Two-dimensional Block Partitionings for the Parallel Sparse Cholesky Factorization : the Fan-in Method

نویسندگان

  • B. Dumitrescu
  • D. Trystram
چکیده

This paper presents a discussion on 2D block mappings for the sparse Cholesky factorization on parallel MIMD architectures with distributed memory. It introduces the fan-in algorithm in a general manner and proposes several mapping strategies. The grid mapping with row balancing, inspired from Rothberg's work 21, 22] proved to be more robust than the original fan-out algorithm. Even more eecient is the proportional mapping, as show the experiments on a 32 processors IBM SP1 and on a Cray T3D. Subforest-to-subcube mappings are also considered and give good results on the T3D. Partitionnements par blocs bi-dimensionnels pour la factorisation parall ele creuse de Cholesky : la m ethode fan-in R esum e : Ce rapport etudie les partitionnements par blocs bi-dimensionnels pour la factorisation parall ele creuse de Cholesky sur des machines MIMD a m emoire distribu ee. Nous introduisons l'algorithme fan-in dans un cadre g en eral et etudions dii erentes strat egies de placement. Le placement sur grille avec equilibrage de charge sur les lignes, inspir e des travaux de Rothberg 21, 22], s'av ere plus robuste que l'algorithme fan-out original. Le placement proportionnel est encore plus eecace, comme le montrent les exp erimentations sur un IBM SP1 a 32 processeurs et sur un Cray T3D. Le placement sous-for^ et vers sous-cube est egalement etudi e et donne de bons r esultats sur le Cray T3D. Mots-cl e : factorisation creuse de Cholesky, algorithmes parall eles, communication fan-in, partitionnement blocs 2D, placement proportionnel.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

A Study of the Effects of Ordering, Partitioning and Factorization Algorithms on Distributed Sparse Cholesky Factorization

In this paper, we perform a comprehensive evaluation of ordering, partitioning, and factorization algorithms under a uni ed framework. Previous research in distributed, sparse Cholesky factorization has considered each of the stages in the factorization process | ordering, partitioning and numerical factorization | in isolation. However, due to the strong dependencies between the stages, it is ...

متن کامل

Task Scheduling using Block Dependency DAG of Block-Oriented Sparse Cholesky Factorizationy

The block-oriented sparse Cholesky factorization decomposes a sparse matrix into rectangular sub-blocks, and handles each block as a computational unit in order to increase data reuse in a hierarchical memory system. As well, the factorization method increases the degree of concurrency with the reduction of communication volumes so that it performs more eeciently on a distributed-memory multipr...

متن کامل

Scalable Parallel Algorithms for Solving Sparse Systems of Linear Equations∗

We have developed a highly parallel sparse Cholesky factorization algorithm that substantially improves the state of the art in parallel direct solution of sparse linear systems—both in terms of scalability and overall performance. It is a well known fact that dense matrix factorization scales well and can be implemented efficiently on parallel computers. However, it had been a challenge to dev...

متن کامل

Efficient Parallel Solutions of Large Sparse Spd Systems on Distributed-memory Multiprocessors

We consider several issues involved in the solution of sparse symmetric positive deenite systems by multifrontal method on distributed-memory multiprocessors. First, we present a new algorithm for computing the partial factorization of a frontal matrix on a subset of processors which signiicantly improves the performance of a distributed multifrontal algorithm previously designed. Second, new p...

متن کامل

Comparative Analysis of High Performance Solvers for 3D Elliptic Problems

The presented comparative analysis concerns two iterative solvers for 3D linear boundary value problems of elliptic type. After applying the Finite Difference Method (FDM) or the Finite Element Method (FEM) discretization a system of linear algebraic equations has to be solved, where the stiffness matrix is large, sparse and symmetric positive definite. It is well known that the preconditioned ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1997