| 1 |
/*! |
| 2 |
\mainpage Anasazi: A Block Eigensolvers Package |
| 3 |
|
| 4 |
\section intro Introduction |
| 5 |
|
| 6 |
%Anasazi is a powerful, extensible and interoperable C++ framework for |
| 7 |
the numerical solution of large-scale eigenvalue problems. The |
| 8 |
motivation for this framework is to provide a generic interface to a |
| 9 |
collection of algorithms for solving large-scale eigenvalue |
| 10 |
problems. %Anasazi's interoperability results from its understanding |
| 11 |
of operators and vectors as opaque objects---these objects are |
| 12 |
accessed only via elementary operations. Hence the user has the |
| 13 |
flexibility to specify the data representation for operators and |
| 14 |
vectors and may so leverage any existing software investment. This |
| 15 |
mechanism is accomplished via the \ref anasazi_opvec_interfaces |
| 16 |
"Anasazi Operator/Vector abstract interface". Current interfaces |
| 17 |
available include Thyra, Epetra and Tpetra. As a result, any existing |
| 18 |
code employing Thyra, Epetra or Tpetra operators and vectors (such as |
| 19 |
other packages in Trilinos) may also be used in conjunction with |
| 20 |
%Anasazi. |
| 21 |
|
| 22 |
The \ref anasazi_solver_framework "Anasazi eigensolver framework" |
| 23 |
describes eigensolver and eigensolver managers that provide efficient, |
| 24 |
convenient and powerful computational methods. %Anasazi's power is a |
| 25 |
result of the distribution of functionality across different |
| 26 |
computational entities in the framework: orthogonalization, |
| 27 |
sorting/selection, eigenvalue iterations, restarting methodologies, |
| 28 |
etc. %Anasazi's extensibility comes via the abstract interfaces |
| 29 |
describing these entities. %Anasazi currently provides a robust set of |
| 30 |
options, and users are able to expand this set to suit specific |
| 31 |
research and/or application needs. The ability to interact directly |
| 32 |
with these objects provides great flexibility in computation, while |
| 33 |
the existence of solver managers allow convenient programming for |
| 34 |
standard use cases. |
| 35 |
|
| 36 |
\section solvers Current Eigensolvers |
| 37 |
|
| 38 |
Unlike <a href="http://www.caam.rice.edu/software/ARPACK/">ARPACK</a>, |
| 39 |
which provides a single eigensolver, %Anasazi provides a framework |
| 40 |
capable of describing a wide variety of eigenproblems and algorithms |
| 41 |
for solving them. %Anasazi can currently solve complex and real, |
| 42 |
Hermitian and non-Hermitian, eigenvalue problems, via the following |
| 43 |
included methods: |
| 44 |
|
| 45 |
<ul> |
| 46 |
<li> |
| 47 |
Block Krylov-Schur method, a block extension of <em>A Krylov-Schur |
| 48 |
Algorithm for Large Eigenproblems</em>, G. W. Stewart, SIAM J. Matrix |
| 49 |
Anal. Appl., 23, pp. 601-614 (2000). |
| 50 |
</li> |
| 51 |
|
| 52 |
<li> |
| 53 |
Block Davidson method described in <em>A Comparison of Eigensolvers |
| 54 |
for Large-scale 3D Modal Analysis Using AMG-Preconditioned Iterative |
| 55 |
Methods</em>, P. Arbenz, U. L. Hetmaniuk, R. B. Lehoucq, |
| 56 |
R. S. Tuminaro, Internat. J. for Numer. Methods Engrg., 64, |
| 57 |
pp. 204-236 (2005). |
| 58 |
</li> |
| 59 |
|
| 60 |
<li> |
| 61 |
LOBPCG, a stable implementation of <em>Toward the optimal |
| 62 |
preconditioned eigensolver: Locally optimal block preconditioned |
| 63 |
conjugate gradient method</em>, SIAM J. Sci. Comput., 23 (2001), |
| 64 |
pp. 517-541, as described in <em>Basis selection in LOBPCG</em>, |
| 65 |
U. L. Hetmaniuk and R. B. Lehoucq, J. Comput. Physics, 218, |
| 66 |
pp. 324-332 (2006) <li>IRTR, an implicit version of the Riemannian |
| 67 |
Trust-Region Eigensolver, orginally described in <em>A truncated-CG |
| 68 |
style method for symmetric generalized eigenvalue problems</em>, |
| 69 |
P.-A. Absil, C. G. Baker, and K. A. Gallivan, J. Computational and |
| 70 |
Applied Mathematics, 189, pp. 274-285 (2006). |
| 71 |
</li> |
| 72 |
|
| 73 |
<li> |
| 74 |
Generalized Davidson method, a non-symmetric variant of the block |
| 75 |
Davidson solver described above. This solver is <em>experimental</em>, |
| 76 |
but is currently in active use. It currently only supports real-valued |
| 77 |
scalar types. |
| 78 |
</li> |
| 79 |
|
| 80 |
<li> |
| 81 |
TraceMin method described in <em>A trace minimization algorithm for |
| 82 |
the generalized eigenvalue problem</em>, A. Sameh and J. Wisniewski, |
| 83 |
SIAM J. Numer. Anal., 19, pp. 1243-1259 (1982). This solver is |
| 84 |
<em>experimental</em>. |
| 85 |
</li> |
| 86 |
|
| 87 |
<li> |
| 88 |
TraceMin-Davidson method described in <em>The trace minimization |
| 89 |
method for the symmetric generalized eigenvalue problem</em>, A. Sameh |
| 90 |
and Z. Tong, J. Computational and Applied Mathematics, 123, |
| 91 |
pp. 155-175 (2000). This solver is <em>experimental</em>. |
| 92 |
</li> |
| 93 |
</ul> |
| 94 |
|
| 95 |
\section contributors Anasazi Contributors |
| 96 |
|
| 97 |
The following people have contributed to the development of %Anasazi: |
| 98 |
|
| 99 |
<ul> |
| 100 |
<li> <a href="http://www.csm.ornl.gov/~cbaker">Chris Baker</a>, Oak Ridge National Laboratory </li> |
| 101 |
<li> Steve Hamilton, Oak Ridge National Laboratory </li> |
| 102 |
<li> <a href="http://www.hetmaniuk.net/Site/Welcome.html">Ulrich Hetmaniuk</a>, University of Washington </li> |
| 103 |
<li> Mark Hoemmen, Sandia National Laboratories </li> |
| 104 |
<li> Alicia Klinvex, Purdue University </li> |
| 105 |
<li> <a href="http://www.sandia.gov/~rblehou">Rich Lehoucq</a>, Sandia National Laboratories </li> |
| 106 |
<li> <a href="http://www.cs.sandia.gov/~hkthorn/">Heidi Thornquist</a>, Sandia National Laboratories </li> |
| 107 |
</ul> |
| 108 |
|
| 109 |
*/ |