87 informatica <
YU tSSN 0350-5596

informatica
JOURNAL OF COMPUTING AND INFORMATICS
Published by Informatika, Slovene Society for Informatics, Parmova 41, 61000 Ljubljana, Yugoslavia
YU ISSN 035&-5596
Editorial Board
T. AleksiC, Beograd; D. Bitrakov, Skopje;	P.
Dragojloviö, Rijeka; S. Hodaar, Ljubljana;	B.
Horvat, Maribor; A. Mandale, Sarajevo;	S. MlhalU, Vara2din; S. Turk, Zagreb
VOLUME 11,1987-No. 4
Edltor-tn-Chlef : Prof. Dr. Anton P. Zeleznikar
CONTENTS
Executive Editor : . Dr. Rudolf Mum
Pubi 1shlng Counc ì1 :
T. Banovec, Zavod SR Slovenije za statistiko, Vozarslcl pot 12, 61000 Ljubljana;
A.	Jerinan-Blaiie, DO Iskra Delta, Parraova 41,
61000 Ljubljana;
B.	KlemenölC, iskra Telematlka, 64000 Kranj;
S. Saksida, Institut za sociologijo Univerze Edvarda Kardelja, 61000 Ljubljana
J. Virant, Fakulteta za elektrotehniko, TrSaSka 25, 61000 Ljubljana.
Headquarters :
Informatica, Parmova 41, 610QQ LJubljana, Yugoslavia. Phone: 61 31 29 88. Telex: 31366 yu
del ta
Annual Subscription Hate: US$ 30 for companies, and US* 15 for individuals
Opinions expressed in the contributions are not necessarily shared by the Editorial Board
Printed by; Tiskarna Kresija, Ljubljana
D. FajCar M. Lokar
A.P.Zeleznikar	8
J. Sllc	27
B,	Robi«
M.B.JocJjovlć 35 D.M.VelaSeviö
S. Presern
40
B.Mihovilovic 54 P. Kol bežen J. Sile
Analysis of Buffered Multistage Interconnection Network for Parallel Processors
Information Determiantions U
Data Flow Based Parallel Inference Machine
One Solution of the Busline Crew Scheduling Problem
A Selected Survey of Parallel Computer Systems
A Paradigm of Transputer System Implementation
M.	OjsterSek 59
V,	Zumer
P.	Kokol
A.	Zorman
D.	Zivkovie 64
A Simulation of Data Flow Computer Models and a Model of Fault Tolerant Dataflow Computer
An Alternative Procedure for Determination of Entity Types and Actions In JSD Method ...
A. Brodnik M. Spegel T. Lasbaher
69 Progracrnii ng with Modula~3 II
P. Markovi <! 77
B.Jerman-	84
Bla^i C M.Kapus-Kolar
implication of Nervous System Organization on Computer System Development
Communication and Application Processes in the Upper Layers of the OSI Reference Model
informatica
ČASOPIS ZA TEHNOLOGIJO RAČUNALNIŠTVA
IN PROBLEME INFORMATIKE
ČASOPIS ZA RAČUNARSKU TEHNOLOGIJU I
PROBLEME INFORMATIKE
SPISANIE ZA TEHNOLOGIJA NA SMETANJETO
I PROBLEMI OD OBLASTA NA INFORMATI KATA
Časopis izdaja Slovensko društvo Informatika, 61000 Ljubljana, Parmova 41, Jugoslavija
ürednUki odbor:
T. A.leksi(!, Beograd; D. Bitrakov, Skopje; P. Dragojloviti, Rijeka; S. Hodzar, Ljubljana; B. Horvat, Maribor; A. Mandzic, Sarajevo; S. Mihalie, Varaždin; S. Turk, Zagreb
YU ISSN 0350-^596
LETNIK 11,1987-ŠT. 4
Glavni in odgovorni urednik; prof. dr, Anton P. Zeleznikar
Tehnični urednik t dr. Rudolf Murn
Zalo&niski sveti
T, Banovec, Zavod SR Slovenije za statistiko, VoZarskI pot 12, 61000 LJubljana;
A.	Jerman-BlažiC, DO lakra Delta, Parmova 41,
61000 Ljubljana;
B.	Klemenäiö, Iskra Telem&tlka, 64000 Kranj t
S. Saksida, Institut za sociologi Jo Univerze Edvarda Kardelja. 61000 Ljubljana;
J. Virant, Fakulteta za elektrotehniko, Trsaska 25, 61000 Ljubljana.
Uredništvo tn uprava:
Informatika, Parmova 41, 61000 LJubljana, telefon (061) 312 988; teleks 31366 YU Delta.
Letna naročnina za delovne organizacije znaSa 11990 din, za zasebiie naročnike 2990 din, za Studente 990 din; posamezna Številka 4000 din.
Številka ziro raCuna: 50101-678-51841
Pri financiranju Časopisa sodeluje Raziskovalna skupnost Slovenije
Na podlagi mnenja Republiškega komiteja za informiranje 9t. 23-85, z dne 29. 1. 1986, Je časopis oproščen temeljnega davka od prometa proizvodov.
Tisk: Tiskarna Kresija, Ljubljana
D. Fajfar M. Lokar
A.P.Zeleznikar	B
J. Sile	27
B.	RobiC
M.B.JockOVie 35 D.M.VelaSeviC
VSEBINA
Analysis of Buffered Multistage Interconnection Network lor Parallel Processors
Information Determiantions II
S. Presern
40
B.MIhov! loviC 54 P. Kolbesen J. Silc
Data Flow Based Parallel Inference Machine
One Solution of the Busline crew scheduling Problem
A Selected Survey of Parallel Computer Systems
A Paradigm of Transputer System Implementation
M.	OJsterSek
V.	Zumer
P.	Kokol
A.	Zorman
D. Zivkovlc
A.	Brodnik 69
M.	Spegel
T.	Lasbaher
P.	Markoviö 77
59 A Simulation of Data Flow Computer Models and a Model of Fault Tolerant Dataflow Computer
64 Alternativni postupak određivanja tipova entiteta ...
Programiranje z Modulo-2 II
B.Jerman-BiaM.e M.Kapus-Kolar
Implikacija organizacije nerv-nih sistema na računarske sisteme
84 Komunikacijski In aplikacijski procesi V vlSJih nivojih refp-rencnega modela OSI
ANALYSIS OF BUFFERED MULTISTAGE INTERCONNECTION NETWORK FOR PARALLEL PROCESSORS
UDK 618.3.02
Dušan Fajfar Institute for Teleinformatics ISKRA Telematika, Kranj Matija Lokar
University E. K. Ljubljana, Department of Mathematics
Abstract In the paper a method for calculating distribution law on the nuaber of neoory requests in the finite buffer in multistage Interconnection network of parallel processors is given. A modified delta network ia used for the connecting processors with niemory modules. Memory roquesta on each processor are generated randomly and independently. Two cases of traffic flow are discussed: the constant average rates of requests and the time dependent average rates.
Povzetek V članku analiziramo večnivojaka povezovalno isrežo med procesorji In pomnllniSklmi moduli, ki jo sestavljajo elementi z vnesnimi pomnilniki omejeno kapacitete. Podana je metoda za izračun porazdelltveoega zakona števila zahtev po podatkih iz pomnilnika v posameznem elementu mreže. Vsak procesor generira zahteve po pomnllnläkih modulih naključno In neodvisno od ostalih. Obravnavana sta dva primera: časovno konstantno in s Saaom spreminjajoče se povprečno število zahtev.
Keywords Buffered network, delta network, multistage interconnection network, buffer length, queuelng theory.
I.Introduction
In the recent years a lot of new multiprocessor architectures have been proposed. The main problem of any multiprocessor system la its interconnection network. A typical configuration of such system is Illustrated In Fig. 1. Many identical processors are connected via an interconnection network to identical memory modules. Each processor should have access to each memory module with requests generated randomly and Independently at each processor.
ini&t'conriec.'i^lon ne.t^or'tc
We have several possible organizations of such, processors - memory interconnection. For example, a shared bus in an unexpehslve, but admits very low
transfer rate. On the other hand maximum transfer rate Is attained by the full crossbar switch (with n«m switches where n Is number of processors and m number of memory modules). However it is far too complicated and expensive for practical application. Thus we have to use interconnection networks with less then nsm switches. One of then Is multistage interconnection network Cl]. k lot of them have been presented in the literature (C2],[33,tl],C5l,C6J). In this paper we study a delta network with some modifications In the first and the last stage (Input process from processors and output process to memory). In section II description of network la given. The network consists of three types of switches. For more detailed presentation see [6]. In sections III, IV and V the distribution law of buffer length for each type of switch is calculated. In section VI results and conclusions are given.
II. The network nodal and ayatem operation
We discuss a multistage network for connection N processors with N memory modules, where N = 2". The model for N = 16 is illustrated in Fig. 2.
The network consists of logg N + 1 stages. Each stage has N identical switches. With regard to the number of input and output links there are three types of switches. Switches at the first stage have one input and two output
links, awitches at the last stage have two input links and only one output link. Switches at all other stages have two input and two output llnka. To achieve higher transfer rate and to avoid blocking there is a finite buffer at each output link, where the Inccnin« requesta wait to be processed further,
he we have finite buffers we propose that in the case when the buffer Is full, the incoming requests are lost.
For the time unit we choose the system cycle time.
At each cycle only one request can be transmitted through
the same link except at the first stage where awre than one request can come by input links from each processor.
In one time unit only the first request In the buffer
can travel from the stage 1 to stage 1 + 1. As we can
see on Fig. 2. there are no links between the switches
of the sane stage. The input link a of stage 1 are the
output links of the stage 1-1. The first stage has Input
links from processors and the last stage has output links
to memory modules. This regularity of the network gives
the possibility to analyae It stage by stage Instead of
the whole network at once. In the next section we give
the analysis of the first stage in the section IV we
give analysis of the stages indexed from 2 to log^N and
In the section V the last stage is analyzed.
®©©®o©®o®®®©
III. Pitches with one Input and two output links
Switches with one input and two output links can be found only in the first stage. The Input links come from processors and the output links are connected to switches of the second stage. Each processor Is connected to only one switch. There are no links between the switches of the sane stage, so we have N identical systems. Each system consists of a processor that sends the requests to switch and two output links, each with finite buffer.
As the Bemory requests are time Independent ([6]), we use the Polsaon distribution with a given average rate a. The average rate can be different for each processor. Let p denote the probability that k requests are aent from processor in one cycle. By Polsson law we get

a" e-^
kl
(U
He analyse the case where a Is constant and the case where a Is tine dependent, denoted by aft).
Each request coming from processor is switched in one of two output buffers. As requests are uniforinly distributed between meoory modules ([6]) there Is an equal probability that the incoming request joins the first or the second buffer. By pjj „, denote the probability that m of k requests enter the first buffer (and k-m requests enter the second buffer). Thus we get
(0.5)
(2)
Let p^ denote the probability that m requests enter the first buffer In one cycle. Then

T k_l = i P P,
kzn
J a^^e-
ksm
/ k) (0.5)"
a-	/ ( 2«
(3)
The output process Is very simple if we compare it with the input procesa. In each cycle only the first request from each buffer leaves system.
Let us first analyze stationary system with constant average rate of input process. Since eventa on both buffers are equal wa could analyze only one of them. Let bl denote the buffer length (we propose that all buffers have the same length but we do not have any difficulties when buffers have different length. We Just have to calculate the distribution for all switches). The balance diagram Is shown In Fig 3.

The nodes In the diagram denote the number of requests in the buffer. If we look at in- and out-coming arcs to each node we get the system of balance equations
where ß(») denotes the probability that we have a
requests in the buffer. m
nco) • I = 0(1) » Pp
IDsl
n(a).[J>Q + I =
]c=1
a = 1, 2, .... bl - 1 (4)
n(bi)»['jf>Q]
bi-1
=	+ I n(k)»p
k=1
bl-k+l
Iw we pot Pjii expressed by (3) In the left side of equations (4) we get the following system of llnaar equations.
a(n s (e"®'^ - 1) I n(o)
0(s).[l -	^
s ^^^	3-k+l	0
a : 1, 2.....bl - 1 (1')
(l<bl) 1 ■ bl-l

-ktl
Aa the system (1') Is trlvlaly solved by putting all fl(a} to 0, we add the normalization equation (5).
will Join the buffer.
q,.^^ = niln (bl, q^ + V^^, - l) q^ ^ 0

= Bin (bl, Vj.^,}
qj. = 0
So 0(8,t) ia expreaaed by
s+l
n(a,t) = Ì (ici,t-1) v'
1=1 ■t- ptO,t-1)
a-l+l
(9)
If a is lesa than buffer length bl. In that case sons of the requests are rejected. The equation Is almoat the same as (9), we must only replace vj with v^, where
1-1
v^ , - I v5 ^	J=0 J
(10)
la the probability that more than i-1 requests are Joining tha buffer at the time t.
s+1
R(a,t)	0(1,t-1) v^.^^,
+ p(0,t-l) v^
(11)
We start with the diatributlon
p<0,0) = 1, pts.O) = 0
0 = t,2.....bl
and repeatedly calculate the probabilities.
I P<s) = 1
(5)
SsO
Now the system can be solved by any of the well-known methods for solving linear system of equations.
For the system where average rate of input process Is time dependent we could not use the same approach. Because the system works In cycles we consider the discrete time. Instead of probabilities nCs) we have n(s,t) where the second variable denotes the cycle counter. With vj we denote the probability that i requests came on input line to the buffer at the time t and la expressed by the following equation
^l-l pj . P'^	(6)
k=l
A Simple calculation gives

(7)
With a given buffer length bl we get the recurrence relation for the number of requests In the buffer at the time t (denoted by qj.) if m this moment V^^ requests
The same approach we could use in the case where average input rate is tloe constant. But as equations (9) and (10) can not be simplified, the calculation with this approach is tine consuming.
IV. Stiltches with two input and two output llnka
Switches with two Input and two output links we find In atages form 2 to log^ N. Two input links are connected to two switches of the previous stage and the output links are connected to two switches of the next stage.
The input process now depends on the output proceas of the previous stage. If the buffers that are connected to Input links of the switch are not empty, we get a request out of them. As we have only two input links, we can not get more than two requests in one cycle. The output process Is Just the same as described in section III.
Again we first analyse the stationary case with constant average rate. As we could sea from the previous section and as is shown in this, probabilities that buffers are not empty (In this case we get a request) are not changing with tine. Let n^ and be the
probabilities that the buffers from where Input links come are empty. Then we have the following probabilities on the number of inconing requests
n, . D,
"2 <1 -
(1 - - n^)
<12)
u = 3, 4,
As the path from the processor to the oeniory module Is completely random, requests Join each of the two buffers with equal probability. So we get
P^ = p] = pf = 0.5	(13)
Pq : Pj s 0.25
For the probability of m requests coming into buffer In one cycle we get

= 0,1,2 (1J|)
The balance diagraai for this type of switch is shown In Fig. 1.
The balance equatations which we obtain from the diagram are
0(0) (p, + p^) = ß(1) Pq
tn(o) + 0(1)> p^ = 0(2) pg 0(s) Pg = 0{a+l) pg
(15)
= 2, 3. bl - 1
How the aysten is already linear. We just have to add the normalization equation (5) and solve it.
For the time dependent system we use a similar procedure as In the previous section. Consider that we analyze the stage I. Then we have
n(0,t) = 1, 0(a,t) : 0, s - 1.....bl,
t : 0, 1-1
(16)
on the stage 1. The nuaber of requests in the buffer may grow maxlnaly by a request per cycle, because of maxlmaly two incoming requests and one outgoing request (except in the case where at the previous moment the buffer was enpty. But we only have a Jump from 0 to 2 and after that the length grows maxlmaly by one). So we get the following system
nCO.t) = [O(0,t-1) + 0(1,t-O] p^
0(2. t-1) P(j
0(2,t) = [0(0,t-1) + 0(1,t-U] Pg + 0(2,t-1) p^ + 0(3,t-1) Pp
(17)
a*l
n(e,t) = Ì 0(s-l,t-i)p k=a-1
a+1-k
s s 3. .... bl - 1
n(bi,t) = n(bi,t-i) (p, + p^) +
n(bi-i,t-i) pg
Equatations (17) are the aaae in the case where average input rate is time dependent and where la not.
V.	Suitchea with tvro input and ona output link
This kind of switches we find at the last stage. The input links come from the previous stage and the output links are connected to memory oodules. From each awltch we can reach only one memory module. The situation Is almost the same as described In the section IV. The difference is only that we have Just one buffer so all requests go in the same buffer. So we have to change the ooefficlenta p^ defined by (12) and (IM).
VI.	Sumary and eoncluaion
In the Table 1 probabilities on number of requests in the buffer for one switch on the first stage with different average Input rates are given.
average rate
3	0.5	1	1.5	1.9
0 0.75000		0.50000	0.25005	0.07687
Ì	0.21302	0.32436	0.28020	0.12169
2	0.03277	0.12260	0.19489	0.12635
3	0.00385	0.03779	0.11706	0.11699
it	0.00037	0.01091	O.O67S7	0.10594
5	0.00004	0.00311	0.03915	0.09569
6	0,00000	o.ooose	0.02258	0.08643
7	0.00000	0.D0025	0.01302	0.07807
8	Q.OOOOQ	0.00007	0.00751	0.07052
9	0.00000	0.00002	0,00433	0.06370
10	0.00000	0.00001	0.00250	0.05753
After 1 cycles the first request can come to the switch
Table 1
		average rate		
s	0.5	1	1.5	1.9
0	0.75000	0.50000	0.25086	0.08057
1	0.22959	0.38889	0.390ÌI7	0.1973'»
Z	0.01999	0^09876	O.2300U	O.ZOkZZ
3	O.OOOÜI	0.01097	0.08251	0.15007
	O.OOOOl	0.00122	0.08251	0.11028
5	0.00000	O.OOOlJl	0.02960	o.oeioi«
6	0.00000	O.O0QO2	0.01062	0.05955
7	0.00000	0.00000	0.00381	0.0^376
8	0.00000	0.00000	0.00137	0.03216
9	0.00000	o.ooooo	o.oooijg	0.02363
10	0.00000	0.00000	0.00006	0.01737
Table 2
In the Table 3 the probabilities for the switches on the last stage are given. All requests are generated with equal average Input rate.
		average rate		-
8	0.5	1	1.5	1.9
0	0.50000	0.02500	0.00000	0.00000
1	0.38889	0.07500	0.00000	0.00000
2	0.09877	0.10000	0.00000	0.00000
3	0.01097	0.10000	0.00000	0.00000
It	0.00122	0.10000	0.00000	0.00000
5	0.00014	0.10000	0.00000	0.00000
6	0.00002	0.10000	0.00014	0.00000
7	0.00000	0.10000	0.00125	0.00000
8	0.00000	0.10000	0.Q1117	0.00008
9	0.00000	0.10000	0.09958	0.00897
10	0.00000	0.10000	0.88784	0.99094
As we can see from the tables It Is obvious that as the input average rates grow to the maximum output rate, which Is 2, the buffers are more and more full, so also grows the number of rejected requests. Acknowledgeitent
This vrark was partly ■ supported by Iskra Delta Computers.
fleferences
[1]	Rettberg B., Thomas fi,, Contention is no obstacle to ahared-memory multiprocessing, CACM, £9(1986), 12.
[2]	Patel J.H., Performance Of Processor-Memory Interconnection for Multiprocessors, IEEE Trans, on CoQp., Vol. C-30 (1981), 10.
[3]Kruskal-C.P.,	Snir M., The Performance Of Multistage Interconnection Networks for Multiprocessors, IEEE Trans, on Coop., Vol. C-32 (1983), 12.
[4}Dia3 D.H., Jump J.R., Analysis and Simulation of Buffered Delta Networks, IEEE Trans, on Coop., Vol. C-30 (1981), i|.
[5]Thanawastien	S., Nelson V.P., Interference Analysis of Shuffle/Exchange Networks, IEEE Trans, on Comp., Vol. C- 30(1981), 8.
[6]BraJak	P., Designing a reconfigurable Intelegent memory module (RIMM) for Performance Enhancement to Large Scale, General purpose Parallel Processor, Informatica 11(1987), 1.
[7]Gelenbe	E., Mitranl I. Analysis and Synthesis of Computer Systems, Academic Press, 1980 London.
Table 3
INFORMATION DETERMINATIONS II	INFORMATICA 4/87
Anton P. Železni kar
UDK 519.72	Iskra Delta, Ljubljana
Abstract. This essay is the continuation of the previously published 'Information Determinations I' (2eleznlkar, IDI) and 'Principles of information' (Zeleznllcar, POI) and investigates some additional higher informational forms and Informational processes. In respect of determination from the informational point ot view, it deals with creativity, language, writing, reading, utterance, hearing, philosophy, ideology, dialectic, science, technology, aesthetics, ethics, rhythm, harmony, art, literature, crisis, capability, incapability, functions of mind, Information systems, social systems, and artificial intelligence. As close as possible, these informational forms and informational processes are investigated through the so-called Informational principles of spontaneity, counter-information. Informing, arising, embedding, counter-Informing, circularity, recurrence, parallelnes, and sequentialness of information. The choice pf very divergent informational forms (for instance, creativity, Ideology, technology, crisis, etc.> and their informational investigation shows appropriateness and suitability of informational comprehension of different subjects.
tStßCDlSffi^ It
Anton P. 2eleznikar I-jkra Delt-j, Ljubljana
Ißx y	rtgfBwiRjl: Ij <" I nfunndt. i on DeLeriiii na I i onü 1": 2e-
Uznikai', IDI) , L T rfillGcolS/ljllj ("Priiicie;- uf 1 nformaHon": 2elezni-kar, POI) è:Vfir.o(2)x/ b-fO^šTS^. ^x yt
ilJj^B^ttèUt.'7D b ^ mti'I ■ tSJBB^^^St»^fiT, &1Ü11, f Ig. S <Ii:. IXVlt. m, GS<It.	tm.	'JXA, SI
m, X^.	titti, m^j, )!PtEo|tta6È,	A
I8'JtS^B 7 a -b 7M i:' /AVa i ò hrz nò	^ fe--1:7. S^i^iftMffili t: Š
SÄti. S-tllB,	ClnFonning) , ÜjrÄ (arising) , if&ü^», S-ISfS
Sffi (cüLinLrii- inronning) . ÖSHISÜ, SÄ, ^'LTtSfficrj^trti^;JiLltS/fttt^^'.WJìt'^tSffi
zoi^òti.	CfjaìffiStt, -("rta
cj!,?^ /y^f^öSSIitJ^-S ž n tè: %.
Keywords. Aesthetics, art, artificial intelligence, asocial, association, autopoietic, beauty, bisociatlon, capability, circular aris Ing-embedding, circular-spontaneous Information, coding, cognition, cognitive, psychology, conmunicat ion, cortical process, counter-Informing, creativity, creative Informing, crisis, culture, cynicism, dialectic, diallctlcal principles, displacement of meaning, ethics, expertise, formal i sat ion, functions of mind, harmony, hearing, hermeneutics, hermeneutic-semantic translation, humor. Ideology, incapability, information, information processing, Information system, Informational arising, Informational circularity, Informational embedding, informational filtering, informational modulation, informational noise, informational path, informational parallelness, informational phenomenology, informational spontaneity, informational transformation. Informing, intelligence. Inventiveness, knowledge, language, language generation, language understanding, linguistic, literature, metalnformation, metaphorlc, metaphysics, methodology, mind, natural language, periodic, philisophy, problem solving, reverse information, rhythm, science, semantics, sensory Information, social action, social Interaction, social system, apealcing, storytelling, subinformation, technology, thinking, uttering, value, work of art, world-view, writing.
10
of creative thinking and creative behavior.
10.1. Creativity
Creativity is not nearly as mysterious as the genius view leads us to believe, but neither is it as trivial as the behaviorist view claims. ... Creative thinking must remain mysterious and unknownable. ... In the case of creative thinking, the myths explain the phenomena, the creative products of artists and scientists, by emphasizing unconscious thought processes, far-ranging leaps of insight, and some unique personal characteristics. ... what we believe about creativity is not based on hard data but Is more or less folklore ... The creative capacity that the ancient Greeks assigned to the mythical gods has in our era been assigned to the unconscious and to other exotic processes.
(H. W. Welsberg, CRY, 3)
Creativity is one of the most regular properties of Information. Creativity as Informing concerns spontaneous arising, embedding, and counter-Informing In an informattonally circular manner. Information Is creative in Its circularly shaped arising, embedding, and Counter-Informing. Creativity is semantlcally and hermeneut1 cally closest to the Greek 'polesls' and Latin 'cogltatio'. Informing of information is creative, whereas counter-Information Is the result of Informational creation.
Creativity Is circularly spontaneous process, which	Is	field-directed,	domain-
character I st ical ly or field-concerning oriented complexity of mutually dependent and circularly connected informational processes. Creativity is Individually structured and organized domain-subjected informational activity which results in creative behavior (thinking, handling, happening).
In comparison to intelligence, creativity Is a highly parallel informational process which Integrates various Intelligent processes, and has the capacity to Inform over and above different Intelligent, activities. Creativity can- Influence the arising of various Intelligent processes by Itself. Therefore it Is similar to Individual expertise, where expertise is understood as the highest possible state of individual capability In a specific domain or discipline. Individual expertise is to be understood as the highest point of a particular Intelligence where expertise Is achieved by passing through at least five developmental stages of,the performer; novice, advanced beginner, competence, proficiency, and expertise (Dreyfus, Dreyfus, MOM). In this way the performer achieves the state of expertise through learning, instruction, experience, Investigation,	and	counter-Informing.
Creativity is grounded in 'self, through Individual expertising, taste, and origin. Creativity Is no less than individually expressed metaphysics of a being, a processing within total information of a living being.
In this respect, creativity Is more or less a normal informational way of circular-spontaneous ly informational arislng-embedding (In the human brain), grounded in a specific Informational realm (background, field, domain, discipline, skill, art, profession). From the mythical point of view, creativity manifests and roots in the mind of a genius in the form
Although creativity Is understood as a particular and characteristic Informational processing of a living being in general, there is no reason to understand it also as an informational phenomenology In non-living or cosmic space. Theoretically, creativity can be divided in several ar1 s 1ng-embedding (parallel, parallel-serial, and sequential) informational subprocesses such as (according to a. w&ll&s, AOT) preparation, Incubation, illumination, and verification of the created subjects (CRY).
Creativity Involves novel combinations of Information and observation, investigation, and recognition of these combinations. Where contrasted to association, the combining of unrelated information is termed blsoclatlon. At this point, a question arises, whether an unrelated information exists at all? Creative information informs by connecting or blsoclatlng an Information tn which some Informational relations are not evident. It arises and comes Into existence by this combinational Informing. In this, process, connective information of previously unrelated informational forms and processes becomes regular information. And this Is the act of creat ion.
The comprehension of designing, planning, and discovering of procedures, by which problems will be solved or objectives achieved leads to the notion of creativity. Most of the functions In the informational and process-generating hierarchy of the brain are learned. Is creativity merely an Illusion which turns operational through informational recurrence of combining and discovering information, this is, through recursive procedure for discovering procedures? What kind of information do we have in mind when something new is created? Obviously, some information is taken from somewhere and then this Information Is processed by another Information Into a new Information giving it a new name or Interpreting It anew. Known Information is taken and changed into another, so that this process guarantees occurrence of a new InformatIon.
Real creativity when inventing a completely new Information from the Informational nothing is considered not possible. Information can only arise from given Information. The question comes from whether a complete creative act could be recognized at all. Information ot a being simply cannot be related to sensory input patterns which are entirely different from patterns known up to now, from patterns memorized or presented as cortical processes. Such sensory informational patterns will be received as Informational noise without meaning or sense, as strange information, or as possible hostile Information. This is the way of autopoietlc Informing in which' a being's total information Is molded when Information processing is taking place In the living substance.
philosophically, every living being is creative In Its entire behavior because Information can never be repeated In an exact manner. However, this Informational change is not understood to be creative if the arisen information Is not essentially shifted, bent, inflected, thus becoming different from averagely changed InformatIon.
Creative Informing is characterized by capability of hypotheses selection, generating realistic expectations, and by precise processing and evaluation In accord with some given rules. In this manner, creative Informing
is the art of a complex objective-oriented behavior. Most creative acts need habitual observation, Investigation and cognition, and yet unusual selection of voluntary objectives.
The Informational value of creativity occurs In Its rareness in the function of Informational process and In Informational (biological, life-concerning) advantages, which It carries, its rareness Is a consequence of non-creative Information processing in human mind. Creativity is oriented into advantage, usefulneaa, benefit, favourableness, and privilege by discovering new information which is useful for the being and its population.
As a case, creativity also offers the possibility of learning the inventiveness and that ot solving problems. A human being learns to be creative in a similar way as it learns adopting of several patterns of problem solving. However, creativity is also a coming, eruptive, and dominant information process, which carries the power of new cognition and development of inventive thinldng. No matter how, creativity remains a regular, circular arislng-embedding Informational process.
According to Maturana (AAC, xxvlii), it Is important to stress that a creative being, which is an observer, an investigator, and a recognizer, always is potentially antisocial. A being can simultaneously be a member of several social systems (a family, fraternity, club, clan, party, religion, army, race, tribe, nation), however, as a creative being It can create a metadomaln from.which It can see its participation In social systems that it Integrates. In this way, through observation, investigation, and cognition, through a being's counter-Informing, a creative being Is standing behavior&lly outside of social systems and may find them contradictory, controversial, so, that it may undergo social interactions that do not confirm these social systems.
10.2. Language
... in thinking Being comes to language. Language Is the house of Being. In its home man dwells. Those who think and those who create with words are the guardians of this home.
(M. Heidegger, BW, 193)
Language as phenomenology of thinking and behavior (speaking,	hearing, writing,
comnunlcating, etc.) Is at the same time a higher informational form and Informational process by which also the complex realm of Informational phenomenology can be governed arbitrarily. Language is founded in the capability of generating and comprehending regular and irregular (new) sequences of phonemes, words, and thoughts. The coding of arising Information can be achieved by different linguistic sequences.
Natural language as informational expression and as informational reception is a living, socially active and developing Informational process, which carries and generates a variable coding (interpreting) of the meaning. Linguistic processing is a fair example of information processing concerning the circular arlslng-embedding principle of symbolic meaning and formal expression. Arbitrary Informational phenomena can be described and expressed In linguistic forms and performed by linguistic processes. However, specific portions of
Informational	phenomenology cannot be
adequately or sufficiently precisely expressed by means of natural language. Such are several forms of sensory Information (concerning sight, hearing, smell, taste, or tactility), experience and other Information. In this respect,	language remains particular
informational form, 1. e. particular Information, or sub-information.
In spite of the fact that It is possible to express Information by language. It Is not possible to express Information In the way as informational forms and Informational processes appear In a living substance. Expressing Information always Involves informational transformation, whereas through language, thinking, a notion, a form, or a process are translated from their original informational appearance In a specifically linguistic manner. Under language an external and environmental, contextual,	and graimiatical-semantIcally
determined natural language can be understood. The Internal, intellectual, mind-concerning Information includes imaginative, picture-like, acoustic,	extraordinarily	shaped,
Individualized properties which are determined by Internal, Individual informational patterns, connectives, relations, processes, and as such cannot be expressed Immediately and adequately by means of a natural language. However, the use of language Is a form of behavior, a means of social action, which are characteristically related with thinking and development of the possibility for future expression of Informat i on.
Language is destined tor specific social Interactions, such as awarding, refusal of punishment, or achieving success In the social hierarchy. Linguistic behavior enables Infortnlng, counter-Informing, communication, receiving of reverse Information and counter-information from the environment. This kind of behavior Is primarily a kind of social interaction, populatlonal culture, and valid behavioral forms, which concerns highly intelligent and creative being. Its conscious and Intelligent Information.
Language generation and language cognition also root in Informational processes of rhythm, periodicity, and patterns of harmony which appear on different Informational levels and in different social situations. Already very simple, however well-structured or metaphoric linguistic entities can influence the creative power and cognition of language generation and language understanding.
Language reflects a being's Internal information. Informing of this Information in itself. Creative, meditative, strategic, contemplative and other kind of thinking Is internally arising and linguistically structured Information. It Is expressed and understood In the realm of a being's total Information by means of Its natural language. Language possesses its Informational depth and penetrates as informational phenomenology into arbitrary depth of philosophy and poetry. Through language It Is thought about Information and language, up to arbitrarily possible abstract Informational levels. In some examples, language reaches the power of information by which it Is coming Into existence In an Information-processing manner. A being with language capacity and with its in-Itself-language Is continually speaking. It gets on the way of this language through acoustic articulation, through listening to other speaking beings, and through thinking by means of linguistic categories. Internal.speech and being's (voice-articulated) speech are
sprouting Internally, developing language, and procesalng It. Being's information processing from one kind of complexity Into another, from less to more complex and vice versa, Is setting on the way to a new meaning and new symbolism. Language Is the home of Informational abstractness, symbolics, meaning, semantics, hermeneutlcs, cognition, understanding, and other 11nguls t ic-Informat1onal modeling and behavior.
Language means speaking of language and Its processing In Itself. Information means circular ari sing-embedding of information. Language is circular arislng-erabedding of language. Several Informational forms of a being are language-characteristic, verbally shaped, or processed by linguistic means on different informational levels. These forms can have	language-characteristic	codes,
abstraction,	symbolism,	meaning, and
understanding. Mastering a language means emphasizing	lingual abstractness and
recurrence, developing lingual thinking and counter-informational processing up to the highest level of behavior-generative hierarchy. It involves exploration of the arising of counter-language and lingual invention, where new semantical and hermeneuti cal categories are constructed. Speaking in verbal language means the arising of language-shaped information, the development of linguistic information processing. Language develops Being of language In the direction of higher or lower diversity, variety, complexity, and I nforma>t lonal performance. Language is at-the-top of a being's informatlonalness. It is a tool, expression, cause, and consequence of Intelligence, the being's creativity. In the cortex of a human being, language is represented as a form of particularly distributed Information processing which is linguistically coded,	abstracted, and
symbol 1 zed.
To some extent, language is informational self-sameness when it speaks about Information, supposing that the information Is comprehended by lingual categories. Language develops information about information. Language develops information about language when language speaks language and speaks about language. In this case, the way to Information is on the way to language where a being's informational processes of a being are verbal. The way to language by language is also the way to the being's Information by information.
Language processing is an essential faculty of being's	informationalness,	which	Is
distributed, essentially rhythmic, harmonic, filtered, and modulated by other Informational processes (for Instance, emotions) of mind. A being Is a developing informational. Ungual system, where language processing is embedded into the being's informatlonalness. Language processing is only one of possible modes of information processing. A being's Imagination can be characteristically verbal, processed linguistically. On several Informational levels language processing can be semantically and hermeneutlcally evened to a being's Informatlonalness. Language is an informational model which models' information, and simultaneously represents and processes information by linguistic means. In a similar way, as information is on the way to information by counter-1nformatIon, language Is on the way to language by counter-language, which Is the arising meaning, expression {concerning linguistic	form),	and
understanding.
Within "classical" cognitive psychology (CPY), language Is defined by notions of semanticlty, arbitrariness, discreteness, displacement, productivity, Iteration, and recursion (CPY, 322). One of the basic questions of language remains how linguistic entitles, units, and fragments are coming into understanding, how the processes of apprehension and comprehension are arising, developlngj coming into existence within a being and in a being's population (environment, culture). In this respect, the approach to understanding within the being's total information is. hermeneutic, focused upon Internal Information and on the internal process of speaking. In contrast to hermeneutic information, the approach to understanding in populational environment or culture is semantic,	informing linguistic	facts
externally, by external processes of speaking cultural languages.
10.3. Writing, Heading, Utterance, and Hearing (In verbal Language)
... Language Is not a defined realm of the speakabie, over against which other realms that are unspeakable might stand. Rather, language is all-encompassing. There Is nothing that is fundamentally excluded from being said, to the extent that our act of meaning Intends It.
(H.-Q. Gadamer, PHH, 6T)
On the one hand the informational forms and Informational processes, like writing, reading, uttering, andhearing all concern language, I. e. live linguistic activity, thus covering acts of Internal and external speaking. On the other hand the said activities also involve Interaction and reaction of speakers (writers) and hearers (readers). Writing, reading, uttering, and hearing are social activities which concern social action of living beings, social interactions within beings' populations.
Writing a text in a (natural, formal, professional) language concerns creative internal shaping of meaning, planning a sequence of meaning as a hermeneutic form and afterward transforming this hermeneutic skeleton Into an adequately granmatically and semantically shaped result, displayed in an appropriate external, linguistic, or cultural form. This process of hermeneutic-semantIc translation causes a displacement of meaning between the initial hermeneutic plan and the final semantic result. Displacement of meaning is the difference occurring between a being's total	information	(Its metaphysical
presentations) and external language-ruled Information. The being's metaphysical fragments, which represent the hermeneutlcally shaped content of guiding ideas are through writing carefully translated Into linguistic (granmatlcal and semantic) entities. However, this is only the first step of the writing Iteration.
Linguistically written text is now presented for further creative Improvement of its meaning and its form. Semantics of the written text can now be analyzed, observed, investigated, and recognized by reading the text. Through these processes, the text may be transformed into hermeneutlcs, giving reason to change, complete, add, or reduce semantics of the given text for the second time. According to this, changes, insertions, additions, or dropping of some textual fragments may occur. Such
Improving iterations of meaning and form can continue until writer is getting satisfied with the final result. One has to bear In mind that the Informational process of writing concerns planning of the meaning on the hermeneut1eal level, translating this meaning Into graimatI cally written text on semantic level, reading this text by translating It on hermeneutic level, correcting the obtained hermeneutics, translating hermeneutic form into semantic one, etc. In this way, writing and correction (Improvement) is an Iterative informational process of text creation.
Writing and writer's skill is a complex informational goal-directed activity, • a particular intelligence, which was acquired through laborious learning, repeated exercise, sell-correctlon, self-mot1 vat 1 on, sensitivity, perception, cognition, mastering of meaning, style, gramnar, composition, expression of writer's thinking, imagination, and factual knowledge. However, no one text is written in such a form that it would not call for Improvement, in order to Increase its linguistic and informational valuableness. In this context, informational valuableness has the meaning of the power which is to generate Intended Informing In the reading auditory, to influence this auditory 1nformatlonally (ideologically).
Reading a text is an Informational process of transforming the 'semantics of the text Into hermeneutics of the reader. !n the process of reading, the linguistic information Informs a reader's total Information by counter-Information which results Into hermeneutics of the given text. In process of understanding, the reader can repeat reading of problematic passages until he obtains the information that is hermeneutleally satisfactory. In the process of reading, the reader's information informs and counter-informs, so the reader is getting Its own new information, its counter-hermeneutics. By reading, an informationally sensitive being can be Informed and counter-informed, for he performs its own Informing concerning the semantics of a text. Of course he develops it in a new hermeneutlcal direction, where the "coming-to-understand" arises. Reading is the individualized process of Informing where text Is Information, giving clues lor the context of a being's entire Information.
Utterance (speech) is the act of speaking (voice articulation) in the context of social Interaction (hearing). Internal speech (utterance within itself) has a form of dialog, of a social interaction with a being itself, with imagined speakers and hearers. In speech and in hearing of the speech, hermeneutic interaction occurs among speaker and hearers. In this social interaction, speech Interferes and depends from the reaction of hearers.
In speech, the speaker informs in several ways: while speaking he plans the meaning of speech fragments as his hermeneutics, articulates this hermeneutics In a linguistically appropriate form	through	several	Informational
trans format ions	(hermeneut ic-semant io,
semant 1 c-gramat 1 cal, 1 ingul st 1 c-artlculatory, etc.), receives reactions (linguistic, gestural, mimic, sighing, etc.) of hearers, and Interacts with adaptation and correction of his speech activity. In this way, hearers can influence the planned utterance of the speaker, signalling and communicating their hermeneutic agreement and disagreement to the speaker. Only a poorly structured and organized speech lacks such a social Interaction of speaker and hearers.
Speech gives evidence how 1 hermeneutic backgrounds of are evened to the appropri not be to far away. If not received as Informational wishful speaking-hearing hermeneutic backgrounds o different, speaker has to bright semantics, which Is a broad or mixed auditory will use general accepted utterance and convenient semantic forms.
11
t Is Important that speaker and hearers ate extent and must , the speech will be noise without a interaction. If t hearer are very use a sufficiently externally valid for a 'good' speaker cultural forms of gramatical and
11.1. Philosophy
... If philosophy is the scientific construction of a world-view, then the distinction between "scientific philosophy" and "philosophy as world-view" vanishes. The two together constitute the essence of philosophy, so that what Is really emphasized ultimately is the task of the world-vIew.
(M. Heidegger, BPP, 7)
Philosophy is a linguistically based (usually literary) form and a linguistically based process of particular	(philosophical)
Information with a specific, uttered, written, esseyistlcally structured, logically organized, and thought-free orientation. It Is dedicated to	some	selected,	philosophically
characteristic objects (questions, problems). Each philosophy has Its own system of Information and its own way of Informing. The orientation of a philosophy Is Its semantical (objectively selective), comprehensìonal (philosophically systematical observation, Investigation, and cognition), and stylistlcal (characteristically organized and formalized) task. In this way, philosophy Is a culture-memorizing, a being-investigating, and a populatlon-concernlng i nformat ion.
In philosophy, poetry, literature, and Ideology, Informational processes can come Into existence informatlonally and counter-Infoùnatlonally through their particular respectability. They are established through bethinking,	sensing, grasping, and
comprehending occurrences, which are separated from the imperative of necessity and can be arbitrarily intormatlonally perverted. The Informational processes are variably playable, truant, magically oriented into the contrariness of the real, unreal, rational, irrational, free, forced; they are involved into the play of themes, images, sounds, colors, notions, events, models, Into self-bethlnklng as a response concerning similar environmental absurdity. Furthermore, they are characterized by the autorefiection of counter-reflection. Is this kind of information autopathologi cal or Is It a healthful discourse which is also self-contrary, evidently autoabsurd, autocorrectIng, in many cases, however, only raataphorlc, paraphrasing, poetically expressive, and philosophical?
Since the ancient understanding, philosophy had been a friendly, methodical Informing, reasoning, wisdom, and basic cognition without a purpose. It could have been explicated as an Information theory in itself. Throughout the history and cultural development, philosophy
became more and more formalized, structured, purposely constitutional,	specialized,
sclentized, artistic, and ideologicai. Eventually it turned Into a systematic, purposeful, ut i 11 tar Istic, metbodological reasoning about being, nature, cognition, mind, life, history, and about Itself alone. From the original, friendly reasoning It developed also Into human hostile, external, forced thlnltlng. In the history of mankind, philosophical orientations were confronted, prosecuted, and prohibited as information of unbound thinking. In this respect philosophy appears to be only a particular, cognitlvely oriented historical phenomenon of information and counter-info rma t ion.
By Its informational nature, philosophy enters the realm of abatractness and theory, where modeling, changing, systematizing, and artificial thinking are exercised, due to Inherent philosophical capacity of performance. In this capacity lies its developmental-progressing and thinking-complex value. Philosophy offers possibilities of divergent and heterogeneous discourse about arbitrary questions of being, existence, essence, and within these questions also about culture, science, technology, economy, democracy, applying philosophical modeling and prevision of answers. Information which is generated by philosophy in this way, can be transformed into counter-Information down to the arbitrary cognitive depths.	In this respect,
philosophical information basically contributes to cognitive Informational processes In science, technology, and management.
Philosophy is the principal informational activity of philosophers (from Plato to Heidegger) and their writings are Informational forms of philosophy. According to A. Flew (DPH), "... philosophy is a matter of standing back a little from the ephemeral urgencies to take an aphoristic overview that usually embraces both value-commitments and beliefs about the general nature of things." Responsive philosophy can always be understood as a being's therapeutic and developmental counter-Informational phenomenon of its creators, producers, and followers. Philosophies can inform therapeutically and developmentally against pathological informational syndromes. A sound being or a sound population understands Ideology as an essentially reduced, specialized Information of already acute, mass Incapability of brains (illness, lesion. Injury, blocking, excessive artificial or natural excitation), whose consequences are pathological, without expected therapeutic etflcaey, descending Into pathologic compensation and momentary balance of survival.
Philosophy of Information Is nothing more than a particular information about Information, an expressively recurrent Informational process, counter-informal Ion about information, which broadens and Innovates comprehension and Informatlonalness of Information. [n any case, philosophy is only a pre!nformatlonal process of general informational arising, since it Is only a particular form of informational and counter-Informational regularity, naturalness, reasonabl11ty,	cognitivlty,	and
progressIveness.
For Martin Heidegger (BAT, 487), philosophy "Is universal phenomenologlcal ontology, and takes its departure from the hermeneutic of Dasein, which, as an analytic of existence, has made fast the guldlng-line for all philosophical inquiry at the point where It arises and to which it returns." ...' Counter-Information especially embraces philosophy as
sufficiently complex informational process, as sufficiently broadened investigation of phenomenology of beings living in the beings' environments. Information departures from explaining life processes or survival, from analytics of survival, from existence of living and non-living, from information itself and via counter-information also	returns	to
informat ion.
Basic questions of philosophical ontology and anthropology cannot surpass informational essence, informational Being of a being, and Informational liveliness which accompany beings from their conception to their death. A being's informational origins and sinkages are autopoletlcally unlimited In thinking and In behavior of beings.
11.2. Ideology
To learn	Ideology	means to step Into
Ideological circle.	To understand ideology
means to observe	informatlonally the
ideological	behavior.
Hitherto, it Is not possible to give an exhaustive, sufficiently determined and Informatlonally we11-embedded notion of characteristic Ideological forms and processes. In this section, only a beginning into the action of opening of informational understanding of Ideology will be given, mostly through putting questions being on the way to Ideology, What is ideology as an Informational form and informational process In a being, population, or society? Which Informational principles does It follow? What Is idelogy by Itself as an Informational principle? In what way Is It characterized through Informational arising, embedding, counter-Informing, and clrculatI on?
Ideology will be the label for a characteristically structured, organized, and semantlcally oriented informational form or Informational process. In general. Ideology as Information violates the most fundamental principle of Information - the principle of spontaneity (Zeleznikar, POI). Ideology concerns the being's everyday experience through Its unpleasant,	disagreeable.
Intruding, officious, and manipulative Influence. Ideology is essentially distinctive from the natural being's metaphysics or from a freely arising population's ontology. It la very close to the informational artifact or to artificial Information, which has Its own way of informational propagation and circulation. Ideology is, for example, the Informational product of an archaic culture, of a civil-backward, hindered population, which Is marked by antifreedom. Ideology stands for ethically, morally, scientifically, and philosophically bound learning, doctrine. It Is Is an evident example of non-spontaneous, strange, fictitious,	objectively false, and
inappropriate Information of the modern epoch.
In general, information is a free, spontaneous, formally unlimited and in no way formalized phenomenon. Every informational formalization excludes some substantial meaning from an informational domain by introduction of a specifically formalized language (Ideological, philosophical, scientific) in a being's realm of thinking. Ideology is characteristically formalized, oriented and limltedly cultural Informational form of population or individual
being. It may be treated as I ntorniat ional process concerning thinking and behavior of beings. From this point of view, the question of Ideology versus information can be questioned upon if It is to become the topic of this section. This questioning will produce interrogated and newly ar 1 s ing-ques11ons which will all recurrently Influence making questioning about the essence of ideology.
It seems that in the realm of a being's thinking. Ideology qua information is formalized in a specific way. Through this thinking, this formalization Is extended also into the realm of the behavior of human beings and their culture. Each formalization, which also holds true for the Informational one, is restricted in the spontaneous Informational arising, embedding, counter-Informing, and circulating (recurring, paralleling, and sequencing). In some cases, this Ideological formalization is presented culturally in a linguistically oriented, written, spoken, or behavioral form as a process of traditional, fraternal, clanlsh, or revolutionary (radically new) orientation. Ideology Is developing out from Its magic, thus conducting and managing a body of Ideas, embedded In the being's metaphysics and^^or in the population's ontology.
Ideology is an ideally oriented metaphysics or ontology, possessing its conducting kernel. In this respect, Ideology has Its Informational grounding, which can be mutually different to the extremes. Extremes given actually represent a list of synonymous antonyms, such as:
ax Iomat i c
magic
suicidal
genti listič
nihilistic
r&tionaUstlc
logical
hard-scientific caste-feel Ing trlbe-1 lable
dogmatic, Indoctr1nat ing, fratricidal, homi cldal, pos i 11 v 1 St i c, behaviorIstlc, ideological, art i St Ic, class-hostile, Indivi dual,
etc. Ideology has its own approved system of rules for the so-called ideological deduction, induction, argumentation, and inference of ideological truth and untruth (ideologically a false Information).
Various Ideological • forms and Ideological processes can be understood, for example, as particular forma of Informatlonlsm, Inforralsm, Informat1vlsm, Informatism, )nformat1 alsm, and Informatlzationism (described In paragraphs 1.4, 2.3, 3.2, 4.3, 5.3, and 6.2). Linguistics (or formalization) of an Ideology can be determined more exhaustively, restrictIvely, paternally, radieally, or when led to the extreme In the form of a totalitarian Ideology. It is more Indefinite, pragmatical, or Ideological-independent when concerned with subtle, concealed, underground, or apparently non-manipulative Ideology. As a totalitarian code system, and through this as a total form of a being's thinking, ideology governs a being's behavior, not only through a being's restraining but also through the ideological forecasting as behavior it Is Ideologically expected. In this way, any Ideology Is being vitally approved or verified as a system of truth and through this, it compells beings to perform ideological behavior.
However, any ideology Is only an Informational form or inlotmatlonal process, so that the arising principle or counter-Informing of Ideology qua information is preserved. In the informational context of a being's total
Information, ideology is appearing as a hardly ever Informationally resolved counter-information producing Information for everyday life. The arising of information In an Ideological environment Is the governing Ideology itself. Further, It seems that ideology, which In some broader Information-embedded specific information Is embedded In its own grounding. In such a manner the being Is self-sufficient, self-producing, self-preserving, self-referencing, and autopoletlc In an Ideologically specific manner. With regard to normal informational arising Inside being's thinking, ideology appears as a characteristic counter-Information, by which the environment Is exerting pressure on a being in the direction of an Ideologically ordered, oriented, non-dlverae, non-alternatlve, non-spontaneous thinking. Through this it keeps the being's ideological, non-natural, artificial, counter-feint, pathogenic. Imitative, mimicking, or generally deceiving behavior. Ideology is an environmentally supported movement of Information which pertains to population and society In general. Obviously, Ideology Is Informing in a specific, informationally and counter-informationally blocking way.
Ideology Is self-perpetual, gaining Its own orientation In amplification, maintenance, and In Its perpetuity. This is possible because ideology la concerned with counter-information, which is a trivial (Ideologically well-structured) form of counter-information, the respective Ideological realm Itself. In this context, the self-perpetuity has the meaning-of Ideological, of ideologically true, of non-spontaneous arising, which Is governed by the perpetuity Itself. Evidently, ideology is non-Intelllgent, non-creative, because In ideology. Intelligence or creativity as a phenomenon of Informational throwness Is superfluous, not needed as an informational process.
In many cases, the so-called traditional background is Ideological to some extent too and rests on a being's (ontic) metaphysics or on a population's ontology. A traditional orientation as Informational background Is cultural. In fact ontlcally outward, however, a being feels it as it It were its own, as to be In a being dwelling orientation. Therefore, a being defends the tradition as Its own information, for which it feels as belonging to itself, as It having been co-produced by a being itself.
How Is It possible to characterize Ideology qua information? Mbeit, the following list of characterizations Is not completed since it only supports further questioning applied to Ideology: ideology as Information is non-autonomous,	alternatively	blocked,
ideologically controlled, filtered, and modulated in its arising or In its multidirectional-arising, ideologically structured and organized, topologically unified, iterating In itself. It Is further ideologically autopoletlc, but generally non-poietlc, dogmatically embedded and ideologically stable, with a being's influencing and supervising. Ideology at the same time covers Ideal, axiomatic, simplifying, and Utopian features. It Is: Ideologically gaining, amplifying, and maintaining;	Informationally	reduced;
semantically hidden, converted or transformed; Informationally or spontaneously non-parallel, ImpractIcally oriented, unfitted for lite (inanimate, lifeless, Immature, unhealthy), etc. Ideology Is a body of characteristically limited information (ideas) ot a being, of a population, or of a culture representing assertions, theories, and aims that constitute an Ideological program, directed thinking and
uniform behavior. Ideology Is structured and orgftnized hierarchically. Its highest, supervisory level Is the ideological kernel, which might incorporate different informational forms as hatred, hate, hostility, fear, anxiety, destruct1veness or destructive irrationality, archaic or fraternal terror, suicide, fratricide, homicide, genocide, envy, revengefulness, Incapability, greed, avidity, covetousness, etc. The kernel of an ideology ia represented by the so-called avant-garde of a given ideology, where this avant-garde is evidently metaphysically blinded, emotionally founded, bacl(ward, and possibly biologically unsuitable and Jnformationally (cortlcally) pathogenic.
Ideology is characteristically periodic and In itself closed informational phenomenon, which repeats Its patterns of thinking and behavior In an exclusive, one-way, one-sided, biassed, partial, un]ust, unilateral, lop-sided, top-heavy, single-track, unitarian, uniform, however rhythmic, harmonic, and thematic manner. Especially, because of Its clearly forecasting possibilities, ideology is performing beings' harmonic, informing also in a killing, apathetic, indifferent, narcotic, subduing, restrained, consciously excluded, and repressive way. All these ways of Informational filtering and modulation give the ideology a fictitious, hypocritical, deceptive, and garbled impression of survival safety within its own protection. Thus, tdelogy is established in self-protective and development-stable way. In the human brain. Ideology Is blocking, preventing, obstructing, frustrating, and artificially separating the left hemispherical Informational activity from the right one, thus destroying the left-right Informational parallelism, connectedness, mutuality, and harmony.
Ideology seems to offer mentally and behaviorlstlcally passive encouragement and turns out on the other hand to be connected with an indefinite feeling of delight, safety, rewarding, efficiency, surviving easiness, and self-motivation which perfectly suits a properly incapable or disabled individual. Ideological feeling changes and varies the Internal world model In the direction of monochromic beliefs, faith, and expectations, which are being harmonically realized and approved, although not using the extrapolation, retrpspectlon, and Introspection concerning the future posslbllties or rewarding plausible forecasts for the future survival. Ideological feeling and will determinating Ideological behavior are conditioned with the Internal Ideological state of a being's central nervous system and with the Ideological environment or culture, which generates and modifies this internal state. The Ideological will selects Ideological behavior, its patterns and generators, which are activated in an Ideological environment and In social hierarchy. They Incorporate Ideological habits and customs (relations, ceremonies, myths, taboos) as well as strategies and tactics of Ideological behavior (Ideological speaking, fai thfulness-and-obedlence	approvIng,
uniformness, ideological enthusiasm, respect for ideologically prohibited and repressive Justness). Ideological behavior makes use of ideological expertness (fluency of ideological feeling, elucidating, simplifying, scientific contemplation,	and manipulation), and
Ideological reflexes (adaptability, submission, docility, humility, servility).
Patterns and generators of ideological behavior start and maintain the world model, the world-outlook of an Ideology-subjected being. This
model is based on Ideological beliefs, mindfulness, and remembering. An ideological world model, which is a wishful and expected Image and Imagination of an ideology subjected environment, Includes the categories of spiritual, social, and physical entities. The spirit of an ideology is its Idolatry, guiding Ideas, ideological doctrine of salvation, an essential theory of the given ideology including entire Ideological abstraction, symbolism, language, sense, and consecration. The sociability of an Ideology roots In an ideological framework, which Is Ideologically correct, and provides Ideologically equitable relations. These relations are reflected from the Ideological environment only. The physics of such an informational model Is, for example, dialectically	ideology-materialized or
Ideology-idealized physical phenomenology, where Ideology Interferes in a totalitarian manner with thinking and with modeling of the physical world.
Sensory processors which are receiving Information and sensations from the ideological environment, are under the filtering and modulation control of the Ideological world model. These Ideologically abstract sensations are in a perfect, prodigious, rhythmic, and harmonic agreement with beliefs, expectations, and forecasts of the world Inner structure and generate feelings, emotions, and values (in fact non-values) of contentment, delight, happiness, safety, and success, but also their controversies. These feelings which are virtually satisfactory do control the will for a rhythmic and harmonic selection of the behavior	according to	Ideological
circumstances. In this way, the ideological circle of happiness, safety and their controversies Is Joined into circulation. In a loop of ideological feeling. In this respect. Ideology Is an evident example of a particularly	composite	informational
circularity (recurrence, parallelness, and serlalism).
The objective of an Ideology Is to suppress the spontaneous arising and spontaneous embedding of counter-information by which the Ideology is observed, examined, and recognized. In the framework of an ideology, this counter-Information is promulgated as an Information of the hostile environment, because it causes the decay of ideological kernel and through this It results In decay downfall of Ideology Itself. For an Ideology, every spontaneous arising of Ideology-related or even Ideology-non-related counter-1nformatIon is dangerous and must stay under the systematical control of ideology 1tself.
In the realm of an Ideology, the entire phenomenology, changing of the world and of Information qua v 11al i za 11 on, qua renewing is comprehended only In the form of vaccination of the existing, governing ideology. It is performed by usage of Ideologically approved rules of thinking and behavior. In this typical looping and ideological harmony, ideology meets. Its informational regression, extemporlaatlon, blindness, diminishing, and decline of Its organism, functionality, structure, and organization. Sooner or later, every Ideology must confront with consequences of its artificial activity, atrophy, changes of Its age, decay, degeneration, and involution. Because ideology qua information does not allow the arising of a spontaneous being's counter-information (natural breakdowns), It remains Being to Itself without an essential informational arising and vanishing. In this context, ideology is à'n unnaturally alive, artificially Inhibited Information.
As a being's Informational process, ideology creates its characteristic world model, which is being supplemented through ideological behavior and also through ideological Imagination ot a being. Are there two -world models in a being, the first one in the form of an authentic imagination and the second one as Ideological behavior? What is the model duality creatcd for, besides the being's authenticity? In general, ideology is an imposed outside informational process, belonging to other beings; this process Interferes with authentic informational subprocesses. The feeling of intruslweness Is more or less ontleally conscious, destined, existential and a being accepts it as an inevitable disturbance, as an indispensabl1 ity and inconvenience or as a pretense and bother. This informational model duality creates the feeling of threat and fear, which by means of will select a counter-feint, artificial, and forced ideological behavior which is lacking liberty and variety in a spontaneous bethinking. A being selects the permissible behavior through a forbidden Imagination. Ideology Is a phenomenon of Informational contrast, of the absurd of simultaneous affectation and spontaneity, where the spontaneity Is imaginative and affectation is behavioral.
A typical Informational side-effect of the ideological repression Is the so-called ideological cynicisin. If a being feels an Ideology as a distortion, extortion, punishment, and torture, the being responds to ideology by moclcery and revengef ulness simultaneously,	individually,	however
populatlon-typical, and expressive, but population-diffused. As the reply to ideology, this consciousness is destructive, untouchable by Ideology itself, private, ideologically non-convertible, generally skeptical, illuminated Into counter-Ideological consciousness, which Is falsified, nihilistic, although it Is rebelllously and reslstantly Imposed. This cynical consciousness Is in contrast to pressing, to rational imperative, it is' not capable to be on the way to this Imperative in the context of Ideology, because it Is a radical irony standing outside of the senses oX necessity and survival. An Ideological cynic ts like a ruler and resembles a subdued, malicious realist. He Is like a socially integrated antlassociate, cynically socialized, lost in his own and populational cynicism, tdeological cynicism Is a degraded consciousness, being ill from repression, for which it has to accept ideological relations, not believing in them, however being forced to be surrounded by them, and at last, for which It must perform Ideologically distorted actions. [delogical cynicism is an Instructive example of a non-spontaneous,	informationally	repressed,
however, individually perverted, protectively reactive behavioral process.
11.3. Dialectic
Dialectical reasoning Is only one of possible forms of counter-Information, which could arise by circularly spontaneous Informing, counter-Informing and embedding of information.
Let now dialectic be illuminated as a particular example of an informational form and an informational process through the following questions: How is dialectic embedded in philosophy and Ideology? What is dialectic as a world-view? What is dialectic as Information and how can it 1nformationally embed Information as information?
The notion of dialectic comes from the Greek 'dialektike tehne' and has the meaning of skill performed in dialog, discussion, or discourse. Its very first meaning concerns intelligence of discourse or art of social interaction in the realm of language. In this sense, at its appearing, dialectic Is the label for art of discussion in processes of questioning and answering.
Through history, dialectic as a principle of philosophy and Ideology develops from Socratlc, Platonic, Aristotelian, Kantian, and Hegelian platform to Marxist platform, that is from discourse, supreme knowledge, reasoning from premises, transcendental reasoning, and logical pattern that thought must follow (thesis, antithesis, synthesis) to the pattern that thought and reality must follow (subjective and objective thinking). In Its late development, one of Its supreme orientation or world-view marked dialectical materialism becomes a metaphysical doctrine or social Ideology which asserts the matter, motion and development of matter as fundamental. An informationally parallel world-view arises as historical materialism, which aa Information deals with developmental laws of human society and thought. Dialectical materialism opposes Idealism stressing that matter is not a product of mind, but that mind is a complex product of matter. In this manner, the dialectical doctrine distinguishes materialism and idealism explaining dialectical principles (rules of dialectical inference) as transformation of quantity Into quality. Interpretation of opposites, and negation of negation.
How can dialectical methodology be understood informationally? Certainly, dialectical principles of thesis, antithesis, and synthesis can be included in principles of information, counter-1nformatIon	and	informational
embedding, respectively. Furthermore, dialectic as a movement and development of things and thinking coincides with Informing and counter-Informing, respectively, however, not In information-embracing entirety.	Counter-
intormation, for instance, is not only an antithesis, but everything which arises, which comes into existence spontaneously from information. Also, Information Is not only a thesis which corresponds to a dialectical category of consciousness, but a phenomenon of free Informational arising in a being's consciousness and also outside of any consciousness. Infortn&tIonally, the dialectical circle of thesis-antithesis-synthesis Is a part of informational circularity (recurrence, parallelness, sequentlalness) of Informing, counter-Informing, and Informational embedding. In this Informational circle. Informational arising as circular informational spontaneity Is stressed as the basic principle of various Informational developments. Dialectical synthesis concerns counter-Informing and embedding simultaneously, in this respect, some essential differences between dialectical and informational approach can be observed, and It may happen that dialectical principles can be broadened Into informational ones in an easy, Informationally natural way.
Let now Information which was developed as a notion of information and a notion Illuminated through previous examples In this essay (through various examples of informational forms and informational processes) be thrown into the realm (informational system) of dialectical	discourse.	Dialectic as
dlalectlcally embedded Informational system of materiallstD, which follows its own rules of Inference now comes to the conclusion that the proposed notion of Information in this essay is
nothing more than Idealism and/or metaphysics (t'his already are the necessary dialectical-Ideological etiquettes of triviality of a notion) .because it stands outside the correct and known inferential realm, and maybe, cannot be "directly" comprehended through basic materialistic principles (for Instance, materialistic particles, materialistic forms and iTialerial Istic processes of reasoning), informationally, metaphysics is the label for the total (entire) information of a being (as determined in section 8.4). Informationally, Idealism is a form of philosophical and/or Ideological	informallonalism.
Materialistically,	idealism	is	the
philosophical counterpart of materialism which does not enter into the materialistic realm as a	materialistically legal inferential
subsystem. Mater t al 1 s t ically, metaphysics opposes' dialectic, for metaphysics is a materialistically non-movable,	static,
idealistic doctrine. In this context, evidently, the exclusion principle of materialistic reasoning takes Its philosophical and/or Ideological power.
11.4. Science
... Heisenberg argues that contemporary thought Is endangered by the picture of nature drawn by physics. This danger lies in the fact that the picture is now regarded as an exhaustive account of nature itself, so that science forgets that In Its study of nature it is merely studying Its own picture. ... Heidegger, however, sees the picturing of nature as a necessary feature of scientific thought. Yet both would agree that so long as science Is allowed to speak ex cathedra, other modes of thought remain Impossible.
(H. Alderman In HMP, 41-4Ì)
How science is revealed as phenomenon of information and how does scientific information arise within Informational process of science?
Science concerns knowledge which is acquired or gathered	sc Ientific-methodologlcal 1 y and
ordered sc Ientific-systematI cally. In this respect, science as a discipline concerns Its own scientific circle of Its own methodology andof Its own systematlcs. Science circularly informs Its own, scientifically legal or scientifically	permissive	information.
Scientific Informational arising, counter-Informing as a scientific contributing, and informational embedding as	scientific
(ac)knowledge(ment) are scientifically and disciplinary bound to the valid scientific system of truth and untruth. Science is informational process by which the scientific knowledge is produced through criteria of scientifically methodological truths. Knowledge generated by the corresponding science and scientifically generated methodology both represent the so-called sc ientifi cally embedded legal information, from which new scientific achievements can be Informed and counter-informed.
Knowledge and methodology are the very valid informational foundation (background) of a science. Scientific Informing proceeds from this knowledge base by application of methodological rules. Thus, science becomes a formal system which is scientifically bound by its knowledge and rules of permissive inference (deduction, induction, inferential Informing). Science investigates its truth and its own Investigation; by the said methods It deals
with discovery of truth, rationality, reliable prediction, generalization from sufficient collection of data, theoretical construction, predictive theory, model, explanation, deduction, induction, Itnowledge, and scientific methodology, revealing the truth about the hidden objective reality. However, science also deals with its own Ideological forms and processes of logic, empiricism, positivism, meaning of new theories, theoretical concepts, and verification. Through its life cycle, science accumulates (enlarges) subjects of its theories, increases its knowledge, suggests theoretical realism, but also imposes doctrines ■of formalism, Instrumental1sm, and sclentlsm.
Science does not allow philosophical spontaneity and philosophical abyss, therefore, in many cases. It cannot recognize its factual possibilities and its attainments, when it alludes behind the possible reality. In the frameworl« of its rlgorousness and correctness, science has to preserve Its characteristic Ideology, autocensoring, disclpl1 nar Ity, and steadiness, to prevent it from the decay and to enable the accumulation of Its scientific Informat i on.
Although sciences are grounded in knowledge and sciences are merely that what Is known, knowledge also exists outside of the sciences. Knowledge Is a result of cognition, Jt Is an objectively founded belief about the truth of a proposition, assumption, and consequence. Cognition, which Is not only scientific. Is an Informational process capable of production of knowledge, where knowledge appears as truth. The notion of knowledge contradicts the notions of belief, opinion, and view. Knowledge by Itself Is an arising, changing, and recurrent Informal ion.
11.5. Technology
... What has the essence of technology to do with revealing? The answer: everything. For every bringlng-forth is grounded In revealing. Brlnglng-Corth, Indeed, gathers within itself the four modes of occasioning - causality - and rules them throughout. Within Its domain belong end and means, belongs instrumentality. Instrumentality is considered to be the fundamental characteristic of technology. If we Inquire, step by step. Into what technology, represented as means, actually Is, then we shall arrive at revealing. The possibility of all productive manufacturing lies in revealing.
(Heidegger, QCT, 12)
Technology as InTormatlon is nothing more than a sort of productive presence of beings in themselves and In their environment. In this context, productive presence has the meaning of action,	development, destruction, and
construction. Technology qua Information arises as beings' thinking and behavior for changing themselves and environmental circumstances, destructing and constructing themselves and environment physically (constructively, methodologically) and biologically (culturally, intellectually). In this respect, a machine or even a living being can be understood as a final physical and/or biological construction, whereas methodology (thinking or ideology) can be comprehended as a final cultural and/or Intellectual construction (or destruction). However,	machines	and beings	as
physical/biological constructions arise through
circles of Informational constructions <cbunter-]nforn)ing, embedding). Technology is information of artificial and living construction and destruction, of machines as forms and processes, of beings as biological structures and organizations, and of methodologies and minds as forms and processes for achievement ot constructive and/or destructive results.
Constructive technological activities can be observed In animal-labor (specific behavioral patterns), for Instance, in constructive activities whose products are bird nests, lairs ot beasts, beaver lodges and dams, spider webs, honeycombs, ant-hills, cocoons, and other forms which arise through behavioral activity. In constructions of this sort, principles can be observed whose nature can be compared with sensefulness,	optimalness,	solldness,
intentness, suitability, rationality, and Imagination. Technological phenomena in microstructure of matter are, for instance, crystals, macromolecules, synthesis of proteins <genetlc Information of DNA as a program of a machine for synthesis), and at last biological growth, which reveals construction (structure and organization of living).
Roughly, technology as Informational phenomenon also materializes and methodizes thought and constructive imagination through specific human behavior which is called labor. The arising technological information Influences and changes beings' behavior and simultaneously and continuously develops thinking Into the direction of more and more compleic technological achievements. On this way of technological development the question comes up if already belng-1lice machines (instruments, processes) can be constructed which could possess intellectual, brain-llice, or mind-lilte properties. Technology as realization of useful machines and methodologies is the developmental way of construction, of technological revealing and creativity, and also a part of strategic thinicing and behavior for survival.
12
12.1. Aesthetics
... How nature pleases us belongs Instead to the context that it stamped and determined by the artistic creativity of a particular time. ... We are justified, therefore, in proceeding from the work ot art rather than from natural beauty if we want to define the relation between aesthetics	and
hermeneut Ics. ... The language of art Is constituted precisely by the fact that It speaks to the self-understanding of every person, and It does this as ever present and by means of Its own contemporaneousness. ... Everything depends on how something is said. But this does not mean we should reflect on the means of saying It.
(Gadamer, PHH, 98, 102)
Experiencing, feeling, sensing, observing, Investigating) recognizing, comprehending, understanding, living, describing, determining, expressing, thinking beauty and value are only a few Informational processes, which can be hardly determined by linguistic thinking and expressed semantically. Do the notions of beauty and value exist hermeneutleally, or even
ontologically, or are they first of all an emotionally and Intuitively conditioned and guided Information? Are beauty and value Information which also concerns ethics, morality, emotion, rationality, logic, reality, reason, behavior?
AS information, beauty is an experience, property, and quality which informs as a feeling of fulfillment, perfection, improvement, and also as pleasure, happiness, delight, enjoyment, influencing pleasurably through Its Informatlonally characteristic filtering. modulation, and harmony the (artistic) emotional domain. In this concern, aesthetics appears as metalnformation on beauty. on value of beauty In thinking, comprehension, and behavior. Beauty and the value of beauty ate the main subjects of aesthetics which deals with questions as What Is beautiful? What are attributes of beauty? What Is the value of beautiful? Is beauty an Irrational value? etc.
Evidently, it would be very difficult or even Impossible to determine beauty qua information by lingual means. Beauty as information is only weakly expressive on the way to language. Beauty is a characteristic informational process of a being's metaphysics, an inward experience of belief which cannot be entirely explicated outwardly. Beauty, and through it easthetica, are evident examples of information and its metalnformatIon, whose transformation from the hermeneutic (ontical) to the semantic (cultural, ontological, linguistic) level is informatlonally weak.
Informatlonally, beauty as information and counter-Information is Informing of beauty through beauty. In this Informational circle, the experience of beauty is arising within metaphysical embedding and within accepted environmental information of a being.
A work of art speaks to a being through Its symbols. Only if it speaks, it Is symbolic. It is beautiful, it Is artistic. It Is aesthetic. If. it does not speak. If Its symbols are not pointing to something which seems fanti liar, rhythmic,	harmonious,	observable,
investigative, or cognitive, this work of art cannot be observed. It represents a non-understandable object whose symbols are evened by Informational noise. In this way. beauty is a regularly relative Information, which is populationally, culturally, and ontologically conditioned. However, beauty also speaks hermeneuti cally,	inwardly,	generating
individual meaning or feeling of a being. The greatest art can appear as nonsense, which in fact does not sense the sensory system of a being at all. At this point, aesthetics, as metalnformation of beauty or value of art, confronts with this particular, beauty-concerning informational relativity. It appears that counter-information of beauty, of the value of beauty, and of the value of art (aesthetics) is informing freely, wildly, spontaneously, and circular-unforeseeably. Evidently, Informational illumination of aesthetics or Informational thinking about aesthetics can bring to the surface observations, which are controversory to a general understanding of art or to teaching of an artistic school, which can reveal the blindness of beauty and art, however, also break down this blindness.
12.2. Ethics
... hypocrisy plays an important role in the realization of human societies, permitting human beings under stress to feign having certain properties which they abandon as soon as the stress is removed. This Is why In a human society a social change takes place as a permanent phenomenon only to the extent that it is a cultural change: a revolution is a revolution only if it is an ethical revolution.
(Maturana, AAC, xxvil)
The Qreek 'ethos■ has the meanings of habit, habitude, wont, use, usage, custom, practice, etc. TheOreelc 'ethlkos' means moral, virtuous, righteous, and noble. Some of the basic ethical questions are What Is right and good? How does ethics or moral imply conformity to established sanctioned codes or accepted notions of right and wrong? What is rigtness, fairness, equity, or moral excellence? What Is moral eminence, particularity, Individuality, and freedom from anything petty, mean, or dubious In thinking and behavior? How moral is social and antisocial?
In general, ethics is a philosophical discipline, which investigates objectives, senses, processes, and phenomena of mora.1 wonts and behavior, criteria for evaluation of moral action, and through these, conception and origin of moral and morality. Within these various orientations, ethics can proceed autonomously (from an authentlcal, ontlcal, or a being's origin of moral) or heteronomously (from a cultural, ontologlcal, or a population's origin of moral) in an a priori, empiric1 Stic, Intellectualistic, axlological, naturalistic or voluntaristic way. Within these conceptions and heterogeneousness of ethics, it is possible to follow eudaemonlsm, hedonism, perfectionism, etc! stressing individualism (egoism, altruism) or social-moral problems.
12.3. Rhythm and Harmony
... One of the most important properties of regularly recurring temporal relationships is that they are predictable. This permits efficient learning and optimization of behavioral trajectories.
(Albus, BBR, 200- 201)
Rhythmic and harmonic informational forms and informational processes appear in the structure of cosmos and in natural forms and processes of matter, space, and time as consequences of different repeating and periodic phenomena; however, they also appear in living substance as particular phenomena of life, for Instance, as oscillations of macromolecules, electrical charges, protein production, biological replication, and phenomena of thinking and behavior in a highly developed being. Rhythm and especially harmony create feelings of pleasure, beauty, and good in arising and changeable autopoletic systems. By their nature, autopoletic processes and autopoietic forms are rhythmic and harmonic processes of self-production, self-reference, and circular spontaneity.
Periodic (oscillating) and harmonic Informational processes appear in a living cell and are characteristic of life processes of a being. Processes of this kind are, tor instance,	collective oscillations	of
cytoskeletal proteins in eukaryotlc cells (this is a comnunlcative function of the injection of viral nucleic acids Into host cells (Hameroff, ULC, 23S)), heart-beat, rhythm of breathing, a being's biological rhythm, harmony of musical, literary, and philosophical expression, etc. Characteristic behavioral patterns using rhythmic and periodic information are walking, dancing, singing, speaking, gesturing, playing, working, thinking, etc. Life activities of a being are synchronized and harmonized (coordinated, adjusted, consonant) according to natural and artificial environmental phenomena; these are lite habits, the rhythm of day and night, cyclic gravitational Influences of the solar system and cosmos In Its entirety.
Poetical, literary artistic, metaphoric expression can have Its hidden or revealed rhythmic semantics or lingual rhythmic measure by which phrases from semantlcaliy rhymed and harmonically selected meanings are composed periodically. Such a harmony of semantic, stylistic, and even syntactic periodicity arises on different informational levels of audio-speaking, sensory-motor 1c, and language-understanding hierarchy by means of metaphoric and diversely hermeneutlc processes which enter In the highest informational areas (emotional, meaning-concerning, planning, decision-making) of cerebral cortex.
A musical expression has evident properties of basic rhythm and harmony or disharmony of sounds. A melody is a composition of harmonically sequenced and dependent sounds, which appear in phrases by means of regularly repeating themas. A number of synchronized rhythmic harmonies can be the measure for musical informational richness. A simple ballade has only one thema, one melody, and one verse. Multlsound, choral music can have many parallel, harmonically dependent melodies and several verses, which can be synchronized Into regularly repeated patterns. Symphonical music is based on a rather various set of Instruments, where each of instruments is playing a different, however, harmonically dependent musical sequences. A symphonical performance and its comprehension are expressive examples of arising of a complex parallel information In brains of musical performers and their hearers.
In. general, rhythm and harmony are characteristic processes of informational Iteration and Informational recurrence, both of which are survival-regular informational forma of life. In the brain structure, rhythmically repeating informational processes entering the associative memory areas within the Internal, a being's model of the world exist. Repeating expectations are caused by these rhythmic informational processes and then these expectations are compared to rhythmically repeated sensory experience. A background of rhythmic patterns of meaning or rhythmic hermeneutlcs always exists and this hermeneutics pervades the entire generative information-processing hierarchy of the nervous system.
Regularly repeating informational phenomena and their relations are predictable. This property of Informational Iteration and expected informational recurrence enables effective learning of a being and through the learned or captured information also a senseful or optimal behavior. In this way, outcomes from various.
also critical situations, can be foreseen and several deviations Trom normal informational patterns can be revealed and learned. The capability of foreseeing and prediction is based on the capability of distinguishing differences which occur between expected and observed information. These differences are the so-called Information of error concerning expectation.	Information of expectant
difference can be pleasant (harmonic, beautiful, good, rewarding, stimulative, healthy) or unpleasant (disagreeable, ugly, bad, punishable, stressful, pathogenic).
Recognizing of essential, although small systematic deviations from the predicted ones, gives rise to learning and to arising of more complex recognitions and	predictions.
Recognizing of deviations between expected and observed phenomena is the basis of a complex learning of a being, of changes In information-processing of brain. Differences between expected and observed informational processes also modify Individual	expectations,
hypotheses, world models, and behavior of a being. In this concern, perception is ah informational process of searching for such hypotheses that can produce expectations which are adopted to the appearing observations, perception qua information is an instructive example of circular Informing, counter-Informing, and embedding of information.
A being masters Its environment through the capability of prediction of its sensory Information and through the capability to select such thinking and behavioral patterns that predominantly, consistently, or reliably produce results of delight (of beauty, good, pleastire, advantage, survival). In a learning environment, repeating of matter Is rewarding or attractive by Itself (repealing in singing, rhythmic movement and gesturing, hearing of known musical records, memorizing of sound phrases, everyday and working habits). In a potentially hostile environment, a being has several essential possibilities of survival If it can master the prediction of outcomes of its future actions. In a periodically changing environment such predictions are much more reliable than in a spontaneously or non-perlodlcally varying world. In this way, every informational environment which demonstrates phenomenal periodicity is Informing a being rewardIngly. In this context, harmonic relations	occur between Informational
hierarchies of behavior generation and of sensory Information processing. Dwelling in a predictable environment appears to be pleasant, non-arduous, manageable, safe, rhythmic, and harmonic.
It is clear that an exact prediction is not always possible. Sudden, abrupt, unforeseen deviations from the expected or predicted ones can produce feeling of surprise, where surprise stirs up at tent iveness and generates information of awalteness, openness, vigilance, lurking, willingness, and readiness. Processes and phenomena which are delightful in a surprising (a new, an unexpected) way are rewarding in an emphasizing way because before all they awake the lurking and then they reward beyond the expectation. Pleasant surprlsings can awake behavioral responses like laughter, crying, and excessive enthusiasm.
A pleasant surprise can constitute the essence of good humor. An interesting story can end In an unexpected,	thoughtful distortion,
perversity,	deformation,	spinning,
metaphorlcalness, ambiguity. Daring, critical, risky humor has Its Interest In a luring action of exceeding, breaking, overstepping boundaries
and limits o( environmental, cultural morality, taboos, ceremonials, and habits. Humor possesses its unexpected, metaphorical rhythm and Its spinning,	semantS cally and
hermeneutically ambiguous harmony.
Unpleasant surprises can awake emotional responses of fear, anger, aggressiveness, reluctance, disgust. In a hostile environment, new and unpleasant events can awake the feeling of danger, risk, or critical state. Deviations from the normal can cause disaster, accident, eatastrophy,	misadventure,	misfortune.
Incapability of prediction means to be surprised, means the possibility to lose the existing advantages of survival. Without predictability not only learning is threatened, but also the survival of a being or a population. A continuing and prolonged decay, disintegration,	disorganization,	and
interruption of regular informational patterns In the Internal rhythm and environmental stimuli of a being destruct predictability, disable normal learning processes, and cause individual and populational stresses (as forms of punishment) that can produce various social pathologic kinds of behavior (neuroses, Intoxication like alcoholism, smoking, drugging, sulcidity).
Rhythmic Informational processes are periodic. Periodicity enables the prediction of information arising In	Informational
processes. Harmonic informational processes are adjusted to experienced and memorized expectations, and In this respect, they are predictable. Informational prediction Is the kernel of the arising of rewarding information. This information embeds (confirms) individual learning and experiencing principles, and changes them. If environmental phenomena are changing. Rhythm and harmony as information Inform rewardingly, arise and are embedded circularly through the information of reward. However, the principle of rewarding is only one of many other aspects of informational rhythm and informational harmony.
11.i. Art
... We shall now attempt a first characterization of Nietzsche's total conception of the essence of art. We will do this by exhibiting a sequence of five statements on art which provide weighty evidence. ...
1.	Art is the moat perspicuous and familiar configuration of will to power;
2.	Art must be grasped In terms of the art I St ; ...
3.	According to the expanded concept of artist, art is the basic occurrence of all beings; to the extent that they are, beings are self-creating, created; ...
4.	Art l9 the distinctive countermovement to niht1 Ism; ...
5.	Art is worth more than "the truth."
(Heidegger, NIE, 67-76)
Art seems to be the sunnilt of unanalyzable, sensitive, expressive, and recipient informational spontaneity, the synonym for the free creative power, however, also for Informational skill, cunning, artifice, craft, and will to power. Literally, in Its characteristic environment, a work of art Is
arising, Is creating and being created circularly,	iteratlvely,	repeatingly,
rhythmtcally and harmonically. The artist creates	and simultaneously observes,
investigates, recognizes, experiences, and admires the beauty, artistry of his worl<. On this way, also ti>e new, upredictable is arising, being created out from the artistic admiration and from the will of artistic power. The arising of a work of art proceeds successively, and from the achieved, new artistic details are growing sensitively and spontaneously.	Thus, the artificially
unforeseeable, unexpectable is. generated. Certainly, the arising of work of art is social-circularly Interactive, that is, interactively soc i al-spontaneous.
Art is a specific people's mental, that is informational activity that carries information of sensitivity, incorporating creative glance, the work of art and its experiencing. Art la the capability of aesthetic creation, expression, reception, of aesthetic Informing in the domains of feeling, thinking, fantasizing by means of spoken or written words, instruments, human voice, color, mimics, plasticity, gesture, etc. In a broader sense, art Incorporates work of literature (drama, lyrics, epic, novel, story, tale, essay), work of music, ballet, acting, painting, sculpture, graphics, architecture. Art is productive (first-hand creativity) and reproductive (making music, acting, reciting), applicative (trade, industrial, professional), and mediative (film, radio, television).
Art Informs in a free and spontaneous way. In symbolic art, information is coming into existence from abstract indefinlteness, from uncertainty of a form or of a process of performance and comprehension. Sculpture is predominant in classical art, and the inward, spiritual dominates over the outward In the art of romanticism (painting, music, and poetry). As information. In principle, also l'art pour l'art is nothing more than a consequence of artist's social and antisocial tendentlousness,	of his informational
obscurity, ambiguity, and en Igmat1calness. Art Is cultural information and has little or no essential Influence outside its cultural doma 1n.
11.5. Literature
of a recipient, a story can create several Informational paths (states, processes, and patterns of them) which can approach the sting or experience of the story in an associative and in a bisociatlve (unrelated) manner. A linguistic Informational path is determined by a sequence of words and is understood to be a program for a distinct informational process or for a story. Each intormatlonal path represents a story for itself. Löwer paths in the story hierarchy represent details of a story, whereas higher paths are stories with more comprehensive vocabulary, however with less details. These paths are actions In the hierarchy of storytelling behavior.
Informational paths can belong to very different informational segments of a being's metaphysics (total Information of a storytelling and/or story-hearing being). For instance, informational paths in the hierarchy of sensory information describe experiences and feelings. Informational paths in the hierarchy of world-view or world mode] describe hopes, expectations, dreams, and beliefs. Each storytelling behavior can be composed of several stories which appear In recurrent, simultaneous, parallel, and sequential relation to different lower and higher paths, to details and programs of a story.
Informational paths as hermeneutleal (sub)9torles possess basic roles, for they influence a storyteller's behavior, determine experience, and cause imagination. Hermeneutic-semantlc translation of Informational paths into sequences of words causes the arising of language. A storytelling strategy selects one single sequence of words from many existing informational paths. Essential effects of an arising story can be achieved by Jumping from one Informational path to the other and from one hierarchical level to the other. A descent to a lower informational path enables the development ot an exciting, actual, or essential detail of the story, whereas an ascent to a higher Informational path enables a fast tt'ansition over unexciting, ordinary, or unessential events.
Informational paths (informational forms and Informational processes) of mind (in brain) are various as much as possible, for instance, they are behavioral, experiential, emotional, faithful, intelligent and they enable Jumping and transition over informational paths In a forward, backward, upward, and downward direction in groups of informational paths and thei r h I erarchi es.
... Literature Is what has been literally written down, and copied, with the intent that it be available to a reading public.
(M. Heidegger, WCT, 134)
13
13.1. Crisis
Literature is Information which dwells In many different forms and processes of Its production and reception in the realm of natural or formalized (musical, progranming, scientific) languages. The fundamental aspect of literature as natural or formalized language usage (speaking, hearing, writing, reading) is the social interaction of storytelling, of verbal expression and perception of stories, tales, adventures, imagination (also In the areas of music, formal sciences, progranming).
Storytelling is the basic process in writing and speaking. The basic component of every literature Is a story which can be philosophical,	artistic,	scientific,
technological, and/or occasional. In the mind
Being oC crisis is informational. Crisis as information is circled, closed into its own realm of cr111calness. On the way out of crisis, Its informational circle has to be broken into a new, releasing perspective."
A being's and a population's capability and incapability to survive, to be successful, to develop rhythmically and harmonically, can be particularly investigated through	the
perspective of an Individual, populatlonal, or social crisis. What is crisis qua information In the context of biological, behavioral, and Informational capability and incapability and how does It arise?
The meaning of the Greek 'krlsts' Is manifold, for Instance: separatton, exposition, decay; quarrel, conflict, strife; claim, suit, truth; decision, breakdown; judgment, arbitration, opinion; ordinance, fixation. Information of crisis arises In different occasions, for instance, when a blind Informational orientation Is broken down by a new, necessary information for surpassing the blindness, when a ruling Ideology is not capable to change Incapable (unsuccessful) Individual and social behavior in sharpened social circumstances, when the immune system of a being cannot resist to or neutralize pathological Invaders, when a disease continuously progresses onto the account of health, when a being is feeling stress, repression, or hostility, etc. and begins to develop counter-inforniatlon for improvement, defense, or change.
Crisis arises informationally from dlfferences occurring among particular information, meanings, understandings, and informational backgrounds In the form of conflicting and decisive information among them. Crisis is differential information which considers informational differences as the most substantial and most essential Informational entitles. In crisis as information, differences are compared, confronted, and developed to serve in searching the way out of crisis, to solve the problems of crisis, and at last, to annul the Information of crisis. If crisis cannot be solved satisfactorily. Its Information is growing in complexity overwhelming other, vital Information and ending In catastrophe in which it ceases to exist.
In general, it could be said that crisis begins to arise In a form of counter-Information from Information within which the critical Incapability Is coming to the surface, through the recognition that this Incapability is a consequence of Informational blindness, Insufficient knowledge, incorrect Inference and cognition, etc. This recognition of blindness simultaneously	represents	Informational
breakdown, the point in which critical information begins to develop. This Information which relates blindness and breakdown of blindness Is called crisis. Crisis is an information of opposites which appear in the critical context of crisis. Simultaneously, crisis Is a corrective information which Insists to correct the ruling blindness into a new illumination and perspective, to solve the problem of crisis, to decide on the way to new orientation, thinking and behavior. In this respect, crisis Is developing as regular counter-information through specific, crisis-concerning Informing, counter-Informing, and embedding.
Crisis as information appears in everyday life, is a consequence of individual or populatlonal cri t iealness , is regular (well-intentioned, democratic,	cultural) or Irregular
(subordinative,	repressive,	radical)
information. Crisis as regular Information gains the development of Individual and populatlonal capability, Its Irregularity leads to the development of Informational Incapability. Regular crisis appears to proceed by small, goal-directed, and successful steps. In an appropriate developmental rhythm and harmony, whereas Irregular crisis acts chaotically,	radically,	archaically,
ideologically,	and in this respect
arhythmlcally and d Isharmonlcally, also destructlng already positive developmental achievements of a being's culture.
13.2. Capability
The basic question of survival of a being and its population remains in the decision: to develop individual and populatlonal capability or to stay incapable.
Capability as information of a being concerns the metalnformatlonal structure and metalnformatlonal organization within the total Information of a being, this is. Informational structuring and organizing of its metaphysics. Information as Informational process Integrates informationally lower and higher components of Information and Information-processing. It Is simultaneously a metalnformatlonal process which creates lower and higher Informational components. A metalnformatlonal hierarchy includes basic informational components, metacomponents and higher metacomponents of a general degree
meta-meta- ... meta-components.
Metalnformatlonal hierarchy has the property of metalnformation, that is, of a circularly (metarecurrently,	metaparallel,	and
met&sequentially) spontaneous metalnformlng, metaarislng,	metacounter-Informlng,	and
metaembeddlng. MetalnformatI on performs as InformatIon.
Capability Is the Information of comprehension and of behavior, by which a being or a population Is observing, investigating, and recognizing its own Information and the outside information. Capability is a degree, a stage, the greatest achievement in the comprehension of the living, also of the human mtnd. The life and the survival perspectives are directly dependent on abilities of observation. Investigation, and cognition of a being, of Its population, and of its environment. There Is a substantial difference between Intelligence and capability. Intelligence is above all information, which is a consequence of exercise, practice, experience, drill, instruction, learning, informational selection, and will. Capability on the other hand is the supreme and creative ability in search and examination of various and complex Informational circumstances. Capability of a being is determined by the ability of Its self-comprehension, where a being Is critically Investigating its entire information and the appropriateness of Its intelligence and behavior. The counterpart of capability is characterized precisely by the deficiency of this	self-comprehensive	Informational
component.
Capability of a being is grounded in its self-comprehensive Counter-Informing. In this context, all Of the four attributes, which are self-reference (survival, ethics, moral), comprehension (self-observation,	self-
Investlgatlon, self-cogn11 Ion), Intensive arising of Informing (creativity, variability, alternation), and resulting counter-information (antagonism,	resistance,	rebellion,
disagreement. Impossibility, etc.) are necessary. By capability a being is changing itself, realizing Itself to a new, more appropriate capability. In this respect, capability as Information concerns capability.
Capability as Information can be an active strategy of Increasing the structuring and organizing principles of a being's total information (Us metaphysics). Through such an
intension strategical thinking and behavior are cđming into existence that support the development of capability. Through capability, a being is intentionally changing, developing, improving its strategic information for its life and survival, and through this information. It is changing Its behavior, developing new Intelligence, and improving Its strategic thlnlclng.
13.3. Incapability
theory of the mind per se. Rather, Issues that are central to the philosophy of mind occu-r throughout metaphysics, epi steinology, logic, aestethetlea, and, particularly, in the contexts of the philosophy of psychology and what ts called mind-body problem. In contemporary philosophy, most discussions tai<e the form of analyses of specific mental concepts: for example, consciousness, emotion, imagination, introspection, intention, thinking, and. the will.
(DPH, 233, 234)
To apply Ideologies Ln practical reasoning and behavior means to follow the way into incapability. Thereupon, informational circle of incapability closes; ideology and incapability produce each other and perform a stable course of development into Incapabl11ty.
Capability or incapability is the highest form and processing of informational structure and organization of the human mind. If informational capabilities lead to survival and development, then incapabilities approach to regression, decay, and dying out. There are some essential informational (also dialectical) opposites between capability and incapability. Whereas capability builds Its decisive information on forms and processes of spontaneity, creativity, and throwness Into factual situations considering developmental and intelligent experiences, incapability rests on philosophical and idelogical farms and processes of Informational orientation, uniform (population-molded, conform, individually stupid) behavior, and predetermination of thinl(ing and behavior in factual situations, considering promissory ideological experience.
In the life of a being and its population, the already acquired capability can Suddenly change to such a degree that it in reality descends Into incapability, for Instance, through the prarctioal use of a radical, revolutionary, unethical, or social-improper ideology. Such a turn in thinking and behavior through ideology prepares the way on which Incapability dominates, so it can tal<e Its triumphant march into a stable, self-referencing, self-productive, and self-satisfactory incapability. In this case, incapability as information develops an Incapability o'f a special kind, penetrating deeper and deeper into its own informational realm. In the course of. time, incapability of this sort, which becomes beings' and a population's total orientation in thinking and behavior actually destroys informal iona 1ly regular relations, capable Informing and counter-Informing and causes Individual and social disaster (decay of sound economy and of skillful labor, disappearance of spontaneous creativity and intellectuality, increasing of lawlessness and social diseases, etc.)
14
14.1. Functions of the Mind
While the principal task of the philosophy of mind may be regarded as the attempt to provide an account of what mind ts, the complexity of the task is such that there is little in philosophical literature that will seem to constitute an integrated
Viewed Informatlonally, mind is a notion which concerns Informational forms and informational processes of the human nervous system, this is, informational structure and informational organization of central and peripheral nervous system and of the body and its various kinds of processes as a whole. Mind concerns known and yet unknown information of body's processing, conscious or integrative and microstructural Information on the level of brain and on the level of processing within a particular living cell. The concept of mind is unrevealed and a being observes its own mind through "looking" Into It experientially, Investigating it in a top-down	(braln-integrative,	global,
psychological)	and In a	bottom-up
(biologically, physically, and biochemically fundamental or neurophyslological) manner.
Several functions of mind have been discussed in this essay, for instance, intelligence, creativity, language, ideology, crisis, incapability, etc. However, still many of basic (unctions have to be analyzed informatlonally (memory, learning, vision, etc. )
14.2. Information Systems
... Advances in Intracellular imaging and molecular biology have Illustrated the complex dynamic organization	of
intracellular cytoplasm. Specifically a dense, parallel, highly Interconnected, solid state network of dynamic protein polymers, the "cytoskeleton," is a medium which appears to be ideally suited for information processing, and which is actively involved in virtually all cell funct ions.
(Hameroff, ULC, 36)
Information	Informs (Influences)	and
informatlzBs (generates actively) in several systems of Information* An Information system (or Jnformational system) does not exist as a singular (individual) system of information, as an informatlonally isolated or informatlonally Independent system. However, information systems exist In the form of Informational plurality. Interacting among themselves, participating in other Information systems, being Informatlonally distributed, associated, and integrated. As a consequence of informational plurality, information systems are to some degree mutually informatlonally embedded in each other. In informatlonally divergent and convergent manner. Within this informational plurality, information arises as a consequence of systems' plurality, this is, circular-spontaneously in the informational sense. In fact, arising of 1nformation'brings new informational systems into existence. Information systems integrate a more or less informatlonally universal realm of information.
In general, information Is always thrown (injected) In an Informatlonal realm tn which It operates and Is operated on Its way of Informing. On this way of Informing, information finds Its informational ground, Its embedding and its distributive Informational power In several developing information systems, information performs as Information system. Information system is only another form of information.
14.3. Social Systems
... An autopoietic system participates In the constitution of a social system only to the extent that it participates in it, that is, only as It realizes the relations proper to a component of the social system. Accordingly, in principle, an autopoietic system may enter or leave a social system at any moment by just satisfying or not satisfying the proper relations, and may participate simultaneously or in succession In many different ones.
(Maturana, AAC, xxv)
By phi losoptit cal terms, an autopoietic system will be presented as belng-in-thc-world (or, more generally, as Dasein). What does a social system mean from the point of view of living Integration? What is the nature of the so-called social Information?
A very important constitutive (integrative) domain of a human social system is the cognitive linguistic behavior of human beings that integrate such a system. In a social system, beings of this system inform (or informatize) among themselves tn a cognltlvely characteristic way using different linguistic, symbolic, behavioral, and environmental forms of Informing, counter-Informing, and embedding of being's and of society's information. Every circularly arising informational spontaneity is socially bounded and It underlies the characteristically bounded autopolesls of a particular society. This informational boundary that constitutes a society's Informational closeness (particularity,	peculiarity,
prohibition), contributes essentially to the constitutive and Integrative power which orders, manages, subjects, and governs beings of a society into main informational perspectives of the social systems within soc i ety.
From the informational point of view, a social system is Integrated on the basis of an instantaneously ruling social consciousness (a state of culture, of economic development, of a population's Ingenuity) - the society Integrative Ideology. This Ideology is no more than a conmonly (socially, culturally, populationally, democratically, ethically) or authoritatively (asocially, Ideologically, totallstlcally,	repressively)	accepted
information which Implements the Informational coupling among beings, realizing a social autopoiesls. In this concern, a social system behaves as a being of beings, as the populatienal being. True and valid social information is only the coirmonly accepted information of beings which Integrate a social system. Otherwise, social information Is disagreeable, disappointing, and disintegrative and has not the power of the being's validity and acceptance.
14.4. Artificial Intelligence
twenty-five years of artificial intelligence research has lived up to very few of Its promises and has failed to yield any evidence that it ever will. The time has come to ask what has. gone wrong and what we can reasonably expect from computer intelligence. How closely can computers processing facts and making inferences approach human intelligence? How can we profitably use the Intelligence that can be given to them? What are the risks of enthusiastic and ambitious attempts to redefine our intelligence in their terms, of delegating to computers key decision-making powers, of adapting ourselves to the educational and business practices attuned to mechanized reason?
(Dreyfus, Dreyfus, MOM, xl)
Artificial Intelligence as a discipline could be hardly classified as a science, art, or even technology. Does it belong to a kind of new era's working philosophy which only searches Its authentic field in relation to concepts and yet . indeterminable Images of natural intelligence?
On its way of development, artificial Intelligence has established Its own sorts of applicability and methodology. Its applications can be classified, for instance, as theorem proving, games, robotics, vision, natural language processing, automatic prograrwnlng, knowledge engineering, etc. Its basic methodology has developed as representation of problems, searching, inference mechanisms, languages tor artificial Intelligence, representation and utlliEation of knowledge, etc. These methods and activities tend to explore and Investigate information processing in mind and machines. in this respect, artificial intelligence is a real and particular information-concerning discipline dealing with information in mind, brain, and in possible future machines. Artificial Intelligence particularly concerns Information, Its forms and processes in the living and the artificial.
By methodologies used up to now, artificial intelligence has not achieved the goals and promises which it has given during Its developmental phases. It seems evident that its philosophy of research was founded on the belief which says that Western-traditional, rationallstically hard, and in some way essentially reduced (simplified, quasi-scientific) techniques could be sufficient for mastering problems In the domain of Intelligence. However, up to now, artificial Intelligence has not succeeded to realize any real or evident intelligence-like product, for Instance, neither on the level of computer architecture nor in the domain of computer progranniing. The progress artificial intelligence has brought lies In the field of Improved algorIthmtcal concepts which accelerate, combine, and reduce data processing, however do not prove to be intelligent in any respect.
Although artificial intelligence can and must use rationalistic experience of the past, it must simultaneously step on and search new ways outside of traditional thinking and technological behavior. In the course of time, it has to establish a new philosophy which will consider results coming from marginal sciences
and from marginal philosophies. Intormation sketched (n this Principles of Information (POI)
15
Philosophy essay and is marginal.
of In
15.1. Conclusion
In this essay, trials have been made how to determine different phenomena and/or their phenomenologEes informationally. There could be understood that it is always possible to observe, Investigate, explain, recognize, and finally also comprehend the entire cosmic phenomenology	through	informational
illumination, where everything (metaphysically, on conscious or other non-cognitive level) appears as Information. In this essay, first of all examples of the so-called integrative, global, cortical, mlnd-concern1ng Informational forms and Informational processes have been revealed. No examples belonging to the field of blomoiecular consciousness (e. g., Hameroff, ULC) were given (for instance, informational processes of molecular computing; of consciousness as particle/wave physics, as a property of protoplasm, as learning and as a metaphysical imposition; of holography, neuronal signalling, microtubules, solitonsj cytoskeletal information processing, viruses, etc.) This kind of informational forms and Informational processes will be treated In a separate essay.
The objective of this essay Is to show the necessity and usefulness of further development of Information philosophy, theory, and technology within a broad, phenomenologically all-embracing Informational realm and not only from the standpoint of existing mathematical and coinnunI cational theory of information. In the future, the arising information philosophy should also essentially contribute to the philosophy of technological possibilities when Intelligent machines will be constructed.
Bow. New York, 1977.
(BPP) M. Heidegger: The Basic Problems of Phenomenology. Indiana University Press. Bloomington, 19S2.
(QCT) M. Heidegger; The Question Concerning Technology and Other Essays. Harper & Row, New York, 1977.
(NIE) M. Heidegger: Nietzsche. Volume I: The Will to Power as Art. Routledge & Kegan Paul. London, 1981.
(WCT) M. Heidegger: What Is Called Thinking? Harper & Row. New York, 1968.
(AAC) H. R. Maturana, F. J. Varela: Autopolesis and Cognition: The Realization of the Living.' D. Reidel P. C. Dodrecht, Holland, 1980.
(HMP) M. Murray (Editor): Heidegger and Modern Philosophy: Critical Essays. Yale University Press. New Haven, 1978.
(AOT) O. Wallas: The Art of Thought. Harcourt Brace. New York, 1926.
(CRY) R. W. Welsbergi Creativity: Genius and Other Myths. W. H. Freeman & Co. New York, 1986.
(OWÌ) A. P. Zeleznikar: On the Way to Information. Informatica 11 (1987), No. 1, 418.
(IDI) A. P. Zeleznikar: Information Determinations 1. Informatica 11 <1987), No. 2, 3-17.
(POD A. P. Zeleznikar: Principles of Information. Informatica 11 (1987), No. 3, 917.
ERRATUM
(BBR) J. S. Rabo t ics. Peteraborough,
Albus: Brains, Byte	Books.
N. H., 1981.
Behavior, and McGraw-Hi11.
(CPY) J. R. Anderson: Cognitive Psychology and Its Implications. Second Edition. W. H. Freeman à Co. New York, 1985,
(MOM) H. L. Dreyfus, S. E. Dreyfus (with T. Athanasiou): Mind over Machine: The Power of Human Intuition and Expertise In the Era of the Computer. The Free Press, Macmillan. New York, 1986.
(DPH) A, Flew (Editor): A Dictionary of Philosophy. Revised Second Edition. St Martin's Press. New York. 1984.
(PHH)	H.-O.	Gadamer;	Philosophical
Hermeneutlcs. University of California Press. Los Angeles, 1977.
(ULC) S. R. Hameroff: Ultimate Computing: Blomoiecular Consciousness and Nanotechnology. in Print: Elsevier Science Pubi. 1987.
(BAT) M. Heidegger: Being and Time. Harper & Row, Pubi. New York, 1962.
(BW) M. Heidegger: Basic Writings. Harper &
In the article A, P, Zeleznikar, Principles of Information, Informatica 11 (1987), Nn 3, pp, 9-17, pages 11 and 12 were printed twice! Instead page 11, the next page (page 26 in this issue) has to be inserted. We appology to our readers for this fatal error.
examples. On ttie other side, the following pFlnciples will depend upon the previous ones, so that all principles will be recurrently interwoven among themselves and simultaneously open for further comprehensional development.
THE CONSEQUENCE OF INFORMATIONAL PRINCIPLES. Principles of information will be counter-Informational too. This means that Irrespective of the principle of Information In this essay, a principle of Information will have the property of Information» of Its circular and spontanebus development (thlnkin^i arising, embedding) in the realm of information.
THE CONSEQUENCE OF INFORMATIONAL OBJECT AMD INFORMATIONAL SUBJECT.	Information is an
Informational object and/or informational subject (B mastery over itself). This fact can be expressed in formalized terms where information can be understood as an informational	operand and/or as an
informational	operator	(operation).
Informational objects are both Informational forms and informational processes which are simultaneously informational operations. Informational objects (operands) can be formally expressed as unities In the forms as Informa11onal_form,	Informatlonal_proces3,
being_ln_the_form, belng_in_the_proces8, etc. Similarly, informational operations (operators) can be expressed by different verbal compositions, e. g., inform, generate, arise, give,	vary,	come_into_exlstence,
eome_ln_presence, modify_intelllgently, embed, etc,
THE CONSEQUENCE OF COUNTER-INFQRMATICW. Counter-informational objects and counter-informational operations are, for example, countep-informatlonal_form,	counter-
Informatlonal_process, etc. as objects and lnform_by_obaepvlng, generate_through_cogni t ion, arÌ8e_by_investigatlon, etc.
THE PRINCIPLE OF INFORMATIONAL FORMULA. Information can be formally expressed by Informational	and counter-informatlonal
formulae which can be constructed by using Informational	operands (objects)	and
Informational operators (subjects) In a linguistically free, unbounded form. Informational operations and their compositions can be used as Informational operands (objects) and as informational operators (subjects). Although an informational formula is a free sequence (a serial expression) of Informational operators and informational operands, it can (formal-llnguistically) express circularity, spontaneity, parallelness, etc. of forms and processes appearing in an informational formula.
Now, some examples of primitive and composite informational formulae can be shown.
AN EXAMPLE OF BASIC INFORMATIONAL INFERENCE. The following formula Is the deductive inference represented in the usual if-then form:
information informs
i ts_own_and_other_lnforming; Informing Informs InformatTon;
Information informs information;
An Individual statement is terminated by a semicolon. In general, a list of statements can represent an 'and', 'or', or even an 'and/or' statement, etc.
AN EXAMPLE OF INFOHMATIC9JAL EQUIVALENCE. The following formula describes the equivalence
information informs intelligence 1s_cqual_to Information informs Informing! Informing informs Intelligence;
Operands and operators of a formula can be constructed by free choice. A statement before an operator is not terminated by a special sign.
AN EXAMPLE OF TWO MEANINGFUL FORMULAE. The following two formulae, where the first one concerns the operation composition and the second one the equivalence of two formulae, are meaningful :
(1)	inform la_equal_to
Inform, generate, modify;
(2)	information Informs intelligence
is_equal_to information Informs, generates,
modifies intelligence i B__equal_to information informs intelligence, information generates intelligence, information modifies Intelligence;
In this example some statements are terminated by the comna to explicate the special ^and* connective, etc.
THE CONSEQUENCE OF THE FORM OF THE OPERATOR 'INFORM'. The most general (and perhaps the strongest) informational operator is the operator 'inform'. This operator can be understood as a composition of an arbitrary sequence of other Informational operators. Each operator composition can be meant to be purely parallel, a mixture of parallel and serial, or purely serial. The order of an individual operator In the composite sequence may or may not be determined by distinguished formal connect 1 ves.
2. Principles of Informing of Information
In this section we shall deal with several principles which originate in the meaning of the verb 'to Inform' and several of its consequent derivatives. Let the verb 'to inform' include the meaning of all possible verbs and verbal compositions. In this way, the verb 'to inform' can represent an arbitrarily complex informational . operator. Let this meaning also be transparent to the noun 'information'. Yet, information is understood as an arbitrary form, an arbitrary process, or as both of them. In this way, information can represent an arbitrarily complex informational operand (object) and informational operator (subject).
Informing of Information as the broadest comprehension which is imaginable concerns four basic forms or processes of information.. These are: Informing, embedding, arising, and Counter-Informing of Information. This reflection yields the following four prInclples:
THE PRINCIPLE OF INFORMING OP INFORMATION. Informing of Information, where Informing is Inherent to information Itself, means that through Informing, information Informs according to the principle of Information, that through Informing counter-information is coming Into existence according to the principle of counter-information, and that this counter-information is embedded in existing information according to the principle of embedding of information.	informing circularly and
spontaneously concerns information according to
DATA FLOW BASED PARALLEL INFERENCE MACHINE
INFORMATICA 4/87
UDK 681.3.001:519.6
Jurij šlic and Borut Roblč Jožef Stefan Institute, Ljubljana
Abstrsot. Execution models of the data flow based parallel inference machine for OR-parallel and AND-paraliel Prolog and the eKperinental Machine architecture are presented. It is shown that two types of logic progrannlng languages with different sins can be supported on this machine. The programs are conpiled into data flow prograa graphs corresponding to machine language codes. Thus, parallslisot in the progran can be exploited naturally. The machine is constructed from processing elenents and structure memories interconnected through a hierarchical network. The processing elements interpret the procedures represented by the data flow program graphs in parallel. Structured data is distributed to structure menories and shared among these procedures,
Ksywords. fifth generation computer systems, parallel Inference machine, OR-parallel Prolog, AND-parallel Prolog, data How mechanism, machine architecture.
1 Introduction
The Fifth Generation Computer Systems (FSCS> research and development aim is to build a prototype ot knowledge Information processing system capable of efficiently performing know-iBdge-based problem solving and inference. Toward this end, a ten-year period has been assigned to the FGCS Project, and this period has been further devided into three stages. The goal of the initial three-year stage is to conduct basic research on individual system cortiponents in order to establish basic configuration technology for subsystems which are to be realized in the intermediate four-year stage.
Fig. 1 shows what has become known as the "basic configuration image" of the fifth generation computer C13. Looked at vertically, it has a hardware layer, a software layer, and an external Interface to applications systems, as might be expected. Looked at horizontally, it becomes clear that each aspect of the functionality of a fifth generation computer - problem solving and inference, kncwledge base management and intelligent interfacing - requires its own hardware and software support mechanisms.
The parallel inference machine and knowledge base machine are the most important hardware components of the FGCS. In the FGCS prototype to be completed as the final product of the project, the two machines will be integrated through a close link. In the Initial stage, however, research and development are proceeding separately for each machine with research themes separately determined, since the initial stage mainly aims to conduct research and development of individual component technologies to establish the basic technology
for the hardware, called the inference subsystem and knowledge base subsystem to be build in the intermediate stage CS3.
2 Knowlmdge bttm and infcrmns* subsyatams
Development of the FGCS hardware and architecture will include the implementation of mechanisms for processing and controlling a knowledge base and efficient execution of problem-solving and inference techniques with these mechanisms. The system will depend on multiprocessing and parallel processing techniques for which two objectives are critical t
(1)	Provide machines with the power to handle the natural parallelism found in problems tackled by humans. The structure of a problem and the necoessary processing for solving it can be shown by rules controlled by an inference nechanism. Thus a major goal of the FSCS project is to devise an execution model for the inference mechanism and to determine a way to configure it.
(2)	Achieve high-speed parallel processing capable of supporting intelligent human activities. For this requirement, the principal research must be concentrated on knowledge base processing algorithm to handle a large number of tacts as well as a mechanism supporting the algorithm.
The inference subsystems, together with the knowledge base subsystem, forms the kernel of the FGCS hardware CB3. The ultimate aim of the FGCS research and developitent project is a machine enabling the execution of parallel inferences C3,7,93. In the following we shall describe some FSCS project "data flow directed" efforts In designing such a machine.
28
HiBh Levet Enquiry Language y
Knowledge Base Managnent Systen
Probien Solving Inference Systen
Natural Language Speech Picture

Intelligent Interface Systen
External interface of the Basic Software Systen
Knowledge Base Managnent Systen
Probien Solving t Inference Systen

Intelligent Interface Systen
Knowledge Base Machine
Probien Solving &
Inference Machine
H
n
Basic
Software
Systen
Intelligent Interface Machine
Hordware Systen
VLSI Archichecture
Fig. 1 The overall structure at a fifth generation computer.
3 P«ralivi inferano« subsyct««
Machine language of parallel inlerenee machine
In F6CS project, logic prograamlng was selected as a bridge tc fill the gap between a highly parallel coaputer architecture and knowledge infcrmatian prccessing. Several logic programming languages, named kernel languages, which define an abstract interface between the hardware and the software, are beeing developed. The kernal language KLO is the maahine language of sequential inference machine fro« which a parallel version KL1 is beeing developed. KL1 is the the machine language of parallel inference machine C23 including two types of basic languages; AND-paraltel Prolog and Oft-parallel Prolog
In the executian of logic programs, a high degree of paraUellsm can be implemented with use of AND-parallel and OR-parallel BKecutions. When OR-parallelism is applied alternative douses of the sane goal are executed in parallel. The alternative clauses have identical initial states and do not interlere with each other, except far possible concurrent initialiiation attempts of a goat variable by multiple' clauses. On the other hand, in AND-paral lei isti the conjunctive goals of a clause body are executed in parallel. In general, the goals nay share variables and thus interfere with each other when the shared variables are accessed concurrently. AND-parallelisB in logic programming involves the simultaneuos execution of subgoals
in a clause. Uhereas OR-parallelism attenpts to achieve increased speed by investigating many possible solutions in parallel, AND-paral-lelism attempts to achieve increased speed by investigating the subparts of a particular solution in parallel.
In conventional sequential Prolog the search and test operations (called unifications) are executed one by one, but parallel search and test operations can be inplenented through parallel machine architecture to obtain a highspeed machine. A few sources of parallelisa which can be distinguished for parallel eMeou-tion of Prolog are ANO parallelisa, and OR parallelism.
<1) OR-parallel Prolog. When a goal literal S is given, the definition of 6 is invoked. A clause C is then selected froa the definition, and unification of 6 and the head H of the clause C is attempted. Generally, when multiple clauses C1, C2, ... Cn exist In the definition, unification of G and each HI, H2,
Hn can be executed in parallel. A unit clause Ci that Is successfully unified with G returns the solution(B>. A nonunit clause Gj initiates the next unification, treating its body as a new goal statement, and waits for the solutions. The resulting solutions of the goal S are merged into streams by streaa «erging primitives (in the order in which they are obtained) and then returned to the goal. Thus, OR-parallel Prolog is suitable for the olas ol "search-for-all-solutions" prob leas.
An example of OR-parallel Prolog Is Parallel Prolog C43.
(2) AMD-parallel Prolog. When a goal is expressed as 61 AND 62 AND ... AND So, AND-parallei Ism can be used to search for conditions for all literals 61 in parallel. The goal statement is satisfied only when solutions are found for all the literals Gi and there Is no inconsistency between these solutions. The consistency checking is easy or even unnecessary in cases when the goal literals Si have no unbound variables shared a»ong AND processes, or where shared variables are bound to the ground instances before invocations of these literals.
Several languages have been proposed to realize AND-parallei ism. They include PARL06, Concurrent Prolog, and 6HC (Guarded Horn Clauses>
WechanlBiBS of ,par.^U.el; inf^pgnpe
Various mechanisms of parallel inierence and architectures based on those mechanisms are beeing studied! data flow meohanisai, reduction mechanise, complete-copying nechanism and, clause unit processing mechanism. In what follows we shall briefly describe these four mechanlsns
cai i
(1) Data flow »leohanls«. In the data flow concept, execution starts when data necessary for the execution arrived. This concept can result in parallelism regardless of whether it is sKpIicitly indicated In the program. This mechanlsa executes kernel language programs in parallel based on the data flow concept.
<2) Reduction mechanism. When executed, an OR-parallel and AND-parallel Prolog program generates resolvents from a goal and clause. This can be regarded as a process in which a goal modifies Itself using a clause as a rule. The reduction mechanism can also he viewed as a kind of self-modification. Thus, there is a close sifflilarity between the execution of OR-parallel end ANO-parallel Prolog programs and the reduction mechanism. Accordingly, the reduction nechanisflt was selected far a machine architecture that executes OR-paraUel and AND-parallel Prolog programs.
t3) Complete-copying mechanism. Complete-copying is type of reduction mechanisa. Even if a process includes several literals (subgoals) and only one literal (subgoal) is reducible, the whole process is copied and transferred to a unit that executes the unification process. This increases the number of copies and the length of a packet in the network, while enhancing the independence of each process.
<A) Clause unit processing mechanism. In response to a request from an idle processing unit, a busy processing unit sends a process. Thus, this mechanism can avoid an explosion of resource requests. However, it takes time for all processing units to become busy.
>1 Data flow based infvrvna* ■mohin*
The parallel inference machine based on data flow mechanism tPin-D) is naturally well suited ta parallel procesing becouse the data flow mechanism is closely related to functional languages.
Data flow computation ...
Programs in the data flow, model are represented by data flow graphs, nodes correspond to operators and directed arcs correspond to data paths along which operands are sent. An operator Is driven by operand arrivals from its input arcs, and it outputs the result operands to its output arcs without affecting the other
operators' execution. This functionality of operators has close similarity to the functional languages.
... and looic orooramiaino
Execution of logic programs is performed in a goal-driven manners a clause in the programs is initiated when a goal is given and returns the solutions to the goal. Logic programming languages make use of the unification operation, which is one of their basic functions. Nondeternlnism is another basic feature of these languages; in particular, "don't-know nondeterninisn" is required for OR-parallel Prolog, while "don't-care nondetmr-minisin" is required for AND-parallel Prolog. The data flow model is also similar to logic programming languages such as OR-parallel and AND-parallel Prolog. The pograms written in OR-parallel or AND-parallel Prolog are compiled into data flow graphs.
Implementation of SHC
6HC was selected as a basic language of KL1 because it has clearer semantics and provides more efficient implementation than Concurrent Prolog, and it has more powerful descriptive power than PARLOS. SHC programs consist of guarded clauses such asi
H :- G1 , G2,
Gm I B1, B2,
Bn.
where, H, Si, and BJ are head, guard and body literals, respectively and "I" is called a commit operator. When a goal literal is given each definition clause is invoked and a semaphore flag shared among these clauses is created. Unification is attempted between the head literal and the given goal literal and if it succeeds then the guard literals are invoked as the new goal literals. Only the clause whose guard literals succed first can execute its body; i.e. the clause whose guard succeeds performs a test-and-set operation to the shared semaphore flag. H the result of this operation is also successful, the clause can execute its body; processing of the other clauses is terminated. Thus, ore clause is exclusively selected for a given goal from all the clauses whose guards succeeded. There are several implementation schemes to support the guard mechanism in GHC C63>
(1) Complete compilation scheme. 'All the unification directions are analysed in compilation time and codes are generated using unidirectional unification primitives. In this scheme the compiler is complicated.
<2) System number scheme. All the environments are managed by guard system numbers. A new guard system number Is allocated, when a new definition is invoked and is restored to its parent number when the commit operator is executed. The guard numbers are associated with all the variables included in the invoked clauses and the invironment to which each variable belongs is compared with the current environment when unification to the variable is attempted.
(3) Pointer coloring scheme. The pointer coloring scheme distinguishes variables belonging to the goal literals from those belonging to the current guard by coloring.If unification is attempted between a goal variable and a variable in the invoked clause, the Bailee's variable is changed to a colored variable, which points to the original variables. If a colored variable is unified with a term, the instance bound to the variable is read before unification. The commit operator restores the colored variables to their original variables.
Read-only tagging scheme. This scheme is an extension of the pointer coloring soheoe, in which every variable has a tag speciiying its read-only level. The read-only levels of the goal variables are incremented by one before the definitions are invoKed, and decreoented by one when the commit operator of each invoked clause succeeds.
Translation Iroffl SHC Droarawi into the d^tfa flQw graph
Ue shall itlustate the translation fron a given GHC program into the corresponding data (law graph. Let us have a sanple progran written in fiHC CFig. 2K It is a list-append program which appends a list specified by the second argument of the head literal to the end of the list specified by the first argument.
the right side of the "=" operator. The "« = " operator specifies a procedure "app" invocation macro. The "wait_instance" instruction reads the instance of the first goal argument which is passed along the input path "arg1". If the goat argument is an unbound variable, it is suspended until the variable is instantiated ("uargl"). This operation will need renote access if the variable cells are distributed over the nemory units in the systen. Then, the "switoh_by_type" Instructions switch all the goal arguments according to the first arguaent "uargl". If the first argument is nil, they put their left operands on their first destinations, otherwise <lf "uacgl" Is a llst> they put their left operands on the second destinations. Thus, one of the subsequent segaents is invoked exclusively. The "wrlte.instance" instruction tries to unify its two operands and
E Hß isuKei pftteKftu 1
»pii<>([],Y,Z):-true:ZsY.
apnđ([HiX),Y,Z]:-true;Z = [H;zi],*pn«(X,Y,Z).
Fi«. 2 €MC so
I (BSUPILKB eSiC 9
ret<<s«pp(ar|l,ar«2,ars3). beitliv
( CLAUSE IKDEXINQ ) uarcl:«»tt„tnftaflc«(ar8l), (argl_L,arBl_R):<w(tct)_by_ty pa (uargl,uargl). (arg2_L,ar«2_R):«iiHtcl)_t)»_type(arg2,uargl), (arg3_L,arg3_R) = *w(tcij_by_typ«(arg3,uargl>. (ret ret _2)=«<»ltci)_t>y_typ« (ret,uargl). ( COHPILED CODE OF THE FIRST CLAUSE ] resls«rlt»_jnft»iice(arg3_L,arga_L), r«turfl(real,ret_i).
t COHPILED CODE OF THE SECOKO CLAUSE I (pl,pa)s<>«coiep<>ie_M«t(argl_R), p3:craat*_0lobal_v«r(argl R). p4i«on»_H«t<pl,p3). pS:«rit*_(nftafic«(arg3_R,p4l. p6<<j»pp{pB,arg2_R,p3). res2:ci)ecH_c<>flflit«flcy(pS,p6). r»turii(res2,ret_e). end.
Fig. 3 Compiled code.
The resulting list is unified with the third argument. The compiled code is depicted in Fig. 3. The first statement of the conpiled code specifies the procedure name "app" and Its arguments "argi", "ar92", and "arg3". The pocedure body is enclosed by "begin" and "end" statements and consists of three segments. The second and the third segment are compiled codes of the first and second source clause, respectively. The role of the first segment is to decide which of the subsequent segments should be invoked. Each body statement corresponds to a node in the data flow graph. The left side of the "=" operator specifies destination paths for the results of the instuction specified by
if one of them is a variable, It will instantiate the variable to another operand. In the second clause of the source prograa, there is a variable "11" In the body which does not appear in the guard. For such a variable, the "orea-te_global_var" Instruction creates a new variable cell and initialiies it. Two body literals in the body of the second source clause will be executed in parallel. The first is the "wrlte-_instance" instruction and the other is the recursive invocation of the predicate. The "cheoV.consistency" instruction tests their results 'whether they terminated successfully. The data flow program graph representation of the compiled code is shown in Fig,
orgl
(wQit „Instance)
uarg!
crgg
arg3

Invoke 'app'
Flg. 4 Data flow graph representatlan of the coaplled code.
Machine architecture
Abstract ■•ohiM arohitvoUir«. The nachl-ne can exploit OR and AND-parallellsa as well as parallellsM in unification. In head unification, if both literals consist of multiple arguments, or 1f both arguments are structured data, the unification of these arguaents or their substructures can be executed in parallel. The machine is constructed from multiple
processing elements and multiple structure Memories interconnected by networks. The abstract machine architecture is shown by Fig, 5 CO.
ExpHiaental ■■china. The eKperinental machine is constructed troa nultiple processing element nodules <PEs) and multiple structure" memory nodules <SHs) interconnected through a hierarchial netwcrk as shown in Fig. 6 There are several hierarchy levels in the
processing element			processing elenent	
				
				
NETWORK				
				
structure nenory			structure nenory	
				
Fl®. 5 The abstract «achine architecture.
interconnection network. Each PE has Its local bus. Four PES and (our SMs are interconnected by an inter-nodule network bus. A set of these nodules is called a cluster. Several clusters are futher interconnected by an Inter-aluster network ijus. The hardware specHlcation lor these intercanneotion busses are the saae, and they are called T-busses ttoken busses).
Actual impleaentation of the experimental machine inoludes two clusters and is currently being expanded to four clusters. Of these clusters, one is specialized, having one SM repleced by a host processor (VAX-11/730>, which is used to initialise or «lonitor the system.
Caokct (anats. Each PE has several stages in order to inplenient pipelined or parallel executicn. Packets transferred between these stages include result packets and executable Instruction packets.
A resul packet <a token which is sent along the directed arc In the prograa graph), consists of three fields«
t1) The activity identifier C16 bits) specifies the invoked procedure instance nane to which the result packet belong.
(2)	The destination field <2A bits) specifies the address of the destination instruction <a node in the data flow graph) of the result packets. It also includes two bits for additional information! one specifies whether the destined instruction receives one or two operands, and the other specifies whether the operand is a left or right operand.
(3)	The data field (32 bits) contains the operand data to be send to the instruction. The machine uses a tagging scheue, in which each operand has a value field <25 bits) and tag field <7 bits), which specifies the data type of the operand. If the operand is a structured data, the value field has a pointer to the structure «eniory tS-bit module nuaber and a 20-bit local address in nettory), and tag field is further divided into two suhfieldi a data type subfield, which specifies the data type of structure (i.e., list, vector,	and a attribute subfield. The attribute subfield contains a non-ground flag, which indicates whether the structure has any sieple variables. The attribute subfield also contains a shared flag, which indicates whether the structure has any shared-type variables (i.e., shared variables, global variables, or read-only variables). The machine recognizes the tag field of
the operand and transfers control to the appropriate firmware routine.
An eKecutable instruction packet consists of five fields!
<1) The current instruction address (20 bits) indicates the instruction address to be executed and is used to obtain the destination
PE...processing element SM...structure nenory NN...network node
<
T-BUS
NN
C
T-BUS
NN
CLUSTER 1
<
T-BUS
>
ES.....[3
PE 1
PE 4
NN
X
SM 1
NN
X
SM 4
Flg. é Configuration cf the experimental «aehine.
address fron the destination specifier field as described below.
(2> The operation code field <B bits) specifies the operation to be executed.
(3) The iett operand t32 bits). The right operand (32 bits).
(S) The destination specifier field (Afi bits) specifies the destination addresses of the results. There are two modes to specify the destination addresses in the destination specifier field) in the full destination «ode the specifier Held contains up to two destinations (each of then Is 2A-bit length), and in the short destination mode the specifier field contains up to four destinations, where each destination is of 12-bit length and contains the relative addresses from the current instruction. The relative addresses are added to the current instruction address to obtain the absolute addresses.
Processing eleaent nodule. Fig. 7 depicts the configuration of each PE, The stages in a PE Include a packed queue unit CPQU) , an instruction control unit CICU), several atomic processing units tAPUs), and a network node (NN). These functional units have their own controllers and are operative in a pipelined manner. Packet transmission via T-bus is controlled by a NN, which has nine-tc-one arbiter to arbitrate the requests fron its lower level units and from its higher level bus. The PE has a local menory unit (LNU), which is used to store local data such as activity fflanegsnent information, and is shared and accessible from APUs. PQU is a FIFO queue memory to store the result packets from the T-bus. ICU receives the result packets from PSU and checks if the destination instructions are executable or not. An instruction is executable if it receives a
■y/
<
T-BUS
1		1				
A		A		PQU		
p		P				L
U		U				M
						U
1		8		ICU		
>
I-BUS
>
APU,..atonic processing untt PQU,„packet queue unit ICU,„instruction control untt LMU,..IocqI nenory unit T-BUS.„token bus I-BUS.„instruction bus
Fig. 7 Configuration of a processing element.
single operand, or if the partner operand is already in the operand memory COM) in the ICI4 when it receives two operands. In the later' case, the ICU searches in its OM whether the partner operand exists or not. If it does, the partner is removed from the memory! otherwise, the result packet is stored in the OH. This searching is performed associatively by hardwa-
re hash tjslng the identifier and the destination address as the key field. If the instruction is executable, the ICU fetches the instruction code in its Instruction memory CIII) and constructs an executable instruction packet and sends the packet to the next stage, one of APUs via the instruction bus (I-bus). The APÜ interprets the instruction packets and sends result packets to the PQU in its PE or other PEs, or sends structure access comand packets tc Sds via the token bus.
Structure aeaory «odul«. The SMs are
responsible for the structure access commands, perform structure manipulation operations, and return resul.ts to the destination specified by the commands. Each SM consists of an structure processing unit (SPU) and structure-menory unit tSMU) for storing the structured data CFig. 8). The SPUs receive the structure manipulation commands from the APUs and interpret them. If the commands need the responses, new result packets are created and sent back to the PEs. Such commands include read commands,' memory allocation commands, and so on.
SPU..,structure processing untt SMU,„structure nenory unit T-BUS...token bus
Pig. ft Configuration of a structure menory.
The specification and typical processing tines of the various units are given in Table 1 and Table 2, respectively.
Table 1 Specification of the units.
! unit I —+---
specif ication*
I IPSUI	FIFO Siie!	UKW x A&b
I II CU I	in size:	?&Kw x 59b
IPEI I	0« sizei	32Kw * 64b
I lAPUl	micro storei	IKu x 32b
t I I	7Kw X 32b
t ILMUl memory size! 512Kw x 32b +—+---+--------------------------
I tSPUl micro store: IKw x 32b ISMl !	TKw X 32b
I ismul memory size: inw x 34b
1 I I 512KW X lOb
+—+---+--------------------------
I NN I FIFO Size;	64w x Bùb
• w ,. word b .. bit
------------+
i
-------------+
<16K tokens)I t96K instr.)l (32K tokens)I ROM	!
RAM	I
I
-------------+
ROM I RAM 1 (data, tags)I (ref. count)I -------------:+
(6A tokens) I
34
Tabli 2 Typical processing tines of the units.
I unit I	item	I
)POUI packet receive	I
I I delay in queue	I
llCUl single operand instruction	I
I I tuo operand instruction	I
PEI I (on arrival of Ist operand)	I
I t two operand instruction	I
I I (on arrival of 2nd operand)	I
lAPUl "copy" instruction	)
I t packet creation	I
SM I Sn-read operation	I
[ SM-write operation	I
NN I packet receive	I ! packet send
time*
2 a
2
5 3 2
2 2
2 a

■ in machine cycles
5 CnnGluding points
Striking progress in computer technology has given us single-chip oonputers whose processing power far exceeds that of the first-generation computers. There are also various high-level languages) operating systems, and data-base systeas. As a result, programs for almost any kind of application can be written, provided that their algorlthins can explicitly be described. This means that computers can replace people in many areas because of their high-speed processing and large memory capability. However, there remain many application fields with hard-to-sol ve problems. One such Is the knowledge-information processing field, where FGCS are eKpected to play an important role.
A machine to cope with knowledge-information processing should support extensive storage of data and high-speed inference using the data. Up to now, inference procesing has involved Implementing functional and logic programming languages on conventional sequential computers. However, the need for processing power of new applications in knowledge-inforaa-tlon processing may exceed the capabilities of sequential computers.
The architecture of parallel Inference mechine makes it a possible candidate for coping with such processing requirements. Computer architectures proposed for parallel inference machines include the high-level language machine C103 as well as the data (low machine.
b Literatur«
Cl^ P. Bishop, Fifth generation cospu-ters; concepts, implementations and uses, (Ellis Horwood, 1986).
C21 K. Furukawa, T. Yokel, Basic software system, inj ICOT, ed., Proc. Inf 1 Conf. Fifth 6ener. Comp. Systems 198*, (North-Holland, 198A) 37-57.
C3] U. 0. Hertzberger, R. P. Van De Riet, Progress in the fifth generation Inleren-ce architectures, Future Generation Computer Systems 1 <2) (198*) 93-102.
N. Ito, H. Shimlzu, M. Klshi, E. Kuno, K. Rokusawa, Data-flow based execution mechanisms of Parallel and Concurrent Prolog, New Generation Computing 3 (3) (198S) 15-*1.
C5] N. Ito, n. Kishi, E. Kuno, K. Rokusawa, The dataflow-based parallel inference machine to support two basic languages in KL1, in: J.V. Moods, ed.. Fifth Gener. Comp. Architectures, (North-Holland, 1986) 123-1A5.
C6] N. Ito, n. Sato, E. Kuno, K. Rokusawa, The architecture and preliminary evaluation results of the experimental parallel Inference machine PIM-D, Proc. 13th Int'l Symp. Comp. Arch,, (IEEE, 1986) U9~156,
m T. Moto-oka, H. Tanaka, H. Aida, K. Hirata, T. Murayama, The architecture of a parallel inference engine - PIE, ini ICOT, ed., Proe. Int'l Conf. Fifth Qener. Comp. Syst-heos 198A, (North-Holland, 198*) *79-*a8.
ca3 K. Murakami, T. Kakuta, R. Onai, Architectures and hardware systems: parallel Inference machine and knowledge base machine, inf ICOT, ed., Proc. Int'l Conf. Fifth Gener. Comp. Systems 198*, (North-Holland, 198*> 18-36.
C93 K. Murakami, T. Kakuta, R. Onai, N. Ito, Research on parallel Bachine architecture for fifth-generation computer systems, Computer 18 (6) (1985) 76-92.
t103 H. Tanaka, A parallel inference machine, Computer 19 (5) (1966) *8-S*.
PODATKOVNO PRETOKOVNl PARALELNI STROJ ZA SKLEPANJE. V dlanku je predstavljen paralelni stroj za sklepanje, ki temelji na podatkovno pretokovnem IzvrBevanju logiCnih programov. Stroj podpira iivrSevanje logiftnih programov, zapisanih v OR ali AND-paralelnem Prologu (Parallel Prolog, PAULOS, Concurrent Prolog, 6HC). TakSni programi se prevedejo v podatkovno pre-takovne programske grafe, ki ustrezajo strojnemu Jeziku. Podan Je primer transformacije programa, zapisanega v jeziku SHC, -v ustrezni podatkovno pretokovnl programski graf. Arhitektura stroja obsega procesne elemente ter strukturne pomnilnike, ki Jih povezuje hierarhična mreia. Procesni elementi izvrSuJejo dele programskega grafa sotiasno, pri tìemer si delijo podatke, zapisane v strukturnih pomnilnikih. Podan so tudi prostorske in Öasovne zahteve posameznih konponent arhitekture.
ONE SOLUTION OF THE BUSLINE CREW SCHEDULING PROBLEM
UDK 681.3:656.132
Miroslav B. Jockovic institute »Mihaiio Pupin«, Beigrade Dušan M. Veiašević Faculty of Eiectricai Engineering, Beigrade
SUMMARY: The paper considers the problem of minimising the number of crews to carry out a specified nuiitoer of trips. A trip represents in this case a inovement between the two adjacent bus stops (each bus stop is treated as a timing point). Two procedures are presented for establishing a number of crews in a period of 24 hours which can handle all trips on one bus route. The concept of finding an optimum was not applied in solving this problem. Instead, smaller sets of trips were extracted from the whole nuuber and the best possible combinations of trips in each set are established. This gave excellent practical results for the test cases.
SADRŽAJ; Rad razmatra problem minimiziranja broja posada pomoću kojih se opslužuje potreban broj vožnji. Vožnja u ovom slučaju pretstavlja kretanje Izmedju dve susedne stanice autobusa (svaka zaustavna tačka se tretira kao vremenska tačka j. Prikazane su dve procedure za izračunavanje broja posada u periodu od 24 Casa koje uključuju sve vožnje na jednoj autobuskoj liniji. Koncept nalaženja apsolutnog optimuma nije primenjen pri reSavanju ovoga problema. Umesto toga, mali skupovi vožnji su izdvojeni iz celog skupa 1 iz njih izračunavane.' najbolje kontinacije vožnji. Ovaj postupak je dao odlične praktične rezultate pri testiranju.
1,	INTRODUCTION
The problem of calculating the necessary number of crews was solved using different techniques (Ref.1,2,3, 4), The heuristic method is employed to minimise the n number of crews serving a large number of trips, while smaller sets of trips allow application of the detei^ ministic method to find the absolute minimum of the number of crews required. So, this problem was never solved in a unique manner for any number of trips.
The intention here is to suggest a new method for solving the crew scheduling problem, and to trigger off further research for development of this model. The solution which will be presented doe& not require special mathematical software (linear programming, etc.) and it is economical in hardware requirement and running time,
2.	DEFINITIONS
Definition 1: A trip is one-way movement of a bus along a route between the two terminal points. However, in the current context each of the bus stops will be treated as a tifting point and the term "trip" will indicate a one-way movement between the adjacent timing points.
We also define the variables for a trip: . - point of departure	BPV,
-	point of arrival	BPM,
-	departure tine	ABF,
-	arrival time	ANK,
-	direction of the trip	R,
-	live running tiine	FZ,
- "link" time (required to move the bus from the arrival point to the departure point}
-	preparation time
-	layover time in case of an early arri val
ZVG,
MWZ.
Definition 2: A duty consists of a set of trips executed by one crew during one day. A duty is defined by:
- working hours (maximum allowed)	ARBMAX,
®ir®i2
ARSVOR, ARBABS, ,,.,a
n.
T,
In'
P*
UMZT,
-	crew preparation time
-	clear time
-	a selected subset of trips
-	nuiribcr of trips
-	meal break
A sequence of trips may constitute a duty if it satisfies the requirement and restrictions above and at least one of the conditions (1° and Z° below) of the meal breaks Tp to which the crew is entitled.
l" The meal break Tp during a duty must not be less than 1/z of the sum of trip times (e.g. z=6)
2° If the above condition (1°) is not satisfied then at least the following must apply:
-	one meal break of P^ min, or
-	two meal breaks of P2 min, or
-	three meal breaks of P^ min.
From experience, these values are P^=30 min, = 20 min and P^-IS min. If a minimum number of breaks exists in a duty (as. in case 2°), then the time distribution of breaks must be as follows:
-	the time from the start of a duty to the start of meal break P^ and from the end of meal break P^ to the end of the duty must not exceed 270 ( 4:30) min,
-	the time from the start of a duty to the start of the second meal break P^ and from the end of the first
meal break P^ to the end of the duty must not exceed 270 min,
- the time from the start of a duty to the start of the third meal break P^ and from the end of the first meal break P^ to the end of the duty must not exceed 270 min.
Meal break may be taken at any stop on route. The shortest possible tneal break is fixed to 10 min. A meal break less than 10 min is ignored.
The problem is to find a simple algorithm for determining the minimum number of crews (or value close to the minimum) on one bus line. In other words, to distribute the total number of trips N chi one line in one day to the ralniraum nunber of duties M, so that each duty satisfies all the previously mentioned restrictions. Also, it should allow for the fact that a new trip of the same duty «annot start before the previous one has finished (this will be referred to as "trip combination").
The number of buses servicing one line has not been considered in the analysis. The number of buses represents a separate previeni.
3. ALGORITHM
This section considers an algorithm which can give a short-cut soluti«1 SC and a general solution GS which is more complex but more accurate. They differ in the number of solutions which can be obtained from ore execution of the algorithm and the way it is implemented on a computer. Two methods can be applied for both solutions. The methods differ from each other in the sequence of selecting duties; metho< F calculates duties in an ascending time flow, while method B does the reverse, i.e. in a descending sequence. Method F will be explained only since there is an analogy between F and B.
The general solution GS
A. Sort all the trips using departure time ABF; resulting in a sorted set of trips
S„ = {a^; 1 = 1, Nl.
e. Extract all the trips from the set which satisfy the Inequality
ABF(a,) < ANK(aj) < ARBMAX - ARBVOR - ARBABS + + ABF(a,},
(aj: rll)	(1)
and let them constitute the trip subset S^,^ = = {a^; i=l, rll} from rll members, with member a in the first position. Trips from this subset only can take part in the selection of the first duties.
C. Combine all the trips from subset S^,^ in accordance with the restrictions from section Z, so that each combination has in its first position the trip a^.
Combinations with the highest time utilisation factor (the ratio of the sum of trip times in a duty and the duration of the duty) are stored as the best solutions for the first crew, hence creating a table C =
""•^ij^nixrll-
Element c^j, for i<j, a^s.umes a meal break p^^
between a^-th and a^.-th trip from the siAset S^jp If the trips a^ and aj are simultaneous, then c^.j=0. The symmetrical element Cj^- assumes the sum of times fy of executing the a^-th trip and moving the bus to the starting point of the new trip a^ at the same station. If and Sj cannot be combined, then f^j=0-
The values of p^j.and f^j are
p^ j=ABF(lij )-ABF(ä^ )-FZ(ä^ )-UMZT(S^ ,ì. )-ZV6(a^ )--G(BPN(S:^). BPV(afjO)	^jj
G Is a table of the times between stops. Element g^^ of this table represents the time required to travel the distance between the arrival point BPN(a^) of the trip and the departure point BVP(aj) of the trip Sj, The user specifies a time interval T in-which trips can be cotibined and this results that the table C contains only those meiibers of c. - and c.., which satisfy the Inequality for trips a^ and aj.
ABF(aj) < ABF(a^)	1<j	(3)
This restriction and condition for combining trips results in a sparse table C, greatly reducing the number of conèinations. The solutions for the first crew are found via the subscripts of the table elements which are greater than zero and are located above the main diagonal. E.g., let us examine a series of this table elements:
'-3,8' "'S, 13' '^IS.Z?' *'27,41' ''41,57* cripts point to the trip cotriblnation a^, a^, ag, a^j,
a^^, a^^, Bgy. When we analyse all the trip chains-, we
get Š^^ solutions for the first crew. Let us label them
as K](a]i). K2C®ii)'-'"''s1i(®ii)' parametar a,, means that these potential solutions for the first crew have been selected from the first extracted subset s^^^; that a,, is the first tHp In all solutions (In the first execution of step C In the algorithm	Hence, Jt
follows that
Wi 1,je{1,S,,}.
(4)
Each calculated duty which has a time utilisation factor greater then the preset thershold time utilisation factor Pp, is entered in one row of the table H. Let us assume that vii solutions satisfy this condition. For instance, let us say that after the first execution of steps B and C of the algorithm we have yl1=3 solutions for the first crew, i.e.
a, 83 a, a^ a^ a^ a^^a^ia^ja^^ 0...0 a^ 10 1 p,
KgO,): a, ag a^ Sg ag ag a,,a,3a,4 0.....0 aj 9 1 p^
K3(a,); a, ajaja^a^aga^^a^^a^^a^^ 0.
.0 a^ 10 1 P3
This shows that the number of trips for three duties is ii^(a^)=10, n2(a^)=9 and n^ta^i'lO, with the first unrelated trips being Spil^ii'Sj' =a^. Tine utilisation factors are p^, p^ P3» respectively.
In general case, after first execution the steps 8 and C we have ^^(a,^),	solutions
for the duties with "^{an''"?'®!!'.....
trips in each, and the unrelated trips in these solutions being api(aii).ap2(aii).....^pvll'^ll'- "^^pecti-
vely.
D.	create sets Vn^ta^^j'Vn^ta,,'} without the trips included in the duties
KgCa,,)....."^vll^^n^' '«spectiveìy.
Extract subsets	S^gj»" • • »S^jvll
sets by executing step B of the algorithm vii times.
Execute step C once for each of these subsets and the best second duties should be extracted.
E.g.,	the solutions for the first and the second duties whixh can be combined would be
+ (K^Ca^i), Vag,),.... K^giCag,))
(5)
*
The nuntoer of solutions is high, hence the user is allowed to select only those pairs of duties which have a high tine utilisation factor, or a high number of trips in duties. If we wish to retain the same number of solutions as in the selection for the first crew, and assuming that the values in brackets in relation (5) are sorted in a descending sequence of the time utilisation factors, the set of solutions for two
duties would be ,l+K^lag,). K^la,,}+K^(a22}.....'
solutions satisfy the condition for combining duties represented by relation
(K.(a^,)AK,(a2^))=0j 1e{1,v11}	(6)
The unrelated trips of the first duties are the first trips of the second duties, i.e.
This explains how a set of solutions for two crews is created. The procedure is repeated until each rcw of the taBle H contains'.a 11 the trips from set Sj^. If a greater number of solutions is obtained (rows in the table) which contain the same minimum values of the number of crews M, then the final choice is made by the user.
The short-cut solution SC
This is a sinplified version of the algorithm providing a single solution of the required nunèer of crews. Because of simplicity and short running time on the computer, this solution is of greater practical value.
a)	Sort all trips of the set Snj by ABF paraneter as in A.
b)	Extract the	i = 1, r11} subset of trips from the Sj^ set according to the inequality condition (1).
c)	Determine the combination of trips frran subset which has the highest time utilisation factor as in
C. Let this be the cotrtbination	with "^(a^^)
trips. Let us declare this combination as the first duty.
d)	Revise the set S^ extracting all the trips which make up the first duly.
We now obtain a new set	1 = 1»
K-n^Ca^^)}. Branch to step b of ihe^ilgorithts.
After k-th pass through steps b, c and d of the algorithm, duties	K^Ca^,).....''il^kl^
established from subsets ^rgp»"» with
"1(3,1).	.....P®*"
vely. Test whether the remaining trip set
algorithm terminates and the total number of crews is M=k, otherwise substitute k+l+k and branch to b,
4. REDUCING THE NUMBER OF CREWS
If we use alternative GS of the algorithm, we can end up with a large nurnber of solutions. Amongst that set of solutions there is a great chance that we obtain the solution with the mi ni mum nwnber of crews. However, the general solution is döonomically justifiable only if it is applied to smaller sets of trips or if few solutions are required. We shall now explain the procedure for reducing the number of crews which will also improve the result obtained from the short cut colution.
Ue proved experimentally that In applying methods F and B of the short cut solution on a set of N trips, we get two Independent solutions R^, Rji.... Rp and L^, Lg,..,, Lq.i> l-q' The number of crews p and q are different in the general case. It becomes evident that there is a tendency for the time utilisation factor of the duties to decrease as the subscript i increases for the row R and vice versa for L. This statement is not always valid for two consecutive duties, but it can be considered as such when the whole set of duties is analysed. This becomes more evident when numerous trips are taking place simultaneously. Also the qualit;y of each duty is dependent on the size of the subset from which it was selected, the mean trip duration, time distribution of trips and on the selection of the previous duties.
Here, we shall try to reduce the number of crews as obtained from solution R and L, to get a series of
38
m crews where m can be represented by
m < min (p,q)	(7)
The follcwirg procedure is recoitmended. let us consider the relationships which may occur between duty and Lp.
1° L^CR^; an trips of duty L^'are included in duty
Based on 1 transform series R and L into the fol-loviiing form
»1-
"z.....%-r \
..... L s
The character (-) appears, for example, in duties L^ and Lj, and means that certain trips have been exi traete d from these duties, as they have been included "ffi duty R|. In further steps, the series which are being compared are on the right hand side of the vertical line.
2° i-pR,; all trips of duty R^ are represented in duty Lp The procedure is the same as in Ue get
■r
■1'
4
'

S
3°	certain trips from duty fi^ are not inclu-
ded in duty L^ and vice versa. Based on the time utilisation factor we select, for example, duty R^ The series now becomes
M'
ft R-ii • • • t


p-r "p
Vi- S
Characters + and - show that some trips have been extracted and added to duties U and L^, for example. For the new values L^' and it ia not evident whether they can represent duties. However, based on the assumption of monotony of the series R and L, will assign in the second place of the L series, hence
Rp R2' ' "2 '
l(^) l
L3 ,..., L
q-1'
By analogy, after two steps, going from right to left the series becomes
Rj » R2»
Va


I » • * i
VrS
Generalising the procedure we arrive at a subscript i for which
+ + (8)
is valid. Here pj^ and p^^ are time utilisation factors of the seHes R and L. If the standard deviation from trip duration is small, we can also suppose that
(a sinplified notation was used. Hence
We repeat the procedure starting from the end of series R and L until we reach subscript j for which

(10)
and


(q)
(11)
are valid.
Hence, i+j duties with high utilisation factors are taken as those finally selected and we extract all their trips from further reduction. These are R^ Rg, ....R^,	The remaining duties are
determined by executing the short cut version of the algorithm on the remaining N^^^ trips
(12)
and then we repeat the reduction procedure with an interactive control from the terminal. After reaching certain values i and j, the selection of duty may have a negative effect on Lj^^ and vioe versa. In such a case the general solution of the algorithm is recommended.
5.	A NUMERICAL EXAMPLE
In order to test the algorithm, a program was written which processed a set of N=2Z7 trips. Variable length trips, between 5 stops of a route were taken, with a mean trip duration of 35.45 rain and with a standard deviation of fi.91 min. The short cut solution gave a result of M=27 crews. By reduction, the total number of crews was reduced to 26. Using a Honeywell, level 6 computer, and 23 K words of store, the processing took 12 min.
6.	CONCLUSION
During the analysis stage of this problem it was quite evident that It is impossible to calculate, in a short time on a computer, all distributions of N trips in M crews allowing for a set of restrictions. Hence, it was impossible to select a solution with a minimum value for M, However, an algorithm was developed, where the nuntier of combinations under investigation is reduced to a minimum by searching through a significantly smaller trip subsets and extracting til e best duty from each.
The weaknesses evident in selecting duties from subsets with a small number of elements or in selecting the last duties, can be eliminated by reduction. This procedure also reorganises duties. On the other hand, the general solution represents a set of sinplified solutions. Also, it gives the user the flexibility in
selecting solutions with a high tire utilisation factor in accordance with the optimisation level it wishes to attain and depending on the computer resources available.
It would be interesting for further research to apply this algorithm to the Inseparable strings of trips. In this case, a string would mean a sequence of trips with a high percentage of live running time.
acknowledgement
The author would like to express his appreciation for the many helpful comnents and suggestions in defining the problem received from Mr, Miodrag Mojsilović, from Hamburger Hochbahn.
references
Ml Jerrod Rubin, A. Technique for the Solution of Massive Set Covering Problems, with Application to Airline Crew Scheduling, Trans.Sci.,1973, 7, No 3.
|2| J.P.Arabeyre, J.Fearnley, F.C. Steiger, W.Teather, The Airline Crew Scheduling Problem, Trans.Sci., 1969,3, No 2.
|3| A.P.Artinov, V.V.Skaleckij, Avtomatizacija i
procesov planirovanlja i upravlenija transportnimi sistemami. Akademija nauk SSSR, Moskva, t98t.
|4| R.A. Chapman, J.F. Michel, Modeling the tendency of buses to form pairs. Trans.Sci., 1978, 12, No.2.
|5| A.G.MiIson, Entropy in urban and regional modelling, 1970, London.
A SELECTED SURVEY OF PARALLEL COMPUTER SYSTEMS
INFORMATICA 4/87
UDK 681.3.02
Saša Prešern Iskra Delta and Jozef Stefan Institute, Ljubtjana
ABSTRACT. This paper is a selected survey o-f parallel computer systems. A classification of parallel computers Is given and some most attractive architectures are discussed. Special attention is paid to massively parallel processors. The organization and interconnection structure of multiprocessor systems is given. By analyseing a trend of research in parallel computer systems over last 10 years some predictions are given about individual ■features which will probably have great influence on -future parallel compiuter systems. An extensive survey of references in parallel computer systems is given.
IZBQR IN PRESLED PftRALELNIH RAČUNALNIŠKIH SISTEMOV. Članek podaja izbor in pregled paralelnih raCunalniékih sistemov. Narejena je klasifikacija paralelnih računalnikov in opis nekaterih najbolj sanimivih arhitektur. Podaina je organizacija mul ti procesor jev in opisane so razliine povezovalne strukture med procesorji ter pomnilniki v posameznih sistemih. Analiza trenda raziskav paralelnih raiunalniékih sistemov v zadnjem desetletju omogofia izločitev posameznih značilnosti, ki bodo predvidoma moćno vplivale na razvoj bodočih paralelnih računalniških sistemov. V bibliografiji Je prilaien obèiren pregled referenc za paralelne računalniške sisteme.
1. INTRODUCTION - EVERVBODV MAKES IT PARALLEL
A few years ago all high developed counties in the world have started projects In developing a parallel computer system. All these projects were financially supported by governments. Many companies and research institutes also started research projects on parallel systems. The falling price of microcomputers and VLSI facilities on universities has encouraged many universities to design and to build parallel computer architectures based on linking many microprocessors or specially designed VLSI chips together to work on one job.
Development of a parallel computer Is an extremely difficult task which includes:
-	development a new concept of parallel computer architectura,
-	design of an operating system that supports parallel architecture,
-	transformation of traditional «equential application programs	to parallel programs either by preprocessor or by a parallel programming language.
We see that by switching from SISD (single instruction single data) machines to MIMD (multiple instruction multiple data) machines one can not simply upgrade an existing SISD computer system but one is faced with problems which are conceptually new. Research and development of a parallel computer system requires a very strong research which often include«!
~ more than lOO specialists,
-	a billon dollar financial support,
-	research and development phase which several years.
lasts
Government financial support is only a fraction of the whole finances which are devoted to projects in paral 1 el computing. Strategy makers in most companies are familiar with market research studies which predict that parallel processing machines will take about SO percent of the market in high-performane» computers by 1990.
2. CLASSIFICATION OF PARALLEL SYSTEMS
Parallel computers are usually divided into three architectural configurations!
-	SIMD pipelined computers
•	early vector processors,
•	attached processors,
» recent vector processors,
•	other vector processors,
-	SIMD array processors,
-	MIMD parallel processors,
•	massively parallel processors,
•	small scale parallel systems.
Another grouping in possible as -for example classification according to distribution of local and global memory into tightly and loosely coupled parallel systems or classification according to application possibilities into general purpose or special
purpofiB computers.
Many eKiatlng computers are now using several parallel »pproaches. Parallelism In pipeline computer» i» per-formed by overlapping computations and ii there-fore temporal paral lei ism. Parallelism in array processors is performed by multiple synchronized ALUs and is there'Fore «patial parallelism. Parallelism in multiprocessor systems Is performed by a set of processors Hith shared resources which Mork in asynchronous mode.
The list of projects in parallel computing is getting longer every day. By comparing the architectural approach in different projects we see that the computer scene in parallel computer systems is particularly varied. It is difficult to classify parallel computers, but helpful in order to concentrate on similarities and differences between the computer architectures.. Because parallel computers are using several different	architectural
principles one might argue a proposed classification.
Some of described computer are "paper machines" that have been etudled theoretically and by simulation, but have not been build. Many of this projects were funded by government agencies, but some of them are industry projects (IBH, Burroughs, CDC,...).
There follows an alphabetic list of the parallel computer systems or projects, each with the name of the chief architect and host institution. A list af references dealing with each project is also given. The most interesting	architectures are briefly
described. The list of parallel computers is grouped according to upper classification.
SIMD PIPELINE COMPUTERS
EARLV VECTOR PROCESSORS
BVM (Boolean Vector Machine), Robert ft. Wagner,
Duke University, North Carolina.
This is a collection of l-bit processing
elements connected as a hypercube with rings at
each corner, using the Cube-Connected-Cycle
topology.
STAR-100, Control Data Corporation , The design of Star started in 196S and was delivered in 1973. This is a processor with two nonhotnogeneous arithmetic pipelines. (HWA8S, LINBZ, PUR74).
TI ASC, (TeKaš Instruments Advanced Scientific Computer), Texas Instrument.
This machine uses 1 to 4 homogeneous pipelines and was delivered in 1972. (MUASS, KOGSl).
ATTACHED PIPELINE PROCESSORS
(CSP
Inc. ,
BILLERICA,
CSPI MAXIM/64, Massachusetts).
Maxim/64 in a minimal configuration includes a là slot chassis, a ó4-bit -floating point array processor, 16 Mbytes of data memory and Micro VAX-11 CPU. The machine is designed for research, scientific and engineering users and coots about »170.000. CNftN86>.
264 which is used in configuration named LCAP (LosBly coupled array of processors). More than 1500 machines had been sold and were used mostly for signal processing. They are quite cost effective in comparison to Cray or Cyber computers. <H0CB1,' HWA85, WIL82).
IBM 3838
IBM 383B is a multiple pipeline scientific processor specially designed to attach to IBM mainframes, like the System/370, for enhancing the vector-processing capability of the host machine. It is microprogrammed pipeline processor which can be supplied with custom-ordered instruction sets for specific vector appiications.
RECENT VECTOR PROCESSORS
Cray-1 Cray Research Inc., Chippewa Falls, Wisconsin, USA.
This is the first successful vector computer. More than 40 computers have been sold and installed, -first in 1976. It comprises 12 special-purpose pipelines for the different arithmetic operations. It is very expensive. (HWA85, J0R82, RUS7a).
An upgrade of this computer Is Cray-2, (H0L6S). Cyber-20a
This computer is an example of pipelined architecture and ts highly competitive with the CRAY-1. It is based on CDC STAR 100. It is based on one, two or four pipelined general-purpose units working always to-and from main memory. It is an expensive machine, designed initially to weapons' calculations and weather simulation. (HOCSl, HMABS, V0Na4).
CĐC/NASF Control Data Corporation Nunerlcal Aerodynamic Simulation Facility. This is a supercomputer to be used in 1990s for aerospace vehicle or superjet designs. The speed requirements was set to be at lest 1000 Mflops and the purpose is to calculate the viscous Navier-Stokes fluid equations for three dimensional modeling of the wind tunnel experiments. (HUA8S, HOCBl).
VP-200, Fujitsu.
This system has a scalar and a vector processor which can operate concurrently and it can be used , as a loosely coupled back-end system. (HMASS, LLUa4, UCHBS).
OTHER VECTOR PROCESSORS
Ahmdal 1200
This computer is a European version of Fujitsu's recent vector processor VP-200. Similar version o-f VP-100 is known in Europe as Ahmdal 1100 computer. (KOCàS).
Siemen* VP200
This is another European version of Fujitsu's vector processor VP-200. Fujitsu's VP-lOO ts as Siemens product known a* Siemens VPIOO. (KOCaS).
YHl
This is Chine's first supercomputer, known also as "Galaxy". The development started in 197B at the University of Defense Science and Technology in Changsa. The machine looks like a Cray computer. <N£W8S/11.
FPS-AP120, Floating point systems, Beaverton, Oregon, USA.
This company produces also a new version of attached pipeline processors FPS-164 and FPS-
SIMO ftRRftY COMPUTERS
MIMO PARALLEL PROCESSORS
BSP,	(Burrouflh. Sci.ntlfic ProcMmor) ,
Burroughs (HOC01, HWft83, KUC02>. BSP h« been largely based on the experiences thÄt Burroughs have gained ae major contractor» on the ILLIAC IV project. The deaiqn principles of the BSP were to provide a machine using a standard technology, «bich vJOul d be programmed in a high level language and «ustavn a continuoua 20-40 M-flop»/s.
ICL OAP (ICL Di »tri but »d Arr»y Proc«Mor) (HOCS1)
This is an array of one-bit processors which are. often called associative processors- The design o-f pilot DAP «as started in 1974 and consisted of a two-dimensional arrays of 1024 1-bit processors.
ILLIAC-IV <BAR60, DRV69, BQU72, HWA85> This computer «as designed for the solution of partial differential equations and can be described as an 9«e array of 64-blt floating po^nt processing elements each (PE) with 2Kwords of memory. It was working with nearest-neighbor connections (fig. D »"d controlled by a single Instruction stream processed in a central control unit.
Jt V SB 51 n	C3
FiQ. 1.« The connectivity between 64 processing elements in ILLIAC IV (HWA05).
MPP, (Massively Parallel Prac*ssor)
Thl4 processor was developed for processing
lat^lli^r imagery M the NASA
Flight Center and has 120«128=lfa384
microprocessors that can be	^ "^-bU
Each processor is associated with a 1024 bit
RArt. (BATBO, BAT82, HWA85)
PEPE (Parallel Element Processor Ensemble». This social purpose computer
floating point processor array «^ich was developed at Bell Laboratories and designed to control a ballistic missile defense *V«tem of radar detectors and missile launchers. This is r loosely coupled system of 288 processing elements, each containing three processors-tKAR02, YAW77, FIN77)
fl'^'thi« processor a bit serial associative memory Is used. Staran consists up to 32 associative array modules each cnntaining 256 processing elements- The first Staran was installed for digital image processing in 1975. <YAW77, RUD72, BAT77, KAR82)
MASSIVELY PARALLEL PROCESSORS
The main accent in this architecture is in interconnection mechanism to connect several hundred processors with .memory modules- A high processing power is achieved even by applying a standard processors.
BUTTERFLY, BBN - Bolt, Beranek & Newman (H0L85/2, H0L86, RET86)
The Butterfly computer is « large scale shar&d memory parallel processor that achieves high performance in configurations as large as 2j6 processors. The processors used are Motorola 60020 with Motorola 68881 floating point hardware. The system has a maximum performance of 236 MIPS of processing power in 1 MIP increment and up to 1 Gbyte of memory in 4 Mbyte increments.
Processor-memory interconnection la real i7ed via a multistage self-routing switch network. Pill processors can access memory simultaneously and in parallel, provided that no two processors try to take the same output path from a particular node. Butterfly network for 16 processors and 16 memories, called barrel-switching network is shown in fig 17. The speedup is nearly linear and is measured in a system with 236 processors ranging form leo to 230 times that of a single processor (fig 22) .
CEDAR, David Kuck, Duncan Cawrie and Daniel eaJski, University of Illinois at Urbana-Champaign, USA. tABUe4, ABU86>.	^ h ^
Cedar is an eight year project that started in 1903, The architecture is hierarchical! siKteen clusters of eight processing elements are connected via an extended tirnega switching network to 256 global memory modules of 4 to 16 Mwords each. Each cluster has eight processing elements, each with 16 kwords of local memory. These processing elements are pipelined and interconnected via a local switching network (fig 2.)
I I I I I i
pO
I »k—	.	I taM« I	IMM	1 lUCPIfWIC«« I ,	I
j -CD- I I -o , I ^ZD■ j	, -cn- j
i.ä.01 lùòl iòQ_öS
r
i SMind I I menortii I ' 1
rO -o -o
-C3-
LM - local memory
CCU - cluster control unit
P - processor
PI - communication processor
DC - disc controller
Fig. 2.» The architecture of CEDAR parallel computer (ABU0S>.
The prototype Cedar 32 has four clusters of eight PEs and uses 400 ns clock period. This
gives a total fflaxiintjin per-f ormane» of 80 Mflaps/s (Comparable to the Cray-1) for the desk-top sized prototype. Cedar 126 will have 16 clusters, giving total maximum performance of 320 Mflops (198S> and Cedar 512 will have 64 clusters, giving total maKimum performance of 1.2 Gflop/s (1970). An alternative engineering is planned using 40 ns clock period giving a four cluster Cedar 32H BOO Mflops/s (1909) and 16 cluster Cedar 128H 3.2 Gflopa/s (1991). Extensive software development project called Parafrase is underway. It Is focused on program transformations to enable standard FORTRAN programs to run on parallel Cedar machine.
ReconfIgurabls Array et. al,, University
Computar t, of TeKaš,
TRAC, (Taxi J.C.Browne Austin.
16 8-bit microprocessors will be connected via a 4-level banyan switch to SI memory modules. <JEN81, JENB2, LIP?7, PREB2, SEJ80),
Dppe7 (Dal ft Parall«l Processor 87)
This computer Is an upgrade of DPP91. The DPP81 consists of one PE—cluster with B processing elements (fig. 3).
DPP87 is a modular MIMD system with up to 16 processing modules each having 32 processing elements. Each processing element consists of a stack oriented arithmetic processor AMD 9311. A PDP 11/23 is a host computer. The DPPa? computer is designed for' simulation of systems {SIPB4).
ESPA (Erlangen Sanerai Purpose Array). W. Handler (HAN85)
ESPA consists of a grid-like array of memorycoupled processor modules. Above the array there is a pyramidal hierarchy of processors for supervising and for. data transport. Each node consists of one processor and one mefflory block (fig.4).
CHIP (Configurable Highly Paralisi Computar), Lawrence Snyder, Purdue University, Indiana (SNYSl/l, SNYBl/2, VALSS)
This computer is an array of processing elements embedded ,in an array fo switching elements such that network connectivity between the processing elements can be reconfigured under program control in one machine cycle. The switch lattice is typically a regular structure such as four neighbor or eight neighbor mesh. Fig. 21. illustrates how the original lattice is reconfigured as a mesh and as a binary tree. The project aims to produce 2eKp8 and 2expl6 processing elements with a few processing elements on a VLSI chip.
COSMIC CUBE (Nearest Neighbor Concurrent Processor, NNCP), Geoffrey Fok and Charles Seitz, CalTech (California Instltutis of Technology), Los Angeles, California. The first machine is 2eKp6 Hypercube hosted by VAXI1/780, with processor Intel 8086 at each node together with 8087 floating point coprocessor and 12B Kbytes RAM. This machine was commercialized by Intel and marked as IPSC. The Intel iPSC Is available with 32, 64 and 128 nodes. Each node Is an Intel B02B6 processor and 80287 coprocessor together with 512 Kbytes of local mamory. The maximal performance of the 2expl0 hypercube Is estimated to be about 100 Mflops/s that is to say about the same as the large supercomputers Cray X-MP and Cyber 205. (CHA86, SEIB5, EMM85, EMMB6/1, EMM86/2).
Citnnir t« putite diti nitw«rlii
'p' ProCessor-Memory-Modul e (PMM)
symmetric mul tiport-memory connections between neighboring PfiMs
asymmetric multi port-memory connections between PMMs of different hierarchical levels
Fig, 3.1. Architecture Processor DPPSl {SIPa4).
of Delft Parallel
I/O communication to elementary pyramid, supported by I/O processor
Fig. 4.1 The EBPA multiprocessor architecture consisting of 85 processor-memory-modules (HANeS).
The project started in 1975. The processor-memory modules are commercially available computers	AEG	80/60.	Interprocessor
communication takes place via common control blocks and mailbox techniques.
FEM (Finita Elamant Machine), David Loendorf and Harry Jordan, NASA Langley Research Center, Hampton, Virginia.
Processing elements are T19900 microcomputers,. controlled by a TI990 minicomputer,
FLEX/32 (Flexible Computer)
This parallel system Is composed of 32 bit processors NS 32032. Up to 20 processor modules are connected by a bus and for a bOK. Common busses link as many as 10 local buses per cabinet (fig 3.).
44
CPU	
	10
	
CPU	
IH 1	■v ,
: CPU	
	to
	
cm	
LH	ID
	
CPU	
M<iMn(	10
	
tloncrr	
Fig. S.: The mul ti computer.
architecture of Fl ex/32
A total system is designed -for 24B0 processors. The system has global and local memory. The price -for minitnal ccrcfiguration is »150.000. <MAN85, ZS0B6).
IBM QFli, IBM T.J. Watson Research Center, Yorktown Heights, NV, USA (BEESS) SFll is a parallel computer with S76 -floating point processors (512 primary processors and 64 spares). Each processor has space for 2 Mbyte of memory and is capable of 20 Mflops, giving the total machine « peak of 1.1 Gbyte of memory and 11.S Bflops. The floating point processors are interconnected by a dynamically reconfigurable non-blocking switching network called the Memphis switch (fig 6.).
-Srifrteemr«-
'1 k ■7T

'i
Tr


•I r
1 r
Heinphii Swttch S76 - STt
fttmuloliin litwcrk
cfte
S7S

Ik
Jwilch tntiol
I t
Ctnlnk		
(OiUil	/	
		
Fig. 6.« The GFH architecture.
The main intended application of GFll is a class of calculations arising from quantum chromodynamics tn nuclear physics where EFll is expected to be 100 times faster than Cray 1.
IBM RP3 (Research Parallel Processor Project), IBM T. J. Watson Research Center, Yorktown Heights, N.V. (PFISà) .
RP3 project is performed in cooperation with the Courant Institute of Mathematical Science at New York University. The goal of RP3 project is « parallel system with 512 32-blt fflleroprocessors and 2 6bytes of main storage. RP3 ha* two multistage Omega-like networks to connect processor-memory elements. The first network is constructed from high-speed bipolar logic and designed for fast interconnection to nonlocal storage. The second network is using a
technique» developed at the NYU Courant Institute Ultracomputer project (EDLS5, G0T82, QOT83, SCHBO)i This network contains more complex functions required to carry out synchronization operation and storage requests. With these techniques the efficiency of parallel system is not degraded as the number of processors is increased.
The language supportIni ti al1y envisaged, consists of minimal extensions of commonly used languages such as Fortran, C, and Pascal. The adaptation of other languages, such as Common Lisp and Ada to the highly parallel environment is also being studied.
The operating system will probably be an extension of BBĐ 4.2 Unix, modified internally to make it a fully distributed, symmetrical system, and extended to provide multiple process shared memory and efficient massage passing.
The RP3 is expected to achieve an aggregate of 1000 MIPS on shared memory scientific applications. A demonstration of the RP3 with 64 processor-memory elements is planned for 1987.
RP3 machine is not associated with any IBM product program, and it is expected that very few of the RP3 prototypes will be constructed. RP3 is target to exploit automatic paral leiiiation of Fortran and compilation of functionally-orlented languages. Development of parallel applications for RP3 is performed before the completion of RP3 hardware construction by means of an experimental emulation system called EPEX (Environment for Parallel Execution).
MIDAS (Modular Interactive System), Creve Maples, California, Berkeley. This is a hierarchical system
Data Analysts University of
of a primary
»Ml
B
		O
Di Ik*	DOHb	BNb
mm/i
llMbl Cuhtmemorf

irvtit»
SEI 32/7/
	mi	i»t	1 1 1 1
	liti	11H	M
1 CCI felK		* ' r tt CT Uk J. b	
		K »I n»'t f	
APint		»mb		«■llOb		tfllOb
II H ri» (U						UMfl^i
[D bit«)		[31 biti)		(»bitii	|3lbiti]	
11Mb/l
m Hemorr
AP 120B - array processor
H5D - high speed data interface with
4 Mbits/s throughput GPIOP - general purpose 1/D processor
Fig. 7.1 The architecture of ONERA parallel computer <LECB6).
cqmputer controlling several secondary computers, each of which.controls a Multiple Processor firray (nPfl). Each MPfl also has an input and an output processor and a crossbar switch connecting the proceaaors to 16 SMitchable memory «nodule» each of 2S6 Kbytes, The system is used to solve problems xn computational physics particularly nuclear science.
ONERA. (Offic« Natlonel d'Etude« at Rechars Aarospatialcs).
This is French project on a mul ti-array processor which started In 1979 and evolved to a loosely coupled architecture. It was designed to solve partial differential equations. The present system ha» as a host a 32-t)lt Gould SEL 32/77 minicomputer. Four array processors AP 120S are connected to the SEU bus. The AP» are connected to a sharable memory of 32Mbyte (fig. 7.) (LEC86, ADE85)
PASM (Partitioned BIMD/MIMD machine), Howard J. Siegel, Purdue University Indiana. Envisaged that a full machine might have 1024 processing elements connected via a multilevel switch network. A prototype will have ló Motorola 68000 processors and four control units fSieSl, SCH8ò>.
PRINGLE, University of Washington and University of Purdue (called RP2).
proCBBBora Intel 8031 are connected via a switch and controlled by an Intel 8086. (KAPB4).
8UPERNÙM. W. K. Biloi, H. Muhlenbein, (SILB6, H0P8Ó, KRAa7)
This is a national project in parati lei processing in Germany. The computer has a hierarchical hardware structure consisting of nodes, clusters and hyperclueters. It is conceived as a 64 cluster machine with 1024 node®. Each cluster has 16 nodes and 140 Mbytes Winchester disc. Motorola 68020 processors are used in nodes. Fig. 8 shows a SUPERNUM architecture with 16 clusters, where each cluster has 4 processors. The processors in the cluster are connected by s bus. The clusters are connected by row and column rings.
				— hMt 1
				
	-m		-m	
				
Fig. 8.: The topology of the BUPRNUM parallel computer (KRA87>.
The system is designed for solving partial differential equations and other numerical appi i cations.
VFPP (Vary Fast P*rall«l Processor), Norman Christ and Anthony Terrano, Columbia University, New York.
This is conceived as a 16x16 array of 256 processing elements with nearest neighbor connections. Each processing element has an Intel 80286 with 80287 coprocessor, 20 Kbytes
of local memory and an 11 stage pipelined mlcroprogratnmable vector processor. VFPP is a special purpose computer designed for lattice-gauge and similar calculations.
SMALL SCALE PARALLEL SYSTEMS
The word small In this context means only a few processors connected together. To get high processing power by these machines one needs powerful processors with interconnection mechanism which need not be as sophisticated as by connecting several hundred processors.
ALLIANT FX/B (Alliant Computer Systems Corp., Acton, MA 01720> (LLUe6/2, HARB6, SIE86, TEC85) An FX/B computer combines vector and concurrent processing In a system consisting from I to B computational elements. Each computational element Is a microprogramed processor which can execute both scalar and vector instructions. Computational element can access via a crossbar switch, two 64 Kbyte caches, Fit/S can have from 2 to 12 interactive processors based on a Motorola 68012 microprocessor. Each interactive processor has 512 Kbyte local memory and is designed for execution of parallel I/O and operating system tasks (fig, 9).
Concurrtncir cantrg( bui

CE - computational element IP - interactive processor IP C - IP cache CP C - CP cache M - multibus
Fig. 9.1 The Alliant FX/B architecture.
FX/B operating system Concentrix is an extension of UNIX Berkeley 4.2 version, A minimal configuration with 1 computational element costs »270.000.
CONVEX C-1 (Convex Computer Corp., Richardson, TX 75081)
C-1 supercomputer is based on a Cray-like architecture. The processing units are interconnected through 64-bit buses and include a dedicated scalar and vector unit. It has a dual ported main memory and up to five 32-bit I/O processors. The operating system ia Convex UNIX operating system, similar to UNIX 4.2 BSD operating system. The price for a basic system is «500.000. New versions are coming to the market, Cl XP processor, a faster version of C-1. (H0L85/3, TEC86)
CRAY X-MP, Steve Chen, Cray Research Inc., Mendota Heights, Minnesota. (HWA85, QEDa6, ERH86, LUB8S)
The Cray X-MP ia a mul t i processor upgrade o4 the Cray-1 architecture. It conipriaee t, 2 or 4 CPUs sharing common memory o.-f up to 8 Mword» in 64 banka of 38 ns ECL memory chips (fig 10).
Fig. 10.1 (HUIABS) .
Cray X-MP ayBtem organization
Each CPU has 13 pipelined -functional units which operate with data in B vector registers, each holding up to 64-bit elements. Threè memory access pipelines or ports are provided, allOHing each CPU to read two vector argument« and store one vector result simultaneously. Each CPU executes its own instruction stream, and rapid synchronization is achieved via 16 shared registers and 32 shared one-bit 'flags. The clock period is 9.5 ns, giving maximum performance o-f a pipeline i05 Mflops/s.
CRAY-3, Seymour Cray, Chippewa Falls, Misconsln. <ERH86, HUIA8S, 0ED86,) The Cray-3 is scheduled for 1987. It consists of 4 CPUs accessing a shared memory of 236 Mworda. It is an implementation of Cray-2 in gallium arsenide technology, and it is •peculated that a clock period of Ins might be obtained leading to a maximal performance of IBflops/s per floating point pipeline.
C.mmp, Carnagie-Mellon University, Pittsburgh, Pennsylvania, (HWABS, JONSO, Mflsa2, OSLBI). This was one of the most ambitious early examples of MIMD computers. This comprised 16 DEC PDP-11 minicomputer* connected to 16 memory modules by a 16x16 crossbar switch. The design started in 1971 and the machine was completed In 1975.
Cm*, Carnagle-Mellon University, Pittsburgh, Pennsylvania.
This computer was a successor of C.mmp and was based on microprocessors that had now been available.	Communication	between the
microprocessors is via « hierarchical packed switching network. A basic computer module was DEC LSI-11 microprocessor and may act as an independent computer or may be linked to a common interclusted bus with up to 14 other modules to form a tightly coupled cluster. The total Cm» 1» built-up by loosely coupled clusters. (SWA77>.
CYBERPLUS, CDC Corporation, Minneapolis, Minnesota, USA.
The architecture of CyberPlus I« based on communication via multiple ring topology. The architecture was derived from the Advanced Flexible Processor which was build for rapid analysis of photographs taken form aircraft. CyberPlus comprises form 1 to 16 CyberPlus processors connected in' -a ring and attached to a channel of a host CDC Cyber 170/800. Up to 4 such rings can be attached to the host. Communication between processors is achieved by sending information packets to the ring. The packets move round the ring at the rate of one station per clock period, until their destination Is reached.
The CyberPlus processor has 256K or S12K words of 64-bit memory for floating point data, 16K words of lò-bit memory for integers and a program memory of 4K words of 240-bit instructions. It has 15 independent functional unite. The clock period is 20 ns , giving a floating point capability of 65Mflops/s in 64-mode and 103 Mflaps/s in 32-bit mode.
ELXSI 6400, Elxsi Corp., San Jose, California. This computer is similar to CyberPlus. It contains 1 to 12 CPUs and 1 to 4 I/O processors accessing 1 to 6 memory syctems via the Bigabus. Potentially the system can achieve 72 Mips. The system incorporates three operating systems: Etnbos, UNIX BSD 4.3 and UNIX System V. 2 - they can all run concurrently. The price for 12-CPU system i» approxametely «3 million.
HEP (Heterogeneous Elament Processor), Burton J i Smith, Denelcor Inc. Aurora, Colorado ,USA. (HMA8S, LINBS, SNESS)
The HEP computer was the first commercial computer to offer the facility O'f programming with multiple Instruction stream«. A full HEP configuration comprises of 16 Process Execution Modules (PEMs) connected to 120 Data Memory Modules via a multilevel packet switching network called shuttle network. Each may have up to SO user instruction streams. But the largest system built at the time of writing has 4 PEMs and 4 DMMs and Is Installed at the NASA Boddard Space Flight Center (fig, 11) .
' Pnckt! Svikli Uli mrl
. T/O (homi»ii
l'iiViViVi'i'i'iri'ivri'i'i'i'iTi
Fig. U.I The architecture of a typical system with four processors (HWA8S).
HEP
IBM LCAP, <Loos«ly Coupled Array of processors). Enrico Clementi, IBM Kingston, USA. <CLE84, NEUB6, SIE86)
LCAP is a powerful parallel system put together
■from parts "o-f+ the shel-f". These parts are ■frtom IBM and Floating Point Systems, Beavertan, Oregon. IBM contributed a host which may be di-f + erent ir» different configurations as for example IBM 4301, IBM 4341, IBM 3081, IBM 3089, IBM 3090, and Floating Point Systems contributed attached pipeline processors FPS li4, FPS-164/MAX and FPS-2t4. Each FPS-» is a single Instruction stream computer. An example of LCAP configuration consists of seven FPS-1Ò4, each with 4f1byte of main memory, attached to an IBM 43B1 host through a 3 Mbyte/s channal, tree FPS-164 are hosted by an lBM-4341 <flg. 12). For favorable problems the system is capable of 60 Mflops/s.
AP - array processor
Fig. 12.1 architecture
Schemati c {CLE04).
diagram of the LCAP
A bottleneck in this architecture is the time needed to transfer information between the constituent computers. It is necessary to decompose a problem into substantial parts that seldom need to communicate Nith each other. Many physical and chemical problems decompose well and for those problems an LCAP i3 a cost effective computer.
A similar project to dementi's is that of Ken Wilson at Cornell University. He has linked a FPS-iOO to VAX 11/750. Hfe plans to expand the system to 4000 164/MAX to give a theoretical maKimum performance of 40 Gflops/s. New versions of LCAP are under construction LCAP-2 and LCAP-3.
MINERVA, Lawrence Widoes, Stanford University, California.
8 Intel 8080 microprocessors and four Intel 3000 microprocessors form a shared memory bus system.
PLURIBU6
This is a symmetric tightly coupled multiprocessor which was based on Lockheed SUE fflinlcomputer, this is a 1& bit computer similar to DEC PDP-11. The system has its beginings in 1972. A lot of attention was paid to software development. <f<ARB2}
8-1, Michael Farmwald, ßeorge Michael et al., Lawrence Livermore Laboratory, Li vermore, California. (HWASS)
This is the largest MIMD project, sponsored by US Navy and Department of Energy. The complete design for S-l computer comprises 16 Cray-1 class pipelined vector computers connected to 16 memory banks by a full cross-bar switch. The S-l can therefore be regarded as a "grown-up" version of C.mmp. An overall performance is expected to be 1 Gflop/s. Each of the uniprocessors is provided with a data cache of
64 Kbytes and instruction cache of 16 Kbytes in order to limit traffic through the switch. Each memory module may contain up to one Gbyte of storage, giving a total physical storage of up to 2 Gbyte. Single instructions are provided for some common mathematical functions e.g. sine, exponential, etc, and operations e.g. matrix multiply, fast Fourier transform, etc.
We have in this survey excluded logic machines that are being proposed to support the aims of Fifth Generation project, and also dataflow and reduction machines. The reason for this Is the fact that these type of computers will probably form a special group of dedicated machines and will not evolve in a general purpose parallel computer of 90'.
3. MULTIPROCESSOR SYSTEM
After we have looked at different parallel computer system architectures, we are going to concentrate on multiprocessors. We are particLilarly interested in multiprocessors because they are almost general purpose parallel computers and have therefore a great potential power to upgrade or even replace some existing computer architectures.
Multiprocessor is a single computer with multiple processors. Processors communicate and cooperate at different levels in solving a given problem. Multiprocessor is classified as MIMD computer which is defined to be a control-flow computer capable of processing more than one stream of instructions. The communication between processors may occur by sending messages from one processor to the other or by sharing a common memory. Processor» have access to common sets of memory modules and peripheral devices.
A mutiprocessor system is	controlled by one
operating system which provides interaction between processors and their programs at the process, data set and data element level. ,
Multiprocessors are classified according to organirational classification into tightly coupled and loosely coupled multiprocessors and according to structural classification into groups which have similar interconnection structures or topology.
3.1 Multiprocessor organization
Multiprocessors can be organized in a tightly coupled organization or in a loosely coupled organization.
Tightly coupled multiprocessors
Tightly coupled systems can tolerate a high degree of interaction between tasks performed on different processors. Processors communicate through a shared main memory. A small local cache memory may exist in each processor. The connectivity may be accomplished by different interconnection structures between the processors and the shared memory. When two or more processors attempt to access the same memory unit concurrently performance degradation occurs due to memory contention (fig. 13).
48
3.2. Interconnection structure
Fig, 13.1 Tightly coupled multiprocessors where a complete Connectivity exists between the processors and memory (HMA85).
Interconnection structure between the memorie» and processors is elthen
-	time shared common bus,
-	switched network,
•	crossbar
•	multistage . omega
, banyan
-	mul ti port memories,
-	interconnection network,
•	mesh,
•	cube,
•	recon-f iqurabla,
•	hierarchical .
ft term "interconnection network" is sometimes used also -for switched networks especially for multistage switched networks because they interconnect processors with memory modules. But we distinguish in this paper switched networks -from all other interconnection networks.
Loosely coupled multiprocessors
Time shared CQfnniofi bus
Loosely coupled »ysteins are efficient when the interactions between tasks are ntinimal. Processes which eKocute on different computer modules communicate by exchanging messages through a RMtssage - transfer system (Fig. 14). In loosely coupled mul ti processors each processor has s set of input-output devices and a large local memory where it accesses most of instructions and data. Sometimes loosely coupled multiprocessors are referred to as a distributed system.
The time shared common bus is the simplest and attaches every processor to every memory board (fig. 15). Sus requester, driver, and receiver perform «11 address and data handling. Because its low cost, low complexity and high maximum throughput, the common bus Interconnection structure is today the most widely used commercial type of parallel computer system, but is limited to small shared memory computers with up to 20 processor modules (ALLIANT FX/B, ELXSI 6400, FLEX/32, MINERVft).
r"
Utat bin
I/O
Onimi and arbiltr ««itch
(b| /kcanputir moduli
_CawpuUr mwhiU Q
rn-J^-F
CampuUf moduU H-1_

Up
___I
Ht»«ii trantftr iptim I HIS I
Fig. 14.1 Loosely coupled multi processors where processes communicate through message transfer system (HUA8S).
The mesaage - transfer system could be a time shared bus or a shared memory system.
Very often computer architecture in parallel processing is a combination of loosely and tightly coupled processors. Loosely coupled systems have often hierarchical organization.
0 a E
<c
tOMHOM BUS
n S

p - processor h - memory moduli
Fig. 15.1 A structure.
Switched network
common bus interconnection
The switched network system i® realized either as a crossbar switch or as a multistage switch.
A further subdivision of switched networks is divided, according to the type of interconnection network, intos
-	cross bar (C.mmp, 3-1)
-	multi-stage (Butterfly, RP3, ULTRA, TRAC).
Cedar, QF-11, HEP,
A switched system can be realized a» shared memory system or as a distributed memory system. In a shared memory system each processor communicates with each memory module through a switch. In a distributed memory
system each meniory tnadiJlB is connected a« local m^mary ta carreeponding processor. Tha role of a switch is now to interconnect the processing elements, and there are no ineaiory «nodules connected directly to the switch.
The crossbar «witch is an extension of the coininan bu3 and implements N buses -for N processors and M buses for M memory modules, ft separate «Hitch unit connects together a number of processors (P) and memory modules (M). The nodal circuits that couple the processor bus to o memory bus are the switch. A crossbar st^itch allows all processors to access memory modules simultaneously, as long as each processor accesses different memory module. Wh»n two or more processors contend for the same memory module, arbitration lets one processor proceed while the others wait by applying the same techniques as used on a common bus architecture. Since most of the logic is concentrated in the switch nodes, the complexity of a crossbar switch and its cost grows as the square of configuration size. The switch is quite likely to be the largest unit in the system and may be as expensive as one ar several of the processors.It is a good choice for systems that are not highly parallel and have about 10 powerful processors.

o-o

l3>-0—ĆH-0-----0
Fig. 16.1 structure.
crossbar switch interconnection
ft multistage network reduces the size and the cost of a crossbar switch by linking multiple crossbars as nodes in a network so that each node in a multistage network resembles a small crossbar switch. For example a multistage network that connects 16 processors to ló memories, realized in two levels consists of 4-by-4 crossbars. The switching elements are distributed throughout the system Cfig. 17).
A cost of multistage network that attaches n processors to n memories grows as n log n. All processors can access memory simultaneously i provided that no two processors try to take the same output path from a particular node. In order to reduce this limitation many multistage network architectures have extra pathways to reduce the potential Of contention.
The switch in a multistage network is complex. A type of multi-stage switching network is usually omega or banyan. It seems that the multistage shared memory switched computer is the most favored currant architecture in par-allel processing and enables efficient parallel systems with up to a few hundred processors.
In most switched systems there is both substantial local memory as well as substantial
global memory. Local mempry is often realized in a form of registers, cache or buffer memory.
PROCESS-DRS
PI PI P3 PI PS Pt P7 P8 PS P!0 PH PlI PI3 Pit PIS P1I
HI H2 H3 HI MS KB H7 Mt H9 H10 Hit Mil H13 Mil HIS HIB
MEMORIES
p - processor rl - memory module
Fig. 17.; A multistage network interconnection structure.
Mul ti port memory
A muti port memory with m ports 1» similar to n-by-m croebar switch. In a multipart memory system is the switching logic simply bounded onto the memory module.
Frequently memory module has two ports, one connected directly as local memory to one of the processors, and the other port connected to the switch. Thus each memory bank Is both local to one of the processors and globally available to the other processors via the switch.
A mul ti port memory system could also be treated as a special technical Implementation of » switched network and not as a topology.
Interconnection networks
This is a large varity of different interconnection network topologies in parallel systems. Networks are constructed in four different topologiesi
-	mesh networks (CYBER PLUS, VFPP)
-	cube networks (COSMIC CUBE, iPSC)
-	reconfIgurable network tCHiP) and
-	hierarchical network (Cm*, EBPA, BUPERNUM).
Mesh networks are one or multidimensional and are realized in square, hexagonal or other geometry. In a square mesh of dimensionality, d, each processing element is connected to 2d neighbors. The number of processing elements N^n exp d, may ba varied independent of of the dimensionality by increasing the linear dimension of mesh n. For example, a square mesh with dimensionality 2 Is connected to four neighbors. Fig. 18. shows a mesh system with 16 processing elements.
SPEEDUP 2E0 -
S 20 U) £0 eo 100 120 1U) ISO 100 200 22 0 2(0 260 PROCESSORS
- MATRIX MÜLTtPLICATtO*
.......... IMAEE CONVOIUTION
----THEORETICAL SPEEDUP
Fig. 22.1 Linear speedup in a multistage SMltched network with 236 processors for Butter-fly parallel «yatem.
4. APPLICATION OF PARALLEL SYSTEMS
Uniprocessor architectures are approaching theoretical limits in processing speed.
In high speed or real time processing tightJy coupled computer systems have to be used.
Plast parallel computer noHadays are designed ■for numerical Mork with -floating point numbers, and are build for the solution o-f large problems In physics, chemistry and engineering.
Large computer parti culary in:
capabilities are
necc
ALL 1 ANT, BUTTERFLY, CEO AR, C. «imp, Csi-», CONVEX, COSMIC CUBE, CRAY X-MP, CRAY-3, CYBERPUUS, DCA, DPP, EGPfl, ELXSI 6400, FLEX/32, FMP, HEP, IBM RP3, IBM LCAP, IBM QFll, MINERVA, ONERA, PLURIBUS, PRINGLE, SUPERNUM, S-1, TRAC, ULTRA.
These multiprocessors are designed -for large sclentl-fic and engineering problems, for CAD automation, real time .voice data multipleKlng and other computationaly Involved problem».
Some multiprocessors are more limited in applications and are considered as special purpose parallel computers designed -for one bit logic operations, or image processing, knowledge based eMpert systems or designed -for other special applications in arti-ficial Intelligence. Special purpose multlprocessors aret
CHIP, DADO, FEM, MANIP, MEIKO, PASM, PUMPS, WFPP.
5. WILL PARALLEL PROCESSINQ WIN?
Many ambitious projects in parallel processing in past have failed. For example ILLIAC IV cost four times the original contract figure and did not come even Mithin a factor of 10 of its originally proposed performance. However its influence was profound and ILLIAC IV was the first to to pioneer the new and faster emlttei— coupled logic (£CL> rather than the established transistor-transistor logic <TTL). ILLIAC IV also pioneered the use of 15-1ayer circuit boards and computer aided layout methods.
Other parallel computer systems of the 70' were also not very successful 1. For example C.mmp and Cm-» had problems with hot memories because their inteconnections structure, which was based on crossbar switch was not intelligent and could not overcome this problem. Nowadays these solutions are given, BSP NASF had a bottleneck in a central control processor and no efficient synchroniiation mechanisms were known at that time. But again ICL DAP was pioneering in an important feature of engineering design that processing element logic is mounted on the same printed circuit board as the memory to which it belongs. VLSI technology can now include processing element and its memory on the same chip.
Now the technology has advanced sufficiently to make parallel architecture practicable. Therefore we see such great interest in parallel processing.
-	complex graphic images,
-	structural analysis,
-	aerodynamics,
-	meteorology,
-	medical diagnostics,
-	research in an oil exploration,
-	research in fusion physics,
-	industrial automatixatlon,
-	processing of sensing signals,
-	genetic engineering,
-	molecular dynamics,
-	quantum mechanical problems,
-	socioeconomic models, etc.
Mathematical problems which parallel systems aret
are salved by
-	Monte Carlo simulation,
-	Hartree-Fock equation in the electron gas,
-	finite element methods, etc.
Many of multiprocessors are almost general purpose as for exampiei
é. CONCLUSION
It is net possible to predict which of these varied computer architecturea will prove the most successful in future on the market.
By analysing the performance of multiprocessors which is primarly dependent on interconnection structure one might got an insight to the development of parallel computer systems and try to predict future trends in parallel computing. It seems that in next decade the most Influence on parallel computing are going to have the projects in massively parallel processori
-	NYU Ultracomputer, whose principles are applied in I6M-RP3 parallel system,
-	Butterfly, produced by the company BBN Bolt, Beranek t« Newman,
-	Cedar, a multiprocessor supercomputer of the University of Illinois.
<a)
(b)
Fig. IB. I A mesh netMork Mith 16 Mlemertt connacted in a lattice connected a» a torua <b).
processing
CA) and
Cube networks have either hypercube architecture or cube-connected-cycle« network architecture. A cube-connscted-cycles network ie a cube where each node of the hypercube is replaced by a ring (or cycle) a-f processing elements. In a d-di menstonal binary hypercube there are d connections to each processing element, n»2, and there-fore the number of processors equals N'"2 exp d. We see that the number of processing element« can not be increaae^J without also increasing the number o-f connections to each processing element. For example, a aiK-dimensional hypercube which has 64 nodes is topologically the same as 4x4x4 three dimensional mesh with triply periodic boundary conditions.
the switch lattice configured as a mesh
the switch lattice con-figured as a binary tree
Fig, 19.1 A cube network with 4, 8 and 16 processing elements.
Hierarchical class of tnulti processor systems is realized as tree network, hierarchy of pyramides or clusters of clusters.
H
W H
U)
(bl
Fig. 20.1 A hierarchical network realized as a tree network.
Fig. 21.1 The original switch lattice in CHIP parallel computer configured as a mesh and as a binary tree.
Let us compare three most widely used topologi esI common bus, crossbar and multistage network. Seven features are going to be comparedI
1	- cost
2	- complexity
3	- max. throughput
4	- interconnect bandwidth
5	- * o-f signal paths
6	- efficiency
7	- max. * of CPU
We see Ctable 1.) that cost of a parallel .system is the lowest in a common bus topology, but e-fficiency drops with increasing the number of processors. A crossbar switch is very powerful in connecting a few processors, but the price and complexity of a system is very high. A multistage network is a good topology to interconnect a large number of processors for a medium cost.
Reconfigurable networks include all cases in which the interconnection pattern between processing elements can be changed. This is usually achieved by interspersing switching elements between the processing elements which may be controlled by a user program.
the original lattice architecture
FEATURE BUS
CROSSBAR
1	low	high (CPU exp2)
2	low	high (CPU exp2)
3	high	no limit
4	fixed by	proportional cycle time	to tt of CPU
3	large	medium
6	drops	linear
7	up to 30	up to 10
MULTISTABE NET.
medium (n legn)
medium
high
proportional to # of CPU medium linear up to 1000
Table l.t Comparison of different features for a common bus, a crossbar and a multistage parallel computer system.
New synchroniiation mechanisms for multistage switched networks are nowadays enabling almost linear speedup for systems with up to 296 processors (fig. 22).
The reasons for success of theaa projects seem td be the fact that they are devoted to development of a GENERAL PURPOSE MIMB parallel computer. Excellent per-formance results are reached particularly because they usei
-	MULTlSTftGE INTERCONNECTION NETWQRKl fl near linear speedup Is reached by 256 processors using as interconnection ctructure between the memories and processor a multistage interconnection network (Omega network)j
-	INTERLEflVINGi that is spread data uni-formaly throughout common memory modules in order to avoid contention for any one memory module;
-	FETCH-AND-ADD:	a	very	effective interprocessor synchronization operation.
An operating system seeir to be a parallel version of a UNIX-like operating system. It i« possible to achieve high performance by connection a large number of processing elements, even with "of the shelf" standard processors. It seems also that some architectural features as foi* example the size of local memory or the size of cache memory at every processor is of secondary importance for high performance of a parallel system.
7. REFERENCES
(ABUa4>Abu-Sufah W., A. Kwok, Performance Prediction Tools for Cedar« a Multiprocessor Supercomputer,	IEEE Canf. on Comp.
Architecture, 1984, p.406-413
(ABU86) Abu-Sufah W., H. Husmann, D. Kuck, On t/0 Speedup in Tightly Coupled Multiprocessor, IEEE Trans, on Computer», June 19B6, p. 520-530
<flDEB5> Adelatado M., D. Comte, P. Siron, Ph. Berger, Expression of Concurency and Parallelisfli in an MIMD environment. Computer Physics Commentar» 37, 1985', p. 63-67, North Holland
<BAR68) Barnes S. at al.. The Illiac IV Computer IEEE Trans. on Comp., August 68, p. 746-736
{BAT77> Batcher K., The multidimensional Access Memory in STARAN, IEEE Trans. on Comp., 1977, p. 174-177
(BAT80> Batcher K. E., Design of a Massively Parallel Processor, IEEE Trans on Comp., Sept. 80, p. 836-844
(BAT82) Batcher K. E., Bit Serial Parallel Processing System, IEEE Trans. on Comp., May B2, p. 377-3B4
(BEEBS) Beetem John, et al.. The GFli Supercomputer, IEEE, pp 108-115, 198S.
tB0U72> Bouknight at al., The Illiac IV System Proc. IEEE, April 1972, p. 369-38S
(CLE84) dementi E. at al. Parallelism in Computations in Quantum and Statistical Mechanics, Proceeding on 2nd International Conf. on Vector and Parallel processors in Comp. Sci., Oxford, August 84, p.287-294
ICHASć) Chamberlain Richard, Experlances with the Intel iPSC hypercube. Supercomputer, p 2429, 1986.
(DAV69) Davis R., The Illiac IV Processing Element, IEEE Trans on Comp., Sept. 69, p. 800816
(EDL85> Edler J., A. Gottlieb at.all. Issues Related to tllMD Shared-memory Computers: the NYU Ultracomputer Approach, IEEE Conf. on Comp. Architecture, 1983, p. 126-135
(EMMBS) Emmen ad, Intel's iPSCi a family of parallel computers based on microprocessors, SUPERCOMPUTER News, May 1985
(EMM86/^l) Emmen Ad, Hypercube-toy or tool?, SUPERCOMPUTER News, July/September 1986
(EMMa6/2) Emmen Ad, Vector extension for the iPSC, SUPERCOMPUTER News, July/September 1986
(ERH86) Erhel J., Parallel programming and applications on Cray X-MP, Supercomputer, Sept. 86, p. 53-60
<FiN77) Finnila, Charles A., H. Love, The Associative Linear Array Processor, IEEE Trans, on Comp., Feb. 77, p. 112-129
(6IL86} Biloi M. K., H. Muhlenbeim, Rationale and Concepts on the Supernuffl Supercomputer Architecture, MIPRO 86, Opatija, Ist Jugoslav Conf. on New Generation of Computers, p. 3.13. 17
(60TB2) Gottlieb A. at al.. The NYU Ultracomputer - Daaigning a MIMD Shared Memory Parallel Computer, IEEE Conf. on Comp. Architecture, 1982, p. 27-42
<60TB3) Gottlieb A. at al., The NYU ULTRA com-putei—^^Designing on MIMD shared tlemory parallel Computer, IEEE Trans, on Comp., 1983 •
(HANSS) Handler W. at al., A tightly coupled and hierarchial Multiprocessor architecture, Computer Physics Comm 37, 1983, p. S7-93
(HARe6> Hars N.
New Systems offer neai—
supercomputer performane®, IEEE, March 86, p. 104-107
(HOCBl) Hockney R.W. and C. R. Jesahope, Parallel Computers, Adam Hilger Ltd, Bristol, p. 126-143, 1981
CH0L8S/1) Hoilenberg Jaap The Cray-2 computer system SUPERCOMPUTER 8/9 , September 1983
(HDL8S/2) Hollenberg J., The Butterfly Parallel Processor Computer System, Supercomputer, Sept.
85,	p. 23-27
<H0LeS/3> Hollenberg J., The C-l; A Hinisuper Supercomputer, March 85, p. 7-8
(H0Pe6) Hoppe H. C., H. Muhlenbetn,	Parallel
adaptive fui 1-mul tigrad methods on	message-based multiprocessor. Parallel Computing, Oct.
86,	p. 269-289
CHUASS) Hwang K. and F. Briggs	Computer
Architecture and Parallel Processing,	McGraw-Hill Book Company, p. 237-24», 1905.
(JENai) Jenevein R., D. Degroot, G. Lipovski, A Hardware Support Mechanism for Scheduling Resources in Parallel Machine Environment, IEEE Conf. on Comp. Architscture, 1981, p. 57-65
(JEN82) Jenevein R.,J.Brown, A Control Processor for a Reconfigurable Array Computer, IEEE Conf. on Comp. Architecture, 1982, p.81-89
<JONBO) Jones A., P Schwarz, Experience Using Multiprocessor Systems! A Status Report, ACM Computing Surveys, June GO, p.121—167
CJ0Ra2) Jordan T. L. A Guide to Parallel Computation and Some Cray-1 Experiences Parallel
Computations ftP, 1932
(KAPBA) Kapauan A., J. Field, D. Gannon, L. Snyder, The PRIN8LE Parallel Computer, IEEE Corcf. on Comp. Architecture, 19S4, p. 12-20
(KARe2) Kartashev 5., S. Kartaehev, Designing and Programming modern Computers and Syoteins, vol. 1, chapter 11, Prentice-Hall, p. 143-154,
1982.
(K0CB3) Koch Wilhelm , First European installation of Si emana VP-200, SUPERCOMPUTER 7, May
1983.
(KOSei) Kogge Peter M. The Architecture o-f Pipelined CooiputerB, McGraw-Hill Bool Company, p. 1S9-162, 1981.
(RUD72) Rudolph J., A production implementation of an associative array processor- STARAN, Fall Joint Computer Conference, 1972, p. 229-241
(RUS7e> Russell Richard M. The CRAY-l Computer System Comm. ftCM, vol. 21, No. t, January 1978
(SCHBO) Schwarts T., U1tracomputer, ACM Trans, on Programming Languages and Systems, Oct. 1900, p. 484-521
(SCHaò) Schwederski Thomas and Siegel Howard Jay, Adaptable Software for Supercomputers, IEEE i Computers, pp.40-4S, February 1986.
(SEJ80) SeJnoMski et «1., Overview of the Texas ReconfIgurable Array Computer-, AFIPS National Computer Conference, 1980, p. 631-642
(KRAS?) Kramer 0. and Muhlenbein H., Mapping Strategies in Message Baaed Multiprocessor Systems (to be published).
{KUC82) Kuck David J. and Richard A. Stokes The Burroughs Scientific Processor (BSP) IEEE TRANSACTIONS ON COMPUTERS vol. C-31, No. 5, May 1982
(LECB6) Lac« P., The ONERA experimental MIMD system. Supercomputer, Sept. 86, p.91-96
(LINB2> Lincoln Nell R. Technology and Design Tradeoffs in the Creation of a Modern Supercomputer IEEE TRANSACTIONS ON COMPUTERS vol. C-31, No. 5, May 1902
(LINB3) Llnebock R., Parallel Processingi Why a Shakeout News, Electronics, Oct. 35, p. 32-34
(LIP77) Lipovski J., of tlicrapracesaors, Feb. 1977, p 125-138
On a Vari structured Array IEEE Trans. on Computers,
(SEieS) Seitz Charles L. , The Cosmic Cube, Communications of the ACM vol.28, No.l, January 198S
(BIEBl) Siegel H., PASMi A Partitionable SIM-D/MIMD System for Image Processing and Pattern Recognition, IEEE Trans on Computers, Dec. 1981, p. 934-947
(SIE86> Sieworsk D., New Trends In Comp. Ai— chltectura, MIPRO Conference, Opatija 1986
(8IPB4) Sips H., The DPPBl- an excrcise in parallel processing. Supercomputer , Nov. 84, p. 31-37
(SNE85) Snelling D., HEP Applications! real time flight Simulation, Computer Physics Comm. 37, 1983, p.261-271
(SNYBl/l) Snyder L., Programming Processor Interconnect!on	Structures,	Technical
Report CDS-TR-3BI, Perdue University, 19B1
(LLU84) Llurba Rossend ,VP-200! Fujitsu's Supercomputer, SUPERCOMPUTER 2, July 1984
<LLUB6> Llurba R., The Alliant FX/9 entry level supercomputer, SUPERCOMPUTER, March B6, p. 7-11
(MAN85) Manuel Tom, Parallel Machine Expanda Indefinitely, Electronics Meek, May 03, p. 4933
(MASB2) Mashburn Henry , The C.mmp/Hydra project! An Architectural Overview, Computer Structures: Reading and Examples, ed D. Siewo-rek. Bell, Newell, p. 350 - 370, McBraw Hill, 1982
(OED86) Oed W. , 0. Lang, Modeling, measurement's and simulation of memory interference in . the Cray X-MP, Parallel Computing, Oct 86, 343-359
(OSL82) Oslund B., P. Hibbard, R. Whiteside, A Case Study in the Application of the Tightly Coupled Multiprocessor to Scientific Computation, Parallel Computations, ed 6. Rodrigue, p. 315-364, Academic Press, 1982
(PFIB6) Pfister S. F., Parallel processor project to link 512 .32-bit micros, IEEE Computer, Jan. 86, p 98-99
(PREa2) Premkumar U., J. Browne, Resource Allocation in Rectangular SW Banyans, IEEE Conf. on Comp. Architecture, 1982, p. 326-333
(PUR74) Porseli Charles J. The control STAR-100- Performance measurements NCC 74
data
<RETB6) Rettberg Randall and Robert Thomas, •Contention is no obstacle to shared-memory multiprocessing. Comm. of the ACM, vol. 29, No. 12 p.1202-1212, December 1986.
(SNY81/2) Snyder L., Introduction to the Configurable, Highly Parallel Computer, IEEE Computer, Jan. 1981, p. 47-56
(SWA77) Swan, Fuller, Sieworek, Cm» - A Modular Multi-microprocessor, AFIPS National Computer Conference, 1977, p637-644
(ÜCH85) Uchida Keiichiro and Mikio Itoh, High Speed Vector Processor in Japan, Computer Physics Communications 37 (1985) -7-13, NH Amstei— dam
(V0N84) Vons Peter Cyber 205 vector-feature» used by vector!zers SUPERCOMPUTER 3, September 1984
(WIL82) Wilson Kenneth 8. Experiences with a Floating Point Systems Array ProcessorParal1 el Computation AP, 1982
(YAW77) Yaw B. S., H. S. Fung, Associative Processor Architecture- A Survey, ACH Computing Surveys, March 77, p. 3-27
(ZS086) Zsohar Leslie et all.. Bus Hierarchy Facilitates Parallel Processing in 32-bit Multicomputer, Computer Technology Review, summer 1986, pp 51-S9.
(NEW8S) News, China's first SUPERCOMPUTER 6, March 1985
supercomputer,
(NEWe6) News, "Supercomputer is just an ad4'ertisement word" - an interview with E. CI ementi, SUPERCOMPUTER, Sept. 86, p. 24-33
(TEC0Ć) Technology to watch", aimed at parallel processing, 86, p. 56-60
Th.i^ Mi ni super is Electronics, Oct.
A PARADIGM OF TRANSPUTER SYSTEM IMPLEMENTATION
INFORMATICA 4/87
UDK 681.3.06
Branko Mihovilovič, Peter Kolbezen, Jurij Sile Jožef Stefan institute, Ljubljana
flb«tr*ct - Ct)« linpleatantAtlon of • r«ytr«cina slaorittim on «n array of tranaputcrs if prccanttđ. »uch syftcin !)■( con« (nportant prop«rtl«*. *roces>tna *pe«ò te tftroctty proportional to tD* nunb«r of transput«rf ui*d. C»« lyit«« (f ramarltably robuft namely, tn4iy(0ual traniputert cati b« ramovad fro« tl)« syct«« «f!)ll« tu« proaraM 1« funning, and tti* (y(t*m «III continua to function «rthouaht «itti re4tic«4 parfortnanca ana toma lot* of «ata.
CEY WORDS - Transputer, oceam, raytraclng, performances.
1. INTRODUCTION
This paper describes the implementation of a computer graphics program on an array of transputers. This program was written to provide a demonstration of the performance obtainable by using large numbers of transputers, we used a technique Known as raytracing wich can generate very realistic images but requires massive amounts of computer power. This Is an Ideal application for transputers as the calculation for each picture element on the screen are Independent of one another an so can be done in parallel on separate transputers. The entire program is written in occam, however, the main part of the program could have been written in any suitable language. Only those parts of the program which deal exspllcitly with concurrency and the distribution of work are easier to write in occam.
1. BACKQKOUND 2.i. Trmnsputer
Concurrent systems can be constructed from a collection of microcomputers which operate concurrently and communicate through point to point serial communication Jinks. The INMOS transputer /i-4/ is the ideal building block of high performance multiprocessor systems which provide maximum speed with minimum hardware,
A transputer contains memory, a processor and a number of standard point-to-point communication links which allow direct connection to other transputers. The processing capability may be general purpose or may be optimized to a specific purpose. . The on-chip memory may be extended of chip by suitable interface. A transputer may also have special purpose Interfaces for connection to specific types of hardware.
The first transputer available was the IMS T4.14, a 32 bit RISC microprocessor with a
throughput of «0 mips. It has 2 kbytes of fast SHAM <50 ns) and four serial links. The 32 bit multiplexed bus allows up to 4 Obytes of external memory to be accessed. Compatible with the T414 is IMS TSOO wich Includes 4 Kbytes SRAH and floating point arithmetic.
2.2, Occam
Occam /6.6/ Is a small and elegant language. It combines the best of contemporary thought about control structures and variable scoping with some radical new structures to handle concurrency. It is based on an model of computation that is different from conventional languages in that It Includes the notion of communication, parallel execution, and synchronization in its very structure.
The basic unit of occam programming is a process that performs a sequence of actions and then terminates. There may be more than one proccess executed at any gl'v^en moment. Occam programs are constructed fron three primitive processes: assignment, Input and output. The assignment
V e
sets the variable v to the value of the expression e. The output
c I e
outputs the value of expression e to the channel c. Similarly, the input
c ? V
sets the variable v to a value input from the channel c. Constructors are used to combine processes to form larger processes. The sequential constructor, SEQ, causes its components to executed one after another. The parallel constructor, PAB, causes its components to be
executed concurrently. Finally the alternative conjtructor, ALT, choose« one component process for execution which is earliest . ready. It Is cle«re that IF and WHILB constructors are also provided In occam program.
2.3. SaxtraeJaff
Kaytracing Is a now welt introduced technique <or realistic image synthesis fron three dlnen-slonai geometric scenes, The basic raytracing algorithm Is described In /7/ and Is briefly given here. Simplifying somewhat, for e&ch picture element of the rastered Image, a ray Is t rased from the viewpoint into the three dimensional scene to calculate the first intersection with an object. If the object is reflecting or refracting, an appropriate ray is dete-mined by the law of reflection and refraction. These new rays are traced analogously, To calculate shadows, the ray-object Intersection points are connected by line segments to the point light sources illuminating the scene. If there Is an object intersecting the line segment, the intersection point lies in the shadow of this light source, and its Intensity is not taken into account for intensity calculations,
3. LOGICAL ABCHITECTUKE
The Key problem with raytracing is the relatively high unavoidable basic amount of computation. In the past, two main strategies were followed to process sets of elementary primitives: image decomposition and scene decomposition. In a former case a subset of picture elements of the Image are assigned to each of several processors. Every processor has access to the relevant primitives of the scene. In the later case a subset of primitives are assigned to each processor. A processor has access to the relevant rays of the scene.
The calculations performed for each picture element on the screen are completely Independent so they can be performed In any ordered and on any number of processors. The example of distributing the work to a number of processors is given in Fig. 1.
from host
to display Iiardirare
Plg.t: Logical architecture
This solution requires three different processes running concurrently on three, or more, processors:
-	control process <controller),
-	Intersect and shading calculation process (calculator I,
-	display process (display).
A controller Interfaces with the user or host computer to provide a description of the scene beelng viewed and allocate work to processors. A calculator can be replicated any number of times, to render the picture elements. A display collect the results from each calculator and drives the graphic display.
Every calculator is first given the description of the scene and then processing work can be allocated by the controller giving each calculator picture element to evaluate. When the calculations have been completed the results are passed out to the display. The display then informs the controller that there Is now a free processor and another picture element is sent out for evaluation.
The amount of computation required varies from picture element to element and this method automatically balances the load amongst the processors and ensures they are all kept busy. An interesting Idea here is that the picture elements do not need to be generated in sequence and, If they are generated in some pseudorandom order, a good impression oof the final picture can be obtained well before every picture element has been evaluated. This could be particularly useful In a CAD system where the user wishes to change his view of an object very rapidly.
Note, that this structure is not related to the raytracing algorlthem and is suitable for any problem which can be broken Into Independent parts.
4. PHYSICAL ARCHITECTURE ♦.i. Basle design
It appears, a first sight, that the above architecture cannot be maped directly into a network of transputers because of the fixed number of communication links available.
There is a partitioning which can aid the understanding and implementation of the structure of parallelism. The data processed on calculators consist a sequence of values (picture element), then all of the processes can be executed concurrently, even those which process the data in sequence. Alternatively the constructed process can be replicated over a number of calculating transputers each of which will execute the construct on a subset of the data structure as illustrated in Fig, 2. This hardware realisation named proces replication is mapped onto a network of transputers, active data structure is mapped onto the reconflgurab-le processor array (RPAK Both are ideal for Occam process vlrtualisatlon. The occam model adopted in the KPA system, uses point to point comnunications to synchronise processes. A processprocessor mapping is Implemented by providing a physical network of transputers, which is isomorflc to the process structure, but only at the chosen point in the hierarchy of the occam program.
It Is very simple to arrange for the controller to communicate with any transputer In a network by passing messages through the Intervening
host^
displa;
cejculator 1
patcbea
picture elamenis
Flg, 2: Physioal architecture
transputers. For simplicity, the raytraclng algorithm was mapped on to a linear array of transputers /A/.
In Fig. 2 where the basic architecture is shown we see that eacb transputer link implements two Occam channels, one in each direction. f here-fore, this mapping uses only two of the four links available on a transputer. Control and display processes are executed in parallel on one transputer (control/display transputer» while the rest of the transputers do the intersection and shading calculation processes (calculating transputers). The control/display transputer also does these calculations and the same, parameterised, program is loaded into every calculating transputer. Such method of mapping processes requires that each calculating transputer also execute routing processes, i.e. commands and data are passed from the controller along the array and results are passed back for display. This linear connection of transputers implies some sort of commaad protocol for identifying the nature and destination of data, consequently the routing process on each transputer only needs to decide whether a message is to be accepted locally or passed on to be dealt with elsewhere.
The controJ/aSspJay transpater
There are two processes executed by the control/display transputer (Fig. 3):
lendPatches, - loadBalance,
screen, called patches, rather than individual picture element, are given to each transputer for two resons;
a> to give better ratio of calculation to conni un ication*,
b) to enable segments of data to be transmitted,
k segment communication transmits an array of words as a single operation. This has two big advantages over the transmission of inctlTidual words, Firstly, there Is the same processor overhead for setting up ' the links to transmit a single word as for a million words. This better exploits the autonomous transputer links and allows the processor to continue calculating at very nearly full speed. Secondly, it gives a better ration of communication to computation,
LoadBalance coordinates the sending of data to the other transputers and the display of the generated picture elements. The first thing it does is to determine the number of transputers in the system. This is done by sending a count, incremented by each transputer, around the loop. If there are n transputers then loadBalance passes on primitives containing 2n picture elements from the process sendPatches and then waits until a result is returned before passing out another request.
4.3, The calealatlttff transputer
The work of each of cal eu latin s transputers is organised as three processes (Fig. 4):
-	thTougput,
-	render,
• feedback.
next
calculating traiusputer
Fig, 3: Processe B on control/display transputer
SendPatches interfaces to the host computer to receive the description of the scene being modelled and other commands. It passes the world model out to all the other transputers and then sends out requests for picture elements to be evaluated. Square areas of the
Fig. 4; Procoaass on th® calculating transputor
The throughput process receives patch requests from the previous transputer and either pass them on to the next or keep them for processing locally. It is also able to hold one request in an Internal buffer so, initially, it accepts two requestes. The first is passed immediately to render process for evaluation and the second is held until needed. Any further patches received are passed on to be evaluated elsewhere. As soon as render has finished the computation of its patch, throughput passes it the buffered patch to work on and is then ready to accept another. Since the time taken to compute a patch is less than the time before throughput receives the next patch request, the render process is kept busy.
The' render process is given patches to evaluate, I.e. it does all the calculations to find intersections, build the bundle of rays and then traverse this bundle to get the final picture element value. When the picture eie-
Dents In the patch are evaluated tben another patcD 15 requested <ron througnput and the picture elements are passed out to tne feedback pt'ocess. This is a completely sequential part of program which can be written in any standard prograiDBlng language.
The feedback process BUltlplezes the local result and those received from other transputers and passes them back towards the display transputer via the shortest route,
*.*. Occam imptemeatation
As we have said the transputer architecture slmplyfles system design by the use of processes as standard software and hardware building blocks. The ability to specify a hardwired function as an occam process provides the architectural framework for transputers with specialized capabilities (e.g. graphics). The required function (e.g. graphics drawing and display engine) is deflnied as an occam process, and implemented in hardware with a standard Occam Channel interface.
The Occam program of the proposed architecture (see Fig, 2) contain:
-	the description of the entire transputer system,
-	the control/display transputer program,
-	the calculating transputer program.
For simplicity, only the essential outline of the mentioned above occam programs is given here.
The system description is as follows:
VAL number.transputers IS 24; VAL last IS liumber.iFanspuiers - i:
CtiAS host, display, loopbaeic,
forwara[number.traiisputers], reiurn[ number.transputers]:
pljk:sđ par
—	transputer 0 i control/display transputer
PXOCSSSOX 0 T*
~~ data from Host PLACE host	AT llnJlOin ;
~ to display PLACE display AT IJiiHOout;
—	patches out
PLACE forward [0} AT unit lout:
—	picture element value back PLACE return[0} AT Unit tin :
control ( bost, display,
forward[0}, returnfO} )
~ the main body of the pipeline of calculators
PLACED PAS 1 s 1 FOR number.transputers - 2 PROCESSOR 1 T4
—	patches In
PLACE forwarđ[i} AT llnKOln :
—	picture elements out PLACE returntl} AT linHlout:
—	patches out
PLACE forward[l*l] AT link lout:
—	picture elements in PLACE return[i+i] AT llnKOln ;
calculate ( for»rarđ[lj, forward[l*ì}, return[i*l}, returnCl) )
—	the last transputer is a special case as it
—	has no one else to talk to. The fact that
—	the channel 'loopback' Is not placed means
—	that an internal ('soft'f channel riii be
—	created. In fact this channel is never used
—	J?ut is required as a parameter.
PROCESSOR last T4
PLACE forwardllast) AT linHOln ; PLACE return[last) AT linkOoutt
calculate ( forward(last), loopbacK, loopback, return[last] )
The program running on the control/display transputer is:
P ROC control ( CHAN fromHost, toDisplay,
tocalculators, pixelsin )
... definltlon of sendPatches procedure ... definitions of loadBalance procedure
CHAN data: PAR
sendPatches ( fromHost, data ) loadBalance ( data, tocalculators, plxelln, toDlsplay )'
Finally, each of the calculating transputers runs the following program:
PROP calculate ( CHAS fromPrev, toNext,
fromXert, toPrav }
... definition of the throughput procedure ... definition of the render procedure ... definition of the feedback procedure
CHAN toLocal, fromLocal, request/tore: PRI PAR
" run these at high priority for
—	fastes response to messages
PAR
throughput ( fromPrev, toNext, toPrev,
toLocal, requestlfore ) feedback ( fromLocal, fromNext, toPrev )
—	and this is at low priority
render ( toLocal, fromLocal, re^ue^ft/fore )
5. CONCLUDING REMARKS S.l. Performances
Whithout doubt, the processing speed of the system is directly related to the number of transputers used. A number of factors contribute to this aspect of the system. Most of these were carefully worked out design decisions but one had to be determined empirically. The transputers require only two words of data to specify the position of the all picture elements in the patch. If the work were distributed on a picture element by element basis then two words of data would be required for every element. This would mean a worst ratio of communication to processing. A more Intelligent approuch use of segment communication for data (paragraph 4.2.). These means less processor overhead per word sent and allows a greater amount of concurrency between the link engines and the processor, It Is important to-say that the. message routing processes are bad to be run at high priority to ensure that incoming messages can be examined and forwarded immediately it Is received. Carefully ordered priority in the ALT constructs of these processes are en-
sure that patches are returned to the control transputer as qulcKly as possible. Withholdings In the work In throughput and ieedbacl processes. are reduced by introducing software buffers Into two Input channels of these processes. Channel buffers are frequently used, and easy to inplenent, in occam programs. Buffers introducing nade a significant difference to the perfornance from
speed : (transputers + 1) » K/2
to very nearly
speed : transputers « E,
where K is the performance of a single transputer.
5.2. SobuđtntMS Mad r^ìlAìtlìity
The system described above is already remarkably robust. It should be possible to exploit the number of transputers with some degree of redundancy. If a calculating transputer fails then the system will progressively deadlock only if the neighbour, on the controller side, attempts to communicate with it. In order to make the system more robust it must be possible to detect when a failure has oecured. This reqlre the using a timeout on all communications. Secondly it must be possible to ensure that, If a communication does fall, all the input and output processes will terminate. These desired function are performed by a number of predefined procedures which allow an input or output to be attempted within a time Unit, an recovery from a failed communication /9A The use of these procedures means that failure of a transputer can be detected by its neighbour. The controlling transputer could then be informed and so take action to recover or regenerate the lost data. Detection of the failure of a transputer implies that facilities could be added to allow the defective transputer to be bypassed, As we remember, on each transputer two
communication links are unused, So this can be done with no extra hardware in such a way that the precedent of fall transputer switches it to the other link to communicate with the next transputer along.
6, REFERENCES
/1/ Vhitby-Strevens, C., The Transputer, Froc, 12th Annual Int'l Symp. Computer Arch., Boston, Massachusetts, 19A5, pp. 292-300.
/2/ Taylor, R., Transputer Communication Link, Microprocessors dnd Microsystems, vol.10, Ho,4, 1980, pp. 211-21S.
/3/ Mlhovllovie, B., S. Mavrie, P, Colbexen, Transputer - The Basic Componnent of Multiprocessor Systems, Informatica, vol.io, .N0.4, 19Ae, pp. «1-6«.
/4/ Mlhovllovl£. B., P. Koibezen, J. silc. Communicating Processes in Transputer Systems, Informatica, Vol.11, No.3, 19A7. pp. 74-77.
/S/ May, D,, S, Taylor, Occam - An Overview, Microprocessors and Microsystems, vol.A, No.2, 1964, pp. 73-79.
/6/ Curry, B. J., Occam Solves Classical Operating System Problems, Microprocessors and Microsystems, Vol.A, No.e, 1984, pp. 260-283.
/7/ Whltted, T., An Inproved Illumination Model for Shaded Displey, Comm. ACM, Vol,23, No,6, i960, pp, 343-349.
/tf Packer, J., Exploiting Concurrency! A Ray Tracing Example, Tech. Note 7, INXOS Ltd, Bristol.
/9/ Shepherd, R., Extraordinary Use of Transputer Links, Tech. Note I, INMOS Ltd, Bristol.
A SIMULATION OF DATAFLOW COMPUTER MODELS AND A MODEL OF FAULT TOLERANT DATAFLOW COMPUTER
UDK 681.3.068
Milan OjftterSek, Viljem Žumer, Peter Kokol, Anton Zorman
Tehnical Faculty Maribor
ABSTRACTi DatifloN architecture is «ccepted aa the «rchitectura for future coaputerf bactuce (t can' exploit potintialy all the parallelita in a prograa. Ttiis architecture is atsuaad to exacute datafloe graphs. The nodes in the datafloM graphs represent asynchronous tasks. The arcs connecting nodes represent coaaunlcation paths for the eessages (tokens) generated by nodes or tuptied froa the external environasnt. Each node i« executed (fired) «hen the required input becoees available. The dataflon arhitecture proposats can be classified as static and dynaaic architectures. 1ft a static architecture the nodes of a progras graph are loaded into aeaorv before the coaputation begins and at aost one instance of a node is enabled for firing at a tiaa. A dynaaic architecture facillitates the firing of several instances of a node at a tiaa and these nodes can be created at runtiae. He have developed prograa language DFLl, siaulator for the siaulation dataflo» coaputer aodels and their key aechanisa«, iaproved harduare, softxare and set of instruction* needed to support an efficient fault-tolarint datafloa coaputer.
t INTRDIIÜCTION
Over the last fen years data flou coaputer architectures have been proposi and sose coeputers have been built. A siepi e data floa Eiecutiun aodel is based on a 'pure* data flan prograe organisation! aith an instruction being enabled when all its input tokens are available. Eaclt instruction aay have any nuaber of inputs and any nuaber of outputs. Discussing the U-lnterpreter, Arvind and Gosteloa suggest that aaxiaua parallelisa can only be extracted froa a prograa if an arc is alloaed to carry aore than « single token - a process achieved by carrying a label with each token that identifies the context of that particular token (ARVI 82). Packet coaaunication aachine organisation and high level coaputer language with exploiting inherent parallelise in data, flow graph is the aost significant for the data flow architecture. Ne have developed prograa language SFLl t)COK S4) and siaulator for the siaulation of five various data floa coaputer aodels, Mhich are suited for today« technology. During the siaulation, siaultaneous execution of a data ft on prograa in distinct units' (token transaission and eatching, foraing activities) and soae taportant statistical paraaeters (i.e. an input streaa, a business period, an idle period, a service tiee ...t are observed (DJST 84, :UM BS-1, ZUn eS-2, OJST Bi, OJST 87-21.
2 DESCRIPTION OF HÖDELS
Dur aodels are based on the packet coeaunication eachlne organisation Nith token eatching and consist of an input section, aeaory sections, a global store, processor sections, an output section and units for coaaunication aaong sections. The input section decoeposes the data floM prograa and supports aeaory sections «ith input data. Me have used two decoaposition eethodst
- Each instruction into its oan aeaory section. If there are aore instructions than aeaory sections, then aore instructions are stored in one eeaory section. Instructions are randoely distributed.
- Each block into its onn aeaory section. If there arc aore blocks than aeaory sections, then aore blocks are stored in one aeaory section.
The aeaory section aatches tokens into sets of tokens. Hhen all of its input tokens having the saae context are available, it foras activity (executable Instruction vhich consists of a set of tokens eith the saae context and a copy of instruction) is foraed and sent into an processor section. The processor section executes the activity and sends output results into aeaory sections. The output section collects final results of cnaputation. The global store is used to prevent deadlock of the systaa. Coaaunication units (netaorks, busses) transsit aessage« (tokens, sets of tokens, activities). They use post office interconnection principle for transaission eessages. Coaaunication units in our siaulation are siaulated as delay units.
2.1 tlodel 1
henory ~ jaeetlan 1
lobil tò f»
	aenkorjr
	■66 fclon Ü

Figure liThe aodel 1
The aodel 1 uses tuo netaarksi one for transaitting activities and one for transaitting data. It has separate aeaory sections and separate processor sections. The aeaory section consists of an input FIFO
60
queue, * aatching «lore, an instruction store, an ataoclative logic and »n output FIFO queue. The procetsor »ectlon consists at an input FIFO queue,a processor, an output FlfQ queue. This configuration is not suitable for super-systees, because netMorks cause too euch del^y tlae in transoitting aessages. An advantage of this aodel is autonoaic Nork of units. The aodel is suitable for seall systeas and for a subsystea of a large systea.
2.2 Models Connected In Clusters
input ne ti on
1	glóbAl
	ntt t («orle
sluitar 1

«lobal «ton
Figure 2;Nodels connected in clusters
Hödels 2,3,4,5 are very siailar. Sections of aodels are connected in clusters.
2.2.1 Hödel 2
The aeaory section and the processor are associated in one eleaant (figure 3). The aeaory section eatches tokens into sets of tokens, foras and eiecutes activities. Output tokens ahich don't have their oMn instructions in the cluster are transaitted into butput FIFO queue. Other tokens are eatched in the cluster. Higher speed of foraing and executing activities is the advantage of this aodel. There is no additional transaitting betiaen aeaory section and processor. This aodel is very suitable for seilt and aediua systeas if the adequate decosposition aethod is used.
2.2.2 nodel 3
The processor and the aeaory section are separated (figure 4). H the processor is idle and the aeaory section hasn't any activities, it can receive activities froa other clusters. II the aeaory section has activities and the processor isn't idle, activities are transaitted into the output FIFO queue. The global netaork transaits tt to the idle processor section.
inilut FIFO ^ueu«_
butput Firo hutu.
Figure 3
Be«ory fflction ^ proo««aor	
	
1	
dotal Ittrt but
bnput FIFC kiteue		Input PEFO
		I
«wiaory + i a«soeS.fttiv« Jogi« [		
loutpul; FIF hiiflij*_
|burf*r 1	
	
|pi>oe*(	sòr i
1	
joutjlut FIFO)	
	1
glstxi a
Figure 4
2.2.3 Hödels 4 And 5
Hödels 4 and S are siailar to aodel 3. tn each cluster they have one or aore processors. The aeaory logic establishes which processor is idle and adresses the activity to it. Nodel 4 is slightly better than aodel S, because it needs less tiae for eriting in FIFO queues.
Figure 3t The cluster of aodel 2 Figure 4( The cluster of aodel 3
kua
Figure St The cluster of aodel 4

Figure &iThe cluster of aodel 5
3 SIHULATION RESULTS
98 dif-fercnt pariiEters can ba varied in aur ti»ulatlon but it i« retsonable to viry just soae of the*. Constant values an fivan to all others. Qur aodelt art (iaulatad ittth four prograas, the blocks of Nhich are coaposed c-f data fio« graphs shoan on figures 7,8,?,If. Prograa block* tre independent and they have no coaaon fio* of data tokens.
In the file of input paraaeter* ne have varied the folloNing paraaetersi
-	the selection of systea (1-51
-	the decDaposition aethod
the delay tiae in transaitting aessages through the netaork If, 1, 2 cycles)
-	the capacity of aeeory section and a capacity of global store
the input algoritha
nuaber of blocks
and aeaory sections.
Figure 7
Figure fi
Figure 9
Figure If
All data fio» graphs have IS instructions. Craphs shoan on Figures 7 and 8 have aany inherent parallelise and need aany input data tokens. The graph shoan on Figure 8 consists of instructians /, ahich are executed 3f tiaes sloaer than instructions *. Such an instruction delays the execution of folloaing instructians. The graph aith fe« inherent parallelisas is shoan on Figure On Figure If the graph aith aany inherent parallelisas and three input data tokens is shaan. Each instruction in the graph shoan on Figure II, has one input operand and one constant. This increases the speedup of firing enabled instructians. Hon are algorithas influenced to our eodels is described and evaluated.
Our siaulation is executed aith these constant input paraMtcfti
an instruction * is executed in 2 cycles an instruction / is executed in 6f cycles aatching one token in a set of tokens is executed in 1 cycle activity foraing is executed in 1 cycle FIFO queues have S eleeents aritting or reading tiae to FIFO queues is negleable
a netaark can transait an infinite nuaber of aessages
blocks of prograa are independent - a aodel 1 has S aeaory sections
He have reduced the nuaber of observed output paraaeters to the ainiaua folloaing onest a prograa execution tiae requested for execution of 2ff instructions, an average nuaber of busy processors, an utitiiaticn of prociisors, an average nuaber of busy aeaory logic in aeaory sections, an average nuaber of sets of tokens, ahich are Matched in aeeory section, an utiliiatian of a aatching store, an average nuaber of sets of tokens, ahich are aatched in global store, and an utiliiation of a global store.
The folloaing abbreviations arc used in diagraasi
UP - the utilization of processors II)
ON - the delay tiae in transaitting aessages through the
netaork (cycles!
MP - nusber of processors
TE - the prograa execution tiae (cycles)
- algoritha shoan on Figure 7
--------algoritha shoan on Figure 8
----- algoritha shoan on Figure Ì
algoritha shoan on Figure If
For the aodel 1 the decoaposition aethod 'each Instruction into its Dan aeaory section* is better than 'each block into its oan aeaory section* , so ae have aadi diagraas ahich evaluate the aodel 1, only for this aethod. For aodeis 2|3,4 ae have aade diagraas only for 'each block into its oan aeaory section* decoaposition aethod because this aethod is batter.
Th« prosru e»ciitlon tla« dependant on th* delay tiat In tnnaalttlat «tuiea ihrouiA tk* aetuork
Z. DK
a.
62
n« utlllKUon of proGHHri 4a)wnd*iit on tli* dtlar tlM in tranaBlttim aHH4H tnroii«l> tA* ntevork
MOUl. 2
tonstruction of « Urgt mnd fut »«tching uaary it a rcafontbic cost,
construction and ««nAgeoent of a itructurc uaory, netnork construction,
intorruption, trror and «xcaption hMdllng, in«fficiancy du« to iniufficltot paralUtis«, developing af^telBnt control, partitioning and schaduling algortthas.
Mc have davalopad fcty atchanisca, iaprovcd hardaara, loftaara and sat of Instructiont oaadail to support an afficlant fault-tolarant datarlo« coaauter lOMT B7-1, (MST 87-31.
Proposed datafloH architacturtt ara vary inafficiant on regular structures because qf fine granularity of their operations. Hhan data is structured (vectors, aatricas, records) the control and data flOH is very regular and predictable and there ii no need to pay high overhead for scheduling. These architectures don't have aachaniies for interruption, error and exception handling, a aechanise nhich reassigns nodes to another unit if faults have been detected, a aechanlsa «hieh stops sending tokens to tha faulty processor and a eechaniia ahtch destroying read/urite raquests In aeaory after fault. Contents of data buffer In the output port of the fail processor are lost.

7. CN
Figure III Task Invocation Structure
----- control <nd InJorurtion flam
■- data, dMi*nd *nd in&triMitiaA flow
The al gor itha shoan on Figure if is the fastest one lit has a aaxiaua degree of inherent paralleiise and a ainiaua nuaber of input dataJ. Proportions in executing tiee of prograa and an utiliiatlon of units are changed by increasing tha delay tile in the transeission through the netaork. An eKCCuting tiae of prograa is dependent on an inherent parallelise of algoritha and a nuaber of input data needed for the e>ecution of this algoritha. The executing tiae of algoritha is decreased to soae liait and then it keeps constant value (alt inherent paral leiisas are used), if the nuaber of aeaory sections or the nuabtr of processor sections are increased or delay tiae in transaission through the coaaunication units is decreased. Oaing to the fact that the processor section and the aeaory section of aodel 2 fore one eleaent, execution tiae of aodel 3 is alaoust taice as long coapared to aodel 3; but its advantage is the best utilixation of units. Because delays of reading and ari ting into F1F(] queues aere not considered, the results of aodels 4 and 3 are identical. Hith this considerations ae have found that aodel 4 is better than aodel 3. The biggest average nueber of sets of tokens in tha aeeory section is in the algoritha shoan on Figura 9, because Instruction / delays tha execution of folloMing Instructions. On eodal 1 ae have siaulated the deadlock of a systea. If aatching stores arc full, than other tokens aust aatch in the global store. This cause* bad utilization of units in a aodel (a lot of tiae is aasted for coaaunication betaen aeaory sections and the global store). He have siaulated the deadlock aith decreasing capacity of aatching stores. A good dacoaposition aethod can prevent deadlock or bad utilisation of a systea.
	
anvtriiCtior	
litor*	---1
	1 1
blob*!	
lock
Jglobal jatructurt	r-1 ! - 1	:lustcr 1
T !		
		:lustw N
T !	i	
t	'1	|l/a CllMttr
Figure IZi Global level of Fault Tolerant Datafloa nachlne (FTDFII)
EnritJr
InstructIc tor*
iluatar ■trucfcwa rtor«
|clu«v	'—•tlu«
pluck : ----Jcont
kiÈl2__"Sailt
luBtor trai
h" IV
Figura 13i Cluster level of FTDFH

Esòu

»tor« f
ESesr nktructlon toro
	hr-h-	--r-!----. I
■siäül		llnotr. 1 , . , iln»tr, [ btoro 1 btor» I
1 1		
* FAULT TOLERANT DATAFLON COBPUTER
Results of slaulation and our knoaledge about datafloa coaputers are shoan that there are aany probleas to be solved. In the design of an actual dataflow coeputer the aain probleas arei
Figure Ut Processing Eleaent of FTDFH
In our coaputer aodel (figures 12,13,14» the coabination of control floa, data fina and deaand driven are used. He adapt to the granularity of thi data structure and treate large structures as one object. He reduce scheduling overhead by coabining together as aany scalar operations as possible and executing thea as one
objett.
Hicrarchlc«! decotposition aethad <ARVI BS) «f-e used. ThI« atthod is based on the concept of resource bounded graphs <R8Gs). The KBG is • prograe frageent for Mhich a bound on the resource requireeent of the fragaent can be derived at coapile tiae as a siaple function Bf a coilectlon of paralieliso paraaeters. A prograa is vieaed «■ a collection of RBG«.
The eiecutlon of a prograa can be represented by a tree of task invocations ( figure 11). This aethod yse breadh-first partitioning algaritha until sufficient 'parallcllsa is generated and then it use depth-first partitioning aigoritha. With depth-first partitioning algoritha deadlock are avoided.
Our systea is « dynaalc architecture for task level datafioa aith facilities to support node reassignaent tthen processors fall. It uses tagging scheae slailar to the HIT dynaaic architecture. It have three hierarchical levels. The purpose of using three hierarchical levels is to execute prograas efficiently by utilizing principles of locality. Each hierarchical level have an instruction . stare, a block table, a structure store and a control unit. The instruction store holds copies of thè instructions «hich ar« executed in this lavel. The structure store holds input data structures IDS) of RBG which are executed In this level. The block table contains adresses of units uhere RBBs- are evecuted. The control unit adresses RBGs and its 03 to units in this level and generates block table. If hardaare faults occur in on« of the units, control unit reassigns RSG« and its DS to healthy units. Mith this aechanisa «e achieve that only the RB& and their DS ithich is assigned in the faulty unit aust be reexecuted. Control unit also deletes RBBs and DS «hen the coaputations of thea are finished. It also controls the aaount of available aeaory and the utilization of units. If deadlock is occured the control unit reassigns RBGs, its sets of tokens and DS froa too busy units to idle units.
Our proposed aodel has high bandwidth (HBI and loa bandaidth (LB) coaaunicatinn paths. HB paths are intended for transsaiting RBBs, DS, activities and tokens betaen units. The LB paths are intended for transsaiting status, diagnostics, control and aeasureaent inforaatlon. Our aodel has I/Q instructions, instructions ahich are executed in exception condition, instructions iapleeenting the special operators for dynaaicaly creating instances of node resulting froa recursive nodes In dataflow graphs, loop and streaas, table-oriented instructions (for readind an entry, writing Into an entry and aodifying parts of an entry), structure-oriented instructions (for selecting an eleaent froa a structure, appending an eleaent to a structure and testing for eaptiness), string aanipulatlon instructions (to search for a substring and a lenght of a string to coapare strings, and concatenate strings), streaa oriented instructions, fixed-point Instructions, logical/shift instructions, floating-point Instructions, coapound instructions (for reducing token anveaent).
S CONCLUSIOK
Today a vast collection of single-board coaputers art available which offers roughly 1 KIPS at loa cost) these are touted as building blocks for aultiprocessors. Can datafloM aachines coapete? It is not clear if a single datafioa processor can achieve the perforaance of a von Neuaann processor at the saae hardware cost. The datafioa Instruction-scheduling aechanisa is clearly aore coeplex than increaenting a prograa counter. An enginearing effort substantially beyond any of the current dataflow projects is required to aake fair coaparison. The SIGHA-i (SHIN 86) project is an iaportant step in this direction. The question becoses aore interesting when we consider aachines with aultiple processors, where the dataflow scheduling aechanisa yields significant benefits. The dataflow approach can be viewed as an extreae solutions to the eeaory latency
problea - the processor never waits far resppnses froa aeaoryi it continues processing other instructions. Instructions are scheduled based on the availability of data, so aeaory responses are slaply routed along with the tokens produced by processors. It is our belief, that dataflow architectures together with iaproved hardware, software, set of instructions needed to support an efficient fault tolerant dataflow coaputer and with powerfull high level functional languages, will show the prograaaing generality, perforaance and cost effectiveness nedded to aake parallel aachines widely applicable.
References
tARVI S2) Arvind, Gosteloa K. P.: •The U-interpreter*
IEEE Coaputer, February 19B2, pp. tZ-4B.
(ARVI 85) Arvind, Culler D. E.:
'Managing Resources in a Parallel Machine*, Proc. of the IFIP TC-ll., Conference on Fifth-generation Coaputer Architecture, Ranchester U.K., July 19BS, pp. If3-121.
IKOK 84) Kokol P., Stlglic B., Zuaer V.i
Pretvorba aplikativnega Jezika v grafe pretaka podatkov*, ETAN, IXVIII Jugoslavenska konferencija Split, 4.- 8. Juna im Split, pp. IV. 347 - IV 5S4, - in slovene.
lOJST B41 OJstertek H.i
'SiaulaciJa podatkovno vodenega raCunalnika*
Tehnièka fakulteta Maribor,
Maribor 1984 - diploasko delo -In slovene.
(ĐJST Bi) Ojstertek lt., tuaer V., Kokol P.: *Data Flow Coaputer Siaulation*, in Proceedings of the 8th International syaposiua Coaputer at the University, Cavtat, pp. 2.07 1 - 2.17 1«, May 1986.
(DJST 87-11 Ojsteriek M., iuaer V., Kokol P., loraan A.t 'Izboljiana strojna in prograaska opreaa ter nabor instrukcij, ki DBogoCajo delovanje ufinko.vite in na napake neobCutljive podatkovno vodene arhitekture*, 11 BosanskohercegovaCki siapoziua iz inforsatike 'Jahorina 87*, zbornik radova, knjiga I, Sarajevo, april 1987, pp. ISf 1-8 - in slovene.
(ĐJST 87-2) Ojsteriek M., luaer V., Koknl P.: *Dataflow Coaputer Models*, COMP EURO 87, VLSI and Coaputers, Haaburg, aay 19S7, pp. 884-885.
lOJST G7-3) Qjsteriek H., Zuaer V., Kokol P.,
Zoraan A.: *An Overview and Coaparison of Data Flow Coaputer Systeas*, Proceedings of the 9th International Syaposiua Coaputer at the University, Cavtat, Yugoslavia, pp. 2R.f6 1 - 2R.B& 6, Hay 1987. (SHIM 861 Shieada T., Hiraki K., Mtshida K.,
Sekiguchi S.i 'Evaluation of a Prototype Data Flow Processor of the SIGMA-1 for Scientific Coaputations*, Proc. of the 13th Annual International Syaposiua on Coaputer Architecture, Hashington, USA, 1986, pp. 226-234.
(ZUM 8S-1) tuaer V., Kokol P., DjsterSek H.t
'Modeli podatkovno vodenih sisteaov in ocena zaoglJivosti*, IX BosanskohercegovaCki siapoziua iz inforsatike 'Jahorina BS*, zbornik radova, knjiga I, Sarajevo, april 198S, pp. 16B-1-8, - in slovene.
(ZUM 86) Zuaer V.,Kokol P., Ojsteriek M., loraan A.: 'Parallel Coaputer Architectures, Prograaaing Languages and Algorithas,' Tehniika Fakulteta Maribor, Maribor 1986, Technical Report - in slovene.
ALTERNATIVNI POSTUPAK ODREĐIVANJA TIPOVA ENTITETA I AKCIJA U JSD METODI PROJEKTOVANJA INFORMACIONIH SISTEMA
UDK 681.3:519.863
Dubravko Živković Savezna direkcija za promet i rezerve proizvoda sa posebnom namenom, Beograd
ABSTRACT. This paper presents alternative way ot determining an initial set o-f entities and actions in the Jackson System Development - JSD - method of information systems design. The proposal is baaed on analysis of system input documents insted o-f determining entities and actions by,studying kinds o-f words from system description as suggested by Jackson in /JACK 83/. This approach has useful application in design o-f busines oriented information systems using JSD method.
SA±ETAK. Ovaj rad pokazuje alternativni postupak kojim se odreduJe inicijalni skup entiteta i akcija u Jackson System Development - JSD - metodi projektovanja informacionih sistema. Predlog se sasniva na analizi ulaznih dokumenata u sistem umeato na određivanju entiteta i akcija proudavanjem vrsta redi iz opisa sistema, ito sugeriie Jackson u /JACK 83/. Ovaj pristup ima korisnu primenu kada se projektovanje poslovno orijentisanih informaci oni h sistema vr±i JSD metodom.
1.
UVOD - JSD METODA PROJEKTOVANJA INFORhACIONIH SISTEMA
Jackson System Development ili skraćeno JSD metoda projektovanja informacionih sistema (/JACK B3/, /CAME B6/Ì nastala je iz JSP metode projektovanja programa, (/JACK 75/, /INGE 79/), primenom istih principa na Airu klasu problema. Po JSD metodi, projektovanje podinje -fazom speci-fi kaci je sistema, koja se odvija kros prvih pet koraka, a zavrtava se -fazo/n implementacije koja je ujedno i sadnji, žesti, korak JSD procedure razvoja projekta.
Koraci JSD procedure su:
1.	Entiteti i akcije
2.	Struktura entiteta
3.	Inicijalni model
4.	Funkcije
5.	Vremensko uskljadi van je
6.	Implementacija
Na slici 1. ovi koraci su prikazani - kao listovi stabla, a redosled njihovog izvođenja definisan je prefiks prolazom kroz stablo.
U praksi, međutim, razvoj sistema ne odvija se striktno po opisanoj proceduri, kretanjem od njenog jednog do drugog koraka, bez potrebe za povremenim povratkom na neki od prethodno predenih. Iteraci je svakako postoje, ali metoda raspolaže mehanizmima da ih svede na minimum.
2. ODREĐIVANJE TIPOVA ENTITETA I TIPOVA AKCIJA
Prvim korakom JSD procedure - entiteti i akcije - poćinje opisivanje realnog sveta koje se okončava u koraku struktura entiteta. Naime, proučavanjem raspoložive dokumentacije, stručne literature, intervjui sanjem korisnika i ostalim načinima upoznavanja sa problemom, projektant nastoji da identifikuje £to vedi broj	grupa,	pojedinaca,	dokumenata,
organizacionih delova, jednom rečju objekte u sistemu, koje moie proglasiti enti teti ma'-'
1) Značenje pojma "entitet" u JSD povezano je sa idejom procesa u modelu i idejom vremenskog uređenja njegovih akcija. Zbog toga se i razlikuje od snaćenja koje se, za isti termin, sreče u vezi sa bazama podataka, gde je entitet statičan, tj. niti provodi niti je Iziolen dejstvu akcija.
(kandidati za entitete). Osim potencijalnih entiteta, proučavanjem događaja koji se odigravaju u realnom svetu, projektant formira i listu mogučih akcija.
Rad na projektu, kao Ato je rečeno, projektant započinje izborom kandidata za entitete i akcije. Da bi spisak potencijalnih entiteta i akcija bio žto ^iri, jer je to vatan zahtev u ovoj fazi rada, a da se . mogučnost pogrebnog izbora svede na minimum, Jackson
Slitta 1. Procedura razvcsja JBD projekta.
projektantu predisse jednostavan postupal:i svaka imenica koja se Javlja u opisu realnog sveta i dent i-fi kuje potencijalan entitet, a svaki glagol mogudu akciju. U toj preporuci ee pod "opisom" podrazumevaju sve -forme upoznavanja sa podruijem interesovanJa. Preporuka se aasniva na sliinosti između JSD modela i prirodnog jezika, tj. na predpostavci da ljudi svoje misli israiavaju u terminima entiteta, akcija koje ti entiteti vr±e i/ili Cl jem su dejstvu izloženi. Međutim, ovaj predlog propraten je komentarom da je potrebno ■obratiti painju na ćinjenicu da mnoge imenice imaju'svoju glagolsku formu (glagolske inseni ce) i obrnuto, 4tD ma±e izazvati odredene probleme, □si m toga, moie se ponekad dogoditi da neki vaian entitet nema ime, tj. da se u opisu pojavljuje u obliku okoli ine, zaobilazne +raEe', ito otetava njegovu identifikaciju.
Ovakav pristup, pogotovu kada su u pitanju složeni sistemi, mote generisati obimnu listu kandidata za entitete i akcije, sa kojom bi dalji rad bio znatno oteian. Problem se moie donekle ublaüti na taj nadin ito se vet u procesu formiranja spiska pristupa sa odredenom dozom kritidnosti, tako da se u spisak ne ukljuiuju objekti ili dogadaji za koje je evidentno da nisu interesantni.
Slićnu proceduru, premda u drugom kontekstu, predlaiu i drugi autori (/'CHEN B3/, /BOOC 83/).
Po okončanju ovog posla, dobi jeni spisak potencijalnih entiteta i akcija treba suziti, imajući u vidu sledeća pravila:
1. Opita pravila sa izbor akcija:
- Akcija se uvek deiava u odredenom vremenskom trenutku i u naCelu nema trajanje. Na primer, u projektu LD, "REDDVAN-RftO" ne moie biti dopustiva akcija entiteta "RADNIK" poito rad
radnika u redovnom radnom vremenu ima atribute koji se direktno odnose na trajanje redovnog rada, osim ako je vremenska skala projekta tako odredena da se ćitav period vremena u kome Je rđdnik radio smatra kao trenutni dogadaJ.
Za akciju moie biti odabran dogadaj koji se deiava u realnom svetu, a ne u sistemu. U tom sveti u, na primer, "OBRAČUN-LD" ne moie biti. va±e<ia akcija jer se ne odigrava van sistema.
Akcije su nedeljive, tj. ne mogu se dekomponovati na podakcije ili neke droge sastavne delove.
2. Opita pravila za izbor entiteta:
ili
biti
- Entitet mora provoditi izložen dejstvu akcija. Entitet mora postojati u realnom svetu van raCunara, a karakteriie ga odredeno trajanje.
Entitet karakteriie uređenost akcija, odnosno definì san redosled po kome se akcije odvijaju u vremenu (struktura entiteta), a koja se prikladnim sredstvima moie opisati. Stanje entiteta se menja tokom vremena.
Osim nabrojanih opitih kriterijuma, projektant je slobodan u izboru posebnih koji su prihvatljivi za projekat koga razvija, imajući u vidu da entitet ili akciju treba uključiti u konaian spisak entiteta i akcija ako se planira da sistem u eksploataciji treba da proizvodi ili koristi in-formaci je o njima.
Odabranim tipovima entiteta i akcija projektant daje kratka i razumljiva imena. Međutim, opisivanje entiteta i akcija samo
niihovim imenima nije davoljno, jer bi razliti te osobe mogle na različite načine ova imena tumađitl. Zbog toge se jai zahteva da akcije, pored naziva, budu dodatno opisane sas
1.	Kratkim neformalnim opisom koji i tna za cilj detaljniju specifikaciju semantike naziva akcije.
2.	Atributima, <ii<ne se znaeenje akcije Jo£ bi i te odreduje. Entiteti, medutiin, ne zahtev«ju dodatan opis jer su dovoljno opisani akcijama u njihovom sastavu.
Blavrti dokumenat koraka entiteti i akcije je (prećiićen) spisak tipova entiteta i tipova akcija. Akcije su, saglasno zahtevu metode, dodatna specificirane opisom i atributima. Sa ovim spiskom kao ulazom, započinje rad na sledečem koraku JSD procedure.
3. ALTERNATIVNI POSTUPAK ZA ODREBIVANJE TIPOVA ENTITETA I AKCIJA
U ovom odeljku biće prezentiran alternativni postupak za odredivanje tipova entiteta i tipova akcija koji se moie koristiti pril.ikom pro j ek tovan j a poslovno or i Jent i sani h informaci onih sistema (IS). U osnovi tog postupka Jè analiza dokumenata koji iz raznih izvora ulaze u sistem. Ovakav pristup ima opravdanje u tlinjenici, da poslovno orijentisani sistemi mogu preći iz jednog stanja u drugo samo ako postoji odgovarajući dokumenat koji opravdava ili regi struje takvu promenu. Sa druge strane, promena stanja sistema u JSD terminima nastaje pod dejstvom akcija odredenih tipova entiteta. Iz toga proizlazi da je razumno predpostaviti postojanje odredenog stepena korelacije između skupa akcija u JSD modelu i skupa dokumenata u posmatranom poslovnom sistemu.	Dalja
razmišljanja u tom pravcu navode na zakljuiak da se, u ovoj fazi, izvori dokumenata mogu poistovetiti sa JSD entitetima. Inicijalan skup tipova entiteta i tipova akcija, koji je dobljen na opisan naćin, potrebno je dalje analizirati koriićenjem koncepta funkcionalne zavisnosti, a pridr±avajući se i opitih JSD kriterijuma za izbor entiteta i akcija, da bi se dobio njihov konaćan skup.
□vakav postupak, nesumnjivo, daleko br±e dovodi do završetka prvog koraka JSD procedure od onoga koji je predlo±en u referenci /JACK 93/, Jer Je broj dokumenata i izvora njihovog nastajanja daleko manji od broja imenica, glagola i drugih vrsta reći. pomoć.u kojih se opisuje funkcionisanje dotićnog poslovnog sistema. Samim tim Je i posao projektanta manje obiman, njegov rad je produktivni ji•
U ovom odeljku, dakle, biće izlo±en postupak za odredivanje tipova entiteta i akcija prilikom projektovanja poslovno orijentisanih informaci oni h sistema koji se zasniva na analizi ulaznih dokumenata. Nakon neophodnih objašnjenja, izlaganja će biti zakljućena definisanjem procedure koju projektant u radu treba da sledi.
izvore, uzroćnike, njihovog nastanka. Na taj naćin biće formirana inicijalna lista tipova akcija i entiteta. Akcije će u ovoj fazi biti poistovetene sa odgovarajućim tipom dokumenta, a entiteti sa uzroćnikom njihovog nastanka. Opravdanost za takav postupak moie se ilustrovati na mnoštvu primera. U sistemu LD, na primer, poveri lac upućuje u RO administrativnu zabranui, pa se zbog toga mole tretirati kao potencijalni tip entiteta koji provodi akciju tipa "zabrana". Sličan je primer, takode, sa dobavljaćem - tip entiteta -koji Je poslao fakturu - tip akcije - za isporućenu robu, itd. itd.
Dokumenat koga projektant izgrađuje dok radi na ovom zadatku, slićan je finalnom dokumentu iz koraka entiteti i akcije. Razlika postoji jedino u naćinu tretiranja atributa akcija, kojima se, kada Je u pitanju metod zasnovan na analizi vrata reći, pridaje manja vainost u odnosu na onu koju imaju kada se entiteti i akcije odreduju na osnovu ulaznih dokumenata. Zbog toga, skup atributa neke akcije mole, u prvom slućaju, biti i nekompletan. Jer sluti prvenstveno žbog toga da poboljša semantiku akcije. Kada se entiteti i akcije određuju na osnovu ulaznih dokumenata, tada se svi podaci koji se nalaze na odredenom dokumentu tretiraju kao atributi akcije koja Je izvedena na osnovu tog dokumenta. Dodatan zahtev, koji se tom prilikom pred projektanta postavlja, sastoji se u tome da treba izvršiti unifikaciju naziva atributa, odstranjivanjem tiv. sinonima i homonima.
Skup potencijalnih entiteta i akcija koji Je odreden pòistovećivanjem ulaznih dokumenata sa akcijama i uzroćnicima njihovog nastanka sa entitetima, u nekim slučajevima nije u dovoljnoj meri precizan. Ta nepreciznost mote se odnositi kako na dobi jene tipove entiteta tako i na tipove akcija. Regularni tipovi entiteta su U takvim sloćajevima podla±ni pogrešnoj identifikaciji, dok Je za skrivene tipove entiteta karakteristiAno da su jednostavno izostavljeni iz razmatranja. Nepreciznosti prilikom odredivanja tipova akcija najžešće su takve prirode, da se manifestuju nerazlikovanjem zajednićkih akcija više tipova entiteta. Od projektanta se, radi toga, oćekuje da sa inicijalnim skupom entiteta i akcija nastavi rad. U tom smislu njegova patnja treba da se usredsredi na atribute a:kcija, sa ciljem da , se identifikuju funkcionalne zavisnosti koje postoje medu atributima. Na osnovu tih zavisnosti, potrebno je	odrediti kljućeve relacija koje
korespondiraju inicijalno odredenim akcijama. Nakon toga, provodi se, eventualna, korekcija identifikovanih tipova entiteta i preciznije odreduju tipovi akcija.
Zbog kontinuiteta izlaganja, u nastavku je dat primer koji ilustruje postupak odredivanja kljuća realaci je kada su poznate funkcionalne zavisnosti, a za detalje s tim u vezi potrebno je konsultovati literaturu, npr. /ULLM 82/, /ALAG S4/ ili /LAZA 05/.
PRlMEft.
Posmatra se relaciona iemas
3.1. ODREBIVANJE TIPOVA ENTITETA I AKCIJA ANALIZOM ULAZNIH DOKUMENATA
Osnovni zadatak projektanta, za vreme dok je angatovan na upoznavanju sa sistemom, sastoji se u tome da identifikuje sve tipove ulaznih dokumenata koji su vatni za rad sistema. Uporedo sa tim, on treba da odredi i objekte iz okrutenja koji predstavljaju
za koju vati sledeći skup zavisnosti F:
funkcionalnhih
1.	A»
2.	At
3.	Aa
4.	Aa
-> A., -> A^
-> Aa -> Ab
Po definiciji, klju4 relacije treba da zadovolji osobine jednoznaenosti i
neredundantnosti■ Ove osobine zadovoljàva Jedino par atributa At Aa, pa oni predstavljaju kljuž relacije R.
Kada .se svim kandidatima za akkcije, odnoeno relacijama koje» ih predstavljaju, u nekom konkretnom primeru u praksi odrede ključevi, dobi jeni rezultati mogu se
prezentirati prikazan na	tabelarno slici 2.	na način	koji je
Ključevi J	Ki ! K»	! K, ; . . 1 1	. ; K„ ■!
Akcije •	. 1		i 1
A» !	X ! >i	) {	1 1
A. !	! K	! >i •	t 1
; 1	1	1 1 1 «	1 1
* . !		1 4	( i
A. !	1		) X !
Slika 2. Tabela koja sumira podatke o kljueevima relacija boje su iivedene iz ulaznih dokumenata u sistem.
Kada su u tabelu uneieni podaci a svim akcijama, projektant je isvriio neophodne pripreme za odlutujuiiru faru odredi van ja entiteta i akcije.
Skup entiteta odreden je skupom kljućeva koji su upisani u zaglavlju tabele (Kt, Ka,..., K„>. Svaki kljui identifikuje jedan entitet, pa u apstraktnom primeru, ćija je tabela prikazana na slici 2. , postoji n tipova entiteta. Na osnovu podataka o ključevima, međutim, nije moguite odrediti da li neki entitet pripada kategoriji regularnih ili skrivenih entiteta. Da hi se napravila ova razlika medu entitetima, mo±e se postupiti na dva naäina.
Prvi se aasniva, na analizi preslikavanja koja postoje između kljućnih atributa. U tom sluiaju, ako između dva kljuća postoji preslikavanje	<l:hl),	tada je entitet
predstavljen kljutem koji u preslikavanju učestvuje sa kardinalnoiću M, kandidat za skriveni tip entiteta. U sistemu LD, na primer, iameda matičnog broja radnika (ključ Ki) i broja partije koji predstavlja deo ključa entiteta ODBITAK (ključ Ka), postoji
preslikavanje Ki <->-> Ka, tj. preslikavanje
(liM), Ovaj tip preslikavanja postoji i iamedu broja partije i ii-fre poverioca (ključ K,) , za
koje se takode moie pisati K» <->—> Ka. Zbog
toga se odbitak mole tretirati kao kandidat za skriveni tip entiteta.
Drugi način u analizu uključuje akcije, odnosna preciznije, akcije i pridrulene ključeve. Po tom postupku, ako neki entitet Ei, koji je predstavljen ključem Kj, nema ni.jednu akciju kojoj je pridruien samo jedan ključni atribut, u konkretnom si utaju ključni atribut Kl, predstavlja kandidat za skriveni tip entiteta. Do tog zaključka se dolazi ako se ima u vidu da skriveni tip entiteta karakteriiu isključivo zajedničke akcije sa drugim tipom ili tipovima entiteta. Zajedničke akcije su, dakle, one Vcojima je u tabeli pridružen sloieni ključ. Ako se kao primer i ovog puta uzme sistem LD i zadrie uvedene oanake za ključeve, tada če admini strativna zabrana (koja je u prethodnim izlaganjima bila predstavljena akcijom ZABRANA) biti identifi kovana ključevima K., Ka i K'3. Slična je stvar i sa ostalim
tipovima akcija za koje je u tabeli naznačeno da sadle Ka. Svi oni, pored Ks, imaju i druge atribute (Ki ili Ka) koji formiraju sloieni ključ, S druge strane, nalog da se poverilac uključi u registar poverilaca (akcija EVIDENTIRANJE) ili odluka o prijemui radnika u radni odnos (akcija ZASNIVANJE-RADNOG-ODNDSA), imaju ključeve koji se sastoje od jednog atributa. Ka i Ki respektivno. Barnim tim, u sistemu LD, projektant'ima osnova da zaključi da su RADNIK i POVERILAC regularni, a da je ODBITAK skriveni tip entiteta.
Vaino je primetiti da entitet tipa ODBITAK nije ni identi+ikovan, niti je to mogao biti, u prvoj fazi rada. Do te spoznaje moglo se doči tek nakon analise funkcionalnih zavisnosti.
Na sličan način bi se provodila analiza imajuči u vidu i ostale tipove dokumenata odnosno akcija, tako' da njenim okončanjem projektant dobija konačnu listu entiteta i akcija sa kojom mole otpočeti aktivnosti iz sledečeg koraka JSD procedure.
Razume se da je projektant slobodan u izboru postupka koje če U' radu koristiti, a, moie ih i kombinovati, tako da rezultate do kojih je doAao na jedan način proveri pomoču drugog. Međutim, bilo kakvo da Je njegovo opredeljenje, on treba u toku rada uvek da ima na umu i opite JSD kriterijume za izbor entiteta i akcija.
Da bi skup entiteta i akcija odreden analiiom ulaznih dokumenata bio kompletan, potrebno je da projektant u obzir uzme sve tipove dokumenata koji au bitni za rad odredenog poslovnog sistema. To neće uvek biti lako, a razlozi koji do toga dovode mogu biti različiti. Jedan od njih Je svakako neiskustvo projektanta, a zatim sledi njegovo nedovoljno poznavanje oblasti poslovanja, za koju treba projektovati IS u fazi kada Je potrebno odrediti entitete i akcije. Međutim, isti problemi postoje i ako se entiteti i akcije određuju postupkom koji je predlolio Jackson. U ■ slučaju poslovno orijentisani h informacionih sistema, oni su čak i izraieniji ako se umesto analize ulaznih dokumenata koristi neki drugi postupak. Problemi takode mogu nastati i u slučaju kada je stepen organizovanosti konkretnog poslovnog sistema niii od uobičajenog, kada u sistemu postoje mnoge priručne evidencije koje nemaju zvanični karakter, kada iz okru±ei;)ja dolazi loie kreirana dokumentacija itd. Istralivanje odnosa apstraktnog skupa ulaznih dokumenata ("Idealan" skup dokumanata), koji bi omogučio jednoznačnu identifikaciju entiteta i akcija u nekom sistemu, prema realnom skupu dokumenata, verovatno hi rezultiralo interesantnim zaključcima koji bi se koristili u procesu odredi van ja entiteta i akcija. U ovom pravcu bi se mogla odvijati dalja istraživanja!
iJz malo dodatnog truda projektant mole, ako entitete i akcije odreduje postupkom analize ulaznih dokumenata, Jednim potezom obaviti dva posla, tj. defini sati skup entiteta i akcija i i zvržiti pripreme' za odredivanje strukture entiteta. Radi se o tome da je potrebno, u toku upoznavanja sa dokumentima i njihovim izvorima, prikupiti podatke i o tragovima potencijalnih entiteta. Do tih. podataka on mote doči paliJivijim proučavanjem postojećeg sistema evidencije koja sluii za registrovanje primljenih dokumenata.
Ako se u postupku određivanja entiteta i akcija analizom ulaznih dokumenata, anga±ovanje projektanta na uočavanju tragova smatra pozitivnim, to se bez izvesne ogrjide ne tno±e prihvatiti u vezi zahteva da je potrebno, ako se postupak ±eli korektno primeniti, proučiti funkcionalne zavisnosti koje postoje medu atributima. Naime, poznavanje funkcionalnih zavisnosti nesumnjivo je korisno i sigurho če biti iskor'iičeno za pro jektovan je baze
paflataka, ali se ange.iavanjE? pirojektanta na tcjin poslu u koraku entiteti i akcije smatra proLiran jenim. Taj ;ahtev je Ebog toga i najveći nedostataU Dvog cfostupka.
3.2.
PROCKDURft Zft ODREDI VAN,JE-: TIPOVA ENTITETA I AKCIJA
Na kraju, postupak za odredivanje entiteta i akcija metodom analize ulaznih dokumenata biie izloien u formi procedure. Dna treba eia Dlakža primenu opisanog poEittipka u praksi, poAto doprinosi da korak entiteti i akcije dobi ja strukturni ju forrau.
koje su tom prilikom doneiene potrebno je na prikladan naiiin dokumentovati ä to se postiie iiradom spiska entiteta i akcija. Ovaj dokumenat treba da udovolji sahtevima JSD metode, ito znaii da svaka akcija i entitet treba da imaju kratko i razumljivo ime i da se akcije, osim atributima,	-dopune i kraćim
ne-formalnim opisom.
Definiranjem procedure za odredivanje entiteta i akcija metodom analize ulaznih dokumenata, okoniava se i izlaganje koje se na ovu temu odnosi.
Procedura BS sastoji iz 6 koraka i' to:
1. korak.
korak.
Proudavanjem dokumenata koji postoje u RO, ćritanjefTi atrucne literature, iz saradnje sa korisnikom i drugih raspoloživih izvora, potrebno je identifi kovati rasliiite	vrste
dcjkumenata koji iz okruienja dolaze u poslovni sistem 4ij i rab treba automatizovati.	Za svaki tip
dokumenta potrebno je odrediti mesto ili Dscbu koja je odgovorna za njegovo kreiranje.
Poistovetiti svaki dokumenat sa odredenim tipom akcije, a entitete sa mesti ma odnosno osobama koji dokumente generiiu. Rezultate treba prezentirati u formi spiska koji je slićan konaćnom dokumentu za korak entiteti i akcije;, ali se u ovoj -fazi od projektanta ne zahteva da akcije dopuni tekstuaJtilm opisom. Međutim, atributima akcija treba pokloniti vedru painju. U tom smislu podaci sa dokumenta postaju atributi akcija, a od projektanta se oiekuje da njihove nadive ujednaći i isbegne pojavu sinonima i hamonima.
3.	korak. 1 dent ifikavati skup funkcionalnih
zavisnosti nad skupom atributa svake akci je.
4.	korak. Na osnovu poznatih funkcionalnih
zavisnosti odrediti kljućeve relacija koje koresp-ondi ra ju	inicijalno
odredenim akcijama odnosno ulaznim dokumentima. Rezultate prezentirati u tabelarnoj formi na takav naiin da SE za svaku akciju naznati njen kljućni atribut ili atributi.
5.	korak. Koristeći se tabelom koja je urađena
u 4. koraku, preciznije definisati skup entiteta i akcija od onoga koji je identifi kovan u drugom koraku procedure. Konaina odluka o tome da li će neki od kandidata za entitete i akcije biti odabran za nastavak rada, koji će od njih biti ukljućen u finalni dokumt^nat csvö procedure, odnosno finalni dokumenat koraka entiteti i akcije, zavisi i od toga da li je on u Bkladu aa ppitim J<iO kriter-i jumima iia oriredivanje entiteta i akcija.
ZAKLJUČAK
U ovom radu predioien je alternativni poatupak za odredivanje tipova entiteta i akcija u JSD metodi projektovanja IS. Dn se zasniva na analizi ulaznih dokumenata u siatem, a ne na analizi vrsta refii sa kojim je sistem opisan, kao ito predlaže Jackson. Kod poslovno orijentisani h IS prediD±eni postupak brie dovodi do reienja, a defini sana procedura prvom koraku JSD procedure projektovanja IS, "Entiteti i akcije", daje strukturni ju formu.
Nedostatak predioienog postupka je preuranjena analisa funkcionalnih zavisnosti medu atributima, koja se provodi da bi se precisnijo odredio skup entiteta i akcija.
LITERATURA
/ALAG 84/ S. Alagić, "Relacione baze podataka", Svijetlost, Sarajevo, 1984.
/EDOC 93/ G. Booch, "Software Engineering with fida",	The	Benjami n/Gumminqs
Publishing Company, 1963.
/CAME Sh/ J. R. Cameron, "An Overview of JSD", IEEE Transsacttons on Software Engineering, No 2, 1986,
/CHEN 83/ P. P. Chen, "English Sentence Structure and Entity-Relationship Diagrams",	Information Sciena,
Elsevier Science Publishers, B, V. (North-Holland), 1983.
/INGE 79/ L. Ingevaldsson, "JSP - A Practical Method	of Program	Design",
Studenti itteratur, 1979.
/JACK 75/ M. Jackson, "Principles of Program Design", Academic Press, 1975.
/JACK S3/ --- , "System Development", Prentice-
Hall, 1963,
/LAZA B5/
B. Lazarević informacioni h knjiga, 1985.
i dr. "Fr o jektovanJe sistema",	Naudna
korak. Kada su de f i ni sane
odabrani entiteti i njihove akcije, odluke
/ULLM 8:2/ J. D. Ullman, "Principles of Database Systems", Pitman Publishing, Ltd., 1983.
PROGRAMIRANJE Z MODULO - 2 II
INFORMATICA 4/87
UDK 681.3.06 MODULA - 2
Andrej Brodnik, Marjan Spegel, Tadej Lasbaher Institut »Jožef Stefan«, Ljubljana
Ta prispevffli obsega primerjavo nodule-2 s programski«i jeziki paecatt ada in c, sploäen opis prevajalnika . za iiiodulo-2 In opi.« najbolj vidnih prevajalnikov t« paiUftne operacijske -Bistene. Sklepni del prispevka je namenjen kratkemu opi^u zanimive združitve znoZnosti funkcionalnega in proceduralnega jezika (Tiiodule-2 in prologa). V dodatk\» navajamo dva pcinera programov v moduli-2, ki naj bi predstavila tako osnovna nafiela sistemskega programiranja, kot postopek nodularnega programiranja. Ta prispevek je nadaljevanje prej objavljenega članka in zaokroluje kratka predstavitev jezika modula-2.
Prograawing Ulth l1sđula-2 In this paper the comparisson of Modula-2 with Pascali Ada and C is given. Further, the general overview of compiler is presented with «orne «pecific examples for several operating systems. Finally, the very emart connection of functional and procedural language (Prolog and 'nodula-2) is given. The appendices contain two eMantples of programmes which should give a flavour of modular and system programming in Modula-2. This paper with prevecusly published one clases the sketching of nQdula-2 programming language.
_um
bil prv.i del prispevka namenjen Jeziku , potem poskušamo v tem delu jezik
Ce je samemu
postaviti nekam v okolje. Tako navajamo nekatere podatke, ki bodo pomagale oceniti razmerje madule-2 do drugih programskih jezikov. Temu so namenjena prva tri poglavja.
V	nadaljevanju opisujemo sam princip prevajanja programov napisanih v moduli-2 in nivojskl ustroj splošnega prevajalnika. V zaključnih dveh poglavjih pa poskuftamo prikazati nekaj zelo zanimivih primerov uporabe jezika na dokaj različnih prinerih zunaj in pri nas.
V	posebnem dodatku vam predstavljamo dva lupini programov, ki :sta napisana v moduli-2 in sicer razporejevalnik' procesorskega časa in program za pogovor med dvema terminaloma.
2, flgfftf^,^-^
2e v prejSnJem sestavku o noduli-2 smo opozorili na to, da Je modula-2 kakovostna nadgraditev pascala. Po njem je podedovala predvsem bločno zgradbo in vse osnovne tipe stavkov, od njega pa se bistveno razlikuje predvsem v naslednjih pogledih
a)	gradnik "modul" omogoCa
-	ločeno prevajanje programskih in podpro-gramsklh modulov
-	enostavno gradnjo modulov s programskimi orodji
-	knjilnice, ki so podprte g sistemom jezika
b)	nizkonivojsko ali sistemska programiranje
-	dostop do poljubnega stalnega mesta v pomnilniku
-	dostop do posameznih procesorskih registrov
-	pisanje preklnltvenih podprogramoy
-	pisanje splolnih sorutln
C) podprogram Je definiran kot poseben tip
d} sintaktične razlike
Razlike pod točkami a, b 1 c so bistvene in so semantične narave, medtem ko so razlike pod točko d predvsem oblikovne oziroma sintaktične narave. Ker pascal ne pozna pojma modula in ker ne omogoča sistenskega programiranja, ga na teh dveh področjih tudi ne moremo primerjati z modulo-2.
2.t Padpragru kot tip
Podprogram kot tip Je v zelo omejenem smislu sioer poznal ie pascal in sicer kot parameter podprograma C&t	ki je moral biti brez
parametrov. V moduli-2 Je povsem pravilna definicija naslednjega tipai
TYPE
ProcType
PROCEDURE <Typel, Type2)t Type3|
Parametri podprogranov so prav tako lahko kakrfnegakoll tipa.
2.2 8int«ktiena razlika
t1adula-2 Je podobna pasoalu tudi po zunanjem videzu. Vendar vseeno obstajajo nekatere razlike ned obema jezikoma in jih navajamo tukajs
a)	mDdula-2 loči med velikimi in malimi črkami. Na primer; ana ni ista spremenljivka kot ENA
b)	pascal pozna dva tipa komentarjev, medtem ko Je v maduli-2 dovoljen samo eden, ki pa ga lahko gnezdimo. Tako v moduli-2 { to ni komentar >, (♦ ampak to •), C* ki Je lahko <* gnezden *).
c)	modulA-2 pozna več osnovnih tipov kot pascal.
d)	v moduli-2 so dovoljene različne okrajšave za veččrkovne operatorje. Znak neenakosti O se zapiie kot logični AND Je dovoljeno okrajšati v 4.
e) modulft-2 ne pozna sestavljenega stavka, kajti likale s«, da Je nepotreben. Tako Jb besedic« twgtn v naslednjon primeru povsem odvefinai
pascal i
for	to 10 do
tiagin
stavekli
Stavek2t
etavekn and;
inadula-2 i
FOR i 1=1 TO 10 DO «taveklI stavQk2|
Btavekn EHD;
n inodula-2 pozna brezpogojno zanko (LOOP stavek).
g)	stavka IF ol THEN si ELSIF c2 THEN s2 ELSE s3 END; pascal ne pozna.
h)	v ■odull-2 se vsi stavki zaklJuCiJo z rezervirarto besedo END, kar Je razvidno is iz primera za sestavljene stavke.
1) v inoduli-2 Je povsen prepovedan goto stavek.
J) raiUk« Je pri definiciji (unkcijskih padprogranov.
k) itiodula-2 ne pozna standardnih podrogranov gat, put, read, readln, write, writeln za delo z vhodno izhodnimi enotami.
U inodula-2 ne pozna tipa file.
Tu naStete razlike seveda niso edine, vendar so po na^en mnenju najpomambneJfie.
BITSET 1« «rray CO..15) at BOOLEAN;
in podatkovni tip ADDRESS, ki ga Uhko v definirano kot
adi
ADDRESS
MORD;
Razlike ned ukaznimi strukturami Jezikav so bistveno manjtte. Ob« Jezika poznata podobne osnovne stavke, vendar ina ada pri ten nekaj ve6 moZnosti. l1adula-2, na primer, ne pozne stavka daolAra (nadomestimo ga lahko i uporabo novega modula) in d«lay stavka <nadomestino ga s klicem ustreznega, podprograma). Nodula-2 tudi povsem prepoveduje uporaba GOTO stavka, ki pa je v adi dovoljen. Naslednji stavek, ki ga najdemo v adi in ne v naduli-2, je null stavek. V resnici obstaja tudi v moduli~2, vendar ga tam ne napifieno, saj Ja predstavljen s praznim stavkom, ki ima obliko saino podpiCJa ali celo niäesar. Kako v moduli-2 nadomestimo raise stavek iz ade, pa bomo opisali na primeru nadomefiftanJa strukture eKoeptlon in njene uporabe.
Ada Je tudi precej mo&nejea kot modula-2 pri uporabi podprogramov, vendar lahko prednosti ade doseiemo s pametnim programiranjem tudi v moduli-2. Tako adino prednost, da Je rezultat funkcijskega podprograma lahko tudi strukturiran tip, v moduli-2 nadomestimo a tam, da izračunani rezultat vrnemo preko parametra, ki Je bil klican po naslovu (referenci). Paö pa v moduli-2 ne obstaja moSnost ponovnega definiranja osnovnih operacij, kot so na primer nnolenje, seetevanjs in podobno Cinfiksne operaoije), in si moramo zato pomagati s klici ustreznih podprogramov.
3. AD* IN IIODULA-2
Ada C21, 2*3 in modula-2 sta po letnici rojstva vrstnici, zato ni nifi öudnega, fie v obeh Jezikih najdemo nekatere bistvena podobnosti, kljub bistvenim eintaktiCnim, zlasti pa semanti&nim razlikam. Zelo pomembna razlika Je 2e kar velikost jezika. Kot navaja Pomberger C203, obsega formalna definicija ade vefi sto strani, medtem ko za Riodulo-2 zadostuje ie dobrih 2S. Temu primerna Je tudi razlika v velikosti prevajalnikov in kode, ki Jo naredi prevajalnik. lato na majhnih računalnikih popolnih prevajalnikov za ado praktično ni, medtem ko Jih za modulo-2 najdemo kar nekaj C9, 223.
Podobnosti med ado in modulo-2 si oglejmo na primerjavi podatkovnih in ukaznih struktur, na primerjavi modularnosti in moZnosti loCenega programiranja ter na primerjavi sposobnosti Jezikov za ni ikonivojsko programiranje in za vzporedno izvajanje postopkov.
3.1 Podobnost podatkovnih In ukaznih struktur
Ada in modula-2 poznata podobne osnovne podatkovne tipe, le da ada obvlada tudi fttevila v takoimenivani stalni vejici. V naduli-2 realiziramo ta fitevila s sestavljenim tipomt
FixedPoint
RECORD
numi CARDINALI base, del tal REAL END;
Razen tega ada pozna podtipe, ki Jih v moduli-2 lahko vodimo kot delno tipe, po drugi strani pa ada ne pozna tipa PROC. Poleg tega iiodula-2 pozna tip mnolice, ki ga v adi realizirano s tabeloi
3.2 Posebnosti in osnovni podprograsl
V adi obstajata dve strukturi, ki Jih modula-2 nima: prva Je «Nosptions (posebnosti) in je namenjena predvsem za obravnavanje posebnih stanj (na primer napak) v izvajanju programa. Te strukture v moduli-2 v bistvu sploh ne potrebujemo; 6e smo pravilno razbili naS program, moduli sami prestreiejo vse napake in jih ostali program sploh rve zazna.
Druga struktura, ki Jo najdemo samo v adi, izhaja iz dejstva, da je nek algoritem (recimo za vodenje sklada) povsem neodvisen od tega, kakSni elementi ao v skladu. V ta namen pozna ada takoimenovane osnovne podprograme (generic), kar Je v moduli-2 fie najtefje nadomestiti. Pri tem si pomagamo z sploSnimi tipi (ADDRESS, WORD, BITSET, ARRAY) in predvsem z molnoatjo loCenega prevajanja, kjer uporabimo za dano razlifiico dani modul.
3.3 tlodulsrnost In loOsno prevajanj*
NaCela modularnosti in mofnosti loCenega prevajanja amo si le ogledali v četrtem poglavju prvega dela nafiega prispevka. Na tem mestu pa ai oglejmo le podobne strukture ade.
Modularno programiranje in lo6eno prevajanje omogo&a ada enako dobro kot inodula-2. Pojem modula nadomeS&s v adi pojem paketa (psokags). Kot modula-2 tudi ada loCi med definicijskim in delovnim modulom. Detinioijski modul Je zaznamovan z besedo paokSQ«, delovni pa z paoksgs body. Kot v noduli-2 so tudi v adi definirane strukture za oddajo in prevzem tipov, spremenljivk In podprogramov Cu>«, privat»). Bistvena razlika med obema jezikom pa Je v programskcn modulu, ki Je pri adi podprogram v moduli-2 pa zopet modu).
V sistemu jezika inodulA-2 po definiDiji obstaja ena bistvena prednost pred ado. Sistem jezika modula-2 namreC v postopku prevajanja in po'i'ezovanja sam preverja, da je modul, ki prsvzena neko strukturo iz drugega . modula, preveden kasneje. Tega ada v osnovi ne pozna.
3.4 Nlikonlvojska programiranJ« In vzporednost
Na tem podroCju je ada v definicijah precej fiibkejfia od module-2. Adino spremeni jivko, .na priner, ni maino namestiti na željeno mesto v pomnilniku in programer ima na voljo le Se dinamiSen nafiin dostopa.
Za programiranje vzporednih postopkov ima ada enoto ta«k, ki je podobna procesu v moduli-2. Prednost pred madulo-2 pri viporednem programiranju pa predstavljajo adine ukazne strukture, ki omogočajo vsklajevanje vzporednih procesov; v moduli-2 mora programer te podprogram napisati sam. Kot modula-2 tudi ada oracgoCa poslutevanje prekinitev.
Collins Cl^] pravi, da bi bila primerjava inodule-2 in ade nekaj podobnega kot primerjava med pascalom In P4-/li kar je prepui&eno v inDduli-2 programerju, je v adi Se vgrajeno v jezik. Tu lahko tudi iSčemo vzrok obsežnosti ad«.
C IH (lpPMl,A-2
noduIa-2 in jezik c sta zelo razliCna jezika. Zato se bomo omejili le na primerjanje moZnosti za programiranja na spodnjem nivoju, torej na nivoju sistemskega programiranja. Razen tega je «miselna primerjava med obema jezikoma v pogledu prenosljivosti programov, medtem ko bi bila primerjava viSjih programskih struktur brezpredmetna, ker jih v o-ju ni. Zato v o-ju ne poznamo principov modularnega programiranja in loCenega prevajanja. Tudi v podatkovnih strukturah Je c dokaj reven.
Jezik 0 je nastal hkrati z operacijskim sistemom Unix C7, 233 ter postal osnovni jeziV. tega moSno razširjenega operacijskega sistema. Zato pcogratn napisan v c-Ju lahko dandanes brez. veCjih sprememb poienemo na poljubnem drugem ra&unalniku z operacijskiin sistemom Unix. Prav prenosljivosti programov je tisti pojem, pa katerem je smiselno primerjati modulo-2 in c. Kot pri jeziku c, je torej tudi pri moduli-2 zagotovljena prenosij'ivost programov, vendar v primeru programov v moduli-2 ne samo med različnimi ra&unalniki z istim operacijskim sistemom, ampak celo med raöL'rtal nik i z razlifinimi operacijskimi sistemi. Dosefii ta cilj v kar se da veliki meri je naloga delovne skupine projekta OSSI C23. Točneje, iidelati ielijo nekakšen vmesnik tned modulo-2 in poljubnim operacijskim sistemom.
Jeziku c obiCajno pripisujejo moč in prednost predvsem pri nlzkonivojskem programiranju, vendar zlahka uvidimo, da ima vse te zmotnosti tudi modula-2, ki omogoča celo večji nabor sistemskih operacij.
Največja moč jezika c je v naslavljanju pomnilnika, vendar prav c ne pozna statične definicije spremenljivke na določenem mestu v pomnilniku. Pač pa c lahko poljubno spremenljivko pretvori v kazalec, na naslov (to doselemo z uporabo posebnega znaka »), kar v moduli-2 doselemo s klicem funkcijo ADDRESS. V c-ju lahko vedno ugotovimo naslov v pomnilniku, kjer se nahaja določena spremenljivka, v modull-2 pa to opravimo s funkcijo ADR.
Ker C izhaja po svojih definicijah iz zbirnika,
ki nu poskuSa ostati čimbolj podoben, je tenu primerno bogat tudi po načinih naslavljanja. Avtomatično dodajanje (k++) ali odvzemanje Cx-=i) v c-ju, je v moduli izvedljivo z dodatnim klicem podprogram« INC.
V c-Ju obstaja poseben način definiranja posameznih eprentenl Jivk, ki prevajalniku pove, naj poskuSa shraniti spremenljivke v procesorske registre (ragi«t«r). V moduli-2 to ni potrebno, ker so registri neposredna dostopni, drugi cilj omenjene c-jeve strukture -optimizacija kode, pa je v bistvu problem prevajalnika in ne jezika samega.
Delo s posameznimi biti, ki je v c-ju zelo enostavno, je v moduli-2 izvedljivo s «premenijivKaaii tipa BXTSET in klicem podprogramov INCL ter EXCL.
Za razliko od module-2 pa v c-ju ne moremo uporabljati procesov in tudi posluisvanje prekinitev po definiciji ni del jezika C1, 7,
Nizkonlvojsko In sistemsko programiranje, za kar Je c dokaj primeren, je enako dobro podprto tudi v madull-2. Ob tem pa nam inodula-2 ponuja Se vse prednosti strukturiranega programiranja, ločenega prevajanja, preverjanja- sovpadanja tipov In Se cele vrste prednosti, ki jih zahteva sodoben, učinkovit programski jezik, s čimer se c ne more ravno pohvaliti.
L.—Nafcf^iiwri^ prfYftJ^tfìtKi^ it»
v tem poglavju si bomo ogledali tri prevajalnike in en interpreter ter z njimi bomo prehodili kratko zgodovino madule-2.
Prevajalnik H2M in MSRTll
Prevajalnik t12RTll C1Q, 11, 12, 153 iz ETH Zurich je bil prvi filroko uporabljani prevajalnik za mcdula-2; pisan je v noduli-2 za računalnik PDP-11 <LS1-11) pod operacijskim sistemom RT-11 . Avtor module-2 Wirth C2S:] je novi jezik dokončno oblikoval in realiziral skozi ta prevajalnik.
Dokumentacija obsega pet zvezkov, ki opisujejo samo sistem jezika brez primerjave s kakšnim drugim Jezikom. V njih tudi zasledimo, da obstaja različica tega prevajalnika za operacijski sistem Unix.
Osnovni prevajalnik CP12RT1'1) je petprehodovni , kakršni so praviloma tudi vsi drugi. Prav počasnost prevajanja, ki izvira i- Števila prehodov, je ena največjih slabosti module-2. Ker so vsi novejSi prevajalniki nastali prav na podlagi tega osnovnega prevajalnika, si ga oglejmo podrobneje.
Sistem jezika sloni na kosu programa, ki je napisan v zbirniku in se Imenuje flun Time System. V nJem so vsebovani vsi podprogrami, ki Jih oddaja nodul SYSTEM in Se nekaj sploSnih podprograitov. Kilo vseh teh podprogramov je izveden preko programskih prekinitev - pasti. Vedno obstaja možnost preprogram!ranja tega dela prograna tako, da lahko potem poganjamo program brez podlage operacijskega sistema (stand alane).
V sistem je vključen prevajalnik, ki obsega naslednjih pet prehodov:
1.1 preverjanje sintaktične pravilnost pre-vajane enote In branje simbolne datoteke vnetenih modulov 2.1 preverjanje pravilnosti deklaracij in izdelava refernčne datoteke
3.)	pr»varJ«njB sovpadanja tipov v telesih podprogramov
A.I generiranje kode za izraze 5.1 ganeriranje kode za posamezne stavke in izdelava izhodne datoteke
Kasneje so defininicali neko oblika vmesne kode, podobno kot p-kodo pri pasoalu, in jo imenovali n-koda. Prvifi je bila ta oblika kode uporabljena pri prevajalniku M2n, kjer sta A. in 5. prehod zdruiena v enega samega in sicer:
4.)	generiranje vmesne, M kode
V dokumentaciji je dodan tudi interpreter te kode in njen opis Č133 Sioer pa se vsi prehodi izvedejo le pri prevajanju delovnih in programskih modulov, nedten ko so za definicijske modula dovolj te prvi trije.
Poleg prevajalnika vsebuje sistem Jezika Se povezovalnik in iskalnik napak (debugger) ter vrsto uporabniških modulov z orodji, ki onogofiajo delo z datotekami in podobno.
Sistem razpolaga z veC vrstami datotek, ki jih loCi po podaljških!
" .MODI programski ali delovni modul
-	.OEF« definicijski modul
-	.LST: izpis prevedenega programa (potrebuje
ga iskalec napak)
-	.REF: refernčna tabela (potrebuje jo iskalec
napak)
-	.LNK: izhodna datoteka iz prevajalnika, ki je
primerna za povezovanje
-	.LODi izhodna datoteka iz povezovaInika, ki
je primerna za izvajanje v okviru siateiiia
-	M2ri in M2Si sistemske datoteke
Vsi naslednji primeri prevajalnikov so nahakäne izpeljanke iz osnovnega M2M,
5.2 HaduU-2 z« ZSO pod CP/H
Tudi ta prevajalnik prihaja iz Svice, je pa precej okrnjena razliCica, saj ne pozna pojma procesa in vzporednosti delovanja. Ražen tega tudi ne omogofia posluievanja prekinitev. Kot nadomestilo pa ohranja povezavo z zbirnim jezikom; uporabnik lahko po ielji daloSene delovne module napiSe v zbirniku.
Sistem sestoji iz podobnih delov kot n2M in zopet Je prilolena mnoiioa delavnih modulov, hi omogoäajo delo s samim operacijskim slstemoin. Dokumentacija te implementacije sistema modu-le-2 CIAD podrobno opisuje razlike med pascalam in fflodulo-2.
5.3 Hodula-2 razvojni slates
Je interpreter za modulD-2 za računalnik v tem priseru IBM-PC. Med predstavljenimi primeri je to tudi edini Interpreter. Nastal je leta 1?34 pri tvrdki Modula Corporation. Ta sistem je pravzaprav celoten prevajalnik, ki pa mu manjka peti prehod, torej generacija kode, ki bi bila izvedljiva na ciljnem raöunalniku. Implementacija module-2 je dokaj celovita, saj iz definicije Jezika manjka samo mofnost poeluZe-vanj« prekinitev (podprogram SYSTEM.lOTRftNS-FER). Ta sistem je Se najbolj primeren za nadaljna razvojno delo na moduli-2, predvsem v smeri razvoja in izdelave petega prehoda, ki bi izdeloval resniftno izvedljivo kodo. Motno pa bi bilo tudi dodati 6e en prehod za optimizacijo H-kode, ki bi ga vstavili med Četrti in peti prehod.
S.4 8D8-XP
Sistem SDS-XP sodi trenutno med najbolJSe prevajalnike ta no(lulo-2 za raùunalnik IBM-PC in operacijski sistem PC^DOS.
Implementacija madule-2 je povsem popolna in vkljuCuJe molnost nadzora nad prekinitvami ter vzporedno poganjanje procesov. Pa tudi sicer lahko o sistemu zaplSemo samo najboJjSe, saj poleg obi&ajnih delov vsebuje Se poseben urejevalnik, v katerem so definirane posebne tipke, ki omogo6ajo enostavno pisanje ukaznih struktur (zato tudi odpade prvi prehod) ter možnost uporabe knjiinice modulov, iesar ne omogoča nobena druga od opisanih implementacij inodule-2. Razen tega 5DS-XP odlikuje z izredno bogatimi orodji, ki so vkljufiena v sistemski knjižnici ITCLlB.
S tem smo zaključili pregled nekaterih vidnih primerov prevajalnikov za modulo-2. Ta pregled pa ni popoln, saj nismo omenili sistemov za veöje rafiunalnike (VAX pod VMS) in podobno.
&a_Hodula-2 In praloa
Zadnjih nekaj vrstic tega pregleda namenimo zelo zanimivemu paketu za povezavo )nodule-2 in prologa. Razvit je bil v Zurichu pod vodstvom Carla Mul terja tl7, IS]. Povezava med obema programskima jezikoma je dvosmerna, saj lahko uporabljamo v prologu podprograme napisane v moduli-2, ali pa prclagova nafiela logifinega Iskanja po podatkovni bazi v moduli-2. Prva molnost je zelo enostavna. DeJinirati moramo le obliko podprograma vsebovanega v nekem modulu. Ta podprogram natt v postopku interpretacije prologovega programa predstavlja nek prologov stavek (klavzulo). Tako lahko reenićno prlSnemo razvijati nek program v prologu in postopoma stavek za stavkom prevajamo v fflodula-2. Na koncu imamo celoten program v moduli-2, ga prevedemo in s tem pridobimo na hitrosti, hi je najbolj Šibka to&ka prologovih interpreterjev.
Obraten, manj pogost primer uporabe, sloni na klicu prologovega Interpreter ja, ki je samo nek podprogram. Temu podprogramu predamo kot parameter nek rtiz, hi je v bistvu prologov stavek, katerega interpreter obdela na enak naSin, kot vsak drug prologov stavek. Opisani sistem je bil razvit na raöunalniku Lilìth, sedaj pa je na volju tudi za računalnike IBH-PC pod MS-DOS in VAX pod VMS.
Enostavna prenosljivost paketa je prav plod uporabe OSSI standarda, ki ga paket uporablja za vse posege proti operacijskemu sistemu.
U_Lastne iikuln.l»
V Odseku za računalništvo in informatiko Instituta Jotef Stefan le nekaj Casa slstema-tifino zbiramo dokumentacijo In orodja za programiranje v modull-2. Do sedaj smo v mcduli-2 realizirali nekaj manjSlh poskusnih projektov, od katerih jedra dveh navajamo v dadatkw. Med drugim smo v moduli-2 razvili razporejevalnik sistemskega Casa in s tem omogočili pisanje resnlCnih soCasno tekoCih procesov na enem procesorju v nasprotju z definicijo module-2, ki pozna samo scrutine.
Naslednji korak pa je implementacija veSih so&asno tekofiih prcoesov na veCih procesorjih. Projekt razvijamo v okviru razvoja poskusnega okolja P5-11 za programiranje vzporedno delujočih procesorjev na vodilu Q C3, Dosedanji rezultati so dokaj vzpodbudni! doslej
suo ufipBli pognati vsÄ vzporednih, vendar zaenkrat noaodelujoftih procesov, ki si delijo skupni pomnilnih..
PoleQ tega nastaja v naSih laboratorijah prva inplementacija OSSI standarda is. operacijski siste» RT-11 in MioroVMS.
I:_2aliHuS»k
Modula-2 S0 Je v preteklih nekaj latih uveljavila kot Jezik za «istenifiko programiranje in kot jezik za pisanje velikih programskih sietenov. Obstaja ocena, da Je v moduli-2 napisane fe okoli 15% noveJSe sistemske proqramske opreme. Razlag za hitro razširjenje jezika je brez dvoma v njegovi enostavnosti in učinkovitosti. Prav, gotova si bo iiiodula-2 s svojo preglednostjo in majhnostjo samega sistema, ter enostavnostjo rokovanja pridobila v prihodnje Se veliko većjo veljavo in že morda kar krtalu nadomestila "stari dobri, pascal".
la -vso pomaö pri iskanju novih znanj In dragocene nasvete pri oblikovanju besedila se najlepše zahvaljujem prof.dr,BoStjanu Vilfanu.
t 13 V.Batagelj, A.Brodnik in ostali: Procge-
tinas—Si__gMwwqr_gchpoi of pompu^pr
Sciences. IJS Report A352, Ljubljana, 19B6.
C E.Biagioni, K.Hinrich, 6.Heiser, C.Mullen
A Portable Ooeratino System Interface_ariä.
Utility Library. IEEE Software, November
1986.
C 3D A,Brodnik, S.Mavriö, R.Trobeči Q-bus Base HqHi,proogssqr_System, CompEuro, Hamburg,
1987.
C 4D
C 53 C 6] C 73 C S3
C
C1D3 C113
[123 C133 C143 ri53 C163 C173
S.Collins, B.Marshall, N.King: Hodula-2 . -A Babel Fish for Chios?. System International, Decembar 19B6.
R.GIeaves; l1odula-2 for Pascal Programmers, Springer-Verlag, IPS'». K.Jansen, N.Uirth: PASCAL User Manual and Report, Springer-Verlag, 1975. B.M.Kernighan, D.M.Ritchie» The C Prograra-ming Language, Prentice-Hall, 1976. S.MavriO, A.Brodnik, R.Trobec, N.Spegel,
P.Kolbezen« ES-1.ÌJ_VeSorocesorski ■ sistem
YQdtlH Q, «ipco 19S7, Opatija, 1987. Wodula-2 PevBlooment System. Modula Corporation Provo, Utah, 198^.
Modula-2	_lElž.Sifui;tij;ps. ETH
Institut fur Informatik, Zurich, 1981. Modula-2 Overview of the Hodula-S Compilar.. ETH Institut fur Informatik, Zurich, 1981 .
Modula-2 Overview_of MQdula-2 Debugger,
ETH Institut fur Informatik, Zurich, 1981. «odula-2 Overview of Modula-Ž Svftem H^tl. ETH Institut fur Informatik, Zurich, 1982. MDdula-2 System for Z8Q CP/M. Hoohstrasser Computing AG, Zurich, Switzerland, 1984. IlDdula-2 User Guide. ETH Institut fur
Informatik, Zurich, 1981.
B.Mohar,	E.ZakrajSek; Uvocf y programira-n.ìe, DMFA, Ljubljana, 1982. .
C.Mullen	Modu^-Prplgg tüsgr	ETH Institut fur Informatik, Zurich, Juli
1985.
C183 C.Huller; Modula—Prolog; A Software Pevelogmant_laaL. I-E^E Software, November
1986.
Qs.nq^- Prfl.q.Eijl.Ht,fikefl.g._...j^_Oelovni
i:t93 C203
C213 I.C.Pylet Iba__Prcarammino_Language.
Prentice-Hall, 1981.
C223 SDS-XP, Extended_Performance t1odula-2
Software_PBY^tBpmeTit .Systqi^., Interface
Technologies, Houston, C233 R.Thomas, L.R.Rogers,
Texas, 1985. J.L.Yatest CHjJygJpgd
gysten_Vi
HcSraw-Hill, 198t. C2A3 P.Wegnert Proorammino with Ada. Prentice-Hall, 198Q.
C253 N.Wirthi	Third,
Corrected Edition,'Springer-Verlag , 198S .
Podatak A - Phone
Program Phone omogoCa pogovor ned dvema terminaloma prikljufienima na rafiunalnik LSI-11 pod operaoijskim «istemom RT-11. Ker je RT-11 enouporabniSki sistem, je lahko hkrati aktiven le en terminal. Tu program poženemo in izberemo drugega. Pri obeh se na zaslonu pojavi obrazec. Na zgornjo polovico zaslona se izpisujejo znaki, sprejeti od drugega terminala, na spodnjo pa zanki, ki jih terminal poSilja drugemu.
Program sestoji iz Šestih modulovi
-	Phonas osnovni program
-	rSor: modul xa d«lo z enim zaslonom
-	xSor: skrbi za drugi zaslon
-	VTLib»: posebni ukazi za delo s terminali tipa VTtOO
-	rPortt poslulevanje prekinitev na enem vhodu
-	xPovt: ppsluSavanja prekinitev na drugem vhodu
Na sliki 1 je Se prikazana podrobnejša shema medsebojne odvisnosti posameznih nodulov.
		Phone		
				
rScr		VTLibx		xSc r
xPort
zvezek Izobraisvalni center Delta, 1984. 5,Pomborgen Software Engineering and Modula-2. Prentice-Hall •, 1984.
stika 1
Definicijski moduli!
DEFINITION MODULE rSorj
(• modul za delo s prvim terminalom •) EXPOftT QUALIFIED
rPutCh, rPripraviEkran, rBrisi, rGetCh|
PROCEDURE rPutCh Ccht CHARj flag; BOOLEAN)t
<• izpiSe znak na prvi zaslon •) PROCEDURE rPriprsviEkran CVAR rVeot,
rCSRs CARDINAL)! (• pripravi zaslon terminala •) PROCEDURE rBrisiJ
('« zbrite zaslon terminala *> PROCEDURE rSetCh (VAR oh: CHARj
VAR OKt BOOLEAN); (< pobere zlog s terminala •)
ENO rSor.
DEFINITION MODULE VTLibHi
(• Ukazi za delo s terminali tipa VT 1Q0 >) EXPORT QUALIFIED sCursor, sErase, sWriteLine, sScroll, sNewLine, entire, PutString, Obrazec, DelCh.ì
CONST
entire = 2; "t* zbriäe ves zaslon »>
TYPE HriteProc = PROCEDURE fCHAR)5 (* podpragc-am 2a izpis zloga
PROCEDURE BScroll (from, to: CARDINAL;
write: UriteProc); (* doloSi obrnoCje premi ksn ja zaslona *> PROCEDURE «Ecase (area: CARDINAL;
writei WriteProc); (* zbriäe doloCen del zaslona ») PROCEDURE sCursor (k, y: CARDINAL;
write» MritePcoc); (• postavi slednik na doloCena mesta PROCEDURE sNewLine (write: WriteProo)5
t* premakne slednik v novo vrsto •) PROCEDURE sUriteLtne (long« CARDINAL i
write» MriteProcli <* izpila firta zahtevane dolilne ») PROCEDURE PutString Cwhsti ARRAY OF CHARj
write; HriteProc);
<« izpiSs nlz •} PROCEDURE Obrazec fwrite: HriteProc);
(t lepite osnovni zaslonski obrazec •> PROCEDURE DelCh (write: UrlteProo);
(• zbriSe zadnji znak v tekoCi vrstici ») END VTLibK.
DEFINITION MODULE KPortt
(* pDSluievanie prekinitev na kanalu *> EXPORT QUALIFIED
xGetByt, xinit, xClose, KputSyt)
PROCEDURE xGetByt (VAR cht CHAR)
VAR OKI BOOLEAN)I (• pobere zlog Iz vmesnika, □B ie na voljo •) PROCEDURE xInit (VAR xVect, xCSR: CARDINAL)i (• nastavi vse preklnitvene podprograme •) PROCEDURE xClase;
(• zapre In vse pospravi okoli kanala PROCEDURE xPutByt (VAR eh: CHAR;
VAR OKI BOOLEAN)j (* vstavi znak v vmesnik, Ce je prostor *) END xPort.
V primeru modulov rSor in xPort srao navedli sano po en modul za vsaka stran, Oruga sta namreC skoraj povsem enaka. Sledi Se podoben pregled programskega in delovnih modulov. Navajano samo delavna modula, ki sta zanimivej-fia (izpueeamo VTLibx).
MODULE Phon«I
(« glavni programski modul <) FROM xScr IMPORT
xGetCh, xPutCh, xPripraviEkran, xBrisii FROM rScr IMPORT
rSetCh, rPutCh, rPripraviEkran, rBrisl; FROM TTIO IMPORT
HrlteString, Read, Hrite, SetMode;
CONST ETX = 003C; (» <CTRL><C> *) VAR xch, rch! CHARj OK, OKt BOOLEAN) CSR, Veot: CARDINAL;
PROCEDURE BsrlCSR (niz: ARRAY OF CHAR;
VAR Vect, CSR: CARDINAL); C» dolafii vektorja in naslove statusnih
registrov obeh terminalov *) VAR Enota« CHAR; BEGIN
WriteStrlng tnlz);
Read (Enota); Hrite (Enota);
CASE Enota OF
"!': Vect i= 310BI CSR != 1765108 1 '2't Vect s= 3508; CSR := 1765SDB END
END BeriCSR; BESIN
SstMode CO, FALSE); C* prepove <CTRL><C> *) BerlCSR (' Od TT (st.): Vect, CSR); XPripraviEkran (Vect, CSR);
BerlCSR (' Proti TT (st.): ', Veot, CSR)j
rPripraviEkran (Vsot, C5R);
REPEAT (• glavna zanka za izmenjavo znakov *) xSetCh (*ch, OK)i IF OK THEN
rPutCh (xoh, TRUE); xPutCh (xoh, FALSE) END)
rGetCh (rch, OK); IF OK THEN
xPutCh <roh, TRUE); rPutCh (rch, FALSE) END
UNTIL (Kch » ETX) OR (rch = ETX); xBrlsi; rBrlsi END Phone.
IMPLEMENTATION MODULE rSor;
(» delo s sprejemnim zaslonom *) FROH rPort IMPORT
rinit, rGetByt, rClose, rPutBytj FRO« VTLibx IMPORT
sCursor, sErase, sUrlteLine, sScroll, sNewLlns, entire, PutStrlng, DelCh, Obrazec ;
CONST CR = OlBCt LF = 012C; DEL = 177Ci
BS = OlOC; REFR =■ 027C; ESC = Q33Ci
VAR KB, Kt, ym, yt, x, y» CARDINAL;
PROCEDURE Send (chi CHAR))
(• pofilje znak na zaslon •) VAR'akt BOOLEAN; BEGIN
REPEAT rPutByt (eh, ok) UNTIL ok END Send;
PROCEDURE rPripraviEkran (VAR rVect,
rCSR: CARDINAL); (•» nastavi terminal in IzpiSe obrazec ») BEGIN
rinit (rVect, rCSH);
XD 1= 1; xt I5 yn 1= 11; yt 1» 22; X I= xm; y :n ym;
Otirazea (Send) END rPripraviEkran;
PROCEDURE rPutCh Cchs CHARj flag: BOOLEAN); (• interpretira sprejeti znak •) PROCEDURE Uredi (VAR a, b« CARDINAL); VAR i: CARDINAL; BEGIN
IF oh = REFR THEN
FOR ii>»l TO ID DO sNewLine (Sfend) EMD ; a t END;
IF oh = DEL THEN DelCh (Send); ai=a-l ELSIF oh = CR THEN sNewLine (Send); a;=l ELSE Send (ch); a 1= a 1 END; END Uredl; BEGIN
IF (flag) THEN IF K » xn THEN
sScroll <3, 11, Send); sCursor (xm, yn. Send); X i" *m END;
Uredi (xm, ym) ELSE
IF X « xt THEN
sScroll (14, 22, Send); sCursor txt, yt, Send); K 1° xt END;
Uredl (xt, yt) END; END rPutCh;
PROCEDURE rGetCh (VAR oh: CHAR;
VAR OK s BOOLEAN); (• dobi znak s terminala «) BEGIN
rGetByt (oh, OK); END rGetCh;
PROCEDURE rBrlsi;
(• zbrise ekran in zaključi komunikacijo »)
BEGIN
sEraea Centire, Send); rClose END rBrisl; EMO rScr,
IMPLEMENTATION MODULE xPort C73f
C» podprogrami za dela i enim ad kanalov •>
FROM SYSTEM IMPORT
ADR, ADDRESS, SIZE, WORD, NEWPROCESS, TRANSFER, lOTRANSFERj
CONST N = 32i
TYPE DLVllJ » RECORD
rCSRi	BlTSETi
reUFi	CARDINALI
xCSR:	BlTSETj
*8UF:	CARDINAL END t
TypeCSR = POINTER	TO DLVllJi
VAR rn, rin, rout, xn, xln, xoutt CARDINAL) rbuf, Kbul; ARRAY CD..N-ID OF CHARt rPRO, xPRO, CON» ADDRESS) rUSp, KUSp! ARRAY CO..177B3 OF UORD;
pxl, px2, px3, px4i UORD)
activef BOOLEAN; Vector: ADDRESS t CSR) TypeCSRf
PROCEDURE xQetByt (VAR cht CHAR;
VAR OK t BOOLEAN); C* pobere zlog i.z «prejennegA vmesnika, če lahka «)
BEGIN
IF rn > □ THEN OK TRUE;
eh != rbuf CroutJ;. rn := rn - ■!{ rout != Crout+D MOO N; ELSE OK 1= FALSE END
END KSetByti
PROCEDURE producer;
(* sprejemni prekinitveni podprogram BEGIN LOOP
lOTRANSFER (rPRQ, CON, Vector)? IF rn < N THEN
rbuf CrinO != CHR <ĆSR-.rBUF)i rin t= (rin+1) MOD Nj rn T= rn + 1 END END
END producer;
PROCEDURE xPutByt CVAR oh; CHAR?
VAR OKs BOOLEAN); (» vstavi zlog v oddajni vmesnik, ee lahko «)
BEGIN
IF *n < N THEN OK ,1« TRUEl
xbuf CxinS t' eh I xn t^ xn -f It Kin 1= Cxin+l) MOO N; IF NOT active THEN
INCL tCSR".xCSR, 6>; active ;= TRUE END
ELSE OK 1= FALSE; END END xPutByt;
PROCEDURE consumer;
(* oddajni prekinitveni podprogram •) BEGIN LOOP
lOTRANSFER CxPRO, CON,
ADDRESS (Vector+4}); IF xn > 0 THEN
CSR'.xBUF !■= ORD <xbuf CxoutD) ;
xn 1= xn - 1; xout m (xout+1) MOD N ELSE
EXCL CCSR'.xCSR, 6); active FALSE END END
END consumer;
PROCEDURE xlnit (VAR xVect, xCSRs CARDINAL); (« inicializlrama registre, ki Jih bomo potrebovali in poienemo nov proces •) VAR at ADDRESS} BEGIN
Vector t= xVecti CSR i= TypeCSR CxCSR);
(* shranimo stare vrednosti *) a f ADDRESS (Vector)( pxl := a~; INC (a, 2); px2 t" INC (a, 2)t px3 t= a"; INC (a, 2); px4 I" a'i
rn 1= 0; rin i= 0; rout i« 0; NEUPROCESS (producer, ADR(rUSp) , SlZECrWSp), rPRO);
xnt^O; xini>Ot xautt^Q; active:°FALSE; NEUPROCESS (consumer, ADR(xH5p), SIZE(xV)Sp>, xPRO);
WITH CSR- DO
INCL CrCSR, &>; TRANSFER (CON, rPRO)s TRANSFER (CON, xPRO) END END xlnit;
PROCEDURE xClose;
(• pofiistlno vse za seboj «) VAR ai ADDRESS; BEGIN
C« obnovimo stare vrednosti •) a !» ADDRESS (Vector)t «- s= pxl; INC C«, 2>i a- i' px2; INC (a, 2) ; a- t= p*3i INC Ca, 2); a" t= px^ END xClose; END xPort.
Na sliki 2 pa podajano Se prikai medsebojnih klicev modulov. Primer je tudi lap način razbitja na «odule in definicije njihovih stifinih toAk - podprogramov.
Podatak B - RazporaJevlnlk
Tukaj navajamo samo primer definicijskega in delovnega modula razporejevlanlka procesorskega Casa in kratek primer z dvema procesoma.
DEFINITION MODULE Scheduler; EXPORT QUALIFIED Proce«sGo} PROCEDURE Procaseeo Cwhicht PR,OCs
space) CARDINAL); (• požene nov proces, ki uporablja space velik prostor v pomnilniku •) ENO Scheduler.
IMPLEMENTATION MODULE SchedulerC73}
FROM SYSTEM IMPORT ADDRESS, PROCESS, NEHPROCESS, TRANSFER, lOTRANSFER; FROM Storage IMPORT ALLOCATE;
CONST N = lOj
(» največje število procesov ClockVec ° 10GB;
C» prekinitveni vektor ure ») TinerSpace > 200B(
(• prostor za prekinitveni
*)
rPripraVtEkii " rQ«ICh rPulCli rBrl>l
S*nd
rScr
(Cur»«r,. Obtax«c
VTLlbx
rinit rOalBft rPwiByl FC lo»
rPorl	rPatSf)
	rttetByt
/
P'»"««:«'

• Ilk* 3
podprogram za uro TioksPerPeriad = SO
(* stttvilo urinih enot, ki so dodeljeni enemu procesu O
VAR StartMewProcessi BOOLEAN; wspt ADDRESSl
FirstProo,	<* starting process »J
TimerAddi PROCESS) tìere Is timer «) Processesi ARRAY C1..N3 OF PROCESS] n, ourrenti CARDINAL;
PROCEDURE Timeri
(* dusa razporejevlnika *) VAR NunbsrOfTlcks« CARDIÜIAL; BESIN
NuBberOfTicks i= TioksPerPeriodi LOOP
IF TI > 0 THEN
IF StartNewProcess THES StartNewProcess t <= FALSE; NuniharOfTioks i= TicksPer Period i currant■ EL31F NumberOfTicks <=0 THEN
NumberOfTicks t= TicksPerPeriodi currentt^Ceurrent MOD n>+l ELSE
NumberOfTioks.1= NumberOfTicks ' 1 END)
lOTRANSFER CTlmerAdd,
ProcessesCcurrent3, ClookVec)
ELSE
NunberOfTloks 1= TicksPerPeriOdj lOTRANSFER CTittifirAdd, FirstPrca, CloofcVec)
END| END ENO Timer)
je prvi «>
OldCuri CARDINAL; BEGIN
IF n < N THEN
ALLOCATE (w«p, fipace>!
<• poiščemo prostor za novi proces •) nt"n+t I
NEHPROCESS (which, wsp, epace,
Processesene) ; StartNewProoess t« TRUE| IF n = t THEN	ali j€
TRANSFER (FirstProo, TimerAdd) ELSE	<• sicer ga požene Timer «>
TRANSFER (ProcessesCourrentS, llmerAdd)
END; END
END ProcessGo) BEGIN
n;>:Q| currenti'l; StartNewProcess ; FALSE; ALLOCATE (w«p, TtmerSpace)i NEMPROCESS (Timer, wsp, TimerSpaoe,
TlmerAdd)) TRANSFER (FirstPrco, TimerAdd) in poxttneo Timer END Scheduler.
NODULE TestScheduleri
FROM Scheduler INPORT ProcessGo) FftOn VTLibx IMPORT sCursor; FROM TTIO IMPORT Write, Readi
CONST SpaceNeedcd " A00&t VAR cht CHAR{
PROCEDURE One;
(• na zaslon Izpisuje "A" »] VAR il CAR01NAL( BEGIN
ProcessGo (Two, SpaceNeeded)t it°lt LOOP
sCucsor (i, Mrite); li^Ci HOD B0)'«-1; Urit» CA') END £ND One I
PROCEDURE Two)
(» ta pa naj izpisuje "8" ») VAR il CARDINAL; BEGIN il-ll LOOP
sCursor (i, S, Write); l:=(i MOD 80)+l( Write CB') END END Two)
BEGIN
ProcessGo (One, SpaceNeeded) END T«8tScneduVsr.
PROCEDURE ProcessGo (whichi PROC;
spaces CARDINAL); (« požene nov proces «) VAR wspt ADDRESS;
IMPLIKACIJA ORGANIZACIJE NERVNIH SISTEMA NA RAČUNARSKE SISTEME
UDK 681.327.2
Marković Pavle Visoke vojne tehničke škole, Zagreb
strukturna I funkc(on»Ua oretnUtclJ« f n»eln intcrprette I Je doltzecih Informacija nervnih slst«na su mcdju najsloženiji" strukturtna koje J« priroda stvorlN/ I Jos uvek su daleko cd moguće t«hnlcke emulaclje« Osnovni načela te organizacij« su od skoro poznata I Iskoriscena su u ovoai radu u InpllkacIJI na računarske sistene. U redu su na osnovu Izvršene enaMze organizacije najviše« stadljuna nervnih sistema (mozga) postavljeni globalni zahtevi koje neki računarski ststem mora Ispuniti da ti mogao emulirati analiziranu nervnu organi zaciJu>
Ktjucna reci) blotoski slteml> upravljački sistemi« obrada Informacija
IHPLICATIQN OF NERVOüS SYSTEMS ON COMPUTER ORGANIZATION, funclonal organization and the manner of Interpretation Information of nervous system« are the aost complicated struc and tticy ere far atiey of possible techntcal esulatlon. Basle known and used In this piper In Implication on eoaputer analyzing the hlgest level of nervous system thumaan bra requlaents which sane computer have to have In order to ei»u)e nervous organization are giveni
Structural and of Incoming tures In nature principles are systems. After nl the global te the analyzed
KeyNordsi biological systems* control systems« information procesing
1. U V 0 0
Brzina racunara se od njihove pojave do danas yisastruko povećala. Ipakf današnji fiapor i I poHusaJI značajnijeg povećanja brzina računanja su teško ostvarivi I veoma skupi. Uglavnom oni su usmeréni u sledečim pravcima'
tehnološko povećanje brzine rada električnih . komponenti
promene u arhitekturi računarskih
s i stems.
Tehnološko povećanje brzine rada električnih komponenti nena vise onakav ostar
U central Izovanom pristupu sve Informacij« o stanju svakog objekta upravljanja I vanjshln dalovanjima dolaze u centralno mesto upravljanja. Tu se na osnovu Informacija o stanju sistema I budućim zadacima odredjuju upravljačke akcije za svaki element sistema. Ovakva struktura se može uclnltl Idealnim za pristup upravljanju« ali ona u sebi nosi i dosta skrivenih poteškoća. Jedna od njih Je I prevelika količina Informacija koje dolaze u centralno sesto upravljanja. To automatski znaci tesko odredjivanje najpovoljnijeg režima rada u realnom vremenu* upravo zbog preobilja Inf oriaac I Ja .
gradijent porasta kao pre> tako da se glavna povećanja brzine rtounanja očekuju u premenam» arhitektura računarskih sistena. Iako« koncepti nultiprocesorskih«	multlracunasklh«
distribuiranih« hijerarhijskih I drugih sistema su odavno poznati« njihova primana ne daje ona povećanja brzine računanja koja su se očekivala. U pomenutlai organizacijama Jos uvek kijune teškoće njihove iskoristivosti predstav 1JaJu)
-	problemi dekompozicije posla ne grupe zavisnih I nezavisnih zadataka
-	problemi komunikacije u sistemi
-	pristup upravljanju (koordinaciji) u s Istemu
Od gornja tri problena centralno mesto zauzima problem pristupa upravljanju u sistesu. Svi dosadasnJi pristupi mogu se svrstati u Jednu od sledeče tri kategorljei
-	centralizovanl pristup
-	samoupravni pristup
-	decentralIzovanl pristup
Za razliku od centra IIzovtnog pristupe« saiouprivnl podrizumevt skoro potpunu autonomiju odlučivanja svakeg eieoienta sistema. Pri toae svaki element sistema samostalno resava vlastite lokalne zadatke« dok u saradnJI sa drugim elementima sistema resava zadatke od zajedničkog Interesa. Ostvarenje sistema u kome bi svaki nt Imao totalnu autonomiju Je praktično neizvodljivo Jer bi tada bila nemoguća uskladjenost rada cttavog sistema u eeiini. Zato ovaj pristup upravljanju nosi u sebi obeležja i prvog [ trećeg pr istupa.
OecentaralIzovanl pristut podrazumeva raščlanjivanje sistema na delove (hijerarhijski sistemi) Izmedju kojth se uspostavlja koordinacija. Tada upravljački element vtseg nivoa upravlja eieaentima nižeg I sve tako do najnižeg. Hijerarhijska struktura« sileno centra 11zovanoJ« podrazumeva izdvajanje upravljačkih elemenata sistema vtseg nivoa od upravljačkih elemenata nizeg nivoa« « funkcionalna podela rada smislena je samo na Istom nivou« kao razdvajanje različitih podzadataka Istog nivoa.
Realizacija bilo kog naYedenog pristupa upravljanju u mu 111procesorskI m sistemima apste namene pokazalt se nedovoljno eflkesnoin Jer Je unosila dodatna ograničenja u prlmeni I upravljanju resursima. Zbog toga s« krenulo u analizu onih bioloških sistema za koje se sigurno moze tvrditi da spadaju u klasu iiuttlprocesorsklh sistema Inervnl sisteni)» Autor smatra dt su ti sistemi najsloženije strukture, koje Je priroda stvorila (aozda cak \ optlualne) I da Je današnja nauka 1 tehnologija Jos uveh Jako daleko od njihove moguće Implementacije. Na osnovu takve analize trebali su proisteći osnovni zahtevi koja neki računarski slsten nora Ispuniti kako bi Iskoristio pristup upravljanju koJI se nalazi uflradjen u nervnin sistemima. U analizu nervnih slsteaia pošlo se od njihovih elenentarnih oblika da b) se završila sa analizom najsloženijih nervnih formacija tj. mozga coveka.
Analiza nervnih sistema Izvršena Je sa stanovista njihove funkcionalne organizacije I načina Interpretacije ulaznih Informacija. Ovako traženom pristupu najbolje Je odgovarao pristup ruskog akademika I osnivača NeuropsIhologiJe A.R.LurlJe« a koJI Je on dao u knjizi *Gsnovi neuropsi ho IogI Je • « (prevod na s rpskohrvatskoai Izdala Je Izdavačka kuca Holit u biblioteci Sazvezdjal. Autor nije Imao naiaeru da prepričava pomenutu knjigu vec su sve diskusije vezene za organizacije nervnih sistenaf a koje se nalaze u ovca radu> III prepisane III saao prliagodjene radu (zbog duzine tekstualnih objašnjenja).
Oa se priliko* čitanja teksta ne bt došlo u koMzlJu važno Je objasniti tuaacenje pojna "nivo hijerarhije"« u organizaciji covekovog BOzga> Zbog tumačenja nivoa hijerarhije u poaenutoj knjizi uzeto Je sledeče znacenjei stari delovt mozga I moždano stablo se nazivaju nizim tako d« retikularna formacija predstavlja naJnlzl nivo hijerarhije ljudskog mozga a aparati nove kore najviši. U istom smislu tumači se značenje nivoa u uporednoj Implikaciji na organizaciju računarskog sIstema.
2. RAZVOJ NERVNIH SISTEHA
U evoluciji J najel ementar n I JI m nI sveta prijem signala ostvaruje dtfuznin S Istenom. Na toj eta Jedinstveni centar Informacije* I protok pr I vremenlfi dotninan stvaraju u ovom Ilio Uporedjujucl takav ne
e poznato da se na voma razvoja životinjskog I organizacija kretanja iti mrežastim nervnifi pi evolucije ne postojf koji prima I obradjuje drazi se odredjuje onim tnlm centrima koji se nom dalu nervnog aparata, rvnl aparat s računarskim
sistemima dolazi se do zaključka da bi se taj sistem sastojao od vise 'izolovanlh* racunara koji svaki radi svoj posao/ a u pojedinim monentlma oni nedjusobno razmenjuju poruke koje bi govorite o ukupnom stanju sistema ili podsistema. Poruke su kratke 1 vezane saao za ukupno stanje sistema. Na današnjem nivou računarske tehnike ovon modelu najbliže su distribuirane računarske mreže (slika 1). U toku datje evolucije dolazi do razvitka prednje ganglije koja polako počinje da preuzim« složenije funkcije. Ona prima i preradjuje Informacije sa receptora I prebacuje nadrazaje na silazne puteve koji se kreću ka motornim aparaUna životinja.
prednja ganglija Je funkcionalnu strukturu linsekto doblji sve si organizaciju (Izdvajaju se Ju samo na pojedi lvidne>mirtsne*tahtiInel). računarske tehnike ovakvi odgovaraju eenlirel Izovane ■ centar podnje da preuzi značajne za slste«.
imala Jednostavnu dok u kasnljln ozenlju funkcionalnu neuroni koji reaguju ne vrste drazI Gledano sa strane M nervnin ststeaiaa reze Is I Ika 2 gde ma pojedine funkcije
suha f. C*Afnili zorano rgfunortlio Atra id
Visim stadijufllaa ovih nervnih aparata odgovaraju takve strukture gđa se centralno« racunaru predodeljuje sve vise funkcija tJ. on počinje da biva sve značajniji 1 uticajniji na ukupni sistem.
U toku dalje evolucije nervni aparati prednje ganglije ne mogu da obezbede pri 1agodjavanje naglo promenjenlm ustovlaa sredine. U takvim slučajevima očuvanje vrste moguće Je III zahvaljujući preobilnoj proizvodnji Individua III zahvaljujući razradi Ind1 vi duaIno-pr omen ijlvog ponašanja.
Po ovoj drugoj ilnljl se odvija razvitak klcmenjaka. Ako kod nizIh klcaenjaka vazi start princip onda se kod ktcaenjaka koji prelaze na egzistenciju na kopnu pojavljuje potreba za takvim nervnla aparatom koji ce obezbedJIvatI maksimalnu Individualnu promeni JIvost ponašanja koja pak odgovara velikoj promeniJIvostI uslova Života na zemlji.
Tak Na ovom s lobodno centar autonomi postoj 1 s Istea nikakvih informac sve Info sesto* aktuator
vim biološkim zadacima odgovara mozak, nivou razvoja nervnih aparata maze se UJavltl da postoji Jedan Jedinstveni koj) ne pruža nikakvu lit aaiu Ju rada. To vodi da zaključka đa (računarski gledano) Jedan centralni koji sve oko sebe kontroMše. Ne*a distribuiranih mreža u kojima se Ije prenose a Jednog na drugi krajf vec ratclje sa senzora dolaze na centralno tu se obradjuju 1 reakcije prenose na e (si ika 31.
SEN ZO RI
eL
è
RAfUNAR
CT •
A K T u A T 0 R I SUko 3. RaCunonlii uprovljon ptxal
Slikal. Oisiribuirana rofononho ariio U ranln stadijima evolucije
(crvll
Dolazi se do zaključka da Je to konačno resenje i Jednim centralnim racurtarskla sisteaom koji obavlja sve funkcije. Nedjutia» postavlja se problem veličine tog računarskog sistema (propusna moc) koji bi kr^lz sve etape evolucije sisara zadovoljio sve zahteve koji se pred njim postavljaju. Očigledno Je daj gledajući evoluciju od najprostijih sisara do covekaf da bi taj računarski sistem verovatno rastao eksponencijalno (propusne moc) llf po nekoj sličnoj funkciji* obzirom na broj I složenost funkcija koje bi on kroz uporedni evolucioni razvoj mozga morao da zadovolji. (Jcigledntm kvantitativnim poredjenjem se vidi da coveK nema toliko proporcionalno veci mozak od nizih sisara* koliki bi po broju funkcija 1 siozenoscu operacija trebalo očekivati. Sama gornja cinjenica navodi na zaključak da resenje problema lezi u unutrašnjoj organizaciji tog nazovimo ga računarskog slsteaa.
2.1 apirats
Funkcionalna organizacija nozdanog
Kao sto je pomato aparati moioa su u potpunosti sposobni da primaju I analiziraju
Iz oKolne sredlnaj nove veze i čuvaju stanju da urodjene zamene složenim« obezbedjuJuci pri
infcrnaclje Ito Je do laze preradjuju ih^ forniraju njihove tragove. Oni su u programe ponašanja Ind.ividiaino-promenljivlm« tome ne samo razradu uslovnih refleksa vec I Tormlranje mnogo složenijih prograna ponašanja.
Velll<e hemisfere mozga I moždana kora postaju kod coveka najvažniji aparat za regullsanje ponašanja. S» prelaskom od visih si&ara ka ooveku evotuclja mozga povezan« je sa povecanjea površine najsloženijih (tercijarnih) rnih) zona kore» doH površina elementarnih delova kore (primarnih I sekundarnih) praktično se ne povećava
'CENZORI
rI I I
1 j

] nivo prijeiiiG
nivo obrode
I I nivo uprovljnnjo [ J
D/A

A K T UAT ORI
Sliko 5. Vlšenivovska organizacijo rofunorshcg sistemo Bilo bi nepravilno smatrati« da ako u coveka vodeću ulogu doblja moždana kora« da se onda sve nervne formacije koje su na n{zjm nivoima evolucij« predstavljate Jedine aparate ponašanja sada u potunosti odstranjeni od rada. TI prijašnji nervnt aparati ostaju u njenu« ustupajući pri tone vodeće mesto novln foraacI Jana I dobijaju novu ulogu. Oni sve vise postaju aparati koji obezbedjuju osnovu ponašanja sto aktivno učestvuje u regullsanju stanja u kome se nalazi organizamt predajući tako funkcije prijema« obrade i čuvanja informac)ja« a isto tako i funkciju stvaranja novih programa ponašanja« regullsanja i kontrole svesne delatnost) visim aparatima
nozdane kore. Sve navedeno« gledana s računarskog stanovista vodi do zaključka da bt računarski sistem koji bi koristio osnovne principe rada mozga morao biti reallzovan kao neka visenivoska arggnizacije (slika S). Sana veza izmedju nivoa nije definisana ali bi ona «orala zadovoljiti osnovne principe evolucije« a to je da svaki nlZl ili visi nivo nora predstavljati Integralni deo ukupnog sistema s odeljeni« funkcijama po nivoima« a opet u cilju sveukupnog delovanja celog sistema. To znaci da taj računarski sistem po.svojoj strukturi mora biti jedan dinamički sistem koji mora obezbediti neku promenjivost (prilagodljivost) funkcija po nivoima. NIzl nivoi bi preuzimali aktivno ucesce u organizaciji 1 pripremi aktivnosti za vise nivoe.
Iz fiziologije i neurologije se zna da najjednostavnije elemente ponašanja Kao sto su segmentami refleksi ostvaruju samo nehanlzml kieinene moždine. Gledano tehnički« refleksi kičmene moždine predstavljaju zaštitu pojedinih organa za koje se uspostavlja refleksni krug. To Je slično kao kad se u krugu za regulaciju brzine tiristorshl upravljanog motora postavi osigurač Za prekostrujnu zaštitu s tim da Je« grubo gledano« ovde refleks izbacivanje
os iguraca.
Takodje se zna da najsloženija urodjena forma ponašanja kao sto je regullsanje ravnoteže razmena naterlja «koju obezbedjuje disanje« hranjenje I te riaor egu i ae i ja « ostvaruje posredstvom mehanizama koji se nalaze u gornjin delovima moždanog aparata (produženoj moždini« hi potaiamusu1. Jos složenije forme ponašanja koje zahtavaju obezbedjivanje tonusa 1 koordinacije tesno su povezane sa radom malog mozga i subkortikalnih notornih centara. Konačno« najsloženije forme delatnosti ne mogu biti obezbedjene bez ucesca moždane kore« koja predstavlja organ najviših formi ponašanja Životinja I svesnog ponašanja coveka. Iz . ovih cinjenica te vidi da uloga nlzIh nivoa moždanog aparata Ima odlucujueu ulogu u onim funkcijama koje su najznačajnije za celi organizam (održanje organizma). Ujedno te funkcije po svom karakteru 1 psihičkim delatnostima su nize. Ka« bi se to pokušalo preslikati na neki računarski sisteai onda bi nlzi nivoi preuzimali funkcije koje su po svojoj kompleksnosti nize ali značajnije za ukupnu organizaciju i funkciju slstena« dok bi visi nivoi vršili detaljnu obredu ' infornacija i preduzinali potrebne akcije koje ona Iziskuje.
2.2 Strukturna I funkcionalna organizacija moždane kore
Posnatranja koja su pokazala	da mozak u
celosti« a dellmicno i njegova	kora . nisu
Jednorodne strukture datiraju .	Iz početka prošlog veka.
Kako su pokazala kasnija Istraživanja nova moždana kora sa sastoji od sest slojeva ćelija. Samo donji od njih predstavljaju aparate koji neposredno povezuju moždanu koru sa periferijom« organima euis (IV- aferentni Sloj) I misicima IV • eferentnl sloj). Ovi slojevi se nazivaju primarne zone kore.
Slika 6. Gruba lopogrofsko karta moždanih iona olspoljno površino bi unutrašnja površina
Kako su pokazala istraživanja nad svakom <prImarnomi« zonom kore nadogradjuje se sistem ■sekundarnih'« zona u kojima dominantno mesto zauzimaju složeniji po svojoj strukturi II I III sloj. TI slojevi sastoje se od cellja sa kratkim aksonlma od kojih najveći deo ili nema veze sa periferijom Ili dobije svoje impulse Iz subkortikalnih formacija« koje se nalaze u dubini mozga« a koje ostvaruju prvu preradu
informacija sto dolaze sa periferije« Struktura tih slojeva omogućuje da se svrstaju u slozeAlJI ssos Ica Jat Ivn ) Ml Integralni aparat moždane Kore«
Hijerarhijska struktura moždane kore tako se moze videti ako se pogleda topografska hart» mozga (slika 6). Kao sto se na karti moze videti* u moždanoj kori coveka se mogu Izdvojiti oblasti koje leze na grenicana Izmedju kortlkalnlh predstavništva pojedlnili osecajnlh xona moždane kore* a koje su dobile naziv tereljerne zone kore (IM zone prepokr I vanja Itor ti kalnl I) delova pojedlnlti analizatora). Te oblasti Kor« se sastoje u celosti od gornjih slojeva celljs I one nemaju direktne veze sa per I fer i Joni< Postoje sve osnove za predpostavku da tereljerne zone kore obezbedjuju zajednički rad grupa analizatorai tj< najsloženiju funkciju mozga.
Sve do sada Izneseno vodi do zaključka da računarski model koji bi koristio osnovne principe rada mozga mora biti hijerarhijska struktura u vise nivoa. 0 Samoj h iJerah I JsKoJ strukturi bit ce detaljno govoreno kasnije ali ono sto Ja važno napomenuti Je to da nizi nivoi hljerahlje sa sCanovIsta ukupnog sistema moraju Izvršavati značajnije funkcije. Ovo proizlazi Iz saznanja da nizl nivoi mozga kontrolisu najznačajnije funkcije organizma« Kao sto su disanje^ pulsj lučenje zlezda ltd. Oni ne izvršavaju vise psihičke deiatnostlf ali Igraju odlueujucu ulogu u funkcijama održavanja organizma sto je sa stanovista ukupnog sistema najznačajnija funkcija.
Kako su pokazala i s trazi v an J a« dete dolazi na svet sa potpuno oformljenim aparatima subkortlkalnlh formacija i naJJednostavnlJIm^ proJeKcionim Ili primarnim zonama kore I sa nedovoljno oformljenim aparatima složenijih sekundarnih! tercljalnlh zona kore. Najburnije sazrevanje sekundarnih delova kore zapaza se u uzrastu od 2-3 godine života dok najsloženije sazrevaju konačno u uzrastu od b-7 Terctjernl delovi kore sazrevaju u periodu do 3-3«9 godina života dok najsloženije Cblast-I produzavaju svoj razvitak do	godine
života. Ova hljerahljska struktura moždane organizacije I same kore govori da su najznačajnije funkcije po organizatn (za njegovo preživljavanje u sredini) kod coveka ugradjene od samog rodjenja* dok sve vise forme psihičke delatnosti se razvijaju kroz razvoj sekundarnih I tercijalnih zona kore g samim tim ( vlsfh nivoa moždane hijerarhije. To znael da nlll nivoi hijerarhije pored svojih urapred definlsanlh funkcija Imaju I dodatnu funkciju 'stinuiIsanja'r 1 odredjlvanja razvoja visih nivoa Jer njihove funkcije se tek kasnije u ontogenez) razvijaju zajedno s razvojem
sekundarnih zona kore. Ako se povuče paralela na model računarskog sistema to znaci da nizl rivol hijerarhije bi morali da pored svojih normalnih (nlzlh) funkcija 1 da aktivno učestvuju u funkcijama vlsth pa cak i da te funkcije odredjuju.
James Albus u svojoj knjizi 'Brains* Behaviors C Robotics'/ Je dao hljerahljskl model mozga (slika 7.1 7.Z) gde nIzi nivoi ekstrahiraju Informacije za vise nivoe. Taj made( bi vazlo kad bi se covek radjao s potpun« formiranim sekundarnim I tercijarnim zonama korei Jer onda bi nizl nivoi hijerarhije Izlučivali Informacije za vise. Ali kako to Izlučivanje izvršiti kad visi nivoi hijerarhije nisu formirani do Kraja I kad nizl nivoi uticu na forulranje visIh?
2.3 Tri osnovna funkcionalna bloka mozga
Hogu se Izdvojiti tri osnovna funkcionalna bloka ili tri osnovna moždana aparata cije Je ucesce neophodno za ostvarivanje bilo kakvog oblika psihičke delatnosti. TI osnovni blokovi su> li blok reguitsanja tonusa ili stanja budnosti« Z. blok prljena« obrade I čuvanja informacija* 3. blok programiranja* reguMsanJa
oblasti godina.
i kontrole psihičke delatnosti*
Svaki od tih osnovnih blokova poseduje hijerarhijsku strukturu I sastoji se od najmanje trt nadgradjene zone tlpa> primarne* u koju stizu Impulsi sa periferije ili odakle se upućuju impulsi prema periferiji; selturtdarne* u kojoj se obavlja obrada primljene informacije Ml se pripremaju cdgovartjuci progranij tercijarne* KoJ« predstavlja aparate vel Ik i h hemisfera koji obezbedjuju najsloženije forme psihičke delatnosti. Da bi računarski nodal opstao morao bi zadovoljit) upravo predloženu hijerarsIjsku strukturu Koja bi na neki nacIn oponašala upravo »Omenuta trt funkcionalna bloka mozga. Taj racunarsi sistem bi morao Imati adekvatne Komponente pojedini* funkcionalnim blokovima I toi
-	Aktivirajucu (uzbudnui konponentu
-	Obradjujucu komponentu
-	Programirajucu konponentu
2.3.1 budnosti
Blok reguHsanja tonusa I stanj«
Jedno Od najvažnijih otkrića za mesto bloka koji reguMs« tonus kore Je otkrivanje cinjenice da se ti aparati ne moraju nalaziti u samoj korif vec u delovina mozga koji se nalaze u delovitna moždanog stabla i subKor 11 k a I n I m formacijama* I da se tl aparati nalaze u dvojnom odnosu prema kori podižući njen tonus a Istovremeno I osecaju na sebi njen regulacioni uticaj.
Godine 1949 dva Istaknuta naučnika* Kegun I Moruci* otkrili su da se u delovina moždanog stabla nalazi posebna nervna fornacija* koja Je po svojoj strukturi i po svojin funkcionalnim svojstvlna prilagodjena zadatku da ostvaruje ulogu mehanizama koji regullsu stanja moždine kore tj. on« Je u stanju da «enja njen tonus I obezbedjuje stanje budnosti.
Ta fornacija Je sagradjena po tipu nervne mreže u koju su ubačena tela nervnih cellJa. Po mrezi te formacije koja Je nazvana RETIKULARNA FCJRrtACtJA* nadrazaj se siri ne pojedinin Izolovanin impulsima* ne po zakonu sve III nista* vec postepeno menjajuci svoj nivo I modulirajući stanje celoKupnog nervnog aparat« (slika 7).
Slika 7. Shemo oKtivirojuiI« retikuiorne formacije
Jedan deo vlakana retikuiarne formacije ide nagore završavajući se u visim nervnin formacijama I na Kraju u foraacijama nove kore cija Je uloga U organizaciji složene psihičke delatnosti. Te formacije su dobile naziv uzlazni rettkuiarni sistem. Samo otkriće da se u evoluclono najstarijim delovlaa «ozgt nalazi takva formacija koja reguiise ukupni tonus kore govori puno/ Jer so to Istovremeno najniži nivoi hijerarhije mozga. Računarski gledano to znaci da nizl nivoi predloženog hijerarhijskog modela deiuju upravljački nt vise nivae. Ovaj
model Je suprotan svim dosadasn J lin h I jer ih I jsn I (d irođellma jer su uveK rìii nivoi samo ekstrahirall informacije za vise a odluke donosili vis) nivoi, Mosuencst opstanka ovakvog modela lezi u cinjenica sto nlii nivoi ne obrađjuju. sve informacijs vec samo ore koje su najinteresantnije za održanje sistema. Ujedno vreme obrade takvih informacija Je minimalno» jer reakcije moraju biti jako brze. Zbog toga nlzi sistemi hijerahlje moraju uticati na vise kako bi ukupna reakcija sistema blia skoro trenutna« a samim tim visi nivoi se moraju podčiniti nizim, Cvo dovodi do zaključka da najnijin nivoima moraju biti dostupne sve Informacije koje doiaze u sistem. Priroda je to resila postavivsl retlkuiarnu for«3elju kao Jednu mrezastu cllindricnu tvorevinu oka svih nervnih vlakana koji doiaze u moždani sistem (slika 6).
Slika a. Aferentni 3ut6vi i osecojne zone Ipo f i. Poljaiiovu)
Druga viakna retikularne formacije kreću se u suprotno« smeru. Ona polaze od visih nervnih fornaclja I Idu ka nizin Strukturana srednjeg mozga« hipotalamusa I moždanog stabla. Te fornaclje su dobile naziv silazni retlkularnl sistem i one stavljaju ostale delove mors« P«đ kontrolom onih prograna koji se stvaraju o kori velikog mozga i za clJe Je Izvršavanja potrebna modifikacija I modulacija stanja budnosti. Ovo
poznata Sto znaci da za nju nije potrebna posebna mobiiiiacija organizma. S»«io takav mehanizam može da obezbed) proces privikavanja kada vise puta ponovljena drai gubi karakter novoB. Očigledno je da je u taj iehaniza« učenja uključen «ehanlzaa pamćenja koji omoeucuje tekvu komparaciju. Znatan deo ljudske aktivnosti uslovljen je naaerana I Planovima/ perspektivama I prograitima koji se formiraju u procesu svesnog života coveka I oni predstavljaju treci Izvor aktivacije. |
1
Tipovi izvoraaktivac I je govore da aozdanl sistem stalno vodi računa o svon unutrašnje« stanju 1 oiiolini I tekuće« Izvršenju prograna psihičke delatnostl. Kod računarskih sistema koji bi koristili principe moždane organizacije bilo bi za očekivati di sva tri Izvora aktivacije budu zastupljen«. Drugi i trecI po samoj prirodi modela moraju biti prisutni ali se postavlja pitanje sta s prvin tipom Izvora. Kod covek« postoji veliki broj redundantnih sistema *koJi preuzimaju funkcije ako glavni nosilac funkcije odkaze. Najbolj) prlner ja disanje koje Je Jedna od kritičnih funkcija za život. Kad se presecanjen nerva koji regulise (upravlja) redom dijafragme ona Isključi Iz rada ne dolazi do gušenja vec njenu funkciju preuzimaju medjurebarnl mlstel koji stre 1 skupljaju grudni kos I tako onogucuju razmenu vazduha u plucina. Ista stvar se dogadja I sa moždanim centrima. Kad neki centar usled nekog oštećenja otkaze» drugi preuzimaju njegove funkcije u potpunom IM degrad iranon obliku. To maci ako se zeli bilo kakva r edundtnci Ja u modelu (neki oblik sistema neosetljivog na greške) ovij Izvor aktivacije nora biti prisutan. U stavrnisti model bi stalno vodio raune u kakvom se stanju (funkcionalnom) nalaza pojedine komponente hijerarhije I sami nivoi hijerarhije. Globalne odluke o stanju sistema I njegovo eventualno rekonfiguri san je vodilo bi se na najniže« nivou hijerarhije
2.3.2 Blok prijema* obrade 1 čuvanja Informici ja
DvaJ blok se nalazi u spoijnjtn delovima nove kore I zauzima njene zadnje delove> ukljucujuci u sebe aparate vidne (pot 11jacne)> slusne (slepoocne) i opsteosecajne (temane) oblasti, isastojl se od subkorti kalnI h neurona I moždane kore. 'ti neuroni rade na principu sve III nista primajući odvojene Impulse I prenose Ih na druge grupe neurona. Drugim recima ovaj funkcionalni blok prima nadrazaje od perifernih receptora« rastavlja Ih u najsitnije delove (vrsi analizu najsitnijih deiova informacija)« I vrsi kombinovanje u potrebne dinamičke funkcionalne strukture (vrsi sintezu u čitave funkcionalne sisteme). U računarskom modelu ovaj blok bi morao biti
govori 0 postojanju 'povratne veze'f od visIh nervnih formacija ka nizlia Ili od vlsih psihičkih procesa ka nizina. Računarski gledano* to znaci da visi nivoi hijerarhije imaju povratni uticaj na nize nivoe I to takav da oni mogu modultratl utieaj nizih nivoa ali sa ogranlcsnjem da on moze postojati samo do Jedne granice. Visi nivoi ne mogu Isključiti funkcije nIzih Jer u tom slučaju bi postojali radi sebe samih. Poznato Je da se nervnl sistem nalazi stalno u stanju odredjene aktivnosti 1 da Je postojanje Izvesnog tonusa neophodno za Ispoljavanje bilo Kakve aktivnosti .
(logu se Izdvojiti tri (zvora aktivacije nervnog sistema. Prvi od Izvora je proces razmene «atertja organizma koji leze u osnovi unutarnje ravnoteže organizma 1 Instiktivnih procesa. Drugi Izvor aktivacije je povezan sa dolaskom u organizam drazi iz okolnog sveta 1 on dovodi do nastanka potpuno drugih formi aktivaclje koje se ispoljavaju o obliku orjentacionog refleksa. Svaka reakcija na ■novotariju'« zahteva poredjenje nove sa sistemom starih drazi. Saao takvo poredjenje moze da odredi da 11 je data draz nova i zahteva II orjentacioni refleks ili Je vec
zastuplje kao analizator I sintezator svih ulaznih Informacija u sistem 1 služio bi kao podloga za buduće reakcije sistema. Organizacija ovog bloka analogno predhodnom bi morala da bude hl Jer ah IJska.
Osnovu ovog bloka sačinjavaju primarne Ili projekclone zone moždane kore« koje karakterise razvoj neurona IV aferentnog sloja« elji znatan deo poseduje najveću modalnu spec IfIcnost> Hpri pojedini neuroni vidnih aparata reaguju samo na nijanse boje« karakter linije« pravac kretanja. U prlnarnirn zonama pojedinih oblasti mogu se sresti neuroni mul 11«oda inog karaktera koje reaguju na nekoliko vrsts drazi ali njihov broj ni Je vec 1 od ^ - 5 X,
Uprava ova cinjenica moze dovesti u sukob s predloženim modelom jer najniži nivo bi morao da vrsi najgrublju obradu InforntcIJa « nadredjeni su visem. Kako to da neki nivo koji vrsi ani Izu najsitnijih deiova Informacij* moze biti na nizem nivou od nivoa koji vrsi sintezu? Odgovor lezi u načinu analize Informacija i reakcija na nju. Rastavljanje Informacija na delove dato je njenom prirodo« i ono samo n« znaci nista. Nedjutim« reakcije sistema na neke elemente Informacija (npr. bljesak« Jak zvuk«
gorak ukus) inortju biti brze (trenutne) I taj niz) nivo mora tu brzu reakciju omogućiti. TeK kasnije se ustanovljava tsinteio« Informacija) sta se dogodilo I da II Je reakcija sistema bila ispravna III ne. To znaci da nizl nivoi hijerarhije nodeta vrse rastavljanje InforaacIJa na sastavne delove za vise nivoe« s reakcije koje oni preduzimaju na osnovu tih delov« infornacije su u svojoj susCInI značajnije za siste« od reakcija visih nivoa.
Nad prlinarnoin nalaze se aparati sekundarnih zona kore u kojima vodeće netto zauzinaju neuroni II I III sloja« I oni nenaju Izraženu nodalnu specifičnost sto omogućuje da se konbinuju drazi koje doleze u potrebne ■funkcionalne oblike* I tako ostvaruje sintetička funkcija.
Saznajna delatnost coveka se nikad ne oslanja samo na jednu vrstu drazI vec predstavlja rezultat poli- modalne detatnostl. Zbog toga sasvin Je prirodno sto on« mora da se oslanja na zajednički rad celog sfsteaa zona kore velikog nozga. Funkciju obezbedjIvanja takvog zajedničkog rada cele grupe- analizatora obezbedjuju I nose tercijarne zone bloka III kako se one nazivaju zone prepokrIvanJa kortikainth delova različitih analizatora.
TercIJerne zone skoro u celesti sastoje stoja kore sto znaci skoro u potpunosti stlmulusa koji dolaze Smatra se da najveći br stoJa Je nultlmodalno na kompleksne oznake sr u prostorIJIf broj e reaguju neuroni prva dv
zadnjth delova mozga se od cellja II I III da se njihova funkcija svodi na Integraciju sa razni h anal Izatora. oj neurona tercijarnog g karaktera I da reaguju edine (oznake raspoerda lenenata ...) na koje ne a s loja.
Delatnost tercijarnih zona zadnjih delova kore Je neophodna ne samo za uspesnu sintezu očigledne Informacije vec I za prelazak od nivo« neposredne očigledne sinteze na nivo slobolicklh procesa» za ' operisanje sa znacenjina reci« slozenin gramatičkim i logičkim strukturama/ sa sistemima brojeva I apstraktnim odnosima. Oruglm recima tercijarne Zone predstavljaju aparate cije Je ucesce neophodno za pretvaranje očiglednog zapazanja u apstraktno mišljenje I za očuvanje u pamćenju organ Izovanog	Iskustva.	Organizacija
sekundarnih I t ercI jer zon I h analizatora govore da za detaljnu obradu prispelih Informacija treba puno vise vremena nego za primarnu analizu. To znaci da reakcije sistema su puno sporije ail puno kvalitetnije I da nivoi hijerarhije koji vrsi sintezu informacija moraju biti u visiR nivoima hijerarhije/ Jer z« fine I visoke akcije po svojoj strukturi moraju pored nizih nivoa biti uključeni i visi. avo sto Je
Aparati trećeg funkcionalnog btoka nalaze se u prednji« deloviaa velikih hemisfera/ Ispred prednje centralne vijuge. Izlazna vrata» tog bloka predstavlja notorna zona kore velikog nozaa* čiji V sloj sadrži Secove gigantske piramidalne cellje/ c I Ja vlakna vode ka motornim Jedrima kičmene nozdine. U toj zoni su u najvećem stepenu zastupljeni organi koji Imaju naJfunkc lonainlJI značaj I kojima je potrebna najfinija moguća regulacija. Prirodno Je d« motorna struktura Impulsa koje salje ovaj deo kore mora biti dobro pripremljena tj> uključena u odredjene programe/ Jer bez takve pripreme se ne bi nogli-obezbediti svrsishodni pokreti. Sa stanovista modela primarni sloj ovog bloka bit ca najviši sloj hijerarhije predloženog modela. Razlog za to je očigledan Jer on bi u računarsko« modelu . upravljao aktuatorima koji upravljaju procesom lil su veza s oko I Inom.
Priprema motornih impulsa ne moze biti ostvarena samo piramidalnim ćelijama vec mora biti obezbedjena kako aparatom prednje cantretne vijuge tako I visim aparatima •sekundarnih*/ zona motorne kore. Prednja centralna vijuga predstavlja projekclonu zonu/ izvrsni aparat moždane kore. Odiucujuci značaj u pripremi motornih impulsa imaju sekundarne I tercijarne zone» koje se potclnjavaju Istim principima hijerarhijske strukture i silazne specifičnosti kao I predhodna dva bioka ali u obrnutom redostedu. l>lagu sekundarnih zona Igraju premotorni deiovi čeone oblasti mozga. Draženje tih delova kore izaziva čitave kompleksne pokrete kao sto su> pokreti oelju* glave i celog tela I pokreti ruke koja nešto uzica... Sa stanovista modela znaci da neposredno nizt nivo vrsf organizaciju Ipr Ipremu) .reakcija sistema/ ali ne samo sa stanovista proračuna (programa) reakcije sistema vec I sa stanovista buduće reakcije s istema.
Najznacajnij deo ovog funkcionalnog bloka predstavljaju prefrontalnl deiovi mozga koji su sastavljeni od velikog broja ceitja II i Ilt sloja 1 predstavljaju tercijarne zone. Oni Igraju najznačajniju ulogu u formiranju namera i programa i regullsanja I kontrole najsloženijih formi ponašanja coveka. (/pravo zbog toga/ prefrontaina obtast Ima najbogatiji sistem veza sa drugim delovima nozga kao i sa retikularnom	formacijom.	Zahvaljujući
dvostrukom karakteru tih veza prefrontalnl deiovi kore se nalaze u posebno povoljnom polo-polozaju I u odnosu na ponovnu preradu najsloženijih uzbuda/ koje dolaze od svih dalova mozga i za organzaclju upravljačkih impulsa koji omogućavaju da se obavija regulaciono delovanje na sve te strukture. Ako najviši nivo hijerarhije Je neposredni
Iziozeno suprotno je svim dosadasnjim model ina analize 1 sinteze Informacija. Kao potvrda ovome pristupu moze poslužiti podatak da na spomenute drazi tipa* bljeska/ gorkog ukusa/ jakih zvukova« reakcije mozga nisu na nivou nove I stare moždane kore vec na ■ nivou timbickog sistema koji se nalazi lispod'« nove I stara kore. Za detaljnu analizu I sintezu tih drazi kad se ustanovljavaju razlozi njihove pojave« reakcije su na nivou nove I stare kore.
2.3.3 Blok progrargi r anja« regullsanja I kontrol e
složenih formi deiatnosti
Prijem« prerada I čuvanje Informacija predstavlja samo Jednu stranu psihičkog života coveka. Njenu drugu stranu predstavlja organizacja	aktivne«	svesne psihičke
deiatnosti. Covek ne reaguje samo pasivno na signale koje prima. Qn formira planove i programe svojih akcija/ prati. njihovo ostvarivanje I regulise svoje ponašanje/ usagiasavJucI ga sa tim planovima i programima. Na kraJu/ on kontroilse svoju svesnu delatnost« poredeci efekat svojh akcija sa polaznim namerama i korigujuci počinjene greška, vka.
Izvrsiiac radnje onda nizl nivoi aoraju pripremati globalne instrukcije za njihovo izvršenje. Ta priprema globalnih radnji je tesno povezana s predhodno dva opisana funkcionalna bloka nadela I ne moze se vršiti bez njih. Nizl nivoi hijerarhije donose globalne odluke o reakciji sistema dok visi nivoi samo Izvršavaju globalne odluke i eventualno vrse njihovo doterivanje. Ovde je važno napomenuti značaj povratnih Infarnacija koje ovim nivoima govore o izvršenju predvidjenih programa I reakcija. Na osnovu njih ovaj blok vrsi korekciju Cdoterlvan Je) svojih programa tj. globalnih akcija.
Od odiucujuceg Cnacaja Je činjenica da čeone regije mozga poseduju jake snopova uzlaznih I silaznih veza sa retlkuiarnoa formacijom. T« oblasti nova kore dobljaju impulse od sistema prvog funkcionalnog bloka dok Istobvemeno one vrse tntezivan noduiacioni uticaj na formacije retikularne formacije« dajući	njenim akt I v IraJucI m Impulsima
diferencirani karakter i usaglasavajuci Ih sa onim dinamičkim seirama ponašanja koje se formiraju neposredno u čeonoj kori mozga. Ovo ističe usku funkcionalnu saradnju prvog i trećeg funkcionalnog bloka 1 vezu sa svim
honponentama nervnoo sistema» Sa stanovista nodela to znaci da su uzbudnl I programirajući sisteni tesno povezani. Programirajući tioze delovati na uzbudnl^ ali konačne odtuke za slxten uvek donosi uzbudnl. Kod modela treba Istaci d« se sva tri funkcionalna bloka
-	aktivirajući
-	obradjujuti
-	. progranirajucl nalaze nadjusobno Iznesanl na svim nivoima hijerarhije I to ne kao odvojene komponente vec integralno u swim homponentatna hijerarhije.
3. Z A K LJ U C A K
tz svega do sada Iznesenog sledi da računarski sistem koji bi pokušao iskoristiti osnovne postulate rada aozgs mora zadovoljiti sledeče globalne zahteve:
-	Da -Je -to dinamička hijerarhijska struktura sa vise nivoa
-	Oa Je stanje takvog sistema odredjeno spoljnom okolln«n» unutrašnjim stanjem i funkcijom koja se trenutno Izvršava
-	Da su svi ulazi u sistem direktno na najnižem nivou a svin ostalim nivoima se prosledjuje samo onaj podskup ulaznih informacije koji nu Je potreban za obavljanje njegove funkcije
-	Oa najniži nivo na osnovu Infornacija Iz spoljnjeg sveta* unutrašnjeg stanja I funkcije koja se obavlja u sistemu donosi odluke o budućoj reakciji sistema (funkciji) I na osnovu nje inici.ra dalju analizu I sintezu ulaznih Infornacija. Istovremeno vrsi distribuciju odabrane reakcija ostalim elementima sistema
Da poseduje sledeče tri komponente ukonponovane na svim nivoima hijerarhije
-	akt i-v i rajucu
-	obradjujucu
-	programirajucu
-	Oa Ina dvosnernu komunikaciju Izmedju nivoa I unutar nivoa hijerarhije
-	Oa na osnovu odredjene buduće reakcije se Izvrši rekonfIguraeiJa slsteaa kako bi sa sistea najbolje prilagodio nastaloj situaciji
-	0« naJnlzl ntvo svoje odtuke i reakcije donosi skoro trenutno» a svi ostali nivoi u zakašnjenju za njin
-	Da postoje mogućnosti uticaja viših nivoa na/nize u procesu odredjlvanja buduće reakcije sistem«
-	Da svi nivoi Inaju mogućnost uticaja na kontroilsani proces stin da sto je nivo visi to njegov uticaj mora biti višeg kavtiteta
-	Oa ako nanizl nivo odredi pogrešnu reakciju sistema ne postoji niko ko bi tu reakciju promenio u drugu
LITERATURA
1.	Janes S. Albus. "Bralnsi Sehavior« and Robotics" BYTE Books, Subsidiary of ncGrawHi I If 1981.
2.	A.R. Lurija. "Osnovi neuropsI ho I eg I Je" prevod sa rusliog, Nollt, Beograd« 19e3i
3.	Lav Vigotski "nisljenje I govor" prevod sa ruskogi hoiit, Beograd« 1963
George 0. Watts. "Dyna«lc Neuroselenee"« Harper and Roh Publisher« 197$.
5.	John F. Jervis. "Robotics"« Conputer* strana 2B3>292> oktobar 196«.
6.	Mlillaa T. PoHers. "The nature of robots" Byte« Jun« Jul« avgustrseptembar 1979
7.	Andrew Filo. "Deslgnl<ng a robot from nature" Byte« februar, mart 1979.
e. Steven t, Svetiana Kartashe*. "Dynamic archicteturesI Probieas and Solving" Computer« Jul! 1976,
KOMUNIKACIJSKI IN APLIKACIJSKI PROCESI V VIŠJIH NIVOJIH REFERENČNEGA MODELA OSI
UDK 681.3
B. Jerman - Blazič in M. Kapus - Kolar Institut Jožef Stefan, Ljubljana
t> s-t. Ä C-fc
The paper deals Hlth the conununlcatlon concepts and functions asaoclated with the upper layers of the ISO/OSI Reference Model. The basic comraunica-tion functions in the upper layers of the model are compared to the information processing functions of a computer aystam. The role and the task of both functions in a distributed system are explained in an understandable way.
rxo^vz; o ^ ole
Prispevek obravnava komunlHaclJake koncepte in funkcije vifijih nivojev referenčnega modela OSI. Primerjali smo naloge komunikacijskih funkcij v modelu a nalogami funkcij za obravnavo in obdelavo Informacij v raćunal-nläkem sistemo. Vlogi obeh funkcij v diatrlbuiranem sistemu sta raaloienl na enostaven natin.
X . UN'ocä
Vsaka komunikacija med dvema ali vet partnerji vključuje komunikacijske funkcije, ki Jih Identificiramo Earadi preglednosti in načrtovanja sistemov kot tri razrade komunikacij-akih funkcij. Ta razdelitev velja ne glede na to, ali komunikacijo iavajajo ljudje s prenaSanJejn pisem ali komunicirajo stroji z elektronskim izmenjavanjem podatkov. Standardi iz referenčnega modela OSI obravnavajo te funkcije za primer komuniciranja preko naprav za prenos podatkov, na primer preko mrc4 s preklapljanje paketov ali vodov, najetih linij, lokalnih mrež ali Javnih mrei za prenos podatkov
Komunikacijske funkcije vsakega od razredov obravnavajo različne, dobro ločljive vidike komuniciranja. TI so:
1.	prenos podatkov - to so komunikacijske funkcije za prenos predstavitev informacij (podatkov) la enega končnega al sterna v drugega, s pogostostjo napak, sprejemljivo za aplikacijski proces.
2.	povezava - to so komunikacijske funkcije, ki omogočajo aplikacijskim procesom, da vzpostavijo dialog. Ker v vsakem računalniškem sistemu pogosto obstaja več aplikacijskih procesov, ni dovolj, da se podatki dostavijo v Izbrani končni sistem. Potrebne so Še funkcije, ki ugotovijo, kateri aplikacijski proces poèilja oz. sprejema informacije.
3.	sodelovanje - to so komunikacijske funkcije, ki omogočajo aplikacijskim procesom smiselno komunikacijo In izvajanje procedur zA porazdeljeno procesiranje informacij.
Te funkcije so del sistemske opreme sodobnih računalniških alatenov, ker Jo le s pomočjo
teh funkcij možna komunikacijo med porazdeljenimi aplikacijskimi procesi z napravami za prenos podatkov. AplikaciJsklra procesom morajo biti na razpolago v računalniških sistemih, v katerih so aplikacije locirane.
Slika 1. ilustrira odnos med informacijsko-procesnlml funkcijami enega računalniškega sistema In komunikacijskimi funkcijami aplikaci jakega procesa. Za uporabnika se komunikacijske funkcije izvajajo kot del aplikacijskega procesa, z vidika sistema pa se te funkcije izvajajo kot funkcije ločenih nivojev. Nekatere od teh funkcij so tesno povezane s specifičnimi komunikacijskimi potrebami uporabnikove aplikacije, zato se te funkcije specificirajo za vsako aplikacijo posebej. Uporabniško specifične funkcije se ve-iejo na funkcije v sistemu, ki so sploSnega pomena in so uporabne pri večjemu itevilu različnih aplikacij. Sploinlh funkcij v računalniškem sistemu, ki Ima vgrajene funkcije za komunikacijo, ni treba posebej specificirati, ker so uporabniku na razpolago kot del sistema. KaJbolJ elementarne komunikacijske funkcije 30 sistemsko specifično funkcije, ki delajo s konkretnimi napravami za prenos podatkov.
2. OpirecLeli-to-v Icomuini-IcaodL Jsslcd-Vi ^vinltcij v OSI RM
OSI RM identificira naštete tri nivoje komunikacijskih funkcij in jih dalje deli v sede« nivojev, ki ustrezajo sedmim nivojem tega modela. Vsak od nivojev nudi del množice vseh funkcij, ki so potrebne za sodelovanje oddaljenih aplikacijskih procesov. OSI RM podaja abstrakten pogled na te funkcije, ki ga prikazuje slika 2. Slika 2. kombinira različne poglede na aplikacijski proces In
prikaauje odnos med tremi nivoji komunikacijskih funkcij in nivoji OSI BK. Tudi v OSI RM so funkcije prenosa podatkov ločene od funkcij poveEovajciJa in sodelovanja. S to lotl-tvljo funkcij na meji med transportnim nivojem in nivojoi» 5ejc je naznaćena razlika med funkcijami prenosa podatkov (spodnjimi nivoji) in funkcijami pora«deljenega procesiranja podatkov (zgornjimi nivoji).
V nadaljevanju privzomamo, da ao funkcije spodnjih nivojev realizirane. Te funkcije

Pranol »Bdel ko I
laliraicijtta
frtcttM
Uakcijt
Siitin
Slika 1 Itilijni vidiki ktrnufiiktcijtlii funkcij
podpirajo funkcije zsornjih nivojev o prenašanjem predstavitev informacij mod kontnimi sistemi v obliki nlaov binarnih oktetov m nudijo servis s kvaliteto, ki je sprejemljiva za dane aplikacijske procese. Funkcije vseh spodnjih nivojev skupaj nudijo aervla transporta podatkov, ki podpira lavaJanje funkcij Egornjih nivojev.
Kot ilustrira slika 3-, Je rezultat izvajanja funkcij spodnjih nivojev prenos nizov binarnih oktetov iz oddajnega v sprejemni konòni sistem (lahko preko poljubnega ätevila sistemov - posrednikov). To je analogno prenosu pisem ali magnetnih trakov. Servis, ki ga nudijo spodnji nivoji, Je kot servis dostave poälljke "od vrat do vrat', ki ga nudi poétni sistem. Slovenski del JUPAK-a na sedanji stopnji razvoja nudi le približek tega servir sa (prve tri, namesto Štirih nivojev}. Navpične puičlce na sliki 3 predstavljajo tok podatkov iz enega kontnega sistema v drugega. Vodoravne puSòlce predstavljajo interakcije ali protokole, potrebne v vsakem od nivojev ra izvajanje funkcij transporta podatkov. Po terminologiji OSI ao to funkcije in protokoli, definirani a OSI standardi za fizični, linijski, tnreini in transportni nivo, ki nudijo zgornjim nivojem servis transporta podatkov.
Preden aplikacijski procesi lahko uporabljajo podatke, jih morajo sprejeti in pravilno interpretirati. V vsakem računalniškem sistemu lahko obstaja veliko aplikacijskih procesov. Zato se morata aplikacijska procesa, ki želita izmenjavati podatke, najprej dogovoriti za komunikacijo in določiti pravila, po katerih bo potekala. Ker vsak proces po svoje predstavlja informacije, se morata dogovoriti tudi se enotno predstavitev informacij pri izmenjavi.
Za to potrebne funkcije in protokoli omogoča-
jo aplikacijskim procesom v različnih računalniških sistemih povečavo in sodelovanje. V OSI RM so definirane kot funkcije nivoja seje, predstavitvenega in aplikacijskega nivoja.
3 .
Fo-vozjovajn. J® — ri±-vo
teaiwìlacijét fwhciji
-Uporabnitki litik* r~| .Siittmiki Yi(ik'
Kivo seje nudi funkcije za poveaavo dveh aplikacijskih procesov v odnos logične komunikacije in funkcije sa organisacljo in sinhronizacijo njunega dialoga. Pri tem se uporabljajo mehanizmi za vzpostavitev, vzdrževanje In razpustitev povezav na nivoju seje. Povezava na nivoju seje Jo dogovor med dvema aplikacijskima procesoma o vključitvi v kontroliran dialog za izmenjavo podatkov. Aplikacijski procesi torej izmenjujejo podatke s poBoćJo povezav na nivoju seje, S pomočjo mehanizmov povezav na nivoju seje lahko aplikacijski procesi pošiljajo podatke, ki jih sprejemni sistem dostavi naslovljenemu aplikacijskemu procesu,
Preden začneta dva aplikacijska procesa lame-njavati podatke, morata vzpostaviti povesovo na nivoju seje. Vsak aplikacijski proces lahko sam sproit pobudo za vzpostavitev povezave na nivoju seje, lahko pa tudi sprejme zahtevo za vzpostavitev povezave od drugega aplikacijskega procesa. Ko Je to opravljeno. Je odnos med aplikacijskima procesoma vzpostavljen in procesa lahko začneta izmenjavati, podatke.
Povezavo na nivoju seje lahko obravnavamo kot povezavo med dvema aplikacijskima procesoma preko nivojev seje njunih kontnih sistemov. Vendar ne smemo pozabiti, da Je povezava na nivoju seje odvisna od povezav, vzpostavljenih na nižjih nivojih za izvajanje funkcij nivoja seje. Te povezave ao potrebne za prenos podatkov in protokolnlh IsforMaclJ med končnima sistemoma. Privzeli bomo, da te povezave obstajajo, to Je da Jih spodnji nivoji vzpostavljajo in roapuSčaJo v skladu s potrebami povezav in funkcij na nivoju seje.
Povezava na nivoju seje se lahko raztegne tudi čez več povezav na spodnjih nivojih. To lahko primerjamo z dvema iahistoma, ki igrata partijo preko telefona. Podobno kot poveaava na nivoju seje, partija teče, dokler se eden od äahlstov ne odloči za konec. Dejanska povezava preko telefona pa se lahko medtem večkrat vzpostavi in spet razpusti, saj Je potrebna samo, kadar eden od iahiatov želi drugemu sporočiti naslednjo potezo.
Med vzpostavljanjem povezave se dva aplikacijska procesa dogovorita za pravila dialoga, ki se bodo uporabljala pri komunikaciji med njima. Pomen pojma dialoga na nivoju seje je analogen kot pri komunikaciji med ljudmi. Prvi tip dialoga označuje pretok informacija samo v eni stneri, na primer od predavatelja k poslušalcem. Fri drugem tipu dialoga tete informacija v obeh smereh, vendar hratl samo v eni smeri, ko na primer predavatelj prekinja predavanje, da lahko poslučalcl postavi jajo vprašanja. V tem primeru al predavatelj in posluéalci Izmenjujejo pravico govora. Primer tretjega tipa je komunikacija dveh ljudi, ki kontrolirata časovno kritičen proces (na primer pristajanje letala), zato morata nujno imeti oba ves čas pravico do govorjenja. Bistveno je, da potreben tip dialoga izhaja iz narave aktivnosti.
Blvo seje nudi ètevllne tipe dialoga za različne tipe aplikaciJakih procesov. Nekateri aplikacijski procesi {na primer tisti s področja kontrole časovno kritičnih procesov)
prlkasuja odnos med tremi nivoji komunikaclj-sliih funkcij in nivoji OSI RM. Tudi v OSI RM ao funkcije prenosa podatkov ločene od funkcij povoBOvanJa In sodelovaoja. S to ioti-tvijo funkcij na meji med transportnim nivojem in nivojem seje je naznačena razlika med funkcijami prenosa podatkov (spodnjimi nivoji) in funkcijami porazdeljenega procesiranja podatkov (agornjlrai nivoji).
V nadaljevanju privzemamo, da ao funkcije spodnjih nivojev realizirane. To funkcije
Jo aplikacijskim procesom v raslićnih računalniških sistemih povečavo in sodelovanje. V OSI BM so definirane kot funkcije nivoja seje, predstavitvenega in aplikacijskega nivoja.
SlliltMtji
finionnjt
Prug« polilltti

llrtr(laiiki-tp«cifi(M

iilti«
turtriMtiliki
ftlHtM
(«•tc ij*
iMoikMi/ltl
iMttiii
^ -8pof»tftilli »ijH-
n .iiitiiMti <i<ii"
Slika 1 UiliEiii vKiki ktmuniktcijiki fitnkcl)
podpirajo funkcije zgornjih nivojev 3 prenašanjem predstavitev informacij med kontnimi sistemi v obliki nizov binarnih oktetov in nudijo servis s kvaliteto, ki Je sprejemljiva sa dane aplikacijske procese. Punkcije vseh spodnjih nivojev skupaj nudijo aervia transporta podatkov, ki podpira izvajanje funkcij sgornjih nivojev.
Kot ilustrira slika 3,, je rezultat izvajanja funkcij spodnjih nivojev prenos niaov binarnih oktetov ia oddajnega v sprejemni kontni sistem (lahko preko poljubnega àtevila sistemov - posrednikov). To Je analogno prenosu pisem ali aagnetnih trakov. Servis, ki ga nudijo spodnji nivoji, je kot servis dostave poSilJke "od vrat do vrat", ki ga nudi poStni sistem. Slovenski del JUFAK-a na sedanji stopnji razvoja nudi le približek tega servisa. (prve tri, namesto Jtlrih nivojev).- Nav-piöne pužčice na sliki 3 predstavljajo tok podatkov iz enega končnega sistema v drugega. Vodoravne puićlco predstavljajo interakcije ali protokole, potrebne v vsakem od nivojev sa iBvajanJe funkcij transporta podatkov. Po terminologiji OSI ao to funkcije in protokoli, definirani z OSI standardi sa flalćnl, linijski, mrežni in transportni nivo, ki nudijo zgornjim nivojem servis transporta podatkov.
Preden aplikacijski procesi lahko uporabljajo podatke. Jih morajo sprejeti in praviino interpretirati. V vsakem raCunalniSkem sistemu lahko obstaja veliko aplikacijskih procesov. Zato se morata aplikacijska procesa, ki želita izmenjavati podatke, najprej dogovoriti za komunikacijo in določiti pravila, po katerih bo potekala. Ker vsak proces po svoje predstavlja informacije, se morata dogovoriti tudi z» enotno predstavitev inforraa-cij pri izmenjavi.
Za to potrebne funkcije in protokoli omogoča-
3 .
Foveao'vanJ o — nivo
Nivo seje nudi funkcijo za poveaavo dveh aplikacijskih procesov v odnos logično komunikacije in funkcije za organizacijo In sinhronizacijo njunega dialoga. Pri tem se uporabljajo mohanlzmi za vzpostavitev, vzdr-ievanje in razpustitev povesav na nivoju seje. Povezava na nivoju seje Je dogovor med dvetna aplikacijskima procesotoa o vkljutltvl v kontroliran dialog za izmenjavo podatkov. Aplikacijski procesi torej izmenjujejo podatke 3 pomočjo povezav na nivoju seje. S pomočjo mehanizmov povezav na nivoju seje lahko aplikacijski procesi požiljajo podatke, ki jih aprejemivl sistem dostavi naslovljenemu aplikacijskemu procesu.
Preden začneta dva aplikacijska procesa izmenjavati podatke, morata vzpostaviti povezavo na nivoju seje. Vsak aplikacijski proces lahko sam sproži pobudo za vzpostavitev povezave na nivoju seje, lahko pa tudi sprejme zahtevo za vzpostavitev povezave od drugega aplikacijskega procesa. Ko Je to opravljeno. Je odnos med aplikacijskima procesoma vzpostavljen in procesa lahko začneta izmenjavati, podatke.
Povezavo na nivoju seje lahko obravnavamo kot povezavo med dvema aplikacijskima procesoma preko nivojev soje njunih končnih sistemov. Vendar ne smemo pozabiti, da Je povezava na nivoju seje odvisna od povezav, vzpostavljenih na nižjih nivojih za Izvajanje funkcij nivoja seje. Te povezave so potrebno za prenos podatkov In protokolnlh InforcaciJ med končnima sistemoma. Privzeli bomo, da te povezave obstajajo, to Je da Jih spodnji nivoji vzpostavljajo In raapuSčaJo v skladu 3 potrebami povezav in funkcij na nivoju seje.
Povezava na nivoju seje se lahko raztegne tudi čer več povezav na spodnjih nivojih. To lahko primerjamo z dvema äahiatoma, ki igrata partijo preko telefona. Podobno kot povezava na nivoju seje, partija teče, dokler se eden od žahistov ne odloči za konec. Dejanska povezava preko telefona pa se lahko medtem večkrat vzpostavi in spet razpusti, saj Jo potrebna samo, kadar eden od iahlatov ieli drugemu sporočiti naslednjo potezo.
Med vzpostavljanjem povezave se dva aplikacijska procesa dogovorita sa pravila dialoga, ki se bodo uporabljala pri komanikaclji med njima. Pomen pojma dialoga na nivoju seje Je analogen kot pri komunikaciji med ljudmi. Prvi tip dialoga označuje pretok informacije samo v eni smeri, na primer od predavatelja k poslušalcem. Pri drugom tipu dialoga teče informacija v obeh smereh, vendar hratl samo v eni smeri, ko na primer predavatelj prekinja predavanje, da lahko poslučalcl postavljajo vpraSanJa. V tem primeru si predavatelj in poslušalci izmenjujejo pravico govora. Primer tretjega tipa Je komunikacija dveh ljudi, ki kontrolirata časovno kritičen proces (na primer pristajanje letala), zato morata nujno imeti oba vos čas pravico do govorjenja. Bistveno Jo, da potreben tip dialoga izhaja iz narave aktivnosti.
Nivo seje nudi Številne tipe dialoga za različne tipe aplikacijskih procesov. Keka-terl aplikacijski procesi (na primer tisti s področja kontrole časovno kritičnih procesov)
odprtimi siatemi in Jih ISO rasvija kot skupna orodja za katerikoli tip aplikacij. Nekatere standarde aplikacijskega nivoja ra-svljajo druga standardlsaciJaka telesa za tipe aplikacij splo&nega pomena (npr. za delo s sporočili in lamenjavo dokumentov), medtem ko se ostali rasvijajo sa uporabo v posebnih déJavnostih (npr, v bančništvu, trgovini, letalskem prometu, knjižnici itd.)
Wr(
Uleti tiHtM
l«tii*l littia
iiiiintitnitiiMiinii
J L
Hprt knl*i littM
r«4it ti
llt||111ll1l1l111|H1TI1
		. «Sf. JdMI- -----T					
ll«l t lin]	IrM^t	-				— 1- Trtmftrt	
	Krilc		r	|-Hr.i. .		—»r— Hrti)	
■ in}	Mittaiiiliniie			fritiiLiiiijii		ridttkdmlinili	
lltil	ritiUi			1 FltU*i 1		riiitbl	
i-- \ y-1 XJiiipi «tdij/							
Slika 3 rrifloi podfttkn
Nekateri standardi aplikacijskega nivoja podpirajo splo&ne tipe aplikacij, ki so Egod<>-vlnsko ' povezane z uporabo rafunalniikih sistemov. To BO standardi za prenos datotek med porazdeljenimi siatemi In delo z njihovo vsebino (prenos datotek, dostop do datotek In upravljanje z datotekami), dostop do ra£unal-niäklh sistemov.preko terminalov (servis vlr-tualnega terminala) in za izvajanje poslov v porazdeljenih sistemih (prenos in manipulacija poslov). Te standarde razvija ISO sa podporo množice rasnih tipov aplikacij, pri katerih so potrebne te splo&ne procesne funkcijo.
S .
s od e 1 ovsLnCL
'eri J.

Standardi aplikacijskega nivoja definirajo procedure In tipe Informacij. potrebne za sodelovanje ned porazdeljenimi aplikacijskimi procesi. DeflaiciJa tipov informacij, kl se uporabljajo v danem aplikacijskem proceau. Je abstraktna sintaksa podatkov. Za sodelovanje aplikacijskih procesov Je potrebno, da uporabljajo isto abstraktno sintakso.
Podatkovni elementi, definirani s standardi aplikacijskega nivoja, so abstraktne definicije informacij, ki ae bodo prenaàale in morajo biti nekako zakodlrane. Na41n kodiranja v posameznem računalniškem sistemu Je odvisen od konvencij, ki ae uporabljajo v tem sistemu, ter od prevajalnikov, zbirnikov itd. , ki so bili uporabljeni za kreiranje aplikacijskega procesa. Vse konvencije za predstavitev informacij v računalniškem sistemu skupaj imenujemo konkretna sintaksa informacij. .Pri sodelovanju heterogenih sistemov Je običajno, da so Informacije lokalno predstavljene v vsakem sistemu po drugačnih konvencijah. Konkretno sintakso za predstavitev informacij v posaiDeznem sistemu imenujemo njegova lokalna sintaksa.
Pri sodelovanju morajo aplikacijski procesi kodirati informacije po Istih pravilih. Čeprav ni nujno, da sodelujoča . procesa uporabljata enako lokalno sintakso, ae morata dogovoriti sa konkretno sintakso za kodiranje Informacij, ki se bodo prenašale med njima.
Konkretno sintakso, ki se uporablja pri prenosu, imenujemo prenosna sintaksa. Prenosna sintaksa je lahko enaka lokalni sintaksi enega ali obeh sistemov, kar pa nI nujno. Bistveno Je, da se sistema dogovorita za določeno prenosno sintakso in sta sposobna prevajati med svojo lokalno sintakso In Izbrano prenosno sintakso.
Mehanizmi predstavitvenega nivoja omogočajo aplikacijskim procesom definicijo in izbor prenosne - sintakse za'-njihovo komunikacijo. Prenosno sintakso izberejo z dogovarjanjem med sistemoma funkcije predstavitvenega nivoja v imenu aplikacijskih procesov iz množice prenosnih sintaks, ki Jih podpirata oba sistema. Hed komunikacijo se funkcije predstavitvenega nivoja lahko dogovorijo za spremembo prenosne sintakse in izberejo novo v skladu s standardi aplikacijskega nivoja. To se na primer zgodi v primeru, ko dva aplikacijska procesa komunicirata po standardih zs delo z datotekami In se dogovorita za določeno prenosno sintakso za izmenjavo znakovnih nizov. V nekem trenutku"se pojavi potreba po prenosu datoteke, ki vsebuje cela fitevila, V tej točki lahko funkcije predstavitvenega nivoja Izberejo novo prenosno sintakso - aa predstavitev celih Števil,
Prenosna sintaksa Je rezultat uporabe množice pravil kodiranja na abstraktni sintaksi, ki jo uporabljata sodelujoča aplikacijska procesa. Kot rezultat standardov aplikacijskega nivoja se definirajo vedno nove kategorije informacij, ki predstavljajo Informacije v raznih tipih aplikacij. Vsak od teh standardov definira za predstavitev informacij podatkovne tipe različne kompleksnosti. Ce naj uporabljena prenosna sintaksa odraža semantike komunikacije, mora biti sposobna predstaviti vse potrebne podatkovne tipe in pripadajoče podatkovne vrednosti. Bilo bi na primer nemogoče predstaviti vse Informacije (t.J. almbole) 6-bitno kodirane znakovne množice, če Je bilo za prenos Izbrano 5-bltno kodiranje.
Vsak aplikacijski proces mora nujno imeti pripravljeno eno ali več prenosnih sintaks za komunikacijo z drugimi sistemi. Zato mora biti vsakemu standardu aplikacijskega nivoja pridružena prenosna sintaksa, ki je dovolj močna za Izražanje kompletne semantike standarda. Ena od možnih režitev bi bila seveda definicija ene same prenosne sintakse za podatkovne elemente določenega standarda aplikacijskega nivoja, morda po kakSnem obstoječem standardu za kodiranje podatkovnih elementov. Tako bi lahko nek standard predpisoval, da se morajo znakovni nizi vedno prena-žati kodirani po ISO 646. To bi pomenilo, da bi dva sistema, ki uporabljata ta standard, nujno morala prevajati svojo lokalno sintakso v ISO 646, tudi če bi bili njuni lokalni sintaksi enaki, kar ne bi bilo gospodarno.
Zato Je bil pri zgornjih nivojih izbran pristop specifikacije potreb po prenosu Informacij na aplikacijskem nivoju v obliki spložne abstraktne sintakse, brez specifikacije specifične prenosne sintakse. Primerno prenosno sintakso mora nuditi predstavitveni nivo.
Definicija in registracija prenosnih sintaks Je posebna področje standardizacije. Vsaka prenosna sintaksa se definira In registrira glede na abstraktno sintakso, ki Jo podpira. Tako lahko predstavitveni nivo Izbere prenosno sintakse, ki so najprimernejše za določen standard aplikacijskega nivoja. Britanski nacionalni računski center (National Computing Center) Je prevzel dolžnosti telesa za
es
registracijo kod oziroma apeclfiòne apllltaclje.
kodne sintakse za
Standardi predstavitvenega nivoja definirajo mehanizroe za dogovarjanje o množici sprejemljivih predstavitvenih sintaks ter izbor in spremembo izbora ene izcnod njih. Pri tem predpostavljajo. da so obravnavane sintakse uradno registrirane, vendar ne izključujejo možnosti dinamične definicije posebnih prenosnih sintaks a pomočjo jeaika za definicijo podatkov,
6.	SltloEJno misli
ISO In druge standardi zacijske organizacije Intenzivno delajo na standardizaciji agornjih nivojev. Z napredovanjem tega delo postaja vedno bolj viden njegov odnos do drugih vidikov procesiranja informacij. Čeprav področje OSI standardov pokriva samo komunika-oijake , vidike povezovanja odprtih sistemov.
morajo ti standardi upoštevati tudi SirSe zahteve.
Zato Je nujno, da načrtovalci in strokovnjaki, ki delajo na sorodnih področjih standardizacije, priznajo uporabnost komunikacijskih servisov, ki se standardizirajo v okviru OSI, strokovnjaki za OSI pa morajo pri razvoju novih servisov upoštevati potrebe aplikacij, ki Jih bodo uporabljale.
Uporabljeni so bili delovni dokumenti in standardi ISO za področje OSI.