pa1081@sdcc13.ucsd.edu (pa1081) (10/18/89)
Does anyone out there know the uucp address of Jack Dongarra of the Argonne Institute? (the guy who does the LINPACK stuff). I need to get the all FORTRAN LINPACK source to test our machine here. Please reply to: pa1081@sdcc13.ucsd.edu
hirchert@uxe.cso.uiuc.edu (10/20/89)
Argonne National Laboratory distributes the software it has developed via a mail based server. Send messages to netlib@mcs.anl.gov (Internet) or research!netlib (UUCP). The message should contain one or more commands of one of the following forms: send index send index from <library> send <routine> from <library> find <keywords> Note that the above does not provide a way to request who libraries. I believe they still prefer to do that by sending tapes rather than over the network. (The above is based on an article I read a couple of years ago. I tried it out then, but I haven't tested it recently to verify that it still works.)
hirchert@uxe.cso.uiuc.edu (10/21/89)
I tested out what I posted earlier. Some of the details have changed.
It's all explained in the reply I received:
===== How to use netlib =====
This file is the reply you'll get to:
mail netlib@research.att.com
send index
Here are examples of the various kinds of requests.
* get the full index for a library
send index from eispack
* get a particular routine and all it depends on
send dgeco from linpack
* get just the one routine, not subsidiaries
send only dgeco from linpack
* get dependency tree, but excluding a subtree
send dgeco but not dgefa from linpack
* just tell how large a reply would be, don't actually send the file
send list of dgeco from linpack
* search for somebody in the SIAM membership list:
who is gene golub
* keyword search for netlib software
find cubic spline
* keyword search in the approximation catalog
find schumaker from approximation
The Internet address "netlib@research.att.com" refers to a gateway
machine at AT&T Bell Labs in Murray Hill, New Jersey. This address
should be understood on all the major networks. For systems having
only uucp connections, use the address uunet!research!netlib. In this
case, someone will be paying for long distance 1200 baud phone calls,
so keep your requests to a reasonable size! An excellent guide to the
mysteries of networks and address syntax is: Donnalyn Frey and Rick
Adams (1989) "!%@:: A Directory of Electronic Mail Addressing and
Networks", O'Reilly & Associates, Inc, 632 Petaluma Ave, Sebastopol CA
95472. Background about netlib is in Jack J. Dongarra and Eric Grosse,
Distribution of Mathematical Software Via Electronic Mail, Comm. ACM
(1987) 30, 403--407.
Bugs reports, comments, and annual lists of recipients will be
forwarded to the code authors when possible. Many of these codes are
designed for use by professional numerical analysts who are capable of
checking for themselves whether an algorithm is suitable for their
needs. One routine can be superb and the next awful. So be careful!
An inventory list is given below and in the indices for the individual
libraries. If you know exactly what you're looking for, these guides
may be enough. An interactive system called "walk" provides a more
systematic list (not limited to netlib) but at present covers only
approximation. Volunteers from other fields are needed. The reference
is Eric Grosse, "A Catalog ...", in Algorithms for Approximation",
Mason and Cox (eds.), Chapman and Hall, 1989. Dialup (at 1200 baud)
201-582-1238 or telnet to research.att.com and login as walk; no
password is required.
-------quick summary of contents---------
a - approximation algorithms
alliant - set of programs collected from Alliant users
amos - special functions by D. Amos. = toms/644
apollo - set of programs collected from Apollo users
benchmark - various benchmark programs and a summary of timings
bihar - Bjorstad's biharmonic solver
blas3 - matrix * matrix BLAS
bmp - Brent's multiple precision package
c - another "misc" library, for software written in C
cascade - analysis and design of linear control systems
cheney-kincaid - programs from the text Numerical Mathematics and Computing.
conformal - Schwarz-Christoffel codes by Trefethen; Bjorstad+Grosse
core - machine constants, vector and matrix * vector BLAS
dierckx - spline fitting
domino - communication and scheduling of multiple tasks; Univ. Maryland
eispack - matrix eigenvalues and vectors
elefunt - Cody and Waite's tests for elementary functions
errata - corrections to numerical books
fishpack - separable elliptic PDEs; Swarztrauber and Sweet
fitpack - Cline's splines under tension
fftpack - Swarztrauber's Fourier transforms
fmm - software from the book by Forsythe, Malcolm, and Moler
fn - Fullerton's special functions
uncon/data - optimization test problems
gcv - Generalized Cross Validation
go - "golden oldies" gaussq, zeroin, lowess, ...
graphics - ray-tracing
harwell - MA28 sparse linear system
hompack - nonlinear equations by homotopy method
itpack - iterative linear system solution by Young and Kincaid
jakef - automatic differentiation of Fortran subroutines
lanczos - Cullum and Willoughby's Lanczos programs
laso - Scott's Lanczos program for eigenvalues of sparse matrices
linpack - gaussian elimination, QR, SVD by Dongarra, Bunch, Moler, Stewart
lp - linear programming
machines - short descriptions of various computers
microscope - Alfeld and Harris' system for discontinuity checking
minpack - nonlinear equations and least squares by More, Garbow, Hillstrom
misc - everything else
na-digest - archive of mailings to NA distribution list
napack - numerical algebra programs
ode - ordinary differential equations
odepack - ordinary differential equations from Hindmarsh
paranoia - Kahan's floating point test
parmacs - parallel programmming macros
pchip - hermite cubics Fritsch+Carlson
pltmg - Bank's multigrid code; too large for ordinary mail
polyhedra - Hume's database of geometric solids
port - the public subset of PORT library
pppack - subroutines from de Boor's Practical Guide to Splines
quadpack - univariate quadrature by Piessens, de Donker, Kahaner
sched - environment for portable parallel algorithms in a Fortran setting.
slap - Seager + Greenbaum, iterative methods for symmetric and unsymmetric
slatec - machine constants and error handling package from the Slatec library
sparse - Kundert + Sangiovanni-Vincentelli, C sparse linear algebra
sparse-blas - BLAS by indirection
sparspak - George + Liu, sparse linear algebra core
specfun - transportable special functions
toeplitz - linear systems in Toeplitz or circulant form by Garbow
toms - Collected Algorithms of the ACM
typesetting - typesetting macros and preprocessors
vanhuffel - total least squares, partial SVD by Van Hufell
voronoi - Voronoi diagrams and Delaunay triangulations
y12m - sparse linear system (Aarhus)
--------a bit more detail--------
The first few libraries here are widely regarded as being of high quality.
The likelihood of your encountering a bug is relatively small; if you do,
we certainly want to hear about it!
CORE Machine constants (i1mach,r1mach,d1mach), blas (level 1 and 2)
EISPACK A collection of Fortran subroutines that compute the eigenvalues
and eigenvectors of nine classes of matrices. The package can
determine the eigensystems of complex general, complex Hermitian,
real general, real symmetric, real symmetric band, real symmetric
tridiagonal, special real tridiagonal, generalized real, and
generalized real symmetric matrices. In addition, there are two
routines which use the singular value decomposition to solve
certain least squares problems.
Developed by the NATS Project at Argonne National Laboratory.
(d.p. refer to eispack, s.p. refer to seispack)
FFTPACK A package of Fortran subprograms for the Fast Fourier
Transform of periodic and other symmetric sequences
This package consists of programs which perform Fast Fourier
Transforms for both complex and real periodic sequences and
certian other symmetric sequences.
Developed by Paul Swarztrauber, at NCAR.
FISHPACK A package of Fortran subprograms providing finite difference
approximations for elliptic boundary value problems.
Developed by Paul Swarztrauber and Roland Sweet.
FNLIB Wayne Fullerton's special function library. (single and double)
GO Golden Oldies: routines that have been widely used,
but aren't available through the standard libraries.
Nominations welcome!
HARWELL Sparse matrix routine MA28 from the Harwell library. from Iain Duff
LINPACK A collection of Fortran subroutines that analyze and solve linear
equations and linear least squares problems. The package solves
linear systems whose matrices are general, banded, symmetric
indefinite, symmetric positive definite, triangular, and tridiagonal
square. In addition, the package computes the QR and singular value
decompositions of rectangular matrices and applies them to least
squares problems.
Developed by Jack Dongarra, Jim Bunch, Cleve Moler and Pete Stewart.
(all precisions contained here)
PPPACK Subroutines from: Carl de Boor, A Practical Guide to Splines,
Springer Verlag. This is an old version, from around the time the book
was published. We will install a newer version as soon as we can.
TOMS Collected algorithms of the ACM. When requesting a specific
item, please refer to the Algorithm number.
----------------
In contrast to the above libraries, the following are collections of codes
from a variety of sources. Most are excellent, but you should exercise
caution. We include research codes that we haven't tested and codes
that may not be state-of-the-art but useful for comparisons.
The following list is chronological, not by merit:
MISC Contains various pieces of software collected over time and:
the source code for the netlib processor itself;
the paper describing netlib and its implementation;
the abstracts list maintained by Richard Bartels.
FMM Routines from the book Computer Methods for Mathematical
Computations, by Forsythe, Malcolm, and Moler.
Developed by George Forsythe, Mike Malcolm, and Cleve Moler.
(d.p. refer to fmm, s.p. refer to sfmm)
QUADPACK A package for numerical computation of definite univariate integrals.
Developed by Piessens, Robert(Appl. Math. and Progr. Div.- K.U.Leuven)
de Donker, Elise(Appl. Math. and Progr. Div.- K.U.Leuven
Kahaner, David(National Bureau of Standards) (slatec version)
TOEPLITZ A package of Fortran subprograms for the solution of systems
of linear equations with coefficient matrices of Toeplitz or
circulant form, and for orthogonal factorization of column-
circulant matrices.
Developed by Burt Garbow at Argonne National Laboratory,
as a culmination of Soviet-American collaborative effort.
(d.p. refer to toeplitz, s.p. refer to stoeplitz)
ITPACK Iterative Linear System Solver based on a number of methods:
Jacobi method, SOR, SSOR with conjugate gradient acceleration
or with Chebyshev (semi-iteration - SI) acceleration.
Developed by Young and Kincaid and the group at U of Texas.
BIHAR Biharmonic solver in rectangular geometry and polar coordinates.
These routines were obtained from Petter Bjorstad,
Veritas Research, Oslo Norway in July 1984.
LANCZOS procedures computing a few eigenvalues/eigenvectors of a large (sparse)
symmetric matrix. Jane Cullum and Ralph Willoughby, IBM Yorktown.
LASO A competing Lanczos package. David Scott.
CONFORMAL contains routines to solve the "parameter problem" associated
with the Schwarz-Christoffel mapping. Includes:
SCPACK (polygons with straight sides) from Nick Trefethen.
CAP (circular arc polygons) from Petter Bjorstad and Eric Grosse.
FITPACK A package for splines under tension. (an early version)
For a current copy and for other routines, contact:
Alan Kaylor Cline, 8603 Altus Cove, Austin, Texas 78759, USA
BENCHMARK contains benchmark programs and the table of Linpack timings.
MACHINES contains information on high performance computers that
are or soon to be made available
MINPACK A package of Fortran programs for the solution of systems of
nonlinear equations and nonlinear least squares problems.
Five algorithmic paths each include a core subroutine and an
easy-to-use driver. The algorithms proceed either from an analytic
specification of the Jacobian matrix or directly from the problem
functions. The paths include facilities for systems of equations
with a banded Jacobian matrix, for least squares problems with a
large amount of data, and for checking the consistency of the
Jacobian matrix with the functions.
Developed by Jorge More', Burt Garbow, and Ken Hillstrom at
Argonne National Laboratory.
(d.p. refer to minpack, s.p. refer to sminpack)
PORT The public subset of the PORT library. Includes the latest version
of Gay's NL2SOL nonlinear least squares. The rest of the PORT3
library is available by license from AT&T.
Y12M calculation of the solution of systems of linear systems of
linear algebra equations whose matrices are large and sparse.
authors: Zahari Zlatev, Jerzy Wasniewski and Kjeld Schaumburg
PCHIP is a fortran package for piecewise cubic hermite inter-
polation of data. It features software to produce a monotone and
"visually pleasing" interpolant to monotone data.
Fred N. Fritsch, Lawrence Livermore National Laboratory
LP Linear Programming - At present, this consists of one subdirectory,
data: a set of test problems in MPS format, maintained by David Gay.
For more information, try a request of the form
send index for lp/data
ODE various initial and boundary value ordinary differential equation
solvers: colsys, dverk, rkf45, ode
A subset of these in single precision is in the library sode.
ODEPACK The ODE package from Hindmarch and others.
This is the double precision verison; to get sp refer to sodepack.
Alan Hindmarch, Lawrence Livermore National Laboratory
ELEFUNT is a collection of transportable Fortran programs for testing
the elementary function programs provided with Fortran compilers. The
programs are described in detail in the book "Software Manual for the
Elementary Functions" by W. J. Cody and W. Waite, Prentice Hall, 1980.
SPECFUN is an incomplete, but growing, collection of transportable
Fortran programs for special functions, and of accompanying test
programs similar in concept to those in ELEFUNT.
W.J. Cody, Argonne National Laboratory
PARANOIA is a rather large program, devised by Prof. Kahan of Berkeley,
to explore the floating point system on your computer.
SLATEC library DoE policy apparently prohibits us from distributing this.
Contact the National Energy Software Center or your congressman.
HOMPACK is a suite of FORTRAN 77 subroutines for solving nonlinear systems
of equations by homotopy methods. There are subroutines for fixed
point, zero finding, and general homotopy curve tracking problems,
utilizing both dense and sparse Jacobian matrices, and implementing
three different algorithms: ODE-based, normal flow, and augmented
Jacobian.
DOMINO is a set of C-language routines with a short assembly language
interface that allows multiple tasks to communicate and schedules
local tasks for execution. These tasks may be on a single processor
or spread among multiple processors connected by a message-passing
network. (O'Leary, Stewart, Van de Geijn, University of Maryland)
GCV software for Generalized Cross Validation from: O'Sullivan,
Woltring (univariate spline smoothing), Bates, Lindstrom,
Wahba and Yandell (multivariate thin plate spline smoothing
and ridge regression), Gu (multiple smoothing parameters).
Cheney-Kincaid programs from: Ward Cheney & David Kincaid, Numerical
Mathematics and Computing.
POLYHEDRA a database of angles, vertex locations, and so on for over a
hundred geometric solids, compiled by Andrew Hume.
GRAPHICS presently just contains some C routines for testing ray-tracing
A approximation algorithms (almost empty, but soon to grow)
lowess: multivariate smoothing of scattered data; Cleveland+Devlin+Grosse
Apollo A set of programs collected from Apollo users.
Alliant A set of programs collected from Alliant users.
parmacs - parallel programmming macros for monitors and send/receive
Rusty Lusk, Argonne National Laboratory, June 5, 1987 (lusk@anl-mcs.arpa)
sched - The Schedule Package is an environment for the transportable
implementation of parallel algorithms in a Fortran setting.
Jack Dongarra and Dan Sorensen, Argonne National Laboratory,
June 5, 1987 (dongarra@anl-mcs.arpa sorensen@anl-mcs.arpa)
NAPACK A collection of Fortran subroutines to solve linear systems,
to estimate the condition number or the norm of a matrix,
to compute determinants, to multiply a matrix by a vector,
to invert a matrix, to solve least squares problems, to perform
unconstrained minimization, to compute eigenvalues, eigenvectors,
the singular value decomposition, or the QR decomposition.
The package has special routines for general, band, symmetric,
indefinite, tridiagonal, upper Hessenberg, and circulant matrices.
Code author: Bill Hager, Mathematics Department, Penn State
University, University Park, PA 16802, e-mail: hager@psuvax1.bitnet
or hager@psuvax1.psu.edu. Related book: Applied Numerical Linear
Algebra, Prentice-Hall, Englewood Cliffs, New Jersey.
Book scheduled to appear in December, 1987.
SPARSPAK Subroutines from the book "Computer Solution of Large Sparse
Positive Definite Systems" by George and Liu, Prentice Hall 1981.
VANHUFFEL
The TLS problem assumes an overdetermined set of linear equations
AX = B, where both the data matrix A as well as the observation
matrix B are inaccurate.
The subroutine PTLS solves the Total Least Squares (TLS) problem by
using a Partial Singular Value Decomposition (PSVD), hereby improving
considerably the computational efficiency with respect to the classi-
cal TLS algorithm.
Sabine VAN HUFFEL
ESAT Laboratory, KU Leuven.
Kardinaal Mercierlaan 94, 3030 Heverlee, Belgium
DIERCKX
A package of spline fitting routines for various kinds of data and
geometries. Written by: Professor Paul Dierckx, Dept. Computer Science,
K. U. Leuven, Celestijnenlaan 200A, B-3030 Heverlee, Belgium.
VORONOI
Algorithms for Voronoi regions and Delaunay triangulations. Currently
contains Fortune's 2d sweepline method.
SPARSE A library of subroutines written in C that solve large sparse
systems of linear equations using LU factorization. The
package is able to handle arbitrary real and complex square
matrix equations. Besides being able to solve linear systems,
it is solves transposed systems, find determinants, multiplies
a vector by a matrix, and estimate errors due to
ill-conditioning in the system of equations and instability in
the computations. Sparse does not require or assume symmetry
and is able to perform numerical pivoting (either diagonal or
complete) to avoid unnecessary error in the solution. Sparse
also has an optional interface that allow it to be called from
FORTRAN programs.
Ken Kundert, Alberto Sangiovanni-Vincentelli. (sparse@ic.berkeley.edu)
SLAP This is the official release version 2.0 of the Sparse Linear
Algebra Package: a SLAP for the Masses! It contains "core"
routines for the iterative solution symmetric and non-symmetric
positive definite and positive semi-definite linear systems.
Included in this package are core routines to do Iterative
Refinement iteration, Preconditioned Conjugate Gradient
iteration, Preconditioned Conjugate Gradient iteration on the
Normal Equations, Preconditioned BiConjugate Gradient iteration,
Preconditioned BiConjugate Gradient Squared iteration, Orthomin
iteration and Generalized Minimum Residual iteration. Core
routines require the user to supply "MATVEC" (Matrix Vector
Multiply) and "MSOLVE" (Preconditiong) routines. This allows the
core routines to be written in a way that makes them independent
of the matrix data structure. For each core routine there are
several drivers and support routines that allow the user to
utilize Diagonal Scaling and Incomplete Cholesky/Incomplete LU
factorization as preconditioners with no coding. The price for
this convience is that one must use the a specific matrix data
structure: SLAP Column or SLAP Triad format.
Written by Mark K. Seager & Anne Greenbaum
UNCON/DATA test problems: unconstrained optimization, nonlinear least squares.
Problems from More, Garbow, and Hillstrom; Fraley, matrix square
root; Hanson, Salane; McKeown; De Villiers and Glasser;
Dennis, Gay, and Vu. Collected by Chris Fraley.
JAKEF is a precompiler that analyses a given Fortran77 source code for
the evaluation of a scalar or vector function and then generates an
expanded Fortran subroutine that simultaneously evaluates the gradient
or Jacobian respectively. For scalar functions the ratio between the
run-time of the resulting gradient routine and that of the original
evaluation routine is never greater than a fixed bound of about five.
The storage requirement may be considerable as it is also proportional
to the run-time of the original routine. Since no differencing is done
the partial derivative values obtained are exact up to round-off errors. A. Griewank, Argonne National Laboratory, griewank@mcs.anl.gov, 12/1/88.
sparse-blas an extension to the set of Basic Linear Algebra Subprograms.
The extension is targeted at sparse vector operations, with the goal of
providing efficient, but portable, implementations of algorithms for high
performance computers.
convex!dodson@anl-mcs.ARPA Mon Aug 31 19:53:21 1987 (Dave Dodson)