Digital PDFs

DEC-08-AFCO-D

March 1970

71 pages

Original

2.0MB

Document:	4K Fortran Pgmr Ref 197003
Order Number:	DEC-08-AFCO-D
Revision:
Pages:	71
Original Filename:	http://bitsavers.org/pdf/dec/pdp8/fortran/DEC-08-AFCO-D_4K_Fortran_Pgmr_Ref_197003.pdf

OCR Text

Digital Equipment Corporation
Maynard, Massachusetts

PDP-8
Programmer's Reference Manual

4K FORTRAN

DEC-08-AFCO-D

4K FORTRAN

PROGRAMMER'S
REFERENCE MANUAL

For additional copies specify Order No. DEC-08-AFCO-D to Program Library,
Digital Equipment Corporation, tv\aynard, tv\ass.
DIGITAL EQUIPMENT

Price: $2. 00

CORPORATION o MAYNARD, MASSACHUSETTS

1st Edition October 1966
2nd Edition Revised May 1967
3rd Edition Revised October 1967
4th Edition Revised May 1968
5th Edition January 1969
6th Edition May 1969
7th Edition March 1970

Your attention is invited to the last two pages of this manual.
The Reader's Comments page, when fi lied in and returned,
is beneficial to both you and DEC. All comments received
are considered when documenting subsequent manuals, and
when assistance is required, a knowledgeable DEC representative wi II contact you. The Software Information page
offers you a means of keeping up-to-date with DEC's software.

The material in this manual is for information purposes and is subject to change without notice.

The following are trademarks of Digital Equipment
Corporation, Maynard, Massachusetts:
PDP
FOCAL
COMPUTER LAB

DEC
FLIP CHIP
DIGITAL

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 Iibraries 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 deve Iopments for the other.

CONTENTS

CHAPTER 1
INTRODUCTION
CHAPTER 2
THE FORTRAN LANGUAGE
Statements

2-l

Program Format

2-l

Types of Statements

2-2

Comments.

2-2

Continuation

2-3

The Character Set

2-4
CHAPTER 3
FORTRAN ARITHMETIC

3-1

Arithmetic Expressions
Constants

3-2

Variables

3-2

Operators

3-3

3-4

Use of Parentheses

3-4

Functions
The Arithmetic Statement
CHAPTER 4
NUMBER REPRESENTATION AND VARIABLE TYPES
Integers and Floating Point Numbers

4-1

Fixed- and Floating-Point Representation

4-1

Types of Variables

4-2
CHAPTER 5
SUBSCRIPTED VARIABLES

5-1

Arrays
Subscripts

5-1

The Dimension Statement

5-1

iii

CONTENTS (Cont)

CHAPTER 6
PROGRAM CONTROL
Program Termination

6-1

END Statement

6-1

Branches and Loops

6-2

The GO TO Statement

6-2

Integer Summation

6-2

DO Loops

6-5

II legal DO Nesting

6-7

The CONTINUE Statement

6-7

Computed GO TO

6-8
CHAPTER 7
INPUT AND OUTPUT

Available Devices

7-1

Input and Output Statements

7-2

Device Selection and Direction of Transfer

7-2

Statement Number of Format Statements

7-3

List

7-3

Format Specifications Statement

7-4

Control Elements E and I

7-4

Input

7-5

Integer Values - Fixed Point - FORMAT (I)

7-5

Decimal Fraction Values - Floating Point - FORMAT (E)

7-5

Correcting Typing Errors

7-6

Output

7-6

Data Word Output - Floating and Fixed Modes FORlv\AT (E) and FORMAT (I)

7-6

7-7

Other Format Control Elements
11

Quote ( ) (Hollerith Output)

7-7

Slash (/)

7-8
CHAPTER 8
FORTRAN WITH DECTAPE OPTION

FORTRAN Compiler with DECtape 1/0 Option

8-1

CONTENTS (Cont)
Page
Use of Symbolprint with FORTRAN

8-1

FORTRAN Operating System with DECtape I/O Option

8-1

DECtape FORTRAN Statements and Operation

8-2

CHAPTER 9
PDP-5/8 FORTRAN SY MBO LPRI NT
APPENDIX A
OPERATING PROCEDURES
FOR RIM AND BIN PAPER TAPE LOADERS
APPENDIX B
PREPARATION OF SYMBOLIC (SOURCE) TAPE
APPENDIX C
FORTRAN OPERATING PROCEDURES
APPENDIX D
FORMAT OF COMPILER OUTPUT
APPENDIX E
ASR-33 8-BIT CHARACTER SET
APPENDIX F
PDP-8 FOR TRAN SOURCE PROGRAM RESTRICTIONS
APPENDIX G
DIAGNOSTICS
i LLUSTRATIO NS

A Fortran Program

2-1

Example of Comments

2-3

The Continued Statement

2-3

Number Representation

4-2

Indexing Statements

5-3

Schematic Representation of Program Branching

6-2

Integer Summation

6-3

Use of IF Statement in Integer Summation Problem

6-3

IF Statement with Substatement Feature

6-4

Fibonacci Series

6-4

Fibonacci Series Calculation Programmed As a DO Loop

6-5

Nested DO Loops

6-6

ILLUSTRATIONS (Cont)
Page
13

DO Loops

6-6

Program Branching in DO Loops

6-7

Input-Output Statement

7-2

A List Example

7-3

Examples of Quote and Slash

7-7

CHAPTER l
INTRODUCTION

Using a digital computer to solve a problem generally involves the fol lowing series of steps:
a. Determining the correct procedures to be used, including mathematical formulas, the
handiing of data, the presentation of resuits, etc.
b.

Arranging the procedures in the proper order.

Determining the sequence of computer instructions that wil I 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 listed 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 cal led 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 famiiiar 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. Non-mathematical operations are specified by English words. 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.

1-1

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 G). The programmer or operator should then take the appropriate corrective
measures and recompile the program. Note, however, that the compiler cannot detect all such errors,
and also that the diagnostic message may not accurately indicate the cause of the error but rather the
symptom.
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 ti me 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.

1-2

CHAPTER 2
THE FORTRAN LANGUAGE

STATEMENTS
Figurn 1 is cm 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
follow.

C;
5;
l O;

30;
32;
3·3;

THIS PROGRAM CALCULATES FACTORIALS
TYPE 200
ACCEPT 300 Ix
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
41;
200;
300;
400;

PAUSE
GO TO 5
FORMAT (/, "PLEASE TYPE A POSITIVE NUMBER", /)
FORMAT (E)
FORMAT (/,E, "FACTORIAL IS",E)
END

Figure l

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 l ).

2-1

The identification field extends from the left-hand margin up to and including a semicolon
character. This field may be left blank, or it may contain one of the fol lowing types of identification:
a. The first digit of a statement number. This number, which may be any positive integer
from l to 2047 inclusive, identifies the statement on that line 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.
b.
Comments).

The letter C. This identifies the remainder of the line as a comment (see Section

The semicolon 0) is necessary only if the statement is numbered or is a comment, (i.e., if
the identification field is blank, the semicolon may be omitted).
The statement field begins immediately after the semicolon and extends through the next carriage return. Although the continuation character ( 1 ) al lows a single statement to extend over two or
more lines, no more than one statement can be written on one line. Note that a comment cannot be
continued by use of the continuation character ( 1 ) .

TYPES OF STATEMENTS
FORTRAN statements are of several types with differing functions distinguished as fol lows:
a. Arithmetic statements resemble algebraic formulas. They specify the mathematical
operations to be performed.
b.

Program control statements direct the flow of the program.

c. Specification statements al locate data storage, determine variable and data types, and
specify input/output formats •
d•

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 every I ine intended as a comment, are not interpreted by the compiler and have no
effect on the executab Ie program •

2-2

C;
C;
C;

46;

CALCULATE PERCENTAGE OF CORRECT RESPONSES
PERCENTAGE= -1 IF THERE ARE NO ITEMS
IN CATEGORY
DO 47 1=1, 57
DO 48 J=l, 6
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 !ine 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.

l I·

2·,

3·,

5·,

Figure 3

FORMAT
IF (N R (l ) -1 ) 2 I 2 I 3
AP=-14.73
GOT06
IF (NR (l)-2)4, 4, 5
AP=-44. 19
GOTO 6
AP=SH (2)*3.0-AG(4)/'
AG (l)+SQTF (AG (14))

The Continued Statement

2-3

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: all letters and numbers - A through Z, 0 through 9; control
characters - semicolon (; }, carriage return (CR), line feed (LF), single quote (' ), double quote ( 11 ) ,
left parenthesis (, right parenthesis), period (.), comma (,);and the operators - plus (+), minus (-),
slash (/), asterisk (*), equa I sign (=). Al I other characters are ignored by the compiler except in
Hollerith information of FORMAT statements where al I 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.

2-4

CHAPTER 3
FORTRAN ARITHMETIC

ARITHMETIC EXPRESSIONS
An algebraic formula such as the fol lowing
[5a + 4b (x

- xo)] /2a sin 9

represents a relationship between literal symbols (a, b, x, x , 9) and constants (5, 4, 2) indicated by
0
mathematical functions and arithmetic signs (+, -, /, multiplication, exponentiation, sine). This same
formula can be written as a FORTRAN arithmetic expression with very I ittle 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 con-

stants, variables, and functions connected by operators.
The following examples demonstrate the properties of arithmetic expressions. Each expression
is shown with its corresponding algebraic form.

Algebraic Expression
ax

FORTRAN Expression

+bx+ c

(A**2 - B**2) I (A+B)** 2

2
-3-

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

4n r

3x n- 2 (x-+y)
4.25
A *SI NF (THT A)+2 .*A *COSF (THT A-0. 7853982)

a ·sin 9 + 2a • cos (9 -45)

2 -IX

2 .* SQTF(X)/3.

-3-

3-1

Constants
Constants are explicit numerical quantities. They may be integers, decimals, or numbers in
decimal exponent form. Some examples follow:
integers

-70

2047

3.14159

-0.00025

5
decimals
18.75

decimal exponent
l .66E-16

(meaning l .66 x l 0

-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

/l. variable name is composed cf one er mere characters accoiding to

the execution of the program..

these three rules:
a.

The only characters which may be used in a variable name are A through Zand 0

The first character must be alphabetic (i.e., A through Z).

through 9.

c. Only the first four characters of any variable name are meaningfu I. Al I characters after
the fourth are ignored by the compiler.
Some examples of acceptable variable names are:

XZRO

LST8

THTA

P51

The name EX IT represents one variable, not two.

DC8B

EPSL

XSUM
(Remember that blank spaces have no function in

FORTRAN.) Thus, EX IT, EXIT, or even EXIT, are identical names as far as the compiler is concerned
because they al I 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.

3-2

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 avai Iable. To i II ustrate 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 - x 0 )] /2a si n9
Normally, a FORTRAN expression is evaluated from the 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:

()

l st.

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 subtraction) is binary.

3-3

The left-to-right rule can now be stated more precisely as fol lows:
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 is
evaluated as A-(B*C), not (A-B)*C. The example below gives a few more illustrations of the left-toright rule.
The expression

is evaluated as
(A/B)*C
(A/B)/C
(A **B)**C

An easy way to check the proper pairing of parentheses is by counting out, i I lustrated in the
following example:
(Z+AM*(ZM+l .}}/((X**2+C**2)*P)
l
2
10 12
l 0
The procedure is this: Reading the expression from left to right, assign the number l to the first left
parenthesis (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 pairing is incorrect.

Use of Parentheses
Note the use of parentheses in the fol lowing example below. They are used to enclose 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(I+J) = (A+B)**C+SINF (X)
In algebra there are severa I devices, such as square brackets ( [] ) , rococo brackets ( {} ) ,
etc., for distinguishing between levels when subexpressions are nested. In FORTRAN, only the curved
parentheses are avaiiabie, so the programmer must be especiai iy carefui to make certain that parentheses are properly paired. In a given expression, the number of left parentheses 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.

3-4

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 function
occurs.
PDP-5/8 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 exampie, the expression:
LOGF(SINF(X/2)/COSF(X/2)) is equivalent to log*tan ;

•

The PDP-5/8 FORTRAN library currently consists of the following functions:
Function Name

Meaning

SQTF

(X)

square root of X

SINF

(X)

sine of X, where Xis expressed in radians

COSF

(X)

cosine of X, where Xis expressed in radians

ATNF

(X)

arc tangent X, where the angle is given in
radians

EXPF

(X)

exponential of X

LOGF

(X)

logarithm of X

THE ARITHMETIC ST ATEt'v~ENT
The arithmetic statement relates a variable V to an arithmetic expression Eby means of the
equal sign (=). Thus:

V=E
Such a statement looks I ike 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.

3-5

Th is means that the value of the left-hand variable wi 11 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)

PI = 3. 14159

d. THT A= 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 wi 11 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

lnterpretati on

Use

(Unary)

-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

(Binary)

Treated this way, the equal sign is considered to have the lowest binding strength of al I 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 following:
CPRM = (CKL - CKG) I (CPG = P*(O + 1.))

3-6

The i nterna ! arithmetic statement, CPG = P *(O + 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 followed 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 Din 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 B" 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
Only single level replacements wi 11 compile correctly in this manner. For example: ·Statements of the
type A(l) = A(2) = R(l) = 0.123 are not allowed and will not compile properly.
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

3-7

CHAPTER 4
NUMBER REPRESENTATION AND VARIABLE 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 fol lows:
a. Integers are positive or negative numbers written without a decimai point. I hese 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 211 -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 integral results.
b. Floating point numbers have two forms. They are simple decimals, such as 0.0025, .4,
57., 2.71828; or numbers in decimal exponent form. Numbers in decimal exponent form are simple
decimals multiplied by a power of 10. Numbers in decimal exponent form must have an explicit decimal
point.
Exampies:
Mathematical Form

FORTRAN Form

6.023 x 10

6.023E23

1 .66 x 10- 16

l.66E-l 6

72. x 10

72.E12

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 Nx 10K. The fol lowing are floating point representations of the number 19:
19.0
• 19E2

1. 9E+l
1900.E-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 the
in 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

4-1

rightmost digit. A FORTRAN integer may not exceed the range of -2047 through +2047. All integers
greater than ±2-47 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 2s complement form.

SIGN

I I

MAGNITUDE

0 1

a. FORTRAN INTEGER

SIGN OF EXPONENT

I !

EXPONENT

MANTISSA

1
SIGN OF
0 1
MANTISSA----,
2

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 conventions
as fol lows:
a.

Integer variable names must begin with one of the letters I, J, K, L, M, or N.

Floating point variables are designated by names beginning with any other ietter.

These are integer variable names: INDX, KDTA, M359. These are floating-point variable
names: ZXRO, CONT, FICA.

4-2

Integers cannot appear in floating point expressions except as exponents or subscripts. Some
examples of illegal and legal expressions are as follows:
Expression

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

---

Legal

Mode

Yes
Yes
No
No
Yes
Yes
Yes
No

Floating
Fixed

4-3

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 I ist, and a
matrix are all arrays. In mathematics, an element of an array is referenced by means of a symbol denoting the array and subscripts identifying the position of the element. For example, the sixth element
in a :ector 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 fol lowing 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 al I the elements in
the array.
The example below gives a few illustrations of the use of subscripted variables.
a. X(I+L)=X(I)+ALPH(l)*P(I)
b. X(I+3)=X(I+2)+X(I+l )/2.
c. C(J)=A(I*J+3)
d. A=B(6)

Subscripts

As the example obove 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:
a.

Which are the subscripted variables?

b. What is the maximum dimension of the subscript? (When an array is used, a certain
amount of storage space must be set aside for its elements; hence this requirement.)

5-1

Al I the above information is provided by the fol lowing specification statement:
DIMENSION A(20), 8(15}
where A and Bare array names, and the integer constants 20 and 15 are the maximum dimensions of
each subscript.
The rules governing the use of array names and the dimension statement are as follows:
Al I 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 an array appears before any statement which references the array.
DIMENSION LIST(30), MAT(lOO}, REGR(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, REGR, 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 ore set aside for the array.
DIMENSION 8(30), 1(15)
This version of the PDP-5/8 FORTRAN does not hove 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.
In this example the matrices are stored column wise in memory, that is, sequential locations
in memory are used as fol lows:
Relative Position
in Memory (INDX)

Element

1
2

all
021
031
041
051
a61

3
4
5
6
7

012
022

056
a66

35
36

If referencing element a 56 in the array, M=S, 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 a22, INDX=2+6*1=8.

5-2

2;
10;

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=1, I
DO 20 N=l, I
INDX=M+I*(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+I *(M-1)
IA=K +I *(M-1)
IB=N+I *(K-1)
C (IC)=C (IC)+A(IA)*B(IB)
CONTINUE
TYPE 15
DO 40 M=l, I
TYPE 21
DO 40 N=l, I
I NDX=N+I *(M-1)
TYPE 2, C(INDX)
CONTINUE
FORMAT (/)
TYPE 15
END

Figure 5

Indexing Statements

5-3

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 statements
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 allow the programmer to alter the
sequence, repeat sections, suspend operations, or bring the program to a complete ha It.

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 stopping point.

Such terminations are indicated by the statement:
STOP
This causes a final, complete halt; no further computation is possible; although the program may be
completely restarted from the beginning.
When a STOP is encountered during program execution at object time, the system signifies
that a stop has occurred by outputting an exclamation point(!) to the Teletype or high speed punch,
whichever is being used as the output device.
The stop statement prevents further computation after it has been executed. There is a way,
however, to suspend operation for a time and then restart the program manua Ily. Th is 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.

END Statement
END occurs alone on a line and indicates the physical end of the program to the FORTRAN
compiler. It must be followed by carriage return and line feed.
statement.

6-1

Every program must contain an END

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).

•
Figure 6

Schematic Representation of Program Branching

All of these branches can be performed in several ways, the simplest of which is by using the
statement:
GO TON
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 following example, 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

6-2

statements preceding th is loop provide the starting conditions; th is is cal led 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.

SUM OF FIRST N INTEGERS BY ITERATION
KSUM=O
INUM=1
KSUM=INUM+KSUM
INUM=INUM+1

GO iO 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 fuifiiis both requirements. It has the following form:
IF (E)K, L,M
where E is any variable name, arithmetic express ion, or arithmetic statement, and K, L, and M are
statement numbers. The statement is interpreted in th is 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.

Figure 8

SUM OF THE FIRST 50 INTEGERS
KSUM=O
INUM=1
KSUM=INUM+KSUM
INUM=INUM+1
IF (INUM-50) 2,2,3
STOP

Use of IF Statement in Integer Summation Problem

6-3

In this example, the initialization and main loop are the same as for the preceding example
except that the GO TO statement of 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 3-6 and 3-7).

SUM OF THE FIRST 50 INTEGERS
KSUM=O
INUM=50
KSUM =INUM +KSUM
IF(INUM=INUM-1) 3,3,2
STOP

2;
3;

Figure 9

IF Statement with Substatement Feature

In th is example, the sum is formed by counting down, but the same resu Its are achieved.

The initialization is changed so that !NU!v\. starts with the value of 50 instead of 0, and the statement
INUM=INUM+l is no longer required.
A loop may also be used to compute a series of values. The fol lowing 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-1 )+ FIB(K-2)
K=K+1
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.

6-4

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 ca lcu lat ion stops.
DO Loops
Iterative procedures such as the loop in the example above are so common that a more concise way of presenting them is warranted. In this example, three statements are required to initialize
the subscript, increment it, and test for termination. The following type of statement combines all
these functions:
DO n I =Kl , K2, K3
Here, n is a statement number, I is a simple (non-subscripted) integer variable, and Kl, K2, and K3
are simple integer variables or integer constants which provide, in order, the initial value to which
I set, the maximum value of I for which the loop will still be executed, and the amount by which I is
incremented at each return to the beginning of the loop. If K3 is omitted from the statement, it is assumed equal to one. Statement n must be a CONTINUE statement.

Figure 11

FIBONACCI SERIES, 100 TERMS
DIMENSION FIB(lOO)
FIB(l) = l
FIB(2) = l
DO 5 K=3, 100
FIB(K)=FIB(K- l)+FIB(K-2)
CONTINUE
STOP

r•
r
I
I
,•
n
A
r""-1"""'\. I
rmonacct• ~er1es
\....a1cu1ar1on
rrogrammeaI 1-\S
a LIV
Loop

r-•I

In words, the DO statement says: Do all statements through statement 5 for K=3, when statement 5 is
encountered. Perform the following test: If K+ 1 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. In this example this is a STOP.
DO loops are commonly used in computations with subscripted variables. In these cases,
it is usually necessary to perform loops within loops. Such nesting of loops is permitted in FORTRAN.

6-5

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

5;
10;

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 B and 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. The algebraic expression for the loop is as follows:

2
A. = X. - C.
I

for i l, 2, 3, .•• 20

where
40

x. =L

(b.-z.)
• "'
! !
1=L

for j = 2, 4, 6, .•• 40

The following general rules about DO loops must be observed.
a. 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.

IF,~
I I I

Figure 13

DO Loops

6-6

b. 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:
.

(1)

Incrementing and testing start with the present value of I.

(2)

Controi was originai iy transferred out of the DO other than by compieting it.

A DO loop must end on a CONTINUE statement.

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 1.

II legal DO Nesting
DO 10 I=l, 20
DO 20 J=l, 100 ,2
SUM=(X(I)-Y(I))**2
CONTINUE
Z(J)=SUM+A(J)
CONTINUE

10;

I
20;

Figure 14

Program Branching in DO Loops

Branches 2, 5, 6, and 7 are permitted; branches l, 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

6-7

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 on a 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 variable, 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 IV AR=7.
When IVAR is less than one or greater than seven, the next sequent ia I statement after the
GO TO is executed.

6-8

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:
a. 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.
b. Which device is being used? Information can be transferred between core and either
of two different input/output devices; each I/O operation must specify which device is involved.
c. 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.
d. 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.
e. 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 fast two items are determined by the FORMAT statement.
PDP-5/8 FORTRAN provides for communication of data to and from a program in the fol lowing
ways.
a.

ASCII Coded Data - (Appendix E)

The Teletype can be used to transfer data to the program either via the keyboard on
which the user types the data, or from previously punched paper tape read via the teletype reader.
Data can be output from a program to the Teletype, producing a printed copy with or
without the corresponding punched paper tape (depending on whether or not the punch is turned on).
tape.

The high-speed reader and punch can also be used for data transfer via punched paper
No printed copy is made when output is to the high-speed punch.
b.

BINARY Coded Data

DECtape can also be used for data transfer in which case the data is stored as a core
image on tape in 128 word blocks of 12-bit binary words. Integers are read and written as single 12bit words, floating point numbers as three words. Alphanumeric information is transmitted as 8-b it
ASCII coded characters right justified in 12-bit words (one character per word).

7-1

INPUT/OUTPUT STATEMENTS
The input/output statements contro I th is transfer of information. As illustrated in Figure 15,
I/O 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 list of
names of the variables whose values are to be output or changed by new inputs.

N ,_y(I), {(I+1) ,

V(I +2}.1

List 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 Teletype and the PDP-5/8.
ACCEPT causes information to be accepted into core memory from either the Teletype papertape reader, the keyboard, or the photo-electric reader, depending on a switch option selected at run
time.
TYPE causes information to be transferred from core memory to the Teletype printer, or the
printer and paper tape punch depending on whether the punch is activated or not, or to the high-speed
punch depending on a switch option selected at run time.
READ causes information to be read into core memory from DECtape. (See Chapter 8 for
details.)
WRITE causes information to be written on DECtape from core memory.
;READ
;WRITE

UNIT I
UNIT I

BLOCK,
BLOCK,

FORMAT,
FORMAT,

LIST
LIST

where UNIT and BLOCK specify the DECtape unit to be used, and the position of information on tape
respectively (UNIT and BLOCK can be either integer constants or simple integer variables). The
balance of these statements is exactly analogous to the corresponding information in a TYPE statement.
FORMAT specifies the format statement, and LIST specifies the variables to be written from, or read
into core.
Bit 0 of the SWITCH REGISTER must be set to l (up) when compiling or running a program
containing READ and WRITE statements and 0 (down) otherwise.
cause error diagnostics. (See Appendix G.)

7-2

Failure to properly set the switch will

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:
ACCEPT 10, A
FORMAT (E)

10;

Every I/O statement must have a reference to a FORMAT statement.

List
The final item of specification in the I/O statement is the LIST of variables. This is a
sequential list of the names of the variables and array elements whose values are to be transferred in
the order transfer. There is no restriction on the number of names which may appear in the I ist of an
I/O statement as long as the tota I statement length does not exceed 128 characters. The modes of the
variables named need not agree with the corresponding FORMAT statement, however, the modes spec ified in the FORMAT statement take precedence.
Example:
TYPE 23, A, J, KAL, BOB
23;

(where A=3. 2 J=27 KAL=302 and BOB=7. 58)

F0 RMA T (I , E, I , E)

The decimal portion of A will be dropped and the 3 typed as an integer; the value of J will
be typed as a normalized number; KAL wi II be typed as an integer; and BOB as a normalized number.
The output will look I ike the following:

+0.2700000E+02

302

+o.7580000E+ol

Array names included in 1/0 lists must be subscripted in one of the following forms: A ,
v

, A
or A where A is the array name, v is a simple integer variable and c is an integer constant.
v+c
v-c
c

10;

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

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

TY pE l 0, A I J I B, C( N)

Figure 16

A List Example

7-3

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 I ist.
For instance in the preceding example when the value of the variable B is typed in the E format, the control character slash (/) causes a carriage-return I ine-feed to occur. Then the FORMAT
statement is reinitialized, and the array element C(I+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. If
there are fewer names in the I ist than there are elements in the FORMAT statement, the program completes the I/O operation and proceeds to the next sequential FORTRAN statement. If this next statement is another I/O 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 a 11 of their elements are exhausted.

FORMAT SPECIFICATIONS STATEMENT

As already mentioned in the previous description of input/output statements, the FORMAT
<:totPmPnt
- ·
· - · · · - · · · rontrol"
- - · · · · - · - thP
· · · - orrnnoP.mP.nt
-· · · -- · ·o - · · · - · · · oncl
-- · · -- mnclP.
· · · - -- - nf
- · thP.
· · · - informotion
· · · - · · · · · · - · · heino
- - · ·v transfP.rred.
- - - ·- - - - - ~

The
the·
- values
- - - ·- of
-

names appearing in the I ist of the I/O 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(/) and quote("). The set of elements must be enclosed in parentheses and separated by commas.
Example:
FORMAT (E,I,/, "HOLLERITH")

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:
10;

TYPE 10, A
FORMAT (I)

The variable A is typed as an integer and the fractional part of A is truncated.
if A has a value of 14.96, only the integer part, 14, would be typed.
less than one, zero would be typed.

7-4

For instance,

If A has an absolute value of

input data words consist of a sign, the decimal value, an exponent value if the data is
fioating point, and a fie id 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 rubout. 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 rubout.
These input words can be transferred into core memory from either the Teletype paper-tape

reader: the keyboard 1 the photo-electric reader 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 follows:
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 'Nord 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 1s represent characters in the data word and n represents the power of ten of the ex··
ponent. 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 dee ima I character.

If the exponent is omitted, the power of ten is taken as zero.

*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.

7-5

Examples of floating-point values are as follows:
Va Iues Accepted

Typed Numbers

16.

0. 16 x 10

. l6E02

0. 16 x 10

1600.E-02

0. 16 x 10

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 ca nee I or erase a word, type a rubout 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 rubout 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.

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

-1043
16
+ 16

+0016
Floating-point values ore printed as per example
SO . dddddddsxxxx
where

S is the sign, minus sign, or space
d is the seven decimal digits of the data word
s is the sign of the exponent value
xxxx is the exponent value

7-6

Decimal Value

Output Format

-8,388,608.0

-0.8388608E+o7

+. 000000 l 192092

0. ll 92092E-06

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 FORlv\AT control elements quote ( 11 ) and slash(/).

Quote ( 11 ) (Hollerith Output)
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 all other elements of that
FORMAT statement previous to the text have been used.

10;

TYPE 10
FORMAT (/,

THIS IS HOLLERITH",/)

TYPE l 00, AMIN, AMAX
100;

210;

200;
250;

FORMAT(/, 11 MINIMUM:::: 11 ,E,/, 11 MAXIMUM:::: 11 ,E,/)
TYPE 210
11
FORMAT(/,/,
CUMULATIVE DISTRIBUTION 11 , / , / 1
II
INCREMENTS
FREQUENCY 11 , / )
DO 220 K::::l, 100
TYPE 250, K,VALU(L). VALU(K+l), COUNT(K)
CONTINUE
11
11
11
11
11
FORMAT (I, 11
,E,
,E,
,E,/)

Figure 17

Examples of Quote and Slash

All legal Teletype characters can be contained within quotes and are output as text (Appendix D).

7-7

If a statement continues on another I ine the Hollerith field must be ended before typing the
continuation character ('); it may be re-opened on the next line. Before text is output, the elements
of the FORMAT statement that appear in front of the Hollerith information must have been used.
Example:
10;

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

In th is example, the text is not typed because one of the E elements was not used.

If a FORMAT statement containing a Hollerith field is referenced by an ACCEPT statement,
the original text will be replaced character by character until the closing quote mark is encountered.
All ASCII characters except I ine-feed wi II be read. Because of the variable number of spaces following
numeric data printed in either E or I format, precautions must be taken when text and data are intermixed in a punched tape which is to be read by a FORTRAN program. The technique has been used
successfu I ly, however.

If a FORMAT statement containing a Hollerith field is referenced by a READ statement, the
original text will be replaced character by character with 12-bit words from the DECtape. Because
the E and I format for DECtape 1/0 are fixed lengths, there is essentially no difficulty here. Note that
it is possible to read text from DECtape with an I format in order to manipulate the numeric equivalent
of the ASCII character.

Slash (/)
The slash character is used for typing a carriage return and line 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:
10;

TYPE 10, A, B
FORMAT (/,/,/,E,/,/,E,/,/)

Three carriage-return line-feeds will be typed before the value of A; then two carriage-return linefeeds w i II be typed before and after the va Iue of B is typed.
The input subroutine of the object time system ignores all non-numeric characters except as
data word delimiters so that input can be labeled and spaced in intermixing the appropriate text and
carriage-return I ine-feeds with the data.

7-8

CHAPTER 8
FORTRAN WITH DECTAPE OPTION

PDP-8 FORTRAN includes provisions for storage of data on DECtape. In this way FORTRAN
programs requiring large amounts of accessible data may be readily written to run on a 4K PDP-5 or
PDP-8.
The standard I ibrary version of the FORTRAN Compiler is written such that the type of DECtape hardware to be used is irrelevant. The standard I ibrary version of the FORTRAN Operating System
is written to handle the TCOl DECtape Control with TU55 Tape Transports since these are more common.
There is, however, an overlay tape to convert the OP SYS to do the same operations using the 552 Control with 555 DECtape Transports. If this overlay is required the user should so specify when making
requests to the I ibrary.

FORTRAN Compiler with DECtape 1/0 Option
When the FORTRAN Compiler is read into core, it is equipped with a switch option governing the compilation of DECtape 1/0 statements (READ and WRITE). If the user wishes to compile a
program containing DECtape 1/0 statements, he must set Switch Register bit 0 to l (up) before starting
any compilation.
The Compiler is designed so that the space occupied by the processing routines for this option
becomes part of the input statement buffer if SR bit 0 is set to 0. This means that the DECtape 1/0
processing routines are destroyed if any compilation is done with bit 0 set to 0 and the Compiler must
be reloaded into core to regain the option.
Any program containing DECtape I/O statements must limit the length of the source statements to l 00 characters per statement.

Use of Symbolprint with FORTRAN
Symbolprint destroys a portion of the DECtape Compiler in core. The compiler must be reloaded if it is to be used to accept a symbolic program containing DECtape 1/0 statements.

FORTRAN Operating System with DECtape 1/0 Option
When the FORTRAN Operating System is read into core, it is equipped with a switch option
governing the execution of DECtape 1/0 statements. If the user wishes to run a program containing
such statements he must set Switch Register bit 0 to l before running his program.

8-1

There is a further condition which must be observed, since DECtape 1/0 requires a
considerable amount of additional processing routines.

Like the Compiler, the OP SYS destroys its

DECtape handling routines if it is used with SR bit 0 set to 0, thus gaining extra space. This requires,
however, that the OP SYS must be reloaded into core to regain the option.

DECtape FORTRAN Statements and Operation
When using DECtape, the FORTRAN Operating System contains a buffer area which is defined as a page of memory reserved to handle transfers to and from a block of DECtape (128

data

words).
The DECtape routines transfer one ful I block from tape to one page of core (and vice versa),
therefore, even if the block contains only one data word, the whole block wi 11 be read into the OP SYS
buffer, overlaying whatever had been there.
To store variables or arrays of data on DECtape requires two steps:
a. From the locations assigned by the OP SYS to the variables or al lotted to the arrays by
a DIME NS ION statement, the programmer must collect the data and put it in the OP SYS buffer. This
is done with pseudo WRITE statements (in a DO loop in the case of arrays). (See below.)
b.
statement.

He must write the buffer onto a block of DECtape. Th is is done by a phys ica I WRITE

(See below.) The programmer must be aware of how much data he has in the buffer and

write it out on DECtape, before he overflows the buffer. Overflow wil I cause an error diagnostic.
To retrieve data from DECtape is also a multiple operation.
a.
to retrieve.
b.

The programmer is responsible for remembering which block contains the data he wishes
He must read this block into the OP SYS buffer using a physical READ.

(See below.)

c. He must remember in which order he stored the variables or arrays and reference them here
in the same order.
d. He must disperse the data from the buffer to the locations assigned by the OP SYS or
allotted by a DIMENSION statement. This is done by pseudo READ (in a DO loop for arrays). (See
below.)
NOTE
Data which has been brought from tape into the buffer is
not yet available for use within the program. It must be
dispersed first.

Pseudo WRITE and pseudo READ statements operate between the user program and the buffer
only. They are used to collect into the buffer data from within the user program and to disperse into
the user program data in the buffer. They hove no effect on the phys ica I DECtape.

8-2

The user specifies pseudo READ or WRITE by specifying UNIT 0 and BLOCK 0 in the READ
or WRITE statement. Specifying any unit other than 0 wil I indicate that the user wishes to read from
DECtape into the buffer or write the buffer out on tape. Pseudo READ and WRITE are of the form;
READ 0 I 0 I FORMAT f LIST
WRITE 0, 0, FORMAT, LIST
Physical READ and physical WRITE statements operate between the buffer of the OP SYS and
the DECtape. They cause the actual reading or writing of tape. The user specifies physical READ or
WRITE by specifying a UNIT number from i-7 and the number of the actual block on which the data
has been stored or is to be stored. They are of the form;
READ UNIT, BLOCK, FORMAT, LIST
WRITE UNIT, BLOCK, FORMAT, LIST
It is not necessary to specify a list on a physical READ or WRITE but it is advisable, since it does no

harm and is an aid to remembering which variables in which order are on which block.
Two examples follow which demonstrate the storage and retrieval of data.

(

DIMENSION iDAT (i28) '"'\
11 O; FORMAT (I)
\......
Part l ~ DO 100 I= l , J
(
WRITE 0, 0, 110, IDAT (I)
'- 100; CO~TINUE
_)

Part 2 {WRITE MU, MBLK, 110

J has been previously defined~ 128
This DO loop will WRITE J number
of elements of IDAT into the buffer

Th is statement wi 11 then store data
from buffer to tape

}

IDAT

BUFFER

C·,
C·,
C·,
C;
C;
C;

.............
MUNT=TAPE UNIT TO BE SELECTED
MBLK=BLOCK TO BE WRITTEN ON
IBLK=INITIAL BLOCK TO BE SEARCHED FOR (TO REWIND TAPE)
VALUES FOR J AND K AS DETERMINED BY USER MUST BE LESS THAN 200

.............
•

•

8-3

DIMENSION IBFl (200), IBF2 (200)
90; FORMAT (I)
199; ACCEPT 90, J, K, MUNT, MBLK, IBLK
DO 10 I=l, J
ACCEPT 90, IBFI (I)
lO;CONTINUE

C·,
C·,
C·,
C;
C;

.............
.......
DATA HAS BEEN ACCEPTED NOW WE READ BLOCK 10 1
THIS INITIALIZES AND REWINDS THE TAPE
A GOOD IDEA TO DO THIS BUT NOT ESSENTIAL

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

READ MUNT, IBLK, 90

C;
C·,
C·,
C·
C·,
C;
I

.............
.
.......
TIME TO TAKE ACCEPTED DATA AND WRITE IT INTO
THE BUFFER, NOT ONTO THE ACTUAL TAPE. SEE PART ONE OF DIAGRAM.
BUFFER IS WRITTEN BY SELECTING UNIT 10 1
DO LOOP NEEDED SINCE DATA IS IN AN ARRAY

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

DO 31 I=l ,J
WRITE 0 1 0 I 90 I IBFl (I)
31;CONTINUE

C;
C·,
C·
C·,
I

.............
.
.......
TAPE IS NOW PHYSICALLY WRITTEN BY SELECTING
A LOGICAL UNIT, OTHER THAN ZERO. SEE PART TWO OF DIAGRAM.

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

WRITE MUNT I MNLK I 90

C·
C;
C·
I

.............
NOW READ BLOCK BACK INTO BUFFER
.............
.............

READ MUNT, MBLK,90
I

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

READ BUFFER AND STORE IT IN ANOTHER ARRAY
DO LOOP NEEDED BECAUSE OF ARRAY

C·
C;
C·
C;

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

DO 32 1=1,K
READ 0 I 0 I 90 I IBF2(1)
32; CONTINUE

C·
C;
C·
C·,
C·
I

*************

ALL 1/0 DONE HALT.

****** * ******

8-4

DO 13 I=l ,K
TYPE 91,IBF2 (I)
91 ;FORMAT (/,I)
13;CONTINUE
PAUSE

C;
C;
C·
I

.............
.
.......
HALT, THEN DO IT ALL AGAIN IF USER PRESSES CONTINUE ON CONSOLE
.............
.
......

GO TO 199
END

8-5

CHAPTER 9
PDP-5/8 FORTRAN SYMBOLPRINT

FORTRAN Symbolprint is a useful aid in finding where a FORTRAN program is stored in
interpretive memory, the exact memory locations assigned to each FORTRAN variable, and the amount
and location of interpretive core memory that is not used by a FORTRAN program.
Symbolprint loads over the FORTRAN Compiler and starts at address 600. The following is a
typical example of the typeout.

List of Variable Names

Assigned Location

HW
TB
G
TF
MC
DSR
Cl
C2
C3
C4

7546
7543
7540
7535
7534
7531
7526
7515
7504
7470

6312

7241

Note that a single word only has been assigned for the fixed point variable MC.
The last two octal constants typed indicate respectively the highest address used by the
program in interpretive memory and the lowest address used for data. Therefore the area of core between these two addresses is available. In the example there are
7241 -6312 - 1 = 726
octal locations free.
A machine language subroutine may occupy this available space. Use the FORTRAN PAUSE
statement to link the FORTRAN program to the subroutine.

If PAUSE is followed by a number (decimal), FORTRAN compiles (in effect) JMS to that
address. For example:
;PAUSE 3328

9-1

effects JMS 6400.

Location 6400 should contain coding such as the following:
SUBR, 0
JMP I SUBR

constituting the desired machine language program.

9-2

APPENDIX A
OPERATING PROCEDURES
FOR RIM AND BIN PAPER TAPE LOADERS

READ-IN-MODE LOADER (RIM)

l. 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 permanent 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 s3 2 l
l 0 0 0 0 . 0 0 0
0 l
Al •
A2
0 0
A3 •
A4
0 0
Xl .
X2
0 0
X3 •
X4
0 l
A3 •
A4
00
A3.
A4
0 0
Xl .
X2
0 0
X3 •
X4
(ETC.)
1 0 0 0 0 • 0 0

Leader/Trailer code
Absolute address to
contain next 4 digits
Contents of previous
4 dig it address
Address
Contents
(ETC.)
Lead/Trailer code

The complete PDP-8 RIM Loader for the ASR33 (SA 7756) is as fol lows:

Abs.
Addr.

Octal
Contents

7756,
7757,
7760
7761
7762,
7763,
7764,
7765,
7766,
7767,
7770,
7771,
7772,
7773,

6032
6031
5357
6036
7106
7006
7510
5357
7006
6031
5367
6034
7420
3776

Symbolic
BEG,

KCC
KSF
JMP .-1
KRB
CLL RTL
RTL
SPA
JMP BEG +1
RTL
KSF
JMP .-1
KRS
SNL
DCA I TEMP

A-1

/clear AC and flag
/skip if flag= 1
/looking for char
/read buffer
/ch 8 in ACO
/checking for leader
/found leader
/OK, ch? in link

/read, do not clear
/checking for address
/store contents

Abs.

Adck.

Octal
Contents

7774,
7775,
7776,
7777,

3376
5356
0
5301

Symbolic

TEMP,

DCA TEMP
JMP BEG
0
0

/store address
/next word
/temp storage
/jump to start of bin loader

Placing the RIM Loader in memory by way of the console switch register is accomplished

as follows:
a.

Set 7756 in the switch register {SR)

Press LOAD ADDRESS

Set the first instruction in the SR {6032)

Press DEPOSIT

Set the next instruction in the SR

Press DEPOSIT

Repeat steps e and f until all 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 follows:

Abs.
Addr.

Octal
Contents

7756

6014

7757
7760
7761

6011
5357
6016

RSF
JMP .-1
RRB RFC

7762
7763
7764
7765
7766
7767
7770
7771
7772
7773
7774
7775
7776
7777

7106
7006
7510
5374
7006
6011
5367
6016
7420
3776
3376
5357
0000
5301

CLL RTL
RTL
SPA
JMP TEMP-2
RTL
RSF
JMP .-1
RRB RFC
SNL
DCA I TEMP
DCA TEMP
JMP BEG +l
0
0

Symbolic
BEG,

TEMP,

RFC

A-2

/clear flag and fetch char. into
buffer
/skip when flag=l
/read buffer into AC, get next
char. into buffer
/rotate channel 8 into
/AC bit 0
/is it leader
/yes c Iear AC
/NO rotate channel 7 to LINK

/link set=origin
/store data
/store address
/next word
/temporary storage
/JMP to start of BIN Loader

BINARY LOADER (BIN)
1. The BIN Loader is used to read in the machine language tapes. A binary-formatted 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 Rrtv·\ tape and is, for th is reason, the more des i rob Ie format to use with the 10 cps
ASR33 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 BIN 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 checksum error appears after readin, the appropriate
maintenance procedure shou Id be fol lowed.

A-3

APPENDIX B
PREPARATION OF SYMBOLIC (SOURCE) TAPE

l. Symbolic tape preparation using Symbolic Tape Editor. {It is to the user's benefit to
use the Editor to put his source program on tape since using the Editor minimizes the chance of extraneous characters getting on the tape and also foci! itates deletion and correction of statements).
a.

Load Symbolic Tape Editor using Binary Loader.

Start Editor at 176 {Load Address, Start).

Type

Hold the CTRL key and press the FORM key; the bell wi II sound.

Type P

Ap and type the symbolic source program.
tJ , fo II owed by F) when punching stops.
NOTE

For complete explanation of Editor, see Symbolic Tape
Editor Manual.
2.

Symbol tape preparation off-I ine using Teletype only.
a.

Turn power on in computer (key on left in PDP-8, switch on right in PDP-5).

b. Turn Teletype LINE-OFF-LOCAL knob to LOCAL to disconnect Teletype from
computer.

Press PUNCH ON button on the Teletype.

Generate leader.*

Type the source program.

Generate trailer.*

Manual symbolic tape editing using the ASR33.
a. An incorrect character might be typed while preparing the symbolic tape: (the error is
detected N characters after typing the incorrect character) press the PUNCH B.SP.
button N+ 1 .times, press rubout N+ 1 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 procedures 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 duplicate 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 duplicate the remainder of the tape.
*To generate leader/trailer (200 code), hold the CTRL and SHIFT keys with the left hand, depress the
REPT key and then the P with the right hand. Release in reverse order or a P will be punched on the
tape.

B-1

APPENDIX C
FORTRAN OPERATING PROCEDURES

COMPILER
1.

Load the Compiler with the Binary Loader (see Appendix A).

2. Put the starting address of the Compiler (0200 octa I) into the switch register and press
LOAD ADDRESS.
3.

Set 1/0 switches. (Conditional) See 1/0 Control.

Place the source language tape in the selected reader and tum on the reader and punch.

Press ST ART.

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 appropriate
reader, turn the reader and punch on, and press CONTINUE. 1/0 selections cannot be changed without reloading compiler.

SYMBOLPRINT
Symbolpr!nt is run immediately after compi ! ing a program and before comp ii ing anothei Oi
loading the Operating System. (It cannot be run if the Operating System has been loaded into core.)
Use of Symbolprint destroys the portion of the Compiler which processes DECtape READ and
WRITE statements. The Compiler must therefore be reloaded if it is to compile a source program containing such statements.
1.

Load Symbolprint with the Binary Loader.

Set 0600 in the switch register.

Press LOAD ADDRESS and ST ART, see Chapter 8.

OPERATING SYSTEM
(OBJECT TIME SYSTEM)
1.

To load a compiled program:
a. Load the FORTRAN Operating System using the Binary Loader. If using 552 DECtape, load the 552 overlay tape using the Binary Loader.
b. Place the Compiler output (interpretive code object tape) in the Teletype or photoelectric reader. Turn on the reader, making sure ASR33 is ON LINE.
c.

Load the SWITCH REGISTER with 0200, and press LOAD ADDRESS.

d. Set switch register bit 1 to read in compiled tape from photo-electric reader or
Teletype (as shown below).
e.

Press START. The Operating System reads the compiler output tape.

C-1

f. The Operating System halts at the end of loading. The loading is correct if the
checksum difference which appears in the AC equals O.
g.
2.

Turn off the reader and remove the compiler output tape from it.

To execute a program after loading
a.

Set SWITCH REGISTER bits 0, l, and 2 {as shown below).

b. If input is to be from paper tape, put the data tape in the appropriate reader and
turn the reader on. If output is to be punched, turn punch on.
c.

Press CONTINUE.

NOTE

l . Once Icoded, a program can be executed any number
of times.

2. The Operating System need not be reloaded to run
more than one program in succession. To do so start at
step b of Section l.

To Re-execute a Loaded Program
a.

Set the SWITCH REGISTER to 0201; press the LOAD ADDRESS key.

Set the SWITCH REGISTER for the 1/0 (as shown below).

Press START.

The FORTRAN Operating System checks each of its internal stacks after the execution of
each interpretive instruction to insure that there is neither stack overflow nor stack underflow. If a
FORTRAN program has been debugged and is known to operate correctly, this test may be NOPed by
changing C(0404) to 7000 (NOP). This wil I speed up the execution of the program by a factor of
about 2.

1/0 CONTROL
The selection of 1/0 devices for both COMPILER and OP SYS is controlled by setting the
switches as shown below:

Bit
Number

Switch
Position

Meaning
The program contains only paper tape l/O statements.
The program contains DECtape 1/0 statements.

Compiler: Use the Teletype reader for input of source tape.
OP SYS: Use the Teletype reader for loading the object
program and the keyboard for ACCEPT statements.

C-2

Bit
Number

Switch
Position

Meaning

Use the high-speed reader.
2

Compiler: Use the Teletype printer/punch for compiler output (interpretive code) tape and error diagnostics.
OP SYS: Use the Teletype printer/punch for TYPE statements.
Use the high-speed punch (error diagnostics still come on
Teletype).

C-3

APPENDIX D
FORMAT OF COMPILER OUTPUT

INTERPRETIVE CODE

200 codes (leader, ignored by loader).

Data blocks, each as follows:
a.

Origin (2 frames, first has bit 7 punched)

Data words (2 frames/word)

Forward referencing table - first frame has bits 7 and 8 (only) punched.

Checksum
a.

First has bit 6, 7, and 8 (only) punched

Next two frames are checksum

200 codes (trailer, ignored by loader which stops after checksum).

Error comments, if any, in ASCII.

D-1

APPENDIX E
ASR33 8-BIT CHARACTER SET

Character
A
B

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

D
E

F
G

H
I
J
K

L
M

N
0
p
Q

u
v

z
0
l

2
3

4
5

6
7
8
9

Character

$
%

&
(
)

I
<
=

>
?
@

[

I,
.J

332
260
261
262
263
264
265
266
267
270
271

Leader/frailer
Line Feed
Carriage Return
Space
Rubout
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

E-1

APPENDIX F
PDP-8 FORTRAN SOURCE PROGRAM RESTRICTIONS

The fol lowing I imits are imposed upon al I FORTRAN source programs for the PDP-8:

l. Not more than 896 10 data cells. Th is includes al I dimensioned variables, user-defined
variabies, constants, and all constants generated by the usage of a DO loop.
2. Not more than 20 undefined forward references to unique statement numbers per program.
An undefined forward reference is a reference to any statement label that has not previously occurred
in the program. Multiple references to the same undefined statement numbers are considered as one
reference.
3.

Not more than 64 different variable names per program.

4. Not more than 128 characters per input statement. (When using the DECtape Compiler,
the input statement size is reduced to l 00 characters.)
5.

Not more than 40 numbered statements per program.

PDP-8 COMPILER AND OPERATING SYSTEM CORE MAP
The Compiler occupies the fol lowing core locations:

- 7600

Compiler itself plus tables

7200

- 7600

Compiler tables (undefined forward reference
table, etc.)

The Operating System occupies locations:
0 - 5200

Operating System for paper tape 1/0

0 - 6000

Operating System for DECtape 1/0

Locations 5200 - 7576 are available for the user's program when using paper tape input/output or locations 6000 - 7576 when using DECtape.

NOTE
The 896

10 data word restriction applies.

F-1

APPENDIX G
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-line tape-teleprinter. These messages indicate the source of the error and direct the
programmer 1s efforts to correct the error.
To speed up the compi fer process, the compi fer 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 reentering the offending I ine via the tape teleprinter without
having to physically correct the symbolic tape and r~compile. This feature is not implemented in the
high-speed reader version of the compiler since the higher speed of the device makes recompilation
easy.
To choose the dynamic correction mode:
1. Load the starting address of the compiler (0200) in the console switches and press
LOAD ADDRESS.

Set SR bit 11to1, press START (can only be used with low-speed paper tape 1/0).

If cm error is detected, the diagnostic piints out in the noima! fashion and the computei halts.
To correct the statement:

Turn READER switch to FREE.

With the READER switch still in the FREE position, press CONTINUE.

3. Type the new line in its entirety,* obeying all rules for the source language and terminating the statement with a carriage-return I ine-feed.
4.

Turn reader on and compilation will continue.

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

*If the statement was numbered, do not reenter the statement number unless it was in error.

G-1

COMPILE TIME DIAGNOSTICS

Format of Diagnostics

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:

10;

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

During compilation cf the above statements the fo!!o\•1ing error cede \•1cu!d be printed,

indicating that a statement which occurs eleven statements octal (eight decimal} after the appearance
of statement 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 compi lotion 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 - Reload Compiler and repeat compilation process.
Contact Software Quality Control, PDP-8 Division if this reoccurs.

A comma has been used i I legally in an arithmetic statement.

Too many operators appear in a single statement.

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

G-2

Table A-1 (Cont)
Diagnostic
Code

Conditions

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 (decimal) 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 illegal character was detected and ignored.

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

Two statements with the same statement number.

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=I(J-K) for A=I*(J-K).

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

A floating-point operand should have been fixed-point, e.g., DO
10 I=l, 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 incomplete
operations, e.g., C=A+(C+(D+(E+ ....

Too many (more than 20) statements have been referenced before they
are defined.

Attempt to compile a READ or WRITE program statement after starting
program without switch 0 set.

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.

G-3

e.g., GO TO 100
The diagnostic would appear as follows where statement 100 is never defined.
100 77 20

OPERATING 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 the word 11 TILT 11 and an error number. The computer then halts. If the CONTINUE
toggle is pressed, the computer takes the action I isted in the fol lowing 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.

Floatino ooint exoonent on input greater than plus or minus
2047

Svstem exec1_1tes 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.

II legal format statement constituent

System examines next constituent.

Attempt to fix large floating
point number

Attempt to take square root of a
negative number

System takes square root of abso I ute va Iue.

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 (0))
still gives an error halt.

G-4

Table A-2 (Cont)
Error
Number
I

Action Taken

Possible Cause

Select error

Phys ica I Tape Error

I DECtape buffer exceeded

The operating system ha its with the cai ied
unit in bits 0-2 of the AC (0-3 is using 552/
555). Recovery is possible by correcting the
logical unit and pressing continue.

The program halts with the error status in the
AC. (The configuration of bits is dependent
upon the tape control being used.)

DECtape control switch set incorrectly

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. Try recompiling the
program.

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

Same as Error 76

G-5

INDEX

Arithmetic

Selection 7-2

Algebraic 3-1

Diagnostics {see Errors)

Constants 3-2

Compiler G-2, -3

Expressions 3-1

Operating System G-4, -5

Functions 3-4
Integers 3-2
Operators 3-3, 3-6, -7

DIMENSION Statement 5-1
Array Names 5-2
DO statement 6-5, -6, -7

Statements 3-4, -5

Ii iegai Nesting 6-7

Variables 3-2

Loops 6-5, -6

Arrays 5-1
Subscripted 5-1

Nesting Loops 6-6
E Format 7-4, -5, -6

Branches and Loops 6-2, 6-8

Editor, Symbolic Tape B-1

Character Set 2-4, E-1

END Statement 6-1

COMMENTS 2-2, -3

Errors (see Diagnostics)

Compiler
Diagnostics 6-1

Corrections 7-6, G-1
Messages G-1

Operating Procedures C-1

Expressions, Arithmetic 3-1

Output D-1

Field

with DECtape 8-1

Identification 2-2

Constants 3-2

Fixed Numbers 4-1, -2

CONTINUE Statement 2-3, 6-7

Floating-point

Core Map
Compiler and OP SYS F-1
DECtape

1/0 Options 8-1
Statement and Operation 8-2
Decimal Integers 3-2
Exponents 3-2
Devices

1/0 7-1

Decimal Fraction and Integer 7-5
Format 4-2
Numbers 4-1
Representation 4-1, -2
Format
Control Elements 7-7, -8

E and I 7-4, -5, -6
Floating-point 4-2
Hollerith 7-1

Input 7-5

Operating Procedures

Program 2-1

BIN A-3

Slash 7-8

Compiler C-1

Statement Number 7-3

I/O Control C-2, -3

FORMAT Statement 7-3, 7-6

Operating System (OPSYS) C-1, -2

Functions, Arithmetic 3-4

RIM A-1

GO TO Statement 6-2

Symbolprint C-1

Computed 6-8
Ho 11 er ith Format
Output 7-7, -8

Operation
Compiler 8-1, C-1
DECtape 8-1, -2

I Format 7-4, -5, -6

Loaders A-1, -2, -3

IF Statement 6-3

Operating System 8-1

Input/output

Symbolprint 8-1

DECtape Options 8-1
Devices 7-1

System 8-1, C-1, -2, -3
Operators

FORMAT 7-5

Arithmetic 3-3, -4

Statements 7-2, -3

Unary Mi nus 3-3

Integers 3-2, 4-1 , 6-5

Output

E and i Format 7-5

Data Word 7-6

Summation 6-2

Floating and Fixed Modes 7-6

Loaders
RIM and BIN A-1, -2, -3
Numbers

FORMAT 7-6
Parentheses, use of 3-4
Operators 3-3

Fixed 4-1, -2

PAUSE statement 6-1

Floating-point 4-1, -2

Program

Integers 4-1

Control 6-1

Positive or Negative 4-1

Example 2-1

Representation 4-2

Format 2-1

Statement 2-2, 7-3

Restrictions F-1

Variables, Types of 4-2

Source Preparation B-1

OP SYS (Operating Procedures) 8-2
Operating Procedures C-1

Termination 6-1
Terminator (END) 6-1

Quote, Hollerith Output 7-7
Single Quote, use of 2-3
READ Statement 8-2, -3
Restrictions
Source Program F-1
Semicolon 2-2
Slash, Control Element 7-8
Format 7-8
Statements 2-1
Arithmetic 3~4
Comment 2-2
Continuation 2-3
DECtape 8-2
END 6-1
GOTO 6-2
Input/Output 7-2, -3
Number 2-2
PAUSE 6-1
READ 8-2, ~3
STOP 6-1
Types of 2-2
Variables 3-2

WRITE 8-2: -3
STOP 6-1
Subscripts 5-1
Arrays 5-1
Variables 5-1
Symbolprint 8-1, 9-1, -2
Operating Procedures C-1
with DECtape 8-1
Tape, Symbolic Preparation 8-1

Variables, types of 4-2, -3
Arithmetic 3-2
Subscripted 5-1, -2
WRITE Statement 8-2, -3

READER'S COMMENTS

PDP-8 4K FORTRAN
PROGRAMMERS REFERENCE MANUAL
DEC-08-AFCO-D

Digitai Equipment Corporation maintains a continuous effort to improve the quality and usefulness of its
publications. To do this effectively we need user feedb.ack - your critical evaluation of this manual.
Please comment on this manual's completeness, accuracy, organization, usability, and readability.

DEC also strives to keep its customers informed of current DEC software and publications. Thus, the following periodically distributed publications are available upon request. Please check the appropriate boxes for a current issue of the
publication(s) desired.

Software Manual Update, a quarterly collection of revisions to current software manuals.

User's Bookshelf, a bibliography of current software manuals.

Program Library Price List, a list of currently available software programs and manuals.

Name - - - - - - - - - - - - - - - Organization
Street-------------- Department-----------------City ___________ S t a t e - - - - - - - - - - - - - - Zip or Country _ _ _ __

- - - - - - - - -

- - -

- - - - - - - - -

- - - - -

Fold Here - - - - - - - - - - - - - - - - - - - -

- - Do Not Tear - Fold Here and Staple -

- - - - -

- - -

FIRST CLASS
PERMIT NO. 33
MAYNARD, MASS.
BUSINESS REPLY MAIL
NO POSTAGE STAMP NECESSARY IF MAILED IN THE UNITED STA TFS

Postage will be paid by:

mamaama
Digital Equipment Corporation
Software Information Services
146 Main Street, Bldg. 3-5
Maynard, Ma~chusetts 01754

Digital Equipment Corporation
Maynard, Massachusetts

printed in U.S.A.