Digital PDFs

DEC-08-AFC

December 2000

84 pages

Original

1.9MB

Document:	dec-08-afco-d
Order Number:	DEC-08-AFC
Revision:
Pages:	84
Original Filename:	https://svn.so-much-stuff.com/svn/trunk/pdp8/src/dec/dec-08-afc/dec-08-afco-d.pdf

OCR Text

0..

M'ag§t.?ard?fiZZ‘SLthué’éfli'am"

EJflEflIEII

FORTRAN

PDP- 8
PROGRAMMER s REFERENCE MANUAL

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, Maynard, Mass.

$2.00

O|GITAL

EOLJIF’IVIENT

CORPORATION

Price:

MAYNARD,

MASSACHUSETTS

lst 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

Your attention is invited to the last two pages of this

manual.

The Reader's Comments page, when filled in and

returned, is beneficial to both you and DEC. All comments
received are considered when documenting subsequent man-

uals, and when assistance is required, a knowledgeable
DEC representative will contact you.

The Software Infor~

mation page offers you a means of keeping up-to-date with

DEC '5 software

Instruction times,

operating speeds and the like are in-

cluded in this manual for reference only; they are not to
be taken as specifications.

The following are registered trademarks of Digital
Equipment Corporation, Maynard, Massachusetts:
DEC

PDP

FLIP CHIP

FOCAL

DIGITAL

COMPUTER LAB

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's Programmed Data Processor-5. This compatibility between

the libraries of the two computers results in three maior advantages:
1.

The PDP—8 comes to the user complete with an extensive selec-

tion of system programs and routines

pability of the new computer

making the full data processing ca-

immediately available to each user, elimi-

nating many of the common initial programming delays.
The PDP-8 programming system takes advantage of the many
man—years of field testing by PDP-5 users.
2.

Each computer can take immediate advantage of the continuing

program developments for the other.

CONTENTS

Page
CHAPTER 1

INTRODUCTION
CHAPTER 2
THE FORTRAN LANGUAGE

Statements

2-i

Program Format

2—1

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
Use of Parentheses

3—4
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

Arrays

Subscripts
The Dimension Statement

CONTENTS (Cont)

CHAPTER 6
PROGRAM CONTROL

Program Termination
END Statement

Branches and Loops
The GO TO Statement

Integer Summation
DO

Loops

Illegal DO Nesting
The CONTINUE Statement

Computed GO TO
CHAPTER 7
INPUT AND OUTPUT

Available Devices

Input and Output Statements
Device Selection and Direction of Transfer

Statement Number of Format Statements
List

Format Specifications Statement

Control Elements E and I

Input
Integer Values

Fixed Point

Decimal Fraction Values

FORMAT (I)

Floating Point

FORMAT (E)

Correcting Typing Errors

Output
Data Word Output

Floating and Fixed Modes
FORMAT (E) and FORMAT (1)
-

Other Format Control Elements

7—6
7—7

Quote (") (Hollerith Output)

7-7

Slash (/)

7-8

CHAPTER 8

FORTRAN WITH DECTAPE OPTION
FORTRAN Compiler with

DECtape I/O Option

CONTENTS (Cont)

Page
Use of Symbolprint with FORTRAN

8—I

FORTRAN Operating System with DECtape I/O Option

8—I

DECtape FORTRAN Statements and Operation

8—2

CHAPTER 9

PDP-5/8 FORTRAN SYMBOLPRINT
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
ILLUSTRATIONS
2—]

A Fortran

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

6-6

Program

Loops

ILLUSTRATIONS (Cont)

Loops

Program Branching in D0 Loops

Input-Output Statement
A List Example

Examples of Quote and Slash

CHAPTER 1

INTRODUCTION

Using a digital computer to solve a problem generally involves the Following series of steps:
Determining the correct procedures to be used, including mathematical formulas, the

handling of data, the presentation of results, etc.

specified

Arranging the procedures in the proper order.

Determining the sequence of computer instructions that will perform the operations

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.
a

Later, a symbolic notation was developed for

computer instruction set, and programs called assemblers were written that could interpret a tape or

card deck punched with this notation.

These assemblers translate each symbolic instruction into a ma-

chine operation and assemble an executable program.

Thus, an assembler can accept input from step c

above.
An assembler requires that the programmer be familiar with the particular instruction set of

the computer being used.

programming

To solve the same on another computer would usually require complete re-

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 ac-

cepts 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:

system

l-l

the compiler and the object time

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 obiect 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 obiect time system is used to execute the program.

This system con-

tains the interpreter, the arithmetic Function subroutines, and the input/output packages.
execution is required, the object time system,
into the computer For solution of the problem

When program

obiect program, and the data it will work with are loaded

This is a one-pass compiler, which means the source language tape must be read only once.

The compiler generates one tape which contains coding in a form that is executable under control of the

object time system.
To use the system, it is only necessary to load the compiler.
source

language tape and generates the obiect program tape.

any time simply by

The compiler then processes the

This obiect program tape can be run at

loading the obiect system, which, in turn, loads and executes the object program.

CHAPTER 2
THE FORTRAN LANGUAGE

STATEMENTS

Figure l is an example of a FORTRAN program, consisting of a title, the body of the program,
and the end statement.

The first line of the program is the title, which may be anything the programmer writes to

identify the program.

It is not

incorporated into the final executable program.

The body of the program is a series of statements, each of which specifies a sequence of

mathematical operations, controls the flow of the program, or performs other tasks related to the proper

working of the program.
The end statement must be physically the last statement of every FORTRAN program.

Its

function is to indicate to the compiler that nothing more connected with the preceding program is to

follow.

c;
5;
10;

THIS PROGRAM CALCULATES FACTORIALS

TYPE 200

ACCEPT 300,x

FACT=Y=T.

(X) 5,32,30
(X-Y) 41,32,33
TYPE 400,X,FACT
IF

30;
32;

GO TO 10

33;

FACT=FACT*(Y=Y+l .)
GO TO 30

41;

PAUSE

200;
300;
400;

FORMAT

(/, "PLEASE TYPE A POSITIVE NUMBER", /)

FORMAT

(E)

FORMAT

(/,E, "FACTORIAL IS",E)

GO TO 5

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—]

The identification field extends from the left-hand margin up to and including a semicolon
This field may be left blank, or it may contain one of the following types of identification:

character.

0.
The first digit of a statement number. This number, which may be any positive integer
from i to 2047 inclusive, identifies the statement on that line for reference by other parts of the pro-

gram.
ments

Statement numbers are used for program control or to assist the programmer in identifying segof his program.

Up to 40 statements can have statement numbers.

The letter C.

This identifies the remainder of the line as a comment (see Section

Comments).
The semicolon

(;) is necessary only if the statement is numbered or is a comment, (i .e.

the identification field is blank, the semicolon may be omitted).

The statement field begins immediately after the semicolon and extends through the next car-

riage return
more

Although the continuation character (' ) allows a single statement to extend over two or

lines, no more than one statement can be written on one line.

continued by use of the continuation character

Note that a comment cannot be

(' ).

TYPES OF STATEMENTS
FORTRAN statements are of several types with differing functions distinguished as follows:
a.

operations

Arithmetic statements resemble algebraic formulas.

They specify the mathematical

be performed.

Program control statements direct the flow of the program

c
Specification statements allocate data storage, determine variable and data types, and
specify input/output formats
.

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 line intended as a comment, are not interpreted by the compiler and have no
effect on the executable program.

C;
C;

CALCULATE PERCENTAGE OF CORRECT RESPONSES
PERCENTAGE

-1 IF THERE ARE NO ITEMS
IN CATEGORY

C
DO 471:1, 57
DO 48 J=i , 6
IF

46;

(ITMS (J)) 46, 46, 51
(I) -].0

PRCN

Figure 2

Example OF Comments

CONTINUATION

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

If the character single quote

(')

appears as the last character of a line before the carriage return, the next line is treated as a continuation of the statement on the line 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;
;

2;
;

3;
4;
;

FORMAT
IF

(NR (1) -1) 2, 2, 3

AP=-l4.73
GO TO 6
IF (NR (I) —2) 4, 4, 5

AP=-44.l9
GO TO 6
AP=SH

Figure 3

(2)*3.0-AG(4)/'

(i)+SQTF (AG (14))

The Continued Statement

THE CHARACTER SET

The characters which are meaningful in FORTRAN belong to the ASCIIset listed in Appendix D.
Of these, the acceptable characters are:

characters

semicolon

all letters and numbers

A through Z, 0 through 9; control

(; ), carriage return (CR), line feed (LF), single quote (‘ ), double quote (" ),

left parenthesis (, right parenthesis ), period

slash (/), asterisk (*), equal sign (=).

(.), comma (, ); and the operators

plus (+), minus (-)

All other characters are ignored by the compiler except in

Hollerith information of FORMAT statements where all Teletype characters are legal.

The character

space has no grammatical function except in FORMAT statements, but may be used freely to make a

program easily readable.

2-4

CHAPTER 3
FORTRAN ARITHMETIC

ARITHMETIC EXPRESSIONS
An algebraic formula such as the Following

[5a + 4b

(x2 x0)] /2a sin 9
-

represents a relationship between literal symbols (0, b, x, x0, 9) and constants (5, 4, 2) indicated by
mathematical functions and arithmetic signs (+,

/, multiplication, exponentiation, sine).

This same

formula can be written as a FORTRAN arithmetic expression with very little change in appearance:

(5.*A + 4.*B*(X**2

XZRO) )/ (2.*A*SINF(THTA))

The construction of both expressions is the same; the differences are notational.
The elements of an arithmetic expression are of four types:
and Functions.

stants,

An expression may consist of a single constant,

constants, variables, operators,

single variable, or a string of con-

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

0x2

FORTRAN

bx + c

Expression

A*X**2 + B*X + C

ELLE?)
(0
b)

(A**2

3w) / (A+B)** 2

4-*PI*(R**2)/3-

4n
3

W)
a

~sin 9 + 20

23“

(3.*X*PI—2.* (X+Y)) /4.25
'

cos

(9 ~45)

A*SINF(THTA)+2.*A*COSF(THTA-0.7853982)

2.*SQTF(X)/3.

Constants
Constants are explicit numerical

decimal exponent Form.

quantities.

They may be integers, decimals, or numbers in

Some examples Follow:

integers
-70

2047

3.14159

-0.00025

decimals
18.75

decimal exponent

(meaning 1.66 x 10—16)

1.66E-16

These different Forms oF numerical representation are described in detail in Chapter 4.

Variables
A variable is a literal symbol whose value is not implicit; its value may be changed during

the execution 0F the program.

A variable name is composed oF one or more characters according to

these three rules:
a.

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

The First character must be alphabetic (i .e.

Only the First Four characters oF any variable name are meaningful

through 9.
,

A through Z).
.

All characters aFter

the Fourth are ignored by the compiler.
Some examples oF acceptable variable names are:

XZRO

LST8

THTA

P5l

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 EXI T, are identical names as Far as the compiler is concerned
because they all reference the same variable.
The name EPSILON would be interpreted by the compiler as EPSI, since only the First Four
characters are meaningFul
identical.

For example, the two names XSUM] and XSUM2 would be considered

Some incorrect variable

names are:

9SRT

(first character not alphabetic)

GO(5

(illegal character included)

CSH$

(illegal character included)

Operators
The operators are symbols representing the common arithmetic operations.
about operators in the FORTRAN arithmetic expressions is this:
resented by an operator.

Every operation must be explicitly rep—

In particular, the multiplication sign must never be left out.

exponentiation is also provided since superscript notation is not available.
are

The important rule

A symbol for

To illustrate the rule, here

the FORTRAN and algebraic forms given in the section on arithmetic expressions:

(5.*A +4.*B*(X**2—XZRO) ) / (2.* A*SINF (THTA))
[50 + 4b (x2

xo)] /2a sing

Normally, a FORTRAN expression is evaluated from the left to right iust 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:
lst

Expressions

( )

within parentheses
2nd
3rd

4th

5th

Unary minus

—-

Exponentiation

Multiplication

Division

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, expo—

nentiation 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

3—3

(indicating a subtraction) is binary.

The left-to-right rule can now be stated more precisely as follows:

binding strength is evaluated from left to right.

A sequence of operations of equal

To change the order of evaluation,

evaluated as A-(B*C),

right rule

not

parentheses are required. Thus, the expression A-B*C is

(A-B)*C. The example below gives a few more illustrations of the left-to—

The expression

is evaluated as

A/B*C
A/B/C

(A/B)*C
(A/Bl/C

A**B**C

(A**B)**C

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

following example:

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

The procedure is this:

I0 12

Reading the expression from left to right, assign the number I to the first left

parenthesis (if you encounter a right parenthesis first, the expression is already wrong!)
count

by one each time a left parenthesis is read, and decrease the count by one when a right paren-

thesis is used.
comes

zero

Increase the

When the expression has been completely scanned, the count should be zero.

less than zero during the scanning, there are too many right parentheses.

If it be

If it is greater than

the end of an expression, then the pairing is incorrect.

Use of Parentheses
Note the use of parentheses in the following 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)
In algebra there are several
etc. , for

(A+B)**C+SINF (X)

devices, such as square brackets ([l ), rococo brackets ( {} ),

distinguishing between levels when subexpressions are nested.

In FORTRAN,

only the curved

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

In a given expression, the number of left parentheses must be equal to the

number of right parentheses.

Functions
Functions are used in FORTRAN
an

arithmetic expression

just as they are in ordinary mathematics

variables in

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:
tangent, exponentiation, and natural

square root, sine,

cosine, arc

logarithm.

The argument of a function can be a simple or subscripted variable or an expression

The

FORTRAN recognizes a term as a function when the term is a pre—

argument must be in floating point.
defined symbol

ending in F followed by an argument enclosed in parentheses (if the F is missing from the

term, the symbol is treated as a subscripted variable).
other function or group of functions.

The argument of a function can consist of an—

For example, 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 X is expressed in radians

COSF

(X)

cosine of X, where X is expressed in radians

ATNF

(X)

arc

tangent X, where the angle is given in

radians

EXPF

(X)

exponential of X

LOGF

(X)

logarithm of X

THE ARITHMETIC STATEMENT

The arithmetic statement relates a variable V to an arithmetic expression E by means of the

equal sign (i).

Thus:
V

Such a statement looks like a mathematical equation, but it is treated differently.

The equal sign is

interpreted in a special sense; it does not represent a relation between left and right members, but it
specifies an operation to be performed

NOTE
In an arithmetic statement, the value of the expression to

the right of the equal sign replaces the value of the variable
on

the left.

This means that the value of the left-hand variable will be different after the execution of
an

arithmetic statement.

A few illustrations of the arithmetic statement are given below.

VMAX

T
P1

VO + AXT

2.*PI*SQTF(T./G)

3. 14159

THTA
MIN

OMGA + ALPH*T**2/2.

MINO

INDX

INDX + 2

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

arithmetic statement.

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

its value will be 42 after execution.

Perhaps another way of looking at the equal sign illustrates its use and interpretation more
fully.

In arithmetic expressions,

binary operator requires an operand on its left and right.

sign of an arithmetic statement is considered to be a binary operator also.

The equal

This interpretation is dem—

onstrated in the following revised table of operators:

Operator

Interpretation

Elie
-A

negate A

A**B

raise A to the Bth power

A*B

multiply A by B

A/B

divide A by B

A+B

add B to A

A-B

subtract B from A

A=B

replace A with B

—

(Unary)

(Binary)

Treated this way, the equal sign is considered to have the lowest binding strength of all the operators.

This means that the expression on the right is evaluated before the operation indicated by

is performed.

The most important result of treating the equal sign as a binary operator is that it may be used
more

Consider the following:

than once in an arithmetic statement.
CPRM

(CKL

CKG) / (CPG

P*(Q +1.))

The internal arithmetic statement, CPG

P*(Q + 1.), is set off from the rest of the statement

by parentheses. The complete statement is a concise way of expressing the Following common type of
mathematical procedure:

th- Ckg
Let

wer

Cpg

:*+i
P(q)

Cpg

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 D in the example, may be an arithmetic expression
statement is

interpreted as follows:

The

”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.
most expression is given to each of the variables in

In other words, the value of the

the string to the left.

right-

A common use for this con-

struction is in setting up initial values:

XZRO=SZRO=AZRO=O
T=T l=T2=T3=6O

P=FP=4 *ATM-AK
.

Only single level replacements will compile correctly in this manner.
type A(l)

A(2)

R(l)

For example:

Statements of the

O.l23 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
necessary

sign is converted to the mode of the left-hand variable before storage, if

Example:

CHAPTER 4
NUMBER REPRESENTATION AND VARIABLE TYPES

INTEGERS AND FLOATINGcPOINT NUMBERS
In mathematics there are many ways to categorize numbers.

rational or irrational, whole numbers or Fractions.
into integers and decimals,

They may be positive or negative,

In FORTRAN, the treatment oF numbers is separated

distinguished as follows:

Integers are positive or negative numbers written without a decimal point. These num9, 17, 147, l024, 2047. The last number, 2047, is the largest quantity that can be
a FORTRAN integer.
as
For Fractional quantities and For numbers larger than i2047 (which
expressed
is 2H -l), the second type oF number is required.
a.

bers are integers:

When using integer arithmetic, any Fractional results are truncated.

expression M=N/3 with N=8 would result in M=2.
addition and subtraction yield integral results.

For example, the

This applies only to division because multiplication,

b.
Floating point numbers have two Forms. They are simple decimals, such as 0.0025, .4,
2.7l828; or numbers in decimal exponent Form. Numbers in decimal exponent Form are simple
decimals multiplied by a power of IO. Numbers in decimal exponent Form
wt have an explicit decimal
point.
57. ,

Examples:
Mathematical Form

FORTRAN Form

l023
10"6
1.66
72.xl0l2

6.023 x

6.023E23
1.66E—16

In general,
an

72.512

Floating point number in decimal exponent Form is NEiK, where N may be

integer or simple decimal, and K is an integer From 0 to 99, inclusive. The construction NEiK is

used to represent the number

leOK.

The Following are Floating point representations oF the number l9:
l9.0
.l9E2

l.9E+l
l900.E-2
l90.0E-l

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 l2-bit computer word.
in high-order bit and the magnitude in the remaining ll bits.

The sign oF the number is kept the

This representation, shown schematically

Figure 4 is called Fixed point, because the decimal point is always considered to be to the right oF the

rightmost digit.

A FORTRAN integer may not exceed the range of -2047 through +2047.

greater than i2-47 are taken modulo 2048 (that is 2049 is taken as OOOl
The floating point format consists of two parts:

All integers

4099 is taken as 3).

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.
to obtain

The exponent represents the power of two by which the mantissa is multiplied

the true value of the number for use in computation.

stored in 25 complement form

The exponent and mantissa each are

SIGN

l
Ll

MAGNITUDE
‘

FORTRAN INTEGER

“

SIGN OF EXPONENT

SIGN OF

MANTISSA-fi‘

EXPONENT
11

0 1
.

MANTISSA

b.FL.OAT|NG 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

follows:
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 letter.

These are integer variable names:
names:

ZXRO, CONT, FICA.

INDX, KDTA, M359. These are floating—point variable

Integers cannot appear in Floating point expressions except as exponents or subscripts.

examples of illegal and legal expressions are as Follows:
Expression

Legal

Mode

A(I)*B(J)**2
I(M)*K(N)

Yes

Floating

Yes

Fixed

4 .* J

I+D

l6.*B

Yes

Floating

(K+l 6)*3

Yes

Fixed

A**(I+2)/B

Yes

Floating

8*A

Some

CHAPTER 5
SUBSCRIPTED VARIABLES

ARRAYS
An array is a grouping of data.
matrix are all arrays.

In mathematics,

A column of figures, the elements of a vector,

For example, the sixth element

v6.

FORTRAN, array elements are similarly identified.

The array is given a name subiect to

the some rules as the names of variables, described in Chapters 3 and 4.
an

list, and a

element of an array is referenced by means of a symbol de—

noting the array and subscripts identifying the position of the element.
in a vector v is designated

element of the array is enclosed in parentheses.

The subscript which identifies

The element referred to in the preceding paragraph

would have the following form in FORTRAN:

V(6)
Such a name designates a subscripted variable, which may be used in computation just like a

simple variable.
the array

The array name determines the mode, integer, or floating point of all the elements in

The example below gives a few illustrations of the use of subscripted variables.
a

X(I+L)=X(I)+ALPH(I)*P(I)

X(I+3)=X(I+2)+X(I+I)/2.

C(J)=A(I*J+3)

A=B(6)

Subscripts
As the example above illustrates,

subscripts may be quite diverse in form.

In fact,

may be any acceptable FORTRAN arithmetic expression as long as it is integerwalued.

subscript

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.
must

Two items of information

be provided in any program using arrays:
a.

Which are the subscripted variables?

What is the maximum dimension of the subscript?

amount of storage space must be set aside for its elements;

5-]

(When an array is used, a certain

hence this requirement.)

All the above information is provided by the following specification statement:
DIMENSION A(20),

B(l5)

where A and B are array names, and the integer constants 20 and I5 are the maximum dimensions of
each subscript.
The rules governing the use of array names and the dimension statement are as; follows:
DIMENSION may be used as many

All array names must appear in a dimension statement.

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(l00), REGR(20)

In the statement in the example above, the names LIST and MAT designate integer arrays;

The third name, REGR, designates a floating-point array.

The

first array is a list of 30 elements maximum, so that 30 words of storage are set aside for its use.

The

that is, each element is an integer.

third array is floating-point and there are 20 elements in it.

Since this array is floating, each element

requires 3 words of storage so that 60 words are set aside for the array
DIMENSION B(30),

1(15)

This version of the PDP-5/8 FORTRAN does not have the facility for double subscripted variables.

To accomplish double subscripting, the programmer has to include indexing statements in the

source

program as illustrated in Figure 5.
In this example the matrices are stored column wise in memory, that is,

sequential locations

in memory are used as follows:

Relative Position
Element

Memory (INDX)

al I

02 ll

ONIO‘tn-P-OJN—I

03 ll
a4l|
05 ll

aéll
a122

a222

056

a6<5

36;

If referencing element

INDX=M+I*(N-l)=5+6*5=35.

056

in the array,

M=5, N=6 (I would be =6 for a 6 by 6 array.), and

If referencing element 022, INDX=2+6*I=8.

5-2

MATRIX MULTIPLY

DIMENSION A(36),

B(36), C(36)

ACCEPT I, I
FORMAT (I)
DO 10 M=I, I
DO 10 N=I, I

INDX=M+I*(N-])
ACCEPT 2, A(INDX)
FORMAT

(E)

CONTINUE
TYPE 15

FORMAT

(/,/,/)

DO 20 M=I, I
DO 20 N=I, I

INDX=M+I*(N-1)
ACCEPT 2, B(INDX)
CUNDXFO
20;

CONTINUE
DO 30 M=I, I
DO 30 N=I, I
DO 30 K=I, I

IC=N+I*(M—1)
IA=K+I*(M-1)
IB=N+I*(K-1)

C(IC)=C(IC)+A(IA)*B(IB)
30;

CONTINUE
TYPE 15

DO 40 M=I , I
TYPE 21

DO 40 N=I, I

INDX=N+I*(M-])
TYPE 2, C(INDX)

40;

CONTINUE
FORMAT (/)
TYPE I5
END

Figure 5

Indexing Sfofemenfs

CHAPTER 6
PROGRAM CONTROL

In this chapter, the FORTRAN statements which have been described as isolated elements
It is obvious that FORTRAN statements

are

discussed in their proper context

are

executed in the order in which they are written unless instructions are given to the contrary.

in program sequences.

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 halt.

PROGRAM TERMINATION
A program arranged so that the last written statement is the final and only stopping place
needs no special
Most programs,

terminating indication.

The end statement automatically determines the final halt.

however, contain loops and branches so that the last executed statement is often some-

Frequently, there may be more than one stopping point.

where in the middle of the written program.

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 (1) 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 manually. This procedure is

frequently necessary when the operator must do such tasks as loading and unloading tapes or card decks
in the middle of a program.

This kind of temporary halt is provided by the following 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.
E N D Statement
END occurs alone on a line and indicates the physical end of the program to the FORTRAN

compiler.
statement.

It must be followed by carriage return and line feed.

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 (l), or it may branch to an entirely different
sequence (2), return to an earlier point (3),

skip to a later point (4).

Us
34
—_|

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 TO N

where N is a statement number used in the program.

The use of this statement is described in the fol-

lowing 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

statements

preceding this loop provide the starting conditions; this is called the initialization. The

partial sum is set to zero, and the First integer is given the value of one.

The loop then proceeds to

add the integer value to the partial sum, increment the integer, and repeat the operation.

SUM OF FIRST N INTEGERS BY ITERATION

KSUM=O
INUM=I

KSUM=INUM+KSUM

INUM=INUM+I
GO TO 2

Figure 7

Integer Summation

Limits and Decisions

The IF Statement

The program shown in the preceding example performs the required computation, but there is
one

flaw:

the

loop is endless.

To get out of the

loop, the user must know when to stop the iteration

and what to do afterwards.

The IF statement fulfills both requirements.

It has the following form:

IF (E)K,L,M

where E is any variable name, arithmetic expression, or arithmetic statement, and K, L, and M are
statement

numbers.

The statement is interpreted in this way:
if the value of E is less than 0, GO TO statement K

value of E is equal to 0, GO TO statement L

value of E is greater than 0, GO TO statement M

Thus, the IF statement makes the decision of when to stop by evaluating an expression, and also provides the program branch choices which depend on the results of the evaluation.

SUM OF THE FIRST 50 INTEGERS
KSUM=O
INUM=I

KSUM=INUM+KSUM
INUM=INUM+I
IF

Figure 8

(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.
ment says:

This state-

If the value of the variable INUM is less than or equal to 50 (which is the same as saying

that it 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=0
INUM =50

KSUM=INUM+KSUM

STOP

IF(INUM=INUM-I) 3,3,2

Figure 9

IF Statement with Substatement Feature

In this example, the sum is formed by counting down, but the same results are achieved.

The initialization is changed so that INUM starts with the value of 50 instead of O, and the statement
INUM=INUM+I is no

longer required.

A loop may also be used to compute a series of values.
to

The Following example is a program

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:

KN"KN-1+KN-2
C;

FIBONACCI SERIES, IOO TERMS

;

DIMENSION

;
;

FIB(1)=1
FIB(2)=I

;

K=3

5;
6;
;

IO;

FIB(I00)

FIB(K): FIB(K-l)+FIB(K-2)
K= K+I
IF (K-tOO) 5,5,10

STOP

Figure I0

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 calculation stops.
DO Loops
Iterative procedures such as the

loop in the example above are so common that a more con-

cise 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=Ki,

K2, K3

Here, n is a statement number, I is a simple (non—subscripted) integer variable, and KT
are

K2, and K3

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.
sumed equal to one.

;
;
;
;
;

5;
;

Figure II

FIBONACCI SERIES, 100 TERMS
DIMENSION

FIB(100)
FIB(i) =1
FIB(2) =1
DO 5 K=3, 100

FIB(K)=FIB(K-i)+FIB(K—2)
CONTINUE
STOP

Fibonacci Series Calculation Programmed As a DO Loop

In words, the DO statement says:

encountered.

If K3 is omitted from the statement, it is as-

Statement n must be a CONTINUE statement.

Do all statements through statement 5 for

Perform the following test:

If K+I is less than or equal to

in the program by executing the first statement after the DO.

sequential statement following statement 5 is executed.
DO

K=3, when statement 5 is

100, set K=K+I and continue

If the K+I is greater than I00, the

In this example this is a STOP.

loops are commonly used in computations with subscripted variables.

it is usually necessary to perform

loops within loops.

In these cases,

Such nesting of loops is permitted in FORTRAN.

;

DO l0 [:1 ,20

;

X(I)=0

;

DO 5 J=2,40,2

X(1)=X(I)+(B(J)-Z(J))**2

;

CONTINUE

A(I)=X(I)**2+C(I)
CONTINUE

l0;

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 sum-

ming every element in a C array and the square of every element in the X array.

The algebraic ex—

pression For the loop is as Follows:
2

Ai =Xi —Ci

For I l, 2, 3,...20

where

X.=2
I

(b.-z.)

.22
l

Fori=2,4, 6,...40

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

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.
schematically illustrates permitted and forbidden arrangements.

‘,______1
'-——2
3
'

”—2

——

5
6

[—7
b.

Figure 13

Loops

Figure 13

lo.

DC, the variable I is not initialized as specified
Transferring into the range of a DO is allowed as long as:

If the user transfers into the range of a

in the DO statement.

(l)

Incrementing and testing start with the present value of I.

(2)

Control was originally transferred out of the DO other than by completing it.

A DO

loop must end on a CONTINUE statement.

Those in a are permitted;
in b are not permitted;

loops 5, 6, and 7 end on the same statement.

The arrangements

loop 3 ends on a statement outside the range of loop l.

Illegal DO Nesting
DO 10 1:1, 20
DO 20 J=t ,ioo,2

SUM=(X(I)-Y(I))**2

;

CONTINUE

10;

Z(J)=SUM+-A(J)

CONTINUE

20;

rwug
(4)
U!

Figure l4

Program Branching in D0 Loops

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

The CONTINUE Statement
Since the DO loop may contain alternate courses of action, programmers frequently wish to

make the last executable statement of a loop, a test to determine which of the alternatives should be

taken next.

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

statement is

provided which is not an executable statement itself, but provides a termination For such

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 (KI ,K2,K3,.

.Kn),J

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

2, 3,

according to the results of some previous computation.

For example,

IVAR =I4*J/2+K
GO TO (5, 7, 5, 7, 5, 7,
causes a

ment

IO), IVAR

branch to statement 5 when IVAR=I ,3, or 5; to statement 7 when IVAR=2,4, or 6; and to state-

I0 when IVAR=7.
When IVAR is less than one or greater than seven, the next sequential statement after the

GO T0 is executed.

6-8

CHAPTER 7
INPUT AND OUTPUT

AVAILABLE DEVICES
So far,

have assumed that all 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
a.

input/output procedure, several items must be specified:

In which direction is the data going?

read into memory; information going out is

In FORTRAN terms, the data coming in is being
being written on whatever medium is specified.

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.
Where in core memory is the data coming from or going to?
c.
its location in the computer storage must be specified.
d.

In what mode is the data represented?

The amount of data and

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

quired to provide all of the information listed above.

device, two statements are re—

The first three items are specified by the input/

output statement and the last two items are determined by the FORMAT statement.

PDP-5/8 FORTRAN provides for communication of data to and from a program in the following
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).

The high—speed reader and punch can also be used for data transfer via punched paper

tape.

No printed copy is made when output is to the high-speed punch.

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—bit
ASCII coded characters right iustified in 12—bit words (one character per word).

IN PUT/OUTPUT STATEME NTS

The input/output statements control this transfer of information.

1/0 statements consist of three basic items of information:

As illustrated in Figure 15,

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,V(I),

V(I+l),

V(I+2)

List of variable names

Statement number of FORMAT statement

Device selection and direction of transfer

Figure I5

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 paper-

tape 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

;READ
;WRITE

UNIT,
UNIT,

DECtape from core memory.

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,

read

into core.
Bit 0 of the SWITCH REGISTER must be set to I

(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

Following 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 [/0 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 m of the variables and array elements whose values are to be transferred in
There is no restriction on the number of names which may appear in the list of an

the m transfer.

[/0 statement as long as the total statement length does not exceed 128 characters.

The modes of the

variables named need not agree with the corresponding FORMAT statement, however, the modes speci—
fied in the FORMAT statement take precedence.

Example:
TYPE 23,

23;

A, J, KAL, BOB

(where A=3.2 J=27 KAL=302 and BOB=7.58)

FORMAT (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 will be typed as an integer; and BOB as a normalized number.

The output will look like the following:
+

+0 2700000E+02
.

302

+0 7580000510]
.

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

+c'

or
_

Av'

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

;

10;

TYPE

10,A,I,B,C(I+K),N(J+L)

FORMAT (E,I,E,/)

;

A=A+(C(J)**é—C(N)**2)

;

TYPE

10, A,J,B,C(N)

Figure l6

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 list.
For instance in the preceding example when the value oF the variable B is typed in the E For-

mat, the control character slash (/) causes a carriage-return line-Feed to occur.
statement is

Then the FORMAT

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.

there are Fewer names in the list than there are elements in the FORMAT statement, the program com—

pletes the I/O operation and proceeds to the next sequential FORTRAN statement.
ment is

IF this next state-

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 all oF their elements are exhausted.

FORMAT SPECIFICATIONS STATEME NT

As already mentioned in the previous description oF
statement
names

input/output statements, the FORMAT

controls the arrangement and mode oF the inFormation being transFerred.

The values oF the

appearing in the list oF the [/0 statement are transFerred in the mode speciFied by the corre-

sponding 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.

input or output

can

Mode conversion on

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

TO, 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

l4.96, only the integer part, 14, would be typed.

less than one, zero would be typed.

For instance,

IF A has an absolute value oF

Input
Input data words consist of a sign, the decimal value, an exponent value iF the data is
floating point, and a Field terminating character such as space. Any character that is not a number,
decimal point, sign, or E can be used to terminate a Field except the character 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.

It 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, the photo-electric reader or DECtape.

floating-point modes For integers or decimal Fractions.

They can be entered in either Fixed- or

The mode in which they will be stored is con-

trolled by the corresponding element in the FORMAT statement.

Integer Values
An

Fixed Point

—

FORMAT (I)

integer data Field consists ot a sign (minus or space) and up to Four decimal characters.

Some examples of integer values are as Follows:

Decimal Fraction Values

A Floating-point

acters,

Typed Numbers

Values Accepted

—2001

-200l

—4O

-OO4O

-0040

0016

-2047

Floating Point

FORMAT (E)

input word consists oF a sign*, the data value of up to seven decimal char-

E if an exponent is to be

included, the sign oF an exponent, and the exponent which is the

power of ten that the data word is multiplied by.

Example:
dddd.dddEnn

The d's represent characters in the data word and n represents the power oF ten of the 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 decimal character.

It 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:
Values Accepted

Typed Numbers

l02
O.i6x 102
O.l6X102
0. l6 X

l6.

.iéEOZ
l600.E-02

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 re-

typing the data word that is in error.
To cancel

or erase a

word, type a rubout character.

When this character is detected during the acceptance of a data word and before the ter—
mination 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 ex—

ponent (minus or plus), and an exponent value of two characters.

Examples:
Integer Values

Output Format

-iO43

—lO43

-0016

—

+0016

l6
16

Floating-point values are printed as per example
SO dddddddsxxxx
.

where
S is the sign, minus sign,
d is the seven decimal
5

space

digits of the data word

is the sign of the exponent value

xxxx

is the exponent value

Decimal Value

Output Format
-O 8388608E+07

-8 , 388 , 608 O
.

+. 0000001 192092

0. l l92092E-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 FORMAT control elements quote (") and slash (/).

Quote (") (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.

TYPE 10

TO;

HOLLERITH",/)

FORMAT (/, ”THIS IS

TYPE 100, AMIN, AMAX

TOO;

FORMAT (/,

"MINIMUM=",E,/, ”MAXIMUM=”,E,/)

TYPE 210

210;

FORMAT (/,/,"
"

DISTRIBUTION",/,/'
FREQUENCY",/)

CUMULATIVE

INCREMENTS

DO 220 K=l ,100
TYPE 250,

200;
250;

K,VALU(L), VALU(K+I), COUNT(K)

CONTINUE
FORMAT (1,"

",,"E

Figure 17

All

pendix D).

”,,"E

",E,/)

Examples of Quote and Slash

legal Teletype characters can be contained within quotes and are output as text (Ap—

If a statement continues on another line 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:
1

TYPE

10;

10, VAR,SD

FORMAT (E,E,E, ”VARIANCE AND STANDARD DEVIATION",/)

In this 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 line-feed will be read.
numeric data printed in either E or I format,

Because of the variable number of spaces following

precautions must be taken when text and data are inter—

mixed in a punched tape which is to be read by a FORTRAN program.

The technique has been used

successfully, 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 I/O are fixed lengths, there is essentially no difficulty here.
it is possible to read text from

Note that

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 TO, A, B
FORMAT (/,/,/,E,/,/,E,/,/)

Three carriage-return line-feeds will be typed before the value of A; then two carriage-return linefeeds will be typed before and after the value of B is typed.

The input subroutine of the obiect 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 line-feeds with the data.

7-8

CHAPTER 8

FORTRAN WITH DECTAPE OPTION

PDP-8 FORTRAN includes provisions for storage of data on
programs requiring

DECtape.

In this way FORTRAN

large amounts of accessible data may be readily written to run on a 4K PDP-5 or

PDP-8.

The standard library version of the FORTRAN Compiler is written such that the type of DEC-

tape hardware to be used is irrelevant.
is written to handle the TCOI

The standard library version of the FORTRAN Operating System

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 library.

FORTRAN Compiler with DECtape I/O Option
When the FORTRAN Compiler is read into core, it is equipped with a switch option govern-

ing the compilation of DECtape I/O statements (READ and WRITE).
program containing

If the user wishes to compile a

DECtape I/O statements, he must set Switch Register bit 0 to I (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 I/O

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 I00 characters per statement.

Use of Symbolprint with FORTRAN

Symbolprint destroys a portion of the DECtape Compiler in core.
loaded if it is to be used to accept a symbolic program containing

FORTRAN Operating System with

The compiler must be re-

DECtape I/O statements.

DECtape I/O Option

When the FORTRAN Operating System is read into core, it is equipped with a switch option

governing the execution of DECtape I/O statements.

If the user wishes to run a program containing

such statements he must set Switch Register bit 0 to 1 before running his program.

There is a further condition which must be observed, since
considerable amount of additional processing routines.

DECtape [/0 requires a

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 de-

fined as a page of memory reserved to handle transfers to and from a block of DECtape

(12810 data

words).
The DECtape routines transfer one full block from tape to one page of core (and vice versa),

therefore, even if the block contains only one data word, the whole block will be read into the OP SYS

buffer, overlaying whatever had been there.
To store variables or arrays of data on

DECtape requires two steps:

From the locations assigned by the OP SYS to the variables or allotted to the arrays by
a.
DIMENSION 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.)

He must write the buffer onto a block of DECtape.

This is done by a physical 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 will cause an error diagnostic.
statement.

To retrieve data from
a.

DECtape is also a multiple operation.

The programmer is responsible for remembering which block contains the data he wishes

to retrieve.

He must read this block into the OP SYS buffer using a physical READ.

He must remember in which order he stored the variables or arrays and reference them here

(See below.)

in the same order.

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 have no effect on the physical 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 will 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

O, 0, FORMAT, 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 0 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
BLOCK, FORMAT, LIST

WRITE UNIT,

It is not necessary to specify a list on a physical READ or WRITE but it is advisable, since it does

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.

f DIMENSION IDAT (128)

no,- FORMAT (1)
Part I

J has been previously detinedS 128

DO IOO I=I ,J

0, IIO, IDAT (I)
CONTINUE
100;
j
k
WRITE 0,

Part 2

WRITE MU,

This DO loop will WRITE J number
of elements of IDAT into the butter

This statement will then store data

MBLK, IIO

From buffer to tape

IDAT

DEC

BUFFER

‘0

0900
~

‘x

\><°

50‘“
1
V0“

TAPE

to‘“

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

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

DIMENSION IBFI

(200), IBF2 (200)

90;FORMAT (I)
I99; ACCEPT 90, J,K, MUNT, MBLK, IBLK
DO IO 1:] , J
ACCEPT 90, IBFI (I)

IO;CONTINUE
C;
C;
C;
C;

:::::::::::::

DATA HAS BEEN ACCEPTED NOW WE READ BLOCK '0'
THIS INITIALIZES AND REWINDS THE TAPE

A GOOD IDEA TO DO THIS BUT NOT ESSENTIAL

c.,
READ MUNT,

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

IBLK, 90

C;
C;
C;
C;
C;
C;

:::::::::::::

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 '0'
DO LOOP NEEDED SINCE DATA IS IN AN ARRAY
:::::::::::::

DO 31 I=I ,J
WRITE O, 0, 90, IBFI

(I)

3I;CONTINUE
C;
C;
C;
C;

C;
C;
C;
C;

:::::::::::::

TAPE IS NOW PHYSICALLY WRITTEN BY SELECTING
A LOGICAL UNIT, OTHER THAN ZERO.

:::::::::::::

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

DO 32 I=I,K

IBF2(I)

32; CONTINUE
.

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

on on
;

ALL I/O DONE HALT.

SEE PART TWO OF DIAGRAM.

:::::::::::::

***~k-k*****-k**

DO 13 I=],K
TYPE 91,1BF2 (I)

91;FORMAT (/,I)
13;CONTINUE
PAUSE

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

7546

7543

7540

7535

7534

DSR

753i

7526

7515

7504

7470

6312

724]

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 be-

tWeen these two addresses is available.

In the example there are

724T -6312 -1 =726
octal locations free.
A machine language subroutine may occupy this available space.
statement to

Use the FORTRAN PAUSE

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

effects JMS 6400.

Location 6400 should contain coding such as the following:

SUBR, o
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
stored
in memory by way of the CONTROL console switches. Once stored, it is considered
initially

permanent occupant of locations 7756 through 7777 (absolute octal addresses) and care should
be taken to keep it from being destroyed.
to be a

A paper tape to be read in by the RIM Loader must be in RIM format:

Tape Channel
G) 7654S321
0 0 0 0

CD<DOC>O<DO—I
.—a

O O 0

Leader/Trailer code

Absolute address to

contain next 4 digits

Contents of previous

4 digit address

Address
Contents

(ETC.)
0 0 0 0

(ETC .)

Lead/Trailer code

0 0

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

Abs
Addr

Octal

Symbolic

Contents

/clear AC and flag
/skip if flag=l

7756,
7757,

6032
603]

KSF

7760

5357

JMP .-I

7761

6036

KRB

7762,
7763,
7764,
7765,
7766,

7106

CLL RTL

7006

RTL

/ch 8 in ACO

7510

SPA

/checking For leader

5357

JMP BEG +l

7006

RTL

/found leader
/OK, ch7 in link

7767,
7770,
777],

6031

KSF

5367

JMP .-l

6034

KRS

/read, do not clear

7772,
7773,

7420

SNL

3776

DCA I TEMP

/checking For address
/store contents

BEG,

KCC

/looking For char
/read buffer

Abs

Symbolic

Contents

7774,
7775,

3376

7776,
7777,

4.
as

Octal

Addr

DCA TEMP

5356

JMP BEG

TEMP,

5301

/store address
/next word

/temp storage
/iump to start of bin loader

0
0

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

follows:

Q a

Set 7756 in the switch register (SR)

0—

Press LOAD ADDRESS

Set the First instruction in the SR (6032)

Press DE POS IT

Set the next instruction in the SR

Press DEPOSIT

Repeat steps e and f until all 16 instructions have been deposited.

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

Abs

Oc_ral

Addr

7755) is as follows:

Symbolic

Contents
RFC

/clear flag and fetch char. into

601 1

RSF

/skip when flag=1

7760

5357

JMP .-I

776 1

60 I 6

RRB RFC

/read buffer into AC, get next

7762

7106

CLL RTL

7763

7006

RTL

7764

7510

SPA

/rotate channel 8 into
/AC bit 0
/is it leader

7765

5374

JMP TEMP—2

/yes clear AC

7766

7006

RTL

/NO rotate channel 7 to LINK

7767

6011

RSF

7770

5367

JMP .-I

7771

6016

RRB RFC

7772

7420

SNL

7773

3776

DCA I TEMP

7774

3376

DCA TEMP

7775

5357

JMP BEG +1

7776

0000

7777

5301

7756

6014

7757

BECL

buffer

char. into buffer

TEMP,

0
0

/link set=origin
/store data
/store address
/next word

/temporary storage
/JMP to start of BIN Loader

BINARY LOADER (BIN)
l.

The BIN Loader is used to read in the machine language tapes.

length oF a comparable RIM Formatted tape.

is about one halF the

It can,

A binary-Formatted tape

therefore, be read in about

twice as Fast as 0 RIM tape and is, For this reason, the more desirable format to use with the IQ cps

ASR33 Reader.

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 following conditions exist:
a.

No checksum error:

halt with AC=0.

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
b.

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

tape.

should be obtained.

IF a diFFerent checksum error appears aFter readin, the appropriate

maintenance procedure should be Followed.

APPENDIX B
PREPARATION OF SYMBOLIC (SOURCE) TAPE

I.
use

Symbolic tape preparation using Symbolic Tape Editor. (It is to the user's benefit to

the Editor to put his source program on tape since using the Editor minimizes the chance of ex-

traneous

characters getting on the tape and also Facilitates deletion and correction of statements).
a.

Load Symbolic Tape Editor using Binary Loader.

Start Editor at I76 (Load Address,

Type Ag and type the symbolic source program.

Hold the CTRL key and press the FORM key; the bell will sound.

Type PD), followed by FA) when punching stops.

Start).

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

Symbol tape preparation off-line using Teletype only.
a.

Turn power on in computer (key on left in

PDP-8, switch on right in PDP-5).

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+l times, press rubout N+l times, and continue.

b.
Characters, words, or statements can be inserted or deleted after the entire symbolic tape has been prepared. Use the following procedures to accomplish such changes.

Duplicate the tape up to the point at which it is desired to make
placing the tape in the reader, starting the
with
READER switch using the printout as a
the
reader
the
and
reader,
stopping
Continue
guide). Next type the insertion.
by pressing the READER switch to start
and duplicate the remainder of the tape.
(I)

Insertions

insertion (by turning the punch on,

(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
Release in reverse order or a P will be punched on the

REPT key and then the P with the right hand.

tape.

APPENDIX C
FORTRAN OPERATING PROCEDURES

COMPILER
1.

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

Put the starting address of the Compiler (0200 octal) into the switch register and press

LOAD ADDRESS.

reader,
out

See I/O Control.

Set I/O switches.

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

Press START.

At the end of compilation, the computer will halt with the run light off.

To compile additional programs,

turn the reader and

(Conditional)

place the source language tape in the appropriate
I/O selections cannot be changed with-

punch on, and press CONTINUE.

reloading compiler.

SYMBOLPRINT

Symbolprint is run immediately after compiling a program and before compiling another or
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

WRITE statements.

DECtape READ and

The Compiler must therefore be reloaded if it is to compile a source program con-

taining such statements.
1.

Load Symbolprint with the Binary Loader.

Set 0600 in the switch register.

Press LOAD ADDRESS and START, see Chapter 8.

OPERATING SYSTEM

(OBJECT TIME SYSTEM)
l

To load a compiled program:
0.

Load the FORTRAN Operating System using the Binary Loader.

If using 552 DEC-

tape, load the 552 overlay tape using the Binary Loader.
b.

Place the Compiler output (interpretive code object tape) in the Teletype or photoTurn on the reader, making sure ASR33 is ON LINE.

electric reader.
c.

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

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.

The Operating System halts at the end of loading. The loading is correct if the
checksum difference which appears in the AC equals 0.
f.

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

To execute a program after loading
a.

Set SWITCH REGISTER bits 0, I, and 2 (as shown below).
If input is to be from paper tape, put the data tape in the appropriate reader and
If output is to be punched, turn punch on.

turn the reader on.
c.

Press CONTINUE.

NOTE

Once loaded, a program can be executed any number

of times.

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

To Re—execute a Loaded Program
a.

Set the SWITCH REGISTER to 020i; 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 will speed up the execution of the program by a factor of

about 2.

I/O CONTROL
The selection of I/O devices for both COMPILER and OP SYS is controlled by setting the

switches as shown below:

.91

Switch

Meaning

Number

Posutlon

The program contains only paper tape I/O statements.

The program contains

Compiler:

OP SYS:

DECtape I/O statements.

Use the Teletype reader for input of source tape.
Use the Teletype reader for loading the object

program and the keyboard for ACCEPT statements.

C-2

Bit

Switch

Number

Position

Use the high-speed reader.

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).

APPENDIX D
FORMAT OF COMPILER OUTPUT

INTERPRETIVE CO DE
200 codes (leader,

ignored by loader).

Data blocks, each as Follows:
Cl.

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.

APPENDIX E
ASR33 8-BIT CHARACTER SET

8-Bit Code

(in Octal)

8-Bit Code

(in Octal)

301

241

302

242

303

304

305

306

307

310

311

(
)

312

313

314

315

316

317

257

320

272

321

;

273

274

322
323

275

324

300

325
326
327

276
277

330

331

E
/
j

333

X
Z

332

336

260

337

261
262

Leader/Trai Ier
Carriage Return
Space

215

Rubout

377*

263
264
265
266

Blank

000*

267

270

271

2
3
4
5

,
-

Line Feed

243
244
245
246
247
250
251
252
253
254
255
256

334
335

200*
212*
240

*Ignored by the operating system

E-1

APPENDIX F
PDP-8 FORTRAN SOURCE PROGRAM RESTRICTIONS

The Following limits are imposed upon all FORTRAN source programs for the PDP—8:
Not more than 896

data cells.

This includes all dimensioned variables, user-defined

variables, constants, and all constants generated by the usage of a DO loop.
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

Multiple references to the same undefined statement numbers are considered as one

in the program.

reference.
3.

Not more than 64 different variable names per program.

Not more than l28 characters per input statement.

(When using the DECtape Compiler,

the input statement size is reduced to 100 characters.)
Not more than 40 numbered statements per program.

PDP-8 COMPILER AND OPERATING SYSTEM CORE MAP

The Compiler occupies the following core locations:
3

7200

—

Compiler itself plus tables

7600

Compiler tables (undefined forward reference

7600

table, etc.)
The Operating System occupies locations:
0

0
Locations 5200
cations 6000

5200

Operating System for paper tape I/O

6000

Operating System for DECtape [/0

7576 are available for the user's program when using paper tape input/output or lo-

7576 when using

DECtape.

NOTE

The

89610 data word restriction applies.

F-i

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 mes-

sages on the on-line tape-teleprinter.

These messages indicate the source of the error and direct the

programmer's efforts to correct the error.
To speed up the compiler process, the compiler prints out only an error code.
mer

The program-

then looks up the error message corresponding to the code in Table A-l and takes the appropriate

corrective measures

DYNAMIC ERROR CORRECTION
A user may choose to compile in either of two modes:
rection mode.

the normal mode or the dynamic cor-

The latter allows the user to correct a statement, which the compiler has determined

contains a source-language error,

by reentering the offending line via the tape teleprinter without

having to physically correct the symbolic tape and recompile.

This Feature is not implemented in the

high-speed reader version of the compiler since the higher speed of the device makes recompilation
easy.

To choose the dynamic correction mode:
I.

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

LOAD ADDRESS.

Set SR bit II to I, press START (can only be used with low-speed paper tape I/O).

IF an error is detected, the diagnostic prints out in the normal Fashion and the computer halts.
To correct the statement:

Turn READER switch to FREE.

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

Type the new line in its entirety,* obeying all rules For the source language and ter-

minating the statement with a carriage-return line-Feed.
4.

Turn reader on and compilation will continue.

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

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

6-]

COMPILE TIME DIAG NOSTICS

Format of Diagnostics

XXXX

l—The 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:

A=I (J+i)

10;

B=A*(B+SINF(THTA))

During compilation of the above statements the following error code would be printed,
l0

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 ll shows that the number of left and

right parentheses in the statement is not equal.

The statement is examined and corrected; then com—

pilation is resumed.
Table A—l

DEZZZC:HC

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 +
02

Compiler error

Reload Compiler and repeat compilation process.

Contact Software Quality Control, PDP-8 Division if this reoccurs.
03

A comma has been used illegally in an arithmetic statement.

Too many operators appear in a single statement.

A function argument is in fixed mode, e.g. ,

SINF(INC).

Table A-l

(Cont)

Diagnostic

Condltlons

Code

A variable subscript is in floating-point mode.
dicate that an operator is missing, e.g. ,

This could also in-

A+B(C+l .) for A=B*(C+l .).

More than 64 (decimal) different variable names have been used in

the program.

requirements have overlapped.

Program too large

There is an unequal number of right and left parentheses in a state-

program and data

ment.

12
l3

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 dimen-

subscripted variable does not appear in a diIt might also indicate that an operator is missing
in a fixed-mode expression, e.g.
A=I(J-K) for A:I*(J-K).

sion statement,

or a

mension statement.

long; more than 128 characters have been counted
including spaces except in format statements where all legal TTY

Statement too
not

characters are counted.
17

A floating—point operand should have been fixed—point, e.g. , DO
TO 1:1, 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
23

Too many
are

C=A+(C+(D+(E+.

(more than 20) statements have been referenced before they

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

where the number usually reserved for the last numbered statement (xxxx) is replaced by the missing
statement

number.

e.g., GO TO TOO

The diagnostic would appear as follows where statement l00 is never defined.
100

OPERATING SYSTEM DIAGNOSTICS
Not all errors are detected by the compiler. Some errors can only be detected by the object
time system.

obiect system.

Also, there are some conditions which indicate errors on the part of the compiler and/or
When such an error occurs during running of a program, the computer types out an error

message containing the word "TILT'I and an error number.

The computer then halts.

If the CONTINUE

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

Table A-2
E rr o r

Possnble Cause

Number
ll

Attempt to divide by zero

Action Taken
Quotient set to plus or minus largest number

representable in computer; then continue executing instructions.
12

Floating point exponent on input greater than plus or minus

System executes next instruction.

2047
13

Illegal operation code (either

System executes next instruction.

compiler error, or data stored
over

program, or transfer to data

section)
14

Transfer to core location zero or

No recovery possible.

one

Non-format statement used for a

System executes next instruction.

format
16

Illegal format statement consti-

System examines next constituent.

tuent

Attempt to fix large floating
point number

Attempt to take square root of a

System takes square root of absolute value.

negative number
2]

Attempt to raise a negative number to a power

Attempt to find the logarithm of
zero or a

negative number

System raises absolute value to the power
specified.
System attempts to find logarithm of absolute
value. Note that log (absolute value (0))
still gives an error halt.

Table A-2 (Cont)
Error

Action Taken

Possible Cause

Number

The operating system halts with the called
unit in bits 0-2 of the AC (0-3 is using 552/

Select error

Recovery is. possible by correcting the
logical unit and pressing continue.

555).
32

The program halts with the error status in the

Physical Tape Error

AC.

(The configuration of bits is dependent

upon the tape control being used.)

DECtape buffer exceeded

DECtape control switch set incorrectly

One of the stacks used by the

system has underflowed.
more

(i.e.

No rec0very possible.
,

program.

data has been requested

than was placed on the stack)
77

One of the stacks has overflowed

(i.e.

data placed on it

than there is storage in the ma-

chine.)

Same as Error 76

Try recompiling the

INDEX

Arithmetic

7-2

Selection

Algebraic

3-I

Diagnostics (see Errors)

Constants

3-2

Compiler G-2, -3

3-I

Expressions
Functions

Operating System 6-4, -5

3-4

DIMENSION Statement

Integers 3-2

Statements

DO statement

3-4, -5

6-5, -6, -7

Illegal Nesting

Variables 3-2

Arrays

5-2

Array Names

Operators 3-3, 3—6, —7

5-]

6-7

Loops 6—5, -6
Nesting Loops 6-6

5-I

Subscripted

5—l

E Format

7-4, -5, -6

Branches and Loops 6-2, 6-8

Editor, Symbolic Tape B—l

Character Set 2-4, E-I

END Statement

COMMENTS 2-2, -3

Errors

6-I

(see Diagnostics)

Corrections 7-6, G-I

Compiler
Diagnostics 6—I

Messages G-I

Operating Procedures C-I

Expressions, Arithmetic 3-I

Output D-I

Field
8-I

with DECtape
Constants

Identification

Fixed Numbers 4-I, -2

3-2

CONTINUE Statement

2-3,

6-7

Floating-point
Decimal Fraction and Integer 7-5

Core Map

Compiler and OP SYS

F—I

Format

4-2

Numbers 4-I

DECtape

I/O Options

8-I

Statement and Operation 8-2

Decimal Integers

Exponents 3—2

3-2

Representation 4-I, —2
Format

Control Elements 7-7, -8
E andI

7—4, -5, -6

Floating-point

Devices

I/O

2—2

7-1

Hollerith 7-I

4-2

Input

Operating Procedures

7-5

A-3

Program 2—]

BIN

Slash 7-8

Compiler

Statement Number

I/O Control C-2, -3

7—3

Operating System (OPSYS)

7-3, 7-6

FORMAT Statement

Functions, Arithmetic 3-4

RIM

6—2

GO TO Statement

Compiler 8-l, C-l
DECtape 8-l, -2

Output 7-7, -8
7-4, -5, -6

Operating System

Input/output

System 8—], C—1, -2, -3

7-]

FORMAT
Statements

Operators

Unary Minus 3-3

7-2, -3
,

6-5

Output

7-5

E and I Format

Floating and Fixed Modes
7-6

FORMAT

Loaders

Parentheses, use of 3-4

A-I, -2, -3

Numbers
Fixed

7-6

Data Word

6-2

RIM and BIN

3-3, -4

Arithmetic

7-5

Integers 3-2, 4-l

Summation

3-3

Operators
4-], -2

PAUSE statement

Floating-point 4-], -2

Positive or Negative

Statement

4-]

4-2

2-2, 7-3

6-I

Example

2-l
2-]

Restrictions

Variables, Types of 4-2

Operating Procedures

Control

Format

OP SYS (Operating Procedures)

C-I

6-]

Program

Integers 4-]

Representation

8-l

Symbolprint

DECtape Options 8—I
Devices

A-l, -2, -3

Loaders

6-3

IF Statement

C-l

Operation

Hollerith Format

I Format

C-l , -2

A-l

Symbolprint

6-8

Computed

C-l

F-I

Source Preparation

8-2

Termination

B-I

6-I

Terminator (END)

6-]

7-6

Quote, Hollerith Output 7-7
2-3

Single Quote, use of
READ Statement

8-2, -3

F-l

Semicolon 2-2

Slash, Control Element
Format

7-8

Statements

2-]

2-2

Comment

Continuation

DECtape

7-8

3-4

Arithmetic

2-3

8-2

6-]

GOTO 6-2

Input/Output 7-2, -3
Number 2-2
PAUSE 6-1
READ

8—2, -3

STOP

6-]

Types of

2-2

Variables 3-2

STOP

3-2

WRITE Statement

Source Program

WRITE

Arithmetic

Subscripted 5-], -2

Restrictions

END

Variables, types of 4-2, -3

8-2, —3

6—]

Subscripts

5-l

Arrays

5—]

Variables 5-l

Symbolprint 8-], 9—l, -2
Operating Procedures

C-l

with DECtape 8-]

Tape, Symbolic Preparation 8-]

8-2, -3

323:;rEfl'SZL‘Z'LthEé’éfl‘é'a‘i°"

printed in USA.

EflEflIEII