SIMULATION DIGEST VOL. 2, NUM. 4
SIMULATION MODELING & ANALYSIS
Volume: 2, Issue: 4, Tue May 10 09:44:06 EDT 1988
+----------------+
| TODAY'S TOPICS |
+----------------+
(1) Referees for SIMULATION
(2) Distributed Simulation in SIMNET
(3) Circuit Analysis
(4) Sparse Linear Equation Solver
(5) CALL: AI and Simulation papers
----------------------------------------------------------------------------
Return-Path: <cd...@boulder.Colorado.EDU>
Date: Sun, 8 May 88 16:43:40 MDT
From: Charles Shub <cd...@boulder.Colorado.EDU>
To: simulatio...@ufl.edu
Subject: Re: SIMULATION.DIGEST...VOLUME..2..NUMBER..3
SIMULATION, the monthly publication for the Society for Computer
Simulation (SCS) can always use referees. If you would be willing to
serve the simulation community (and SCS) in this important role, please
drop a note to:
Mary Steketee, Managing Editor
SCS
Box 17900
San Diego, CA 92117
The note should include a brief description of your qualifications,
areas of expertise, and a mailing address. Mary has a sheet with a
whole page of keywords we use to match manuscripts with referees.
You might also include information on how many papers you would be
willing to referee per year.
As technical editor, my job is to assign manuscripts to referees and
then make some sense out of the reviews to make a final decision on
publication.
cdash aka cd...@boulder.colorado.edu aka ...ncar!boulder!cdash
aka ...nbires!boulder!cdash aka (719) 593-3492
----------------------------------------------------------------------------
Date: Mon, 9 May 88 09:49:54 EDT
From: rab...@EDDIE.MIT.EDU (Warren J. Madden)
To: simul...@ufl.edu
Subject: Re: SIMULATION DIGEST VOL. 2, NUM. 3 (Distributed Simulation)
Newsgroups: comp.simulation
Organization: USAF Human Resources Lab, Wright-Patterson AFB
For an inteesting example of a distributed simulation system that
does work, check out the SIMNET project funded jointly by the Army and
DARPA and developed by BBN Labs and Perceptronics. Each node in the
network represents one combat vehicle. Currently they have simulators for
M1 Abrams Main Battle Tanks, M2/M3 Bradley Fighting Vehicles, A-10
Thunderbolt II Close Air Support aircraft, and I believe Apache
helicopters.
The gist behind the distribution is that each node maintains its
own dead-reckoning model of all objects in the world. The only time a node
must broadcast a message is when it detects a significant difference
between its own dead-reckoning model of itself and its actual state. It
then broadcasts a message telling all other objects its new state, from
which those other objects update their own dead-reckoning models. In this
manner, overall network traffic is reduced.
I have driven the tank simulator, and found it VERY impressive. It
had Battle Zone (on which I wasted many a quarter in my college days) beat
by many a country mile. I'm still trying to find a way to get back to Fort
Knox and drive it again!
Warren J. Madden
USAF Human Resources Lab
Wright-Patterson AFB
Dayton, OH
wma...@wpafb-af-hrl.arpa
DISCLAIMER: These opinions are my own and in no way reflect those of the
US Air Force or the United States Government.
----------------------------------------------------------------------------
Date: Tue, 10 May 88 07:39:36 +0200
From: unido!christine!sys...@uunet.UU.NET (Sys111)
To: sys...@uunet.UU.NET, fish...@fish.cis.ufl.edu
Subject: query: benchmark-circuits for simulators, CAD-systems
Hello netters,
we are a groupe belonging to Siemens, Corporate Research and Technology.
We are working on architectures for electrical simulation analysis and
modelling, especially hardware acceleration for digital, analog and hybrid
simulation. Within our team one of my special tasks is the question, how can
performance of CAD/systems, analog simulators digital simulators, hybrid
simulators, supported or not supported by hardware, be messured and compared?
My idea is to get a representative collection or a set of small circuits for
electricalm simulation. These circuits should be able to test simulation at any or at several levels like gate-, register-transfer-, switch-, physical level.
One example is a circuit which consists of a flip-flop chain. This circuit is
easy to model with hierachical describtion. It can be used with a chosen
complexity depending on the length of the chain and the number of the
hierachical levels of the describtion, and it can be modelled with flip-flop-
elements of the register-transfer-level or with gates (=elements for creating
a flip-flop) for simulation at gate level.
Another example is an inverter-chain; this circuit can simulated at gate-,
switch- or physical level.
Please contact me if you have similar interest in benchmarking or if you know
such a circuit or if you have used some circuits for performance messurement.
Thanks,
Evelyn Pfeuffer
e-mail: sys111@christine@ztivax.siemens.com (ARPANET)
or UUCP: unido!ztivax!christine!sys111
o-mail: Siemens Ag, ZT ZTI SYS111
Otto-Hahn-Ring 6
D - 8000 Munich 83
West Germany
----------------------------------------------------------------------------
Date: Mon, 18 Apr 88 17:33:01 PST
From: kun...@ic.berkeley.edu (Ken Kundert)
To: fish...@fish.cis.ufl.edu
Subject: simulation software
This is in reply to your request for simulation tools. I have written and
am distributing a sparse matrix package written in C. It was originally
designed for use in circuit simulation, but has since found use in other
applications as well, in particular, device simulation. The currently
released version is sparse1.2, but by summer the faster and more flexible
1.3 version should be released. A small advertizment follows.
Ken Kundert
kun...@ic.berkeley.edu
Sparse1.3
A Sparse Linear Equation Solver
Kenneth S. Kundert
Alberto Sangiovanni-Vincentelli
Sparse1.3 is a flexible package of subroutines written
in C used to quickly and accurately solve large sparse sys-
tems of linear equations. The package is able to handle
arbitrary real and complex square matrix equations. Besides
being able to solve linear systems, it is also able to
quickly solve transposed systems, find determinants, and
estimate errors due to ill-conditioning in the system of
equations and instability in the computations. Sparse also
provides a test program that is able read matrix equations
from a file, solve them, and print useful information about
the equation and its solution.
Sparse1.3 is generally as fast or faster than other
popular sparse matrix packages when solving many matrices of
similar structure. Sparse does not require or assume sym-
metry and is able to perform numerical pivoting to avoid
unnecessary error in the solution. It handles its own
memory allocation, which allows the user to forgo the hassle
of providing adequate memory. It also has a natural, flexi-
ble, and efficient interface to the calling program.
Sparse was originally written for use in circuit simu-
lators and is particularly apt at handling node- and
modified-node admittance matrices. The systems of linear
generated in a circuit simulator stem from solving large
systems of nonlinear equations using Newton's method and
integrating large stiff systems of ordinary differential
equations. However, Sparse is also suitable for other uses,
one in particular is solving the very large systems of
linear equations resulting from the numerical solution of
partial differential equations.
The Sparse1.3 package is currently available from the
Department of Electrical Engineering and Computer Sciences
of the University of California, Berkeley. It was written
in the C programming language by Kenneth Kundert and ver-
sions exist for the UNIX and VAX/VMS operating systems. Be
sure to specify the version when ordering. Sparse1.3 has
replaced Sparse1.2; providing greater capability and speed,
and a more refined interface to the calling program.
Sparse is available for a $150.00 charge. The package
includes the source code on tape, the user's guide, and a
large selection of test matrices. To obtain a copy of
Sparse, send a check or money order payable to the Regents
of the University of California to:
EECS Industrial Liaison Program
461 Cory Hall
University of California
Berkeley, CA 94720
Please allow four weeks for delivery.
The University often does not have the resources to
consult with users on how to use or modify these programs.
We would, however, like to be notified of any problems or
errors in the material provided and appreciate copies on
tape of any troublesome matrices. If the programs are con-
verted to run on other systems, we would like to receive
copies of the modified programs so that these versions can
be made available to the public.
Sparse1.3 Timing Comparisons
Sparse1.3 is compared to Sparse1.2, Harwell's MA28, and
Yale's YSMP. Comparisons are made based on the time
required to perform each of three tasks: factor the matrix
given an unordered matrix, factor a previously ordered
matrix, and compute the solution to a matrix equation given
a factored matrix. Times were measured on a vax8650 under
Ultrix2.2 and are given in seconds. The last digit of the
given times is uncertain.
Sparse is configured to use diagonal pivoting with a pivot
threshold of 0.001. It is not using its modified Markowitz
pivoting algorithm.
YSMP uses a symbolic ordering algorithm that does not take
into account the numerical values in the matrix, which
explains why it failed on the grid10 matrix. As a result,
YSMP cannot be fairly compared with the other packages on
the basis of time required to factor an unordered matrix and
so it was not included in that table.
Most matrices were generated during the course of computing
the DC operating point or the transient solution of some
circuit using Relax, Advice, or the simulator from Zycad.
The exceptions are mat8; which was taken from a force
directed placement problem; and grid10, grid33, grid66,
karti, and karti2; which were generated when trying to solve
a system of partial differential equations.
________________________________________________________________________
Order and factor times
________________________________________________________________________
matrix size entries/row sp1.3 sp1.2/sp1.3 ma28/sp1.3
________________________________________________________________________
digfi.dc 378 3.85 0.4 0.93 0.74
eprom 630 4.92 1.32 1.01 0.46
timem.dc 1957 3.42 2.55 0.93 3.05
timem.tr 1957 5.08 3.7 0.97 0.36
zy1137 1137 11.14 8.37 1.05 0.21
zy3315 3315 4.82 9.07 0.9 0.34
zy450 450 4.52 0.58 1.24 0.55
zy1195 1195 6.73 43.18 1.07 1.79
matrix4000 4000 3.20 29.6 0.99 0.90
vish 2806 8.36 80.48 1 0.31
decode03 14 4.86 0.02 1 oo
decode04 24 6.67 0.03 1 1
decode05 42 8.86 0.05 1.4 1
decode06 76 11.26 0.20 1.1 0.4
decode07 142 13.72 0.72 1.04 0.34
decode08 272 16.12 2.9 1.08 0.32
decode09 530 18.42 12.5 1.01 0.21
decode10 1044 20.64 53.85 0.99 0.19
adv2806 2806 8.36 74.55 1.01 0.29
adv3388 3388 11.97 118.22 1 0.08
adv3776 3776 7.31 82.92 1 0.087
adv920 920 8.10 6.65 0.94 0.18
joeyADC.dc 5355 4.63 34.03 1 0.54
joeyADC.tr 5355 6.02 126.95 0.99 0.2
mat8 300 13.33 13.73 0.97 0.69
grid10 100 9.22 0.17 1.08 1
grid33 1089 4.88 11.18 1.02 2.88
grid66 4356 4.99 135.48 1.02 5.62
karti 171 9.39 0.85 1.02 0.72
karti2 1073 9.20 84.83 1.48 3.24
________________________________________________________________________
____________________________________________________________________________
Factor times
____________________________________________________________________________
matrix size entries/row sp1.3 sp1.2/sp1.3 ma28/sp1.3 ysmp/sp1.3
_____________________________________________________________________________
digfi.dc 378 3.85 0.024 1.13 3.52 2.78
eprom 630 4.92 0.093 1.42 2.7 1.27
timem.dc 1957 3.42 0.084 1.46 4.79 1.99
timem.tr 1957 5.08 0.134 1.25 4.35 1.49
zy1137 1137 11.14 0.206 1.37 7.55 2.11
zy3315 3315 4.82 0.264 1.35 4.08 1.64
zy450 450 4.52 0.026 1.27 5 1.92
zy1195 1195 6.73 2.794 1.66 0.92 6.82
matrix4000 4000 3.20 5.907 1.38 1.3 0.75
vish 2806 8.36 4.178 4.06 1.56 0.77
decode03 14 4.86 0.001 0 2.5 0
decode04 24 6.67 0.002 1.5 oo 8.5
decode05 42 8.86 0.004 1.75 2.33 0
decode06 76 11.26 0.014 1.93 3.47 1.7
decode07 142 13.72 0.031 2.26 3.4 1.67
decode08 272 16.12 0.084 3.51 4.37 1.19
decode09 530 18.42 0.225 6.06 3.71 0.96
decode10 1044 20.64 0.565 11.1 3.41 0.94
adv2806 2806 8.36 4.455 3.54 1.21 0.67
adv3388 3388 11.97 1.355 2.07 17.2 1.49
adv3776 3776 7.31 1.612 1.65 9.94 1.16
adv920 920 8.10 0.150 1.35 12.4 1.35
joeyADC.dc 5355 4.63 0.250 2.2 6.27 2.73
joeyADC.tr 5355 6.02 4.370 4.22 1.15 0.84
mat8 300 13.33 3.961 1.35 0.66 0.77
grid10 100 9.22 0.012 1.5 3.06 zero pivot
grid33 1089 4.88 1.078 1.65 1.63 0.94
grid66 4356 4.99 12.03 1.80 2.10 0.85
karti 171 9.39 0.062 1.42 2.12 1.34
karti2 1073 9.20 11.14 2.09 0.77 1.05
______________________________________________________________________________
_______________________________________________________________________________
Solve times
_________________________________________________________________________________
matrix size entries/row sp1.3 sp1.2/sp1.3 ma28/sp1.3 ysmp/sp1.3
_________________________________________________________________________________
digfi.dc 378 3.85 0.011 1.36 1.5 0
eprom 630 4.92 0.019 1.05 1.3 0.94
timem.dc 1957 3.42 0.04 0.93 2.24 0.87
timem.tr 1957 5.08 0.04 1.08 2.05 1.68
zy1137 1137 11.14 0.037 1.03 2 1.52
zy3315 3315 4.82 0.072 1.11 1.79 1.39
zy450 450 4.52 0.008 0.88 2.57 0
zy1195 1195 6.73 0.081 1.17 1.43 3.7
matrix4000 4000 3.20 0.16 1.13 2.14 1.04
vish 2806 8.36 0.297 1.01 0.92 0.62
decode03 14 4.86 0 oo 0.5 oo
decode04 24 6.67 0.001 0 0.5 0
decode05 42 8.86 0.001 3 oo oo
decode06 76 11.26 0.002 0 1.33 0
decode07 142 13.72 0.005 2 3.33 5.67
decode08 272 16.12 0.016 0.82 1.2 1.31
decode09 530 18.42 0.035 0.71 1.27 1.22
decode10 1044 20.64 0.073 0.79 1.28 1.06
adv2806 2806 8.36 0.147 1.09 1.43 0.90
adv3388 3388 11.97 0.122 1.12 1.89 1.37
adv3776 3776 7.31 0.138 1.12 1.41 1.33
adv920 920 8.10 0.023 1.23 1.85 2.17
joeyADC.dc 5355 4.63 0.1 0.83 1.67 1.5
joeyADC.tr 5355 6.02 0.233 1.0 1.21 1.07
mat8 300 13.33 0.108 1.04 0.6 0.62
grid10 100 9.22 0.005 0.6 3.5 zero pivot
grid33 1089 4.88 0.057 1.28 1.95 1.75
grid66 4356 4.99 0.34 1.29 1.70 1.27
karti 171 9.39 0.017 1 0.77 1.31
karti2 1073 9.20 0.173 1.92 1.04 1.65
_________________________________________________________________________________
----------------------------------------------------------------------------
Date: Tue, 10 May 88 09:37:28 EDT
From: Paul Fishwick <fish...@fish.cis.ufl.edu>
To: simul...@ufl.edu
P l e a s e P o s t
==========================
*** CALL FOR PAPERS ***
==========================
ARTIFICIAL INTELLIGENCE & SIMULATION
==========================
DEADLINE: JULY 1, 1988
==========================
First Notice
_________________________________________________________________
Publication: ACM SIGSIM SIMULETTER
Issues # 3 and 4, 1988
Topic: Artificial Intelligence and Simulation
Paper Deadline: July 1, 1988 (issue #3), September 1, 1988 (issue #4)
Paper Length: 800 words to 5000 words
Spacing: Single
Leave pages *unnumbered* - publisher will do numbering
Send 3 copies of the manuscript on 8 1/2 x 11 paper
Send to:
Prof. Paul A. Fishwick
SIGSIM Associate Editor for Tech. Themes
Dept. of Computer and Information Science
University of Florida
Bldg. CSE, Room 301
Gainesville, FL 32611
NOTE: Your submission must be camera ready (i.e. include
all figures and tables within text)
_________________________________________________________________
Request:
All research associated with knowledge based simulation and
bridging the gap between AI methodology and topics within
simulation.
________________________________________________________________
Policy:
Contributions can flexibly range between 800 and 5000 words.
All contributions must be original. Papers will be reviewed and
a letter of acceptance or rejection will be sent to the first
author. If an accepted paper cannot be fit into a special issue
(due to size limitations or other considerations) then it will
be forwarded to the editor for inclusion in a later issue.
________________________________________________________________
SIGSIM:
The SIGSIM (ACM Special Interest Group in SIMulation) publication
"SIMULETTER" has had a technical theme associated with each issue
beginning with issue # 2 for 1988. SIMULETTER has recently
combined with IEEE TC MODELLING to form a single, joint newsletter
promoted by both ACM and IEEE.
________________________________________________________________
+--------------------------+
| END OF SIMULATION DIGEST |
+--------------------------+