[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
| Title: | Mathematics at DEC |
|
| Moderator: | RUSURE::EDP |
|
| Created: | Mon Feb 03 1986 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 2083 |
| Total number of notes: | 14613 |
368.0. "REDUCE Newsletter" by TOOLS::STAN () Thu Oct 24 1985 23:01
From: ROLL::USENET "USENET Newsgroup Distributor 24-Oct-1985 2101" 24-OCT-1985 21:57
To: @[.net.math.symbolic]NEWS.DIS
Subj: USENET net.math.symbolic newsgroup articles
Newsgroups: net.math.symbolic
Path: decwrl!greipa!pesnta!qumix!ittatc!dcdwest!sdcsvax!sdcrdcf!randvax!lseward
Subject: Reduce Newsletter #10
Posted: 16 Oct 85 18:39:58 GMT
Organization: Rand Corp., Santa Monica
REDUCE
Newsletter
______________________________________________________________________________
Number 10 September 1985
______________________________________________________________________________
_E_d_i_t_o_r'_s _F_o_r_e_w_o_r_d
In a paper I gave at the 1980 IFIP Congress, I predicted that by the mid-
1980's it would be possible to obtain a complete computing system (both
hardware and software) that could run programs like REDUCE for prices in the
$10,000 - $20,000 range. Not only do some of the workstation offerings
described in this issue meet that target, but, more excitingly, there is the
promise of such systems being available soon for under $5000. This means that
researchers who are spending large amounts of money running computer algebra
problems on central computing systems may find themselves better served by
purchasing personal systems. It is important to recognize that such worksta-
tions are not toys; in many cases, the available systems can run REDUCE appli-
cations at half the speed of a DEC VAX 11/780, which many researchers now use.
To compare this speed with the machine you are using, we give a table later in
this issue comparing the time it takes to run the REDUCE standard test file on
a number of different computers.
_N_e_w _R_e_l_e_a_s_e _o_f _R_E_D_U_C_E
A new version of REDUCE (3.2) was released earlier this year. Following our
usual tradition, the system date of this version is April 15, although a cou-
ple of months of field testing followed before the system was announced for
general use. Updates consisted mainly of bug fixes, and a list of these was
included in a mailing to registered system holders of version 3.1, which also
offered the new version for a reduced fee.
_R_E_D_U_C_E _T_h_i_r_d _P_a_r_t_y _D_i_s_t_r_i_b_u_t_i_o_n _P_r_o_g_r_a_m _B_e_g_u_n
For over a decade now, REDUCE has been distributed from the Editor's home in-
stitution to a large number of sites throughout the world. From time to time,
we have also made informal arrangements with other organizations to distribute
versions of REDUCE for machines that we did not support ourselves. With more
and more machines becoming available which can run such software, we have de-
cided that the time has come to make such arrangements more formal. We have
therefore instituted a REDUCE third party distribution program for this pur-
pose with appropriate licensing arrangements. Listed below are some new dis-
tributors of REDUCE who are part of this program. We expect to add several
more to this list in the near future.
Recipients of REDUCE are reminded that the basic source and documentation have
been copyrighted and that distribution of the software or documentation to
other sites is a violation of that copyright. Furthermore, system holders
should register with us once the software is running. As we pointed out in the
last newsletter, the REDUCE User's Manual and other documentation now carry a
copyright statement requiring such registration as a condition for their local
distribution. We are not imposing these conditions to make it harder for
users to access REDUCE. However, changes in the software market place force
us to take such steps in order to ensure that there is no unauthorized exploi-
tation of the software.
There are several new versions of REDUCE now available:
1. Apollo Version. This is designed for use on Apollo workstations under
Aegis SR8.0 and above. It is based on the Portable Standard LISP (PSL)
system developed at the University of Utah. The tape includes the PSL
files needed to run REDUCE. For further details, contact
Utah Portable AI Support Systems Project
Attn: Loretta Cruse
Department of Computer Science
3160 MEB
University of Utah
Salt Lake City, UT 84112
Telephone: (801) 581-5017
Electronic Mail (Arpanet): [email protected]
2. Metacomco Version. This version, for stand-alone Motorola 68000 based
microcomputers, is a full implementation of REDUCE in Cambridge LISP
currently running under either CP/M-68K or Metacomco's operating sys-
tem TRIPOS. A minimum of 1 MB of RAM and a Winchester fixed disk are
required for a comfortable performance. For further details, contact
Michael Taraniuk
Metacomco
26 Portland Square
Bristol, BS2 8RZ
GREAT BRITAIN
Telephone: (272) 428781
Telex: 265871 MONREFG (please quote MEA001 in all telexes)
Electronic Mail (Telecom Gold): 84:MEA001
3. Orion Version. This version, for the High Level Hardware 32 bit user
microprogrammable superminicomputer running UNIX 4.2bsd and Cambridge
LISP, has been developed at the University of Bath in collaboration
with High Level Hardware. For further details contact
Sales Office
High Level Hardware Ltd
P.O. Box 170
Windmill Road
Oxford OX3 7BN
GREAT BRITAIN
Telephone (865) 750494
4. StaffLISP Version. This version, which runs under CP/M-68K on a
Motorola 68000 based machine with at least 1 MB of main memory and a 1
MB floppy disk, uses the StaffLISP interpreter and compiler developed
by B U G, Inc. StaffLISP has been carefully optimized so that REDUCE
runs very efficiently on this configuration. B U G, Inc. can also pro-
vide help to anyone interested in implementing REDUCE on other sys-
tems. For further details contact
B U G, Inc.
Kitayama Bldg.,
4F S6W11, Chuo-ku
Sapporo 064
JAPAN
Telephone (11) 562-2706
_U_s_e_r_s _L_i_b_r_a_r_y _F_o_r_m_e_d
In REDUCE Newsletter No. 8, we asked users to contribute programs for a users
library to be included on the REDUCE distribution tape. I am happy to an-
nounce that four such packages have now been submitted, and will be made
available with the next distribution of REDUCE, scheduled for April 1986. A
description of these packages is given below, together with a reference.
BAS (Buchberger Algorithm System ) is a package written by R. Gebauer, A.
C. Hearn and H. M. Moeller for constructing a Groebner Basis of an
ideal _I, when an arbitrary basis of _I is given. Operations are based
on a distributive polynomial representation package developed by R.
Gebauer, A. C. Hearn and H. Kredel. Reference: R. Gebauer and H. M.
Moeller "A Fast Variant Of Buchberger's Algorithm" submitted to Jour-
nal of Symbolic Computation.
EXCALC is a system written by Eberhard Schruefer for performing calcula-
tions in modern differential geometry. Geometrical objects like exte-
rior forms or vectors can be introduced by declarations. Operations
such as exterior multiplication and differentiation, inner products,
Lie derivatives, variations etc. can then be performed on these ob-
jects. Moving frames and easy calculation of connections are support-
ed. Reference: E. Schruefer, F.W. Hehl and J.D. McCrea, "Application
of the REDUCE Package EXCALC to the Poincare Gauge Field Theory of
Gravitation", University of Bonn preprint (1985).
GENTRAN is a code generation and translation package written by Barbara L.
Gates which generates complete numerical programs directly from REDUCE
by transforming REDUCE forms into formatted FORTRAN, RATFOR or C code.
Reference: B.L. Gates, "GENTRAN: An Automatic Code Generation Facil-
ity for REDUCE", SIGSAM Bulletin (to appear).
SPDE is a package written by Fritz Schwarz for determining symmetries of
partial differential equations. The package provides a set of REDUCE
functions which perform the various operations for determining the
symmetry generators and the structure of the symmetry group for the
given system of PDE's. Reference: F. Schwarz, "Automatically Deter-
mining Symmetries of Partial Differential Equations", Computing, 34
(1985) 91-106 and Addendum, to appear in same journal.
Readers are encouraged to submit other packages to this library. As we stated
in the cited newsletter, to qualify for inclusion in this library, a program
must have:
1. A header comment listing the title of the program, a brief abstract,
the version of REDUCE (including machine class) on which it was test-
ed, and the name and address of the author.
2. A separate, machine-readable document describing the program and its
use.
3. A test file with expected results included as comments.
4. A letter from the author giving us nonexclusive rights to unlimited
distribution of the code.
_R_E_D_U_C_E _N_e_t_w_o_r_k _M_a_i_l_i_n_g _L_i_s_t _n_o_w _i_n _O_p_e_r_a_t_i_o_n
On the mailing list information form in the last issue of this newsletter, we
asked recipients to list a networking address if they had one. The REDUCE dis-
tribution form also requests this information. We recently sent an online
copy of this newsletter to those who gave us such an address, and shall be
sending other mailings as appropriate from time to time. If you have already
given us your network address, and haven't heard from us, it means either that
we failed to add your address to our mailing list, or we couldn't reach your
address from our site. The networks we can currently reach are: Arpanet, Bit-
net, Csnet, Earn, Mailnet, uucp, and those United Kingdom sites that can be
reached via the Arpanet node at University College London (ucl-cs). Please
let us know if you wish to be on this list and have not already received a
message.
_R_e_v_i_e_w _o_f _R_E_D_U_C_E _P_u_b_l_i_s_h_e_d
There are so many papers being published these days that reference REDUCE that
it is impossible to acknowledge them all in this newsletter. However, papers
which appeal to a wide audience or contain useful tutorial material will be
referenced as they appear. In particular, I'd like to draw your attention to a
recent review of REDUCE by John Fitch. By means of application examples, this
paper provides a comprehensive introduction to the system for a prospective
user. The paper is entitled "Solving Algebraic Problems with REDUCE" and is
published in The Journal of Symbolic Computation 1 (1985) 211-227.
_R_E_D_U_C_E _T_i_m_i_n_g_s
During the past few years, we have asked recipients of the REDUCE distribution
tape to report the time it takes to run a standard test file. In a paper re-
cently sent to the SIGSAM Bulletin, Jed Marti and the Editor have summarized
the results of the analysis of timings of REDUCE 3.0 performance from ninety
installations involving a wide range of different machines. The results show a
reasonable correlation between manufacturers' claims of processor performance
and the collected execution times. We show below an alphabetical summary of
the timing tests on 40 different machines. The four columns contain the aver-
ages for the number of CPU seconds to run the test, the total elapsed time,
the total number of megabytes the test used, and the size of the physical
memory in megabytes. Occasionally, numbers were not reported. These cases
were either ignored, or are represented by an asterisk. The SIGSAM Bulletin
article should be consulted for a description of the LISP systems used and the
portions of REDUCE exercised by the test file.
_T_a_b_l_e _o_f _R_E_D_U_C_E _T_i_m_i_n_g_s
Machine CPU Time Wall Clock Region Size Total Size
(seconds) (seconds) (megabytes) (megabytes)
____________________________________________________________________________________________________________________________________________________________
Amdahl 470 V7/A 10.3 11.9 0.7 8.0
Amdahl 470 V8 7.2 179.6 0.7 12.1
Apollo DN 300 78.7 108.5 2.0 1.4
Apollo DN 320 80.4 121.0 2.0 1.4
Apollo DN 400 125.5 189.0 2.0 1.0
Apollo DN 600 89.9 110.0 2.0 2.0
CDC Cyber 170/825 106.8 106.8 0.6 *
DEC 1099 17.7 240.0 1.0 4.5
DEC 2020 122.8 192.0 0.4 1.1
DEC 2060 22.5 125.0 0.7 5.1
DEC KL-10 24.6 * * 0.6
DEC KL-1091 44.1 234.0 * *
DEC VAX 11/750 78.7 132.6 1.3 1.3
DEC VAX 11/780 50.3 60.1 2.6 5.5
Facom M-180N 21.3 25.6 0.5 8.0
Facom M-200 7.2 8.2 1.3 7.0
Facom M-380 3.8 25.0 2.4 32.0
Facom M-382 3.6 4.1 4.0 64.0
Hewlett-Packard 9836 65.3 65.3 4.4 4.4
Hitachi M-200H 6.5 8.4 0.7 0.8
Hitachi M-280H 4.0 5.0 2.9 32.0
Hitachi S-810 2.8 4.5 1.5 1.5
IBM 3031 40.7 48.7 0.5 5.4
IBM 3033 9.6 11.4 1.1 13.1
IBM 3081 6.6 8.3 1.2 24.0
IBM 3083 7.6 8.2 2.0 9.0
IBM 3084 5.4 6.3 0.7 32.0
IBM 370 158 49.9 58.0 2.0 6.0
IBM 4341 Model 1 52.0 60.9 1.0 1.7
IBM 4341 Model 2 30.1 45.7 0.9 7.2
MRP ES-1060 68.4 113.4 0.3 *
NAS AS6 18.2 * 4.0 6.0
NAS NAS7000 23.4 * * *
Robotron ES-1040 149.2 175.5 0.5 0.5
Sage IV 224.8 224.8 1.0 1.0
Siemens 7865 19.4 21.9 1.0 *
Siemens 7890 3.8 4.2 0.9 14.0
SML Darkstar 227.9 227.9 1.7 1.7
Symbolics 3600 45.0 45.0 160.0 6.0
Xerox Dolphin 322.0 322.0 1.2 1.2
_________________
The REDUCE Newsletter is published on an occasional basis as a service to the
REDUCE user community. Contributions or inquiries should be addressed to the
Editor, Anthony C. Hearn, The Rand Corporation, 1700 Main Street, P.O. Box
2138, Santa Monica, CA 90406-2138, telephone (213) 393-0411.
| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 368.1 | | AURORA::HALLYB | | Fri Oct 25 1985 12:15 | 24 |
| Well I don't know anything whatsoever about REDUCE; I may even find it a fun
product, but there is something definitely odd about the reported timings.
Consider this 36-bit excerpt:
Machine CPU Time Wall Clock Region Size Total Size
(seconds) (seconds) (megabytes) (megabytes)
DEC 1099 17.7 240.0 1.0 4.5
DEC 2060 22.5 125.0 0.7 5.1
DEC KL-10 24.6 * * 0.6
DEC KL-1091 44.1 234.0 * *
All four of these CPUs are substantially the same. The "1099" is a dual-CPU
version of the -10, and all indications are that only one job was run at a time,
hence there would be no benefit from a second CPU. Yet the 1099 took 40% the
CPU time of a 1091, but the elapsed time was the same! Furthermore, the 1091
and 2060 are identical CPUs (different cabinets, O/S's), but the 2060 took only
half the times of the 1091!
This isn't sour grapes -- I don't even like grapes -- it just makes me wonder
about the numbers reported for CPUs that I don't know anything about. Is there
more here than meets the eye?
John
|