------------------------------ Dec 2004 * improved the organization of the examples, introduces more detailed comments and some README files. Deleted some outdated examples. * bug fixes to the auxiliary matrix algorithm for problems defined on stretched grids. ------------------------------------------------------------------------------ Nov 2004 (current version 3.1) * ML now supports Zoltan as coarsening scheme (RCB only). Zoltan can be used to create the aggregates, but also to redistribute the next-level operator among the specified number of processors (which can be less than the available number of processoes) * Several minor and major fixes to the example suite * Added a test to verify the correctness of compiled examples * Extended the test harness scripts and files * a variable-block partiononing scheme (based on METIS) is now available. ------------------------------------------------------------------------------ Oct 2004 * support for new IFPACK smoothers. ML now supports IFPACK's variable block Jacobi, Gauss-Seidel and symmetric Gauss-Seidel preconditioners, as well as the IFPACK-Amesos preconditioners (subdomain-based exact LU factorizations) * fixes support for Amesos solver (introducing LAPACK as coarse solver) ------------------------------------------------------------------------------ May 2004 (current version 3.0, distributed with Trilinos 4.0) * Update the ml user's guide (in the doc subdirectory) * Improved interface to C++. Introduced two classes to use ML as a preconditioner for Epetra_LinearProblem (both in the ML_Epetra namespace): MultiLevelOperator and MultiLevelPreconditioner. * Basic feaures to solve Maxwell equations via MultiLevelPreconditioner are now supported * Improved interface to Amesos. Now ML can use Anasazi to estimate the maximum eigenvalue for both symmetric and non-symmetric problems. * Improve interface to Anasazi * Add power method to estimate lambda max for non-syymmetric problems * Improve the MLS preconditioner for block systems * New options for configure. Drop --with-ml_epetra (and similar for other Trilinos packages), and use --enable-epetra (and similar) * Fixed various compiler warnings for LINUX (GNU and Intel), SGI 64, DEC, west-cplant, ASCI Red ------------------------------------------------------------------------------ Apr 2004 * Change to support new Teuchos methods' name * Change to support new Amesos methods' name * Now support to ML native interface to SuperLU and SuperLUDIST is deprecated in favor of Amesos interface ------------------------------------------------------------------------------ Jan 2004 * Add METIS and ParMETIS aggregation schemes * Add Teuchos support * Add interface to Amesos * Add interface to IFPACK * Add interface to Anasazi * Add configure options --with-ml_epetra, --with-ml_parmetis2x, --with-ml_parmetis3x, --with-ml_ifpack, --with-ml_anasazi * Add two-level additive and hybrid domain decomposition preconditioners (not officially supported) ------------------------------------------------------------------------------ Sep 2003 (version 2.0, distributed with Trilinos 3.1) * ML has now been put "autotooled", i.e., autoconf and automake. * Haim Waisman's GGB method has been implemented. * Bill Cochran's cache optimizations for multiple RHSs have been implemented. * Bill also has added code so that ML will compile with c++ compilers. ------------------------------------------------------------------------------ ml2_1_0_0-branch-patch_1 * 1st patch for release ml2_1_0_0 (trilinos-release-3-01-branch). * Bug fix for eigenvalue estimate that was causing Alegra sinmag benchmark to fail. * Added definitions for make macro CXX in architecture-dependent makefiles in order for Alegra to be able to build epetra interfaces. * Minor changes to block matrix (equivalent real form) capabilities. * Fixed compiler warnings. ------------------------------------------------------------------------------ mystring = ml2_1_0_0 * Many minor fixes to get the commuting relationship T_h P_n = P_e T_H exactly right. * Serial implementation of enriched prolongator. * New norm for convergence of Maxwell solver. This norm measure and scales properly the range space and null space components of the (curl,curl) operator. ------------------------------------------------------------------------------ mystring = ml2_0_0_5 * Changes in aggregation post-processing to improve aggregate quality. * Added wrappers so that ML can use Epetra objects & so that AztecOO can use ML as a preconditioner. * Added block Maxwell capabilities for real block systems that are equivalent to the complex systems arising from Maxwell in the frequency domain. ------------------------------------------------------------------------------ mystring = ml2_0_0_3 * Added in smoothing of Reitzinger/Schoeberl prolongator. * Least squares and smoothed l.s. prolongators in serial for R/S AMG. * Enhanced detection of degenerate coarse grid matrices in building R/S AMG hierarchy. ------------------------------------------------------------------------------ mystring = ml2_0_0_2 * Symmetrize Hiptmair smoother and finished off MLS within Hiptmair. ------------------------------------------------------------------------------ ml2_0_0_1 * MLS polynomial smoother * matrix does not need to be scaled. * Added more print control: - "ml_inputfile" now takes an additional argument in the general - problem specs section: - "Print Level = [an integer]" - 0 => minimal output - 1 => some output (default) - 2,3,... => increasing amounts of output ------------------------------------------------------------------------------ ml2_0_0_0 * AMG for Maxwell's equations * Reitzinger Schoeberl AMG. * Hiptmair smoother.

Generated on Thu Sep 18 12:42:34 2008 for ML by 1.3.9.1