new

[Krylov_multi.git] / krylov_multi.tex

\documentclass{article}
\usepackage[utf8]{inputenc}
\usepackage{amsfonts,amssymb}
\usepackage{amsmath}
\usepackage{graphicx}
\usepackage{algorithm}
\usepackage{algpseudocode}

\algnewcommand\algorithmicinput{\textbf{Input:}}
\algnewcommand\Input{\item[\algorithmicinput]}

\algnewcommand\algorithmicoutput{\textbf{Output:}}
\algnewcommand\Output{\item[\algorithmicoutput]}


\title{A scalable multisplitting algorithm for solving large sparse linear systems} 
\date{}



\begin{document}
\author{Raphaël Couturier \and Lilia Ziane Khodja}

\maketitle


%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%


\begin{abstract}
In  this paper  we  revisit  the krylov  multisplitting  algorithm presented  in
\cite{huang1993krylov}  which  uses  a  scalar  method to  minimize  the  krylov
iterations computed by a multisplitting algorithm. Our new algorithm is based on
a  parallel multisplitting  algorithm  with few  blocks  of large  size using  a
parallel GMRES method inside each block and on a parallel krylov minimization in
order to improve the convergence. Some large scale experiments with a 3D Poisson
problem  are presented.   They  show  the obtained  improvements  compared to  a
classical GMRES both in terms of number of iterations and execution times.
\end{abstract}


%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%


\section{Introduction}

Iterative methods are used to solve  large sparse linear systems of equations of
the form  $Ax=b$ because they are  easier to parallelize than  direct ones. Many
iterative  methods have  been proposed  and  adapted by  many researchers.   For
example, the GMRES method and the  Conjugate Gradient method are very well known
and  used by  many researchers  ~\cite{S96}. Both  the method  are based  on the
Krylov subspace which consists in forming  a basis of the sequence of successive
matrix powers times the initial residual.

When  solving large  linear systems  with  many cores,  iterative methods  often
suffer  from scalability problems.   This is  due to  their need  for collective
communications  to  perform  matrix-vector  products and  reduction  operations.
Preconditionners can be  used in order to increase  the convergence of iterative
solvers.   However, most  of the  good preconditionners  are not  sclalable when
thousands of cores are used.


Traditionnal iterative  solvers have  global synchronizations that  penalize the
scalability.   Two  possible solutions  consists  either  in using  asynchronous
iterative  methods~\cite{ref18} or  to  use multisplitting  algorithms. In  this
paper, we will  reconsider the use of a multisplitting  method. In opposition to
traditionnal  multisplitting  method  that  suffer  from  slow  convergence,  as
proposed  in~\cite{huang1993krylov},  the  use  of a  minimization  process  can
drastically improve the convergence.




%%%%%%%%%%%%%%%%%%%%%%%
%% BEGIN
%%%%%%%%%%%%%%%%%%%%%%%
The key idea  of the multisplitting method for  solving a large system
of linear equations $Ax=b$ consists  in partitioning the matrix $A$ in
$L$ several ways
\begin{equation}
A = M_l - N_l,~l\in\{1,\ldots,L\},
\label{eq01}
\end{equation}
where $M_l$ are nonsingular matrices. Then the linear system is solved
by iteration based on the multisplittings as follows
\begin{equation}
x^{k+1}=\displaystyle\sum^L_{l=1} E_l M^{-1}_l (N_l x^k + b),~k=1,2,3,\ldots
\label{eq02}
\end{equation}
where $E_l$ are non-negative and diagonal weighting matrices such that
$\sum^L_{l=1}E_l=I$ ($I$ is an identity matrix).  Thus the convergence
of such a method is dependent on the condition
\begin{equation}
\rho(\displaystyle\sum^L_{l=1}E_l M^{-1}_l N_l)<1.
\label{eq03}
\end{equation}

The advantage of  the multisplitting method is that  at each iteration
$k$ there are $L$ different linear sub-systems
\begin{equation}
v_l^k=M^{-1}_l N_l x_l^{k-1} + M^{-1}_l b,~l\in\{1,\ldots,L\},
\label{eq04}
\end{equation}
to be solved  independently by a direct or  an iterative method, where
$v_l^k$  is   the  solution  of   the  local  sub-system.   Thus,  the
calculations  of $v_l^k$  may be  performed in  parallel by  a  set of
processors.   A multisplitting  method using  an iterative  method for
solving the $L$ linear  sub-systems is called an inner-outer iterative
method or a  two-stage method.  The results $v_l^k$  obtained from the
different splittings~(\ref{eq04}) are combined to compute the solution
$x^k$ of the linear system by using the diagonal weighting matrices
\begin{equation}
x^k = \displaystyle\sum^L_{l=1} E_l v_l^k,
\label{eq05}
\end{equation}    
In the case where the diagonal weighting matrices $E_l$ have only zero
and   one   factors  (i.e.   $v_l^k$   are   disjoint  vectors),   the
multisplitting method is non-overlapping  and corresponds to the block
Jacobi method.
%%%%%%%%%%%%%%%%%%%%%%%
%% END
%%%%%%%%%%%%%%%%%%%%%%%

\section{Related works}


A general framework  for studying parallel multisplitting has  been presented in
\cite{o1985multi} by O'Leary and White. Convergence conditions are given for the
most general case.  Many authors improved multisplitting algorithms by proposing
for  example  an  asynchronous  version  \cite{bru1995parallel}  and  convergence
conditions  \cite{bai1999block,bahi2000asynchronous}   in  this  case   or  other
two-stage algorithms~\cite{frommer1992h,bru1995parallel}.

In  \cite{huang1993krylov},  the  authors  proposed  a  parallel  multisplitting
algorithm in which all the tasks except  one are devoted to solve a sub-block of
the splitting  and to send their  local solution to  the first task which  is in
charge to  combine the vectors at  each iteration.  These vectors  form a Krylov
basis for  which the first task minimizes  the error function over  the basis to
increase the convergence, then the other tasks receive the update solution until
convergence of the global system. 



In \cite{couturier2008gremlins}, the  authors proposed practical implementations
of multisplitting algorithms that take benefit from multisplitting algorithms to
solve large scale linear systems. Inner  solvers could be based on scalar direct
method with the LU method or scalar iterative one with GMRES.

In~\cite{prace-multi},  the  authors  have  proposed a  parallel  multisplitting
algorithm in which large block are solved using a GMRES solver. The authors have
performed large scale experimentations upto  32.768 cores and they conclude that
asynchronous  multisplitting algorithm  could more  efficient  than traditionnal
solvers on exascale architecture with hunders of thousands of cores.


%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%


\section{A two-stage method with a minimization}
Let $Ax=b$ be a given sparse  and large linear system of $n$ equations
to  solve  in  parallel   on  $L$  clusters,  physically  adjacent  or
geographically distant, where $A\in\mathbb{R}^{n\times n}$ is a square
and  nonsingular matrix, $x\in\mathbb{R}^{n}$  is the  solution vector
and   $b\in\mathbb{R}^{n}$  is   the  right-hand   side   vector.  The
multisplitting of this linear system is defined as follows:
\begin{equation}
\left\{
\begin{array}{lll}
A & = & [A_{1}, \ldots, A_{L}]\\
x & = & [X_{1}, \ldots, X_{L}]\\
b & = & [B_{1}, \ldots, B_{L}]
\end{array}
\right.
\label{sec03:eq01}
\end{equation}  
where for  $l\in\{1,\ldots,L\}$, $A_l$ is a rectangular  block of size
$n_l\times n$ and $X_l$ and  $B_l$ are sub-vectors of size $n_l$, such
that  $\sum_ln_l=n$.  In this  case,  we  use  a row-by-row  splitting
without overlapping in  such a way that successive  rows of the sparse
matrix $A$ and  both vectors $x$ and $b$ are  assigned to one cluster.
So,  the multisplitting  format of  the  linear system  is defined  as
follows:
\begin{equation}
\forall l\in\{1,\ldots,L\} \mbox{,~} \displaystyle\sum_{i=1}^{l-1}A_{li}X_i + A_{ll}X_l + \displaystyle\sum_{i=l+1}^{L}A_{li}X_i = B_l, 
\label{sec03:eq02}
\end{equation} 
where $A_{li}$ is  a block of size $n_l\times  n_i$ of the rectangular
matrix  $A_l$, $X_i\neq  X_l$ is  a sub-vector  of size  $n_i$  of the
solution vector  $x$ and $\sum_{i<l}n_i+\sum_{i>l}n_i+n_l=n$,  for all
$i\in\{1,\ldots,l-1,l+1,\ldots,L\}$.

The multisplitting method proceeds by iteration for solving the linear
system in such a way each sub-system
\begin{equation}
\left\{
\begin{array}{l}
A_{ll}X_l = Y_l \mbox{,~such that}\\
Y_l = B_l - \displaystyle\sum_{i=1,i\neq l}^{L}A_{li}X_i,
\end{array}
\right.
\label{sec03:eq03}
\end{equation}
is solved  independently by a cluster of  processors and communication
are required  to update the  right-hand side vectors $Y_l$,  such that
the  vectors  $X_i$  represent   the  data  dependencies  between  the
clusters. In this work,  we use the parallel GMRES method~\cite{ref34}
as     an     inner      iteration     method     to     solve     the
sub-systems~(\ref{sec03:eq03}).  It  is a well-known  iterative method
which  gives good performances  to solve  sparse linear  systems in
parallel on a cluster of processors.

It should be noted that  the convergence of the inner iterative solver
for  the  different  linear  sub-systems~(\ref{sec03:eq03})  does  not
necessarily involve  the convergence of the  multisplitting method. It
strongly depends on  the properties of the sparse  linear system to be
solved                 and                the                computing
environment~\cite{o1985multi,ref18}.  Furthermore,  the multisplitting
of the  linear system among  several clusters of  processors increases
the  spectral radius  of  the iteration  matrix,  thereby slowing  the
convergence.  In   this  paper,  we   based  on  the   work  presented
in~\cite{huang1993krylov} to increase  the convergence and improve the
scalability of the multisplitting methods.

In  order  to  accelerate  the  convergence, we  implement  the  outer
iteration  of the multisplitting  solver as  a Krylov  subspace method
which minimizes some error function over a Krylov subspace~\cite{S96}.
The Krylov  space of  the method that  we used  is spanned by  a basis
composed  of   successive  solutions  issued  from   solving  the  $L$
splittings~(\ref{sec03:eq03})
\begin{equation}
S=\{x^1,x^2,\ldots,x^s\},~s\leq n,
\label{sec03:eq04}
\end{equation}
where   for  $j\in\{1,\ldots,s\}$,  $x^j=[X_1^j,\ldots,X_L^j]$   is  a
solution of the  global linear system. The advantage  of such a Krylov
subspace   is  that   we  need   neither  an   orthogonal   basis  nor
synchronizations  between  the  different  clusters to  generate  this
basis.

The  multisplitting   method  is  periodically   restarted  every  $s$
iterations  with   a  new  initial   guess  $\tilde{x}=S\alpha$  which
minimizes  the error  function $\|b-Ax\|_2$  over the  Krylov subspace
spanned  by  the vectors  of  $S$.  So,  $\alpha$  is  defined as  the
solution of the large overdetermined linear system
\begin{equation}
R\alpha=b,
\label{sec03:eq05}
\end{equation}
where $R=AS$  is a  dense rectangular matrix  of size $n\times  s$ and
$s\ll n$. This leads us to solve the system of normal equations
\begin{equation}
R^TR\alpha=R^Tb,
\label{sec03:eq06}
\end{equation}
which is associated with the least squares problem
\begin{equation}
\text{minimize}~\|b-R\alpha\|_2,
\label{sec03:eq07}
\end{equation}  
where $R^T$ denotes the transpose  of the matrix $R$.  Since $R$ (i.e.
$AS$) and  $b$ are  split among $L$  clusters, the  symmetric positive
definite  system~(\ref{sec03:eq06}) is  solved in  parallel.  Thus, an
iterative method would be more  appropriate than a direct one to solve
this system.  We use  the parallel conjugate  gradient method  for the
normal equations CGNR~\cite{S96,refCGNR}.

\begin{algorithm}[!t]
\caption{A two-stage linear solver with inner iteration GMRES method}
\begin{algorithmic}[1]
\Input $A_l$ (local sparse matrix), $B_l$ (local right-hand side), $x^0$ (initial guess)
\Output $X_l$ (local solution vector)\vspace{0.2cm}
\State Load $A_l$, $B_l$, $x^0$
\State Initialize the minimizer $\tilde{x}^0=x^0$
\For {$k=1,2,3,\ldots$ until the global convergence}
\State Restart with $x^0=\tilde{x}^{k-1}$: \textbf{for} $j=1,2,\ldots,s$ \textbf{do}
\State\hspace{0.5cm} Inner iteration solver: \Call{InnerSolver}{$x^0$, $j$}
\State\hspace{0.5cm} Construct the basis $S$: add the column vector $X_l^j$ to the matrix $S_l^k$
\State\hspace{0.5cm} Exchange the local solution vector $X_l^j$ with the neighboring clusters
\State\hspace{0.5cm} Compute the dense matrix $R$: $R_l^{k,j}=\sum^L_{i=1}A_{li}X_i^j$ 
\State\textbf{end for} 
\State Minimization $\|b-R\alpha\|_2$: \Call{UpdateMinimizer}{$S_l$, $R$, $b$, $k$}
\State Local solution of the linear system $Ax=b$: $X_l^k=\tilde{X}_l^k$
\State Exchange the local minimizer $\tilde{X}_l^k$ with the neighboring clusters
\EndFor

\Statex

\Function {InnerSolver}{$x^0$, $j$}
\State Compute the local right-hand side: $Y_l = B_l - \sum^L_{i=1,i\neq l}A_{li}X_i^0$
\State Solving the local splitting $A_{ll}X_l^j=Y_l$ using the parallel GMRES method, such that $X_l^0$ is the initial guess
\State \Return $X_l^j$
\EndFunction

\Statex

\Function {UpdateMinimizer}{$S_l$, $R$, $b$, $k$}
\State Solving the normal equations $(R^k)^TR^k\alpha^k=(R^k)^Tb$ in parallel by $L$ clusters using the parallel CGNR method
\State Compute the local minimizer: $\tilde{X}_l^k=S_l^k\alpha^k$
\State \Return $\tilde{X}_l^k$
\EndFunction
\end{algorithmic}
\label{algo:01}
\end{algorithm}

The  main key points  of the  multisplitting method  to solve  a large
sparse  linear  system  are  given in  Algorithm~\ref{algo:01}.   This
algorithm is based on a two-stage method with a minimization using the
GMRES iterative method as an  inner solver. It is executed in parallel
by  each cluster  of processors.   The matrices  and vectors  with the
subscript  $l$ represent  the local  data for  the cluster  $l$, where
$l\in\{1,\ldots,L\}$. The two-stage solver uses two different parallel
iterative algorithms:  the GMRES method  to solve each splitting  on a
cluster of processors, and the CGNR method executed in parallel by all
clusters  to minimize  the  function error  over  the Krylov  subspace
spanned by  $S$.  The  algorithm requires two  global synchronizations
between the $L$  clusters. The first one is  performed at line~$12$ in
Algorithm~\ref{algo:01}  to exchange  the local  values of  the vector
solution $x$ (i.e. the  minimizer $\tilde{x}$) required to restart the
multisplitting  solver. The  second  one is  needed  to construct  the
matrix $R$ of  the Krylov subspace.  We choose  to perform this latter
synchronization $s$  times in every  outer iteration $k$  (line~$7$ in
Algorithm~\ref{algo:01}). This is a straightforward way to compute the
matrix-matrix    multiplication     $R=AS$.     We    implement    all
synchronizations   by   using   the   MPI   collective   communication
subroutines.


\section{Experiments}

In order  to illustrate  the interest  of our algorithm.   We have  compared our
algorithm  with  the  GMRES  method  which  a very  well  used  method  in  many
situations.  We have chosen to focus on only one problem which is very simple to
implement: a 3 dimension Poisson problem.

\begin{equation}
\left\{
                \begin{array}{ll}
                  \nabla u&=f \mbox{~in~} \omega\\
                  u &=0 \mbox{~on~}  \Gamma=\partial \omega
                \end{array}
              \right.
\end{equation}

After discretization, with a finite  difference scheme, a seven point stencil is
used. It  is well-known that the  spectral radius of  matrices representing such
problems are very close to 1.  Moreover, the larger the number of discretization
points is,  the closer to 1  the spectral radius  is.  Hence, to solve  a matrix
obtained for  a 3D Poisson  problem, the number  of iterations is high.  Using a
preconditioner  it  is   possible  to  reduce  the  number   of  iterations  but
preconditioners are not scalable when using many cores.

\section{Conclusion and perspectives}

Other applications (=> other matrices)\\
Larger experiments\\
Async\\
Overlapping


%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%

\bibliographystyle{plain}
\bibliography{biblio}

\end{document}