diff options
author | Gael Guennebaud <g.gael@free.fr> | 2015-06-26 10:49:40 +0200 |
---|---|---|
committer | Gael Guennebaud <g.gael@free.fr> | 2015-06-26 10:49:40 +0200 |
commit | 555b9c684346129057b14d73db64a75274fdfe0f (patch) | |
tree | d35749e34b390790a9cd12c7e9309391baf69309 /Eigen/src/IterativeLinearSolvers/ConjugateGradient.h | |
parent | 53b930887d118af5204840231f08b3307addce4e (diff) |
Doc: explain perf and multithreading issues in sparse iterative solvers
Diffstat (limited to 'Eigen/src/IterativeLinearSolvers/ConjugateGradient.h')
-rw-r--r-- | Eigen/src/IterativeLinearSolvers/ConjugateGradient.h | 12 |
1 files changed, 9 insertions, 3 deletions
diff --git a/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h b/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h index 5f0159e52..8f33c446d 100644 --- a/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h +++ b/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h @@ -114,14 +114,20 @@ struct traits<ConjugateGradient<_MatrixType,_UpLo,_Preconditioner> > * * \tparam _MatrixType the type of the matrix A, can be a dense or a sparse matrix. * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower, - * Upper, or Lower|Upper in which the full matrix entries will be considered. Default is Lower. + * \c Upper, or \c Lower|Upper in which the full matrix entries will be considered. + * Default is \c Lower, best performance is \c Lower|Upper. * \tparam _Preconditioner the type of the preconditioner. Default is DiagonalPreconditioner * * The maximal number of iterations and tolerance value can be controlled via the setMaxIterations() * and setTolerance() methods. The defaults are the size of the problem for the maximal number of iterations * and NumTraits<Scalar>::epsilon() for the tolerance. * - * The tolerance is the relative residual error: |Ax-b|/|b| + * The tolerance corresponds to the relative residual error: |Ax-b|/|b| + * + * \b Performance: Even though the default value of \c _UpLo is \c Lower, significantly higher performance is + * achieved when using a complete matrix and \b Lower|Upper as the \a _UpLo template parameter. Moreover, in this + * case multi-threading can be exploited if the user code is compiled with OpenMP enabled. + * See \ref TopicMultiThreading for details. * * This class can be used as the direct solver classes. Here is a typical usage example: \code @@ -129,7 +135,7 @@ struct traits<ConjugateGradient<_MatrixType,_UpLo,_Preconditioner> > VectorXd x(n), b(n); SparseMatrix<double> A(n,n); // fill A and b - ConjugateGradient<SparseMatrix<double> > cg; + ConjugateGradient<SparseMatrix<double>, Lower|Upper> cg; cg.compute(A); x = cg.solve(b); std::cout << "#iterations: " << cg.iterations() << std::endl; |