Digital PDFs

XX-98A1C-01

January 1965

62 pages

Original

2.5MB

Document:	8-2-S FORTRAN Programming Manual 196501
Order Number:	XX-98A1C-01
Revision:
Pages:	62
Original Filename:	http://bitsavers.org/pdf/dec/pdp8/fortran/8-2-S_FORTRAN_Programming_Manual_196501.pdf

OCR Text

DIGITAL

DIGITAL EQUIPMENT CORPORATION

•

8-2-S

MAYNARD, MASSACHUSETTS

PDP-8

FORTRAN
PROGRAMMING
MANUAL

PREFACE
The program discussed in this manual, though written for use on the Programmed Data Processor-8 computer, can also be used without change on
Digital 1s Programmed Data Processor-5. This compatibility between the
libraries of the two computers results in three major advantages:

l . The PDP-8 comes to the user complete with an extensive selection of system programs and routines making the ful I data processing capability of the new computer immediately available to
each user, eliminating many of the common initial programming
delays.
2. The PDP-8 programming system takes advantage of the many
man-years of field testing by PDP-5 users.
3. Each computer can take immediate advantage of the continuing program developments for the other.

iii

CONTENTS
Page

Section
INTRODUCTION

............................................

...................................
Statements
..........................................
Program Format
.......................................

THE FORTRAN LANGUAGE

Types of Statements
Comments

.............................................

Continuation
The Character Set

.......................................

........................................

.....................................
Arithmetic Expressions
.................................

FORTRAN ARITHMETIC

The Arithmetic Statements

.................................

...............
lntergers and Floating Point Numbers
.....................
Fixed-And Floating-Point Representation ....................
Types of Variables .......................................

NUMBER REPRESENTATION AND VARIABLE TYPES

SUBSCRIPTED VARIABLES

.....................................

...............................................
Subscripts ..............................................

Arrays

......................................
Program Termination .....................................
Branches and Loops .......................................

PROGRAM CONTROL

........................................
Available Devices .......................................

INPUT AND OUTPUT

3
3
4
4

5
5
6
7

7
12
15
15
16
17

18
18
18

22
22

23
31
31

CONTENTS (continued)
Section

Page
Input-Output Statements

. . . . . . . . . . . . . . . . . ••. . . . . . . . . . . •. . .

Format Specifications Statement . . • . . • . . . . . • • . • . . . . • . . . . . . . .

Other Format Control Elements . • . . . • . . . . • . . . . . . • . . • . • . . . . . .

DIAGNOSTICS ........................•.•.......•....•......

Appendix

Compile Time Diagnostics

. . . . . •• •. . •. . . • . . . . •. . . . . . . . . . . . .

Diagnostics . . . . . . . . . . . . . . . • . . • . . . • . . . . • . . . . . . . . . • • . . • • . . .

Object System Diagnostic . • • • • . . . . • . • . . • . • . . • . . . • • . • • • . • . • .

TAPE TELEPRINTER OPERATING PROCEDURES .•••••....•••...•..

OPERATING PROCEDURES FOR RIM AND
BIN PAPER TAPE LOADERS .................................. ..

Read-In-Mode Loader (RIM) . . . • • . • . • . • • • . • . . . . • . • . . • • . • • . . .

Binary Loader (Bl N) . • . • . . . . . . . • . • • • . . • • • . . . . . . . . • • . . . . . • . •

ASR-33 8-BIT CHARACTER SET ............................... .

OPERATING PROCEDURES .................................. ..

Compiler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Object Time System

• . •. •. . •. • •• . •••. • . • . . . . • . . . . •. •. . •• . .

FORMAT FOR COMPILER OUTPUT ........................... ..

Interpretive Code

. . . . . . . . •. . . •••. •••. . •. •. •. • •. . . . . . •. . •.

ILLUSTRATIONS
Figure

Page
A Fortran Program ••••••••.•••••••••••.••.••..•.•..•..•..••••••

Example of Comments

The Continued Statement

......................................
Number Representation ........................................

Indexing Statements •.•••••.••••••.••••••••

Schematic Representation of Program Branching ••••••••••••.••••••

Integer Summation

Use of IF Statement in Integer Summation Problem .•••••••.••••••••

IF Statement with Substatement Feature

..........................

Fibonacci Series

............................

Fibonacci Series Calculation Programmed As a DO Loop ••••••••••••

.............................................
DO Loops ...................................................

Program Branching in DO Loops ..•.••..•..•.•.••••.•.•..••..••..

15
16
17

Nested DO Loop

.......................................
A List Example ...............................................
Examples of Quote and Slash ...................................
Input-Output Statement

27
28

29
32
33

CHAPTER 1
INTRODUCTION

Using a digital computer to solve a problem generally involves the following series of steps:
a.

Determining the correct procedures to be used, including mathematical

formulas, the handling of data, the presentation of results, etc.
b. Arranging the procedures in the proper order.
c.

Determining the sequence of computer instructions that will perform the

operations specified.
d. Converting the sequence of instructions into binary notations in a physical medium capable of being entered into the computer for execution.
Much of the progress and development in programming has been made in discovering ways to
make the computer perform more of the steps Iisted, above, and leaving the programmer free
to concentrate more on the problem itself. At first, programmers entered instructions manually
from the computer console or prepared the binary program for direct input.

Later, a symbolic

notation was developed for a computer instruction set, and programs called assemblers were
written that could interpret a tape or card deck punched with this notation.

These assemblers

translate each symbolic instruction into a machine operation and assemble an executable program. Thus, an assembler can accept input from step c above.
An assembler requires that the programmer be familiar with the particular instruction set of the
computer being used. To solve the same on another computer would usually require complete
reprogramming.

To free the programmer from the need of learning a given computer's language before using
the machine to solve problems, compilers were developed which accept input more closely related to the problem and convert the input into an executable program. FORTRAN is such a
compiler.

It accepts input in the form of statements which resemble mathematical formulas

(hence its name, which stands for FORmula TRANslation) and compiles sequences of instructions
necessary to perform the procedures specified.
English words.

Non-mathematical operations are specified by

In terms of the steps given above, a FORTRAN source program is a product of

step b; the computer performs steps c and d.
A FORTRAN compiler can thus be written for any digital computer to convert a source program
into an executable program. Extensive reprogramming is made unnecessary, since the same
source program can be compiled on different machines with only minor changes.
The PDP-8 FORTRAN System consists of two subsystems: the compiler and the object time system.
The FORTRAN Compiler contains the instructions the computer requires to perform the clerical
work of translating the FORTRAN version of the problem statement into an object program in
the language of the object system.
When the compiler detects errors in statement format or usage, it prints out diagnostic messages
(see Appendix A). The programmer or operator may then take the appropriate corrective measures.
After compilation, the object time system is used to execute the program. This system contains
the interpreter, the arithmetic function subroutines, and the input-output packages. When
program execution is required, the object time system, object program, and the data it will
work with are loaded into the computer for solution of the problem.
This is a one-pass compiler, which means the source language tape must be read only once.
The compiler generates one tape which contains coding in a form that is executable under control of the object time system.

To use the system, it is only necessary to load the compiler. The compiler then processes the
source language tape and generates the object program tape.

This object program tape can

be run at any time simply by loading the object system, which, in turn, loads and executes
the object program.

CHAPTER 2
THE

FORTRAN LANGUAGE

STATEMENTS
Figure 1 is an example of a FORTRAN program, consisting of a title, the body of the program,
and the end statement.
The first line of the program is the title, which may be anything the programmer writes to identify the program. It is not incorporated into the final executable program.
The body of the program is a series of statements, each of which specifies a sequence of mathematical operations, controls the flow of the program, or performs other tasks related to the
proper working of the program.
The end statement must be physically the last statement of every FORTRAN program. Its function is to indicate to the compiler that nothing more connected with the preceding program is
to fol low.

5;
10;

30;
32;
33;
41;
200;
300;
400;

THIS PROGRAM CALCULATES FACTORIALS
TYPE 200
ACCEPT 300,X
FACT=Y=l.
IF (X) 5, 32, 30
IF(X-Y)41,32,33
TYPE 400, X, FACT
GO TO 10
FACT=FACT* (Y=Y+l .)
GO TO 30
PAUSE
GO TO 5
FORMAT(/, 11 PLEASE TYPE A POSITIVE NUMBER 11 , / )
FORMAT (E)
FORMAT (/, E, 11 FACTORIAL IS 11 , E)
END
Figure 1

A Fortran Program

PROGRAM FORMAT
Each line contains two fields: the first, which begins at the margin, is an identification field;
the second contains the statement proper (see Figure 1).
The identification field extends from the left-hand margin up to a semicolon character. This
field may be left blank, or it may contain one of the following types of identification:
1. The first digit of a statement number. This number, which may be any
positive integer from 1 to 2047 inclusive, identifies the statement on that
Iine for reference by other parts of the program. Statement numbers are
used for program control or to assist the programmer in identifying segments
of his program. Up to 40 statements can have statement numbers.
2. The letter C. This identifies the remainder of the line as a comment (see
Section Comments).
The statement field begins immediately after the semicolon and extends through the next carriage return. Although the continuation character(') allows a single statement to extend over
two or more Iines, no more than one statement can be written on one Iine.

TYPES OF STATEMENTS
FORTRAN statements are of several types with differing functions distinguished as fol lows:
1 . Arithmetic statements resemble algebraic formulas. They specify the
mathematical operations to be performed.
2. Program control statements direct the flow of the program.
3. Specification statements allocate data storage, determine variable and
data types, and specify input-output formats.
4.

Input-output statements control the transfer of information into and out

of the computer.

COMMENTS
Although a FORTRAN program using English words and mathematical symbols can be read and
understood more easily than a symbolic language program, it is helpful to provide comments
freely throughout the program to explain the procedures being used.

Such comments, identified

by a C in the first position of the line, are not interpreted by the compiler and have no effect
on the executable program.

C;
C;
C;

CALCULATE PERCENTAGE OF CORRECT RESPONSES
PERCENTAGE= -1 IF THERE ARE NO ITEMS
IN CATEGORY

DO 47 l=l, 57
DO 48 J=l, 6
46;

IF (ITMS (J)) 46, 46, 51
PRCN (I)= -1 .0

Figure 2

Example of Comments

CONTINUATION
Frequently, a statement may be too long to fit on one line.

If the character single quote(')

appears as the last character of a line before the carriage return, the next line is treated as
a continuation of the statement on the Iine above (see Figure 3). A statement may be continued
on as many lines as necessary to complete it, but the maximum number of characters in the statement may not exceed 128.

1.,

,
2·,
3·,

4·,

5·,

FORMAT
IF(NR(l)-1))2, 2, 3
AP=-14.73
GO TO 6
IF (NR (1) -2) 4, 4, 5
AP=-44.19
GO TO 6
AP=SH (2)•3. 0-AG (4)/ 1
AG (1 )+SQTF (AG (14))

Figure 3

The Continued Statement

THE CHARACTER SET
The characters which are meaningful in FORTRAN belong to the ASCII set Iisted in Appendix D.
Of these, the acceptable characters are: al I letters and numbers - A through Z, 0 through 9;
control characters - semicolon (;), carriage return (CR), line feed (LF), single quote ( 1 ) , double
quote (11 ) , left parenthesis(, right parenthesis), period(.), comma (,);and the operators plus(+), minus(-), slash(/), asterisk(•), equal sign(=). All other characters are ignored by
the compiler except in Hollerith information of FORMAT statements where all Teletype characters are legal. The character space has no grammatical function except in FORMAT statements, but may be used freely to make a program easily readable.

CHAPTER 3
FORTRAN ARITHMETIC

ARITHMETIC EXPRESSIONS
An algebraic formula such as the fol lowing
[5a + 4b (x 2 - xo)] /2a sin Q
represents a relationship between literal symbols {a, b, x, x 0 , Q) and constants {5, 4, 2) indicated by mathematical functions and arithmetic signs {+, -, /, multiplication, exponentiation,
sine). This same formula can be written as a FORTRAN arithmetic expression with very little
change in appearance:
{5.*A + 4.*B*(X**2 - XZRO) )/ (2.*A*SINF(THTA))
The construction of both expressions is the same; the differences are notational.
The elements of an arithmetic expression are of four types: constants, variables, operators, and
functions. An expression may consist of a single constant, a single variable, or a string of constants, variables, and functions connected by operators.
The fol lowing examples demonstrate the properties of arithmetic expressions. Each expression
is shown with its corresponding algebraic form.

Algebraic Expression

FORTRAN Expression

2
ax +bx+ c

(A**2 - B••2) I (A+B) **2

4 ir r
3

2
4.* Pl*R**2/3.

3x n- 2 (x+y)
4.25

(3.•X•Pl-2.• (X+Y)) /4.25

a ·sin 9+ 2a . cos (9 -45)

A•SINF(THTA)+2. *A*COSF(THTA-0. 7853982)

2 .p:3

2. *SOTF(X)/3.

Constants
Constants are explicit numerical quantites. They may be integers, decimals, or numbers in
decimal exponent form.

Some examples follow:

integers

-70

2047

3.14159

-0.00025

decimals
18.75
decimal exponent
l . 66E-l 6

(meaningl.66xl0

-16

)

These different forms of numerical representation are described in detail in Chapter 4.

Variables
A variable is a literal symbol whose value is not implicit; its value may be changed during the
execution of the program. A variable name is composed of one or more characters according
to these three rules:
l. The only characters which may be used in a variable name are A through
Zand 0 through 9.
2. The first character must be alphabetic (i.e., A through Z).
3. Only the first four characters of any variable name are meaningful. Al I
characters after the fourth are ignored by the compiler.

Some examples of acceptable variable names are:

XZRO

LST8

THTA

P51

DC8B

EPSL

XSUM

The name EX IT represents one variable, not two.

(Remember that blank spaces have no func-

tion in FORTRAN.) Thus, EX IT, EXIT, or even EXIT, are identical names as far as the compiler is concerned because they a II reference the same variable.
The name EPSILON would be interpreted by the compiler as EPSI, since only the first four characters are meaningful. For example, the two names XSUMl and XSUM2 would be considered
identical.
Some incorrect variable names are:
9SRT

(first character not alphabetic)

G0(5

(illegal character included)

CSH$

(illegal character included)

Operators
The operators are symbols representing the common arithmetic operations. The important rule
about operators in the FORTRAN arithmetic expressions is this: Every operation must be explicitly represented by an operator.

In particular, the multiplication sign must never be left

out. A symbol for exponentiation is also provided since superscript notation is not available.

To ii lustrate the rule, here are the FORTRAN and algebraic forms given in the section on arithmetic expressions:
(5.*A + 4.*B*(X**2-XZRO)) I (2.*A*SINF (THTA))
[5a + 4b (X 2 - x0 )J /2a sin9
Normally, a FORTRAN expression is evaluated from left to right just as an algebraic formula
is. There are exceptions to this rule. Certain operations are always performed before others
regardless of order. This priority of evaluation is as follows:

()

1st.

Expressions
within parentheses

2nd.

Unary minus

3rd.

Exponentiation

4th.

Multiplication
Division

5th.

Addition
Subtraction

The term binding strength is sometimes used to refer to the relative position of an operator in a
table such as the one above, which is in the order of descending binding strength. Thus, exponentiation has a greater binding strength than addition, and multiplication and division have
equal binding strength.
The unary minus is simply the operator which indicates a quantity whose value is less than zero,
such as -53, -GAMME, -K.

It refers only to the operand which it precedes as opposed to a

binary operator, which refers to operands on either side of itself, as in the expression a-b. A
unary minus is recognized by the fact that it is preceded by another operator, not by an operand. For example:
A+ B**-2/C-D
The first minus (indicating a negative exponent} is unary; the second (indicating a subraction}
is binary.
The left-to-right rule can now be stated more precisely as follows:
A sequence of operations of equal binding strength is evaluated from left to right.

To change the order of evaluation, parentheses are required. Thus, the expression A-B*C 1s
evaluated as A-(B*C), not (A-B)*C. The example below gives a few more i I lustrations of the
left-to-right rule.
The expression

is evaluated as

A/B*C
A/B/C
A**B**C

(A/B)*C
(A/B)/C
(A**B)**C
10

An easy way to check the proper pairing of parentheses is by counting out, illustrated in the
following example:

(Z+AM*(ZM+l .)) I ((X**2+C**2)*P)
1

1 0

The procedure is this: Reading the expression from left to right, assign the number 1 to the first
left parentheses (if you encounter a right parenthesis first, the expression is already wrong~) Increase the count by one each time a left parenthesis is read, and decrease the count by one
when a right parenthesis is used. When the expression has been completely scanned, the count
should be zero. If it becomes less than zero during the scanning, there are too many right parentheses. If it is greater than zero at the end of an expression, then the pa iring is incorrect.
Use of Parentheses
Note the use of parentheses in the fol lowing example below. They are used to enc lose the
subscript of the dimensioned variable D, to specify the order of operations of the expression
involving A, B, C, and to enclose the argument of the function.

D(l+J) = (A+B)**C+SINF (X)
In algebra there are several devices, such as square brackets ( [ ] ), rococo brackets ({ } ),
etc., for distinguishing between levels when subexpressions are nested.

In FORTRAN, only

the curved parentheses are available, so the programmer must be especially careful to make
certain that parentheses are properly paired.

In a given expression, the number of left paren-

theses must be equal to the number of right parentheses.

Functions
Functions are used in FORTRAN just as they are in ordinary mathematics -- as variables in an
arithmetic expression.
The function name represents a special subprogram which performs the calculation necessary to
evaluate the function; the result is used in the computation of the expression in which the functi on occurs.
PDP-5 FORTRAN provides several mathematical functions: square root, sine, cosine, arc tangent, exponentiation, and natural logarithm.

The argument of a function can be a simple or subscripted variable or an expression. The argument must be in floating point. FORTRAN recognizes a term as a function when the term is a
predefined symbol ending in F fol lowed by an argument enclosed in parentheses (if the F is
missing from the term, the symbol is treated as a subscripted variable). The argument of a function can consist of another function or group of functions.

For example, the expression:

LOGF(SINF(X/2)/COSF(X/2)) is equivalent to log* tan~.
The PDP-5 FORTRAN Iibrary currently consists of the fol lowing functions:
Function Name
SQTF (X)
SINF

(X)

Meaning
square root of X
sine of X, where X is expressed in radians

COSF (X)

cosine of X, where X is expressed in radians

ATNF (X)

arc tangent X, where the angle is given
in radians

EXPF

(X)

LOGF (X)

exponential of X
logarithm of X

THE ARITHMETIC STATEMENT
The arithmetic statement relates a variable V to an arithmetic expression E by means of the
equal sign (=). Thus:

V=E
Such a statement looks like a mathematical equation, but it is treated differently. The equal
sign is interpreted in a special sense; it does not represent a relation between left and right
members, but it specifies an operation to be performed.
NOTE: In an arithmetic statement, the value of the expression to
the right of the equal sign replaces the value of the variable on the
left.
This means that the value of the left-hand variable wil I be different after the execution of an
arithmetic statement. A few illustrations of the arithmetic statement are given below.

a. VMAX =VO+ AXT
b. T = 2.*Pl*SQTF(l ./G)
c.

Pl = 3. 14159

d. THTA = OMGA + ALPH*T**2/2.
e. MIN= MINO

INDX =- INDX +2

With the interpretation of the equal sign defined previously, example f becomes meaningful as an
arithmetic statement.

If, for example, the value of INDX is 40 before the statement is executed,

its value will be 42 after execution.
Perhaps another way of looking at the equal sign illustrates its use and interpretation more fully.
In arithmetic expressions, a binary operator requires an operand on its left and right.

The equal

sign of an arithmetic statement is considered to be a binary operator also. This interpretation
is demonstrated in the fol lowing revised table of operators:
Operator

Use

Interpretation

-A

negate A

A**B

raise A to the Bth power

A*B

multiply A by B

A/B

divide A by B

A+B

add B to A

A-B

subtract B from A

A=B

replace A with B

(Unary}

(Binary)
=

Treated this way, the equal sign is considered to have the lowest binding strength of all the
operators. This means that the expression on the right is evaluated before the operation indicated by = is performed.
The most important result of treating the equal sign as a binary operator is that it may be used
more than once in an arithmetic statement. Consider the fol lowing:
CPRM = (CKL - CKG) / (CPG = P*(Q + l .))

The internal arithmetic statement, CPG = P*(Q + 1 .), is set off from the rest of the statement
by parentheses. The complete statement is a concise way of expressing the fol lowing common
type of mathematical procedure:

Let

c' =

where

= P*(q+l)

The stating of a relation fol lowed by the conditions for evaluating any of the variables can be
expressed in a single arithmetic statement in FORTRAN.
Another important result of treating the equal sign as an operator is that the operations may be
performed in sequence. Just as there may be a series of additions, A+B+C, so may there be a
series of replacements, A=B=C=D. Note that since the operand to the left of an equals sign
must be a variable, only the rightmost operand, represented by D in the example, may be an
arithmetic expression. The statement is interpreted as follows:

Let the value of the expression

D replace the value of the variable C, which then replaces the value of the variable 8 11 and
so on.

In other words, the value of the rightmost expression is given to each of the variables

in the string to the left. A common use for this construction is in setting up initial values:
XZRO=SZRO=AZRO=O
T=Tl=T2=T3=60
P=FP=4.*ATM-AK
Another useful result in treating the equal sign as an operator is that the value of an expression
on the right of an equal sign is converted to the mode of the left-hand variable before storage,
if necessary.
Example:

A=M
K=B

CHAPTER
NUMBER

REPRESENTATION

AND

VARI AB LE TYPES

INTEGERS AND FLOATING POINT NUMBERS
In mathematics there are many ways to categorize numbers. They may be positive or negative,
rational or irrational, whole numbers or fractions.

In FORTRAN, the treatment of numbers is

separated into integers and decimals, distinguished as follows:
l . Integers are positive or negative numbers written without a decimal point.
These numbers are integers: 9, 17, 147, 1024, 2047. The last number, 2047,
is the largest quantity that can be expressed as a FORTRAN integer. For fractional quantities and for numbers larger than ±2047 (which is 2 11 -1), the second type of number is required.
When using integer arithmetic, any fractional results are truncated. For example, the expression M=N/3 with N=8 would result in M=2. This applies
only to division because multiplication, addition and subtraction yield integra I resu Its.
2. Floating point numbers have two forms. They are simple decimals, such
as 0.0025, .4, 57., 2.71828; or numbers in decimal exponent form.

Num-

bers in decimal exponent form are simple decimals multiplied by a power of 10.
Examples:
Mathematical Form

FORTRAN Form

6.023 x 1023

6.023E23

1 .66 x 10- 16

1 . 66E-16

72. x 10 12

72El 2

In general, a floating point number in decimal exponent form is NE±K, where
N may be an integer or simple decimal, and K is an integer from 0 to 99, inclusive. The construction NE±K is used to represent the number NxlOK

The

fol lowing are floating point representations of the number 19:
19 .0
.19E2
1 . 9E+l
1900E-2
190.0E-1

FIXED-AND FLOATING-POINT REPRESENTATION
The difference between integers and real numbers in FORTRAN is the way in which each is
represented in core memory.
A FORTRAN integer is stored in one 12-bit computer word. The sign of the number is kept in
the high-order bit and the magnitude in the remaining 11 bits. This representation, shown
schematically in Figure 4 is cal led fixed point, because the decimal point is always considered
to be to the right of the rightmost digit. A FORTRAN integer may not exceed the range of
-2047 through +2047. All integers greater than ±2047 are taken modulo 2048 (that is 2049 is
taken as 0001 , 4099 is taken as 3).
The floating point format consists of two parts: an exponent (or characteristic) and a mantissa.
The mantissa is a decimal fraction with the decimal point assumed to be to the left of the leftmost digit. The mantissa is always normalized; that is, it is stored with leading zeros eliminated so that the leftmost bit is always significant. The exponent represents the power of two
by which the mantissa is multiplied to obtain the true value of the number for use in computation. The exponent and mantissa each are stored in 2 1s complement form.

SIGN

I I

MAGNITUDE

0 1

o. FORTRAN INTEGER

SIGN OF EXPONENT

I l

SIGN OF
MANTISSA

EXPONENT

t01

MANTISSA
b.FLOATING POINT

Figure 4

Number Representation

TYPES OF VARIABLES
Since variables represent numeric quantities, the type of representation must be specified in
some manner.

In normal programming, variable types are specified using the FORTRAN con-

ventions as fol lows:
l.

lntegervariablenamesmustbeginwithoneofthe letters I, J, K, L, M, or N.

2. Floating point variables are designated by names beginning with any other letter.
These are integer variable names: INDX, KDTA, M359. These are floating-point variable
names: ZXRO, CONT, FICA.
Integers cannot appear in floating point expressions except as exponents or subscripts. Some
examples of illegal and legal expressions are as follows:
Exeression

Legal

Mode

A(l)*B(J)**2
l(M)*K(N)
4.*J
l+D
16. *B
(K+l6)*3
A**(l+2)/B

Yes
Yes
No
No
Yes
Yes
Yes
No

Floating
Fixed

B*A

Floating
Fixed
Floating

CHAPTER 5
SUBSCRIPTED VARIABLES

ARRAYS
An array is a grouping of data. A column of figures, the elements of a vector, a list, and a
matrix are al I arrays.

In mathematics, an element of an array is referenced by means of a sym-

bol denoting the array and subscripts identifying the position of the element. For example,
the sixth element in a vector v is designated v 6 .
In FORTRAN, array elements are similarly identified. The array is given a name subject to
the same rules as the names of variables, described in Chapters 3 and 4. The subscript which
identifies an element of the array is enclosed in parentheses. The element referred to in the
preceding paragraph would have the following form in FORTRAN:

V{6)
Such a name designates a subscripted variable, which may be used in computation just like a
simple variable. The array name determines the mode, integer, or floating point of all the
elements in the array.
The example below gives a few illustrations of the use of subscripted variables.

a. X(l+L)=X(l)+ALPH(l)*P{I)
b. X{l+3)=X(1+2)+X(I+ 1)/2.
c. C(J)=A{l*J+3)

d. A=B(6)

Subscripts
As the example above illustrates, subscripts may be quite diverse in form.

In fact, a subscript

may be any acceptable FORTRAN arithmetic expression as long as it is integer-valued. This
means that there may not be any floating-point quantities in a subscript expression.

The Dimension Statement
Array names must be identified as such to the FORTRAN compiler. Two items of information
must be provided in any program using arrays:
l. Which ore the subscripted variables?
2. What is the maximum dimension of the subscript? (When on array is used,
a certain amount of storage space must be set aside for its elements; hence
this requirement.)
All the above information is provided by the fol lowing spec ificotion statement:
DIMENSION A(20), B(l5)
where A and Bore array names, and the integer constants 20 and 15 ore the maximum dimensions of each subscript.
The rules governing the use of array names and the dimension statement are as follows:
All array names must appear in a dimension statement. DIMENSION may be used as many
times as desired and may appear anywhere in the FORTRAN program, provided that the DIMENSION of on array appears before any statement which references the array.
DIMENSION LIST(30) I MAT(l 00) I RE GR(20)
In the statement in the example above, the names LIST and MAT designate integer arrays; that
is, each element is an integer. The third name, RE GR, designates a floating-point array.

The

first array is a I ist of 30 elements maximum, so that 30 words of storage are set aside for its use.
The third array is floating-point and there are 20 elements in it.

Since this array is floating,

each element requires 3 words of storage so that 60 words are set aside for the array.
DIMENSION B(30), 1(15)
This version of the PDP-8 Fortran does not have the facility for double subscripted variables.

To accomplish double subscripting, the programmer has to include indexing statements in the
source program as illustrated in Figure 5.

C·
I

1I.

2·
I

1O;
I

15;

20;

30;

40;
21;

MATRIX MULTIPLY
DIMENSION A(36), B(36), C(36)
ACCEPT 1, I
FORMAT (I)
DO 10 M=l, I
DO 10 N=l, I
INDX=M+l*(N-1)
ACCEPT 2, A(INDX)
FORMAT (E)
CONTINUE
TYPE 15
FORMAT (/,/,/)
DO 20 M=l, I
DO 20 N=l, I
INDX=M+l*(N-1)
ACCEPT 2, B(INDX)
C(INDX)=O
CONTINUE
DO 30 M=l, I
DO 30 N=l, I
DO 30 K=l, I
IC=-N+l*(M-1)
IA=K+l*(M-1)
IB=N+l*(N-1)
C(IC)=C(IC)+A(IA)*B(IB)
CONTINUE
TYPE 15
D040M=1, I
TYPE 21
DO 40 N=l, I
INDX=N+l*(M-1)
TYPE 2, C(INDX)
CONTINUE
FORMAT(/)
TYPE 15
END

Figure 5

Indexing Statements

In this example the matrices are stored column wise in memory, that is, sequential locations in
memory are used as fol lows:

Element

Relative Position
in Memory (INDX)

all
a21
a31
a41
a51
a61

1
2
3
4
5
6

al2
a22

7
8

a56
a66

35
36

------

If referencing element a 56 in the array, M=5, N=6 (I would be =6 for a 6 by 6 array.), and
INDX=M+l*(N-1)=5+6*5=35. If referencing element a 22 , INDX=2+6*1=8.

CHAPTER 6
PROGRAM CONTROL

In this chapter, the FORTRAN statements which have been described as isolated elements are
discussed in their proper context -- in program sequences.

It is obvious that FORTRAN state-

ments are executed in the order in which they are written unless instructions are given to the
contrary. Such instructions are provided by the program control statements, which al low the
programmer to alter the sequence, repeat sections, suspend operations, or bring the program to
a complete halt.

PROGRAM TERMINATION
A program arranged so that the last written statement is the final and only stopping place needs
no special terminating indication. The end statement automatically determines the final halt.
Most programs, however, contain loops and branches so that the last executed statement is often
somewhere in the middle of the written program.

Frequently, there may be more than one stop-

ping point. Such terminations are indicated by the statement:
STOP
This causes a final, complete halt; no further computation is possible.
When a STOP is encountered during program execution at object time, the system signifies that
a stop has occurred by typing an exclamation mark (~) on the tape teleprinter.
The stop statement prevents further computation after it has been executed. There 1s a way,
however, to suspend operation for a time and then restart the program manually. This procedure is frequently necessary when the operator must do such tasks as loading and unloading
tapes or card decks in the middle of a program.

This kind of temporary halt is provided by the

fol lowing statement:
PAUSE

This brings the program to a halt, but the operator may restart it at any time by pressing the
CONTINUE key on the computer console.

BRANCHES AND LOOPS

The GO TO Statement
There are various ways in which program flow may be directed. As shown schematically in
Figure 6, a program may be a straight-line sequence (1), or it may branch to an entirely different sequence {2), return to an earlier point {3), or skip to a later point (4).

-- l
2

-J4
,,
l

Figure 6

Schematic Representation of Program Branching

All of these branches can be performed in several ways, the simpliest of which is by using the
statement:
GOTO N
where N is a statement number used in the program. The use of this statement is described in
the following example, which also illustrates the construction of a loop, the name given to
program branches of the type shown in Figure 6, No. 3.

Integer Summation
In the example below, the sum of successive integers is accumulated by repeated addition. The
main computation is provided by the three-instruction loop beginning with statement 2. The
statements preceding this loop provide the starting conditions; this is called the initialization.
The partial sum is set to zero, and the first integer is given the value of one. The loop then
proceeds to add the integer value to the partial sum, increment the integer, and repeat the
operation.

(.
I

2·
I

SUM OF FIRST N INTEGERS BY ITERATION
KSUM=O
INUM=l
KSUM=INUM+KSUM
INUM=INUM+l
GO TO 2
Figure 7

Integer Summation

Limits and Decisions - The IF Statement
The program shown in the preceding example performs the required computation, but there is
one flaw: the loop is endless. To get out of the loop, the user must know when to stop the
iteration and what to do afterwards.
The IF statement fulfi I ls both requirements.

It has the fol lowing form:

IF (E)K, L,M
where E is any variable name, arithmetic expression, or arithmetic statement, and K, L, and
Mare statement numbers. The statement is interpreted in this way:
if the value of E is less than 0, GO TO statement K
value of E is equal to 0, GO TO statement L
value of E is greater than 0, GO TO statement M
Thus, the IF statement makes the decision of when to stop by evaluating an expression, and
also provides the program branch choices which depend on the results of the evaluation.

3;
Figure 8

SUM OF THE FIRST 50 INTEGERS
KSUM=O
INUM=l
KSUM=INUM+KSUM
INUM=INUM+l
IF (INUM-50) 2,2,3
STOP

Use of IF Statement in Integer Summation Problem

In this example, the initialization and main loop are the same as for the preceding example
except that the GO TO statement of the earlier program has been replaced by an IF statement.
This statement says: If the value of the variable INUM is less than or equal to 50 (which is
the same as saying that if the value of the expression INUM-50 is less than or equal to zero),
go to statement 2 and continue the computation.

If the value is greater than 50, stop.

A further improvement on the example above can be made if the feature of substatements within an expression is incorporated (refer to pages 13 and 14).

C·
I

2·
I

3·
I

Figure 9

SUM OF THE FIRST 50 INTEGERS
KSUM=O
INUM=50
KSUM=INUM+KSUM
IF(INUM=INUM-l) 3,3, 2
STOP
IF Statement with Substatement Feature

In this example, the sum is formed by counting down, but the same results are achieved. The
initialization is changed so that INUM starts with the value of 50 instead of 0, and the statement INUM=INUM+ 1 is no longer required.
A loop may also be used to compute a series of values. The following example is a program to
generate terms in the Fibonacci series of integers, in which each succeeding member of the
series is the sum of the two members preceding it:

;
5;
6;
10;

FIBONACCI SERIES, 100 TERMS
DIMENSION FIB{lOO)
FIB(l)=l
FIB(2)=1
K=3
FIB(K)=FIB(K-l)+FIB(K-2)
K=K+l
IF (K-100) 5,5, 10
STOP
Figure 10

Fibonacci Series

In this program, the initialization includes a dimension statement which reserves space in storage, and two statements which provide the starting values necessary to generate the series.
Each time a term is computed, the subscript is indexed so that each succeeding term is stored
in the next location in the table. As soon as the subscript reaches 100, the calculation stops.

DO Loops
Iterative procedures such as the loop in the example above are so common that a more cone ise
way of presenting them is warranted.

In this example, three statements are required to initia-

1ize the subscript, increment it, and test for termination. The following type of statement com-

bines al I these functions:
DO n l=Kl, K2, K3
Here, n is a statement number, I is a simple integer variable, and Kl, K2, and K3 are indexing parameters which provide, in order, the initial value of I, the final (terminating) value of
I, and the indexing increment of I.

If K3 is omitted from the statement, it is assumed equal to

one. Statement n must be a CONTINUE statement.
C;

Figure 11

FIBONACCI SERIES, 100 TERMS
DIMENSION FIB(lOO)
FI B(l)=F IB(2)= 1
DO 5 K=3, 100
F IB(K}=FI B(K-1 )+Fl B(K-2)
CONTINUE
STOP

Fibonacci Series Calculation Programmed As a DO Loop

In words, the DO statement says: Do al I statements through statement 5 for K=3, when statement 5 is encountered.

Perform the following test: If K+l is less than or equal to 100, set

K=K+ l and continue on in the program by executing the first statement after the DO.

If the

K+l is greater than l 00, the next sequential statement following statement 5 is executed.

this example this is a STOP.
DO loops are commonly used in computations with subscripted variables.
usually necessary to perform loops within loops.

5;
l O;

In these cases, it is

Such nesting of loops is permitted in FORTRAN.

DO 10 l=l, 20
X(l)=O
DO 5 J=2, 40, 2
X (I )=X (l)+(B(J)-Z(J) )** 2
CONTINUE
A(l)=X(l)**2+C(I)
CONTINUE
Figure 12

Nested DO Loops

In the previous example, sequential elements in the X array are formed by summing the square
of the difference of every second element in the Band Z arrays.

Then the A array is formed

by summing every element in a C array and the square of every element in the X array.
algebraic expression for the loop is as follows:

A. =X. - C.
I

where

for i l, 2, 3, ••. 20

xi = fO~ (b.-z.) 2 for j = 2, 4, 6, ... 40
j=2

J J

The following general rules about DO loops must be observed.
l . DO loops may be nested, but they may not overlap.

Nested loops may

end on the same statement, but an inner loop may not extend beyond the
last statement of an outer loop. Figure 13 schematically illustrates permitted and forbidden arrangements.

The

2. If the user transfers into the range of a DO, the variable I is not initialized as specified in the DO statement. Transferring into the range of a DO
is allowed as long as:
a.

Incrementing and testing start with the present value of I.

b. Control was originally transferred out of the DO other than
by completing it.
3. A DO loop must end on a CONTINUE statement.

------1

---2

---3

[[[~
b.

Figure 13

DO Loops

Those in a are permitted; loops 5, 6, and 7 end on the same statement. The arrangements in b are not permitted; loop 3 ends on a
statement outside the range of loop l.

Illegal DO Nesting

l O;
20;

DOlOl=l,20
DO 20J=l,100,2
SUM=(X(l)-Y(l))**2
CONTINUE
Z(J)=SUM+A(J)
CONTINUE

Figure 14

Program Branching in DO Loops

Branches 2, 5, 6, and 7 are permitted; branches 1, 3, and 4 are not.

The CONTINUE Statement
Since the DO loop may contain alternate courses of action, programmers frequently wish to
make the last executable statement of a loop, a test to determine which of the alternatives
should be taken next.

However, Rule 3 above forbids a DO loop to end on an IF or GO TO;

so a special statement is provided which is not an executable statement itself, but provides a
termination for such a DO loop. The statement is:
CONTINUE
DO loops must be terminated ona CONTINUE statement.

Computed GO TO
The GO TO statement described in the section on branches and loops is unconditional and provides no alternatives. The IF statement offers a maximum of three branch points. One way of
providing a greater number of alternatives is by using the COMPUTED GO TO, which has the
following form:
GO TO (Kl, K2, K3, ... Kn), J

where the K's are statement numbers, and J is a simple integer, which takes on values of l,

2,3, ... n according to the results of some previous computation. For example:
IVAR= 14*J/2+K
GO TO (5, 7, 5, 7, 5, 7, 10), IVAR
causes a branch to statement 5 when IVAR=l,3, or 5, to statement 7 when IVAR=2,4, or 6, and
to statement l 0 when IVAR=7.
When IVAR is less than one or greater than seven, the next sequential statement after the GO

TO is executed.

END Statement
END occurs alone and indicates the physical end of the program to FORTRAN.
lowed by CR-LF.

It must be fol-

CHAPTER 7
INPUT AND

OUTPUT

AVAILABLE DEVICES
So far, we have assumed that al I information (programs, data, and subprograms) was in memory,
without regard to how it got there.

Programs, of course, are read in by a special loader, but

the programmer is responsible for the input of data and the output of results by including these
operations in his program.
For any input-output procedure, several items must be specified:

l . In which direction is the data going? In FORTRAN terms, the data coming in is being read into memory; information going out is being written on
whatever medium is specified.
2. Which device is being used?

Information can be transferred between

core and either of two different input-output devices; each 1/0 operation
must specify which device is involved.
3. Where in core memory is the data coming from or going to? The amount
of data and its location in the computer storage must be specified.
4.

In what mode is the data represented?

In addition to floating- and

fixed-point modes for numeric data, there is the Hollerith mode for transferring alphanumeric or text information.
5. What is the arrangement of the data?

In FORTRAN terms, the format

of incoming or outgoing data is specified.
For every data transfer between core memory and an external device, two statements are required to provide all of the information listed above. The first three items are specified by
the input-output statement and the last two items are determined by the FORMAT statement.

PDP-5 FORTRAN provides ways of communicating with either the on-line tape teleprinter or
DECtape.

Information is transferred between these devices and core memory in the 8-bit ASCII

code {Appendix D).
The tape teleprinter can be used to transfer information to the computer from paper tape through
the paper tape reader or manually from the operator by typing on the keyboard.
Information can be transferred from the computer to the tape teleprinter. A printed copy, with
or without a punched paper tape, can be obtained.
When transferring information with DECtape, the same ASCII code format is used.

INPUT-OUTPUT STATEMENTS
The input-output statements control this transfer of information. As illustrated in Figure 15,
1/0 statements consist of three basic items of information: The device being accessed and the
direction of transfer, the number of the FORMAT statement that controls the arrangement of
data, and the I ist of names of the variables whose values are to be output or changed by new
inputs.
ACCEPT

N, V(I),

V{l+l),

V{l+2)

Lust of variable names

Statement number of FORMAT statement
Device selection and direction of transfer
Figure 15

Input-Output Statement

Device Selection and Direction of Transfer
ACCEPT and TYPE transfer information between the tape teleprinter and the PDP-5.
ACCEPT causes information to be accepted into core memory from either the paper tape reader
or the keyboard depending on which is activated.

TYPE causes information to be transferred from core memory to the printer or the printer and
paper tape punch depending on whether the punch is activated or not.
READ causes information to be read into core memory from DECtape.
WRITE causes information to be written on DECtape from core memory.

Statement Number of Format Statement
Fol lowing the instruction that selects the device and direction of transfer is the statement number of the FORMAT statement that controls the arrangement of the information being transferred.
Example:

10;

ACCEPT 10, A
FORMAT(E)

Every 1/0 statement must have a reference to a FORMAT statement.

List
The next item of specification in the 1/0 statement is the names of the variables or array elements that are involved in the transfer. These names are arranged in a sequential I ist in the
order that their values are transferred. There is no restriction on the modes of the variables in
the list or the number of names that can appear in a list as long as they are compatible with
the corresponding FORMAT statement.

10;

TYPE 10, A, I, B, C(l+K), N(J+L)
FORMAT (E, 1,E,/)

A=A+(C(J)**2-C(N)**2)

TYPE 10, A, JIB, C(N)

Figure 16

A List Example

If the list contains more names than there are elements in the FORMAT statement, when the
elements are exhausted the FORMAT statement is reinitialized, and the first element in the
FORMAT statement corresponds to the next name in the Iist.
For instance in the preceding example when the value of the variable B is typed in the E format,
the control character slash V) causes a carriage-return Iine-feed to occur. Then the FORMAT
statement is reinitialized, and the array element C(l+K) is typed in the E format and the array
element N(J+L) in the I format.
Correspondingly, the list does not have to exhaust the elements of a FORMAT statement.

there are more names in the Iist than there are in the FORMAT statement, the program completes
the 1/0 operation and proceeds to the next sequential FORTRAN statement.

If this next state-

ment is another 1/0 statement that references a previously unexhausted FORMAT statement,
that FORMAT statement is reinitialized.

In other words, FORMAT statements are reinitialized

when they are first referenced or when all of their elements are exhausted.

FORMAT SPECIFICATIONS STATEMENT
As a Iready mentioned in the previous description of input-output statements, the FORMAT statement controls the arrangement and mode of the information being transferred.

The values of

the names appearing in the list of the 1/0 statement are transferred in the mode specified by
the corresponding element in the FORMAT statement. These controlling elements consist of
the characters E, I, slash V) and quote ( 11 ) . The set of elements must be enclosed in parentheses
and separated by commas.
Example:

FORMAT (E, I,/, 11 HOLLERITH 11 )

Control Elements E and I
The control elements E and I are used for defining the mode of the data being transferred. When
a variable is transferred in the E format, it is stored or output in floating point.

If the variable

is transferred in the I format, it is stored or output in fixed point. Mode conversion on input
or output can be accomplished because the elements in the FORMAT statement define the mode
of the data and the mode of the variable is overriden.

Example:
l O;

TYPE l 0, A
FORMAT (I)

The variable A is typed as an integer and the fractional part of A is truncated. For instance,
if A has a value of 14. 96, only the integer part, 14, wou Id be typed.

If A has an absolute

value of less than one, zero would be typed.

Input data words consist of a sign, the decimal value, an exponent value if the data is floating
point, and a field terminating character such as space. Any character that is not a number,
decimal point, sign, or E can be used to terminate a field except the character rub out. When
typing data, any number of spaces or other non-numeric characters can be typed before the sign
or decimal value is typed to make the data sheet more readable. If a mistake is made when
typing data words, the last word or partial word can be erased from core memory by typing the
character rub out.
These input words can be transferred into core memory from either the paper tape reader, the
keyboard or DECtape. They can be entered in either fixed- or floating-point modes for integers or decimal fractions. The mode in which they will be stored is controlled by the corresponding element in the FORMAT statement.

Integer Values - Fixed Point - FORMAT (I)
An integer data field consists of a sign (minus or space) and up to four decimal characters.
Some examples of integer values are as fol lows:
Typed Numbers

Values Accepted

-2001
-40
-0040
16
-2047

-2001
-0040
-0040
0016
-2047

Decimal Fraction Values - Floating Point - FORMAT (E)
A floating-point input word consists of a sign*, the data value of up to seven decimal characters, an E if an exponent is to be included, the sign of an exponent, and the exponent
which is the power of ten that the data word is multiplied by.
Example:
dddd.dddEnn
The d's represent characters in the data word and n represents the power of ten of the
exponent. Either the sign, the decimal point, or the entire exponent part can be omitted. If
the sign is left out, the number is assumed to be positive; if the decimal point is left out, it is
assumed to appear after the rightmost decimal character.

If the exponent is omitted, the power

of ten is taken as zero.
Examples of floating-point values are as follows:
Typed Numbers

Values Accepted

0.16x 10

.16E02

0.16x 10

1600.E-02

0.16x 10

2
2
2

Correcting Typing Errors
If a mistake is made when typing data words into a FORTRAN program, the mistake can be corrected by canceling or erasing the data word before typing the terminating character and then
retyping the data word that is in error.
To cancel or erase a word, type a rub out character.
When this character is detected during the acceptance.of a data word and before the termination
character has been transmitted, the data word appearing before the character rub out is erased
from memory. Operations on the names in the list do not advance to the next sequential name
until a complete data word and the terminating character have been received.
*Plus sign can be represented by a plus or space character. Minus is represented by a minus
character. If a sign character is absent from the data word, the data is stored as positive.

Output

Data Word Output - Floating and Fixed Modes - FORMAT (E) and FORMAT (I)
Integer values are always printed as the sign and a maximum of four characters with spaces replacing leading zeros. Floating-point values are printed in a floating-point format which consists of sign, leading zero, decimal point, seven decimal characters, the character E, the sign
of the exponent (minus or plus), and an exponent value of two characters.
Examples:
Integer Values

Output Format

-1043

-0016

+0016

+ 16

Floating-point values are printed as per example
SO . dddddddsxx
where
S is the sign, minus sign, or space
dare the seven decimal digits of the data word
s is the sign of the exponent value
xx is the exponent value
Output Format

Decimal Value

-0.8388608E+07

-8,388,608.0

0.1192092E-06

+.0000001192092

OTHER FORMAT CONTROL ELEMENTS
In most cases when data is to be presented it must be labeled and arranged properly on a data
sheet. In order that this can be accomplished with FORTRAN, a provision has been made so
that text information and spacing can be typed out along with the data words. These features
are provided by the special FORMAT control elements quote (11 ) and slash (/).

Quote (11 )
When text information is contained as part of a FORMAT and this information is enclosed in
quotes, it is output to the specified device as it appears in the statement. This output occurs
when a TYPE or WRITE statement references a FORMAT statement containing text and al I other
elements of that FORMAT statement previous to the text have been used.

10;

TYPE l 0
FORMAT(/, 11 THIS IS HOLLERITH 11 , / )

l 00;

TYPE l 00, AMIN, AMAX
FORMAT (/, 11 MIN IMUM= 11 IE,/, 11 MAXIMUM= 11 IE,/)

;
21 O;

220;
250;

TYPE 210
FORMAT (/,/, 11
CUMULATIVE DISTRIBUTION",/,/'
11
INCREMENTS
FREQUENCY",/)
DO 220 K= l I l 00
TYPE 250, K, VALU(K), VALU(K+l), COUNT(K)
CONTINUE
FORMAT(l, 11 11 ,E, 11 11 ,E, 11 11 ,E,/)

Figure 17 Examples of Quote and Slash
All legal Teletype characters can be contained within quotes and are output as text (Appendix D).
Before text is output, the elements of the FORMAT statement that appear in front of the Hollerith
information must have been used.
Example:
l
10;

TYPE 10, VAR,SD
FORMAT (E,E,E, 11 VARIANCE AND STANDARD DEVIATION 11 , / )

In this example, the test is not typed because one of the E elements was not used.

Slash V)
The slash character is used for typing a carriage return and Iine feed for advancing the paper
of the tape teleprinter. A carriage-return line-feed will be typed for every slash that appears
in the statement.
Example:
TYPE 10, A, B

1O;

FORMAT V,/,/, E,/,/, E,/,/)

Three carriage-return line-feeds will be typed before the value of A; then two carriage-return
line-feeds will be typed before and after the value of Bis typed.
The input subroutine of the object time system ignores al I non-numeric characters except as data
word delimiters so that input data can be labeled and spaced by intermixing the appropriate
text and carriage-return line-feeds with the data.

APPENDIX A
DIAGNOSTICS

COMPILE TIME DIAGNOSTICS
Diagnostic procedures are provided in the compiler to assist the programmer in program compilation. When the compiler detects errors in a FORTRAN source program, it prints out error messages on the on-I ine tape-teleprinter. These messages indicate the source of the error and direct
the programmer in the efforts to correct the error.
To speed up the compiler process, the compiler prints out only an error code. The programmer
then looks up the error message corresponding to the code in Table A-1 and takes the appropriate corrective measures.

Dynamic Error Correction
A user may choose to compile in either of two modes: the normal mode or the dynamic correction mode. The latter allows the user to correct a statement, which the compiler has determined
contains a source-language error, by re-entering the offending Iine via the tape teleprinter without having to physically correct the symbolic tape and recompile. This feature is not implemented
in the high-speed reader version of the compiler since the higher speed of the device makes
recompi lotion easy.
To choose the dynamic correction mode:
1.

Load the starting address of the compiler (0200) in the console switches

and press LOAD ADDRESS.
2.

Set all console switches on (7777), press START (setting the switches to

7777 before pressing START is the only departure from normal operation).
If an error is detected, the diagnostic prints out in the normal fashion and the computer halts.

To correct the statement:

1 . Turn READER switch to FREE.
2. With the READER switch stil I in the FREE position, press CONTINUE.
3. Type the new I ine in its entirety,* obeying al I rules for the source language and terminating the statement with a carriage-return line-feed.
4.

Turn reader on and compilation will continue.

To leave the dynamic correction mode, restart the compiler in the normal fashion.

DIAGNOSTICS

Format of Diagnostics

xx xx

xx
LThe identifying condition code
The number of statements since the appearance of a numbered
statement (octal value).

The statement number of the last numbered statement
Example:

1 O;

A=I (J+l)
B=A*(B+SINF(THTA)

During compilation of the above statements the following error code would be printed,

*If the statement was numbered, do not re-enter the statement number unless it was in error.

indicating mat a statement which occurs eleven statements octal (8 decimal) after the appearance of statement number 10 is in error. The message corresponding to code 11 shows that the
number of left and right parentheses in the statement is not equal. The statement is examined
and corrected; then compilation is resumed.
TABLE A-1
Diagnostic
Code

Conditions

Fixed- and floating-point modes have been mixed in an expression.

Two operators appear adjacent to each other (i.e. a variable
has been left out of an expression) e.g. A=C + * D.

Compiler error - Temporary diagnostic until compiler has been
completely field tested. Contact Applied Programming if this
occurs.

A comma has been used illegally in an arithmetic statement.

Too many operators appear in a single statement.

A function argument is in fixed mode. e.g. SINF(INC).

A variable subscript is in floating point mode. This could also
indicate that an operator is missing. e.g. A+B(C+l .) for
A=B*(C+l.).

More than 64 different variable names have been used in the
program.

Program too large - program and data requirements have overlapped.

There is an unequal number of right and left parentheses in a
statement.

An ii legal character was detected and ignored.

The compiler is unable to recognize or process this statement
due to some error in its format.

Not used at th is time.

A subscripted variable is defined before the appearance of a
dimension statement or a subscripted variable does not appear
in a dimension statement. It might also indicate that an operator is missing in a fixed-mode expression. e.g. A=l(J-K)
for A=l*(J-K).

TABLE A-1 (continued)
Diagnostic
Code

Conditions

Statement too long - more than 128 characters have been counted
not including spaces except in format statements where al I legal
TTY characters are counted.

A floating-point operand should have been fixed-point, e.g.
DO 10 I= 1 I 7. 3.

A statement number that has been referenced does not appear
in the program. See the paragraph on the next page.

There are more than 40 numbered statements in the source program.

A statement cannot be compiled because it has too many incompleted operations, e.g. C=A+(C+(D+(E+ ....

Too many statements. have been referenced before they are defined.

If a statement number is referenced but does not appear in the source program, the diagnostic
code will be printed as follows:
xxxx 77 20
where the number usually reserved for the last numbered statement (xxxx) is replaced by the
missing statement number.
e.g. GOTOlOO
The diagnostic would appear as fol lows where statement 100 is never defined.
100 77 20

OBJECT SYSTEM DIAGNOSTICS
Not all errors are detected by the compiler. Some errors can only be detected by the object
time system. Also, there are some conditions which indicate errors on the part of the compiler
and/or object system. When such an error occurs during running of a program, the computer
types out an error message containing an error number. The computer then halts.

If the CON-

TINUE toggle is pressed, the computer takes the action listed in the following table.

TABLE A-2
Error
Number

Possible Cause

Action Taken

Attempt to divide by zero

Quotient set to plus or minus largest
number representable in computer;
then continue executing instructions.

Floating point exponent on input
greater than plus or minus 2047

System executes next instruction.

Illegal operation code (either
compiler error, or data stored
over program, or transfer to data
section}

System executes next instruction.

Transfer to core location zero or
one

No recovery possible.

Non-format statement used for a
format

System executes next instruction.

Illegal format statement constituent

System examines next constituent.

Attempt to take square root of a
negative number

System takes square root of absolute value.

Attempt to raise a negative number to a power

System raises absolute value to the
power specified.

Attempt to find the logarithm of
zero or a negative number

System attempts to find logarithm of
absolute value. Note that log (absolute value (O)) still gives an error halt.

One of the stacks used by the
system has underflowed. (i.e.,
more data has been requested
than was placed on the stack)

No recovery possible. Since this may
be a system error, communicate program and circumstances to DEC.

One of the stacks has overflowed.
(i.e., more data placed on it
than there is storage in the machine.}

Same as Error 76

APPENDIX B
TAPE TELEPRINTER OPERATING

PROCEDURES

1 . Symbolic tape preparation using the ASR-33.
a. Turn the LINE switch to OFF-LINE to disconnect the ASR-33 from
the PDP-5.
b. Turn the POWER switch on.
c. Check quantity and positioning of typewriter paper and paper tape
for the punch.
d. Press the PUNCH ON button.
e. Generate leader.*
f.

Type the symbolic program.

g. Generate trailer.*
h. Verify tape as fol lows {suggested procedure):
1 . Turn READER switch to FREE.
2. Place tape in reader.
3. Press PUNCH OFF button {so the tape read is not dup-

1icated).
4. Turn READER switch to START.
5. The entire content of the tape is printed for visual verification and the reader stops automatica 11 y at the end of the tape.
*To generate leader/trailer {200 code) press the character keys P, SHIFT, CTRL and REPT
simultaneously.

For computer-aided symbolic tape preparation and editing, refer to PDP-5

symbolic tape editing program (DEC 5-37-S).
3. Manual symbolic tape editing using the ASR-33.
a.

An incorrect character might be typed while preparing the sym-

bolic tape.

Use the fol lowing procedures to correct the tape: (the

error is detected N characters after typing the incorrect character)
press the PUNCH B.SP. button N+l times, press rubout N+l times,
and continue.
b. Characters, words, or statements can be inserted or deleted after
the entire symbolic tape has been prepared.

Use the fol lowing pro-

cedures to accomplish such changes.
l.

Insertions - Duplicate the tape up to the point at which

it is desired to make an insertion (by turning the punch on,
placing the tape in the reader, starting the reader, and
stopping the reader with the READER switch using the printout as a guide).

Next, type the insertion. Continue by

pressing the READER switch to start and dupl icote the remainder of the tape.
2.

Deletions - Duplicate the tape up to the point at which

it is desired to make a deletion (see Insertions).

Next, turn

the punch off; start the reader; and using the printout of the
information to be deleted as a guide, stop the reader. Continue by turning the punch on and starting the reader to dup1icate the remainder of the tape.

APPENDIX C
OPERATING
AND

PROCEDURES FOR RIM

BIN PAPER TAPE LOADERS

READ-IN-MODE LOADER (RIM)
1. The RIM Loader is a minimum-length, basic paper tape loader for the PDP-8.
It is initially stored in memory by way of the CONTROL console switches. Once
stored, it is considered to be a permenent occupant of locations 7756 through
7777 (absolute octal addresses) and care should be taken to keep it from being
destroyed.
2. A paper tape to be read in by the RIM Loader must be in RIM format:

Tape Channel
8 7 6 5 4

1 0 0 0 0
0 1
Al
A3
0 0
Xl
0 0
X3
0 0
Al
0 1
A3
0 0
Xl
0 0
X3
0 0

s3 2
0 0 0

Leader/Trailer code
Absolute address to
contain next 4 digits
Contents of previous
4 digit address

A2
A4
X2
X4
A2
A4
X2
X4

Address
Contents
(ETC.)
Leader/Trailer code

(ETC.)

0 0 0 0

0 0

3. The complete PDP-8 RIM Loader for the ASR-33 (SA=7756) is as fol lows:
Abs.
Addr.

Octal
Contents

7756,
7757,
7760
7761,
7762,
7763,
7764,
7765,
7766,
7767,
7770,
7771,
7772,
7773,
7774,
7775,
7776,
7777,

6032
beg,
6031
5357
6036
7106
7006
7510
5357
7006
6031
5367
6034
7420
3776
3376
5356
0
temp,
jmp start of bin loader

Instruction i z

Comments

kcc
ksf
jmp . -1
krb
c II r+ l
rtl
spa
jmp beg +1
rtl
ksf
jmp . -1
krs
snl
dca i temp
dca temp
jmp beg
0
0

/clear AC and flag
/skip if flag= l
/looking for char
/read buffer
/ch 8 in ACO
/checking for leader
/found leader
/OK, ch7 in link

/read, do not clear
/checking for address
/store contents
/store address
/next word
/temp storage

4. Placing the RIM Loader in memory by way of the CONTROL console switches
is accomplished as fol lows:
a. Set 7756 in the switch register (SR)
b.

Press LOAD ADDRESS

c. Set the first instruction in the SR (6032)
d.

Press DEPOSIT

e. Set the next instruction in the SR

Press DEPOSIT

g. Repeat steps e. and f. unti I a II 16 instructions have been deposited.

5. To load a tape in RIM format, place the tape in the reader, set the SR to
7756, press LOAD ADDRESS, press START, and start reader.
6. The complete PDP-8 RIM Loader for the high-speed reader 750 (SA=7755)
is as fol lows:
Abs.
Addr.

Octal
Contents

7755,
7756,
7757,
7760,
7761,
7762,
7763,
7764,
7765,
7766,
7767,
7770,
7771,
7772,
7773,
7774,
7775,
7776,
7777,

6014
6011
5356
7200
6012
7106
7006
7510
5355
7006
6014
6011
5370
6012
7420
3777
3377
5355
0

Tag

Instruction

Comments

beg,

rfc
rsf
jmp . -1
cla
rrb

/fetch char
/skip when flag = l

/OR buffer & AC

ell

tern,

rtl
rtl
spa
jmp beg
rtl
rfc
rsf
jmp . -1
rrb
snl
dca i tern
dca tern
jmp beg
0

/ch 8 in AC0
/check for leader
/found leader
/OK, ch 7 in I ink

/OR buffer & AC
/check for addr
/store contents
/store addr
/get next word
/temp storage

BINARY LOADER (BIN)
l. The BIN Loader is used to read in the machine language tapes. A binaryformatted tape is about one half the length of a comparable RIM formatted
tape.

It can, therefore, be read in about twice as fast as a RIM tape and is,

for this reason, the more desirable format to use with the l 0 cps ASR-33
Reader.
2. To load a tape in BIN format, place the tape in the reader, set the SR
to 7777; press LOAD ADDRESS, press START, and start reader.

3. After a Bl N has been read in, one of the two fol lowing conditions exist:
a. No checksum error: halt with AC=O.
b. Checksum error: halt with AC= (computer checksum) - (tape
checksum).

If a checksum error exists, a character was misread

from the binary tape or is mispunched on the tape. The operator
should reload the binary tape; and if the same checksum error appears in the AC indicator after readin, the binary tape was mispunched and a new copy should be obtained.

If a different check-

sum error appears after readin, the appropriate maintenance procedure should be fol lowed.

APPENDIX D
ASR-33 8-BIT CHARACTER SET

Character

8-Bit Code
(in Octal)

301
302
303
304
305
306
307
310
311
312
313
314
315
316
317
320
321
322
323
324
325
326
327
330
331
332

0
1
2
3
4
5
6
7
8
9

260
261
262
263
264
265
266
267
270
271

D
E
F
G
H

J
K

L
M
N

0
p
Q

u
v
w
x
y

Character

$
%
&

(
)
*
+

I
<
=

>
?
@

[

]
t

Leader/Trailer
Line-Feed
Carriage-Return
Space
Rub-out
Blank

8-Bit Code
(in Octal)
241
242
243
244
245
246
247
250
251
252
253
254
255
256
257
272
273
274
275
276
277
300
333
334
335
336
337
200*
212*
215
240
377*
000*

*Ignored by the operating system

APPENDIX E
OPERATING

PROCEDURES

COMPILER
1.

Load the compiler system with binary format loader (see Appendix C).

2. Put the starting address of the compiler (0200 octal) into the switch register and press LOAD ADDRESS.
3. Turn the tape teleprinter on and on line with the punch on.
4. Place the source language tape in the reader and turn on the reader.
5. Press START.
6. At the end of compilation, the computer will halt with the run light off.
7. To compile additional programs, place the source language tape in the
reader, turn the reader on, and press CONTINUE.

OBJECT TIME SYSTEM
1.

Load the FORTRAN object time system using the binary loader.

2. Place the compiler output (interpretive code) in the ASR-33. Turn the
ASR-33 on I ine. Turn on the reader.
3. Start at 200 (i.e. load the switch register with 200, press LOAD ADDRESS, and press START.
4. The object time system will now load the interpretive code. When the
tape has been read, the loader halts with the checksum difference in the AC.
A zero AC indicates a correct load.

5. Turn off the reader.
6. Press CONTINUE. The program wil I now be executed.
7. At the end of the program or at a stop statement execution, an exc lamation mark (!) is typed out.
8. To restart the program without reloading, start at 201.

APPENDIX F
FORMAT OF COMPILER OUTPUT

INTERPRETIVE CODE
1 . 200 codes (leader, ignored by loader)
2. Data blocks, each as fol lows:
a. origin (2 frames, first has bit 7 punched)
b. data words (2 frames/word)
3. Forward referencing table - first frame has bits 7 and 8 (only) punched
4. Checksum
a. first has bits 6, 7, and 8 (only) punched
b. next two frames are checksum
5. 200 codes (trailer, ignored by loader which stops after checksum)
6. Error comments, if any, in ASCII

mnmnoma

DIGITAL EQUIPMENT CORPORATION

5234

•

PRINTED IN U.S.A.

MAYNARD, MASSACHUSETTS

50·1/65