PUBLICATION ON AUTOMATIC PROGRAMING AT THE COMPUTING CENTER OF THE SOVIET ACADEMY OF SCIENCES
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP80T00246A060200410001-7
Release Decision:
RIPPUB
Original Classification:
S
Document Page Count:
30
Document Creation Date:
December 22, 2016
Document Release Date:
January 12, 2012
Sequence Number:
1
Case Number:
Publication Date:
September 26, 1961
Content Type:
REPORT
File:
Attachment | Size |
---|---|
![]() | 849.03 KB |
Body:
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
This material contains information affecting the National Defense of the United States within the meaning of the Espionage Laws, Title
18, U.S.C. Secs. 793 and 794, the transmission or revelation of which in any manner to an unauthorized person is prohibited by law.
CENTRAL INTELLIGENCE AGENCY
COUNTRY USSR
REPORT
SUBJECT Publication on Automatic Programing DATE DISTR. cZ G September 1961
at the Computing Center of th
S
e
oviet
Academy of Sciences m
DATE OF
INFO.
PLACE &
DATE ACQ.
THIS IS
a 28-page; English.-language c.oemRwi' ;n; by A.P. Yershovaentitled
e works of the Computing Center of the Academy of
in the Field of Automatic o ralnin
The publication contains
scuss on of compu.?er programing, programing for the STRELA-3
and problems arlsinrr in / ,
STATE I X ARMY NAVY
(Note: Washington distribution indicated by "X"; Field distribution by ".#".)
3
50X1-HUM'-
50X1-HUM
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Academy of Sciences
of the USSR
Computing Centre
FOR OFFICIAL USE ONLY
THE WORKS OF TAE COMPUTING CENTRE
OF THE ACADEMY OF SCIENCES OF
::7 THE FIELD OF AUTOMATIC PROGRAMM11r ;x
by
A.P. ERSHOV
MOSCOW
1 9 5 8
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR G~,` r 1CIrIC USE
011/ ,,
THE WORKS OF THE COMPUTING CENTRE
OF THE ACADEYII OF SCIENCES OF THE USSR
IN TEE FIELD OF AUTOMATIC PROGRAMMING
by Ershov A.p.,
chief of the theoretical programming department
of the Computing Centre of the AS of the USSR
1. Preliminary researches
The real beginning of the automatic programs=,ing develop-
ment in the Computing Centre (like as in the Soviet Union)
is the year 1954. Just this year the first preliminary works
appeared which contain the main ideas and their realization
about which I am going to say at first.
The main necessity, which x'ised in programming from the
very beginning, is the separation of the programming process
into two main stages. On the first stage the programmer
writes a detailed, but sufficiently lockable, computing
plan of problem solving. On the second stage the direct
programming (coding in your terminology) is carried out,
which presents itself the writing of this detailed computing
plan on the computer's language.
Thus, in the course of programming, the necessity of
construction of the intermediate language rises out. This
language is the language, by means of which the results of
he first stage of programming (the detailed computing plan
of problem solving) are written down.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
At the beginning in the Soviet Union as well as in any
other countries such an intermediate language was the flow-
charts language, which was used still by John von Neumann
and Goldstine.
But in the years 1952-53 prof.A.A.Ljapunov developed a
new programming method, which was first described on his
lectures in the Moscow University. This method had a great
influence on the development of the programming in the USSR,
is perfectly acknowledged nowadays and is known as the
operator programming method.
The main conception of the operator method is the
consideration of the program as a complex operator which
acts with initial aatas and consists of elementary operator*
of various types. The main types of elementary operators,
acting with the initial datas, are arithmetical operators
which fulfill the direct data'stransformation and logical
operators which determine the next direction of computations.
The conception of the depending of the operator on some
integer parameter (for instance the depending of the adress
of the indexed number on the values of these indexes) was
also very important.
Besides this, the control operators were considered.
These operators do not act with the initial datas but deal
with the program itself. I shall mention at first the re-
adressing operators which exchange the depending on the
parameter adresses whilE the parameter's value is exchanging.
The restoration operator establishes the depending on the
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFiCiAL USE ONLY
parameter operator to its initial view according to the
:r;itial value of the parameter.
The detailed computing plan's record written as the
sequence of the operators of various types is called the
operator scheme of program (or the logical scheme of
program).
This conception was briefly described at first in the
book /2/ (pp.193-206).
,Ljapunov's operator method was very important for the
development of the automatic programming because of the
four main reasons:
1. For the first time the program's components (element-
ary operators) were formally classified according to their
functional appointment, independable on the sense of the
problem programmed. This permited to formalize the logical
scheme program's language easily and make the language
universal for writing problems to be solved.
2. The language of logical schemes of programs was easy
to algebraize. The latter was very important for the direct
input of the program's logical scheme into the computer.
3. The logical scheme of program as an intermediate
result of the programmer's work, successfully separates the
programming process. The composition of the logical scheme
accords to the "intellectual" work of the programmer while
the most labour and mechanical work falls on the stage of
he transition from the logical scheme to the object program.
Thus, the stage of the programming process, which is the
3
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
ONLY
easiest to automatize, was separated. i:i i :'. : ?~ LJ~
4. The partition of the whole problem on the sequence
of separate operators permits the programmer to divide the
programming process on programming of separate elementary
operators. This also makes the coding more easy.
Still in 1953 prof.Ljapunov supposed that some parts of
the second stage of programming could be automatized.
The first works in the field of automatic programming
in the Soviet Union were carried out in the Mathematical
Institute of the AS of the USSR and in the Computing Centre
of the AS of the USSR, which at that time was a division of
the Institute of Exact Mechanics and Computing Technique of
the AS of the USSR.
In summer 1954 in the Computing Center y.M.gurocn
composed a program for the BESM computer, which permited to
fulfill the symbolic programming. L.M.Korolev composed the
first version of the program for BE, which carried out the
programming of arithmetical formulas.
G.S.Bagrinovskaja from the Mathematical Institute
described a project of the program, fulfilling the programming
of readressing operators. In summer 1954 E.Z.Lubimaky and
S.S.Saminin constructed the programming program for the
STRELA computer (Pp-1), which performed the programming of
readressing, restoring and logical operators.
All these works were only experimental but it was a
ziecessary stage on the way to the design of big programming
programs.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
f iG(AL USE ONLY
2. The first programming programs
The first project of the programming Program (PP-2),
which fully automatize the transition from the logical
scheme to the object program, was described by E.Z.Lubimsky
and S.S.Kaminin on the prof.Ljapunov,s seminar. in October
1954 /3/.
Source information for this PP consists of the logical
scheme of the program and the table of the storage distribut-
ion. The logical scheme may include arithmetical and logical
operators, readressing and restoring operators and so called
non-standard operators, which presents some peaces of the
object program written in commands with symbolic adresses.
In February 1955 a group of the Mathematical Institute
stuff-workers constructed a working pp for the,STRELA computer
on the basis of this project. This PP was described in a
series of articles /4/.
In December 1954 a project for another pp was carried
out. Some versions of the PP were developed on the basis of
this project; the final one was constructed to March 1956.
This PP is described in my book /5/.
I want to point out the main differences of the Pp for
BESM from the PP-2 for STRELA.
a) In the PP for BESM the hierarchy of arithmetical
operations when programming arithmetical formulas is taken
into consideration.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
b) In PP for BESH the logical operators used are of
another type than those in the PP-2.
c) There are no readressing and restoring operators
in the source information for the PP for BESM, but some
special notations, pointing out the necessity of the
.Syclic repetition of some operator's group, take place.
d) In the PP for BESM the storage is distributed
automatically.
The first public communications about the first
programming programs were made by the authors in their
,.reports on the All-Union Conference *The ways of the
development of the-Soviet mathematical computing
machinery" held in Moscow on the 12-17 of March, 1956.
3.The programming program for the STRELA-3 computer
The main result of the Computing Centre's works in
the field of automatic programming is nowadays the
programming program for the STREI+A-3 computer. This
compute! is installed in the Computing Centre.
In autumn 1956 the wbrk on the construction of the
programming program for the STREI,A-3 (PPS) began. It has
been finished in autumn 1957 and performed by six stuff-
workers of the theoretical programming department.
When the requirements for the source information were
developed we were intended to realize the following aims:
a) The representation of the source information has
to be close to the mathematical formulation of the problem.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
FOR OFFICIAL USE ONLY
b) The size of the auxiliary and technical work, not
I"nnected with the mathematical formulation, has to be
=educed to minimum.
a) The source information must give the full inform-
ation about the structure of the object program.
d) The source information has to be maximally compact-
i-le and lockable.
The source information prepared by the programmer
?onsists of five parts:
1. The program's scheme (the main part of information)
2. The carried away operators
3. Information about variables
4. Information about storage blocks
5. Blocks.
Nome parts of the source information may be absent.
A little about terminology. A variable is a letter
symbol which is contained in the source information. It
may denote some mathematical variable or constant or, in
Jeneral, any machine word. A storage block means any group
locations which has to be placed in the storage one
a`ter another.
First about the 3-th and 5-th parts. In the 3-th part
'1l the variables, which require some additional inform-
!tion, are written down. Such an additional information
lay be:
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
FOR OFFICIAL USE ONLY
Name of the operator
a) A constant, if the variable represents a constant.
Four types of constants may exist; decimal flow-point
constant, binary flow-point constant, command with sym-
bolic adresses and command with real adresses.
b) An adress, if you wah to prescribe the location
of the variable in the storage.
The 5-th part inoludes any datas placed into blocks
and introduced into the storage by the object program
itself.
The yroltram's scheme. In the program's scheme all the
operators of the object
program are written down one after
another. The order of placing of the operators in the
scheme corresponds to their placing in the object program.
The scheme may include operators of the folowing types:
symbol denoting
the operator's type
1. Arithmetical A
2. logical J1
3. Restoring B
4. Non-standard H
5. Readressing
6. Doubl-counting AC
As a rule, every operator is represented by a letter
denoting the operator's type, after which (in parenthesis)
the information about the operator follows_ only for the
arithmetical operators the letter symbol and parenthesis
are omitted.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
FOR OFFICIAL USE ONLY
F (xi,...,xn) -
scheme into the second part of the source information. All
the non-standard operators and some complicated arithmetical
ones are carried away into the second part. In this case on
the place of the operator carried away only a letter-denot-
ing the type with the number is left in the scheme.
The main cause, which makes the source information look-
able and obvious, is the successful solution of the problem
of the recording of arithmetical formulas and the use of the
obvious symbolism.
Arithmetical operators. Arithmetical operator fulfills
the evaluation by the help of the sequence of formulas as a
It is possible to require in the program's scheme the
syclic repetition of some sequence of operators. For this
purpose such a sequence is put into a figure brackets: f
is an opening bracket of the sycle (a loop in your termi(no-
logy) and 1 is a closing bracket of the sycle.
Beside this, special symbols may appear in the source
information; they denote the command, which transmit an
information between computer's parts, and transfers.
Every operator may posers a number. The number is put
by a letter denoting the operator's type as a subscript.
Every operator can be carried away from the program's
a_nere F is a superposition of operations from some fixed
list of operations and xl,...9xn are variables and con-
stants. In particular, this list contains all the operations
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
of the STRELA-3. Y denotea'the result of the evaluating of th
formula. The constants in formulas are represented-either
by letters or by numbers. The variables may have any number
of letter subscripts. In this case every subscript must be
a parameter of some loop.
I shall give (with some abbreviations) the list of
operations used in the arithmetical operators.
The list of the operations
operations with two arguments
addition
substraction
addition by modulo 2 (digit by digit) +
logical addition (digit by digit)
v
multiplication
x or
logical multiplication (digit by digit) A
operations with one argument
integer part
fractional part
absolut value
signum
sinus
cosinss
aresinus
arccosinuo
arctangent
exponent
E
D
Mod
sin
cos
aresir.
arccos
arctg
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFEiC1A . USE ONLY
Name
natural logarithm
binary to decimal conversion
decimal to binary conversion
logical negation (digit by digit)
1 1 =p
x-1' x2, a3' x7' x-7, x2~ (1--pliiS1i, Ocq, 7)
The usual hierarchy of operations useful in mathematics
is used for the determination of the order of fulfilling of
the operations. There is only one difference between the
marked and non-marked multiplication ( x and
which is
obvious from th, 'ollowing example:
tg C05_2- X2' + s in (a + b)(a - b) + sin(a+b)x(a-b) -
-Mo(f y sign (x+y) tg at.
u
In this example the terms, which are arguments of the
corresponding operations, are underlined. Besides this,
the following record is allowed: sin m x, which denotes
sin x)r.
Our experience shows that all the transformations of
formulas, which are carried out by the preparation of
source information, are thus: the omitting of radixes and
lines, denoting the division, and replacing them by ap-
propriate powers.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
Logical operators. One logical operator realizes the
verifying of one of five standard logical relations. The
standard logical relations are following:
a < b
a6 b
tat < lbl
lad-jr. IbI
a - b ,
where a and b are variables or constants.
In the program's scheme logical operators are represent-
ed as
N1
Jl (a-b I )
where N1 and N2 are operators' numbers, a - b is one
of the five logical relations. This operator acts in such a
way: if the relation a - b is true then we transfer to the
operator N1, otherwise we transfer to the operator N2. If
one of the operators' numbers is omitted, this means, that
we transfer to the operator next after the logical operator.
Bon-standard operators. A non-standard operator is an
any part of object program, written in commands with symbolic
adresses. There are many rules of writing down the non-
standard operators but they are of no interest, and I shall
omit them.
As it was mentioned above, the non-standard operator is
always written down in the second part of the source inform-
ation (carried-away operators). In the program's scheme the
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR O F.CfAL .USE ONLY
non-standard operator is represented by a letter H with
the number. In the second part a "head" is placed before the
non-standard operators, which has a form
where N is the non-standard operator's number, n is a
number of words the operatorcontains.
The loops. A loop realizes the'repeated fulfillment of
some sequence of operators for all given values of the para-
meter of the loop.
In the program's scheme the loops are represented in
such a form
{A}
where { and } are the opening and closing brackets of
the loop, A is the sequence of the operators repeated. The
loop's parameter i and its initial value I. , if it is not
equal to zero, are placed under the opening bracket
{
(if iK . 0)
If the number of repetitions of the loop is determined
by the final value of the parameter iK then the latter is
placed over the opening bracket
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
W: Ur" iCIAL USE ONLY
In such a case the loop will be repeated from the value iK
up to the value 1K includingly by the step of the size 1
0.Cc 1K).
iK and J. are either variables or non-negative integer
constants. If'iM and 1K are variables, then it is supposed
that they are-evaluated by the object program before running
of the loop as non-negative integer normalized numbers.
It is also possible to set the number of the loop's
repetitions by means of one of five logical relations a - b.
The readressing operators. A readressing operator is
recording in the program's scheme as
R (N, n, h),
where N is an operator's number; n is either a number
of a command of a non-standard operator or a variable; h
is either a variable or an integer constant.
The readressing operator is fulfilled thus:
a) If n is a number of a command of the non-standard
operator N , then the?readressing constant h is added to
the command.
b) If n is a variable and h is a constant, then h
is added to all the adresses of the variable n in the
operator N .
c) If n is a variable and h is also a variable, then
it is supposed that the value of the h needed is evaluated
in the third adress of the location for h as an integer
number.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
FOR OFFICIAL USE ONLY
The restore operators. A restoring operator is recor-
ded in the program's scheme as
B(N)
where N is an operator's number.
If there are some variable commands in the operator N
which are transformed by readressing operators, the restoring
operator will transfer the initial values of these variable
commands into their places in the operator N
The double-counting operator. While composing the program
with the help of the PPS we may prescribe the control of
the computation by means of double computations of separate
peaces of the program. Therefore the double-counting operat-
ors' symbols are arranged in the program's scheme. The
computer will compute every peace of the program between two
such symbols (following one after another) and accumulate
all the contents of the internal storage. If the sums are
equal, then the whole content of the storage is recorded
on the magnetic tape, and the computation is carried on
further. If the control cams are not equal,the computer
stops and, after a new start, repeats the computation once
more comparing after that three sums already.
Special symbols for transfer and transmission of infor-
mation between computer's parts. The following notation,
which denotes the transmission of information between the
different computer's parts, is used:
A n 0. 2
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
where A is the symbolic adrese of the beginning of the
place, from which they are going to transfer the informatio
B is the symbolic adress of the beginning of the place,
where they are going to transfer, and n is a number of
words in the information transferred.
The unconditional transfer is denoted by
C '
N
where N is a number of an operator to which we shall
transfer.
The conditional transfer is denoted by
N.
NZ
If the signal CO for the proceeding command is equal
to 1 , we shall transfer to the operator N. Otherwise
we shall transfer to the operator N.
For transfer to subroutines the following notation is
used:
NJ
which means; to fulfill the 'subroutine beginning from the
operator Ni and up to the operator
NZ, exclusively.
Information about stor a blocks. This information, at
first, shows how many blocks there have to be in the storage
and how long they should be. If it is necessary, it is
possible to point out*the position of some blocks in the
storage. Such an information about every block has the
following form
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
FOR OFFICIAL USE ONLY
Mk ) A words, Va.. .
where k is the block's number, A is its length and
V is the adress of the first location of the block (if
it is given).
Then the information about the variables concerned to
this block is written down.
If the adress of a variable a or aLi... depends on
parameters 1,j ,... it is recorded in such a form
a(or aij...)=H(or K)+e+h1i+h2j+...
where 14 denotes the beginning of the block (its first
location) and K denotes the end of the block (its last
location). Number A La a "shift" in our terminology, h1,
ht,... are the steps of readresaing according to parameters
i,j,... . A and h may be either integer constants or
variables.
If the adress of the variable a from the block does
not depend on single parameter then the information about
ita place in the block is recorded by
a - H(orK ) +a.
If some variable appears in the program's scheme with
different sets of equal number of subscripts then the depend-
uses of the adress of the variable on the. subscripts is
aentioned in the information only once. For example, if the
variables aisk?dkl1'aksL appear in the program's scheme,
then in the information about blocks only one row is written
down
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246A060200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
aiik - H(or K) + + h11 + h2J + h3 .
That is all about source information.
I shall briefly describe the structure of PPS. The PPS
as a whole consists of 30 separate blocks recorded on the
magnetic tape. Every block works only once during the
programming program's running. Tue information transformed
is placed in the internal storage. Every block has the
length of about 240 locations. A part of the storage (about
130 locations) is intended for some working materials, which
the PPS works out and overgives from one block to the other.
Thus the length of the simultaneously transformed informat-
ion may arise up to 1500 words. The blocks of the PPS change
one after another automatically. The PPS punches on punch-
cards the ready object program and some additional inform-
ation. Besides this, it is possible (if you wish) to punch
some intermediate information. ,
PPS is able to discovery automatically many formal mis-
takes in the source information.
In conclusion of this part let us-consider an example of
preparation of the source information.
Suppose we have to integrate a parabolic partial
differential equation
at = 0,75 x(l-x)(t x+z) 2
aX
z(x,0) = 0; z(O,t) = 0;
from t-0 to t=T, M=h-1
Z(1,t) = t
(h is the step by x, r is the step by t).
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
The evaluation formulas are following
ZOn?f
Zm,n?i`Zm n+ T.0,7 5x'(1-0 Zx+zmn) Zm?9,n 2Zmnt2R,_i n
hz
m?1,2,...,M-1
Z'M n?i=(n+1)T
n-0,1,2,...
Let us'pick out one block of the storage, containing
1000 number for storing all the values of z for each layer
of t . The new z. we shall place on the places of the old
ones with delay for one step, because the old values of Z
will be needed for the evaluation of the new ones. Every time
three points Zm ,z, and Zr*,l will be used, which
correspond to Zm,n ' Zm-1,n , Zm+1,n '
The source information is thus:
1. The program's scheme
M M
-M --r AC(5) { 0W+z,? l tn'r.*t J,TI(m-OC)
m'O n m Ni
if, N
0-+r[ fl (m-MC) (n+1)r zm[[ J MM-i-#x
NZ N1 N3 N4 N3
Zm +0,75(x(1-x)(tZx t zm)) (zt-2 Zmt z;r,)M2.
tu~ (n,u)~v (m,v)~w ~pcl,w)iz, punch z
n -n. m mo 1 10
would be, of course, much more profitable. That is because,
according to this scheme, only that part of the function
F is being evaluated in every loop, which is essentially
depending on the parameter of this loop. The evaluation of
the left part of F is carried out into the external loops,
which repeat more seldom. Therefore the program as a whole
works faster.
In this case it is also probably possible to find an
algorithm, which would perform the transformation of the
program's scheme mentioned above.
2. The increasment of the velocity of the PP's work. The
new PP obtain the more and more opportunities in order to the
quality of the object programs by the expansion of the
complication of the programming algorithms. But the increased
complication and the size of PP have their shadow side,
which turns out in the increasment of the programming time.
When the running of Pp offers much time, it reduces the
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
effect of using of the PP (in particular, when the computer
works unstable). Therefore I would like to stress the fact,
that it is necessary while constructing new algorithms of
the programming, to aspire in every possible way the in-
creasing of the velocity of the algorithm's work. One my
work /6/ testifies some possibilities in this direction.
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7
FOR OFFICIAL USE ONLY
R e f e r e n c e
/1/ A.A.Ljapunov. "0 log1 eskikh ekhemakh program" (On
logical schemes of programs). Problemy kiberne-
tiki, Vypusk 1 (Cybernetics probleme, issue 1).
;.:oskva, Fiziko-matemati6eskoje izdatalstvo, 1958.
/2/ A.I.Kitov. "Elektronnye cifrovye mashyny" (Electronic
digital computers), Moskva, Sovetskoe radio,1956.
/3/ S.S.Kaminin,-E.Z.Ljubimskiy, M.R.Sht a-Hura. "Avtoma-
tizacija programmirovanija c pomog6ju progr i-
rujuc6ej programmy" (Automatization of programming
by means of a programming program). Problemy kiber-
netiki, Vypusk 1.
/4/ 2.Z.Ljubimskiy. "Arifineti6eskij Blok v PP-2" (The
arithmatical block in the PP-2).
S.S.Kaminin. "Blok pereadresacii v programme PP-2"
(The readressing block in the PP-2 program).
3.S.Lukhovickaja. "Blok obrabotki logi6eskikh usloviy
v PP-27 (The. block of the treatment of logical
conditions in the PP-2).
V.S.Shtarkman. "Blok ekonomii rabo6ikh ja6eek v PP-2"
(The block of the economy of temporary locations
in Pp-2).
Problemy kibernetiki, Vypusk 1.
/5/ A.l.-rshov. "ProgrammirujuMaja programme dlja bystro-
dejstvujuX6ej elektronnoj s6tstnoj mashyny" (The
programming program for.the BESM).
:oskva, izdatelstvo Akademii Nadk, 1958.
/6/ A.P.Ershov. "0 rogramrirovanii arifineti6eskikh opera-
torov" (On programming of arithmetical operators),
:ok-lady Akademii Naak SSSR, v.118, N93 (1958),
42?-430.
::.ere is a translation: Communications of the
C::, v.1. :;48 (1958).
Declassified in Part - Sanitized Copy Approved for Release 2012/01/12 : CIA-RDP80T00246AO60200410001-7