Fortran_Extended_4_Ref_Man_60497800J_Jun83 Fortran Extended 4 Ref Man 60497800J Jun83

Fortran_Extended_4_Ref_Man_60497800J_Jun83 Fortran_Extended_4_Ref_Man_60497800J_Jun83

User Manual: Pdf Fortran_Extended_4_Ref_Man_60497800J_Jun83

Open the PDF directly: View PDF .
Page Count: 546 [warning: Documents this large are best viewed by clicking the View PDF Link!]

60497800

/ff^y

CONTROL DATA

CORPORATION

FORTRAN EXTENDED

VERSION 4

REFERENCE MANUAL

CDC® OPERATING SYSTEMS:

NOS 1

NOS 2

NOS/BE 1

SCOPE 2

REVISION RECORD

/f*3S*j\

Revision

A (11/01/75)

B (03/05/76)

C (04/15/77)

0 (03/31/78)

E (07/20/79)

F (08/22/80)

G (01/15/81)

H (08/13/82)

J (06/10/83)

Description

Original release.

This revision documents Version 4.6 of FORTRAN Extended. Features documented include

CP155, Compiler Enhancements, and CP079, Math Library Upgrade.

Revised to include feature F7540, CYBER 170 Model 176 Support, as well as miscellaneous

technical corrections, at PSR level 446.

This revision documents Version 4.7 of FORTRAN Extended. Features documented include

CP091 and CP162, CRM products BAM and AAM, 191, Math Library Upgrade, CP184, Fast Overlay

Loading, and 66, CYBER Interactive Debug interface. Also documented is the implementa

tion of STATIC mode memory management, as well as miscellaneous technical changes and

corrections.

This revision documents Version 4.8 of FORTRAN Extended. The Post Mortem Dump facility

is documented with this release, as well as numerous technical changes.

This revision documents changes to Post Mortem Dump, adds the FORTRAN Interface to Common

Memory Manager, and adds the STATIC option to FORTRAN Extended. Numerous technical

changes are included. PSR level 524.

This revision documents release of Post Mortem Dump and STATIC option under SCOPE 2.

Numerous technical changes are included. PSR level 533.

This revision documents numerous technical changes and corrections. PSR level 552.

This revision documents numerous technical and editorial corrections. PSR level 577.

REVISION LETTERS I, 0, Q, AND X ARE NOT USED

©COPYRIGHT CONTROL DATA CORPORATION

1975, 1976, 1977, 1978, 1979, 1980, 1981, 1982, 1983

Printed in the United States of America

Address comments concerning this manual to:

CONTROL DATA CORPORATION

Publications and Graphics Division

P. 0. BOX 3492

SUNNYVALE, CALIFORNIA 94088-3492

or use Comment Sheet in the back of this manual

n60497800 J

LIST OF EFFECTIVE PAGES

J0B&K, New features, as well as changes, deletions, and additions to information in this manual are indicated by bars

in the margins or by a dot near the page number if the entire page is affected. A bar by the page number

indicates pagination rather than content has changed.

ijfpSN

/*$RN

/0S\

Page Revision Page Revision Page Revision

Front Cover 5-11 thru 5- ■19 8-7

Inside Front Cover 5-20 8-8

Title Page 5-21 8-9

ii 5-22 8-10

iii 5-23 8-11 thru 8-16

iv 5-24 8-16.1/8-16.2

vd5-25 thru 5- ■28 8-17

vi 6-1 8-18

v i i / v i i i 6-2 8-19

ix 6-3 8-20

XJ6-4 8-21

xi 6-5 8-22

xii 6-6 thru (5-8 8-23

1-1 thru 1-4 6-9 8-24

1-5 6-10 8-25

1-6 thru 1-9 6-11 thru 6- ■17 8-26

1-10 6-18 8-26.1/8-26.2

1-11 thru 1-17 6-18.1/6-18.2 8-27

2-1 6-19 thru 6- ■21 8-28

2-2 6-22 8-29

2-3 6-23 8-30

2-4 6-24 8-31

2-5 6-25 thru 6- 27 8-32

2-6 6-28 8-32.1 /8-32.2

2-7 thru 2-20 6-29 8-33 thru 8-35

3-1 thru 3-9 6-30 8-36

3-10 6-31 8-37

3-11 6-32 8-38

3-12 6-33 8-39

3-13 6-34 8-40

3-14 6-35 thru 6- 37 8-41

3-15 7-1 8-42

3-16 7-2 8-43

3-17 7-3 8-44

3-18 7-4 8-45 thru 8-47

3-19 7-5 thru 7-S 8-48 thru 8-51

3-20 7-10 8-52

3-21 7-11 8-53

4-1 7-12 8-54

4-2 7-13 thru 7- 15 8-55

4-3 7-16 8-56

4-4 7-17 8-56.1/8-56.2

4-5 thru 4-9 7-18 8-57 thru 8-61

4-10 7-19 9-1

4-11 7-20 9-2 thru 9-4

4-12 7-21 9-5

4-13 7-22 9-6

4-14 thru 4-16 7-23 9-7

5-1 7-24 thru 7- 28 9-8

5-2 8-1 9-9

5-3 8-2 9-10

5-4 thru 5-6 8-3 9-11 thru 9-29

5-7 8-4 10-1

5-8 8-5 10-2

5-9 8-6 10-3

5-10 8-6.1/8-6.2 10-4

5-10.1

5-10.2

60497800 J i i i •

Revision Page Revision Page Revision

10-5 thru 10-7

10-8

10-9

10-10

11-1

11-2

11-3

11-4

11-5

11-6

11-7

12-1

12-2

12-3

12-4

12-5

13-1

13-2

13-3

13-4

13-5 thru 13-8

13-9

13-10

13-11

13-12

13-13

13-14

13-15

13-16 thru 13-20

13-21

14-1

14-2

14-3

15-1 thru 15-3

16-1

16-2

16-3

16-4

16-5

16-6

16-7

16-8

16-9

16-10

16-11

16-12

16-13 '

17-1

17-2

17-3

17-4

17-5

17-6

18-1

18-2 thru 18-6

18-7

18-8

18-9 thru 18-11

19-1

19-2

19-3

19-4 thru 19-7

19-8

19-9 thru 19-11

19-12

19-13 thru 19-18

19-19

19-20 thru 19-27

.19-28

19-29 thru 19-32

19-33

19-34

20-1

A-l

A-2

A-3

B-l thru B-4

B-5

B-6

B-7

B-8

B-9

B-10

B-ll thru B-14

B-15

B-16

B-17

B-18

B-19

B-20

B-21

B-22

B-23 thru B-26

B-27

B-28

B-29

B-30

B-31 thru B-33

B-34

B-35

B-36

B-37

B-38 thru B-41

B-42

B-43

B-44 thru B-49

B-50

B-51 thru B-64

B-65

B-66

B-67

B-68

B-69 thru B-76

B-77

B-78

B-79

B-80

B-81

B-82

B-83

B-84

B-85

B-86

B-87

B-88

B-89

B-90

B-91

B-92

B-93

B-94

B-95

B-96

B-97

C-l thru C-8

C-9 thru C-ll

D-l

D-2

D-3 thru

D-6

D-7

D-8

E-l thru

Index-1

D-5

E-6

Index-2

Index-3

Index-4

Index-5

Index-6

Comment Sheet

Mailer

Back Cover

yfS&Sv

.■**^^\

• I V 60497800 J

/<*^j.

PREFACE

f This manual describes the FORTRAN Extended 4.8 language. FORTRAN Extended is designed to comply with

American National Standards Institute FORTRAN language, as described in X3.9-1966. It is assumed the reader

has knowledge of an existing FORTRAN language and is familiar with the computer system on which the

/^^ language is used.

The FORTRAN Extended compiler operates in conjunction with the COMPASS 3 assembly language processor

^p^ under control of the following operating systems:

NOS 1 for the CONTROL DATA® CYBER 170 Series, CYBER 70 Models 71, 72, 73, 74, and 6000 Series |

Computer Systems

NOS 2 for the CDC® CYBER 170 Series, CYBER 70 Models 71, 72, 73, 74, and 6000 Series Computer |

Systems

NOS/BE 1 for the CDC CYBER 170 Series, CYBER 70 Models 71, 72, 73, 74, and 6000 Series Computer |

0s\ Systems

SCOPE 2 for the CONTROL DATA CYBER 170 Model 176, CYBER 70 Model 76, and 7600 Computer

/^ Systems

All references in this manual to NOS 1 refer to both NOS 1 and NOS 2.

Due to capsule loading, relocatable binaries compiled by versions of FORTRAN Extended prior to version 4.7

cannot be run with CRM BAM 1.5 or AAM 2; they must be recompiled.

Control Data extensions to the FORTRAN language are indicated by shading. Example programs or parts of

programs are shaded in their entirety if they contain lines using extensions to the ANSI standard (unless the only

such extension is the PROGRAM statement). Shading is used only in sections 1 through 8, which contain the

f speciﬁcation of the FORTRAN Extended language; later sections describe the implementation of these

speciﬁcations and shading is not used.

/*"5Rx

r Extended memory for the CYBER 170 Model 176 is large central memory (LCM) or large central memoryextended (LCME). Extended memory for the CYBER 170 800 Series Computer Systems is uniﬁed extended

memory (UEM). Extended memory for all other NOS or NOS/BE computer systems is extended core storage

(ECS) or extended semiconductor memory (ESM). In this manual, the acronym ECS refers to all forms of

0^ extended memory unless otherwise noted. Programming information for the various forms of extended memory

can be found in the COMPASS reference manual and in the appropriate computer system hardware reference

manual.

\. Related material is contained in the listed publications. These publications are listed alphabetically and

grouped according to their importance to the FORTRAN user. The NOS 1, NOS 2, and NOS/BE 1 manual

abstracts are pocket-sized manuals containing brief descriptions of the contents and intended audience of all

{■^ operating system and product set manuals. The abstracts manuals can be useful in determining which manuals

are of greatest interest to a particular user.

rThe Software Publications Release History is a guide for determining which revision level of software

documentation corresponds to the Programming System Report (PSR) level of installed site software.

/*"fSS

/^n

60497800 J

The following publications are of primary interest:

Publication

Publication Number

FORTRAN Common Library

Mathematical Routines

Reference Manual 60498200

FORTRAN Extended Version 4

DEBUG User's Guide 60498000

FORTRAN Extended Version 4

User's Guide 60499700

NOS Version 1 Reference Manual

Volume 1 of 2 60435400

NOS Version 2 Reference Set

Volume 3, System Commands 60459680

NOS/BE Version 1

Reference Manual 60493800

SCOPE Version 2

Reference Manual 60342600

The following publications are of secondary interest:

Publication

Number

Common Memory Manager

Version 1 Reference Manual 60499200

COMPASS Version 3

Reference Manual 60492600

CYBER Interactive Debug

Version 1 Reference Manual 60481400

CYBER Interactive Debug

Version 1 Guide for Users

of FORTRAN Extended Version 4 60482700

CYBER Loader Version 1

Reference Manual 60429800

CYBER Record Manager

Advanced Access Methods

Version 2 Reference Manual 60499300

CYBER Record Manager

Advanced Access Methods

Version 2 User's Guide 60499400

CYBER Record Manager

Basic Access Methods

Version 1.5 Reference Manual 60495700

CYBER Record Manager

Basic Access Methods

Version 1.5 User's Guide 60495800

NOS1 NOS 2 NOS/BE 1 SCOPE 2

X X

X0£&$$\

NOS1 NOS 2 NOS/BE 1 SCOPE 2 ,<*^*\

X.y<S^^V

/•/5-^V

VI 60497800 H

0^\

Publication

DMS-170

DDL Version 3 Reference Manual

Volume 1: Schema Deﬁnition

for Use With:

COBOL

FORTRAN

Publication

Number NOS1 NOS 2 NOS/BE 1 SCOPE 2

Query Update 60481900

FORTRAN Data Base Facility

Version 1 Reference Manual 60482200

INTERCOM Interactive Guide

for Users of FORTRAN Extended 60495000

INTERCOM Version 5

Reference Manual 60455010

Loader Version 1 User's Guide 60482300

Network Products

Interactive Facility Version 1

Reference Manual 60455250

NOS Version 1 Diagnostic Index 60455720

NOS Version 1 Manual Abstracts 84000420

NOS Version 2 Diagnostic Index 60459390

NOS Version 2 Manual Abstracts 60485500

NOS Version 2 Reference Set

Volume 1, Introduction to

Interactive Usage 60459660

NOS/BE Version 1

Diagnostic Index 60456490

NOS/BE Version 1

Manual Abstracts 84000470

SCOPE Version 2 Loader

Reference Manual 60454780

SCOPE Version 2 Record Manager

Reference Manual 60495700

Software Publications

Release History 60481000

Sort/Merge Versions 4 and 1

Reference Manual 60497500

CDC manuals can be ordered from Control Data Corporation, Literature and Distribution Services,

308 North Dale Street, St. Paul, Minnesota 55103.

This product is intended for use only as described in this document. Control

Data cannot be responsible for the proper functioning of undescribed features

or parameters.

60497800 H vii/viii

/"PN

CONTENTS

1. FORTRAN LANGUAGE ELEMENTS

Coding FORTRAN Statements

FORTRAN Character Set

Column Usage

Comments

Statement Labels

Continuation

Columns 73-80

Statement Separator

Blank Lines

Data

Ordering of Statements

Constants

Integer Constant

Real Constant

Double Precision Constant

Complex Constant

Octal Constant

Hollerith Constant

nHf and #?*

nRf and nLf

Logical Constant

Variables

Integer Variables

Real Variables

Double Precision Variables

Complex Variables

Logical Variables

Arrays

Subscripts

Array Structure

1-1

1-3

1-4

1-5

1-6

1-7

1-8

1-9

1-10

1-11

1-12

1-13

1-15

1-16

IMPLICIT Type Statement

DIMENSION Statement

COMMON Statement

EQUIVALENCE Statement

EQUIVALENCE and COMMON

LEVEL Statement

EXTERNAL Statement

DATA Statement

Implied DO in Data List

4. FLOW CONTROL STATEMENTS

GO TO Statement

Unconditional GO TO Statement

Computed GO TO Statement

ASSIGN Statement

Assigned GO TO Statement

Arithmetic IF Statement

Three-Branch Arithmetic IF Statement

Two-Branch Arithmetic IF Statement

Logical IF Statement

Standard-Form Logical IF Statement

Two-Branch Logical IF Statement

DO Statement

DO Loops

Nested DO Loops

CONTINUE Statement

PAUSE Statement

STOP Statement

END Statement

RETURN Statement

3-3

3-4

3-5

3-8

3-11

3-12

3-14

3-15

3-19

4-1

4-2

4-3

4-4

4-5

4-6

4-7

4-8

4-9

4-12

4-13

4-14

4-15

2. EXPRESSIONS AND ASSIGNMENT

STATEMENTS 2-1

Expressions

Arithmetic Expressions

Evaluation of Expressions

Type of Arithmetic Expressions

Exponentiation

Relational Expressions

Logical Expressions

Masking Expressions

Assignment Statements

Arithmetic Assignment Statements

Conversion to Integer

Conversion to Double Precision

Conversion to Complex

Conversion to Real

Logical Assignment

Masking Assignment

Multiple Assignment

3. SPECIFICATION STATEMENTS

Type Statements

Explicit Type Declarations

INTEGER

REAL

COMPLEX

DOUBLE PRECISION

LOGICAL

2-1 5. INPUT/OUTPUT STATEMENT

2-1

2-2 Formatted Input/Output

2-5 Formatted Output Statements

2-6 PRINT

2-7 PUNCH

2-9 WRITE

2-12 Formatted READ

2-14 Unformatted Input/Output

2-15 Unformatted WRITE

2-16 Unformatted READ

2-16 List Directed Input/Output

2-17 List Directed Input

2-18 List Directed Output

2-18 NAMELIST

2-19 Input

2-19 Output

Arrays in NAMELIST

Buffer Statements

3-1 BUFFER IN

BUFFER OUT

3-1 ENCODE and DECODE

3-1 ENCODE

3-1 DECODE

3-2 File Manipulation Statements

3-2 REWIND

3-2 BACKSPACE

3-3 ENDFILE

5-1

5-2

5-3

5-4

5-5

5-7

5-8

5-10.2

5-13

5-14

5-15

5-17

5-20

5-22

5-25

5-27

60497800 H

/^$$K

6. INPUT/OUTPUT LISTS AND FORMAT Miscellaneous Utility Subprograms 8-8

STATEMENTS 6-1 Random Number Generator 8-8

Operating System Interface Routines 8-9

Input/Output Lists 6-1 Debugging Aids 8-14

Implied DO in I/O List 6-2 Input/Output Status Checking 8-23

FORMAT Statement 6-5 Other Input/Output Subprograms 8-25

Data Conversion 6-6 ECS/LCM/UEM Subprograms 8-26.1/

Conversion Speciﬁcation 6-7 8-26.2

Iw and Iw.z Input 6-7 Terminal Interface Subprograms 8-27

Iw and Iw.z Output 6-8 Mass Storage Input/Output 8-29

Ew.d, Ew.dEe and Ew.dDe Output 6-9 Random File Access 8-29

Ew.d, Ew.dEe and Ew.dDe Input 6-10 Mass Storage Subroutines 8-30

Fw.d Output 6-13 Opening a File 8-30

Fw.d Input 6-13 Writing Records 8-31

Gw.d Input 6-14 Reading Records 8-32

Gw.d Output 6-14 Closing a File 8-32

Dw.d Output 6-16 Specifying a Different Index 8-32.1/

Dw.d Input 6-16 8-32.2

Ow Input 6-17 Index Key Types 8-33

Ow Output 6-17 Master Index 8-33

Zw Input and Output 6-18 Sub-Index 8-33

Aw Input 6-18.1/ Multi-Level File Indexing 8-33

6-18.2 Compatibility With Previous Mass

Aw Output 6-19 Storage Routines 8-39

Rw Input 6-20 FORTRAN-CYBER Record Manager Interface 8-39

Rw Output 6-21 Parameters 8-39

Lw Input 6-21 Subroutines 8-42

Lw Output 6-21 Error Checking 8-44

Scale Factors 6-21 Multiple Index Processing 8-45

Fw.d Scaling 6-22 FORTRAN-Sort/Merge Interface 8-46

Ew.d and Dw.d Scaling 6-23 FORTRAN-CYBER Interactive Debug Interface 8-51

Gw.d Scaling 6-23 Control Statement 8-51

X Speciﬁcation 6-24 User-CID Interaction 8-52

nH Output 6-25 CID Output 8-52

nH Input 6-26 Batch Debugging 8-52

End of Record Slash 6-28 Interface to Common Memory Manager 8-52

Repeated Format Speciﬁcation 6-29 Post Mortem Dump 8-53

Printer Control Character 6-31

Tn Speciﬁcation 6-32

V Speciﬁcation 6-34 9. DEBUGGING FACILITY 9-1

Equals Sign 6-34

Execution Time Format Speciﬁcation 6-36 Debugging Statements 9-3

7. PROGRAM UNITS, PROCEDURES,

AND OVERLAYS

Main Programs

PROGRAM Statement Format

PROGRAM Statement Usage

Block Data Subprogram

Procedures

Subroutine Subprogram

Function Subprogram

Basic External Function

Intrinsic Function

Statement Function

Procedure Communication

Passing Values to a Procedure

Using Arguments

Using Common

Using Arrays

Referencing a Function

Calling a Subroutine Subprogram

Using the ENTRY Statement

Overlays

Overlay Communication

Creating an Overlay

Calling an Overlay

8. FORTRAN EXTENDED SUPPLIED

PROCEDURES

Intrinsic Functions

Basic External Functions

7-1

7-2

7-3

7-5

7-6

7-8

7-9

7-10

7-12

7-14

7-15

7-16

Continuation Line

ARRAYS Statement

CALLS Statement

FUNCS Statement

STORES Statement

Variable Names

Relational Operators

Checking Operators

Hollerith Data

GOTOS Statement

TRACE Statement

NOGO Statement

Debug Deck Structure

DEBUG Statement

AREA Statement

OFF Statement

Printing Debug Output

STRACE Entry Point

10. FTN CONTROL STATEMENT

9-3

9-5

9-7

9-10

9-11

9-12

9-13

9-14

9-15

9-17

9-22

9-23

9-26

9-27

9-28

10-1

7-18 Parameters 10-1

7-19 A Exit Parameter 10-2

7-21 B Binary Object File 10-2

7-21 BL Burstable Listing 10-2

7-23 C COMPASS Assembly

CC Control Statement Continuation 10-2

Parameter 10-2

8-1 D Debugging Mode Parameter 10-3

DB CYBER Interactive Debug Parameter 10-3

8-1 E Editing Parameter 10-3

8-1 EL Error Level 10-4

60497800 J

/<^!v

/*e$!\

>^V

/<SsV

0^*-

ER Error Recovery 10-4

G Get System Text File 10-4

GO Automatic Execution (Load and Go) 10-4

I Source Input File 10-5

L List Output File 10-5

LCM Level 2 and Level 3 Storage Access 10-5

ML Modlevel 10-5

OL Object List 10-5

OPT Optimization Parameter 10-6

P Pagination 10-6

PD Print Density 10-6

PL Print Limit 10-6

PMD Post Mortem Dump 10-6

PS Page Size 10-6

PW Page Width 10-7

Q Program Veriﬁcation 10-7

R Symbolic Reference Map 10-7

ROUND Rounded Arithmetic Computations 10-7

S System Text (Library) File 10-8

SEQ Sequenced Input 10-8

SL Source List 10-8

STATIC Static Loading 10-8

SYSEDIT System Editing 10-8

T Error Traceback 10-9

TS Timesharing Mode 10-9

UO Unsafe Optimization 10-9

X External Text Name 10-9

Z Zero Parameter 10-9

FTN Control Statement Examples 10-10

Time-Sharing Mode

R=l Maps

R=2, R=3 Maps

Common Blocks

Entry Points

External References

Statement Labels

Variables

14. OBJECT CODE

Optimizing Mode

Subroutine and Function Structure

Main Program Structure

Renaming Conventions

External Procedure Names

Listing Format

Time-Sharing Mode

Listing Format

15. EXECUTION CONTROL STATEMENT

Alternate File Name Speciﬁcation

Print Limit Speciﬁcation

Post Mortem Dump Parameters

13-15

13-16

13-17

13-18

13-19

13-20

14-1

14-2

14-3

15-1

15-2

11. COMPILATION MODES AND

OPTIMIZATION 11-1

Optimizing Mode 11-2

/0S\ Object Code Optimization 11-2

OPT=0 11-2

OPT=l 11-2

OPT=2 11-2

UO 11-3

Source Code Optimization 11-4

Time-Sharing Mode 11-6

TS Listings 11-7

/f|S\ Sequenced Line Format 11-7

12. COMPILER LISTINGS 12-1

0^\ Optimizing Mode Listings 12-1

Time-Sharing Mode Listings 12-2

0 S \

Listing Control Directives 12-2

13. CROSS REFERENCE MAP 13-1

/i^N Optimizing Compilation Mode 13-1

Source Program 13-2

R=l Maps 13-3

R=2/R=3 Maps 13-4

0O&\ Entry Points 13-5

Variables 13-6

File Names 13-7

External References 13-8

Inline Functions 13-9

J*$^N Namelists 13-10

Statement Labels 13-10

DO Loops 13-11

Common Blocks 13-12

/""S"N Equivalence Classes 13-13

Program Statistics 13-14

Error Messages 13-14

Debugging (Using the Reference Map) 13-14

16. INPUT/OUTPUT IMPLEMENTATION

Execution-Time Input/Output

File and Record Deﬁnitions

Structure of Input/Output Files

Sequential Files

Mass Storage Input/Output

FILE Control Statement

Sequential File Operations

BACKSPACE/REWIND

ENDFILE

Input/Output Restrictions

Record Manager Error Suppression

Compile-Time Input/Output

Source Input File Structure

Coded Output File Structure

Binary Output File Structure

17. COMPASS SUBPROGRAM LINKAGE

Call by Name and Call by Value

Call by Name Sequence

Call by Value Sequence

Intermixed COMPASS Subprograms

Entry Point

Restrictions on Using Library Function

Names

18. SAMPLE DECK STRUCTURES

FORTRAN Source Program with Control

Statements

Compilation Only

TS Mode Compilation Only

Compilation and Execution

FORTRAN Compilation With COMPASS

Assembly and Execution

Compile and Execute With FORTRAN Sub

routine and COMPASS Subprogram

16-1

16-2

16-6

16-8

16-10

16-11

16-12

16-13

17-1

17-2

17-4

18-1

18-2

18-3

18-4

18-5

60497800 H xi

Compile and Produce Binary Cards

Load and Execute Binary Program

Compile and Execute With Relocatable

Binary Deck

Compile Once and Execute With Different

Data Decks

Preparation of Overlays

Compilation and 2 Executions With Overlays

19. SAMPLE PROGRAMS

Program OUT

Program B

Program MASK

Program EQUIV

Program COME

Program UBS

Program PIE

Program ADD

DECODE (READ)

ENCODE (WRITE)

Program PASCAL

Program X

Program VARDIM

Program VARDIM2

SUBROUTINE IOTA

SUBROUTINE SET

Function AVG

Function PVAL

Function MULT

Main Program - VARDIM2

Program CIRCLE

Program OCON

List Directed Input/Output

20. STATIC OPTION

18-6

18-7 APPENDIXES

A Standard Character Set

18-8 B FORTRAN Diagnostics

C Statement Forms

18-9 D Arithmetic

18-10

18-11

E Glossary

19-1 INDEX

A-l

B-l

C-l

D-l

E-l

/f**"^?*\

19-1

19-3

19-5 FIGURES

19-7

19-9 1-1 Program PASCAL 1-2 s ^ \

19-11 9-1 Example of Interspersed Debugging

19-13 Statements 9-18

19-15 9-2 External Debugging Deck 9-19 /*^\

19-15 9-3 Example of Internal Debugging Deck 9-20

19-15 9-4 Example of External Deck on Separate

19-18 File 9-21

19-19

19-21 /<*S"X

19-23 TABLES

19-23

19-23 2-1 Mixed Type Arithmetic Expressions ^*^\

19-23 With + - * / Operators 2-5

19-24 7-1 Differences Between a Function and

19-25 Subroutine Subprogram 7-1

19-25 7-2 Procedure and Subprogram Inter /Cc£$V

19-28 relationships 7-2

19-30 8-1 Intrinsic Functions 8-2

19-33 8-2 Basic External Functions 8-6

8-3 LABINFO Block Content 8-26 *r*^\

16-1 Defaults for FIT Fields Under FORTRAN

20-1 Extended 16-3

/®ft\

>«"»fe\

Xll 60497800 H

FORTRAN LANGUAGE ELEMENTS

A FORTRAN program contains executable and non-executable statements. Executable statements specify actions

the program is to take, and non-executable statements describe characteristics of operands, statement functions,

arrangement of data, and format of data.

CODING FORTRAN STATEMENTS

The FORTRAN source program is written on the coding form illustrated in ﬁgure 1-1. Each line on the coding

form represents an 80-column source line (terminal line or card image). The FORTRAN character set is used to

code statements.

FORTRAN CHARACTER SET

Alphabetic A to Z

Numeric 0 to 9

Special = equal ) right parenthesis

+ plus , comma

- minus . decimal point

* asterisk $ dollar sign

/ slash blank

(left parenthesis •£ or " quote

In addition, any character (Appendix A) may be used in Hollerith constants and in comments. Blanks are not

signiﬁcant except in Hollerith ﬁelds.

COLUMN USAGE

Column 1 C or $ or * indicates comment line

Columns 1-2 C$ indicates a debug directive if In DEBUG mode.

Columns 1-2 C/ indicates a list directive.

Columns 1-5 Statement label.

Column 6 Any character other than blank or zero denotes continuation; does not

apply to comment lines or list directives. A debug continuation line

must contain C$ in columns 1-2.

Columns 7-72 Statement.

Columns 73-80 Identiﬁcation ﬁeld, not

processed by compiler.

Can contain information for

debug AREA directive.

/*SS\

60497800 A 1-1

,-sS^v

/^^V

j t c

« ta

tc S

Ul 3

ﬁ

IN lil

u.£

Z -1

o <

• ■

a S

N <

a •

ota.

ﬁ

t-l

■i

i"-

*•»

•*

**2

•*

8•a

•J)

. "5

^ -

.n"

■ST

■i

-SS

*■»

»' /■■;

*>»

-4

•*

1 ~

•*+

* 1

<<" ■»

'-5

jLo

■ y±

» —

•a

i-i 61 tv

. i n

• ri

IS)

■^

11 13

.*2 . Z

'■ Oi

•5

o o z t-= tﬂ- ID

s ▶*■

fesz. . — cvt "2

f > tu 111 •o

^$*\

<^*ﬁ$\

,/t^^v

1-2 60497800 A

/ﬂ^sv

0S\

COMMENTS

^^ In column I a C, *, or $ indicates a comment line. Comments do not affect the program; they can be

written in column 2 to 80 and can be placed anywhere within the program. If a comment occupies more

than one line, each line must begin with C, *, or $ in column 1. In a comment line a character in column 6

/^ is not recognized as a continuation character. Comments can appear between continuation lines; they do not

interrupt the statement continuation.

/^ Comment lines following an END line are listed at the beginning of the next program unit unless the END

line is continued.

f STATEMENT LABELS

A statement label (any 1- to 5-digit integer) uniquely identiﬁes a statement so it can be referenced by

^^ another statement. Statements that will not be referenced do not need labels. Blanks and leading zeros are

not signiﬁcant. Labels need not occur in numerical order; however, a given label must not be used more than

once in the same program unit. A label is known only in the program unit containing it; it cannot be refer-

/^ enced from a different program unit. Any statement can be labeled, but only FORMAT and executable state

ment labels can be referenced by other statements. A label on a continuation line is ignored.

/fSV

/fp**»

CONTINUATION

Statements are coded in columns 7-72. If a statement is longer than 66 columns, it can be continued on as

many as 19 continuation lines. A character other than blank or zero in column 6 indicates a continuation line.

Column 1 can contain any character other than C, *, or $; columns 2, 3, 4, and 5 can contain any character.

Any statement except a comment or a list directive can be continued, including the END statement.

COLUMNS 73-80

Any information can appear in columns 73-80 because they are not part of the statement. Entries in these

columns are copied to the source program listing. They are generally used to order the lines in a deck, but

can contain information for DEBUG AREA processing.

STATEMENT SEPARATOR

Several statements can be written on one line if they are separated by the special character $. Each state

ment following a $ sign is treated as a separate statement. For example:

ACUM=24.$I=0 $ IDIFF-1970-1626

is the same as

ACUM =. 24.

I -. '0'

IDIFP = 1970-1626

$ can be used following any statement except FORMAT statements or list and debug directives. The state

ment following $ cannot be labeled; the information following $ is treated exactly as if it were in column 7 on

the next line.

/(PS,

60497800 A 1-3

BLANK LINES

Blank lines can be used freely between statements to produce blank lines on the source listing. Unlike a

comment line; a blank line interrupts statement continuation, and the line following the blank line is the be

ginning of a new statement. This line can optionally have the form of a continuation line.

DATA

No restrictions are imposed on the format of data read by the source program. Data input on cards is limited

to 80 characters per card, but a record can span more than one card. The maximum length in characters for

formatted, list directed, and NAMELIST records must agree with the length, r, speciﬁed in the PROGRAM

statement. If r is not speciﬁed, a default value of 150 is used.

^^s

/""^v

ORDERING OF STATEMENTS

The following table shows the general form of a FORTRAN program unit. Statements within a group can

appear in any order, but groups must be ordered as shown. Comment lines can appear anywhere within the

program.

STATEMENTS

■"""3^

1OVERLAY

PROGRAM*

FUNCTION*

SUBROUTINE*

BLOCK DATA

3IMPLICIT

type

COMMON

DIMENSION

EQUIVALENCE

EXTERNAL*

LEVEL

5Statement function*

deﬁnitions

§§

lit

iii

ENTRY*

Executable

statements*

7 END

F<S^N.

Not allowed in BLOCK DATA Subprograms

Namelist group name must be deﬁned before it is used

,**fs§\

x*"S"X

1^1 60497800 A

■ffspN

CONSTANTS

A constant is a ﬁxed quantity. The seven types of constants are: integer, real, double precision, complex,

octal, Hollerith, and logical.

INTEGER CONSTANT

n1 n2 • ■ ' nm

n is a decimal digit (0-9)

1 <m < 18

pies:

237 -74 +136772 -0024

An integer constant is a string of 1-18 decimal digits written without a decimal point. It may be positive,

negative or zero. If the integer is positive, the plus sign may be omitted; if it is negative, the minus sign

must be present. An integer constant must not contain a comma. The range of an integer constant is

- (259-l) to 2s9-1 (2S9-1 = 576 460 752 303 423 487).

Examples of invalid integer constants:

46. (decimal point not allowed)

23A (letter not allowed)

7,200 (comma not allowed)

When an integer constant is used as a subscript, or as an index in a DO statement or implied DO, the maximum

value is 217-1 (217-1 = 131 071), and the minimum is 1.

Integers used in multiplication, division, and exponentiation, whether constant or variable, should be in the range

- (248 -1) to 248 -1 (248 -1 = 281 474 976 710 655). The result of such operations also should be in this

range. If an integer constant exceeding this range is used, a fatal diagnostic is issued. Any other cases are not

diagnosed, and the results are unpredictable. For integer addition and subtraction (where both operands are

integers), the full 60-bit word is used.

When values are converted from real to integer or from integer to real (in an expression or assignment state

ment), the valid range is also from - (248-l) to 248-l . For values outside this range, the high order bits

are lost and no diagnostic is provided.

REAL CONSTANT

n.n .n n. n.nE±s .nE±s n.E±s nE±s

/*$ps\

n Coefﬁcient < 15 decimal digits

E±s Exponent (base 10)

60497800 B 1-5

/*""SsS\

A real constant consists of a string of decimal digits written with a decimal point or an exponent, or both.

Commas are not allowed. If the exponet is positive, the plus sign is optional. >*«%

The range of a real constant is IO"293 to 10+322; if this range is exceeded, a diagnostic is printed. Precision is

approximately 14 decimal digits, and the constant is stored internally in one computer word.

Examples:

7.5 -3.22 +4000. 23798.14 .5 - .72 42.El 700.E-2

Examples of invalid real constants:

3,50. (comma not allowed)

2. 5A (letter not allowed)

Optionally, a real constant can be followed by a decimal exponent, written as the letter E and an integer con

stant indicating the power of ten by which the number is to be multiplied. If the E is present, the integer

constant following the letter E must not be omitted. The sign may be omitted if the exponent is positive, but

it must be present if the exponent is negative.

/**S"j\

*^*^s

Examples:

42.El (42. X IO1 = 420.)

.00028E+5 (.00028 X IO5 = 28.)

6.205E12 (6.205 X IO12 = 6205000000000.)

8.0E+6 (8. x IO6 = 8000000.)

700.E-2 (700. X 10~2 = 7.)

7E20 (7. X IO20 = 70 000 000 000 000 000 0000.)

Example of invalid real constants: ^1

7. 2E3.4 exponent not an integer

DOUBLE PRECISION CONSTANT

| n.nD±s .nD±s n.D±s nD±s|

n Coefﬁcient )

D±s Exponent (base 10)

1-6 60497800 A

/SfPV,

Double precision constants are written in the same way as real constants except the exponent is speciﬁed by

the letter D instead of E. Double precision values are represented internally by two computer words, giving

extra precision. A double precision constant is accurate to approximately 29 decimal digits. If the exponent

is positive, the plus sign is optional.

Examples:

f* 5.834D2 (5.834 X IO2 = 583.4)

14.D-5 (14. X IO"5 = .00014)

9.2D03 (9.2 x 103 = 9200.)

^ -7.D2 (-7. X 102 = -700.)

3120D4 (3120. X IO4 = 31200000.)

0^\

Examples of invalid double precision constants:

0*^ 7.2D exponent missing

D5 exponent alone not allowed

1 2,1.3D2 comma illegal

i 3.141592653589793238462643383279 D and exponent missing

f* COMPLEX CONSTANT

(M,r2)

rl Real part

r2 Imaginary part

Complex constants are written as a pair of real constants separated by a comma and enclosed in

parentheses.

FORTRAN Coding Complex Number

(1., 7.54) 1. + 7.54i i = yf~A~

(-2.1E1, 3.24) -21. + 3.24i

( 4 . 0 , 5 . 0 ) 4 . 0 + 5 . 0 i

(O., -l.) 0.0 - l.Oi

60497800 A 1-7

The ﬁrst constant represents the real part of the complex number, and the second constant represents the

imaginary part. The parentheses are part of the constant and must always appear. Either constant may be

preceded by a plus or minus sign. Complex values are represented internally by two consecutive computer

words.

Both parts of complex constants must be real; they may not be integer.

Examples of invalid complex constants:

(275, 3.24) 275 is an integer

(12.7D-4 16.1) comma missing and double precision not allowed

4.7E+2,l.942 parentheses missing

(0,0) 0 is an integer

Real constants which form the complex constant can range from IO"293 to 10+322. Division of complex

numbers might result in underﬂow or overﬂow (see Appendix D) even when this range is not exceeded.

/*"3!"\

OCTAL CONSTANT .«*\

0f&$\

n1 •••nmB

i m

n is an octal digit, 0 through 7. 1 < m < 20 octal digits

An octal constant consists of 1 to 20 octal digits suffixed with the letter B. .-*..• /*m*\

Examples:

777777B

52525252B

500127345B

Invalid octal constants:

892777B 8 and 9 are non-octal digits. ■'■■ '■,;.:■;•■; **•»&

77000000007777752525252B exceeds 20 digits i. . .

07766 O not allowed

1-8 60497800 A

An octal constant must not exceed 20 digits nor contain a non-octal digit. If it does, a fatal compiler

diagnostic is printed. When fewer than 20 octal digits are speciﬁed, the digits are right justiﬁed and zero

ﬁlled. Octal constants can be used anywhere integer constants can be used, except that they cannot be used

as statement labels or statement label references, in a FORMAT statement, or as the character count when a

Hollerith constant is speciﬁed.

They can be used in DO statements, expressions, and DATA statements, and as dimension speciﬁcations.

,|p\

Examples:

BAT - (I * 5252B) . OR. JAY masking expression

J = jMAXO (I, iooob , J, K+40B) octal constant used as parameter in function

masking expression

arithmetic expression

dimension speciﬁcation

NAME -.1 ..AND. 77700000B

J = (5252B + N)/K

DIMENSION BUF(IOOOB)

When an octal constant is used in an expression, it assumes the type of the dominant operand of the

expression (Table 2-1, section 2).

HOLLERITH CONSTANT

j*$pn

>ip£\

/*^\

0&>\

nHf

nRf

nLf

=£f-v±

Unsigned decimal integer representing number of characters in string including blanks;

must be greater than zero.

String of characters; must contain at least one character

String delimiter

Left justiﬁed with blank ﬁll

Left justiﬁed with binary zero ﬁll

Right justiﬁed with binary zero ﬁll

A Hollerith constant has two forms: one is an unsigned decimal integer followed by the letter H, L, or R and a

string of characters; the other is a •?■= delimited string.

Hollerith constants can be used in DATA statements, as arguments in subroutine calls or function references, and

in expressions. In an expression, they are limited to 10 characters; and in a DATA statement they should be

limited to 10 characters (see section 3). If a Hollerith constant is used as an operand of an arithmetic operation,

an informative diagnostic is given. If a Hollerith constant is used as an argument in a subprogram call, it is

followed by a zero word.

60497800 A 1-9

,<«*r«t*X

The Hollerith speciﬁcation in a FORMAT statement (see section 6) is not the same as a Hollerith constant.

Example:; "V./>- '":-"" ."* 'v-••-*•■.• •"-.''''' ,l' •' ;■' ;,'*.'* '"- - "" ' - -' ""-* "-/•"' "; - • •-'-- /

PROGRAM HOLL (OUTPUT)

I * 6HABCDEF ***)

J-=6LABCDEF

K •*' 6RABCDEF A

L = * A B C D E F # ^

PRINT It ItI»J»JfK»K»L»L ^' '

1 FORMAT <024tA15) ^

- S T O P ^

END

/**"^"\

Stored Internally: Display Code:

01020304050655555555 ABCDEF H format ^

01020304050600000000 ABCDEF:::: I-format ?

00000000010203040506 ::::ABC0EF R format'

01020304050655555555 ABCDEF •* format ^

nHf and •J6f*?fe ^\

These two forms produce left-justiﬁed display code constants with 10 characters per word. If the string length

is not a multiple of 10, the ﬁnal word is blank ﬁlled.

y-^-W^V

nHf Examples:

18HTHIS IS A CONSTANT

7HTHE END ^

19HRESULT NUMBER THREE

0 t E x a m p l e s : " - . , , ■ " • . ' ';" '''V'-.-V'.y:..'_ " -".r « ■• . " ' _ ■ ' " ^

.F(V.EQ.¥Y,ES-# Y»Y+1. '_','.,■ ? •*.'..,." ."' .".„."•-•- '" '-■'•''•

PRINTI^SQRT^.SQRTR) =,. . ;; .

1 FORMAT (A10.F10.2) " "*

PRI NT 2, =# TEST PASSED-?*- ..■'■: ■■ > • . ; ' — ■* • • ^

2 FORMAT (2A10)

INTEGER LINEC7), N1THRU9 . /=%>

LOGICAL NEWPAGE, ', '

IF (NEWPAGE) LINE{7);= -£PAGE0-£ + N1 THRU 9

/***^\

-**s^^.

/cr^\

1-10 60497800 E '

nRf and nLf

These two forms produce display code constants with 10 characters per word. If the string length is not a

multiple of 10, the ﬁnal word is zero-ﬁlled, and justiﬁed; nRf indicates right justiﬁcation and nLf indicates left

justiﬁcation.

nRf Example:

I VAL (IJ;- 1RA

nLf Example:

INDEX {■} - 3LLGO

LOGICAL CONSTANT

A logical constant takes the forms:

.TRUE. or .T. representing the value true

.FALSE, or fF. representing the value false

The decimal points are part of the constant and must appear.

Examples:

LOGICAL XI, X2

XI = .TRUE.

X2 = .FALSE.

VARIABLES

A variable represents a quantity whose value can be varied; this value can be changed repeatedly during

program execution. Variables are identiﬁed by a symbolic name of one to seven letters or digits, beginning

with a letter. A variable is associated with a storage location; whenever a variable is used, it references the

value currently in that location.

A variable can have its type speciﬁed in a type statement (see section 3) as integer, real, double precision,

complex, or logical. In the absence of an explicit declaration, the type is implied by the ﬁrst character of

the name: I, J, K, L, M, and N imply type integer and any other letter implies type real, unless an IMPLICIT

statement (see section 3) is used to change this normal implicit type.

60497800 A 1-11

«"-^\

-ExjaMpleV" 1 ' *;" , : .'--.„,.

" ^: /|^lPL(C^:pOMiUiE^f^ECISION (A)

\:CQWPL'EX>lprtA \, >'L . /

APPLE«ORANGES+PEARS

Ah, explicit declaration overrides an IMPLICIT declaration. Therefore, ALPHA is type complex; APPLE is

type double precision.

Default typing of variables:

A-H,0-Z Real

I-N Integer

INTEGER VARIABLES

An integer variable is a variable that is typed explicitly or implicitly as described under Variables.

The value range is - (2S9-1) to 2S9-1. When an integer variable is used as a subscript, the maximum value is

217-1. The resulting absolute value of conversion from integer to real, or real to integer must be less than

248. The operands, as well as the result, of an integer multiplication or division must be less than 248 in

absolute value. If any of these restrictions are violated, the results are unpredictable. For integer addition

and subtraction, the full 60-bit word is used; the resulting absolute value must be less than 2s9.

See section 4 for restrictions or integers used in DO statements.

An integer variable occupies one word of memory.

Examples:

ITEM1 NSUM JSUM N72 J K2S04

REAL VARIABLES

A real variable is a variable that is typed explicitly or implicitly as described under Variables.

The value range is IO"293 to 10+322 with approximately 14 signiﬁcant digits of precision. A real variable

occupies one word of storage.

Examples:

A V A R S U M 3 R E S U L T T O T A L 2 B E T A X X X X

1-12 60497800 A

/**^S*v

/-"SsX

/f"SS"v

00$$^

DOUBLE PRECISION VARIABLES

Double precision variables must be typed by a type declaration. The value of a double precision variable can

range from IO"293 to 10+322 with approximately 29 signiﬁcant digits of precision.

Double precision variables occupy two consecutive words of memory. The ﬁrst word contains the more

signiﬁcant part of the number and the second contains the less signiﬁcant part.

Example:

IMPLICIT DOUBLE PRECISION(A)

DOUBLE PRECISION OMEGA.X.IOTA

The variables OMEGA, X, IOTA and all variables whose ﬁrst letter is A are double precision.

COMPLEX VARIABLES

Complex variables must be typed by a type declaration. A complex variable occupies two words of memory;

each word contains a real number. The ﬁrst word represents the real part of the number and the second

represents the imaginary part.

Example:

COMPLEX ZERA,MU.LAMBDA

LOGICAL VARIABLES

V Logical variables must be typed by a type declaration. A logical variable has the value true or false and

occupies one word of memory.

/**P\

\ Example:

r LOGICAL L33,PRAVDA,VALUE

_ A R R A Y S

A FORTRAN array is a set of elements identiﬁed by a single name composed of one to seven letters and

digits beginning with a letter. Each array element is referenced by the array name and a subscript. The type

(^ of the array elements is determined by the array name in the same manner as the type of a variable is deter

mined by the variable name (see Variables in this section). The array name and its dimensions must be de

clared in a DIMENSION or COMMON statement or a type declaration. Arrays can have one, two, or three

dimensions.

The number of dimensions in the array is indicated by the number of subscripts in the declaration.

dimension STOR(6) declares a one-dimensional array of six elements

60497800 A 1-13

REAL STOR{3,7) declares a two-dimensional array of three rows and seven columns

LOGICAL STOR(6.6,3) declares a three-dimensional array of six rows, six columns and three planes

The entire array may be referenced by the unsubscripted array name when it is used as an item in an input/

output list, as an actual parameter, or in a DATA statement. In any other context, only the ﬁrst element

of the array is implied by the unsubscripted, array name.

Example 1:

The array N consists of six values in the order: 10, 55, 11, 72, 91, 7

i*C5*^y

N(l) value 10

N(2) value 55

N(3) value 11

N(4) value 72

N(5) value 91

N(6) value 7

Example 2:

The two-dimensional array TABLE (4,3) has four rows and three columns.

Column 1 Column 2 Column 3

Row 1 44 10 105

Row 2 72 20 200

Row 3 11 30

Row 4 91 76 714

To refer to the number in row two, column three write TABLE(2,3).

TABLE(3,3) = 30 TABLE(1,1) = 44 TABLE(4,1) = 91

TABLE(4,4) would be outside the bounds of the array and results are unpredictable.

Example 3: ,' .

. PROGRAM VARDIM <OUTPUT•TAPE6=bUTPUT>

COMMON X<4*3)

■ '"' • ' -REAL Y<6)

CALL IOTA <Xil2>

CALL IOTA <Y*6>

WRITE (6ﬂ00) X*Y

100 FORMAT (* ARRAY X « *tl2E9„l»5X»*ARRAY Y * *6E9.1>

STOP

E N D • ■ • • .

The program declares and references two arrays;' X is a two-dimensional'.array of 12 elements and Y

is a one-dimensional array of six elements.

1_14 60497800 A

i*lip\

SUBSCRIPTS

A subscript indicates the position of a particular element in an array. A subscript consists of a pair of

parentheses enclosing from one to three subscript expressions which are separated by commas. The subscript

follows the array name. A subscript expression can be any valid arithmetic expression. If the value of the

expression is not integer, it is truncated to integer.

If the number of subscript expressions is less than the number of declared dimensions, the compiler assumes

the omitted subscripts have a value of one. The number of subscript expressions in a reference must not

exceed the number of declared dimensions.

The value of a subscript must never be zero or negative. It should be less than or equal to the product

of the declared dimensions, or the reference will be outside the array. If the reference is outside the bounds

of the array, results are unpredictable.

The amount of storage allocated to arrays is discussed under DIMENSION declarations in section 3.

Valid subscript forms:

A(l.K)

B(l+2,J-3,6*K+2)

LAST(6)

ARAYD(1,3.2)

STRING (3*K*ITEM+3)

Invalid subscript forms:

ATLAS (0)

D(1 .GE. K)

A(,l) or A(I„K)

zero subscript causes a reference outside of the array

relational or logical expression illegal

commas can only be used to separate adjacent subscript expressions

Example:

Plane 1

C o M C o l 2 C o l 3

Plane 2

Col 1 Col 2 Col 3

22 51

098

3207 99

the single

subscript

NEXT (3):

represents

NEXT (3,1,1)

NEXT (3,2)V-NEXT (2,2)

represents represents

NEXT (3,2,1) NEXT (2,2,1)

Plane 3

Col 1 Col 2 Col 3

2 1 552

77 60

85 100

Row 1

Row 2

Row 3

In the three-dimensional array NEXT when only one or two subscripts are shown, the remaining subscripts are

assumed to be one.

60497800 A 1-15

ARRAY STRUCTURE

Arrays are stored in ascending locations: the value of the ﬁrst subscript increases most rapidly, and the value of

the last increases least rapidly.

Example:

/"""SIS

In an array declared as A(3,3,3), the elements of the array are stored by columns in ascending locations.

Plane 1

Col 1 Col 2 Col 3

Plane 2

Col 1 Col 2 Col 3

/■Ss?K

Row 1 A111 [-A121 i*A131

\1 1

Row 2 A211 A221 A231

j,1 i

Row 3 A311-1 A 3 2 1 J A331

Row 1

Row 2

Row 3

A112 rA122 r*A132

J1\

A212 t A222 , A232

\1\

A312-J A322-1 A332

Plane 3

Col 1 Col 2 Col 3

N*.

Row 1

Row 2

Row 3

A113 r~A123 r-A133

1\\

A213 A223 A233

11 1

A313-1 A323J A333

The array is stored in linear sequence as follows:

Location Relative

Element to ﬁrst Element

A(l,l,l)

A(2,l,l)

A(3,l,l)

A(l,2,l)

A(2,2,l)

A(3,2,l)

A(l,3,l)

A(2,3,l)

A(3,3,l)

A(l,l,2)

A(2,l,2) 10

A(3,l,2) 11

A( 1,2,2) 12

A(2,2,2) 13

Location Relative

Element to ﬁrst Element

A(3,2,2) 14

A( 1,3,2) 15

A(2,3,2) 16

A(3,3,2) 17

A(l,l,3) 18

A(2,l,3) 19

A(3,l,3) 20

A( 1,2,3) 21

A(2,2,3) 22

A(3,2,3) 23

A(l,3,3) 24

A(2,3,3) 25

A(3,3,3) 26

0*&fa\

^m$\

1-16 60497800 A

To ﬁnd the location of an element in the linear sequence of storage locations the following method can be

used:

Number of

Dimensions

Array

Dimension Subscript

Location of Element

Relative to Starting Location

1ALPHA(K) ALPHA(k) (k-1)XE

2ALPHA(K.M) ALPHA(k,m) (k-1+KX(m-1) )XE

3ALPHA(K,M,N) ALPHA(k,m,n) (k-1+KX(m-1+MX(n-1)))XE

K, M, and N are dimensions of the array.

k,m, and n are the subscript expression values of the array.

1 is subtracted from each subscript value because the subscript starts with 1, not 0.

E is length of the element. For real, logical, and integer arrays, E = 1. For complex and double

precision arrays, E = 2.

Examples:

Subscript

Location of Element

Relative to Starting Location

INTEGER ALPHA (3) ALPHA(2) (2-1)X1 = 1

REAL ALPHA (3,3) ALPHA(3,1) (3-1+3X(1-1))X1 =2

COMPLEX ALPHA (3,3,3) ALPHA(3,2,1) (3-1+3X(2-1+3X(1-1)))X2 = 10

jjp\

60497800 A 1-17

■^

iffS^. EXPRESSIONS AND ASSIGNMENT STATEMENTS

EXPRESSIONS

FORTRAN expressions are arithmetic, masking, logical and relational. Arithmetic and masking, expressions

yield numeric values, and logical and relational expressions yield truth values.

ARITHMETIC EXPRESSIONS

An arithmetic expression is a sequence of unsigned constants, variables, array elements, and function references

separated by operators and parentheses. For example,

(A-B)*F + C/D**E

is a valid arithmetic expression.

The FORTRAN arithmetic operators are:

+ addition

subtraction

* multiplication

/ division

** exponentiation

An arithmetic expression may consist of a single constant, variable, array element, or function reference. If X

is an expression, then (X) is an expression. If X and Y are expressions, then the following are expressions:

X + Y

X*Y

-X

+ X

X-Y

X/Y

60497800 A 2-1

All operations must be speciﬁed explicitly. For example, to multiply two variables A and B, the expression

A*B must be used. AB, (A)(B), or A.B will not result in multiplication.

Expression Value

3.78542 Real constant 3.78542

A ( 2 * J ) Array element A (2*J)

BILL Variable BILL

SQRT(5.0) xTT

A+B Sum of the values A and B

C*D/E Product of C times D divided by E

J**I Value of J raised to the power of I

(200 - 50)*2 300

EVALUATION OF EXPRESSIONS

The sequence in which an expression is evaluated is governed by the following rules, listed in descending precedence:

1. References to external functions are evaluated.

2. Arithmetic statement functions and intrinsic functions are expanded.

3. Subexpressions delimited by parentheses are evaluated, beginning with the innermost subexpressions.

4. Subexpressions deﬁned by arithmetic, relational, and logical operators are evaluated according to the

following precedence hierarchy:

(exponentiation)

/(division or multiplication)

+(addition or subtraction)

GT. .GE. XT. .LE. .EQ. .NE. (relational)

.NOT. (logical)

.AND. (logical)

.OR. (logical)

-f**^*\\

Ajﬁ^\

'e^\

^ S K

2-2 60497800 A

0*^>

5. Subexpressions containing operators of equal precedence are evaluated from left to right. However,

individual operations that are mathematically associative and/or commutative may be reordered by the '

compiler to perform optimizations such as removal of repeated subexpressions or improvement of

functional unit usage. The evaluation of the expression A/B*C is guaranteed to algebraically equal

AC -5- B, not A -h BC, but the speciﬁc order of evaluation here is indeterminate. Subexpressions

containing integer divisions are not reordered within the * / precedence level because the truncation

resulting from an integer division renders these operations non-associative.

Unary addition and subtraction are treated as follows:

+n the same as n

- n n e g a t e n

An array element (a subscripted variable) used in an expression requires the evaluation of its subscript. The

type of the expression in which a function reference or subscript appears does not affect, nor is it affected

by, the evaluation of the arguments or subscripts.

The evaluation of an expression having any of the following conditions is undeﬁned:

Negative-value quantity raised to a real, double precision, or complex exponent

Zero-value quantity raised to a zero-value exponent

Inﬁnite or indeﬁnite operand (Appendix D)

Element for which a value is not mathematically deﬁned, such as division by zero

If the error traceback option (T) is selected on the FTN control statement (section 10), the ﬁrst three

conditions produce informative diagnostics during execution. If the traceback option is not selected, a mode

error message is printed (Appendix D).

In the case of invalid exponentiation, a diagnostic might be issued by one of the library routines ALOG, EXP, or

DEXP when the exponent is real, complex, or double precision, and the base is integer, real or double precision.

Two operators must not be used together. A*-B and Z/ + X are not allowed. However, a unary + or - can

be separated from another operator in an expression by using parentheses. For example,

A*(-B) andZ/(+X) Valid expressions

B*-A andX/-Y*z Invalid expressions

Each left parenthesis must have a corresponding right parenthesis.

Example:

(F + (X * Y) Incorrect, right parenthesis missing

(F + (X * Y)) Correct

Examples:

In the expression

A-B*C

B is multiplied by C, and the product is subtracted from A.

60497800 E 23

^^\

0^$$\

The expression A/B-C*D**E is evaluated as follows:

D is raised to the power of E.

A is divided by B.

C is multiplied by the result of D**E.

The product of C*D**E is subtracted from the quotient of A divided by B.

The expression -A**C is evaluated as 0-A**C; A is ﬁrst raised to the power of C and the result is

then subtracted from zero.

The expression A*B*C may be evaluated as ((A*B)*C), ((A*C)*B) or (A*(B*C)), since the operator * is

associative.

The expression A**B**C is evaluated as ((A**B)**C), since the operator ** is not associative. ,**■*■■*

Dividing an integer by another integer yields a truncated result; 11/3 produces the result 3. Therefore,

when an integer expression is evaluated from left to right, J/K*I may give a different result than I*J/K. '^

Example:

I « 4 J = 3 K = 2 J

J/K*I I*J/K }

3 / 2 * 4 = 4 4 * 3 / 2 - 6

An integer divided by an integer of larger magnitude yields the result 0.

Example: '•*■%

N = 2 4 M = 2 7 K = 2

R/M*K

24/27*2 - 0

Examples of valid expressions:

3.14159

B + 16.427

( X B A R +( B ( I , J + I , K ) /3 . 0 ) )

- ( C + D E L T A * A E R O ) ^

2-4 60497800 A

/*"S\

(-B - SQRT(B**2-(4*A*C)))/(2.0*A)

GROSS - (TAX*0.04)

TEMP + V(M,AMAX1(A,B))*Y**C/ (H-FACT(K+3))

TYPE OF ARITHMETIC EXPRESSIONS

An arithmetic expression may be of type integer, real, double precision, or complex. The order of

dominance from highest to lowest is as follows:

Complex

Double Precision

Real

Integer

Table 2-1. Mixed Type Arithmetic Expressions with + - * / Operators

\ . 2 n d

1st^vjoperand

operand n^

Integer Real Double

Precision Complex Typeless

Operand

Integer Integer Real Double

Precision Complex Integer

Real Real Real Double

Precision Complex Real

Double

Precision

Double

Precision

Double

Precision

Double

Precision Complex Double

Precision

Complex Complex Complex Complex Complex Complex

Typeless

Operand Integer Real Double

Precision Complex Integer

When an expression contains operands of different types, type conversion takes place during evaluation. Before each

operation is performed, operands are converted to the type of the dominant operand. Thus the type of the value of

the expression is determined by the dominant operand. For example, in the expression A*B-I/J, A is multiplied by

B, I is divided by J as integer, converted to real, and subtracted from the result of A multiplied by B.

Octal and Hollerith constants, as well as references to shifting or masking functions, are typeless operands. When

these operands are used, type is not converted. When these operands are the only operands in an expression, they

are treated as if they were type integer, and the result is type integer.

Variables into which Hollerith constants are stored should be of type INTEGER to ensure proper results when used

in subsequent arithmetic or logical expressions. For example, if the variables are REAL, expressions involving these

variables are evaluated using ﬂoating point arithmetic.

60497800 E 2-5

EXPONENTIATION

In exponentiation, the following types of base and exponent are permitted:

Base

Integer

Real

Double Precision

Complex

Typeless

Exponent

Integer, Real, Double Precision, Complex, Typeless

Integer, Typeless

Integer, Real, Double Precision, Complex, Typeless

/""SSi"!

.f**"StN

The exponentation is evaluated from left to right. The expression A**B**C is evaluated, as ((A**B)**C)

In an expression of the form A**B the type of the result is determined as follows:

Type of A Type of B Type of Result

of A**B

Integer Integer

Real

Double

Complex .

Typeless

Integer

Real

Double

Complex

Integer

Real Integer

Real

Double

Complex

"Typeless

Real

Double

Cpmplex

Real -,'.

Double Integer

Real

Double

Complex

Typeless

Double

Complex

Double

Complex Integer

Typeless

Complex

Typeless Integer

Real ,

Double

Cornplex >

Typeless

Integer* .

Real

Double

Complex

Integer

The expression -2**2 is equivalent to 0-2**2. An exponent may be an expression. The following examples are all

acceptable.

B**2.

B**N

B**(2*N-1)

(A+B)**(-J)

A negetive exponent must be enclosed in parentheses:

A**(-B)

NSUM**(-J)

2-6 60497800 D

**f*^a\

/*f^\

When the exponent is of a type other than integer, exponentiation is performed by means of a call to FORTRAN

Common Library routines. The value of the result in these cases is determined according to the formula:

xy = ey<Mx))

/*f^\

where In is the natural logarithm function.

•fjf^V Examples:

Expression Type

0$^s CVAB**(I-3) Real**Integer

Jp\ D**B

C**I

Real** Real

Complex** Integer

JsP^V BASE(M,K)**2.1 Double Precision

**Real

/fPN K**5 Integer* *Integer

314D-02**3.14D-02 Double Precision

**Double Precision

Result

Real

Complex

Double Precision

Integer

Double Precision

RELATIONAL EXPRESSIONS

'1 o p a .

a|,a2

Arithmetic or masking expression

Relational operator

A relational expression is constructed from arithmetic or masking expressions and relational operators.

Arithmetic expressions may be type integer, real, double precision, or complex. The relational operators are:

.GT. Greater than

.GE. Greater than or equal to

.LT. Less than

.LE. Less than or equal to

.EQ. Equal to

.NE. Not equal to

The enclosing decimal points are part of the operator and must be present.

60497800 A 2-7

Two expressions separated by a relational operator constitute a basic logical element. The value of this

element is either true or false. If the expressions satisfy the relation speciﬁed by the operator, the value is

true; if not, it is false. For example:

X+ Y .G T. 5 . 3

If X + Y is greater than 5.3 the value of the expression is true. If X + Y is less than or equal to

5.3 the value of the expression is false.

A relational expression can have only two operands combined by one operator, a, op a2 op a3 is not valid.

Relational operands may be of type integer, real, double precision, or complex, but not logical. With complex

operands, the relational operators .EQ. and .NE. test for equality oh both the real and imaginary parts; for all

other relational operators only the real parts are compared.

Examples:

J.LT.ITEM

580.2 .GT. VAR

B . GE. (2.7,5. 9E3) real part of cornplex number is used in evaluation

E.EQ..5

(I) .EQ. (J(K))

C . LT. 1 . 5 D 4 m o s t s ig n i ﬁ c an t p ar t o f d o u b l e p r e c i s i o n n u mb e r i s u s ed i n

evaluation

Relational expressions are evaluated according to the rules governing arithmetic expressions. Each

expression is evaluated and compared with zero to determine the truth value. For example, the expression

p.EQ.q is equivalent to the question, does p - q = 0? q is subtracted from p and the result is tested for zero. ^%

If the difference is zero or minus zero the relation is true. Otherwise, the relation is false.

I f p i s 0 a n d q i s - 0 t h e r e l a t i o n i s t r u e . . - * * * %

Expressions are evaluated from left to right. Parentheses enclosing an operand do not affect evaluation; for

example, the following relational expressions are equivalent: **®\

A.GT.B

<**^K

A.GT.(B) '

(A).GT.B

(A).GT.(B)

ySSsEX

>^*^\

2-8 60497800 A

Examples:

JPN

REAL A

A.GT.720

INTEGER I,J

I . E Q . J ( K )

(I).EQ.(N*J)

B.LE.3.754

Z.LT.35.3D+5

Examples of invalid expressions:

A .GT. 720 .LE. 900

B .LE. 3.754 .EQ. C

DOUBLE PRECISION BILL, PAY

BILL .LT. PAY

A+B.GE.Z**2

300.+B.EQ.A-Z

.5+2. .GT. .8+AMNT

2 relational operators must not appear in a relational expression

LOGICAL EXPRESSIONS

L. op L, op L, op ...L,

L,...Ln

logical operand or relational expression

logical operator

A logical expression is a sequence of logical constants, logical variables, logical array elements, or relational

expressions separated by logical operators and possibly parentheses. After evaluation, a logical expression

has the value true or false.

Logical operators:

.NOT. or .N.

.AND. or ,A.

.OR. or .0.

logical negation

logical multiplication

inclusive OR

The enclosing decimal points are part of the operator and must be present.

60497800 A 2-9

The logical operators are deﬁned as follows (p and q represent LOGICAL expressions):

.NOT.p

p.AND.q

p.OR.q

If p is true, .NOT.p has the value false. If p is false, .NOT.p has the

value true.

If p and q are both true, p.AND.q has the value true. Otherwise, false.

If either p or q, or both, are true then p.OR.q has the value true. If both

p and q are false, then p.OR.q has the value false.

Truth Table

p q p .AND. q p .OR. q .NOT. p

TT T T F

TF F T F

FT FT T

FFF F T

If precedence is not established explicitly by parentheses, operations are executed in the following order:

.NOT. .AND. .OR.

Example:

PROGRAM L0GIC,(0UTPUTtTAPE6=0UTPUT)

THIS PROGRAM PRINTS OUT A TRUTH TABLE FOR LOGICAL

OPERATIONS WITH P AND Q

LOGICAL P»Q»L0GNEG,L0GMLT»L0GSUM»TABLE(4»2)

DATA TABLE/.TRUE.t.TRUE.♦.FALSfc.*.FALSE.».TRUE.».FALSE.t.TRUE.9

1.FALSE./

WRITE(6»10)

1 0 F 0 R M A T < 6 1 H 1 P Q . N O T . Q P . A N D Q P . 0

1R. Q /10X» 51UH-))

DO 20 I = l*«f

LOGNEG = .NOT. TAbLE<I*2)

LOGMLT = TA6LE<I»D -AND. TABLE(I»2>

LOGSUM = TA6LEU»1) .OR. TABLE(I»2>

20 WRITE(6»30) (TABLE(I 9J),J=l»2>t LOGNEG, LOGMLT* LOGSUM

30 FORMATUH0* 5(L1D)

STOP

END

>**^v

/*""*s*\

2-10 60497800 A

''""a^k

0SS,

Output:

p U . N O T . Q P . A N D U P . O K . Q

T T F T T

T F T F T

F T F F T

F F T F F

The operator .NOT. which indicates logical negation appears in the form:

.NOT. p

.NOT. can appear in combination with .AND. or .OR. only as follows (p and q are logical expressions):

p .AND..NOT. q

p .OR..NOT. q

p .AND.(.NOT. q )

p .OR.(.NOT. q )

.NOT. can appear adjacent to itself only when the second operator is enclosed in parentheses, as in .NOT. (.NOT.p).

Two logical operators can appear in sequence only in the forms .OR..NOT. and .AND..NOT.

Valid logical expressions, where M, L, and Z are logical variables, are:

.NOT.L

.NOT.(X .GT. Y)

X .GT. Y .AND..NOT.Z

(L) .AND. M

Invalid logical expressions, where P and R are logical variables, are:

.AND. P .AND. must be preceded by a logical expression

K .EQ. 1 .OR. 2 .OR. must be followed by a logical expression

P .AND. .OR.R .AND. always must be separated from .OR. by a logical expression

60497800 A 2-11

>#sv

Examples:

A, X, B, C, J, L, and K are type logical.

Expression h§^;i^i^

A .AND. .NOT. X -:. A|^A;.:;^i;N :|;^Z-^H;10-3 ©f

.NOT.B S»^l|g§^

A.AND.C HSSiiiSJSSSS^SB

J.OR.L.OR.K iiiiiii§#gig^5^^ig

Examples:

B-C < A < B+C is written asB-C .LE. A .AND. A .LE. B+C

FICA >176. andPAYNB = 5889. is written FICA .GT. 176. .AND. PAYNB .EQ. 5889.

MASKING EXPRESSIONS

Masking expressions are similar to logical expressions, but the elements of the masking expressions are any type

of variable, constant, or expression other than logical.

Examples:

J .AND. N .NOT. 55 ,NOT. (B) KAY .OR. 63

Masking operators are identical in appearance to logical operators but meanings differ. In order of

dominance from highest to lowest, they are:

.NOT. or .N.

.AND. or .A.

.OR. or .0.

Complement the operand

Form the bit-by-bit logical product (AND) of two operands

Form the bit-by-bit logical sum (OR) of two operands

The enclosing decimal points are part of the operator and must be present. Masking operators are

distinguished from logical operators by non-logical operands.

Examples:

Expression Alternative Form

B .OR. D B .0. D

A .AND. .NOT. C A .A. .N. C

BILL .AND. BOB BILL..A. BOB

I .OR. J- .OR. K .OR. N I . 0 . J . 0 . K . 0 . N .

( . N O T. ( . n o t . ( , : n o t . A .OR. B))) (.N.(.N.(.N. A.OR. B)))

/fSofev

/r*^\

'*t*^?X

2-12 60497800 A

The operands may be any type variable, constant, or expression (other than logical).

■ Examples:

0m\ TAX .AND. INT

\ . N O T . 5 5

734 .OR. 82

rA .AND. 77B Extract the low order 6 bits of A

B . O R . C L o g i c a l s u m o f t h e c o n t e n t s o f B a n d C

M .A N D. . N O T. 77B C l ea r t he l o w o rd e r 6 bi t s of M .

,***S\

v In masking operations operands are considered to have no type. If either operand is type COMPLEX,

operations are performed only on the real part. If the operand is DOUBLE PRECISION only the most

0$zs signiﬁcant word is used. The operation is performed bit-by-bit on the entire 60-bit word. For simplicity,

V only 10 bits are shown in the following examples. Masking operations are performed as follows:

^ J = 0 1 0 1 0 111 0 1 a n d I = 11 0 0 11 0 1 0 1

J .AND. I ■ '

V The bit-by-bit logical product is formed

^•N J 0101011101 ■/■_,'■,

I 1100110101

ifs^V

0*&S

0100010101 Result after masking

J .OR. I

The bit-by-bit logical sum is formed

J 0101011101

I 1100110101

101111101 Result after masking

.NOT. Complement the operand

.NOT. I

I 1100110101

0011001010 Result after masking

.NOT. must not immediately precede .AND. or .OR.

60497800 A 2-13

,<eS8\

Using vthe following values:

•';-:;;'/•.;-*'■/. v?X 77770060600000000000

^.V.\''JV". .".'.;'^-.- D 00000000777777777777

.'.,_,. B 00000000000000001763

; , ; C 20045000000000000000

Masking operations produce the following octal results:

octal constant ;

octal constant

octal form of integer constant

octal form of real constant

y«£#&\

^ N O T ^ A / - ^ •

; A A N D i . C " -

' ■A\aND;";N,QT..C"

' '^^OJt.:'^OT.''D- ;

Invalid example:

;:'; LOGICAL. A

A .AND. B .OR. C

is. : 00007777777777777777

is . -20040000000000000000

is 57730000000000000000.

is 77777777000000001763

masking expression must not contain logical operand

Example:

PROGRAM MASK CINPUT,OUTPUT)

FORMAT <iHi»5X»^HNAME»///>

PRINT i

2 FORMAT <3A1Q,I1>

3 READ 2,LNAME,FNAHE,ISTATE,KST0P

IFCKSTOP.EQ.DSTCP

C IF FIRST TWO CHARACTERS OF ISTATE NOT EQUAL TO CA REA0 NEXT CARO

IFf(ISTATE»AND.77770000000000000000B).NE.(2HCA.ANOo777700000000000

KOOO00B)) GO TO 3

11 FORMAT<5X,2A10>

10 PRINT 11,LNAM6,FNAME

GO TO 3

END

ASSIGNMENT STATEMENTS

An assignment statement evaluates an expression and assigns this value to a variable or array element. The

statement is written as follows:

v = expression

v is a variable or an array element

^3k£V

2-14 60497800 A

The meaning of the equals sign differs from the conventional mathematical notation. It means replace the

value of the variable on the left with the value of the expression on the right. For example, the assignment

statement A=B+C replaces the current value of the variable A with the value of B+C.

ARITHMETIC ASSIGNMENT STATEMENTS

v = arithmetic expression

Replace the current value of v with the value of the arithmetic expression. The variable or array element

can be any type other than logical.

Examples:

A=A+1

N=J-100*20

WAGE=PAY-TAX

VAR=VALUE+(7/4)*32

B(4)=B(1)+B(2)

replace the value of A with the value of A + I

replace N with the value of J-100*20

replace WAGE with the value of PAY less TAX

replace the value of VAR with the value of VALUE+ (7/4)*32

replace the value of B(4) with the value of B(l) + B(2)

If the type of the variable on the left of the equals sign differs from that of the expression on the right, type

conversion takes place. The expression is evaluated, converted to the type of the variable on the left, and

then replaces the current value of the variable. The type of an evaluated arithmetic expression is

determined by the type of the dominant operand. Below, the types are ranked in order of dominance from

highest to lowest:

Complex

Double Precision

Real

Integer

In the following tables, if high order bits are lost by truncation during conversion, no diagnostic is given.

J0S\

60497800 A 2-15

'*^i\

>*"*S\

CONVERSION TO INTEGER

Value Assigned Example

Value of 1 FORM

After Evaluation

Integer = Integer Value of integer

expression re

places v.

IFORM = 10/2

Integer = Real Value of real

expression, trun

cated to 48-bit

integer, replaces

IS^Hl^^^^^^^S

Integer = Double Precision Value of double

precision expres

sion, truncated to

48-bit integer,

replaces v.

IFORM = 3141.593D3 3141593

9St^UlU^S&^fMl8SiﬁlliliHBiiiift

Blii^iiiiiiiBi

p||^^i|i|ii|§i

^^Si^^^BMii^K

;S^Bitl9^S^S^S

r5$"N

CONVERSION TO DOUBLE PRECISION

Value of SUM

Value Assigned Example After Evaluation

Double Precision = Integer Value of integer

expression, trun

cated to 48 bits,

is converted to

real and replaces

most signiﬁcant

part. Least sig

niﬁcant part set

toO.

SUM = 7*5 35.D0

Double Precision = Real Value of real

expression re IIBttl^Siﬁli^iBlIililSSiS^M 15.D0

places most fﬁSSv^iﬁiiRll^^^^iS^^?

signiﬁcant part; ^i^%y;::H':^y-;)'^

least signiﬁcant

part is set to 0.

2-16 60497800 A

CONVERSION TO DOUBLE PRECISION (CONTINUED)

Value of SUM

Value Assigned Example After Evaluation

Double Precision Value of double SUM = 7.322D2 - 32.D -1 7.29D2

= Double Precision precision expres

sion replaces v.

Double Precision = Complex Value of real

part of complex

expression re

places v. Least

signiﬁcant part

is set to 0.

SUM = (3.2,7.6)+ (5.5,1.0) 8.7D0

CONVERSION TO COMPLEX

Value of AFORM

Value Assigned Example After Evaluation

! Complex-3 Integer Value of integer

expression, trun

cated to 48 bits,

is converted to

real, and replaces

real part of v.

Imaginary part is

set to 0.

AFORM=2+3 (5.0,0.0)

Complex = Real Value of real

expression re

places real part

ofv. Imaginary

part set to 0.

AFORM=2.3 + 7.2 (9.5,0.0)

Complex = Double Precision Most signiﬁcant

part of double

precision expres

sion replaces real

part ofv. Imag

inary part set to

AFORM = 20D0 + 4.4D1 (64.0,0.0)

Complex = Complex Value of complex

expression replaces

AFORM = (3.4,1.1) + (7.3,4.6) (10.7,5.7)

variable.

60497800 A 2-17

CONVERSION TO REAL

Value Assigned Example

Value of AFORM

After Evaluation

Real = Integer Value of integer

expression, trun

cated to 48 bits,

is converted to

real and replaces

AFORM =200+ 300 500.0

Real = Real Value of real

expression re

places V.

AFORM = 2.5+ 7.2 9.7

Real = Double Precision Value of most

signiﬁcant part

of expression re

places v.

AFORM = 3421.D - 04 .3421

:! Real % Complex

i * ■' ' ' ^ Value of real

part of complex/

.expression re

places v.

AFORM = (9.2,1.1}- (2.1,5.0) .

1 y 1 )

7,1

LOGICAL ASSIGNMENT

Logical variable or array element = Logical or relational expression

<<*K§\

<"**»S^

>*3ftX

Replace the current value of the logical variable or array element with the value of the expression.

Examples:

LOGICAL L0G2

1 = 1

L0G2 = I .EQ.O

LOG2 is assigned the value .FALSE, because h*0

LOGICAL NSUM.VAR

BIG = 200.

VAR = .TRUE.

NSUM = BIG .GT. 200. .AND. VAR

2-18 60497800 A

/^\

0$B\

jf|S5\

NSUM is assigned the value .FALSE.

LOGICAL A,B,C,D,E,LGA,LGB,LGC

REAL F.G.H

A = B.AND.C.AND.D

A = F.GT.G.OR.F.GT.H

A = .NOT.(A.AND..NOT.B).AND.(C.OR.D)

LGA = .NOT.LGB

LGC = E.OR.LGC.OR.LGB.OR.LGA.OR.(A.AND.B)

MASKING ASSIGNMENT

/?$^y

v = masking expression

Replace the value of v with the value of the masking expression, v can be any type other than logical. No

type conversion takes place during replacement. If the type is double precision or complex, the value of the

expression is assigned to the ﬁrst word of the variable; and the least signiﬁcant or imaginary part set to

zero.

Examples:

B = D .AND. Z .OR. X

SUM - (1.0,2.0) .OR. (7.0,7.0)

NAME - INK .OR. JAY .AND. NEXT

J(3) =- N .AND. I

A•- B.OR. (C.AND. Z)

INTEGER I,J,K,L,M,N(16)

REAL B,C,D,E,F(15)

N(2) •= I.AND.J

B - C.AND.L

F(J) - I.OR..NOT.L.AND.F(J)

I - . . N O T. I

N(l) - I.OR.J.OR.K.OR.L.OR.M

MULTIPLE ASSIGNMENT

V , = V o = v = expression

60497800 A 2-19

Replace the value of several variables or array elements with the value of the expression. For example,

X = Y = Z = ( 1 0 + 2 ) / S U M ( l ) i s e q u i v a l e n t t o t h e f o l l o w i n g s t a t e m e n t s : ^

Z = ( 1 0 + 2 ) / S U M ( l ) • r

Y = ' Z "....,,'_",

X - Y ' ^

The value of the expression is converted to the type of the variable or array element during each

replacement. ^N

Examples:

NSUM :- BS U M - ISUM - TOTAL = 10.5 ~ 3 . 2

1,. TOTAL is assigned the value 7.3

2.. >JSUM is ^assigned, the value 7

/ 3; 3SUMis; assigned the value 7.0 - * !

4. NSJJM is. assigned the value 7

Multiple assignment is legal in .all types of! assignment statements.

/^$\

^S^rv

ytC*^\

2-20 60497800 A

SPECIFICATION STATEMENTS

J0$\

0$t$$?\

Speciﬁcation statements are non-executable; they deﬁne the type of a variable or array, specify the amount

of storage allocated to each variable according to its type, specify the dimensions of arrays, deﬁne methods

of sharing storage, and assign initial values to variables and arrays. The speciﬁcation statements are:

IMPLICIT

Type

DIMENSION

COMMON

EQUIVALENCE

EXTERNAL

ievei;'

Tte IMPLICIT statement must precedeJother/speciﬁcation statements.

If any of these statements appears after the ﬁrst executable statement or

statement function deﬁnition, the speciﬁcation statement is ignored and a

fatal diagnostic is printed.

The DATA statement, which is not a speciﬁcation statement, is also described in this section. The DATA state

ment must follow all other speciﬁcation statements except statement function deﬁnitions and FORMAT statements;

it can occur after the ﬁrst executable statement.

TYPE STATEMENTS

A type statement deﬁnes a variable, array, or function to be integer, real, complex, double precision, or logical.

An explicit type statement can be used to supply dimension information. The word TYPEmay be used as a

preﬁx.

In the absence of an explicit type statement, the type of a symbolic name is implied by the ﬁrst character

of the name: I, J, K, L, M, or N imply type integer and any other letter implies type real, unless ah

IMPLICIT statement is used to change this normal implied type.

Basic external and intrinsic functions are implicitly typed, and need not appear in a type statement in the

user's program. The type of each library function is listed in section 8.

EXPLICIT TYPE DECLARATIONS

There are ﬁve explicit type statements: INTEGER, REAL, COMPLEX, DOUBLE PRECISION, and LOGICAL.

INTEGER

nINTEGER name., name0 ,.. ..name

1 2 n

60497800 A 3-1

The symbolic names listed are declared as type integer.

Example:

INTEGER SUM, RESULT, ALIST

The variables SUM, RESULT and ALIST are all declared as type integer.

REAL

REAL name., name.,,..., name„

1 2 n

The symbolic names listed are declared as type real.

Example:

REAL NEXT<7), ITEM

NEXT is declared as an array with 7 real elements, and ITEM is declared as a real variable.

COMPLEX

/ COMPLEX name1, name2,..., namen

The symbolic names listed are declared as type complex.

Example:

COMPLEX ALPHA, NAM, MASTER, BETA

The variables ALPHA, NAM, MASTER, BETA are declared as type complex.

DOUBLE PRECISION

DOUBLE PRECISION name,, name0,..., name

l 2 n

The symbolic names listed are declared as type double precision. DOUBLE can be used instead of

DOUBLE PRECISION.

/*S"v

>f*St^*\

/tUe^.

y*=^£y

^*^?s

3-2 60497800 A

Jf»S

0$&K

0S$s

0^\

Example:

DOUBLE PRECISION ALIST, JUNR, BOX4

The variables ALIST, JUNR, BOX4 are declared as type double precision.

LOGICAL

LOGICAL name., name,,,..., name

1 ' 2 ' ' n

The symbolic names listed are declared as type logical.

Example:

LOGICAL P,Q,NUMBR4

The variables P, Q and NUMBR4 are declared as type logical.

IMPLICIT TYPE STATEMENT

type

IMPLICIT type, (ac, acn) type^ac, acn)

LOGICAL, INTEGER, REAL, DOUBLE PRECISION, DOUBLE, or COMPLEX

aCj, Single alphabetic characters, or ranges of characters represented by the ﬁrst and last character

separated by a minus sign.

This statement speciﬁes the type of variables, arrays, and functions beginning with the letters ac. Only one

IMPLICIT statement may appear in a.program unit, and it must precede other speciﬁcation statements. An

IMPLICIT statement in a function or subroutine .subprogram affects the type associated with dummy arguments

and the function name, as well as other variables in the subprogram.

Explicit typing of a variable name or array element ih a type statement or FUNCTION statement overrides an

ILLICIT speciﬁcation. /'

Example 1:

COMPLEX FUNCTION RHO (CDHS)

IMPLICIT INTEGER (A-D.R)

REAL ASUM

ASUM-BOR + ROR * ANEXT

60497800 A 3-3

,£*3!5\

E x a m p l e 2 : , _ , ; , - . - '■ _ • ; , \ V , ,■. . - ' , , , ' . . / - * < . * ' / , ' ' - " • "

tHPUfclT tNTEOER CA-ﬁH)

DIMENSION "E>t3*4? ^

OOHMON All>tBfCfO» Ft6»H }

EQUIVALENCE <A#£#I>

NANE:tISt/VLISft/AtBtCfqt£»Ft6tH»I ^

00 I J-m,U 12

WRITE (6»yU$T>

:' end ".":"• ■'

The arrays A and E and the variables B» C, D, F, H, and I are of type' integer; G is type real. ^

DIMENSION STATEMENT

rDIMENSION name,(d,),...,namen(dn)

/rfS^sv

dj Array declarator, 1-3 integer constants separated by commas. If name is a dummy param- "^

eter, d can be 1-3 integer constants or integer dummy parameters intermixed.

namej Symbolic name of an array. ^i

The DIMENSION statement is a nonexecutable statement which deﬁnes symbolic names as array names and

speciﬁes the bounds of the array. More than one array can be declared in a single DIMENSION statement. "^

Dummy parameter arrays speciﬁed within a procedure subprogram can have adjustable dimension speciﬁcations.

(A further explanation of adjustable dimension speciﬁcations appears under Procedure Communication in section

7). Within the same program unit, only one deﬁnition of an array is permitted. '"^

The number of computer words reserved for an array is determined by the type of the array and the product

of the subscripts. For real, integer and logical arrays, the number of words in an array equals the number of "*^

elements in the array. For complex and double precision arrays, the number of words reserved is twice the

product of the subscripts. No array can exceed 131,071 words.

Example:

COMPLEX BETA ^

DIMENSION BETA (2,3)

BETA is an array containing six elements; however, BETA has been deﬁned as COMPLEX and two words /"s3^

are used to contain each complex element; therefore, 12 computer words are reserved.

>*S!s*\

3-4 60497800 A

Example:

REAL NIL

DIMENSION NIL (6,2,2)

These statements could be combined into one statement with 24 words reserved for array NIL:

REAL NIL (6,2,2)

Example:

DIMENSION ASUM(10,2)

DIMENSION ASUM (3), VECTOR (7,7)

The second speciﬁcation of ASUM is ignored, and an informative message is printed. The speciﬁcation for

VECTOR is valid and is processed.

COMMON STATEMENT

COMMON/ N. v

i n

COMMON/blkname^v,, . . . ^/blknanr^/v.., . . .vn. . ./blknamen/v1, . . . ,vp

/fRv

0Sfc\

blkname;

C O M M O N v v

1 ' ' n

Block name or number. A block name is a symbolic name of 1-7 letters

or digits beginning with a letter. A block number is 1-7 digits; it must

not contain any alphabetic characters. Leading zeros are ignored. 0 is a

valid block number. The same block name or number can appear more

than once in a COMMON statement or a program unit; the loader links all

variables in blocks having the same name or number into a single labeled

common block.

Variable or array name which can be followed by constant subscripts

that declare the dimensions. The variable or array names are assigned to

blkname. The COMMON statement can contain one or more block

speciﬁcations.

60497800 A 3-5

// Denotes a blank common block. If blank common is the ﬁrst block in the

statement, slashes can be omitted.

Variables or arrays in a main program or subprogram can share the same storage locations with variables or

arrays in other subprograms by means of the COMMON statement. Variables and array names are stored in the

order in which they appear in the block speciﬁcation.

Example:

C0MM0N/BLACK/A(3)

DATA A/l.,2.,3./

COMMON ; 1,00/1 (4.)

DATA l/4;5"r,B,7/

COMMON M,N/CAT/ALPHA,BINGO//ADD

f**§S^k

y<*"$^\

/■*s*\

COMMON is a non-executable statement. See section 1 for proper location of COMMON statements relative

to other statements in the program unit. The COMMON speciﬁcation provides up to 125 storage blocks that ^\

can be referenced by more than one subprogram. A block of common storage can be labeled by a name or a

number. A COMMON statement without a name or number refers to a blank common block. Variables and

array elements can appear in both COMMON and EQUIVALENCE statements. A common block of storage can /^

be extended by an EQUIVALENCE statement; however, no common block can exceed 131,071 words..

All members of a common block must be allocated to the same level of storage; a fatal diagnostic is issued if /^m)

conﬂicting levels are declared. If only some members of a common block are declared in a LEVEL statement,

the remaining members of that common block are allocated automatically to the same level, and an informative

diagnostic is issued. "^

Block names can be used elsewhere in the program as variable or array names, and they can be used as.sub

program names. Numbered common is treated like labeled common. Data stored in common blocks by the '^

DATA statement is available to any subprogram using these blocks.

The length of a common block, other than blank common, must not be increased by a subprogram using /^

the block unless the subprogram is loaded ﬁrst.

Data may not be entered into blank common blocks by the DATA declaration.

The COMMON statement may contain one or more block speciﬁcations:

COMMON/X/RAG,TAG/APPA/Y,Z,B(5) ^

RAG and TAG are placed in block X. The array B and Y.Z are placed in block APPA.

Any number of blank common speciﬁcations can appear in a program. Blank, named and numbered

common blocks are cumulative throughout a program, as illustrated by the following example:

COMMON A,B,C/X/Y,Z,D//W,R

.♦•flC^MtSV

3-6 60497800 A

/f*S5s

ZfpN

/*$S\

These statements have the same effect as the single statement:

COMMON A,B,C,W,R,M,N,ADD/X/Y,Z,D/CAT/ALPHA,BINGO

Within subprograms, dummy arguments are not allowed in the COMMON statement.

If dimension information for an array is not given in the COMMON statement, it must be declared in a

type or DIMENSION statement in that program unit.

Examples:

C0MM0N/DEE/Z(10,4)

Speciﬁes the dimensions of the array Z and enters Z into labeled common block DEE.

COMMON/BLOKE/ANARAY,B,D

DIMENSION ANARAY(10,2)

COMMON/Z/X,Y,A

REAL X(7)

C0MM0N/HAT/M,N,J(3,4)

DIMENSION J(2,7)

In the last example. J is deﬁned as an array (3.4) in the COMMON statement. (2,7) in the

DIMENSION statement is ignored and an error message is printed.

The length of a common block, in computer words, is determined by the number and type of the variables

and array elements in that block. In the following statements, the length of common block A is 12 computer

words. The origin of the common block is Q( 1).

REAL Q,R

COMPLEX S

C0MM0N/A/Q(4),R(4),S(2)

Block A

origin Q(l)

Q(2)

Q(3)

Q(4)

R(l)

R(2)

R(3)

R(4)

S( 1) real part

S( 1) imaginary part

S(2) real part

S(2) imaginary part

If a program unit does not use all locations reserved in a common block, unused variables can be inserted

in the COMMON declaration to ensure proper correspondence of common areas.

60497800 A % 3_7

Example:

COMMON/SUM/A,B,c,D main program ■'

C0MM0N/SUM/E(3) ,D subprogram ^

If the subprogram does not use variables A,B, and C, array E is necessary to space over the area

reserved by A,B, and C.

Alternatively, correspondence can be ensured by placing unused variables at the end of the common list.

COMMON/SUM/D,A,B,C main program sas^

COMMON/SUM/D subprogram

If program units share the same common block, they may assign different names and types to the members "^

of the block; but the block name or numbers must remain the same.

Example: /«%

COMPLEX C

C0MM0N/TEST/C(2O) ^

The block named TEST consists of 40 computer words.

The subprogram may use different names for variables and arrays as in: 1

SUBROUTINE ONE J

COMPLEX A |

C0MM0N/TEST/A(10),G(10),K(10)

The length of TEST is 40 words. The first 10 elements (20 words) of the block represented by A are /4£^j

complex elements. Array G is the next 10 words, and array K is the last 10 words. Within the

subprogram, elements of G are treated as ﬂoating point; elements of K are treated as integer.

EQUIVALENCE STATEMENT

/! EQUIVALENCE (glist--) (glistn)

3-8 60497800 A

Each glist| consists of two or more variables, array elements, or array names, separated by commas.

Array elements must have integer constant subscripts. Dummy arguments must not appear in an equivalence „r^\

statement. Equivalenced variables must be in the same level of storage. If members of an equivalence class are

assigned to conﬂicting storage levels, a fatal error results. If a member of an equivalence class is declared in a

LEVEL statement, the other members of the class are automatically allocated to the same level, and an informa- z-*^

tive diagnostic is issued.

EQUIVALENCE is a non-executable statement and must appear before all executable statements in a pro- ^s^

gram unit. If it appears after the ﬁrst executable statement, a fatal diagnostic is printed. )

*c^\

**Si\

0^:

^^^ EQUIVALENCE assigns two or more variables in the same program unit to the same storage location (as

f^ opposed to COMMON which assigns two variables in different program units to the same location). Variables

or array elements not mentioned in an EQUIVALENCE statement are assigned unique locations.

^^ Example:

DIMENSION JAN(6),BILL(10)

f^ EQUIVALENCE (IRON,MAT,ZERO), (JAN(5),BILL(2)),(A,B,C)

The variables IRON. MAT and ZERO share the same location, the ﬁfth element in array JAN and the

^ second element in array BILL share the same location, and the variables A,B and C share the same location.

Wh.en an element of an array is referred to in an EQUIVALENCE statement, the relative locations of the

^ other array elements are, thereby, deﬁned also.

0^\

0$>\

Example:

DIMENSION Y(4), B(3,2)

EQUIVALENCE (Y(1),B(1,2)), (X,Y(4))

This EQUIVALENCE statement causes storage to be shared by the ﬁrst element in Y and the fourth

element in B and, similarly, the variable X and the fourth element in Y. Storage will be as follows:

B(l,l)

B(2,l)

B(3,l)

B(l,2) Y(l)

B(2,2) Y(2)

B(3,2) Y(3)

Y ( 4 ) X

The elements of a glist constitute an equivalence group. When an equivalence group contains an element that

appears in another equivalence group, these groups are merged and their elements constitute an equivalence

class.

Example:

DIMENSION A(100)

EQUIVALENCE (A.B), (C,A(50J), (D,E), (F,C)

These statements establish the following equivalence groups:

{A,B}, Ja,C}, {C,F}, JD,E[

and the following equivalence classes:

{a,b,c,f}, |d,e}

The statement EQUIVALENCE (A,B),(B,C) has the same effect as EQUIVALENCE (A,B,C).

60497800 A 3.9

^3\

Wjtjiii^pte .it/&a_ssurfted tojbe 1. ., : _. . .^ .'• ':,, /..,;.-.".-''.]"'.,-,,,' ••«- ,': V1 ;\:,- :■.•-*•;.

Di'MENSiqN.LZEBRACiQ,) .'• . .,••'"'>*]' ."'-.,', >'r j -'^ ' ' , \,'\. • , -■ •! J " •- *.. l' -'

EQUIVALENCE (ZEBRA,TIGER) ' " . - •**^%\

means the same as the statements:.

DIMENSION ZEBRA(10) , ' ^

EQUIVALENCE (ZEBRA(l),TIGER)

A logical, integer, or real entity equivalenced to a double precision or complex entity shares the same location yﬁ^\

as the real or most signiﬁcant part of the complex or double precision entity. If an array with single-word

elements is equivalenced to an array with double-word elements, the single word elements are stored contiguously,

so that each half of a double-word element is equivalent to a different single-word element. **%

An array with multiple dimensions may be referenced with a single subscript. The location of the element

in the array may be determined by the following method: *s\

DIMENSION A(K,M,N)

The position of element A(k,m,n) is given by: >

A(k+K*(m-1+M*(n-1)))*E)

E is 1 if A is real, integer or logical; E is 2 if A is complex or double precision.

Example:

DIMENSION AVERAG(2,3,4),TERM(7)

EQUIVALENCE (AVERAG(8),TERM(2))

Elements AVERAG (2,1,2) and TERM(2) share the same locations.

Two or more arrays can share the same storage locations.

Example:

DIMENSION ITIN(10,10),TAX(100)

EQUIVALENCE(ITIN(1),TAX(1))

500 READ (5,40)ITIN (1)

600 READ (5,70) TAX (1)

3~10 60497800 D

■***^\

J*lpsv

jf$S^

/fS^v

The EQUIVALENCE declaration assigns the ﬁrst elements of arrays ITIN and TAX to the same

location. READ statement 500 stores the array ITIN in consecutive locations. Before READ

statement 600 is executed, all operations involving ITIN should be completed; as the values of array

TAX are read into the storage locations previously occupied by ITIN.

Lengths of arrays need not be equal.

Examples:

DIMENSION ZER01(10,5),ZER02(3,3)

equivalence ( zeroi , ZER02) is a legal EQUIVALENCE statement

EQUIVALENCE (ITEM,TEMP)

The integer variable ITEM and the real variable TEMP share the same location; therefore, the same

location may be referred to as either integer or real. However, the integer and real internal formats

differ; therefore the values will not be the same.

EQUIVALENCE AND COMMON

Variables, array elements, and arrays may appear in both COMMON and EQUIVALENCE statements. A

common block of storage may be extended by an EQUIVALENCE statement.

Example:

COMMON/HAT/A(4),C

DIMENSION B(5)

EQUIVALENCE (A(2),B(1))

Common block HAT will extend from A(l) to B(5):

/HAT/ Origin A(l)

A(2) B(D

A(3) B(2)

A(4) B(3)

CB(4)

B(5)

EQUIVALENCE statements which extend the origin of a common block are not allowed, however.

Example:

COMMON/DESK/E,F,G

DIMENSION H(4)

EQUIVALENCE (E,H(3))

60497800 A 3-11

/*I"SN

The above EQUIVALENCE statement is illegal because H(l) and H(2) extend the start of the common

block DESK: <**s^v

/DESK/

Origin

H(l)

H(2)

H(3)

H(4)

An element or array is brought into COMMON if it is equivalenced to an element in COMMON,

in COMMON must not be equivalenced to each other.

Two elements

Examples:

COMMON A,B,C

EQUIVALENCE (A,B) illegal

COMMON /HAT/ A(4),C /X/ Y,Z

EQUIVALENCE (C,Y) illegal

As stated in section 1, the result of indexing outside of array bounds is unpredictable. Since the compiler attempts

to minimize the size of equivalence classes in common blocks to the smallest subset of the block that includes all

members named in associated EQUIVALENCE statements, all members of a common block will not necessarily be

considered as one array. The programming practice of intentionally referencing locations outside a known array

can produce unintentional results as shown in the following example.

/«**^k

COMMON/ /A(4), B, D, E

DIMENSION AA(4>

EQUIVALENCE (AA, A{2))

D=2.

E=2.

DO 10 1=1,6

AA(l)=D*E

PRINT \ E

When these statements are compiled under OPT=0, E will have a value of 8. on exit. Under OPT=l or 2,

the evaluation of D*E will be moved out of the loop since AA and D (or E) are not recognized as being

in the same equivalence class. If the program is to produce the same results under all OPT levels, AA

must be dimensioned to include the entire common block in the equivalence class.

LEVEL STATEMENT

LEVEL n, at ,...,an

'< !

<**^s!\

3-12

<****^V

60497800 C

00&\

'_..--Vi"-!a| : Variables or array names. " -i-v..\ \--: ■

n Unsigned integer 1, 2, or 3 indicating level to which list is to be allocated.

Small core memory resident (SCM)

Large core memory resident (LCM). Directly addressable (or word addressable)

/-m I 3 Large core memory resident, accessed by block transfer to or from small core memory

■ ^ through MOVLEV subroutine call (section 8)

Central memory resident

0&\ ^ 3 Extended core storage resident, accessed by block transfer to or from central memory

through MOVLEV subroutine call

The LEVEL statement speciﬁes the storage level of variables or array names. The storage level indicates the

f storage residence and mode of access for entities in a common block or for actual arguments associated with

dummy arguments. LEVEL statements must precede the ﬁrst executable statement in a program unit. Names

of variables and arrays which do not appear in a LEVEL statement are allocated to central memory.

/*sSSy

/*PN

/$$?\

{

/$^\

No dimension or type information is included in the LEVEL statement.

Variables and arrays appearing in a LEVEL statement can appear in DATA, DIMENSION, EQUIVALENCE,

COMMON, type, SUBROUTINE and FUNCTION statements. Data assigned to levels 2 and 3 must be

members of common blocks or dummy arguments.

For all levels, no single array or common block can exceed 131,071 words. If the total ECS, LCM, or UEM ﬁeld I

length accessed by the entire program exceeds 131,071 words, the LCM = I parameter must be speciﬁed on the

compiler call. When ECS and LCM blocks are loaded, the length assigned at compile time is rounded up to a

multiple of eight words.

_v Data assigned to level 3 can be referenced only in: COMMON, DIMENSION, type, EQUIVALENCE, DATA,

f CALL, SUBROUTINE, and FUNCTION statements. Level 3 items cannot be used in expressions.

No restrictions are imposed on the way in which reference is made to variables or arrays allocated to levels 1

and 2.

If the level of any variable is multiply deﬁned, the level ﬁrst declared is assumed; and a warning diagnostic

is printed.

All members of a common block or equivalence class must be assigned to the same level; a fatal diagnostic is

issued if conﬂicting levels are declared. If some, but not all, members of a common block or equivalence class

are declared in a LEVEL statement, all are assigned to the declared level, and an informative diagnostic is printed.

§ Applies only to Control Data CYBER 170 Model 176, CYBER 170 800 Series, CYBER 70 Model 76, I

and 7600 computers.

t Applies only to Control Data CYBER 70 Models 71, 72, 73 and 74, CYBER 170 Models 171, 172, 173,

0®s 174, 175, and 6000 Series computers.

60497800 J 3_13

/•"SK

If^a ^variable pr^array ..Same'peclared;in a LEVEL statement appears as art .actual argument in a CALL

statem^n^ W-cQrr^s^d r>e allocated tolthe same, level in the called

■"subprosram. <* - •-»•.?*. ."f-'h "**, » >*• v' > r** " "•'.-' \- - ■ ,ji' ; ■ , . "■"

^.dimension BC6O9),B(5o6),,0M(^boo)

, AycOMMbX/^CSBL'K/'y^V - \ t " .

/***SSy

' CALL' MOVLEV.'(CM»E, 1000)

The LEVEL' statement allocates arrays E and B to extended core storage. They are assigned to a named

common block, ECSBLIL Starting at location CM (the ﬁrst word address of the array CM), 1000 words of

central memory are'transfetred to the two arrays E and B in extended core storage by the library routine

MOVLEV, ''*•,>"*• ' -'••--- ' •■ « . . - ' - ' -

EXTERNAL STATEMENT

0£w%\

nEXTERNAL name , name.

name,,...,namen Subprogram names

^'^^^k

Before a subprogram name is used as an argument to another subprogram, it must be declared in an

EXTERNAL statement in the calling program. The subprogram can be user-supplied or can reside in the

FORTRAN library. If an actual argument is the name of an intrinsic function, and the user supplies a sub

program with the same name as the intrinsic function, the user-supplied function will be used. Otherwise,

the subprogram is supplied by the library. If the subprogram does not exist on the FORTRAN library and

is not supplied by the user, a loader error occurs.

In a chain of nested subroutine calls, where a,subprogram name is passed as an argument to lower level sub

routines, the EXTERNAL statement need only appear in the highest level program unit in which a CALL

statement containing the subprogram name appears.

Any name used as an actual argument in a call is assumed to be a variable or array unless it appears in an

EXTERNAL statement. An EXTERNAL statement must be used even if the subprogram concerned is a

basic external function, such as SQRT.

0^%K

3-14 60497800 H

Example:

Calling Program

EXTERNAL SIN, SQRT

CALL SUBRT(2.0,SIN,RESULT)

WRITE (6,100) RESULT

100 FORMAT (F7.3)

CALL SUBRT(2.0,SQRT,RESULT)

WRITE (6,100)RESULT

STOP

END

Subprogram

SUBROUTINE SUBRT (A,B,C)

X=A+3.14159/2.

C=B(X)

RETURN

END

First the sine, then the square root are computed; and in each case, the value is returned in RESULT.

The EXTERNAL statement must precede the ﬁrst executable statement, and always appears in the calling

program. (It cannot be used with statement functions.)

A function call that provides values for an actual argument does not need an EXTERNAL statement.

Example:

Calling Program Subprogram

C A L L S U B R T ( S I N ( X ) , R E S U LT ) SUBROUTINE SUBRT(A,B)

/f$B\

B=A

END

An EXTERNAL statement is not required because the function SIN is not the argument of the

subprogram; the evaluated result of SIN(X) becomes the argument.

DATA STATEMENT

DATA vlist1/dlistl/,vlist2/dlist2/ vlistn/dlistn/

vlist

DATA (vlist == dlist),..., (vlist = dlist)

List of array names, array elements, variable names, and implied DO loops, separated by commas.

Unless they appear in an implied DO loop, array elements must have integer constant subscripts.

60497800 H 3-15

^*e^K

dlist One or more of the following forms separated by commas:

/»"*"3!**»

X~-u&&\

constant

(constant list) -

rf*constant

rf*(consfarit list)

rf(constarit list)

constant list List of constants separated by commas.

rf Positive integer constant. The constant or

constant list is repeated the number of times

indicated by rf.

The data statement is non-executable and must follow all speciﬁcation statements except statement function

deﬁnitions, NAMELIST statements, and FORMAT statements. It can occur after the ﬁrst executable statement.

It assigns initial values to variables or array elements. Only variables assigned values by the DATA statement have

speciﬁed values when program execution begins. The DATA statement cannot be used to assign values in blank

common or to dummy arguments.

The number of items in the data list should agree with the number of variables in the variable list. If the data

list contains more items than the variable list, excess items are ignored, and an informative diagnostic is printed.

If the data list contains fewer items than the variable list, remaining variables are riot deﬁned, arid an informative

diagnostic is printed.

The type of the constant in the data list should agree with the type associated with the corresponding name

in the variable list. If the types do not agree, the form of the value stored is determined by the constant used

in the DATA statement rather than by the type of the name in the variable list.

When constants in a data list are enclosed in parentheses and preceded by an integer constant, the list is >

repeated the number of times indicated by the integer constant. If the repeat constant is not an integer, a

fatal errbr message is printed.

The forms: !••...-.. , ,

rf * (real constant, real constant) , ! "^

and: "'.•'.*•;,._. ^.

rf (real constant, real constant) ; . /^

are interpreted as a repeated speciﬁcationr'of two fe& constants," not as a single complex constant. In order to

specify the repetition of a complex constant, another set of parentheses must be used:

„ rf * ((real, constant, real constant))

-or: ..."'."

rf £(real constant, real constant)) /

Example: ." ; *.'■•: *'.,.__

2*(1.0, 2.0) Means repeat the real constants l.Q and 2,0 twice

2*((1.0, 2,0)) Means repeat the complex constant (1.0, 2.0) twice ;

3-16 60497800 A '

J0Wr\.

/jSES^

00^\

An unsubscripted array name implies the entire array in the order it is stored in memory.

Example:

INTEGER B(10)

DATA B/000077B,000064B,3*000005B,5*000200B/

The following octal constants are stored in ARRAY B:

77B

64B

200B

0&\

Example*

/"15\

PROGRAM OATA C (OUTPUT,TAPE6=0UTPUT)

COMPLEX Z(3),Z1

REAL Mk)

LOGICAL L

N4MEL I S T / 0 UT/I,L,X,Z1 , A , Z

DATA I,L,X,Zl,A,Z/5,.TRUE.,3.i'+159£6536,(2.i,-3.),2*(i.,2.)>

i 3*(<l.,-i.5*)/

W*ITE<6,0UT)

STOP

END

J0$&\ SOUT

j0&\ \L-

;'x

00$£\ Zl

L:A'

J0^\ $ -

T^eno1

= 5f

s T ' » • ' " : . - ; .

■■ •3l41S926536E*01t

«■ (.21E*>01»-.3E*01)»

* •lE*>pit »2E***01» .1E*01, .2E*0l»

= «.iE*0i»-.l5E*01>*» <.lE*01»-.l5E*0l>t (.lE*01t-.l5E*0l) t

60497800 A 3-17

0$3a\

Example: ''" - ', . 'A-.,.' ;' '• , '*' • ''-'"■' "'- '•-;••'•_"•.

the following are examples of alternative (nonstandard) forms of the DATA statement:

. DATA (X-3.)*(Y»5.)

Subsequent explicit setting of elements in the array, however, overrides implicit setting by means of a lengthy

Hollerith.constant. , , ' . .'-'■" - , ,

INTEGER ARAY(5)

DATA (A-=7.)»(B=200.),(ARAY-1,2,7,50,3) **k

COMMON/BOX/ARAY4(3,4,5)

DATA (ARAY4(1,3,5)=22.5) ^

Hollerith constants of any length can be included in the data list. If the constant corresponds to a one-word

variable (integer, real, or logical), the variable is set to the ﬁrst 10 characters of the Hollerith constant. If the "**')

constant corresponds to an array element, that element, and as many succeeding elements as necessary, are

ﬁlled by the constant.

Example:

DIMENSION JEF(IO) ^

DATA JEF(3) / 40HI THINK, THEREFORE I AM, I THINK/

T h e t h i r d t h r o u g h s i x t h e l e m e n t s o f t h e a r r a y J E F a r e s e t a s f o l l o w s : ' ' ]

, . - i

JEF(3) * 10HI THINK, T j

'JEF(4) - 10HHEREF0RE I '-• ■

" f J E F ( 5 ) > X 0 H A M , . , . . * ' " : , ; - ' - • • \

, J E F ( 6 ) • = 1 G H . . I T H I N K . ■ ' " * , . ^

•Example:' ' .'.. , ■*.__' , ; ',,'-' '-,'''""■,"..

: : I N T E G E R A B C ( 5 ) ' - J ' . - ' " ' . ' - ' . ' \ . ' ' - V . :

: ■' '• DATA ABC(3), ABG(4) /30HONCE UPONA MIDNIGHT DREARY , ^

, +10HA TIME .../ ' .--..-.•-•-,..:-.> - „ ;

"he third arid fourth elements of the array ABC are set as follows: ,. -*

ABC(3) = 10HONCE UPON

ABC(4) ■= 10HA TIME ...

If an unsubscripted array name appears in a variable list, each item in the data list sets exactly one array element, "^

even if Hollerith constants occur that exceed the array element length. The ﬁrst data list item sets the ﬁrst array

element, the second data list item sets the second array element, and so forth.

r**&£\

3-18 60497800 E

00$?\

yg^y

/5^KuJrr\

/p$\

/£$^N

Initially, the full length of each Hollerith constant is assigned to successive array elements; subsequent data list

items might cause overwriting of all or part of the constant.

Example:

DIMENSION J(4)

DATA J/20H1234567890ABCDEFGHIJ, 20HKLMNOPQRSTUVWXYZ+-*$/

The elements of array J are set as follows:

J(l) = 10H1234567890

J(2) = 10HKLMNOPQRST

J(3) •=• 10HUWXYZ+-*$

J (4) == undeﬁned

The ﬁrst Hollerith constant in the data list sets elements J(l) and J(2), and the second constant sets elements

J(2) and J(3), overriding the previous setting of J(2). J(4) remains undeﬁned.

Because of this method of storing Hollerith constants appearing in DATA statements, there are only two ways

in which an array can be set to a long character string:

The string can be speciﬁed as one long constant:

DATA J/40H1234567890ABCDEFGHIJKLMNOPQRSTUVWXYZ*f-*$/

or the string can be broken up into constants of 10 characters each:

DATA J/10H1234567890, 10HABCDEFGHIJ, 10HKLMNOPQRST, 10HUWZYZ+-*$/

If variables containing Hollerith data are to be compared with Hollerith constants, the variables should be of

type INTEGER so that the actual bit value is used and no conversion is performed.

IMPLIED DO IN DATA LIST

The implied DO can be used as a shortened notation for specifying items in the variable list of a DATA

statement. The implied DO in a DATA statement has the following form:

(varlist,i=m j ^2^3)

where:

varlist an array element name or another implied DO. If it is an array element name,

its subscript expressions must be of the form

M*i±N

where M and N are unsigned non-zero integer constants. ±N can be omitted.

i a simple integer variable called the index variable

mj.n^m^ unsigned integer constants specifying the initial value, terminal value, and

increment, respectively, for the index variable; if nv-j and the preceding comma

are omitted, the value of n^ is assumed to be 1.

60497800 E 3-19

The range of the implied DO is varlist. Within the range, the value of the variable i must not be redeﬁned.

If varlist contains more implied DOs, those implied DOs are considered to be nested within the containing ^

implied DO; the nested implied DO is completely processed for each value of i in the containing implied DO.

Implied DOs can be nested a maximum of three deep.

When an implied DO is encountered in a DATA statement, the elements in its range are initialized for index

variable i with the value mj. The index variable is then increased by m3 and, if i is less than or equal to

m2, the range of varlist is initialized for the new value of i. This procedure continues until the value of the ^

index variable exceeds m>-..

Example 1: '•^

REAL ANARAY(IO)

DATA (ANARAY(I),I «■ 1,10)/l.,2.,3.,7*2.5/

The values stored in array ANARAY are:

ANARAYO) I.

2.5

ir2.5

ANAR/-YY(IO) 2.5

When an implied DO is used, to store values into arrays, only one array name can be used within the implied

DO nest. ,'■' . [ ' * ' ' - ^ , * ■ ' ",'","

E x a m p l e 2 : " ' ■•,'■" " ^ ' * • . . . .

DIMENSION UNIT (10,10)

.DATA (UNtTU, \), l=1,1p)/f10*1:/

These two statements declare a matrix and preset the diagonal elements to ones.

Example 3:

DIMENSION ARdO}

DATA (AR(2*I+1). 1=1, 4)/4*3.5/

These two statements declare a ten-word array and preset elements AR(3), AR(5), AR(7), and AR(9) to

3.5,.,

Example 4:

DIMENSION AMASS(1Q,10,10),. A(10), B(5)

DATA (AMASS(6,K,3),K=l,10)/4*(-2.,5.139),6.9|,10./

DATA (A(I),I=5,7)/2*(4.1),5.0/ ***\

DATA B/5*0.0/

3-20 60497800 A

f»SB\

ijfJI^N

These statements dimension arrays AMASS, A, and B and preset elements as follows:

ARRAY AMASS: ARRAY A

i#^

AMASS(6,1,3) = -2. A( 5) = 4.1

AMASS(6,2,3) = 5.139 A ( 6 ) - 4.1

AMASS(6,3,3) = -2. A ( 7 ) - 5.0

AMASS(6,4,3) = 5.139

AMASS(6,5,3) = -2. ARRAY B

AMASS(6,6,3) = 5.139

AMASS(6,7,3) = -2. B ( l ) = 0 . 0

AMASS(6,8,3) = 5.139 B ( 2 ) - 0 . 0

AMASS(6,9,3) =6.9 B ( 3 ) = 0 . 0

AMA SS(6, 10,3) - 10 . B ( 4 ) = 0 . 0

B ( 5 ) = 0.0

ypp-\

Example 5:

Invalid: DATA (A(l). B(l), 1=1, 3)/1. .2. ,3. ,4., 5. ,6./

Example 6:

The statements:

DIMENSION D3(4),P0Q(5,5)

DATA (D3 = 5.,6.,7.,8.),(((P0Q(I,J),I=l,5),J=l,5)=25*O. )

Initialize:

D3(l) = 5.

D3(2) = 6.

D3(3) = .7.

D3(4) =8.

and set the entire array POQ to zero.

/*"S"\

0^*S

0^\.

60497800 A 3-21

FLOW CONTROL STATEMENTS

FORTRAN ﬂow control statements provide a means of altering, interrupting, terminating, or otherwise modifying

the normal sequential ﬂow of execution:

ASSIGN PAUSE

GO TO STOP

IF END

DO RETURN

CONTINUE

Control can be transferred only to an executable statement. .

A statement can be identiﬁed by an integer, 1-99999, with leading zeros and embedded blanks ignored. Each

statement label must be unique in the program unit (main program or subprogram) in which it appears.

GO TO STATEMENT

The three types of GO TO statements are unconditional, computed, and assigned. The ASSIGN statement is

used in conjunction with the assigned GO TO and is therefore described in the GO TO statement group.

UNCONDITIONAL GO TO STATEMENT

GO TO sn

sn is a label of an executable statement.

fjfPN

This statement transfers control to the statement labeled sn which must be an executable statement in the

current program unit. -*

Example:

10 A=B+Z

100 B=X+Y

IF(A-B)20,20,30

20 Z=A

GO TO 10-*

30 Z=B

STOP

END

•Transfers control to statement 10

60497800 A 4-1

COMPUTED GO TO STATEMENT

nGO TO (sn1 ,sn2 ,..., snm), iv

>f*^**\

GO TO (sn1,sn2,...,snfT>),eam

snj is a label on an executable statement.

iv is an integer variable.

, earn ~ is an' arithmetic of masking expression.

The computed GO TO statement transfers control to one of the statements referenced in the parentheses. If

the variable iv has a value of one, control transfers to the statement labeled snj; if the value is i, control

transfers to the statement labeled sn^.

The variable iv can .be replaced by an expression. The value of the expression is'truncated Jand converted to

an integer, if necessary,, and/used in'place of iv. The comma separating the- statement label from the variable

or iexprVsaton] is'optional, . [-^ ' \ ".!;.""* '"'.,*" . " " '',

The variable must not be speciﬁed by an ASSIGN statement. If it is speciﬁed by an ASSIGN statement, the

object code is incorrect, but no compilation error message is issued.

If the value of the variable'or expression is less than one or larger than the number of statement numbers in

parentheses, the transfer of control is undeﬁned and a fatal error results at execution time.

Example 1:

GO T0(10,20,30,20),L

.^^*s

/4SS\

'*aa%i

4-2 60497800 A

/*$p^\

The next statement executed is:

10 if L = 1

20 if L = 2

30 if L = 3

20 if L = 4

Example 2:

K=2

GO T0(100,150,300),K

Example 3:

Statement 150 is executed next.

K=2

X=4.6

GO T0(10,no,11,12,13),X/K Control transfers to statement 110, since the integer value of the

expression X/K equals 2.

Example 4:

M=4

GO TO (100,200,300),M

Execution of the last example causes a fatal error during execution because fewer than four numbers are

speciﬁed in the list of statement labels.

ASSIGN STATEMENT

ASSIGN sn TO iv

sn is a label of an executable statement.

iv is an integer variable.

The ASSIGN statement assigns a statement label to a variable used in an assigned GO TO. The integer

constant assigned to iv represents the label of an executable statement to which control may be transferred

by an assigned GO TO statement. Once iv is used in an ASSIGN statement, it must not be referenced in

any statement, other than an assigned GO TO or another ASSIGN, until it has been redeﬁned.

/0$v\

60497800 E 4-3

The assignment must be made prior to the execution of the assigned GO TO statement and sn (the label of

an executable statement) must be in the same program unit as both the ASSIGN and assigned GO TO

statements.

Example:

ASSIGN 10 TO LSWIT

GO TO LSWIT,(5,10,15,20) Control transfers to statement 10

ASSIGNED GO TO STATEMENT

/ ^ ^ i

GO TO iv, (sn1 ,...,snm)

". ) •7.-..K ->

M ^ :

. t V*- GO TO jv (sry. •vsn') t.

♦\;> '. •;

(sn-,. ..,snm)

is an integer variable.

is a list of all the statement labels to which control can be passed by this assigned

GO TO. Upon execution of the assigned GO TO, iv must be assigned to one of the

labels in the list.

The assigned GO TO statement transfers control to the statement label last assigned to iv by the execution of

a prior ASSIGN statement. All the statement labels in the list must be in the same program unit with both

the ASSIGN and the assigned GO TO statements. Omitting the list of statement labels causes a fatal error.

If a statement label is omitted from the list or the value of iv is deﬁned by a statement other than an

ASSIGN statement, the results are unpredictable. (Control is transferred to the absolute memory address

represented by the low order 18 bits of iv.) The corririia after iv is optional.

Example:

-•"^ ■v

ASSIGN 50 TO JUMP

10 GO TO JUMP,(20,30,40,50) Statement 50 is executed immediately after statement 10.

20 CONTINUE

30 CAT=ZERO+HAT

40 CAT=10.1-3.

50 CAT=25.2+7.3

4-4 60497800 F

'*-•*%

/*|PN

ARITHMETIC IF STATEMENT

The arithmetic IF statement has a three-branch and a two-branch form. In both cases, zero is deﬁned as a

word containing all bits set to zero or all bits set to one (+0 or -0). If the type of the evaluated expression

is complex, only the real part is tested.

THREE-BRANCH ARITHMETIC IF STATEMENT

IF (earn) sn-j. ai.2-.sn3

earn is an arithmetic or masking expression.

snj,sn2,sn-j are labels on executable statements.

The three-branch IF statement transfers control to the statement labeled snj if the value of the expression is

less than zero, to the statement labeled sn2 if it is equal to zero, or to the statement labeled s^ if it is

greater than zero.

Example:

PROGRAM IF (INPUT,OUTPUT,TAPE5=INPUT,TAPE6=0UTPUT)

READ (5,100) I,J,K,N

100 FORMAT (10X,4I4)

IF(I-N) 3,4,6

3 ISUM=J+K

6 CALL ERR0R1

WRITE (6 ,2) ISUM

2 FORMAT (110)

4 STOP

END

TWO-BRANCH ARITHMETIC IF STATEMENT

/SPN IF (earn) sn1#sn2

earn is an arithmetic or masking expression.

sni,sn2 are labels on executable statements.

60497800 A 4-5

^«fesJ\

The two-bjranch IF statement transfers control to one of two executable^statem]etttsr„ Control is transferred to

the,

sri2

Example-- v-'*,,V

statement labeled sn-j if the value of the expression is not equal to zero and 'to the statement labeled,

if ib'is* equal to-zero. *-. ', !" ^ * *\{ "-1' *.*■,~ * V,; .*<"!' " .-", !~ " -"

H* s -V5

- vll (I*J*DATA(K))100,101

ipO.IF<(I*Y*K)l65,lG6 -'r^'/,^

LOGICAL IF STATEMENT

The logical IF statement has a standard form and a.tytoTbranch form.

STANDARD-FORM LOGICAL IF STATEMENT

IF (elr) stat

elr is a logical or relational expression.

stat is any unlabeled executable statement other than DO, END, or another standard-form

logical IF.

The standard-form logical IF allows for conditional execution of a statement. If the logical or relational

expression is true, stat is executed. If the expression is false, stat is skipped.

Examples:

IF (P.AND.Q) RES=7.2

50 TEMP=ANS*Z

If P and Q are both true, the value of the variable RES is replaced by 7.2; otherwise, the value of RES

is unchanged. In either case, statement 50 is executed.

IF (A.LE. 2.5) CASH=150.

70 B=A+C-TEMP

If A is less than or equal to 2.5, the value of CASH is replaced by 150. If A is greater than 2.5, CASH

remains unchanged.

IF (A.LT.B) CALL SUB1

20 ZETA=TEMP+RES4

If A is less than B, the subroutine SUB1 is called. Upon return from this subroutine, statement 20 is

executed. If A is greater than or equal to B, statement 20 is executed and SUB1 is not called.

y^x

>«^x

4-6 60497800 A

TWO-BRANCH LOGICAL IF STATEMENT

/0^s

f 1- - .

IF (elr) sn|fsn2

elr is a logical or relational expression.

snj,sn2 are labels on executable statements.

The two-branch logical IF allows for transfer of control to one of two executable statements. If the value of

the logical or relational expression is true, control is transferred to the statement labeled snj. If the value of

the expression is false, control is transferred to the statement labeled s^J

Example: ./

'/ IF(K.ll(Si>100)60,70 ! •-_.-■.-

If K is equal'to 100, statement 60 is executed; otherwise statement 70 is executed.

DO STATEMENT

DO sn iv=m1,m2,m3

DO sn iv=m..,m2

/$^N

sn Terminal statement label; an executable statement that must physically follow and reside in

the same program unit as its associated DO statement. The terminal statement must not be

any arithmetic or two-branch logical IF, a GO TO, RETURN, END, STOP, PAUSE, or

another DO statement.

0^\

m-,

Control variable; an integer variable.

Initial parameter.

Terminal parameter.

Incrementation parameter.

Indexing parameters: unsigned integer or octal constants or

integer variables with positive non-zero values at execution such

that neither mi+m? nor mo+m? is larger than 2-1. If the

indexing parameters exceed these constraints, the results are

unpredictable. If m? is not speciﬁed, its value is assumed to be 1.

0S\

60497800 A 4-7

The DO statement makes it possible to repeat groups of statements and to change the value of an integer

variable during the repetition.

DO LOOPS

The range of a DO loop consists of all executable statements, from and including the ﬁrst executable state

ment after the DO statement to and including the terminal statement. Execution of a DO statement causes

the following sequence of operations: /CSSy

1. iv is assigned the value of m j.

2. The range of the DO loop is executed.

3. iv is incremented by the value of m^.

4. iv is compared with rc^. If the value of iv is less than or equal to the value of n^, the sequence

of operations starting at step 2 is repeated. If the value of iv is greater than the value of n^,

the DO is said to have been satisﬁed, the control variable becomes undeﬁned, and control passes

to the statement following sn. If mj is greater than or equal to n^, the range of the DO loop

is executed once. 1

A transfer out of the range of a DO loop is permissible at any time. When such a transfer occurs, the control

variable remains deﬁned at its most recent value in the loop. If control eventually is returned to the same range, x-stn

the statements executed while control is out of the range are said to deﬁne the extended range of the DO. The

extended range should not contain a DO statement. Subroutines or functions invoked within the range of a DO

can contain DO statements, however.

The control variable must not be redeﬁned in the range of a DO; such redeﬁnition causes a fatal-to-execution

diagnostic to be issued. The control variable should likewise not be redeﬁned in the extended range; such

redeﬁnition causes the results of execution to be unpredictable. 1

The indexing parameters should not be redeﬁned in either the range or the extended range of a DO. In ^^^

either case, the results of execution are unpredictable. Redeﬁnition in the range of the DO causes an 7

informative diagnostic to be issued.

Example 1: "^^

DO 10 1=1,11,3

IF(ALIST(I)-ALIST(I+1))15,10,10

15 ITEMP=ALIST(I)

10 ALIST(I)=ALIST(I+1)

300 WRITE(6,200)ALIST 1

The statements following DO up to and including statement 10 are executed four times. The DO

loop is executed with I equal to 1, 4, 7, 10. Statement 300 is then executed. ^

4-8 60497800 A

0$$$£\

Example 2:

DO 10 1=1,5

CAT=BOX+D

10 IF (X.GT.B.AND.X.LT.H)Z=EQUATE

6 A=ZERO+EXTRA

Statement 10 is executed ﬁve times, whether or not Z = EQUATE is executed. Statement 6 is

executed only after the DO loop is satisﬁed.

Example 3:

IVAR = 9

DO 20 I = 1,200

I F ( I - I VA R ) 2 0 , 1 0 , 1 0

20 CONTINUE

10 IN = I

An exit from the range of the DO is made to statement 10 when the value of the control variable I

is equal to IVAR. The value of the integer variable IN becomes 9.

Example 4:

K=3 :!.

J=5

DO 100 I=J,K

RACK=2.-3.5+ANT(I)

100 CONTINUE

The DO loop is executed only once (with 1 = 5) because J is larger than K.

NESTED DO LOOPS

When a DO loop entirely contains another DO loop, the grouping is called a DO nest. DO loops can be

nested to 50 levels. The range of a DO statement can include other DO statements providing the range of

each inner DO is entirely within the range of the containing DO statement.

The last statement of an inner DO loop must be either the same as the last statement of the outer DO loop

or must occur before it. If more than one DO loop has the same terminal statement, a transfer to that

statement can be made only from within the range (or extended range) of the innermost DO, and the label

cannot be referenced in any GO TO or IF statement in the nest except in the range of the innermost DO.

A DO loop can be entered only through the DO statement. Once the DO statement has been executed, and

before the loop is satisﬁed, control can be transferred out of the range and then transferred back into the

range of the DO.

60497800 A 4-9

/fP*,

"^^•V

A transfer from the range of an outer DO into the range of an inner DO loop is not allowed; however, a

transfer out of the range of an inner DO into the range of an outer DO is allowed because such a transfer

is within the range of the outer DO loop.

0Illegal -3 Legal

The use of and return from a subprogram within a DO loop are permitted. A transfer back into the range

of an innermost DO loop is allowed if a transfer has been made from the same loop.

Illegal

The extended range of an inner DO loop must be outside the outermost DO loop.

Example 1:

DIMENSION A(5,4,4), B(4,4)

DO 2 I - 1,4

DO 2 J = 1,4

DO 1 K - 1,5

1 A(K,J,I) = 0.0

2 B ( J , I ) - 0 . 0

This example sets arrays A and B to zero.

Example 2:

r—D1 r-D1

i—D2

•—n2

[n4

L-r»1

r—D1

— D2

— n2

— D3

— n3

I— n1

L— n1=n2 = n3

r*^\

(••s^v

/•^V

4-10 60497800 E

0$0£\

DO loops can be nested completely within an outermost loop or can share a terminal statement. The diagrams

in example 2 might be represented by the following code:

0S&\

DO 1- 1 = 1,10,2

DO 2 J=l,5

DO 3 K=2,8

•-3 CONTINUE

2 CONTINUE

CDO 4 L=l,3

4 CONTINUE

1 CONTINUE

DO 100 L=2,LIMIT

DO 10 J=l,10

1-10 CONTINUE

DO 20 K=K1,K2

1-20 CONTINUE

100 CONTINUE

DO 5 1=1,5

DO 5 J=I,10

DO 5 K=J,15

1—5 A = B*C

Example 3:

DO 10 J=l,50

DO 10 1=1,50

DO 10 M=l,100

GO TO 10

/fPN

10 CONTINUE

Since statement 10 is the terminal statement for more than one DO loop, it can be referenced in

a GO TO or IF statement in the range of the innermost DO. If 10 is referenced in one of the outer

loops, control is transferred out of the range with undeﬁned results.

60497800 A 4-11

Example 4:

DO 10 K=l,100

IF(DATA(K)-10.)20,10,20

20 DO 30 L=l,20

IF(DATA(L)-FACT*K-10.)40,30,40

40 DO 50 J=l,5

GO TO (101,102,50),INDEX

101 TEST=TEST+1

GO TO 104

103 TEST=TEST-1

DATA(K)=DATA(K)*2.0

JS^\

f**^S\

/*"H&v

50 CONTINUE

30 CONTINUE

10 CONTINUE <***^£\

GO TO 104

102 DO 109 M=l,3

109 CONTINUE

GO TO 103

104 CONTINUE

When an IF statement is used to bypass several inner loops, different terminal statements are required for each

loop.

CONTINUE STATEMENT

5 7

/T sn CONTINUE

sn is a statement label.

4-12 60497800 A

^an\

The CONTINUE statement performs no operation. It is an executable statement that can be placed anywhere

in the executable statement portion of a source program without affecting the sequence of execution. The

CONTINUE statement is most frequently used as the last statement of a DO loop. It can provide loop termina

tion when a GO TO or IF would normally be the last statement of the loop. If the CONTINUE statement does

not have a label, an informative diagnostic is provided.

Example 1 Example 2:

DO 10 I = 1,11

IF (A(I)-A(I+1))20,10,10

20 ITEMPP = A(I)

A (I ) = A ( 1 + 1 )

10 CONTINUE

PAUSE STATEMENT

DO 20 1=1,20

1 IF (X( I ) - Y(I))2,20,20

2 X(I)=X(I)+1.0

Y(I)=Y(I)-2.0

GO TO 1

20 CONTINUE

fPAUSE

fPAUSE n

rPAUSE •* c.c #

n is a string of 1-5 octal digits,

c.c is a string of 1-70'characters.

When a PAUSE statement is encountered during execution, the program halts and PAUSE n, or PAUSE c... c,

appears as a dayﬁle message on the operator console, and at the user terminal''if the job is executing interactively.

For batch originated programs, the console operator can continue or terminate the program with an entry from

the console.

For programs executing interactively through INTERCOM, the user types GO to continue execution or DROP to

terminate. For any other type-in, a diagnostic message is issued and INTERCOM waits for a correct type-in.

For programs executing interactively through the NOS 1 Time-Sharing System, the user types STOP '''

to terminate execution. Any other type-in causes execution to continue.

+Only 40 characters for SCOPE 2.

^Does not apply to SCOPE 2.

'"•"' Applies to TELEX only. For IAF, a terminating character must be used; for most terminals the terminating

character is CTRL/T or the ")" character.

60497800 F 4-13

STOP STATEMENT

nSTOP

STOP n ff-^SsfX

STOP *c..,c*

n is a string of 1-5 octal digits.

c.c - is a string of 1-70 characters,

The STOP statement terminates program execution. When a STOP statement is encountered during execution,

STOP n or STOP c . c is displayed in the dayﬁle, the program terminates, and control returns to the

operating system. If n is omitted, blanks are implied. A program unit can contain more than one STOP

statement.

,/3\

END STATEMENT

END

The END statement indicates the end of the program unit to the compiler. Every program unit must

physically terminate with an END statement. The END statement can follow a $ statement separator, be

labeled, and be continued. If control ﬂows into or branches to an END statement, it is treated as if a

RETURN statement had preceded the END statement.

If the END statement is not continued (all three characters are on the same line) no scanning for possible con

tinuation information is performed and any information,after the END statement is considered part of the next

program unit. If the END statement is continued (all three characters not on one line), any comment statements

and blank lines following the END statement are listed with the current program unit.

The following examples are interpreted as the end of one program unit, followed by another, program unit

beginning with an illegal continuation line of either „ FILE 3 or . - 4.

/^iV

END

.FIDT3

END

-s 4

r*£$>Z\

4-14 60497800 A

/8»fcV

RETURN STATEMENT

/ffS^i. /T

RETURN

RETURN i

0$r\

is a dummy argument which appears in the RETURNS list in the SUBROUTINE statement.

/0S\

The RETURN statement terminates the execution sequence within a program unit and normally returns

control to the current calling program unit. In a main program, execution of the program terminates and

control returns to the operating system when a RETURN is encountered.

When a RETURN statement is encountered in a function subprogram, control returns to the referencing

program unit and the evaluation of the expression is completed using the value returned from the function.

Since control must return to the referencing expression, a RETURN i statement in a function subprogram

causes a fatal error at compilation time.

/*S©\

In a subroutine subprogram, a RETURN statement transfers control to the next executable statement

following the CALL statement in the calling program unit.

A RETURN i in a subroutine transfers control to the calling program statement label corresponding to i

in the RETURNS list. It allows control to return to an executable statement other than the one immediately

following the CALL statement and can only be used in a subroutine subprogram.

The RETURNS list is described in more detail under Subroutine Subprogram and Calling a Subroutine Sub

program in section 7.

Example 1:

A = SUBFUN (D,E)

10 DO 200 1-1,5

FUNCTION SUBFUN(X,Y)

SUBFUN = X/Y

RETURN

END

When the RETURN statement is encountered in the function subprogram, control is returned to

the statement referencing the subprogram, and the value calculated by SUBFUN is stored in A.

60497800 A 4-15

. Example 2: ;

Calling Program Subprogram

CA1X;?GMI(A,B,C)>

XBﬁTUR^S (5,10)

5 B*SQRT(A*C)

10'1 >G^'fJ3W-&>iy >

SUBROUTINE PGM1(X,Y»Z),

, XRETURNS <M,N) '-'>'/' . :' \

> . <u*x**r , ' r ' -.'•""- : y;y y ',"* ' " :-

' ' x =z +x * Y. " • ♦ ,, ' - " - *" y * , * ' - • A

, 2 0 I F ( X l + x ) , 2 5 , 3 0 . , 3 5 , - . y ' ' j :

; 25 RETURN M Return is to statement 5 in calling program.

.30 RETURN,» Return is to statement 10 in calling program.

35 Z*Z+(X*Y) .

RETURN , Jteturn is to statement following CALL PGML

••' "END- y y* • >.* .*.-•".- * ^ '••■", y

* y .♦ -j *y'vV '*

f *.*

* ^*,t---^r • «■ ", i "i >-'

,^a^le^"Jshows,boA :forms o'f th/fefURN statement k'A MMj^lff? #P^:

/*^\

/*S^y

4-16 60497800 A

>^®\

INPUT/OUTPUT STATEMENTS

JPN

Processing resulting from input/output statements depends on the type of statement used. For each category,

there are one or more input statements and corresponding output statements. The categories are:

Formatted (READ, WRITE, PRINT and PUNCH statements with format designator)

Unformatted (READ and WRITE without format designator)

List-directed (READ, WRITE, PRINT and PUNCH with an asterisk as the format designator)

Namelist (READ, WRITE, PRINT and PUNCH with the NAMELIST group name replacing the format

designator)

Buffered (BUFFERIN and BUFFEROUT)

Mass storage input/output (Subroutines READMS, WRITMS, OPENMS, CLOSMS, and STINDX; see

section 8)

CYBER Record Manager interface routines (see section 8)

In addition, there are the core-to-core transfer statements ENCODE and DECODE and the ﬁle motion

statements REWIND, BACKSPACE, and ENDFILE, all discussed in this section.

\ Subprograms used in connection with input/output, besides the mass storage routines and the CYBER Record

Manager routines, include EOF, IOCHEC, UNIT, LENGTH, and LENGTHX. These subprograms are discussed

zp^ in section 8. Format speciﬁcations and input/output lists are discussed in section 6.

Input and output involve reading records from ﬁles and writing records to ﬁles. Every ﬁle must have a logical

rﬁle name of one to seven letters and digits, the ﬁrst a letter. The logical ﬁle name is deﬁned only for the

current job, and is the name by which the ﬁle is referred to in control statements.

rFor batch jobs (jobs not executed interactively at a terminal), certain ﬁle names have a predeﬁned origin or

destination. These ﬁle names are:

INPUT Data from user's source deck PUNCH Punched in Hollerith format at job termination

V OUTPUT Printed at job termination PUNCHB Punched in binary format at job termination

The ﬁles INPUT, OUTPUT, and PUNCH should be processed only by formatted, list-directed, or namelist input/

(' output statements.

The predeﬁned meaning of any ﬁle name except INPUT can be overridden by control statements.

( All ﬁles used by input/output statements or the mass storage subroutines must be declared on the PROGRAM

statement (discussed in section 7). Files processed by CYBER Record Manager routines, however, must not

^p^ be declared on the PROGRAM statement. The PROGRAM statement also allows the user to specify maximum

( record length and buffer size for a ﬁle. In the absence of user speciﬁcation, default values are provided.

60497800 B 5-1

/$$&2y

Mixing types of operations on the same ﬁle can sometimes lead to destruction of ﬁle integrity. In particular,

ﬁles processed by mass storage or CYBER Record Manager subroutines should be processed only by these /s-q-x

routines. Files processed by buffer statements should be processed only by these statements in a given pro

gram (REWIND, ENDFILE, and BACKSPACE are permitted for these ﬁles).

iolist Input/output list specifying items to be transmitted (section 6).

FORMATTED INPUT/OUTPUT

5-2 60497800 C

XS£*\

A ﬁle should not be processed both by unformatted operations on the one hand and by formatted, namelist,

or list directed operations on the other. However, if a ﬁle is rewound, it can then be rewritten in a different

mode.

If formatted, list directed, or namelist input/output is performed on a 7-track S or L tape, a FILE control

s t a t e m e n t t h a t s p e c i ﬁ e s C M = N O ( s e e s e c t i o n 1 6 ) m u s t b e i n c l u d e d i n t h e j o b . , ^

After every formatted, list directed, namelist, or unformatted READ, end-of-ﬁle status should be checked by

a call to the EOF function (section 8). If end-of-ﬁle is encountered, and EOF is not called, the contents of ^

the input/output list are undeﬁned.

Record length on card ﬁles should not exceed 80 characters. Record length on print ﬁles should not exceed ^

137 characters; the ﬁrst character is always used as carriage control and is hot printed. (Under the NOS 1 Time- 1

Sharing System, the ﬁrst character, is printed.) The second character appears in the ﬁrst print position. Carriage

control characters are listed in section 6. y^

The following mnemonics are used throughout this manual in descriptions of input/output statements and

subprograms: "^

u Input/output unit designator, used to determine the logical ﬁle name of the ﬁle to be used j

for input and output. The ﬁle name is derived from u depending on its value. The value i

can be one of the following:

Integer constant of one or two digits (leading zeros are discarded). The compiler associates ^^,

these numbers with ﬁle names of the type TAPEu, where u is the ﬁle designator (refer to

PROGRAM statement, section 7).

Simple integer variable name with a value of: )

0 - 99 or

A display code ﬁle name (L format, left justiﬁed with binary, zero ﬁll). This is the /^

logical ﬁle name.

fn Format designator; a FORMAT statement number or the name of an array, Variable, joi" /<^

array element containing the format speciﬁcation. The statement number must identify a

FORMAT statement in the program unit containing the input/output statement.

/ﬁS^v

For formatted input/output, a format designator must be present in the input/output statement. The input/ "^

output list is optional. Each formatted input/output statement transfers one or more records.

FORMATTED OUTPUT STATEMENTS

/T PRINT fn,iolist

/"S5\

PRINT fn

PRINT(u.fn) iolist

/^t\

PRINT(u.fn)

This statement transfers information from the storage locations named in the input/output list to the ﬁle

named OUTPUT (if u is omitted) or the ﬁle speciﬁed by u, according to the speciﬁcation in the format

designator fn. At the end of a job, if the user has not speciﬁed an alternate assignment, the ﬁle OUTPUT

is sent to the printer.

PROGRAM PRINT (OUTPUT)

A=1.2

L=3HYES

N=19

PRINT 4,A,L,N

4FORMAT (G20.6,A10,I5)

The iolist can be omitted. For example,

PRINT 20

20 FORMAT (30H THIS IS THE END OF THE REPORT)

/fsS\

60497800 F 5-3

,jf£^JV

PUNCH ..-f^X

PUNCH fn.iolist

rPUNCH fn

PUNCH(u,fn) iolist

PUNCH (u,fn)

Data is transferred from the storage locations speciﬁed by iolist to the ﬁle PUNCH (if u is omitted) or the

ﬁle speciﬁed by u. At the end of a job, if the user has not speciﬁed an alternate assignment, the ﬁle PUNCH

is output on the standard punch unit as Hollerith codes, 80 characters or less per card in accordance with

format speciﬁcation fn. If the card image is longer than 80 characters, a second card is punched with the

remaining characters.

0%8$>\

,XS^!\

ys3£?\

5 7 .

PROGRAM PUNCH (INPUT,OUTPUT,PUNCH)

2 READ 3,A,B,C

3 FORMAT (3G12.6)

ANSWER - A + B - C

IF (A .EQ. 99.99) STOP

PRINT 4, ANSWER

4 FORMAT (G20.6)

PUNCH 5,A,B,C,ANSWER

5 FORMAT (3G12.6,G20.6)

GO TO 2

END

i-rf-^S.

.<^v

^^v

The iolist can be omitted. For example,

PUNCH 30

30 FORMAT (10H LAST CARD)

5-4 60497800 C

WRITE

WRITE (u,fn) iolist

WRITE (u,fn)

WRITE fnriolist

0 S \

>*pS\

0S\

WRITE fn

The formatted WRITE statement transfers information from the storage locations named in the input/output

list to the ﬁle named OUTPUT (if u is omitted) or the ﬁle speciﬁed by u, according to the FORMAT

speciﬁcation, fn. At the end of a job, if the user has not speciﬁed an alternate assignment, the ﬁle OUTPUT

is sent to the printer.

PROGRAM RITE (OUTPUT,TAPE6=0UTPUT)

X=2.1

Y=3.

M=7

WRITE (6,100) X,Y,M

100 FORMAT (2F6.2.I4)

STOP

END

The iolist can be omitted. For example.

WRITE (4,27)

27 FORMAT (32H THIS COLUMN REPRESENTS X VALUES)

FORMATTED READ

READ (urfn) iolist

60497800 C 5-5

READ (u,fn)

fjf READ fhjolist

READ fn

These statements transmit data from unit u, or the ﬁle INPUT (if, u is omitted), to storage locations

named in iolist according to FORMAT speciﬁcation fn. The number of words in the list and the FORMAT

speciﬁcations must conform to the record structure on the input unit. If the list is omitted, one or more

FORTRAN records will be bypassed. The number of records bypassed is one plus the number of slashes

interpreted in the FORMAT statement. Except for information read into H speciﬁcations in the FORMAT

statement, the data in the records skipped is ignored.

The user should test for an end-of-ﬁle after each READ statement to avoid input/output errors. If an

attempt is made to read on unit u and an EOF was encountered on the previous read operation on this

unit, execution terminates and an error message is printed. (Refer to section 8, EOF Function.)

Example 1:

^^\

PROGRAM IN (INPUT,OUTPUT,TAPE4=INPUT,TAPE7=0UTPUT)

READ (4,200) A,B,C

200 FORMAT (3F7.3)

A = B*C+A

WRITE (7,50) A

50 FORMAT (50X,F7.4)

STOP

>*"8"*\

"^S *\

The READ statement transfers data from logical unit 4 (externally, the ﬁle INPUT) to the variables A,

B, and C, according to the speciﬁcations in the FORMAT statement labeled 200.

Example 2

PROGRAM RLIST (INPUT.OUTPUT)

READ 5,X,Y,Z

5 FORMAT (3G20.2)

RESULT = X-Y+Z

PRINT 100, RESULT

100 FORMAT (10X.G10.2)

STOP

END

<*=5n

5-6

The READ statement transfers data from the ﬁle INPUT to the variables X, Y, and Z, according to

the speciﬁcations in the FORMAT statement labeled 5.

60497800 C

,-fﬂSv

UNFORMATTED INPUT/OUTPUT

Unformatted READ and WRITE statements do not use format speciﬁcations and do not convert data in any

way on input or output. Instead, data is transferred as is between memory and the external device. Each

unformatted input/output statement transfers exactly one record. If data is written by an unformatted

WRITE and subsequently read by an unformatted READ, exactly what was written is read; no precision is

lost. Unformatted input/output cannot take place with coded tapes.

UNFORMATTED WRITE

WRITE (u) iolist

WRITE (u)

Example:

PROGRAM OUT(OUTPUT,TAPE10)

DIMENSION A(260),B(4000)

WRITE (10) A,B

END

This statement is used to output binary records. Information is transferred from the list variables, iolist, to

the speciﬁed output unit, u, with no format conversion. One record is created by an unformatted WRITE

statement. If the list is omitted, the statement writes a null record on the output device. A null record has

no data but contains all other properties of a legitimate record.

j-*$iSS\

UNFORMATTED READ

READ (u) iolist

P^^N,

READ (u)

60497800 A 5-7

One record is transmitted from the speciﬁed unit, u, to the storage locations named in iolist. Records are

not converted; no FORMAT statement is used. The information is transmitted from the designated ﬁle in

the form in which it exists on the ﬁle. If the number of words in the list exceeds the number of words in

the record, an execution diagnostic results. If the number of locations speciﬁed in iolist is less than the

number of words in the record, the excess data is ignored. If iolist is omitted, READ (u) spaces over one

record.

PROGRAM AREAD (INPUT,OUTPUT,TAPE2)

READ (2) X,Y,Z

SUM - X+Y+Z/2.

,-jS^JV

END

LIST DIRECTED INPUT/OUTPUT

list directed input/output involves the processing of coded records without a FORMAT statement. lEach

record consists of a list of values in a freer format than is used for formatted input/output. This type of

input/output is particularly convenient when the exact form of data is not important.

/>3™\

LIST DIRECTED INPUT

Data is transmitted from unit u or the ﬁle INPUT (if u is omitted) to the storage locations named in iolist. The

input data items are free-form with separators rather than in ﬁxed-size ﬁelds.

A list directed READ following a list directed READ that terminated in the middle of a record continues with

the same data record when no formatted READ statements have intervened. If, however, a formatted READ

statement has intervened, the remainder of the original record is destroyed. This is because a single working stor

age area is created for all formatted, list directed, and NAMELIST input/output ﬁles. For ﬁles referenced in list

directed READ statements, the maximum record length in characters should be speciﬁed in the PROGRAM state

ment (section 7). This speciﬁcation creates a separate working storage area for the ﬁle, which is different from

the default area.

When a list directed READ follows a formatted READ or a formatted READ follows a list directed READ, a new

data record is always read. Unformatted input/output does not require a working storage area and therefore does

not affect list directed input/output.

Input data consists of a string of values separated by one or more blanks, or by a comma or slash, either of which

may be preceded or followed by any number of blanks. Also, a line boundary, such as end of record or end of

card, serves as a value separator; however, a separator adjacent to a line boundary does not indicate a null value.

>*S^v

,+^^$£\

5-8 60497800 D

f*sev

j0&*\

00Z\

0$$\

0^.

Embedded blanks are not allowed in input values, except Hollerith values. The format of values in the input

record is as follows:

Integers Same format as for mrcgcr constants.

Real numbers Any valid FORTRAN formal for real numbers. In addition, the deciihal point can be

omitted; it is assumed to be to the right of the mantissa.

Complex numbers Two real values, separated by a comma, and enclosed by parentheses. The parentheses

are not considered to be a separator. The decimal point can be omitted from either of

the real constants.

Hollerith values A string of characters (which may include blanks) enclosed by two =£ characters. The =£

character can be represented within the string by two successive -?-= characters. Hollerith

values can only be read into integer variables or arrays. Values less than 10 characters

long are left justiﬁed and the word blank ﬁlled. Values longer than 10 characters are

truncated to 10 characters.

To repeat a value, an integer repeat constant is followed by an asterisk and the constant to be repeated. Blanks

cannot be embedded in the speciﬁcation of a repeated constant.

A null may be input in place of a constant when the value assigned to the corresponding list entity is not to

be changed. A null is indicated by the ﬁrst character in the input string being a comma or by two commas

or slashes separated by an arbitrary number of blanks. Nulls may be repeated by specifying an integer repeat

count followed by an asterisk and any value separator. The next value begins immediately after a repeated null.

A null cannot be used for either the real or imaginary part of a complex constant; however, a null can repre

sent an entire complex constant.

When the value separator is a slash, remaining list elements are treated as nulls; when the next input statement

is executed for this speciﬁed unit, the character following the slash becomes the ﬁrst input character for the

second READ. When the list is exhausted and no slash has been encountered, the next list directed read on

the same unit begins after the following value separator.

Input values must correspond in type to variables in the input/output list. Although the format of a real value

can be the same as that of an integer value, a repeated integer constant should not be read into variables of

different types.

For example:

READ(5,*) I. J. X, Y

0^ can read correctly:

2*7, 2*7 but not 4*7

f0^ assuming that I and J are integer and X and Y are real.

Repeated constants or repeated null values should be used entirely by one read.

The only Hollerith constants permitted are those enclosed in the symbol -7=. Hollerith constants can contain

embedded blanks. The paired symbols -?*•£ can be used to represent a single -»= within a character constant.

0xk A character cannot have a repeat count associated with it, and it must be read into an integer variable or

array. A character constant of less than 10 characters is padded on the right with blanks to ﬁll the word.

Only the ﬁrst 10 characters are used if the constant exceeds 10 characters.

60497800 E 5-9

Example.1:. ..,.,>

PROGRAM LDR <INPUT ^OUTPUT tTAPE5=INPUT♦TAPE6«0UTPUT)

NAMELlST/OUT/CAT»8IRDtDO(5

READ<5t*>CAT*6IRD»DQG

WRITE(6tOUT)

STOP

END

Input:. -,;,..

13>3, -5.2, »01

Output: : \ • ,

SOUT

■■'•■"/" CAT",'' •'.= .133E4-0 2,

BIRO *'r>52£>0.it

DOG = .IE-01,, ^S*s

Example. 2'-" ■'"', y ■ •• • "•. ' • ': '. --• '• '<

PROGRAM LDiN.(INPUT=100/80,OUTPUT=100) " J

CALL CONNEC(5LINPUT) •- , (

CALL CONNEC(6LOUTPUT)

. . , W R I T E , 8 . ,

8 FORMATP INPUT:'M4X,"OUTPUT:"//) ^

1 1 , R E A D * , J , K " ,

,IF(ECF(5LINPUT).NE.0) STOP

PRINT 22, J,K ***)

2 2 , F O R M AT ( T 2 0 , 2 1 5 )

,, GO TO 11

END .-<***.

>*£^^\

5-10 60497800 E

j*jj3S*\

INPUT: OUTPUT:

1 2 A pair of numbers

3/ Comma divides 2 and 3

3 2

4/ Null before comma (after previous slash)

Two nulls

One delimiter, one null

Second input value

Terminates for the null and 6

Null list

1 * , 7 One null, then a 7

2*8 Two8's

3 6

3 7

8 8

s0&\

0$$£\

1 2 ,3/, 4/,, , , , 5 , , 6/1* , 7 2*8 All together how

60497800 E 5-10.1

^"*^\

/^^K

/^§\

5-10.2 60497800 E

WRITE*,iolist

PRINT (u,*) iolist

PRINT*,iolist

rPUNCH(u,*) iolist

PUNCH*,iolist

Data is transferred from storage locations speciﬁed by the iolist to the designated ﬁle in a manner consistent

with list directed input.

PRINT and WRITE both output data to the ﬁle OUTPUT if no unit designator is speciﬁed. PUNCH outputs to

the ﬁle PUNCH if no unit designator is speciﬁed.

/*$^\

For ﬁles referenced in list directed WRITE, PRINT, and PUNCH statements, the maximum record length in char

acters should be speciﬁed in the PROGRAM statement (section 7).

List directed output is consistent with the input; however, null values, as well as slashes and repeated constants

are not produced. For real or double precision variables with absolute values in the range of 10 to 10 , an

F format type of conversion is used; otherwise, output is of the 1PE type. Trailing zeros in the mantissa and

leading zeros in the exponent are suppressed. Values are separated by blanks.

/3 S \

/^\

If a list item is an integer variable, array or array element, it is output in either integer or Hollerith format,

depending on its contents. If the upper 12 bits are all zeros (for positive numbers) or all ones (for negative

numbers) the item is output in integer format. In all other cases, the item is output in Hollerith format

(delimited by =-£ characters). For an array, only the first element is checked, and the whole array is output

accordingly. Therefore, a left-justiﬁed character string stored in an integer variable is usually output in

Hollerith format. Also, no integers with absolute value greater than 2 -1 can be written by list directed

output statements.

For list directed PRINT statements, a blank is output as the ﬁrst character (carriage control) of each record and

also as the ﬁrst character when a long record is continued on another line; for list directed WRITE statements, a

blank is output as the ﬁrst character of each record only.

List directed WRITE statements include the -?-= symbols with the character output; therefore, they should be used

if the list directed record output is to be input subsequently with a list directed READ statement

60497800 C 5-11

y^^\

On a connected ﬁle under NOS 1, if the iolist of a list directed output statement ends with a comma, no carriage

control or line feed takes place after the line is output. Under SCOPE 2 and NOS/BE 1, a comma as the last

character oi* an iolist is ignored.

Example 1:

PROGRAM H(0UTPUT*/80)

X » 3*6

PRINT«**THE VALUE OF SQRT<*t Xt *> IS »#t SQRT(X)

WRITE»t*SAME WITH WRITE, SQRT(*» Xt #) IS -* tSQRT(X)

STOP

END

Output:

1 -2 3 -<»

<7.,-i.) «-3.,2.) .00001

Example 3:

PROGRAM K (lNPUT,OUTPUT)

PRINT *," TYPE IN X",

READ*, X

PRINT*, " TYPE IN Y"

READ *, Y

END

Terminal listing under NOS 1:

TYPE IN X ? 1.234

TYPE IN Y

? 5.678

5-12 60497800 C

Output: ^

THE VALUE OF SQRT (3,6) IS =1.897366596131

/SAME WITH WRITE, SQRTC* 3.6 *> IS -t 1.897366596101 ^

Example 2:

PROGRAM LUW (OUTPUT=/80*TAPE6»OUTPUT) ^

INTEGER J(4)

COMPLEX Z<2)

OOUBLEPRECISION Q

OATA J»ZtQ/l»-2»3»-4»<7.»-l.>tt-3.f2.>»WD-5/

WRITE<6,*>J

WRITE(6,«)Z»Q

STOP

END

,/^KS

NAMELIST

The NAMELIST statement permits input and output of groups of variables and arrays with an identifying name.

No format speciﬁcation is used.

rNAMELIST/group name1 /a1 ,...,an/ ... /group namep/a1,... ,a

V group name Symbolic name which must be enclosed in slashes and must be unique within the program unit.

a1 ,...,an List of variables or array names separated by commas. Arrays with adjustable dimensions can-

#^ not appear.

The NAMELIST group name identiﬁes the succeeding list of variables or array names.

(^ A NAMELIST group name must be declared in a NAMELIST statement before it is used in an input/output state

ment. The group name may be declared only once, and it may not be used for any purpose other than a NAME-

LIST name in the program unit. It may appear in any of the input/output statements in place of the format

{ number. The group name cannot, however, be used in an ENCODE or DECODE statement in place of the format

number. When a NAMELIST group name is used, the list must be omitted from the input/output statement.

{ A variable or array name may belong to one or more NAMELIST groups.

Data read by a single NAMELIST name READ statement must contain only names listed in the referenced

/psv NAMELIST group. A set of data items may consist of any subset of the variable names in the NAMELIST. The

j value of variables not included in the subset remain unchanged. Variables need not be in the order in which they

f appear in the deﬁning NAMELIST statement.

{ Example:

_ PROGRAM NMLIST <INPUTt0UTPUTtTAPE5*INPUT,TAPE6»0UTPUT)

f NAMELIST/SHIP/A,8»C»11,12

READ<5tSHIP)

A IF <E0F(5>> 10t20

f^ 10 PRINT*, * NO DATA FOUND *

STOP

20 IF (C .LE. 0.) 40,30

f * 3 0 A . « 8 ♦ C

II =12 ♦ II

_ WRITE (6tSHIP)

f* ■"■ 40 STOP-.-.'

END

Input record:

'" 2 •:>1-^;' - .-)'-'■' ■...■■■ ' . -

fj$SHIP 'i&?$2 • 2, ?=20., C*3.4,11=8,12=50$

60497800 C 5-13

/$fe\

Output:

SSHIP

A•234E+02,

B=•2E*02,

Cs.34E*01t

11 58t

12 50,

SEND

INPUT

READ (u,group name)

j*«S|\

ﬁREAD group name

When a READ statement references a NAMELIST group name, input data in the format described below is

read from the designated ﬁle. If the speciﬁed group name is not found before ehd-of-ﬁle, a fatal error occurs.

If the ﬁle is empty, control returns to the statement following the READ; however, a subsequent read on the

same ﬁle will result in a fatal error. Consequently, a NAMELIST read should be followed by a test for end-of-

ﬁle, '

Data items succeeding $ NAMELIST group name are read until another $ is encountered.

Blanks must hot appear:

Between $ and NAMELIST group name

Within array names and variable names

Blanks may be used freely elsewhere.

More than one record can be used as input data in a NAMELIST group. The ﬁrst column of each record is

ignored. All input records containing data should end with a constant followed by a comma; however the last

record may be terminated by a $ without the ﬁnal comma. >^^\

5-14 60497800 C

A sample format for NAMELIST data on cards is as follows:

■terminator

00$t\

0&\

/£$&\

/a$

/karray name(n) = constant,... .constant,

Aarray name=constant,..., constant,

/a vanabie=constant,

/*$group name

data items

Data items separated by commas may be in three forms:

variable = constant

array name = constant,...,constant

array name (unsigned integer constant subscripts)=constant,...,constant

Omitting a constant constitutes a fatal error.

Constants can be preceded by a repetition factor and an asterisk.

Example:

5 * (l"-. 7, -2.4) ﬁve complex constants.

Constants may be integer, real, double precision, complex or logical. Logical constants must be of the form:

.TRUE. T. T .FALSE. .F. or F. A logical variable may be replaced only by a logical constant. A complex

variable may be replaced only by a complex constant. A complex constant must have the form (real

constant, real constant). Any other variable may be replaced by an integer, real or double precision

constant; the constant is converted to the type of the variable.

OUTPUT

rr WRITE(u,group name)

60497800 C 5-15

/!PSy

/T PRINT(u,group name)

/"^^\

-^^V

,**~%\

PUNCH(u.group name)

WRITE group name

ﬂ

PRINT group name

PUNCH; group name

All variables and arrays, and their values, ih the list associated with the NAMELIST group name are output on

the designated ﬁle, OUTPUT, or PUNCH. They are output in the order of speciﬁcation in the NAMELIST

statement; Output consists of at least three records. The ﬁrst record is a $ in column 2 followed by the group

name; the last record is a $ in column 2 followed by the characters END. Each group begins at the top of a

new page. ■ ,-'',"■':'-'*" "'

Example:

PROGRAM NAME (INPUT,OUTPUT, TAPE5«= INPUT , TAPE6«0UTPUT)

NAMELIST/VALUES/TOTAL,QUANT,COST

OATA OUANT,COST/15.,3.02/

TOTAL = QUANT»COST»1.3

WRITE (6,VALUES)

STOP

END

5-16 60497800 C

*£^t

Output

• SVALUES

f^ TOTAL = .58889999999999E*02»

QUANT = .15E*02,

f COST « .302E*01,

/^ SEND

No data appears in column 1 of any record. If a constant would cross column 80, the columns up to and

f" including 80 are ﬁlled with blanks instead and the constant begins in column 87; therefore, card boundaries will |

not be crossed if data is punched. The maximum length of any record is 136 characters (unless a smaller maxi

mum record length has been speciﬁed in the PROGRAM statement). Logical constants appear as T or F. Elements

of an array are output in the order in which they are stored.

y^f^N,

ui aii array arc uuipui m uie uiuei in wiuun uiey aie siuicu.

Records output by a WRITE (u, group name) statement may be read by a READ (u, group name)

statement using the same NAMELIST name.

Example:

NAMELIST/ITEMS/X,Y,Z

WRITE (6,ITEMS)

/f5?V

Output record:

pSITEHS

X » ,73*2E*03»

/SSN Y » ,23749E*0*»

Z * ,222SE*02»

$£N0

This output may be read later in the same program using the following statement:

READ(5,ITEMS)

f * * A R R A Y S I N N A M E L I S T

In input data the number of constants, including repetitions, given for an array name should not exceed the

number of elements in the array.

60497800 C 5-17

j ® \

Example:

INTEGER BAT(10) y

NAMELIST/HAT/BAT,DOT

READ (5,HAT) ,,,,.> ■ - \ ^ v

/j$ HAT BAT=2,3, 8*4, D0T= 1.05 $END

I ' ' . / * S n

The value of DOT becomes 1.05, the array BAT is as follows:

B A T ( 1 ) 2 y « x

B A T ( 2 ) 3

B A T ( 3 ) 4

B AT ( 4 ) 4

B A T ( 5 ) 4

B AT ( 6 ) 4

B AT ( 7 ) 4

B AT ( 8 ) 4

BAT(9) .4

BAT(10) 4 .

Example:

**^^\

■ DIMENSION GAY(5) ;

NAMELIST/DAY/GAY,BAY,RAY j

READ (5,DAY) j

Input Record:

2 . _ _

/ j$DAY GAY(3)~7.2,GAY(5)=3.0,BAY=2.3,RAY»77.2$

array element = constant,.,Mconstant

When data is input in this form, the constants are stored consecutively beginning with the location given by

the array element. The number of constants need not equal, but may not exceed, the remaining number of ^^\

elements in the array.

Example: ^^

DIMENSION ALPHA (6) ,

NAMELIST/BETA/ALPHA,DELTA,X,Y ^

READ (5,BETA)

Input record: ; . , ^^\

rJ$BETA ALPHA(3)-=7.,S.,9.,,DELTA=2.$ . „ '.- 1 ■-,.-•, ,.

5-18 60497800 C

J0&S

0S\

y$^N

0$^\,

0$®*\

In storage

ALPHA(3) 7.

ALPHA(4) 8.

/fﬂ-*v ALPHA(5) 9.

D E L T A 2 .

0m, Data initialized by the DATA statement can be changed later in the program by the NAMELIST statement.

Example:

PROGRAM COSTS (INPUT,OUTPUT,TAPE5=INPUT,TAPE6-0UTPUT)

DATA TAX,INT,ACCUM,ANET/23.,10,500.2,17.0/

NAMELIST/RECORDS/TAX,INT,ACCUM,ANET

FIRST - TAX + INT

SECOND - FIRST * SUM

READ(5; RECORDS)

Input Record

Example:

•$RECORDS TAX-27.', ACCUM=666.2$

DIMENSION Y(3,5)

LOGICAL L

COMPLEX Z

NAMELIST/HURRY/I1,I2,I3,K,M,Y,Z,L

READ (5,HURRY)

Input record:

$HURRY 11=1,L=.TRUE.,12=2,13=3.5,Y(3,5)=26,Y(1,1)=11,

12.0E1,13,4*14,Z=(1.,2.),K=16,M=17$

produce the following values:

11=1 Y(l,2)=14.0

12=2 Y(2,2)=14.0

13=3 Y(3,2)=14.0

Y(3,5)=26.0 Y(l,3)=14.0

Y(l,l)=11.0 K=16

Y(2,l)=120.0 M=17

*s± Y(3,l)=13.0 Z=(l.,2.) The rest of Y is unchanged.

L-.TRUE.

60497800 C 5-19

BUFFER STATEMENTS

Buffer ihpiit/output statements (BUFFER IN and BUFFER OUT) allow input/output operations to occur simul- <

taneously with other processing. They differ from formatted and unformatted READ and WRITE statements in

the following ways: ^^

A buffer statement initiates data transmission and then returns control to the program so that it can perform

other tasks while data transmission is in progress. A READ or WRITE statement completes data transmission

before returning control to the program. '

In a buffer statement, parity must be speciﬁed by a parity indicator. In.a READ or WRITE statement, parity

is implied by the form of the statement: an unformatted READ or WRITE implies binary mode, and a for- '

matted READ or WRITE implies coded mode.

READ and WRITE are associated with an input/output list. Buffer statements are not associated with a list;

data is transmitted to or from a block of storage.

A ﬁle processed by buffer statements must not be processed in the same program by formatted or unformatted

input/output statements, or by mass storage or CYBER Record Manager subroutines. ENDFILE, REWIND, and

BACKSPACE are valid, however. The user buffer size speciﬁcation in the PROGRAM statement (section 7) should

be set to zero for these ﬁles; ﬁeld length requirements are reduced for each ﬁle for which this is done.

Each buffer statement deﬁnes the location of the ﬁrst and last words of the block of memory to or from which

data is to be transmitted. The address of the last word must be greater than or equal to the address of the ﬁrst

word. The relative locations of the ﬁrst and last word are deﬁned only if they are the same variable or are in the

same array, common block, or equivalence class. If the ﬁrst and last words do not satisfy one of these relation

ships, their relative position is undeﬁned and a fatal error might result at execution time.

If the ﬁrst word and the last word are in the same common block but not in the same array or equivalence class,

optimization might be degraded.

BUFFER IN

BUFFER IN (u,p) (a,b)

5-20 60497800 J

After execution of a buffer statement has been initiated, and before referencing the same ﬁle or any of the contents

of the block of memory to or from which data is transferred, the status of the buffer operation must be checked 'i

by a reference to the UNIT function (section 8). This status check ensures that the data has actually been trans

ferred, and the buffer parameters for the ﬁle have been restored. If a second buffer operation is attempted on the

same ﬁle without an intervening reference to UNIT, the results are undeﬁned, "]

Under SCOPE 2, the block of memory speciﬁed by the buffer statement is not used as the actual buffer; all buffers

are maintained in an operating system buffer area in LCM. Therefore, the execution of a buffer statement results ")

in transfer of data between system buffers in LCM and the SCM area speciﬁed in the buffer statement.

On a CYBER 170 Model 176, a FILE control statement (section 16) specifying SBF=NO must be provided if a ****)

I level 2 or 3 (LCM) variable is used in a buffer statement.

/••K5N

.fﬂ^

u Unit designator.

p Integer constant or simple integer variable specifying the magnetic tape data conversion mode. The param

eter is applicable only when tape is assigned to the unit. Zero designates conversion (coded mode) and one

designates no conversion (binary mode). For the tape ﬁle characteristic, such as parity, refer to the NOS/BE

Reference Manual. Use of coded SI tapes under NOS results in a job abort by the system. Under NOS/BE,

coded SI 9-track tapes are written in binary mode.

a First variable or array element of block of memory to which data is to be transmitted.

b ' Last variable or array element of block of memory to which data is to be transmitted. If u is a unit

designator for a tape or mass storage device, the block of memory to which data is to be transmitted should

be one word larger than logically required. The additional word is needed to receive an error status from the

operating system if an input/output error occurs. Under SCOPE 2, the additional word is not needed because

no error status word is written.

BUFFER IN transfers one record from the ﬁle indicated by the unit designator u to the block of memory beginning at a

and ending at b. If the record is shorter than the block of memory, excess locations are not changed. If the record is

longer than the block of memory excess words in the record are ignored, except when the record type is ﬁxed (RT=F on

FILE statement), in which case an error occurs.

After UNIT has been referenced, the number of words transferred to memory can be obtained by a call to the function

LENGTH or the subroutine LENGTHX (section 8). If records do not terminate on a word boundary (in a ﬁle not

written by BUFFER OUT), the exact length of the record is returned by LENGTHX in terms of words and excess bits.

For ﬁles with record type S, if the end of a system-logical-record (end-of-section) is encountered on a ﬁle other than

INPUT, no data is transferred, and the length returned by LENGTH is zero. The next BUFFER IN begins reading after

/*BS the end-of-section. (On INPUT, end-of-section is treated as end-of-partition). End-of-section is transparent to ﬁles

j with other record types.

| The UNIT function should be used to test for end-of-ﬁle after BUFFER IN; the EOF function cannot be used for this

purpose.

/Pn Example 1:

DIMENSION CALC(51)

BUFFER IN (1,0) (CALC(1),CALC(50))

/£ S \

/^&N

/!^\

/"*S\

Coded information is transferred from logical unit 1 into storage beginning at the ﬁrst word of the array,

CALC(l), and extending through CALC(50).

Example 2:

PROGRAM TP (TAPF1«0,OUTPUT I

TMTCGER RFC(513>#RNUMR

RF.WTNn 1

nn a rnmmp « 1,10000

1 rijcff/q tn *1»1> IRECC1 !,R?CC512I )

2 TF (UNTTﬂM 3,5,5

3 k-st CNGTHf 1 \

' 1ENGTH RFTIJPNS (MUMPER OF ViORDS TR/-MS FFRRFP BY fMJFFFR TN

t\ prtk't 1CCRNUMP, (RFC ft>» I" 1*K)

30^ FORMAT f7HCRECPPDtT5/«1*,10U0»1

5 S T n p

60497800 J 5-21

Binary information is transferred from logical unit 1 into storage beginning at the ﬁrst word of the array,

REC(l), and extending through the last word of the array, REC(512). The UNIT function tests the status

of the buffer operation. If the buffer operation is completed without error, statement 3 is executed. If an

end-of-ﬁle or a parity error is encountered, control transfers to statement 5 and the program stops.

BUFFER OUT

j BUFFER OUT (u,p) (a,b)

u,p,a,b are the same as for BUFFER IN

BUFFER OUT writes one record by transferring the contents of the block of memory beginning at a and

ending at b to the ﬁle indicated by the unit designator u at the parity (even or odd) indicated by p. The

length of the record is the terminal address of the record (LWA) — starting address (FWA) +1, except for

ﬁxed-length records (RT=F on FILE statement), in which case the record length is the length (characters)

speciﬁed on the FILE statement (FL parameter). If FL is greater than (LWA - FWA + 1) x 10, an error

occurs.

The UNIT function must be referenced before another reference is made to the ﬁle or to the contents of the

block of memory.

ENCODE AND DECODE

The ENCODE and DECODE statements are used to reformat data in memory; information is transferred under

FORMAT speciﬁcations from one area of memory to another.

ENCODE is similar to a formatted WRITE statement, and DECODE is similar to a formatted READ statement.

Data is transmitted under format speciﬁcations, but ENCODE and DECODE transfer data internally; no peri

pheral equipment is involved. For example, data can be converted to a different format internally without the

necessity of writing it out on tape and rereading under another format.

ENCODE

r*€~«t™Qr\

• ® \

ENCODE (c,fn,v) iolist

Variable or array name which supplies the starting location of the record to be encoded.

Unsigned integer constant or simple integer variable specifying the length of each record.

The ﬁrst record starts with the leftmost character of the location speciﬁed by v and continues for c characters, 10 char

acters per computer word. If c is not a multiple of 10, the record ends before the end of the word is reached; and the

remainder of the word is blank ﬁlled. Each new record begins with a new computer word. There is no intrinsic limit

on c, except if v is a level 2,variable c must be less than or equal to 150. (ENCODE does not write a zero-byte termi

nator.) Additionally the target variable area v is blank ﬁlled for c characters prior to transfer.

fn Format designator, statement label or integer variable, which must not be a NAMELIST group

name or an *.

iolist List of variables to be transmitted to the location speciﬁed by v.

5-22 60497800 F

Example:

/*9\

PROGRAM ENCDE (OUTPUT)

INTEGER A(2),ALPHA<4)

DATA A,B»C/10HABCDEFGHIJ,10HKLMNOPQRST,5HUVWXY»7HZ123456/

ENCODE (40,1,ALPHA)A,B,C

FORMAT (2A4,A5,A6)

PRINT 2,ALPHA

FORMAT (20H1CONTENTS OF ALPHA = *,4AlO)

STOP

END

/iS^s

In memory after ENCODE statement has been executed:

Xpw\

ABCDKLMNUV WXYZ12345

ALPHA (1) ALPHA (2) ALPHA (3) ALPHA (4)

If the list and the format speciﬁcation transmit more than the number of characters speciﬁed per record, an

execution error message is printed. If the number of characters transmitted is less than the length speciﬁed

by c, remaining characters in the record are blank ﬁlled.

For example, in the following program which is similar to program ENCDE above, the format statement

has been changed, so that two records are generated by the ENCODE statement. A(l) and A(2) are written

with the format speciﬁcation 2A4, the / indicates a new record, and the remaining portion of the 40

character record, c, is blank ﬁlled. B and C are written into the second record with the speciﬁcation A5 and

A6, and the remaining characters are blank ﬁlled. The dimensions of the array ALPHA must be increased

to 8 to accommodate two 40-character records.

PROGRAM

INTEGER

DATA A,

ENCODE

FORMAT

PRINT 2

FORMAT

STOP

EMU

TWO (OUTPUT)

A(2)»ALPHA<8)

B»C/10HABCDEFGHIJ,l0HKLMNOPQRST,5HUVWXY,7HZ123456/

(40,1,ALPHA)A,B,C

(2A4/A5,A6)

,ALPHA

(20H1CO'NTENTS OF ALPHA =,6A10)

60497800 E 5-23

UVWXYZ12345

Output:

CONTENTS OF ALPHA =ABCOKLMN

If this same ENCODE statement is altered to:

ENCODE (33,1,ALPHA)A,B,C

1 FORMAT (2A4/A5.A6)

the contents 6f ALPHA remain the same. When a record ends in the middleof a word the remainder

of the word is blank ﬁlled (each new record starts at the beginning of the word).

y^Sv

>5l^?V

Record 1 Record 2

ABCDKLMN blank

ﬁll UVWXYZ1234 5 blank

ﬁll

ALPHA(i) ALPHA(2) ALPHA(3| ALPHA(4) ALPHA(5) ALPHA(6) ALPHAS) ALPHA{8»

end of record end of record

The array in core must be large enough to contain.the total number of characters speciﬁed in the ENCODE

statement. For example, if 70 characters are generated by the ENCODE statement, the array starting at

location v (if v is a single word element) must be dimensioned at least 7. If 27 characters are generated, the

array must be dimensioned 3. If only 6 characters are generated, v can be a 1-word variable.

ENCODE may be used to calculate a ﬁeld deﬁnition in a FORMAT speciﬁcation at object time. Assume

that in the statement FORMAT (2A 10,1m) the programmer wishes to specify m at some point in the

program. The following program permits m to vary in the range 2 through 9.

IF(M.LT.10.AND.M.GT.1)1,2

1 ENCODE (10,100,SPECMAT)M

100 FORMAT (7H(2A10,I,I1,1H))

PRINT SPECMAT,A,B,J

/^S\

<*«?\

5-24 60497800 D

/sS^I

M is tested to ensure it is within' limits; if it is not, control goes to statement 2, which could be an

error routine. If M is within limits, ENCODE packs the integer value of M with the characters

(2A10J ). This packed FORMAT is stored in SPECMAT. SPECMAT contains (2A10,Im).

A and B will be printed under speciﬁcation A10, and the quantity J under speciﬁcation 12, through

19 according to the value of m.

The following program is another example of forming FORMAT statements internally:

PROGRAM IGEN (OUTPUT,TAPE6=OUTPUT)

DO 9 J=l,50

ENCODE (10,7,FMT)J

7 FORMAT (2H(I,I2,1H))

9 WRITE (6,FMT)J

STOP

END

In memory, FMT is ﬁrst (I 1) then (I 2), then (I 3), etc.

A variable should not be encoded or decoded upon itself, as this gives unpredictable results.

DECODE

DECODE (c,fn,v) iolist

c, fn, and v are the same as for ENCODE.

iolist is the list to receive variables from the location speciﬁed by v. iolist conforms to the

syntax of an input/output list.

DECODE is a core-to-core transfer of data similar to formatted READ. Display code characters in a variable or

an array, v, are converted under format speciﬁcations and stored in the list variables, iolist. DECODE reads from

a string of display code characters in an array or variable in memory; whereas the READ statement reads from

an input device. Both statements convert data according to the format speciﬁcation, fn. Using DECODE, how

ever, the same information can be read several times with different DECODE and FORMAT statements.

Starting at the named location, v, data is transmitted according to the speciﬁed format and stored in the list

variables. If the number of characters per record is not a multiple of 10 (a display code word contains 10 dis

play code characters) the balance of the word is ignored. However, if the number of characters speciﬁed by

the list and the format speciﬁcation exceeds the number of characters per record, an execution error message

is printed. DECODE processing an illegal character for a given conversion speciﬁcation produces a fatal error.

If DECODE is processing an A or R format speciﬁcation and encounters a zero character (6 bits of binary

zero), the character is treated as a colon under 64-character set or as a blank under 63-character set.

(Note: Internal records are not processed as zero-byte terminated records.)

60497800 E 5-25

Example: ;;

c-j* multiple of 10

DECODE (16,1,GAMMA) IX,L,M,N

1 FORMAT (2A8)

beginning of new record

Record 1

W o r d 1 W o r d 2

GAMMA

Record 2

W o r d 1 W o r d 2

A = blank

Data transmitted under this DECODE speciﬁcation would appear in storage as follows

Last 4 characters of the second

word in each record are ignored.

IX=HEADER 1

L=21HEAD

M=HEADER 1

N=22HEAD

The following illustrates one method of packing the partial contents of two words into one. Information is

stored in core as:

LOC(l)SSSSSxxxxx

LOC(6)xxxxxDDDDD

To form SSSSSDDDDD in storage location NAME:

DECODE(10,1,LOC(6) ) ITEMP

1 F0RMAT(5X,A5)

ENCODE(10,2,NAME)LOC(1) , ITEMP

2 FORMAT(2A5)

The DECODE statement places the last 5 display code characters of LOC(6) into the ﬁrst 5 characters of

ITEMP. The ENCODE statement packs the ﬁrst 5 characters of LOC(l) and ITEMP into NAME.

Using the R speciﬁcation, the example above could be shortened to:

ENC0DE(10,1,NAME)L0C(1),L0C(6)

1 FORMAT(A5,R5)

5-26 60497800 E

/«S^\

ifsP^S

0S>S\

FILE MANIPULATION STATEMENTS

Three statements can be used to manipulate ﬁles: REWIND, BACKSPACE, and ENDFILE.

REWIND

/T REWIND u

The REWIND operation positions a ﬁle at beginning of information so that the next input/output operation

references the ﬁrst record in the ﬁle, even though several ENDFILE statements may have been issued to that

unit since the last REWIND. If the ﬁle is already at beginning of information, the statement acts as a do-nothing

statement. (Refer to BACKSPACE/REWIND, section 16 for further information.)

Example:

REWIND 3

BACKSPACE

BACKSPACE u

Unit u is backspaced one record. When the ﬁle is positioned at beginning of information, this statement acts

as a do-nothing statement. If BACKSPACE is the ﬁrst operation on a ﬁle positioned at beginning-of-information,

a non-fatal Record Manager error results. A backspace operation should not follow a list directed read on a ﬁle.

Example:

DO 1 LU N = 1 , 1 0 , 3

1 BACKSPACE LUN

Files TAPE1, TAPE4, TAPE7, and TAPE 10 are backspaced one record.

ENDFILE

ENDFILE u

An end of partition is written on the designated unit.

Note: When ENDFILE is used on a ﬁle deﬁned with W type records, an end-of-partition is not physically

written but is marked in the control word. For all other record types a level 17 zero-length PRU is written.

60497800 E 5-27

To ensure ﬁle integrity, ENDFILE should not be the ﬁrst operation on a ﬁle.

Meaningful results are not guaranteed if ENDFILE is used on a ﬁle processed by mass storage subroutines. '

Example:

IOUT - 6L0UTPUT .

END FILE IOUT

End of partition is written on the ﬁle OUTPUT.

End of partition is the ﬁle boundary recognized by the EOF function (section 8).

For records written by an unformatted WRITE statement, an end-of-partition boundary is detected as an

end of section (end-of-record) by the operating system.

/f^*^"X

5-28 60497800 E

j0&\ INPUT/OUTPUT LISTS AND FORMAT STATEMENTS

This chapter covers input/output lists and FORMAT statements. Input/output statements are covered in

\ section 5.

C INPUT/OUTPUT LISTS

^!Ss^ The list portion of an input/output statement speciﬁes the items to be read or written and the order of

[ transmission. The input/output list can contain any number of items. List items are read or written sequentially

from left to right.

\ If no list appears on input, a record is skipped. Only Hollerith information from the FORMAT statement can

be output with a null (empty) output list.

f A list item consists of a variable name, an array name, an array element name, or an implied DO list. On output

the data list can include arbitrarily long Hollerith constants and arithmetic expressions. No expression in an

^^ input/output list can reference a function if such reference would cause any input/output operations (including

DEBUG output) to be executed or would cause the value of any element of the input/output statement to be

changed.

\ Multiple lists can appear, separated by commas, each enclosed in parentheses.

An array name without subscripts in an input/output list speciﬁes the entire array in the order in which it is

f" stored. The entire array (not just the ﬁrst word of the array) is read or written.

Subscripts in an input/output list may be any valid subscript (section 1).

Examples:

f READ (2,100)A,B,C,D

READ (3,200)A,B,C(I).D(3,4),E(U,7),H

READ (4,101 )J.A(J),I,B(U)

f^ READ (2,202)DELTA

READ (4,102)DELTA(5*J+2,5*l-3.5*l<),C/D(l+7)

READ <3,2)A,(B.C,D),(X,Y)

An implied DO list is a list followed by a comma and an implied DO speciﬁcation, all enclosed in parentheses.

^^ An implied DO speciﬁcation takes one of the following forms:

i = m . . . m ^ r r i . , i = m 1 , m „

0&*\

The elements i, mj, n^, and m^ have the same meaning as in the DO statement. The range of an implied DO

speciﬁcation is that of the implied DO list. The values of i, mj, n^, and nv^ must not be changed within the

fm\ range of the implied DO list by a READ statement.

60497800 A 6-1

Jps.

yf^^V

On input or output, the list is scanned and each variable in the list is paired with the ﬁeld speciﬁcation

provided by the FORMAT statement. After one item has been input or output, the next format speciﬁcation

is taken together with the next element of the list, and so on until the end of the list.

Example:

READ (5,20)L,M,N

20 FORMAT (13,12,17)

Input record

IOC 223456712/lOC22

x^^x

./^fsX

^J^y

/ * ^ y

100 is read into the variable L under the speciﬁcation 13, 22 is read into M under the speciﬁcation

12, and 3456712 is read into N under speciﬁcation 17.

Reading more data than is in the input stream produces unpredictable values. The EOF function described in

section 8 can be used to test for the end of the data.

IMPLIED DO IN I/O LIST

Input/output of array elements may be accomplished by using an implied DO loop. The list of variables

followed by the DO loop index, is enclosed in parentheses to form a single element of the input/output list

Example:

READ (5,100) (A(l),l=1,3)

has the same effect as the statement

READ (5,100) A(1),A(2),A(3> _

The general form for an implied DO loop is:

( . . . ((Iist,i1=m1,m2,m3),i2=j.|, J2. J3) 'n=k1'k2'k3' '*>

m,j,k are unsigned integer constants or predeﬁned positive integer variables. If m3, j3 or k3 is omitted, /<_

a one is used for incrementing.

*l"**n are integer control variables. A control variable should not be used twice in the same implied DO

nest, but array names, array elements, and variables may appear more than once. The value of a control l

variable within an implied DO speciﬁcation is deﬁned only within that speciﬁcation; it should not be

referenced outside the speciﬁcation. ^^^

The ﬁrst control variable (ij) deﬁned in the list is incremented ﬁrst, ij is set equal to mj and the

associated list is transmitted; then ij is incremented by m3, until m2 is exceeded. When the ﬁrst

control variable reaches m2, it is reset to m^; the next control variable at the right (i2) is incremented; ^

and the process is repeated until the last control variable (in) has been incremented, until k2 is exceeded.

6-2 60497800 D '

The general form for an array is:

(^ (((A(l,J,K)#l=m1,m2,m3),J=n1,n2,n3),K=kl,k2,k3)

Example:

READ (2,100) ((A(JV,JX),JV=2.20,2),JX=1,30)

READ (2,200) (BETA(3*JON+7),JON=JONA,JONB,JONC)

READ (2.300) (((ITMLIST(I,J+1.K-2),I=1,25)0J=2,N),K=IVAR,IVMAX,4)

An implied DO loop can be used to transmit a simple variable more than one time. For example, the list

^^^ item (A(K),B,K=1,5) causes the variable B to be transmitted ﬁve times. An input list of the form

(; K,(A(I),I= 1,K) is permitted, and the input value of K is used in the implied DO loop. The index variable

in an implied DO list must be an integer variable.

Examples of simple implied DO loop list items:

READ (1,400) (A(l),l=1,10)

400 FORMAT (E20.10)

The following DO loop would have the same effect:

DO 5 1=1,10

5 READ (1,400) A(l)

■■ '- Example:

CAT,DOG, and RAT will be transmitted 10 times each with the following iolist

(CAT, DOG, RAT, 1=1.10)

Implied DO loops may be nested.

Example:

DIMENSION MATRIX(3,4,7)

READ (3.100) MATRIX

100 FORMAT (16)

Equivalent to the following:

(^ DIMENSION MATRIXC3.4.7)

READ (3,100) (((MATRIX(U,K),I=1,3),J=1,4),K=1,7)

The list is similar to the nest of DO loops:

DO 5 K=1.7

DO 5 J=1,4

DO 5 1=1,3

5 READ (3,100) MATRIX(I,J,K)

60497800 A 6.3

ys^v

Example:

The following statement transmits nine elements into the array E in the order: E(l,l), E(l,2), E(l,3), "^)

E(2,l), E(2,2), E(2,3), E(3,l), E(3,2), E(3,3)

READ (1,100) ((E(U)^J=1,3).I=1,3)

Example:

READ (2,100) (((((A(I,J,K),B(I,L).C(J,N),I=1,10),J=1,5),

X K=1.8),L=1,15),N=2,7)

Data is transmitted in the following sequence:

READ (5,100) (MATRIX(I),I=1,10)

100 FORMAT (F7.2)

In the above statements, the implied DO loop is part of the READ statement; therefore, the FORMAT

statement speciﬁes the format of the data input and determines when a new record will be read.

READ (12,13) B

is equivalent to

READ (12,13) ((BO,J),I=1,10),J=1,15)

6-4 60497800 A

A(l,l,l), B(l,l), C(l,2), A(2,l,l), B(2,l), C(l,2)...

...A(10,l,l), B(10,l), C(l,2), A(l,2,l), B(l,l), C(2,2)...

...A(10,2,l), B(10,l), C(2,2),...A(10,5,1), B(10,l), C(5,2)... 1

...A(10,5,8), B(10,l), C(5,2),...A(10,5,8), B(10,15), C(5,2)...

Data can be read into or written from part of an array by using the implied DO loop. ""^

Examples:

Data (consisting of one constant per record) is read into the ﬁrst 10 elements of the array MATRIX. "^

The following statements would have the same effect:

DO 40 i = 1,10

40 READ (5,100) MATRIX(l)

100 FORMAT (F7.2)

In this example, numbers are read from unit 5, one from each record,into the elements MATRIX(l) /*se\

through MATRIX(IO) of the array MATRIX. The READ statement is encountered each time the DO

loop is executed; and a new record is read for each element of the array. Each execution of a READ

statement reads at least one record regardless of the FORMAT statement. >«5*\

READ (5,100) (MATRIX(I),I=1,10)

100 FORMAT (F7.2)

If statement 100 FORMAT (F7.2) had been 100 FORMAT (4F20.10), only three records would be read.

To read data into an entire array, it is necessary only to name the array in a list without any subscripts. ^^.

Example:

DIMENSION B (10,15) ^

The entire array B will be transmitted in both cases. ***$,

FORMAT STATEMENT

Input and output can be formatted or unformatted. Formatted information consists of strings of display code

characters. Unformatted information consists of strings of binary word values in the form in which they

normally appear in storage. A FORMAT statement or variable format speciﬁcation is required to transmit

formatted information.

5 7

f s K

sn FORMAT (fsr .K)

Statement label which must appear

Set of one or more ﬁeld speciﬁcations separated by commas and slashes and

optionally grouped by parentheses

Note that the syntax sn FORMAT (, that is, a statement label followed by the word FORMAT followed by

a left parenthesis, is understood by the FORTRAN compiler to be a FORMAT statement, regardless of

previous conditions or uses of the word FORMAT in the user program.

/*fp\,

Example:

READ (5,100) INK.NAME.AREA

100 FORMAT (10X,I4,I2,F7.2)

FORMAT is a non-executable statement which speciﬁes the format of data to be moved between input/output

device and main memory. It is used in conjunction with formatted input and output statements, and it may

appear anywhere in the program.

The FORMAT speciﬁcation is enclosed in parentheses. Blanks are not signiﬁcant except in Hollerith ﬁeld

speciﬁcations.

Generally, each item in an input/output list is associated with a corresponding ﬁeld speciﬁcation in a FORMAT

statement. The FORMAT statement speciﬁes the external format of the data and the type of conversion to

be used. Complex variables always correspond to two ﬁeld speciﬁcations. Double variables correspond to one

ﬂoating point ﬁeld speciﬁcation (D, E, F, G) or two of any other kind. The D ﬁeld speciﬁcation corresponds

to exactly one list item or half of a complex item.

The type of conversion should correspond to the type of the variable in the input/output list. The

FORMAT statement speciﬁes the type of conversion for the input data, with no regard to the type of the

variable which receives the value when reading is complete.

For example:

INTEGER N

READ (5,100) N

100 FORMAT (F10.2)

A ﬂoating point number is assigned to the variable N which could cause unpredictable results if N is

referenced later as an integer.

60497800 E 6-5

/*P*N

-iCS^\

DATA CONVERSION

The following types of data conversions are available:

srEw.d Single precision ﬂoating point with exponent

srEw.dEe «... Single precision floatingpoint with explicitly specified exponent length *1

sr Ew.dDe Single precision ﬂoating point with explicitly speciﬁed exponent length

srFw.d Single precision ﬂoating point without exponent *j

srGw.d Single precision ﬂoating point with or without exponent

srDw.d Double precision ﬂoating point with exponent "^)

rlw Decimal integer conversion

rlw.z Decimal integer with minimum number of digits speciﬁed 'l

rLw Logical conversion

rAw Character conversion ,/*5S^

rRw Character conversion

rdw , Octal integer conversion <-SB^

rOw.z Octal integer with minimum number of digits speciﬁed

rZw Hexadecimal conversion

srVw.d Variable type conversion

E, F, G, D, I, L, A, R, O, and Z are the codes which indicate the type of conversion.

w Non-zero, unsigned integer constant specifying the ﬁeld width in number of character pos

itions in the external record. This width includes any leading blanks, + or - signs, decimal

point, and exponent. /^^1)

d Unsigned integer constant specifying the number of digits to the right of the decimal point

within the ﬁeld. On output all numbers are rounded. >a^

Non-zero, unsigned integer constant specifying the number of digits in the exponent.

Non-zero, unsigned integer constant less than 2 -1 specifying the number of times the con

version code is to be repeated.

Optional scale factor.

z Unsigned integer constant specifying the minimum number of digits to be output. ^^)

The ﬁeld width w must be speciﬁed for all conversion codes. If d is not speciﬁed for w.d, it is

a s s u m e d t o b e z e r o , w m u s t b e > d . ^

.■»* ■■* IWJfV

6-6 60497800 A

Field separators are used to separate speciﬁcations and groups of speciﬁcations. The format ﬁeld separators

are the slash (/) and the comma. The slash is also used to specify demarcation of formatted records.

CONVERSION SPECIFICATION

Leading blanks are not signiﬁcant in numeric input conversions; other blanks are treated as zeros. Plus

signs can be omitted. An all-blank ﬁeld is considered to be minus zero, except for logical input, where an

f^ all-blank ﬁeld is considered to be FALSE. When an all-blank ﬁeld is read with a Hollerith input speciﬁcation,

each blank character is translated into a display code 55 octal.

f* For the E, F, G, and D input conversions, a decimal point in the input ﬁeld overrides the decimal point

speciﬁcation of the ﬁeld descriptor.

r^ The output ﬁeld is right-justiﬁed for all output conversions. If the number of characters produced by the

conversion is less than the ﬁeld width, leading blanks are inserted in the output ﬁeld. The number of

characters produced by an output conversion must not be greater than the ﬁeld width. If the ﬁeld width is

f^ exceeded, asterisks are inserted throughout the ﬁeld.

Complex data items are converted on input/output as two independent ﬂoating point quantities. The

f^ format speciﬁcation uses two conversion elements.

/"P&s Example:

COMPLEX A.B,C,D

WRITE (6,10)A

10 FORMAT (F7.2,E8.2)

READ (5,11) B,C,D

11 FORMAT (2E10.3(F8.3,F4.1))

Data of differing types may be read by the same FORMAT statement. For example:

10 FORMAT (I5.F15.2)

speciﬁes two numbers, the ﬁrst of type integer, the second of type real.

READ (5,15) NO,NONE,INK,A,B,R

15 FORMAT (3I5.2F7.2.A4)

reads three integer values, two real values, and one character string.

Iw and Iw.z INPUT

The I conversion is used to input decimal integer constants.

I w I w . z

w is a decimal integer constant designating the total number of characters in the ﬁeld including signs and

blanks, z is ignored on input.

60497800 A 6-7

The plus sign may be omitted for positive integers. When a sign appears, it must precede the ﬁrst digit in

the ﬁeld. Blanks are interpreted as zeros. An all blank ﬁeld is considered to be minus zero. Decimal points

are not permitted. The value is stored in the speciﬁed variable. Any character other than a decimal digit,

blank, or the leading plus or minus sign in an integer ﬁeld on input will terminate execution.

Example:

READ(2,10)I,J,K,L,M,N

10 FORMAT (I3,I7,I2,I3,I2,I4)

Input Record:

f.59lbb-15b1ﬂ 8^7^11341"

3 ? 2 3 2 4

In storage:

I contains 139

J contains -1500

K contains 18

L contains 7

M contains -0

N contains 104

Iw and lw,z OUTPUT

The I speciﬁcation is used to output decimal integer values.

I w I w. z

w is a decimal integer constant designating the total number of characters in the ﬁeld including signs and

blanks. If the integer is positive the plus sign is suppressed. Numbers in the range of-(259-l) to 259-l

(2s9-1=576 460 752 303 423 487) are output correctly.

z is a decimal integer constant designating the minimum number of digits output. Leading zeros are generated

when the output value requires less than z digits, if z=0, a zero value will produce all blanks. If z=w, no blanks

will occur in the ﬁeld when the value is positive, and the ﬁeld will be too short for any negative value. Not

specifying z produces the same results as z=l.

The speciﬁcation Iw or Iw.z outputs a number in the following format:

ba...a

b Minus sign if the number is negative, or blank if the number is positive

a...a May be a maximum of 18 digits

The output quantity is right justiﬁed with blanks on the left.

r^%S\

<^SS\

.■^^"*\

If the ﬁeld is too short, all asterisks occupy the ﬁeld.

Example:

PRINT 10,U,K

10 FORMAT (19,110.15.3)

I contains -3762

J contains +4762937

K contains + 13

/*5gv

6-8 60497800 A

Result:

/IPN

/f|St*s

/|P^\

Example:

WRITE (6.100)N,M,I

100 FORMAT (15,16,19)

Result:

1st blank taken

as printer control

character

Ew.d, Ew.dEe and Ew.dDe OUTPUT

1st blank taken as

printer control character

N contains +20

M contains -731450

I contains +205

bb20[••••••bbbbbb205I

speciﬁcation too

small5* indicates ﬁeld

is too short

E speciﬁes conversion between an internal real value and an external number written with exponent.

Ew.d Ew.dEe Ew.dDe

w is an unsigned integer designating the total number of characters in the ﬁeld, w must be wide enough to

contain digits, plus or minus signs, decimal point, E, the exponent, and blanks. Generally, w > d + 6

or w > d + e + 4 for negative numbers and w>d + 5orw>d + e + 3for positive numbers. Positive

numbers need not reserve a space for the sign of the number. If the ﬁeld is not wide enough to contain the

output value, asterisks are inserted throughout the ﬁeld. If the ﬁeld is longer than the output value, the quan

tity is right justiﬁed with blanks on the left. If the value being converted is indeﬁnite, an I is printed in the

ﬁeld; if it is out of range, an R is printed.

d speciﬁes the number of digits to the right of the decimal within the ﬁeld.

e speciﬁes the number of digits in the exponent and is limited to 6 or fewer digits.

The Ew.d speciﬁcation produces output in the following formats:

b.a...aE ± ee For values where the magnitude of the exponent is less than one hundred

b.a...a ± eee For values where the magnitude of the exponent exceeds one hundred

b is a minus sign if the number is negative, and a blank if the number is positive

a...a are the most signiﬁcant digits of the value correctly rounded

60497800 E 6-9

/**sgjw

When the speciﬁcation Ew.dEe or .Ew.dDe is,used, the exponent is denoted by E or D and the number of

digits, used for the exponent ﬁeld not counting the letter and sign is determined by e. If e is speciﬁed too

small for the value being output, the entire ﬁeld width as speciﬁed by w will be ﬁlled with asterisks.

Examples:

WRITE (2,10)A A contains -67.32 or +67.32

10 FORMAT (E10.3)

Result: -.673E+02 or b.673E+02

WRITE (2,10)A

10 FORMAT (E13.3)

Result: bbb-.673E+02 or bbbb.673E+02

6-10 60497800 C

>ﬂ3?^s

X^tv

If an integer variable is output under the Ew.d speciﬁcation, results are unpredictable since the internal format >sa\

of real and integer values differ. An integer value does not have an exponent and will be printed, therefore, as

a very small value or 0.0.

Ew.d, Ew.dEe and Ew.dDe INPUT

E speciﬁes conversion between an external number written with an exponent and an internal real value.

Ew.d Ew.dEe Ew.dDe

w is an unsigned integer designating the total number of characters in the ﬁeld, including plus or minus

signs, digits, decimal point, E and exponent. If an external decimal point is not provided, d acts as a -^

negative power-of-10 scaling factor. The internal representation of the input quantity is:

(integer subfield)x 10"d x 10 <exP°nent subfieid) ^

For example, if the speciﬁcation is El0.8, the input quantity 3267E + 05 is converted and stored as:

3267X10"8X 105 = 3.267. ^

If an external decimal point is provided, it overrides d. If d does not appear it is assumed to be zero. % if

speciﬁed, has no effect oh input. "^

In the input data, leading blanks are not signiﬁcant; other blanks are interpreted as zeros. ^^

An input ﬁeld consisting entirely of blanks is interpreted as minus zero.

The following diagram illustrates the structure of the input ﬁeld: )

^~*S

input ﬁeld

digit E or D

integer .

subﬁeld

fraction

subﬁeld

exponent

The integer subﬁeld begins with a + or - sign, a digit, or a blank; and it may contain a string of digits. The

integer ﬁeld is terminated by a decimal point, E, +, - or the end of the input ﬁeld.

The fraction subﬁeld begins with a decimal point and terminates with an E, +, - or the end of the input

ﬁeld. It may contain a string of digits.

The exponent subﬁeld may begin with E, + or -. When it begins with E, the + is optional between E and

the string of digits in the subﬁeld.

For example, the following are valid equivalent forms for the exponent 3:

E + 03 E 03 E03 EO E3 •> 3 +3 D3 D*3 D+ 3

The range, in absolute value, of permissible values is IO"293 to IO322 approximately. Smaller numbers are treated

as zero; larger numbers cause a fatal error message.

Valid subﬁeld combinations:

+ 1.6327E-04

-32.7216

+ 328 + 5

.629E-1

+ 136

136

.07628431

E-06 (interpreted as zero)

Integer-fraction-exponent

integer-fraction

integer-exponent

fraction-exponent

integer only

fraction only

exponent onlv

If the ﬁeld length speciﬁed by w in Ew.d is not the same as the length of the ﬁeld containing the input

number, incorrect numbers may be read, converted, and stored. The following example illustrates a

situation where numbers are read incorrectly, converted and stored; yet there is no immediate indication

that an error has occurred:

READ (3,20) A,B,C

20 FORMAT (E9.3,E7.2,E10.3)

60497800 A 6-11

On the input record, quanities are in three adjacent ﬁelds, columns 1-24:

+6 .47E-01J-2.36 f 5. 321E+02

, * ■— * *

9 5 1 0

9 7 1 0

|+6.47E-oT|

+6.47E-0lF2". 56+51

+6.47E-01-2. 36+5|.321E+02bb|

First, +647E-01 is read, converted and placed in location A. The second speciﬁcation E7.2 exceeds

the width of the second ﬁeld by two characters. The number -2.36 + 5 is read instead of -2.36. The

speciﬁcation error (E7.2 instead of E5.2) caused the two extra characters to be read. The number

read (-2.36 + 5) is a legitimate input number. Since the second speciﬁcation incorrectly took two

digits from the third number, the speciﬁcation for the third number is now incorrect. The number

.321E + 02bb is read. Trailing blanks are treated as zeros; therefore the number .32IE+ 0200 is read

converted and placed in location C. Here again, this is a legitimate input number which is converted

and stored, even though it is not the number desired.

Examples of Ew.d input speciﬁcations:

Input Field Speciﬁcation Converted Value Remarks

+143.26E-03 E11.2 .14326 All subﬁelds present

327.625 E7.3 327.625 No exponent subﬁeld

4.376 E5 4.376 . No d in speciﬁcation

-.0003627+5 El 1.7 -36.27 Integer subﬁeld only a minus

sign and a plus sign appears

instead of E

-.0003627 E5 E11.7 -36.27 Integer subﬁeld left of decimal

contains minus sign only

blanks Ew.d -0. All subﬁelds empty

E+06 El 0.6 0. No integer or fraction subﬁeld;

zero stored regardless of expo

nent ﬁeld contents

LbEbl

1 ,

E6.3 10. Blanks are interpreted as zeros

f**s3$&v

6-12 60497800 A

<**,*\

/■Jfcv

Fw.d OUTPUT

The F speciﬁcation outputs a real number without a decimal exponent.

Fw.d

w is an unsigned integer which designates the total number of characters in the ﬁeld including the

sign (if negative) and decimal point, w must be > d + 2.

d speciﬁes the number of places to the right of the decimal point. When d is zero, only the digits to the

left of the decimal and the decimal point are printed.

The plus sign is suppressed for positive numbers. If the ﬁeld is too short, all asterisks appear in the output ﬁeld.

If the ﬁeld is longer than required, the number is right justiﬁed with blanks on the left. If the value being

converted is indeﬁnite, an I is printed in the ﬁeld; if it is out of range, an R is printed.

The speciﬁcation Fw.d outputs a number in the following format:

■ decimal point

b...afa...a

b Minus sign if the number is negative, or blank if the number is positive.

Examples:

Value of A FORMAT Statement PRINT Statement Printed Result

+32.694 10 FORMAT (1H,F6.3) PRINT 10,A 32.694

+32.694 11 FORMAT (1H.F10.3) PRINT 11,A bbbb32.694

-32.694 12 FORMAT (1H,F6.3) PRINT 12,A ******

.32694 13FORMAT(1H,F4.3,F6.3) PRINT 13,A,A .327bb.327

The speciﬁcation 1H is the carriage control character.

Fw.d INPUT

On input F speciﬁcation is treated identically to the E speciﬁcation.

60497800 A 6-13

Examples of the F format speciﬁcation:

Input Field Speciﬁcation Converted Value Remarks

367.2593 F8.4 367.2593 Integer and fraction ﬁeld

-4.7366 F7 -4,7366 No d in speciﬁcation

.62543 F6.5 .62543 No integer subﬁeld

.62543 F6.2 .62543 Decimal point overrides d of speci

ﬁcation

+144.15E-03 F11.2 .14415 Exponents are allowed in F input,

and may have P scaling

5bbbb F5.2 500.00 No fraction subﬁeld; input number

converted as 50000x10"2

bbbbb F5.2 -0.00 Blanks in input ﬁeld interpreted as

-0

««3^\

.-^^*\

Gw.d INPUT

Input under control of G speciﬁcation is the same as for the E speciﬁcation. The rules which apply to the E

speciﬁcation apply to the G speciﬁcation.

Gw.d

Example:

Unsigned integer which designates the total number of characters in the ﬁeld including

E, digits, sign, and decimal point

Number of places to the right of the decimal point

READ (5,11) A,B,C

11 FORMAT (G13.6.2G12.4)

/^^\

<^S5\

Gw.d OUTPUT

Output under control of the G speciﬁcation is dependent on the size of the ﬂoating point number being

converted. The number is output under the F conversion unless the magnitude of the data exceeds the range

which permits effective use of the F. In this case, it is output under E conversion with an exponent.

6-14 60497800 A

>S^V

Jp"*»\

Gw.d

wUnsigned integer which designates the total number of characters in the ﬁeld including

digits, signs and decimal point, the exponent E, and any leading blanks.

d Number of signiﬁcant digits output.

If a number is output under the G speciﬁcation without an exponent, four spaces are inserted to the right of

the field (these spaces are reserved for the exponent field E±00). Therefore, for output under G conversion

w must be greater than or equal to" d + 6. The six extra spaces are required for sign and decimal point plus

four spaces for the exponent ﬁeld.

Example:

WRITE (7.200) YES

200 FORMAT (G10.3)

YES contains 77.132

Output: b77.1bbbb b denotes a blank

If the decimal point is not within the ﬁrst d signiﬁcant digits of the number, the exponential form is used

(G is treated as if it were E).

Example:

WRITE (4,100) EXIT

100 FORMAT (G10.3)

EXIT contains 1214635.1

Output: .121E+07

Example:

READ (5,50) SAMPLE

WRITE (6,20) SAMPLE

20 FORMAT (1X,G17.8)

Data read by

READ statement Data Output Format Option

.1415926535bE-10 .14159265E-10 E conversion

.8979323846 .89793238 F conversion

2643383279. .26433833E+10 E conversion

-693.9937510 -693.99375 F conversion

0^r\ 60497800 A 6-15

Dw.d OUTPUT

Type D conversion is used to output double precision variables. D conversion corresponds to E conversion

except that D replaces E at the beginning of the exponent subﬁeld. If the value being converted is

indeﬁnite, an I is printed in the ﬁeld; if it is out of range, an R is printed.

Examples of type D output:

DOUBLE PRECISION A,B,C

A = 111111.11111

B = 222222.22222

C = A+ B

WRITE (2,10) A,B,C

10 FORMAT (3D23.11)

.11111111111D+06 .22222222222D+06 .33333333333 D+06

The speciﬁcation Dw.d produces output in the following format:

- decimal point

bta...a±eee -308 < eee -Z 337

b.a...aD±ee 0< ee-s 99

Minus sign if the number is negative, or blank if the number is positive

Most signiﬁcant digits

Digits in the exponent

a...a

Dw.d INPUT

D conversion corresponds to E conversion except that D replaces E at the beginning of the exponent

subﬁeld.

The following diagram illustrates the structure of the input ﬁeld:

^^K

0*3^k

^§N

r^*^\

input ﬁeld

digit D or E .

integer

subﬁeld

fraction

subﬁeld

exponent

6-16 60497800 A

/ S ^ v

Ow INPUT

Octal values are converted under the O speciﬁcation.

w is an unsigned integer designating the total number of characters in the ﬁeld. The input ﬁeld may contain

a maximum of 20 octal digits (0 through 7). Blanks are allowed and a plus or minus sign may precede the ﬁrst

octal digit. Blanks are interpreted as zeros and an all blank ﬁeld is interpreted as minus zero. A decimal point is

not allowed.

Example:

INTEGER P.Q.R

READ 10,P,Q,R

10 FORMAT (010,012,02)

Input Record:

/3737373737t666b6644b444r-0

I 1 0 1 2 2

Input storage (octal representation):

P00000000003737373737

00000000666066440444

77777777777777777777

Ow OUTPUT

The O speciﬁcation is used to output the internal representation in octal.

Ow O w.d

w is an unsigned integer designating the total number of characters in the ﬁeld. If w is less than 20, the

rightmost digits are output. For example, if the contents of location P were output with the following

statement the digits 3737 would be output.

WRITE (6,1) P

100 FORMAT (1X,04)

location P 00000000003737373737

If w is greater than 20, the 20 octal digits (20 octal digits = a 60- bit word) are right justiﬁed with blanks

on the left.

For example, if the contents of location P are output with the following statement

WRITE (6,200) P

200 FORMAT (1X,022)

60497800 A 6-17

Output would appear as follows:

bb000Q0000003737373737 b = blank

A negative number is output in one's complement internal form.

If d is speciﬁed, the number is printed with leading zero suppression and with a minus sign for negative

numbers. At least d digits will be printed. If the number cannot be output in w octal digits, all asterisks

will ﬁll the ﬁeld.

Example:

I =-11

WRITE (6,200) I

{AC53F93D

,/r^^\

Output would appear as follows:

bb77777777777777777764 ^

The speciﬁcation Ow produces a string of up to 20 octal digits. Two octal speciﬁcations must be used for variables

whose type is complex or double precision. /s^

Zw INPUT and OUTPUT ^

Hexadecimal values are converted under the Z speciﬁcation.

w is an unsigned integer designating the total number of characters in the ﬁeld. The input ﬁeld contains digits 0 ^

through. 9 and the letters A through F. A maximum of 15 hexadecimal digits is allowed; blanks and a plus or

minus sign may precede the ﬁrst hexadecimal digit.

Input values are stored right justiﬁed with zero ﬁll. On output, if w is greater than 15, leading blanks are

written. If w is less than 15, digits are taken from the right. If w is greater than the number of digits, ^

leading zeros are written.

Example:* - / s

READ 20, A,B

20 FORMAT (26.Z2)

Input record:

1 6 8

/ ^ s \

6-18 60497800 E

JP\

Location A:

O0OOOOO0OAC53F9

<§&&*.

WRITE 30,A,B

30 FORMAT (1X,Z3.1X,Z3)

0&\

Output:

0^\ 3F9 03D

Aw INPUT

/SEN

0iS\

jfftSav

Location B:

00000000000003D

The A speciﬁcation is used to input character data

w is an unsigned integer designating the total number of characters in the ﬁeld.

Character information is stored as 6-bit display code characters, 10 characters per 60-bit word. For example, the

digit 4 when read under A speciﬁcation is stored as a display code 37. If w is less than 10, the input quantity is

stored left justiﬁed in the word; the remainder of the word is ﬁlled with blanks.

Example:

READ (5,100) J

100 FORMAT (A7)

60497800 E 6-18.1/6-18.2

/&^.

/$^ \

Input record:

jp\

/ﬁ^K

/example

When EXAMPLE is read it is stored left justiﬁed in the 10 character word

1234567890

MfattlEMII

If w is greater than 10, the rightmost 10 characters are stored and remaining characters are ignored.

Example:

READ (5,200)K

200 FORMAT (A13)

Input record:

1 1 3

f'SPECIFICATION

In storage:

12345678910

|(Ml|C|VW0iN|

READ (5,10) L,M,N

10 FORMAT (A10.A10.A5)

Input record:

THIS IS AN EXAMPLE

| 1 0 1 0

In storage:

12345678910

Me|x|a[mf^e

N| tKJNfOJW

Aw OUTPUT

The A speciﬁcation is used to output alphanumeric characters.

60497800 E 6-19

/e£&$S.

w is an unsigned integer designating the total number of characters in the ﬁeld. If w is less than 10,

the leftmost characters in the word are printed. For example, if the contents of location M in the Aw

input example are output with the following statements:

WRITE (6,300)M

300 FORMAT (1X,A4)

In storage:

r"^S5\

Characters EXAM are output

If w is greater than 10, the characters are output right-justiﬁed in the ﬁeld, with blanks on the left. For

example, if M in the previous example is output with the following statements:

WRITE (6,400)M

400 FORMAT (1X,A12)

<<<R"\

Output is as follows:

bbEXAMPLEbbb b = blank

Rw INPUT

w is an unsigned integer designating the total number of characters in the ﬁeld. The R speciﬁcation is the

same as the A speciﬁcation unless w is less than 10. If w is less than 10, the input characters are stored

right-justiﬁed, with binary zero ﬁll on the left.

Example: . ,

READ (5,600) MOO.JAY "

600,FORMAT(R10,R5)

Input record: , .,

/!^V

'RESULTS m

\'ib

In storage:'.

rJTEST|

MQO, |E}E|SlUrjIlSjb|0|F|

JAY [ft b = blank

/K ? \

6-20 60497800 E

Rw OUTPUT

w is an unsigned integer designating the total number of characters in the ﬁeld.

This speciﬁcation is the same as the A speciﬁcation unless w is less than 10. If w is less than 10, the right

most characters are output. For example, if JAY from the previous example is output with the following

statements:

WRITE (6,700) JAY

700 format (1X,R3) Characters EST are output.

Lw INPUT

The L speciﬁcation is used to input logical variables.

w is an unsigned integer designating the total number of characters in the ﬁeld.

If the ﬁrst non-blank character in the ﬁeld is T, the logical value TRUE, is stored in the corresponding list

item, which should be of type logical. If the ﬁrst non-blank character is F, the value .FALSE, is stored. If the

ﬁrst non-blank character is not T or F. a diagnostic is printed. An all blank ﬁeld has the value .FALSE.

Lw OUTPUT

w is an unsigned integer designating the total number of characters in the ﬁeld.

Variables output under the L speciﬁcation should be of type logical. A value of .TRUE, or .FALSE, in

storage is output as a right justiﬁed T or F with blanks on the left.

^ Example:

/m\ LOGICAL l,J,K

\ I = . T R U E .

J = .FALSE.

K = .TRUE.

r WRITE (4,5) l,J,K

5 FORMAT (3L3)

0^* Output:

bTbbFbbT

SCALE FACTORS

The scale factor P is used to change the position of a decimal point of a real number when it is input or

output. Scale factors may precede D. E. F and G format speciﬁcations.

60497800 E 6-21

nPDw.d nPDw.dEe nPDw.dDe

nPEw.d nPEw.dEe nPEw.dDe

nPFw.d

nPGw.d.

rip

Fw.d SCALING

INPUT

The number in the input ﬁeld is divided by 10n and stored. For example, if the input quantity 314.1592 is

read under the speciﬁcation 2PF8.4, the internal number is 314.1592 X IO"2 = 3.141592. However, if an

exponent is read the scale factor is ignored.

rffrsifev

n is the scale factor which can be any integer constant, w is an unsigned integer constant designating the

total width of the ﬁeld, d determines the number of digits to the right of the decimal point. ''*%.

A scale factor of zero is established when each FORMAT statement is ﬁrst referenced; it holds for all F,E,G,

and D field descriptors until another scale factor is encountered. ■'m\

Once a scale factor is speciﬁed, it holds for all D, E, F, and G speciﬁcations in that FORMAT statement

until another scale factor is encountered. To nullify this effect for subsequent D, E, F, and G speciﬁcations, ***\

a zero scale factor (OP) must precede a speciﬁcation.

Example: ^

15 FORMAT(2PE14.3,F10.2.G16.2,0P4F13.2)

The 2P scale factor applies to the El4.3 format speciﬁcation and also to the F10.2 and G16.2 format

specification. The OP scale factor restores normal scaling (10° = 1) for the subsequent specification I

4F13.2. |

"*l

6-22 60497800 A

OUTPUT

The number in the output ﬁeld is the internal number multiplied by 10n. In the output representation, the

decimal point is ﬁxed; the number is adjusted to the left or right, depending on whether the scale factor is plus or

minus. For example, the internal number-3.1415926536 may be represented on output under scaled F speciﬁcations

as follows:

(

-1PF13.

F13.

1PF13.

3PF13,

6 > -

•«

-3«

-31.

3141.

>314159

.141593

•415927

,592654

Ew.d and Dw.d SCALING

INPUT

Ew.d scaling on input is the same as Fw.d scaling on input.

OUTPUT

The effect of the scale factor nP is to shift the output coefﬁcient left n places and reduce the exponent by n.

In addition, the scale factor controls the decimal normalization between the coefﬁcient and the exponent such

that: if n < 0, there will be exactly -n leading zeros and d + n signiﬁcant digits after the decimal point; if

n > 0, there will be exactly n signiﬁcant digits to the left of the decimal point and d - n + 1 signiﬁcant digits

to the right of the decimal point. For example, the number -3.1415926536 is represented on output under the

indicated Ew.d scaling as follows:

(-3PE21U 4) -.0003E+04

(-1PE20. 4) -.0314E+02

( E 2 0 . < ♦) -.3142E+C1

( 1PE20. 4) -3.1416E+G0

( 3PE20, 4) -314.16E-C2

Gw.d SCALING

INPUT

Gw.d scaling on input is the same as Fw.d scaling on input.

OUTPUT

(^ The effect of the scale factor is nulliﬁed unless the magnitude of the number to be output is outside the range

that permits effective use of F conversion (namely, unless the number N < 10"' or N > 10 ). In these cases,

the scale factor has the same effect as described above for Ew.d and Dw.d scaling. For example, the numbers

-3.1415926536 and -.00031415926536 are represented on output under the indicated Gw.d scaling as follows:

60497800 C 6-23

^C^N

(-3PG20. 6)

(-1PG20. 6)

( 6 2 0 . 6 1

( 1PG20. 6)

( 3PG20, 6)

( 5PG20, 6)

( 7PG20. 6)

3.14159

3.1*tl59

3.14159

3.1<»i59

-3.1*1159

3.1^159

3.1<ii59

<-3PG20. 6)

(-1PG20. 6)

( G 2 0 . 6 )

C 1PG20. 6)

( 3PG20. 6)

( 5PG20. 6)

-.0G0314E+Q0

-.031416E-02

-.314159E-03

-3.141593E-04

-314.1593E-06

-31415.93F-08

X SPECIFICATION

The X speciﬁcation is used to skip characters in an input line or output line. On output, any character

positions not previously ﬁlled during this record generation will be set to blank. It is not associated with a

variable in the input/output list.

nX Number of characters, n, to be skipped. An optional plus sign may precede n.

OX is ignored, X is interpreted as IX. The comma following X in the speciﬁcation list is optional.

-nX Back up n characters, will not back up beyond the ﬁrst column.

Example:

WRITE (6,100) A,B,C

100 FORMAT (F9.4,4X,F7.5,4X,I3)

Output record:

-342.743bbbb1.53190bbbbb22

on input n columns are skipped.

Example:

READ (3,11) R,S,T

11 FORMAT (F5.2, 3X, F5.2, 6X, F5.2)

11 FORMAT (F5.2, 3XF5.2, 6XF5.2)

Input record:

A = -342.743

B = 1.53190

C = 22

b is a blank

/ ^ £ \

/l4.62bb$13. 78bC0STbl5.97

6-24 60497800 D

.rfS^V

/f^S

In storage:

R 14.62

S 13.78

T 15.97

Example:

INTEGER A

WRITE (1,10) A,B,C

10 FORMAT (I2.6X,F6.2,6X,E12.5)

Result:

A contains 7

B contains 13.6

C contains 1462.37

7bbbbbbb13.60bbbbbbb.146237E+04

0S\

nH OUTPUT

The H speciﬁcation is used to output strings of alphanumeric characters and, like X, H is not associated with

a variable in the input/output list.

n Number of characters in the string including blanks.

H Denotes a Hollerith ﬁeld. The comma following the H speciﬁcation is optional.

For example, the statement:

WRITE (6,1)

1 FORMAT (15HbENDbOFbPROGRAM)

can be used to output the following on the output listing.

END OF PROGRAM

Examples:

Source program:

WRITE (3,20)

20 FORMAT (28HbBLANKSbCOUNTblNbANbHbFIELD.)

produces output record:

BLANKSbCOUNTbl NbANbHbFI ELD.

Source program:

WRITE (2,30)A

30 FORMAT (6HbLMAX=.F5.2)

A contains 1.5

/8sSS>

60497800 A 6-25

produces output record:

LMAX=b1.50

nH INPUT

bTHIS IS A VARIABLE HEADING

After a READ statement, the FORMAT statement labeled 10 contains the alphanumeric information read from the

input record; a subsequent reference to statement 10 in an output statement acts as follows:

WRITE (6,10)

produces the output line:

THIS IS A VARIABLE HEADING

Character strings delimited by a pair of * or =£ symbols can be used as alternate forms of the H specification

for output. The paired symbols delineate the Hollerith ﬁeld. This speciﬁcation need not be separated from

other speciﬁcations by commas. If the Hollerith ﬁeld is empty, or invalidly delimited a fatal execution error

occurs, and an error message is printed.

An asterisk cannot be output using the speciﬁcation * *. For example,

PRINT 1

1 FORMAT (»ABC*DE«)

The second * in the FORMAT statement causes the speciﬁcation to be interpreted as * ABC* and

DE*, which is not valid. , /

6-26 60497800 A

/rf>~3$\

The H speciﬁcation can be used to read Hollerith characters into an existing H ﬁeld within the FORMAT

statement. '^

Example:

Source program:

READ (2,10) '^

10 FORMAT (27Hbbbbbbbbbbbbbbbbbbbbbbbbbbb)

Input record: "^

/*^\

/ﬁS^s

The H speciﬁcation or-?*..; -^ could be used to output this correctly:

/0Ofc\t

PRINT 1

1 FORMAT (7H ABC*DE)

Output appears as follows: ABC*DE

PRINT 2

2 FORMAT (=#= ABC*DE=#

Output appears as follows: ABC*DE

^ can be represented within #... # by two consecutive -j* symbols.

Example:

PRINT 3

f^ 3 FORMAT (^ DON =£#T=£)

Output examples:

0^\

PRINT 10

10 FORMAT { * SUBTOTALS*)

produces the following output:

SUBTOTALS

WRITE (6,20)

20 FORMAT ( -V^bRESULT OF CALCULATIONS IS AS FOLLOWS =£)

produces the following output:

RESULT OF CALCULATIONS IS AS FOLLOWS

PRINT 1, =5*SQRT£ SQRT(4.)

1 FORMAT (A10.E10.2)

produces the following output:

S Q R T 2 . 0

Note: =£ is output as " on some printers.

The *...* or #...=£ specification can be used to read alphanumeric data; however, the effect differs depending

on whether *...* or ^...-v^ occurs in an actual FORMAT statement or in a format speciﬁcation contained in a

variable or array. When the READ statement contains a constant specifying a FORMAT statement, alphanu

meric characters are read into the *...* or =£...=£ specification. When a name occurs in the READ statement

to specify the format information (variable format), characters in the input stream are skipped and no change

is made in the *...* or -v*...-^ speciﬁcation.

60497800 A 6-27

Jh IfOkMAT statements,^Ihe";*...* or $•& specification is changed to"rill;., at compile jtoe;.'T^s;c^nversion

does^ not;ocbur with-,variable format s^pecificatioris. '"*',, ' ■'V , y

END OF RECORD SLASH

The slash indicates the end of a record anywhere in the FORMAT speciﬁcation. When a slash is used to separate

ﬁeld speciﬁcation elements, a comma is allowed but not required. Consecutive slashes can be used and need not

be separated from other elements by commas. When a slash is the last format speciﬁcation to be processed, it "^l

causes a blank record to be written on output or an input record to be skipped. Normally, the slash indicates the

end of a record during output and speciﬁes that further data comes from the next record during input.

Example:

r**m$\

WRITE (2,10) 1

10 FORMAT (6X, 7HHEADING / / / 1X, 5HINPUT, 7H OUTPUT)

Output:

HEADING line I

(bl a nk) l ine 2

(bl a nk) l ine 3

input OUTPUT line 4

Each line corresponds to a formatted record. The second and third records are blank and produce the line

spacing illustrated.

Example:

1=5

J=6

K=7

WRITE (2,1)I,J,K

FORMAT (315/)

WRITE (2,2)

FORMATC A BLANK LINE SHOULD PRECEDE THIS LINE*)

^*$"\

Output:

A BLANK LINE SHOULD PRECEDE THIS LINE

The variable list (I, J, K) is exhausted and processing continues until a variable conversion is encountered. /sm^

Since the slash has been processed, it causes a blank line to be printed.

/t*Sl$s

6-28 60497800 F

Example:

/S^\

/|K\

DIMENSION B(3)

READ (5,100)IA,B

100 FORMAT (I5/3E7.2)

These statements read two records; the ﬁrst contains an integer number, and the second contains three real

numbers.

WRITE (3,11) A,B,C,D

11 FORMAT (2E10.2/2F7.3)

In storage:

A -11.6

B .325

C 46.327

D -14.261

Output:

b-.12E+02bbb.33E+00

46.327-14.261

WRITE (1,11) A,B,C,D

11 FORMAT (2E10.2 / / 2F7.3)

Output:

b-.12E+02bbb.33E+00 line I

(bl a nk) lin e 2

46.327-14.261 line 3

The second slash causes the blank line.

REPEATED FOR/WAT SPECIFICATION

Format speciﬁcations can be repeated by preﬁxing the control characters D, E, F, G, I, A, L, R, Z, and O

/pn with a non-zero, unsigned integer constant specifying the number of repetitions required.

100 FORMAT (3I4,2E7.3) is equivalent to: 100 FORMAT (I4,I4,I4,E7.3,E7.3)

/#sn 50 FORMAT (4G12.6) is equivalent to: 50 FORMAT (G12.6,G12.6,G12.6,G12.6)

A group of speciﬁcations can be repeated by enclosing the group in parentheses and preﬁxing it with the repeti-

zpg\ tion factor. If no integer precedes the left parenthesis, the repetition factor is assumed to be one.

1 FORMAT (I3.2(E15.3,F6.1.214))

60497800 B 6-29

f*^\

is equivalent to the following speciﬁcation if the number of items in the input/output list does not

exceed the format conversion codes: ^^

1 FORMAT (I3,E15.3,F6.1,I4,I4,E15.3,F6.1,I4,I4)

A maximum of nine levels of parentheses is allowed in addition to the parentheses required by the FORMAT

statement.

If the number of items in the input/output list is fewer than the number of format codes in the FORMAT

statement, excess format codes are ignored.;

If the number of items in the input/output list exceeds the number of format conversion codes when the ﬁnal

right parenthesis in the FORMAT statement is reached, the line formed internally is output. The format control

then scans to the left looking for a right parenthesis within the FORMAT statement. If none is found, the scan ,

stops when it reaches the beginning of the format speciﬁcation. If a right parenthesis is found, however, the

scan continues to the left until it reaches the ﬁeld separator which precedes the left parenthesis pairing the

right parenthesis. Output resumes with the format control moving right until either the output list is exhausted

or the ﬁnal right parenthesis of the FORMAT statement is encountered.

A repetition factor can be used to indicate multiple slashes, n(/), where n is an unsigned integer constant /^S

indicating the number of slashes required and n-1 is the number of lines skipped on output.

Example: ^

WRITE (3,15)(A(I),I=1,9)

15 FORMAT (8HbRESULTS4(/),(3F8.2) )

Format statement 15 is equivalent to: 15 format (SHbRESULTS / / / / (3F8.2) )

Output:

RESULTS _____ line 1

(blank) line 2

3.62 -4.03 -9.78

-6.33 7.12 3.49

6.21 -6.74 -1.18

(blank) line 3

(blank) line 4 ^^s

line 5

line 6

l i n e 7 ^ ^ \

Example:

READ (5,300)I,J,E,K,F,L,M,G,N,R

300 FORMAT (I3,2(I4,F7.3),I7)

is equivalent to storing data in I with format 13, J with 14, E with F7.3, K with 14, F with F7.3, and L

with 17. A new record is then read; data is stored in M with the format 14, G with F7.3, N with 14,

and R with F7.3.

READ (5,100) NEXT, DAY, KAT, WAY, NAT, RAY, MAT

100 FORMAT (I7,(F12.7.I3) )

6"30 60497800 A

NEXT is input with format 17, DAY is input with F12.7, KAT is input with 13. The FORMAT state

ment is exhausted (the right parenthesis has been reached), a new record is read, and the statement is

rescanned from the group (F12.7,I3). WAY is input with the format F12.7, NAT with 13, and from a third

record, RAY with Fl2.7, and MAT with 13.

PRINTER CONTROL CHARACTER

0s\

The ﬁrst character of a printer output record is used for carriage control and is not printed. It appears in other

forms of output as data. Carriage control also applies to records listed at a terminal under INTERCOM; the

meaning of carriage control characters depends on the type of terminal (see the INTERCOM reference manual).

Carriage control does not apply to records listed at a terminal under the NOS 1 Time-Sharing System; the

ﬁrst character is listed as data.

The printer control characters are as follows':

J0^\

Character Action

Blank Space vertically one line then print

0Space vertically two lines then print

1Eject to the ﬁrst line of the next

page before printing

+No advance before printing; allows

overprinting

Any other

character

Refer to the operating system

reference manual

For output directed to the card punch or any device other than the line printer or terminal, control characters

are not required. If carriage control characters are transmitted to the card punch, they are punched in column one.

Carriage control characters are required at the beginning of every record to be printed, including new

records introduced by means of a slash. Carriage control characters can be generated by any means.

Examples:

FORMAT (1H0,F7.3,I2,G12.6)

/|SS!\

FORMAT (1H1,I5*RESULT = \F8.4)

FORMAT (*1\!4,2(F7.3) )

FORMAT (1X,I4,G16.8)

'This chart applies only to NOS/BE 1 and SCOPE 2. For corresponding information under NOS 1, refer to

the reference manual for the subsystem under which the program is executed.

60497800 F 6-31

Example:

PROGRAM CHARCON (OUTPUT)

PRINT 10

10 FORMAT (1HU 5Xt »HERE WE ARE AT THE TOP OF A NEW PAGE*) ^

PRINT 20

20 FORMAT <3</>)

D O 3 0 I * 2 t 8 ^

IF (I ♦EQ* 4 »OR» I .EQ. 6) 40*50

50 PRINT 60

6 0 F O R M A T ( 2 1 X * * X X # / 1 H * » 2 0 X t * 5 5 # > * * v

60 TO 30

40 PRINT 70

70 FORMAT (20X» # XXXXXXXXXX t / lH->» 19X» * ====55==== #> ^

30 CONTINUE

PRINT 80

9 0 F O R M A T ( l H O t 5 X » * B E G 1 N T I C T A C T O E * ) ^

STOP

END

Output

HE*E WE ARE AT THE TOP OF A NEW PAGE

BEGIN TIC TAC TOE

Tn SPECIFICATION

This speciﬁcation is tabulation control.

n Unsigned integer. If n = zero, column 1 is assumed.

8 8

S18S8S88SS

.■ 8■ 8 . / ' s % .

SSSS888S88 ■'

8 8 ySSs

//j8»8S

When Tn is used, control skips columns right or left until column n is reached; then the next format speciﬁ

cation is processed. Using card input, if n > 80,the column pointer is moved to column n, but a succeeding

specification would read only blanks. ' j*®*.

6-32 60497800 A

,«ffl-s

READ 40, A, B, C

40 FORMAT (Tl, F5.2, Til, F6.1, T21, F5.2)

Input:

84.73bbbbb2436.2bbbb89.14.

A is set to 84.73, B to 2436.2, and C to 89.14.

0^s,

0^*>.

WRITE (31, 10)

10 FORMAT (T20,*LABELS*)

The ﬁrst 19 characters of the output record are skipped and the next six characters, LABELS, are written

on output unit number 31 beginning in character position 20.

With T speciﬁcation, the order of a list need not be the same as the input or output record, and the same

information can be read more than once.

When a T speciﬁcation causes control to pass over character positions on output, positions not previously

ﬁlled during this record generation are set to blanks; those already ﬁlled are left unchanged.

Example:

5 7

PROGRAM TEST (0UTFUT)

1FORMAT <12(10H0123456789)1

PRINT 1

PRINT 60

60 FORMAT <T8Q,*COMMENTS*,T60,*HEA0ING4*,T40t

( *HEADING3*,T20,*HEADING2*,T2>*HEAOING1*>

FRIMT 10

10 FORMAT <20X»THIS IS T^E ENO OF THIS RUN*T52*.«

FRIKT 1

STOP

END

•HONEST*)

lf3f-»S67Mtimm69O1234S6ramm:WI9liC34S6?690m^

HEAQXN61 HEADINGS HEA0INSI HEAOXMC* COMENTS

1HXS XS THE CN0 OP THIS RUN ...HONEST

123%5678901234567«90123<.S67e90123*m8901234S67a90123W

Since the ﬁrst character in a line output to the printer is used for printer control, T2 is output in the ﬁrst

print position.

The following example shows that it is possible to destroy a previously formed ﬁeld. The speciﬁcation T5

destroys part of the Hollerith speciﬁcation 10H DISASTERS.

1 FORMAT (10H DISASTERS, T5, 3H123)

PRINT 1

60497800 A 6-33

produces the following output: DIS123ERS

V SPECIFICATION*

When V is encountered in a FORMAT statement, the rightmost 6 bits from the current variable in the input/output '

list are interpreted as display code for a character to be used in place of the V as the conversion speciﬁcation for the

next variable in the input/output list. V can be used as a dummy speciﬁcation for the following conversions: A, D, ^_

E, F, G, I, L, 0, P, R, T, X and Z. It cannot be used as the E or D explicitly specifying exponent length; Ew.dVe is )

illegal.

Example:

PROGRAM V (OUTPUT)

INTEGER AFORMAT* RFORMAT

AFORMAT * IRA

RFORMAT * 1RR

NUM « 10H01234S6789 ^\

PRINT 109 AFORMAT* NUM

10 FORMAT (T8» "FORMAT SPECIFICATION TAKEN FROM VARIABLE AFORMAT8 NUI*

• O U T P U T S A S " • • V 5 / ) ■ _

PRINT 20» RFORMAT* NUM ^

20 FORMAT (T8* "FORMAT SPECIFICATION TAKEN FROM VARIA6LE RFORMAT* NUM

- OUTPUTS AS ♦* V5)

STOP <*>^\

E N D J

Output: i

FORMAT SPECIFICATION TAKEN FROM VARIABLE AFORMAT! NUM OUTPUTS AS 01234

FO R M AT S P E C I FI C AT IO N TAK E N F R O M VA R I A B LE RF O R M AT I N U M OU T P U T S A S 5 6 7 8 9 ^

EQUALS SIGN*

When = is encountered in a FORMAT statement, the current variable in the input/output list supplies a positive

integer value to be used in place of the = in the conversion speciﬁcation for the next variable in the input/output

list. The = can be used in place of a number anywhere within a FORMAT statement. Such use of = precludes com

pilation syntax checking of the FORMAT statement. V and = can be combined in one conversion speciﬁcation.

'.■■ " ' ' ^ \

Example: ■

PROGRAM EQUALS (OUTPUT)

0*^\

'If a V or = appears in a format speciﬁcation ﬁeld containing a repetition factor which does not modify a parenthesized /"aB^

speciﬁcation list, the V or = assumes the value supplied by the ﬁrst repetition, and retains that value until the current

repetition count is completed.

6-34 60497800 H

PROGRAM EQUALS (OUTPUT)

INTEGER WHO)

DATA »iVl*2*3*4*5»6*7*8*9*«0/

NUM = 1 OHO123456789

00 10 I « 1*10

10 PRINT 20* WU>9 NUM

20 FORMAT (T30» A>)

STOP

END

/$ffiv»

Output:

012

0123

01234

012345

0123456

01234567

012345678

0123456789

A variable must exist in the input/output list each time an = or V is processed in the format statement.

Example:

DIMENSION A(5).B(5)

-.13 ».3-- .'. "

PRINT 1,f3,A,l3,B

1 FORMAT{1X,5F10.=)

^^ Two lines of ﬁve values each are printed; however, 13 must be repeated in the input/output list or the ﬁrst value

of B is used to replace the =.

Example:

PROGRAM VEQUALS (OUTPUT)

INTEGER FORMATS)* W(10)

DATA FORMAT/IRA* 1RR/* W/lt2t394f5*6t79899»]0/

/SSN

NUM a 10H0123456789

DO 10 I = 1*2

DO 10 J = 1*10

K = J

IF (I *EQ. 2) K = 11-J

10 PRINT 20* FORMAT(I)9 W(K>*

20 FORMAT (T20* V*>

STOP

END

NUM

0^s

60497800 A 6-35

Output: 0

01 t*^mP\

0 1 2 " >

0123

012345 1

0123456

01234567 ^

012345678 1

0123456789

0123456789 ^

123456789 1

23456789

3456789

456789

56789

6789

789

EXECUTION TIME FORMAT SPECIFICATION

Variable format speciﬁcations can be read in as part of the data at execution time and used by READ, WRITE,

PRINT, PUNCH, ENCODE, or DECODE statements later in the program. The format is read in as alphanumeric

text under the A speciﬁcation and stored in an array, simple variable or array element, or it may be included in

a DATA statement. The format must consist of a list of format speciﬁcations enclosed in parentheses, but

without the word FORMAT or the statement label.

For example, an input record could consist of the characters:

f<E7.2.G20.5,F7.4,I3)

The name of the array containing the speciﬁcations is used in place of the FORMAT statement number in the

associated input/output statement. The array name speciﬁes the location of the ﬁrst word of the format

information.

For example, assume the following format speciﬁcations:

(E12.2,F8.2.I7,2E20.3,F9.3,I4)

This information on an input record can be read by the statements of the program such as: ^^s.

DIMENSION IVAR(3)

READ (2,1) IVAR ^^

1 FORMAT (3A10) '

6-36 60497800 A

/I-*\

/fﬂfﬁ^^.

/sp^\

j^\

The elements of the input record are placed in storage as follows:

IVAR(1) (E12.2.F8

IVAR(2) 2.I7.2E20.

IVAR<3) 3.F9.3.I4)

A subsequent output statement in the same program can refer to these format speciﬁcations as:

WRITE (2.IVAR) A,B,I,C,D,E,J

Which produces exactly the same result as the statements:

WRITE (2,10) A,B,I,C,D,E,J

10 FORMAT (E12.2,F8.2,I7,2E20.3,F9.3,I4)

60497800 A 6-37

<*%

PROGRAM UNITS, PROCEDURES, AND OVERLAYS

j**8fcv A program unit consists of FORTRAN statements, with optional comments, terminated by an END statement.

A main program is a program unit that does not begin with a SUBROUTINE, FUNCTION, or BLOCK DATA

statement. Normally, a main program begins with a PROGRAM statement, but this statement can be omitted.

A subprogram is a program unit that begins with a SUBROUTINE, FUNCTION, or BLOCK DATA statement.

An executable program contains one main program with or without subprograms. A program unit containing

no FORTRAN statements other than an END statement is considered a null program; it is diagnosed and

ignored.

/P^v

A subprogram is deﬁned separately and can be compiled independently of a main program. If the subprogram

begins with a SUBROUTINE or FUNCTION statement, it is a procedure subprogram and can accept and use

zero, one, or more values through a list of arguments, through common, or both. If the subprogram begins

with a BLOCK DATA statement, it is a speciﬁcation subprogram.

A procedure is a procedure subprogram, statement function, intrinsic function, or basic external function.

Intrinsic functions and basic external functions are FORTRAN supplied procedures and are available to any

programmer (see section 8). Statement functions and procedure subprograms are supplied by the programmer.

The differences between function and subroutine speciﬁcation and use are summarized in table 7-1.

TABLE 7-1. DIFFERENCES BETWEEN A FUNCTION AND SUBROUTINE SUBPROGRAM

/0^\

Function Subroutine

How Used The name appearing in an expression is A CALL statement is used as

used as the reference. the reference.

Arguments One or more arguments must be included. Arguments need not be present.

How Typed Name is typed implicitly by ﬁrst letter No type is associated with the

or explicitly by the type designation name.

appearing before the word FUNCTION.

Functions return a single value through the function name. Function subprograms deﬁned by the programmer

also can return values through a list of arguments, through common, or both.

Table 7-2 summarizes the terminology of the overlapping categories of procedures and subprograms.

60497800 A 7-1

TABLE 7-2. PROCEDURE AND SUBPROGRAM INTERRELATIONSHIPS

/*^5\

Statement Intrinsic Basic

External

Function

Function Subroutine Block Data

Function Function Subprogram Subprogram Subprogram

Procedure yes yes yes yes yes no

External procedure no no yes yes yes N/A

Subprogram no no no yes yes yes

Function yes yes yes yes no no

External function no no yes yes N/A N/A

Who deﬁnes user compiler compiler user user user

Where deﬁned within

program unit

compiler library external to

calling pro

gram unit

external to

calling pro

gram unit

external to

calling pro

gram unit

N/A = not applicable

/£*~*$\

S0*S*$\

Programmer written procedures (statement functions, function subprograms, and subroutine subprograms) are

discussed below as a group. FORTRAN supplied procedures (intrinsic functions and basic external functions)

are discussed in detail in section 8. The only subprogram that is not a procedure is the block data subpro

gram. Since it is not executable, it is discussed separately.

MAIN PROGRAMS

A main program can contain any FORTRAN statements except FUNCTION, SUBROUTINE, ENTRY or BLOCK

DATA; it should have a PROGRAM statement and an END statement: it must have at least one executable state

ment. One main program is required in any executable FORTRAN program. No program can have more than one

main program, except an overlay structured program, which has one main program in each overlay.

PROGRAM STATEMENT FORMAT

PROGRAM name (fpar1ffpar2,.... ,fpar_)

/ffSigjf.

name Symbolic name (1 to. 7 letters and digits, beginning with a letter). Cannot be used elsewhere

in the program as a user-deﬁned name.

fparj The fpar can be any of the following forms:

7-2 60497800 D

/SB\

j#3N

/^ﬂ^\

ﬁle File name (1-6 letters or digits beginning with a letter) for each ﬁle required by the main

program or its subprograms; the maximum number of ﬁle names is 50.

ﬁle=n n' is an integer or octal constant specifying the buffer length; default length is 2003 octal

words.

ﬁle=/r r is the maximum length in characters for list directed, formatted, and NAMELIST records;

default limit is 150 characters.

ﬁle=n/r n/r deﬁnes both buffer and record lengths.

ﬁle^ﬁle^ Filea is made equivalent to previously deﬁned ﬁleu.

In a program structured for overlays, the fparj parameter list is used only in the PROGRAM statement for the

main overlay. It is not used in primary and secondary overlay PROGRAM statements.

PROGRAM STATEMENT USAGE

The PROGRAM statement deﬁnes the program name that is used as the entry point name and the object deck

name for the loader. Optionally, the PROGRAM statement can declare ﬁles that are used in the program and

0^ in any subprograms that are called. If this statement is omitted from the main program, the program is as

sumed to have the name START, and two ﬁles named INPUT and OUTPUT.

rAll ﬁle names used in standard FORTRAN input/output statements (including mass storage subroutines) must

be listed in the PROGRAM statement. File names referenced by CYBER Record Manager interface subroutines

t must not be listed in the PROGRAM statement. If a ﬁle name is referenced in a standard FORTRAN input/

[ output statement in a main program, but is not speciﬁed in the PROGRAM statement, a warning diagnostic is

issued at compile time. If a ﬁle name is referenced in a standard FORTRAN input/output statement in a sub

program, but is not speciﬁed in the PROGRAM statement of the main program, a diagnostic is issued when

f^ the ﬁle is used at execution time.

File names on the PROGRAM statement must satisfy the following conditions:

0&\

' The ﬁle name INPUT must be declared if a READ statement without a unit designator is included

in the program.

f The ﬁle name OUTPUT must be declared if a PRINT statement without a unit designator is

included in the program.

0^ The ﬁle name PUNCH must be declared if a PUNCH statement without a unit designator is in

cluded in the program.

The ﬁle name TAPEu (u is an integer constant 0-99) must be declared if any input/output state

ment involving unit u appears in the program, and u is an integer value. At execution time, if u

is a variable, there must be a ﬁle name TAPEu for each integer value u assumes.

FORTRAN input/output routines add the characters TAPE as a preﬁx to the unit number to form the ﬁle

name. TAPE3 is the ﬁle name assigned to unit;number 3 and TAPES is the ﬁle name assigned to unit

number 5. TAPE5 and TAPE05 do not specify the same ﬁle name.

rn is ignored if speciﬁed in a program run under SCOPE 2.

60497800 B 7-3

ySSsiy

TAPEu refers to a ﬁle located on rotating mass storage unless speciﬁed otherwise in the job deck before the

program is executed. The ﬁle is temporary unless made permanent by the user.

FORTRAN input/output statements use the buffer areas established by the ﬁle name speciﬁed in the PROGRAM

statement. The buffer length can appear only with the ﬁrst reference to the ﬁle in the PROGRAM statement.

A buffer length of zero should be speciﬁed for a ﬁle referenced by a buffer statement, unless the ﬁle is a con

nected ﬁle or the ﬁle description has been changed by a FILE control statement. Since buffered records are

transmitted directly into and out of central memory, ﬁeld length of the program is reduced for each ﬁle

declared with zero buffer length in the PROGRAM statement.

For ﬁles not referenced by BUFFER statements, the following values of n^ are minimal if default record and

block types are usedT^

n=number of words in the largest record plus one.

Large records and sequential reading/writing execute faster with a larger

Minimum Value of n

For terminals: n=number of

| For mass storage

input/output ﬁles:

n>65. Large

buffer.

For sequential ﬁles: Format

SI tape

I, X tape

S tape

L tape

mass storage

128 for formatted.

. 513,for unformatted,

513 for unformatted.

513 for formatted or unformatted.

> maximum block length.

65 for formatted.

513 for unformatted.

Record length, r, should always be speciﬁed for ﬁles referenced in list-directed input/output statements. This

speciﬁcation creates a separate working storage area for the ﬁle, which is different from the default area. If

the default area is used, input/output to other ﬁles destroys any data remaining after a list directed read.

Care should be taken when specifying a record length for ﬁle OUTPUT that the length is not so short as to

inhibit output of any run-time diagnostics.

When ﬁle names are made equivalent, the buffer length and record size speciﬁed apply to both ﬁles.

Examples:

PROGRAM ORB (INPUT,OUTPUT=1000,TAPE1 = INPUT.TAPE2=OUTPUT,TAPE4=1000/2000)

All input/output statements that reference TAPE1 reference INPUT instead, and all listable output normally

recorded on TAPE2 is transmitted to the ﬁle named OUTPUT. TAPE4 has a buffer length of 1000 words

with a maximum of 2000 characters per record.

PROGRAM JIM(INPUT,TAPE19=INPUT)

TAPE19=INPUT must be preceded in the same statement by INPUT. TAPE19 becomes the name for the ﬁle

INPUT.

^Z£%\

fDoes not apply to SCOPE 2.

ttSee section 16 Input/Output Implementation.

7-4 60497800 J

PROGRAM SAMPLE (INPUT,OUTPUT,TAPE5=INPUT,TAPE6=0UTPUT)

READ(5,100)A,B,C

100 FORMAT (3F7.3)

This statement reads from logical unit 5; it is declared in the

PROGRAM statement as TAPE5 which is equivalent to INPUT.

WRITE(6,200)A,B,C

200 FORMAT (1H1.3F7.3)

Logical unit 6 is declared as TAPE6 in the PROGRAM state

ment and equivalent to OUTPUT.

BLOCK DATA SUBPROGRAM

/fp^v

BLOCK DATA

BLOCK DATA name

name identiﬁes the BLOCK DATA subprogram if more than one is compiled.

The block data subprogram is a nonexecutable speciﬁcation subprogram that can be used to enter data into

labeled or numbered common (but not blank common) prior to program execution. The name BLKDAT. is

assigned to the block data subprogram if it is not named by the user.

The block data subprogram contains only IMPLICIT, LEVEL, type, DIMENSION, COMMON, EQUIVALENCE,

DATA, and END statements. A valid BLOCK DATA subprogram must contain at least one COMMON state

ment and one DATA statement. Any executable statements are ignored and a warning is issued. All DATA

statements must follow the speciﬁcation statements. Data can be entered into more than one block of

common in a block data program. The speciﬁcations in a BLOCK DATA subprogram take effect when the

binary output ﬁle (speciﬁed by the control statement B option) is loaded.

Example:

j"*-"**.

BLOCK DATA ANAME

COMMON/CAT/X,Y,Z/DEF/R.S.T

COMPLEX X,Y

DATA X,Y/2*((1.0,2.7))/,R/7.6543/

END

Z is in block CAT and S and T are in DEF, although no initial data values are deﬁned for them.

/fSsv 60497800 A 7-5

PROCEDURES

The category of procedure to be used is determined by its particular capabilities and the needs of the program

being written. If the program requires the evaluation of a standard mathematical function, a FORTRAN

supplied intrinsic function or a basic external function can be used. If a single computation is needed

repeatedly, a user-written statement function can be included in the program. If a number of statements are

required to obtain a single result, a function subprogram is written. If a number of calculations are required

to obtain several values, a subroutine is written.

Procedure Communication (later in this section) contains details on how to use procedures and how procedures

use arguments or common to communicate.

SUBROUTINE SUBPROGRAM

r— SUBROUTINE name {prp2 pn)

SUBROUTINE name (pvp2r.'..#pn), RETURNS (b^b. V'bm>

SUBROUTINE name, RETURNS (b^b- ,..., bj

y^^v

(•^iaN

z<^sy

name

Symbolic name of the subroutine.

Dummy arguments that must agree in order, number, type, and level with the actual argu

ments passed to the subprogram at execution time.

/ ^ \

bp..., bm Dummy statement label arguments, that must agree in order and number with the ictiial

statement labels passed to the subroutine at execution time.

The argument lists are optional and limited to a maximum combined total of 63 parameters, f

A subroutine subprogram is executed when a CALL statement is encountered in a program unit. A sub

routine subprogram must not directly or indirectly call itself. The subroutine subprogram communicates with

the calling program unit through a list of arguments passed with the CALL statement or through common.

Calling a Subroutine Subprogram later in this section contains more CALL statement details.

7-6 60497800 A

The SUBROUTINE statement contains the symbolic name that is used as the main entry point of the sub

program. (The ENTRY statement speciﬁes an alternate entry point in the subprogram.) The subprogram name

,f^ is not used to return results to the calling program, does not determine the type, and must not appear in any

other statement in the same subprogram.

^s Subroutine subprograms can contain any statements except PROGRAM, BLOCK DATA, FUNCTION, or

another SUBROUTINE statement. They begin with a SUBROUTINE statement, should have at least one

RETURN statement, and end with an END statement. If control ﬂows into the END statement, then a

f^ RETURN is implied. Control is returned to the calling program when a RETURN, RETURN i or END is

encountered.

1^** Dummy arguments which represent array names must be dimensioned within the subprogram by a DIMENSION

or type statement. If an array name without subscripts is used as an actual argument in a CALL statement

and the corresponding dummy argument is not declared an array in the subprogram, the ﬁrst element of the

/^ array is used in the subprogram. Adjustable dimensions are permitted in subroutine subprograms (details are

given later in this section under Using Arrays).

/^ lhe RETURNS list allows control to be returned to the calling program somewhere other than at the

executable statement immediately following the CALL statement. The CALL statement speciﬁes actual

statement labels to replace the dummy statement label arguments in the RETURNS list. The actual statement

(^ labels must correspond in order and number with the dummy statement label arguments. The

dummy statement label argument i is the statement to which control transfers when RETURN i is executed.

I^ The RETURN statement in section 4 and the CALL statement in this section give further details.

Example 1:

Calling Program Subprogram

SUBROUTINE ERR0R1

WRITE (6,1)

1 FORMAT (5X,22H NUMBER IS OUT OF RANGE)

IF (A-B) 10,20,20 RETURN

10 CALL ERR0R1 END

20 RESULT=(A*CAT) +375.2-ZER0

The subroutine ERROR1 is called and executed if A-B is less than zero. Control returns to

statement 20. This example also illustrates that arguments need not be used.

60497800 A 7-7

Example 2:

Calling Program Subprogram yf^^V

.CALL PGM1(A,B,C)

.XRETURNS (5,10)

5 B=SQRT(A*C)

10 CALL PGM2 (D,E)

SUBROUTINE PGM1(X,Y,Z),

XRETURNS (M,N)

U=X**Y

X=Z+X*Y

20 IF (U+X) 25, 30, 35

25 RETURN M Return is to statement 5 in calling program

30 RETURN N Return is to statement 10 in calling program

35 Z=Z+(X*Y)

RETURN Return is to statement following CALL PGM 1

END

0^$&K

This example illustrates the use of the RETURNS list as well as the use of the; normal

:RETURN'-statefnerit.'-:;' ""■'"-';■

FUNCTION SUBPROGRAM

name

FUNCTION name (p1 pn)

type FUNCTION name (p, ..... pn)

Symbolic name of the subprogram.

XSs^V

pj,...,pn Dummy arguments that should agree in order, number, type;and;leyel with the actual

arguments in the calling program. At least one argument is required; a maximum of

63 is allowed.

type The type may be REAL, INTEGER, D&UBLE, DOUBLE PRECISION, COMPLEX, or

LOGICAL.

A function subprogram performs a set of calculations when its name appears in an expression in a referencing

program unit. Execution of the function subprogram must result in a value being deﬁned for the function

name. A function subprogram can modify the value of one or more of its arguments or store data in common.

7-8 60497800 A

Dummy arguments which represent array names must be dimensioned within the subprogram by a DIMENSION

or type statement. If an array name without subscripts is used as an actual argument in the function reference

and the corresponding dummy argument has not been declared an array in the subprogram, the ﬁrst element

of the array is used in the subprogram. Adjustable dimensions are permitted in function subprograms (details

are given in Using Arrays later in this section).

/"J"-a\

The FUNCTION statement contains the subprogram symbolic name that is used as the entry point when the

function is referenced. The ENTRY statement speciﬁes an alternate entry point in the function. The func

tion name must not appear in any nonexecutable statements other than the FUNCTION statement in the sub

program. The type of the function name must be the same in the referencing program and the referenced

function subprogram. When type is omitted, the type of the function result is determined by the ﬁrst char

acter of the function name. Implicit typing by the IMPLICIT statement takes effect only when no explicit

type appears in the FUNCTION statement.

The function subprogram can contain any statements except PROGRAM, BLOCK DATA, SUBROUTINE,

another FUNCTION statement, or any statement that directly or indirectly references the function being

deﬁned. The function subprogram begins with a FUNCTION statement, should have at least one RETURN

statement, and has an END statement that is treated as a RETURN if executed. Control is returned to the

referencing program when either a RETURN or END is encountered. A RETURN i in a function subprogram

causes a fatal error at compilation time.

A function subprogram can have the same name as that of an intrinsic or basic external function supplied by

FORTRAN. Section 8 deﬁnes the conditions under which programmer supplied routines override the

FORTRAN supplied routines.

Example:

Calling Program Subprogram

DIMENSION ARY (5,5)

10 RES=DIAG(ARY,5)**2

FUNCTION DIAG (A,N)

DIMENSION A(N,N)

DIAG=A(1,1)

DO 70 1=2,N

70 DIAG=DIAG*A(I,I)

RETURN

END

The statement labeled 10 contains the reference to function DIAG. The statement labeled 70 sets the

function name to a value. At the end of the function subprogram execution, RES will have the value of

DIAG squared.

BASIC EXTERNAL FUNCTION

A basic external function is a predeﬁned procedure included with the system. Section 8 contains further

details.

60497800 A 7-9

/fSraPy

INTRINSIC FUNCTION

An intrinsic function is a compiler-deﬁned procedure that is inserted in the referencing program at compile

time. Section 8 contains further details.

STATEMENT FUNCTION

^f^v

name

Pi' .P,

expression

name (p1 ,p2,p3,..., pn) = expression

Type of the function is determined by the type of the function name.

Dummy arguments must be simple variable names. At least one argument is required; a

maximum of 63 is allowed. These arguments should agree in order, number, type, and

level with the actual arguments used in the function reference.

Any expression may be used. It may contain references to intrinsic or basic external func

tions, statement functions, or function subprograms. Names in the expression that do not

represent arguments are normal variables having the same value as they have outside the

function

XS"\

A statement function is a user-deﬁned, single-statement computation and applies only to the program unit containing

the deﬁnition. Since the statement function only deﬁnes the function, the value is computed when the function is

referenced and the actual arguments are substituted for the dummy arguments in the deﬁnition.

During compilation, the statement function deﬁnition is retained by the compiler. Whenever the function is referenced,

instructions are generated in-line to evaluate the function (as opposed to FUNCTION subprograms for which an external

procedure is used at each reference). The expansion of a statement function is the same as writing the expression in

place of the reference. Thus the statement function does not reduce execution speed or efﬁciency.

Statement function names must not appear in DIMENSION, EQUIVALENCE, COMMON or EXTERNAL statements;

they can appear in a type declaration but cannot be dimensioned. Statement function names must not appear as actual

or dummy arguments. If the function name is type logical, the expression must be logical. If the function name is not

type logical, the expression must not be a relational or logical expression. For other types, if the function names and

expression differ, conversion is performed as part of the evaluation of the function. For example, in the program

segment:

LSUM(I,J) = OR(I,J)

A = OR(15,50)

B = LSUM(15,50)

OR is typeless and LSUM is a statement function of type INTEGER. In the ﬁrst function evaluation, no

conversion takes place; the binary value is assigned to A. In the second function evaluation, the value is

converted to ﬂoating point before being assigned to B.

A statement function must precede the ﬁrst executable statement and it must follow all speciﬁcation state

ments. A statement function must not reference itself either directly or indirectly.

/*^i^\

0^^K

7-10 60497800 E

/p*\

Examples:

Statement Function Deﬁnitions Statement Function References

ADD(X,Y,C,D)=X+Y+C+D RES1=GR0SS-ADD(TAX,FICA,INS,RES3)

AVERGE(0,P,Q,R) = (0+P+Q+R) / 4 GRADE=AVERGE(TEST1,TEST2,TEST3,

TEST4)+MID

LOGICAL A,B,EQV

EQV(A,B)=(A.AND.B).OR. TEST=EQV(MAX,MIN).AND.ZED

(.NOT.A.AND..NOT.B)

COMPLEX Z RESULT=(Z(BETZ,GAMMA(I+K))**2-l.)

Z(X,Y)=(1.,0.)*EXP(X)*COS(Y) /SQRT(TWOPIE)

+(0.,1.)*EXP(X)*SIN(Y)

Example 1:

The statement function can be used to substitute a FORTRAN supplied function name in a program con

taining an alternate name for this function.

SINF(X)=SIN(X) Statement function deﬁnition.

A=SINF(3.0+B) +7. Statement function reference.

The above sequence generates exactly the same object code as:

A=SIN(3.0+B)+7.

Example 2:

To compute one root of the quadratic equation ax^+bx-h^O, given values of a, b and c, an arithmetic

statement function can be deﬁned as follows:

ROOT (A,B,C)=(-B+SQRT(B*B-4.*A*C))/(2.0*A)

When the function is used in an expression, actual arguments are substituted for the dummy arguments

A, B, C.

RESA = ROOT (6.5,7.,1.)

is equivalent to writing:

RESA = (-7.+SQRT(7.*7.-4.0*6.5*1.0))/(2.0*6.5)

Wherever the statement function ROOT (A, B, C) is referenced, the deﬁnition of that function - in this

case (-B+SQRT(B*B-4.*A*C))/(2.*A) — is evaluated using the current values of the arguments A, B, C.

60497800 A 7-11

/t^\

PROCEDURE COMMUNICATION

The procedures deﬁned by a statement function or a procedure subprogram are executed when they are

referenced in a program unit.

PASSING VALUES TO A PROCEDURE

Values can be passed between a calling program unit and a procedure as actual arguments in an argument '^

list or through common. Arrays with adjustable dimensions can be used to pass values of arguments.

Arguments passed to a procedure must agree with the procedure deﬁnition in order, number, type, length,

and level.

USING ARGUMENTS

Arguments used for communication between procedures are either actual or dummy (formal). The arguments

appearing in a subroutine CALL statement or a function reference are the actual arguments. The corre

sponding dummy arguments appear in the SUBROUTINE or FUNCTION statement. If a RETURNS list is

used/the actual statement label arguments appear in1 the CALL statement and the dummy statement label

arguments appear in the SUBROUTINE and RETURN statements.

The actual arguments allowed for a particular procedure are given in the discussion of the procedure

reference.

Dummy arguments are used as variable, array or external procedure subprogram names within the subprogram

and can be used to return values to the calling program. The dummy arguments are replaced by the actual

arguments when the procedure is executed. Since all names are local to the program unit containing them,

the same dummy argument name can be used in more than one program unit. A dummy argument must

not appear in COMMON, EQUIVALENCE, or DATA statements within a program unit. *^

Dummy arguments representing array names must appear within the subprogram in a DIMENSION or type

statement giving dimension information. If dummy arguments are not dimensioned, they cannot be referenced /*^\

as arrays in a subprogram.

In a subprogram, the deﬁnition of a dummy argument that is associated with a constant actual argument or with ^^

any expression except a variable or array element is prohibited.

If a subprogram reference causes a dummy argument to be associated with an entity in common in the refer

enced subprogram, deﬁnition of the dummy argument or of the entity in common is prohibited. If a subpro

gram reference causes two dummy arguments to be associated, the deﬁnition of either in the referenced subpro

gram is prohibited.

7-12 60497800 C

Example 1:

Calling Program Subprogram

/|S!S

W(I,J)=FA+FB-GRATER(C-D,3*AX/BX)

FUNCTION GRATER(A.B)

IF (A.GT.B)1,2

1 GRATER=A-B

RETURN

2 GRATER=A+B

RETURN

END

This example shows the normal use of arguments in a function subprogram. The actual argument C-D is

used in place of the dummy argument A and 3*AX/BX is substituted for dummy argument B when the

function subprogram is executed.

Example 2:

CALL SUBA(1.5) SUBROUTINE SUBA(R)

IF (R.NE.O) R = 0

This example contains a prohibited deﬁnition of a dummy argument, R, which is associated with a con

stant actual argument.

Example 3:

CALL SUBB (X, X) SUBROUTINE SUBB (A, B)

A = Y

2 = B

This example contains a prohibited deﬁnition of a dummy argument, A, which has been previously

associated with another dummy argument, B, in the referencing program unit.

Example 4:

COMMON x

CALL SUBC (X)

SUBROUTINE SUBC (B)

COMMON A

A = Y

Z = B

This example contains a prohibited deﬁnition of an entity in common, A, which is associated with a

dummy argument, B, in the same subprogram.

jp*n

60497800 A 7-13

USING COMMON

Example:

PROGRAN 1 CMN <INPUTtOUTPUT)

COMMON NED (10)

READ 3* NED

3 FORMAT. (10X3) • •:-•'- •■*.':'.'-■• ■'■. •" - .-

- >CALti:JA\/&: ,'.•' '<'•-'., .' K

-.•'. STOP * % ; • , ^ ( i <

-: END-;-!

SUBROUTINE JAVG

CTHIS, SUBROUTINE COMPUTES THE AVERAGE

cCOMMON (I * * ' *<'■**''■**' „

COMMON N(IO) ;

ISTORE » 0 - ' ' -

DO 11 =M^10':-''1' '; '■ - J-V i;. -•

1 ISTORE a ISTORE ♦ N(I)

ISTORE s ISTORE/10

PRINT l»t ISTORE

Z FORMAT (*1AVERAGE = *tI10)

RETURN -, . '• <■'

END

OF THE FIRST 10 ELEMENTS IN

AVERAGE s 4£>

The array NED in program CMN and the array N in subroutine JAVG share the same locations in common.

NED(l) shares the same location with N(l), NED(2) with N(2), etc. The values read into locations

NED(l) through NED(IO) are availably to subroutine JAVG. JAVG computes and prints the average of

these values. *

USING ARRAYS

Common can be used to transfer values between a calling program unit and a subprogram. Passing values )

through common is more efﬁcient than passing values through arguments in a CALL statement or function reference.

If a dummy argument in a subprogram is associated with an entity in a common block in the same subpro-

gram, the definition of either is prohibited. ■

/ssx

"SJ5ijk\

/S^s

,««%.

The array dimensions in a subprogram must be the same as those in the calling routine if the subscripts are

to agree between the called and calling program units. If a dummy argument is not dimensioned, it cannot ^

be referenced as an array in the subprogram.

If any of the entries in a subscript of a type or DIMENSION statement is an integer variable name, the 'l

array is called an adjustable array. The variable names are called adjustable dimensions. Such an array

can only appear in a procedure subprogram. The dummy argument list of the subprogram must contain

the array name and the integer variable names that represent the adjustable dimensions. The values of the ^^)

actual arguments that represent array dimensions in the argument list of the reference must be deﬁned

prior to calling the subprogram and cannot be redeﬁned during execution of the subprogram. The absolute

7-14 60497800 A '

/8Jft\

size of the actual array may not be exceeded. For every array appearing in an executable program, there

must be at least one constant array dimension associated through subprogram references.

In a subprogram, an array name that appears in a COMMON statement must have ﬁxed dimension

speciﬁcations.

REFERENCING A FUNCTION

A function is referenced when the name appears in an expression. A function must not directly or indirectly

reference itself. The reference can appear anywhere in an expression that an operand of the same type can

be used.

When a statement function or intrinsic function is referenced, instructions are generated in-line to evaluate the

function. The value is computed with the actual arguments substituted for the dummy arguments in the

deﬁnition.

When a function subprogram or a basic external function is referenced, control is transferred to the function

subprogram and the values of the actual arguments are substituted for the dummy arguments. Control is

returned to the referencing program unit when a RETURN is encountered.

Actual arguments in a function subprogram reference may be an expression, constant (including Hollerith),

variable, array name, array element name, subroutine subprogram name, external function name (not intrinsic

function or statement function), or function reference (the function reference is a special case of an arith

metic expression).

Example:

Calling Program

Z=A+B-J0E(3.*P,Q-1)

R=S+JAM(Q,2.5*P)

Function Subprogram

FUNCTION JOE(X,Y)

10 J0E=X+Y

RETURN

ENTRY JAM

IF(X.GT.Y)10,20

20 J0E=X-Y

RETURN

END

Function subprogram JOE is executed as a result of its name appearing in another program unit.

00&S

60497800 A 7-15

CALLING A SUBROUTINE SUBPROGRAM

,<fi^«S\

CALL name

CALL name (p1 ,..., pn)

CALL name (pt,...,pn), RETURNS (b., ,...,bm)

CALL name, RETURNS (^ ,....bm)

name

Pl« »Pr

,bm

Name of subroutine called.

Actual arguments which must correspond in order, number, type, and level -with those

speciﬁed in the SUBROUTINE statement.

Actual statement labels in the calling program unit that correspond in order and number

with the dummy statement label arguments in the SUBROUTINE statement. This

speciﬁcation can be omitted if control returns to the statement immediately following

the CALL statement.

The total number of arguments of both kinds must not exceed 63.

A subroutine subprogram is executed when a CALL statement is encountered in a program unit. A subroutine

must not directly or indirectly call itself. The CALL statement transfers control to the subroutine and either

a RETURN or a RETURN i in the subroutine returns control to the calling program unit. If a RETURN is

encountered, control is transferred to the ﬁrst executable statement following the CALL statement. If

RETURN i is encountered, control is transferred to the statement corresponding to i in the RETURNS list.

(The RETURN statement is section 4 and Subroutine Subprogram in this section contain further details on

the RETURNS list.)

The CALL statement can contain actual arguments and statement labels. They must correspond in order,

number, type, and memory level to those in the subroutine subprogram deﬁnition.

The name in the CALL statement can be an alternate entry point in a subroutine subprogram, as speciﬁed in

an ENTRY statement, or a subroutine name. The subroutine name must not appear in any speciﬁcation state

ment in the calling program except an EXTERNAL statement.

Actual arguments in a subroutine subprogram call can be any of the following: expression, constant, variable,

array name, array element name, subroutine subprogram name, basic external function name (not an intrinsic

y<^\

7-16 60497800 C

or statement function name), function reference (the function reference is a special case of an arithmetic

expression).

Example 1:

Calling Program

DO 5 I = 1,20

5 CALL GRATER (STACK(I),TEMP(I))

Subprogram

SUBROUTINE GRATER (A,B)

IF (A.LE.B) GO TO 2

1 B = A - B

RETURN

2 B = A + B

RETURN

END

#j*$^s

The subroutine subprogram GRATER is called 20 times since the CALL statement as the last statement in

a DO loop causes looping to continue until the DO loop terminal parameter, 20, is satisﬁed.

Example 2:

Calling Program

PROGRAM MAIN(INPUT.OUTPUT)

Subroutine Subprogram

10 CALL XC0MP(A,B,C),

XRETURNS(101,102,103,104)

101 CONTINUE

SUBROUTINE XCOMP (B1.B8.G),

XRETURNS(Al,A2,A3,A4)

. IF(B1*B2-4.159)10,20,30

10 CONTINUE

GO TO 10

102 CONTINUE

RETURN Al

20 CONTINUE

Return to 101

GO TO 10

103 CONTINUE

RETURN A2

3Q CONTINUE

Return to 102

0^*. GO TO 10

104 CONTINUE

END

IF (Bl)40,50

40 RETURN A3

50 RETURN A4

END

Return to 103

Return to 104

The values of A, B, and C in the CALL statement replace Bl, B2, and G in the SUBROUTINE statement

for use in the subprogram XCOMP. Statement numbers 101, 102, 103, and 104 replace Al, A2, A3, and

A4 in the SUBROUTINE and RETURN i statements.

60497800 A 7-17

USING THE ENTRY STATEMENT

ENTRY name

name is an entry point in a procedure subprogram.

The ENTRY statement deﬁnes an alternate entry point, which is other than the ﬁrst executable statement, in

a procedure subprogram. The ENTRY statement can appear anywhere an executable statement can appear in

the subprogram except as the object of a logical IF or within the range of a DO where it is ignored and a

warning diagnostic is issued. A procedure subprogram can contain any number of ENTRY statements. The

ﬁrst executable statement following ENTRY becomes the alternate entry point to the subprogram. ENTRY

statements cannot be labeled and cause a fatal-to-execution error in a main program unit.

In the subprogram, the entry name can appear only in the ENTRY statement and each name must appear

in a separate statement. A function entry name must be the same type as the name in the FUNCTION

statement. The entry name must be unique within the program.

In the calling program, the reference to the entry name is made just as if reference were being made to

the function subprogram or subroutine subprogram in which the entry name is contained. itie nahie can appear

in an EXTERNAL statement, and if it is a function subprogram entry name, in a type statement.

The dummy arguments, if any, appearing with the FUNCTION statement or SUBROUTINE statement do not

appear with Ae ENTRY statement, but.are assumed to be the same as for the main entry, point. .

to a function subprogram, the value of the function is the last value assigned to the name of the function,

regardless of which ENTRY statement was used to enter the subprogram. The function name is used to

returii results to the calling program even, though the reference was through an entry -name,, r

j*S^Sv

/*^\

. Example;;!:

Calling Program

COMMON VsiTi; (25)

Subroutine Subprogram

*]Main .entry point-

SUBROUTINE CLEAR (ARAY) .

\ » : . , ; V ' - y y " : - ' \ y , ; . ' / . ; - P p e n s i o n ; a r a y : ( 2 5 j ■

;. *\ ■'.'-■•--''","...' v" '''"*"•,' do 10G ± - 1,25'^ —---

CALL CLEAR (SET1) .100 ARAY (I) - 0.0 '„"..-

.-"•-'.''-*•/ .' ""■ ■/ '," '■ ENTRY F I L L '- - ' ' , _ - : "' " - " ~ • ' _.-" . '

".,.'..' '. 3."READ 2, VALUE, IPLACE-*——— Alternate entry point

CALL FILL (SET1)

2' FORMAT (10X, F7.2, 14)

ARAY (IPLACE) - VALUE

IF (IPLACE .GT. 24) RETURN

GO TO 3

END

7-18 60497800 A

00**\

00&\

00*^\

At some point in the calling program, a call is made to the subroutine: CALL CLEAR (SET1). The

array SET1 is set to zero and values are read into the array. Later in the program, a call is made again

to the subroutine CLEAR; but this time it is entered at the entry point FILL. When FILL is called,

further values are read into the array SET1 without ﬁrst setting the array to zero.

Example 2:

Calling Program Subprogram

RESULT=FSHUN(X,Y,Z) FUNCTION FSHUN(A,B,C)

RES2=FRED(R,S,T) 3 FSHUN=A*B/C**2

RETURN

ENTRY FRED

IF(A .LE. 702.) GO TO 3

FSHUN=-(C+A)/B

RETURN

END

When the FUNCTION FSHUN is entered at the beginning of the function, or through the ENTRY FRED,

the result will be returned to the calling program through the function name FSHUN.

Example 3:

FUNCTION CAT(A,B)

D0G-1O.+3.2

ENTRY DOG

The entry point name DOG is not valid because it has been used as a variable.

OVERLAYS

To reduce the amount, of storage required, and to make efﬁcient use of ﬁeld length, a user can divide a

program into overlays. Prior to execution, the object modules of an overlay program are linked by the loader

^^^ arid placed on a mass storage device in their absolute form; no time is required for linking at execution time.

f (See Loader Reference Manual for more details.)

An overlay is a portion of a program written on a ﬁle in absolute form and loaded at execution time

C without relocation. As a result, the size of the resident loader for overlays can be reduced substantially.

Overlays can be used when the organization of core can be deﬁned prior to execution.

S*v&\

f When each overlay is generated, the loading operation is completed by loading library and user

subprograms and linking them together. The resultant overlay is in ﬁxed format, in that internal references

are ﬁxed in their relationship to one another. The entire overlay has a ﬁxed origin address within the ﬁeld

f" length and, therefore, is not relocatable. The overlay loader simply reads the required overlay from the

overlay ﬁle andloads it starting at its pre-established origin in the user's ﬁeld length.

60497800 E 7-19

Overlays are loaded into memory at three levels: zero, primary, and secondary.

Fixed starting

address for

primary overlays

Fixed starting

address for (1,n)

secondary overlays

Zero overlay (0,0)

Primary overlay (1,0)

Secondary overlay (1,1)

The zero or main overlay is loaded ﬁrst and remains in core at all times. A primary overlay may be loaded

immediately following the zero overlay, and a secondary overlay immediately following the primary

overlay. Overlays may be replaced by other overlays. For example, if a different secondary overlay is

required, the overlay loader simply reads it from the overlay ﬁle and places it in memory at the same

starting address as the previously loaded overlay.

Fixed starting address

for primary overlay

Starting address for

secondary overlay

(4,2)

Fixed starting address

Secondary

overlay

(3,1)

for secondary overlay

Zero (0,0)

Primary

overlay (4,0)

7-20 60497800 E

00^\_

/0$&\

/ ^ N

/sP^v

For example, (1.0) (2,0) (3,0) (4.0) are primary overlays. (3,1) (3,2) (3,5) (3,7) are secondary overlays

associated with primary overlay (3,0). Secondary overlays are denoted by the primary number and a non

zero secondary number. For example, (1,3) denotes that secondary overlay number 3 is related to primary

overlay (1,0). (2,5) denotes secondary overlay 5 is related to primary overlay (2,0).

A secondary overlay can be called into core only by its primary overlay. Overlay (1,0) can call (1,2); but overlay

(2,0) should not call (1,2), nor should overlay (0,0) call overlay (1,1).

Overlay numbers (0,n) are not valid (n > 0). For example, (0,3) is an illegal overlay number.

Execution is faster if the more commonly used subprograms are placed in the zero overlay, which remains in main

memory at all times, and the less commonly used subprograms are placed in primary and secondary overlays which

are called into memory as required.

An overlay can consist of one or more FORTRAN or COMPASS program units. Each overlay must contain

one FORTRAN main program; it need not be the ﬁrst program unit in the overlay. The program name in

the PROGRAM statement becomes the primary entry point for the overlay when the overlay is called.

OVERLAY COMMUNICATION

Data is passed between overlays through labeled or blank common. Any element of a labeled or blank

common block in the main overlay (0,0) may be referenced by any higher level overlay. Any labeled or

blank common declared in a primary overlay may be referenced only by the primary overlay and its associ

ated secondary overlays - not by the zero overlay. If blank common is used for communicating between

overlays, the user must ensure that sufﬁcient ﬁeld length is reserved to accommodate the largest loaded over-

lav in addition to blank common.

Blank common is located at the top (highest address) of the ﬁrst overlay in which blank common is declared.

For example, if blank common is declared in the (0,0) overlay, it is located at the top of the (0,0) overlay

and is accessible to all higher level overlays. If blank common is declared in the (1,0) overlay, it is allocated

at the top of the (1,0) overlay and is accessible only to the associated (l,k) overlays. Labeled common blocks

are generated in the overlay in which they are ﬁrst encountered; data may only be preset in labeled common

blocks in this overlay.

CREATING AN OVERLAY

An overlay is established by an OVERLAY directive preceding the program units for that overlay. An overlay

consists of all program units appearing between its OVERLAY directive and the next OVERLAY directive or

the end of source input. The directive must be punched starting in column 7 or later.

The PROGRAM statement for the zero or main overlay (0,0) must specify all ﬁle names such as INPUT,

OUTPUT, TAPE1, etc., required for all overlay levels. File names should not appear in PROGRAM statements

for other than the (0,0) overlay. The compile-time warning or informative message I/O FILE NOT DEFINED

should be ignored for programs outside the (0,0) overlay.

60497800 G 7-21

Loading overlays from a ﬁle requires an end-around search of the ﬁle for the speciﬁed overlay; this can be

time-consuming in large ﬁles. Under NOS and NOS/BE, a Fast Overlay Loading facility (FOL) is available.

When FOL is enabled, an FOL directory is created in space allocated in the (0,0) overlay. When a higher

level overlay is loaded, the directory is used to locate the overlay, and the overlay is loaded with a single disk access.

This is the fastest method available for overlay loading and is recommended for applications where speed is essential.

The FOL facility requires that all overlays in the overlay structure reside on the same ﬁle in the same order in which

they were generated. FOL mode is speciﬁed by the presence of the OV parameter on the OVERLAY directive.

If the FOL facility is not used, and speed is essential, each overlay should be written on a separate ﬁle, or the overlays

should be called in the same order in which they were generated.

The group of relocatable decks to be processed by the loader to create an overlay-structured program must be

presented to the loader in the following order. The main overlay must be loaded ﬁrst. Any primary group followed

by its associated secondary group can follow, then any other primary group followed by its associated secondary

group, and so forth.

The OVERLAY directive format is:

fname

origin

OVERLAY (fname,i,j,origin,OV=m)

Name of the ﬁle on which the generated overlay is to be written.

Overlay level numbers in octal without the B sufﬁx. The numbers speciﬁed are not checked or

converted by FORTRAN.

Optional parameter specifying the origin of the overlay; not allowed for (0,0) overlay. The

CYBER loader (NOS/BE, NOS) accepts any of the following forms; the SCOPE 2 loader allows

only Cnnnnnn. If the origin parameter is omitted, the preceding comma must also be omitted.

Cnnnnnn The overlay is loaded nnnnnn words from the start of blank common, nnnnnn

must be an octal number of up to six digits.

0=nnnnnn The overlay is loaded at the address speciﬁed; nnnnnn must be an octal

number >110o.

0=eptname The overlay is loaded at the address of the entry point speciﬁed, which must

have been declared in a lower level overlay.

0=eptname The overlay is loaded at the address of the entry point speciﬁed but the address

± nnnnnn is biased by the amount of the offset.

OV=m Optional parameter specifying the total decimal number, m, of higher level overlays in the over

lay structure. Valid only on a (0,0) overlay directive and only under NOS and NOS/BE. Signals

the overlay generator and overlay loader to go into Fast Overlay Loading (FOL) mode. The

value of m must not exceed 20000 (decimal). If the OV parameter is omitted, the preceding

comma must also be omitted.

The ﬁrst overlay directive must have a ﬁle name and ij must be 0,0. Subsequent directives can omit ﬁle name indicating

that the overlays are to be written on the same ﬁle. All overlays need not reside on the same ﬁle, unless in fast overlay

loading mode. The second overlay directive must be of a primary overlay such as 3,0.

If the origin parameter is omitted, the overlay is loaded in the normal way directly after the zero overlay. The origin

parameter cannot be included on the zero overlay directive. It is used on primary and secondary overlay directives to

allow the programmer to change the size of blank common at overlay generation time.

7-22 60497800 H

Example:

OVERLAY(FNAME,0,0 ,0V=4)

PROGRAM CAT(INPUT,OUTPUT,TAPE5=INPUT)

OVERLAY(1,0)

PROGRAM A

OVERLAY(1,1)

PROGRAM B

/*3N OVERLAY(1,2)

PROGRAM C

OVERLAY(1,3)

PROGRAM D

All the above overlays are written on the ﬁle FNAME. The OV parameter signals FOL mode and indicates

that there are 4 higher level overlays.

CALLING AN OVERLAY

Overlays are called with a CALL OVERLAY statement; only the zero overlay (0,0) can be loaded when a pro

gram call control statement is encountered. The format of the CALL OVERLAY statement is:

CALL OVERLAY (fname,i,j,recall,k)

/SSN

fname

recall

The name of the ﬁle or overlay in H format.

Overlay level numbers in octal with the B sufﬁx, or the decimal equivalent.

Optional recall parameter.

Optional parameter specifying where the overlay is located; can be zero, non-zero,

or L format Hollerith constant.

60497800 D 7-23

■^^\

/*^§v

If the k, parameter is zero or not speciﬁed, the overlay is included in the ﬁle named by fname. If a non-zero .

k parameter is speciﬁed, fname is the name of the overlay to be loaded. If k is an L format Hollerith con

stant, the overlay is loaded from the library named ih the constant (not applicable'. under &OS 1). If k is

any other non-zero value, the overlay is loaded from the global library set (refer to the appropriate operating

system reference manual or the Loader Reference Manual).

If 6HRECALL is speciﬁed as the recall parameter, the overlay is not reloaded if it is already in memory. If the

overlay is already in memory and the recall parameter is not used, the overlay is actually reloaded, thereby changing /lS^

the values of variables in the overlay. ;

The three parameters fname, i, and j must be speciﬁed or the results are unpredictable.

When a RETURN or END statement is-encountered in the main program of a zero overlay, execution of the

program terminates and control returns to the operating system. When either of the statements is encountered ^^

in a primary or secondary overlay, control returns to the next executable statement after the CALL OVERLAY /

statement that invoked the current overlay.

E x a m p l e 1 : • - • ' ~ " . : • > . . • ' y ; " _ " ' ' " " . ' )

"CALLOyERLAY(lHA,l,0) '"'.'•',, '.'-■' ' • ' • , V ; ! ' " " < " %

This statement causes a primary; overlay to be,loaded from the ﬁle harried A.

Example 2; ' '"■*' , ' •' • <' • -' ' y '- * V ' , . ' ; ' •■ - ', '* , f, " ;: . . -. . ' .1 V •.

CALL OVERLAY(3HBJR,0,0,0,1) ...

This statement, which speciﬁes the k parameter as a non-zero value, causes a main overlay with

' , ' the name BJ& to.be loaded from the global, library set. ,

7-24 60497800 A

Example 3:

OVERLAY(XFILE,0,0)

PROGRAM ONE(INPUT,OUTPUT.PUNCH)

CALL OVERLAY(5HXFILE,1,0,0)

STOP

END

OVERLAY(XFILE,1,0)

PROGRAM ONE ZERO

CALL OVERLAY(5HXFILE,1,1,0)

RETURN

END

OVERLAY(XFILE,1,1)

PROGRAM ONE ONE

RETURN

END

Execution of RETURN in the 1,1 overlay returns control to the statement in the 1,0 overlay following the

1,1 call. Execution of RETURN in the 1,0 overlay returns control to the statement in the main overlay

following the 1,0 call.

60497800 A 7-25

j^N

/<"S5s.

Example:,

Secondary Overlay

(1.1)

Source Deck

Primary Overlay

(1.0)

Source Deck

Main Overlay

(0,0)

Source Deck

Call to

Primary Overlay

FRANK 1,0

/S5N

Preparation of Overlay 0,0; 1,0; and 1,1

7-26 60497800 A

The above example illustrates the preparation of zero, primary and secondary overlays. The zero overlay

0^ FRANK 0,0 consists of a main program LEO and a subroutine GROUCH. The primary overlay FRANK 1,0

consists of a main program RDY. The secondary overlay FRANK 1,1 is a main program MLT. All three over

lays reside on the ﬁle FRANK.

The LOAD(LGO) and NOGO control statements request the loader to load the program from the ﬁle LGO.

As the loader reads ﬁle LGO, it encounters the overlay directive OVERLAY (FRANK,0,0) which instructs it

/#** to create a main overlay from the program and write it on ﬁle FRANK. When the absolute form of all the

* overlays has been generated, execution begins when the control statement FRANK, is encountered. FRANK,

causes the main overlay to be loaded from ﬁle FRANK and executed.

V During execution of the main overlay, the CALL OVERLAY (5HFRANK, 1,0,0) statement is

encountered and the primary overlay 1,0 is loaded into central memory. The CALL OVERLAY

/r-*s (5HFRANK,1,1) statement in the primary overlay causes the secondary overlay to be loaded into

memory.

/s=\ The primary and secondary overlays can reside on ﬁles other than FRANK. For example, the

• primary overlay could be on file JIM and the secondary overlay on file JOHN-

FTN.

f^ LOAD(LGO)

NOGO.

FRANK.

f^ 7/8/9

; OVERLAY (FRANK,0,0)

( PROGRAM LEO (INPUT.OUTPUT,TAPE1)

0 S \ CALL OVERLAY (3HJIM,1,0,0)

OVERLAY (JIM,1,0)

PROGRAM RDY

CALL OVERLAY (4HJ0HN,1,1,0)

END

OVERLAY (JOHN * 1,1)

PROGRAM MLT

END

Example:

The following program, for execution under NOS/BE 1 or SCOPE 2, contains several subroutines and func

tions and is to be used repeatedly. The entire program can be generated, therefore, as a main overlay and

placed on the ﬁle in the absolute form. The control statement CATALOG creates a permanent ﬁle OVRLY

60497800 A 7-27

where the absolute form of the program will be kept. When the program is required again, the permanent ﬁle

OVRLY is called by an ATTACH control statement.

Control

Cards

FTN.

LOAD(LGO)

NOGO.

CATALOG(REPEAT,OVRLY,ID=IBB)

7/8/9 .

OVERLAY (REPEAT,0,0)

PROGRAM A (INPUT,OUTPUT,TAPE1)

END.

SUBROUTINE B

""I

Main

Overlay

E N D ,

FUNCTION C

END

SUBROUTINE D

**\

END

REAL FUNCTION E

END

6/7/8/9

Main program A and the subroutines and functions B-E reside on the ﬁle REPEAT in absolute f< rm

They can be called and executed without recompilation by the control statements:

job card

ATTACH(REPEAT,OVRLY,ID=IBB)

REPEAT.

7/8/9

data

6/ 7 / 8 / 9

The operating system and Loader reference manuals give full details of the control statements which appear in

the program above.

7-28

V^Sv

/Ej^SV

y^^v

>lute f( rm.

yfjCWT^X

;h appear in

60497800 A

FORTRAN EXTENDED SUPPLIED PROCEDURES 8

J0$$&\

FORTRAN Extended provides certain procedures that are of general utility or difﬁcult to express in

FORTRAN; they are referenced in the same way as user-written procedures. The three classes of FORTRAN

Extended supplied procedures are: intrinsic functions, basic external functions, and utility subprograms.

Utility subprograms include operating system interface routines and subroutine packages such as the mass

storage input/output routines, the CYBER Record Manager interface routines, and the Sort/Merge interface

routines.

INTRINSIC FUNCTIONS

0i^\

\ An intrinsic function is a compiler-deﬁned procedure that returns a single value. It is inserted in the refer

encing program at compile time. The form of the intrinsic function reference is the same as the statement

-g^ function reference outlined in section 7.

When a variable, array, or statement function is deﬁned with the same name as that of an intrinsic function,

-^ the user-supplied deﬁnition prevails.

t0£"s.

When a function subprogram is deﬁned with the same name as that of an intrinsic function, the user deﬁnition

prevails only if, in the calling program unit, the name of the function appears either in an EXTERNAL state

ment or in an explicit type statement that changes the type associated with the intrinsic function.

Table 8-1 lists the intrinsic functions provided by FORTRAN Extended. The results of functions with the

type listed as no mode assume the type of the expression in which they are used; such functions cannot be

used as operands in logical expressions.

In tables 8-1 and 8-2, a single argument is denoted by A and multiple arguments by Al, A2, .... The

symbol 0 denotes a very small positive quantity; 0~ denotes a very small negative quantity.

BASIC EXTERNAL FUNCTIONS

A basic external function is a predeﬁned procedure included with the FORTRAN Common Library. These

procedures are used to evaluate standard mathematical functions such as sine, cosine, square root, etc. A

basic external function is referenced by the appearance of the function name with appropriate arguments in

an expression.

A basic external function ordinarily is called by value; however, it is called by name if, in the calling program

unit, the name of the function appears either in an EXTERNAL statement or in an explicit type statement

that overrides the type associated with the function, or if option T, D, or OPT=0 is speciﬁed on the FTN

control statement. (Section 17 contains a description of call by value and call by name.)

When a variable, array, or statement function is deﬁned with the same name as that of a basic external func

tion, the user deﬁnition overrides the system deﬁnition.

60497800 D 8-1

/*^\

<_>

Y=ABS(X) J=IABS(I) DOUBLE A,B B=DABS(A) Y=AINT(X) l=INT(X) DOUBLE Z J=IDINT(Z) B=AM0D(A1,A2) J=M0D(I1,I2) X=AMAXO(l,J,K) A=AMAX1(X,Y,Z) L=MAX0(l,J,K,N) l=MAX1(AfB) DOUBLE W,X.Y,Z W=DMAX1(X,Y,Z) Y=AMIN0(l,J) Z=AMIN1(X,Y) L=MIN0(l,J) J=MIN1(X,Y) DOUBLE A,B,C C=DMIN1(A,B)

IX.

T—

o o

^ " I L Real Integer Double Real Integer Integer Real Integer

l _ _ O J

9 1 0 1 —

_ _ en o> -Q

( 0 C O O J » 3

<» cu +■" ■{_ o

oc cc £ £ q Real Real Integer Integer Double

Type of Argument

Real Integer Double Real Real Double Real Integer

_ _ . C D

c u a > - r

e n _ O ) _ - 3

Q J m O J C O - J

■•_ 3 ♦_ 8 o

£ cc £ oc Q Integer Real Integer Real Double

■f

^ 2

•2 •=

E «

> z

CO ABS IABS DABS 1NI0IINI

IN IV

4 - 4 —

Q 4 —

O Q

2 O

< 2 AMAXO AMAX1 MAXO MAX1 DMAX1 AMINO AMIN1 MINO MINI DMIN1

4—

-J

Number of Arguments

—"

Sign of A times largest integer < | A1 for |A|< 248-1

Al (mod A2)

T— "*"■■

< ;

* X «

CO CM

2 <

< :

|«

2 < Conversion from integer to real

. _ §

f . _

c *->

— S

t->

~ ; o >

v t _ ;

_ 3 c o

< >

't*

_. Remainder ing t Choosing largest value

O J >

.E tt

V I O i

O —

O C O

C J i n

li.

3ro

cu S3

OCO

c~"

v> ..

OS3

"D ro

Ti '■'

/*s?*v

cu O f^

O) is

*- s. 00

C O I M •«* 00

JV>

CU <»-

■c £

-i*

• f- f c •M

'" E ro

rr 3

X o> D"

CO CT

u. CD

2: •_

91 C

■c o

5 w ro

fCO

-° a, v>

</>

« c vt

ro-O CO

TJ o>

t u c

a> V)

\zz o)

CD T3

Er-

« o

-l

— C

s s

r o r o

— «/>

O si

: > _

+- ■

CU <JJ

< c u fc

of o> _.

§*' ro

_ * C O

SZ sz

\- \-

^^v

8-2 60497800 J

/0^\

/fS\

(X.

Ul LL

—. N —

— CM yT >-

> - £ x

O — m z

2 ^ -■ o

__ S2 e° 55

w_2q

I I ~ O i i

n - ■■ > a n A=DIM(C,D) J=IDIM(I1,I2)

o o

l_T 3

• ^ U L

*-•

t - O J

o> ~z

_ a> -2

C O 0 ) 3

cu «_ o

DC £ Q

_ O )

C O «

0 ) £

oc £

Type of Argument 75

_ c u

® X 3

~ 5 0 ? 3

OJ *_ O

OC £ Q

"co cij

£_=

' 4 -

4—

• 4 -

z z

Z g (5 _i

59 OC

-1

Number of Arguments

<- CM CM

1—

E ,_•

o j» 1—

§ = 'O — r o

_ B 0)

§ J_ E

S i C O C O

O _ w

.9?

'so

4-"

5 4-

Ze CM

C U T -

< 0 J

A CM £

< J- <

it: < V/

o <

Q £

.12 Z o)

I **■ o

>■ — —•

-?.<_£

__<

o <

SCO).

•? ;§ J.

T * O J —

__<

DC <.

"~ ° 4t

S 2 =

I .-.2 o

> OT _. -

"._: x. «-

03 LU <

X t T -

o o

° +*

C O c

•V a>

b3 8

.2 §

(A .2".

£ S

C 3

— u_

0 1 -

t n O )

rrj W

1— O

._. ^

' 3 5 $ Z

O 1 ^

a . Q

__ *-

— 1 O -

o> E

- 1 C / 5

'tn

§c

(1) x:

c-$.

oOJ

OT OJ *->

a</9

cO)

oOT

■f-f" OJ

r sz

cu ft

F»

-j

O) CD

eo CO

CO CO

o„

uZ3

oCO

t_ 91

oSZ

OT u

Q. CO

01 CD

3♦->

OuCO

T7 CD

OOF

U . L L < 5

60497800 F 8-3

«**^v

I—I

' / > • „ - " „ . * ' ' - y . > ■ * " . < ' * ' . "

■"*»■- * j v ( * J . t * ! %

JM ,<?' .' .'•'"■' • ' y , ! • tf^*''. 1 .^ ' x<

UJ _l

"5. . u . > - . t , y < • , . t o * ' • ' . " " s -1 <

111

~ 3 C y « " , ' A ' • ■Q - i n

O I I

' « , . "■> " - ' ' ' • * , ' ' , • s

.,< • *. *t O 03

* > y. - s * „ '- 7 < x * ■ .' ; y «

o o " ' - " t o - ' ' ' - • V. , * ' , ' \ . . * ' < ; » • ' ,

■"^■so. • : - < i * - y * * ■ * »* •' ><- ".to* "

' S - X *

. o -r\

: i . . . « •

"5i

OC OC

' y*4*'\ * y '»'♦'•*** " '"** * -. * v"*- - <.~'* • y « * i , . ' < <

' " t ~ ~ " i ' ' V » * < ~ » " . " ' " ' - i " * ' ' . ' > .

! " O f . J * ' 1 * * ^ - 1 ' > ' < > , ' ' " * ' > * > ' -.

O cu * > . * ~ ^ * / f 1 ^ J ~ i < / t « V , V J . * f . * * " . * . *

Type Argum . to ! !.S'y .▶ .',-'." *'** > „r v ^

* * ' ' ' s

S i . ; . " 0J

U* " % . ' - ' J » * ' ^ * ' ' * - ? ' ' • * ' '

' i " *

.§ E

i_' . _J -J

-2? ^ <- ., UJ

CO - j w y ^ i y , ' * « _ , , ! • , ; 2 . ; , \ \ CO QC

,* r"y>:7* -.', '' ■ - - _ -, - \;r> "' i *! " r

!i c - i ' ' « ' > ! ' ^ - , < '

- s i

Ii' c m ■ - ; ' ' / , ' : . . - ; « - c ^_ __

, . i

3 w

z < •.; '",'." :.'

,cm " ,11 i*- c!,t 3 » - " _ Tf

< , « . ? = § * j » - s q » 4 ,

( 0 J O S - < / C 3 . c j j « c f , 0 ' . ; _ ; ,

*2J cu _ -to o' fS ,OT *0 J_ tn

E ot ct +; $

ft ~" (0 4 - 11

' - ( ,

cu cu

.2 §

OT .—.

c *■*O T 3 " C O E

9 1 O i - i

•c 2

_ §

— u.

M £ 11

.— h- o ot —:

iS "E +- "o 2.

g S S _ ff

O CO CL CL <

cc o g

S tr a

o _ ! <

/ f S % .

/!^S

^S^v

S 3

3*—

OTJ

-j

_. aj

_. +->

(1)

TJ OJ

CD O)

(0 ro

-1

<'o

2CD

*- O.

oCD

a> TJ

—■ro

cu i_

-Z

HLL

/<f<^S

/^^v

8-4 60497800 C

z*-fe».

jJS8\.

0m*K

0^\

N_^

>—l

t—<

C/3

COMPLEX A D=AIMAG(A)

>- X

" J U J ,

ni ■-

3 Q

O ii

Q > COMPLEX C C=CMPLX(A1,A2) COMPLEX X,Y Y=C0NJG(X)

a .

LL.

-5

o o

91 **P

•^ u. CO

-3

CJ OC

•M

Type of Argument

"<0

+-»

>•

■M

^ c

_5 E

E z

—I

-J

—j

1—

U-

Number of Arguments

-■CM -■—•»

^ 4 n *

|<S>

—- 1

*N_

CM CU

< t » -

C O —■•

. _ ; C - ~

_ , a >

O T £ j . O J C - - .

, 3> ._"_: E .-

§ > . - * p 5 ? .- * X T " 5 > C O T 3

2 E o o >. g>

■ 1 o "fi a, E g

K§_.*g;.., ;

'■'• «''& '■'• ■ '

-,S2 •*; 2 _. J3

* J ° m o ~

c oj ro ° *"

fi- E _£ '£ a £

E <2 jo .» £ 2

|jglB.I«g

< . + 3 > a > c u Q .

= 1

Ol i+-

co o

E " x c

- C O J |

.£ a- a E

5 ro o «-

O c ( J <

1 C U t -

9 1 3 — ; c

CT £• 3 o

.E < O u_

co c Q c

8 o c o

9 1 O T ^ O T

Q . O C C J

H- CD CU CD

- a . E d .

OT 4-

o 22

*=: E x

l* 5 -2r o i a

Jft < E

£ _ o p

a ro o E

X O J _ o

LU cc £ u.

ﬁ

3 E g

o c j |

C J r o o ,

f °<

2 CU X

O §> Q.

1 o3 +-■

z 2

C O J

E c

O O J

T J C O

r o c u

O C j Q

OT « ^_.

* •_ J > g OJ

« = C O O J U ; O J 0

T O > > C X Q .

■M C O C O O J c u _ 5 *

i ; « 4 - t » _ > * - 3

O O c o O O o t

60497800 A 8-5

.0 )

' O Q

_ x x x <

£, uji ST uj 2:

STrJ X -J x

X * U J S : j * j

ii O ii O ii

N Q >- CJ oo

> xxx <

CD UJ CD. UJ (j

O * - O - J o

_j co _3 a. 3

< 2 Q ,2 o

ii 2 ii On

N Q >- O CQ

< LU 3

O Q,Q

CD UJ, 5

odo

-i 2} -J

II O ii,

00 Q LU

]-' '. LL

X UJ. s UJ Z

* —" — ' _ . _ J T Z

_=S°CO%CJ

CO 2 Q g n

ii O ii o q

*> a lu o o

- • H"_

8s°8g8

° 5 a il »

ii Q ii O o

XOtUUU

: i

"JlJJ

<QX

XUJZ

z ^ <

i-2 Q

ii 2 A

OQ Q Q

— d?u|

^uj75

S 2-55

w 2 Q

"2 A

m a q

' ui —

— S"i

< . 5c

xUjco

CO « O

o 2 Q

n O ii

CQ. Q Q

Q« e

H fa

O J o >

- _ P

C O 3 C

c u O O

O C Q O

0 J O J

_ 5 a

c o = £

0 J O o

O C D o

« 2

_ O

OC o

0 ) C D

re 2 E

0 J o o

O C Q ( J

0 ) 0 )

c o = E

O J o o

O C Q C J

« 2

QJ O

QC Q

0 J •

•J3

ro 2

C D O

oc a

';-' cu

15 3

C D O

oc a

■S 1

91 I

t o O J

c o = E

Q J O O

O C Q O

t o 9 1

_ 3 a

c o 3 E

C D O O

0 C Q O

« 2

a j O

0C Q

O J O J

r o 2 E

Q J O O

O C Q C J

C D 0 J

C O = E

C D O o

O C Q C J

o j O

O C Q

« 2

c u O

OC Q

0 J i

» 2

cu O

O C Q -

X) o >

51 Q . Q _

Q . X X

X u j u j

O l d C J

CD CD "'cb

O o o

-l-l _J

< a o

O O

CD CD

< Q

z z

_■ CO CO

c o Q o

CO CO

C O O o

o O C J

C J Q C J

X z

z < X ',.,

X z

£: CO

CO O

'" X *

X C O

co O

- c u

■2 i

C O C O _ _

C ' ^ ' .

•v

•a

— - — -

«o- <**

C M C M X

0 0 C O ^

£ >- £v/

C D C U

*C cn

o j < c t X

_ 2 £ O J <

C M C M C O

s v/ iv/v

O T O T

* * f c

C M C M C D

x £ x 5

I ? : x f c : 1 ^

< V/ x V/ V

* £ < " S T X >

o — o

o o

.JZ

4->

• E <

," co

*_t

C Q f a

'£<

i - _ .

3 ro

• P O J

co o

Z _ l

b o i

o o

O - J

o *£

• - C O

+-• ._

oi o>

'i- .E

r- CO

E c

c ' g _

O c r5

o> „> •-

r - O 2

•9 c

_ S,

1 =

1 1 =

—. 0 > •

q. a>

> . E

X CO

•e »

c u c

> o

X O

CO 91

OJ CO

ECO

cx:

i->

OT c

_.3CO

■M

0J a.

_•

aro

XO)

OCO

^OT

u. _,

M- 4->

a> QJ

_. l_

XCD

_.o-J

,~3

—_fe CO

fe fc:

c91

■f

r■M

O T §

?* u.

CO f"

cCO

CO O)

i- CO

F4-

(1>

CD CD

/£^5V

/C^V

^ci«^v

CJ CJ

8-6 60497800 F

(rfCS^w

0$£\

>—i

—

1- UL CJ

LL CO

OC OC \-

sUJ LU CO H-

xII II II II II II

fa >- >-

<~ S

o .2

•«— ro CO ro CO ro CO

4 J O OJ OJ CD CD CD CD

OC OC OC QC OC OC

H fa

■»—

«- s

° s ro CO CO ro ro ro

OJ a> CD CD CD CD

a, 3, CC cc CC OC OC OC

>. s

H <

.o 4—

"o CJ

x> <u LL LL CO

E | rr OC

& z LU LU CO CJ 1-

o ■*—

^ s

x> E ,_ ,- —. X- *- T - '

§ s,

£X

i-i

-rj *->

oCN

t - »

T J C O

C M C M

r-f

<*4*

•43 < * C D >°

C M | ^

* - > © X? r- CM CM CM

ta 8. oj in"

<—^<CM

CM|^^

r 5 V

a <C O —

< V

•E <

t o —

< v

i / > —

O <

< v

c —

ro <.

?• <=

■aC O O

X B

*zt +=

c o

1 i _! -

o o>

■f

+-<

E .E 9 3 C

E z

u .2

■i— si CD LL

9 r -

E 2

g c JS

o S

O ) J D D )

O O T

C O J 9 3

O c o >

O ) ' J ﬂ O )

g c 3

0 S, o)

0 1 c O )

c 3 3 t 3 o £ > _ * Z c Q J ~ O 0 > " _ . r o Q >

oa fa LU LL O LU X < I-WQ r- CJ Q 1- 1- Q

I-

60497800 F 8-6.1/8-6.2

/*%

igKN

/f"^\

S § 8 g 8

co 2 Q ^ ii

i i O t i o c j

>- Q UJ O CJ Y=ATAN(X) DOUBLE D,E E=DATAN(D) B=ATAN2(A1,A2) DOUBLE D,D1,D2 D=DATAN2(D1,D2) DOUBLE DM.D1.D2 DM=DM0D(D1,D2) COMPLEX C CM=CABS(C)

couj o

O J o

oca <-

<2 o

" O H

X Q L U

—.UJQ

£.o"z

E|_co

ﬂ O I'

X O U I X=TAN(Y) DOUBLE D,E E = DTAN(D)

o .2

*->

C D U

a c

> 3

1- u-

C D O J

i 3 E

Q J O O

O C Q O

Q J Q J

_ - C J _ X I

C O 3 C O " 3

Q j O O J O

O C Q O C Q

_ J _

c o 3

CD O

Q C Q

m 2

C D O

oc a

« 2

cu O

OC O

Type of Argument

C D O J

_ 5 a

C O 3 E

O J O o

O C Q O

Q J O J

_ - Q _ S 3

r o 3 r o " 3

0 ) 0 Q J O

OC Q oc a

« 2

Q J O

oc a

.■ .OJ ■■:'

: xi '■

« 2

o j O

O C Q

CD -

15 2

cu O

oc a

S 3 0 )

C O Z

•*—

1 - H -

h- on oc

oc a o

a c o c o

c o a o

4—

•*- ±~

4 — 4 — 4 —

. $ _ 4 — 4 — C M

4 — Z C M Z

z < z <

< b < b

r - < 1 - <

< q < a

aas

•Up MP

MP CO

co O

o <

< a

ajjtMﬂ

MP MP

M P Z

..co

co <

< O

z <

< t

h* o

Number of Arguments

t - t - C M C M CM t—• ■-

I"j

. 1 1

l - t - " o - — ■

8 $ S _ J

S c 2 3 §

£ A \

,-• cm

< < <

c c "

c o c o ^

t - i t - t «

O C J ^

i - —

C O C O

^ +

S3 CO

4- I I

CN <£

OTO <

o _

■ i_

.E < .

OT —_

I s -

Basic External Function

co oc

I-

'ot

—.<

| _ OT

E C D C

O OJ CO

— co T2

. < u

CM OT

zx:

<+-•

r- •i

fc:

4- OJ

II ■o,__,

OJ fcs|CM

3 C

O) Ncm

E —

t=|(M

. fc:

fc:|CM_.

■*-»OJ

c u x

1 _ t - i

t n C D

_ o >

5 3 a >

- = c

♦■" ._

O O T

v _•

_ »

o ' S

- ; » . 2

^ OJ '—■

=! X

« > _ ■

Q J c o

O >

s £

5 3

f c : _ ! X

- 5 „

. <

O MP

" MP

J, CU OT §

«* S

O ) r o

C U £ «

_§ QJ *-

Q. ■£ E

■S C «"

CO — CU

• C O T 3

£35

."■ ro _

- > £

£ c =

x : r - f

i-t bz to

c 5 -

- _ C M

OT —' 1^

QJ _ 2.

3 Z <

"co < I-

3 T J

!- ro cm

oc z z

a < <

co 1- \-

o < <

Ex.

»- ro

"x ro

1 * O T

O c o

3.2

^~ 3

<-o

CO Xl^

• « co'

fc: —

+ 8

* s

cd Q

ro 0

CO CJ

3 C U

■♦-, x;

§ O

E "

O J O )

ro ro

c cu

0 x

cj *-

_ C D

I 2

5 E

.E £

*« CO

T J O

V t °

3 C O

OT £ CO

• 8

60497800 G 8-7

y*^^\

^^K

When a function subprogram is deﬁned with the same name as that of a basic external function, the user

deﬁnition overrides the library deﬁnition only if, in the calling program unit, the name of the function appears

either in an EXTERNAL statement or in an explicit type statement that overrides the type associated with

the library function, or if option T, D, or OPT=0 is speciﬁed on the FTN control statement.

Table 8-2 lists the basic external functions.

/-s®\

Arguments for which a result is not mathematically deﬁned, or those of a type other than that speciﬁed, "^

should not be used. Arguments of the trigonometric functions SIN, COS, and TAN are in radians; those of

SIND, COSD, and TAND are in degrees. The inverse trigonometric functions return principal values in radians.

If the name of the function appears either in an EXTERNAL statement or in an explicit type statement that

overrides the type associated with the library function, or if option T, D, or OPT=0 is speciﬁed on the FTN

control statement, the arguments of all external functions are checked to ensure that they are neither indeﬁnite <^*\

nor inﬁnite and fall within the limits listed in the Deﬁnition column of table 8-1. Argument checking is pro

vided unconditionally for all single and double precision math functions except DSIN, DCOS, DLOG, and

DLOG10. An informative diagnostic is provided when an argument is found to be invalid. /«"%

MISCELLANEOUS UTILITY SUBPROGRAMS ^

The utility subprograms described below are supplied by the system and are always called by name (section >^

17 deﬁnes call by name). A user-supplied subprogram with the same name as a library subprogram overrides

the library subprogram. Other utility routines, such as the mass storage routines, CYBER Record Manager

interface routines, Sort/Merge interface routines and Post Mortem Dump routines are described later in this section.

In the deﬁnitions listed under the routines:

i and n are integer variables, constants, or expressions.

j is an integer variable. ^

a and b are variable or array names of any type.

u is a unit designator (as deﬁned in section 5). <

H is a Hollerith speciﬁcation. /!^_

RANDOM NUMBER GENERATOR ^

RANF (n)t

Random number generator. Returns values uniformly distributed over the range (0,1); the value 0 and 1 are .'

excluded, n is a dummy argument which is ignored. Result is type real.

/^"§s

'RANF is an intrinsic function. /!y

8-8 60497800 E

/t*9$t

CALL RANSET(n)

Initializes seed of RANF. n is a one-word bit pattern. Bit 0 will be set to 1 (forced odd), and bits 59

through 48 will be set to 1717 octal.

' CALL RANGET(n)

/pas Obtains current seed of RANF between 0 and 1. n is a symbolic name to receive the seed. It is not neces

sarily normalized. The value returned may be passed to RANSET at a later time to regenerate the same

sequence of random numbers.

OPERATING SYSTEM INTERFACE ROUTINES

f DATE (a) or CALL DATE(a)^

rThe current date is returned as the value of argument a or of the function in the form lOHbmm/dd/yyb (under

NOS/BE 1, SCOPE 2) or lOHbyy/mm/dd. (under NOS 1), where B denotes a blank, mm is the number of the

month, dd is the number of the day within the month, and yy is the year. The value returned is Hollerith data

^p^ and can be output using an A format speciﬁcation.'*

The default type of the function DATE is real; thus if J and K are integer variables as in:

J = DATE(K>

J will not be useful because the value returned will have been converted from real to integer.

JDATE(a) or CALL JDATE(a)+ *

SN. The current date is returned as the value of argument a or of the function in the form SRyyddd, where yy

is the year and ddd is the number of the day within the year. The value returned is Hollerith data and can

be output using an R format speciﬁcation. The type of the function JDATE is integer.

JS~*N

SECOND(t) or CALL SECOND(t)^

The central processor time is returned from start-of-job in seconds as a real number, usually accurate to two

decimal places, t is a real variable.

Example:

DPTIM = SECOND (CP)

'These routines can be used as functions or subroutines. The value is returned via the argument and the

normal function return.

'* Not available under SCOPE 2.

''The date format can be changed by the installation.

60497800 D 8-9

/^^v

Example:

CALL DISPLA (7H TIME = , STOP-START)

CALL REMARK (H)

Places a message in the dayﬁle. Under SCOPE 2, the maximum message length is 90 characters displayed on one line.

Under NOS/BE 1, the maximum message length is 80 characters displayed 40 characters per line. Under NOS 1, the

message length is one line of 30 characters. A message exceeding the maximum length is truncated. A message

shorter than the maximum must have all zeros in the lower 12 bits of the last word. These zeros are automatically

supplied when a Hollerith constant is used as the parameter. Characters with display code greater than 57 octal are

listed in the dayﬁle, but they are replaced by blanks when displayed at the operator's console. If the ﬁrst character

is $, the message will ﬂash at the console, except under NOS 1, which allows ﬂashing messages only for system

origin jobs.

CALL SLITET(i,j)

'These routines can be used as functions or subroutines. The value is returned via the argument and the

normal function return.

8-10 60497800 E

/^^\

TIME(a) or CALL TIME(a)t

CLOCK(a) or CALL CLOCK(a)t

The current reading of the system clock is returned as the value of argument a or of the function in the form

10Hbhh.mm.ss., where b denotes a blank, and hh, mm, and ss are the number of hours, minutes, and seconds, <&®\

respectively. The value returned is Hollerith data and can be output using an A format speciﬁcation.

The default type of the functions TIME and CLOCK is real; thus if J and K are integer variables in the following ^^.

statement, J is not useful because the value returned will have been converted from real to integer. }

Example:

J = TIME(K)

CALL DISPLA (H,k) ^

A name and a value are placed in the dayﬁle. H is a Hollerith speciﬁcation of not more than 50 characters; k is

a real or integer variable or expression and is displayed as an integer or real value. Characters with display code -«^n

greater than 57 octal are replaced by blanks when displayed at the operator's console. If the ﬁrst character is $,

the message will ﬂash at the console except under NOS 1, which allows ﬂashing messages only for system origin jobs.

/s^s

/^^\

Example:

CALL REMARK (9HLAST DECK)

CALL SLITE(i) "^

Sense light i is turned on. If i = 0, all sense lights are turned off. If i is other than 0 through 6, an

informative diagnostic is printed and sense lights are not changed. '"^

,fr*$^EV

Sense light i is tested. If sense light i is on, j = 1; if sense light i is off, j = 2. If i is other than 1-6, an infor

mative diagnostic is printed, all sense lights remain unchanged, and j = 2. Execution turns off sense light i if

i t i s o n . " ^

^■^N

/J-*N

0^S

(Note: Logical variables generally provide a more efﬁcient method of testing a condition than do calls to

SLITE or SLITET.)

CALL SSWTCH(ij)

If sense switch i is on, j is set to 1; if sense switch i is off, j is set to 2. i is 1 to 6. If i is out of range,

an informative diagnostic is printed, and j is set to 2. The sense switches are set or reset by the computer

operator or by the control statements SWITCH (NOS and NOS/BE), ONSW (NOS only), and OFFSW

(NOS only).

CALL OVERLAY(fname, primary, secondary, recall,k)

See section 7.

CALL EXIT

Program execution is terminated and control is returned to the operating system. (Note: use of the STOP

statement is preferable to CALL EXIT.)

CALL CHEKPTX(ﬁlelist,SP)

A checkpoint dump of the ﬁles speciﬁed is taken. The arguments are:

ﬁlelist Array in the following format:

/*f^\

Word 1

Word 2

Word 3

Word n + 1 lfn.

1 7 1 1

noooo

lfn1 f 1 *

lfn2 f 2 *

' n *

FDoes not apply to SCOPE 2.

60497800 J 8-11

lfn.

Example:

Flag indicating whether or not all ﬁles assigned to the job are to be checkpointed. If SP is

zero, all local ﬁles are checkpointed. If SP is nonzero, the ﬁles speciﬁed by ﬁlelist are

checkpointed.

Number of ﬁles in following list, to a maximum of 42.

Name (in left-justiﬁed display code) of user mass storage ﬁles to be processed.

Number indicating speciﬁc manner in which lfn is to be processed.

0 Mass storage ﬁle is copied from beginning of information to its position at

checkpoint time, and only that portion will be available at restart. The ﬁle

is positioned at the latter point.

1 Mass storage ﬁle is copied from its position at checkpoint time to end of

information, and only that portion will be available at restart. The ﬁle is

positioned at the former point.

2 Mass storage ﬁle is copied from beginning of information to end of informa

tion; the entire ﬁle will be available at restart time. The ﬁle is positioned at

the point at which the checkpoint was taken.

3 The last operation on the ﬁle determines how the mass storage ﬁle is copied.

DIMENSION IFILES(4)

IFILESO) = 30000B

IFILES(2) = 5LTAPE1 .OR. 10000B

IFILESO) = 5LTAPE2 .OR. 30000B

IFILES(4) = 5LTAPE3

CALL CHEKPTX(IFILES,1)

/■C-JP^V

j^$$\

The names deﬁned in the array passed to CHEKPTX must be the actual ﬁle names used at run time.

For more information, refer to the operating system reference manual checkpoint/restart discussions.

CALL RECOVR(name,ﬂags,checksum,addr1,addr2)

**^i^?\

name

ﬂags

Name of subroutine to be executed if ﬂagged conditions occur (must be speciﬁed in an

EXTERNAL statement).

Octal value for conditions under which recovery code is to be executed, as outlined below.

Conditions can be combined as desired, with octal values up to 377 allowed.

001 Arithmetic mode error.

002 PP call or auto-recall error.

/iSS\

:Not available under SCOPE 2.

8-12 60497800 J

/$p°^\

/ ^ \

004 Time or storage limit exceeded.

010 Operator drop, kill, or rerun.

020 System abort.

040 CP abort.

100 Normal termination.

200 Terminal interrupt.

checksum Last word address of recovery code to be checksummed; 0 if no checksum is desired; a

negative number if the address parameters follow, specifying the area where checksumming

is desired.

addrl First word address of recovery code to be checksummed (used only when checksum is a

negative number).

addr2 Last word address of recovery code to be checksummed (used only when checksum is a

negative number).

The RECOVR subroutine allows a user program to gain control at the time that normal or abnormal job

termination procedures would otherwise occur. Initialization of RECOVR at the beginning of a program

establishes the conditions under which control is to be regained and speciﬁes the address of user recovery

code. If the stated condition occurs during program execution, control returns to the user code. If necessary,

the system increases the CP time limit, input/output time limit, or mass storage limit to provide an installation

#^ deﬁned minimum of time and mass storage for RECOVR processing. No limit is increased more than once in

a job. RECOVR can be called more than once during program initialization to reference different user recovery

subroutines. These calls to RECOVR can use different combinations of conditions for the same or different

user recovery subroutines.

No more than ten routines can be speciﬁed by RECOVR in one program. If an error occurs and more than one

J^n routine has been established for that error, the routines are called successively, with the routine most recently

speciﬁed called ﬁrst.

jsn The second speciﬁcation of a subroutine overrides its previous parameters. This override can be used to remove a

subroutine from the RECOVR list by passing a mask of zero.

jPN A checksum of the user recovery code can be requested during initialization. If ﬂagged conditions subsequently

occur, RECOVR again checksums the code before returning control to it. This gives some assurance of user code

integrity before it is executed.

■ If the checksum parameter is zero, no checksum is done.

/pn If one of the user's selected error conditions occurs, RECOVR gains control, performs internal tasks, and then

v transfers control to the user's recovery subroutines. The following three arguments are passed to the user's

recovery subroutine:

/*ﬁ"*\

1. A 26-word integer array. The ﬁrst 16 words are an image of the exchange package; the seventeenth

word is the contents of RA+1. The ﬁrst word of the exchange package contains the value of B0;

/sp5\ bits 0 through 17 of B0 contain the error ﬂag. The 18th word contains a zero value unless it is

v changed by user's recovery subroutine (discussed later in this section). The 19th through 26th words

contain a copy of the 1st through 9th words of the reprieve parameter block. Refer to the appro-

jiSN priate operating system reference manual for detailed information.

/ | E \

60497800 J 8-13

If the ﬂags are not set by any of the recovery subroutines, the job continues as if RECOVR had not been

called; that is, the original system error code is reset and processed.

In an overlay structured program, calls to RECOVR as well as the user recovery subprograms should be in the

(0,0) overlay.

For further information about RECOVR, refer to the appropriate operating system reference manual.

Example:

PROGRAM MAlN(INPUT,OUTPUT)

EXTERNAL REPREV.CHKSUM

CALL RECOVR(REPREV,72B,LOCF(CHKSUM))

STOP

END

SUBROUTINE REPREV(IXCHNG.IFLAG.IFLDLN)

DIMENSION IXCHNG(17), IFLDLNHOOOOB)

IFLAG = 1

PRINT 10, IXCHNG, (IFLDLN(I), 1=1,64)

10 FORMAT (3(6X, O20)>

RETURN

entry chksum <« determines end of code to be checksummed

END

DEBUGGING AIDS

A number of calls and functions useful in debugging are described here. Many users ﬁnd CYBER Interactive

Debug and/or Post Mortem Dump more useful. They are described near the end of this section.

8-14 60497800 J

#*%$\

2. A ﬂag that, upon return, determines the type of program termination. This ﬂag is initially set to

zero. The possible types of program termination are discussed later in this section. //SS!^

3. An array, starting at RA+1, that allows a FORTRAN subroutine to access all of the user's ﬁeld

length. ^

If the recovery subroutine was called because of normal termination, the subroutine, before returning, should

ﬂush the buffers of all output ﬁles. Buffers can be ﬂushed by an ENDFILE or REWIND statement. ^

Upon return from a user's recovery subroutine, the 18th word of the reprieve package and the argument ﬂag

contain values set by the user's recovery subroutine. If the 18th word of the reprieve package is set to 4, '^

the program immediately resumes normal execution, and the argument ﬂag is ignored. If the 18th word is not

set to 4, the argument ﬂag, which is set by the user's recovery subroutine, is checked. If the ﬂag is set to

nonzero, the job terminates normally, as if no errors had occurred. This happens after all the recovery sub- '^

routines have been processed. If the ﬂag remains zero for all the recovery subroutines, the job continues as

if RECOVR had not been called; that is, the original system error code is reset and processed.

/fS^K

0^^\

0£~$?K

ygSN

CALL DUMP (a1,b1,f1 ^M

CALL PDUMP (a1,b1,f1 a^'V

Dumps central memory on the OUTPUT ﬁle in the indicated format. PDUMP returns control to the calling

program; DUMP terminates program execution, aj and bj specify the beginning and the end of the storage

area to be dumped. 1 < n < 20. f is a format indicator, as follows:

f = 0 or 3 octal dump

f = 1 r e a l d u m p

f = 2 i n t e g e r d u m p

For f values 0 through 3, aj and bj are the ﬁrst and last words dumped. If 4 is added to any f value, the

values of aj and bj are used as the addresses of the ﬁrst and last words dumped within the job's ﬁeld length.

An ASSIGN statement or the LOCF function can be used to get addresses for the a: and bj parameters.

Examples:

CALL PDUMP(A(D.A(100), 1) Dumps from A(l) to A(100) as real numbers

CALL PDUMP (0, 1000B.4) Dumps from location 0 to 1000B in octal

CALL STRACE

Provides traceback information from the subroutine calling STRACE back to the main program. Traceback

information is written to the ﬁle DEBUG. To obtain traceback information interspersed with the source

program, DEBUG should be equivalenced to OUTPUT in the PROGRAM statement. (Refer to STRACE,

section 9).

LEG VAR (a)

Checks the value of variable a. Returns the result -1 if variable is indeﬁnite, +1 if out of range, and 0

otherwise. Variable a is type real; result is type integer.

CALL SYSTEM(errnum,rnesg)

errnum Error number. An integer value from 0 to 9999 decimal. Error numbers used by the

compiler (listed in appendix B) retain the severity associated with them. Error numbers

51 (non-fatal) and 52 (fatal) are reserved for the user. If an error number greater than

the highest number deﬁned in appendix B is speciﬁed, 52 is substituted. If errnum is

negative, SYSTEM returns immediately with no action taken.

mesg Error message; entered as a Hollerith constant with the ﬁrst character used as a carriage

control character and not printed.

The subroutine SYSTEM enables the user to issue an execution-time error message.

If error number zero is entered, the message is ignored, the output buffers are ﬂushed, and control is

returned to the calling program.

60497800 J 8-15

/<^\

The ﬁle OUTPUT should be declared before SYSTEM is called. Otherwise, no errors are printed; and a message

to this effect is entered in the dayﬁle.

Each line is printed unless the line limit of the OUTPUT buffer is exceeded, in which case the job is terminated.

Example:

CALL SYSTEM (3. ■?= CHECK DATA # )

CALL SYSTEMC (errnum.speclist)

errnum Error number for which non-standard recovery is to be implemented. Error numbers are

listed in table B-5 of appendix B.

speclist Integer array containing error processing speciﬁcations in consecutive locations:

word 1 F/NF (1 = fatal, 0 = non-fatal)

word 2 Print frequency

word 3 Frequency increment

word 4 Print limit

word 5 User-speciﬁed error recovery routine address

word 6 Maximum traceback limit applicable to all errors; this limit is 20 unless

changed by a call to SYSTEMC

SYSTEMC enables the user to replace the standard fatal/non-fatal error table with an enhanced error table,

which contains speciﬁcations that regulate error processing. The error table, whose format is described below,

is ignored for erroneous data input from a connected (terminal) ﬁle.

In an overlay program, if SYSTEMC is not called in the (0,0) overlay, the routine might not be available to

higher level overlays.

In the error table, the ﬁrst entry corresponds to error number 1, the second to error number 2, and so on.

Each entry has the following format:

X^\

y£I~£V

y^^sv

59 51 43 31 2 0 1 7

frequency

increment

print frequency

frequency increment

print limit detection

total NN user-speciﬁed

recovery address

By default, print frequency value is 0. If the value is changed to n by a call

to SYSTEMC, diagnostic and traceback information is listed every nth time until

the print limit is reached.

By default, frequency increment value is 1. This speciﬁcation can be changed

by a call to SYSTEMC if the call speciﬁes print frequency as 0. When

frequency increment is 0, diagnostic and traceback information is not listed;

when it is 1, such information is listed until the print limit is reached; when

the frequency increment is n>l, such information is listed only the ﬁrst n

times unless the print limit is reached ﬁrst.

jr-CE&SV

>+^$\

I 8-16 60497800 J

/eSfﬂ.

print limit By default, print limit value is 10. It can be changed by a call to

SYSTEMC.

detection total Detection total is a running count of the number of times an error occurs.

0®^ The final value is reported in the error summary issued at end of job if

SYSTEMC is called during execution.

F/NF This bit speciﬁes the severity of the error: 1 indicates a fatal error; 0, non

fatal. The severities of system deﬁned errors are given in appendix B. All

errors deﬁned by the user with these numbers in a call to SYSTEM retain

0^- the speciﬁed severity. The severity of any error can be changed by a call

to SYSTEMC, however.

vjfpN

0$^\

60497800 J 8-16.1/8-16.2 |

./*%

0^\

rA/NA The A/NA bit is ignored unless a non-standard recovery address is speciﬁed;

it can be set only during assembly of SYSTEMC. When this bit is set, the

address in an auxiliary table is passed in the third word of the secondary

r argument list to the recovery routine. Each word in the auxiliary table must

have the error number in its upper 10 bits, so that the address of the ﬁrst

error number match is passed to the recovery routine. An entry in the aux-

^sn iliary table for an error is not limited to any speciﬁc number of words.

user-speciﬁed This address is speciﬁed in a call to SYSTEMC.

0m*. recovery address

A negative value for any word in the speclist indicates that the current value of that speciﬁcation is not to be

r changed. A user-speciﬁed error recovery routine activated by a call to SYSTEMC can be canceled by a sub

sequent call with word 5 of the speclist set to zero.

rlf SYSTEMC has been called, an error summary is issued at job termination indicating the number of times

each error occurred since the ﬁrst call to SYSTEMC.

^^ For an error detected by a routine in the math library, a user-supplied error recovery routine should be a

f function subprogram of the same type as the FORTRAN function detecting the error. For any other error,

a user-supplied error recovery should be a subroutine subprogram.

0!®\

/!p"N

/ ^ \

i # ^ N

The error recovery subprogram must not invoke any code during recovery that was also involved with the

error being handled. In particular, if an input/output related error is being processed, no input/output

operations of any type should be done, unless the recovery routine intends to terminate execution.

When SYSTEMC is called from an overlay or segment, it must reside in the (0,0) overlay or the root segment.

When an error previously referenced by a SYSTEMC call is detected, the following sequence of operations is

initiated:

1. Diagnostic and traceback information is printed in accordance with the speciﬁcation in the pertinent

error table entry. The traceback information is terminated for any of the following conditions:

Calling routine is a program

Maximum traceback limit is reached.

No traceback information is supplied.

2. If the SYSTEMC call references a user-speciﬁed error recovery routine address, SYSTEMC,

FORSYS=, and the routine detecting the error are delinked from the calling chain, and the

user-supplied error recovery routine is entered.

3. If the error is non-fatal, control returns to the routine that called the routine detecting the error.

An error summary is printed at job termination.

4. If the error is fatal, all output buffers are ﬂushed, an error summary is printed, and the job is

terminated.

60497800 H 8-17

If a non-standard recovery address is speciﬁed in the SYSTEMC call, the following information is available to

the user recovery routine:

Al Address of argument list passed to routine detecting the error for errors detected by a

math library routine.

Address of the FIT for error 103.

Undeﬁned for all other errors.

XI Address of the ﬁrst argument in the list for errors detected by a math library routine.

Undeﬁned for all other errors.

AO Address of argument list of routine that called the routine detecting the error.

Bl Address of a secondary argument list containing, in successive words:

Error number associated with this error.

Address of message associated with this error.

Address within auxiliary table if A/NA bit set; otherwise 0.

In upper 30 bits, instruction consisting of RJ to SYSERR.j; in lower 30 bits, address of

traceback information for routine detecting the error.

Information in the secondary argument list is not available to user supplied error recovery

routines coded in FORTRAN.

A2 Address of error table entry for this error.

X2 Contents of error table entry for this error.

Example 1:

PROGRAM EXPECT(OUTPUT)

DIMENSION IRAY(6)

DATA IRAY /6 * (-0)/

C SET PRINT LIMIT TO ZERO

IRAY(4)=0

X = EXP(800.0>

X = EXP(-800.0)

C CALL SYSTEMC TO INHIBIT PRINTING OF ERROR 115

C. AND START ERROR SUMMARY ACCUMULATION

CALL SYSTEMC (115.IRAY)

PRINT *. jtjt

PRINT *,******SYSTEMC IS CALLED TO SUPPRESS PRINTING*,

+ j * O F E R R O R 1 1 5 *

X = EXP(800.0)

X = EXP'-SOO.O)

8-18 60497800 C

y^Ss

/*~Is

/^ZlK

^**\

••-®*N

Jﬂ~*\

/*J"©\

J^fei&S.

PRINT *,# *

PRINT *.****"ERROR 115 DETECTED BUT NOT PRINTED*

END

ARGUMENT TOO LARGE, FLOATING OVERFLOW

ERROR NUMBER 30 DETECTED BY EXP

ARGUMENT TOO SMALL

ERROR NUMBER 115 DETECTED BY EXP

•••••SYSTEMC IS CALLED TO SUPPRESS PRINTING OF ERROR 115

ARGUMENT TOO LARGE, FLOATING OVERFLOW

ERROR NUMBER 30 DETECTED BY EXP

•••••ERROR 115 DETECTED BUT NOT PRINTED

ERROR SUMMARY

ERROR TIMES

0030 0001

0115 0001

Program EXPECT illustrates a standard error recovery in a math library routine and how to suppress the print

ing of error message 115.

Example 2:

PROGRAM FXAMPL(TAPEUOUTPUT)

EXTERNAL ITSOK

DIMENSION NARRAY(6)

DATA NApqAY/6<M-1)/

NARRAY(l) * 0

NARRAY<5) = LOCF(ITSOK)

NARRAY(6) ~ 1

CALL SYSTEMC<66«NARRAY)

NAMELIST/DATAI/A-jB

READ (It DATA!)

REWIND 1

NAMELIST/DATA2/AtB

READ (!• DATA2)

NAMELIST/DATAOUT/AtB

PRINT DATAOUT

STOP

END

SUBROUTINE ITSOK

PRINT 10

10 FORMAT (oODATA SET NAMED ABOVE NOT USED»)

RETURN

END

Input:

»DATA2

A - 3 . - *

B = 4 . »

60497800 C 8-19

j^SN

Output:

NAMELIST NAME NOT FOUND - DATA!

ERROR NUMBER 0066 DETECTED BY NAMIN« AT ADDRESS 000435

DATA SET NAMED ABOVE NOT USED

0*~$\

SDATAOUT

A " .3E*01»

B - .4E+01t

SEND

ERROR SUMMARY ^

ERROR TIMES

0066 0001

CALL LIMERR(lim) ^

lim Integer value; the program does not terminate when data errors are encountered until the

number of errors occurring after the call exceeds the value of lim.

NUMERR(n) A function that returns the number of errors since the last LIMERR call. Result type is

integer, n is a dummy argument which is ignored.

The subroutine LIMERR and function NUMERR enable the user to input data without the risk of termination

when improper data is encountered.

LIMERR can be used to inhibit job termination when data is being input with a formatted, NAMELIST, or list

directed read, or with DECODE statements. It operates only when data is encountered that would ordinarily

cause job termination under error number 78 ("ILLEGAL DATA IN FIELD") or error number 79 ("DATA

OVERFLOW"). LIMERR has no effect on the processing of errors in data input from a connected (terminal) ﬁle.

LIMERR initializes an error count and speciﬁes a maximum limit (lim) on the number of data errors

allowed before termination. LIMERR continues in effect for all subsequent READ statements until the limit

is reached. LIMERR can be reactivated with another call, which will reinitialize the error count location and

reset the limit. A LIMERR call with lim speciﬁed as zero nulliﬁes a previous call; improper data will then

result in job termination as usual.

When improper data is encountered in a formatted or NAMELIST read or in a DECODE statement with LIMERR

in effect, the bad data ﬁeld is bypassed, and processing continues at the next ﬁeld. When improper data is

encountered in a list directed read, control moves to the statement immediately following the READ statement. "^)

NUMERR returns the number of errors since the last LIMERR call. The previous error count is lost when

LIMERR is called, and the count is reinitialized to zero.

8-20 60497800 G

Example:

00?\

The following example illustrates the use of LIMERR and NUMERR to suppress normal fatal termination

when large sets of data are being processed.

/"~N

CALL LIMERR (200)

READ(1,125)(ARAY(I),I=1,1500)

125 FORMAT (3F10.5,E10.1)

IF (NUMERR(O).GT.O) GO TO 500

500 CALL LIMERR(200)

READd,125)(BRAY(I),I=1,1500)

IF (NUMERR(O).GT.O) GO TO 600

/"~J\

600 CALL LIMERR(100)

READ(1,230)(LRAY(I),I=1,500)

PRINT 99, NUMERR(O)

READ(4,127)(MRAY(I),I=1,500)

PRINT 99, NUMERR(O)

READ(4,225)(NRAY(I),I=1,50)

IF (NUMERR(O).GT.O) GO TO 700

700 STOP

END

When LIMERR is called, a limit of 200 errors is established. The number of errors is reset to zero. After

ARAY is read, NUMERR(O) is checked. If errors occur, the following statements are not processed and a

branch is made to statement 500. Had LIMERR not been called, fatal errors would have terminated the pro-

rgram before the branch to statement 500. At statement 500, LIMERR once more initializes the error count,

and ftYftmitinn rnntinni»c

i^^V

\ 60497800 E 8-21

Example:

/^§\

program exahpluape1,output)

dimension acard(5)

oata acard /-i. ,-2.,-3. , -4. ,-5./

:all limerr(2)

read(1,10) (acard(i),1=1,5)

FORMAT (F4.1)

PRINT 2u, NUMERR(ﬂ)

FORMAT

STOP

END

MHu, II, ♦ DATA ERRORS

;G, (ACARDCI),I=i,5)

( IX, F4.1)

FOUND*//) /^3^\

Input:

4 7 . 1

25./

48.3

24,6

91.2

Output: 25./

...t 12345a739u1234567S9012345679901234567ft90

♦■ ERROR OATA INPUT ♦ ILLLGAL DATA IN FIELD

ERROR NUMBER 7ft DETECTED BY INCOM= AT

3ALLEO FROM KRAKER= AT ADDRESS ii0Q345

CALLED FROM INPC= . AT ADDRESS 300075

CALLED FROM EXAMPL AT .INE 5

4 2 4 , 6

ADDRESS 300215

i^^\

,♦••....•12345b739C123456789u12345678931234567890

• ERROR DATA INPUT ■

ERROR NUMBER 78

CALLED FROM KRAKER=

FROM INPC=

FROM EXAMPL

GALLED

CALLED

ILLEGAL DATA IN FIELD

DETECTED BY INCPM= AT

AT ADDRESS 000345

AT ADDRESS S0QG75

AT L I N E 5

*!•**

ADDRESS 000215

2 3ATA ERRORS FOUND

4 7 . 1

-2.Q

49.3

-4. ii

9 1 . 2

8-22 60497800 D

j|~Sy

„ I N P U T / O U T P U T S TAT U S C H E C K I N G

rFORTRAN Extended provides the capability of checking for an end-of-ﬁle or a parity error condition following

read operations via the functions UNIT, EOF, and IOCHEC

Any of the following conditions encountered during a read returns an end-of-ﬁle status via the functions UNIT

or EOF:

End of section (in the case of ﬁle INPUT only)

^pes End of Partition

End of information

Non-deleted W format ﬂag record

f Embedded tape mark

Terminating double tape mark

^ Terminating end-of-ﬁle label

Embedded zero length level 17 block

v The functions UNIT and IOCHEC return a parity error indication for every record within or spanning a block

containing a parity error; however, such an indication does not necessarily refer to the immediately preceding

^ operation because of the record blocking/deblocking performed by the Record Manager input/output routines,

8 Parity status can be checked on write operations that access mass storage ﬁles when the write check option

has been speciﬁed on the REQUEST statement for the ﬁle (SCOPE 2 Reference Manual). Write parity errors

for other types of devices (such as staged/on-line tape) are detected by the operating system, and a message

to this effect is written in the dayﬁle.

f^ UNIT(u,a.b)

The UNIT function is used to check the status of a BUFFER IN or BUFFER OUT operation for an end-of-

^ ﬁle or parity error condition on logical unit u. Parameters a and b are the ﬁrst and last variable or array

elements, respectively, of the block of memory speciﬁed in the preceding BUFFER IN or BUFFER OUT

statement. When UNIT is referenced, the user program does not regain control until input/output operations

f* on the unit are complete. The function returns the following values:

-1. Unit ready, no end-of-ﬁle or parity error encountered on the previous operation

+0. Unit ready, end-of-ﬁle encountered on the previous operation

+1. Unit ready, parity error encountered on the previous operation

Example:

BUFFER IN(5,1)(B(1),B{100))

IF (UNIT(5,B(1),B(100)) 12,14,16

'Applies only to SCOPE 2.

60497800 J 8-23

/t^\

Control transfers to the statement labeled 12, 14 or 16 if the value returned was -1., 0., or +1., respectively.

Parameters a and b should always be included. If these parameters are omitted (as in most older programs),

there is a slight chance that at OPT=2, execution errors will result or that the elements between a and b will

not contain the expected values. j

If 0. or +1. is returned, the condition indicator is cleared before control is returned to the program. UNIT

should only be called for a ﬁle processed by buffer statements. "^

EOF(u)

The EOF function is used to test for an end-of-ﬁle condition on unit u following a formatted, list-directed,

NAMELIST, or unformatted read. Zero is returned if no end-of-ﬁle is encountered, or a non-zero value if

end-of-ﬁle is encountered. '"i

Example:

IF (EOF(5)) 10.20

returns control to the statement labeled 10 if the previous read encountered an end-of-ﬁle; otherwise,

control goes to statement 20.

If an end-of-ﬁle is encountered, EOF clears the indicator before returning control.

The EOF function returns a zero value following read or write operations on random access ﬁles (ﬁles

accessed by READMS/WRITMS), and also following write operations on all types of ﬁles, regardless of

whether an end-of-ﬁle condition has been detected; therefore, the EOF function should not be used in

those circumstances.

The user should test for an end-of-ﬁle after each READ statement to avoid input errors. If an attempt

is made to read on unit u and an EOF was encountered on the previous read operation on ﬁle, execution

terminates and an error message is printed.

IOCHEC(u)

The IOCHEC function tests for parity error on unit u following a formatted, list-directed, NAMELIST, or

unformatted read. The value zero is returned if no error has been detected.

Example:

J = IOCHEC(6)

IF (J) 15,25

Zero value would be returned to J if no parity error occurred and non-zero if an error had occurred;

control would transfer to the statement labeled 25 or 15 respectively.

If a parity error occurs, IOCHEC clears the parity indicator before returning. Parity errors are handled in

this way regardless of the type of the external device.

/3§H

/£Ss\

8-24 60497800 J

OTHER INPUT/OUTPUT SUBPROGRAMS

LENGTH(u) or CALL LENGTHX(u,nw,ubc)

^ Returns information regarding the previous BUFFER IN or READMS call of the ﬁle designated by u. nw or

the value of LENGTH is set to the number of 60-bit words read, ubc is set to the number of unused bits in

the last word of the transfer, nw, ubc, and value returned are type integer.

After an unformatted BUFFER IN on a 9-track S or L tape, the unused bit count parameter of LENGTHX

is rounded down so as to indicate a whole number of 6-bit characters. For example, a BUFFER IN of a

l#^ record of 23 8-bit frames returns a length of four words with an unused bit count of 54, even though the

actual unused bit count is 56.

/^ If an odd number of words is written to a 9-track S or L tape by an unformatted BUFFER OUT, the record

on the tape contains four additional zero bits at the right so as to be a whole number of 8-bit characters. If

such a record is subsequently read by BUFFER IN, the length indication in LENGTH or LENGTHX is one

/pn word greater than the number of words originally written.

For a ﬁle accessed by buffer statements, LENGTH or LENGTHX should be called only after a call to UNIT

/pa* ensures that input/output activity is complete; otherwise, ﬁle integrity might be endangered.

Example:

NW = LENGTH(5>

CALL LENGTHX(5,NW,NUBC)

CALL LABEL(u,labinfo) *

u Logical unit number.

0$$$*\

labinfo Name of 4-word array containing label information in the format shown in table 8-3.

This subroutine passes label information to the operating system.

The control statement that requests the tape for the job must have speciﬁed that the tape has labels before the

CALL LABEL statement can be used.

CALL LABEL should not be used with ﬁles accessed with CYBER Record Manager interface routines.

/#^ ^Recognized but ignored under SCOPE 2.

60497800 H 8-25

TABLE 8-3. LABINFO BLOCK CONTENT

Word Bit

Positions

Character

Length

Contents Default for Output

1 59 Ihru 0 10 File identiﬁer None. Entire ﬁeld must be sup

plied by the user.

2 59 thru 18

17 thru 0

File identiﬁer continued

File sequence number

None. Entire ﬁeld must be sup

plied by the user.

001+

3 59 thru 48

47 thru 30

29 thru 0

Generation version number

Retention cycle

Creation date (YYDDD)

oot

000+

Today's Julian date.

4 59 thru 24

23 thru 0

Set identiﬁer (VSN)

File section number

Blanks (ﬁeld can be blank or

zero ﬁlled.

0001+

'Field must be display code zero ﬁlled.

../S^V

Default output values are speciﬁed for a ﬁeld by ﬁlling it with binary zeros. If any of the ﬁelds are improperly

initialized, NOS generates an LX BUFFER/FET PARAMETER ERROR and aborts the job on output. For

further information concerning the ﬁelds, consult the appropriate operating system reference manuals concerning

words 9-12 of the ﬁle environment table (FET).

The actions that occur at the time of the LABEL call are as follows:

Passing the address and length of the user label block to CYBER Record Manager

Forcing a rewind on any succeeding open of the logical unit

The LABEL reference occurs before any access is made to the logical unit. If a WRITE subsequently occurs on

the logical unit, the Record Manager routines write a label formatted from the LABINFO area onto the tape as

an HDR1 label. Or, when a close or endﬁle operation is performed at job termination, the routines write an

EOF1 label formatted from LABINFO. If a READ subsequently occurs on the logical unit, and a label is

encountered on the opening read or an EOF condition, the label information is reformatted and placed into

LABINFO.

/ ^ ^ V

/^^V

^^V

8-26 60497800 H

^*^\

ECS/LCM/UEM SUBPROGRAMS

0^\

CALL MOVLEV (a,b,n)

Transfers n consecutive words of data between a and b. a and b are variables or array elements; n

is an integer value, a is the starting address of the data to be moved and b is the starting address of the

receiving location.

Example:

(^ CALL MOVLEV(A,B,1000)

No conversion is done by MOVLEV. If data from a real variable is moved to an integer type receiving ﬁeld,

^ t h e d a t a r e m a i n s r e a l .

Example:

CALL MOVLEV (A, I, 1000)

0^ After the move, I does not contain the integer equivalent of A.

Example:

DOUBLE PRECISION D1(500), D2(500)

CALL MOVLEV (D1, D2, 1000)

Since DI is deﬁned as double precision, n should be set to 1000 to move the entire DI array.

"T|N

r CALL READEC(a,b,n)

/P-n Transfers data from extended core storage to central memory.

a is a simple variable or array element located in central memory, b is a simple variable or array element located

/^n in an extended core storage block or LCM block, n is an integer constant or expression, n consecutive words

of data are transferred beginning with a in central memory and b in extended core storage.

/*J~**

/*~*\

0^\

60497800 J 8-26.1/8-26.2

^%^

/«%

0^\

/jpN

0$£S

CALL WRITEC(a,b,n)

Transfers data from central memory to extended core storage or LCM.

No type conversion is done.

Example:

LEVEL 3,B

CALL READEC(A,B,10)

CALL WRITEC(A,B,10)

TERMINAL INTERFACE SUBPROGRAMS*

CALL CONNEC (u,cs)

u unit designator.

cs optional character set designator (applicable to NOS/BE only): cs is an integer with a value from 0

to 2, in accordance with the character set to be used for the data entered or displayed at the terminal:

0 display code (default)

1 ASCII-128

2 ASCII-256 code

If a FORTRAN program to be run under INTERCOM for NOS/BE, under the NOS Time-Sharing System, under

the NOS Interactive Facility, or under HELL07 for SCOPE 2, calls for input/output operations through the user's

remote terminal, all ﬁles to be accessed through the terminal must be formally associated with the terminal at the time

of execution.

In particular, the ﬁle INPUT must be connected to the terminal if data is to be entered there and an alternate logical

unit is not designated in the READ statement. The ﬁle OUTPUT must be connected to the terminal if execution diag

nostics are to be displayed or printed at the terminal, or if data is to be displayed or printed there and an alternate unit

is not designated in the WRITE or PRINT statement. These ﬁles are automatically connected to the terminal when the

program is executed under the NOS FTNTS subsystem or under the RUN command of the EDITOR utility of

INTERCOM.

Under HELL07, any ﬁle can be connected by providing a FILE control statement specifying CNF = YES.

Under INTERCOM, any ﬁle can be connected to the terminal by the CONNECT command.

Under all operating systems, the user can connect any ﬁle from within the program by using the CALL CONNEC

statement.

A ﬁle n is considered still connected if a CALL CONNEC(n) has been made by a program running at a terminal and if

the program terminates under normal or abnormal circumstances without a CALL DISCON(n). Any subsequent

input/output on n will still be through the terminal unless the ﬁle is returned.

Under NOS, if CONNEC speciﬁes an existing local ﬁle, the buffers for the ﬁle are ﬂushed (if it is an output ﬁle) and the

ﬁle is returned. A subsequent DISCON for the ﬁle causes the connected ﬁle to be returned, but the pre-existing ﬁle is

not reassociated with the ﬁle name.

#^ More information about INTERCOM is in the INTERCOM reference manual and the INTERCOM Interactive Guide

for Users of FORTRAN Extended. More information about NOS is in the Interactive Facility reference manual.

More information about HELL07 is in the SCOPE 2 reference manual.

60497800 H 8-27

If cs is not speciﬁed, it is set to 0. If display code is selected, input/output operations must be formatted, list-directed,

NAMELIST, or buffered. **)

If either of the ASCII codes is selected, input/output operations must be either formatted or buffered. When buffer

input/output is used, either a FILE control statement (section 16) specifying RT=S must be provided, or blanks cannot ^

terminate a line.

When a CALL CONNEC speciﬁes a ﬁle already connected with the character set speciﬁed, the call is ignored. If the ﬁle

speciﬁed is already connected with a character set other than that speciﬁed, cs is reset accordingly. "^

Data input or output through a terminal under INTERCOM is represented ordinarily in a CDC 64-character or ASCII

64-character set, depending on installation option. For these sets, ten characters in 6-bit display code are stored in .^

each central memory word. As described above, a terminal user can specify from within a FORTRAN program that

data represented in an ASCII 128-character set (providing the capability for recognizing lowercase letters and control

codes) or an ASCII 256-character set (providing the capability for recognizing lowercase letters, control codes, and >■*&

parity) be input or output through the terminal. For the ASCII 128-character and 256-character sets, characters are

stored in ﬁve 12-bit bytes in each central memory word. Characters in the ASCII 128-character set are represented in

7-bit ASCII code right justiﬁed in each byte with binary zero ﬁll; characters in the ASCII 256-character set are ^

represented in 8-bit ASCII code right justiﬁed in each byte with binary zero ﬁll. When data represented in either

ASCII character set code is transferred with a formatted input/output statement, the maximum record length should be

specified in the PROGRAM statement as twice the number of characters to be transferred (see section 7). Allowance /«■«&

should also be made in input/output operations for the fact that internal characters require twice as much space as

external characters.

If ASCII information needs to be read from or sent to a terminal a formatted READ or WRITE statement should be

used. This will cause trailing blanks to be truncated from the line.

If the exact length of the data input at the terminal is required, BUFFER IN and BUFFER OUT should be used.

LENGTHX can be used after a BUFFER IN to determine how many characters were entered. BUFFER IN and

BUFFER OUT ensure that trailing blanks are not added or deleted from the line. The ﬁle used for BUFFER IN and

BUFFER OUT must be of record type S. A

For a FORTRAN program run under NOS, any ﬁle can be connected to the terminal by the ASSIGN command.

In addition, the user can connect any ﬁle from within the program by using the statement: /=

CALL CONNEC (u)

Data input or output through a terminal under NOS in NORMAL mode is represented ordinarily in 6-bit display

code. However, the user can elect to have data represented in an ASCII 128-character set (which provides the

capability for recognizing control codes and lowercase, as well as uppercase, letters) by entering the ASCII com

mand to input in ASCII mode. Characters contained in the standard set are stored internally in 6-bit display code,

whether or not the ASCII command has been entered. The additional characters which complete the ASCII

128-character set are stored internally in 6/12-bit display code if the ASCII command has been entered.

Under any system, if a ﬁle speciﬁed in a CALL CONNEC exists as a local ﬁle but is not connected at the time of

the call, the ﬁle's buffer is ﬂushed before the ﬁle is connected to the terminal; under NOS, the ﬁle is returned.

/ d ^ \

8-28 60497800 H

CALL DISCON (u)

This subroutine disconnects a ﬁle from within a FORTRAN program.

0^\ This request is ignored if the speciﬁed ﬁle is not connected. After execution of this statement under NOS/BE 1,

the speciﬁed ﬁle remains local to the terminal. In addition, if the ﬁle existed prior to connection, the ﬁle name

is re-associated with the information contained on the device where the ﬁle resided prior to connection. Data

j~n written to a connected ﬁle is not contained in the ﬁle after it is disconnected. Under NOS 1, a CALL

* DISCON causes the connected ﬁle to be returned; the disconnected ﬁle name is not re-associated with the

pre-existing information.

j^N

All ﬁles to be connected or disconnected during program execution must be declared in the PROGRAM state

ment. An attempt to connect or disconnect an undeclared ﬁle results in a fatal diagnostic.

Calls to CONNEC and DISCON are recognized and ignored when programs are not executed under INTERCOM

or interactively under NOS 1.

Examples:

CALL CONNEC (6)

K = PLAGES

^ CALL CONNEC (K)

CALL COiMNEC (6,2)

^ CALL CONNEC <4LDATA,1)

CALL DISCON (6)

MASS STORAGE INPUT/OUTPUT

V Mass storage input/output (MSIO) subroutines allow the user to create, access, and modify ﬁles on a random basis

without regard for their physical positioning. Each record in the ﬁle can be read or written at random without

jpe^ logically affecting the remaining ﬁle contents. The length and content of each record are determined by the user.

A random ﬁle can reside on any mass storage device. Record Manager word addressable ﬁle organization is used to

implement MSIO ﬁles. The Record Manager reference manual contains details of word addressable implementation.

( A ﬁle processed by mass storage subroutines should not be processed by any other form of input/output.

f* RANDOM FILE ACCESS

Random ﬁle manipulations differ from conventional sequential ﬁle manipulations. In a sequential ﬁle, records

{^ are stored in the order in which they are written, and can normally be read back only in the same order.

This can be slow and inconvenient in applications where the order of writing and of retrieving records differ

and, in addition, it requires a continuous awareness of the current ﬁle position and the position of the required

f^ record. To remove these limitations, a randomly accessible ﬁle capability is provided by the mass storage

input/output subroutines.

f^ In a random ﬁle, any record may be read, written or rewritten directly, without concern for the position or

structure of the ﬁle. This is possible because the ﬁle resides on a random-access mass storage device that can

be positioned to any portion of a ﬁle. Thus, the entire concept of ﬁle position does not apply to a random

0s* ﬁle. The notion of rewinding a random ﬁle is, for instance, without meaning.

60497800 D 8-29

To permit random accessing, each record in a random ﬁle is uniquely and permanently identiﬁed by a record

key. A key is an 18- or 60-bit quantity, selected by the user and included as a parameter on the call to ^

read or write a record. When a record is ﬁrst written, the key in the call becomes the permanent identiﬁer

for that record. The record can be retrieved later by a read call that includes the same key, and it can be

updated by a write call with the same key. /S*^

When a random ﬁle is in active use, the record key information is kept in an array in the user's ﬁeld length.

The user is responsible for allocating the array space by a DIMENSION, type, or similar array declaration ^

statement, but must not attempt to manipulate the array contents. The array becomes the directory or index

to the ﬁle contents. In addition to the key data, it contains the word address and length of each record in

the ﬁle. The index is the logical link that enables the mass storage subroutines to associate a user call key /S~.

with the hardware address of the required record. '

The index is maintained automatically by the mass storage subroutines. The user must not alter the contents

of the array containing the index in any manner: to do so may result in destruction of the ﬁle contents.

(In the case of a sub-index, the user must clear the array before using it as a sub-index, and read the sub-

index into the array if an existing ﬁle is being reopened and manipulated. However, individual index entries

should not be altered.)

,/^N

When a permanent ﬁle that was created by mass storage input/output routines is to be modiﬁed it must be ^^^

attached with modify and extend permissions (append permission under NOS 1). Under NOS/BE 1 and 1

SCOPE 2, the EXTEND control statement should be used after the ﬁle is modiﬁed. Failure to extend the

ﬁle can render it unusable. ^-^

In response to a call to open the ﬁle, the mass storage subroutine automatically clear the assigned index array.

If an existing ﬁle is being reopened, the mass storage subroutines locate the master index in mass storage and

read it into this array. Subsequent ﬁle manipulations make new index entries or update current entries.

When the ﬁle is closed, the master index is written from the array to the mass storage device. When the ﬁle

is reopened, by the same job or another job, the index is again read into the index array space provided, so

that ﬁle manipulation may continue.

MASS STORAGE SUBROUTINES

0 bject time input/output subroutines control the transfer of records between central memory and mass storage. ^_^

OPENING A FILE ^

OPENMS opens the mass storage ﬁle and informs the system that it is a random (word addressable) ﬁle.

CALL OPENMS (u,ix,lngth,t) J

u Unit designator. <e^s^

ix Name of the array containing the master index.

lngth Length of master index "1

for a number index: lngth > (number of entries in master index) + 1

f for a name index: lngth > 2 * (number of entries in master index) + 1

8-30 60497800 D

/PSN

J0PrA.l

/P^S

t T y p e o f i n d e x .

t = 0 ﬁle has a number master index

t = 1 ﬁle has a name master index

The array (ix) speciﬁed in the call is automatically cleared to zeros. If an existing ﬁle is being reopened, the

master index is read from mass storage into the index array.

Example:

DIMENSION 1(11)

CALL OPENMS (5,1,11,0)

These statements prepare for random input/output on the ﬁle TAPE5 using an 11-word master index of the

number type. If the ﬁle already exists, the master index is read into memory starting at address I.

WRITING RECORDS

WRITMS transmits data from central memory to the ﬁle.

CALL WRITMS (u,fwa,n,k,r,s)

u Unit designator.

fwa Name of the array in central memory (address of ﬁrst word).

n Number of 60-bit words to be transferred.

k Record key.

for number index: 1 < k < lngth - 1

for name index k = any 60-bit quantity except +0

Rewrite,

r = 1

r = -1

Rewrite in place. Unconditional request; fatal error occurs if new record

length exceeds old record length.

Rewrite in place if new record length does not exceed old record length,

otherwise write at end-of-data.

r = 0 No rewrite; write at end-of-data (default value).

Sub-index ﬂag,

s = 1

s = 0

Write sub-index marker ﬂag in index control word for this record.

Do not write sub-index marker ﬂag in index control word (default value).

End-of-data (for r = -1 and r = 0) is deﬁned to be immediately after the end of the data record which is

closest to end of information. The ﬁrst record written at end-of-data overwrites the old index.

Except under SCOPE 2, Record Manager operates more efﬁciently if n is always a multiple of 64. The r

parameter can be omitted if the s parameter is also omitted. The s parameter is for future ﬁle editing

routines. Current routines do not test the ﬂag, but the user should include this parameter in new programs

(when appropriate) to facilitate transition to a future edit capability.

Example:

CALL WRITMS (3,DATA,25,6,1)

60497800 E 8-31

This statement unconditionally rewrites in place of ﬁle TAPE3, starting at the address of the array named

DATA, a 25-word record with an index number key of 6. The default value is taken for the s parameter.

READING RECORDS

READMS transmits data from the ﬁle to central memory. /=*%.

CALL READMS (u,fwa,n,k)

u Unit designator • ^^

fwa Name of the array in central memory (address of ﬁrst word)

n Number of 60-bit words to be transferred. If n is less than the record length, n words /**%

are transferred without diagnostic.

k Record key

f o r n u m b e r i n d e x : k = 1 < k < l n g t h - 1 ' ^

for name index: k = any 60-bit quantity except +0

/<'3*\

Except under SCOPE 2, Record Manager operates more efﬁciently if n is always a multiple of 64. 1

Example: ^~

CALL READMS (3,DATAMOR,25,2) )

This statement reads the ﬁrst 25 words of record 2 from unit 3 (TAPE3) into central memory starting at the ss$\

address of the array DATAMOR. I

CLOSING A FILE

CLOSMS writes the master index from central memory to the ﬁle and closes the ﬁle. CLOSMS is provided to 7

close a ﬁle so that it can be returned to the operating system before the end of a FORTRAN run, to preserve

a ﬁle created by an experimental job that might subsequently abort, or for other special purposes. In an over

lay program that is STATICly loaded, a mass storage ﬁle must be closed explicitly by CLOSMS. 1

Since new data records can overwrite the old index, a ﬁle which has had new data records added is invalid

unless the ﬁle is closed. (Under NOS/BE 1 and SCOPE 2 permanent ﬁles must also be extended.) Jobs which ^^

might abort before closing the ﬁles should use RECOVR to recover and terminate normally (i.e. STOP) to

cause the ﬁles to be closed.

When using mass storage input/output subroutines in overlays or segments, care should be taken to close a

ﬁle before program termination. If this is not possible, the mass storage input/output routines must reside in

the (0,0) overlay or root segment. This can be done by including a call to an MSIO routine in the (0,0) /m%

overlay or root segment (the call need not be executed), or by using the LIBLOAD control statement.

CALL CLOSMS (u) ^

u Unit designator

8-32 60497800 D

J0$\

/^ﬁ\

Example:

CALL CLOSMS (2)

This statement closes the ﬁle TAPE2.

SPECIFYING A DIFFERENT INDEX

STINDX selects a different array to be used as the current index to the ﬁle. The call permits a ﬁle to be

manipulated with more than one index. For example, when the user wishes to use a sub-index instead of

the master index, STINDX is called to select the sub-index as the current index. The STINDX call does not

cause the sub-index to be read or written; that task must be carried out by explicit READMS or WRITMS

calls. It merely updates the internal description of the current index to the ﬁle.

CALL STINDX (u,ix,lngth,t)

u Unit designator.

ix Name of the array in central memory containing the sub-index (ﬁrst word address).

lngth Length of sub-index

for a number index: lngth > (number of entries in sub-index) + 1

for a name index: lngth > 2 * (number of entries in sub-index) + 1

t Type of index. If omitted, t is the same as the current index.

t = 0 File has a number sub-index

t = 1 File has a name sub-index

Example 1:

DIMENSION SUBIX (10)

CALL STINDX (3,SUBIX, 10,0)

These statements select a new index, SUBIX, for ﬁle TAPE3 with an index length of 10. The records ref

erenced via this sub-index use number keys.

Example 2:

DIMENSION MASTER (5)

CALL STINDX (2,MASTER,5)

These statements select a new index, MASTER, from ﬁle TAPE2 with an index length of 5 and index type

unchanged from the last index used.

60497800 C 8-32.1/8-32.2

/Si^V

/ ^ \

f^l\

l/ﬁ^1

/^ts

.s^tf.

0&&\

INDEX KEY TYPES

There are two types of index key, name and number. A name key may be any 60-bit quantity except +0

or -0. A number key must be a simple positive integer, greater than 0 and less than or equal to the length

of the index in words, minus 1 word. The user selects the type of key by the t parameter of the OPENMS

call. The key type selection is permanent. There is no way to change the key type, because of differences

in the internal index structure. If the user should inadvertently attempt to reopen an existing ﬁle with an

incorrect index type parameter, the job will be aborted. (This does not apply to sub-indexes chosen by

STINDX calls; proper index type specification is the §ole responsibility of the user.) In addition, key types

cannot be mixed within a ﬁle. Violation of this restriction might result in destruction of a ﬁle.

The choice between name and number keys is left entirely to the user. The nature of the application may

clearly dictate one type or the other. However, where possible, the number key type is preferable. Job

execution will be faster and less central memory space will be required. Faster execution occurs because it is

not necessary to search the index for a matching key entry (as is necessary when a name key is used). Space

is saved due to the smaller index array length requirement.

MASTER INDEX

The master index type for a given ﬁle is selected by the t parameter in the OPENMS call when the index is

created. The type cannot be changed after the ﬁle is created; attempts to do so by reopening the ﬁle with

the opposite type index are treated as fatal errors.

SUB-INDEX

The sub-index type can be speciﬁed independently for each sub-index. A different sub-index name/number

type can be speciﬁed by including the t parameter in the STINDX call. If t is omitted, the index type

remains the same as the current index. Intervening calls which omit the t parameter do not change the most

recent explicit type speciﬁcation. The type remains in effect until changed by another STINDX call.

STINDX cannot change the type of an index which already exists on a ﬁle. The user must ensure that the t

parameter in a call to an existing index agrees with the type of the index in the ﬁle. Correct sub-index type

speciﬁcation is the responsibility of the user; no error message is issued.

MULTI-LEVEL FILE INDEXING

When a ﬁle is opened by an OPENMS call, the mass storage routines clear the array speciﬁed as the index

area, and if the call is to an existing ﬁle, locates the ﬁle index and reads it into the array. This creates the

f" initial or master index.

The user can create additional indexes (sub-indexes) by allocating additional index array areas, preparing

(^ the area for use as described below, and calling the STINDX subroutine to indicate to the mass storage

routine the location, length and type of the sub-index array. This process may be chained as many times as

required, limited only by the amount of central memory space available. (Each active sub-index requires an

(^ index array area.) The mass storage routine uses the sub-index just as it uses the master index; no distinc

tion is made.

8-33 60497800 A

/"Sfjv

A separate array space must be declared for each sub-index that will be in active use. Inactive sub-indexes

may, of course, be stored in the random ﬁle as additional data records.

If the user wishes to retain the sub-index, it must be written to the ﬁle after the current index designation

has been changed back to the master index, or a higher level sub-index by a call to STINDX.

Example 1 creates and modiﬁes a random ﬁle using a number index:

PROGRAM MSI (TAPE3)

C CREATE RANDOM FILE WITH NUMBER INDEX.

DIMENSION INDEX(ll), DATA(25)

CALL OPENMS (3,INDEX,11,0)

DO 99 NRKEY=1,10

C (GENERATE RECORD IN ARRAY NAMED DATA.)

99 CALL WRITMS (3.DATA,25,NRKEY)

STOP

END

8-34 60497800 A

The sub-index is read from and written to the ﬁle by the standard READMS and WRITMS calls, since it is

indistinguishable from any other data record. Although the master index array area is cleared by OPENMS /ts^

when the file is opened, STINDX does not clear the sub-index array area. The user must clear the sub-index ■■

array to zeros. If an existing ﬁle is being manipulated and the sub-index already exists on the ﬁle, the user

must read the sub-index from the ﬁle into the sub-index array by a call to READMS before STINDX is a

called. STINDX then informs the mass storage routine to use this sub-index as the current index. The ﬁrst '

WRITMS to an existing ﬁle using a sub-index must be preceded by a call to STINDX to inform the mass

storage routine where to place the index control word entry before the write takes place. /^v^V

z^^^V

PROGRAM MS2 (TAPE3)

C MODIFY RANDOM FILE CREATED BY PROGRAM MSI.

C NOTE LARGER INDEX BUFFER TO ACCOMMODATE TWO NEW

C RECORDS. ^

DIMENSION INDEX(13), DATA(25), DATAM0R(40)

CALL OPENMS (3,INDEX,13,0) ^

>c^v

/£?p&2\

C CLEAR THE SUBINDEX AREA.

DO 77 1=1,10

77 SUBIX(I)=0

C CHANGE THE INDEX IN CURRENT USE TO SUBIX.

CALL STINDX (2,SUBIX,10)

C GENERATE AND WRITE NINE RECORDS.

DO 88 MIN0R=1,9

C WRITE A RECORD.

88 CALL WRITMS (2.RECORD,50,MINOR)

C CHANGE BACK TO THE MASTER INDEX.

CALL STINDX (2,MASTER,5)

C WRITE THE SUBINDEX TO THE FILE.

CALL WRITMS (2,SUBIX,10,MAJOR,0,1)

99 CONTINUE

C READ THE 5TH RECORD INDEXED UNDER THE 2ND SUBINDEX.

CALL READMS (2,SUBIX,10,2)

CALL STINDX (2,SUBIX,10)

CALL READMS (2,RECORD,50,5)

C (MANIPULATE THE SELECTED RECORD AS DESIRED.)

STOP

END

Example 4:

PROGRAM MS5 (INPUT,OUTPUT,TAPE9)

C CREATE FILE WITH NAME INDEX AND TWO LEVELS OF SUBINDEX.

DIMENSION STATE(lOl), C0UNTY(501), CITY(501), ZIP(IOO)

INTEGER STATE, COUNTY, CITY, ZIP

10 FORMAT (A10.I10)

11 FORMAT (110)

12 FORMAT (5X.8I15)

CALL OPENMS (9,STATE,101,1)

60497800 A 8-37

C READ MASTER DECK CONTAINING STATES, COUNTIES, CITIES

C AND ZIP CODES.

DO 99 NRSTATE=1,50

READ 10.STATNAM, NRCNTYS

C CLEAR THE COUNTY SUBINDEX.

DO 21 1=1,501

21 C0UNTY(I)=O

DO 98 NRCN=1,NRCNTYS

READ 10, CNTYNAM, NRCITYS

C CLEAR THE CITY SUBINDEX.

DO 31 1=1,501

31 CITY(I)=0

CALL STINDX (9,CITY,501)

DO 97 NRCY=1,NRCITYS

READ 10, CITYNAM, NRZIP

C CLEAR THE ZIP CODE LIST

DO 41, J=l,100

ZIP(J)=0

DO 96 NRZ=1,NRZIP

READ ll.ZIP(NRZ)

CALL WRITMS (9,ZIP,NRZIP,CITYNAM)

CALL STINDX (9 , COUNTY,501)

CALL WRITMS (9,CITY,501,CNTYNAM)

CALL STINDX (9.STATE,101)

CALL WRITMS (9,COUNTY,501,STATNAM)

C FILE IS GENERATED. NOW PRINT OUT LOCAL ZIP CODES.

CALL STINDX (9,STATE,101)

CALL READMS (9,COUNTY,501, ^CALIFORNIA?*)

CALL STINDX (9,COUNTY,501)

CALL READMS (9,CITY,501, -*SANTACLARA#)

CALL STINDX (9,CITY,501)

CALL READMS ( 9 , ZIP, 100, ^SUNNYVALE?*)

PRINT 12, ZIP

CALL STINDX (9.STATE,101)

STOP

END

8-38 60497800 A

10^

/ﬂS\

0mHi\

J0^K

C READ 8TH RECORD FROM FILE TAPE3.

CALL READMS (3,DATA,25,8)

C (MODIFY ARRAY NAMED DATA.)

C WRITE MODIFIED ARRAY AS RECORD 8 AT END OF

C INFORMATION IN THE FILE

CALL WRITMS (3,DATA,25,8)

C READ 6TH RECORD.

CALL READMS (3.DATA,25,6)

C (MODIFY ARRAY.)

C REWRITE MODIFIED ARRAY IN PLACE AS RECORD 6.

CALL WRITMS (3,DATA,25,6,1)

C READ 2ND RECORD INTO LONGER ARRAY AREA.

CALL READMS (3,DATAMOR,25,2)

C (ADD 15 NEW WORDS TO THE ARRAY NAMED DATAMOR.)

C CALL FOR IN-PLACE REWRITE OF RECORD 2. IT WILL

C DEFAULT TO A NORMAL WRITE AT END-OF-INFORMATION

C SINCE THE NEW RECORD IS LONGER THAN THE OLD ONE,

C AND FILE SPACE IS THEREFORE UNAVAILABLE.

CALL WRITMS (3,DATAM0R,4O,2,-1)

C READ THE 4TH AND 5TH RECORDS.

CALL READMS (3,DATA,25,4)

CALL READMS (3,DATAM0R,25,5)

C (MODIFY THE ARRAYS NAMED DATA AND DATAMOR.)

60497800 A 8-35

/*^L

C WRITE THE ARRAYS TO THE FILE AS TWO NEW RECORDS,

CALL WRITMS (3,DATA,25,11)

CALL WRITMS (3,DATAM0R,25,12)

STOP

END

Example 2 uses a name index for a random ﬁle:

PROGRAM MS3 (TAPE7)

C CREATE A RANDOM FILE WITH NAME INDEX.

DIMENSION INDEX(9), ARRAY(15,4)

DATA REC1 ,REC2/7HREC0RD1,t*REC0RD2t*V

CALL OPENMS (7,1NDEX,4,1)

(GENERATE DATA IN ARRAY AREA.)

WRITE FOUR RECORDS TO THE FILE. NOTE THAT

KEY NAMES ARE RECORD(N).

CALL WRITMS (7,ARRAY(1,1),15,REC1)

CALL WRITMS (7,ARRAY(1,2),15,REC2)

CALL WRITMS ( 7 , ARRAY( 1, 3 ) , 15 ,.7HRECORD3)

CALL WRITMS (7, ARRAY(1,4) , 15 , t*REC0RD4#)

CLOSE THE FILE.

CALL CLOSMS (7)

STOP

END

Example 3:

PROGRAM MS4 (TAPE2)

C GENERATE SUBINDEXED FILE WITH NUMBER INDEX. FOUR

C SUBINDEXES WILL BE USED, WITH NINE DATA RECORDS

C PER SUBINDEX, FOR A TOTAL OF 36 RECORDS.

DIMENSION MASTER(5), SUBIX(IO), REC0RD(50)

CALL OPENMS (2,MASTER,5,0)

DO 99 MAJ0R=1,4

/*SS\

8-36 60497800 E

COMPATIBILITY WITH PREVIOUS MASS STORAGE ROUTINES

0$&\.

FORTRAN Extended mass storage routines and the ﬁles they create are not compatible with mass storage

routines and ﬁles created under versions of FORTRAN Extended before version 4. Major internal differences

f^ in the ﬁle structure were necessitated by adding the Record Manager interface. However, source programs are

fully compatible. Any source program that compiled and executed successfully under earlier versions will do

so under this version, provided that all ﬁle manipulated by mass storage routines are manipulated only by

f^ these routines.

** FORTRAN-CYBER RECORD MANAGER INTERFACE

0 S \

The CYBER Record Manager interface subroutines correspond closely to the CYBER Record Manager COMPASS

macros. The names are different in some cases, and the parameters are not necessarily speciﬁed in the same order,

but the processing performed by each subroutine is for the most part the same as the corresponding COMPASS

macro.

Only a summary of the format, parameters, and purpose of each subroutine is given here. The differences in

usage of these routines among the ﬁve ﬁle organizations are not discussed. In order to use these routines, it is

necessary to refer to the CYBER Record Manager publications listed in the preface.

The user can either allocate buffers within a program block or allow CYBER Record Manager to allocate them

dynamically when the ﬁle is opened.

To allocate a buffer within the program block, an array must be dimensioned and the length and position of

the array speciﬁed by the BFS and FWB ﬁelds of the ﬁle information table. If either of these ﬁelds is zero

when the ﬁle is opened, CYBER Record Manager allocates a buffer in central memory following the execut

able code and blank common (if declared). In an overlay program, dynamically allocated buffers are assigned

to memory beyond the last word address of the longest overlay chain.

These routines are available under NOS/BE 1 and NOS 1, but not under SCOPE 2.

PARAMETERS

The ﬁrst parameter in the call to every subroutine is the name of the array containing the ﬁle information table

being processed. This array should be dimensioned 35 words long; 20 words for the ﬁle information table

itself and 15 for the ﬁle environment table. Any other parameters can be omitted; default values are supplied

by CYBER Record Manager. With the exception of FILExx, parameters are identiﬁed strictly by position;

thus, parameters can be omitted only from the right.

When a program is compiled with OPT=2, wsa must be speciﬁed on all calls to GET, GETP, and GETN. Also,

ka must be speciﬁed on calls to GETN and PUT for indexed sequential, direct access, and actual key ﬁles.

Most of the parameters establish values for ﬁle information table ﬁelds. CYBER Record Manager always uses

the most recent value established for a ﬁeld; if a parameter is omitted, the previous contents of the ﬁeld are

used instead.

If the same subroutine is called twice in the same program unit with a different number of parameters, an

informative diagnostic is issued by the compiler.

60497800 F 8-39

/ ^ v

Values for parameters can be: "^

Array or variable names, identifying areas used for communication between the user program and

CYBER Record Manager ^

Subprogram names for user owncode exits (must be speciﬁed in an EXTERNAL statement)

Integer values

L format Hollerith constants, used to express symbolic options and to identify file information table ■'**%

ﬁelds

The following mnemonics are used in the subroutine formats below. The precise meaning of any parameter '"^

depends on the ﬁle organization of the ﬁle being processed, as well as the subroutine being called. Not all

parameters are applicable to all ﬁle organizations.

ﬁt Name of array containing ﬁle information table. Linked to the actual ﬁle by means of the LFN ﬁeld.

aﬁt Name of an array that contains a list of addresses of FITs terminated by a word of zeros. /^

wsa Working storage area. A variable, array, or array element name indicating the starting location

from which data is to be read or into which data is to be written. ^^

pd Processing direction established when ﬁle is opened:

5LINPUT Read only

6LOUTPUT Write only

3LI-0 Read and write '"^

3LNEW File creation (indexed sequential, direct access, actual key only)

of File positioning at open time: '")

1LR Rewind

1LN No ﬁle positioning

ILE Extend; ﬁle is positioned immediately before end of information

cf File positioning after close:

1LR Rewind ^

1LN No positioning

1LU Unload /*-**.

3LRET Return

3LDIS Disconnect (terminal ﬁles only)

3LDET No positioning; release buffer space and remove from active ﬁle list

8-40 60497800 F

/SpSy

type Type of close (not a ﬁle information table ﬁeld):

4LFILE File close

6LVOLUME Volume close

ka Location of key for access to record in a direct access, indexed sequential, or actual key

ﬁle. For GETN, key is returned to this location.

wa Location of word address for read or write of record in a word addressable ﬁle.

kp Character position (0 through 9) within word designated by ka in which key begins (direct

access, indexed sequential only).

mkl Major key length (indexed sequential only).

rl Record length in characters for record to be read or written,

ex Name of user owncode error exit subroutine,

dx Name of user owncode data exit subroutine,

post For duplicate key processing:

1LP Write record preceding current record

1LN Write record as next record

1LC Delete or replace current record

0 Delete or replace ﬁrst record in duplicate key chain

count Number of records to be skipped; positive count indicates forward skip, negative count indi

cates backward skip, zero count should not be used.

ptl Number of characters to be used for a partial read or write.

skip Positioning before execution of GETP:

0 Continue reading at current position

4LSKIP Skip to beginning of next record before reading

lev Level number for end of section; 0 to 17.

id FIT identiﬁer.

'Applies only to Initial Indexed Sequential ﬁles.

60497800 E 8-41

SUBROUTINES

In the subroutine formats below, braces are used to indicate that more than one parameter occupies the same

position. In all cases, these parameters are applicable to mutually exclusive ﬁle organizations.

CALL FILExx (ﬁt, keyword-,, value-,, ... ,keywordn, valuen)

xx is SQ (for sequential ﬁles), IS (for indexed sequential ﬁles), DA (for direct access ﬁles), AK (for actual

key ﬁles) or WA (for word addressable ﬁles).

All parameters, with the exception of ﬁt, are paired. The ﬁrst parameter in each pair is the name of a ﬁle information

table ﬁeld, in L format. The second parameter of each pair is the value to be set in that ﬁeld. CALL FILExx must be

executed before the ﬁle is opened. CALL FILExx ensures that the object libraries BAMLIB and AAMLIB are made

available to the job.

CALL STOREF (ﬁt, keyword, value)

STOREF speciﬁes a value for a single ﬁle information table ﬁeld. It can be called before or after the ﬁle is opened.

The keyword is the name of a ﬁle information table ﬁeld, in L format, and value is the value to be placed in that ﬁeld.

IFETCH(ﬁt,ﬁeld) or CALL IFETCH(ﬁt,ﬁeld,value) ^

IFETCH is an integer function that returns the current value of a single ﬁle information table ﬁeld. A one-bit ﬁeld is

returned in the sign bit; if the bit is 1, the value of the function is negative; if the bit is 0, the value of the function is

positive.

IFETCH can also be called as a subroutine; in which case, the value is returned in the integer variables speciﬁed as the

third parameter.

CALL OPENM(ﬁt,pd,of)

OPENM opens a ﬁle and prepares it for further processing. Only FILExx, STOREF, and IFETCH can precede

execution of CALL OPENM.

CALL CLOSEM (ﬁt,cfftype)

CLOSEM closes the ﬁle after all processing has been completed. Only STOREF and IFETCH can follow execution

of CLOSEM.

CALL GET(fit,wsa, <ka >,kp,mkl,rl, |®X 1)

GET reads a record and returns it to the working storage area (wsa). The last parameter speciﬁes dx for

sequential ﬁles, ex for all other ﬁles.

CALL PUT(ﬁt,wsa,rl,< a >,kp,pos,ex)

lwa J

PUT writes a record to the ﬁle from the working-storage area (wsa).

CALL GETP(ﬁt,wsa,ptl,skip,dx)

GETP reads a partial record. The number of characters to be read is indicated by ptl.

8-42 60497800 F

r^^\

/ftw^TV

/{SN

CALL PUTP(ﬁt,wsa,ptl,rl,ex)

PUTP writes a partial record. The number of characters to be written by this write is indicated by ptl; the

total number of characters to be written is given by rl (required only for record types U, W, and R).

CALL GETI\l(ﬁt,wsa,ka,ex)

GETN reads the next record in sequential order from an indexed sequential, direct access, or actual key ﬁle.

The key of the record read is placed in ka after the read.

CALL DLTE(ﬁt,ka,kp,pos,ex)

DLTE deletes a record from an indexed sequential, direct access, or actual key ﬁle. The key of the record

to be deleted is in the location speciﬁed by ka.

CALL REPLC(ﬁt,wsa,rl,ka,kp,pos,ex)

REPLC replaces a record on a sequential, indexed sequential, direct access, or actual key ﬁle. The key of the

record to be replaced is in the location speciﬁed by ka; the new record is in the working storage area indicated

by wsa. For sequential ﬁles, the last record read is replaced by a record of exactly the same size.

CALL WEOR(ﬁt,lev)

WEOR terminates a section or partition, or S type record.

CALL WTMK(ﬁt)

Writes a tape-mark (equivalent to end of partition).

CALL ENDFILE(ﬁt)

Writes an end of partition.

CALL REWND(ﬁt)

REWND positions a tape ﬁle to the beginning of the current volume. It positions a mass storage ﬁle to the

beginning of information.

CALL GETNR(ﬁt,wsa,ex,ka)

GETNR transfers the next record in sequential order to the working storage area, unless an input/output

operation is required, in which case control returns to the user before the input is complete. The user

must continue to call GETNR until the transfer is complete (FP ﬁeld of the FIT is set to 0).

60497800 E 8-43

^o$!\

CALL FLUSHM(aﬁt)

FLUSHM performs all ﬁle close operations (such as buffer ﬂushing), but the ﬁle remains open.

CALL FLUSH 1 (ﬁt)

FLUSH 1 performs the same function as FLUSHM, but for a single ﬁle instead of a list of ﬁles.

CALL FITDMP(ﬁt,id)

FITDMP dumps the contents of the ﬁle information table to the error ﬁle ZZZZZEG. The CRMEP control

statement (see the CYBER Record Manager AAM reference manual) can then be used to print ﬁle ZZZZZEG.

CALL SEEKF(ﬁt,ka,kp,inkl,ex)

SEEKF initiates block transfer to the ﬁle buffer. The program can continue processing while the transfer

occurs. This overlapping of central memory processing and input/output activity can shorten program execu

tion time.

CALL SKIP(ﬁt,count)

SKIP repositions an indexed sequential or actual key ﬁle in a forward or backward direction a speciﬁed num

ber of records. It does not return a record to the working storage area. A positive value for count indicates

a forward move; a negative value indicates a backward move.

CALL STARTM(ﬁt,ka,kp,mkl,ex)

STARTM positions an indexed sequential or alternate key index ﬁle to a record that meets a speciﬁc con

dition; the record is not transferred to the working storage area. The ﬁle is positioned according to the key

relation ﬁeld in the ﬁle information table and the current value at the key address location. A new key

address value is stored in ka.

ERROR CHECKING

CYBER Record Manager interface routines perform limited error checking to determine whether the call can

be interpreted, but actual parameter values are not checked.

The following fatal error conditions are detected at execution time, and a message appears in the dayﬁle:

FIT ADDRESS NOT Array name was not speciﬁed.

SPECIFIED

FORMAT ERROR Parameters were not paired (FILExx), or required parameters were not speci

ﬁed (STOREF, IFETCH or SKIP).

UNDEFINED SYMBOL A ﬁle information table ﬁeld mnemonic or symbolic option was speciﬁed

incorrectly; for example, an incorrect spelling, or the of_ parameter in

OPENM was not speciﬁed as R, N or E.

8-44 60497800 H

/ ^ t x

«£^»

0^s

0$Z\

/a$&\

MULTIPLE INDEX PROCESSING

FORTRAN Extended provides the capability of multiple indexing for IS, DA, and AK ﬁles via CYBER

Record Manager.

Each multiple-indexed ﬁle has an associated alternate key index ﬁle. An alternate key index is a cross-

reference table of alternate values and IS, DA, or AK primary key values. The key-ﬁeld position identiﬁes

each table, which consists of all the different alternate key values that occur in the records of the ﬁle.

Associated with each alternate key value is a list of primary keys, each of which identiﬁes a record contain

ing the alternate key value.

To utilize this capability, the index ﬁle is speciﬁed in the XN ﬁeld of the ﬁle information table. To open

the index ﬁle, the following statement is used:

CALL RMOPNX(ﬁt,pd,of)

0g* The parameters are the same as those of CALL OPENM. The ﬁle may be opened by a CALL OPENM

V instead of CALL RMOPNX if XN was speciﬁed on a FILE control statement rather than by a CALL

FILExx.

The following subroutine should be called to describe a key ﬁeld when creating a new IS, DA, or AK

ﬁle. It must be called once for each key ﬁeld in the record.

CALL RMKDEF(ﬁt,kw,kp,kl,ki,kt,ks,kg,kc)

ﬁt Name of an array containing the ﬁle information table

kw Word of record in which key starts (0 = ﬁrst word)

kp Starting character position of key (0 through 9)

V kl Key length in characters (1 through 255)

ki Summary index; reserved (0)

f^ kt Key type: 0 = symbolic, 1 = signed integer, 2 = unsigned

ks Substructure for each primary key list in the index; I = index-sequential; F = FIFO;

^ U (default) = unique; speciﬁed as L format Hollerith constant

kg Size of repeating group in which key resides (default = 0)

kc Occurrences of group (default = 0)

To position a multiple index ﬁle, the following subroutine is used:

(^ CALL STARTM(ﬁt,ka,kp,mkl,ex)

If the RKW and RKP parameters are set to indicate the primary key, STARTM positions the data ﬁle and

f^ subsequent calls to GETN retrieve records in sequential order. If RKW and RKP indicate an alternate key,

STARTM positions the index ﬁle, and subsequent calls to GETN retrieve records in their order on the index

ﬁle.

0^\

60497800 J 8-45 |

/fW$\

mrl Maximum length in characters of records to be sorted.

ba Number of words of central memory to be used by Sort/Merge for working storage. If

omitted, amount is computed by Sort/Merge.

8Applies only to SCOPE 2.

Applies only to NOS 1 and NOS/BE 1.

| 8-46 60497800 J

ir^&)\

FORTRAN-SORT/MERGE INTERFACE ^

FORTRAN Extended provides the capability for processing data records under the Sort/Merge system from

within a FORTRAN program. The FORTRAN user of this feature should be familiar with the autonomous "^

functioning of the Sort/Merge- system as described in the Sort/Merge Reference Manual.

Sort/Merge uses the unused part of the ﬁeld length as a scratch area; if this is not adequate, additional ﬁeld +*^

length is obtained from the system. For this reason the STATIC control statement parameter must not be

used for programs using SORT/MERGE.

The FORTRAN subroutines interfacing with Sort/Merge are listed below. The series of calls to Sort/Merge

subroutines must begin with a call to SMSORT, SMSORTB, SMSORTP, or SMMERGE. If a ﬁle is processed

by CYBER Record Manager subroutines, OPENM should be called before any of these routines. The ^%

Sort/Merge subroutines are on the library SRTLIB.

In an overlay structured program using blank common, the Sort/Merge interface routines must not be called <*^.

from the (0,0) overlay.

CALL SMSORT(mrl,ba) ^

mrl Maximum length in characters of records to be sorted.

ba§ LCM buffer area in decimal for intermediate scratch files constructed by Sort/Merge.

ba* Number of words of central memory to be used by Sort/Merge for working storage. If

omitted, amount is computed by Sort/Merge.

SMSORT calls for a sort on rotating mass storage. .-«a\

CALL SMSORTB(mrl,ba)*

mrl Maximum length in characters of records to be sorted.

ba Number of words of central memory to be used by Sort/Merge for working storage. If jsmx

omitted, amount is computed by Sort/Merge.

SMSORTB calls for a balanced tape sort. SMTAPE (see below) must also be called. /^,

CALL SMSORTPtmrLba)*

<<*~|\

0$S\

SMSORTP calls for a polyphase tape sort. SMTAPE must also be called.

' CALL SMMERGE(mrl,ba)

j#s\ mrl Maximum length in characters of records to be merged.

ba» LCM buffer area in decimal for intermediate scratch files constructed by Sort/Merge.

ba Number of words of central memory to be used by Sort/Merge for working storage. If

omitted, amount is computed by Sort/Merge.

SMMERGE calls for merge-only processing.

The MRL option on the SMSORT, SMSORTB, SMSORTP, and SMMERGE calls is used only when one or

more ﬁles are FORTRAN ﬁles. If all ﬁles are processed through direct calls to CYBER Record Manager

(access = 0 in SMFILE calls), the MRL option is required only as a dummy parameter.

CALL SMFILE(dis,i/o,lfn,action)

dis File disposition:

-^SORT* File to be sorted.

=£MERGE=£ File to be merged.

-v^OUTPUT-J6 File to receive output.

i/o Mode of ﬁle input/output:

^ F O R M AT T E D * ) _ . , J . u r

>CODED ,= t Fde accessed with formatted input/output.

-v^-BINARY^ File accessed with unformatted input/output.

0* File accessed with interfacing CYBER Record Manager subroutines

(see this section above).

J?"-X

/"™N

JP*N

£0&i"\

lfn File name indicator:

( u Logical unit number, 0 to 99.

nLﬁlename File name left justiﬁed with zero ﬁll.

jssn ﬁt* When i/o is speciﬁed as 0, an array containing the ﬁle information

f table.

sg^ action File disposition following sort or merge:

=£REWIND*

^UNLOAD^

=£NONE=£ (default)

8Applies only to SCOPE 2.

^Applies only to NOS 1 and NOS/BE 1.

60497800 J 8-47

SMFILE must be called for each ﬁle to be sorted or merged, and once for the ﬁle to receive the output (unless SMOWN

is called). If a ﬁle is to be accessed with formatted or unformatted FORTRAN input/output, its name must be declared

in the PROGRAM statement. Files should be properly positioned before they are sorted or merged.

CALL SMKEY(charpos,bitpos,nchar,nbits,code,colseq,order)

charpos Integer specifying position of ﬁrst character of sort key, considering the ﬁrst characters as position

number 1.

bitpos Integer specifying position of ﬁrst bit of sort key in character (or 6-bit byte) speciﬁed by charpos,

considering the ﬁrst bit as position number 1.

nchar Integer specifying number of characters or complete 6-bit byte in sort key.

nbits Integer specifying number of bits in sort key in excess of those indicated by nchar.

code Coding identiﬁer:

t^SEPARATE-t-^TRAILINGt6 Numeric data in display code; sign is a separate character

at end of ﬁeld.

/f^\

■v^-DISPLAYS Internal display code.

t^FLOAT^ Floating point data.

^INTEGER-* Signed integer data. . ^

^LOGICAL-?5- Unsigned integer data (default).

The following identiﬁers must be speciﬁed in pairs separated by a comma, as indicated. Each pair /=^s

is positionally interchangeable:

T^SIGN-y*,^-LEADINGS Numeric data in display code; sign present as an overpunch /ee^

at beginning of ﬁeld. j

■v^SIGN-^^TRAILlNG^ Numeric data in display code; sign present as an overpunch I

a t e n d o f ﬁ e l d . )

t^SEPARATE^-v^LEADINGt6 Numeric data in display code; sign is a separate character at

beginning of ﬁeld. /*s~\

,/*^%y

colseq Collating sequence (applicable only if code is specified as =£DISPLAY-v^):

-v* ASCI^t6 6-bit ASCII collating sequence (default for installations using "^%

ASCII character set).

=£COBOL6=?-: 6-bit COBOL collating sequence (default for installations using

CDC character set) ,&%

¥= DISPLAYS Internal display collating sequence.

=fc INTBCD^ Internal BCD collating sequence,

sequence Name of a collating sequence speciﬁed in a call to SMSEQ (see below). ^^

If a code identiﬁer other than DISPLAY is used, this ﬁeld must be omitted; otherwise, run time

error 165 is issued. /SSI*\

order Order of sort processing.

t^A-v* Ascending (default). •

zfrD^z Descending.

One SMKEY call is required to describe each sort key to be used. The ﬁrst SMKEY call indicates the major key; ^

subsequent calls indicate additional or minor keys in the order encountered.

8-48 60497800 F

ys^X

—^ CALL SMSEQ (seqname,seqspec)

r ■

seqname Name of user supplied collating sequence.

' seqspec Name of integer array, terminated with a negative number, containing entire sequence of

characters in order of collation.

(^ SMSEQ speciﬁes a user's collating sequence, or redeﬁnes the default to be a user collating sequence or a

standard collating sequence other than the system default.

.-^v The characters in seqspec can be speciﬁed as their octal equivalents in the form ijB or as Hollerith constant:.

■ in the form IRx. Characters to collate equal are specified in a call to SMEQU (sec below). Unspecified char

acters collate high (following the last character speciﬁed in seqspec) and equal.

id-N

CALL SMEQU (colseq.equspec)

0xS~*>\

f colseq Collating sequence determined by a previous call to SMKEY (and perhaps SMSEQ).

equspec Name of an integer array, terminated with a negative number, containing characters to collate

{ equal to the last character, which must be included in colseq.

SMEQU speciﬁes that two or more characters in the collating sequence are equal for comparison purposes.

CALL SMOPT (opt., optn)

\ opt Non-ordered options separated by commas:

=£ VERIFY^ Check output for correct sequencing (important for insertions

(^ " during output and merge input).

=£ RETAINS Retain records with identical sort keys in order of appearance

on input ﬁle.

^VOLDUMP=£ * Checkpoint dump at end-of-volume.

-y^DUMP-^* Checkpoint dump after 50,000 records.

_ ^DUMP-v^n* Checkpoint dump after (decimal) n records.

f -^NODUMP-v** No checkpoint dumps.

^NODAY-v** Suppress dayﬁle messages.

-p-x ^ORDER=£,mo* Merge order = mo (default: mo = 5).

t^COMPAREt^ The key comparison sorting technique is to be used.

¥-EXTRACT=£ The key extraction sorting technique is to be used.

C =t-€OMPARE=£ and #EXTRACT^ are mutually exclusive. If both are omitted, Sort/Merge decides which to

use. =5*€OMPARE=£ usually decreases elapsed time while increasing central processor time, whereas -v^EXTRACT^

usually decreases central processor time while increasing elapsed time.

j0&\.

ifjP^V

SMOPT speciﬁes special record handling options. If SMOPT is called more than once, the last call will override all

previous calls. If SMOPT is called, it must be done immediately after the call to SMSORT or SMMERGE.

* Applies only to NOS 1 and NOS/BE 1.

60497800 F 8-49

CALL SWITAPE (taplist)*

taplist List of logical ﬁle names, each in the form nLﬁlename, to be used in balanced or polyphase

tape merge.

The ﬁle names in taplist must not be declared in the PROGRAM statement. A balanced merge requires a mini

mum of four tapes; a polyphase merge, a minimum of three tapes.

CALL SMOWN (exitnum.,,subname.,,..., exitnumn,subnamen)

exitnum Number of the owncode exit.

subname Name of the user-supplied owncode exit subroutine

Each subname speciﬁed in a call to SMOWN must appear in an EXTERNAL statement in the calling program.

For each subname speciﬁed, the user must supply a subroutine which exits through a call to system subroutine

SMRTN, in accordance with the owncode exit number and return address as follows:

exitnum entry exit

1 or 3 SUBROUTINE subname (a,rl) CALL SMRTN (retaddr), for retaddr = 1 or 3

CALL SMRTN (retaddr,b,rl), for retaddr = 0 or 2

2 or 4 SUBROUTINE subname CALL SMRTN (retaddr), for retaddr = 0

CALL SMRTN (retaddr,b,rl), for retaddr = 1

5 SUBROUTINE subname (a^rl-,^,^) CALL SMRTN (b1,rl1,b2,rl2), for retaddr = 0

CALL SMRTN (b, .rl,), for retaddr = 1

retaddr Return address:

Normal return address

Normal return address + 1

Normal return address + 2

Normal return address + 3

a Integer array of length (rl + 9)/10 in which Sort/Merge stores a record when subname is called.

Storing into a causes indeterminate results.

b Integer array of length (rl + 9)/10 in which the user stores a record when subname is called,

b should not be the same as a.

rl Record length in characters.

No parameters are needed on SUBROUTINE subname for exit number 1 if there are no input ﬁles.

CALL SMEND

Required as the last in a series of Sort/Merge interfacing subroutines, SMEND initiates execution of the sort or merge.

CALL SMABT

Terminates a sequence of SORT/MERGE interface calls without calling Sort/Merge. The state of the interface

is the same as if no calls had been made.

* Applies only to NOS 1 and NOS/BE 1.

8-50 60497800 F

/Cff^5\

<*C^V

.»^S^!\

^ s i - N

FORTRAN-CYBER INTERACTIVE DEBUG INTERFACE

CYBER Interactive Debug (CID) is a debugging facility, available under NOS 1 and NOS/BE 1, which allows the user

to monitor and control the execution of programs from an interactive terminal. CID is on the library DBUGLIB.

A brief discussion of CID is presented here. For more information, refer to the CYBER Interactive Debug

reference manual.

FORTRAN Extended provides the capability of interfacing with CID. The CID features allow the user

to:

Suspend program execution at speciﬁed locations called breakpoints.

.j*i~"\

- Set traPs which cause program execution to be suspended on speciﬁc events, such as the loading of an

overlay.

0^\

Display values stored into variables and arrays while program execution is suspended.

^x Enter data into the program.

Interrupt and restart the program from the terminal.

Deﬁne and save sequences of CID commands to be executed automatically when a breakpoint or trap is

encountered during program execution.

j CONTROL STATEMENT

{ In order to make use of all the CID facilities, a FORTRAN program must be compiled, loaded and

executed in debug mode. Debug mode is activated by the control statement

V DEBUG or DEBUG(ON)

When a source program is compiled in debug mode, the compiler produces a line number table and a symbol

^s table along with the binary object code. The CID package is loaded along with the compiled code and

becomes part of the user's ﬁeld length.

CID is deactivated by the control statement

/SIN

f DEBUG(OFF)

As an alternative to compiling with DEBUG(ON), the necessary compiler tables can be produced by

f specifying DB or DB=ID on the FTN control statement. Subsequent executions with DEBUG(ON) can

make use of CID.

09&K.

v If debug mode has been activated with DEBUG(ON), it can be subsequently turned off for the duration of a

compilation by specifying DB=0 on the FTN control statement. The default is DB=0.

'. A program that has been compiled with DEBUG(ON) or DB=ID can subsequently be executed with

DEBUG(OFF), but CID cannot be used.

60497800 F 8_51

/tt'ims$\

/^^\

USER-CID INTERACTION ^

In debug mode, after the user's program has been loaded, but before execution is initiated, CID requests input

of commands. Typically, the user initially sets breakpoints and traps which specify debugging options to be

performed during program execution.

When a breakpoint or trap is encountered during execution, execution is suspended while CID performs the

sequence of commands speciﬁed in the body of the breakpoint or trap deﬁnition. With certain breakpoints

or traps, the user has the option of entering debug commands at the terminal before execution is resumed.

CID OUTPUT

Output from CID consists of informative messages, diagnostics, and the results of commands. Certain

informative messages always appear at the terminal; other messages are arranged into classes, and the user

can specify which message classes are to be sent to the terminal.

BATCH DEBUGGING

CID is primarily intended to be used interactively, but can be used in batch mode. In this case, the user must

place CID commands as the ﬁrst record in the ﬁle DBUGIN.

Output from CDD is written to a ﬁle called DBUGOUT. The type of output written to this ﬁle is controlled

in the same manner in which output is sent to the terminal when CID is used interactively.

INTERFACE TO COMMON MEMORY MANAGER

Common Memory Manager (CMM) is used for the management of ﬁeld length, except when using the static loading

options. CMM ensures that the ﬁeld length is increased or decreased properly to accommodate assigned blocks.

Interface to CMM can be done to assign blocks of memory for arrays. This assignment is completely dynamic,

and for efﬁcient use, the blocks should be returned to the system when ﬁnished.

The Common Memory Manager reference manual should be read for a detailed description of CMM usage.

The following descriptions are for simple CMM usage.

CMMALF is called to allocate a ﬁxed position block. The dynamic array to be assigned is deﬁned in the *^

FORTRAN program as an array of length 1. The proper offset to the base address of the array is calculated

by using the LOCF function on the array name, and subtracting this value from the ﬁrst word address of the

block returned by CMM. The array element n can then be referenced by array name (offset + n). For /m>\

example, the following statements assign a block and set the ﬁfth element to 1:

PROGRAM CMM1

DIMENSION CMMAR(l)

ILEN=10

CALL CMMALF(ILEN,0,0,IFWA)

IOFF=IFWA-LOCF(CMMAR(l))

CMMAR(IOFF +5)=1.0

y^^\

8-52 60497800 H

iHPN

0®>*

The calling sequence for CMMALF is:

CALL CMMALF(IBLKSZ,ISZCDE.IGRPID,IBLFWA)

IBLKSZ Number of words required for the block.

ISZCDE Size code:

0 Fixed size block (should be used in most cases).

1 Block can grow at last word address.

2 Block can shrink at last word address.

4 Block can shrink at ﬁrst word address.

5 Block can grow at last word address and shrink at ﬁrst word address.

6 Block can shrink at ﬁrst and last word addresses.

7 Block can shrink at ﬁrst and last word addresses and grow at last word

address.

IGRPID Group identiﬁer:

0 Item does not belong to a group (normal usage).

>0 The block is assigned to this group. The group number is determined by

^ calling CMMAGR (see the Common Memory Manager reference manual).

The group number may be any value greater than 0.

The value returned from a call to CMMALF is:

IBLFWA First word address of block allocated by CMM.

(^ CMM FRF is called to free the ﬁxed-position block when it is no longer needed. When the block is freed, the

contents of the block are no longer accessible.

(; The calling sequence for CMMFRF is:

_ CALL CMMFRF(IBLFWA)

IBLFWA First word address of block (must have been returned by CMMALF).

/P\ Other routines are available to accomplish other tasks, such as determining maximum ﬁeld length and other statistics,

assigning blocks to groups, and releasing groups of blocks (see the Common Memory Manager reference manual). All

CMM interface routines for NOS and NOS/BE are on the library SYMLIB. Therefore, the statement LDSET

/""n (LIB=SYMLIB) must be included in the loader directives for a run using the CMM interface routines, or the user

\ should include a CALL SYMLIB subroutine call in the main program. SCOPE 2 users must specify SYMIO in the

LDSET statement instead of SYMLIB.

/"PN

POST MORTEM DUMP

Post Mortem Dump (PMD) analyzes the execution time errors in FORTRAN Extended Version 4 programs.

PMD provides interpreted output in a form which is more easily understood than the octal dump normally

output following a fatal error; PMD prints a summary of the error condition and the state of the program at

the time of failure in terms of the names used in the original program. The names and values of the variables

60497800 G 8-53

Use of PMD does not affect the use of FORTRAN Extended DEBUG or CYBER Interactive Debug. PMD is

activated by a hardware or software fatal error and can also intentionally be invoked by the user. PMD overrides

any user-supplied load map directive or MAP(ON) control statement. For example, the following statements do

not produce a load map if PMD was speciﬁed:

LDSET(IVIAP=SBEX)

LOAD (LGO)

EXECUTE.

However, the loader always writes a block and statistics map to ﬁle ZZZZZMP for PMD's use. It is the user's

responsibility to rewind and copy this ﬁle to output. If nonfatal loader errors occur, a summary of the errors is

included in the PMD output.

When PMD is used, the FORTRAN Extended compiler generates a loader request to preset all memory to a

special value for initialization testing. This preset is similar to that produced by the following load sequence:

LDSET(PRESETA=60000000000433400000)

LOAD (LGO)

EXECUTE.

in the routine in which the error was detected are printed; this process is repeated, tracing back through the calling **±

sequence of routines until the main program is reached.

//Ss\

/^*5^N

^^\

Any user LDSET(PRESET=) loader speciﬁcation is overridden.

PMD reloads the user ﬁeld length before it aborts to allow a subsequent octal dump of the user's program if one has

been speciﬁed.

To use PMD, the PMD parameter must be speciﬁed on the FTN control statement. PMD will then be activated by a

fatal execution error or by one of the user-callable subroutines PMD LOAD or PMDSTOP. Information provided by

the dump includes the following, where applicable:

A summary of all nonfatal loader errors.

A list of all COMMON block length clashes.

The nature of the error that activated PMD.

The array-dumping parameters selected and the ﬁeld length required to load and run the user program.

The activity of each ﬁle used by the user program at the time of the error.

The overlays in memory at the time of the error.

The location of the error in terms of statement labels and line numbers, if possible.

An annotated register dump; an attempt is made to associate each address register with a variable or array

referenced within the routine in which the error occurred.

An alphabetical list of all variables and their values, accessible from the current routines.

A printout of arrays according to speciﬁed parameters.

A message-tracing call beginning at the previous routine and ending when the main program is reached.

A completion message upon reaching the main program. *^s

Variables are printed alphabetically. The column labeled RELOCATION is left blank for local variables. It contains

the block name for COMMON variables and F.P. nn for formal parameters, where nn indicates the parameter number. ^^

In addition to being printed as numbers, INTEGER variables are interpreted as masks or characters in H, L, or R

format. In character representation, binary zeros are converted to blanks within a word, but a word with binary

zeros at each end has the ﬁrst binary zero printed as a colon. "^

,_^~^y

8~54 60497800 G •

^^v

jm*

/P*v

The column headed COMMENTS ﬂags undeﬁned local variables as *UNDEF, which indicates a potential source of error.

Variables passed as parameters to the previous routine in the traceback tree are labeled PARAM nn in the COMMENTS

column. The COMMENTS column contains F.P. nn where the same variable occurs more than once in an argument

string; nn points to the last occurrence. Constants passed in the previous routine are also printed at the end of the list

and given the symbolic name CONSTANT. Untraceable functions and subroutines passed as arguments are printed.

Full checking is carried out on subroutine or function arguments, and a warning message is issued if:

A routine is called with the wrong number of arguments.

A type conﬂict exists between actual and formal arguments.

The argument was a constant and the called routine either treated it as an array or corrupted it.

A conﬂict in the use of EXTERNAL arguments is detected; note that the results given for EXTERNAL arguments

can be imprecise because several utilities can reside within the same routine and PMD cannot differentiate

between them. For example, both SIN and COS reside within the routine SINCOS=.

A warning message is also issued if a real variable contains an unnormalized value, for example, integer.

For batch jobs, the dump is written to ﬁle OUTPUT. For jobs executed from an interactive terminal, the disposition

of the dump is determined by options speciﬁed on the execution control statement (typically LGO) as follows:

LGO,*OP=option [option] [option].

where option is one of the following:

l^v T A condensed form of the dump is displayed at the terminal.

A The variables in all active routines are included in the dump. An active routine is a routine that has been

executed but is not necessarily in the traceback chain. This option is valid for batch, as well as interactive,

jobs.

00^\

F A full dump is written to the ﬁle PMDUMP when the job is executed with the ﬁle OUTPUT connected.

This option is valid for interactive jobs only and is the default if the *OP parameter is omitted.

PMD can be used with overlay programs. In this case, only variables deﬁned in the overlay currently in memory are

dumped. The overlay numbers of the current overlay appear in the PMD output.

PMD output produced by a program compiled under a given optimization level can differ from that produced by the

same program compiled under a different optimization level. This occurs because different optimization levels

generate different sequences of object code. At the actual time of an abort, the machine instruction being executed for

a speciﬁed optimization level might be different from the instruction being executed for a different optimization level.

Variable values printed by PMD might differ for successive executions of the same program on certain computer systems.

This can occur on systems with parallel functional units such as the 6600, 6700, CYBER 70 models 74 and 76, and the

CYBER 170 models 175,176,750, and 760.

In addition to the ﬁle ZZZZZMP created by the loader, PMD requires a ﬁle called ZZZZZSY, which is created

by the compiler. If a program that uses PMD is to be compiled, loaded, and executed in separate jobs, those

ﬁles must be saved and must be available to PMD during execution.

PMD can, in most cases, trace back from the point at which an error occurs to the main program. Traceback

occurs repeatedly through the calling sequence of routines until the main program is reached. In PMD trace-

back, PMD prints the names and values of the variables in the routine in which an error is detected.

0&\ 60497800 H 8-55

Traceback is abandoned under the following conditions:

PMD discovers a routine name that occurred earlier in the traceback. For example, the descendent of an

ancestor routine has called the ancestor again.

/ﬁt^k

Traceback ﬂows through a routine that was not invoked by standard FORTRAN calling conventions, such

as those used in a function or subroutine reference. For example, if an abort occurs while an input/output

operation is being performed in a CRM routine, the traceback is abandoned because CRM uses the "^

sequence:

SB6 address ^

JP B6

to perform the equivalent of the standard calling sequence: /S3?^

RJ address

that a function or subroutine reference generates.

/S g \

Multiple entry points exist in a system resident routine. If multiple entry points are in a system resident /SS^

routine, traceback might be abandoned because PMD continues traceback from the ﬁrst entry point (which

is assumed to be the last active entry point for the routine). If traceback continues at an entry point

other than the last active entry point for the routine, traceback continues up a calling chain other than the ■*

one that produced the error. CPU.SYS is an example of a routine with multiple entry points that is

subject to this condition.

When traceback is abandoned the following error message is issued:

***TRACEBACK ABANDONED ^

The formats of the optional PMD subroutine calls are as follows:

CALL PMDARRY(i) '

CALL PMDARRY(ij)

CALL PMDARRY(i,j,k)

The last subroutine call listed causes dump of arrays to be limited to elements whose subscripts do not exceed i, j, and

k for their respective dimensions;!, j, and k represent the ﬁrst, second, and third dimensions, respectively.

If k is omitted, three-dimensional arrays are not printed. If j and k are omitted, two- and three-dimensional arrays 1

are not printed; only one-dimensional arrays are printed.

>*^N

Array dumping parameters can also be speciﬁed on the LGO call card. The three formats are:

LGO,*DA=I Corresponds to call PMDARRY(I). ^\

LGO,*DA=I+J Corresponds to call PMDARRY(I,J).

LGO,*DA=I+J+K Corresponds to call PMDARRY(I,J,K). ^

where I, J, and K represent the ﬁrst, second, and third dimensions, respectively.

8-56 60497800 H

/|KN

/I-N

If neither CALL PMDARRY nor LGO,*DA= is used, the default array dimensions of I, J, and K are assumed

to be 20, 2, and 1, respectively.

Once PMDARRY has been called, the established conditions apply to all program units in the user program.

Any number of PMDARRY calls can be included; the most recent call determines the effective conditions.

Example:

DIMENSION RAY (10,10,10)

CALL PMDARRY (3,4,1)

Array elements are printed with the ﬁrst subscript varying fastest and with a maximum of six values per line for real,

integer, and logical arrays, and a maximum of three values per line for double precision and complex arrays.

The following twelve elements of array RAY will be printed:

(1.1.1)(2,1,1)(3,1,1)(2,3,2)(2,2,1)(3,2.1)

(1,3,1)(2,3,1)(3,3,1)(1,4,1)(2,4,1)(3,4,1)

If all the requested elements of an array have the same value, PMD will print the message:

ALL REQUESTED ELEMENTS OF THIS ARRAY WERE ....

If several consecutive elements of an array subblock have the same value, PMD will print .the message:

ALL THREE ELEMENTS WERE ....

CALL PMDDUMP causes a dump of variables in the calling routine, not at once, but when an abort occurs or

when PMDLOAD or PMDSTOP is called. PMDDUMP and PMDLOAD or PMDSTOP need not be called from

the same routine. The dump includes an analysis of all active routines that have called PMDDUMP. These

active routines have been executed but are not necessarily in the traceback chain. Following an abort or call

to PMDSTOP, all routines in the traceback chain are dumped. A limit of ten successive calls to PMDDUMP

is imposed. The tenth call to PMDDUMP is converted to a PMDSTOP call.

CALL PMDLOAD causes an immediate dump of variables in the calling routine and in any routines that have called

PMDDUMP. Program execution continues normally after the dump unless PMDLOAD is called more than 10 times,

in which case a nonreturnable call to PMDSTOP occurs.

CALL PMDSTOP causes an immediate dump of variables in the calling routine, all routines in the traceback chain,

and any routines that have called PMDDUMP. The job is then aborted. Programs cannot recover from a call to

PMDSTOP.

An example of Post Mortem Dump output follows:

60497800 H 8-56.1/8-56.2

■^

/$PN

0H&S.

^8? $^*\

rTiKXTfxrTrXTffxxn^Txx

uju*uj^ujuj^>ti;ujujujuitij^^U'UJU;u.uju*UiUi

Ulltf,

ujiu

I T .

<rir«EH«r>wr-oO'r><

o o e o o

V*VM/<'/,V)UJU.'UiUIVfV<V*l*JVl

u.u.u.u.u.zzzzuu.u.zu.

U'tUUHUtUMfl— MI1.U. lU—UI

Of'fcoccrau.u.u.u.ercrcru.at

ttU^UJlU UJ

e c e e o

- « Z 3 u -

r — t u

f— -JfX

~ _ » 3

-.cr-t—t*.

/PSN

—i- w<r—ma o— •

« ^^.orv—>-r- «o • oaso

TTltl 0fO--*«J «»—.*_HJJf- O-*"—

<IL *4^»OUjaf • • UJ«/*yl»l».*.*l/l

tt » JU«»Okj42Z»K-l

oa oJifiiN »T*tx«ifmec

cr o«»Lui-i-----«u.o_i«T-«fc-

are mjw _i«~—~» ,H(fiifi«rx

as uj«^Z'*«»«m■ ■ ■ i ■ ■ ■ « u j

a JU2C> v^^—«Mf*l*-*

c o a . * - u j X r - * * a c r r a i c r x a a i ( - l a .

—UJ rOIt3h«444«fl4irjao

xx co"-oo<a>»»»«4t-r

«-i- u jcuiuc 4ua^MMi»Nir uvtiu

fVCM

ee<rtrtr

fV—o*

«j-»^f»»

TfOf

orof

U l U l

o T-i -ftatataacr-iat

T ** uiujuJujUIUIUI^lLi

Ul -J Ul JOOUOOUUO

ac a^-jaiuujujuuujuj—u>^^_i

ui u> X4ri-^Ki->-i-OK<i«

u. ui i-ujiuoi rriTTCtiuujui

ui o aatUHMfM.MMjM«a(a

om o-"4

nrui •KU1

Ul uiuia Ull/l

aa.o>- XIM -13

>• >Ulf- f—«>

»— »—f— OIM ZJT

ZO CO

—SE Ul E l O

Vt • •••• to em

0 4 U r > o

J _ i - M M V « M t — V t — I O * t r i S O N

O oa V>^4>f4Nm«riMXH ^3i4«1 « t- KY JTuiZUim

co •-* -««acafcraccirararor waCf* 3awui or o s* ou »-_.ui-j.o

r at"« <xrr*atac* + M*Mm**et v\a.t-a.a. t- r>-irt oo wio ->uj

V M» morac>«o>»»«rx WZ344 vim 3T4 _i_ <_>♦» - c d

vt Out ui«<uMMwwMH^ra(0(> ChBp-i- _juj u,»-C coco wec«or«

a j « « k . 4 w u i - i

c o z a t u i X v i f c - o u .

vn> «o«t-fr^ir\e^<\jcn«^«r>4r o-4-o<r ac ar o «»iou.x

atf* — -fin.O'O'O.e-o.Atn.o.cjtfi.o ui inmin ui « *c r-xcsu

f-«M OT m<Mf\IIMr\lf\lfMf\JfMf\lfMfMf>l _ 0-.0 f-l tC ««/»atOV»

Z 4 ■* 4)'0<o<e«<e4)4)««««4) t- «v^»v w ar o i—

u i a » u . u i * • o v »

60497800 F 8-57

y ^ ^ w

-Jf-f_l JJ -JM-JMM

0.-J0. a a a a u i a~*a^-»

•••-ar rxrrtrfrff 1-hIKrt

4 V - 4 4 44444VI4I/1ISI _l VIVf4VI4 4

MU.-M xyyHrxuiKlbili —J UJUIMUJMM

Ujfr —Ul — . u - u r u i i - u . ^ uj f c * * — ( • i - t t - * - u . t - u j u i

z s

>IJJV1

mxm

u j v - X

o » -

aU . Z

ac 0 ( M •

e1—

X>**x

aOJUI

T t - O

a▶—ac

"" »7UI

-i •U.UI

-i Ul .UfO

4-a •o

Oaazuv

V» O m

Of XUIt

Ul Ul a»o

o(9 —4UI

Ul Ch«

l/l ac M M M

Z3 «V V»V»M

Tfc. _iOi-3

Z) _ » a a

*> or 4*>uiac

IM Ul O 3

>■<* ▶-«JZ

or or c* 4C40

at Ul SM/1

4ar t_>«- M

IZOVtOC

1▶-MOM4

ZVI orOT-

Ul u ' O i - r

VUl i s o o

4or OOU.U

at •o •—OV>

or orOSUI

4CM VI <o u> -04T

Veot—

•Man ri • O U I

^oror u. ><car

• 4 4 V) a r * i m O

4 ) f - » U l

U. U.U.

Ul UIUI

e o o

o e o o

i/ii/iuiviviujvii/icoviiuiui/i coo

UUZU.UZU.U.*4U7?U. ﬂ

IX-Ui—UlUI*—>UIU> UJMMUI

arctuacoruacar aruu._e

u i u i u j u i

o e c o

ui ui

- > » T

— e . i - o r « - > o x . a r m

V I " « 1 9 4 O — O r « e *

4 z z z v t a — o u i o u . o

O r C ? — U l » » M O 1 » I Z I —

> - - . > - _ J 4 - » T • » 4 K M O

xi«« _; ot- (M »-ac>-fc.ui O

ui—cy _» o O w4J 4iuUil/*v c»

c a r o C o * 4 9 f W t v o a — o f

u!iu4 o _i-> m«» a oro _icu»-jui z,—

z a r > - - j c c • « . 4 o - % u . a x u i o

fc-a <o >.—../j*. >vi*4 — uj uio vi •

ft* ri w-» MO* Zt-UI_l ceo

S u j u i Z X — O f ■ »M Mwagiui 4uiX

c_o o a zui4 on—i >ov>z_j4ce—m .oc

vcui x tc-h't .ocarm 1—41— ■* . + _ j - » 3

*3r- ac r«i-B("(fi44«i m.>ak.x>«j~f-o

: o Z O O - u a a i i » m z < 0 ■ • fCauiZ

vom o uairs.H>iHoa^^M4>7><xuHVUi

0 _ t « • • •

m o a a a a

m c . . . »

4-1 U.U.U. U.

oco

— O f U l

I M U J t «- O C

■ U . I M f V > - U .

O C U l 4 4 4 U l

— o r o r e v e r o

a r a c a r z

v > v > o z o r

QUI ZtﬂtftCO M UJ-OO

41- UJtVMM _l U.<Uf->

u j m a t u i

o r o r U l u » o r

* J » U . U l

U l o

arat

orac

arm

yft^^v

OOHri 19-4

orotorororor or

UlUlUIUIUIUIUI Ul Ul

ui —lOtSOisiflo 19 ujf— a.

a_iceuiuiujuiuiuj u>_i_j-j os v

V4 3l-Kf'r-l-p> h444 OlLr> f

▶- U I O Z Z Z Z Z Z Z U J U I U I C Z C

aeHHHHHM —arococ su.

C90

X U I

u.i/1

-13

:v»

V I u .

fat

T O C S

O U l 4 ( 9 O

O ▶_(_)«> 1/14 fc-ZO"M

Ul f*4» Z»- f-*« 4 *-? XuiUItt*

v i u i u i w a r u 4 - j 0 0 0 h j j «

3 v » o . a v > o x a r e 0 0 a o v i e u i

Mar UI44 VKOVIM 3_i »->oo -j-i o*--cp

zasx r^i> -I40VI u.11. z-*-4 ccco wo4«

4 4 U l ^ - 4 0 t _ l

a a e x u i a e z v > f c > c 9

▶ki <o«ooooe ohoo or x ui*»m o hioz

O C M 4 9 ) 4 > f » f * » 4 9 U l U l f - l b - M I O X W > - O T O

* • O C _ » • - I 4 X « V K i M

- . V f M

V-4 4V<ﬂ4)

4oc arorocat

x T H o r 4 o r o r 4 4

>> mm viar> 44»>

(/I QUI UI4SHHHHH

Ul M V I VI

>c^\

8-58 60497800 F

M O M U l U l

f M I u . w M

I U l Z m M

• V M 4 Z X O

t r o o r m m

«•» O 4 U, U.

cc in a z z

M 1 * 1 M M

in cm u.

i M j f t r u i u i

• y f a . >

CM N> V» M M

m w> v* m t—

o O- ui M M

m 4'ac vi vi

> . « c O O

> i a a a

f * > i * - o x

M I * . O M

m «r> o u.

i m m o z

M O M

- 4 - O

K O U l

n o r *

O M

o m

O M

. 1/1

« m o *

> o a o

>■ M 4 > u -

< o r o i < v n r

a r 4 a a r a

o r > 4 4 r »

/$^\

CO CC CC GQ CC CO OJT CO cocococccccccccc

UJ r^oo4>Mooo OOmOMOOO

OOOONOOO OOOOl-OOO

4••-oOf-if^-ooo oomof»ooo -1

•OOOmOOOO © O f - 0 . 0 0 0 0

O OOOCMOOOO oooooooo

CuiOOO<0OOOO ec ujoooooooo

VI 3000<n0000 CT300000000 4:

v-jooomoooo u. -joooooooo f»

44ooomoooo 4 0 0 . T O O O O O 1-

>ooo-.oooo >oooooooo 19

ooiriMoooo U: oooooooo

-tooor-oooo VI -JOOOOOOOO UJ

4 0 0 M . O O O O O 400000000 _l

h-OO^-MOOOO MOOOOOOOO

oooomoooo ooooooooo VI

OOOOMOOOO V>C000004300 -J

00©«r©h-f>-0 ar r-6©00-«i».o UJ

000430r>-*.Q Ul r—OOOOCMf-O

oo*ooi«-r-o h-OOOO^f-O u.

OOOOOmmo «/l cnOOOO-.«no

Orfrvf-ftn^N Ul

OOM«\.c«*iri«h-

-1

V U l

— VI

tr re 4 ~ . r * f a t i w _ c

0H^\

O U .

4 U l 1 9 Z or

C " » X m

-1 UJ

-1 - 1 X or

- 1 4 »—

4 Z ««. «

I- or Z X

4 U l r i u j

U M x -i Ul

1 ■»* x • VI

Z UJ TT CM

r» i > tn

z o ■«» CM

•JJ Ul

19 m 19 co

X U .

— VI •-. ■

31 m * X Ul

«"• 1— e* »- 1-

- 1 4 -1 19

-J V t -1 z

e z O U l

U . 3 u. _

O M X

• - o 3 e

v i t u a x

4 V » * » M

- I 3 1 - - »

Z 3 u i

3 a a r

v»

O 4 * -

X » - 3 4 } m

3 a ui ui

x a m a a

4X344

o c M O f f

V I O 4

0 9 • » O

e 4 m

MOM m O M m m

o r o r or

o O

VI O o. • . o

4 X f/> V» 4

VI Ul C o

V) 1— f- O O M

Ul VI Z Z VI

oc M 4 «M

o-J VI VI -J

oU.

4or O er ec

Ul X X Ul Ul

XJf-inMMM

O ui vt x x Ul Ul

4ca X Ul M M X

MX 4 0( M m 4 M M

Xvxaisoi or

o vi 4 a

oa. 4

jocjoeddcisd

c/i 4 * ' r » ^ t n c * t . f - * a 4

or » 4 M 45 N 4» -r -r ■».

>-1- « r» M M 4) I*- P. 4>

VI OOOM M O O O O

OOOOOOOOO

Ul O O M «M m 4» m 4) f*.

1Ul 4 4

oc 4 4 4 4 4

— ocimooocMO

4 m C M H H

> 4 9 O - O

_COCOC0CDCCCO0CD

40tnooo«M~ro

>OOOOOMOO

OMOOOOOO

MO^JSOOOM©

OOOiwOOOmO

ooooooooo

m •

r » c

— O 4 - a .

4 J ﬁ a k

. r ^ * ﬁ 4 m

4 : f f » — - » - j

. P U o — — 4

■C 4) ri -i —

„ • { * ~ » « - I —

C M I —

>■ 4 - J - >■

4 M 4 4 u . ^ >

Or 4- ar a r 3 O

or . * v a r — i 7

4 M 4 4 4

5 » •

rllﬂlﬂ"

a . a . a . 7

u . u . u . X

u i a o c a c i

x u i U J u j u i

m J H O O

M UJ —I * UJ UJ UJ I

3 a 4 3 » - f t - i

O > - u i O Z Z Z

ac m tt o M M M I

Q MM M X M

0&\ 60497800 F 8-59

/*"~**SL

m — z a 3 m

NOOOCZ

m m o o r m

_ l O u . u .

M C T o r 4

4 a c

- i a c

O 4

a u. u. u.

I Ui Ul

l c.fj o

I Ul Ul

u i O

a m

o 3

M O

M Ul

4 4 4

Ul Ul Ul

> or or oe

:m -» xo. 3 m m

>- V M 4' VMIM«n«Mf4 SC

4 4 o r o r o r o r o r o r o r o r 5 » * v >

a t e r 4 a c o r 4 4 4 4 4 v > 4

O t O t > 4 4 > > > > > 4 X

V I V I

z X

o o

c v t

4 4

/"S^\

/f^^V

I (

•D SO

4 * *

* »

# JJ>

CM IM

A A

at or

ui ui

o o

u i u i f a .

f f X r ,

z z 4 c r

m m o r o r

4 ) O — X U l 1 z

• O c m m r w CM O • * M 01

4 5 — O • f M M O o> t* r - i -

• 0003-1 o o 0 » 3

K r n O M O 4 m o 4f m o UJ

— w o — a t m rM m ec —1

O M in _l

O O m cm e

O U l X UJ 49 Ul

V > - O V - J M fW VI 0 » _ l

4 4 . 4 - 1 -J 4 5 _ Ul

o r o t m a t 4 » - » 4

ar or nuaij U. CMO

4 4 m 4 T t a f t

Ul Ul U l o z

x 3 X M

a r c

u.*

V I K .

u i m x

k i o r u i

M 4 U l

MU l V I ac Ul 4 9 M

Ul O X • V I

U l M Ul 4) u .

xatMototatotoi_ior4 r- at ar • 3

X Ul lb X Ul Ul ID 111 ID 4 IV r> V U l U l Ul r^. e » f * l

4 — O m O O i I O O o O X 4 O O — U l

OC Ul Jjaui uiuiuiuiuiMuio_vi-i -i x ui uj m o r CM

o o. 44XMXH-MMMh-0MO4 4 X 4 M M >»

o > » U l U I a X 4 Z X . X X X O X U l U l 4 U l X X 4 - 1

or » - a c a t o m m m m m m j m u j o r u i o t a c MMM at a t — i or

a V I _ -J 1/1 OC 4 oc

m c a CO

4 4 ac

X M l/l >■>•

Xu i a t ac Ul X f - l -

t - 4 O T X ar ar

f 4 N ^ M N > O 4 4 ot ac

vt >VM4 VHNn#Hf4 X M M

Ul Ul 44atot»-ocatotatararirr-v»*-or^-MfﬂAM

-IX « a f 4 0 ( M e r 4 4 4 4 4 V I M 4 M 4 M o r M X X

c e 4 a r a t > 4 u i 4 > > > » » 4 u j x Ul > u i 4 u i e o VI

4X 44oMXMMMMM^xzatza(Z>xoo

jttcfSSsV

i .

j * t f ^ V

0ti$$\

8-60 60497800 F

0mS,,

/SSN.

/0^\

00^\ V " . V )

I — I -

X X

U l U l

X X

U l U l

- I - I

u i u i

V I v >

u i u i

o t — —

0$^\

60497800 F 8-61

I"$~\

/I-*\

DEBUGGING FACILITY

The debugging facility allows the programmer to debug programs within the context of the FORTRAN language.

Using the statements described in this section, the programmer can check the following:

A r r a y b o u n d s F u n c t i o n r e f e r e n c e s a n d t h e v a l u e s r e t u r n e d

Assigned GO TO Values stored into variables and arrays

Subroutine calls and returns Program ﬂow

The debugging facility, together with the source cross reference map, is provided speciﬁcally to assist the pro

grammer develop or convert programs.

The debugging mode is selected by specifying D or D = lfn on the FTN control statement. This option auto

matically selects fast compilation (OPT=0) and full error traceback (T option). If any other optimization level

is speciﬁed, it will be ignored. Speciﬁcation of both D and TS results in a fatal error. The following examples

are equivalent:

FTN (D)

FTN (D=INPUT,OPT=0,T)

FTN (D, 0PT=2) OPT=2 is ignored, OPT=0 and T are automatically selected.

Debug output is written on the ﬁle DEBUG. The DEBUG ﬁle, which must be on a queue device, is given print

disposition and printed separately from the output ﬁle upon job termination. To obtain debugging information

on the same ﬁle as the source program, or any other queue device resident'*' ﬁle, DEBUG must be equivalenced

to that ﬁle in the PROGRAM statement.

Examples:

PROGRAM EX (INPUT,OUTPUT,DEBUG=OUTPUT)

Debug output is interspersed with program output on the ﬁle OUTPUT.

PROGRAM EX(INPUT,OUTPUT,TAPEX,DEBUG = TAPEX)

Debug output is written on the ﬁle TAPEX.

The following control statement sequence causes the debug output to be printed on the output ﬁle at termination

of the job. It is not interspersed with the results of program execution.

FTN(D)

jjﬂl^ LGO.

REWIND(DEBUG)

COPYCF(DEBUG,OUTPUT)

y^®V EXIT(S)tor EXIT.ft

REWIND(DEBUG)

COPYCF(DEBUG,OUTPUT)

"•"NOS/BE 1 and SCOPE 2

"•"•"NOS 1*

"•""•""•"NOS/BE 1

J0\

60497800 E

Abnormal termination

9-1

■^^v

In debug mode, programs execute regardless of most compilation errors. Execution, however, terminates when

a fatal error is detected, and the following message is printed:

FATAL ERROR ENCOUNTERED DURING PROGRAM EXECUTION

DUE TO COMPILATION ERROR

Partial execution of programs containing fatal errors allows the programmer to insert debugging statements to

assist in locating fatal and non-fatal errors. Partial execution is prohibited for only four classes of errors:

Any declarative error (any error encountered before at least one valid executable statement is found)

Any fatal compilation error (deﬁned in Appendix B)

Any missing (undeﬁned) DO termination

Any illegal transfer into an innermost DO loop that is not an extended range loop

When a program is compiled in debug mode, at least 15000 (octal) words are required beyond the minimum

ﬁeld length for normal compilation. To execute, at least 2500 (octal) words beyond the minimum are re

quired. The CPU time required for compilation is also greater than for normal OPT=0 compilation.

If the D option is not speciﬁed on the FTN control statement, all debugging statements are treated as com

ments; therefore, it is not necessary to remove debugging statements from a program.

All debugging options are activated and deactivated at compile time only. This compile time processing is

not to be confused with program ﬂow at execution time.

Example:

PROGRAM TEST (OUTPUT,DEBUG=OUTPUT)

GO TO 4

C$ (DEBUGGING OPTION)

4 CONTINUE

END

Even though a section of code may never be executed, the debugging options are processed at compile time

and are effective for the remainder of the program. In the above example, the code between the GO TO

statement and the CONTINUE statement may never be executed. However, debugging statements between

these statements are processed at compile time and are effective for the remainder of the program, or until

deactivated by a C$ OFF statement.

9-2 60497800 C

/ ^ 3 \

DEBUGGING STATEMENTS

/pp^\

C$ids(p., pn)

Type of option, beginning after column 6: DEBUG, AREA, ARRAYS, CALLS,

FUNCS, GOTOS, NOGO, OFF, STORES, TRACE

Argument list; details extent of the option, ds (not used with NOGO, GOTOS; required

for AREA, STORES; optional for other options)

/fp3Sy

CONTINUATION LINE

1 6 7

I<Pr rPj

Debugging statements are written in columns 7-72, as in a normal FORTRAN statement, but columns 1 and

2 of each statement must contain the characters C$. Any character, other than a blank or zero, in column

6 denotes a continuation line. Columns 3, 4, and 5 of any debugging statement must be blank. A maximum

of 19 continuation lines is allowed.

Comment lines may be interspersed with debugging statements. The statement separator ($) cannot be used

with debugging statements. When the debug mode is not selected, all debugging statements are treated as

comments.

Example:

C$ ARRAYS (A, BNUMB,Z10, C, DLIST, MATRIX,

C$ *NSUM, GTEXT,

C$ *TOTAL)

■fJSN

ARRAYS STATEMENT

/c$"

ARRAYS (aifa0f...,a )

1 ' 2 ' ' n

ARRAYS

a,,...,an array names

60497800 C 9-3

The ARRAYS statement initiates subscript checking on speciﬁed arrays. If no argument list is speciﬁed, all arrays in

the program unit are checked. Each time a speciﬁed or implied element of an array is referenced, the calculated

subscript is checked against the dimensioned bounds. The address is calculated according to the method described in

section 1. Subscripts are not checked individually. If the address is found to be greater than the storage allocated

for the array or less than one, a diagnostic is issued. The reference then is allowed to occur. Bounds checking is

not performed for array references in input/output statements, or in ENCODE/DECODE statements. In a subprogram,

the bounds that are checked are those in effect in the subprogram, including variable dimensions.

Example:

PROGRAM ARRAYS <OUTPUT,DEBUG=OUTPUT)

INTEGER AC2)* B<*-»>« C<6>» D(2t3*<»>

PRINT 1

1 FORMAT<«0 ARRAYS EXAMPLE'///)

* T U R N O N A R R A Y S F O R A R R A Y S A A N D D ^

4»

Cl ARRAYS (A* 0*)

* A(3) IS OUT OF BOUNDS AND ARRAYS IS ON FOR A* SO A DIAGNOSTIC ^

* IS PRINTED.

♦>

A(3) * 1

* 8(5) IS OUT UF BOUNDS BUT ARRAYS IS NOT On FOR 8» SO NO

* DIAGNOSTIC IS PRINTED.

B<5> - 1

C<2) - A(A(3)>

* EVEN THOUGH A(3) WAS OUT OF BOUNDS* THE ASSIGNMENT TOOK PLACE.

* A(AU)) IS EQUIVALENT TO A(l). THIS SUBSCRIPT IS IN BOUNDS*

* HOWEVER THE REFERENCE TO A<3> WILL CAUSF A DIAGNOSTIC

D(-5*0*6> =99 ***)

* FUR THE ARRAY D(L.M*N) THE STORAGE ALLOCATED IS L * M » N.

* THE SUBSCRIPT FOR THE ELEMENT D(I»J*K) IS COMPUTED AS FOLLOWS

* < I ♦ L«(J-I ♦ MMK-1)))

* FOR THE ELEMENT U<-S*U»b) IhE SUBSCRIPT APPEARS TO

* BE OUT OF BOUNDS btCAUSE Tht INDIVIDUAL SUoSCRIPTS ARE OUT

* OF BOUNDS. HOWEVER* 22* THt COMPUTEO ADDRESS* IS (_ESS THAN .-^

* 24* THE STORAGE ALLOCATED. AND NO DIAGNOSTIC IS ISSUED. ^

* T U R N O N A R R A Y S F O R A L L A R R A Y S A

C$ ARRAYS

* W I T H T H I S F O R M A L L A R R AY R E F E R E N C E S W I L L B E C H E C K E D . T H E R E w I L L / m ^

* BE DIAGNOSTICS I OR B<5>> C<-1>. AND D(O.O.O). BECAUSt A(2) 1

* IS IN BOUNDS AND A<<0 IS IN AN I/O STATEMENT* THERr WILL BE

* NO DIAGNOSTICS FOR £ITHER OF THESE REFERENCES.

A(2) * 1

B<5) = 2 ♦ C(-l)

D<0*0»0) = 1

P R I N T 2 * A ( 4 > ' _ .

FORMATtlX* A10) "y

E N D '

9-4 60497800 C

/jls\

ARRAYS EXAHPLC

/DEBUG/ ARRAYS AT LINE

/0E8UG/ AT ITNC

/OEOUG/ .. AT Line

/OEBUG/ AT LT'ir.

itxrtaic,/ «t i tnp

u - t h *

?0- T»"£

*.r- TMt

W - T H "

i.a-j«£

■;ii3?':».:',T

"5!J'»sr<i^T

'.uti-.r^iPT

SUP<>Mt"»T

SUB'.CP.IPT

v » l t r

viL'ir

v»ui«"

v»u«r

VALUE

p«"

-1

-B

!»

I'M

I»1T

»?°»Y

»o»»Y

ARUAY

""VCEtl'S

r»rrios

EXCEEDS

fxfErns

EJ>CEEOS

OI-tNSinurO

PiriHSinm n

oif-tNsinnto

nifENSioNfn

DIMENSIONED

» m m o r

pound or

qouNO of

bound or

BOUND OF

<•

CALLS STATEMENT

c§T

c$ CALLS (a 'an>

CALLS

/fss

/*JS!\

a,,...,an subroutine names

The CALLS statement initiates tracing of calls to and returns from speciﬁed subroutines. If there is no

argument list all subroutines will be traced. Non-standard returns, speciﬁed in a RETURNS list, are

included. To trace alternate entry points to a subroutine, either the entry points must be explicitly named in

the argument list, or the form with no argument list must be used (all external calls traced). The message

printed contains the names of the calling and called routines, as well as the line and level number of the

call and return.

A main program is at level zero; a subroutine or a function called by the main program is at level 1,

another subprogram called by the subprogram at level 1, is at level 2, and so forth. Calls are shown in order

of ascending level number, returns in order of descending level number.

level 0

level 1

level 2

MAIN -▶call

return SUB A - call

return SUBB

For example, subroutine SUB A is called at level l and a return is made to level 0. SUB B is called at

level 2 and a return is made to level 1.

/^*V

60497800 A 9-5

Example:

PROGRAM CALLStOUTPUT»OEBUG=OUTPUT)

PRINT 1

1 FORMAT<*0 CALLS TRACING*)

* TURN ON CALLS FUR SUBROUTINES CALLS1 AND CALLS2

C* CALLS<CALLS1» CALLS2)

X = 1.

CALL CALLS1 <X.Y)» RETURNS (10)

10 IF (X .EQ. 1.) CALL CALLS2U)

CALL SUBNOT

CALL CALLS1E (X*V) ^

• DEBUG MESSAGES WILL BE PRINTED FOR CALLS TO AND RETURNS FROM

* CALLS1 AND CALLS2. SINCE l"HE CALLS ARE FROM THE MAIN PROGRAM*

* THEY ARE AT LEVEL 0. THE CALLS TO SUBNOT AND THE ALTERNATE

• ENTRY POINT CALLS1E ARt NOT TRACED BECAUSE THEY DO NOT APPEAR

• IN THE ARGUMENT LIST OF THt C* CALLS STATEMENT.

* TURN ON CALLS FOR ALL SUBROUTINES

C» " CALLS

CALL SUBNOT ^%

CALL CALLS2CX)

CALL CALLS1E (X*Y)

* DEBUG MESSAGES WILL BE PRINTED FOR CALLS TO AND RETURNS FROM

• SUBNOT* CALLS2* AND CALLS1E* SINCE ALL CALLS ARE TO BE >^*\

• TRACED.

E N D J

SUBROUTINE CALLS1U,Y), RETURNS(A)

Y = -X

IF (Y .NE. X) RETURN A

RETURN

ENTRY CALLS1E

RETURN

END

SUBROUTINE CALLS2CX)

CALL CALLSHXtY)* RETURNSC5) sms*

5 R E T U R N 1

END

SUBROUTINE SUBNOT

X - -1.

CALL CALLSKXtY)* RETURNS(S) ^

5 R E T U R N ^

END

9-6 60497800 H

/DEBUG/ CALLS AT LINE 9- ROUTINE CALLS1 CALLED AT LEVEL

/DEBUG/ AT LINE 10- ROUTINE CALLS1 RETURNS TO LEVEL

/DE8UG/ AT LINE 10- ROUTINE CALLS2 CALLED AT LEVEL

/DEBUG/ AT LINE 11- ROUTINE CALLS2 RETURNS TO LEVEL

/DEBUG/ AT LINE Zk- ROUTINE SUBNOT CALLED AT LEVEL

/DEBUG/ AT LINE 25- ROUTINE SUBNOT RETURNS TO LEVEL

/DEBUG/ AT LINE 25- ROUTINE CALLS2 CALLED AT LEVEL

/DEBUG/ AT LINE 26- ROUTINE CALLS2 RETURNS TO LEVEL

/DEBUG/ AT LINE 26- ROUTINE CALLS1E CALLED AT LEVEL

/OEBUG/ AT LINE 27- ROUTINE CALLS1E RETURNS TO LEVEL

AT STATEMENT 10

In this example, only calls from the main program are traced. To trace calls from subprograms, a C$

CALLS statement must appear in the subprograms.

FUNCS STATEMENT

/c$

/c$ FUNCS (a an>

FUNCS

If no function names (a,,...,an) are listed, all external functions referenced in the program unit are traced.

Alternate entry points must be named explicitly in the argument list, or implicitly in the C$ FUNCS

statement with no paramenters.

Function tracing is similar to call tracing, but the value returned by the function is included in the debug

message. Each time a speciﬁed external function is referenced, a message is printed which contains the

routine name and line number containing the reference, function name and type, value returned, and level

number. The level concept is the same as for the CALLS statement.

Statement function references and intrinsic function references are not traced, nor are function references in

input/output statements.

60497800 C 9-7

Example:

The following program, VARDIM2, illustrates both the C$ FUNCS and C$ CALLS statements. All

function references in the main program are traced because C$ FUNCS appears without an

argument list; references to functions PVAL, AVG and MULT and the values returned to the main ^^\

program (level 0) are traced. All subroutine calls in the main program are traced also because a C$ CALLS

statement without an argument list appears.

Function references within the FUNCTION subprograms PVAL, AVG and MULT are traced since

C$ FUNCS statements appear within these subprograms. If no C$ FUNCS statements appear in the

s u b p r o g r a m s , o n l y m a i n p r o g r a m f u n c t i o n r e f e r e n c e s w i l l b e t r a c e d . / - a a ^

PROGRAM VAR0IM2COUTPUT,TAPE6=0UTPUT,DEBUG=0UTPUT>

C THIS PROGRAM USES VARIABLE DIMENSIONS AND MANY SUBPROGRAM CONCEPTS

COMMON X(i»,3l _

REAL YC6) ^%

5 E X T E R N A L M U L T , A V G

PVALSF<X,Y) = PVALCX,Y>

C $ C A L L S ^

CALL SETCY,6,0.1 ^

CALL I0TA(X,12)

10 CALL INC(X,12,-5.)

C ALL EXTERNAL CALLS ARE DIAGNOSED. ^

C$ FUNCS

15 AA = PVALSF«12,AVG)

AM a PVALSF<12,MULT>

C PVALSF IS A STATEMENT FUNCTION, SO THE FUNCS STATEMENT DOES NOT

C APPLY TO IT ANO NO MESSAGE IS PRINTED. HOWEVER, THE EXTERNAL -,

2 0 C F U N C T I O N P V A L I S R E F E R E N C E D W I T H I N T H E C O D E F O R P V A L S F, ™

C AND THOSE REFERENCES ARE OIAGNOSED.

C MULT AND AVG ARE NAMES AS ARSUHENTS TO PVALSF, HOWEVER, THE

C FUNCTIONS ARE NOT ACTUALLY REFERENCED AND MESSAGES ARE NOT

C PRINTEO.

2 5 C STOP

END ^^5\

SUBROUTINE SET (A,M,V) ^

C SET PUTS THE VALUE V INTO EVERY ELEMENT OF THE ARRAY A

DIMENSION AfH)

0011*1,M

5 1 A ﬂ ) = 0 . 0 ^

C ■•'

ENTRY INC

C I N C A D O S T H E VA L U E V T O E V E R Y E L E M E N T I N T H E A R R AY A _

0021=1,M ^

10 2 A<I)=A<I)+V

RETURN

END

9"8 60497800 A

/*f"^\

/pftv

0^\

SUBROUTINE IOTA CA,M>

C IOTA PUTS CONSECUTIVE INTEGERS STARTING AT 1 IN EVERY ELEMENT OF

C T H E A R R A Y A

DIMENSION ACM)

0011=1,M

1 A<I)=I

RETURN

END

FUNCTION PVAL(SIZE,HAY)

C PVAL COMPUTES THE POSITIVE VALUE OF WHATEVER REAL VALUE IS RETURNED

C BY A FUNCTION SPECIFIED WHEN PVAL WAS CALLEO. SIZE IS AN INTEGER

C VALUE PASSED ON TO THE FUNCTION.

5 INTEGER SIZE

C$ PUNCS<ABS)

PVAL-=ABS(WAY(SIZE>)

C WAY DOES NOT APPEAR IN THE ARGUMENT LIST FOR THE FUNCS STATEMENT,

10 C SO ONLY THE REFERENCE TO ABS IS DIAGNOSED.

0O&\

RETURN

END

1 0 C

FUNCTION AVG(J)

C AVG COMPUTES THE AVERAGE OF THE FIRST J ELEMENTS OF COMMON.

COMMON A(100)

AVG=0.

D01I=1,J

1 AVG=AVG«*A(I>

CS FUNCS

C ALL EXTERNAL FUNCTION REFERENCES WILL BE DIAGNOSED.

1 0 C

AVG*AVG/FLOAT<J)

RETURN

END

REAL FUNCTION MULTCJ)

C MULT COMPUTES A STRANGE AVERAGE. IT MULTIPLIES THE FIRST AND 12TH

C ELEMENTS OF COMMON AND SUBTRACTS FROM THIS THE AVERAGE CCOMPUTEO

C BY THE FUNCTION AVG) OF THE FIRST J/2 WOROS IN COMMON.

COMMON ARRAYC12)

C$ FUNCS

C ALL EXTERNAL FUNCTION REFERENCES WILL BE DIAGNOSED.

MULT=ARRAY(12) "ARRAY<1)-AVGfJ/2)*

RETURN

E N D

60497800 A 9_9

/^^v

/OEeuf./ VAPUTM? AT LINE 3- ROL'TTNF. SET CALLED AT L E V E L

/OFftUG/ AT LINE 9- KOLTINE SET RE TURNS TO LEV EL

/OEOUG/ AT LINE 9- ROUTINE I O TA CALLED AT L E V E L

/OEOUG/ AT L I N E 1 U - cOUTINE I O TA RETURNS TO L EV EL

/OEP.UG/ AT LINE 10- FOUTINE INC CALLED A T L c V t l

/OEOUG/ AT L I N E 1 1 - ROUTINE INC .<ETUKNS T O L f V E L

/HEPUG/ AT LINE .'.5- f-EiV. FUNCTION P VAL C A L L E D AT LEVEL

/OEOUG/ AT LINE lS- REAL FUNCTION P VA L RETURNS VALUE OF l.SUUUUQodO AT LEVEL

/DEBUG/ AT LINE 16- ttCM FUNCTION P VA L C A L L t G .\T LEVEL

/OECUG/ MULT AT LINE 1 1 - REAL FUNCTION AV G CALLEO ■ULEVEL

/OEPUG/ AT LINE 1 1 - RtAL" FUNCTION AV3 Kf T'MHS VALUE OF -1.500000000 AT LEVEL

/OEBUG/ VAROIM2 AT LINE ltS- REAL FUNCTION P VA L RtT URN S VA LUE OF 2b.i>G0iiiiuJ0 AT LEVEL

STORES STATEMENT

S T O R E S ( c v c 2 c n ) ^^\

An argument list must be speciﬁed for the STORES statement.

(c,,...,cn) are variable names or expressions in the forms:

variable name

variable name .relational operator, constant

variable name .relational operator, variable name

variable name .checking operator.

Relational operators are .EQ., .NE., .GT., .GE., .LT., .LE.

Checking operators are .RANGE., .INDEF., .VALID.

Example:

C$ STORES(SUM,DGAMP,AX,NET.LT.4,ROWSUM.RANGE.)

C$ STORES(Al,AGAIN,I,A2.EQ.5.O,IAGAIN.LE.IVAR)

C$ STORES(C.EQ.(l.,1.),L.VALID.,D.NE.10.004)

C$ STORES(G.RANGE..TR.EQ..FALSE.)

9-10 60497800 C

^ifev

The STORES statement is used to record changes in value of speciﬁed variables or arrays. The STORES

statement applies only to assignment statements. Values changed as a result of input/output, or use in

DATA, ASSIGN, and COMMON statements, or argument lists to subroutines and functions are not detected.

The STORES statement does not apply to the index variable in a DO loop.

If the value of a variable in an EQUIVALENCE group is changed, the STORES statement will not detect

changes to the value of other variables in the group.

VARIABLE NAMES

In the ﬁrst form of the STORES statement, a message is printed each time the value of a variable or an

array element changes. The variable and name of the array must appear as arguments in the C$ STORES

statement.

Example:

PROGRAM STORES (INPUT,OUTPUT,0E8UG = OUTPUT)

LOGICAL L1,L2

C$ STORES (NSUM,DGAMP,AX)

NSUM = 20

OGAMP = .5

AX = 7.2 + DGAMP

LI = .TRUE.

L? = .FALSE.

PLANT = 2.5

A = 7.5

PRINT 3

3 FORMAT (1H0)

STOP

END

0^\

Each time the value of the variables NSUM, DGAMP and AX changes, a message is printed. The

values of PLANT. A. L1 and L2 are not printed, since they do not appear in the argument list.

/ D E B U G / S T O R E S A T L I N E < » - T H E N E W V A L U E O F T H E V A R I A B L E N S U M I S 2 0

/OEBUG/ AT LINE 5- THE NEW VALUE OF THE VARIABLE DGAMP IS .5008000000

/DEBUG/ AT LINE 6- THE NEW VALUE OF THE VARIABLE AX IS 7.700000008

/*?i&£\

60497800 A 9-11

Array elements should not be speciﬁed in the parameter list of a STORES statement; the array name must be

used. If an array element appears, an informative diagnostic is printed. Changes to any element of the array

are noted; only the array name without subscript is listed.

Example:

PROGRAM STORAR <INPUT,OUTPUT,DEBUGsOUTPUT>

REAL AI101, B(t»,2)

CS STORES CA,B>

B(i,2) - 5.5

B<<t,2) = 0.

DO 4 N * i,3

4 A 1 N ) s N + 1

PRINT 5

5 FORMAT (1H0I

STOP

END

/*^\

/ ^ \

/OEBUG/ STORAR AT LINE 4- THE NEW VALUE OF THE VARIABLE B

/DEBUG/ AT LINE 5- THE NEW VALUE OF THE VARIABLE B

/OEBUG/ AT LINE 7- THE NEW VALUE OF THE VARIABLE A

/DEBUG/ AT LINE 7- THE NEW VALUE OF THE VARIABLE A

IS 5.500000000

I S 0 .

IS 2.000000000

IS 3.000000000

IS *♦.ooooooooo

•^^K

The values stored into array elements B(l,2) and B(4,2) appear in the debug output under the array

name B in both cases, and array elements A(l), A(2), and A(3) appear under the array name A.

RELATIONAL OPERATORS

In the second form of the C$ STORES statement, a message is printed only when the stored value satisﬁes the

relation speciﬁed in the argument list. The two components of the relational expression must be of the same type.

/*^S\

PROGRAM ST3 (INPUT♦OUTPUTtDEBUG*OUTPUT)

5 FORMAT (1H0)

PRINT 5

M a 5

CS STORES Cl.EQ.3tN.LE.MffANT)

1 = 3

I s 4

N = 4

N = 6

J B 10

ANT = 77.0

END

/DEBUG/ ST3 AT LINE 6- THE NEW VALUE OF THE VARIABLE I

/DEBUG/ AT LINE 8- THE NEW VALUE OF THE VARIABLE N

/DEBUG/ AT LINE 11- THE NEW VALUE OF THE VARIABLE ANT

I S 3

I S 4

IS 77.00000000

0*~^\

9-12 60497800 A

0z?i&§?\

/"asss

ygfe\

I appears in the debug output when it is equal to 3; N appears when it is less than or equal to M.

Since no relational operator is speciﬁed with ANT, it is printed whenever the value changes.

CHECKING OPERATORS

In the third form of the STORES statement, a message is issued only when the stored value is out of range,

indeﬁnite, or invalid as speciﬁed by the checking operator.

Out of range

Indeﬁnite

Out of range or indeﬁnite

RANGE

INDEF

VALID

For example:

C$ STORES (ROWSUM .RANGE., COLSUM .VALID.)

Whenever the value to be stored into ROWSUM is out of range, a message is printed. Whenever the

value to be stored into COLSUM is out of range or indeﬁnite, a message is printed.

J$0&\

HOLLERITH DATA

Hollerith data stored in a variable of type integer is interpreted by the STORES statement as an integer

number. Hollerith data stored in a variable of type real or double precision is interpreted as a real or

double precision number.

In the following example, the three integer variables IHOLL, IRIGHT and ILEFT contain the characters

PA in display code (20 and 01).

IHOLL 20015555555555555555

P A b l a n k ﬁ l l

IRIGHT 00000000000000002001

z e r o ﬁ l l P A

ILEFT 20010000000000000000

P A z e ro ﬁll

60497800 A 9-13

Example:

PROG RAM OEri O L UNf UT.OUTPU T.D EaUG«O UTPUT>

CS DEBUS

CS STGRGSdMOLL.IRIGHT.ILEfTthOLLJ

IMOLL«2rn?A

1RIGHT»2«PA

ILEFT»2lPA

H0LL«2m?A

PaiNT 1

1 FORMAT (IhO)

STOP

END

/OEoUG/ DEMOi.

/DtaUG/

/OtSOG/

/OcdUG/

AT LIKE 6- TMt NEW VALUE Of Trl£ VARIABLE IhOcL Ii ••••••••<•••••

AT LINE 7- TrlE NEW VALUE Of TMSi VAkUBLE 1HIGHC IS 102S

AT LINE 8- ThE NEW VALUE OF THE VARIABLE ILEFT IS •••••••*<.•••••

AT LIME 9- THE MEU VALUE OF THE VARIABLE KOLL IS .<fC21071096E«fl5

The variables IHOLL, IRIGHT, and ILEFT are interpreted as integer numbers. Since the ﬁeld width allocated

by the STORES option (14 digits) is insufﬁcient to contain the converted quantities represented by IHOLL

and ILEFT, these ﬁelds are ﬁlled with asterisks. The variable IRIGHT is converted and printed out by the

STORES option as 1025.

The variable HOLL is interpreted as a real number, and its value is printed out.

/*^*\

GOTOS STATEMENT

GOTOS

No argument list can be speciﬁed with the C$ GOTOS statement. The GOTOS statement initiates checking

of all assigned GO TO statements to ensure that the statement label assigned to the integer variables is in the

GO TO statement list. If no match is found, a message is printed and transfer of control continues.

PROGRAM GO TO«5 (OUTPUT,OERUGsOUTPUTI

INTEGER A

CS GOTOS

• (GOTOS NEVER USES AN ARGUMENT LIST)

ASSIGN 1 TO A

GO TO A (1, 2, 3)

• IN THIS CASE NO MESSAGE IS PRINTED SINCE THE LABEL ASSIGNED TO

» A I S I N T H E G O T O L I S T.

*» PRINT 10

10 FORMAT*.** —CONTROL TRANSFERED TO STATEMENT LABEL «•--•)

STOP

1 ASSIGN «♦ TO A

GO TO A (1, 2, 3)

• IN THIS CASE A MESSAGE TS PRINTED SINCE THE LABEL «• IS NOT IN

• THE GOTO LIST. CONTROL THEN TRANSFERS TO LABEL *♦.

2 CONTINUE

3 CONTINUE

END

/OEBUG/ GOtOS AT LINE 16- ASSIGNEO GOTO INDEX CONTAINS THE A0ORESS 002151. NO MATCH FOUNO IN STATEMENT LABEL ADOPFSS LIST

—CONTROL TRANSFEREO TO STATEMENT LAOEL *♦--

9-14 60497800 A

TRACE STATEMENT

/c$" TRACE (Iv)

TRACE

/$P^N

Iv is a level number 0-49. If lv = 0, tracing occurs only outside DO loops. If Iv = n, tracing occurs up to

and including level n in a DO nest. If no level is speciﬁed, tracing occurs only outside DO loops.

The C$ TRACE statement traces the following transfers of control within a program unit:

GOTO

Computed GO TO

Assigned GO TO

Arithmetic IF

True side of logical IF

Transfers resulting from a return speciﬁed in a RETURNS list are not traced. (These can be checked by the

CS CALLS statement.)

If an out-of-bound computed GO TO is executed, the value of the incorrect index is printed before the job

is terminated.

Messages are printed each time control transfers during execution. The message contains the routine name,

the line where the transfer took place, and the number of the line to which the transfer was made, as well

as the statement number of this line, if present.

A message is printed each time control transfers at a level less than or equal to the one speciﬁed by lv. For

example, if a statement CS TRACE(2) appears before a sequence of DO loops nested four deep, tracing

takes place in the two outermost loops only.

TRACE messages are produced at execution time, but TRACE levels are assigned at compile time;

therefore, the compile time environment determines the tracing status of any given statement. For example,

a DO loop TRACE statement applies only to control transfers occurring between the DO statement and its

terminal statement at compile time (physically between the two in the source listing).

60497800 A 9-15

Example:

level 0

r level 1

r level 2

PROGRAM P<OUTPUT,OEBUG=OUTPUT>

OATA J/0/

C $ T PA C E t l )

IF (J .EQ. 0) GO TO 11

11 DO 1 II = 1, 3

IF ( <J*1» .EQ. II 1 GO TO 12

12 J = 1

OO 2 12 = 1, 5

J = J ♦ 12

GO TO 2

2 CONTINUE

C$ TPACE(3>

OO 20 12 = 1, 3

IF ( 12 .EQ. 3 ) GO TO 20

J = 2

DO 3 13 = 1, 4

IF < J .GT. 13 ) GO TO 31

T31 OO 4 14 = 1» ?

level 4 I GO TO 4

L 4 CONTINUE

3 CONTINUE

20 CONTINUE

J = 0

1 CONTINUE

ENO

r- level 3

><S^V

/0E8UG/ P AT LINE 4- CONTROL

/DEBUG/ AT LINE t*' CONTROL

/DEBUG/ AT LTNE 6- CONTROL

/DEBUG/ AT LINE 6- CONTROL

/DEBUG/ AT LTNE 17- CONTROL

/DEBUG/ AT LINE 17- CONTROL

/OEBUG/ AT LINE 17- CONTROL

/DEBUG/ AT LINE 17- CONTROL

/DEBUG/ AT LINE 14- CONTROL

/DEBUG/ AT LINE 14" CONTPOL

/DEBUG/ AT LINE 17- CONTROL

/DEBUG/ AT LINE 14- CONTROL

/OEBUG/ AT LINE 14- CONTPOL

/DEBUG/ AT LINE 17- CONTROL

/OEBUG/ AT LINE 17- CONTROL

/DEBUG/ AT LINE 17- CONTROL

/DEBUG/ AT LINE 14- CONTROL