PDA

View Full Version : Introduction to NETLIB



Ton Van Den Bogert
05-28-1993, 05:47 AM
Dear Biomch-L subscribers,

It has been a long time since NETLIB was mentioned on this list.
Yesterday, I found this posting on Usenet (sci.math.num-analysis)
with a good introduction to this collection of software.

The NETLIB software, as far as I have used it, is of excellent
quality. Documentation is good and, most importantly, everything
is free. The software comes mostly as subroutines for basic
mathematical operations, not as application programs for the end
user. So, you need some programming skills and mathematical
background to make good use of it.

I have made extensive use of the routines LMDIF/LMDIF1 (from the
MINPACK library) to solve non-linear least squares problems, and
have been very satisfied with the performance.

-- Ton van den Bogert
Human Performance Laboratory
University of Calgary, Canada
++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++++++++++++++++
>From: fc03@ns1.cc.lehigh.edu (Frederick W. Chapman)
>Newsgroups: sci.math.num-analysis
>Subject: Introduction to NETLIB (Revised July 1992)
>Message-ID:
>Date: 27 May 93 15:19:13 GMT
>Organization: Lehigh University



----------------------
Introduction to NETLIB
----------------------


Frederick W. Chapman
(fc03@Lehigh.Edu)
Senior User Consultant
Lehigh University Computing Center


July 7, 1992




1. General Description
----------------------

NETLIB is a network-based facility for the automated
distribution of the source code for public domain scientific
software. Most of this software is written in FORTRAN, but some
software is available in C, C++, and PASCAL. Software is
available for a wide variety of applications. The following list
-- which is far from exhaustive -- should give the reader an idea
of the scope of the NETLIB collection.


NUMERICAL METHODS:

* Linear Algebra (eigenvalue and eigenvector computations,
matrix factorizations, least squares)
* Sparse Matrix Calculations
* Numerical Optimization
* Spline Interpolation
* Ordinary and Partial Differential Equations
* Fast Fourier Transforms
* Special Functions

WELL-KNOWN SOFTWARE:

* Collected algorithms of the ACM "Transactions on
Mathematical Software" (TOMS)
* LAPACK, LINPACK, EISPACK, and BLAS subroutine libraries
* AMS TeX and SIAM typesetting macros
* MATLAB applications packages

SOFTWARE FOR ADVANCED COMPUTER ARCHITECTURES:

* Libraries for supporting parallel computation
* Libraries for computation on vector-processor machines

MISCELLANEOUS:

* Packages supporting multiple-precision floating-point arith-
metic (e.g., Brent's MP; Smith's FM, from TOMS algorithms)
* Benchmark programs for comparing computing platforms
* Collections of problems for testing numerical software
* Programming aids such as single-precision-to-double-
precision and FORTRAN-to-C conversion utilities
* Companion software to various textbooks (e.g., Cheney and
Kincaid; Forsythe, Malcolm, and Moler)
* Bibliographies


2. NETLIB Mail Servers
----------------------

Information (such as NETLIB usage instructions) and software
(such a routine from LAPACK or an algorithm from the ACM TOMS)
can be obtained by sending requests -- via electronic mail -- to
a NETLIB mail server at one of the following Internet addresses:

netlib@research.att.com (AT&T Bell Labs, New Jersey, USA)
netlib@ornl.gov (Oak Ridge Nat. Lab, Tenn., USA)
netlib@unix.hensa.ac.uk (Univ. of Kent, UK)
netlib@nac.no (Oslo, Norway)
netlib@cs.uow.edu.au (U. of Wollongong, NSW, Australia)

Upon receipt of an appropriately worded request, the NETLIB
mail server responds by sending the information or software
requested to the electronic mail address of the requestor.
Depending on your location, the amount of network traffic, and
other factors, you may actually receive a reply within minutes of
submitting your request! Note that the software available may
differ slightly from one NETLIB site to another.


3. Accessing NETLIB via Electronic Mail
---------------------------------------

To obtain more detailed instructions on the use of NETLIB,
as well as an index of the software currently available from a
particular NETLIB site, send an electronic mail (e-mail) message
containing the line

send index

to a NETLIB mail server listed in Section 2 (usually the site
which is geographically closest).

The collected algorithms of the ACM journal, "Transactions
on Mathematical Software" (TOMS), constitute a particularly
worthwhile part of the NETLIB collection. To obtain an index to
the ACM TOMS algorithms available from NETLIB, include the
following line in your e-mail message to a NETLIB mail server:

send index from toms

Many of the NETLIB packages make use of a collection of
routines referred to as the CORE library, which contains routines
for determining machine-dependent constants (e.g., the machine
epsilon), and a set of Level 1, 2, and 3 Basic Linear Algebra
Subprograms (BLAS). To obtain an index to the CORE library,
include this line in your e-mail message:

send index from core

Note that some mail trailers may "confuse" the NETLIB mail
servers. If you encounter problems, you might consider
suppressing the use of your usual signature file when sending
requests to NETLIB via e-mail.


4. Accessing NETLIB via FTP
---------------------------

In addition to the standard e-mail method for accessing the
NETLIB software, the AT&T Bell Laboratories site also provides
access to the NETLIB software via FTP; simply run FTP locally and
connect to the Internet node "research.att.com", login with user-
name "netlib" (in lowercase), and give anything as the password.
The individual NETLIB libraries are arranged in separate subdi-
rectories; for example, the ACM TOMS algorithms are stored in a
subdirectory named "toms", and the CORE library is stored in a
subdirectory named "cor" (*not* "core", oddly enough).

Note that accessing NETLIB via FTP makes it easier to obtain
*all* of the routines in a particular library, but requires a
greater degree of sophistication on the part of the user; i.e.,
the user must possess a certain amount of network and UNIX savvy,
and a solid understanding of how the NETLIB collection is
organized. Accessing NETLIB via e-mail rather than FTP is
recommended for new users, since the NETLIB mail servers
automatically resolve subroutine dependencies (which must
otherwise be resolved manually).


5. X-Windows Version of NETLIB
-------------------------------

In addition to the standard e-mail based NETLIB, the Oak
Ridge National Laboratory site supports "XNETLIB" -- a new, X-
Windows version of NETLIB. According to a recent announcement,
"XNETLIB uses an X-Window graphical user interface and a socket-
based connection between the user's machine and the XNETLIB
server machine to process software requests". Of course, XNETLIB
requires that the user's computer or terminal be capable of
supporting X-Windows. The user will therefore need an engineering
workstation that runs X-Windows, or an X-terminal, or a personal
computer with X-terminal emulation software in order to take
advantage of the benefits of XNETLIB.

To obtain a copy of XNETLIB, send e-mail to the NETLIB mail
server "netlib@ornl.gov" containing the lines

mailsize 300k
send xnetlib.shar from xnetlib

Upon receiving the shar file via e-mail, save the message to a
file (say, a file named "xnetlib.temp"), edit the file to remove
the mail header, and then issue the UNIX commands

sh xnetlib.temp
sh xnetlib.shar

After this, follow the instructions in the README file. Comments
on XNETLIB should be directed via e-mail to "xnetlib@cs.utk.edu".

--

o ------------------------------------------------------------------------- o
| Frederick W. Chapman, User Services, Computing Center, Lehigh University |
| Campus Phone: 8-3218 Preferred E-mail Address: fc03@Lehigh.Edu |
o ------------------------------------------------------------------------- o
| Ecstasy is transitory, but a theorem is forever! |
o ------------------------------------------------------------------------- o