LaBRI LaBRI

    Parallel Sparse matriX package

    SopalinScotch



    PaStiX is a Parallel Direct Solver for Very Large Sparse SPD Systems


    This work is supported by the French Commissariat à l'Energie Atomique CEA/CESTA and by the GDR ARP (iHPerf group) of the CNRS.

    Solving large sparse symmetric positive definite systems Ax=b of linear equations is a crucial and time-consuming step, arising in many scientific and engineering applications.

    This work is a research scope of the new INRIA project (UR Futurs).
    PaStiX is a scientific library that provides a high performance solver for very large sparse linear systems based on direct and ILU(k) iterative methods.
    Many factorization algorithms are implemented with simple or double precision (real or complex): LLt (Cholesky), LDLt (Crout) and LU with static pivoting (for non symetric matrices but with symetric structures).
    The library PaStiX uses the graph partitioning and sparse matrix block ordering package SCOTCH.
    PaStiX is based on efficient static scheduling and memory management to solve problems with more than 10 millions unknowns.
    An available version of PaStiX is currently developped.


    Portability:

    PaStiX can be used on any parallel computer or network of workstations equipped with MPI and BLAS libraries, and C language compilers. PaStiX functions can be called from both C and Fortran 90 programs. It has been extensively tested on IBM SP2.

    Related Software:

    PaStiX uses SCOTCH as its default ordering libraries. For the latest version of SCOTCH and to find more about it refer to the SCOTCH site .

    PaStiX uses the standard BLAS, LAPACK, and MPI library calls for its functionality. Tuned versions of these libraries are recommended to get good performance out of PaStiX. Public domain vanilla implementations of BLAS and LAPACK routines are available on the web. A public domain implementation of MPI standard is available at MPICH site .

    Concurrent Software:

    PSPASES(WSSMP) software based on a multifrontal approach with a 2D block mapping.
    MUMPS.
    SuperLU.


    Page Creator: Pierre Ramet
    Last modified: Thu Sep 16 10:39:00 CEST 1999