Program name:       LINPACK LIBRARY

Author name:        Argonne National Laboratory

Address:            None given

Telephone Number:   None given 

Suggested Donation: None 

Program Description: 

The Linear Equation Package (LINPACK) is a set of routines which solves 
systems of linear equations and related problems.  This package was 
developed by the Applied Mathematics Division of Argonne National Laboratory 
through funding by the National Science Foundation and Department of Energy.  
The original package was developed and tested on large machines over the 
years 1976 through 1979.  The source included on these diskettes was taken 
from a distribution disk provided by International Mathematics and 
Statistical Libraries, Inc.  (IMSL). 

An excerpt from the LINPACK USER'S GUIDE Introduction: "Many of the 
subroutines deal with square coefficient matrices, where there are as many 
equations as unknowns.  Some of the subroutines process rectangular 
coefficient matrices, where the system may be over- or under-determined.  
Such systems are frequently encountered in least squares problems and other 
statistical calculations.  Different subroutines are intended to take 
advantage of different special properties of the matrices and thereby save 
computer time and storage."

The total package will handle equations in single, double, complex and 
complex double precision arithmetic.  For each data type, there are two 
diskettes: test driver and linpack source.  This diskette contains the 
single precision version of the LINPACK library.
   
The driver disk contains a FORTRAN source and link file list (.LNK) for each 
test set.  The source for these drivers can be used to study the calling 
sequences for the various routines.  Since each driver has been successfully 
tested under MS-DOS using Microsoft FORTRAN (V3.31), it is not necessary for 
you to rebuild these executable modules.  If you should, however, decide to 
alter any routines, it would be best to test them using these drivers.  The 
.LNK files are used with the BUILDS.BAT file, so examine that file if you 
wish to rebuild the test executables. 

Each of the following test files was successfully constructed and executed:  
SCH, SGT, SP, SQR, SS, SSV, ST, SG.  Each driver evaluates those single 
precision routines whose names begin with the same characters as the name of 
the driver.  The drivers SUD and SEX test the update and exchange routines 
respectively.

The source disk contains the LINPACK routines along with several of the BLAS 
routines.  Each source file contains some information about usage, but a 
fair amount of knowledge about the different techniques for solving systems 
of equations is required.  If you are unsure where to begin, try 
implementing the SG* (single precision general) routines first since they do 
not utilize any special storage techniques.

I offer the following list of references as a sample with which I am most 
familiar; not necessarily the best, just the ones I know :

LINPACK USER'S GUIDE by Dongarra, Moler, Bunch and Stewart, published by 
    Society for Industrial and Applied Mathematics 
        This is "THE BOOK" written by the authors of the LINPACK system.

NUMERICAL METHODS by Ake Bjorck, published by Prentice-Hall 
        This book has everything, but it is definitely not easy reading. Any 
        university level book on the subject of numerical methods should 
        suffice.
         
THE THEORY OF MATRICES IN NUMERICAL ANALYSIS by Alston S. Householder, 
    published by Dover 
        This is an older, even more theoretical, work than the previous.

If you have any doubts about using these routines, you should at least have 
the LINPACK USER'S GUIDE.  Don't limit yourself, however, to these books as 
there are a number of new, very readable textbooks which can be found at 
most technical or university book stores.

See the "index" directory for an idea of the contents of this disk. If you 
have any questions or criticisms of this package, please write me (Jeffrey 
Fried) through the PC-SIG offices.

No system requirements are given.