fdurastante/amg4psblas
1
0
Fork 0
mirror of https://github.com/sfilippone/amg4psblas.git synced 2026-10-06 14:44:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Updated descriptions

Browse Source
This commit is contained in:
Cirdans-Home
2021-03-08 13:44:46 +01:00
parent ea6a260253
commit 4531532a07
6 changed files with 31 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/src/bibliography.tex
+1 -1
View File
@@ -76,7 +76,7 @@ for Scientific and Engineering Applications}, Appl. Algebra Engrg. Comm. Comput.
%
\bibitem{DDF2020}
P.~D'Ambra, F~Durastante, S.~Filippone,
\emph{AMG preconditioners for Linear Solvers towards Extreme Scale}, 2020, arXiv:2006.16147v2.
\emph{AMG preconditioners for Linear Solvers towards Extreme Scale}, 2020, \href{https://arxiv.org/abs/2006.16147v3arXiv:2006.16147v2}{arXiv:2006.16147v3}.
%
\bibitem{UMFPACK}
T.~A.~Davis,
docs/src/building.tex
+7 -7
View File
@@ -32,9 +32,9 @@ The following base libraries are needed:
\item[BLAS] \cite{blas3,blas2,blas1} Many vendors provide optimized versions
of BLAS; if no vendor version is
available for a given platform, the ATLAS software
(\url{math-atlas.sourceforge.net})
(\href{http://math-atlas.sourceforge.net}{math-atlas.sourceforge.net})
may be employed. The reference BLAS from Netlib
(\url{www.netlib.org/blas}) are meant to define the standard
(\href{http://www.netlib.org/blas}{www.netlib.org/blas}) are meant to define the standard
behaviour of the BLAS interface, so they are not optimized for any
particular platform, and should only be used as a last
resort. Note that BLAS computations form a relatively small part of
@@ -43,14 +43,14 @@ The following base libraries are needed:
libraries. Note that UMFPACK requires a full LAPACK library; our
experience is that configuring ATLAS for building full LAPACK does not
work in the correct way. Our advice is first to download the LAPACK tarfile from
\url{www.netlib.org/lapack} and install it independently of ATLAS. In this case,
\href{http://www.netlib.org/lapack}{www.netlib.org/lapack} and install it independently of ATLAS. In this case,
you need to modify the OPTS and NOOPT definitions for including -fPIC compilation option
in the make.inc file of the LAPACK library.
\item[MPI] \cite{MPI2,MPI1} A version of MPI is available on most
high-performance computing systems.
\item[PSBLAS] \cite{PSBLASGUIDE,psblas_00} Parallel Sparse BLAS (PSBLAS) is
available from \url{github.com/sfilippone/psblas3} ?? {\bf Mettere tutto sotto psctoolkit}; version
3.5.0 (or later) is required. Indeed, all the prerequisites
available from \href{https://psctoolkit.github.io/products/psblas/}{psctoolkit.github.io/products/psblas/}; version
3.7.0 (or later) is required. Indeed, all the prerequisites
listed so far are also prerequisites of PSBLAS.
\end{description}
Please note that the four previous libraries must have Fortran
@@ -102,7 +102,7 @@ in the main directory to generate the necessary makefile.
%\textbf{Sono necessarie le parentesi intorno a s?}
\bf{DA RISCRIVERE}
\textbf{DA RISCRIVERE}
As a minimal example consider the following:
\begin{verbatim}
@@ -280,7 +280,7 @@ it to find libraries and programs with nonstandard names/locations.
Report bugs to <https://github.com/sfilippone/mld2p4-2/issues>.
\end{verbatim}
For instance, if a user has built and installed PSBLAS 3.5 under the
For instance, if a user has built and installed PSBLAS 3.7 under the
\verb|/opt| directory and is
using the SuiteSparse package (which includes UMFPACK), then MLD2P4
might be configured with:
docs/src/distribution.tex
+1 -1
View File
@@ -5,7 +5,7 @@
\noindent
AMG4PSBLAS is available from the web site
\begin{quotation}
\texttt{https://github.com/psctoolkit/amg4psblas}
\href{https://psctoolkit.github.io/products/amg4psblas/}{https://psctoolkit.github.io/products/amg4psblas/}
\end{quotation}
where contact points for further information can be also found.
docs/src/gettingstarted.tex
+8 -8
View File
@@ -33,7 +33,7 @@ The following steps are required:
\item \emph{Build the preconditioner for a given matrix}. If the selected preconditioner
is multilevel, then two steps must be performed, as specified next.
\begin{enumerate}
\item[4.1] \emph{Build the aggregation hierarchy for a given matrix.} This is
\item[4.1] \emph{Build the AMG hierarchy for a given matrix.} This is
performed by the routine \verb|hierarchy_build|.
\item[4.2] \emph{Build the preconditioner for a given matrix.} This is performed
by the routine \verb|smoothers_build|.
@@ -56,26 +56,26 @@ Examples showing the basic use of AMG4PSBLAS are reported in Section~\ref{sec:ex
\begin{table}[h!]
\begin{center}
%{\small
\begin{tabular}{|l|p{1.8cm}|p{8.2cm}|}
\begin{tabular}{|l|p{2cm}|p{8cm}|}
\hline
\textsc{type} & \textsc{string} & \textsc{default preconditioner} \\ \hline
No preconditioner &\verb|'NONE'|& Considered to use the PSBLAS
Krylov solvers with no preconditioner. \\ \hline
Diagonal & \verb|'DIAG'| or \verb|'JACOBI'| & Diagonal preconditioner.
Diagonal & \verb|'DIAG'|, \verb|'JACOBI'|, \verb|'L1-JACOBI'| & Diagonal preconditioner.
For any zero diagonal entry of the matrix to be preconditioned,
the corresponding entry of the preconditioner is set to~1.\\ \hline
Gauss-Seidel & \verb|'GS'| & Hybrid Gauss-Seidel (forward), that is,
Gauss-Seidel & \verb|'GS'|, \verb|'L1-GS'| & Hybrid Gauss-Seidel (forward), that is,
global block Jacobi with
Gauss-Seidel as local solver.\\ \hline
Symmetrized Gauss-Seidel & \verb|'FBGS'| & Symmetrized hybrid Gauss-Seidel,that is,
Symmetrized Gauss-Seidel & \verb|'FBGS'|, \verb|'L1-FBGS'| & Symmetrized hybrid Gauss-Seidel, that is,
forward Gauss-Seidel followed by
backward Gauss-Seidel.\\ \hline
Block Jacobi & \verb|'BJAC'| & Block-Jacobi with ILU(0) on the local blocks.\\ \hline
Block Jacobi & \verb|'BJAC'|, \verb|'L1-BJAC'| & Block-Jacobi with ILU(0) on the local blocks.\\ \hline
Additive Schwarz & \verb|'AS'| & Additive Schwarz (AS),
with overlap~1 and ILU(0) on the local blocks. \\ \hline
Multilevel &\verb|'ML'| & V-cycle with one hybrid forward Gauss-Seidel
(GS) sweep as pre-smoother and one hybrid backward
GS sweep as post-smoother, basic smoothed aggregation
GS sweep as post-smoother, decoupled smoothed aggregation
as coarsening algorithm, and LU (plus triangular solve)
as coarsest-level solver. See the default values in
Tables~\ref{tab:p_cycle}-\ref{tab:p_smoother_1}
@@ -100,7 +100,7 @@ such as those implemented in UMFPACK, MUMPS, SuperLU, and SuperLU\_Dist,
usually lead to smaller numbers of preconditioned Krylov
iterations than inexact solvers, when the linear system comes from
a standard discretization of basic scalar elliptic PDE problems. However,
this does not necessarily correspond to the smallest execution time
this does not necessarily correspond to the shortest execution time
on parallel computers.
docs/src/overview.tex
+1 -2
View File
@@ -29,8 +29,7 @@ two different coarsening strategies, based on aggregation, are available:
\end{itemize}
Either exact or approximate solvers can be used on the coarsest-level system. Specifically, different sparse LU factorizations from external
packages, native incomplete LU factorizations, weighted Jacobi, hybrid Gauss-Seidel,
and block-Jacobi solvers are available. All the smoothers can be also exploited as one-level
packages, native incomplete LU and approximate inverse factorizations, weighted Jacobi, hybrid Gauss-Seidel, block-Jacobi solvers and recursive call to preconditioned Krylov methods are available. All the smoothers can be also exploited as one-level
preconditioners.
AMG4PSBLAS is written in Fortran~2003, following an
docs/src/userinterface.tex
+13 -59
View File
@@ -65,13 +65,7 @@ This method allocates and initializes the preconditioner
\end{tabular}
\vskip1.5\baselineskip
For compatibility with the previous versions of MLD2P4, this method can be also invoked
as follows:
\begin{center}
\verb|call amg_precinit(p,ptype,info)|
\end{center}
\clearpage
@@ -132,13 +126,6 @@ contained in \verb|val|.
underlying objects.
\end{tabular}
\vskip1.5\baselineskip
For compatibility with the previous versions of MLD2P4, this method can be also invoked
as follows:
\begin{center}
\verb|call amg_precset(p,what,val,info)|
\end{center}
\noindent
However, in this case the optional arguments \verb|ilev|,
@@ -238,13 +225,9 @@ See also Sec.~\ref{sec:adding}.
\verb|what| & \textsc{data type} & \verb|val| & \textsc{default} &
\textsc{comments} \\ \hline
\verb|'ML_CYCLE'| & \verb|character(len=*)|
& \texttt{'VCYCLE'} \par \texttt{'WCYCLE'} \par \texttt{'KCYCLE'} \par
\texttt{'MULT'} \par \texttt{'ADD'}
& \texttt{'VCYCLE'} \par \texttt{'WCYCLE'} \par \texttt{'KCYCLE'} \par \texttt{'ADD'}
& \texttt{'VCYCLE'}
&Multilevel cycle: V-cycle, W-cycle, K-cycle, hybrid Multiplicative Schwarz,
and Additive Schwarz. \par
Note that hybrid Multiplicative Schwarz is equivalent to V-cycle and
is included for compatibility with previous versions of MLD2P4. \\ \hline
&Multilevel cycle: V-cycle, W-cycle, K-cycle, and additive composition. \\ \hline
\verb|'OUTER_SWEEPS'| & \texttt{integer} &
Any integer \par number $\ge 1$ & 1 &
Number of multilevel cycles. \\ \hline
@@ -290,7 +273,7 @@ be applied.
if the number of levels reaches this value (see Note). \\ \hline
\verb|'PAR_AGGR_ALG'| & \verb|character(len=*)| \hspace*{-3mm}
& \texttt{'DEC'}, \texttt{'SYMDEC'}
& \texttt{'DEC'}, \texttt{'SYMDEC'}, \texttt{'COUPLED'}
& \texttt{'DEC'}
& Parallel aggregation algorithm. \par Currently, only the
decoupled aggregation (\verb|DEC|) is available; the
@@ -301,7 +284,8 @@ be applied.
\verb|'AGGR_TYPE'| & \verb|character(len=*)| \hspace*{-3mm}
& \textbf{\texttt{'SOC1'}} &
\textbf{\texttt{'SOC1'}},
\textbf{\texttt{'SOC2'}}
\textbf{\texttt{'SOC2'}},
\textbf{\texttt{'MATCHBOXP'}}
& Type of aggregation algorithm: currently,
we implement to measures of strength of
connection, the one by Van\v{e}k, Mandel
@@ -582,7 +566,7 @@ made by the user through the methods \verb|init| and \verb|set|.
& The sparse matrix structure containing the local part of the
matrix to be preconditioned. Note that \emph{x} must be chosen according
to the real/complex,
single/double precision version of MLD2P4 under use.
single/double precision version of AMG4PSBLAS under use.
See the PSBLAS User's Guide for details \cite{PSBLASGUIDE}.\\
\verb|desc_a| & \verb|type(psb_desc_type), intent(in)|. \\
& The communication descriptor of \verb|a|. See the PSBLAS User's Guide for
@@ -615,7 +599,7 @@ hierarchy produced by a previous call to \verb|hierarchy_build|
\verb|a| & \verb|type(psb_|\emph{x}\verb|spmat_type), intent(in)|. \\
& The sparse matrix structure containing the local part of the
matrix to be preconditioned. Note that \emph{x} must be chosen according
to the real/complex, single/double precision version of MLD2P4 under use.
to the real/complex, single/double precision version of AMG4PSBLAS under use.
See the PSBLAS User's Guide for details \cite{PSBLASGUIDE}.\\
\verb|desc_a| & \verb|type(psb_desc_type), intent(in)|. \\
& The communication descriptor of \verb|a|. See the PSBLAS User's Guide for
@@ -659,7 +643,7 @@ unnatural when dealing with simple one-level preconditioners.
\verb|a| & \verb|type(psb_|\emph{x}\verb|spmat_type), intent(in)|. \\
& The sparse matrix structure containing the local part of the
matrix to be preconditioned. Note that \emph{x} must be chosen according
to the real/complex, single/double precision version of MLD2P4 under use.
to the real/complex, single/double precision version of AMG4PSBLAS under use.
See the PSBLAS User's Guide for details \cite{PSBLASGUIDE}.\\
\verb|desc_a| & \verb|type(psb_desc_type), intent(in)|. \\
& The communication descriptor of \verb|a|. See the PSBLAS User's Guide for
@@ -679,13 +663,6 @@ unnatural when dealing with simple one-level preconditioners.
components; this allows e.g. running on GPUs. \\
\end{tabular}
\vskip1.5\baselineskip
For compatibility with the previous versions of MLD2P4, this method can be also invoked
as follows:
\begin{center}
\verb|call amg_precbld(p,what,val,info[,amold,vmold,imold])|
\end{center}
\noindent
The method can be used to build multilevel preconditioners too.
@@ -703,7 +680,7 @@ This method computes $y = op(B^{-1})\, x$, where $B$ is a previously built
preconditioner, stored into \verb|p|, and $op$
denotes the preconditioner itself or its transpose, according to
the value of \verb|trans|.
Note that, when MLD2P4 is used with a Krylov solver from PSBLAS,
Note that, when AMG4PSBLAS is used with a Krylov solver from PSBLAS,
\verb|p%apply| is called within the PSBLAS method \verb|psb_krylov|
and hence it is completely transparent to the user.
@@ -713,11 +690,11 @@ and hence it is completely transparent to the user.
\verb|x| & \emph{type}\verb|(|\emph{kind\_parameter}\verb|), dimension(:), intent(in)|.\\
& The local part of the vector $x$. Note that \emph{type} and
\emph{kind\_parameter} must be chosen according
to the real/complex, single/double precision version of MLD2P4 under use.\\
to the real/complex, single/double precision version of AMG4PSBLAS under use.\\
\verb|y| & \emph{type}\verb|(|\emph{kind\_parameter}\verb|), dimension(:), intent(out)|.\\
& The local part of the vector $y$. Note that \emph{type} and
\emph{kind\_parameter} must be chosen according
to the real/complex, single/double precision version of MLD2P4 under use.\\
to the real/complex, single/double precision version of AMG4PSBLAS under use.\\
\verb|desc_a| & \verb|type(psb_desc_type), intent(in)|. \\
& The communication descriptor associated to the matrix to be
preconditioned.\\
@@ -732,16 +709,9 @@ and hence it is completely transparent to the user.
& Workspace. Its size should be at
least \verb|4 * psb_cd_get_local_| \verb|cols(desc_a)| (see the PSBLAS User's Guide).
Note that \emph{type} and \emph{kind\_parameter} must be chosen according
to the real/complex, single/double precision version of MLD2P4 under use.\\
to the real/complex, single/double precision version of AMG4PSBLAS under use.\\
\end{tabular}
\vskip1.5\baselineskip
For compatibility with the previous versions of MLD2P4, this method can be also invoked
as follows:
\begin{center}
\verb|call amg_precaply(p,what,val,info)|
\end{center}
\clearpage
@@ -761,14 +731,6 @@ This method deallocates the preconditioner data structure \verb|p|.
& Error code. If no error, 0 is returned. See Section~\ref{sec:errors} for details.\\
\end{tabular}
\vskip1.5\baselineskip
For compatibility with the previous versions of MLD2P4, this method can be also invoked
as follows:
\begin{center}
\verb|call amg_precfree(p,info)|
\end{center}
\clearpage
@@ -793,14 +755,6 @@ or \verb|build|, have been called.
will be printed; the default is the standard output.\\
\end{tabular}
\vskip1.5\baselineskip
For compatibility with the previous versions of MLD2P4, this method can be also invoked
as follows:
\begin{center}
\verb|call amg_precdescr(p,info [,iout])|
\end{center}
\subsection{Auxiliary Methods\label{sec:auxil}}
Various functionalities are implemented as additional methods of the
@@ -843,7 +797,7 @@ Create a (deep) copy of the preconditioner object.
\verb|pout| & \verb|type(amg_|\emph{x}\verb|prec_type), intent(out)|.\\
& The copy of the preconditioner data structure. Note
that \emph{x} must be chosen according
to the real/complex, single/double precision version of MLD2P4 under use.\\
to the real/complex, single/double precision version of AMG4PSBLAS under use.\\
\verb|info| & \verb|integer, intent(out)|.\\
& Error code. If no error, 0 is returned. See Section~\ref{sec:errors} for details.\\
\end{tabular}