SVS_Fortran_CPM_68K_Sep83 SVS Fortran CPM 68K Sep83
SVS_Fortran_CPM_68K_Sep83 SVS_Fortran_CPM_68K_Sep83
User Manual: SVS_Fortran_CPM_68K_Sep83
Open the PDF directly: View PDF 
.
Page Count: 192
| Download | |
| Open PDF In Browser | View PDF |
Fortran
Reference Manual
SIlicon Valley Software, Inc.
10011 N. FoothIll Blvd .. Suite III
CupertIno. CA 95014
FORTRAN
Language Reference Manual
Version 2.0 of 1st September 1983
Silicon Valley Software, Inc.
10011 North Foothill Blvd. Suite 111
Cupertino
California 95014
This PORTRAN Reference Manual was produced by:
Jeffrey Barth, R. Steven Glanville, and Henry McGilton.
Silicon Valley Software, Inc.
Publication Number: 810715-02
Copyright 1983 by Silicon Valley Software, Inc.
All rights reserved. No part of this PORTRAN Reference Manual
may be reproduced, translated, transcribed or transmitted in any
form or by any means manual, electronic, electro-magnetic, chemicalor optical without explicit written permission from Silicon
valley Software, Inc.
Preface
PREFACE
This manual is a reference manual for SVS FORTRAN-77
an
implementation of the full ANSI FORTRAN-77 computer programming
language for Motorola MC68000 computer systems.
FORTRAN is one of the most widely used programming languages to
date. FORTRAN is primarily oriented towards scientific computing
applications.
FORTRAN's evolution has developed from its early beginnings as
FORTRAN-II through to FORTRAN-IV, for which an ANSI standard was
adopted in 1966. By the middle of the 1970's decade, it was
apparent that many of the most widely used extensions to
FORTRAN-IV could form the basis for a new FORTRAN language standard.
Atl updated language (called FORTRAN-77 by consensus) was
announced in 1977. The formal standard was issued by the American National Standards Institute (ANSI) in 1978.
The ANSI standard for FORTRAN-77
actually
defines
two
languages: a "full" language and a "subset" language. The subset
omits certain items such as the complex data-type and listdirected input-output. SVS FORTRAN implements the full language.
SCOPE OF THIS MANUAL
This is a reference manual, as opposed to a "how to write FORTRAN programs" manual. There are several tutorial-style FORTRAN
books on the market for novice FORTRAN users.
OVERVIEW OF THIS MANUAL
Chapter 1 - "Introduction" is a general overview of FORTRAN.
Terms and concepts of FORTRAN are introduced here, as is the
metalanguage that this manual uses to describe FORTRAN.
Preface
Chapter 2 - "Lines, Statements, and Control Flow" introduces
the ideas of lines, statements and execution seguence.
Chapter 3 - "Data Types and Constants" introduces FORTRAN
and constants.
data
~
Chapter 4 - "FORTRAN Names, Arrays, and
FORTRAN ~, arrays and substrings.
Substrings"
discusses
Chapter 5 - "Expression" describes the rules for expressions in
FORTRAN.
Chapter 6 - "Specification Statements" is a description of
specification statements used to declare data variables and their
attributes.
Chapter 7 - "Data Initialization" discusses data initialization, that is, static data initialization via DATA statements.
Chapter 8 - "Assignment Statements" describes assignment statewhereby variables are assigned new values.
~
Chapter 9 - "Control Statements" presents
that direct the flow of program execution.
control
statements
Chapter 10
"FORTRAN Input Output" covers input outputstatements which are the means whereby a program communicates
with the world external to the computer.
Chapter 11 - "Format Specifications" is about format statements
which describe the conversion process between internal data
representations and external formats.
Chapter 12 - "Program and Subprogram Structure" describes P1 9and sUbProaram structure where~y large programs can be sp lt
lnto smaller an more manageable unlts.
~ram
Chapter 13 - "FORTRAN Compile Time Options" describes the
piler options which control the actions of the compiler.
com-
Appendix A is a list of FORTRAN error messages.
Appendix B contains a table of FORTRAN intrinsic functions
the "built-inn functions for performing mathematical computations.
Appendix C describes FORTRAN's data representation methods
parameter passing mechanism.
and
Appendix D is an ASCII character set chart.
Appendix E - "Operating the SVS FORTRAN System" describeS the
system independent aspects of operating the system and the
Preface
considerations involved in linking programs
languages.
written
in
several
Appendix F - ·Operating System Specific Information" contains a
description of how to run the FORTRAN compiler on the host
operating· system, and also covers details of specific dependencies and interfacing requirements (if any) of the host operating
system.
CONTENTS
Chapter 1 - Introduction.....................................
1.1 Overview of the FORTRAN Language........................
1.2 Notation and Terminology Used in this Manual............
1.3 Basic Elements of FORTRAN...............................
1.3.1 FORTRAN Character Set............................
1.3.2 Collating Sequence and Graphics..................
1.3.3 Use of Spaces or Blanks or Tabs..................
1
1
4
5
5
6
6
Chapter 2 - Lines, State.ents, and Control Plow..............
2.1 Lines...................................................
2.1.1 Comment Lines....................................
2.1.2 Initial Lines ••••••••••••••••••••••••••••••••••••
2.1.3 Continuation Lines •••••••••••••••••••••••••••••••
2.1.4 Compiler Directive Lines •••••••••••••••••••••••••
2.2 Statements ••••••••••••••••••••••••••••••••••••••••••••••
2.2.1 Statement Labels •••••••••••••••••••••••••••••••••
2.2.2 Order of Statements and Lines ••••••••••••••••••••
. 2.3 Execution Sequence and Control Transfer •••••••••••••••••
9
9
9
10
10
11
11
12
13
14
Chapter 3 - Data Types and Constants •••••••••••••••••••••••••
3.1 Data Type Rules •••••••••••••••••••••••••••••••••••••••••
3.2 Constants •••••••••••••••••••••••••••••••••••••••••••••••
3.3 Integer Data Type •••••••••••••••••••••••••••••••••••••••
3.4 Real Data Type ••••••••••••••••••••••••••••••••••••••••••
3.4.1 Infinite and Indeterminate Real Values...........
3.5 Double Precision Data Type ••••••••.•••••••••••••••••• ~.
3.6 Complex Data Type.......................................
3.7 Character Data.Type ••.••••••••••••••••••••••••••••.•••••
3.8 Logical Data Type •••••••••••••••••••••••••••••••••••••••
17
17
17
18
19
20
20
20
21
22
Chapter 4 - PORTRAN Names, Arrays, and Substrings ••••••••••••
4.1 FORTRAN Names •••••••••••••••••••••••••••• ~ • • • • • • • • • • • • • •
4.1.1 Scope of FORTRAN Names...........................
4.1.2 Undeclared FORTRAN Names •••.••••••••••••.••••••••
4.2 Array Declarations ••••••••••.•••••••••••••••••••••••••••
4.2.1 Dimension Declarations •••••••••••.•••••.•••••••••
4.2.2 Kinds of Array Declarations ••••••••••••••••••••••
4.2.3 Actual Arrays and Dummy Arrays •••••••••••••••••••
4.2.4 Referencing Array Elements - Array Subscr ipts. • • • • • • • • • • • • • • • • • • • • . • • • • • • • • • • • .• • • • • • • .
4.2.5 Using Unsubscripted Array Names ••••••••.••••••.••
4.3 Character Substrings •••.••••.••••••.•••••••••••.••••••••
23
23
24
25
25
26
26
27
Chapter 5 - Expressions ...•••••••••••••••..•.•••••...••••••.•
5.1 Arithmetic Expressions ••••••••••••••••••••••••••••••••.•
5.1.1 Arithmetic Operators ••••••••••••••••••.•••••••••
5.1.2 Arithmetic Operands •••.•.•••.•••.•.•••...••••••.•
5.1.3 Constant Expressions •...•••••.••••••••••.••••••.•
5.1.4 Type Conversion Rules for Arithmetic Expressions ••••••••••••••••••••••••••••••••••••••••••.•
31
31
31
32
33
-
i -
27
28
29
33
Rules for Add, Subtract, Multiply and
Di vide. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
5.1.4.2 Rules for Exponentiation Operator.......
Coercion Rules for Integers of Different
5.1.4.1
5.1.5
34
34
Size.............................................
35
Integer Division •••••••••••••••••••••••••••••••••
Character Expressions •••••••••••••••••••••••••••••••••••
5.2.1 Restrictions on the use of String Expressions. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
Relational Expressions •••..•••••••••••.•.••..•.•••••••••
5.3.1 Arithmetic Relational Expressions ••••••••••••••••
5.3.2 Character Relational Expressions •••••••••••••••••
Logical Expressions •••••••••••••••••••••••••••••••••••••
5.4.1 Precedence of Logical Operators..................
Precedence of Operators.................................
Evaluation Rules and Restrictions for Expressions •••••••
5.6.1 Restrictions on Function References ••••••••••••••
36
36
36
37
37
37
38
38
39
39
40
Chapter 6 - Specification Statements •••••••••••••••••••••••••
6.1 Type Statements - Declaring Data Types ••••••••••••••••••
6.1.1 Arithmetic Type Statements •••••••••••••••••••••••
6.1.2 CHARACTER Type Statement •••••••••••••••••••••••••
6.1.3 LOGICAL Type Statement •••••••••••••••••.•••••••.••
6.2 DIMENSION - Declare Data Dimension ••••••••••••••••••••••
6.2.1 Dimension Declarators ••••••••••••••••••••••••••••
6.3 COMMON - Declare a COMMON Block •••••••••••••••••••••••••
6.4 PARAMETER - Make a Symbolic Association •••••••••••••••••
6.5 IMPLICIT - Establish Default Data Type ••••••••••••••••••
6.6 EXTERNAL - Declare External or Dummy Procedure ••••••••••
6.7 INTRINSIC - Declare Intrinsic Function ••••••••••••••••••
6.8 SAVE - Retain Definition Status •••••••••••••••••••••••••
6.9 EQUIVALENCE - Share Storage Between Elements ••••••••••••
6.9.1 Restrictions on EQUIVALENCE Statements •••••••••••
41
41
42
42
43
44
44
44
45
46
47
47
48
49
50
Chapter 7 - Data Initialization •••••••••••••••••••••••.••••••
7.1 Initializing Character Variables ••••••••••••••••••••••••
7.2 Initializing Non CHARACTER Variables to CHARACTER
Values. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
7.3 Implied DO in DATA Statements...........................
51
52
Chapter 8 - Assignment Statements ••••••••••••••••••••••••••••
8.1 Arithmetic Assignment •••• ~ ••••••••••••••••••••••••••••••
8.2 Logical Assignment......................................
8.3 Statement Label Assignment ••••••••••••••••••••••••• ~ ••••
8.4 Character Assignment •••••.••..••••••••••••.•••••••••••••
57
57
58
59
60
Chapter 9 - Control Statements..............................
9.1
Block IF THEN ELSE Statement.........................
9.1.1
Block IF Statement...........................
9.1.2
ELSEIF Statement.............................
9.1.3
ELSE Statement...............................
9.1.4
ENDIF Statement..............................
61
62
63
64
64
65
5.1.6
5.2
5.3
5.4
5.5
5.6
- ii -
53
54
9.2
9.3
9.4
9.5
9.6
9.7
9.8
9.9
9.10
Logical IF Statement •••••••••••••••••••••••••••••••••
Arithmetic IF Statement ••••.•••••••••••••••••••••••••
DO Statement - Loop Control ••••••••••••••••••••••••••
9.4.1
DO Loop Initialization Sequence ••••••••••••••
9.4.2
DO Loop Incrementation Processing ••••••••••••
9.4.3
Events Which Terminate a DO Loop •••••••••••••
CONTINUE Statement - Null Statement ••••••••••••••••••
STOP Statement - Stop Program Execution •••••••.••••••
PAUSE Statement - Suspend Program Execution ••••••••••
Unconditional GO TO Statement ••••••••••••••••••••••••
Computed GO TO Statement •••••••••••••••••••••••••••••
Assigned GO TO Statement •••••••••••••••••••••••••••••
Chapter 10 - Input and Output •••••••••••••••••••••••••••••••
10.1
Overview of the Input-Output System..................
10.1.1 Records......................................
10.1.2 Files........................................
10.1.3 Properties of Files..........................
10.1.3.1 File Name..........................
10.1.3.2 File Position......................
10.1.3.3 Formatted and Unformatted
Files •••••••••••••••••••.••••••••••
10.1.3.4 Sequential and Direct Access
Files ............................. .
10.1.4 Internal Files •••••••••••••••••••••••••••••••
10.1.4.1 Special Properties of Internal
File s ............................. .
10.1.5 Units ••••••••••••••••••••••••••••••••••••••••
10.2
General Discussion of the Input Output System ••••••••
10.2.1 Pre-Connected Files •••••••••••••••.••••••••••
10.2.2 Examples of Common Input Output Opera-
tions ....................................... .
10.2.3
10.2.4
Less Common File Operations •••••••.••••••••••
Limitations of FORTRAN Input Output System ••••••••••••••••••••••••••••••••••••••••••
o
73
73
73
74
74
74
75
75
75
76
76
76
77
78
78
79
80
10.2.4.1
10.3
Direct Files must be on Blocked
Dev ices ••••••••••••••••••••••••••••
10.2.4.2 No Character Compression in Direct
Files ••••••••••••••••••••••••••••••
10.2.4.3 BACKSPACE Only Applies to Files on
Blocked Devices •••••••••.••••••••••
10.2.4.4
Length Limitations on Formatted
Records ••••••••.•••••••••••••••••••
10.2.4.5 BACKSPACE may not be used on
Unformatted Sequential Files •••••••
10.2.4.6 Side Effects of Functions Used in
Input Output Statements •.•.••••••••
Elements of Input and Output Statements •••.••••••••••
10.3.1 The Unit Specifier 'u' •••••••••••••••••••••••
10.3.2 The Format Specifier 'f' •••••••••••••••••••••
10.3.3 The Record Number 'rn' •••••••••.•••••••••••••
10.3.4 The End of File Exit Specifier •..•..•.•..••.•
65
65
66
67
68
68
69
69
69
69
70
71
- iii -
80
80
81
81
81
81
81
81
82
83
83
10.3.5
The Error Exit Specifier •••••••••••••••••••••
The Input Output Status specifier ·ios· ••••••
10.3.7 The Input-Output List 'iolist' •••••••••••••••
10.3.7.1 Input and Output Objects •••• ~ ••••••
10.3.7.2 Implied DO Lists •••••••••••••••••••
The Specific Input and Output Statements •••••••••••••
10.4.1 OPEN Statement •••••••••••••••••••••••••••••••
10.4.2 CLOSE Statement ••••••••••••••••••••••••••••••
10.4.3 READ, WRITE and PRINT Statements •••••••••••••
10.4.4 File Positioning Statements ••••••••••••••••••
10.4.4.1 BACKSPACE Statement - Backspace a
File...............................
10.4.4.2 ENDFILE Statement - Write an Endfile Record........................
10.4.4.3 REWIND Statement - Rewind a
10.3~6
10.4
10.4.5
10.5
File...............................
INQUIRE Statement - Obtain File Properties ........................................ .
List Directed Input and Output •••••••••••••••••••••••
10.5.1 List Directed READ •••••••••••••••••••••••••••
10.5.1.1 List Directed Value Separators...............................
10.5.1.2 List Directed Input Values.........
10.5.2 List Directed WRITE and PRINT................
Chapter 11 - Pormat Specifications ••••••••••••••••••••••••••
11.1
FORMAT Specifications and the FORMAT Statement •.•••••
11.2
Interaction Between Format Specifications and I/O
Lis t ..•....•....•.•••.•..••........•.....•..•..•....•
11.3
Edit Descriptors •••••••••••••••••••••••••••••••••••••
11.3.1 Repeatable Edit Descriptors ••••••••••••••••••
11.3.1.1 Numeric Editing ••••••••••••••••••••
11.3.1.2 I - Integer Editing ••••••••••••••••
11.3.1.3 F - Real Editing •••••••••••••••••••
11.3.1.4 E and D - Real Editing •••••••••••••
11.3.1.5 G - Real Editing •••••••••••••••••••
11.3.1.6 Formatting Extreme Values ••••••••••
11.3.1.7 L - Logical Editing ••••••••••••••••
11.3.1.8 A - Character Editing ••••••••••••••
11.3.2 Non Repeatable Edit Descriptors ••••••••••••••
11.3.2.1 'xxx' - Apostrophe Editing •••••••••
11.3.2.2 H - Hollerith Editing ••••••••••••••
11.3.2.3 X and T - Positional Editing ••••.••
11.3.2.4 Slash Editing - End of Transfer on
Record ••••••••..•••••••••••••••••••
11.3.2.5 Backs1ash or Dollar Editing Inhibit End of Record ••••••••••••••
11.3.2.6 Colon Editing - Conditional Termination •••••••••••••••••••••••••••••
11.3.2.7 P - Scale Factor Editing •••••••••••
11.3.2.8 BN and BZ - Blank Interpretat ion •.•...••••.••.•••••..•••.•..•.•
- iv -
83
83
84
84
85
85
85
87
88
89
90
90
90
91
95
95
95
95
97
99
99
101
103
103
103
104
104
105
105
106
107
107
107
107
108
108
109
109
110
110
III
S, SS and SP - Sign Control Editing. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . .
III
Chapter 12 - Program and Subprogram Structure •••••••••••••••
12.1
Main Program •••••••••••••••••••••••••••••••••••••••••
12.2
Access To Command Line Arguments .•••••••••••••••••••••
12.3
Formal Arguments and Actual Arguments ••••••••••••••••
12.3.1 Argument Association •••••••••••••••••••••••••
12.4
Subroutines ••••••••••••••••••••••••••••••••••••••••••
12.4.1 SUBROUTINE Statement •••••••••••••••••••••••••
12.4.2 CALL Statement •••••••••••••••••••••••••••••••
12.5
Functions ........................................... .
12.5.1 External Functions •••••••••••••••••••••••••••
12.5.2 Intrinsic Functions ••••••••••••••••••••••••••
12.5.3 Statement Functions ••••••••••••••••••••••••••
12.6
ENTRY Statement ••••••••••••••••••••••••••••••••••••••
12.6.1 Restrictions on the ENTRY Statement ••••••••••
12.7
RETURN Statement •••••••••••••••••••••••••••••••••••••
Definition Status ••••••••••••••••••••••••••••••••••••
12.8
12.9
BLOCK DATA Subprogram ••••••••••••••••••••••••••••••••
12.10 The FORTRAN Intrinsic Functions ••••••••••••••••••••••
113
113
114
115
115
117
117
118
119
120
121
121
122
123
124
125
125
126
Chapter 13 - FORTRAN Compile Time Options •••••••••••••••••••
13.1
$INCLUDE - Include Source File •••••••••••••••••••••••
13.2
$XREF - Generate Cross Reference •••••••••••••••••••••
13.3
$SEGMENT - Designate Segment Name ••••••••••••••••••••
13.4
$COL72 - Restrict Source Lines to 72 Columns •••••••••
13.5
FORTRAN-66 Compatibility Options •••••••••••••••••••••
13.5.1 $F66oo - Implement FORTRAN-66 DO Loops •••••••
13.5.2 $CHAREQU - Character and Numeric Data
Equivalence ••••.•••••••••.•••••••••••••••••••
13.5.3 $INT2 - Make Integers l6-Bits ••••••••••••••••
129
129
129
129
130
130
130
Appendix A - Messages from the FORTRAN System •••••••••••••••
A.1 Compile-Time Error Messages ••••••••••••••••••••••••••••
A.2 Run-Time Error Messages ••••••••••••••••••••••••••••••••
133
133
139
Appendix B - Intrinsic Functions ••••••••••••••••••••••••••••
B.1 Notes on the Intrinsic Functions •••••••••••••••••••••••
B.2 Restrictions on Ranges of Arguments....................
B.3 Non Standard Intrinsic Functions and Subroutines •••••••
B.3.1 POKE - Store Into Arbitrary Memory Location. • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • •• • • • • • • • • •
B.3.2 IPEEK - Read From Arbitrary Memory Loca-
143
146
147
148
11.3.2.9
130
130
148
VERS - Print Date and Version •••••••••••••••••••
RAN - Random Number Generator •••••••••••••••••••
IARGC - Number of Arguments •••••••••••••••••••••
GETARG - Access an Argument •••••••••••••••••••••
148
148
149
149
149
Appendix C - Data Representations •••••••••••••••••••••••••••
C.l Storage Allocation •••••••••••••••••••••••••••••••••••••
C.2 Data Representations .••••.••••.•••.••.••••••••••••.••••
151
151
152
tion ........................................... .
B.3.3
8.3.4
B.3.5
8.3.6
- v -
C.2.l
C.2.2
C.2.3
C.2.4
Representation of Integers ••••••••••••••••••••••
Representation of REAL and DOUBLE PRECISION. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • . • • • • • • • • •
Representation of Extreme Numbers •••••.•••••••••
Hexadecimal Representation of Selected
153
153
154
Numbers. • • • • • • • • • • . • • • • . • • • • . • • • • • • • • • . • • • • • • • • •
155
C.2.5 Deviations from the Proposed IEEE Standard......
C.2.6 Arithmetic Operations on Extreme Values •••••••••
C.2.7 Representation of Logicals ••••••••••••••••••••••
C.2.8 Storage of Arrays •••••••••••••••••••••••••••••••
Argument Passing Mechanism •••••••••••••••••••. ~ ••••••••
Function Results •••••••••••••••••••••••••••••.•••••••••
Register Conventions •••••••••••••••••••••••••••••••••••
155
155
159
159
159
160
160
Appendix D - ASCII Character Set Table ••••••••••••••••••••••
161
Appendix E - Operating the SVS FORTRAN Syste••••••••••••••••
E.l System Components ••••••••••••••••••••••••••••••••••••••
E.l.l Compiler Front End ••••••••••••••••••••••••••••••
E.l.2 Code Generator ••••••••••••••••••••••••••••••••••
E.l.3 Linker ••••••••••••••••••••••••••••••••••••••••••
E.l.4 Libraries •••••••••••••••••••••••••••••••••••••••
E.l.5 Error Messages ••••••••••••••••••••••••••••••••••
E.2 Command Line Directives and Compiler Options •.•••••••••
E.3 Linking Programs which Utilize Pascal and C ••••••••••••
E.3.l What Language must Supply the Main Program ••••••
E.3.2 Referring to the Command Line Arguments •••••••••
E.3.3 Dynamic Memory Allocation and Deallocation •.••••
E.3.4 Parameter Conventions •••••••••••••••••••••••••••
E.3.4.l· Calling FORTRAN from Pascal ••••••••••••
E.3.4.2 Calling Pascal from FORTRAN ••••••••••••
E.3.4.3 Calling C from PORTRAN •••••••••.•••••••
E.3.4.4 Calling FORTRAN from C .••.•••••.•••••••
E.3.S Run Time Libraries •••••.••••••••••••••••••••••••
E.3.6 Upper and Lower Case External Naming Convent ions ••••••••••••••••••••.••••••••••••••••.•••••
163
163
163
164
164
164
165
165
166
167
167
167
167
168
168
168
169
170
Appendix F - CPM Operating Syste. Specific Infor.ation ••••••
F.l Compiling a Simple Program •••••••••••••••••••••••••••••
F.2 Error Message Files ••••••••••••••••••••••••••••••••••••
F.3 Ulinker ••••••••••••••••••••••••••••••••••••••••••••••••
P.3.l Ulinker Inputs ••••••••••••••••••••••••••••••••••
F.3.2 Ulinker Outputs •••••••••••••••••••••••••••••••••
F.3.3 Running Ulinker from the Command Line •••••••••••
F.3.4 Running Ulinker Interactively •••••••••••••••••••
F.3.5 Running Ulinker with Standard Input
Redirected......................................
F.3.5.l Symbol Table Information Placed in
Output File............................
F.3.6 Treatment of Unresolved External References...........................................
F.3.1 Segments........................................
F.3.8 Errors Detected by Ulinker.......................
F.4 Linking to CPM Assembly Code...........................
P.5 Access to Command Line Arguments.......................
113
173
114
174
114
174
177
177
C.3
C.4
C.5
- vi -
171
178
178
178
118
178
179
180
Chapter 1
Introduction
Cbapter 1 - Introduction
FORTRAN is a.computer programming language oriented towards
numerical co.putations.
PORTRAN-77 is the latest Cat the time
of this writing) offering of the ANSI standardization committee.
This FORTRAN reference manual describes the language called
FORTRAN-77, as iaplemented by Silicon Valley Software, Inc.
(SVS). Pro. now on, the word FORTRAN is used to mean this implementation of PORTIAN-77.
1.1
Overview of the FORTRAN Language
A FORTRAN program is (ultimately) composed of characters.
Characters are grouped into lines. Lines are grouped into program units. Program units are grouped into programs.
A line is either a comment line, an initial line of a statement
or a-continuation line of a statement. Lines appear in columns 1
thru 120. For compatibility with older FORTRAN implementations,
the SVS FORTRAN-77 compiler will ignore lines past column 72 if
the user selects the $COL72 compiler option (see Chapter 13
"FORTRAN Compile Time Options" for a description of the compiler
options.)
Comment lines are blank lines, as are lines with the letter C
(upper-case or lower-case) or the asterisk character "*" in
column one. Comment lines can appear anywhere in a FORTRAN program, including between initial lines and continuation lines of a
statement.
The initial line of a statement has a zero or a space character
in column 6.
A continuation line of a statement has any other
character in column 6. A continuation line is also signaled by.
an ampersand character
C&) appearing in column one of a source
line.
Statements may have up to 19 continuation lines.
The initial
line of a statement may have a statement label in columns 1 thru
5. A statement label is one to five digits-rn-length. At least
one of the digits must be non-zero. A statement label serves to
"tag" a statement so that it can be referenced by other statements.
Fortran
R~ference
Manual
Page 1
Introduction
Chapter 1
Statements are broadly divided into the two groups of executable and non-executable. Executable statements perform program
actions that assign values to variables, evaluate expressions,
affect flow of execution and perform data transmission. Nonexecutable statements generally are those that specify the forms
and attributes of program objects. Statements are discussed in
more detail a few paragraphs further on.
Program objects include constants and variables. A constant is
a string of digits or other characters defIning a value that does
not change. Variables occupy storage and have values that can be
changed during program execution. Variables and constants can
have both a name and a data~. The name serves to identify
that object--rn a program. The type of a data object defines,
among other things, the amount of storage it occupies, its range
and precision, and in some cases, the operations that can be performed on it. FORTRAN names can have default data types derived
from a naming convention or the default rules can be overridden
by explicit specifications.
A variable can be a single object or it can be an aggregate.
There are two forms of aggregate data objects, namely arriY variables and character variables. An array variable is a co lection
of data occupying consecutive storage units. Arrays can have up
to seven dimensions. A character variable represents string data
and is a sequence of characters, which can be accessed individually, or collectively, in the form of a substring.
A complete FORTRAN program is composed of a main program and
any number of subprograms. Subprograms fall into the categories
of SUBROUTINE subprograms which can be activated via the CALL
statement to perform out-of-line groups of statements, FUNCTION
subprograms which compute and return a value in the context of an
expression, and BLOCK DATA subprograms which serve to initialize
data declared in COMMON blocks. The main program and subprograms
form what are called program units. In general, the terms "subprogram" and ·program unit" can be used interchangeably.
Userdefined subroutines ~nd functions are also called "procedures".
A variable may be given more than one name by a process of
association. There are several ways to associate data. The COMMON statement provides a way to share data between separate program units.
The EQUIVALENCE statement associates variables in
the same program unit. Variables may also be associated through
the argument passing mechanism when subroutines or functions are
referenced.
Names of variables have a scope which is dependent on the way
that they are defined. In general, most names (except names of
program units, common areas and certain other names) have a scope
that is local to the program unit in which they are defined. A
name defined by being called as an external function has a global scope by default.
Names defined in a common area are local
Page 2
Fortran Reference Manual
Chapter 1
Introduction
to the program unit in which they are declared. The name of the
common area itself is global. Formal parameters to statementfunctions have a scope which is local to the statement-function
statement itself.
Specification statements are one of the two major groupings of
'statements.
Specification statements serve to declare variables
and symbolic constants.
Specification statements include the
~
statement for defining the data type of a variable, the
DIMENSION statement to define the size of array variables, the
COMMON and EQUIVALENCE statements to provide association of variables, the PARAMETER statement to give a symbolic name to a constant and the EXTERNAL and INTRINSIC statements to define attributes of other program units.
The DATA statement provides a mechanism for static initialization of data.
The DATA statement includes an implied DO 190P
construct to facilitate initializing array variables in a conC1se
manner.
Expression~ combine data objects and operators
to create new
values. FORTRAN supports arithmetic, character, logical and relational expressions. Mixed-mode expressions are permitted, with
well defined rules for conversions between the operands and generation of the result.
The assignment statement assigns the value of an expression to
a variable.
There are three variations of assignment, namely
arithmetic, character and logical. The ASSIGN statement serves
to assign the value of a statement label to an integer variable.
Control statements are those that control the flow of execution
in a program. Various kinds of IF statements select other statements for execution, depending on the result of evaluating a logical or arithmetic expression.
The DO statement provides for
repetition of a block of statements while a control variable is
assigned a sequence of values.
The CALL and RETURN statements
provide for subroutine and function execution. Variations of the
GO TO statement provide for transfer of control within a program
unit.
Statement-function statements are characterized by a singlestatement "template" defined in a program unit, with operations
on dummy arguments. The statement function is referenced in a
program unit just like a function, with actual arguments supplied. The arguments are combined according to the statementfunction definition to yield a result that can be used in an
expression.
FORTRAN provides a powerful iaPut and output capability.
A
file can be external (connecte to an external device) or internal (refers to a character variable). Files can be formatted or
unIormatted.
Files can be accessed sequentially or randomly.
Fortran Reference Manual
Page 3
Introduction
Chapter 1
files can be the subject of data conversion operations
from internal storage representations to external character
string representations and vice-versa.
Formatt~d
Format conversion is performed via READ, WRITE or PRINT statements.
There is a rich set of format specifications to control
the form and layout of converted data. There is a list-directed
input-output capability, where default formatting--tules are
applied to the conversion process.
SUBROUTINE and FUNCTION program units may have arguments which
a calling routine passes to them for processing. At the time a
subroutine or function is declared, its formal arguments are
declared.
At the time the subroutine or funct10n is referenced,
actual arguments are substituted. for the formal arguments.
A
subroutine or function may have multiple ENTRY points. An ENTRY
statement can cause execution of a subroutine or function to
begin at a statement other than the first executable statement.
Control is returned from a subroutine or function program unit
either by encountering the END statement, or by executing a
RETURN statement.
FORTRAN provides for an alternate return
specification for ~ubroutines, such that a subroutine can return
to a different place in the caller than the statement following
the CALL statement.
FORTRAN supplies a comprehensive set of intrinsic functions
which perform data type conversion and provide an extensive collection of arithmetic and transcendental functions.
1.2
Notation and Terminology Used in this Manual
This section defines the notation that is used in
to define FORTRAN language constructs.
this
manual
Upper-case letters and special characters are written as shown
in programs.
Lower-case letters and words indicate objects for
which there is a substitution in actual statements described in
the text. Once a lower-case object is define~, it can be assumed
to retain that meaning for the remainder of the construct being
defined.
Example of
~-case
and.
lower-~
usage
The format specification which describes integer editing is
denoted Iv, where v is a non-zero, unsigned integer constant. In
an actual. FORMAT statement, the editing specification might be
written as IS or 121. The editing specification for real numbers
is Fw.d, where d is an unsigned integer constant.
An actual
FORMAT statement might contain an edit specification like PS.4 or
P14.0. Note that the period character is a special character as
defined above, and is taken literally.
Page 4
Fortran Reference Manual
Introduction
."pter 1
Brackets "[" and ")" enclose optional items. For example, A[w)
indicates that either of the forms A or Al2 are valid (as a means
of specifying a character format).
The ellipsis notation " ••• " indicates that the optional item
preceding the ellipsis may appear one or more times. For example, the computed GOTO statement is described by the form:
GOTO ( s
[, s ] ••• )
[, ]
i
which indicates that the syntactic item ',Sf may be repeated
number of times.
any
Spaces (blanks) normally have no significance in describing
FORTRAN statements. The general rules for spaces, supplied later
in this chapter, provide the interpretation of spaces in all contexts.
Throughout this manual, "space" and "blank" are considered synonymous. In general, the word "space" is used.
1.3
Basic Elements of FORTRAN
This section covers the basic lexical and syntactic
that go towards constructing a FORTRAN program.
1.3.1
elements
FORTRAN Character Set
The FORTRAN character set consists of
letters, digits, and special characters.
upper
and
lower
case
A letter is one of the S2 characters:
ABC D E F G H I J K L M N 0 P 0 R STU V W X y Z
abc d e f 9 h i j kIm n 0 p q r s t u v w x y z
A digit is one of the ten characters:
o 1 234 S 6 789
An alphanumeric character is a letter or a digit.
The special characters consist of the following characters:
Fortran Reference Manual
Page S
Chapter 1
Introduction
+-----------+-------------------+
I Character I Name of Character I
+-----------+-------------------+
Blank or Space
+
*
/
\
(
)
.
$
•
,
Equals sign
Plus sign
Minus sign
Asterisk
Slash
Reverse Slash
Left Parenthesis
Right Parenthesis
Comma
Decimal Point
Currency Symbol
Apostrophe
Colon
Ampersand
+-----------+-------------------+
1.3.2
Collating Sequence and Graphics
SVS FORTRAN uses the
sequence in ASCII is:
ASCII
character
set.
The
collating
• Space (blank) collates lowest, followed by:
• Digits "0" thru "9", followed by:
• Upper case letters "A" thru "Z", followed by:
• Lower case letters "a" thru "z".
The special characters appear in between digits and upper-case
letters and before and after lower-case letters. There is an
ASCII character set chart in the appendices.
Within each of the ordered sets, digits, upper-case letters,
and lower-case letters, the characters in those sequences are
contiguous - there are no "holes" in those sequences.
1.3.3
Use of Spaces or Blanks or Tabs·
The tab character is interpreted in all cases by SVS FORTRAN as
a series of one or more blanks, sufficient to fill columns up to
and including the next column position evenly divisible by eight.
All references to blanks in this manual may refer to actual
blanks or blanks resulting from the interpretation of tab characters.
All references to columns in this manual refer to the
columning after the interpretation of tabs. Limitations on the
Page 6
Fortran Reference Manual
Chapter 1
Introduction
total number of characters per line and per statement are after
the interpretation of tabs.
The space (also called blank) character has no meaning in
a
program unit, with the exceptions listed below. Otherwise,
spaces can be used freely to improve the layout and readability
of a program. Spaces are significant in the following cases:
Within string constants, and within Hollerith fields.
On compiler directive lines,
"Using the FORTRAN Compiler".
discussed
in
the
chapter
In column 6, where a space distinguishes an initial line
a continuation line.
Counting in the total number of characters per
statement.
Fortran Reference Manual
line
on
from
and
per
Page 7
Introduction
Page 8
Chapter 1
FORTRAN Reference Mar
Lines, Statements, and Control Flow
Chapter 2
Chapter 2 - Lines, Statements, and Control Flow
This chapter consists of three sections.
The first section
describes the notions of lines in a FORTRAN program. The second
section covers the rules for FORTRAN statements. The final section in this chapter covers the concept of execution sequence or
control flow - the order in which FORTRAN statements are executed.
2.1
Lines
A line in a program unit is a sequence of characters in columns
thru 120 (1 thru 72 if the $COL72 compiler option is selected,
as described in Chapter 13). All characters in a line must be
selected from the character set described in Chapter 1. Comment
lines (described below), character constants and Hollerith fields
can contain any printable ASCII character.
1
The character positions on a line are called columns and are
numbered consecutively from 1 thru 120 (1 through 72), left to
right on a line.
The FORTRAN compiler ignores characters which appear to the
right of column 120 (column 72 if the $COL72 compiler option is
selected) on a line, thus the user may use these columns for any
purpose (such as sequence information).
2.1.1
Comment Lines
Comment lines can appear anywhere in a program unit, including
before the first statement or after the last statement of a program unit. Comment lines may appear between an initial line and
its first continuation line, or between two continuation lines.
FORTRAN .Reference Manual
Page 9
Chapter 2
Lines, Statements, and Control Flow
Examples of Comment Lines
C
*
C
This is a comment line.
This is also a comment line.
The line following this one is all blank •••••
c
..... and is therefore considered to be a comment.
**
*
Comment lines are for documentary purposes and
have no effect on compilation or execution.
2.1.2
Initial Lines
An initial line generally indicates the start of a statement
line.
An initIal line is any non-comment line containing the
space character or the digit 0 in column 6. Columns I thru 5 of
the initial line of a statement may contain a statement label.
Since a line which begins with a tab character is processed as
if it had 8 initial blanks, such lines are interpreted as initial
lines. If a tab character follows a statement label which occupies some or all of the first 5 columns, the line will also be
processed as if it had a blank in column 6.
Examples of Initial Lines
Here are initial lines without statement labels.
C
C
GO TO 999
OGO TO 999
C
C
Here are initial lines with statement labels.
C
379 GO TO 999
4850GO TO 999
2.1.3
Continuation Lines
A continuation
character from
character or the
character ,
in
statement label.
statement.
Page 10
line is any non-comment line containing any
the FORTRAN character set (other than a space
digit 0) in column 6, or having the ampersand
column one. A continuation line must not have a
There may be up to 19 continuation lines in a
FORTRAN Reference Manual
Lines, Statements, and Control Flow
Chapter 2
Examples of Continuation Lines
C
C
C
*
*
,
2.1.4
These contrived statements each span two lines.
GO TO
$ 999
843 GO TO
+ 999
GO TO
999
Compiler Directive Lines
Compiler Directives are an SVS extension to FORTRAN-77. Compiler directives provide additional controls over the compiler's
actions. A compiler directive line is a line with a dollar sign
"$" in column 1. A compiler directive line can appear anywhere
that a comment line can appear, although certain directives must
appear in certain restricted places in the program. Spaces are
signi~icant in compiler directive lines and serve to delimit keywords and filenames. The compiler directives are listed in the
chapter on "Running the FORTRAN Compiler".
Examples of Compiler Directives
*
The following directive
*
include the body of the
*
program source code.
*
*
$INCLUDE rasp. text
2.2
instructs the compiler to
file 'rasp.text' into the
Statements
FORTRAN language statements are described in
12. Statements are used to form program units.
Chapters
6
thru
Statements are written in columns 7 thru 72 of an initial line,
and as many as 19 continuation lines.
An END statement is the exception to the above rules.
An END
statement must appear on an initial line on its own. No other
statement in a program unit may have an initial line that looks
like an END statement.
FORTRAN Reference Manual
Page 11
Lines, Statements, and Control Flow
Chapter 2
In general, statements must begin on new lines, that is, a
statement may not begin on the same line as another statement.
The exception to this rule is the logical IF statement.
Spaces before, within, and after statements have no effect,
except within character constants and Hollerith constants, where
they indicate blank characters.
Examples of Statements
C
An assignment statement
C
A
= 5.0
C
C
A subroutine call statement
C
CALL COLECT(PAY, PHONE)
C
C
A logical IF statement
C
IF (DAY .EO. 'FRIDAY')
2.2.1
RETURN
Statement Labels
Statement labels provide the means to "tag"
that other statemen~s may refer to it.
a
statement
such
Any statement may be labeled, but only labels associated with
executable statements and FORMAT statements can be referenced by
other s~atements.
A statement label is one to five digits appearing anywhere in
columns 1 thru 5 of the initial lfne of a statement. At least
one of the digits in a statement label must be non-zero.
In any given program unit, statement labels must
duplication of statement labels is an error.
be
unique
Examples of Statement Labels
123
C
C
C
FORMAT('The result is', IS)
An example of a DO block.
DO
110 ICON
1, 100
DESK (ICON) = 0.0
110 CONTINUE
Page 12
FORTRAN Reference Manual
Chapter 2
2.2.2
Lines, Statements, and Control Flow
Order of Statements and Lines
In any given program unit, the order in which statements appear
must obey certain rules. These rules are detailed below.
A PROGRAM statement may appear only as the first statement of a
main program. The first statement of a subprogram must be either
a FUNCTION, SUBROUTINE, or a BLOCK DATA statement.
In a program unit, statements
order.
must
appear
in
the
following
1.
FORMAT statements can appear anywhere.
2.
All specification statements must precede all DATA statements, statement-function statements, and executable statements.
3.
All statement-function statements must precede
able statements.
4.
DATA statements may appear anywhere after the
statements.
5.
ENTRY statements may appear anywhere except between a block
IF statement and its corresponding END IF statement or
between a DO statement and the terminal statement of its
DO-loop.
In other words, a subprogram must not be entered
(via an ENTRY statement), in the middle of a block IF or DO
block.
all
execut-
specification
within a program unit's specification statements, IMPLICIT
statements must precede all specification statements other than
PARAMETER statements. Any specification statement that defines
the type of a symbolic name must precede a PARAMETER statement
that defines the symbolic name of a constant.
PARAMETER statements that define symbolic constant
precede all uses of such names.
names
The last (non-comment) line of a program unit must
statement.
be
The diagram below pictorially describes the manner
statements and comment lines may be interspersed.
FORTRAN Reference Manual
an
in
must
END
which
Page 13
Chapter 2
Lines, Statements, and Control Flow
I+--------------+--------------+-----------------+I
I+-----------------+I
~----------+-----------------------------------------------+
PROGRAM, FUNCTION, SUBROUTINE, or
BLOCK DATA statement.
IMPLICIT
Statements
comment
Lines
FORMAT
and
ENTRY
Statements
PARAMETER
Statements
Other
Specification
Statements
+--------------+-----------------+
Statement
DATA
Statements
Function
statements
+-----------------+
Executable
I
I
Statements
~----------+--------------+--------------+-----------------+
I
I
END Statement
~----------------------------------------------------------+
Figure 2-1
Ordering of Lines and Statements in a FORTRAN Program
In the diagram, vertical lines separate statement groups that
may be mixed. For example, FORMAT statements can be mixed with
statement-function statements and executable statements.
Horizontal lines separate statement groups that must not be
mixed.
For example, statement-function statements cannot be
mixed with executable statements. Note that an END statement, in
addition to being executable, must be the last statement in a
program unit.
2.3
Execution Sequence and Control Transfer
Normal execution sequence means that executable statements are
executed in the order in which they appear in the program unit.
Program execution starts with the first executable statement in
the main program. When an external procedure is referenced, execution proceeds with the first executable statement that follows
the FUNCTION, SUBROUTINE or ENTRY statement in the subprogram.
A control transfer means that the normal execution sequence
altered. Statements that cause control transfer are:
1.
unconditional GO TO statement, computed GO TO
assigned GO TO statement,
Page 14
statement
is
or
FORTRAN Reference Manual
Chapter 2
Lines, Statements, and Control Flow
2.
Arithmetic IF statement,
3.
RETURN statement,
4.
STOP statement,
5.
input-output statement containing an error specifier
end-of-file specifier,
6.
CALL with an alternate return specifier,
7.
logical IF statement containing any of the
its subordinate statement,
8.
Block IF and ELSE IF statements,
9.
the last statement (if any) of an IF-block or ELSE IF-block,
10.
DO statement,
11.
the terminal statement of a DO-loop,
12.
END statement.
above
or
an
forms
as
Normal execution sequence is not affected by non-executable
statements,
comment lines or compiler directives appearing
between executable statements in the source code.
Executing a function reference or a CALL statement is not considered a control transfer in the program that makes the reference, except when control is returned to a statement identified
by an alternate return specifier in a CALL statement.
Executing a RETURN or an END statement in a referenced procedure, or a control transfer in a referenced procedure, is not
considered a control transfer in the program unit that makes the
reference.
Function and subroutine subprograms cannot be invoked recursively in SVS FORTRAN. Note, however, that FORTRAN subprograms
can reference subprograms written in other languages (such as
Pascal) that themselves can be recursive.
FORTRAN Reference Manual
Page 15
Lines, Statements, and Control Flow
Page 16
Chapter 2
FORTRAN Reference Manual
Data Types. and Constants
Chapter 3
Chapter 3 - Data Types and Constants
There are six data ~ in FORTRAN, namely: INTEGER, REAL
(floating point), DOUBLE PRECISION
(extended precision REAL),
COMPLEX (elements of the complex number domain), CHARACTER (character string data)
and LOGICAL (able to assume the values .TRUE.
or .FALSE.). The various data types are discussed in the sections to follow. SVS FORTRAN extends ANSI FORTRAN in such a way
that the user can specify the amount of computer storage which a
particular data type consumes. This extension is covered in the
appropriate sections below.
3.1
Data Type Rules
A symbolic name -(associated with a constant, variable, array,
external function or statement-function) can have its type specified in a type statement.
The possible types are those listed
in the paragraph above.
If no explicit type statement is supplied for a program element, the type is implied by the first letter of the name. A
first letter of I, J, X, L, M or N implies type integer.
Any
other first letter implies type real. These default type rules
can be overridden either by explicit type statements or by the
IMPLICIT statement, which changes the default type-rules.
The data type of an array element is the same as that of the
array.
The data type of a function name specifies the type of
the value which that function returns. Intrinsic functions have
a type that is specified in the chart in the chapter on "Program
Structure". Generic intrinsic functions do not have a default
type.
The type of a generic intrinsic function depends on the
type of its argument(s). An external function reference is given
a default type, based upon the first letter of its name, in the
same way as variables and arrays.
3.2
Constants
A constant is a value that defines itself and does not change.
A constant can be an arithmetic value, a logical value, or a
character string value.
The representation of a constant
FORTRAN Reference Manual
Page 17
Chapter 3
Data Types and Constants
specifies both its value and its data type. A PARAMETER statement associates a symbolic name with a constant.
Arithmetic constants are INTEGER, REAL, DOUBLE
COMPLEX values.
PRECISION,
and
For the purposes of definition in this manual, an unsigned constant is a constant without any leading sign. A signed constant
is a constant with a leading plus or minus sign. An optionallysigned constant is a constant that can be either signed or
unsIgned. Integer, real, and double precision constants may be
optionally-signed except where otherwise noted.
3.3
Integer Data Type
Integers, as represented in the finite word size of a computer,
are only a subset of the infinite set of integers. An integer
value as represented in FORTRAN is an exact representation of the
corresponding integer.
An integer data value occupies two words (four bytes or 32
bits)
of storage and can
represent values in the range
-2,147,483,648 thru +2,147,483,647. Integers can be designated
as fitting in 16 bits if the $INT2 compiler option is selected.
Integer numbers are internally stored in two's complement
representation.
As a consequence, there is one more negative
integer value than there are positive integer values.
SVS FORTRAN also provides a means to specify the amount of
storage which integers occupy_
This is an extension to ANSI
FORTRAN. The extended forms of integer are:
INTEGER*l
occupies one byte (8 bits) and can assume
in the range -128 thru +127.
values
INTEGER*2
occupies one word (16 bits) and assumes values in
the range -32768 thru +32767.
INTEGER*4
9ccupies two words (32 bits) and is the same
the standard integer type discussed above.
as
An integer constant consists of a sequence of decimal digits,
preceded by an optional sign. Alternatively, an integer constant
can be expressed in the hexadecimal radix by a dollar sign ($)
followed by
a sequence of digits or letters in the range 'A'
through 'F' or 'a' through If'. Note that hexadecimal numbers
are considered unsigned.
To obtain a signed hexadecimal constant, the user must explicitly code a 1 in the sign bit position.
Page 18
FORTRAN Reference Manual
Data Types and Constants
Chapter 3
Examples of INTEGER Constants
3.4
72
-32768
32767
o
$123
$ffffffff
$Oa
$3e8
+56
Real Data Type
Real Data Types are intended to represent the set of real
values which comprise the continuum.
Because of a finite
representation imposed by a finite word size in the computer, the
real data type in FORTRAN can only represent a finite subset of
the entire set of reals.
A basic real constant has an optional sign, an integer part,
a
decimar--point, and a fractional part. Both the integer and the
real part are sequences of digits. Either part can be omitted,
but not both. Real constants are assumed to be decimal numbers.
Examples of Basic Real Constants
3.14159
.7071
5.
0.0
+2.236
+.5
+8.
-1.4142
-.618034
-6.
O.
.0
A real exponent consists of the letter E followed by an optionally--signed integer constant. A real exponent indicates a power
of ten.
Examples of Real Exponents
E14
E+12
E-10
EO
A real constant is anyone of: a basic real constant, a basic
real constant followed by an exponent part, or an integer constant followed by an exponent part.
SVS FORTRAN provides a means to specify the amount of data
storage which a real data type is to occupy. The forms are:
REAL*4
occupies two words (32 bits) and is the same as a
basic real datum as described above.
FORTRAN Reference Manual
Page 19
Data Types and Constants
REAL*8
Chapter 3
occupies four words (64 bits) and is the same as
the DOUBLE PRECISION data type discussed below.
Examples of Real Constants
+1.52E-l
299193.5E3
20E-3
A real constant containing an exponent part is the product of
the constant preceding the E and the power of ten indicated by
the integer following the E.
A real value occupies four bytes of storage. The range of real
values is approximately -3.4E38 thru +3.4E38. The precision is
about seven decimal places.
3.4.1
Infinite and Indeterminate Real Values
The representation of real values in FORTRAN allows for positive and negative "infinity", and for indeterminate values. This
is primarily of interest when formatting such values for output.
When output by a FORTRAN program, these values appear as strings
of either plus signs '++++.+++' or minus signs ,----.---' for
positive and negative infinity, and as question marks '7771.111'
for indeterminate values. There is more on this subject in the
chapter on input and output.
3.5
Double Precision Data Type
The Double-Precision data ~ is intended for applications
where the range and precision of Single-Precision data is inadequate.
Double-Precision extends the range to approximately
-10E308 thru +10E308
and the precision to about 16 decimal
digits.
A double precision exponent is the letter D followed by an
optionally-signed integer constant. The forms of a double precision constant are either: a basic real constant followed by a
double precision exponent, or an integer constant followed by a
double precision exponent.
3.6
Complex Data Type
The Complex data
domain.
~
represents values from the complex number
A complex number consists of an ordered pair of numbers, each
of which is either an integer or a single-precision real number the first representing the "real" part of the number, and the
Page 20
FORTRAN Reference Manual
Data Types and Constants
Chapter 3
second representing the -imaginary- part.
A complex constant is
written as two integers or single-precision real numbers enclosed
in parentheses and separated by a comma.
Examples of Complex Constants
(1, 1)
3.7
(0.707, -0.707)
(-1.5ElO, 2.6E-5)
Character Data Type
A character data element is a string of characters. The string
can contain any of the printable ASCII characters. Spaces are
significant in character strings.
The length of a character
string is the number of characters in the str1ng.
The form of a character constant is a non-empty string of characters enclosed in apostrophe "I" signs. The apostrophes serve
to delimit the string constant, but are not part of it.
An
apostrophe in the string is represented by two juxtaposed apostrophes.
The length of a character constant is the number of characters
in the string, except that each pair of juxtaposed apostrophes in
the string is counted as one character.
The delimiting apostrophes are not part of the string and are not counted in the
string length.
The maximum length of a character constant is 255 characters.
There is no provision for
strings.
expressing
empty
(null)
character
Examples of Character String Constants
'x'
I
,
'The time is One O"Clock'
'The hunting of the Snark'
"
..
The last two examples illustrate the representation
apostrophes.
of
embedded
FORTRAN source lines may extend up to column 120 (or 72 if $COL
72 set) on a line.
Shorter lines are not space-filled to 72
columns but are left as typed. When a character constant extends
across a line boundary, its value is as if the portion of the
continuation line starting in column 7 abuts the last character
on the preceding line. Thus the FORTRAN source statement:
FORTRAN Reference Manual
Page 21
Chapter 3
Data Types and Constants
200
example. 'First string
~art
$ Second string part'
(where is carriage-return) is equivalent to the statement:
200
example
= 'First
string part Second string part'
where the single space between the "t" at the end of the first
line and the "S~ at the start of the secdnd line is the space in
column 7 of the continuation line. Long character constants can
be represented in this way.
3.8
Logical Data Type
A logical data element represents a Boolean quantity.
only take on the values true or false.
It
can
The form of a logical constant is either: ".TRUE." (representing the truth value) or ".FALSE." (representing the false value).
SVS FORTRAN provides a means to specify logical data items
which occupy less data storage than the standard logical type.
These forms are:
LOGICAL*l
occupies one byte (8 bits).
LOGICAL*2
occupies one word (16 bits).
LOGICAL*4
occupies two words (32 bits) and is the
the standard LOGICAL data type.
Page 22
same
as
FORTRAN Reference Manual
FORTRAN Names, Arrays, and Substrings
Chapter 4
Chapter 4 - FORTRAN Names, Arrays, and Substrings
This chapter introduces the rules for FORTRAN ~ - symbolic
names which may be used to identify program objects.
The second
section describes the way that arrays are defined and referenced.
The third section discusses the ideas of character variables and
substrings of character variables.
4.1
FORTRAN Names
A FORTRAN name or identifier consists of one through six
alphanumeric characters, and must start with a letter. A FORTRAN
name 'can have embedded spaces in it - the spaces have no significance and are ignored. The FORTRAN compiler makes no distinction
between upper-case letters and lower-case letters
the names
PASCAL, PaSeAl, pAsCaL and pascal are all equivalent as far as
FORTRAN is concerned.
A name is used to denote a user-defined variable, a systemdefined variable, array variable, subroutine or function. FORTRAN does not have any reserved words - the compiler recognizes
keywords in context.
For reasons of clarity and readability
though, users are recommended to use names that are distinct from
those of FORTRAN.
Examples of Valid FORTRAN Names
XPos
Eatup
FilSet
MAXO
L5
Shell
Bubble
Examples of Invalid FORTRAN Names
2ndTime
Begins with digit
TooLarge
More than six characters
No Good
Non-alphanumeric character
FORTRAN Reference Manual
Page 23
FORTRAN Names, Arrays, and Substrings
4.1.1
Chapter 4
Scope of FORTRAN Names
The scope of a FORTRAN name is that region of a program over
which the name is known or can be referenced. In general, the
scope of a name is either local to a program unit or global to
the entire FORTRAN program. There are certain exceptions which
are described later.
A name with global scope can be used in more than one program
unit (subroutine, function or the main program) and still refer
to the same object. Names with global scope can only be used in
a single, consistent manner within the same program. The names
of all subroutine, function and block-data program units, the
names of common areas, and the program name, have global scope.
Therefore, there cannot be a subroutine program unit that has the
same name as a function program unit or a common ~rea.
Similarly, two function program units cannot have the same name.
A name with local scope is only known within a single program
unit.
A name with local scope can therefore be used in other
program units with the same or different meanings everywhere it
is used. Within a specific local scope, a name must be used consistently and refer to the same objecte The names of variables,
arrays, constants, arguments and statement-functions all have
local scope. A name with local scope can be used in the same
compilation- as the same name with global scope as long as the
global name is not referenced within the program unit containing
the local name.
For example, there can be a function called
PARTY, and a local variable called PARTY in another program unit,
as long as the program unit containing the variable called PARTY
does not try to reference the function called PARTY. The FORTRAN
compiler detects all such scope errors and issues diagnostics
concerning them.
Common block names are an exception to the scope rules. It is
possible to refer to a globally scoped common block name in a
program unit containing a locally scoped name identical to that
of the common block. This situation is allowed because common
block names always appear in slashes, such as /COLD/, and therefore the compiler can always distinguish such names.
Formal arguments to statement-functions are another exception
to the scope rules. The scope of formal arguments of statementfunctions is the body of the statement function itself.
Any
other use of those names in the statement-function is not allowed
and neither is any other use of such names outside
the
statement-function.
For example, if a formal argument to a
statement-function has the same name as that of a function subprogram, that function subprogram may not be referenced from
within the body of the statement-function. References to formal
argument names of a statement-function from outside the body of
the statement-function refer to objects which are different from
the arguments of the statement-function.
Page 24
FORTRAN Reference Manual
Chapter 4
FORTRAN Names, Arrays, and Substrings
Names used as implied-DO control variables in DATA statements
and input-output statements have a scope which is local to the
DATA statement or input-output statement.
4.1.2
Undeclared FORTRAN Names
When a user name that has not previously appeared in a program
unit is referenced in an executable statement, FORTRAN decides
how to classify that name from the context in which it appears.
If the name appears to be a variable, FORTRAN creates a symboltable entry for that name.
Its type is inferred from the first letter of the name.
Variables starting with the letters I, J, K, L, M and N are considered to be of type integer; all others are considered to be of
type. real.
If an undeclared name appears in the context of a
function reference, the function's type is inferred from its name
in the same manner as for variables.
In both cases, these
default type rules can be overridden by previous IMPLICIT statements (see the chapter on "Specification Statements"). Similarly, a name appearing in the context of a subroutine call has
an entry created for it. If a symbol table entry exists for a
subroutine or function name, its attributes are coordinated with
those of the newly created entry. Inconsistencies such as a subroutine name used in the context of a function or vice versa give
rise to error diagnostics.
In general, users are encouraged to declare all names used in
each program unit, since it helps to assure that FORTRAN associates the proper definition with the name. Letting FORTRAN decide
on the default can sometimes result in logical errors that are
hard to find, usually at execution time when strange results or
forms of behavior are exhibited.
4.2
Array Declarations
Arrays provide the means to deal with data aggregates where the
elements of the aggregates are homogeneous.
An array declaration specifies a symbolic name that identifies an array in a program unit. The declaration also serves to specify properties of
the array, such as its dimension and, optionally,
the type of
its elements.
In any given program unit only one array declaration is allowed for any given array - duplicate declarations are
flagged as errors. The form of an array declaration is:
array name (dim [, dim] •••
Each 'dim' above is a dimension declarator as defined below.
The number of dimensions for an array is equal to the number of
dimension declarations given when the array is declared.
FORTRAN Reference Manual
Page 25
Chapter 4
FORTRAN Names, Arrays, and Substrings
4.2.1
Dimension Declarations
A dimension declarator serves to define the bounds of a
specific dimension in an array. FORTRAN-11 provides for defining
both the lower and the upper bound of a dimension.
The form of
a dimension declaration is:
[lower bound :] upper bound
The optional lower bound, and the upper bound are arithmetic
expressions, called dimension bound expressions, in which all
constants, symbolic constant names, and varIables are of type
integer.
The upper bound of the last dimension declaration can
be an asterisk (see 'assumed size arrays', later).
A dimension bound expression must not contain any function or
array element references.
Integer variables can appear in
dimension bound expressions only in adjustable array declarations
(see 'adjustable arrays', later).
If a symbo~ic constant name
or variable in a dimension bound expression is not of default
implied integer type, it must be specified as integer via a type
statement or an IMPLICIT statement before its use in a dimension
bound expression.
Either dimension bound may have a positive, negative or zero
value. The upper bound must not be less than the lower bound. If
only the upper bound is specified, the lower bound has the value
one (1). An upper bound of • is always greater than or equal to
the lower bound~
4.2.2
Kinds of Array Declarations
There are three basic forms of array declarations.
A constant array declaration is one in which all the
bound expressions are integer constant expressions.
dimension
An adjustable array declaration is one in which the dimension
bounds contain integer variables. Adjustable arrays may be used
as dummy arguments in subroutines and functions. Variables which
define the bounds of adjustable arrays must either be formal
arguments themselves, or they must be in common blocks.
An assumed size array declaration is one in which the upper
bound of the last dimension is an asterisk character '.'.
Assumed size arrays may also only be used as dummy arguments to
subroutines and functions.
Using assumed size arrays in procedures circumvents any range checking which the FORTRAN system
can perform.
Page 26
FORTRAN Reference Manual
FORTRAN Names, Arrays, and Substrings
Chapter 4
4.2.3
Actual Arrays and Dummy Arrays
An actual array declarator "actually" declares an array there
and then. Each actual array declarator must be a constant array
declarator as defined above. An actual array declarator can be
used in the type statement, the DIMENSION statement and the COMMON statement, as defined in the chapter on "Specification Statements".
A dummy array declarator defines a dummy argument for a subroutine or function.
A dummy array declarator can be any of the
forms given above: constant, adjustable or assumed size. A dummy
array declarator can only appear in subroutines and functions. A
dummy array declarator may not appear in a COMMON statement.
Examples of Array Declarations
**
*
C
C
C
C
C
**
*
*
*
*
4.2.4
Constant array Declarations in type statements
a 100 element vector with bounds 1 - 100
INTEGER VECTOR(lOO)
a 20 element matrix with 5 rows and 4 columns
REAL MATRIX(S, 4)
a 256 element array with bounds 0 - 2S5
CHARACTER~2 CHARS(O : 255)
a 3 element array
LOGICAL*2 BOOLS(-l : +1)
a constant expression dimension
REAL WOOD (2*4)
Adjustable Array Declaration in a DIMENSION statement
DIMENSION SCREEN(l : CHARS, 1 : LINES)
Assumed size array declaration in a type statement.
REAL VARIAB(5, *)
Referencing Array Elements - Array Subscripts
An array sUbscrirt is the means to reference an element of
array. The form 0 an array subscript is:
(subexpr [, subexpr]
the
••• )
Note that the term "subscript" includes
enclose the subscript expression list.
the
parentheses
that
A subscript expression is an integer expression.
A subscript
expression can contain array element references and function
FORTRAN Reference Manual
Page 27
FORTRAN Names, Arrays, and Substrings
Chapter 4
references. If a subscript expression contains a function reference, the function must not change the value of any other subscript expression in the same subscript.
In any given program unit, the value of each subscript expression should not be less than the lower bound for the dimension
and should not be greater than the upper bound for the dimension.
If the upper dimension bound is an asterisk, the subscript
expression must not be greater than the size of the dummy array.
Examples of Arrays with Subscripts
SCREEN (2, 3)
VARIAB(N+l, MAX(3, 4»
4.2.5
Using Unsubscripted Array Names
Generally speaking, array names must be followed by subscripts.
There are some exceptions where the array name alone can be used.
An unsubscripted array name can be used in the following places:
a list of dummy arguments for a subroutine or function program
unit,
a COMMON statement when declaring that the
that common block,
array
resides
in
a type statement when the type of the array is established,
an array declaration when the array dimensions are being established,
an EQUIVALENCE statement,
a DATA statement,
the list of actual arguments in a
procedure,
reference
to
an
the list of an input-output statement if the array is
assumed size dummy array,
external
not
an
a unit identifier for an internal file in an input-output
statement if the array is not an assumed size dummy array,
the format identifier in an input-output
array is not an assumed size dummy array,
Page 28
statement
if
the
FORTRAN Reference Manual
FORTRAN Names, Arrays, and Substrings
Chapter 4
a SAVE statement.
4.3
Character Substrings
A Character Substring is a contiguous portion of a character
object.
The type of a character substring is of type CHARACTER.
A character substring can be identified by a symbolic name, and
The
it can be referenced and assigned values by that name.
forms of a substring ~ are:
character variable([start) :
[finish))
where 'start' and 'finish' are substring expressions.
A character variable may be an element of a character array.
'start'
specifies the leftmost character position of the substring.
'finish' specifies the rightmost character position of the substring. The values of 'start' and 'finish' must be such that:
1 <= start <= finish <= length
where 'length' is the length of the character variable or character array element.
If 'start' is omitted, the value one (l) is
used.
If 'finish' is omitted, the value 'length' is used.
Both 'start' and 'finish' can be omitted. In such a case, a substring reference of the form s(:) is equivalent to s.
The
length of a character substring is 'finish' - 'start' + 1.
A substring expression is any integer expression which
can
contain array element references, and function references. The
same restrictions (with regard to side effects) apply to substring expressions as apply to array subscripts.
Examples of Character Substrings
ROPEY(1:3)
THELOT(:)
ACHAR(5:5)
FOURCH(:4)
FORTRAN Reference Manual
Page 29
FORTRAN Names, Arrays, and Substrings
Paqe 30
Chapter 4
FORTRAN Reference Manual
Expressions
Chapter 5
Chapter 5 - Bxpressions
This chapter describes the rules for expressions.
An expression is formed from operands, operators, and parentheses. FORTRAN has four classes of expressions:
• Arithmetic expressions,
• Character expressions,
• Relational expressions,
Logical expressions.
5.1
Arithmetic Expressions
An arithmetic expression expresses a
generates a numeric value.
5.1.1
numeric
computation
and
Arithmetic Operators
The arithmetic operators are as follows:
+----------+-------------------------+
I Operator I
Meaning
I
+----------+-------------------------+
**
Exponentiation
I
Division
Multiplication
Subtraction or Negation
Addition or Identity
*
+
+----------+-------------------------+
The **, I, and
operators can be
* operators are binary operators.
u~ary
The +
and
or binary operators.
The ** operator has the highest precedence, then the * and I
operators, and lastly the + and - operators.
Parentheses may be
used freely to change the order of evaluation.
FORTRAN Reference Manual
Page 31
Chapter 5
Expressions
5.1.2
Arithmetic Operands
An arithmetic operand consists of a primary, a factor, a term
or an arithmetic expression. These various kinds of operands are
discussed below.
The primary operands are:
•
•
•
•
•
Unsigned arithmetic constant,
Symbolic name of an arithmetic constant,
Arithmetic variable reference,
Arithmetic array element reference,
Arithmetic function reference,
Arithmetic expression enclosed in parentheses.
The factor operands are:
• Primary,
• Primary ** factor.
A factor is formed from a sequence of one or more primaries
separated by an exponentiation operator. The second form means
that an expression such as:
2**3**4
is to be interpreted as:
2** (3**4)
A term operand is:
• Factor,
• Term I factor,
• Term * factor.
A term is formed from one or more factors separated by the multiply or divide operator. Factors are combined left to right.
An arithmetic expression consists of:
•
•
•
•
Term,
+term or -term,
Arithmetic expression + term,
Arithmetic expression - term.
An arithmetic expression consists of a series of
terms
separated by plus or minus operators.
The first term in an
expression can be preceded by a plus or minus sign.
Terms are
combined left to right. Note that the rules for expressions mean
that two consecutive operators form an incorrect expression. Thus
A**-B is wrong, whereas A**(-B) is correct.
Page 32
FORTRAN Reference Manual
Expressions
Chapter 5
5.1.3
constant Expressions
Constant expressions are used in many language constructs
throughout
FORTRAN, especially in specification statements.
There are two forms of constant expressions, namely arithmetic
constant expressions and integer constant expressions. These are
discussed below.
An arithmetic constant expression is an expression in which
each primary is an arithmetic constant, the symbolic name of an
arithmetic constant or a constant expression
enclosed
in
parentheses.
Exponentiation is only allowed if the exponent is
of type integer.
Examples of Arithmetic Constant Expressions
5.0*2
2**31-1
2*(4.5, 9.8)
-16/4
3.141592/2
5**(3+2)
An integer constant expression is an arithmetic
expression in which each constant is of type integer.
constant
Examples of Integer Constant Expressions
3*5
5.1.4
-10
4+5*(9-2)
Type Conversion Rules for Arithmetic Expressions
The data type of an expression is ultimately derived from the
data types of its operands according to the rules stated below.
When operands of mixed data types appear in an expression, FORTRAN performs implicit type conversion on the operands according
to well-defined rules in order to generate the result.
When the plus "+" operator or the minus "-" operator operate
upon a single operand (they are used as unary operators), the
data type of the result is the same as the data type of the
operand.
When an arithmetic operator applies to a pair of operands, the
type of the results is as shown in the tables below.
The letter
I stands for an operand or result of type Integer, the letter R
for Real, the letter D for Double-precision and the letter C for
Complex. The rules are given in the form of assignments.
The
result type is indicated by the letter to the left of the equals
sign and the derivation of that result is given by the expression
to the right of the equals sign. The function names REAL, DBLE
FORTRAN Reference Manual
Page 33
Chapter 5
Expressions
and CMPLX are as defined in the table of intrinsic
the Appendix on "Intrinsic Functions".
5.1.4.1
functions
in
Rules for Add, Subtract, Multiply and Divide
The two tables below define the types and interpretations
the +, -, * and I operators. For example, to obtain the rule
Il.C2 where 'II' is an integer and 'C2' is a complex, look in
second part of the table; find the 'II' entry under 'Xl'
the 'C2' entry across from 'X2', the rule is then:
for
for
the
and
'C • CMPLX(REAL(Il), 0.0) + C2'
which is interpreted as:
'the result is of type complex; the first operand is
obtained by converting the integer to a real, then converting that to a complex with the imaginary part 0.0; the two
complex numbers are then added'.
The rules for subtraction, multiplication and division are
obtained by replacing the "+" signs with the desired operator.
+------+-----------------------------+-----------------------+
XlX2
12
R2
+------+-----------------------------+-----------------------+
11
I • 11 + 12
R • REAL(Il) + R2
Rl
R = Rl + REAL(I2)
R = Rl + R2
Dl
0 = 01 + OBLE(I2)
0 • 01 + OBLE(R2)
Cl
C = C1+CMPLX(REAL(I2), 0.0)
C = C1+CMPLX(R2, 0.0)
+------+-----------------------------+-----------------------+
I
I
I
I
+------+---------------------+-------------------------------+
XlX2
02
C2
+------+---------------------+-------------------------------+
11
0 a OBLE(Il) + 02
C = CMPLX(REAL(Il), 0.0)+C2
Rl
0 = OBLE(Rl) + 02
C • CMPLX(R1, 0.0)+C2
Dl
0 = 01 + 02
Not Allowed
C1
Not Allowed
C a Cl+C2
+------+---------------------+-------------------------------+
I
5.1.4.2
I
I
Rules for Exponentiation Operator
The tables below define the
expressions of the form Xl**X2.
Page 34
I
types
and
interpretations
for
FORTRAN Reference Manual
Expressions
Chapter 5
+------+-----------------------------+--------------------------+
X2
I
I
X1
I
12
I
R2
+------+-----------------------------+--------------------------+
11
I - 11 ** 12
R
REAL (11) ** R2
Rl
Dl
Cl
R·· Rl ** 12
D • Dl ** 12
C - Cl**I2
R • Rl ** R2
D = Dl ** DBLE(R2)
C • Cl**CMPLX(R2, 0.0)
+------+-----------------------------+--------------------------+
+------+----------------------+---------------------------------+
I
X1X2
I
I
D2
C2
I
+------+----------------------+---------------------------------+
11
D = DBLE(Il) ** 02
C = CMPLX(REAL(Il), O.0)**C2
Rl
Dl
Cl
D • DBLE(R1) ** 02
0
01 ** D2
Not Allowed
=
C = CMPLX(Rl, 0.0)**C2
Not Allowed
C = Cl**C2
+------+----------------------+---------------------------------+
Four of the entries in the above table specify what happens
when a complex argument is raised to a complex power. In these
cases, the value of the expression is the principal value, determined by the formula:
Xl**X2
= EXP(X2 *
LOG(X2»
where EXP and LOG are the exponential and natural logarithm
intrinsics described in the chapter on "Program Structure".
Except for values raised to an integer power,
in mixed mode
expressions, the operand which differs from the type of the
result is converted to the type of the result according to the
rules given in the tables above. The operator then operates on a
pair of operands of the same type. When a primary is raised to
an integer power, the integer does not need to be converted.
5.1.5
Coercion Rules for Integers of Different Size
In expressions involving INTEGER*l, INTEGER*2 and INTEGER*4
(INTEGER), the smaller sized operand is always "promoted" to the
size of the larger operand, and the arithmetic operation is performed in the larger sized field. In all cases, elements of type
INTEGER*l are always promoted to INTEGER*2. In any case, assigning the result of an expression to a variable of smaller size
produces an undefined result if the value stored exceeds the
range of values allowed for that specific variable. Note also
that many FORTRAN statements and functions specifically require
arguments of type INTEGER. In such cases, neither arguments of
size INTEGER*l nor INTEGER*2 may be used.
FORTRAN Reference Manual
Page 35
Chapter 5
Expressions
5.1.6
Integer Division
If an integer operand is divided by another integer operand,
the result is not the strict mathematical quotient. Instead, the
quotient is obtained by truncating towards zero. Thus 1/2 is 0
and (-a)/3 is -2.
5.2
Character Expressions
~ character expression operates on character strings
and generates character values. The simplest form of character expressions are:
•
•
•
•
•
•
Character
Character
Character
Character
Character
Character
constant,
variable,
array element reference,
substring reference,
function reference,
expression enclosed in parentheses.
There is only one character operator - the
concatenation.
"II" sign,
meaning
The result of a character concatenation operations such as:
II
Xl
X2
is a value which is 'Xl' concatenated on the right with 'X2'.
The length of the result is the sum of the lengths of the individual operands.
5.2.1
Restrictions on the use of String Expressions
Formal arguments to procedures can be character strings whose
length is specified as (*). This designates the string as an
assumed size character string whose length is determined at the
time an actual string argument is associated with that formal
argument. ~ character string expression involving concatenation
of such a string argument may not be passed as an actual argument
to any procedure, nor may it appear in the format specification
of an input-output statement, nor may it appear as an item in the
'iolist' of an input-output statement.
Example of String Concatenation
'Left Side' II 'Right Side'
Page 36
FORTRAN Reference Manual
Expressions
Chapter 5
5.3
Relational Expressions
Relational expressions compare arithmetic expression values or
character expression values. Relational expressions yield logical values. The relational operators are:
+----------+--------------------------+
I Operator I
Meaning
I
+----------+--------------------------+
.LT.
Less than
.LE.
.EO.
.NE.
.GT.
.GE.
Less than or equal to
Equal to
Not equal to
Greater than
Greater than or equal to
+----------+--------------------------+
5.3.1
Arithmetic Relational Expressions
An arithmetic relational expression expresses a relationship
between arithmetic operands.
The form of an arithmetic relational expression is;
El relop E2
where 'El' and 'E2' are arithmetic operands, and 'relop' is
of the operators selected from the table above.
one
Only the .EO. (equality) and .NE.
(inequality)
allowed for operands of complex type.
are
operators
If the operands are of different types, the relational
sion is treated as if it were in the form:
«El) -
(E2»
expres-
relop 0
where 0 (zero) is the same type as the expression.
Comparison of a double precision value and a complex
not allowed.
5.3.2
value
is
Character Relational Expressions
A character relational expression is of the form:
El relop E2
where 'El' and 'E2' are character expressions and 'relop' is one
of the relational operators selected from the table above. The
order ing of character .expressions is as defined in the ASCII
character set table in the appendices. The .EO. (equality) and
FORTRAN Reference Manual
Page 37
Chapter 5
Expressions
.NE. (inequality) operators do not use the ordering. If the
operands in a character relational expression are of different
lengths, the shorter operand is considered to be padded on the
right with spaces until the operands are of the same length.
5.4
Logical Expressions
A logical expression operates on values of
type logical and
generates a result of type logical. The simplest forms of logical expressions are:
Logical constant,
Logical variable reference,
Logical array element reference,
Logical function reference,
Relational expression.
•
•
•
•
Other logical expressions are built up from these simple
by using parentheses and the logical operators as follows:
forms
+----------+------------------------+
I Operator I
Meaning
I
+----------+------------------------+
.NOT.
Logical Negation
.AND.
oOR.
.EOV.
.NEOV.
Logical Conjunction
Inclusive Disjunction
Logical Equivalence
Logical Nonequivalence
+----------+------------------------+
5.4.1
Precedence of Logical Operators
The precedence of the logical operators is shown
page:
on
the
next
+-----------------+------------+
I
Operator
I Precedence I
+-----------------+------------+
.NOT.
Highest
.AND •
• OR •
• EOV. or .NEOV.
Lowest
+-----------------+------------+
The .AND. and .OR. operators are binary operators and must
appear between their operands.
The .NOT. operator is a unary
operator and appears before its operand. Operators of equal precedence associate left to right.
Page 38
FORTRAN Reference Manual
Expressions
Chapter S
A .AND. B .AND. C
is equivalent to:
(A .AND. B)
.AND. C
.NOT. A .OR. B .AND. C
is equivalent to:
( • NOT. A)
• OR.
Two .NOT. operators must not appear
The expression:
(B • AND. C)
adjacent
to
each
other.
A .AND. .NOT. B
is an example of an allowable expression with two adjacent operators.
5.S
Precedence of Operators
When arithmetic, relational and logical operators appear in the
same expression, their relative precedence is:
+------------+--------------+
I Operator
I Precedence I
+------------+--------------+
Arithmetic I Highest
I
I
+------------+--------------+
Rel~tional
Logical
5.6
Intermediate
Lowest
Evaluation Rules and Restrictions for Expressions
Any variable, array element or function referenced in an
expression must be defined at the time it is referenced. Integer
variables must be defined with an arithmetic value rather than a
statement label set by an ASSIGN statement. If a character
string or substring is referenced in an expression, all the
referenced characters should be defined at the time of the reference.
It is an error to divide by zero. It is also an error to raise
a zero value to a zero or negative power. It is also an error to
raise a negative value to a real or double precision power.
FORTRAN Reference Manual
Page 39
Expressions
5.6.1
Chapter 5
Restrictions on Function References
In any given statement, it is an error if a function reference
within that statement changes any other object in the statement.
If a function reference causes an actual argument to the function to become defined, it is an error to reference that object
anywhere else in the statement containing the function reference.
Page 40
FORTRAN Reference Manual
Specification Statements
Chapter 6
Chapter 6 - Specification Statements
This chapter describes SVS FORTRAN specification statements.
Specification statements are non-executable. They are used to
define properties of user-defined variables, arrays and functions. There are nine types of specification statements:
•
•
•
•
•
•
•
•
•
Type statements,
DIMENSION statements,
COMMON statements,
PARAMETER statements,
IMPLICIT statements,
EXTERNAL statements,
INTRINSIC statements,
SAVE statements,
EQUIVALENCE statements.
Specification statements must precede all executable statements
in a subprogram unit. If any IMPLICIT statements appear in the
subprogram, ,they must precede all other specification statements.
Other than that, specification statements can appear in any order
within their own group.
6.1
Type Statements - Declaring Data Types
!YP! statements specify the data type of user-defined names. A
type statement either confirms or overrides the default type
rules for names.
Type statements can also convey dimension
information when declaring arrays. A user-defined name for a
variable, array,
formal
argu~ent,
external
function
or
statement-function can appear 1n a type statement. Such an
appearance defines the type of that name for the entire program
uni~
that contains the type statement.
In any given program
unit, a user-defined name may only appear once in a type statement •.
A type statement can confirm the type of an intrinsic function,
but it is not required to do so.
A main program name or a subroutine name must not appear in a type statement.
A type statement can define the dimensions of an array, or the dimensions can
be declared in a DIMENSION statement (see below), independently
of the type statement.
FORTRAN Reference Manual
Page 41
Specification Statements
6.1.1
Chapter 6
Arithmetic Type Statements
Arithmetic type statements are used to declare arithmetic
objects. The form of an arithmetic type statement is:
type
var [, var J ••
data
0
is one
of
INTEGER,
INTEGER*l,
INTEGER*2,
INTEGER*4, REAL, REAL*4, REAL*8, DOUBLE PRECISION
or COMPLEX.
'type'
INTEGER and INTEGER*4 are the same.
REAL and
REAL*4 are the same. DOUBLE PRECISION and REAL*8
are the same.
evart
is a variable name, array name, formal argument
name, function name or array declarator. See the
definition of array declarators in the chapter on
"FORTRAN Names, Arrays and Substrings".
Examples of Arithmetic !r2! Statements
C
declare some integer variables.
C
INTEGER
CLOCK, HANDS(2), TIME(24)
C
C
declare some real and double precision variables.
C
REAL RADIO, K10l, VARBLS(lO, 10, 5)
DOUBLE PRECISION TWOS(50), TWICE, SECOND
C
C
declare some complex data items
C
COMPLEX
6.1.2
FUNKS, ROCK, BACH(48)
CHARACTER Type Statement
The character type statement is used to declare CHARACTER data
objects. The form of a CHARACTER type statement is:
CHARACTER [*nnn [,1] var [*nnn] [, var [*nnn] J •••
'var'
is a variable name, array name, formal argument
name or an array declarator. For a definition of
an array declarator, see the chapter on "FORTRAN
Names, Arrays and Substrings".
'nnn'
is the length, in characters, of a character
variable or character array element. The length
Page 42
FORTRAN Reference Manual
Specification Statements
Chapter 6
must be an unsigned integer in the range 1 to 255
or a constant expression enclosed in parentheses,
whose value lies in the range 1 to 255.
The
length can also be specified as (*), when the
name is being defined either as a formal argument
which is an assumed size character string, or for
the purpose of establishing a type for later use
in a PARAMETER statement.
The length 'nnn', following the type name CHARACTER,
is the
default length for any name in the list that does not have its
length specified explicitly.
In the absence of a length specification, the default length is one (1).
A length immediately
following a variable or array element overrides the default
length for that item only. For an array, the length specifies
the length of each element of that array.
A formal argument defined as CHARACTER*(*) cannot be used as an
actual argument to a procedure if it is concatenated in a character string expression, whereas a symbolic name of a constant can
be used in such a place.
Examples of CHARACTER
!YE! Statements
CHARACTER FLIP*lO, FLOP*20
CHARACTER WILD(15)*20
CHARACTER*80 LINE(24)
CHARACTER*(lO*20) LSTR
CHARACTER* (*)
VARBLE
6.1.3
LOGICAL Type Statement
The logical type statement is used to declare
objects. The form of a LOGICAL type statement is:
type
'type'
logical
data
var [, var]
is one of LOGICAL, LOGICAL*l, LOGICAL*2 or
CAL*4.
LOGI-
LOGICAL is the same as LOGICAL*4.
evart
is a variable name, array name, formal argument
name, function name, or an array declarator. For
a definition of an array declarator, see the
chapter
on "FORTRAN Names, Arrays and Substrings".
FORTRAN Reference Manual
Page 43
Chapter 6
Specification Statements
Examples of LOGICAL !Ie! Declarations
LOGICAL
LOGICAL*2
LOGICAL'
6.2
SONG
BLACK, WHITE
YES (lO) , NO(lO)
DIMENSION - Declare Data Dimension
A DIMENSION statement specifies the number of dimensions
user-defined array. The form of a DIMENSION statement is:
DIMENSION
of
a
var (dim) [, var (dim)] •••
where each one of the 'var(dim)' pairs is an array declarator
the form:
of
name (d [, d ] ••• )
'name'
is the user-defined name of the array,
'd'
is a dimension declarator.
6.2.1
Dimension Declarators
The number of dimensions in the array is the number of d~men
sion declarators in the array declarator. The maximum number of
dimensions is seven. The rules for array and dimension declarators are defined in the chapter on "FORTRAN Names, Arrays and
Substrings".
Examples of DIMENSION Statements
6.3
DIMENSION
forth(lO, 5:15, 0:99)
DIMENSION
axis(6)
COMMON - Declare a COMMON Block
Common blocks provide a means to share variables between multiple independently-compiled program units.
Common blocks and
their contents are defined via the COMMON statement. The form of
the COMMON statement is:
COMMON [/[cname]/] nlist [[,] /[cname]/ nlist] •••
Page 44
FORTRAN Reference Manual
Specification Statements
Chapter 6
'cname'
is a common block name. If any 'cname' is
ted, the blank common block is implied.
omit-
'nlist'
is a list of variable names, array names and
array declarators, all separated by commas. Formal argument names and function names must not
appear in a COMMON statement.
In each COMMON statement, all variables and arrays appearing in
each 'nlist', following a common block name, are declared to be
in that common block. If the 'cname' is omitted, all elements
appearing in the 'nlist' are specified to be in the blank common
block.
Any common block name can appear more than once in COMMON
statements in the same subprogram unit. All elements in all
'nlists' for the same common block are allocated storage, sequentially in that common block, in the order of their declaration.
All elements in a single common area must be all of type
acter or none of type character.
char-
The size of a common block is equal to the number of bytes of
storage needed to hold all elements in that common block. If the
same named common block is referenced in several subprogram
units, the size must be the same in all those units.
Examples of COMMON Statements
COMMON
6.4
/horde/
TOKEN(lOO), SYMBOL (lOO)
PARAMETER - Make a Symbolic Association
A PARAMETER statement associates a symbolic name with a constant value.
That constant is thereafter associated with that
symbolic name, such that using the name is synonymous with a use
of the constant. The form of a PARAMETER statement is:
PARAMETER (name=expr (,name=expr] ••• )
'name"
is the symbolic name to be defined,
'expr'
is an expression that is to be associated with
the name.
The expression noted in the definition above must be a constant expression.
FORTRAN Reference Manual
Page 45
Chapter 6
Specification Statements
Examples of PARAMETER Statements
C
6.5
PARAMETER
(TODAY
PARAMETER
(BASE
= 'FRIDAY')
s
1,
LIMIT
= 100)
IMPLICIT - Establish Default Data Type
FORTRAN normally assigns a default type to a variable depending
on the first letter of that variable.
The IMPLICIT statement
overrides the default type rules and establishes a new default
type for variables. The form of the IMPLICIT statement is:
IMPLICIT type (letter-list) [,' type (letter-list)] •••
'type'
is one of the data types: INTEGER,
INTEGER·l,
INTEGER*2, INTEGER*4, REAL, REAL*4, REAL*B, LOGICAL, LOGICAL*l, LOGICAL*2, LOGICAL*4, DOUBLE
PRECISION, COMPLEX or CHARACTER[*nnn]
'letter-list'
is a list of single letters or ranges of letters.
A range of letters is indicated by the first and
last letters in the range, separated by a minus
sign.
If a range is specified, the letters must
be in alphabetical order.
'nnn'
is only applicable to a character data type, and
is the size of the character type that is to be
associated with that letter or letters.
'nnn'
must be an unsigned integer in the range 1 thru
255. If 'nnn' is not specified, a value of one
(1) is assumed.
An IMPLICIT statement defines the type and size for all userdefined names that begin with the letter or letters appearing in
the specification. An IMPLICIT statement only applies to the
program unit in which it appears. IMPLICIT statements do not
change the type of any intrinsic functions.
Implicit types can be overridden or confirmed for any specific
user-defined name if that name appears in a subsequent type
statement. An explicit type in a FUNCTION statement also takes
precedence over an IMPLICIT statement. If the type in question
is a character type, the length is also overridden by any later
type specification.
A program unit can have more than one IMPLICIT statement, but
all IMPLICIT statements must precede all other specification
statements.
Page 46
FORTRAN Reference Manual
Specification Statements
Chapter 6
Examples of IMPLICIT Statements
*
*
**
*
*
**
*
6.6
declare all names beginning with A as integer.
IMPLICIT
INTEGER
(A)
declare all names starting with the letters
0, X, Y or Z to be complex.
IMPLICIT
COMPLEX
(0, X-Z)
declare all names starting with C as CHARACTER.
IMPLICIT
CHARACTER*255
(C)
EXTERNAL - Declare External or Dummy Procedure
An EXTERNAL statement specifies that a user-defined name is the
name of an external procedure or a dummy procedure. It also
allows such a name to be used as an actual argument to a subroutine or function reference. The form of an EXTERNAL statement
is:
EXTERNAL proc-name [, proc-name] •••
where each 'proc-name' is the name of an external procedure,
dummy procedure or block data subprogram.
A name appearing in
an EXTERNAL statement declares that name to be an external procedure.
Statement-function names must not appear in an EXTERNAL
ment.
state-
If an intrinsic function name appears in an EXTERNAL statement,
that name becomes the name of an external procedure and the
corresponding intrinsic function can no longer be called from
that program unit.
A user-defined name can only appear once in an EXTERNAL
ment.
6.7
state-
INTRINSIC - Declare Intrinsic Function
An INTRINSIC statement declares that a name is an intrinsic
function.
It also allows a specific intrinsic function name to
be used as an actual argument to a subroutine or function reference. The form of an INTRINSIC statement is:
INTRINSIC
name [, name]
FORTRAN Reference Manual
Page 47
Specification Statements
Chapter 6
where 'name' is an intrinsic function name.
An intrinsic function is anyone of a specific set of predefined functions in the FORTRAN language, such as SIN or MAX.
Since these names are in no way reserved, user functions as well
as variables can have the same name as an intrinsic function, but
not in the same scope. The intrinsic statement, though normally
not required, is used to specifically state that the name in
question refers to an intrinsic function.
Each name may appear only once in an INTRINSIC statement. If a
name appears in an INTRINSIC statement, it may not appear in an
EXTERNAL statement.
All names used in an INTRINSIC statement must be system-defined
intrinsic functions. For a list of intrinsic functions, see the
Appendix on "Intrinsic Functions".
If a specific name of an intrinsic function is used as an
actual argument in a program unit, that name must be declared in
an INTRINSIC statement in that program unit.
If a generic function name appears in an INTRINSIC
that function still retains its generic properties.
statement,
In a given program unit, a name must not appear
one INTRINSIC statement.
more
Certain intrinsic functions may not be
ments. These are:
used
as
in
than
actual
argu-
The type-conversion functions: INT, IFIX, IDINT, FLOAT,
REAL, DBLE, CMPLX, ICHAR and CHAR.
SNGL,
The lexical relationship functions: LGE, LGT, LLE and LLT.
The functions for choosing largest or smallest values: MAX,
MAXO, AMAXl, DMAXl, AMAXO, MAXl, MIN, MINO, AMINl, DMINl,
AMINO and MINl.
6.8
SAVE - Retain Definition Status
A SAVE statement is used to retain the definition of a program
object after returning from the procedure which defines that program object. Within a subroutine or function subprogram, a program object specified in a SAVE statement remains defined after
exit from the subroutine or function. The form of a SAVE statement is:
SAVE
[thing [, thing] ••• ]
where 'thing' is a common
Page 48
block
name
enclosed
in
slashes,
a
FORTRAN Reference Manual
Specification Statements
Chapter 6
variable name or an array name. Any given name may only appear
once in a SAVE statement. The names of dummy arguments, procedures and objects appearing in common blocks must not appear in
a SAVE statement.
If a SAVE statement appears without an associated list of program objects, it is the same as if all objects in that program
unit which could appear in the SAVE statement actually had
appeared in the SAVE statement.
Specifying a
as saving all
mentioned in a
ment in every
SAVE statement
optional.
common block name in a SAVE statement is the same
the elements in that common block. A common block
SAVE statement must be mentioned in a SAVE statesubprogram in which that common block appears.
A
has no effect in the main program, and is
Examples of SAVE Statements
C
C
Save everything in the subprogram with
C
SAVE
C
C
Save some variables
C
SAVE
dimes, nickels, pennies
C
C
Save all of common blocks
C
SAVE
6.9
/Stamps/, /Lettrs/
EQUIVALENCE - Share Storage Between Elements
An EQUIVALENCE statement specifies that two or more variables
or arrays are to share the same storage. If the shared variables
are of different types, the EQUIVALENCE statement does not cause
any kind of automatic type conversion. The form of an EQUIVALENCE
statement is:
EQUIVALENCE
where 'nlist'
(nlist)
[, (nlist) ]
is a list of at least two variable names, array
names, array element names or character substring
names.
Argument names must not
appear
in
EQUIVALENCE
statements.
Subscripts must be
integer constant expressions and must be within
the bounds of the array that they reference.
FORTRAN Reference Manual
Page 49
Chapter 6
Specification Statements
An EQUIVALENCE statement specifies that the storage sequences
of the elements that appear in the list 'nlist' have the same
first storage location. Two or more variables are said to be
associated if they refer to the same actual storage. Thus an
EQUIVALENCE statement causes its list of variables to become
associated.
If an array name appears in an EQUIVALENCE list, it
refers to the first element of the array.
6.9.1
Restrictions on EQUIVALENCE Statements
An EQUIVALENCE statement must not specify that the same storage
location is to appear more than once. For example:
REAL R, S (10)
EQUIVALENCE (R, S(l», (R, S(5»
is in error because it forces the variable "R" to appear
distinct memory locations, namely at S(l) and S(5).
in
two
An EQUIVALENCE statement must not specify that consecutive
array elements be stored out of sequential order. For example:
REAL R(lO), S(lO)
EQUIVALENCE (R(l), S(l», (R(5), S(6»
is in error because, having defined R(l) and S(l) to be associated, the statement then attempts to define R(5) and S(6) to be
associated, and this means that the array "R" has somehow been
"stretched".
Names of dummy arguments must not appear in an EQUIVALENCE
statement.
Also, if a variable name is also a function name,
that name must not appear in an EQUIVALENCE statement.
When EQUIVALENCE statements and COMMON statements are used
together, there are further restrictions. An EQUIVALENCE statement must not try to associate storage elements in different common blocks.
An EQUIVALENCE statement can extend a common block
by adding storage elements following the common block, but not
preceding the common block •. For example:
COMMON /MASSES/ R(lO)
REAL S(lO)
EQUIVALENCE (R(l), S(lO»
is in error because it tries to extend the common block by adding
storage before the start of the block. That is, when R(l) and
S(10) are associated, it means that S(l) would be nine locations
before the defined start of the block.
Page 50
FORTRAN Reference Manual
Chapter 7
Data Initialization
Chapter 1 - Data Initialization
The DATA statement is used to (statically) initialize data
variables.
The DATA statement is non-executable in the sense
that the compiler does not generate any code for it.
If a DATA statement is present within a subprogram, it may
appear anywhere after the specification statements (if there are
any). The form of a DATA statement is:
DATA
nlist /clist/ [[,] nlist /clist/]
enlist'
is a list of variables, arrays, array element
names, substring names and implied-DO lists.
'clist'
is a list of constants, or constants preceded by
an integer-constant repeat-factor and an asterisk. Examples of repeated data items are:
5*3.14159
3*'Help'
100*0
There must be the same number of values in each 'clist' as
there are variables or array elements in the corresponding
enlist'. The appearance of an array in an enlist' is equivalent
to a list of all the elements in that array in order of storage
sequence. Array elements and substrings may be indexed by
integer constant expressions (but see the implied-DO loop below).
The type of each element in a 'clist' must be the same as the
type of the corresponding variable or array element in the accompanying enlist'. If necessary, the 'clist' constant is converted
to the type of the enlist' object according to the rules for
arithmetic conversion given in the table in the chapter on
-Assignment Statements-.
A DATA statement can be used to initialize any variable,
element or substring that is ~ one of the following:
array
a dummy argument,
an object in blank common or any object
with an object in blank common,
FORT~N
Reference Manual
which
is
associated
Page 5L
Data Initialization
Chapter 7
a variable in a function subprogram whose name is also the
same name as that of the function or one of its alternate
entry-point names.
Objects may only be initialized once in any given program unit.
DATA statements in BLOCK DATA subprograms may
objects in named COMMON areas.
only
initialize
Examples of DATA Statements
**
*
Declare some variables
*
*
initialize some reals
*
**
**
*
**
*
7.1
REAL FIRST, SECOND
INTEGER NIG, NOG
COMPLEX WEIRD(lO)
DOUBLE PRECISION VECT(5)
DATA
FIRST, SECOND
/1.0, 2.0/
initialize some integers
DATA NIG /10/, NOG /20/
initialize two elements of the complex array
DATA
WEIRD(2), WEIRD(5)
/2
*
(0.0,0.0)/
initialize all the double precision array
DATA
VECT
/0.0, 0.0, 0.0, 0.0, 0.0/
Initializing Character Variables
If an 'nlist' item is of type character, the corresponding
'clist' item must be a character constant expression.
If the 'clist' item is shorter than the length of the 'nlist'
item, the initial characters occupy the leftmost positions of the
character data item and the remaining character positions are
filled with spaces.
If the 'clist'
item is longer than its
corresponding 'nlist' variable, only the characters needed to
initialize the 'nlist' item are used and the remaining characters
are ignored.
Page 52
FORTRAN Reference Manual
Chapter 7
Data Initialization
Examples of Character Initialization
DATA
STRING
DATA
SVECT
/'Old Rope'/
/6 * 'Attached'/
The second example assumes that SVECT is a 6 element
array.
7.2
character
Initializing Non CHARACTER Variables to CHARACTER Values
The FORTRAN-77 standard explicitly forbids initializing variables of any non-CHARACTER data type with CHARACTER values.
Almost all FORTRAN-66 compilers do allow such initializations
(they had to since there was no explicit CHARACTER data type in
FORTRAN-66). SVS FORTRAN-77 therefore allows such initialization, provided the $CHAREQU option has been selected. For a
description of the $CHAREQU option, see Chapter 13
"FORTRAN
Compile Time Options".
When the $CHAREQU compiler option is selected, non-CH~RACTER
variables may be initialized with CHARACTER constants. Each
character constant in the 'clist' initializes precisely one variable in the enlist'. If the length of the CHARACTER constant is
longer than the number of bytes which the target variable occupies, the CHARACTER constant is truncated on the right to the
same size as the target variable. If the CHARACTER constant is
shorter in length than the number of b¥tes which the tar~et variable occupies, the CHARACTER constant is padded on the right with
trailing spaces.
In all cases, one ASCII character is stored in each byte of the
target variable.
Thus a variable of type REAL*4 would receive
exactly four bytes, and a variable of type LOGICAL*l would
receive exactly one byte.
FORTRAN Reference Manual
Page 53
Data Initialization
Chapter 7
Example of Initializing
~-CHARACTER
Variables
•
Select the option
$CHAREQU
•
Declare some variables
LOGICAL· 1 ARG
DOUBLE PRECISION
•
•
LARJ
Initialize the variables
DATA
ARG, LARJ
/'SING', 'SONG'/
In the example above, the variable ARG would receive the single
character value'S' (because of truncation), while the variable
LARJ would receive the value 'SONG
'(because of space padding) •
7.3
Implied DO in DATA Statements
As an added convenience, the DATA statement can incorporate a
form of DO loop for initializing arrays (for example) in a regular and concise way_ This is known as an "implied-DO loop" and
has the same form as a DO statement (see the chapter on "Control
Statements"). The control variables of an implied-DO loop are
declared implicitly and only for the duration of the DO loop.
The following example should clarify the use of implied-DO loop
initialization. In this case, the form of an implied-DO list in
the DATA statement is:
(dlist, i
'dlist'
, i'
= first, last [, inc])
is a list of array element names
lists.
and
is the name of an integer variable, the
implied-DO
implied-
DO-variable~
'first', 'last' and 'inc'
are each integer constant expressions.
The
expressions can contain implied-DO-variables of
other implied-DO lists whose range includes this
implied-DO list.
The range of an implied-DO list is the list 'dlist'. An iteration count and the values of the implied-DO-variable are established from 'first', 'last' and 'inc' just as for a DO loop, but
the iteration count must be positive.
Page 54
FORTRAN Reference Manual
Data Initialization
Chapter 7
When an implied-DO list appears in a DATA statement, the items
in 'dlist' are specified once for each iteration of the impliedDO list with the appropriate substitution of values for any
occurrence of the implied-DO-variable tit.
The implied-DO-variable can have the same name as a variable in
the subprogram unit containing the DATA statement - there is no
conflict of such names.
Examples of Implied DO Loop Initialization
C
C
Declare some large arrays.
INTEGER PRIMES (1000)
INTEGER UPRTRI(20, 20)
REAL MATR~X(25, 80)
C
C
C
C
Now initialize with DATA statements
containing an implied-DO loop.
= 1,
DATA
(PRIMES(I), I
DATA
«MATRIX(J, K), J
C
1000) /1000*1/
= 1,80),
K
= 1,
25) /2000*1.0/
C
C
C
The last DATA statement initializes the upper triangle
of the array called~UPRTRI.
C
DATA
«UPRTRI(I, J), J=l, 21-1),1= 1, 20) /210 * 0/
FORTRAN Reference Manual
Page 55
Data Initialization
Page 56
Chapter 1
FORTRAN Reference Manual
Assignment Statements
Chapter 8
Chapter 8 - Assignaent Stateaents
An assignment statement computes a value which is then assigned
to a program object. Because FORTRAN does not require that variables be declared ahead of time, it is possible that assignment
actually causes that object to become allocated. There are four
distinct types of assignment statements:
•
•
•
•
8.1
Arithmetic,
Logical,
Statement Label assignment (the ASSIGN) statement,
Character assignment.
Arithmetic Assignment
Arithmetic assignment evaluates aR arithmetic
assigns the result to a variable. The form is:
variable
expression,
and
expression
'variable'
is a variable or an array element name, of type
INTEGER, INTEGER*l, INTEGER*2, INTEGER*4 REAL,
REAL*4, REAL*8, DOUBLE PRECISION or COMPLEX.
'expression'
is an expression compatible
types.
with
one
of
those
If the type of 'variable' and the type of 'expression' are not
compatible, the value of 'expression' is automatically converted
to the type of 'variable' according to the following table.
FORTRAN Reference Manual
Page 57
Chapter 8
Assignment Statements
I
I
I+------------------+-------------------+
+------------------+------~------------+
Type of variable
or array element
I
Integer
Value Assigned
I
INT(expression)
I
+------------------+-------------------+
I Real
I REAL (expression) I
+------------------+-------------------+
I Double Precision I DBLE(expression) I
+------------------+-------------------+
I Complex
I CMPLX(expression) I
+------------------+-------------------+
Table 8-1
Type Conversion for Arithmetic Assignment Statements
The functions in the "Value Assigned" column in the above table
are generic intrinsic functions described in the table of intrinsic functions in the appendices.
If an integer expression is of type INTEGER·l, INTEGER*2 or
INTEGER*4, the result is automatically converted to the correct
type. In general, an operand is "promoted" to the larger size
when evaluating expressions.
Operands of type INTEGER*l are
always promoted to INTEGER*2.
In the assignment statement, converting a longer INTEGER value
to a shorter one is done by truncation. The FORTRAN-77 compiler
does not issue any warning about it, and the running program
issues no error m~·ssages because of truncation.
Examples of Arithmetic Assignment
INTEGER whole
REAL party
DOUBLE PRECISION
COMPLEX house
*
•
8.2
DATA
huge
whole /10/, party /5.5/
whole = 50 * party
party = 3.1415926388 * 5.7
huge = 547.987654e243
house = (1.0, -4.0)
Logical Assignment
Logical assignment assigns the value of an expression to a logical variable. The value of the expression must therefore evaluate to either of the values .TRUE. or .FALSE. The form of a logical assignment statement is:
Page 58
FORTRAN Reference Manual
Assignment .Statements
Chapter 8
logical variable - logical expression
If a logical variable is one of the types LOGICAL*l, LOGICAL*2
and LOGICAL*4, a logical expression is automatically converted to
the correct type.
Examples of Logical Assignment Statements
LOGICAL
TELLME, NONO
TELLME - .TRUE.
NONO
8.3
II:
.FALSE.
Statement Label Assignment
Statement label assignment is used to assign the value of a
format labey--Q[ a statement label to an integer variable. The
form of statement label assignment is:
ASSIGN statement-label TO integer-variable
'statement-label'
is a format label or a statement label.
'integer-variable'
is an integer variable.
The integer ·variable
must not be of type INTEGER*l or INTEGER*2.
Executing an ASSIGN statement sets the integer variable to the
value of the label. The label can be either a format label or a
statement label. The label must appear in the same program unit
unit as the ASSIGN statement. When used in an assigned GO TO
statement, a variable must currently have the value of a statement label.
When used as a format specifier in an input-output
statement, a variable must have the value of a format statement
label.
The ASSIGN statement is the only way to assign the value
of a label to a variable.
Examples of Statement Label Assignment
INTEGER
SLAB, FORLAB
ASSIGN
666 TO SLAB
ASSIGN
905 TO FORLAB
FORTRAN Reference Manual
Page 59
Chapter 8
Assignment Statements
8.4
Character Assignment
Character assignment evaluates a character expression and
assigns the result to a character variable, character arrayThe form of a character
element or a character substring.
assignment statement is:
character-variable - character-expression
None of the character positions being defined in the left-hand
side of the assignment may appear on the right-hand side of the
assignment. If they do, the results are undefined.
The left-hand side and the right-hand side of the assignment
may be of different lengths. If the left-hand side is longer,
the effect is to extend the right-hand side value to the right
with spaces until it is the same length as the left-hand side.
If the left-hand side is shorter, the effect is to take a substring starting at position 1 of the right-hand side, short
enough to assign to the left-hand side.
Only as much of the right-hand side need be defined as is
necessary to define the left-hand side. For example, consider
the following program fragment:
CHARACTER fred*4, bill*8
fred
= bill
The assignment of 'bill' to 'fred' above requires that the substrin? 'bi1l(l:4)' be defined, since that is enough to define
'fred. It is not required that the rest of 'bill', 'bil1(5:8)',
be defined.
If the left-hand side of the assignment is a substring reference, the right-hand side is assigned only to the substring. The
definition status of the character positions not defined on the
left-hand side does not change.
Page 60
FORTRAN Reference Manual
Chapter 9
Control Statements
Chapter 9 - Control Statements
Control statements are used to direct the sequence of execution
of a FORTRAN program. Control statements include constructs to
execute statements selectively depending on the outcome of a logical expression (the block IF and logical IF statements), perform
blocks of statements repetitively (the DO statement), to select
one of a number of statements to execute depending on the value
of an integer expression (the computed GO TO statement, and to
terminate or suspend program execution (the STOP and PAUSE statements). This chapter covers the control statements of FORTRAN,
in this order:
Block IF THEN ELSE
a "structured coding" construct which was
introduced to FORTRAN with FORTRAN-77,
newly
Logical IF
which executes or does not execute a subordinate
statement depending on the truth or falsity of a
logical expression,
Arithmetic IF
which executes a three-way branch depending
the value of an arithmetic expression,
DO
which is FORTRAN's principal means of
trol,
CONTINUE
which acts as a "null" statement,
STOP
to stop
PAUSE
to suspend program execution,
progra~
loop
on
con-
execution,
Unconditional GO TO
which unconditionally
transfers
another part of the program unit,
control
to
Computed GO TO
which selects a statement label to execute,
depending on the value of an expression,
Assigned GO TO
which uses the value of an integer variable as
statement label.
FORTRAN Reference Manual
a
Page 61
Control Statements
9.1
Chapter 9
Block IF THEN ELSE Statement
The block IF THEN ELSE statement group (described in the subsections below) represent "structured coding" constructs that
control program execution flow without the need for indiscriminate jumping around via GO TO statements. As an overview of the
subsections to follow, the three code skeletons below illustrate
the basic ideas of the IF THEN ELSE statement groups.
Skeleton 1 - a simple block IF which skips a
ments if the expression is false:
group
of
state-
IF (I .LT. 10) THEN
some statements that are executed
••••• only if I < 10
ENDIF
Skeleton 2 - block IF with a series of ELSEIF statements.
IF (J .GT. 1000) THEN
••••• some statements executed only
••••• if J > 1000
ELSE IF (J .GT. 100) THEN
••••• some statements executed only
••••• if J > 100 and <= 1000
ELSEIF (J .GT. 10) THEN
••••• some statements executed only
••••• if J > 10 and <= 100
ELSE
••••• some statements executed only if
••••• none of the above condit10ns were true
ENDIF
Skeleton 3 - shows that the constructs can be nested and that
an ELSE statement can follow an IF block without intervening
ELSEIF statements. The indentation is here to enhance readability - FORTRAN does not require it.
IF (I .LT. 100) THEN
some statements executed only
•• if 1 < 100
IF (J .LT. 10) THEN
•. some statements executed only
•• if I < 100 and J < 10
ENDIF
•• some more statements executed only if
•• I < 100
ELSEIF (I .LT. 1000) THEN
some statements executed only
•• if 1 => 100 and I < 1000
IF (J .LT. 10) THEN
•• some statements executed only
Page 62
FORTRAN Reference Manual
Control Statements
Chapter 9
•• if I => 100 and I < 10000 and J < 10
ENDIF
•• some more statements executed only if
•• I -> 100 and I < 1000
ENDIF
To provide a detailed understanding of the block IF and its
associated statements, the concept of the IF-level is introduced.
For any statement, its IF-level is:
nl - n2
where 'nl' is the number of block IF statements
(including the
current statement) from the beginning of the current program unit
and 'n2' is the number of ENDIF statements (not including the
current statement) from the beginning of the current program
unit. The IF-level of every statement must be greater than or
equal to zero, and the IF-level of every block (IF, ELSEIF, ELSE
and ENDIF) must be greater than zero. The IF-level of every END
statement must be zero. The IF-level is used to define the nesting rules for the block IF and its associated statements and to
define the extent of IF blocks, ELSEIF blocks, and ELSE blocks.
9.1.1
Block IF Statement
The general form of the Block IF statement is:
IF (logical expression)
THEN
Execution of the block IF statement involves evaluating the
'logical expression'. If the value is true and there is at least
one statement in the IF block, the next statement executed is the
first statement of the IF block. After the last statement in the
IF block is executed, the next statement to be executed is the
next ENDIF statement at the same IF-level as this IF statement.
If the value is true and there are no executable statements in
the IF block, the next statement to be executed is the next ENDIF
statement at the same IF-level as this IF statement.
If the value of the 'logical expression'
is false,
the next
statement to be executed is the next ELSEIF, ELSE or ENDIF statement at the same IF-level as this IF statement.
Note that transfer of control into an IF block from outside the
IF block is not allowed.
The bloqk IF statement, in fact, looks no different from the
logical IF statement described later; however, the presence of
the THEN keyword as the next statement indicates that a structured block IF statement group follows.
FORTRAN Reference Manual
Page 63
Control Statements
9.1.2
Chapter 9
ELSEIF Statement
The form of the ELSEIF statement is:
ELSEIF (logical expression) THEN
The ELSE IF block associated with an ELSEIF statement consists
of the (possibly zero) executable statements up to, but not
including, the next ELSEIF, ELSE or ENDIF statement that has the
same IF-level as this ELSE IF statement.
Execution of an ELSEIF statement starts by evaluating the 'logical expression'. If the value is true and there is at least one
statement in the ELSEIF block, the next statement executed is the
first statement of the ELSEIF block. After the last statement in
the ELSEIF block is executed, the next statement to be executed
is the next ENDIF statement at the same IF-level as this ELSE IF
statement.
If the value of 'logical expression' is true and there are no
executable statements in the ELSEIF block, the next statement to
be executed is the next ENDIF statement at the same IF-level as
this ELSEIF statement.
If the 'logical expression' evaluates to false, the next statement to be executed is the next ELSEIF, ELSE or ENDIF statement
that has the same IF-level as this ELSE IF statement.
Note that transfer of control into an ELSEIF block from outside
of that ELSEIF block is not allowed.
9.1.3
ELSE Statement
The form of an ELSE statement is:
ELSE
The ELSE block associated with an ELSE statement consists of
the (possibly zero) statements that follow the ELSE statement, up
to, but not including, the next ENDIF statement that has the same
IF-level as this ELSE statement. The "matching" ENDIF statement
must appear before any intervening ELSE or ELSEIF statements at
the same IF-level.
In other words, there may only be one ELSE
statement in a block IF statement. There is no effect in executing an ELSE statement.
Note that transfer of control into an ELSE block
that ELSE block is not allowed.
Page 64
from
outside
FORTRAN Reference Manual
Control Statements
Chapter 9
9.1.4
ENDIF Statement
The ENDIF statement marks the end of a
form of the ENDIF statement is:
block
IF
group.
The
ENDIF
There is no effect in executing an ENDIF statement. There must
be a -matching- ENDIF statement for every block IF statement in a
program unit in order to specify which statements are in a particular block IF statement.
9.2
Logical IF Statement
The logical IF statement evaluates an expression for logical
TRUE or FALSE, then either executes or does not execute a following statement based on the truth or falsity, respectively, of the
expression. The form of the logical IF statement is:
IF (logical expression) statement
where 'logical expression' is a logical expression, and 'statement'
is any executable statement except a DO statement, Block
IF, ELSEIF, ELSE, ENDIF statement, an END statement, or another
logical IF statement.
The 'logical expression' is evaluated, and if the value of the
expression is true, the 'statement' is executed. If the 'logical
expression' evaluates to false, the 'statement' is not executed
and the execution sequence proceeds as if a CONTINUE statement
had been encountered.
Note that functions in the
objects in the 'statement'.
'logical
expression'
can
affect
Example of Logical IF Statement
IF (Token (1)
9.3
.NE.
32)
RETURN
Arithmetic IF Statement
The arithmetic IF statement performs a GO TO to one of three
statement labels depending on the value of an expression being
negative, zero, or positive.
The form of the arithmetic IF
statement is:
IF (expression) sl, s2, s3
FORTRAN Reference Manual
Page 65
Chapter 9
Control statements
The 'expression' in the description must be an INTEGER,
INTEGER*l, INTEGER*2, INTEGER*4, REAL, REAL*4, REAL*S or DOUBLE
PRECISION expression.
'sl', 's2' and 's3' are the statement labels of executable
statements that appear in the same program unit as does the
arithmetic IF statement itself. The same statement label can
appear more than once among the three labels.
The effect of the arithmetic IF statement is to
'expression' and select a label based upon the value.
evaluate
The statement labeled by 'sl' is executed if the value
expression is less than ~.
of
the
The statement labeled by 's2' is executed if the value
expression is ~.
of
the
The statement labeled by 's3' is executed if the value
expression is greater than ~.
of
the
The next statement executed is the statement labeled by the
selected label.
None of the labels may appear within the range
of a DO loop or inside an IF, ELSEIF or ELSE block, unless the
arithmetic IF statement itself is also in the same range or
block.
Example of the Arithmetic IF Statement
IF (I - 100) 20, 30, 40
C
20 PRINT * , , I is less than 100'
STOP
C
C
9.4
30 PRINT *, , I is exactly 100'
STOP
40 PRINT *, , I is more than 100'
STOP
DO Statement - Loop Control
A DO statement block is used for "loop control" purposes such
as applying some operation to each element of an aggregate. The
form of a DO statement is:
DO s [,] i-e1, e2 [,
, s'
Page 66
e3]
is the statement label of an executable statement.
The label must follow this DO statement
FORTRAN Reference Manual
Chapter 9
Control Statements
and be contained in the same program unit.
, i'
is an integer, real or double precision variable,
called the DO-variable.
'el', 'e2' and 'e3'
are each an integer,
expression.
real
or
double
precision
The statement labeled by's' is called the "terminal
statement" of the DO statement. The terminal statement must not be an
unconditional GO TO, assigned GO TO, arithmetic IF, block IF,
ELSEIF, ELSE, ENDIF, RETURN, STOP, END or another DO statement.
If the terminal statement is a·logical IF,
it may contain any
executable statement except those not permitted inside a logical
IF statement.
A DO statement is said to have 'a "range",
beginning with the
statement which follows the DO statement and ending with (and
including) the terminal statement. If a DO statement appears in
the range of another DO loop, its range must be entirely contained within the range of the enclosing DO loop. If a DO state~ent
appears within an IF block, ELSEIF block or ELSE block, the
range of the associated DO loop must be entirely contained in the
particular block.
If ,a block lF statement appears within the
range of a DO loop, its associated ENDIF statement must also
appear within the range of the DO loop. More than one DO loop
can have the same terminal statement.
The DO control variable, 'it, must not be set by any statement
within the range of the DO loop associated with that control
variable.
Transfer of control into the range of a DO
the range of a DO loop is not allowed.
9.4.1
loop
from
outside
DO Loop Initialization Sequence
Execution of a DO statement causes the following events to happen in order:
the expressions 'el',
'e2'
and 'e3' are evaluated, with
conversion to the type of the DO variable if necessary,
according to the rules specified for type conversion in the
chapter on "Assignment Statements".
If le3' is omitted, a
default value of 1 is used.
If 'e3' is present, it must not
evaluate to zero,
the DO variable 'it is set to the value of 'ell,
FORTRAN Reference Manual
Page 67
Control Statements
Chapter 9
The iteration count for the DO loop is computed as:
MAX(INT«e2 - el + el) / ell, 0)
which may be zero,
The iteration count is tested: if it exceeds zero, the statements in the range of the DO' loop are executed.
DO' Loop Incrementation Processing
9.4.2
After the terminal statement of the DO' loop
following events occur in order:
is
executed,
the
The value of the DO' control variable 'i! is incremented by the
value of 'e3' which was computed when the DO' statement was
executed,
The iteration count is decremented by one (1),
The iteration count is tested; if it exceeds zero, the statements in the range of the DO loop are executed again.
The value of the DO control variable is well-defined, whether
the DO' loop is exited as a result of the iteration count being
zero, or as a result of an explicit transfer of control out of
the DO' loop.
•
Example of the final value of a DO' control variable:
This program fragment prints the number
1 thru lIon the console
C
C
DO'
200
9.4.3
200
I
= 1,
10
WRITE(*, '(IS)') I
WRI TE (*, '( I 5) ') I
Events Which Terminate a DO' Loop
A DO' loop is exited under the following conditions:
when its' iteration count is zero, as described above under "DO'
Loop Execution Sequence",
a RETURN statement is executed within
loop,
the
range
of
the
DO'
control is transferred to a statement in the same program unit
but outside the range of the DO' loop,
Page 68
FORTRAN Reference Manual
Control Statements
Chapter 9
a subroutine called from within the range of the DO loop
returns via an alternate return specifier to a statement which
is outside the range of the DO loop,
the program terminates for any reason.
9.5
CONTINUE Statement - Null Statement
The CONTINUE statement is a "null" or "no operation" statement.
It can appear just as any other statement in a program. CONTINUE
has no effect on program execution. CONTINUE is almost always
used as the final statement in a DO block, when the DO block
would otherwise end in a statement that is disallowed in
that
context. The form of a CONTINUE statement is:
CONTINUE
9.6
STOP Statement - Stop Program Execution
The form of a STOP statement is:
STOP
[n]
where Inl is either a character constant or a string of not more
than five digits. The effect of executing a STOP statement is to
cause the program to terminate. If the argument Inl
is present
on the STOP statement, it is displayed on the console upon termination.
9.7
PAUSE Statement - Suspend Program Execution
The form of a PAUSE statement is:
PAUSE
[n]
where Inl is either a character constant or a string of not more
than five digits. The effect of executing a PAUSE statement is
to cause the program to be suspended until there is an indication
from the console that the program should proceed~ If the argument Inl ,is present on the PAUSE statement, it is displayed on
the console as part of a prompt, requesting input from the console. If the indication is received from the console, program
execution resumes as if a CONTINUE statement had been executed.
9.8
Unconditional GO TO Statement
The ~nconditional GO TO causes an unconditional transfer of
control to a specific labeled statement. The form of the unconditional GO TO statement is:
FORTRAN Reference Manual
Page 69
Chapter 9
Control Statements
GO TO statement label
The 'statement label' which is the target of the GO TO statement must be defined within the same program unit as the GO TO
statement. The 'label' must be the statement 1abel of an executable statement.
Transfer of control into a DO, IF, ELSEIF or
ELSE block from outside such a block is not allowed.
9.9
Computed GO TO Statement
A computed GO TO statement acts as a means of transferring control to one out of a set of labeled statements, depending on the
value of an expression. The form of the computed GO TO statement
is:
GO TO ( s [, s ] ...) [, ] i
where 'i' is an integer expression and each's' is a statement
label of an executable statement in the same program unit as the
computed GO TO statement. The same statement label may appear
more than once in the list of statement labels.
The effect of the computed GO TO statement is to evaluate the
integer expression indicated by 'if to a value, say n. Control
is then transferred to the n'th statement label in -the list,
counting from 1.
If the value of n is less than 1, or if the
value of n is greater than the number-of statement labels in the
list, the computed GO TO statement has no effect, and program
execution proceeds as if a CONTINUE statement had been executed.
The same restrictions on transfer of control apply to the computed GO TO statement as those that apply to the unconditional GO
TO statement.
Example of Computed GO TO Statement
WHERE = 3
GO TO (100, 200, 300, 400, 500), WHERE
.•••• execution continues here when 5 < WHERE < 1
C
100
statements executed if WHERE .. I
200
statements executed if WHERE
300
statements executed if WHERE = 3
400
statements executed if WHERE = 4
500
statements executed if WHERE =- 5
C
2
C
C
C
Page 70
FORTRAN Reference Manual
Chapter 9
9.10
Control Statements
Assigned GO TO Statement
The assigned GO TO statement uses the value of an integer variable as a statement label which is to be the target of a GO TO
statement. The effect is as if an unconditional GO TO had been
made to that statement label. The form of the assigned GO TO is:
GO TO i [[,]
(s[, s] ••• )]
where Ii' is the name of an integer variable and each's'
is a
statement label of an executable statement in the same program
unit as the assigned GO TO statement.
The same statement .label
may appear more than once in the list of statement labels in an
assigned GO TO statement.
At the time the assigned GO TO is executed, the integer variable 'i' must be defined with the value of a statement label of
an executable statement which appears in the same program unit as
the assigned GO TO statement.
That variable must have been
defined with an ASSIGN statement.
The integer variable 'i' must
be of type INTEGER or INTEGER*4.
It must not be of type
INTEGER*l or INTEGER*2.
If the optional, parenthesized list of statement-labels is
present, the value of the integer variable Ii' must be that of
one of the statement-labels in the list.
The same restrictions
apply to the assigned GO TO statement as those that apply to the
unconditional GO TO statement.
Example of Assigned GO TO Statement
ASSIGN 645 TO SYSTEM
GO TO SYSTEM, (360, 370, 635, 645, 1108)
645
statements starting here will be executed
because 'SYSTEM' has the value 645.
FORTRAN Reference Manual
Page 71
Control Statements
Page 72
Chapter 9
FORTRAN Reference Manual
Input and Output
Chapter 10
Chapter 10 - Input and Output
This chapter and Chapter 11 describe the FORTRAN input-output
system. This chapter covers the basic concepts of input and output in FORTRAN.
Chapter 11 describes the FORMAT statement.
Topics covered in this chapter and Chapter 11 are:
an overview of the input-output system. Covers the basics of
the FORTRAN file system. Defines the ideas of records, units
and various forms of file access,
a general coverage of the input-output system,
input-output statements are covered, with the exception of the
FORMAT statement, ~hich is covered in Chapter 11.
10.1
Overview of the Input-Output System
This section introduces the basic terms and concepts of the
FORTRAN input-output system. Most tasks related to input-output
can be done without a full understanding of this section, so the
reader can skip to the next section on first reading and use this
section for subsequent reference purposes.
10.1.1
Records
A record is the building block of the FORTRAN input-output system.
A record is a sequence of characters or a sequence of
values. There are three forms of records:
• Formatted records,
• Unformatted records,
• Endfile records.
A formatted record is a sequence of characters terminated by
the character value corresponding to the "end-of-line" key on a
terminal (character value 13, or 10, or both, depending on the
particular operating system). Formatted records are interpreted
consistently on input the same way that the underlying operating
system and any text editor interprets characters. A formatted
record therefore corresponds to the notion of a "line" from a
FORTRAN Reference Manual
Page 73
Input and Output
Chapter 10
device.
Formatted files are normally transportable between different language processors and other text-processing applications.
An unformatted record is a sequence of values. The system does
not alter or interpret such records in any way; neither is there
any representation for an end-of-record as is the case with a
formatted record. Unformatted files are generally not transportable between different language processors or computers because
of differences in the external representations of data.
An endfile record has no physical existence in a file, but the
underlying input-output system supplies an indication of one, as
if there had been some actual record after the last record in a
file.
10.1.2
Files
A FORTRAN file is a sequence of
either Externar-or Internal.
records.
FORTRAN
An external FORTRAN file is a file on a physical
device or it is an actual peripheral device.
files
are
peripheral
An internal FORTRAN file is a character variable which is to
serve as the source or destination of some input-output action.
From here on, FORTRAN files and the files known to the underlying operating system and any text processors are simply called
"files". The correct meaning is determined by the context.
The
OPEN statement provides the association between the two notions
of files, and in most cases, there is no ambiguity after the file
is opened - the two notions of a file being the same.
10.1.3
Properties of Files
A file which is being operated upon by a FORTRAN program has a
number of properties which are described in the paragraphs below.
File properties which are discussed below encompass:
•
•
•
•
Name of the file,
Position of the file,
Record format,
Access method.
10.1.3.1
File Name
A file can have a name. If the name is present, it is a character string identical to that by which it is known to the
operating system.
A
file
can
have
a
name
such
as
Page 74
FORTRAN Reference Manual
Chapter 10
Input and Output
'/source/fourier.text'.
10.1.3.2
File Position
A file has the property of position which is usually established by the preceding input-output operation. There is the
notion of the beginning-of-file, endfile, the current record, the
preceding record, and the next record in the file.
It is possible to be between records, in which case the next record is the
successor to the previous record, and there is no current record.
The position of the file after a sequential write is the endfile
but not beyond the endfile record. Executing an ENDFILE statement positions the file beyond the endfile record. A READ statement executed at end-of-file (but not beyond the endfile record)
positions the file beyond the endfile record. The user can trap
reading of an endfile record via the END= option in a READ statement.
10.1.3.3
Formatted and Unformatted Files
An external file is opened as either formatted or unformatted.
Internal files are always formatted. Formatted files consist
entirely of formatted records.
Unformatted
files
consist
entirely of unformatted records. Formatted files have the structural properties of being a sequence of lines with end-of-line
indicators (usually carriage-return).
10.1.3.4
Sequential and Direct Access Files
An external file is opened as either sequential access or
direct access.
Sequential files contain records in an order
determined by the order in which they were written.
Sequential
files must not be read or written using the REC= option which
specifies a position for direct-access input-output. The system
attempts to extend sequential access files if a record is written
beyond the old endfile, if there is enough room to do this on the
external device.
Direct-access files can be written in any order
(random
access).
Records in a direct-access file are numbered sequentially with the first records having the number one (1).
All
records in a direct-access file have the same length, specified
when the file is opened. Each record in the file is uniquely
referenced by its record number, specified at the time the record
is written. Records can be written out of order, with holes in
the sequence if desired. For example, records 9, 5 and 11 could
be written in that order without writing the intermediate
records.
Once written, a record cannot be deleted, but a record
can be rewritten with a new value. It is an error to read a
record that has not yet been written, but the system can only
FORTRAN Reference Manual
Page 75
Input and Output
Chapter 10
detect this if the attempted read is to a record beyond the
highest numbered record in the file. Direct-access files must
reside on block-structured storage devices, such that a position
in the file is meaningful.
The system attempts to extend a
direct-access file if a write is made to a position beyond the
current highest numbered record in the file, but the success of
this depends on the amount of space on the storage device.
10.1.4
Internal Files
Internal files provide a means for using the formatting capabilities of the input-output system to convert values to and from
their external character representations, within FORTRAN's internal storage structures. That means that reading from a character
variable converts the character values into numeric, logical or
character variables.
Writing to a character variable converts
values into their external character representation.
10.1.4.1
Special Properties of Internal Files
An internal file is a character variable, character array element, character array or character substring.
A record of an internal file is a character variable, character
array element or character substring.
If an internal file is a character variable, character array
element or character substring, such a file has exactly one element, whose length is that of the character variable, character
array element or character substring. If an internal file is a
character array, each element of that array is a record of the
file, with each record being the same length.
If less than an entire record is written by a WRITE
the remainder of the record is filled with spaces.
statement,
The position of an internal file is always at the beginning of
file prior to executing any input-output statement. Only sequential, formatted input-output is allowed on an internal file.
Only the READ and WRITE statements can reference an internal
file. List-directed input-output is not allowed on an internal
file.
10.1.5
Units
A unit is a means of specifying a file.
input-output statement is either of:
A unit specified in an
• External unit specifier,
• Internal unit specifier.
Page 76
FORTRAN Reference Manual
Chapter 10
Input and Output
An external unit specifie~ is either a positive integer expression or the character * which stands for the Standard Input and
Standard Output files for the running program.
In most cases,
external unit specifiers are bound to physical devices (or files
on those devices) by name when the OPEN statement is executed.
Once this binding of a value to a system file name has occurred,
FORTRAN input-output statements refer to the unit number as a
means of referring to the external object. Once opened, the
external unit specifier is uniquely associated with the external
device or file until an explicit CLOSE statement is executed or
until the program terminates.
Unit specifier 0 (zero) is initially associated with the Standard Input and Standard Output files for reading and writing
(respectively) and an explicit OPEN statement is not needed. The
system interprets the character * as specifying unit O. Similarly, the unit specifier 1 (one) is associated with the Standard
Error file if one exists in the system, or the Standard Output
otherwise. No other unit specifiers are initially available
(without explicit OPEN statements) and in particular, unit
specifiers 5 and 6 have no special or initial properties as they
had in certain earlier FORTRAN dialects.
An internal unit identifier is the name of a character variable, character array, character array element or character substring.Internal unit specifiers may only be used in the READ,
WRITE and PRINT statements - they are not allowed in any of the
auxiliary input-output statements such as OPEN, CLOSE, INQUIRE
and so on.
10.2
General Discussion of the Input Output System
FORTRAN supplies a rich selection of possible file structures.
At first there might seem to be a confusing array of choices.
However, two sorts of files will probably suffice for most applications.
The Standard Input and Standard Output files.
These are
referred to as unit 0 or by the character *. The Standard
Input and Output files are sequential, formatted files.
When
reading from the Standard Input, the backspace and line-delete
keys familiar to the user have their normal editing functions.
The Standard Output file has the special property that it is
possible to write partial lines to it (lines which are not
terminated by a carriage-return) by using the "\" or "$" edit
descriptors. These edit-descriptors are described in the next
chapter.
Writing to any other unit does not have this property, even if that unit is explicitly bound to the Standard
Output file by an OPEN statement.
Explicitly-opened
These files are
external,
bound to
FORTRAN Reference Manual
sequential,
formatted
files.
a system file by name in an OPEN
Page 77
Chapter 10
Input and Output
statement. These files can be processed by system editors and
other text-processing utilities.
10.2.1
Pre-Connected Files
The FORTRAN run-time system preconnects unit 0 to the Standard
Input and Standard Output files, as mentioned above. On those
operating systems which support a standard error file, the FORTRAN run-time system also preconnects unit 1 to that stream. On
those systems which do not support a standard error file, the
FORTRAN run-time system preconnects unit 1 to the same place as
the Standard Output.
10.2.2
Examples of Common Input Output Operations
Here is an example program reading and writing the sorts of
files discussed in the paragraph above. The specific inputoutput statements are discussed in detail in the next section.
C
C
C
C
C
Copy a file containing three columns of integers,
each column being 1 characters wide. The user
supplies the name of the input file. The output
file is called OUT.TEXT. The first and second
columns are swapped.
C
PROGRAM SWITCH
CHARACTER*23 FNAME
INTEGER FIRST, SECOND, THIRD
C
C
Prompt to the Standard Output by writing to
*
C
WRITE(*, 900)
900 FORMAT ( , I npu t file name
C
C
C
_ '\)
Read the file name from the Standard Input by reading
READ(*, 910)
910 FORMAT (A)
FNAME
C
C
C
Use unit 3 for input - any
uni~
but 0 will do.
OPEN(3, FILE=FNAME)
C
C
C
Page 78
Use unit 4 for output - any but 0 and 3 are ok.
OPEN(4, FILE='OUToTEXT', STATUS='NEW')
FORTRAN Reference Manual
*
Chapter 10
C
C
Input and Output
Read and write until end-of-file.
C
100 READ(3, 920, END=200) FIRST, SECOND, THIRD
WRITE(4, 920) SECOND, FIRST, THIRD
920 FORMAT(3I7)
GO TO 100
C
200 WRITE(., 910) 'Done Copying'
END
10.2.3
Less Common File Operations
The less common file structures are suitable for certain
classes of applications. In general, the appli9ation areas are
these:
Direct-access files are suitable for random access applications
such as maintaining a database.
Unformatted files are more efficient in input-output overhead
and in file space requirements. Unformatted access can be used
if data is to be writtep and read by FORTRAN on the same processor.
The combination of direct-access and unformatted files is ideal
for a database management facility to be accessed exclusively
through the FORTRAN system.
If data is to be transferred without any system interpretation
(especially if all 256 character combinations are needed), unformatted input output is necessary, since .TEXT files are limited
to the printable character set. A good example of unformatted
input-output usage is to control a device with a single-byte
binary interface.
In this situation, formatted input-output
would interpret certain characters (such as carriage-return) and
would change their meanings in certain ways.
Internal files are not input-output in the conventional sense
but they provide valuable character string operations and conversion via a standard mechanism.
Formatted direct-access files require special care.
FORTRAN
formatted files try to comply with the operating system's rules
for ASCII files. This allows standard system utilities such as
editors to be used on these files. The FORTRAN input-output system is able to enforce these rules for sequential files, but cannot always enforce them for direct-access files. Direct-access
files are not necessarily valid ASCII files and may not work
properly with other system utilities, since any unwritten records
leave "holes" which do not follow the operating system's rules
for ASCII files.
Direct-access files do, of course, follow
FORTRAN Reference Manual
Page 79
Input and Output
Chapter 10
FORTRAN'S input-output rules.
A file opened in FORTRAN is either "old" or "new". There is no
notion of "open for reading" as distinct from "open for writing".
Therefore ·old" (existing) files may be opened and written on,
with the effect of changing an existing file. Similarly the same
file may be alternately read from and written to, providing that
no attempts are made to read beyond end-of-fi1e or to read
unwritten records in direct-access files. A write to a sequential file effectively deletes any records which existed beyond
the fresh record. Normally, when a device (such as a terminal)
is opened as a file, it makes no difference whether it is "new"
or "old". with disk files, opening "new" creates a new file. If
that file is closed with the "keep" option, or the program terminates before a CLOSE is performed on that file, a permanent
file is created with the name given when the file was opened. If
a file is closed using the "delete" option, the newly created
file is deleted. In either case, if a file existed previously
with the same name, the prior version is deleted. Opening a disk
file as "old" generates a run-time error if the file does not
exist1 if it does exist, any writes to that file change its contents.
10.2.4
Limitati,ons of FORTRAN Input Output System
This subsection discusses specific limitations
input-output system imposes.
10.2.4.1
that
FORTRAN's
Direct Files must be on Blocked Devices
The operating system underlying FORTRAN has two kinds
ices, namely blocked and sequential.
of
dev-
Sequential files can be considered a stream of characters with
no explicit positioning;
the only operations are reading and
writing. The Standard Input and Standard Output are examples of
sequential devices.
Blocked devices, such as disks, have the additional capability
to seek to a specific position. Blocked devices can be accessed
either sequentially or randomly and therefore can
support
direct-access files.
Since there is no notion of seeking to a
position on an unblocked file, FORTRAN prohibits direct access
to sequential devices.
10.2.4.2
No Character Compression in Direct Files
Direct-access formatted files must not
compression information.
Page 80
contain
any, character
FORTRAN Reference Manual
Chapter 10
10.2.4.3
Input and Output
BACKSPACE Only Applies to Files on Blocked Devices
BACKSPACE can only be performed on files associated with
blocked devices, it cannot be applied to sequential devices.
10.2.4.4
Length Limitations-on Formatted Records
Formatted records must not be greater than 512 characters
length, including the terminating carriage-return character.
10.2.4.5
in
BACKSPACE may not be used on Unformatted Sequential
Files
It is not possible to apply the BACKSPACE statement to an
unformatted sequential file,
because such a file contains no
indications of its record boundaries. In principle it is possible to place end-of-record marks in such a file format, but this
conflicts with the notion of an unformatted file as a "pure"
sequence of values.
Any structure imposed upon unformatted
sequential files would interfere with their most common application, which is direct control of external instruments.
10.2.4.6
Side Effects of Functions Used in Input Output Statements
During execution of any input-output statement, expression
evaluation can reference functions. Any function so referenced
must not execute any other input-output statement.
10.3
Elements of Input and Output Statements
This section describes the elements that comprise input and
output statements that FORTRAN supports. The next section covers
the individual input-output statements in detail.
FORTRAN input-output statements require certain arguments and
parameters
which specify sources and destinations of data
transfer, as well as other aspects of the operation.
10.3.1
The Unit Specifier 'u'
The unit seecifier indicates the unit to be used in
output operatlons. The format of a unit specifier is:
input-
[UNIT=] u
where 'u' is an external or internal unit identifier.
FORTRAN Reference Manual
Page 81
Input and Output
Chapter 10
The unit identifier,
input-output statement:
•
lUi,
can take one of these
refers to the Standard Input or
filese
forms
in
Standard
an
Output
positive integer expression
refers to an external file whose unit number is
that of the expression. Unit 0 is the same as *
- the Standard Input or Standard Output files.
name
the name of a character variable, character
array, character array element or character substring.
In all input-output operations, if the UNIT: keyword is omitted, the unit specifier, 'UI, must be the first item in the list
of specifiers.
10.3.2
The Format Specifier If I
The format specifier is used to designate format lists when
performing formatted input-output.
The form of the format
specifier is:
[FMT=] f
where 'f' is the format identifier, and can
forms in an input-output statement:
take
statement label refers to the FORMAT statement
statement label.
one
labeled
of
these
by
that
integer variable name
refers to the FORMAT label assigned to that
integer variable using the ASSIGN statement.
character expression
the current value of the character expression
the format specifier.
is
asterisk character '*'
the asterisk is
formatting.
used
to
specify
list-directed
If the FMT= keyword is omitted in a list of specifiers, the
format specifier, 'fe, must be the second item in the list and
the unit specifier, lUi, must be the first item in the list, with
Page 82
FORTRAN Reference Manual
Chapter 10
Input and Output
the UNIT= keyword omitted.
See Chapter 11 for a complete description of
and the elements it may contain.
10.3.3
the
format
list
The Record Number ern'
The record number is used in direct-access input-output only.
It specifies the number of the record to be read or written. The
format of the record number specifier is:
REC=rn
where ern' is the record number.
positive integer.
10.3.4
The record
number
must
be
a
The End of File Exit Specifier
The end-of-file exit specifier designates a statement label at
which execution is to start when an end-of-file condition occurs
while reading from a file. The format is:
ENO=s
where 's' is a statement label in the same program
READ statement.
10.3.5
unit
as
the
The Error Exit Specifier
The error exit specifier designates a statement label at which
execution is to start when any errors occur during execution of
an input-output statement. The format of the error exit specifier is:
ERR=s
where IS' is a statement label in the same program
input-output statement.
10.3.6
unit
as
the
The Input Output Status specifier 'ios'
The input-output status specifier specifies an integer variable
into which an input-output
status value is placed when the
current input-output statement completes.
The. format of the
specifier is:
IOSTAT=ios
where 'ios' is the name of an integer variable or
FORTRAN Reference Manual
integer
array
Page 83
Input and Output
Chapter 10
element.
The integer variable specified by 'ios' must be an
INTEGER (or INTEGER*4). When an input-output statement containing this specifier terminates, 'ios' becomes defined with the
following values:
zero (0)
indicates that the input-output operation completed normally.
There were no errors and endof-file was not encountered,
negative value
indicates that an end-of-file
during a READ statement,
positive value
indicates that an error condition occurred during
the input-output statement.
10.3.7
was
encountered
The Input-Output List 'iolist'
The input-output list - 'iolist' - specifies the objects whose
values are transferred by READ and WRITE statements. An 'iolist'
can be empty. If an Violist' is present, it is a list of elements separated by commas. Items in the list consist of:
• Input and output objects,
• Implied DO lists.
These two forms of 'iolist' are discussed in the two paragraphs
below.
10.3.7.1
Input and Output Objects
An input or output object can be specified in the 'iolist' of a
READ, WRITE or PRINT statement and is one of these forms:
Variable name,
Array element name,
Character substring name,
Array name. This form is a means of specifying all the elements of the array, in the order in which they are stored
internally,
On output only, any other expression except a character
expression which concatenates an assumed length character
string.
Example of Input and Output
WRITE (0, 100) 'Results are:
Page 84
widget, blivet(j, 4)
FORTRAN Reference Manual
Chapter 10
Input and Output
100 FORMAT (A, IS, FIO.S)
10.3.7.2
Implied DO Lists
Implied-DO lists can be specified as items in the 'iolist'
READ, WRITE and PRINT statements and are of the form:
of
(dlist , i - e l , e2 [, e3l)
where the 'dlist' is an 'iolist' as defined previously ••
In a READ statement, the DO variable, 'it,
(or an associated
object) must not appear as an input list item in the embedded
'dlist', but can be read in the same READ statement outside of
the implied-DO list.
The embedded 'dlist' is effectively repeated for each iteration
of Ii' with appropriate substitution of values for the DO control
variable 'it.
Example of
WRITE (0, ISO)
ISO FORMAT (1017)
!!! Implied DO List
(jinx(i), i = 1, 100)
In the example above, the variable Ii'
iterates through 100
elements of the array 'jinx'. The format specified in the FORMAT
statement causes the results to be placed 10 per line on the output.
10.4
The Specific Input and Output Statements
The following input-output statements are supported in the FORTRAN system.
The possible form of each statement is specified
first, with an explanation of the meanings of the following
forms.
Certain items are specified as required if they must
appear in the statement and are specified as optional if they
need not appear in the statement. Optional items normally result
in a default which is indicated if the item is omitted.
10.4.1
OPEN Statement
The format of the OPEN statement is:
OPEN(open list)
where 'open list' is a list of specifiers as described below.
The argument list,
'open list' must contain one unit specifier
FORTRAN Reference Manual
Page 85
Input and Output
Chapter 10
and may contain one of each of the other specifiers listed.
(UNIT=]U
'u' is the unit specifier. It must not be an
internal unit specifier. If the UNIT= keyword is·
omitted, the 'u' argument must be first in the
argument list to OPEN.
IOSTAT=ios
'ios' is an input-output status specifier as
defined above in "Elements of Input and Output
Statements".
ERR=s
is an error exit specifier as defined above
"Elements of Input and Output Statements".
FILE=fname
The file name 'fname' is a character expression.
If the 'fname' argument is omitted, OPEN opens a
an anonymous file with a status of 'SCRATCH' (see
below) •
This file is automatically deleted when
CLOSE'd or upon program termination.
STATUS=sta
'stat is the status of the file when opened.
'stat
is a character expression whose ~alue must
be one of 'OLD', 'NEW', 'SCRATCH' or 'UNKNOWN'.
'OLD'
is the default for reading or writing
existing files.
'NEW' may be used for writing
new files.
If the 'OLD' or 'NEW' keywords are
used, the FILE= argument must be supplied.
If
'SCRATCH'
is specified, the file is deleted when
a CLOSE is performed on that file. If 'UNKNOWN'
is specified, it is treated the same as if 'OLD'
had been specified.
ACCESS=acc
specifies the access mode for this file.
'acc'
is a character expression whose value must be
either 'SEQUENTIAL' or 'DIRECT'.
'SEQUENTIAL' is
the default.
FORM=fm
specifies whether the file is formatted or unformatted.
'fm'
is a character expression whose
value must be either 'FORMATTED' or 'UNFORMATTED'.
'FORMATTED' is the default.
RECL=rl
The record length 'rl' is an integer expression.
This argument to OPEN is for DIRECT-access files
only and is required for that file type.
BLANK=blnk
'blnk' controls the default treatment 6f blanks
(spaces)
in
formatted reads, which can be
altered, in a particular formatted READ, by a 8N
or 8Z edit-descriptor in a format specification.
'blnk' is a character expression whose value must
be either 'NULL' or 'ZERO',
The default is
'NULL'. If 'NULL' is specified, all spaces are
Page 86
in
FORTRAN Reference Manual
Chapter 10
Input and Output
ignored in numeric input fields. If 'ZERO' is
specified, all spaces other than leading spaces
are treated as zeros.
BUFFERED='buffered'
The BUFFERED option selects buffered or unbuffered input or output on a unit in the system.
If the option is 'BUFFERED', buffered input output is selected. If the option is 'UNBUFFERED',
unbuffered input output is selected.
Note that
the operating system might override the option.
Some operating systems
show
a
substantial
improvement
in throughput if the 'BUFFERED'
option is selected. All files are opened with
the 'BUFFERED' option by default, except for the
* unit and unit I (the standard error).
The OPEN statement binds a unit number with an external device,
or file on an external device, by specifying its file na~e. If
the file is to be a direct access file, the REeL=rl option specifies the length of records in that file. If the unit specified
in an OPEN statement is already open, it is closed before being
bound to a file.
10.4.2
CLOSE Statement
The format of the CLOSE statement is:
CLOSE(close list)
where 'close list' is a list of specifiers.
'close list' must
contain an external unit specifier and at most one of any of the
other specifiers, which are as follows:
[UNIT=]u·
'u' is the unit number of an external unit.
If
the UNIT= keyword is omitted, the unit specifier,
lUI, must be the first specifier in the list.
IOSTAT=ios
'ios' is an input-output status specifier as
defined above in "Elements of Input and Output
Statements".
ERR=s
's' is an error exit specifier as defined above
in "Elements of Input and Output Statements".
STATUS=sta
'stat is an optional argument which dictates the
disposition of the file after it is CLOSE'd.
'stat is a character expression whose value must
be either 'KEEP' or 'DELETE'.
FORTRAN Reference Manual
Page 87
Input and Output
Chapter 10
CLOSE disconnects the specified unit and prevents input-output
from being directed to that unit (unless the same unit number is
re-opened, possibly bound to a different file or device).
Files
are discarded if STATUS='DELETE' is specified. Normal te~mina
tion of a FORTRAN program automatically closes all open files as
if a CLOSE with STATUSa'KEEP' had been specified. STATUS-'KEEP'
should not be specified for a file which was opened as 'SCRATCH'.
10.4.3
READ, WRITE and PRINT Statements
The READ statement transfers data into storage. The WRITE and
PRINT statements transfer data from storage. All three statements have similar forms, as defined here:
READ
(control list) [iolist]
READ
f
WRITE
(control list) [iolist]
PRINT
f [, iolist]
[, iolist]
where 'f' and 'iolist' are a format identifier and an inputoutput list as previously described in "Elements of Input and
Output Statements". Note that the PRINT statement has no connection with "printing" on the system printer device (even if such a
device exists).
The 'control list' is a list whose elements may be any
following:
of
the
[UNIT=] u
is a unit specifier. If the optional UNIT= keyword is omitted from the specifier, the unit
specifier, 'u', must be the first item in the
list.
[FMT=] f
is a format specifier. If the optional FMT= keyword is omitted from this specifier, the format
specifier, Iff, must be the second item in the
list and the first item must then be the unit
specifier, 'u', without the UNIT= keyword.
REC=rn
a record nu~ber specifier. If this specifier
included 1n the list, the statement is a
direct-access data transfer statement, otherwise
the
statement
is a sequential access data
transfer.
~s
1S
IOSTAT=ios
Page 88
'ios' is an input-output status specifier as
defined above in "Elements of Input and Output
Statements".
FORTRAN Reference Manual
Chapter 10
Input and Output
's' is an error exit specifier as defined above
in "Elements of Input and Output Statements".
is an end-of-file exit specifier as defined above
in "Elements of Input and Output Statements·.
This specifier is only applicable to the READ
statement.
If the 'control list' contains a format specifier, the statement is a formatted input-output statement, otherwise it is an
unformatted input-output statement.
If the format identifier is an asterisk character,
'*', the
statement is a list-directed input-output statement. In this
case, the record specifier must not appear in the 'control list'.
List-directed input-output must not be done on an internal
file.
If the unit specifier designates an internal file, the
'control list' must not contain a record specifier.
The end-of-file specifier must not appear in the 'control list'
for a WRITE or PRINT statement.
The 'control list' must not contain both a record specifier and
an end-of-file specifier.
10.4.4
File Positioning Statements
FORTRAN supplies three statements which position
citly:
files
BACKSPACE
to backspace a file by one record.
ENDFILE
to write an endfile record on a file.
REWIND
to position or re-position a file
record.
Each of the
forms:
file
positioning
statements
has
at
two
its
expli-
first
different
BACKSPACE u
BACKSPACE (alist)
ENDFILE
ENDFILE
(alist)
REWIND
REWIND
(alist)
u
u
In each case, the 'u' is a unit number.
The
parenthesized list of specifiers of the form:
FORTRAN Reference Manual
'alist'
is
a
Page 89
Input and Output
Chapter 10
[UNIT s 1u
'u' is a unit specifier for the unit. This argument is required. If the UNIT= keyword is omitted, the 'u' argument must be first in the list.
IOSTAT=ios
'ios' is an integer variable which is assigned
the status of the specific input-output statement. 'ios' is defined previously in "Elements
of Input and Output Statements".
ERR=s
IS' is a statement
label -to which control is
passed if there are any errors in the inputoutput statement. The statement label '9' must
appear in the program unit that contains the
input-output statement. 'Sf is defined previously in "Elements of Input and Output Statements".
10.4.4.1
BACKSPACE Statement - Backspace a File
BACKSPACE positions the file connected to the specified unit,
before the preceding record. If there is no preceding record,
the file position is not changed. If the preceding record is the
endfil~ record, the file is positioned before the endfile record.
A non-connected unit may be BACKSPACE'd without any error.
In
other words a BACKSPACE command issued against a non-connected
unit has no effect.
BACKSPACE can only be issued on units which are
formatted. BACKSPACE must not be applied to a
unformatted file.
10.4.4.2
sequentialsequential-
ENDFILE Statement - Write an Endfile Record
ENDFILE "writes" an endfile record as the next record on the
file connected to the specified unit 'u'. The file is then positioned after the endfile record, so that further sequential data
transfers are prohibited until either a BACKSPACE or a REWIND
statement is executed.
10.4.4.3
REWIND Statement - Rewind a File
The REWIND statement positions the file associated with
unit 'u' to its initial point. The unit must be sequential.
the
A non-connected unit may be rewound without any error.
In
other words a REWIND command issued against a non-connected unit
has no effect.
Page 90
FORTRAN Reference Manual
Input and Output
Chapter 10
10.4.5
INQUIRE Statement - Obtain File Properties
The INQUIRE statement is used to obtain information about the
properties of a particular named file or about the connection to
a particular unit. There are two forms of INQUIRE, namely,
inquire by file and inquire by unit.
The INQUIRE statement can be executed before, while or after a
file is connected to a unit. Any values assigned as a result of
INQUIRE are values which are current at the tim. the INQUIRE is
executed.
The two forms of the INQUIRE atatement are:
INQUIRE (iflist)
2! INQUIRE (iulist)
where 'iflist', used for the inquire by file form, is a list of
specifiers containing exactly one file specifier and 'iulist',
used for the inquire by unit form, is a list of .pecifiers containing exactly one unit specifier. Both for •• may then contain
other specifiers as described below.
The format of the file
specifier is:
FILE-fin
where 'fin' is a character expression which, when trailing spaces
are removed, is the name of the file which is the subject of the
inquiry. The named file does not need to exist or to be connected to a unit.
The form of the unit specifier is as· described previously in
"Elements of Input Output Statements". Tne unit specified does
not have to exist, nor be connected to a file. If the unit is
connected to a file, the inquiry is being made about the connection and the,jile connected to it.
The remainder of the list in the INQUIRE statement is the
inquiry specifiers.
This is a list
of one or more inquiry
specifiers as described below. There may be only one of each
inquiry specifier.
Furthermore, each variable name may only
appear once in the list of specifiers, in other words, the same
variable name must not be given to more than one specifier.
The inquiry specifiers are summarized
more detail below.
FORTRAN Reference Manual
here
and
described
in
Page 91
Input and Output
Chapter 10
+-------------------+
IOSTAT = ios
ERR =- s
EXIST = ex
OPENED = od
NUMBER =- num
NAMED =- nmd
NAME = fn
ACCESS =- acc
SEQUENTIAL =- seq
DIRECT =- dir
FORM =- fm
FORMATTED = fmt
UNFORMATTED = unf
RECL =- rcl
NEXTREC = nr
BLANK = blnk
+-------------------+
In every case where an integer variable is specified, it must
be an INTEGER*4 (which is the same as INTEGER). In all cases
where a logical variable is specified, it must be a LOGICAL*4
(which is the same as LOGICAL). The specifiers given here are as
defined previously in "Elements of Input and Output Statements".
The meaning of each of the inquiry specifiers is as follows:
IOSTAT=ios
is an input-output status specifier.
The variable 'ios' is set to zero (0) by the INQUIRE
statement.
ERR=s
IS' is a statement
label to which control is
passed if an error occurs. The INQUIRE statement
never causes any error conditions.
EXIST=ex
'ex' is a LOGICAL*4 variable. If a file with the
specified name exists (in the FORTRAN milieu)
'ex' is set to .TRUE., else 'ex' is set to
.FALSE.
For an inquire by unit, 'ex' is set to
true if and only if the unit actually exists.
'ex' always gets defined by the INQUIRE statement.
OPENED=od
cod' is a LOGICAL*4 variable. If the specified
file is opened (connected to a unit), lode is set
to .TRUE., else 'ode 1S set to .FALSE. 'ode
always gets defined by the INQUIRE statement.
NUMBER=num
'nurn' is an INTEGER*4 variable. 'num' is set to
the
the external unit number for the unit
currently connected to the file. If the file is
not connected to a unit, 'num' is undefined.
Page 92
FORTRAN Reference Manual
Input and Output
Chapter 10
NAMEO=nmd
If the file has a name, the LOGICAL*4 variable
'nmd' is set to .TRUE., else 'nmd' is set to
.FALSE.
NAME=fn
'fn' is a character variable that is set to the
name of the file if the file has a name. If the
file does not have a name, 'nmd' is undefined.
ACCESS=acc
'acc' is a character variable that is assigned
the string 'SEQUENTIAL' if the file is connected
for sequential access, and is assigned the value
'DIRECT' if the file is connected for direct
access. If the file is not connected to a unit,
the value of 'acc' is undefined.
SEQUENTIAL=seq
'seq' is a character variable which is assigned
the value 'YES' if this file can be connected for
sequential access. 'seq' receives the value 'NO'
if the file cannot be connected for sequential
access. If FORTRAN cannot determine what access
methods are allowed for the file, 'seq' receives
the value 'UNKNOWN'.
DIRECT=dir
'dirt is a character variable which is assigned
the value 'YES' if this file can be connected for
direct access.
'dirt receives the value 'NO' if
the file cannot be connected for direct access.
If FORTRAN cannot determine what access methods
are allowed for the file, 'dirt receives the
value 'UNKNOWN'.
FORM=fm
'fm' is a character variable which is assigned
the value 'FORMATTEO' if the file is connected
for formatted input-output, and is assigned the
value 'UNFORMATTED' if the file is connected for
unformatted input-output. If the file is not
connected, 'fm' is undefined.
FORMATTEO=fmt
'fmt' is a character variable which is assigned
the value 'YES' if this file can be connected for
formatted input-output. 'fmt' receives the value
'NO' if the file cannot be connected for formatted input-output. If FORTRAN cannot determine if
the file can be connected for formatted inputoutput, 'fmt' is assigned the value 'UNKNOWN'.
UNFORMATTED=unf 'fmt' is a character variable which is assigned
the value 'YES' if this file can be connected for
unformatted input-output.
'unf' receives the
value 'NO' if the file cannot be connected for
unformatted input-output. If
FORTRAN
cannot
determine if the file can be connected for unformatted input-output, 'unf' is assigned the value
FORTRAN Reference Manual
Page 93
Chapter 10
Input and Output
'UNKNOWN' •
RECL-rcl
'rcl' is an INTEGER*4 variable which is assigned
the record length of the file connected for
direct access. If the connection is not for
direct-access or the the file is not connected at
all, 'rcl' is undefined.
NEXTREC=nr
'nr' is an INTEGER*4 variable which is assigned
the number of the next record to be read or written on a direct-access file. If the file is connected but no data transfer has been done, Inri
is assigned the value 1. If the file is not a
direct-access file or if the position cannot be
determined (possibly because of
a
previous
error), Inri is undefined.
BLANK=blnk
'blnk' is a character variable which is assigned
the value 'NULL' if the BN edit-descriptor is the
default for blank control and is assigned the
value 'ZERO' if the BZ edit-descriptor is the
default for blank control (see the OPEN statement
and the chapter on "Format Specifications"). If
the file is not connected or if the file is connected but not for formatted input-output, 'blnk'
is undefined.
The INQUIRE statement never returns an error condition.
The
specifiers 'ex'
(exists) and 'ad' (opened) always get defined.
When a specifier is said to be "undefined" in the descriptions
above, it means that certain of the specifiers are meaningless in
certain contexts. For example, if the access method for a unit
is SEQUENTI~L, the 'rcl' (record length) and Inri (next record)
specifiers have no meaning, and are thus said to be undefined (in
the context that a program should not be examining those specifiers in such a case).
For example, to see if a file named
statement:
'xyzzy'
exists,
use
the
INQUIRE(FILE='xyzzy', EXIST=L)
The value of the LOGICAL*4 variable L is set to .TRUE. or .FALSE.
depending upon the existence of that file. To see if UNIT 10 is
currently OPEN, use:
INQUIRE (UNIT=lO, OPENED=L)
Page 94
FORTRAN Reference Manual
Chapter 10
10.5
Input and Output
List Directed Input and Output
List-directed input-output serves to process formatted records
without the use of a FORMAT statement. Values in the records are
in a "free-form". List-directed input-output is convenient for
those cases where the precise layout of the data is not important.
10.5.1
List Directed READ
The elements in the list-directed READ statement are the same
as those for the READ statement previously described, with the
exception that an asterisk character "*" is supplied as the format specifier.
Data is read into storage locations as specified in the 'iolist'.
Input data consists of a number of values and value
separators. The next paragraph describes value separators--and
the following paragraph describes values.
10.5.1.1
List Directed Value Separators
Value separators are used to delimit values in a list-directed
input list. A value separator is one of the following:
a comma, with optional spaces on either side.
a slash, with optional spaces on each side.
one or more contiguous spaces between constants or
last constant in the list.
after
the
end-of-record, which appears as a space between two constants.
A comma is used to separate values.
cate a null value.
Two commas in a row
A slash has the effect that the remalnlng items in
input iist are discarded and null values substituted.
A value separator adjacent to
sidered to be a null value.
10.5.1.2
an
end-of-record
is
indiin
not
the
con-
List Directed Input Values
A value is a constant, a null
forms:
FORTRAN Reference Manual
(empty)
value
or
one
of
the
Page 95
Chapter 10
Input and Output
r*c
r*
where 'r' is a repeat factor which must be an unsigned non-zero
integer constant.
The form of 'r*c' represents 'r' successive
appearances of the constant 'ct. The form 'r*' is th~ same as
'r' successive null values. Neither of these forms may contain
embedded spaces except where 'c' is a character constant.
Individual values in list-directed input are in the forms
described below.
Values may not have embedded spaces, except
character constants, where a space stands for itself.
The
acceptable forms of input values are:
Null
A null value is indicated by the presence of two
contiguous value separators, no characters or
spaces preceding the first value separator in an
input list, or by an 'r·' form. A null value
cannot be used as the real or imaginary part of a
complex constant but can be used as an entire
complex constant. List items which are null have
no effect on the definition status of variables
in the corresponding 'iolist'. Variables in the
'iolist' which are already defined stay defined
and variables which are not defined stay undefined.
Integers
Must be the same form as for integer constants.
Real Numbers
Any valid format for a FORTRAN real number.
Furthermore the decimal point can be omitted from
a real number, in which case the number is
assumed not to have a fractional part.
Complex Numbers A complex value is represented by an ordered pair
of real numbers in the format described above for
real numbers. The pair of reals are separated
from each other by a comma, and enclosed in
parentheses. There can be spaces surrounding the
numbers.
An end of record can appear before or
after the comma in a complex number.
Character string values
A string of characters enclosed in apostrophes.
Embedded spaces are significant and are part of
the constant. An apostrophe is represented by
two juxtaposed apostrophes in the string. Character constants can span record boundaries.
A
record boundary in a character constant does not
generate any gratuitous characters in the value.
Character constants can be continued over as many
Page 96
FORTRAN Reference Manual
Chapter 10
Input and Output
records as needed. Note that unterminated character constants can lead to disastrous results.
If the character constant as read is longer than
the length of the corresponding 'iolist' item, it
is truncated to fit.
If the character constant
is shorter than the corresponding 'iolist' item,
it is placed left justified in the variable and
the remaining positions in the variable are
filled with spaces.
Logical values
10.5.2
Logical values are represented by a T (.TRUE.) or
an F (.FALSE.). An optional period character "."
may appear before the T or F character. The T or
F can be followed by optional characters, but
these characters must not be value separators
slashes or commas.
List Directed WRITE and PRINT
The list-directed WRITE and PRINT statements are similar to the
formatted WRITE and PRINT statements described previously, with
the exception that the FORMAT specifier is a *.
Data is transferred from the variables specified by the 'iolist'
to the specified unit. In general, values are written on
the output device in a manner
consistent with list-directed
input, but there are exceptions, the most notable being that
character string data written out by a list-directed WRITE or
PRINT statement cannot be re-read by a list directed input statement. FORTRAN starts new records when necessary. Values in the
output are separated by three spaces, except that character
values are not preceded or followed by any spaces.
FORTRAN
never generates slashes or null values on list-directed output.
The forms of the different data types are as follows:
Integer values
are generated as for an Iv edit-descriptor.
'w'
is the minimum number of characters required to
print the integer value.
Real and Double Precision
are generated with the effect of either an F
edit-descriptor or an E edit-descriptor, depending on the magnitude of the numbers involved.
The
specific
edit-descriptors
used are as
described in the table below.
FORTRAN Reference Manual
Page 97
Input and Output
Chap~er
10
+------------------+------------------------------+
Range of Number I
Edit-Descriptor Used
I
I
Real
I Double Precision
+------------------+-----------+------------------+
11.0 < val < 10.0 I
OPP9.6 I
OPP17.l4
I
+------------------+-----------+------------------+
I ~:~ : io~oor
I
lPBl3.6B2
I
IPB22.14E3
I
+------------------+-----------+------------------+
I Not A Number
1'1111111' I
'1111111'
I
+------------------+-----------+------------------+
Plus Infinity
I '+++++++' I '+++++++' I
I Minus Infinity
,-------,
,-------,
+------------------+-----------+------------------+
Complex numbers are generated as a pair of real numbers enclosed
in parentheses with a comma separating the real
and imaginary parts.
Character string values
are generated as strings of characters. However,
character constants are not surrounded by apostrophes, meaning that they cannot be read back in
using list-directed input.
Logical values
Page 98
are generated as T for .TRUE. and P for .FALSE.
FORTRAN Reference Manual
Format Specifications
Chapter 11
Chapter 11 - Format Specifications
This chapter describes formatted input-output and the FORMAT
statements available from FORTRAN. The reader is assumed to be
familiar with the FORTRAN file system, units, records, access
methods and input-output statements as described in the previous
chapter.
A format specification is used in conjunction with formatted
input-output to supply information which directs the editing or
conversion between the internal representations of the machine
and the representations of character strings in a file or character data item.
11.1
FORMAT Specifications and the FORMAT Statement
If a READ, WRITE or PRINT statement specifies a format, it is
considered to be a formatted input-output statement as opposed to
an unformatted input-output statement.
To reiterate what was
described in the previous chapter, a format can be specified by
any of the following forms:
giving the label of a FORMAT statement,
the name of an integer variable containing the label of a FORMAT statement,
a character variable, character array element or character
expression which ha~ the correct form of a format specifier,
an asterisk character, "*",
output.
indicating
list-directed
input-
The following examples illustrate valid ways to specify a
mat:
FORTRAN Reference Manual
for-
Page 99
Format Specifications
*
*
*
*
*
*
*
Chapter 11
Example the first
reference to a FORMAT statement
WRITE(*, 990) IGOR, JOHANN, KLUTZ
990 FORMAT(2I5, Il)
Example the second •••••
assigned FORMAT label to integer
ASSIGN 880 TO MORFAT ,
880 FORMAT(2I5, Il)
WRITE(*, MORFAT) IVAN, JANUS, KLONG
*
*
*
*
*
*
*
*
*
*
*
*
Example the third
using a character variable
CHARACTER*9 FORCH
FORCH ~ '(215, Il)'
WRITE(*, FORCH) IVOR, JACKO, KELP
Example the fourth ••••.
using a character expression
CHARACTER*7 CHAREX
DATA CHAREX 1'215, Il'l
WRITE(*, '(' II CHAREX II I)') IRENE, JANET, KLARG
Example the fifth •••••
List-directed write
WRITE(*, *) INEZ, JACKIE, KRON
The format specification itself must begin with a
parenthesis "C" and must end with a right parenthesis H)".
acters after a matching right parenthesis are ignored.
left
Char-
A FORMAT statement must have a label. Like all non-executable
statements, a FORMAT statement may not be the target of a branching statement.
Between the "C" and ")" characters, the format specifications
appear. The format specifications are a list of items, separated
by commas. Each of the items ~n the format list is one of:
[rl ed
repeatable edit-descriptors.
ned
non-repeatable edit-descriptors.
[r)
fs
a nested format specification. At most ten levels of nested parentheses are allowed within the
outermost level.
where Ir' is an optional,
Page 100
non-zero,
unsigned
integer
constant
FORTRAN Reference Manual
Format Specifications
Chapter 11
called a repeat specification.
The comma separating two list
items may be omitted if the resulting format is unambiguous, such
as after a P edit-descriptor or before or after a / editdescriptor.
The repeatable
below) are:
edit-descriptors
Xv and Iv ••
Fv.d
Bv.d and Bv.dEe
OW.d
Gv.d and Gv.dEe
Lv
A and Av
for
for
for
for
for
for
for
(described
in
more
detail
Integer editing
Real editing
Real editing
Real editing
Real editing
Logical editing
Character-editing
The wand e are unsigned non-zero integer
and ~ are unsigned integer constants.
constants.
The non-repeatable edit-descriptors (also described
below) are:
The
in
2
detail
'xxxxxxx'
apostrophe editing
nBxxxxxxxxxxxx Hollerith editing
Te, TLe and TRe Tabbin~ to column
nX
insertlng spaces
/
starting a new record
\ or $
inhibits starting new record
conditionally terminates format list
S, SS and SP
optional Sign control
Scale factor editing
kP
Blank control
BN and BZ
x is one of the characters from the character set that FORTRAN
supports. nand c are unsigned, non zero integer constants.
k
is an optionally sIgned integer constant.
Note that the FORTRAN compiler performs as much checking of
FORMAT statements as is possible at compilation time. There can
however, be incorrect FORMAT statements which the compiler cannot
detect, and such incorrect statements will not manifest themselves until the program is actually run.
11.2
Interaction Between Format Specifications and I/O List
Before going into the full details of how the various editdescriptors control the editing, it is necessary to describe how
the format specification interacts with the input-output list
('iolist') in a given READ, WRITE or PRINT statement.
If an 'iolist' contains any
edit-descriptor must appear
FORTRAN Reference Manual
items, at least one repeatable
in the format specification. In
Page 101
Format Specifications
Chapter 11
particular, the empty edit specification "()" may only be used if
there are no items in the 'iolist', in which case the only action
the input-output statement performs is the implicit record skipping action associated with formatted input-output. Each item in
the 'iolist' is associated in turn, with a repeatable editdescriptor during the execution of the input-output statement.
In contrast, the remaining non-repeatable format-control items
interact directly with the record and are not associated with
items in the 'iolist'.
Items in a format specification are interpreted from left to
right. Repeatable edit-des6riptors act as if they were present
'r' times - if 'r' is omitted it is treated as a repeat factor of
1. Similarly, a nested format specification is treated as if its
list of items appears Ir' times.
Format specifications are interpreted at execution time.
The
term "format controller" is used here to describe the entity that
interprets the list. The formatted input-output process proceeds
as follows:
The format controller scans the format items in the order noted
above. When a repeatable edit-descriptor is found, either:
a corresponding item appears in the 'iolist'.
In this case
the item and the edit-descriptor are associated and input or
output of that item proceeds under format control of the
edit-descriptor, or:
the format controller terminates the input-output process.
If the format controller encounters a colon edit-descriptor,
":" in the format list and there are no further items in the
'iolist', the format controller terminates input-output.
If the format controller encounters the matching final ")" of
the format specification and there are no further items in the
'iolist', the format controller terminates input-output.
If
there are further items in the 'iolist', the file is positioned
at the beginning of the next record and the format controller
continues by re-scanning the format, starting at the beginning of
the format specification terminated by the last preceding right
parenthesis.
If there is no such preceding right parenthesis,
the format controller re-scans the format from the beginning.
Within the portion of the re-scanned format, there must be at
least one repeatable edit-descriptor. Should the re-scan of the
format specification begin with a repeated, nested, format
specification, the repeat factor indicates the number of times to
repeat that nested format specification. The re-scan does not
change the previously set scale-factor, BN or BZ blank control or
S, SP or SS sign control.
Page 102
FORTRAN Reference Manual
Chapter 11
Format Specifications
When the format controller terminates, the remaining characters
of an input record are skipped or an end-of-record is written on
output, except as noted under the \ or $ edit-descriptors.
11.3
Edit Descriptors
Edit-descriptors are used to specify a field of a record.
A
detailed description of the various edit-descriptors follows.
The repeatable edit-descriptors appear first, followed by the
non-repeatable edit-descriptors.
11.3.1
Repeatable Edit Descriptors
Repeatable edit-descriptors are associated with items from an
'io1ist l •
Repeatable edit-descriptors are concerned with the
editing of numeric, logical, and character data items. These are
described in the paragraphs to follow.
11.3.1.1
Numeric Editing
The I, S, F and G edit-descriptors are used for formatting
integer, real
and complex data. The following general rules
apply to all of those edit-descriptors:
On input, leading spaces are not significant.
Other spaces
are interpreted differently depending on the BN or BZ flag in
effect, but all blank fields are always treated as "the value
zero (0). Plus signs are optional.
On input, with E and F editing, an explicit decimal point
appearing in the input field overrides position of the decimal
point as specified in the edit-descriptor.
On output, generated characters are right
field with leading spaces if required.
justified
in
the
On output, if the number of characters produced exceeds the
field width, or the exponent exceeds its specified width, the
entire field is filled with asterisks, but also see the next
list item, concerning infinite and indeterminate values.
On output, a value of plus infinity fills the field with plus
signs, a value of minus infinity fills the field with minus
signs, and an indeterminate value fills the field with question marks.
.
Editing of complex numbers is controlled by two E, F or G
edit-descriptors in succession, each of which controls the
editing of "ha1f
of the complex value.
The two editdescriptors for a given complex number do not have to be the
ft
FORTRAN Reference Manual
Page 103
Chapter 11
Format Specifications
same.
11.3.1.2
I - Integer Editing
The edit-descriptor Iv must be associated with an 'iolist' item
which is of type INTEGER (which is the same as INTEGER*4),
INTEGER*l or INTEGER*2. The field width is 'we characters long.
On input, an optional sign may appear in the field.
The Iv •• edit-descriptor specifies the field width 'we as
above, but the em' part specifies a minimum field width for the
integer value.
On input, the 'me specification has no effect.
On output, if the converted integer is shorter than 'm' characters, leading zeros are placed in the field. If em' is zero and
the integer value to be formatted is also zero, the output field
consists of 'w' spaces, regardless of any sign control in effect
(see later). 'm' must not be greater than 'W'g
11.3.1.3
F - Real Editing
The edit-descriptor Fv.d must be associated with· an 'iolist'
item which is of type real, double-precision or one half of a
complex number. The width of the field is 'w' characters.
The
fractional part is 'de digits long.
The input field begins with an optional sign, followed by a
string of digits optionally containing a decimal point. If the
decimal point is present, it overrides the 'de specified in the
edit-descriptor. If the decimal point is not present, the rightmost 'de digits of the string are interpreted as following the
decimal point - leading spaces are converted to zeros if necessary. The number may be followed by an optional exponent which
is either of the forms:
plus or minus followed by an integer, or
E or D, followed by zero or more spaces, followed by an
optional sign, followed by an integer. Band D are treated
identically.
The output field occupies 'w' digits, Ide of which follow the
decimal point. The output value is controlled both by the 'iolist' item and by the current scale-factor (see the paragraph
later,
on "Scale Factor Editing", in the discussion on nonrepeatable edit-descriptors). The output value is rounded, not
truncated.
Page 104
FORTRAN Reference Manual
Chapter 11
11.3.1.4
Format Specifications
E and D - Real Editing
The B or D edit-descriptors control formatting of real elements.
These edit-descriptors must be associated with an 'iolist' item which is of type real, double-precision or one half of
a complex number.
An a or D edit-descriptor takes one of the forms aw.d, Dw.d or
Bw.dEe.
The forms Ew.d and Dw.d have identical editing effects.
In each case, the field width is 'w' characters.
The 'e' has no effect on input.
The input field for an a
edit-descriptor is identical to that described by an P edi~
descriptor which has the same 'w' and 'd' fields.
The form of the output field depends on the scale-factor
(set
by the P edit-descriptor) in effect. For a scale-factor of 0,
the output field contains a minus sign (if needed), followed by a
decimal point, followed by a string of digits, followed by an
exponent field for an exponent 'exp' of one of the following
forms:
Ew.d -99
<= 'exp'
<= 99
H, followed by plus or minus,
two digit exponent.
followed
by
the
Ew.d -999 <= 'exp' <= 999
plus or minus, followed by three digit exponent.
Ew.dEe -lOe- l <= 'exp' <= 10e- l
a, followed by plus or minus, followed by 'e'
digits of exponent with possible leading zeros.
The form Bw.d must not be used if the
exponent to be printed exceeds 999.
absolute
value
of
the
The scale-factor controls decimal normalization of the printed
'E' field. If the scale-factor 'k'
is in the range -d < k <= 0,
the output field contains exactly 'd'-'k' leading zeros after the
decimal point and 'd'+'k' significant digits after that. If 0 <
k < d+2, the output field contains exactly 'k' significant digits
before the decimal point and 'd'-'k'-l places after the decimal
point. Other values of 'k' are errors.
11.3.1.5
G - Real Editing
The G edit-descriptor
descriptors except that
"adaptable" output format
FORTRAN Reference Manual
is similar to the E and P editthe G edit-descriptor provides for an
depending on the magnitude of the
Page 105
Chapter 11
Format Specifications
number being converted - it gives the user a flexible choice of
output formats without the need to pre-check on the size of the
numbers ahead of time.
The G edit-descriptor must be associated with an 'iolist' item
which is of type real, double-precision or one half of a complex
number.
The G edit-descriptor takes one of the forms Gw.d or Gw.dBe.
In either case, the final output field width is 'Wi characters.
The number of digits after the decimal point is 'd' digits unless
a scale-factor greater than 1 is in effect.
On input, G editing acts the same as F editing (see above).
On output, the format of the converted number is dependent on
its magnitude.
If N is the number to be converted, the table
below describes the action of the G edit-descriptor:
+---------------------------+--------------------------------+
I
Magnitude of N
I
Equivalent Conversion
I
+---------------------------+--------------------------------+
N < 0.1 or
Gw.d is the same as kPBw.d
N > 10**d
Gw.dEe is the same as kPEw.dEe
+---------------------------+--------------------------------+
0.1 < N < 1
F(w-n) .d, n(Ib')
I
I
I
1 < N < 10
F(w-n).(d-l), n('b l
10**(d-2) < N < 10**(d-l)
F(w-n) .1, n('b')
)
10**(d-1) < N < 10**d
F(w-n) .0, n('b')
+---------------------------+--------------------------------+
where 'b' stands for a blank (space) in the above table, and
is 4 for Gw.d editing and 'e'+2 for Gw.dEe editing.
11.3.1.6
'n'
Formatting Extreme Values
The floating point system used in SVS FORTRAN contains several
extreme value representations. These are infinity, both positive
and negative, and Not a Number, call NaN.
Infinities are produced by floating-poInt overflow and NaN's are produced by certain invalid operations such as zero ~ivided by zero.
When extreme values are printed in either D, E, F or G format,
they are represented by replacing all digits in the field with
either plus signs '+++.+++' for positive infinity, minus signs
'---.---' for negative infinity, or question marks '111.111' for
NaN.
Page 106
FORTRAN Reference Manual
Chapter 11
11.3.1.7
Format Specifications
L - Logical Editing
The edit-descriptor for a logical item is Lw, indicating that
the field width is 'Wi characters. The 'iolist' element associated with an L edit-descriptor must be of type LOGICAL (which is
the same as LOGICAL*4), ·LOGICAL*l or LOGICAL*2.
On input, the field consists of optional spaces, followed by an
optional decimal point, followed by a T (for .TRUE.) or P (for
.FALSE.). Any further characters in the field are ignored but
accepted on input, so that the strings .TRUE. and .FALSE. a.re
valid inputs.
or
On output, 'w'-l spaces are followed by either the character
P as·appropriate.
11.3.1.8
T
A - Character Editing
The forms of the
Aw.
The straight
from the number of
ated with.
The
associated with an
edit-descriptor for character items are A or
A format acquires an implied field width, 'Wi,
characters in the 'iolist' item it is associ'io1ist' item must be of type character to be
A or Awedit-descriptor.
On input, if 'w' equals or exceeds the number of characters. in
the 'io1ist' element, the rightmost characters of the input field
are used as the input characters, otherwise the input characters
are left-justified in the input 'io1ist' item, and trailing
spaces are added.
On output, if 'w' exceeds the characters produced by the 'io1ist' item, leading spaces are provided, otherwise the leftmost
'Wi characters of the 'iolist' item are output.
It is also possible to read and write non-character items with
the A edit-descriptor.
When this is done, the variable is
treated as if it were a CHARACTER variable whose length is the
number of bytes that that variable occupies.
11.3.2
Non Repeatable Edit Descriptors
Non-repeatable edit-descriptors are format list items which are
not associated with any 'io1ist' items.
11.3.2.1
'xxx' - Apostrophe Editing
APostro~he editing has the form of a
character constant.
It
causes c aracters to be written from the enclosed characters,
including spaces, of the edit-descriptor itself.
Within the
string, two ' signs in a row are interpreted as one apostrophe.
FORTRAN Reference Manual
Page 107
Chapter 11
Format Specifications
Apostrophe edit-descriptors cannot be used on input.
11.3.2.2
H - Hollerith Editing
The nB edit-descriptor (called Hollerith editing) transmits the
In' characters (including spaces) following the descriptor to the
output. Hollerith editing cannot be used for input.
Examples of APostrophe and Hollerith Editing
C
C
C
C
Each of the following WRITE statements writes
the characters
ASC'DEF
to the output.
C
WRITE(*, 970)
970 FORMAT ( , ABC' 'OEF' )
C
WRITE(*,
C
C
• (' i
ABC' • i , OEF' , ) , )
WRITE(*, , (7HASC' , OEF) , )
WRITE(*, 960)
960 FORMAT (7HABC'DEF)
11.3.2.3
X and T - positional Editing
The X and T edit-descriptors described below have the effect of
positioning the format controller within a record. They do not
by themselves transmit any characters to or from a record.
When a formatted record is read on input, it is treated as if
it were of infinite length, with as many trailing spaces as
needed to satisfy input requests. Positioning using the X and T
edit-descriptors determines the position of the next character to
be read from the record. These edit-descriptors may therefore be
used to skip portions of the input record or to re-read the same
positions in the record more than once.
On output, it is as if the input-output system initially
creates a record which is potentially of infinite length and
filled with spaces. As formatted output transmits characters to
the record, the final length of the record is determined by the
rightmost position to which a character is transmittede Changing
the pOSition with the X or T edit-descriptors does not directly
affect the length of the record, but does affect the position at
which the next character is transmitted to the output record.
Using the X or T edit-descriptors, positions in the record may
never have any characters transmitted to them (they are skipped),
which means that those positions retain their original blank
Page 108
FORTRAN Reference Manual
Format Specifications
Chapter 11
values, providing, of course, that characters are transmitted
after the skipped positions so that those character positions are
eventually included in the output record. It is also possible to
overwrite positions of formatted output records using the X and T
edit~descriptors
by positioning to a place where data was previously written. In this case, only the last value written to a
given character position becomes part of the final formatted
record.
nX edit-descriptor advances the record position by In'
If In' is omitted, the value 1 is used.
spaces.
The Te, TLc and TRc edit-descriptors position the record to a
specified column. The Tc edit-descriptor positions the record to
absolute column position 'ct. The TLc edit-descriptor moves the
column position to 'c' characters to the Left (backwards), relative to the current position. The TRc edit=descriptor moves the
column position 'c' characters to the Right (forwards) relative
to the current position.
On input, the T edit-descriptors have the effect of skipping or
allowing re-reading of portions of the input record. If the TLc
edit-descriptor moves the character position to a place where
input fields were previously transmitted, those i~ems can be retransmitted.
On output, if a character is transmitted to a position where
another character has already been transmitted, the earlier
transmission is replaced.
11.3.2.4
Slash Editing - End of Transfer on Record
The slash character
current record.
"I"
indicates
end
of
transfer
on
On input, the file is positioned to the beginning of
record.
th~
On output, an end-of-record is written and the
tioned to write at the start of the next record.
is
11.3.2.5
file
the
next
posi-
Backslash or Dollar Editing - Inhibit End of Record
The backslash "\" and dollar "$" edit-descriptors
when writing to the * device (the Standard Output).
only
apply
Normally, when the format controller terminates a format list,
data transmission to the current record ceases and the file is
positioned so that a new WRITE starts a new record.
If, while
scanning
the
format list, the format controller finds a
backslash character "\" or a dollar character "$", the automatic
FORTRAN Reference Manual
Page 109
Chapter 11
Format Specifications
end-of-record action is inhibited. This means that subsequent
input-output statements can continue reading from or writing to
the sa~e recordo
The most common use for this mechanism is to
prompt to the Standard Input and read a response on the same
line. For example:
WRITE(*, • (A) ') 'Please type your weight -> '
READ(*, '(BN, 16)') LIGHT
The backslash or dollar edit-descriptor does not inhibit the
automatic end-of-record generated when reading from the * unit.
Input from the Standard Input must always be terminated by a
carriage-return.
This permits proper functioning of the backspace and line-delete keys.
11.3.2.6
Colon Editing - Conditional Termination
The : character appearing in a format list has the effect of
terminating processing of the format list if there are no more
items in the 'iolist'. If there are more items in the 'iolist'
when the colon is encountered, the colon is ignored. There may
be more than one colon in a format list.
The colon edit-descriptor is useful for terminating extraneous
textual data that might otherwise be printed after all appropriate numeric items have been transferred. It is also useful for
preventing further / edit-descriptors on input.
11.3.2.7
P - Scale Factor Editing
The kP edit-descriptor sets the scale-factor for subsequent B,
F and G edit-descriptors until another kP edit-descriptor is
encountered. At the start of each input-output statement, the
scale-factor is zero CO). The scale-factor affects format editing as follows:
On input with E, P or G editing, providing that an explicit
exponent does not appear, the externally represented number is
equal to the internally represented number multipled by lO**k.
On input with E, P and G editing, the scale-factor has no
effect if there is an explicit exponent in the input field.
On output with B editing, the real part of the quantity is
multiplied by 10**k and the exponent is reduced by 'k', effectively altering the column position of the decimal point, but
not the actual output value.
On output with P and G editing, the externally represented
number is equal to the internally represented number multiplied by lO**k.
Page 110
FORTRAN Reference Manual
Chapter 11
11.3.2.8
Format Specifications
BN and BZ - Blank Interpretation
The DN and DZ edit-descriptors specify
blanks (spaces) in numeric input fields.
the
interpretation
of
The initial setting of this edit-descriptor is dependent on the
BLANK= parameter to the OPEN statement when this file was opened.
See the OPEN statement for a description of the BLANK= parameter.
If BZ editing is in effect, leading blanks are ignored and
embedded blanks are treated as zeros. This edit-descriptor stays
in effect until a BN edit-descriptor is encountered in the format
list.
If BN editing is in effect, blanks in subsequent input fields
are ignored until a BZ edit-descriptor is processed.
The effect
of ignoring blanks is to treat all the non-blank characters in
the input field as if they were right justified in the field,
with the number of leading blanks equal to the number of ignored
blanks.
In the example below, the READ statement treats the
characters shown between the vertical bars as the value 123,
where the term indicates a carriage-return:
READ(·, 100) 1
100 FORMAT(BN, 16)
123
1
456
123
1231
123\
Using the BN edit-descriptor in conjunction with the infinite
blank padding at the end of formatted records makes interactive
input very convenient.
11.3.2.9
S, SS and SP - Sign Control Editing
An output field generated by I, D, E, P or G editing includes
an optional sign immediately preceding the digits of the value.
The sign always appears if the number is negative, but if the
number is positive, FORTRAN omits the plus sign.
At the start of a format list, FORTRAN opts to omit plus signs.
An S, SS or SP edit-descriptor controls the option. Any option
chosen remains in effect until another one is found in the format
list.
An SP format code specifies that "optional" plus signs are
always printed. An SS edit-descriptor specifies that they are to
be suppressed always. An S option restores FORTRAN's option to
omit plus signs.
FORTRAN Reference Manual
Page 111
Format Specifications
Chapter 11
On input, these format codes have no effect and are ignored.
Page 112
FORTRAN Reference Manual
Chapter 12
Program and Subprogram Structure
Chapter 12 - Program and Subprogram Structure
A complete executable FORTRAN program consists of the following
elements:
a main program,
any number of SUBROUTINE subprograms,
any number of FUNCTION subprograms,
any number of BLOCK DATA subprograms.
A subprogram is a program unit which begins with a SUBROUTINE,
FUNCTION or BLOCK DATA statement.
A subprogram is defined
separately and can be compiled independently of the main program.
Subroutine and function subprograms are procedures which can
share values and results through argument lists, common blocks or
both. A BLOCK DATA subprogram acts as a container for initializing common blocks.
A procedure can be a subroutine, external
function or statement-function.
12.1
function,
intrinsic
Main Program
A main program is any program unit that does not start with a
SUBROUTINE, FUNCTION or BLOCK DATA statement. A main program may
start with a PROGRAM statement. Execution of a FORTRAN program
starts with the first executable statement in a main program.
Consequently there must be precisely one main program in every
executable FORTRAN program. The form of a PROGRAM statement is:
PROGRAM
progname
where 'progname' is a user-defined name of the main program.
The name 'progname' is a global name. Therefore it must not be
the same as the name of another external procedure or the name of
a common block. The name 'progname' is also local to the main
program and must not be the same as any other local name in the
main program. The PROGRAM statement may only appear as the first
FORTRAN Reference Manual
Page 113
Chapter 12
Program and Subprogram Structure
statement of a main program.
Example of ! PROGRAM Statement
PROGRAM
12.2
BESSEL
Access To Command Line Arguments
The SVS FORTRAN-77 system provides for access
line which called up the running program.
to
the
command
The FORTRAN-77 run-time library contains two routines which
enable a FORTRAN-77 program to access its command line arguments.
IARGC (Argument Count) is an INTEGER FUNCTION which returns the
number of arguments actually typed on the command line.
GETARG is a subroutine which retu·rns a specified argument.
definition of GETARG is:
The
SUBROUTINE GETARG(ARGNUM, ARGCH)
INTEGER ARGNUM
CHARACTER *(*) ARGCH
ARGNUM is the number of the argument which is to be accessed
from the command line. Arguments are numbered from 1 (not from
o as on some operating systems). Indexing from 1 is done for
compatibility with the Pascal numbering, and with FORTRAN-77's
default lower array bounds. The value passed to ARGNUM must be
in the range 1 through IARGC. If it is not, the results are
undefined.
Under some operating systems, the first argument
is the name of the program.
(argument
1)
The receiving variable is treated just as it is in a character
assignment statement in FORTRAN-77.
If the source character
string is shorter than the target variable, it is padded with
spaces on the right. If the source character string is longer
than the receiving variable, it is truncated.
Rere is a short example of using the argument access
to echo the command line to the standard output.
Page 114
facility
FORTRAN Reference Manual
Chapter 12
Program and Subpr?gram Structure
PROGRAM ECHO
CHARACTER*100 ARG
INTEGER I
200
12.3
00 200 I • 1, IARGC()
CALL GETARG(I, ARG)
WRITE(*, *) ARG
END
Formal Arguments and Actual Arguments
This section covers the relationship between formal arguments
and actual arguments in function and subroutine subprograms.
Throughout the discussion, the terms "formal argument" and "dummy
argument" are synonymous.
A formal argument is the name (local to the subprogram) by
which the argument is known during the execution of the subprogram.
The actual argument is the actual value (variable, expression,
array and so on) passed to the subprogram in question at the time
a caller references the subprogram.
There are a number of ways to pass values into and out of subprograms.
One way is via common blocks. Another way is to use
the argument mechanism of subroutines and functions.
It is this
second way that this section covers.
Arguments are used to pass values into and out of subprograms.
The number of actual arguments must be the same as the number of
formal arguments. The types of the corresponding formal and
actual arguments must also agree.
12.3.1
Argument Association
On entry to a subroutine or function, the actual arguments
become associated with the formal arguments. The association
remains effective until execution of the subprogram terminates.
Thus assigning a value to a formal argument while executing a
subprogram can change the value of the corresponding actual argument.
If an actual argument is a constant, function reference,
or an expression other than a simple variable, assigning a value
to the formal argument should not be done, and can lead to
strange and hard to diagnose side effects.
A formal argument that is a variable can be associated with an
actual argument tha~ is a variable, an array element or an
expression. The length attributes of integer and logical arguments must match exactly, that is, an actual argument of type
INTEGER*2 must only be associated with a formal argument of
FORTRAN Reference Manual
Page 115
Program and Subprogram Structure
Chapter 12
INTEGER*2.
Actual arguments which are integer expressions must be associated with formal arguments of type INTEGER*4 or INTEGER*2 depending on the default integer size.
Similarly, actual arguments
which are logical expressions must only be associated with formal
arguments of type LOGICAL*4 or LOGICAL*l depending on the default
integer size.
An "expression" in this context is any construct
which is not a variable, array or array element.
Actual arguments which are manifest constants, (i.e. names set
to constant values in PARAMETER statements) are treated exactly
as if the constant (a constant expression) were the actual parameter.
Names set to constant values in PARAMETER statements do
not become typed by their initial letter or otherwise except by
the form of the constant expression to which they were set.
Thus, actual arguments which are manifest constants will result
in 4 byte values with the $INT2 option not set or 2 byte integer
(1 byte logical) values with the $INT2 option set.
If an actual argument is an expression (anything not a variable, array, or array element), it is evaluated immediately prior
to the association of actual and formal arguments. If an actual
argument is an array element, its subscript expression is
evaluated just prior to the association and remains constant during execution of the subprogram, even if the subscript expression
contains variables that are re-defined while the subprogram executes.
A formal argument that is an array can be associated with an
actual argument that is an array or an array element. The number
and size of dimensions in the formal argument can differ from
those of the actual argument, but any re~erence to the formal
argument must be within the limits of the actual array's storage
sequence.
While the FORTRAN system cannot detect such an outof-bounds reference, the results are generally unpredictable and
undesirable.
A formal argument which is an asterisk character, ".", may only
appear in the argument list of a subroutine or in an ENTRY statement in a subroutine. The actual argument is an alternate return
specifier in the subroutine CALL statement (see below).
Formal arguments which are arrays or character strings may have
adjustable dimensions. This enables writing more general subprograms which can accept objects of varying size. A formal argument which is an array may have its dimensions specified by variables passed as actual arguments. This is an adjustable array.
An array formal argument may also have the upper bound of its
last dimension specified as an asterisk character, ".", which
declares it to be an assumed size array. In this case, the value
of that dimension is not passed as an actual argument, but is
determined by the number of elements in the array. If an array
Page 116
FORTRAN Reference Manual
Program and Subprogram Structure
Chapter 12
is dimensioned as *, it is the programmer's responsibility to
ensure that the calling program unit has provided an array big
enough to contain all the elements·stored into it in the subprogram.
Character strings may have their length specified as (*). This
declares the string to be of varying size. The length of the
string is not passed explicitly as an argument, but is determined
by the system from the length of the actual argument.
A formal argument which is of type character must not .be
greater than the length of the actual argument. If the length
of the actual argument is greater than that of the formal argument, the actual argument is truncated on the right.
If a formal argument is of type character whose length is
specified as (*), a character expression involving concatenation
of that argument must not be used as an actual argument to any
other procedure, format specification or input-output list in an
input-output statement.
12.4
Subroutines
A subroutine is a program unit that is called from other program units via the CALL statement. When inVOked, the subroutine
performs the actions that its executable statements define, and
then returns control to the program unit that called it. A subroutine does not directly return a value, although values can be
passed to the caller via the subroutine's arguments or via common
variables.
12.4.1
SUBROUTINE Statement
A subroutine starts with a SUBROUTINE statement and ends with
the first END statement that follows. A subroutine can contain
any kind of statement except a PROGRAM statement, a FUNCTION
statement or a BLOCK DATA statement. The form of a SUBROUTINE
statement is:
. SUBROUTINE
subname [ ( farg [r farg]
••. ) ]
'subname'
is the user-defined name of the subroutine.
'farg'
is a formal argument specification.
A formal
argument can be the user-defined name of a variable, array, dummy procedure, or it can be an
alternate-return
specifier designated by the
asterisk character n*"
The subroutine name 'subname' is a global name.
It is also
local to the subroutine it names. The list of argument names
FORTRAN Reference Manual
Page 117
Program and Subprogram Structure
Ghapter 12
defines the number (and with any subsequent IMPLICIT, type or
DIMENSION statements) the type of arguments to that subroutine.
Argument names must not appear in COMMON, DATA, EQUIVALENCE or
INTRINSIC statements.
I f a subro·utine does not have any formal arguments, an empty
argument list indicated by a pair of parentheses may follow the
name, as shown in the examples below.
Examples of SUBROUTINE Statements
SUBROUTINE
NOARGS
SUBROUTINE
ZILCH()
SUBROUTINE
ONEARG (RILEY)
SUBROUTINE
ALTRET(LIMIT, *)
12.4.2
CALL Statement
A subroutine is executed when a CALL statement in another subprogram references that subroutines by name. The form of a CALL
statement is:
CALL
subname [ ( [arg (, arg] ••• ) ) )
'subname'
is the name of the subroutine to call.
'arg'
is an actual argument.
described below.
The actual arguments
are
The actual arguments in the CALL statement must agree in type
and number with the corresponding formal arguments specified in
the SUBROUTINE statement of the referenced subroutine.
Actual
arguments may be one of the following:
An expression,
An array name,
An intrinsic function name,
An external procedure name,
A dummy procedure name,
An alternate-return specifier of the form *s, where IS' is the
statement-label of an executable statement in the same program
unit as the CALL statement.
Page 118
FORTRAN Reference Manual
Program and Subprogram Structure
Chapter 12
If there are no arguments in the SUBROUTINE statement, a CALL
statement that references that subroutine must not have any
actual arguments. A pair of parentheses following the subroutine
name is optional in the CALL statement. A formal argument can be
used as an actual argument in another subroutine call.
Execution of a CALL statement proceeds as follows:
1.
All actual arguments that are expressions are evaluated.
2.
All actual arguments are then
corresponding formal arguments.
3.
The body of the specified subroutine is executed.
4.
Control is returned to the subroutine's caller when a RETURN
or and END statement is executed in the subprogram, either
at the statement following the CALL statement, or at the
alternate return specifier designated in a RETURN statement.
associated
with-
their
A subroutine specified in any program unit can be called from
any other subprogram within the same executable program. Recursive subroutine calls are not allowed in FORTRAN.
That is, a
subroutine cannot call itself directly, nor ~ay a subroutine
called by the current subroutine subsequently call the current
subroutine.
12.5
Functions
A function is referenced in the context of an expression and
returns a value that is used in the evaluation of that expression. There are three kinds of functions, namely: external functions, intrinsic functions and statement functions. The subsections to follow describe the three kinds of functions.
A function reference can appear in an expression.
Referencing
a function in the context of an expression causes that function
to be executed. The resulting value that the function returns is
used as an operand in the expression that references the function. The form of a function reference is:
funcname
([arg [, arg]... ] )
'funcname'
is the name of an external, intrinsic, or
ment function.
larg'
is an actual argument to the function. The forms
of the actual arguments are described below.
state-
The number of actual arguments must be the same as the number
of formal arguments. Except for generic intrinsic functions, the
types of the actual arguments must agree with the types of the
FORTRAN Reference Manual
Page 119
Program and Subprogram Structure
Chapter 12
corresponding formal arguments. An actual argument can be anyone
of:
An expression,
An array name,
An external procedure name,
An intrinsic function name,
A dummy procedure name.
12.5.1
External Functions
An external function is specified by a function subprogram. It
starts with a FUNCTION statement and ends with an END statement.
A function can contain any kind of statement except a SUBROUTINE
statement,
PROGRAM statement or BLOCK DATA statement. The form
of a FUNCTION statement is:
(type]· FUNCTION
function_name ( (farg (, farg] ••• ] )
'type'
defines the return type of the function.
'type'
is
one
of
INTEGER,
INTEGER*l~
INTEGER*2,
INTEGER*4, REAL, REAL*4, REAL*B, DOUBLE PRECISION, CHARACTER [*len] , LOGICAL, LOGICAL·l, LOGICAL*2, LOGICAL*4 or COMPLEX.
'function~name'
is the user-defined name of the function.
'farg'
is a user-defined name of the formal argument.
The function name 'function name' is a global name.
It must
not be the same as the name of any other PROGRAM, SUBROUTINE,
FUNCTION or BLOCK DATA subprogram. The function name is also
local to the function it names.
If the 'type' specification is omitted from the function
declaration, the function's type is determined by default and any
subsequent IMPLICIT statements that would determine the type of
an ordinary variable.
If the 'type' specifier is present, the
function name must not appear in any subsequent type statements.
The list of argument names determines the number (and with any
subsequent IMPLICIT, type or DIMENSION statements) the type of
the arguments to the function. Neither the argument names nor
the function name may appear in any COMMON, DATA, EQUIVALENCE or
INTRINSIC statements.
The function name must appear as a variable in the subprogram
that defines the function. Each execution of the function must
Page 120
FORTRAN Reference Manual
Chapter 12
Program and Subprogram Structure
assign a value to that variable. The final value of this variable upon execution of a RETURN or END statement, defines the
value of the function. After this variable is defined, its value
can be referenced in an expression just like any other variable.
In addition to the value returned, an external function can
return values via as~ignment to one or more of its formal arguments or through variables in common areas.
A function which is declared as type CHARACTER*(*) derives its
length from the specification (declaration) of the function in
the calling program unit.
12.5.2
Intrinsic Functions
Intrinsic functions are those functions that the FORTRAN compiler pre-defines. Intrinsic functions are available for use in
a FORTRAN program. The table in the appendix on "Intrinsic Functions" gives the name, definition, number of arguments, and type
of the intrinsic functions available in SVS FORTRAN-77. Those
intrinsic functions which accept several types of arguments must
have all those arguments of the same type in any given reference.
An intrinsic function can appear in an INTRINSIC statement, but
only those intrinsic functions listed in the table in the appendix on "Intrinsic Functions" may do so.
An intrinsic function
may also appear in a type statement, but only if the type is the
same as the standard type of that intrinsic function.
Certain intrinsic functions limit the range of their arguments
in certain ways determined by the definition of the function
being computed. For example, the logarithm of a negative number
is mathematically undefined, and is therefore not allowed.
12.5.3
Statement Functions
A statement-function is a function that is defined by a single
statement.
It is similar in form to an assignment statement. A
statement-function statement must appear in a subprogram after
any specification statements and before any executable statements. A statement-function statement is not executable - rather
the body of the statement-function statement serves to define
the meaning of the statement-function. A statement-function is
executed (in the body of the subprogram in which it is defined)
by referencing it just like a function.
The form of
a
statement-function statement is:
function_name ( [arg [, arg) ••• ] ) = expression
'function name' is the name
defined.
FORTRAN Reference Manual
of
the
statement-function
being
Page 121
Chapter 12
Program and Subprogram Structure
name
of
the
formal
'arg'
is the
user-defined
argument(s), if any.
'expression'
is an expression that defines how the formal
arguments are to be combined to generate a function result when the function is referenced.
The type of the 'expression' must be assignment compatible with
the type of the statement-function name. The list of formal
argument names serves to define the number and types of arguments
to the statement-function. The scope of the formal arguments is
the statement-function. Therefore the formal argument names may
be used as other user-defined names in the rest of the program
unit that contains the statement-function statement. The name of
the statement-function is local to the containing program unit,
and therefore must not be used for any other purpose, other than
as the name of a common block, or as the name of a formal argument to another statement-function statement. The type of all
such other uses must be the same. If a formal argument to a
statement-function statement is the same as a local name in the
program unit, a reference to that name within the statementfunction always refers to the formal argument, never to the other
usage.
Within the 'expression', references to variables, formal arguments of the containing subprogram, other functions; array elements and constants, are all allowed. Statement-function references, however, must refer to statement-functions defined prior
to the statement-function in which they appear.
Statementfunctions must not be called recursively, either directly or
indirectly.
A statement-function can only be referenced in the subprogram
in which it is defined.
A statement-function name must not
appear in any specification statement other than a type statement
or a COMMON statement. If a statement-function name appears in a
type statement, that name must not be defined as an array name.
If a statement~function name appears in a COMMON statement, that
name can only be the name of the common area.
12.6
ENTRY Statement
A subroutine or function subprogram has a primary entry-point
which is established via the SUBROUTINE or FUNCTION statement
which declares that program unit. A subroutine call or a function reference normally activates that subprogram at its primary
entry-point, and the first statement which is executed is normally the first executable statement in the subprogram.
It is possible, however, to define alternate entry-points in a
subroutine or function subprogram. These alternate entry-points
are the start of sequences of statements which are different from
Page 122
FORTRAN Reference Manual
Program and Subprogram Structure
Chapter 12
the sequence executed by entering the subprogram at its primary
entry-point. In addition, such alternate entry-points can have
formal argument lists which differ in number and type from those
found in the primary entry-point, and from those of other ENTRY
statements in the same subprogram.
The format of the ENTRY
statement is:
ENTRY
entname [ ( farg [, farg] ••• ) ]
'entname'
is the user-defined name of the entry-point
the subroutine or function subprogram.
for
'farge
is a formal argument specification.
A formal
argument can be the user-defined name ~f a variable, array, dummy procedure, or, if the subprogram is a subr~utine subprogram, it can be an
alternate-return specifier designated by
the
asterisk character "*".
The entry-point name 'entname' is a global name.
It is also
local to the subprogram in which it appears. The list of argument names defines the number (and with any IMPLICIT, type or
DIMENSION statements) the type of arguments to that subroutine.
Argument names must not appear in COMMON, DATA, EQUIVALENCE or
INTRINSIC statements.
If the entry-point name, 'entname' is in a function subprogram,
the name can appear in a type statement.
An ENTRY statement must not appear within the
block or a DO block.
body
of
an
IF
As with SUBROUTINE and FUNCTION statements, if there are no
arguments to the ENTRY statement, an empty argument list can be
supplied.
When a subprogram is referenced or called via an alternate
entry-point, the actual arguments must agree in number, order and
type with the formal arguments (except for subroutine names and
alternate return specifiers which do not have a type).
12.6.1
Restrictions on the ENTRY Statement
An entry name must not appear as a dummy argument in a FUNCTION, SUBROUTINE or another ENTRY statement, and must not appear
in an EXTERNAL statement.
In a function subprogram, the only place the entry-point name
may be used prior to the ENTRY statement is in a type statement.
to
If a function subprogram is of type character, all entry-points
that function must also be of type character. If the length
FORTRAN Reference Manual
Page 123
Program and Subprogram Structure
Chapter 12
of the character function is specified as (*), all entry-points
to that function must also have a length of (*), otherwise all
entry-points must have the same length specification.
An argument in an ENTRY statement cannot appear prior
ENTRY statement unlp.ss it:
to
that
is either a type statement,
is an argument in the SUBROUTINE or FUNCTION statement which
begins the procedure containing the ENTRY statement,
appears in a prior ENTRY statement in the same procedure.
12.1
RETURN Statement
A RETURN statement returns control from a subprogram to the
program unit which called it. A RETURN statement can only appear
in a function or subroutine subprograms The form of a RETURN in
a function subprogram is:
RETURN
The form of a RETURN· statement in a subroutine subprogram is:
RETURN
(!.]
Where the optional!. is an integer expression.
Execution of a RETURN statement terminates the execution of the
containing function or subroutine subprogram.
If the RETURN
statement is in a function subprogram, the value of the function
is the current value of the variable with the same name as the
function. If the function variable has not been assigned to
prior to executing a RETURN or an END statement, the function
value is undefined.
The RETURN statement is optional in a subprograms Executing an
END statement is equivalent to executing a RETURN statement. .
If e is supplied on the RETURN statement, it indicates an
alternate return from the subroutine. If e lies between 1 and
'n', where 'n' is the number of asterisks in the SUBROUTINE or
ENTRY statement, the value of e selects the e'th asterisk from
the formal argument list. ControT then returns-to the caller at
the label specified by the !.'th alternate return specifier.
If e is omitted, or if e lies outside the range 1 to 'n', the
effect is to execute a-normal return. Control then returns to
the caller at the statement after the CALL statement that invoked
the current subroutine.
Page 124
FORTRAN Reference Manual
Chapter 12
12.8
Program and Subprogram Structure
Definition Status
When a RETURN statement or an END statement is executed in a
subprogram, all objects within the subprogram become undefined,
with the following exceptions:
Objects specified by SAVE statements.
Objects in blank common.
Anything in a named common block that appears in the current
subprogram and also appears in at least one other subprogram
that directly or indirectly references the current subprogram.
Initially-defined objects that have
nor become undefined.
neither
been
re-defined
If a named common block appears in the main program,
in that common block does not become undefined.
12.9
anything
BLOCK DATA Subprogram
A BLOCK DATA subprogram is a non-executable subprogram which is
used to initialize the values of variables and array elements in
named common areas. There may be more than one block data subprogram in a FORTRAN program, but if there is more than one block
data subprogram, only one of them can be un-named. The for.mat of
a BLOCK DATA statement is:
BLOCK DATA
[blockname]
where 'blockname' is the optional name of the block data
gram.
subpro-
The BLOCK DATA statement must appear as the first statement of
the block data subprogram. The name, 'blockname', if present,
must not be the same as any the name of any any external procedure, main program, common area or other block data subprogram.
The name, 'blockname' must not be the same as any local name
in
the subprogram.
A block data subprogram can contain type statements, IMPLICIT,
PARAMETER, DIMENSION, COMMON, SAVE, EQUIVALENCE or DATA statements. A block data subprogram ends with an END statement.
More than one named common block can be initialized in the same
block data subprogram.
All the variables in a given named common block must be specified, even if they are not all initialized.
A given named common block may only be specified in
data subprogram in the same executable program.
FORTRAN Reference Manual
one
block
Page 125
Chapter 12
Program and Subprogram Structure
Examples of BLOCK DATA Subprogram
BLOCK DATA
•
*
Whammo
Declare a cornmon block with variables
•
COMMON /DRINKS/
REAL Beer
COMPLEX Wine
DOUBLE PRECISION
•
•
•
Beer, Wine, Scotch
Scotch
Declare another common block with variables
COMMON /FOODS/
LOGICAL Burger
REAL Dogs
COMPLEX Fries
•
•
Burger, Dogs, Fries
Now initialize some of the variables.
•
•
DATA
Beer /3.2/,
Wine /(11.5, 1.5)/
DATA
Burger /.TRUE./, Fries /(1.1, 2.8)/
END
In the example above, note that not all the variables were initialized~
12.10
The FORTRAN Intrinsic Functions
Intrinsic functions are those system supplied ("built-in")
functions which are otherwise difficult to express in FORTRAN.
An intrinsic function is supplied by FORTRAN. An intrinsic function returns a single value and is referenced in the same way as
a user-defined function.
If a variable, array or statement-function is defined with the
same name as that of an intrinsic function, the name is local to
the program unit in which it is declared and the intrinsic function of that name is no longer available to that program unit.
If a function subprogram is defined which has the same name as
that of an intrinsic function, use of that name references the
intrinsic function, unless the name is declared as the name of an
external function via the EXTERNAL statement.
Certain intrinsic functions are generic. In general, if a generic name exists, a generic name can be used in place of a
specific name and permits greater flexibility than a specific
name.
Except for the type conversion functions, the type of the
Page 126
FORTRAN Reference Manual
Chapter 12
Program and Subprogram Structure
argument to a generic function determines the type of the result.
For example, the generic function LOG computes the natural logarithm of its argument, which may be real, double precision or
complex.
The type of the result is the same as the type of its
argument. The specific functions ALOG, DLOG and CLOG also compute the natural logarithm.
ALOG computes the log of a real
argument and returns a real result.
Likewise, DLOG and CLOG
accept double precision and complex arguments and return double
precision and complex results, respectively.
Only the specific name can be used as an actual argument when
an intrinsic function name is passed to a user-defined procedure
or function.
The table in the appendix, "FORTRAN Intrinsic Functions",
shows the intrinsic functions, their generic and specific names,
their number of arguments and their argument types and result
types.
FORTRAN Reference Manual
Page 127
Program and Subprogram Structure
Page 128
Chapter 12
FORTRAN Reference Manual
FORTRAN Compile Time Options
Chapter 13
Chapter 13 - FORTRAN Compile Time Options
Compiler Directives are a SVS extension to ANSI FORTRAN. SVS
FORTRAN comp11er d1rectives provide additional controls over the
compiler's actions.
A compiler directive line is a line with a dollar sign $
in
column one. A compiler directive line can appear anywhere that a
comment line can appear.
Spaces are significant in compiler
directive lines, where they delimit keywords and filenames.
Some of the compiler directives listed
FORTRAN-77 cater to FORTRAN-66 features.
13.1
below
are
to
make
$INCLUDE - Include Source File
$INCLUDE filename
the file specified by "filename" is textually included in the
program source, as if the actual contents of the included file
had been written there. The file name is not quoted.
Included files may be nested to a maximum depth of five.
13.2
$XREF - Generate Cross Reference
$XREF
generate a cross-reference listing at the end
subprogram.
13.3
of
each
compiled
$SEGMENT - Designate Segment Name
$SEGMENT [identifier]
the generated object-code of subsequent procedures is placed into
the segment named by 'identifier'. If the $SEGMENT directive
appears without any 'identifier' field, the generated object-code
is placed in a segment whose name is '
, (eight spaces).
FORTRAN Reference Manual
Page 129
Chapter 13
FORTRAN Compile Time Options
The SVS linker imposes a limit of 32K bytes of object code per
segment.
However, the linker will automatically split larger
segments as required. Even for very large programs, a user normally should not have to explicitly partition his program into
segments.
13.4
$COL72 - Restrict Source Lines to 72 Columns
$COL72
indicates that source lines are to end in column 72.
If this
option is not specified, source lines can be up to 120 characters
long. But, the ANSI FORTRAN-77 standard restriction of a maximum
of 1360 characters per statement still applies. This corresponds
to 20 lines of 66 columns.
13.5
FORTRAN-66 Compatibility Options
The FORTRAN compiler accepts options which change features of
the language in a manner compatible with FORTRAN-66. These
options are listed in the following paragraphs.
13.5.1
$F66DO --Implement FORTRAN-66 DO Loops
If the $F66DO option is used, DO loops always execute at
once.
13.5.2
least
$CHAREQU - Character and Numeric Data Equivalence'
The $CHAREQU option means that CHARACTER and numeric data can
now be assigned to the same COMMON areas. Using this option,
CHARACTER and numeric data can also be EQUIVALENCE'd.
In· addition, the $CHAREQU option indicates that non-CHARACTER
variables can be initialized with CHARACTER data constants via
the DATA statement. See Chapter 7 - "Data Initialization" for
details.
13.5.3
$INT2 - Make Integers l6-Bits
If the $INT2 option is used, the INTEGER data type is INTEGER.2
by default, although all the length specifications are still
available if explicitly used in specification statements.
If the $INT2 option is used, LOGICAL variables default to LOGICAL·l.
Just as for INTEGER, all the length specifications for
LOGICAL are still available if explicitly used in specification
statements.
Page 130
FORTRAN Reference Manual
Chapter 13
FORTRAN Compile Time Options
It should be noted that these last two features
the ·storage unit· standards of
FORTRAN-77,
nevertheless.
conflict with
but are useful
Rote: Although the $INT2 option changes the default size of
INTEGER and LOGICAL variables, the FORTRAN system still expects
to see 4-byte variables in those contexts where an INTEGER*4 (or
a LOGICAL*4) is required. For example, the assigned GO TO statement still expects a 4-byte variable as the subject of the ASSIGN
statement and any GO TO statement which references that variable.
If the $INT2 option is set, actual parameter integer expressions (not variables or array elements but any constant or computed expression) are coerced to two byte values.
Similarly,
constant and computed logical expressions are coerced to one byte
values if the $INT2 option is set.
This makes expressions
default to matching lengths with parameters of default length.
A fairly common programming error is to compile a subprogram
with $INT2 set but to call it from a compile in which $INT2 is
not set. This leads to incompatible actual and formal arguments
if all length specifications are defaulted by the FORTRAN system.
FORTRAN Reference Manual
Page 131
FORTRAN Compile Time Options
Page 132
Chapter 13
FORTRAN Reference Manual
Appendix A
Messages from the FORTRAN System
Appendix A - Messages from the FORTRAN System
A.l
o
Compile-Time Error Messages
1
2
3
4
5
6
8
9
10
Unknown error
Fatal error reading source block
Non-numeric characters in label field
Too many continuation lines
Fatal end-of-file encountered
Labeled continuation line
Missing field on $ compiler directive line
Unrecognizable $ compiler directive
Input sourc~ file not a valid text file format
Maximum depth of include file nesting exceeded
11
12
13
14
15
16
17
18
19
20
Integer constant overflow
Error in real constant
Too many digits in constant
Identifier too long
Character constant extends to end of line
Character constant is zero length
Illegal character in input
Integer constant expected
Label expected
Error in label
21
Type name expected (INTEGER [*n) , REAL[*n], DOUBLE
COMPLEX, LOGICAL [*n] , or CHARACTER[*n)
INTEGER constant expected
Extra characters at end of statement
'(' expected
Letter IMPLICIT'ed more than once
1)1 expected
Letter expected
Identifier expected
Dimension(s) required in DIMENSION statement
Array dimensioned more than once
22
23
24
25
26
27
28
29
30
31
32
33
Maximum of 7 dimensions in an array
Incompatible arguments to EQUIVALENCE
Variable appears more than once in
statement
FORTRAN Reference Manual
a
type
PRECISION,
specification
Page 133
Appendix A
Messages from the FORTRAN System
34
35
36
37
38
39
40
This identifier has already been declared
This intrinsic function cannot be passed as an argument
Identifier must be a variable
Identifier must be a variable or the current FUNCTION name
'I' expected
Named COMMON block already saved
Variable already appears in a COMMON block
41
42
50
Variables in different COMMON blocks cannot be EQUIVALENCE'd
Number of subscripts in EQUIVALENCE statement does not agree
with variable declaration
EQUIVALENCE subscript out of range
Two distinct cells EQUIVALENCE'd to the same location in a
COMMON block
EQUIVALENCE statement extends a COMMON block in a negative
direction
EQUIVALENCE statement forces a variable to two distinct locations, not in a COMMON block
Statement number expected
Mixed CHARACTER and numeric items not allowed in same COMMON
block
CHARACTER items cannot be EQUIVALENCE'd to non-CHARACTER
items
Illegal symbols in an expression
51
52
53
54
55
56
57
58
59
60
Cannot use SUBROUTINE name in an expression
Type of argument must be INTEGER or REAL
Type of argument must be INTEGER, REAL or CHARACTER
Types of comparisons must be compatible
Type of expression must be LOGICAL
Too many subscripts
Too few subscripts
Variable expected
'.' expected
Size of EQUIVALENCE'd CHARACTER items must be the same
61
62
63
64
65
70
Illegal assignment - types do not match
Can only call SUBROUTINES
Dummy arguments cannot appear in COMMON statements
Dummy arguments cannot appear in EQUIVALENCE statements
Assumed-size array declarations can only be used for dummy
arrays
Adjustable-size array declarations can only be used for dummy
arrays
Assumed-size array dimension specifier, "*", must be the
upper bound of the last dimension
Adjustable bound must be either a dummy argument or be in
COMMON prior to appearance
Adjustable bound must be simple integer expression containing
only constants, COMMON variables, or PARAMETER constant names
Cannot have more than one main program
71
The size of a named COMMON block must
43
44
45
46
47
48
49
66
67
68
69
Page 134
be
the
same
in
all
FORTRAN Reference Manual
Appendix A
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
93
94
95
96
98
99
Messages from the FORTRAN System
subprograms
Dummy arguments cannot appear in DATA statements
Variables in blank COMMON cannot appear in DATA statements
Names of SUBROUTINES, FUNCTIONS, INTRINSIC FUNCTIONS and such
cannot appear in DATA statements
Subscripts out of range in DATA statement
Repeat count must be integer value greater than zero
Constant expected
Type conflict in DATA statement
Number of variables does not match the number of values in
DATA statement list
Statement cannot have a label
No such INTRINSIC function
Type declaration for INTRINSIC function does not match actual
type of INTRINSIC function
Letter expected
Type of FUNCTION does not agree with previous usage
This subprogram has already appeared in this compilation
This procedure has already been defined as appearing in
another compilation unit via a $USES command
Error in type of argument to INTRINSIC function
SUBROUTINE/FUNCTION previously used as a FUNCTION/SUBROUTINE
Unrecognizable statement
Expression not allowed
Missing END statement
Fewer actual arguments than formal arguments in a FUNCTION or
SUBROUTINE reference
More actual arguments than formal arguments in a FUNCTION or
SUBROUTINE reference
Type of actual argument does not agree with formal argument
The following procedures were called but not defined
Size of type CHARACTER item must be between 1 and 255
INTEGER*4 variable required
100
Statement out of order
101
102
103
104
105
106
107
108
109
110
Unrecognizable statement
Illegal jump into block
Label already used for FORMAT
Label already defined
Jump to FORMAT label
DO statement forbidden in this context
DO label must follow a DO statement
ENDIF forbidden in this context
No matching IF for this ENDIF
Improperly nested DO block in IF block
III
112
113
114
ELSEIF forbidden in this context
No matching IF for ELSEIF
Improperly nested DO or ELSE block
I (I expected
FORTRAN Reference Manual
Page 135
Messages from the FORTRAN System
Appendix A
115
116
117
118
119
120
'). expected
THEN expected
Logical expression expected
ELSE statement forbidden in this context
No matching IF for ELSE
Unconditional GOTO forbidden in this context
121
122
123
124
125
126
127
128
129
Assigned GOTO forbidden in this context
Block IF statement forbidden in this context
Logical IF statement forbidden in this context
Arithmetic IF statement forbidden in this context
I,' expected
Expression of wrong type
RETURN forbidden in this context
STOP forbidden in this context
END forbidden in this context
131
132
133
134
135
136
137
138
139
140
Label referenced but not defined
DO or IF block not terminated
FORMAT statement not permitted in this context
FORMAT label already referenced
FORMAT must be labeled
Identifier expected
Integer variable expected
'TO' expected
Integer expression expected
Assigned GOTO but no ASSIGN statements
141
142
143
144
145
146
147
148
Unrecognizable character constant as option
Character constant expected as option
Integer expression expected for unit designation
STATUS option expected after ',' in CLOSE statement
Character expression as filename in OPEN
FILE= option must be present in OPEN statement
RECL= option specified twice in OPEN statement
Integer expression expected for RECL= option in OPEN statement
Unrecognizable option in OPEN statement
Direct access files must specify RECL= in OPEN statement
149
150
151
152
153
154
155
156
157
158
159
160
Assumed-sized arrays not allowed as input-output list elements
End of statement encountered in implied DO, expressions
beginning with' (' not allowed as input-output list elements
Variable required as control for implied DO
Expressions not allowed as reading input-output list elements
REC= option appears twice in statement
REC= options expects integer expression
END- option only allowed in READ statement
END- option appears twice in statement
Unrecognizable input-output unit
Unrecognizable format in input-output statement
Page 136
FORTRAN Reference Manual
Appendix A
.161
162
163
164
165
166
167
169
170
171
Messages from the FORTRAN System
Options expected after ',' in input-output statement
Unrecognizable input-output list element
Label used as format but not defined in FORMAT statement
Integer variable used as assigned format but no ASSIGN
statement
Label of an executable statement used as format
Integer variable expected for assigned format
Label defined more than once as format
FUNCTION references require "()"
INTEGER expression expected for array dimension bound
172
Lower dimension bound must be less than or equal to upper
dimension bound
DATA statement cannot initialize arrays of unknown size
200
201
202
203
204
205
206
207
208
209
210
Variable name of named COMMON name expected
This variable already SAVE'd
Cannot SAVE dummy arguments
Cannot SAVE COMMON variables
INTEGER and LOGICAL *1, *2, or *4 only
No *n allowed for DOUBLE PRECISION
Only REAL*4 or REAL*8 allowed
No *n allowed for COMPLEX
Size expression only allowed for CHARACTER
INTEGER constant expression expected
INTEGER constant or INTEGER constant expression expected
211
212
213
214
215
216
217
218
219
220
CHARACTER substring expression out of range
CHARACTER substring expression must be of type INTEGER
Error in CHARACTER substring expression
CHARACTER expression expected
LOGICAL expression expected
CHARACTER*(*} only allowed for dummy arguments
Undeclared PARAMETER constant
Constant expression not allowed
Arithmetic operators only apply to numeric values
Malformed COMPLEX constant
221
222
223
224
229
230
Maximum of seven levels of implied-DO allowed
Error in DATA statement variable list
Error in implied DO list in data statement
Variables in named COMMON can only appear in a DATA statement which is in a BLOCK DATA subprogram
Integer subscript expected
Subscript error
This identifier is already in use as an implied-DO control
variable
Integer constant expression or implied DO control variable
expected
Integer expression required
Division by zero
231
Error in COMPLEX primary
225
226
227
228
FORTRAN Reference Manual
Page 137
Messages from the FORTRAN System
232
233
234
235
236
237
238
239
240
241
242
2t3
244
245
246
247
248
249
250
251
252
253
254
255
256
270
271
272
273
274
275
276
277
278
279
Appendix A
Numeric expression or CHARACTER expression expected
COMPLEX can only compare for equality
COMPLEX is not compatible with DOUBLE PRECISION
Constant expression expected
ENTRY statements must appear in SUBROUTINE or FUNCTION subprograms
ENTRY statements cannot be within a block IF or a 00 statement range
Concatenation only applies to CHARACTER values
'I' expected
Substring operations only apply to CHARACTER variables or
CHARACTER array elements
Error in implied 00 expression in a DATA statement
Implied DO iteration count is zero in a DATA statement
Error in formal argument list
Alternate return is not allowed in a FUNCTION subprogram
Substring error in EQUIVALENCE statement
EQUIVALENCE statement must not require *2, *4, or *8 variables to be allocated on odd byte addresses
EQUIVALENCE statement must not require a COMMON block to be
allocated on odd byte addresses
CHARACTER arguments cannot contain concatenation of values
that are of size *(*)
Numeric expression expected
SUBROUTINE or FUNCTION name has already been used as a COMMON name
Recursive calls are not allowed
Statement-FUNCTIONS require variable or value arguments
Alternate ENTRY in CHARACTER FUNCTION must be of type CHARACTER and must be the same size as the FUNCTION
This INTRINSIC FUNCTION cannot be passed as an argument
Executable statements cannot appear in BLOCK DATA subprograms
An argument to an ENTRY statement has already appeared as a
local variable
280
Assigned GO TO variable must be INTEGER or INTEGER*4
INTEGER, REAL, or DOUBLE PRECISION variable expected
INTEGER, REAL, or DOUBLE PRECISION expression expected
Unrecognizable element in option list
Option appears more than once in an option list
Incorrect type for variable
Variable must be *4 in size
CHARACTER variable or CHARACTER array element required
CHARACTER expression expected
Cannot have FILE and UNIT specifier in same INQUIRE statement
Must have a FILE or UNIT specifier in INQUIRE statement
281
282
Must have UNIT specifier
PRINT statement requires no option list - use WRITE
Page 138
FORTRAN Reference Manual
Appendix A
Messages from the FORTRAN System
283
284
285
286
287
288
289
290
WRITE statement must have an option list
READ statement must not have both RECz and END= options
Must not specify REC= option with * format specifier
Cannot do internal input-output with * format specifier
Cannot use REC- specifier with internal input-output
Malformed implied DO loop
Implied DO loop must have simple variable for loop control
Wrong number of arguments to intrinsic function
291
292
293
294
295
296
297
350
Unit set more than once in input-output statement
No unit specified in input-output statement
Error in FORMAT statement
Hexadecimal constant expected
Too many characters in statement
Can't find $INCLUDE file
Sub arrays cannot exceed 32766 bytes in size
Procedure too large
400
401
402
403
407
408
409
Code file write error
Error in rereading code file
Error in reopening text file
Procedure too large (code buffer too small)
Not enough room for intermediate code file
Error in writing code file
Error in reading intermediate code file
A.2
Run-Time Error Messages
These messages are issued by the input-output run time system,
and represent the possible values of 'iostat' in an 'iolist'.
-1
End of file found on a READ with no END= option.
600
601
602
603
604
605
606
607
608
609
610
FORMAT statement missing final ')'
Sign not expected in input
Sign not followed by digit in input
Digit expected in input
Missing N or Z after B in format
Unexpected character in format
Zero repetition factor in format not allowed
Integer expected for w field in format
Positive integer required for w field in format
'.' expected in format
Integer expected for d field in format
611
612
613
614
615
616
617
Integer expected for e field in format
Positive integer required for e field in format
Positive integer required for w field in A format
Hollerith field in format must not appear for reading
Hollerith field in format requires repetition factor
X field in format requires repetition factor
P field in format requires repetition factor
FORTRAN Reference Manual
Page 139
Messages from the FORTRAN System
Appendix A
618
619
620
Integer appears before '+' or '-' in format
Integer expected after '+' or '-' in format
P format expected after signed repetition factor in format
621
622
623
624
625
626
627
628
Maximum nesting level (10 levels) for formats exceeded
')' has repetition factor in format
Integer followed by',' illegal in format
'0' is illegal format control character
Character constant must not appear in format for reading
Character constant in format must not be repeated
'I' in format must not be repeated
'I, '$', I:', IS', eSP' and ISS' in format must not be
repeated
8M or BZ format control must not be repeated
Attempt to perform input-output on unknown unit number
629
630
631
632
633
634
635
636
637
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
Formatted or list-directed input-output attempted
opened as unformatted
Format fails to begin with '(I
I format expected for integer read
P, 0, G or B format expected for real read
Two '.' characters in formatted real read
Digit expected in formatted real read
L format expected for logical read
T or F expected in logical read
A format expected for character read
on
file
I format expected for integer write
v field in P format not greater than d field + 1
Scale factor out of range'of d field in E format
E, 0, G or P format expected for real write
L format expected for logical write
A format expected for character write
Attempt to do unformatted input-output to a file opened as
formatted
Unable to write blocked output - possibly no room on output
device
Unable to read blocked input
Error in formatted text file - no carriage-return in last
512 bytes
Integer overflow on input
T, TL or TR in format must not be repeated
positive integer expected for c field in T, TL or TR format
Attempt to open direct-access unit on unblocked device
Attempt to do external input-output on a unit beyond endof-file record
Attempt to position a unit for direct-access on a nonpositive record number
Attempt to do direct-access on a unit opened as sequential
Attempt to position direct-access unit on an unblocked device
Attempt to position direct-access unit beyond end-of-file
Page 140
FORTRAN Reference Manual
Appendix A
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
697
698
699
Messages from the FORTRAN System
for reading
Attempt to backspace unit connected to unblocked
unformatted file
device
or
Attempt to backspace sequential, unformatted unit
Argument to ASIN or ACOS out of bounds - ABS(x) > 1.0
Argument to SIN or COS too large - ABS(x) > 10@+(6)
Attempt to do unformatted input-output to internal unit
Attempt to put more than one record into an internal unit
Attempt to write more characters to an internal unit than
its length
EOF called on unknown unit
Direct-access formatted input files must not use DLE blank
compression
Error in opening file
Error in closing file
Can't specify KEEP in close if file opened SCRATCH
Unrecognizable option specified as character value
input-output statement
File name required unless status is SCRATCH
Must not name file if status is SCRATCH
Record length not allowed for sequential files
Record length must be positive
Record length must be specified for direct-access files
BLANK option only for formatted files
Rewind only allowed on sequential files
Endfile only allowed on sequential files
in
Backspace only allowed on sequential files
Formatted records must be less than or equal to 512 characters
More characters written to internal file record than record
length
Incorrect number of characters read in formatted record of
direct-access file
Attempt to write too many characters into formatted record
of direct-access file
No repeatable edit descriptor found and format exhausted
Digit expected in input field exponent
Too many digits in input real number
Numeric field expected in input
Unexpected character encountered in list-directed input
Repeat factor in list-directed input must be positive
',' between reals for complex expected in list-directed
input
')' expected to terminate complex in list-directed input
Attempt to do list-directed input-output to direct-access
file
Integer variable not currently assigned a FORMAT label
End-of-file encountered on a read with no END= option
Integer variable not assigned a label used in assigned GOTO
FORTRAN Reference Manual
Page 141
Messages from the FORTRAN System
Appendix A
statement
701
702
703
704
705
706
707
708
Integer input item expected for list-directed input
Numeric input item expected for list-directed input
Logical input item expected for list-directed input
Complex input item expected for list-directed input
Character input item expected for list-directed input
Incorrect number of bytes read or written to direct-access
unformatted file
Substring index range error
Unable to perform FCHAIN
1000+
Compiler debug error messages - should never appear in
correct programs.
These normally are an indication that
the wrong file was specified as the input to the code generator.
Page 142
FORTRAN Reference Manual
Appendix B
Intrinsic Functions
Appendix B - Intrinsic Functions
+------------+------------+---------+----------+----------------------+
Intrinsic
Number of
Generic
Specific
Type of
I
Function
I
Arguments
I
Name
I
Name
I
Argument
I
Function
I
~~~:~:~
Integer
I
+------------+------------+---------+----------+----------+-----------+
~~n~~~:!~~
I
1
I
INT
I
I
~:ix
IDINT
I
:::i
Double
I
+------------+------------+---------+----------+----------+-----------+
~~n~:~~ion
1
REAL
~~iT
~~~:~:~ :::i
.
SNGL
Double
Real
+------------+------------+---------+----------+----------+-----------+
Integer
Double
I
I
I
I
I
I
I
Conversion
to Double
DBLE
1
Real
Double
Complex
Double
Double
Double
Real
Double
Complex
Complex
Complex
Complex
+------------+------------+---------+----------+----------+-----------+
Integer
Complex
Conversion
to Complex
1 or 2
CMPLX
+------------+------------+---------+----------+----------+-----------+
Conversion I
l '
'ICHAR
I Character \ Integer I
I
I
+------------+------------+---------+----------+----------+-----------+
I
I
I
I
I
I
I+------------+------------+---------+----------+----------+-----------+
to Integer
+------------+------------+---------+----------+----------+-----------+
Conversion I
1
I
I CHAR I Integer I Character I
to Character
Truncation
I
Nearest
Whole
AINT
1
I
1
I
ANINT
AINT
DINT
I
ANINT
DNINT
Real
Double
I
Real
Double
Real
Double
I
Real
Double
I
+------------+------------+---------+----------+----------+-----------+
FORTRAN Reference Manual
Page 143
Appendix B
Intrinsic Functions
+------------+------------+---------+----------+---------------------+
Intrinsic I Number of I Generic I Specific I
Type of
I
I
I
I
+------------+------------+---------+----------+----------+----------+
Function
Arguments
Name
Name
Argument
Function
IDNINT
Double
Integer
lABS
ABS
DABS
CABS
Integer
Real
Double
Complex
Integer
Real
Double
Real
+------------+------------+---------+----------+----------+----------+
Nearest
I 1 I NINT I NINT I Real I Integer I
Integer
Absolute
Value
ABS
1
+------------+------------+---------+----------+----------+----------+
~:~~~;2
MOD
~gD
~~!~ger ~~!~ger
DMOD
Double
Double
+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I
I
~~a~~!~r
I
2
I
SIGN
I
~~~~N
I
DSIGN
I
~~!~ger
I
Double
~~!~ger
I
Double
+------------+------------+---------+----------+----------+----------+
Positive
Difference
I
2
I
DIM
I
I
101M
DIM
DDIM
I
Integer
Real
Double
I
Integer
Real
Double
I
+------------+------------+--------~+----------+----------+----------+
I:~~m~on I
2
I
I
DPROD
I
Real
I
Double
I
+------------+------------+---------+----------+----------+----------+
Choosing
Largest
Value
2 or more
MAX
I ~~l I i~!~ger I i~!~ger I
+----------+----------+----------+
I
I
I
DMAXl
, AMAXO
MAXl
Double
Double
Integer
Real
Real
Integer
+------------+------------+---------+----------+----------+----------+
Choosing
Smallest
Value
2 or more
MIN
I
MINO
AMINl
DMINl
I Real
Integer I Integer I
Real
Double
Double
+----------+----------+----------+
I
I
I
I
+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I
I
+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
Length
1
Index of
Substring
2
AMINO
MINl
Integer
Real
Real
Integer
LEN
Character
Integer
'INDEX
Character \ Integer
.
+------------+------------+---------+----------+----------+----------+
Page 144
FORTRAN Reference Manual
Appendix B
Intrinsic Functions
+------------+------------+---------+----------+---------------------+
Intrinsic I Number of 'GeneriC' Specific I
Type of
I
I
I+------------+------------+---------+----------+----------+----------+
Function
Arguments
Imaginary
Part of
Complex
Argument
Name
Name
AIMAG
Argument
Function
Complex
Real
1
+------------+------------+---------+----------+----------+----------+
Com~lex
I l'
I CONJG I Complex I Complex I
I
ConJugate
+------------+------------+---------+----------+----------+----------+
:~~~re
1
SORT
~~~iT
~~~~le
~~~~le
CSORT
Complex
Complex
+------------+------------+---------+----------+----------+----------+
~~~~nen1
EXP
I ~~~p
I ~~~~le I .~~~~le
CEXP
Complex
Complex
I
+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I
I
I
I
I
~~~~~~~hm
I
1
I
1
I
I
LOG
I
LOG10
I
I
~g~
CLOG
I
I
~~~~le
Complex
I
I
~~~~le
Complex
I
I
+------------+------------+--~------+----------+----------+----------+
I
Common
Logarithm
ALOG10
DLOG10
Real
Double.
Real
Double
+------------+------------+---------+----------+----------+--~-------+
Sine
I
1
I
SIN
I
I'~~~N
CSIN
I
~~~~le
Complex
I
~~~~le
Complex
I
+I-::::::-----+I-----:------+I-:::-----+I-~~:-----+I-~~:~::---+I-~~:~::---+I
CCOS
Complex
Complex
+------------+------------+---------+----------+----------+----------+
Tangent
I 1 I TAN I TAN I Real I Real I
I
I
+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I
DTAN
Double
Double
DASIN
Double
Double
+------------+------------+---------+----------+----------+----------+
Arcsine
I 1 I ASIN I ASIN I Real I Real I
Arccosine
1
ACOS
Real
Real
.\
DACOS
Double
Double
+---------~--+------------+---------+----------+----------+----------+
FORTRAN Reference Manual
ACOS
Page 145
Appendix B
Intrinsic Functions
+------------+------------+---------+----------+---------------------+
Intrinsic, Number of I Generic , Specific \
Type of
I
I
I+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I+------------+---------+----------+----------+----------+
I
I
I
I
+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I
I+------------+------------+---------+----------+----------+----------+
I
I
I
I
I
I
I+----------~-+------------+---------+----------+----------+----------+
I
I
I
I
I
I
I
Function
Arguments
Name
1
ATAN
Argument
Function
ATAN
DATAN
Real
Double
Real
Double
2
ATAN2
'ATAN2
DATAN2
Real
Double
, Real
Double
Hyperbolic
Sine
1
SINH
SINH
DSINH
Real
Double
Real
Double
Hyperbolic
Cosine
1
COSH
COSH
DCOSH
Real
Double
Real
Double
Hyperbolic
Tangent
1
TANH
TANH
DTANH
Real
Double
Real
Double
Arctangent
Name
j-~~:~~~~~;--j-----:------j---------j-:::------j-:::::::::j-:::::::--j
+------------+------------+---------+----------+----------+----------+
,~~:~~:;lY
I
2'
, LGT
I
Character' Logical
,
+------------+------------+---------+----------+----------+----------+
,~~~::~~~~,
2
.,
, LLE
, Character' Logical
,
+------------+------------+---------+----------+----------+----------+
LeXiCallY,
2
I
, LLT I Character I Logical I
Less Than
+------------+------------+---------+----------+----------+----------+
B.l
'
Notes on the Intrinsic Functions
The INT function truncates real or double precision arguments
towards zero.
If the argument to INT is a complex number, the
function is applied to the real part of the complex argument.
IFIX is the same as INT for real arguments.
If the REAL or DBLE functions are applied to a complex
ment, the result is the real part of the argument.
argu-
CMPLX can have one or two arguments. If there is one argument,
it can be of type integer, real, double precision or complex. If
the argument is of type integer, real or double precision, the
result is a complex value whose real part is that of the argument, and whose imaginary part is zero.
Page 146
FORTRAN Reference
Manu~l
Appendix B
Intrinsic Functions
If CMPLX has two arguments, they must both be of the same type.
The arguments can be of type integer, real or double precision.
The result is a complex value whose real part is the first argument and whose imaginary part is the second argument.
ICHAR converts from character to integer. The first character
in the collating sequence is position a and the last character in
the sequence is 'n'-l, where 'n' is the number of, characters in
the character set.
In the trigonometric functions, all angles are in radians.
Functions of complex arguments yield
principal value of the function.
a
result
which
is
the
The INDEX function returns the index where its second argument
starts in its first argument. If the first argument does not
contain the second argument, or if the second argument is longer
than the first argument, the INDEX function returns a value of
zero.
All arguments in an intrinsic function reference must be of the
same type.
B.2
Restrictions on Ranges of Arguments
When intrinsic functions are referenced by their specific
names, the restrictions on ranges of arguments and results are as
follows:
Remaindering
MOD, AMOD and DMOD are
second argument is zero.
undefined
when
their
Transference of Sign
If the first argument to ISIGN, SIGN or DSIGN
zero, the result is zero.
is
Square Root
SQRT and DSQRT require an argument which is not
less than zero. CSQRT returns a value which is
the principal value and is greater than or equal
to zero. If the real part of the result is zero,
the imaginary part is greater than or equal to
zero.
Logarithms
ALOG, DLOG, ALOG10 and DLOG10 require an argument
greater than zero. The argument to CLOG must not
be (0.0, 0.0). If the real part of the ~rgument
is less than zero and the imaginary part 1S zero,
the imaginary part of the result is 'pi', otherwise the imaginary part of the result lines in
the range:
FORTRAN Reference Manual
Page 147
Appendix B
Intrinsic Functions
-pi < imaginary part <= pi
Arcsine and Arccosine
ASIN, DASIN, ACOS and DCOS requir~ that the absolute value of their argument be not greater than
one. The result of arcsine lies in the range
-pi/2 <= result <= pi/2
and the result. of arccosine lies in
o
B.3
<=
the range
result <= pi
Non Standard Intrinsic Functions and Subroutines
The functions and subroutines described here are non-standard
SVS FORTRAN-77 extensions to the FORTRAN-77 language.
B.3.l
POKE - Store Into Arbitrary Memory Location
The POKE subroutine stores a byte into
location. The interface definition is:
an
arbitrary
memory
SUBROUTINE POKE(IADDR, IVAL)
INTEGER IADDR, IVAL*l
The POKE subroutine sets the memory location addressed by IADDR
to the ~yti value of the variable IVAL. IVAL must be a variable
that is ec ared INTEGER*l in the calling procedure or POKE will
not work as expected.
B.3.2
IPEEK - Read From Arbitrary Memory Location
The IPEEK function gets a byte from an arbitrary
tion. The interface definition is:
memory
loca-
INTEGER*4 FUNCTION IPEEK(IADDR)
INTEGER IADDR
IADDR is the address of a memory location. The IPEEK function
returns the signed value of the byte stored at that location.
B.3.3
VERS - Print Date and Version
The VERS subroutine prints the date and version of the
FORTRAN-11 run-time system. The interface definition is:
SUBROUTINE
Page 148
SVS
VERS
FORTRAN Reference Manual
Intrinsic Functions
Appendix B
There are no arguments to the VERS subroutine.
B.3.4
RAN - Random Number Generator
The function RAN generates· pseudo random numbers in the
val [0.0,1.0). The definition is:
inter-
REAL FUNCTION RAN(I)
INTEGER*4 I
If the value of the parameter 'I' is zero, then a new random
result is returned. If 'I' is greater than zero, a new sequence
of random numbers is stated, and the first value in that sequence
is returned.
If 'I' is negative, the same number that was
returned for the last call to RAN is returned.
B.3.5
IARGC - Number of Arguments
The function IARGC returns the number of command line arguments
passed to the program. Its declaration is:
INTEGER*4 FUNCTION IARGC()
INTEGER*4 FUNCTION IARGC()
The exact meaning of the value returned depends
operating syst~m under which the program is running.
B.3.6
upon
the
GETARG - Access an Argument
The subroutine GETARG is used to fetch
line arguments. The form is:
the
value
of
command
SUBROUTINE GETARG(I,C)
INTEGER*41
CHARACTER*(*)C
The value of the I'th command line argument is returned in the
variable C. If there is no argument corresponding to 'I', then C
is set to blanks. If the length of the argument is greater than
the length of C, then only that part tpat fits is returned, and
if it is smaller, than the rest of C is filled with blanks.
FORTRAN Reference Manual
Page 149
Intrinsic Functions
Page 150
Appendix B
FORTRAN Reference Manual
Appendix C
Data Representations
Appendix C - Data Representations
This appendix describes the ways that SVS FORTRAN represents
data in storage and the mechanisms for passing arguments to subroutines and functions. This appendix is intended as a guide to
those programmers who wish to write modules in languages other
than FORTRAN and have those modules interface to FORTRAN.
C.l
Storage Allocation
This section describes the way in which storage is allocated to
variables of various types.
In general, any word value (a value which occupies 16 bits) is
always aligned on-a-word boundary. Anything larger than a word
is also aligned on a word boundary. Values that can fit into a
single byte are aligned on a byte boundary.
INTEGER, REAL and LOGICAL data types all occupy the same amount
of storage,namely 32 bits (four bytes or two words). DOUBLE
PRECISION occupies 64 bits (eight bytes or four words).
COMPLEX
is represented as a pair of single precision real data values and
so occupies 64 bits (eight bytes or four words). There are provisions for indicating that integer and logical data types occupy
less storage.
INTEGER*l
occupies a bits (one byte),
boundary.
INTEGER*2
occupies 16 bits (two bytes or one word), aligned
on a word boundary.
aligned
on
a
byte
INTEGER and INTEGER*4
occupy 32 bits (four bytes or two words), aligned
on a word boundary.
REAL and REAL*4
occupy 32 bits (four bytes or two words), aligned
on a word boundary. A REAL element has a sign
bit, an a-bit exponent and a 23-bit mantissa. SVS
FORTRAN Reference Manual
Page 151
Data Representations
Appendix C
FORTRAN REAL elements conform to the IEEE standard for reals as defined in the March 1981 Computer magazine. The layout of a REAL element is
shown below.
DOUBLE PRECISION and REAL*8
elements occupy 64 bits (eight bytes or four
words), aligned on a word boundary. A DOUBLE
PRECISION element has a sign bit, an ll-bit
exponent and a 52-bit mantissa. SVS FORTRAN DOUBLE PRECISION elements conform to the IEEE standard for double precision data as defined in the
March 1981 Computer magazine. The layout of a
DOUBLE PRECISION element is shown below.
COMPLEX
elements are represented by two REAL elements.
The first element represents the real part of the
number, the second represents the imaginary part.
LOGICAL*l
occupies one byte (8 bits) of storage, aligned on
a byte boundary. A value of 0 represents the
value .FALSB. • A value of 1 represents the
value .TRUB. • Any other value is an "undefined"
logical value.
LOGICAL*2
occupies two bytes (16 bits) of storage, aligned
on a word boundary. A value of 0 represents the
value .FALSB.. A value of 1 represents the
value .TRUB. • Any other value is an "undefined"
logical value.
LOGICAL and LOGICAL*4
occupies four bytes (32 bits) of storage, aligned
on a word boundary. A value of 0 represents the
value .FALSB. • A value of 1 represents the
value .TRUE. • Any other value is an "undefined"
logical value.
C.2
Data Representations
Whatever the size of the data element in question, the most
significant bit of the data element is always in the lowest numbered byte of however many bytes are required to represent that
object. The diagrams below should clarify this.
Page 152
FORTRAN Reference Manual
Appendix C
C.2.1
Data Representations
Representation of Integers
bit --> 7
0
INTEGER*l
+--------+
I byte 0 I
+--------+
INTEGER*2
+--------+--------+
I byte 0 I byte 1 I
+--------+--------+
INTEGER*4
+--------+--------+--------+--------+
I byte 0 I byte 1 I byte 2 I byte 3 I
+--------+--------+--------+--------+
o
15
o
31
C.2.2
Representation of REAL and DOUBLE PRECISION
REAL and DOUBLE PRECISION data elements are represented according to the proposed IEEE standard described in Computer magazine
of March, 1981. The diagrams below illustrates the representation.
31
30
o
23 22
+---+----------------+----------------------------------------+
I S I
Exponen~
I
Mantissa
I
+---+----------------+----------------------------------------+
I
I
Exponent, biased by 127
M!ntissa (23 + 1 bits)
Sign
~
62
63
Representation
52 51
o
+---+----------------+----------------------------------------+
I S I
Exponent
I
Mantissa
I
+---+----------------+----------------------------------------+
I
M~ntissa (52 + 1 bits)
I
Exponent, biased by 1023
Sign
FORTRAN Reference Manual
Page 153
Appendix C
Data Representations
DOUBLE PRECISION Representation
The parts of REAL and DOUBLE PRECISION numbers are as follows:
a one-bit sign bit designated by "S" in the diagrams above.
The sign bit is a 1 if, and only if, the number is negative.
a biased exponent. The exponent is eight bits for a REAL
number, and is eleven bits for a DOUBLE PRECISION number. The
values of all zeros, and all ones, are reserved values for
exponents.
a normalized mantissa, with the high-order 1 bit "hidden".
The mantissa is 23 bits for a REAL number, and is 52 bits for
a DOUBLE PRECISION number. A REAL or DOUBLE PRECISION number
is represented by the form:
2ex ponent-bias
* l.f
where If' is the bits in the mantissa.
C.2.3
Representation of Extreme Numbers
zero (signed)
is represented by an
mantissa of zero.
exponent
of
zero,
and
a
denormalized numbers
are a product of "gradual underflow".
They
non-zero numbers with an exponent of zero.
form of a denormalized number is:
are
The
2exponent-bias+l * O.f
where If' is the bits in the mantissa.
signed infinity (that is, affine infinity) is represented by the
largest value that the exponent can assume (all
ones), and a zero mantissa.
When infinity is
printed by a FORTRAN program, it appears as
either plus '+++.+++' or minus signs '---.---'
depending on the sign.
Not-a-Number (NaN)
is represented
exponent can
mantissa. The
ted printing
question marks
Page 154
by the largest value that the
assume (all ones), and a non-zero
sign is usually ignored.
Formatof NaN appears as a sequence of
'111.111'.
FORTRAN Reference Manual
Appendix C
Data Representations
Normalized REAL and DOUBLE PRECISION numbers are said to contain a "hidden" bit, providing for one more bit of precision than
would normally be the case.
C.2.4
Hexadecimal Representation of Selected Numbers
+-----------+----------+------------------+
I Value I REAL I
PRECISION
I
+-----------+----------+------------------+
+0
00000000
0000000000000000
-0
80000000
8000000000000000
+1.0
-1.0
3F800000
BF800000
3FFOOOOOOOOOOOOO
BFFOOOOOOOOOOOOO
+2.0
+3.0
40000000
40400000
4000000000000000
4008000000000000
+Infinity
-Infinity
7F800000
FF800000
7FFOOOOOOOOOOOOO
FFFOOOOOOOOOOOOO
NaN
7F8xxxxx
7FFxxxxxxxxxxxxx
+-----------+----------+------------------+
C.2.S
Deviations from the Proposed IEEE Standard
Deviations from the proposed IEEE standard in this
tion are as follows:
implementa-
affine mapping for infinities,
normalizing mode for denormalized numbers,
rounds approximately to nearest - 7 or
computed, but the "sticky" bit is not,
more
guard
bits
are
exception flags are not implemented,
conversion between binary and decimal is not implemented.
C.2.6
Arithmetic Operations on Extreme Values
This subsection describes the results derived from applying the
basic arithmetic operations on combinations of extreme values and
ordinary values.
No traps or any other exception actions are taken.
All inputs are assumed to be positive.
Overflow,
and cancellation are assumed not to happen.
FORTRAN Reference Manual
underflow,
Page 155
Data Representations
Appendix C
In all the tables below, the abbreviations have
meanings:
the
following
+--------------+---------------------------------+
I Abbreviation I
Meaning
I
+--------------+---------------------------------+
Den
Denormalized Number
Num
Inf
NaN
Uno
Normalized Number
Infinity (positive or negative)
Not a Number
Unordered
+--------------+---------------------------------+
+---------------------------------------------------+
I
Addition and Subtraction
I
+-----------+---------------------------------------+
Left
I
Right Operand
I
Operand
0
I Den
I Num
I
Inf
I NaN
+-------+-------+-------+-------+-------+
o
I 0
I Den I Num I Inf I NaN I
+-----------+-------+-------+-------+-------+-------+
I
Den
I Den I Den I Num I Inf I NaN I
+-----------+-------+-------+-------+-------+-----~-+
I
Num
I
Num
I
Num
I
Num
I
Inf
I
NaN
I
+-----------+-------+-------+-------+-------+-------+
I
Inf
I Inf I Inf I Inf I Note 11 NaN I
+-----------+-------+-------+-------+-------+-------+
I
NaN
I NaN I NaN I NaN I NaN I NaN I
+-----------+-------+-------+-------+-------+-------+
Note 1: Inf + Inf
= Inf;
Inf - Inf = NaN
+---------------------------------------------------+
I
Multiplication
I
+-----------+---------------------------------------+
Left
\
Right Operand
I
Operand
o
0
I
Den
I
Num
I
Inf
I
NaN
+-------+-------+-------+-------+-------+
0
0
0
NaN
NaN
I
I
I
I
I
I
+-~---------+-------+-------+-------+-------+-------+
I
Den
I
0
I
0
I Num
I Inf
I NaN
I
+-----------+-------+-------+-------+-------+-------+
Num
0
Num
Num
Inf
NaN
+-----------+-------+-------+-------+-------+-------+
Inf
NaN
Inf
Inf
Inf
NaN
+-----------+-------+-------+-------+-------+-------+
NaN
NaN
NaN
NaN
NaN
NaN
+-----------+-------+-------+-------+-------+-------+
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Page 156
FORTRAN Reference Manual
Data Representations
Appendix C
+---------------------------------------------------+
I
Division
I
+-----------+---------------------------------------+
Left,
Right Operand
I
Operand
0
I
Den
I
Num
I
lnf
I
NaN
+-------+-------+-------+-------+-------+
o
NaN
0
0
0
NaN
+-----------+-------+-------+-------+-------+-------+
I
Den
I lnf I Num I Num I 0 I NaN I
+-----------+-------+-------+-------+-------+-------+
Num
lnf
Num
Num
0
NaN
+-----------+-------+-------+-------+-------+-------+
lnf
Inf
lnf
lnf
NaN
NaN
+-----------+-------+-------+-------+-------+-------+
NaN
NaN
NaN
NaN
NaN
NaN
+-----------+-------+-------+-------+-------+-------+
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
+---------------------------------------------------+
I
Comparison
I
+-----------+---------------------------------------+
Left,
Right Operand
I
Operand
0
I
Den
I
Num
I
lnf
I
NaN
+-------+-------+-------+-------+-------+
o
=
<
<
<
Uno
+-----------+-------+-------+-------+-------+-------+
Den
>
<
<
Uno
+-----------+-------+-------+-------+-------+-------+
Num
>
>
<
Uno
+-----------+-------+-------+-------+-------+-------+
I
lnf
I > I > I > I
I Uno I
+-----------T-------+-------+-------+-------+-------+
I
NaN
I Uno I Uno I Uno I Uno I Uno I
+-----------+-------+-------+-------+-------+-------+
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Notes:
NaN compared with NaN is Unordered, and also results in inequality.
+0 compares equal to -0.
FORTRAN Reference Manual
Page 157
Appendix C
Data Representations
+---------------------------------------------------+
I
Max
I
+-----------+---------------------------------------+
Left
\
Right Operand
I
Operand
0
I
Den
I
Num
I
lnf
NaN
+-------+-------+-------+-------+-------+
0
Den
Num
lnf
NaN
+-----------+-------+-------+-------+-------+-------+
Den
Den
Den
Num
lnf
NaN
+-----------+-------+-------+-------+-------+-------+
Num
Num
Num
Num
lnf
NaN
+-----------+-------+-------+-------+-------+-------+
I
lnf
I lnf I lnf I lnf I lnf I NaN I
+-----------+-------+-------+-------+-------+-------+
NaN
NaN
NaN
NaN
NaN
NaN
+-----------+-------+-------+-------+-------+-------+
o
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
+---------------------------------------------------+
I
Min
I
+-----------+---------------------------------------+
Left
I
Right Operand
I
Operand
0
I
Den
I
Num
I
lnf
I
NaN
+-------+-------+-------+-------+-------+
o
0
0
0
0
NaN
+-----------+-------+-------+-------+-------+-------+
I
Den
I 0 I Den I Den I Den I NaN I
+-----------+-------+-------+-------+-------+-------+
I
Num
I 0 I Den I Num I Num I NaN I
+-----------+-------+-------+-------+-------+-------+
lnf
0
Den
Num
lnf
NaN
+-----------+-------+-------+-------+-------+-------+
NaN
NaN
NaN
NaN
NaN
NaN
+-----------+-------+-------+-------+-------+-------+
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Page 158
FORTRAN Reference Manual
Appendix C
C.2.7
Data Representations
Representation of Logicals
bit --> 7
o
+--------+
LOGICAL*1
I byte 0 I
+--------+
15
LOGICAL*2
LOGICAL*4
C.2.B
o
+--------+--------+
I byte 0 I byte 1 I
+--------+--------+
o
31
+--------+--------+--------+--------+
I byte 0 I byte 1 I byte 2 I byte 3 I
+--------+--------+--------+--------+
Storage of Arrays
Arrays are stored with their elements in a specific storage
order.
The elements are actually stored in a linear sequence of
storage elements.
FORTRAN arrays are stored in column major order, such that
first subscript in a multi-dimensional array varies fastest.
position of an arbitrary element in an array is:
the
The
l+SUM( (Si-Li*PRODUCT(Uj-L j »
where 'SkI is the value of the subscript expression specified for
dimension bound of dimension eke. The subscript 'j' in the product above varies between land 'i'-l for any given dimension.
C.3
Argument Passing Mechanism
This section describes the way in which arguments are passed in
SVS FORTRAN.
All arguments to FORTRAN subroutines and functions are passed
by reference.
For every argument except a CHARACTER object, a
32-bit pointer to the object is pushed onto the stack.
When CHARACTER objects are passed in FORTRAN-77, a 32-bit
pointer to the CHARACTER object is pushed onto the stack, followed by a l6-bit value which is the length of the CHARACTER
object.
Pointers to actual arguments are pushed onto the stack in the
order in which they are declared in a subroutine or function
FORTRAN Reference Manual
Page 159
Appendix C
Data Representations
declaration.
Actual arguments which are expressions are evaluated before the
subroutine or function call.
The result of the expression is
assigned to a temporary storage area and a pointer to the temporary is pushed onto the stack. Normally 4 bytes are utilized
for the temporary created to store a numeric expression (8 bytes
for double precision). The number of bytes utilized will be different if the $INT2 option is set. In this case, integer expressions (including constants) are placed into 2 byte temporaries
and logical expressions (including constants) are placed into 1
byte temporaries.
In the exit code of a procedure, all arguments
from the stack before the routine returns.
C.4
are
discarded
Function Results
Functions return their values in register DO (or 00/01 for double precision and complex return values).
C.S
Register Conventions
Registers AO, Al, DO, 01, and 02 are available· as scratch
registers in called routines. That is, they may be clobbered by
functions and subroutines. All other registers must be preserved
across calls. In addition, register A4 and AS must contain their
original values whenever any external routine is called.
A4 is
used in addressing external entry points and AS is used to access
the standard input and output, command line arguments, etc.
Page 160
FORTRAN Reference Manual
Appendix D
ASCII Character Set Table
Appendix D - ASCII Character Set Table
+-------------+-------------+-------------+--------------+
I hex char I hex char ~ hex char I
hex char I
+-------------+-------------+-------------+--------------+
00
NUL
20
SP
40
@
60
01
02
03
04
05
06
07
08
09
OA
OB
OC
OD
OE
OF
SOH
STX
ETX
EOT
ENQ
ACK
BEL
BS
HT
LF
VT
FF
CR
SO
S~
21
I
22"
23.
24
$
25'
26
&
27'
28
(
29)
2A
*
2B
+
2C
20
2E.
2F
/
41
42
43
44
45
46
47
48
49
4A
4B
4C
40
4E
4F
11
12
13
14
15
16
17
18
19
A
B
C
D
E
DC1
DC2
DC3
DC4
NAK
SYN
ETB
CAN
EM
SUB
ESC
FS
GS
RS
US
31
1
32
2
33
3
34
4
35
5
36
6
37
7
38
8
39
9
3A
3B
3C
<
3D:I:
3E
>
3F?
51
0
52
R
53
S
54
T
55
U
56
V
57
W
58
X
59
Y
SA
Z
5B
[
5C
\
50)
5E"
SF
F
G
H
I
J
K
L
M
N
0
61
62
63
64
65
66
67
68
69
6A
6B
6C
60
6E
6F
a
b
c
d
e
f
9
h
i
j
k
1
m
n
0
+-------------+-------------+-------------+--------------+
10
DLE
30
0
50
P
70
P
LA
1B
1C
1D
IE
1F
71
72
73
74
75
76
77
78
79
7A
7B
7C
70
7E
7F
q
r
s
t
u
v
w
x
Y
z
1
DEL
+-------------+-------------+---------=---+--------------+
FORTRAN Reference Manual
Page 161
ASCII Character Set Table
Page 162
Appendix 0
FORTRAN Reference Manual
Appendix E
Operating the SVS FORTRAN System
Appendix B - Operating the SVS PORTRAN Systea
This appendix will describe those characteristics of the SVS
PORTRAN system which are similar among the various environments
in which the system operates. The appendix which follows this
one describes the implementation specific details of the PORTRAN
system under your operating system.
The information in this
appendix describes the PORTRAN system in the form it is released
by SVS. Some of the vendors of the system provide additional
utilities which can used in conjunction with SVS PORTRAN and
which may alter the appearance of the system.
E.I
System Components
In order to most effectively utilize the SVS PORTRAN system, it
is necessary to understand the function and operation of its
various components. In all environments a completely straight
forward procedure is provided for compiling and executing simple
FORTRAN programs (see next appendix). The information provided
here, will only be necessary for more complicated situations
involving separate compilation and multiple source languages.
E.l.l
Compiler Front End
FORTRAN source programs (actually PORTRAN compilation units
consisting of one or more procedures with or without a main program and/or a block data subprogram) are accepted by the compiler
front end, syntax checked, and an intermediate representation of
the program (compilation unit) is written to a file.
This file
is passed to the code generator which generates object code. The
input source program may Rinclude" other files (see Chapter 13).
In addition to the input source file, the FORTRAN compiler front
end accepts certain directives from the command line, which are
described in the ·Command Line Directives and Compiler Options R
section of this appendix.
Input files to the PORTRAN front end generally are files with
names which end in ·.for R, although this differs among operating
environments. The output file from the FORTRAN compiler front
end is an intermediate representation of the program which is
placed in a file which generally ends in R.i".
There is
FORTRAN Reference Manual
Page 163
operating the SVS FORTRAN ,System
Appendix E
virtually nothing which can be done with this ".i" file except
provide it as input to the code generator.
E.l.2
Code Generator
The code generator for the FORTRAN system accepts as input
the
".i" file produced by the front end and generates linkable object
code in a file with a name which generally ends in ".obj".
The same code generator is utilized in compiling SVS Pascal,
FORTRAN, and SVS C and the resulting ".obj" files are linkable providing the applicable rules are followed.
SVS
E.l.3
Linker
A utility is provided with SVS FORTRAN for linking ".obj" files
with each other and with run time libraries which are part of the
language system. The linker is highly specific to the operating
environment and its operation is described in detail in the following appendix. There is, however, certain general information
which applies to all of the linkers.
Each linker accepts as inputs ".obj" files and produces an output which is acceptable to the operating system as an object
file. In some operating environments, the linker's output file
is further linkable in the target environment with object code
generated by the operating system assembler, etc. In all cases,
the linker may be run only once per executable image. The input
to the linker must contain exactly one main program but may contain many object files derived from separate compilations.
E.1. 4
Librar ies
Object files in ".objn format mayor may not be libraries. The
result of a run of the code generator is an ".obj" which is not a
library, although it is possible that such a file contains object
code with corresponds to many subroutines. The main difference
between ".obj" files which are libraries and those which are not
libraries is that the linker includes all of the object code from
non-library input files but only that object code which is referenced from library input files. The determination of what is
referenced is made based on unresolved external code references
in previous input files to the linker. Therefore the order that
files are presented to the linker is important.
When linking FORTRAN programs, the two run time libraries provided with the system must be among the input files to the
linker. One of these libraries is paslib.obj which is a library
which is common for SVS Pascal, SVS FORTRAN, and SVS C. It contains a variety of low level support routines used by all three
Page 164
FORTRAN Reference Manual
Operating the SVS FORTRAN System
Appendix E
languages.
It must be linked as the last w.objW input file to
the linker. The other library is ftnlib.obj which contains run
time support specific to the FORTRAN language and which must be
linked as the second to last w.obj" input file to the linker.
E.l.S
Error Messages
The FORTRAN system contains a file of compile time error messages.
If this file is given the appropriate name, the compiler
will generate English error messages along with error numbers.
If not, the compiler will only give error numbers. The FORTRAN
system also contains a file of run time error messages. If this
file is given the appropriate name, most run time errors detected
in FORTRAN application programs will print English error messages
in addition to the run time error number. The names to be given
to these two error message files differs from one implementation
to another and can be found by referring to the following appendix.
E.2
Command Line Directives and Compiler Options
The FORTRAN compiler front end is invoked to compile a source
file named "prog.for w (other file name endings required on other
systems) with a command line of the form:
fortran prog.for
{options ••• }
Any number of command line options may appear and they may appear
in any order. The possible command line options are:
+q -q
Show more (-q) or less (+q) information on the
progress of the compile to the user. The default
setting varies among different implementations.
+p -p
Prompt (+p) or don't prompt (-p) to the standard
input in the case of a compile time error. The
default setting varies among different implementations.
Prompting mode is useful so that error
messages do not fly off CRT screens but is awkward when compiling in background mode.
+x
Generate a cross reference in the listing file.
Same as setting the $XREF option (see Chapter
13) •
+c72
Truncate input lines to 72 columns. Same as setting the $COL72 option (see Chapter 13).
+f -f
Generate code for the Sky floating point hardware
board (+f) or generate code for software floating
point (-f).
This option is only enabled in
FORTRAN Reference Manual
Page 165
Appendix E
Operating the SVS FORTRAN System
systems which support the Sky board and will
result in an error if not enabled.
The default
is -f, no floating point hardware. Note: If the
Sky floating point hardware interface is to be
used,. the entire program must be compiled with
the +f flag set and the resulting object code
must be linked with sky.paslib.obj instead of
paslib.obj.
-lfname
Create a listing file of the
the file named fname.
source
-efname
Place a summary of the
file named fname.
-ifname
Name the ".i" file fname. If this option is not
provided, the ".i" file when compiling a source
program named prog.for is named progoi.
compile
time
program
in
errors
on
Under certain operating systems the code generator is directly
invoked by the FORTRAN compiler front end. In this case, there
is an additional command line option.
-of name
Name the ".obj" file fname. If this option is
not provided, the ".obj" file when compiling a
source program named prog.for is named prog.obj.
Under systems in which the code generator is not directly
invoked by the FORTRAN compiler front end, the code generator is
invoked using a command of the form:
code prog.i
{optionalfname}
where leaving off the optional file name results in an output
file named prog.obj. If the optional file name is provided, the
output file is named optionalfname.
See the appendix which follows for a description
line arguments and options related to the linker.
E.l
of
command
Linking Programs which Utilize Pascal and C
There are certain rules which must be observed by programmers
wishing to combine object code compiled under more than one
language processor. Throughout the following discussion, Pascal,
FORTRAN, and C refer to the SVS implementations of these
languages.
Page 166
FORTRAN Reference Manual
Operating the SVS FORTRAN System
Appendix E
E.3.l
What Language must Supply the Main Program
In all cases in which FORTRAN code is present, the main program
must be FORTRAN.
In the case where Pascal and C are to be
present, either language may supply the main program. If the C
system is not SVS C, then the main program must be Pascal.
E.3.2
Referring to the Command Line Arguments
In all cases in which the command line arguments are to be
referenced from C, C must provide the main program. This is a
consequence of the fact that command line arguments are "parameters" to the C main routine. Command line arguments are available from Pascal and FORTRAN regardless of which language provides the main program.
E.3.3
Dynamic Memory Allocation and Deallocation
A program may utilize the C library memory allocation and deallocation package (malloc, free, etc.) providing that Pascal components of the program do not call release.
Similarly, Pascal
components should not call release if FORTRAN components performing any I/O are present. If the C system is not SVS C, then the
C routines must not utilize any dynamic memory. allocation or
deallocation directry-or through the operating system run time
library.
E.3.4
Paramete~
Conventions
The calling convention in C is such that parameters are pushed
in "reverse" order from the order in which they appear and the
calling routine is responsible for popping parameters off the
stack after the call returns. Pascal and FORTRAN push parameters
in order and the exit code of the called routines is responsible
for popping off its parameters. Pascal contains a "cexternal"
declaration (similar to Pascal "external") which generates calls
to C routines in which the parameters are popped off at the calling site after the subroutine returns.
The parameters must
appear in reverse order in the Pascal call as compared to the
order expected by C. There is no direct language support for
calling C from FORTRAN or Pascal and FORTRAN from C, but parameterless routines or assembly language interfacing routines can be
utilized for these purposes. It is often easiest to go through
Pascal when calling C from FORTRAN (a complete explanation of
which can be found in the Pascal Reference Manual) •
FORTRAN Reference Manual
Page 167
Operating the SVS FORTRAN System
E.3.4.1
Appendix E
Calling FORTRAN from Pascal
It is straight forward to call FORTRAN subroutines from Pascal.
The called routines should be declared to be external in the
Pascal compilation with formal parameter declarations which match
PORTRAN parameter conventions. In particular, Pascal var parameters will match the FORTRAN call by reference convention. If the
receiving FORTRAN routine expects a character parameter, it will
be necessary to pass the length of the packed array of char as an
explicit two byte value parameter (as described in the parameter
passing section of the FORTRAN reference manual).
Note: Pascal
strings are not compatible with the FORTRAN character datatype.
E.3.4.2
Calling Pascal from FORTRAN
When calling an external routine from FORTRAN, it is merely
invoked without any special declaration. This called routine may
have been written in Pascal. In the event that it is, the routine should be written with formal parameters declared in the
manner which is consistent with what FORTRAN would expect from a
receiving routine written in FORTRAN. Pascal formal parameter
declarations are adequate for expressing all of the interfaces
expected by FORTRAN calling sites.
E.3.4.)
Calling C from FORTRAN
FORTRAN programs call external routines without declaration as
functions in expressions and as procedures in CALL statements.
fORTRAN generally passes all parameters by reference, so the
receiving routine should expect pointer parameters. Assuming a
fORTRAN function call as illustrated below:
100
INTEGER I,J
DOUBLEPRECISION 0
I a CFUNCT2(I,J,D)
the receiving C function might be as follows
cfunct2(d,j,i)
int
*i,*j:
double
*d·,
{
if (*d =a 0.0) return(*i+*j): else return(*i-*j):
}
An assembly language interfacing routine, called a "wrapper",
will be necessary to provide a proper interface between the calling site and the C routine since FORTRAN has no way of knowing
to pop the parameters off at the calling site. The wrapper would
be as follows:
Page 168
FORTRAN Reference Manual
Operating the SVS FORTRAN System
Appendix E
.text
.globl
.globl
CFUNCT2:
movl
jsr
addl
movl
rts
savera: .bss
CFUNCT2
cfunct2
sp@+,savera
cfunct2
t12,sp
savera,-sp@
• • • + 4
There are several important points to note: The FORTRAN external
reference is in upper case letters whereas the C entry point is
in the same upper/lower case letters as specified in the C source
code.
Under some operating systems, the C entry point will
require a prepended underscore to adhere to the conventions in
that environment. The wrapper will not work if the interlanguage
call is recursive, although a more sophisticated version of the
wrapper can be made to work in this situation. The primary role
of the wrapper is to pop off the 12 bytes of parameters (3
pointers) which PORTRAN expects to be popped off by the called
routine and which C expects to be popped off by the caller.
The above procedure is not guaranteed ~ work with £ systems
other than SVS C since the parameter, register, and return value
location-conventIons are not necessarily the same in other C
implementations.
In general, these incompatibilities can be
adjusted for by enhancin9 the wrapper.
The exact syntax of the assembly language will vary from system
to system.
In general the object code for wrappers is linked
into the executable program at the last linkin9 step of the compile.
Normally, a wrapper is required for each FORTRAN to C
call.
Is is particularly difficult to pass character variables from
FORTRAN to C since C has no method of receiving a two byte value
parameter correspondin9 to the length portion of the character
parameter.
E.3.4.4
Calling PORTRAN from C
When calling FORTRAN from C, the actual parameters should
evaluate to pointers to properly map into the FORTRAN reference
parameter conventions. There is no way to tell the C system that
an external reference is to a non C routine. Therefore, assuming
that i and j are 4 byte integers and that d is an eight byte
floating point variable, a C call of the form:
i = ifunc(&d,&j,&i)1
would require an assembly language "wrapper" of the form:
FORTRAN Reference Manual
Page 169
Appendix E
Operating the SVS FORTRAN System
.text
.globl
.globl
ifunc:
movl
jsr
subl
movl
rts
.bss
savera: • s • +
ifunc
IFUNC
sp@+,savera
IFUNC
tl2,sp
savera,-sp@
4
to call a FORTRAN function declared with the header
FUNCTION IFUNC(I,J,O)
DOUBLE PRECISION 0
INTEGER I,J
The important items to note are: FORTRAN entry point is in upper
case, C external reference is in the same case as the programmer
specified. The .globl for the C entry point may need a prepended
underscore on some operating systems. The wrapper will not work
if the inter language call is recursive.
The C calling site
expects to pop off 12 bytes of parameters after the call returns
(3 pointers), but the FORTRAN function has already popped off the
parameters.
Therefore, the wrapper decrements the stack pointer
by the amount the calling site expects to pop off.
The exact syntax of the assembly language will vary from system
to system.
In general the object code for wrappers is linked
into the executable program at the last linking step of the compile.
Normally, a wrapper is required for each C to FORTRAN
call.
The above procedure will not work with C ssstems other than SVS
C -oicause- other C systems expect carIed-su [Outlnes-tO preserve
different registers then FORTRAN functions preserve.
In this
case, the wrapper must be enhanced to preserve the registers
required by the calling C language subroutine.
E.3.5
Run Time Libraries
When linking multiple languages, the last input file provided
to the linker must always be paslib.obj. Immediately preceding
paslib.obj must be clib.obj and ftnlib.obj, in either order. The
former must be present if C is present and the latter must be
supplied if FORTRAN is contained in the program being linked.
Page 170
FORTRAN Reference Manual
Appendix E
E.3.6
Operating the SVS FORTRAN System
Upper and Lower Case External Naming Conventions
It is the convention in Pascal and PORTRAN to upper case all
external names except routine names which are declared cexternal
in Pascal. These names are passed directly to the linker as they
appeared in the cexternal declaration. In C, upper and lower
case letters are distinct, so it is the convention to pass
letters directly through as they were supplied by the programmer.
For interfacing purposes, use upper case names in C, or use
cexternal in Pascal, or use assembly language to bridge the naming conventions.
FORTRAN Reference Manual
Page 171
Operating the SVS FORTRAN System
Page 172
Appendix E
FORTRAN Reference Manual
Appendix F
CPM Operating System Specific Information
Appendix F - CPM Operating System Specific Information
Although the SVS FORTRAN system appears to be almost identical
under a wide variety of operating systems, there are minor
differences, particularly related to the linker and in operating
procedures, among the various environments.
This appendix will
provide the implementation dependent details related to SVS
FORTAN running under the CP/M-68K operating system.
F.l
Compiling
~
Simple Program
Appendix E of this manual described in some detail the
components of the SVS FORTRAN system.
For most FORTRAN programs
the follow i ng simple procedure will be compl ete 1 y adequa te for
compiling, linking, and loading program:
FORTRAN $l.FOR
CODE $1.1
ULINKER -L Sl.O $1.0BJ FTNLIB.OBJ PASLIB.OEJ
ERA $l.OBJ
L068 -S -0 Sl.68K -T10l00 S.O Sl.0 CLIB
To compile a FORTRAN program in a file named 'test.for', using
the above supplied example submit file, type:
FOR TEST
The submit file assumes that the FORTRAN compiler front end
(FORTRAN.68K), the code generator (CODE.68K), the linker
(ULINKER.68K), and the libraries (FTNLIB.OBJ, PASLIB.OBJ, and
CLIB.) reside on the current default disk and directory.
You may
wish to change this to reflect the location of these files on
your system.
FORTRAN Reference Manual
Page 17 3
CPM Operating System Specific Information
Appendix F
The lines of the submit file do the following:
The first two
lines run the front end and code generator on files whose names
are derived from the command line in which the submit file is
invoked.
The linker is run (in its simplest form, see below for
more details) with -1 inhibiting a linkmap listing file, with
output file $1.0, and with three input files, including the SVS
supplied libraries. U1inker produces a file which is then linked
to those CPM system calls which are utilized by the program in
the 1068 step (which invokes the CPM system linker).
P.2
Error Message Files
The SVS FORTRAN system includes two files called ftncterr.src
and ftnrterr.src which should be placed in either the a: or b:
drive. This will allow the compiler to display English messages
for errors which it detects and will allow the FORTRAN run time
system to display English error messages for most detected run
time errors in application programs.
F.3
U1inker
Under CPM, u1inker is the SVS linker. The general operation of
the linker is described in Appendix E.
This section will
describe in detail the modes of operation of u1inker and its load
map listing option.
P.3.1
U1inker Inputs
U1inker links object code in
".obj"
format,
including
libraries.
In addition, u1inker accepts input from the command
line or interactively as described below.
P.3.2
U1inker Outputs
U1inker creates partially linked object code in CPM ".0" format
as its primary output. Optionally, u1inker can produce a listing
file which is a load map of global entry points in the created
".0" file. The form of this map and information contained in it
is best described by the following example with subsequent explanations:
Page 174
FORTRAN Reference Manual
Appendix F
CPM Operating System Specific Information
Example of U1inker Listing File
I,
Linking segment •
I
(4310)
Linking segment
F77RTS' (3376)
MC68000 CPM Object-Code Formatter
05-0ct-83
File: sma11f.o
Memory map for segment
ONE
TWO
SMALLF
$START
, FWRITE
-
':PREAD
- ':PREAO
I
OOOOOE
000120
000174
000174
0001C4
0001F2
ONE
TWO
SMALLF
$START
, FWRITE
001194
00119E
Memory map for segment "_F77RTS'
$2000000 - ERROR
$5000000 - FOUTPUTR
$6000000 - FINPUTRE
0011B4,' F77RTS
0011DA,'-F77RTS
001222,':F77RTS
001E12,'_F77RTS
Static Data Areas:
I'F77RTI
IASC
I
ONE
SMALLF
TWO
No:
o.
at
at
at
at
at
000000
0003C4
000000
0003BO
0003B8
Segment:
I
1. "_F77RTS'
BSS
8SS
Data
BSS
BSS
area
area
area
area
area
relative,
relative,
relative,
relative,
relative,
size 0003BO
size 000008
size oooooe
size 000008
size oooooe
Size:
001006
000030
Start Loc = 000174
Code Size = 001E06
Global Size = 000000
FORTRAN Reference Manual
Page 175
CPM Operating System Specific Information
Appendix F
Explanation of Ulinker Listing File
The listing file was generated from the
gram:
following
FORTRAN
pro-
subroutine one
common/abc/icomml,icomm2
data ii,jj,kk/l,2,3/
write(*,*)ii,jj,kk
icomml • 99
icomm2 • 999
end
subroutine two
common/abc/icomml,icomm2
ii • 1
jj • 2
kk • 3
write(*,*)ii,jj,kk
write(*,*)icomml,icomm2
end
program smallf
i • 17
j • 33 * i
write(*,*)i,j
call one
call two
end
The segment named by 8 blanks had 4310 (decimal) bytes in it.
The segment named , F77RTS had 3376 (decimal) bytes in it. Under
CPM there is no reason for programmers to explicitly deal with
segments, since ulinker handles this automatically.
There were a large number of entry points in the linked files,
most of which were extracted from the run time libraries, only a
few of which are shown above and the remainder have been omitted
in order to keep the example listing short. Three of these entry
points are recognizable as user procedure names.
The addresses
of these entry points are given in hex and are text area relative, but will be further relocated ~ the 1068 step of the compilation: --,ne--relative addresses (dTitance be ween-them) WIII
remain intact through the 1068 step.
For each FORTRAN procedure, for each FORTRAN common area, and for
the FORTRAH run time system there is a static data area listed.
Each such data area is mapped to the data or bss area depending
upon whether the area is initialized at compile time using the
data statement. Note: an area which is partially initialized
using data statements is mapped to the data (as opposed to bss)
Page 176
FORTRAN Reference Manual
Appendix F
CPM Operating System Specific Information
area, even if the area is large and only sparsely initialized.
Initialized data areas are represented in the CPM object file
format as an image and large data areas which are initialized
result in large object files. In the example, the common area is
indicated by having a data area name which begins and ends in
slashes.
Sizes and locations of these data area listings are in
hex and relative to the start of the data or bss area as
appropriate.
F.3.3
Running Ulinker from the Command Line
The command line form of running u1inker is:
ulinker 1istfname outputfname inputfname
{inputfname ••• }
where the optional listing file is created on a file named
1istfname providing that 1istfname is not equal to -1 (no listing
file to be created directive). The command line arguments are
positional.
No file name suffixes are enforced by u1inker in
this mode so complete file names must be entered.
F.3.4
Running U1inker Interactively
It is often not convenient or not possible to have a command
line which is long enough to have all of the input files listed.
In this event, u1inker can be run interactive.
Execute u1inker
without any command line arguments and it prompts:
Listing file
~
Any file name provided creates the listing file.
Enter just
return to suppress the optional listing file. The next prompt
is:
Output file [.0] Ulinker requires an output file. If the file name provided does
not end in w.o·, ulinker will append this file name extension
onto the name which is input.
Following this prompt, u1inker
will repeatedly prompt:
~
Input file [.obj] for its input files, until a plain return is typed, indicating
that the input file list is completed. U1inker will append the
·.obj· suffix onto input file names if it is not supplied by the
user.
Running in this mode, there is no limit on the number of
input files which ulinker can process.
FORTRAN Reference Manual
Page.l71
CPM
Operati~g
P.3.S
System Specific Information
Appendix P
Running Ulinker with Standard Input Redirected
With many input files, it is most convenient to operate ulinker
in its interactive mode with standard input redirected. Por
example, run ulinker as follows:
ulinker
Source Exif Data:
File Type : PDF
File Type Extension : pdf
MIME Type : application/pdf
PDF Version : 1.3
Linearized : No
XMP Toolkit : Adobe XMP Core 4.2.1-c043 52.372728, 2009/01/18-15:56:37
Create Date : 2013:05:03 08:40:27-08:00
Modify Date : 2013:05:03 08:16:56-07:00
Metadata Date : 2013:05:03 08:16:56-07:00
Producer : Adobe Acrobat 9.53 Paper Capture Plug-in
Format : application/pdf
Document ID : uuid:7d6eb343-9129-c342-825a-46481cc243cd
Instance ID : uuid:8985f920-8963-c542-a4cf-0022cf70b3d4
Page Layout : SinglePage
Page Mode : UseNone
Page Count : 192
EXIF Metadata provided by EXIF.tools