sherman-andy@YALE.EDU (andy sherman) (02/14/90)
SCIENTIFIC Computing Associates, Inc. is pleased to announce the availability of PCGPAK2, its new package of subroutines for the iterative solution of large, sparse systems of linear equations. PCGPAK2 offers a choice of solution methods based on a collection of preprocessing, preconditioning, and iterative techniques that includes some of the most robust and efficient methods known. The entire package is written in portable Fortran 77, so it can be easily merged with the large amount of existing scientific and engineering software that depends on solving sparse linear systems. Four basic iterative methods are available in PCGPAK2: --- the conjugate gradient method (CG); --- the generalized minimal residual method (GMRES(k)); --- ORTHOMIN(k); --- the restarted generalized conjugate residual method (GCR(k)). CG is applicable only to symmetric, positive definite systems; the others are general methods designed mainly for systems having nonsymmetric or non-positive-definite symmetric coefficient matrices. PCGPAK2 includes several options that can enhance the performance of the basic iterative methods. Among these are: 1. Incomplete factorization preconditioning 2. Reduced system preprocessing 3. Block iteration PCGPAK2 is applicable to a wide range of engineering and scientific problems that depend on the solution of large sparse systems of linear equations. Examplesof application areas include structural engineering analysis, aerodynamic and hydrodynamic modeling, oil reservoir simulation, ocean acoustics, simulation of VLSI circuit designs and combustion physics. For many problems, PCGPAK2 is substantially faster and uses far less storage than alternative banded or sparse Gaussian elimination methods. For example, on one relatively-small nonsymmetric system of order 3969 arising from a nine-point discretization of an elliptic partial differential equation on the unit square, PCGPAK2 required less than one-fourth of the time and less than one-fifth of the storage required by the band Gaussian elimination routines from LINPACK. For larger two-dimensional and three-dimensional partial differential equations, the savings are far greater. The standard Fortran version of PCGPAK2 will run on essentially any computer. Optimized versions of PCGPAK2 are available for a number of vector machines, including the Cray 1, Cray XMP, Cray YMP, Cray 2, IBM 3090, Convex C-1, Convex C-2, and DEC VAX 9000. More details are available in a posting on sci.math.num-analysis. For further information, contact SCIENTIFIC at SCIENTIFIC Computing Associates, Inc. 246 Church Street, Suite 307 New Haven, CT 06510 Tel.: (203) 777-7442 FAX: (203) 776-4074 Email: sca@yale.edu or yale!sca PCGPAK2 is a registered trademark of SCIENTIFIC Computing Associates, Inc. Computers mentioned may be trademarks of their respective manufacturers.