P_SPARSLIB is a library of portable FORTRAN routines for parallel sparse matrix computations. The current thrust of the library is in iterative solution techniques. In this paper we present the "accelerators" part of the library, which comprise the best known of the Krylov subspace techniques. The iterative solution module is implemented in reverse communication mode in order to allow any preconditioner to be combined with the package. In addition, this mechanism allows us to ensure portability, since the communication calls required in the iterative solution process are hidden in the dot-product, the matrix-vector product and preconditioning operations.
|Original language||English (US)|
|Number of pages||15|
|Journal||Applied Numerical Mathematics|
|State||Published - Dec 1995|
Bibliographical noteFunding Information:
where the coefficient matrix A can be very large and sparse. For such systems, iterative methods that exploit sparsity are often used because of their advantage in storage and computational requirements. Iterative methods are especially attractive for the new class of very large three-dimensional problems that are emerging because of the dramatic improvement in computational power provided by parallel processing. In a multiprocessing environment, iterative methods have the further advantage that they are far simpler to implement than sparse direct methods. Because of the gain in popularity of these methods and their importance for multiprocessing environments, the development of a parallel sparse *Work supported in part by ARPA under grant number NIST 60NANB2D1272, in part by NSF under grant number NSF/CCR-9214116, and in part by AHPCRC (University of Minnesota) under Army Research Office grant number DAAL03-89-C-0038. * Corresponding author.