SIMULATION DIGEST VOL. 2, NUM. 4

SIMULATION DIGEST VOL. 2, NUM. 4

Post by SIMULATION MODELING & ANALYS » Wed, 02 May 1990 14:44:00



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: simulation-requ...@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: simulat...@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
wmad...@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, fishw...@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@christ...@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: kund...@ic.berkeley.edu (Ken Kundert)
To: fishw...@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
kund...@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 <fishw...@fish.cis.ufl.edu>
To: simulat...@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 |
+--------------------------+