1 \documentclass{article}
2 \usepackage[utf8]{inputenc}
3 \usepackage{amsfonts,amssymb}
7 \title{A scalable multisplitting algorithm for solving large sparse linear systems}
12 \author{Raphaël Couturier \and Lilia Ziane Khodja}
17 %%%%%%%%%%%%%%%%%%%%%%%%
18 %%%%%%%%%%%%%%%%%%%%%%%%
22 In this paper we revist the krylov multisplitting algorithm presented in
23 \cite{huang1993krylov} which uses a scalar method to minimize the krylov
24 iterations computed by a multisplitting algorithm. Our new algorithm is based on
25 a parallel multisplitting algorithm with few blocks of large size using a
26 parallel GMRES method inside each block and on a parallel krylov minimization in
27 order to improve the convergence. Some large scale experiments with a 3D Poisson
28 problem are presented. They show the obtained improvements compared to a
29 classical GMRES both in terms of number of iterations and execution times.
33 %%%%%%%%%%%%%%%%%%%%%%%%
34 %%%%%%%%%%%%%%%%%%%%%%%%
37 \section{Introduction}
39 Iterative methods are used to solve large sparse linear systems of equations of
40 the form $Ax=b$ because they are easier to parallelize than direct ones. Many
41 iterative methods have been proposed and adapted by many researchers. When
42 solving large linear systems with many cores, iterative methods often suffer
43 from scalability problems. This is due to their need for collective
44 communications to perform matrix-vector products and reduction operations.
45 Preconditionners can be used in order to increase the convergence of iterative
46 solvers. However, most of the good preconditionners are not sclalable when
47 thousands of cores are used.
51 On ne peut pas parler de tout...\\
56 %%%%%%%%%%%%%%%%%%%%%%%
58 %%%%%%%%%%%%%%%%%%%%%%%
59 The key idea of the multisplitting method for solving a large system of linear equations
60 $Ax=b$ consists in partitioning the matrix $A$ in $L$ several ways
62 A = M_l - N_l,~l\in\{1,\ldots,L\},
65 where $M_l$ is a nonsingular matrix, and then solving the linear system by the iterative method
67 x^{k+1}=\displaystyle\sum^L_{l=1} E_l M^{-1}_l (N_l x^k + b),~k=1,2,3,\ldots
70 where $E_l$ is a non-negative and diagonal weighting matrix such that $\sum^L_{l=1}E_l=I$ ($I$ is the identity matrix).
71 Thus the convergence of such a method is dependent on the condition
73 \rho(\displaystyle\sum^L_{l=1}E_l M^{-1}_l N_l)<1.
77 The advantage of the multisplitting method is that at each iteration $k$ there are $L$ different linear
80 y_l=M^{-1}_l N_l x_l^{k-1} + M^{-1}_l b,~l\in\{1,\ldots,L\},
83 to be solved independently by a direct or an iterative method, where $y_l$ is the solution of the local system.
84 A multisplitting method using an iterative method for solving the $L$ linear systems is called an inner-outer
85 iterative method or a two-stage method. The solution of the global linear system at the iteration $k$ is computed
88 x^k = \displaystyle\sum^L_{l=1} E_l y_l,
91 In the case where the diagonal weighting matrices $E_l$ have only zero and one factors (i.e. $y_l$ are disjoint vectors),
92 the multisplitting method is non-overlapping and corresponds to the block Jacobi method.
93 %%%%%%%%%%%%%%%%%%%%%%%
95 %%%%%%%%%%%%%%%%%%%%%%%
97 \section{Related works}
100 A general framework for studying parallel multisplitting has been presented in
101 \cite{o1985multi} by O'Leary and White. Convergence conditions are given for the
102 most general case. Many authors improved multisplitting algorithms by proposing
103 for example a asynchronous version \cite{bru1995parallel} and convergence
104 condition \cite{bai1999block,bahi2000asynchronous} in this case or other
105 two-stage algorithms~\cite{frommer1992h,bru1995parallel}
107 In \cite{huang1993krylov}, the authors proposed a parallel multisplitting
108 algorithm in which all the tasks except one are devoted to solve a sub-block of
109 the splitting and to send their local solution to the first task which is in
110 charge to combine the vectors at each iteration. These vectors form a Krylov
111 basis for which the first tasks minimize the error function over the basis to
112 increase the convergence, then the other tasks receive the update solution until
113 convergence of the global system.
117 In \cite{couturier2008gremlins}, the authors proposed practical implementations
118 of multisplitting algorithms that take benefit from multisplitting algorithms to
119 solve large scale linear systems. Inner solvers could be based on scalar direct
120 method with the LU method or scalar iterative one with GMRES.
125 %%%%%%%%%%%%%%%%%%%%%%%%
126 %%%%%%%%%%%%%%%%%%%%%%%%
129 \section{A two-stage method with a minimization}
130 Let $Ax=b$ be a given sparse and large linear system of $n$ equations
131 to solve in parallel on $L$ clusters, physically adjacent or geographically
132 distant, where $A\in\mathbb{R}^{n\times n}$ is a square and nonsingular
133 matrix, $x\in\mathbb{R}^{n}$ is the solution vector and $b\in\mathbb{R}^{n}$
134 is the right-hand side vector. The multisplitting of this linear system
135 is defined as follows:
139 A & = & [A_{1}, \ldots, A_{L}]\\
140 x & = & [X_{1}, \ldots, X_{L}]\\
141 b & = & [B_{1}, \ldots, B_{L}]
146 where for all $l\in\{1,\ldots,L\}$ $A_l$ is a rectangular block of size $n_l\times n$
147 and $X_l$ and $B_l$ are sub-vectors of size $n_l$, such that $\sum_ln_l=n$. In this
148 case, we use a row-by-row splitting without overlapping in such a way that successive
149 rows of the sparse matrix $A$ and both vectors $x$ and $b$ are assigned to a cluster.
150 So, the multisplitting format of the linear system is defined as follows:
152 \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,
155 where $A_{li}$ is a block of size $n_l\times n_i$ of the rectangular matrix $A_l$, $X_i\neq X_l$
156 is a sub-vector of size $n_i$ of the solution vector $x$ and $\sum_{i<l}n_i+\sum_{i>l}n_i+n_l=n$,
157 for all $i\in\{1,\ldots,l-1,l+1,\ldots,L\}$. Therefore, each cluster $l$ is in charge of solving
158 the following spare sub-linear system:
162 A_{ll}X_l = Y_l \mbox{,~such that}\\
163 Y_l = B_l - \displaystyle\sum_{i=1,i\neq l}^{L}A_{li}X_i,
168 where the sub-vectors $X_i$ define the data dependencies between the cluster $l$ and other clusters.
172 %%%%%%%%%%%%%%%%%%%%%%%%
173 %%%%%%%%%%%%%%%%%%%%%%%%
175 \bibliographystyle{plain}
176 \bibliography{biblio}