Digital PDFs

AA-DO35A-TE

August 1978

196 pages

Original

9.1MB

view download

OCR Version

8.3MB

view download

Document:	VAX-11 Fortran IV-Plus User's Guide
Order Number:	AA-DO35A-TE
Revision:	0
Pages:	196
Original Filename:

OCR Text

August 1978

This manual describes how to compile, link, debug, and execute programs

written in the VAX-11 FORTRAN 1V-PLUS language, using the facilities of the
VAX/VMS operating system. It also contains other information of interest to
FORTRAN programmers, such as: error processing, programming efficiency,

compatibility with PDP-11 FORTRAN, and FORTRAN input/output.

VAX-11 FORTRAN IV-PLUS
User’s Guide
Order No. AA-DO35A-TE

SUPERSESSION/UPDATE INFORMATION:

This is a new document for this release.

OPERATING SYSTEM AND VERSION:

VAX/VMS V01

SOFTWARE VERSION:

VAX-11 FORTRAN IV-PLUS V01

To order additional copies of this document, contact the Software Distribution
Center, Digital Equipment Corporation, Maynard, Massachusetts 01754

digital equipment corporation - maynard, massachusetts

First Printing,

The
and

information in this document

August

1978

subject to change without notice

should not be construed as a commitment by Digital Equipment
Corporation.
Digital Equipment Corporation assumes no responsibility
for any errors that may appear in this document.
The software described in this
and may only be used or copied

license.

document

is furnished under a license
accordance with the terms of such

responsibility is assumed for the use or reliability of software on
equipment that is not supplied by DIGITAL or its affiliated companies.

(C)

1978 by Digital Equipment Corporation

The

postage~prepaid READER'S
document requests the user's
paring future documentation.

COMMENTS

The

following

are

trademarks

critical

form on the last page of this
evaluation to assist us in pre-

Digital

Equipment

Corporation:

DIGITAL

DECsystem-~10

DEC

DECtape

OMNIBUS

PDP

DIBOL

0s/8

DECUS

EDUSYSTEM

UNIBUS

FLIP

COMPUTER LABS- -

MASSBUS

PHA

CHIP

~FOCAL

RSTS
—~—

—

—RSX

—

COMTEX

INDAC

TYPESET-8

DDT

LAB-8

TYPESET-11

DECCOMM

DECSYSTEM-20

ASSIST-11

TMS-11

RTS-8

VAX

ITPS-10

VMS

SBI

DECnet

IAS

—

iii

HHERWOYWOWWOWWOWWOoOO-IOOTUL
&N
- e
o
N HOOOO

HFHHERHRHEHERRREHERRHERRRRRE

[

aouUTU > WH

DEBUG A PROGRAM

SHOW LANGUAGE Command
SHOW, CANCEL MODULE Commands
SHOW, CANCEL SCOPE Commands

SET,
SET,
SET,

wN
-

PREPARING TO

w N

OVERVIEW OF THE VAX-11l SYMBOLIC DEBUGGER
Sample Debugging Terminal Session
Debugger Command Syntax
Debugger Symbol Table

DEBUGGING FORTRAN PROGRAMS

NNNDNDDNDDNNN

1-17
1-18
1-19
1-19
"1-21
1-23

XS]

> wWw N

1-15
1-17

NDNDDNDNDN

Source Listing Section
Machine Code Listing Section
Storage Map Section
Other Listing Information

1-13
1-14
1-14
1-14
1-14
1-14
1-15

= o

BPHOYOJOAUTES
WD

[
&

[}

[

CORRECTING ERRORS

Error-Related Command Qualifiers
SHOW CALLS Command
SAMPLE TERMINAL SESSION
COMPILER LISTING FORMAT

FINDING AND

NANNNOUMTU

Linker Command Qualifiers
Image File Qualifiers
Map File Qualifiers
Debugging and Traceback Qualifiers
Linker Input File Qualifiers
/LIBRARY Qualifier
/INCLUDE Qualifier

EXECUTION

CHAPTER

Specifying Output Files
FORTRAN Qualifiers
CHECK Qualifier
CONTINUATIONS Qualifier
DEBUG Qualifier
D_LINES Qualifier
I4 Qualifier
LIST Qualifier
MACHINE_CODE Qualifier
OBJECT Qualifier
OPTIMIZE Qualifier
WARNINGS Qualifier
WORK_FILES Qualifier

LINKING

[}

*
@
e
9
®
e
°
Q
@
s

COMPILATION

[]

[

CREATING AND EXECUTING A PROGRAM
File Specifications
Qualifiers

MR WWWWWWWWNNNDNDNNDDNDNNDDNDNDNDND

[

USING VAX-11l FORTRAN IV-PLUS

CHAPTER

HFRRFHEFRHERHERERERERERRERRFRRERRRFRERERRRRRRRERRRRRE

PREFACE

]

[

g
o

CONTENTS

Commands
Commands

SET, SHOW,
SHOW CALLS

CANCEL WATCH
Command

Commands

GO,

STEP

Commands

CTRL/Y Command
EXIT

Command

EXAMINING AND

MODIFYING LOCATIONS

EXAMINE

Command

DEPOSIT

Command

EVALUATE Command
SPECIFYING ADDRESSES

Lines, Labels, and Absolute Addresses
Specifying Scope

2.5.3
2.5.4
2.6

CALLING SUBROUTINES

2.7

DEBUGGER COMMAND QUALIFIERS

2.8

NUMERIC

DATA TYPES

2.9
2.9.1

EFFECTS

Previous,

Current,

Use

FROM THE

DEBUGGING

Codes

FORTRAN

Flow
of /NOOPTIMIZE

and /OPTIMIZE

INPUT/OUTPUT

FILE

LOGICAL NAMES

SPECIFICATION

FORTRAN Logical Names
Implied FORTRAN Logical Unit Numbers
OPEN

Statement NAME

Keyword

Assigning Files to Logical Units
Assigning Logical Names with MOUNT
‘FILE

CHARACTERISTICS

File Organization
Sequential Organization
Relative Organization
Access to Records
Sequential Access
Direct Access

3.3.2.1
3.3.2.2
3.4
RECORD STRUCTURE
3.4.1
Fixed Length Records
3.4.2
Variable Length Records
3.4.3
Segmented Records
3.5
~ OPEN STATEMENT KEYWORDS
3.5.1
BLOCKSIZE Keyword
3.5.2
BUFFERCOUNT Keyword
3.5.3
INITIALSIZE and EXTENDSIZE
3.5.4
ORGANIZATION Keyword
3.5.5
READONLY Keyword
3.5.6
RECORDSIZE Keyword
3.5.7
3.5.8
3.5.9
3.6

DEBUGGER

3.2
3.2.1

3.2.4
3.2.5
3.3
3.3.1
3.3.1.1
3.3.1.2
3.3.2

Symbolically

OPTIMIZATION

Condition

3.1

3.2.2
3.2.3

and Next Locations

Defining Addresses

RECORDTYPE Keyword
SHARED Keyword
USEROPEN Keyword
AUXILIARY I/O OPERATIONS

Commands

2.9.2
2.9.3
2.9.4
CHAPTER

CANCEL BREAK
CANCEL TRACE

2.5.2

SHOW,
SHOW,

|O WOWYWORVOVOJIJ~INIJOAOAUTDdWWN

2.4.2
2.4.3
2.5
2.5.1

SET,
SET,

Keywords

2.3.7
2.4
2.4.1

CONTROLLING PROGRAM EXECUTION

WWWWLwWWwWwwWwwwwwwwwwww

2.3
2.3.1
2.3.2
2.3.3
2.3.4
2.3.5
2.3.6

(Cont.)

CONTENTS

(Cont.)

>
N

CHARACTER ARGUMENTS

DATA EXAMPLES
FUNCTIONS

CHARACTER LIBRARY

[

CHAR Function
ICHAR Function
INDEX Function
LEN Function

EXAMPLES

Argument List Built-In Function Examples
Character Functions
ERROR PROCESSING AND
RUN-TIME

LIBRARY

CONDITION

DEFAULT

ERROR

HANDLERS
PROCESSING

Using ERR= and END= Transfers
Run-Time Library Error Processing Control
Using the ERRSNS Subroutine
OVERVIEW OF THE

VAX-11

FACILITY

Definitions
Condition Signals

CONDITION

BB R WWWWNDNND

HWOYOOOJIJaUn

CODE

Argument Passing Examples

oottt
|

Reference
Calling System Services as Subroutines
Passing Arguments to System Services
Input and Output Address Arguments
Defaults for Optional Arguments
Passing Character Arguments

(+)

W
-

%DESCR

oottt
orn
|
I

o
o

= WN
-

L
*

N wWwWwwwwn
-+

$DESCR

Examples of %VAL, $REF,
Function Return Values

wN
-

$REF

$LOC Built-In Function
CALLING VAX/VMS SYSTEM SERVICES
Calling System Services by Function

o
o

*
*
*
Ld
Ll
WWwwwn

$ VAL

MACHINE
*

CALLING STANDARD

Argument Lists
Argument Passing Mechanisms
Argument List Built-In Functions

(o))

CALLS

VAX-11l PROCEDURE

PROCEDURE

CONVENTIONS

wWwN
N

CALL

(52

CHARACTER I/O

ANy O

CHARACTER

HANDLING

~NoO oy

PASSED LENGTH

DATA

CHARACTER VARIABLES

CHARACTER

INITIALIZING

|
—

CONSTANTS

DECLARING

CHARACTER STRINGS

CHARACTER

3-15

1
WOOOOIJTNHBEWWNH

NETWORKS

3-15

A OO

e
e
*

e
o

S
o

A
s

S
¢
e

o
o

o
o
o

[

o
o
¢
o
o
o
o
o
o
e
*

oottt

COMPUTERS:

DATA

Oy OV OY OY OV

oot

L]
.
L]
*
L] L]
.
*
L]
bRk WWWWW

REMOTE

3-14
3-14

CHARACTER SUBSTRINGS

FORTRAN

(<2 W)}

CHAPTER

CHARACTER

BUILDING

19,

\O 000000000 ~JOYUI>»WN K

USING

CHAPTER

MAILBOXES

Sending and Receiving Data Using Mailboxes
COMMUNICATING WITH

WwhitdhdBddddNDNDDDDNDND
-

CHAPTER

LOCAL INTERPROCESS COMMUNICATION:
Creating a Mailbox

Wwww

Page

CONTENTS

O W ©
-0
O

|
=
B

|
R
B
AN

R
B

N
B

N
B
P
B

LOOPS

ARGUMENTS
REPRESENTATION

DATA

T
B

1
R

T
B

A
B B

i
[

[0}

N
D
I
T

I/0

FORTRAN

DATA

SYSTEM CHARACTERISTICS
REPRESENTATION

INTEGER*2

FORMAT

INTEGER*4

FORMAT

FLOATING-POINT

FORMATS

Real Format (4-Byte Floating Point)
Double Precision Format (8-Byte Floating
Point)

Compiler Optimization Example
FORTRAN

3.1
3.2

Program Blocks

Eliminating Common Subexpressions

Characteristics of Optimized Programs
Compile-Time Operations on Constants
Source

Statement

INCLUDE

Allocating Variables in Common Blocks
Conditional Branching
COMPILER OPTIMIZATIONS

APPENDIX

PROGRAMS

HO®JOUdWWWNH

SOURCE

Statement

00 00 00 OO0 00 00 00 CO OO0 00 0O OO

CONSIDERATIONS

N
e
.

W
N e

PARAMETER

0.0

YRy

wWN

[
.
.

Representation of
Sign Bit Tests

[}
[}
*
*
L

Single Precision Floating Point Data
Double Precision Floating Point Data
Floating Point Data Characteristics
Reserved Operand Faults

WWwwwn
H

POINT

CREATING EFFICIENT

L)
.
*
.
Ld

FOR

STATEMENT

FLOATING

L4
[)
]

Statement

Cautions Concerning Program Interchange
Iteration Count Computation
ENTRY

WNHhRNMNNDNDNMNDNNDHRFRERFPE

MODEL

1
2

COUNT

the EXTERNAL

e B

Use

PROGRAMMING

00 0O 00 00 CO CO 00 0O CO 0O OO 0 O

CHAPTER

INTEGER*4

Integer Constant Typing
Integer-Valued Processor-Defined Functions
Byte (LOGICAL*1l) Data Type
FUNCTIONS SUPPLIED WITH VAX~1ll FORTRAN
Generic Functions
ITERATION

NNNNNNNNNNNN NN
L]
L
L]
.
L]
Ao
UTIR R BBWWWNDNN

1
2

DATA TYPES

AND FIXED-POINT

Integer Data Types Supported
Relationship of INTEGER*2 and
Values

.2
.3

USAGE

ALLOCATION

STORAGE

DY N

PROGRAM SECTION

SYSTEM ENVIRONMENT

1
2
2.1.1

HANDLER EXAMPLES

BMBHEREREEREOVYOYOoOIOOAANMTUIULE B

CONDITION

(o)W e W e Wa Mo WerTe We )

HANDLERS

CONDITION

&
&

e
8

&
*

*
L]

dWwwwwwwN

Responses

Establishing and Removing Handlers
FORTRAN Condition Handlers
Handler Function Return Values
Condition Values and Symbols
Floating Overflow, Zero Divide Exceptions

FORTRAN

NNa

CHAPTER

Handler

USER-WRITTEN

[e) e Mo We Mo We ) We) eyt

g
o)
Q
®

(Cont.)

CONTENTS

(Cont.)

Page

APPENDIX

A.3.3
A.4
A.5
A.6
A.7

Complex Format
LOGICAL*1l FORMAT
CHARACTER FORMAT
HOLLERITH FORMAT
LOGICAL FORMAT

A-3
A-4
A-4
A-4
A-5

DIAGNOSTIC

B-1

B.1l

DIAGNOSTIC MESSAGES

OVERVIEW

B.2

DIAGNOSTIC

FROM THE

B.2.1

Compiler-Fatal
Compiler Limits

B.3.2

B-1
COMPILER

Program Diagnostic Messages

B.2.3

)

APPENDIX

MESSAGES

B.2.2
B.3
B.3.1

APPENDIX

Source

MESSAGES

Diagnostic

Messages

B-~1

B-~1
B-17
B-19

RUN-TIME

DIAGNOSTIC MESSAGES
Run-Time Library Diagnostic Message

B-20

Presentation
Run-Time Library

B-20
B-20

Diagnostic Messages

SYSTEM

C.1l

SYSTEM SUBROUTINE SUMMARY

C-1

C.2

DATE

C-1

C.3
C.4
C.5
C.6
C.7

IDATE
ERRSNS
EXIT
SECNDS
TIME

C-2
C=-2
C-3
C-3
C-4

COMPATIBILITY

D~-1

COMPATIBILITY: OVERVIEW
DIFFERENCES
Logical Tests
Floating Point Results

D=1
D-1
D~1
D-2

D.1
D.2
D.2.1
D.2.2
D.2.3

D.2.4
D.2.5
D.2.6
D.3
D.3.1
D.3.2
D.3.2.1
D.3.2.2
D.3.2.3
D.3.3
D.4
D.4.1
D.4.2
D.4.3
D.4.4
D.4.5

D.4.6
D.4.7
D.4.8
D.4.9
D.4.10
D.4.11

SUBROUTINES

LANGUAGE

Character and Hollerith Constants
Logical Unit Numbers
Assigned GO TO Label List
DISPOSE='PRINT' Specification

RUN-TIME

SUPPORT DIFFERENCES
Run-Time Library Error Numbers
Error Handling and Reporting
Continuing After Errors
I/0 Errors with ERR= Specified
OPEN/CLOSE Statement Errors
OPEN Statement Keywords
UTILITY SUBROUTINES
ASSIGN Subroutine
CLOSE Subroutine
ERRSET Subroutine
ERRTST Subroutine
FDBSET Subroutine
IRAD50 Subroutine
RAD50 Function
RAN Function
RANDU Subroutine
R50ASC Subroutine
USEREX Subroutine

vii

Cc-1

D-2
D~-3
D=3
D-3

D-3
D-4
D=5
D-5
D-5
D-5
D-5
D=5
D-6
D-7
D-7
D-8
D-9
D~10
D-11
D-11
D-12
D-12
D-13

CONTENTS

(Cont.)

Page

INDEX

Index~-1

RELAX

Source

Program

RELAX Machine Code (Optimized)
Sample Diagnostic Messages (Terminal Format)
Sample Diagnostic Messages (Listing Format)

WNOWONANKFNNHEEND
NO VWO

Program Development Process
Traceback List
Source Listing Section
Machine Code Listing Section
Storage Map Listing
Sample FORTRAN Program: CIRCLE
Sample Debugging Terminal Dialog
Character Data Program Example
Output Generated by Example Program

wrna>m4>¢nrh:wbdkflHb~
11

NHEEHEPFENMFEFMDRPOBWNODH

FIGURE

WKflGDmdhhdrh)dedepa
[

FIGURES

/DEBUG and /TRACEBACK Qualifiers
Debugger Commands and Keywords
Debugger Command Qualifiers
Predefined System Logical Names
Implicit FORTRAN Logical Units
RECORDSIZE Limits
Function Return Values
Variable Data Type Requirements
Summary of FORTRAN Run-Time Errors
Condition Handler Function Return Values
PSECT Names and Attributes
PSECT Attributes
Source Program Diagnostic Messages
Compiler-Fatal Diagnostic Messages
Compiler Limits
Run-Time Diagnostic Messages
Default Logical Unit Numbers

viii

WOHPFRMWNHAARHAWHMAHMIW
o
N
~
W
0=

File Specification Defaults
FORTRAN Command Qualifiers
Linker Qualifiers

Utflu1wtn~chna\TLnu)ucuuawrapawra
!
I
i1

(|
U(fiUJWtfl\JQChO\T(nUJwL»hJNIkohJH
|
I

TABLE

HE2WNDNFHFNFEFNMFENMNRFRPWONENDEBRWND
-

TABLES

PREFACE

MANUAL OBJECTIVES

The VAX-1l FORTRAN
IV-PLUS
User's
Guide
1is
intended
for
wuse
1in
developing
new FORTRAN programs, and compiling and executing existing
FORTRAN
programs
on
VAX/VMS
elements
supported
on
VAX-11
IV-PLUS

Language

systems.
FORTRAN
IV-PLUS
language
are
described
in the VAX-11 FORTRAN

Reference Manual.

INTENDED AUDIENCE

This manual is designed for programmers who have a
working
knowledge
of
FORTRAN.
Detailed
knowledge
of
VAX/VMS
1is
helpful
but
not
essential;
familiarity with the VAX/VMS Primer is recommended.
Some
sections
of
this
book,
however, (condition handling, for instance)
require more extensive understanding of the operating system.
1In such
sections,
you
are
directed
to
the
appropriate
manual(s) for the
required additional information.

STRUCTURE

This manual

THIS

is organized

Chapter
and

DOCUMENT

contains

execute

follows:

the

FORTRAN

information needed

compile,

1link,

program.

Chapter 2 covers the debugging process;
Symbolic Debugger is described.

Chapter 3 provides
information
about
FORTRAN
input/output,
including
details
on
the
use
of
1logical
names,
file
conventions,
record
structure,
and
use
of
certain
OPEN
statement

use

the

VAX-11

keywords.

Chapter 4 discusses character data, and
how character data can be manipulated.

includes

examples

Chapter
5
procedures,

Chapter 6 describes
error
processing;
in
particular,
the
condition
handling
facility and how to use it.
This chapter
is intended for users with in-depth knowledge of VAX/VMS.

Chapter 7 describes the relationship
between
VAX-11
FORTRAN
IV-PLUS
and
the
VAX-1ll
system, with particular emphasis on
program section usage, data types, functions,
DO
1loops,
and
floating point data representation.

discusses
the
conventions
followed
1in
calling
especially the argument-passing conventions.

Chapter

typical

covers

FORTRAN

Appendixes A

programming

considerations

relevant

applications.

through

summarize

internal

data

representation,

diagnostic
messages,
system-supplied
functions,
compatibility between VAX-11l FORTRAN and PDP-11 FORTRAN.

ASSOCIATED

DOCUMENTS

The
following
programming:

documents

are

relevant

VAX/VMS

Primer

VAX-11

FORTRAN

VAX/VMS

VAX-11

Common

Run-Time

Procedure

VAX-11l

Linker

Reference

Manual

VAX-11l

Symbolic Debugger

VAX/VMS

VAX-11/780 Architecture Handbook

IV-PLUS

Command

Language

USED

VAX-1l1l

THIS

1IV-PLUS

Reference Manual

Library

Reference

software

are

observed

and

letters,

you

the

word

should

type

Lowercase words
that you are to

Brackets

FORTRAN

User's Guide

Uppercase words

VAX-1l1

Reference

Manual

({

Ellipses
repeated

([

})

])

documents,

see

the

VAX-11

DOCUMENT

The following conventions
VAX-11 documents:

Braces

System Services Reference Manual

For a complete list of
Information Directory.

CONVENTIONS

and

and letters,
substitute a

indicate

this manual,

used

letter
used
word

optional

examples,

exactly
in
or

the

other

indicate

that

shown

format examples, indicate
value of your choice

elements

are used to enclose lists from which one

chosen

(...) indicate that
one or more times

the

preceding

item(s)

element
can

CHAPTER 1

USING VAX-11

FORTRAN

IV-PLUS

VAX-11l FORTRAN IV-PLUS is based on American National Standard
FORTRAN
X3.9-1966.
It
1is
also
a
compatible
superset
of
PDP-11 FORTRAN
IV-PLUS.
VAX-11 FORTRAN IV-PLUS provides the following extensions:
e

Symbolic names up to 15 characters,
and dollar

sign characters

FORTRAN

character

FORTRAN

block

Relative

Standard CALL

Hexadecimal constants

Symbolic debugging

data

including

the

underline

type

constructs

file organization

facility

and

field descriptors

facility

) Increased’file manipulation facilities
Because VAX-11 FORTRAN IV-PLUS is
a
compatible
superset
of
PDP-11
FORTRAN,
you
can
execute existing PDP-11 FORTRAN programs on VAX-11
hardware.
(Note that throughout
the
rest
of
this
manual,
unless
explicitly
stated
otherwise,
VAX-11l FORTRAN IV-PLUS will usually be
referred

1.1

simply as FORTRAN.)

CREATING AND EXECUTING A PROGRAM

Figure 1-1 shows how a program proceeds from inception
to
execution.
You
specify the steps shown in Figure 1-1 by entering commands to the
VAX-11l
$
S
S
$

system.

shown,

the

commands

are:

EDIT file-spec
FORTRAN file-spec
LINK file-spec
RUN file-spec

With each command, you include information that further
defines
what
you
want
the
system
to
do.
Of prime
importance
1is
the
file
specification, indicating the file to
be
processed.
You
can
also
specify qualifiers that modify the processing performed by the system.

USING VAX-1l

FORTRAN

IV-PLUS

COMMANDS

INPUT/OUTPUT FILES

$ EDIT AVERAGE.FOR
Use the file type of FOR to
indicate the file contains a
VAX-11

FORTRAN IV-PLUS

Create a
|

AVERAGE.FOR

<ource program

program.

$ FORTRAN AVERAGE
The FORTRAN command

assumes the file type of an

input file is FOR.

Compile the
source program

(If you use the /LIST

AVERAGE.OBJ

—_—

(AVERAGE.LIS)

qualifier, the compiler

libraries

creates a listing file.)

$ LINK AVERAGE
The L/INK command assumes
the file type of an input file

AVERAGE.EXE

Link the

is OBJ.

(AVERAGE.MAP)

object modute

(If you use the /MAP qualifier,
the linker creates a map file.)

$ RUN AVERAGE

Run the

The RUN command assumes

exe:cutable

the file type of an image is

image

EXE.

Figure

1-1

Program Development Process

l1.1.1

File Specifications

A file
output

specification indicates the input file to be processed, or
file to be produced.
File specifications have the form:

the

node: :device: [directory]l filename.filetype.version

node

Specifies
systems

a network

that

device
Identifies
written.

node

support VAX-11l

the device

name.

This

1is

DECnet.

on which
a file

applicable

stored
:

only

to
o

is
S

directory
Identifies the name of the directory
under
which
the
file
is
cataloged,
on
the
device
specified.
(You
can
delimit
the
directory name with either square brackets, as
shown,
or
angle
brackets < >).

filename
Identifies
characters

the file
1long.

its

name;

filename

can

USING VAX-11

filetype
Describes the kind
characters long.

of data

FORTRAN

the

IV-PLUS

file;

filetype

can

version
Defines which version of
the
file
1is
desired.
Versions
are
identified
by
a
decimal number, which is incremented by 1 each
time a new version of a file is created.
Either a semicolon or a

period can be used to separate filetype and version.

You need not explicitly state all elements
of
a
file
specification
each
time
you compile, link, or execute a program.
The only part of
the file specification that is usually required is the file name.
If
you
omit any other part of the file specification, a default value is
used.
Table 1-1 summarizes the default values.

Table 1-1
Specification Defaults

File

Optional
Element

Default
Value

node

Local

network

node

device

User's

current default device

directory

name

of FOR

file meets

the highest version

these

conditions,

number.

the

compiler

chooses

the

USING VAX-11]

FORTRAN

IV-PLUS

For example, assume that your
default
device
is
DBAQ, your default
directory is SMITH, and you supply the following file specification to
the

compiler:
CIRCLE

The compiler will search device DBAO, in directory SMITH, seeking
the
highest
version of CIRCLE.FOR.
If you do not specify an output file,
the compiler will generate the file CIRCLE.OBJ,
store
it
on
device
DBAO
in directory SMITH, and assign it a version number 1 higher than
any other version of CIRCLE.OBJ currently cataloged in directory SMITH
on DBAO.

1.1.2

Qualifiers

Qualifiers specify special actions to be performed, and can be
either
command qualifiers or file qualifiers.
Qualifiers have the form:
/qualifier

Many qualifiers have a corresponding negative form
that
negates
the
action
that
the
qualifier
specifies.
The
negative
form
is
/NOqualifier.
For example, the qualifier /LIST tells the compiler
to
produce
a
listing file;
/NOLIST tells the compiler not to produce a
listing

file.

Defaults have been
established
for
each
qualifier,
based
actions
that
are appropriate in most cases.
Sections 1.2.2

which describe each command's qualifiers,

contain

tables

on
the
and 1.3,
1indicating

the defaults.

You can specify qualifiers so that either all files
included
command
are
affected,
or
only
certain files are affected.
qualifier immediately follows the command
name,
it
applies

in
the
If the
to
all

files.
For

example:

If you
and

FORTRAN/LIST ABC,XYZ,RST

specify the

If you
(with

for ABC,

XYz,

include a qualifier as part of a file
specification,
it
certain
exceptions)
affect
only
the
file
with which

will
it is

assocliated.

As a

above,

you will

receive

listing

files

RST.

For

example:

FORTRAN/LIST ABC,XYZ/NOLIST,RST

result of

this command,

RST, but not for XYZ.

listing

files

are

created:

for

7 ABC

and

Qualifiers included with
file
specifications
that
are
part
of
a
concatenated
1list
of
input
files are exceptions to this rule.
See
Example 5 in Section 1.2.1, below.

USING VAX-11]

1.2

FORTRAN

IV-PLUS

COMPILATION

To compile a source program,
the FORTRAN command is:

use

FORTRAN|[/qualifiers]

/qualifiers
Codes indicating

the

FORTRAN

command.

The

format

file-spec-list[/qualifiers]

special

actions

to be performed by the compiler.

file-spec-list
Specification of

the source file(s) containing the program to
be
compiled.
You can specify more than one source file.
If source
file specifications are separated by
commas,
the
programs
are
compiled separately.
If source file specifications are separated

by plus

signs,

the

files

are

concatenated

and

compiled

one

program.,

In interactive mode, you can also enter the file
separate line by typing a carriage return after $§
responds with the prompt:

specification
FORTRAN.
The

on
a
system

S_File:

Type

the

1.2.1

file

specification

immediately after

the

§ File:

prompt.

Specifying Output Files

The output produced by the compiler includes object files and
1listing
files.
You
can
control
the production of these files by using the
appropriate qualifiers with
the
FORTRAN
command.
If
you
do
not
specify
otherwise,
the
compiler
generates
an
object
file.
In
interactive mode, the compiler generates no listing file, by
default.
In
interactive
mode,
you must use the /LIST qualifier to generate a
listing file.
1In batch mode, however, just the opposite is true:
by
default,
the
compiler
will produce a listing file.
To suppress the
listing file, you must specify the /NOLIST qualifier.

During the early stages
of
program
development,
you
may
£find
it
helpful
to
suppress the production of object files until your source
program compiles without errors.
Use the /NOOBJECT qualifier.
1If you
do
not
specify
/NOOBJECT,
the
<compiler
generates object files as
follows:

you

If you specify multiple source files, separated
by
plus
the
source
files
are concatenated and compiled, and one
file is generated

If you

source

specify one

source

specify multiple

file

generated

for

1is

each

file,

source

compiled
source

one

object

files,

file

is generated

separated by

separately,

and

signs,
object

commas,

object

each

file

@ You can use both plus signs and commas in the same
command
1line
to
produce
different
combinations of concatenated and separate
object

files

(see

Example

below)

To produce an object file with an
explicit
file
specification,
you
must
use
the
/OBJECT
qualifier, in the form /OBJECT=file-spec (see
Section 1.2.2.8).
Otherwise, the object file will have
the
name
of
its
corresponding
source
file, and a file type of OBJ.
By default,

USING VAX-11 FORTRAN

IV-PLUS

the object file produced from concatenated source files has
the
name
of
the
first
source
file.
All other file specification attributes
(node,
device,
directory,
and
version)
will
assume
the
default
attributes.
Examples:
l.

FORTRAN/LIST AAA,BBB,CCC

Source files AAA.FOR, BBB.FOR, and CCC.FOR
are
compiled
as
separate .
files,
producing
object
files
named
AAA.OBJ,
BBB.OBJ, and
CCC.OBJ;
and
1listing
files
named
AAA.LIS,
BBB.LIS,

and

FORTRAN

CCC.LIS.

)

XXX+YYY+Z7ZZZ

Source files XXX.FOR, YYY.FOR,
and
compiled
as
one
file,

and Z2Z.FOR
are
concatenated
producing an object file named

XXX.0BJ.

FORTRAN/OBJECT=SQUARE

$ FILE:

CIRCLE

The source file CIRCLE.FOR is compiled, producing
an
file
named
SQUARE.OBJ,
but no listing file.
(This
applies to interactive mode only.)

4.,

object
example

FORTRAN AAA+BBB,CCC/LIST

Two object files are produced:
AAA.OBJ
(comprising
AAA.FOR
and
BBB.FOR), and CCC.OBJ (comprising CCC.FOR).
One listing
file is produced:
CCC.LIS.
5.

FORTRAN ABC+CIRC/NOOBJECT+XYZ

When you include a qualifier in a list of files that
are
to
be concatenated, the qualifier affects all files in the list.
Thus, in the command shown, you will completely suppress
the
object
file.
That
1is, source files ABC.FOR, CIRC.FOR, and
XYZ.FOR will be concatenated and compiled, but no object file
will be produced.

1.2.2

FORTRAN Qualifiers

in many cases, the simplest form of the FORTRAN command is
for file processing.
In some cases, however, you will need
FORTRAN qualifiers that specify special processing.
A FORTRAN

qualifier

/aal=yl
where

has

the

form:

L
the

qualifier's

o
name,

L
and

represents

Note
that
many
qualifiers
accept
no
value;
qualifiers is simply to activate or deactivate a
processing.
To
For

specify

sufficient
to use the

list of qualifier

example:

/CHECK= (BOUNDS ,OVERFLOW)

values,

enclose

qualifier

wvalue.

the purpose of these
particular
form
of

them

parentheses.

USING VAX-11 FORTRAN IV-PLUS

command.
Table 1-2 lists the qualifiers you can use with the FORTRAN detail.
Sections 1.2.2.1 through 1.2.2.11 describe each qualifier in
Table

1-2

FORTRAN Command Qualifiers

Default

Negative Form

Qualifier

/NOCHECK

/CHECK=0VERFLOW

None

/CONTINUATIONS=19

/NODEBUG

/DEBUG=TRACEBACK

/D_LINES

/NOD_LINES

/14

/NOI4

/14

/LIST[=file-spec]

/NOLIST

/NOLIST (interactive)

[NO] BOUNDS

[NO]OVERFLOW

/CHECK=)

ALL

NONE

/CONTINUATIONS=n
[NO] SYMBOLS

[NO] TRACEBACK

/DEBUG=)

ALL

NONE

/LIST

(batch)

/MACHINE_CODE

/NOMACHINE_CODE| /NOMACHINE_CODE

/OBJECT [=file-spec]

/NOOBJECT

/OBJECT

/OPTIMIZE

/NOOPTIMIZE

/OPTIMIZE

/WARNINGS

/NOWARNINGS

/WARNINGS

/WORK_FILES=n

None

/WORK_FILES=2

the
1.2.2.1 CHECK Qualifier - At run time, this qualifier causes ons
conditi
the
for
program
your
compiler to produce code to check
It has the form:

indicated.

/CHECK =(

[NO]BOUNDS

[NO]OVERFLOW

ALL
NONE

BOUNDS

Array references are checked to ensure that they are within the
array address boundaries specified. Note, however, that array
bound checking 1is not performed for arrays that are dummy
arguments, and for which the last dimension bound is specified as
1.

For

example:

DIMENSION A(1)
OVERFLOW

BYTE, INTEGER*2,

and

arithmetic overflow.

INTEGER*4

calculations

are

checked

for

USING

VAX-11

FORTRAN

IV-PLUS

ALL
Both

OVERFLOW

and

BOUNDS

checks

are

performed.

NONE

Neither
The

default

/CHECK=ALL,

check
is

and

performed.

/CHECK=OVERFLOW.

/NOCHECK

the

If you specify /CHECK=BOUNDS
implicitly canceled.

Note

/CHECK

CONTINUATIONS Qualifier
continuation lines allowed in

This

the

equivalent

/CHECK=NONE.

/CHECK=OVERFLOW,

1.2.2.2
of

that

equivalent

the

other

check

qualifier

specifies the number
program.
It has the form:

source

/CONTINUATIONS=n
You

can

specify

/CONTINUATIONS,

value

the default

from

value

19.

for

you

omit

1.2.2.3
DEBUG Qualifier - This qualifier specifies
that the
compiler
is
to provide information for use by the VAX-11
Symbolic Debugger and
the run-time error traceback mechanism.
It has the form:

/DEBUG

[NO]SYMBOLS
[NO] TRACEBACK
ALL

NONE

SYMBOLS
The

compiler

provides the debugger with local symbol
user-defined
variables,
arrays
(including

for

information),

and

labels

executable

definitions
dimension

statements.

TRACEBACK

The

compiler
provides
an
address
correlation
table
so
the
debugger and the run-time error traceback mechani
sm can translate
absolute
addresses
into
source
program
routine
names
and
compiler-generated line numbers.
ALL

The

compiler
provides
both
address correlation table.

local

symbol

definitions

and

NONE

The

compiler

provides

debugging

information.

The

default

/DEBUG=ALL, and/NODEBUG
is the equivalent
of /DEBUG=NONE. .If you

specify

either

implicitly
For

Chapter

/DEBUG=TRACEBACK.

/DEBUG=TRACEBACK

canceled.

information

debugging

Note

and

that

/DEBUG

the

/DEBUG=SYMBOLS,

traceback,

see

the

Section

egquivalent

other

1.5

and

USING VAX-11l FORTRAN IV-PLUS

with

D LINES Qualifier - This qualifier specifies that lines

1.2.2. 4

a D in column 1 are to be compiled.

It has the form:

/D_LINES

The default is /NOD_LINES, which means that lines with a D in column 1
are treated as comments.

compiler
1.2.2.5 I4 Qualifier - This qualifier controls how a Ehe
length is not
interprets

specified.

INTEGER and LOGICAL declarations for which

It has the form:

/14,

The default is /I4, which causes the compiler to interpret INTEGER and
If you specify
LOGICAL declarations as INTEGER*4 and LOGICAL*4.
/NOI4, the compiler interprets them as INTEGER*2 and LOGICAL*2,

1.2.2.6
file.

LIST Qualifier - This qualifier

It has the form:

produces

listing

source

/LIST[=file-spec]

You can include a file specification for the listing file. If you do
not, it defaults to the name of the first source file, and a file type
of

LIS.

The compiler does not produce a listing file in interactive mode
unless you include the /LIST qualifier. In batch mode, the compiler
produces a listing file by default. 1In either case, the listing file
You must use the PRINT command to
is not automatically printed.
obtain a line printer copy of the listing file.
See Section 1.7 for a sample listing.

1.2.2.7
listing

MACHINE CODE Qualifier - This qualifier specifies that the
file 1is to include a listing of the object code generated by

the compiler.

It has the form:

/MACHINE_CODE

This qualifier is ignored if no listing file is being generated.
The default is /NOMACHINE CODE.

1.2.2.8

OBJECT Qualifier - This qualifier can be used when

to specify the name of the object file.

It has the form:

you

want

/OBJECT [=file-spec]

The default is /OBJECT.
suppress object code;

The negative form, /NOOBJECT, can be used to
for example, when you only want to test the

source program for compilation errors.

If you omit the file specification, the object file

defaults

name of the first source file, and a file type of OBJ.

the

USING VAX-1ll

1.2.2.9
produce

FORTRAN

IV-PLUS

OPTIMIZE Qualifier - This qualifier
optimized code.
It has the form:

tells

the

compiler

/OPTIMIZE
The default is /OPTIMIZE.
The negative form (/NOOPTIMIZE)
should
be
used
to
ensure
that the debugger has sufficient information to help
you locate errors in your source program (see Section 2.9).

1.2.2.10
compiler

WARNINGS Qualifier is
to
generate

warning-level

(W)

errors.

This qualifier
specifies
diagnostic
messages
in

It has

the

whether
response

form:

the
to

/WARNINGS
The compiler generates warning (W) diagnostic messages by default.
A
warning
diagnostic
message
indicates that the compiler has detected
acceptable but nonstandard syntax, or has
performed
some
corrective
action;
in either case, unexpected results may occur.
To suppress W
diagnostic messages, specify
the
negative
form
of
this
qualifier

(/NOWARNINGS).

Appendix

B discusses

1.2.2.11
work

The default

files

compiler

/WARNINGS.

diagnostic messages.

WORK_FILES Qualifier used by the compiler.

This qualifier changes
It has the form:

the

number

/WORK_FILES=n
The value

specified

Note

while

that

for

a value

can be

may

increase

3.
the

restricts
the
size
of
programs that can be
allows larger programs to be compiled, but may
default is /WORK_FILES=2.

1.3

compilation,

compiled.
A value
slow compilation.

of 3
The

LINKING

Before

file

compiled program

produce

references

symbolic

speed

the
$

can

executable

executed,

you

image

file.

in the object code, and establishes
locations.
To link an object module,

following general

form:

LINK[/command-qualifiers]

must

absolute

issue

the

input object

/file-qualifiers
Specify input file options.

the

object

resolves

addresses

LINK

file

to be

linked.

all

for
command,

file-spec[/file-qualifiers]...

/command=qualifiers
Specify output file options.
file-spec
Specifies

1link

Linking

USING VAX-11] FORTRAN IV-PLUS

no
with
In interactive mode you can issue the LINK command
accompanying file specification. The system responds with the prompt:
$§_File:

the prompt.
The file specification must be typed on the same line as
If there are too many file specifications to fit on one line, you can
the
continue the line by typing a hyphen (-) as the last character of
line, and continuing on the next line.

You can enter multiple file specifications separated from each other
When used with the LINK command, the comma
by commas or plus signs.
a
is wused,
no matter which
sign:
plus
the
as
effect
same
has the
If
single executable image is created from the input files specified.

no output file is specified, the linker produces an executable image
with the same name as the first object module, and a file type of EXE.
interest to
Table 1-3 lists the 1linker qualifiers of particular

users.

FORTRAN

the linker.

See the VAX-11 Linker Reference Manual for details on

Table 1-3
Linker Qualifiers

Command Qualifiers

Negative Form

Default

/EXECUTE [=file~spec]

/NOEXECUTE

/EXECUTE

/SHAREABLE [=file-spec]

/NOSHAREABLE

None

/MAP [=file-spec]

/NOMAP

/NOMAP (interactive)

/BRIEF

None

Not applicable

/FULL

None

Not applicable

/CROSS_REFERENCE

/NOCROSS_REFERENCE

/DEBUG

/NODEBUG

/TRACEBACK

/NOTRACEBACK

/TRACEBACK

/MAP (batch)

Input File Qualifiers

/LIBRARY

(s)
/INCLUDE=module-name

1.3.1

Linker Command Qualifiers

You can specify qualifiers for the LINK command to modify the output
You can also define whether the debugging or the
of the 1linker.
traceback facility is to be included.

USING VAX-11

FORTRAN

IV-PLUS

Linker output consists of an image file and, optionally, a
map
file.
The
following
qualifiers
control
the
image
file generated by the
linker:

/EXECUTE=file-spec
/NOEXECUTE
/SHAREABLE=file-spec
These

Map

qualifiers

file

are

qualifiers

described

Section

1.3.1.1.

Section

1.3.1.2.

include:

/MAP [=file-spec]
/BRIEF
/FULL
/CROSS_REFERENCE
These
The

qualifiers

debugger

and

are

described

traceback

qualifiers

are:

/DEBUG

/TRACEBACK
These

qualifiers

1.3.1.1

Image

are

described

File Qualifiers

Section

Image

1.3.1.3.

file

qualifiers

include:

/EXECUTE
/SHAREABLE
If you

not

/EXECUTE;
the
production of an

specify

1linker
image,

image

file

qualifier,

produces
an
executable
specify the negative form,

the

image.
as:

default

1is

suppress

/NOEXECUTE
For

example:

LINK/NOEXECUTE

CIRCLE

The file CIRCLE.OBJ
is
1linked,
but
no
image
is
generated.
The
/NOEXECUTE
gqualifier
is
wuseful if you want to verify the results of
linking an object file, without actually producing the image.

To designate a file specification
/EXECUTE qualifier in the form:

for

executable

image,

use

the

/EXECUTE=file~spec

For

example:
$

The

LINK/EXECUTE=TEST CIRCLE

file

named

CIRCLE.OBJ

TEST.EXE.

linked,

and

the

executable

image

generated

USING VAX-11] FORTRAN IV-PLUS

different
of
number
A shareable image is one that can be used in a
for your own
use
you
1image
private
a
be
It may
applications.
system
the
by
system
the
in
applications, or it may be installed
manager
for
use
by all users.
To create a shareable image, specify
the /SHAREABLE qualifier.

For

example:

LINK/SHAREABLE CIRCLE

To include a shareable image as input to the linker, you must
wuse
an
options
file, and specify the /OPTIONS qualifier in the LINK command.
Refer to the VAX-11 Linker Reference Manual for details.
If you specify /NOSHAREABLE, the effect is similar to /NOEXECUTE.
The
linker processes the object code and the input as though it were going
to produce a shareable image, but in fact no image is generated.

1.3.1.2

whether
is

Map File Qualifiers - The map file qualifiers

a map file

include.

is to be generated,

Map file qualifiers

and,

include:

if so,

the

tell

the linker

information it

/MAP
/BRIEF

/FULL
/CROSS_REFERENCE

The map qualifiers are specified as follows:
/MAP [=file-spec]

/BRIEF

[/CROSS_REFERENCE]

/FULL

The linker uses defaults to generate
or
suppress
a
interactive
mode, the default is to suppress the map;
the default

to generate

map
file.
In
in batch mode,

the map.

If no file specification is included with /MAP, the map file
has
the
name of the first input file, and a file type of MAP.
It is stored on
the default device, in the default directory.

The qualifiers /BRIEF and
included

the map file,

/FULL

define

follows:

the

amount

information

/BRIEF produces a summary of the
a list of contributing modules.

/FULL produces a summary of the image's characteristics and
a
list
of contributing modules (as produced by /BRIEF);
plus a
list of global
a
summary
of
image.

image's characteristics,

and

symbols and values, in symbol name order;
and
characteristics of image sections in the linked

By default, if neither /BRIEF nor /FULL is
specified,
the
map
file
contains
a
summary
of
the
1image's
characteristics
and a list of
contributing modules (as produced by /BRIEF), plus a
1list
of
global
symbols

and values,

symbol

name order.

The /CROSS_REFERENCE qualifier can be used with either the default
or
/FULL
map
qualifiers,
to
request
cross
reference information for
global symbols.
This cross reference information indicates the object

modules

that

linking.

define

USING VAX-11

FORTRAN

and/or

The default

1.3.1.3

Debugging

and

refer

IV-PLUS

global

/NOCROSS_REFERENCE.

Traceback

symbols

Qualifiers

The

indicates
that the VAX/VMS debugger (see Chapter 2)
in the executable image.
The default is /NODEBUG.
When

the

/TRACEBACK

accompanied

qualifier

symbolic

1is

traceback

specified,
showing

the

encountered

/DEBUG

error
sequence

during

qualifier

to be

included

messages
of

are

calls

that

transferred control to the program unit in which the
error
occurred.
If
you
specify
/NOTRACEBACK, this information is not produced.
The
default
1is
/TRACEBACK.
If
you
specify
/DEBUG,
the
traceback
capability
is automatically included, and the /TRACEBACK qualifier is
ignored.
Figure 1-2
illustrates
a
typical
traceback
1list.
(See

Section

1.3.2

1.5.1.)

Linker

Input File Qualifiers

File qualifiers affect the input file
be
object
files;
shareable
files
files.

1.3.2.1

/LIBRARY Qualifier

The

specification.
1Input
previously
1linked;

/LIBRARY

qualifier

has

files
can
or library

the

form:

/LIBRARY
This qualifier specifies that
the
input
library that is to be searched to resolve
in other input modules.
The default file

1.3.2.2

/INCLUDE

Qualifier

The

file
is
an
object-module
undefined symbols referenced
type is OLB.

/INCLUDE

qualifier

has

the

form:

/INCLUDE=module~name
(s)
The qualifier specifies
that
the
input
file
is
an
object-module
library,
and
that
the
modules
named
are the only modules in that
library that are to be explicitly included as
input.
At
least
one
module name is required.
To specify more than one, enclose the module
names in parentheses, and separate them with commas.
The default file
type
1is
OLB.
The /LIBRARY qualifier can also be used with the same
file specification, to indicate that the same library is
also
to
be
searched

1.4
The

The
omit
type
even

for

unresolved

references.

EXECUTION
RUN

command

RUN[/DEBUG]

file

initiates

execution

of your

program.

has

the

form:

file-spec

name must be

specified;

default values

are

applied

you

optional
elements
of the file specification.
The default file
is EXE.
The /DEBUG qualifier allows you
to
use
the
debugger,
though
you
omitted
this
qualifier
from the FORTRAN and LINK

commands.

See Section

1.5

for

details.

USING VAX-11l

1.5

FINDING AND CORRECTING

FORTRAN

IV-PLUS

ERRORS

Both the compiler and the
Run-Time
Library
include
facilities
for
detecting
and
reporting errors.
VAX/VMS also provides the debugger,
to help you

locate

and

correct

errors.

addition

there
errors

1s
a
traceback
facility
that can also be
that occur during program execution.

1.5.1

Error-Related Command Qualifiers

the

used

debugger,

track

down

At each step in compiling, linking, and executing
your
program,
you
can
specify
command qualifiers that affect how errors are processed.
At compile time, you can use
the
/DEBUG
qualifier
to
ensure
that
symbolic information is created for use by the debugger.
At link time
you can also
specify
the
/DEBUG
qualifier
to
make
the
symbolic
information
available
to
the
debugger.
The same qualifier can be
specified with the RUN command, to invoke the debugger.
Table

1-4

summarizes

the

/DEBUG

Qualifier

/DEBUG

and

/TRACEBACK

Table 1-4
and /TRACEBACK Qualifiers

Command

FORTRAN

Effect

LINK

Default

The FORTRAN compiler
creates symbolic data
needed by
debugger.

/DEBUG

qualifiers.

the

TRACEBACK)

Symbolic data created
by FORTRAN compiler is
passed

the

LINK

Traceback information
is passed to the debugger.
Traceback will be produced.

/DEBUG

RUN

Invokes the debugger.
The
DBG> prompt will be displayed.
Not needed if $ LINK/DEBUG
was specified.

/NODEBUG

RUN

If /DEBUG was specified in
the LINK command, RUN/NODEBUG
the

/NODEBUG

debugger.

/TRACEBACK

suppresses

/DEBUG=
(NOSYMBOLS,

DBG>

/TRACEBACK

prompt.

If
you
use
none
of
these
qualifiers
at
any
point
in
the
compile-link-execute sequence, and an execution error occurs, you will
receive a traceback list by default.
However, you will not be able to
invoke the debugger.

To perform symbolic debugging, you must use the /DEBUG qualifier
with
both
the
FORTRAN
command
and
the
LINK
command.
It then becomes
unnecessary to specify it with the RUN command.
If
you
omit
/DEBUG
from
either
the FORTRAN or LINK command, you can use it with the RUN
command, to invoke the debugger.
However, any debugging
you
perform
must
then
be
done
by specifying addresses in absolute form, rather
than symbolically.

USING VAX-11

If you linked your program with
program without intervention by

FORTRAN

IV-PLUS

the debugger,
the debugger,

but wish
specify

execute

the

RUN/NODEBUG program
If you

specify LINK/NOTRACEBACK,

event

shown

error.

in Figure

example

you will
of

receive

source

program

traceback
and

traceback

the
is

1-2.

0001
0002
0003

I=1
CONTINUE
J=2

0004
0005
0006
0007
0008

CONTINUE
K=3
CALL SUB1
CONTINUE
END

0001

SUBROUTINE

0002

I=1

0003
0004
0005

J=2
CALL
END

0001
0002
0003

SUBROUTINE
COMPLEX W
COMPLEX 2

0004

DATA W/(0.,0.)/

0005

0006

END

SUB1l

SUB2

LOG (W)

$MTH-F-LOGZERNEG, logarithm of zero or negative value
user PC 00000449
$TRACE-F-TRACEBACK, symbolic stack dump follows

module

name

routine

name

line

SUB2
SUB1

5
4

T1$SMAIN

Figure

The

traceback

interpreted

1-2

Traceback

relative

absolute

0000074cC

0000081C
00000011
00000017

0000081cC
00000449
00000437

0000001B

0000041B

List

follows:

When the error condition is
detected,
you
receive
the
appropriate
message,
followed
by
the traceback information.
1In this example, a
message is displayed by the Run-Time Library, indicating the nature of
the
error,
and
the
address
at which the error occurred (user PC).
This is followed by the traceback information, which is
presented
1in
inverse
order
to
the
<calls.
Note that values may be produced for
relative and absolute PC, with no
corresponding
values
for
routine
name and line.
These PC values reflect
procedure
calls
internal to
the Run-Time Library.

USING VAX-11l

FORTRAN

IV-PLUS

Of particular interest to you are the
values
1listed
under
"routine
name"
and "line", the first of which shows what routine or subprogram
called the Run-Time Library, which subsequently
reported
the
error.
The
value given for "line" corresponds to the compiler-generated line
number in
the
source
program
listing
(not
to
be
<confused
with
editor-generated
line
numbers).
Using
this
information,
you can
usually isolate the error in a short time.

If you specify either LINK/DEBUG, or RUN/DEBUG, the
debugger
assumes
control
of
execution.
If
an
error
occurs,
you do not receive a
traceback list.
To display traceback information,
you
can
use
the
debugger command SHOW CALLS, as described in Section 1.5.2.

1.5.2

SHOW CALLS

Command

When an error occurs in a program that is executing under the
control
of
the
debugger,
no
traceback
1list
1is
produced.
To generate a
traceback list, use the SHOW CALLS command, which has the form:
DBG>SHOW

1.6

CALLS

SAMPLE TERMINAL SESSION

typical

dialog

between you and

the

system might

appear

(ReT
)
Username:

SMITH

Password:
WELCOME

(Your password is not displayed)

TO VAX/VMS

RELEASE

$ EDIT CIRCLE.FOR
Input:DBA2: [SMITH]CIRCLE.FOR
00100
(enter source program)

*E
(terminate edit session and write
[DBA2: [SMITH]CIRCLE.FOR;1]
$ FORTRAN/NOOPTIMIZE/LIST/DEBUG CIRCLE
$

LINK/DEBUG CIRCLE

$ RUN CIRCLE

file to disk)

follows:

USING VAX-11l

1.7

FORTRAN

IV-PLUS

COMPILER LISTING FORMAT

The listing
produced
sections, as follows:

)

Source

listing

)

Machine

code

Storage

map

Sections 1.7.1
detail.

the

compiler

consists

two

three

section

listing

section

(optional)

section

through 1.7.3

describe

the compiler

listing

sections

USING VAX-11l

1.7.1

Source Listing

FORTRAN

IV-PLUS

Section

The source listing section shows the
the
input file with the addition of
by the compiler.
Figure 1-3 shows a

source program as it
appears
in
sequential line numbers generated
sample source listing section.

Note that line numbers are generated
only
for
statements
that
are
compiled;
comment
1lines
are
not numbered, nor are lines with D in
column 1 unless you specified /D_LINES.
Compiler-generated
use

them

for

line

numbers

debugging

appear

using

the

RLINE

left

margin.

specification

You
in

can

debugger

commands (see Chapter 2).
1If the editor you use to create the
.source
file
generates
line
numbers,
these numbers will also appear in the
source listing.
In
this
case,
the
editor-generated
1line
numbers
appear in the left margin, and the compiler-generated line numbers are
shifted
to
the
right.
The
SLINE
specification
applies
to
the

compiler—-generated

line

numbers,

not

the

editor-generated

numbers.

Compile-time and run-time error messages
that
contain
line
refer
to
the
compiler-generated
1line numbers in the source
section.
See Appendix B for a summary of error messages.

foeol

SUBROUTINE

2002

PARAMETER

eea3
geed

DIMENSION X(@gM,02N)
COMMON X

goes

LOGICAL

Boeé

DONE

oea?
peos

Ms4Q,

Nsb@

DONE

,TRUE,

12
10

J =
I &

{,N=}
{, M=}

2812

(.NOT,

DONE)

2013

RETURN

2014

END

ANEW 8 ( X(I=otpJ)eX(Ieg,J)eX(IpJe1)eX(I,J®1) ) /7
IF ( ABS(XNEWeX(I,J)) ,GT, EPS ) DONE = ,FALSE,
X(I,J) = XNEW
GO

Figure 1-3

1.7.2

numbers
listing

RELAX2(EPS)

00
0O

eaa9
eeio
2011

1line

Source Listing Section

Machine Code Listing Section

The machine code listing section provides a symbolic represen;a?ion of
the compiler-generated object code.
This representation is similar to
a VAX-11l MACRO assembly listing for the generated code and data.

The machine code listing section is optional.
listing file,
$

you must specify:

include

FORTRAN/LIST/MACHINE_CODE

Figure 1-4

shows a sample machine code listing section.

the

USING VAX-11

4000

+TITLE
+ JDENT

RELAXR
o1

+PSECT

SBLANK

9000

P200
0002

+PSECT
RELAX213

0000
00ee

2009

00029
1119

eooF
0016
2216

,113

+WORD

®M<IV,RS,R6,RT,R8,R9,R10,R11>

MOVAL

SLOCAL,

R1i

MNEGL

#1,

DONE(R1Y)

MOVL

¥i,

RY?

MOVL

¥1,

MOVAL
LSIANE

$BLANK,

29210

ADDLY

R9,

goe29
ga2F
203s

ADDF2
ADOF?2
MULF3

X=164(RS) (R12], RO
X+164(RS) [R12]), R@
#¥“X3F8M, RO, RS

°oa30

SUBF3

X(RS)[R1AA],

ADDF3

BICw?
CHMPF

%41,

R7,

R6,

2Qae6

2ea?

20208

)

20n9

401

)

001}

]

2412

X+d4(RS) [R10),

#*X80003, RO
RO, OEPS(AP)

0948
@240
QO4F

BLEQ
CLRL
LSIAP]s

0O4F

MOVF

R8,

2057

AOBLEQ

#59,

p0%8
POSF
P063

MOVL
MOVF
MOVL

R7,
R8,
R9,

2067
BR6A

BLBC
RET

DONE(R11),

20583

MULLS
LSIAGGH

2042
0047

IV-PLUS

sCODE

0919
2010

on21

FORTRAN

X=d4(RS5)([R1Q]),

RS,

LSIAP]
NDONE(R1Y)

AOBLEQ

439,

X(RS)[R10Q)

R9,

LSIAGG

R7,

LSIANE

J(R11)
XNEW(R1})

I(R11)

+END

Figure 1-4

Machine Code Listing Section

The first line of the machine code listing contains a .TITLE assembler
directive,
indicating
the
program
unit
to
which the machine code
corresponds.
For a main program, the title is either as declared in a
PROGRAM
statement, or filenameS$SMAIN, if you did not specify a PROGRAM
statement.
For subprograms, the title is the name of
the
subroutine
or

function.
For
a
BLOCK DATA subprogram, the title is either
name declared in the BLOCK DATA statement, or
filename$DATA,
if
did not specify a name in the BLOCK DATA statement.

the
vyou

The linesfollowing
.TITLE provide information such as the contents of
storage initialized for FORMAT statements, DATA statements, constants,
and
subprogram
argument
call
1lists.
Machine
instructions
are
represented
by VAX-11 MACRO mnemonics and syntax.
Compiler-generated
line numbers corresponding to generated code lines are listed
at
the
right margin before the machine code generated for the line.

The VAX-11 general registers

(0 through

12)

are

represented

through
R12.
When register 12 is used as the argument pointer, it is
represented by AP;
the frame pointer (register 13) is FP;
the
stack
pointer
(register 14) is SP, and the program counter (register 15) is

USING VAX-11]

PC.

Note

listed

that

the

relative

left margin,

Variables and arrays
machine code listing
variables and arrays
FORTRAN

the

source

machine

source

program

for

each

IV-PLUS

instruction

hexadecimal.

data

item

defined in the source program are
shown
in
as they were defined in the source.
Offsets
are shown in decimal.

labels

code

FORTRAN

referred

listing

refers

dot

with

label

300,

the

source

(.)

the

program

prefix.

label

For

appears

are

shown

example,
in

the

the
from

the

machine

code
1listing
as .300.
Labels that appear in the source program, but
that are not referred to or are deleted during compiler
optimization,
are
ignored and do not appear in the machine code listing, unless you
specified /NOOPTIMIZE.

The compiler may generate labels for its own use.
as
L$xxxx, where the value of xxxx is unique for
program unit.
Integer

constants

precision,

values

and

preceded

Addresses

are

shown

complex

“X.

signed

constants

represented

integer

are

the

shown

program

These labels appear
each such label in a

values;

real,

unsigned

section

name

hexadecimal
offset
within
that
program
section.
Changes
program section to another are indicated by PSECT lines.

1.7.3

the

)

Program

)

Entry

)

Statement

)

Variables

]

Arrays

)

Labels

Functions

()

Total memory allocated

Figure

1-5

summary

for

plus

the

from one

Storage Map Section

The storage
map
section
following information:

double

hexadecimal

that

compiler

listing

summarizes

the

sections

points

shows
section

section.

functions

and

subroutines

sample
heading

storage
is

not

map

section.

printed

entries

were

generated

USING VAX-11l FORTRAN IV-PLUS

PROGRAM

SECTIONS

Name

Bytes

Attributes

2 $CODE
€ SLOCAL

107
16
10004

EXE
SHR
PIC CON REL LCL
PIC CON REL LCL NOSHR NOEXE

3 SBLANK

PIC OVR REL GBL

SHR NOEXE

RD NOWRT LONG
WRT LONG
RD
WRT LONG
RD

ENTRY POINTS

Tvype Name

Address

d=00002000

RELAX2

VARIABLES

Address

Tvype Nanme

2~000000089

Lx4 OONE

2=00000804

Type Name

Address

Type Name

APegoooo0n04e L4 EPS

2=Qno0a0@8

Ix4 I

Address

2*0000m2AC

Ix4 J

Rwy XNEW

ARRAYS

Address

3=00000000

Tvpe Name

Bytes

Dimengions

Rweg X

100084

(0140,0360)

LABELS

0=00000009
Tota)

Address

Lebe!

Address

Space Allocated =

10127 Rytes

Label
10

COMPILER OPTIONS
/CHECK =(NOBOQUNDS, OVERFLOW)
/DEBUG=(NOSYMBOLS, TRACEBACK)

/OPTIMIZE

/WARNINGS

/14

/NOD,LINES

Figure 1-5

Storage Map Listing

Data
in decimal.
Sizes are printed as a number of bytes, expressed
from the start of a program
an offset
as
specified
are
addresses
section, expressed in hexadecimal.
The symbol AP can
appear
instead
of
a
program
section;
in this case, the address refers to a dummy
(AP).
argument pointer
the
from
argument, expressed as the offset
Indirection is indicated by an at sign (@) following an address field.

(or AP)
In this case, the address specified by the program section
to the address of the data, not to the data
the offset points
plus
itself.

each program
The program section summary describes
The descriptions include:
generated by the compiler.
°

PSECT number

Name

Size

Attributes

section

(used by most of the other summaries)

(PSECT)

USING VAX-11 FORTRAN IV-PLUS

Chapter 7 describes PSECT usage and attributes.

addresses.
their
and
The entry point summary lists all entry points
If the program unit is a function, the declared data type of the entry
point

is also

included.

The statement function summary lists the entry point address and

data

The variable summary lists all simple variables, with

type

type of each statement function.

and address of each.

the

data

In addition to
The array summary is similar to the variable summary.
data type and address, it gives the total array size and dimensions.
the size is shown as double
If the array is an adjustable array,
and each adjustable dimension bound is shown as a
(**),
asterisks
single asterisk

(*).

FORMAT
labels.
statement
user-defined
The label summary lists all
If the label
suffixed with an apostrophe.
statement labels are
not
was
address field contains double asterisks (**), then the label
used or referred to by the compiled code.

external

routine

total
the
prints
compiler
the
summaries,
the
Following
allocated for all program sections compiled, in the form:

memory

The functions and

subroutines

references made by the

summary

source program.

1lists

all

Total Space Allocated = nnn Bytes

1.7.4

Other Listing Information

The final entries on the compiler listing are the compiler
in effect for

that compilation.

For example:

COMPILER OPTIONS

/CHECK= (NOBOUNDS ,OVERFLOW)
/DEBUG= (NOSYMBOLS ,TRACEBACK)

/OPTIMIZE

/WARNINGS

/I4

/NOD_LINES

qualifiers

CHAPTER 2

DEBUGGING FORTRAN PROGRAMS

Debugging
is
the process
of
finding
and
correcting
errors
in
executable programs;
that is, in programs that have been compiled and
linked without diagnostic messages, but that produce invalid results.
This chapter shows you how to use
debug

2.1

FORTRAN

the

VAX-1ll

Symbolic

Debugger

programs.

OVERVIEW OF THE VAX-1ll

SYMBOLIC DEBUGGER

The VAX-11 Symbolic Debugger Reference
Manual
describes
the VAX-11
debugger
in
detail.
This
section
provides
an
overview
of
the
debugger, showing a
sample
debugging
session
and
introducing
the
debugger

2.1.1

command syntax and

symbol

table.

sSample Debugging Terminal Session

Figure 2-1 illustrates a program that requires debugging.
The program
has been compiled and linked without diagnostic messages from either
the
compiler
or
the
linker.
(Appendix B
summarizes
compiler
diagnostic
messages.) However, the program produces erroneous results
because
of
the
missing
asterisk
1in
the
exponentiation
operator
(RADIUS*2
should
be
RADIUS**2),
This error is so obvious that you
hardly need the services of the debugger to
find
it.
However,
for
purposes
of
illustration,
this
example will deal with the error as
though it were shrouded
C
C

0001
0002

0003

0004
0005
0006
0007
0008

0009

in obscurity.

PROGRAM TO FIND THE
OF A CIRCLE

AREA

PROGRAM CIRCLE
TYPE 5

FORMAT

enter

radius value

FORMAT

area of circle equals

ACCEPT 10,RADIUS
FORMAT (F6.2)
PI = 3.1415927
AREA = PI*RADIUS*2
TYPE 15,AREA

0010
0011

STOP
END

Figure

2-1

Sample FORTRAN Program:

2-1

CIRCLE

',F10.3)

DEBUGGING

The

key

your

debugging

program.

selected
if
they

find

this,

locations, and
contain
the

FORTRAN

out

what

you

look at
correct

PROGRAMS

happens

need

way

critical
stop

points

execution

the contents of these locations to see
values.
Points at which execution is

stopped

are called
breakpoints.
specify where you want to stop the

The

SET
program.

BREAK

command

lets

vyou

To look at the contents of a location, use the
EXAMINE
command.
To
resume
execution,
use
either
the
GO
or
STEP command.
All DEBUG
commands relevant to FORTRAN are discussed in subsequent
sections
of
this chapter.

Figure

2-2

session.
that

The

an example

circled

the

figure

typical

numbers

and

explain

$ FORTRAN/LIST/NOOFTIMIZE/DERUG
$
$

LINK/DEBUG
RUN CIRCLE
DERUG

Version

0,.5-1

Arril

for

debugging

are keyed to notes

dialog.

1978

’

DRG>=GO

start

radius

the

CIRCLE

ZDEBUG~-I-INITIAL» language is FORTRANy
routine

dialog

CIRCLE

DEBG>SET BREAK ZLINE 7 é
enter

terminal

(for example,

score and module set to CIRCLE

‘a

CIRCLENCIRCLE

value

24,

bresk at rc = CIRCLENCIRCLE Xline 7

‘,

NBGEXAMINE PI 6
CIRCLENFI?

3.141593

DRGEXAMINE

RADIUS

CIRCLENRADIUS?

24.00000

NDBEG>EXAMINE AREA
CIRCLENAREA?
0.0000000
DEG>GO

start
area

rpc
of

FORTRAN

CIRCLENCIRCLE Xline 7
eauals
150.796

circle
STOF

ZDEBUG-I-EXITSTATUSy is “XSYSTEM~-S—-NORMALs normal successful comrletion’

flb

DBGHEXIT
$

Figure

2-2

Sample Debugging Terminal

Invoke the FORTRAN compiler,
You

should

symbolic
Link
In

your

include

debugging
program,

response

the

(see

Section

including

the

specifying

/NOOPTIMIZE

RUN

the

qualifier

Dialog

qualifiers
if

you

intend

shown.
to

use

2.9).

the debugger.

command,

the

debugger

displays

its

identification,
indicating
that
your
program will be executed
under
the
debugger's
control.Following
the
identification
message,
the
debugger
displays an INITIAL message, telling you
the mode settings it has assumed for
the
1language,
scope,
and
module.
The
debugger
derives the mode settings from the first
module specified in the LINK command.
If this message
does
not
appear,
or
1if the settings assumed are not appropriate, use the
SET

LANGUAGE,

Sections

SET

2.2.1,

SCOPE,

2.2.2,

and

SET

2.2.3.

MODULE

commands,

described

DEBUGGING FORTRAN

PROGRAMS

Place a breakpoint at an appropriate point in the program.
This
point
should
be
one
at
which you will be able to examine key
variables.
Note:
Breakpoints suspend execution just before
the
point specified.

Begin program execution.
The debugger displays
at which execution starts.
The debugger announces
specified breakpoint.

that

it has

suspended

the

1line

execution

number

the

Check the variable PI to make sure the correct
value
1is
stored
there.
The
debugger
displays the contents of PI, showing that
its scope is in module CIRCLE.
Check the variable RADIUS.
The debugger
value has been properly stored.
Examine

the

variable AREA

to make

Resume execution.
The debugger
point of program resumption.

sure

shows

its

displays

that

the

specified

contents

are

zero.

a message

Successful
completion
of
the
program
1is
message.
However, as you can see, the result

flb

indicating

indicated
by
is incorrect.

the

this

Exit from the debugger.

By examining the variables PI, RADIUS, and
AREA
as
the
program
is
executing, you can determine that the correct values are being stored.
It follows, then, that the error is probably in the expression of
the
formula for computing the area.
To correct the problem, you must edit
and recompile the source program,
with
the
exponentiation
operator
properly specified in the formula expression.

2.1.2

Debugger

Debugger
is:

Command

commands

DBG>command

Syntax

resemble

other

[keyword]

[operand

command
Specifies

the

keyword
Specifies

the qualifiers

command

VAX/VMS

commands.

[,operand]

The

general

form

...]

name.

for

SET,

SHOW,

CANCEL

commands.

operand

Specifies the object of the command.
The operand may consist
of
constants, names of variables or array elements, or expressions.
The

command,

keyword,

and

operand

fields

are

separated

one

can process integer, real, double precision,
but not complex or character expressions.

and

spaces.

The debugger
expressions,

1logical

The

FORTRAN

decreasing

DEBUGGING

FORTRAN PROGRAMS

supported

operators

order

the

debugger

are

listed

below,

concatenation,

precedence.

*,/

+,-

.NOT.
.AND.

.OR.
.XOR.,

The

.EQV.

debugger

relational

does

not

support

exponentiation,

operators.

The debugger
evaluates
expressions
in
the
same
way
as
FORTRAN.
However,
the syntax of expressions is slightly different.
Spaces are
used to separate elements of debugger commands, and are significant to

the
debugger;
(such as .OR.)
The
the

therefore, variable names and multicharacter
must contain no embedded spaces.

debugger accepts constants in the same form used
following
exceptions:
Hollerith,
Radix-50,

constants

Debugger
forms.

(for

example

"777)

commands observe

are

not

standard

accepted

VAX/VMS

the

for
and

operators

FORTRAN, with
octal integer

debugger.

conventions

for

abbreviated

Table 2-1 lists the debugger commands and keywords discussed
in
this
chapter,
and
shows
their full and abbreviated forms.
Abbreviations
are in parentheses next to the full form.

Table
Debugger

2-1

Commands

and

Keywords

Command
Names
SET

Keywords

(SE)

SHOW

LANGUAGE

(SH)

(CAN)

SCOPE

(SC)

EXAMINE

(E)

BREAK

(B)

TRACE

(T)

WATCH

(W)

(EV)

DEPOSIT

(D)

DEFINE

(DEF)

EXIT

(EXI)

STEP

(S)

(G)

CALL

Debugger

(LA)
(MODU)

CANCEL

EVALUATE

2.1.3

MODULE

(CA)

Symbol

Table

The debugger maintains a table of symbols defined by the program
with
which
it
1is
1linked.
This
table
provides the name of each symbol
defined in the program, its data type, and
its
address.
The
table
also
provides
dimension
bound
information
for
arrays, and length
information for character data.
The debugger's
2000
symbols.
defined in the
more
than one

active symbol table
provides
room
for
approximately
Thus,
you
should
pay heed to the number of symbols
programs you are debugging.
If your
program
contains
program unit, use the SET MODULE command to be sure the

DEBUGGING

FORTRAN

PROGRAMS

symbol table contains symbols from
the
program
units
you
wish
to
debug.
Use
the
CANCEL
MODULE command to remove symbols defined in
program units that no longer
need
debugging.
The
SET
MODULE
and
CANCEL MODULE commands are defined in Section 2.2.2.

2.2

PREPARING TO DEBUG A PROGRAM

The following sections describe the commands
proper

environment

for

debugging

used

establish

FORTRAN programs.

These

the

commands

are:

SET LANGUAGE
SHOW LANGUAGE
SET

MODULE

SHOW MODULE
CANCEL

SET
SHOW

MODULE

SCOPE
SCOPE

CANCEL

SCOPE

These commands can be used if
debugger are not appropriate.

2.2.1

SET,

SHOW LANGUAGE

the

1initial

settings

assumed

the

Command

The SET LANGUAGE command tells the debugger that the debugging
dialog
is
to
be
conducted
according
to
the conventions of the specified
language.
For example, if
you
specify
SET
LANGUAGE
FORTRAN,
the
debugger will accept and display numeric values in decimal radix.
The

command
SET

has

the

LANGUAGE

language
Specifies

the

form:

language

To determine which language is
LANGUAGE command.
This command
SHOW

used.

«currently
in
has the form:

effect,

wuse

the

SHOW

effect.

For

LANGUAGE

The debugger

responds

example:
DBG>SHOW LANGUAGE
language: FORTRAN

displaying

the

1language

DEBUGGING

2.2.2
The

SET,

MODULE

SHOW,

CANCEL MODULE

Commands
the

debugger's

consists of
following

SET

MODULE

let you

PROGRAMS

active
symbol
table
when
the program you want to debug
multiple
program
units.
These
commands
perform
the
functions:
e

commands

FORTRAN

places

unit
into
the active

the

symbols

contents

defined

the

specified

program

the active symbol table.
The debugger initializes
symbol table to include
all
global
symbols,
and

local

symbols
command.

The

control

the

first

program

unit

specified

the

LINK

SHOW MODULE displays the names
of
all
program
units
whose
symbols
are
potentially
available.
"Yes" means the symbols
for that module are set;
"no" means they are not set.

CANCEL MODULE removes the
specified
from the active symbol table.

SET MODULE

command

SET MODULE

has

the

program

unit's

symbols

form:

program-unit
/ALL

[,program-unit]

...}

program-unit
Specifies the name of the program unit whose
included in the active symbol table.

symbols

are

/ALL
Requests the debugger to set the symbols of
all
known
modules.
If
there
1is
insufficient space, the debugger displays an error
message.

The

SHOW MODULE
SHOW

command

has

the

form:

MODULE

This command takes no parameters.
The debugger responds by displaying
the
names
of
the
modules
1linked with the debugger, indicating the
modules whose symbols are included in
the
image,
and
their
sizes.
Only
those module names marked "Yes" have their symbols in the active
symbol table.
The

CANCEL

MODULE

command

CANCEL MODULE

has

/ALL

2.2.3

/ALL

removed.

Specifies that
symbol table.

SET,

SHOW,

form:

program-unit [,program-unit]

program-unit
Specifies the name of
be

the

all

CANCEL

...}

the program unit for which
'

information

SCOPE

purged

symbols

from

the

are

active

Commands

The SCOPE commands
let
you
control
the
default
used
to
resolve
references
to
symbols.
When you use a command such as EXAMINE, you
can either specify or omit the name of the module in which the
symbol

2-6

DEBUGGING FORTRAN PROGRAMS

is defined.

If you omit the name,

the debugger uses a default.

If it

cannot find the symbol in the default scope, the
debugger
creates
a
scope,
based
on
the
<current
value
of the PC.
This indicates the
module or routine in which your program stopped.
If that
fails,
the
debugger
attempts
to
find
an
unambiguous
symbol in the remaining
program units.
A message is displayed if the debugger cannot
resolve
the reference.

The SCOPE commands perform the following functions:
®

SET SCOPE defines the

SHOW SCOPE displays the current default program unit name

CANCEL SCOPE revokes the default program unit named previously
in

a SET SCOPE

The SET SCOPE

program-unit
Specifies
For

the default

command

command has

SET SCOPE

specified program unit to be

the

form:

program-unit

the name of

the program unit

to be used

the default.

example:

SET SCOPE

MAXI

The SHOW SCOPE command has

the

form:

SHOW SCOPE

This command takes no parameters.
current

The CANCEL SCOPE
CANCEL

This command

2.3

The

symbol displayed

indicates

the

scope.

command

has

the

form:

SCOPE

takes no parameters.

CONTROLLING

Scope becomes

<null>.

PROGRAM EXECUTION

To see what happens during execution of your program, you must be able
to
suspend
and resume the program at specific points.
The following
commands are available for these purposes:
SET

BREAK

SHOW

CALLS

SHOW BREAK
CANCEL

BREAK

GO
STEP

SET TRACE

SHOW TRACE
CANCEL

CTRL/Y

TRACE
EXIT

SET WATCH
SHOW WATCH
CANCEL

WATCH

DEBUGGING

2.3.1
The

SET,

BREAK

SHOW,

the

program.

SET

let

you

The

you

select

can examine

BREAK

commands

BREAK defines

suspend

PROGRAMS

CANCEL BREAK Commands

commands

suspension,

FORTRAN

specified

and/or
perform

address

1locations

for

modify ‘variables
the

following

1line

program

arrays

functions.

number

which

the

execution

SHOW BREAK

displays

all

breakpoints

currently

set

program

The

CANCEL

BREAK

SET

BREAK

command

SET

removes

has

BREAK[/AFTER:n]

selected

the

breakpoints

form:

address

[DO(debugger

command(s))]

address
Specifies the address at which the breakpoint is to occur.
Note
that
execution
1is
suspended just before the specified address.
Section 2.5 describes how addresses are specified.
DO (debugger command(s))
Requests that the debugger perform
the
any, when the breakpoint is reached.
For

commands,

example:
SET

BREAK

The result
breakpoint
You

can
a

time

is
at

use

effect.
of

specified

3%LINE

that
line

the

following

if you

and

through

the
100

DO (EXAMINE

TOTAL;

variables TOTAL
is reached.

/AFTER qualifier

Thus,
loop,

100

EXAMINE

and AREA

control

are

when

set a breakpoint on a line

you want

the loop,

the

then

breakpoint

specify

the

AREA)

examined

breakpoint

that

effective

/AFTER

when

switch

the
the

shown

the

takes

range
third

the

example:

DBG>SET

BREAK/AFTER:3

$LINE

Note that if you use the /AFTER qualifier, the breakpoint is
reported
the
nth
time
it
1is
encountered,
and every time it is encountered
thereafter.

The

SHOW

BREAK

SHOW

command

has

the

form:

BREAK

This command takes no parameters.
the location of breakpoints.
The

CANCEL

BREAK

CANCEL

BREAK

command

has

address

the

The

debugger

responds

by displaying

i
form:

[,address...]

/ALL
address
Removes

the

breakpoint(s)

all

breakpoints

the

specified

/ALL
Removes

the

program.

address(es).

DEBUGGING

2.3.2

SET,

SHOW,

CANCEL TRACE

FORTRAN

PROGRAMS

Commands

The TRACE

commands let you set, examine,
and
remove
tracepoints
in
your
program.
A
tracepoint
1is
similar to a breakpoint in that it
suspends program execution, and displays the address at the
point
of
suspension.
However,
program
execution resumes immediately.
Thus,

tracepoints let you follow the sequence of program execution to ensure
that execution is being carried out in the proper order.

Note that tracepoints and breakpoints are
mutually
exclusive.
That
is,
1if
you
set
a
tracepoint
at
the
same
1location as a current
breakpoint, the breakpoint will be canceled, and vice versa.
The TRACE

commands perform the

following

functions:

SET TRACE establishes
points
within
execution is momentarily suspended

the

program

which

SHOW TRACE displays the locations
tracepoints are currently set

the

program

which

CANCEL TRACE
the

removes one or more

tracepoints currently set

program

The SET TRACE

command

SET TRACE

has

the

form:

address

Specifies
The

SHOW TRACE

the address at which
command

has

the

tracepoint

to occur.

form:

SHOW TRACE

This

command

takes

The CANCEL TRACE

CANCEL TRACE

address
Removes

/ALL

2.3.3

parameters.

command has

address

/ALL

the

form:

[,address

the

tracepoint(s)

Removes

all

tracepoints

SET,

SHOW,

CANCEL WATCH

the

...]}

specified

address(es).

the program.

Commands

The WATCH commands let you monitor specified
when
attempts
are
made
to
modify
their

1locations
contents,

to
determine
and
take the
appropriate actions.
These locations are called watchpoints.
When an
attempt
1is
made
to
change a watchpoint, the debugger halts program
execution, displays the address of the instruction, and prompts for
a
command.
Watchpoints
are
monitored
continuously.
Thus,
you can
determine whether locations are being
modified
inadvertently
during
program execution.

DEBUGGING

The WATCH

commands

SET WATCH

SHOW WATCH

CANCEL

perform
defines

the
the

displays

WATCH

FORTRAN

PROGRAMS

following

functions:

location(s)

the

disables

locations

monitoring

monitored

currently

being monitored

specified

Specifies the location to be monitored.
variables and array elements.

You can

monitor

breakpoints

are

locations

The SET WATCH command has fhe form:
SET WATCH

var

For

scalar

example:

SET WATCH AREA
Note

that

watchpoints,

exclusive.
The

SHOW WATCH

command

tracepoints,

has

the

and

mutually

form:

SHOW WATCH

This

The

command

takes

CANCEL WATCH

no parameters.

command

has

CANCEL WATCH

(var }
/ALL

Specifies

location

the

All

watchpoints

are displayed.

form:

var

the

for which monitoring

be disabled.

/ALL
Removes
For

all

watchpoints

from

the

program.

example:

CANCEL WATCH AREA

2.3.4

SHOW CALLS Command

This command
particularly

CTRL/Y

can be

SHOW CALLS

The

debugger

leading
most

used

to produce

useful when you have
command.
It has the form:

recent

described

traceback

returned

the

calls,

debugger

and

following

is
a

[n]

displays

the
calls

traceback

current module.
are

displayed.

in Section 1.5.

list,

showing

If you
The

form

the

include
of

the

sequence

a value

for

traceback

calls

the

1list

n
is

DEBUGGING FORTRAN PROGRAMS

2.3.5

GO,

STEP Commands

These commands let you initiate and resume program execution.

GO initiates execution at a specified location, and
to conclusion or to the next breakpoint

STEP

execution

initiates

from

the

continues

1location,

current

continues for a specified number of statements

and

The form of the GO command is:
GO

[address]

address

Specifies the address at which program execution is to begin.

The address parameter is optional;
at the current location.

if you omit it,

execution

starts

NOTE

the
You must not restart a program from
beginning unless you first exit from the
be
Unspecified results will
debugger.
produced.

The form of the STEP command is:
STEP[/qualifiers]

[n]

be
to
statements
The value specified for n determines the number of
specify 0, or omit n, a default of 1 is assumed.
If you
executed.

Note, however, that if you issue a STEP command while your program is
in a module whose symbols are not set in the active symbol
stopped
table, then n instructions

(not statements) will be executed.

You can specify the following qualifiers for the STEP command:
/[NO] SYSTEM
/OVER
/INTO

/LINE
/INSTRUCTION

/ [NO] SYSTEM

If
you
specify
/SYSTEM,
debugger
to
count
steps

you
are
wherever

including system address space.

telling
the
they
occur,
The
default
is

/NOSYSTEM.

/OVER

subprograms
- Tells the debugger to ignore calls to
That is, it is
steps through the program.
it
as
to step over

the call.

This is the default.

/INTO

to
calls
recognize
to
debugger
the
- Tells
That
subprograms as it steps through the program.
is, it is requested to step into the subprogram.

/LINE

- Tells the debugger to step through the program
a line-by-line basis (default for FORTRAN).

/INSTRUCTION

on
Tells the debugger to step through the program
an
1instruction-by-instruction
basis (default for
VAX-11

MACRO).

DEBUGGING

You
or

can
you

specify
can

SET

use

STEP

FORTRAN

PROGRAMS

each

time

you

shown

these qualifiers
a

SET

STEP

command,

INSTRUCTION,

INTO,

issue
the

STEP

following

command,
example:

SYSTEM

This
command
specifies
that
all
defaults
applicable
to
FORTRAN
programs
are
to
be
overridden.
When you subsequently issue a STEP
command with no qualifiers, these qualifiers
are
assumed
to
be
1in

effect.

with

You

a STEP

can,

however,

command.

STEP/LINE

supersede

them by

including a qualifier

Thus,

tells the debugger
command.

execute

lines,

regardless

the

SET

STEP

It is advisable to use STEP to execute only a few
instructions
at
a
time.
To
execute
many instructions, and then stop, use a SET BREAK
command to set a breakpoint, and then issue a GO command.

2.3.6
You

CTRL/Y Command

can

use

the

CTRL/Y

command

any

time

return

the

system

command level.
This command is issued when you press the CTRL key and
the Y key at the same time.
The $ prompt will
be
displayed
on
the
terminal.
To
return
to
the debugger, type DEBUG.
You can use the
CTRL/Y command if your program loops or otherwise fails to stop
at
a
breakpoint.
To
find
out
where
you were at the instant CTRL/Y was

executed,

use the SHOW CALLS command after you return to the debugger.

See Section

2.3.7
The

2.3.4.

EXIT Command

EXIT

terminate

command

lets

a debugging

you

exit

session.

from

the

It has

debugger

the

when you

are

ready

form:

EXIT

This command
your program

2.4

takes no parameters.
terminates to return

You must
to system

use the
command

EXIT command
level.

when

EXAMINING AND MODIFYING LOCATIONS

Once you have set breakpoints and begun program
execution,
the
next
step
1is
to
see
whether
«correct
values
are
being generated and,
possibly, to change the contents of locations as
execution
proceeds.
You may also want to calculate the value of an expression that appears
in your program.
The debugger provides
the
following
commands
for
these

purposes:
EXAMINE

DEPOSIT
EVALUATE

DEBUGGING FORTRAN

2.4.1
The

EXAMINE

locations.

Command

command

It has

EXAMINE

PROGRAMS

1lets

the

you

1look

the

contents

specified

form:

[address|:address]]

address
Specifies the address whose contents are to
usually
given
symbolically
as a variable

be examined;
it
1is
name or array element

name.

Examples:
EXAMINE

The

IZZY

contents of variable
EXAMINE

The

IARR(I)

contents of
EXAMINE

IZZY are displayed.

the

Ith element

the

first

through

You can also specify that
displayed.
For example:
EXAMINE

600

contents

2.4.2

IARR are displayed.

IARR(1l) :IARR(10)

The contents of
examined.

The

in array

DEPOSIT

absolute

the

tenth

elements of

contents

address

600

are

the

array

absolute

IARR

are

address

displayed.

Command
L]

The

DEPOSIT

locations.

command

It has

DEPOSIT

lets

the

you

change

contents

specified

form:

address=value{,value

address
Specifies

the

address

the

value

...

into which

]

the

value

deposited.

value
Specifies

deposited.

You

can change the contents of a
specific
1location,
or
consecutive locations, as shown in the following examples.
DEPOSIT

This

This
1,

the decimal

value

100

into

the

variable

IZZY.

IARR(1)=100,150,200

command places
and

several

I1I2ZY=100

command places
DEPOSIT

array

the decimal values
IARR.

100,

150,

and

200

into elements

DEBUGGING FORTRAN

2.4.3
The

EVALUATE Command

EVALUATE

determine

command

the

value

EVALUATE

value

EVALUATE

the

use

the

debugger

expression whose

has

the

calculator,

form:

value

be determined.

PI*RADIUS

this

command

expression will
to

determine

EVALUATE /ADDRESS
For

you

expressions.

example:

EVALUATE

The

lets

expression

expression
Specifies
For

PROGRAMS

displayed.

addresses,

You

can

also

use

the

such

follows:

expression

example:

EVALUATE /ADDRESS

This

calculates

the

address

the

variable

in decimal.

EVALUATE/ADDRESS A (J)

This

calculates

You can also
computing an

the

address of

element of

array A.

use EVALUATE
to
perform
address
arithmetic,
offset or array element address.
For example:

EVALUATE /ADDRESS

2.5

the Jth

I+4

SPECIFYING ADDRESSES

The debugger allows you to express addresses in symbolic form.
Thus,
to examine a location, you need only refer to it by its symbolic name.
You don't have
to
concern
yourself
with
its
1location
in
memory
(unless,
of course, you omitted the /DEBUG qualifier from the FORTRAN
and LINK commands).
Simply specify the variable,
array
element,
or
function name in the debugger command.
You
also
need
to
tell
the
debugger
where
to
set
breakpoints,
watchpoints,
and tracepoints.
The following sections describe how to
specify line numbers, statement labels, and absolute addresses.

2.5.1

Lines,

Labels,

and Absolute Addresses

Addresses can
be
specified
by
1line
number,
statement
1label,
or
absolute
value.
To
specify a line number or a statement label, use
either a $LINE prefix or a $LABEL prefix, respectively.
For example:
SET

BREAK

3SLINE

This command sets
a
breakpoint
at
1line
6,
corresponding
to
the
compiler-generated
1line
numbers shown in the listing.
Note that the
debugger does not recognize all
1line
numbers,
in
particular
those
associated with non-executable statements.
If you specify such a line
number, the debugger responds with a message indicating that
no
such

DEBUGGING FORTRAN PROGRAMS

line exists. Simply retry the command, specifying the line number of
an executable statement. To specify a statement label, specify a

command such as:

SET BREAK $%LABEL 7

label

This command sets a breakpoint at statement

identified by the current scope (see Section 2.2.3).

To specify an absolute address, do not use a prefix.

module

the

For example:

SET BREAK 700

You can also enter absolute addresses in symbolic form.

have

must

defined

(see Section 2.5.4).

To do so, you

them symbolically, by means of the DEFINE command

Specifying Scope

2.5.2

If the program you are debugging consists of

one

than

program

unit, you must be sure that your symbol references are unambiguous.
For example, if your main program calls a subroutine, and the symbols
from both program units are in the debugger's symbol table, you must
distinguish between duplicate symbols.

For example,

assume

that

you

want

set

subroutine, and you issue the following command:

breakpoint

the

1in

SET BREAK %LINE 10

Because you do not specify a program unit name in this command, the
debugger uses a default to decide which line 10 you mean. If you used
a SET SCOPE command, the debugger uses the program unit specified in
the SET SCOPE command (see Section 2.2.3). To override this default,

you must specify a command in the following general form:
SET BREAK $LINE program-unit\1l0
For

example:

SET BREAK %LINE ARGO\1l0

This command specifically calls for a breakpoint to be set at line

in the program unit named ARGO.

Unambiguous references are also required when you specify variables.
If there are duplicate variable names (for instance, X) you should
specify which X you want, as in the following example:
EXAMINE SUB3\X

Previous, Current, and Next Locations

2.5.3

The debugger provides a quick method for referring

any

locations:

The previous location

The current location

The location at the next higher address (next location)
2-15

three

DEBUGGING

specify

().

the

FORTRAN

location,

For example:

PROGRAMS

type

up-arrow

(%)

the

location.

circumflex

EXAMINE

This
To

command

specify

displays

the

DEPOSIT

the

current

contents

location,

type

dot

(.).

For

example:

.=100

This command puts a decimal value of
100
in
the
current
location.
This
method
is
most
useful after you have looked at a location and
decided to change it;
or when you
want
to
verify
that
a
DEPOSIT
command

has

been

specify the
entirely.
For

executed

next higher
example:

expected.

location,

simply

omit

the

address

value

EXAMINE

The

2.5.4
You

location's

Defining Addresses

may

do
so,
symbolic

occasionally

address
DEFINE

command

The
the

Symbolically

need

access

absolute

addresses.

help

you

has

the

form:

For

TOP=1036

references

this

address

example:

DEPOSIT

2.6

displayed.

example:

Subsequent

The

name=address

DEFINE

TOP.

will

the
debugger
provides
the
DEFINE command, which creates a
reference for an absolute address.
Then you can refer to the
by its symbolic name, rather than by its absolute value.
The

DEFINE

For

contents

can

made

using

the

symbol

TOP=256

address

1036

will

changed

256.

CALLING SUBROUTINES FROM THE DEBUGGER
CALL command
form:

lets

you

call

Specifies

the

subroutine

Specifies

one

or more

subroutine

name.

actual

arguments.

from

the debugger.

has

DEBUGGING FORTRAN PROGRAMS

On return from the subroutine, control returns to the debugger, at the
point
at which
the
CALL
command was issued.
The context (general
registers, etc.) that
existed
at
the
time
of
the
CALL
1is
also
restored.

When calling FORTRAN routines, you must adhere to the FORTRAN
5.

conventions described in Chapter

2.7

calling

DEBUGGER COMMAND QUALIFIERS

Qualifiers can be used to modify some debugging commands.
form in which qualifiers are specified is:

The general

command/qualifier

Qualifiers

change

the

defaults

the

debugger

uses

1in

processing

when you deposit a value, the debugger uses
example,
For
commands.
You can override the default by specifying
decimal radix by default.
Table 2-2 summarizes the command qualifiers of
either /HEX or /OCT.
particular significance in FORTRAN debugging.
Table

2-2

Debugger Command Qualifiers

Qualifier

Function

Commands

/ADDRESS

Indicates that an
address value 1is

EVALUATE

desired

/HEX
/OCT

Refer to the

VAX-11

Override the
default radix
(decimal)

Symbolic

information on gqualifiers.

2.8

Debugger

EVALUATE
DEPOSIT

Reference

Manual

for

NUMERIC DATA TYPES

FORTRAN
1in VAX-1ll
The debugger supports all numeric data types used
and
(Complex values can be deposited
except complex.
IV-PLUS,

examined, however.) Furthermore, if you attempt to deposit a numeric
a variable or array element that does not have a matching
into
value
data type, the value is converted to the data type of the variable or
array element.

To deposit a complex value,
real part,
For

specify it in two parts as:

imaginary part

example:

DEPOSIT CPLX=3.4,-4.7

When you examine a complex variable or
array
element,
the data
displayed as a complex constant, as (real part, imaginary part).

DEBUGGING

When you

deposit

real

numbers,

you

distinguish
single
precision
D, respectively.
For example:
Number

2.9
You

Data

must

specify a decimal

point.

use

and

Type

Single

precision

24.1E0

Single

precision

24.1D0

Double

precision

241E0

Invalid

(default)

(no decimal

point)

EFFECTS OF OPTIMIZATION ON DEBUGGING
should

include

program

necessary
programs,

the

bugs

compiler

/NOOPTIMIZE
may

need

VAX-11

default;

optimization

eliminating

qualifier

liable

while

when

debugged.

FORTRAN

and,

from programs

uses

IV-PLUS

you
This

create

difficulty

the development

optimization

central

processor

condition

codes

frequently-used

variables

registers

control

proceeds

Binding

Assuming
sequence,

These techniques
described below.

that

the

based

and

flow

source

some

a
is

performs

for

and

techniques:

code

the

bug-free

finding

stage.

following

Using

compile
qualifier

compiler

highly desirable

the

2.9.1

the
that

because

optimizations

implications

for

certain

debugging

are

the

Use of Condition Codes

This optimization

technique

central

the

PROGRAMS

and double precision numbers,

24.1

FORTRAN

The

FORTRAN

processor's
following source
X

=X+

2.5
.LT.

takes

condition

advantage
codes

code:

are

the way

set.

For

which

example,

consider

Rather than test the new value of X to determine
whether
to
branch,
the
optimized
object
code
bases its decision on the condition code
settings after 2.5 is added to X.
Thus,
if
you
attempt
to
set
a
breakpoint
at
the second line, and deposit a different value into X,
you will not achieve the intended result, because the condition
codes
no
1longer
reflect
the
value of X.
1In other words, the decision to
branch is being made without regard to the new value of the variable.

2.9.2

Binding

This technique is used to reduce the number of
memory
references
or
load-and-store
instructions
needed.
The
values of frequently-used
variables are kept in general registers, and the registers
are
used,

rather

than

the

variables.

Therefore,

if you deposit

new value

DEBUGGING FORTRAN PROGRAMS

variable that has been bound to a register, the new value will have no
if you examine the variable, the current value
Moreover,
effect.
register) may not be displayed.
the
in
kept
is
(which

2.9.3

Control Flow

The compiler assumes that statements will be executed in the sequence
in which they appear in source code, if there are no intervening
labels. Optimization of such code sequences will not let you use the
"GO address" version of the GO command.

2.9.4

Effects of /NOOPTIMIZE and /OPTIMIZE

The /NOOPTIMIZE qualifier tells the compiler not to assume that
not to keep the values of variables in
condition codes are valid;
across statement boundaries.
optimize
to
not
and
general registers;
In

short,

the

object

program directly reflects the source program.

When /NOOPTIMIZE is in effect, you can

issue

any

debugging

the

commands.

When /OPTIMIZE is in

effect,

you

should

not

use

the

address

However, you can set and clear breakpoints and examine
command.
COMMON variables. If you need to debug a program that was compiled

with /OPTIMIZE in effect, you may need a compiler listing of the
generated machine code. Thus, if you do not suppress optimization,
you should specify /LIST and /MACHINE_CODE in the FORTRAN command.

2-19

CHAPTER 3

FORTRAN

INPUT/OUTPUT

This chapter describes FORTRAN input/output (I/0) as
implemented
for
VAX-11
FORTRAN IV-PLUS.
1In particular, it provides information about
FORTRAN IV-PLUS I/0 in relation to VAX-1l Record
Management
Services
(RMS).

The

topics

covered

file

include:

VAX/VMS

Logical names as used

FORTRAN

FORTRAN record formats

OPEN

Auxiliary I/O operations

Local

3.7)

(Section 3.2)

(Section 3.3)

(Section

statement features

interprocess

(Section 3.1)

in FORTRAN

file characteristics

(Section

specifications

3.4)

(Section 3.5)

(Section 3.6)

communication

means

Remote communication by means of DECnet-VAX

mailboxes

(Section 3.8)

The FORTRAN I/O statements are:
READ, WRITE, ACCEPT, PRINT, and TYPE.
The
device or file to or from which data is transferred is designated

by a logical unit number, specified or implied
as
part
of
statement.
Logical unit numbers are integers from 0 to 99.

For

the

1I/0

example:

READ

(2,100)

I,X,Y

This statement specifies that data is to be entered from the device or
file
corresponding
to logical unit 2, in the format specified by the
FORMAT

statement labeled

100.

The association between the
1logical
unit
number
and
the
physical
device
or
file
occurs
at
execution
time.
You
can
change this
association at execution time, if necessary, to match the needs of the
program
and the available resources.
You need not change the logical
unit numbers specified in the program.
inherently device independent.

Thus,

FORTRAN

programs

are

You can use standard FORTRAN I/0
statements
to
communicate
between
processes
on
either
the
same
computer
or
different
computers.
Mailboxes permit interprocess
communication
on
the
same
computer.

DECnet
network
facilities are used for interprocess communication on
different computers.
DECnet can also be
used
to
process
files
on
different computers.

FORTRAN

3.1
A

FILE

INPUT/OUTPUT

SPECIFICATION

complete

VAX/VMS

file

specification

has

the

form:

node::device: [directory]filename.filetype.version
For

example:

BOSTON: :DBAQO: [SMITH]TEST.DAT.2
node
BOSTON

device
DBAO

(unit

disk

DBA)

file

name

directory
SMITH

(the

cataloged

the

disk

directory

named

SMITH)

filename
TEST

filetype
DAT

version number
2

you

omit

default

elements

values,

the

file

follows.

specification,

you

omit

the

node,

system
the

supplies

local

computer

is used;
if you omit
the
device
or
directory,
the
current
user
default
1is
used;
if
you
omit
the file name, the system supplies
FOROnn, where nn is the logical unit number;
if
you
omit
the
file
type,
the
the system

input)
For

example,

directory
READ
The

system
supplies DAT;
and if you omit the version number,
supplies either the highest
current
wversion
number
(for
the highest current version number plus 1 (for output).

suppose

your

SMITH,

and you

default

device

DBAO:

and

your

default

specify:

(8,100)

default

file

specification

is:

DBAO: [SMITH] FOR0O08.DAT.n

The
value
FOR008 .DAT.
Then,

The

equals

suppose you

specify:

WRITE

(9,200)

default

file

the

specification

highest

current

version

number

is:

DBAO: [SMITH]FOR0O09.DAT.m
Where m is 1
FOR009 .DAT.

greater

than

the

highest

existing

version

number

FORTRAN INPUT/OUTPUT

3.2

LOGICAL NAMES

a
The VAX/VMS operating system provides the logical name mechanism as
means
of
associating
1logical
units with
file
specifications.
A
logical name is an alphanumeric string, up to 15 characters long, that
is specified instead of a file specification.
names,
1logical
predefined
The operating system provides a number of
Table 3-1
file specifications.
already associated with particular
interest to FORTRAN users.

lists the logical names of special

Table

3-1

Predefined System Logical Names

Name

Meaning

Default

SYS$DISK

Default device and directory

As specified by the
user

SYSSINPUT

Default input stream

User's terminal

SYSSOUTPUT

Default output stream

User's terminal

batch
(interactive);
command file (batch)

batch
(interactive);
log file (batch)

You can create a logical name dynamically, and

associate

with

file
specification
by means
of
the VAX/VMS ASSIGN command.
Thus,
before program execution, you can associate the logical names in your
program with the file specification appropriate to your needs.
For

example:

$ ASSIGN

DBAQ:[SMITH]TEST.DAT.2

LOGNAM

it with
This command creates the logical name LOGNAM and associates
This will be the file
file specification DBAO: [SMITH]TEST.DAT.2.
the
specification used when the logical name LOGNAM is encountered during

program execution.

Logical names provide great flexibility because they can be associated
not
only with
a
complete
file specification, but with a device, a
device and a directory,

3.2.1

or even another logical name.

FORTRAN Logical Names

Usually, FORTRAN I/O is performed by associating a logical unit number
The VAX/VMS logical name concept allows one
or file.
a device
with
name can be
a user-specified 1logical
more level of association:
associated with a logical unit number.

VAX-11 FORTRAN IV-PLUS provides predefined logical names,

in the form:

FOROnNnn

By default,
The value of nn corresponds to the logical unit number.
FORTRAN logical name is associated with a file named FOROnn.DAT,
each

3-3

FORTRAN

which is assumed to be
default directory.
For
WRITE

located
example:

INPUT/OUTPUT

your

default

disk,

under

(17,200)

If you enter
this
statement,
without
including
an
explicit
specification,
the
data
will
be written to your default disk,
file named FOR017.DAT, under your default directory.
You can
logical

file
to a

change
wunit

associated

your

the
file
specification
associated
with
a
FORTRAN
number
by
wusing the ASSIGN command to change the file
with the corresponding FORTRAN logical name.
For example:

ASSIGN

DBAO:[SMITH]TEST.DAT.2

FOR017

This command associates the FORTRAN logical name FOR017 (and therefore
logical
wunit
17)
with
file TEST.DAT.2 on device DBAO, in directory

SMITH.

You can also associate the FORTRAN
logical
names
with
predefined system logical names.
Two examples follow.

ASSIGN

SYSSINPUT

$ ASSIGN
This

Many

VAX-11l

assignments
example.

3.2.2

SYSSOUTPUT

command

output

systems

for

Implied

(for

FORTRAN

logical

example,

provide

1logical

the

5 with the default
stream).

input

FOR006

associates

device

FORO0O05

This command associates logical unit
device (for example, the batch input

any

unit
batch

system-wide

wunits

Logical

the

and

with

output
default

shown

the

default

stream).
logical

the

name

preceding

Unit Numbers

The READ, ACCEPT,
PRINT,
and
TYPE
statements
do
not
include
an
explicit
logical
unit number.
Each of these FORTRAN statements uses
an implicit logical unit number
and
logical
name.
Each
of
these
logical
names
1is,
in
turn,
associated
with
one
of the system's
predefined
1logical
names,
by
default.
Table
3-2
shows
these
relationships.

Implicit

Table 3-2
FORTRAN Logical

FORTRAN

Statement

READ

f,list

ACCEPT
PRINT

TYPE

f,list
f,1list

f,list

Logical

Name

Units

System

Logical

Name

FORSREAD

SYSSINPUT

FORSACCEPT

SYSSINPUT

FORSPRINT

SYSSOUTPUT

FORSTYPE

SYSSOUTPUT

FORTRAN

INPUT/OUTPUT

As with any other FORTRAN
1logical
name,
you
can
change
the
file
specifications associated with these FORTRAN logical names by means of
the ASSIGN command.

SASSIGN

For

example:

DBAO:[SMITH]TEST.DAT.2

FORSREAD

Following execution of this command, the READ statement's logical name
(FORSREAD)
will
refer
to
the
file
TEST.DAT.2, on device DBAO, in

directory SMITH.

3.2.3

OPEN Statement NAME Keyword

You can use the NAME keyword of the
OPEN
statement
to
specify
the
particular
file
to
be
opened
on
a
logical
unit.
(Section 3.5
describes the OPEN statement in greater detail.) For example:
OPEN

(UNIT=4,

NAME='DBAO: [SMITH]TEST.DAT.2', TYPE='OLD')

In this example, the file TEST.DAT.2, on device
DBAO:,
in
directory
SMITH,
is
opened
on
logical
unit
4.
Neither
the
default file
specification (FOR004.DAT) nor the
FORTRAN
logical
name
FOR004
is
NAME keyword can be a character constant,
the
of
value
The
used.
variable,

or expression.

You can also specify a logical name as the value of the NAME
keyword,
if
the
1logical
name
1is
associated with a file specification.
For
example:

SASSIGN

DBAO:[SMITH]TEST.DAT

LOGNAM

This command assigns the logical name LOGNAM to the file specification
DBAQO: [SMITH] TEST.DAT.
The
logical
name can then be used in an OPEN
statement,

OPEN

follows:

(UNIT=19,NAME='LOGNAM',TYPE='OLD"')

When an I/0 statement refers to logical unit 19, the system
file specification associated with logical name LOGNAM.

uses

the

If the value specified for the NAME keyword
has
no
associated
file
specification,
it
is
regarded
as a true file name rather than as a
logical name.
That is, if LOGNAM had not been
previously
associated

with the file specification DBA(O: [SMITH]TEST.DAT by means of an ASSIGN
command, then the following statement would indicate that a file named
LOGNAM.DAT is located on the default device, in the default directory:
OPEN

(UNIT=19,NAME='LOGNAM',TYPE='OLD"')

A logical name
specified
in
an OPEN
statement must
not
contain
brackets
or
periods.
The
system
treats any name containing these
punctuation marks as a file specification, not as a logical name.

FORTRAN INPUT/OUTPUT

3.2.4

Assigning Files

You can

assign

files

to Logical Units

logical

using default
READ

Logical

units

logical

any of

names;

two

three ways:

examples

follow.

(7,100)

unit FOR007

associated with

the

file

FOR007.DAT

default.

TYPE

Logical
default.
2.

unit

FORSTYPE

By specifying
example:
OPEN

logical

1is

associated

name

with

SYS$SOUTPUT

OPEN

statement.

For

an OPEN

statement.

For

(UNIT=7,NAME='LOGNAM')

By supplying
example:
OPEN

You
and

100

file

specification

(UNIT=7 ,NAME='LOGNAM.DAT')

use the ASSIGN command
file specifications.

change

the

association

logical

names

A logical name used with the NAME keyword of the OPEN
statement
must
be
associated with a file specification, and the character expression
specified for the NAME keyword
must
contain
no
punctuation
marks.
Otherwise,
the
logical
name
will
be
treated
as
a
true
file
specification.
Use

the

VAX/VMS

associations of

SHOW

logical

LOGICAL

names

To remove the association
use the DEASSIGN command,

$SDEASSIGN

3.2.5
You
The

of
in

and

command

file

a logical
the form:

determine

the

current

specifications.
name

and

file

specification,

logical=-name

Assigning Logical Names with MOUNT Commands

can specify a
MOUNT command

logical
has the

name as
form:

MOUNT device-name,...

parameter

[volume-label,...]

the

TAPE2 MYTAPE

command.

[logical-name[:]]

If your program refers to devices by means of logical
change
the
association
between the device name and
when you mount the device.
For example:

$ MOUNT MT:

MOUNT

names,
you
the logical

can
name

FORTRAN

INPUT/OUTPUT

This command associates the logical name MYTAPE with
device
name
MT
and
volume label TAPE2.
Whenever your program refers to logical name
MYTAPE, access will be to the volume
1labeled
TAPE2
mounted
on
the
default
magnetic
tape
unit.
If you subsequently mount a different
tape to be referenced by the logical name MYTAPE, you can
change
the
logical
name
association
when
you
issue
the
MOUNT command.
For
example:

$ MOUNT MT:

3.3

TAPE7 MYTAPE

FILE CHARACTERISTICS

A clear distinction must
organized and the manner

be made between the way in
which
in which records are accessed.

files

are

The term "file organization" applies to the way records are physically
arranged
on
a
storage device.
"Record access" refers to the method
used to read records from or write records to a
file,
regardless
of
its organization.
A file's organization is specified when the file is
created, and cannot be changed.
Record access is specified each
time
the file is opened, and can be different each time.

3.3.1

File Organization

VAX-11

FORTRAN

IV-PLUS

Sequential

Relative

supports

two

file

The organization of a file is specified
OPEN statement, as described in Section

3.3.1.1
Sequential Organization sequential.

organizations:

by means
3.5.4.

The default

file

keyword

organization

the

1is

Sequential files consist of records arranged in the sequence in
which
they
are
written
to the file (the first record written is the first
record in the file, the second record written is the second record
1in
the file, etc.).
As a result, records can be added only at the end of
the file.
Sequential file organization is permitted
on
all
devices
supported by the VAX-11 FORTRAN system.

3.3.1.2
Relative Organization - Relative files are permitted only
on
disk
devices.
A relative file consists of numbered positions, called

cells.
These cells are of
consecutively
from
1
to
last available cell in the

fixed,
equal
n, where 1 is
file.

1length,
the first

and
are
numbered
cell, and n is the

This arrangement lets you place records into
the
file
according
to
cell
number;
the
cell
number becomes the record's relative record
number;
that is, its location relative to the beginning of the
file.
As
a
result,
you
can retrieve records directly by specifying their
relative record number, because the actual location of the
record
1is
easily
calculated relative to the beginning of the file.
You can add
records to,
or
delete
them
from,
the
file
regardless
of
their

FORTRAN

location,
the

long

as you

keep

INPUT/OUTPUT

track

the

relative

record

numbers

records.

3.3.2

Access

Records

can be

to Records

accessed

Sequential

Direct

two ways:

access

The access mode chosen is unrelated to the file organization.
You can
access
records
in both relative and sequential files sequentially or
directly (with certain restrictions, described below).

3.3.2.1
Sequential Access - If you
records
are
written
to
or
read
beginning and continuing through the

select
sequential
access
mode,
from
the
file,
starting at the
file one record after another.

Sequential access to a file means that
a
particular
record
can
be
retrieved
only
when
all
the
records
preceding it have been read.
Writing records by means of sequential access varies according to
the
file
organization.
New
records can be written only at the end of a
sequentially
organized
file.
For
a
relative
organization
file,
however,
a
new
record
can
be
written at any point, replacing the
existing record in that cell.
For example, if two records
are
read,
and
then
a
record is written, the new record occupies cell 3 of the
file.

3.3.2.2
Direct
Access - If
you
select
direct
access
mode,
you
determine
the
order in which records are read or written.
Each READ
or WRITE statement must include the relative record number
indicating
the record to be read or written.

You can access relative files directly,
and
you
can
also
directly
access
a
sequential
file
1if
it
contains fixed length records and
resides on disk.
Because direct access
uses
cell
numbers
to
find
records,
you can issue successive READ or WRITE statements requesting
records that either precede or follow previously requested records.
For

3.4

example:

READ

(12'24)

read

record

file

READ

(12'20)

read

record

file

RECORD STRUCTURE

Records

are

stored

Fixed

Variable

Segmented

in one

length
length

three

formats:

FORTRAN

INPUT/OUTPUT

Fixed length and variable length formats can be used
with
sequential
or relative file organization.
Segmented format is unique to FORTRAN,
and can be used only with sequential file organization, and
only
for
unformatted
sequential
access.
You should not use segmented records
for files that will be read by programs
written
in
languages
other
than FORTRAN.
‘

3.4.1

Fixed Length Records

When you specify fixed length records (see
Section
3.5.7),
you
are
specifying
that
all
records
in the file contain the same number of
bytes.
When you create
a
file
that
1is
to
contain
fixed
length
records,
you
must
specify
the
record size (see Section 3.5.6).
A

sequentially organized file
fixed
1length
records,
to
correctly.
Note that in
length record contains an

3.4.2

opened
allow

for
the

a
relative
extra byte,

direct
record

access
must
contain
number to be computed
organization
file
each
fixed
the deleted-record control byte.

Variable Length Records

Variable length records can contain any
number
of
bytes,
up
to
a
specified
maximum.
Variable
1length records are prefixed by a count
field, indicating the number of bytes in the record.
The count
field
comprises two bytes on a disk device, and four bytes on magnetic tape.
The value stored in the count field indicates the number of data bytes
in the record.
Variable length records in relative files are actually
stored in fixed length cells, the size of which must be
specified
by
means

the

RECORDSIZE

keyword

the

OPEN

statement

(see

Section

3.5.6).
This value specifies the largest record that can be stored in
the
file.
Each
variable
1length record in a relative file contains
three extra bytes, two for the count field and one for deleted
record
control.
The count field of a variable length record is available when you read
the
record;
issue a READ statement with a Q format descriptor.
You
can then use the count field information to determine how
many
bytes
should be in an 1/0 list.

3.4.3

Segmented Records

A segmented record is a single logical record
consisting
of
one
or
more variable length records.
Each variable length record constitutes
a segment.
The length of a segmented record is arbitrary.
Segmented
records
are useful when you want to write exceptionally long records,
and
are
especially
appropriate
to
sequentially
organized
files.
Unformatted sequential records written to sequentially organized files
are, by default, stored as segmented records.

Because there is no set limit on the size of a segmented record,
each
variable
length
record
in
the
segmented
record
contains control
information to indicate that it is one of the following:
@

The

first

segment

The

last

segment

the

segmented

record

The

only

segment

the

segmented

record

None

the

above

the

segmented

record

FORTRAN INPUT/OUTPUT

This control information is contained in the first two bytes
of
each
segment
of
a
segmented
record.
Thus,
when you wish to access an
unformatted sequential file that contains
fixed
1length or
variable
length
records,
you must specify RECORDTYPE='FIXED' or 'VARIABLE'
(as
appropriate) when you open the file.
Otherwise the first two bytes of
record will be misinterpreted as control information, and errors
each
will

probably result.

3.5

OPEN STATEMENT KEYWORDS

The following sections supplement the OPEN statement description that
1In
the VAX-11 FORTRAN IV-PLUS Language Reference Manual.
in
appears
particular,

certain

implementation-dependent and/or

system-dependent

aspects

statement keywords are described as affected by the

OPEN

more
For
implementation.
(RMS)
Services
VAX-11 Record Management
to the VAX-11 Record Management Services Reference
refer
information
Manual.

3.5.1

BLOCKSIZE Keyword

The BLOCKSIZE keyword specifies the physical I/O transfer size for the
It has the

file.

BLOCKSIZE

form:

= bks

For magnetic tape files, the value of bks specifies the physical
The default value is 2048
record size in the range 18 to 32767 bytes.
’

bytes.

For sequential disk files,

the value

bks

1is

rounded

integral
number of 512-byte blocks and used to specify RMS multiblock
The
127.
to
1
The number of blocks transferred can be
transfers.
is 2048

default value

For relative files,

bytes.

the value of bks is

rounded

integral

of 512-byte blocks and used to specify the RMS bucket size, in
number
The default is the smallest value capable
the range 1 to 32 blocks.
of holding a

single

record.

BUFFERCOUNT Keyword

3.5.2

The BUFFERCOUNT keyword specifies the number of memory buffers to use.
It

has

the

form:

BUFFERCOUNT =

The range of values for bc is from 1 to 255.

The size of each

buffer

Thus, if BUFFERCOUNT=3 and
is determined by the BLOCKSIZE keyword.
Iis
buffers
for
BLOCKSIZE=2048, the total number of bytes allocated
3%2048,

or 6144.

one buffer for

3.5.3

The default is two buffers for sequential files and

relative files.

INITIALSIZE and EXTENDSIZE Keywords

The INITIALSIZE keyword specifies the initial storage allocation for a
disk
file, and the EXTENDSIZE keyword specifies the amount by which a
disk file is extended each time more space is needed for the file.
3-10

FORTRAN INPUT/OUTPUT

If
INITIALSIZE is effective only at the time the file 1is created.
EXTENDSIZE is specified when the file is created, the value specified
is the default value used to allocate additional storage for the file.
If you specify EXTENDSIZE when you open an existing file, the value
you specify supersedes any EXTENDSIZE value specified when the file
was created, and remains in effect until you close the file. Unless
specifically overridden, the default EXTENDSIZE value is in effect on
of the file.
subsequent openings

If
The system attempts to allocate contiguous space for INITIALSIZE.
not enough contiguous space is available, noncontiguous space is
allocated.

ORGANIZATION Keyword

3.5.4

has

the

When an existing file is opened, the actual organization of

the

file

The ORGANIZATION keyword specifies

file

organization.

form:

ORGANIZATION ={'RELATIVE'

'SEQUENTIAL'

}

The default file organization is sequential.
is

used.

The relative file organization is applicable mainly when creating
files to be used in non-FORTRAN applications, or when reading relative
files created by programs written in languages other than FORTRAN.

READONLY Keyword

3.5.5

The READONLY keyword specifies that write operations are not allowed
The FORTRAN I/O system's default file
on the file being opened.
access privileges are read-write, which can cause run-time I/O errors
The READONLY
file protection does not permit write access.
the
if
Its
keyword has no effect on the protection specified for a file.

to allow a file to be read simultaneously by two or
is
main purpose
the purpose of
Thus, if you wish to open a file for
more programs.
reading the file, but do not want to prevent others from being able to
specify the READONLY
it open,
read the same file while you have
keyword.

RECORDSIZE Keyword

3.5.6

The RECORDSIZE keyword specifies how much data can be contained
record.

It has the form:

RECORDSIZE

a
.

The value specified for rl indicates the length of the logical records
For files that contain fixed length records, rl
in the file.
for files that contain variable
specifies the size of each record;
length records, rl specifies the maximum length for any record.

3-11

FORTRAN

INPUT/OUTPUT

The value
present),

and

of rl does not include the two
segment
control
bytes
(if
or the bytes that RMS requires for maintaining record length
deleted-record control information (two
or
four
for
sequential

organization,
The

value

and

one or

three

interpreted

for

relative organization).

either

bytes

longwords,

depending

on
whether
the
records
are
formatted
(bytes)
or
unformatted
(longwords, that is, 4-byte units).
Table 3-3 summarizes the
maximum
values
that
can
be specified for rl, based on file organization and
record format.

Table 3-3
RECORDSIZE Limits

File Organization

Record
Formatted

Unformatted

(bytes)

(longwords)

32766

8191

Sequential

Sequential and
variable length
on ANSI

9999*

records

magnetic

Limit

2499%*

tape

Relative

Format

16380

imposed

4-byte ASCII

count

4095

field.

If you are opening an existing file containing fixed length
that has relative organization, and you specify a value for
that is different from the actual length of the records in
an

error

file,

the

occurs,

record

default.

length

you

omit

RECORDSIZE

specified when

the

when

opening

file was

The

created

You must specify RECORDSIZE when you create a file that
fixed length records or that has relative organization.

3.5.7

records or
RECORDSIZE
the
file,

existing

used

contain

RECORDTYPE Keyword

RECORDTYPE keyword
has the form:

RECORDTYPE

specifies

the

structure

records

file.

= { 'VARIABLE'

'SEGMENTED'
This keyword is particularly useful
default
record
structure
used to
structure is:
FIXED

- direct

VARIABLE

formatted

SEGMENTED

unformatted

access,

when
you
want
to
create a file.
The

sequential,

sequential

3-12

and

relative

override
the
default record

FORTRAN INPUT/OUTPUT

The default used
when
accessing
an
existing
file
1is
the
record
structure
of
the
file,
except
for
unformatted
sequential
files
containing fixed or variable length records.
1In this case,
you
must
explicitly
override
the
default
(SEGMENTED)
by
specifying
the

appropriate RECORDTYPE value in the OPEN statement.
You cannot use an
unformatted
READ
statement
to
access
an
unformatted sequentially
organized file that contains fixed length or variable length
records,
unless
you
specify
the
corresponding RECORDTYPE value in your OPEN
statement.
Files containing segmented records can be accessed only by
unformatted sequential FORTRAN I/O statements.

3.5.8

SHARED Keyword

The SHARED keyword specifies that the file can
be
accessed
by
more
than
one
program
at a time, or by the same program on more than one
logical
unit.
The
forms
of
sharing
permitted
depend
on
the
organization of

the

file.

For sequential files, both
read
and
write
sharing
are
permitted.
Because
RMS
does not prevent two or more programs from accessing the
same file simultaneously, however,
user
programs
that
share
write
access
to
a
file
must
provide
interprocess
communication
and
coordination to ensure reliable performance.
Otherwise, problems
may
develop.
For
example,
if
two programs write to a shared file that
contains records that cross block boundaries, records containing
data
written
by two different programs can result.
This can happen if the
co-operating programs
do
not
coordinate
their
read,
modify,
and
rewrite
sequerices,
which
are
otherwise
asynchronously
and
independently performed.

Furthermore, RMS usually tries to minimize disk activity by postponing
a
rewrite
in
case a subsequent read or write can be performed using

the program's buffer image.
Thus, the file's disk image may be out of
date
for
arbitrary
time intervals.
This problem can occur for both
sequential and direct access I/0.

You can encounter a similar problem involving the logical
end-of-file
on disk.
When a file is extended, the logical end-of-file in the disk
image is not updated until the file is closed.
This means that
1if
a
file
1is open and program A is adding new records to it, and program B
opens the same file before program A has closed the
file,
program
B
cannot
read
the new records even after program A finishes and closes
the file.
Program B can read the new
records
only
by
<closing
and
reopening
the file.
Only then will the file's disk image reflect the
updated end-of-file.

Relative files permit no write sharing.
For shared reading to
occur,
all programs that open the file must specify the READONLY keyword.

3.5.9

USEROPEN Keyword

The USEROPEN keyword provides access
to
supported
by
the
FORTRAN
I/O system.

RMS
features
not
directly.
That is, this keyword allows
access to RMS capabilities, while retaining the ease
and
convenience
of
FORTRAN
programming.
The
USEROPEN
keyword
1is
intended
for
experienced users.

For the interface specification for a USEROPEN routine,
Common Run-Time Procedure Library Reference Manual.

see the VAX-11

FORTRAN

3.6

AUXILIARY I/0 OPERATIONS

This

section describes

INPUT/OUTPUT

implications of

the

following

I/0

statements:

FIND
BACKSPACE

ENDFILE

A FIND statement is similar to a direct access READ statement with
I/0
1list,
and
can
result
1in
an
existing
file being opened.
associated variable will be set to the specified record number.

no
An

A BACKSPACE statement cannot be performed on a file that is opened for
append
access, because of the manner in which backspacing is done.
A
backspace
operation
requires
that
the
current
record
count
be
available to the FORTRAN I/0 system, because backspacing from record n
is done by rewinding to the start of the file and then performing
n-1
successive
reads
to
reach the previous record.
If the file is open
for append access, the current record count is not
available
to
the
FORTRAN I/0O system.

The ENDFILE
statement
writes
an
end-file
record.
The
following
convention
has
been adopted, since RMS does not support the embedded
end-file concept:
an end-file record is a l-byte record that contains
the
hexadecimal
code 1A (CTRL/Z).
An end-file record can be written
only to sequentially organized files that are
accessed
as
formatted
sequential
or
unformatted
segmented
sequential.
End-file records
should not be written in files that will be read by
programs
written
in a language other than FORTRAN,

3.7

LOCAL

INTERPROCESS

COMMUNICATION:

MAILBOXES

It is often useful to exchange data between processes;
to synchronize execution, or to send messages.

for

example,

A mailbox is a record-oriented pseudo I/0 device that allows
data
to
be
passed
from one process to another.
Mailboxes are created by the
Create Mailbox system service.
The following sections describe how to
create mailboxes and how to send and receive data using mailboxes.

3.7.1

Creating a Mailbox

Use the Create Mailbox

system service

ICHAN
INTEGER*4 SYS$CREMBX
MAILBX= SYS$CREMBX(,ICHAN,

to create a mailbox,

follows:

INTEGER*2

,'MAILBOX')

The INTEGER*2 variable ICHAN is
used
to
store
the
number
mailbox,
which
is
returned by the Create Mailbox and Assign
system services.
This argument is
required
by
the
Create
systems
service,
so
you
must specify an INTEGER*2 variable
ICHAN.
However, all subsequent
references
to
the
mailbox
logical

of
the
Channel
Mailbox
such as
are
by

name.

For more information about calling system
services,
see
Chapter
5.
For
more
information
about
the
arguments
supplied
to the Create
Mailbox system service, see
the
VAX/VMS
System
Services
Reference
Manual.

FORTRAN INPUT/OUTPUT

3.7.2

Sending and Receiving Data Using Mailboxes

Sending or receiving data to or from a mailbox is
no
different
from
other
forms
of
FORTRAN
1I/0.
The
mailbox
1is simply treated as a
record-oriented I/0 device.

Use FORTRAN formatted sequential I/0 statements to
messages.

Use

receive data.

WRITE

send

and

receive

statements to send data and READ statements to

Data transmission by means of mailboxes is performed synchronously, so
That is,-a program
that communicating processes can be synchronized.

1is read,
that writes a message to a mailbox waits until the message
and a program that reads messages from a mailbox waits until a message
an
the mailbox,
When the writing program closes
is written.
program.
reading
the
to
returned
is
condition
end-of-file

The sample program below reads messages from a mailbox
known
by
the
logical
name MAILBOX.
The messages comprise file names, which the
program reads.
The program then prints the file associated
with
the
file

names.

CHARACTER FILNAM*64,TEXT*133

OPEN (UNIT = 1, NAME = 'MAILBOX', TYPE =
READ (1,100,END=12)FILNAM

100

FORMAT

(A)

OPEN (UNIT = 2, NAME = FILNAM, TYPE =
OPEN (UNIT = 3, NAME = 'SYSSOUTPUT')

3.8

'OLD')

READ (2,100, END =
WRITE(3,100) TEXT
GO TO 2

CLOSE (UNIT = 2)
CLOSE (UNIT = 3)
GO TO 1

END

10)

'OLD')

TEXT

COMMUNICATING WITH REMOTE COMPUTERS:

NETWORKS

If your system supports DECnet-VAX facilities, and your
computer
is
one of
the
nodes
in a DECnet-VAX network, you can communicate with
1/0
standard FORTRAN
in the network by means of
nodes
other
statements.
These statements let you exchange data with a program at
the remote computer (task-to-task communication), and to access
files
at the remote computer

(resource sharing).

Both task-to-task communication and

transparent.
intersystem

exchanges.

That
1is,
exchanges,

file access

between

systems

are

the

first

there is no apparent difference between these
and
1local
interprocess
and
file
access

To invoke network communication, specify a
element of a file specification.

node

For example:

name

BOSTON: :DBAQO: [SMITH] TEST.DAT.2

For remote task-to-task communication, you must use a special form of
you must use TASK_ in place of the device name,
specification:
file
and use the task name in place of the file name.
For example:
BOSTON:
: TASK_ :UPDATE

FORTRAN

The

following

from

remote

example
program

100

shows how messages can be sent to
and
received
by means of standard FORTRAN I/0 statements.

OPEN

(UNIT=7,NAME='BOSTON::TASK

READ

(7,100)IARRAY

FORMAT
CALL

:UNA',ERR=200)

(2018)

STATS (IARRAY)

WRITE
CLOSE

200

INPUT/OUTPUT

(7,100)IARRAY
(UNIT=7)

END

The effect of these
(task) named UNA at

statements is to
the node BOSTON,

establish
and receive

a
1link
with
data from the

a
job
logical

unit (7) associated with
20-element
array, and a

the remote program.
The data is stored
call is issued to the subroutine STATS,

processes

results

link

the

data.

The

are

then

sent

back

BOSTON,

in
a
which

and

the

broken.

The following example shows how a remote
of standard FORTRAN I/0 statements.
CHARACTER*64
OPEN (UNIT

1l
1
100

ACCESS

NAME

'DIRECT',
=

WRITE (2'IREC)
GOTO 1

DATA

=
=

can

updated

means

DATA
2,

READ(1,100,END
FORMAT(I10, A)

CLOSE (UNIT
CLOSE (UNIT

file

'DENVER::MASTER.DAT',

TYPE

IREC,

'OLD')

DATA

1)
2)

END

This program reads local data describing
new records into the remote file.

transactions,

and

writes

the

If you use logical names in your program, you can equate
the
1logical
names
with
either
1local
or
remote
files.
Thus, if your program
normally
accesses
a
remote
file,
and
the
remote
node
becomes
unavailable,
you
can bring the volume set containing the file to the
local site.
You can
then
mount
the
volume
set,
and
assign
the
appropriate logical name.
For example:
Remote Access

ASSIGN REM::APPLIC_SET:file—name

LOGIC

Local Access
§$

MOUNT device-name APPLIC_SET

$ ASSIGN APPLIC_SET:file-name
The

MOUNT

and ASSIGN

Command

Language

DECnet

facilities

Manual.

commands are
User's Guide.

are

described

LOGIC
described

fully

the

detail

the

DECnet-VAX

VAX/VMS

Reference

CHAPTER 4
USING CHARACTER DATA

The FORTRAN character
data
type
allows
you
to
easily
manipulate
alphanumeric
data.
You
<can
use
character
data
in
the
form of
character variables, arrays, constants, and expressions.
A
character
operator
(//) is available to form character strings by concatenating
the character elements in a character expression.
See Section 4.2.

4.1

CHARACTER SUBSTRINGS

You can select certain segments (substrings) from a character variable
or
array
element
by
specifying
the
variable
name,
followed
by
delimiter values indicating the leftmost and/or
rightmost
characters
in
the
substring.
For
example,
1if
the
character
string
NAME
contained:
ROBERTAWILLIAMABOBAJACKSON

and you wished to extract
following:

the

substring BOB,

you

would

specify

the

NAME (16:18)

If you omit the
first
value,
character
of
the
substring
For example, if you specify

you
are
indicating
that
the
first
is the first character in the variable.

NAME (:18)

the

resulting

substring

ROBERTAWILLIAMABOB

If you omit
the
second
value,
you
are
specifying
the
rightmost
character to be the last character in the variable.
For example:
NAME (16:)
encompasses BOBAJACKSON

USING

4.2

CHARACTER DATA

BUILDING CHARACTER STRINGS

It is sometimes useful to create strings from
two
or
more
separate
strings.
This
is
done by means of the concatenation operator;
the
double slash (//).
For example, you might wish to create
a
variable
called NAME, consisting of the following strings:
FIRSTNAME
MIDDLENAME
NICKNAME
LASTNAME

To do
For

so,

define

each

character

variable

specified

1length.

example:

CHARACTER*42 NAME

CHARACTER*12 FIRSTNAME ,MIDDLENAME , LASTNAME
CHARACTER*6 NICKNAME
Concatenation

Thus,

NAME

the

strings

FIRSTNAME

are

follows:

FIRSTNAME//MIDDLENAME//NICKNAME//LASTNAME

MIDDLENAME
NICKNAME =
LASTNAME =

which

accomplished

contained

the

values:

'ROBERT'
= 'WILLIAM'
'BOB'
'JACKSON'

stored

individually

ROBERTAAAAAA
WILLIAMAAAAA

BOBAAA
JACKSONAAAAA

then,

when

concatenated

and

stored

in NAME,

they

become

the

string:

ROBERTAAAAAAWILLIAMAAAAABOBAAAJACKSONAAAAA
Applying
nickname

the substring extraction
by specifying:

facility,

you

can

get

the

stored

NAME (25:30)

which picks
blanks).

the

6-character NICKNAME

Thus BOBAAA

retrieved

substring

the

(including

substring.

trailing

USING CHARACTER DATA

4.3

CHARACTER CONSTANTS

are
Strings of alphanumeric characters enclosed 1in apostrophes
character constants. You can assign a character value to a character
variable in much the same way as you would assign a numeric value to a
real or integer variable.
XYZ =

For example:

'ABC'

are stored in
As a result of this statement, the characters ABC three
bytes, the
than
less
is
length
XYZ's
if
that
location XYZ. Note
Thus 1if you
character string will be truncated on the right.
specified

CHARACTER*2 XYZ

XYZ =

'ABC'

longer than
The result is AB. If, on the other hand, the variable is example:
For
blanks.
with
right
the
on
padded
is
it
,
the constant
CHARACTER*6 XYZ

XYZ =
results

‘'ABC'

ABCAAA

being stored in XYZ. If the previous contents of XYZ were CBSNBC, the
result would still be ABCAAA: the previous contents are overwritten.
You can give character constants symbolic names by using the PARAMETER
statement.

For example:

PARAMETER TITLE = 'THE METAMORPHOSIS'

The symbolic name TITLE can then be used anywhere a character constant
is allowed.

Note that an apostrophe can be included as part of the constant.
do so, specify two consecutive apostrophes. For example:
PARAMETER TITLE =

'FINNEGAN''S WAKE'

results in the character constant FINNEGAN'S WAKE.

4.4

DECLARING CHARACTER DATA

To declare variables or arrays as character type, use the CHARACTER
type declaration statement, as shown in the following example:
CHARACTER*10 TEAM(12) ,PLAYER

This statement defines a l2-element character array (TEAM), each,
element of which is 10 bytes long; and a character variable (PLAYER)
which is also 10 bytes long.

USING

You

can

specify

statement

different

including

lengths

1length

example:
CHARACTER*6

CHARACTER DATA

for
value

variables
for

CHARACTER

variables.

For

NAME ,AGE*2,DEPT

In this example, NAME and DEPT are defined
to
while AGE is defined to be a 2-byte variable.

4.5

specific

6-byte

variables,

character

variable.

INITIALIZING CHARACTER VARIABLES

Use

the

For

example:

DATA

statement

CHARACTER*10
DATA

preset

the

value

NAME,TEAM(5)

NAME/'

'/,TEAM/'SMITH','JONES',

'DOE','BROWN','GREEN'/
Note

that NAME will contain 10 blanks, while
TEAM will contain the corresponding character
blanks.

initialize an array so that
value, use a DATA statement of

each
array
element
1in
value, right-padded with

each of its elements
the following type:

contains

the

same

CHARACTER*5 TEAM(10)
DATA TEAM/10*'WHITE'/

The

4.6

result

l0-element

array

in which

each

element

contains WHITE.

PASSED LENGTH CHARACTER ARGUMENTS

Subprograms that manipulate character data
character
actual
arguments of any length

the dummy argument as passed length.
dummy argument, use an asterisk (*) as
SUBROUTINE

can be
written
to
accept
by specifying the length of
To
indicate
a
passed
length
follows:

REVERSE
(S)

CHARACTER*
(*)

The

passed length notation indicates that the
length
of
the
actual
argument
is
used
when
processing
the dummy argument string.
This
length can change from one invocation of the subprogram to
the
next.

For

example:

CHARACTER

A*20,B*53

CALL

REVERSE
(A)

CALL

REVERSE
(B)

the

first

call

second

call,

its

The

FORTRAN

the

string

REVERSE,

length

will

the
be

function LEN can be
(see Section 4.8.4).

length

will

53.

used

determine

the

20;

actual

the

length

USING CHARACTER DATA

4.7

CHARACTER DATA EXAMPLES

example
The
shown below.
1is
usage
data
An example of character
of the
1letters
the
that manipulates
a program
1is
4-1)
(Figure
The results are shown in Figure 4-2.
alphabet.

USING CHARACTER DATA

CHAREXMPL ,FOR

CHARACTER DATA TYPE EXAMPLE PROGRAM FOR VAX FORTRAN IVepPLUS

CHARACTER C,

DATA
99

100

ALPHABET/’ABCDEFGMIJKLMNOPQRSTUVWXYZ®/

WRITE(T7,99)
FORMAT(’1 CHARACTER

ALPMNABET»26

100

EXAMPLE PROGRAM OUTPUT?/)

Isi,26

WRITE(7,10)

ALPHABET

FORMAT(IX,

ALPMABET

ALPMABET(2:1)

ALPHABET(!1Y)

CONTINUE

CALL

REVERSE(ALPHABET)

WRITE(?,10)

ALPMARET

CALL REVERSE(ALPHABET(1:113))

WRITE(Y,10)

ALPHABET

CALL FIND, SUBSTRINGS(®UVWN’, ALPHABET)
CALL FINDSUBSTRINGS(°A’, °DAJHDMAJDANDJA4E CEUEBCUEIAWSAWQLG®)

WRITE(7,98)
FORMAT(®@

END

CHARACTER

EXAMPLE PROGRAM

OUTPUT’)

81w

END

SUBROUTINE REVERSE(S)
CHARACTER T,

Sa(w)

J = LEN(S)
IF ¢J ,GT, §) THEN
DO 10 I=y, J/2
T = S$(Is)
SCl11) = 8¢JtJ)
$(JtJ) o T
J n Jey
CONT INUE
ENDIF
RETURN

END
SUBROUTINE FIND, SUBSTRINGS(SUB,

CHARACTERw(w»)

8UB,S

CHARACTERw#
{32 MARKS

s
MARKS
K

J = INDEX(S(I1), 8UB)
IF (J (NE, 0) THEN
I 81 ¢ (Je})
MARKS(ItI)

ENDIF

,LE,

WRITE(T7,100)

100

FORMATC

4-1

*»°

LEN(S))

2(/71X,

END

Figure

GO TO

MARKS(1K)

)

Character

Data Program Example

USING CHARACTER DATA

CHARACTER EXAMPLF PROGRAM QUTPUT
ABCDEFGHIJKILMNOPQRSTUVWXYZ
BCDEFGHIJKLMNOPORSTUYWXYZA
COEFGHIJKLMNOPQRSTUVWXYZAB
DEFGHIJKLMNOPGRSTUVWXYZABC
EFGHIJKLMNOPQRSTUVWXYZABCD
FGHIJKLMNOPQRSTUVWXYZABCDE
GHIJKLMNOPQRSTUVWXYZABCDEF
HIJKLMNOPGRSTUVWXYZABCDEFG

IJKLMNOPGRSTUVWXYZABCDEFGH
JKLMNOPQRSTUVWXYZABCDEFGH]
KLMNOPQRSTUVWXYZABCUEFGHIJ
LMNOPQRS TUVWXYZABCDRFGHIJK

MNOPQRSTUVWXYZABCDEFGHIJIKL
NOPQRSTUVWXYZABCDEFGHIJKLM

OPQRSTUVWXYZABCDEFGHIJKLMN

PGRSTUVWXYZABCDEFGHIJKLMNO

QRSTUVWXYZABCDEFGHIJKLMNOP

RSTUVWXYZABCNDEFGHTIJXLMNOPQ
STUVWXYZABCDEFGHIJKLMNOPQR
TUVWXYZABCDEFGHIJKLTMNOPGRS

UVWXYZABCDEFGHIJKLMNOPQRST
VWXYZABCDEFGHIJKLMNOPQRSTU

WXYZABCDEFGHIJKLMNOPQRSTUV
XYZABCDEFGHIJKLMNOPGRSTUVW

YZABCDEFGHIJKLMNOPQRSTUVHX
ZABCDEFGMIJKLMNOPQRSTUVWXY
ZYXWVUTSRGPONMLKJIHGFEDCBA
NOPQRSTUVWXYZMLKJIHGFEDCBA

NOPQRSTUVWXYZMLKJIHGFENCBA
#

DAJHOMAJDAHDJAUE CEUEBCUEIAWSAWOLG
£

“

END OF CHARACTER EXAMPLE PROGRAM QUTPUT
Figure

4.8

4-2

Output Generated by Example Program

CHARACTER LIBRARY FUNCTIONS

The VAX-11 FORTRAN IV-PLUS system provides four character functions:
CHAR
ICHAR
INDEX
LEN

4.8.1

CHAR Function

The CHAR function returns a l-byte character value equivalent
integer ASCII value passed as its argument.

It has

the

CHAR(1)

An integer expression equivalent to an ASCII code.

form:

the

USING CHARACTER DATA

4.8.2
The

ICHAR Function

ICHAR

character

function

returns

expression passed

integer

its

ASCII

code

argument.

equivalent

It has

the

form:

ICHAR(c)

A character expression.
code
equivalent to the
bytes are ignored.

4.8.3

If c
first

is longer than one byte, the
ASCII
byte is returned, and the remaining

INDEX Function

The INDEX function
substring:
it has

is used to determine
the form:

the

starting

position

INDEX (cl,c2)

cl
A character expression that specifies the string
searched for a match with the value of c2.

that

which

c2
A character expression
match is desired.
If

INDEX

finds

integer

instance

representing

the

substring

the

value corresponding

specified

the

substring

starting

(c2),

location

(cl).
For example, if the substring sought is
CAT,
and
that is searched contains DOGCATFISH, the return value of

If INDEX cannot find the specified substring,

4.8.4

LEN

the

the
INDEX

returns

string
string
is

it returns the value 0.

Function

The

for

LEN function returns
character expression.

an
It

integer
has the

value
form:

that

indicates

the

length

LEN (¢)

4.9

character

expression

CHARACTER 1/0

The character data type simplifies the
transmission
data.
You
can read and write character strings of
to 32767 characters.
For example:

100

CHARACTER*24

TITLE

READ (12,100)

TITLE

FORMAT (A)

of
any

alphanumeric
length from 1

USING

CHARACTER DATA

These statements cause 24 characters read from logical unit 12
to
be
stored in the 24-byte character variable TITLE.
Compare this with the
code necessary if you used Hollerith data stored in numeric
variables
or

arrays:

INTEGER*4

100

TITLE (6)

READ(12,100)
FORMAT (6A4)

TITLE

Note that you must divide the data into lengths suitable for
real
or
(in
this case) integer data, and specify I/0 and FORMAT statements to
match.
In this example, a l-dimensional array comprising
six
4-byte
elements is filled with 24 characters from logical unit 12.

CHAPTER 5

FORTRAN CALL CONVENTIONS

VAX-11 FORTRAN IV-PLUS provides a mechanism you can use to gain access
to services external to your FORTRAN programs. By including CALL
statements or function references in your source program, you can

procedures

such

use

mathematical functions, VAX/VMS system services,

and procedures written in languages other than FORTRAN.

Refer to the VAX-11l Common Run-Time Procedure Library Reference Manual
for descriptions of the procedures included in the Run-Time Library,
and information on how they are called.

5.1

PROCEDURE CALLS

A procedure is a program, such as a FORTRAN function or subroutine,
that performs one or more computations for other programs. In many
cases, procedures perform calculations that are used widely t and
It is more efficien to
repeatedly in many FORTRAN applications.
write these procedures (subprograms) once, and make them available to
other programs, than to reinvent them every time the need arises.

Subprograms can be either functions or subroutines. A function is a
subprogram that returns a value to the calling program, by assigning
the value to the function's name. A subroutine may return values, but
a value is not associated with the subroutine name. Functions and
subroutines can return values by storing values in elements of the
argument 1list, or in COMMON blocks. See the VAX-11l FORTRAN IV-PLUS
Language Reference Manual for information on defining and 1invoking
subprograms.

5.2

VAX-11l PROCEDURE CALLING STANDARD

Programs compiled by the VAX-11l FORTRAN IV-PLUS

compiler

conform

the standard defined for VAX-11l procedure calls (see Appendix C of the
VAX-11 Architecture Handbook, Vol. 1l). This standard prescribes how
arguments are passed, how function values are returned, and how
VAX-11
procedures (such as subprograms) receive and return control.
IV-PLUS also provides features that allow FORTRAN programs to
call system services and procedures written in other native-mode

FORTRAN

languages supported in VAX/VMS.

The information in this section pertains to calling system service
If you want to write routines that can be
routines from FORTRAN.
n to
called from FORTRAN programs, you should pay particular attentioformat
the argument list descriptions and to the machine code
description in Section 5.4.

FORTRAN

5.2.1

sequence

entry

procedure
of

argument

calling

longword

count.

The

argument

the

1list.

Memory

for FORTRAN
usually
allocated
argument lists are
compile time, when

argument

list.

DIZ(Y)

=X+

*
+

list

Section 5.4.1
argument lists.

which

byte

the

first

how many

arguments

first

the

example:

(for

example

that

for

has

examples

Argument Passing

are passed

use

CALL

X(A,,B))

value

the

same

argument

are

list.

represented

machine

code

generated

Call-by-value

Call-by-reference
the value

Call-by-descriptor
of a descriptor of

standard

defines

three

mechanisms

the

value

entry

to procedures:

By default, VAX-11
call-by-descriptor

will

for

FORTRAN

Mechanisms

The VAX-11 procédure calling
arguments

the

1list

first

lists and VAX-1ll standard
descriptors
is
statically.
To
optimize
space
and
time,
the
pooled and argument list entries are initialized at
possible.
Sometimes several calls can use the same

references

entry

See

5.2.2

argument

DIZ(Z)

arguments

argument

count

indicates

the

DIZ(Y)

these DIZ(Y)

Omitted

defines

entries,

argument

For

standard

(4-byte)

list.

which

CONVENTIONS

Argument Lists

The VAX-1l1l

All

CALL

the

argument

the

list

argument

- the argument
the value

entry
list

list

entry

the

address

the

address

FORTRAN

IV-PLUS
uses
the
call-by-reference
and
mechanisms.
The
call-by-reference
mechanism is
used to pass all numeric actual arguments:
logical,
integer,
real,
double
precision,
and
complex.
The call-by-descriptor mechanism is
used to pass all character actual arguments.

5.2.3
By

Argument List Built-In Functions

default,

FORTRAN

mechani
for
sm
type.

1In

some

uses

passing
cases,

the

call-by-reference

arguments,

however,

depending

function

may

require

call-by~descriptor

on-the

reference

argument's
or

call

data
to

arguments

non—-FORTRAN

IV-PLUS

a different
Therefore,

form.
Calls to VAX/VMS system services are such
a
case.
FORTRAN
provides
three
built-in
functions
for passing

arguments

built-in

when you

procedure

cannot

functions are:

3VAL
$REF

¥DESCR

use

the

FORTRAN

that

you

default

supply

mechanism.

These

FORTRAN CALL CONVENTIONS

These functions can appear only in actual argument lists.
The following sections describe the use of these functions.
the
argument
1list
built-in
functions
are
never
used
procedure written

in FORTRAN.

5.2.3.1
S8VAL - This function forces the argument list
the call-by-value mechanism.
It has the form:

Note that
to
call a

entry

use

$VAL (arg)

The argument list entry (arg) is the
wvalue
of
the
entry.
Because
argument
1list
entries
are
longwords, the argument value must be an
integer, logical,
or
real
constant,
variable,
array
element,
or
expression.

5.2.3.2
$REF - This function forces the argument list
the call-by-reference mechanism.
It has the form:

entry

use

$REF (arg)

The argument list entry (arg)
1is
the
address
of
the
value.
The
argument
value can be a numeric or character expression, array, array
element, or procedure name.
This is the default
FORTRAN
method
for
passing all numeric values.

5.2.3.3
$DESCR - This function forces the argument list entry to
the call-by-descriptor mechanism.
It has the form:

use

$DESCR (arg)

The argument list entry (arg) is the address of a
descriptor
of
the
value.
The argument value can be any type of FORTRAN expression.
The
compiler can generate VAX-11l descriptors for all FORTRAN data types.

Call-by-descriptor
1is
the
default
FORTRAN mechanism
for
passing
character
arguments,
because
the
subprogram
may
need to know the
length of the

generates

addresses

code

character

refer

argument.

in their descriptors.

character

particular,

FORTRAN

5.2.3.4
Examples
of
$%VAL,
$REF,
$DESCR - The
following
illustrate the use of the argument list built-in functions.
CALL

always

dummy arguments through the

examples

SUB(2,%VAL(2))

The first constant is passed by reference.
passed by value.

CHARACTER*10 A,B
CALL SUB(A,%REF(B))

The

second

constant

1is

FORTRAN CALL CONVENTIONS

The first character variable is
character variable is passed by

passed
by
reference.

descriptor.

The

second

is passed

second

INTEGER IARY (20), JARY (20)
SUB(IARY,%DESCR(JARY))

CALL

The first array
descriptor.

See Section
code.

5.2.4

5.4.2

for

reference.

examples

that

The

include

the

array

generated

machine

Function Return Values

The method

type

is passed

the

returning

value,

function procedure

summarized

in Table

values
5-1.

depends

the

data

Table 5-1
Function Return Values

Data

Type

Return

Logical

General

Method

Integer

Real

Double

Precision

RO:
Rl:

High-order result
Low-order result

Complex

RO:
Rl:

Real part
Imaginary part

Character

An
of

extra
the

argument

entry is
argument
entry

descriptor.
allocated
to
and the
the descriptor.

$LOC

points

At
run
contain

result,

5.2.5

added as the first
1list.
This
new

proper

character

entry
first
string

time,
storage
the
wvalue
of
address

stored

is
the
in

Built-In Function

The SLOC built-in function computes the address of a storage
element,
as
an
INTEGER*4
value.
This
value
can
be used in an arithmetic
expression.
This has
particular
applicability
for
certain
system
services
or
non-FORTRAN
procedures
that may require arguments that
contain the addresses of storage elements.

5.3

CALLING VAX/VMS

The VAX/VMS
you

can

describe
and
to
form.

operating

call

from

SYSTEM SERVICES
system provides
FORTRAN

number

programs.

The

the methods you can use to
ensure
that
you pass and

service

following

procedures
subsections

call
system
service
procedures,
receive information in the correct

FORTRAN CALL CONVENTIONS

You can invoke system services from a FORTRAN program by

including

function reference or a subroutine CALL statement in your program,
To specify a system
specifying the system service you want to use.
service,

use the form:

SYSS$service~-name

(a,...,a)

the
to
according
You pass arguments to the system services
a value, an
requirements of the particular service you are calling:

address, or the address of a descriptor may be needed, as described in

See the VAX/VMS System Services Reference
Section 5.3.3, below.
Manual for a full definition of all services.

Calling System Services by Function Reference

5.3.1

In most cases, you should check the return status after calling a
Therefore, you should call system services by
system service.
function reference rather than by issuing a subroutine CALL.
NOTE

must
functions
service
System
declared as INTEGER*4 functions.
For

example:

INTEGER*4

INTEGER*2

SYSS$CREMBX
ICHAN

MBX = SYS$SCREMBX(,ICHAN,,,,, 'MAILBOX')
IF

(.NOT.

MBX)

GO TO 100

In this example, the system service referred to is the Create Mailbox
then an
the system service name 1is declared,
First,
service.
INTEGER*2 variable (ICHAN) is declared, to receive the channel number.

The function reference allows a return status value to

stored

or .FALSE.

the variable MBX, which can then be checked for .TRUE.

return. If the function's return status is .FALSE., indicating
failure, control transfers to statement 100, at which point some form

of error processing can be undertaken.

particular

return

status,

You

can

check

for

the

system.

the return status to one of the status codes defined

For example:

(MBX

.EQ.

also

such as an access violation, by comparing

SS$_ACCVIO)

THEN

Refer to the VAX/VMS System Services Reference Manual for information
concerning return status codes. The return status codes are included
in the description of each system service.

5.3.2

Calling System Services as Subroutines

Subroutine calls to system services are made in the same way that
calls are made to any other subroutine. For example, to call the
Create Mailbox system service, issue a CALL to SYSSCREMBX, passing to
it the appropriate arguments,

as:

CALL SYS$CREMBX (,ICHAN,,,,, 'MAILBOX')

FORTRAN

This CALL corresponds to
difference
1is
that the
not

tested.

system

For
services.

5.3.3

Passing Arguments

Generally,
value, and
must

use

this

system
output

the

%VAL

CALL

CONVENTIONS

the function reference shown above.
The main
status code returned by the system service is

reason,

you

should

avoid

this

method

calling

System Services

services require input
arguments to be passed

arguments
to
by reference.

built-in

passing

function when

be
passed
Therefore,

by
you

arguments

input

certain
system
services,
to
make
sure they are passed the
data.
Some system services require character arguments
(see
5.3.3.3).
To
determine
the
argument
requirements
for
a
service, see the VAX/VMS System Services Reference Manual.

5.3.3.1

tell

Input

the

and Output Address Arguments - You will often
system service where to find input values and where

output values.
Thus, you must determine the hardware
argument:
byte, word, longword, or quadword.

data

correct
Section
system

need

type

store
of

the

For
can

input arguments that refer to byte, word, or longword values,
you
specify either constants or variables.
If you specify a variable,
you must declare it to be equal
to
or
1longer
than
the
data
type
required.
Table 5-2 lists the variable data type requirements.
For

output

required,
system

arguments

returns

you

must

declare

avoid

including

value

byte

variable

extraneous data.
a

word-length

exactly

1If,

for

variable,

the

length

example,
the

eight
bits
of the variable will not be overwritten on output;
the variable will not contain the data you expect.

VAX/VMS

Input Argument

Required

Output Argument

Declaration

Byte

BYTE,

Word

INTEGER*2,

Longword

INTEGER*4

Quadword

Properly

thus,

Table 5-2
Data Type Requirements

Variable

Type

the

leftmost

INTEGER*2,

Declaration

INTEGER*4

BYTE

INTEGER*4

INTEGER*2
INTEGER*4

dimensioned

array

Properly

dimensioned

array

Indicator

For

LOGICAL

arguments

routine,"
example,

want

the

"address

argument value

entry

mask"

the

"address

procedure.

For

routine

you

and

PROGA

declarations portion of
of

an external

if a system service requires the address
specify the routine PROGA, specify

EXTERNAL

address

described

declare

routine

for

use

the
as

5-6

FORTRAN
input

program

argument.

define

the

FORTRAN

store

output

produced

CALL

CONVENTIONS

system

services,

vyou

must

allocate

sufficient
space
to
contain
the
output.
You do this by declaring
variables of the proper size.
For example, the Create Mailbox
system
service produces a two-byte value.
Thus, you set up space as follows:
INTEGER*2

INTEGER*4

MBX =

ICHAN
SYS$SCREMBX

SYSSCREMBX(,ICHAN,,,,,'MAILBOX')

If the output is a quadword value, you must declare an
array
of
the
proper
dimensions.
For'' example,
the
Get
Time
system
service
(SYSSGETTIM) returns the time as a quadword binary value.
Thus,
you
would need to specify the following:
INTEGER*4
INTEGER*4

ISTAT =

SYSTIM(2)
SYSSGETTIM

SYSSGETTIM(SYSTIM)

DECLARE ARRAY

GET TIME

The type declaration INTEGER*4
SYSTIM(2) sets up a vector
of two longwords, into which the time value will be stored.

consisting

5.3.3.2
Defaults for Optional Arguments - All optional arguments have
default
values.
You
must use commas in place of omitted arguments.
For example, the Translate Logical Name
(SYSSTRNLOG)
system
service
takes
five
arguments.
If
you omit the last two arguments you must
include two commas in their place, as follows:
ISTAT =
An

invalid

reference would

ISTAT =

This

SYSSTRNLOG('LOGNAM',LENGTH,BUFFA,,)
result

if you

specified:

SYSSTRNLOG ('LOGNAM',LENGTH,BUFFA)

reference provides only

three arguments,

not the

requisite

five.

5.3.3.3
Passing Character Arguments - Some
VAX/VMS
system
services
(for
example,
the
Translate
Logical
Name
system service) require
character arguments for either input or output.
The Translate Logical

Name

system service

(SYSSTRNLOG)

accepts a logical name as input,

and

returns the associated logical name or file specification, if any,
as
output.
VAX/VMS
system services that process character data require
that arguments be passed by descriptor.
FORTRAN passes all
character
data
by
descriptor.
On input, a character constant, variable, array

element,

or expression is passed to the system service by

descriptor.

On
output, two arguments are needed:
the character variable or array
element to hold the output string, and an INTEGER*2 variable, which is
set to the actual length of the output string.
For example:
CHARACTER*64 BUFFA
INTEGER*2 LENGTH

INTEGER*4

ISTAT =

SYSSTRNLOG

SYSSTRNLOG ('LOGNAM',LENGTH,BUFFA,,)

5-7

FORTRAN

CALL

CONVENTIONS

The logical name LOGNAM is translated to its associated name -:or
file
specification,
and
the
output values, length and associated name or
file
specification,
are
stored
in
the
locations
you
specified -- LENGTH and BUFFA, respectively.

5.4

MACHINE CODE EXAMPLES

The following sections present examples
corresponding machine code.

5.4.1

FORTRAN

calls

and

their

Argument Passing Examples

The format used in the following examples shows FORTRAN
followed by generated object code argument lists.
Example

FORTRAN

Source

REAL

source

lines,

Code

INTEGER J(10)
CHARACTER*15 C
CALL SUB (X,J(3),C)

Object Code
ARGLST:

L$l:

.LONG

;

COUNT

.ADDR X

;

ADDRESS

.ADDR J+8

;

ADDRESS OF J(3)

+.ADDR LS$1

;

C DESCRIPTOR ADDRESS

.WORD 15
.BYTE 14

;
;
;
;

LENGTH OF C
CHARACTER TYPE CODE

.BYTE 1
.ADDR C
This

example

shows

how

the

SCALAR CLASS
ADDRESS OF C

compiler

generates

CODE

argument

list

for

the

arguments
specified
in
the
CALL
statement.
The
compiler
can
initialize the addresses of real variable X
and
array
element
J(3)
because
these
are
-explicitly
specified
in
the
CALL
statement.
Similarly,
the
compiler
has
enough
information
to
generate
an
initialized descriptor for the character string C.

FORTRAN CALL CONVENTIONS

Example 2:

FORTRAN Source Code
REAL X (10)
CHARACTER*15 C

CALL SUB

(X(I),

C(J:K))

Object Code

LS$1:

.LONG 2

0
.ADDR LS1

. LONG

-s

ARGLST:

COUNT

X(I) INITIALIZED AT RUN TIME
C(J:K) DESCRIPTOR ADDRESS

.WORD O

; C(J:K)

.LONG 0

;

;
;

.BYTE 14
.BYTE 1

LENGTH, SET AT RUN TIME

CHARACTER TYPE CODE
SCALAR CLASS CODE

BASE ADDRESS OF C(J:K), SET AT RUN
TIME

Run~time argument list initialization code
MOVL
MOVAF
SUBL3

I,RO
X-4[R0O] ,ARGLST+4
#1,J,R0O

; COMPUTE ADDRESS OF X (I) AND
; STORE IT IN THE ARGUMENT LIST
; COMPUTE THE LENGTH OF C (J:K)

MOVW

R1,LS1

; STORE LENGTH OF C(J:K)

MOVAB

C[RO] ,LS1+4

CALLG

ARGLST,SUB

SUBL3

RO,K,R1

H
:

LIST

ARGUMENT LIST

IN ARGUMENT

: STORE BASE ADDRESS OF C(J:K)
:

CALL SUBROUTINE SUB

In this example, the FORTRAN source

lines

real

define

array

and a character variable C, comprising 15
comprising 10 elements;
elements. The actual arguments passed to subroutine SUB are the I-th
element of array X, and the substring of C from the J-th to the K-th
character. The compiler generates an argument list consisting of
three

longwords;

the

first

1is

the

and

count,

the next two are

(respectively) the address of the I-th element of X and the address of
the descriptor of substring C(J:K). Note that the address of X(I) and

C(J:K), and the length of C(J:K)

are

compile time these values are unknown.

5.4.2

initialized

because

Argument List Built-In Function Examples

In the following examples, the FORTRAN source
followed by the generated object code.

code

1is

shown

first,

FORTRAN

Example

FORTRAN

Source

CALL

CONVENTIONS

$VAL

SUB

Code

(4,

%VAL(6))

Object Code
ARGLST:

CON4 :
As

shown,

first

the

entry,

.LONG 2
.ADDR CON4

;
;

ADDRESS

.LONG

;

VALUE

.LONG 4

;

compiler

but

generates

the

entry.

Example

FORTRAN

Source

COUNT
OF

address

actual

CONSTANT

for

the

value

constant

(6)

the

following

$REF
Code

CHARACTER*10

C,D

CALL

$REF (D))

SUB

(C,

Object Code
ARGLST:

.LONG

L$l:

.ADDR LS$1

’

COUNT

;

ADDRESS

.ADDR D

;

ADDRESS

.WORD

;

LENGTH

.BYTE

TYPE CODE
CLASS CODE
ADDRESS

.BYTE

;
;

.ADDR

;

DESCRIPTOR

As shown, the argument list entry for D is
the
address
of
D.
The
compiler
does not generate a descriptor for D, as it does for C, even
though C and D are both specified in the source program
as
character
variables.
Example

FORTRAN

Source

CALL

$DESCR

SUB

Code

(X,

%DESCR(X))

Object Code
ARGLST:

.LONG

.ADDR

.ADDR LS$1
L$l:

.WORD 4
.BYTE 10

COUNT
;

ADDRESS

;

ADDRESS OF X DESCRIPTOR

;

LENGTH
TYPE CODE
CLASS CODE
ADDRESS

.BYTE

;
;

.ADDR

In this example,
while the second

the first argument list entry
contains
an
entry contains a pointer to a descriptor.

5-10

address,"

FORTRAN CALL CONVENTIONS

5.4.3

Character Functions

The following example
lists

illustrates

are generated.

FORTRAN

Source

how

character

function

argument

Code

CHARACTER*10
= C(1,J)

C,D

Object Code
ARGLST:

LS$1:

.LONG 3
.ADDR LS1

;

COUNT
ADDRESS OF FUNCTION DESCRIPTOR

.ADDR I
.ADDR J

;s
;

ADDRESS OF 1
ADDRESS OF J

.WORD 10

LENGTH

.BYTE

TYPE

.BYTE

CLASS

.LONG

;

ADDRESS

ALLOCATE SPACE FOR 10 CHARACTERS

CODE
CODE

SUBL2

$#10,SP

;

MOVL

SP,LS1+4

SET ADDRESS

CALLG

ARGLST,C

CALL

FUNCTION

MOVC3

#10, (Sp) ,D

MOVE

RESULT TO D

MOVL

R1l,SP

;

REMOVE RESULT FROM STACK

In this example, an additional argument list entry is allocated;
descriptor of the return value of the character function C.

the

CHAPTER 6
ERROR PROCESSING AND

CONDITION HANDLERS

During program execution,
your
program
may
occasionally
encounter
errors
or
exception
conditions.
These
conditions may result from
errors that occur during I/0 operations, invalid input data,
argument
errors
1in
«calls
to
the mathematical library, arithmetic errors, or
system-detected errors.
VAX-11l FORTRAN IV-PLUS provides three methods
of controlling and recovering from errors:
1.

Run-Time Library default

ERR=

VAX-11 Condition Handling
condition handlers)

and

END=

error-processing procedures

specifications

I/0

Facility

statements

(including

user-written

These error-processing methods are
complementary,
and
can
be
used
together
in
the same program.
Thus, you have a number of options to
choose from in dealing with errors.
You can,
of
course,
allow
the
Run-Time
Library to handle errors for you.
This is the default case.
Or, you can use ERR= and/or END= specifications in I/0O
statements
to
provide special processing if an error or end-of-file occurs while the
I/0 statement is
being
executed.
Or,
you
can
provide
condition
handlers
to
tailor
error
processing to the special requirements of
your applications.
Note:
This option should be
undertaken
only
by
more
experienced
users, in particular those familiar with the VAX-1l1
Condition Handling Facility.

The Run-Time Library provides default
processing
for
all
exception
conditions
that occur during FORTRAN program execution.
It generates
appropriate messages, and takes action to recover
from
errors
where
possible.
Section
6.1
describes
FORTRAN-specific Run-Time Library
error processing.
Appendix
B
describes
FORTRAN-specific
Run-Time
Library

error messages.

The VAX-11l Condition Handling Facility provides all
error
processing
in
the
Run-Time
Library.
You
<can also use it directly to provide
procedures (user-written condition handlers) for processing
exception
conditions
that
occur during your program's execution.
For example,
you can include
that
1is
more

code to respond
appropriate
to

to certain kinds of errors
a
particular
application

1in
a
than

way
the

processing performed by the Run-Time Library.
The
use
of
condition
handlers
requires considerable experience:
it is not appropriate for
novice users.
You should be
familiar
with
the
condition
handling
description
in the VAX-11l Common Run-Time Procedure Library Reference
Manual, and in the VAX-11/780 Architecture Handbook before you attempt
to write a

condition handler.

ERROR PROCESSING AND

6.1

RUN-TIME

LIBRARY DEFAULT

The Run-Time Library contains

CONDITION

HANDLERS

ERROR PROCESSING

condition

handlers

that process

number

of
errors that may occur during FORTRAN program execution.
A default
action is defined for each FORTRAN-specific exception
condition
that

the
Run-Time
Library
recognizes;
Table
6-1
These default
actions
occur
wunless
overridden
condition handler (see Section 6.2).

1lists these actions.
by
a
user-written

Some errors result in recovery action.
For example, you can
use
the
ERR=
specification
to
code
an
error
statement
label
in
an I/0
statement.
In some cases, the error may be ignored and execution will
continue.
Or,
recovery
may be initiated based on the nature of the
error.
You can use the ERRSNS system subroutine
to
determine
which
error
occurred.
ERRSNS
returns a FORTRAN-specific
system status codes for the last error that occurred;
The ERRSNS routine is described in Appendix C.

6.1.1

error number and
see Table
6-1.

Using ERR= and END= Transfers

By including
an
ERR=label
or
END=label
specification
in
an
1I/0
statement,
you
can transfer control to error processing code in your
program.
If you use an END= or ERR= transfer to process an I/0 error,
no
error message is printed and execution continues at the designated
statement.
However, if a severe error occurs while an
I/0O
statement
is
being
processed,
and
you
did not specify an ERR= transfer, the
default action is to print an error message and terminate execution.
If you specify ERR=label in an
during I/0 statement execution,
control to the statement at the

I/0
statement
and
an
error
the Run-Time Library transfers
label specified.
For example:

occurs
program

WRITE (8,50 ,ERR=400)
If an error occurs during the
the statement at label 400.
You

can

also

statement.

write

specify ERR=label
For

operation,

keyword

control

transfers

OPEN

example:

OPEN (UNIT=INFILE,TYPE='OLD',ERR=100,NAME=FILN)
If errors
transfers

are
detected
during
to statement 100.

The END=label
condition, as

specification
follows:

can

OPEN

statement

used

execution,

handle

control

end-of-file

READ (12,70 ,END=550)

end-of-file

executed,

control

detected

transfers

If an
end-of-file
executed,
and
you

while

this

statement

550.

I/O

statement

1is

being
)

1is
detected
while
a
READ
statement
1is
being
did
not
specify
END=label,
an error condition
specified ERR=label, control
is
transferred
to
the

occurs.
If you
specified statement.

ERROR PROCESSING AND

6.1.2

Run-Time Library Error

CONDITION

HANDLERS

Processing Control

The Run-Time Library's error processing depends on three factors:
severity
of
the error;
whether the error permits continuation;

whether

Table

6-1

transfer

lists

Library.
symbolic
code

ERR=

the

was

provided

FORTRAN-specific

the

errors

case

processed

I/O

the

and

error.

the

Run-Time

For each error, the table shows the error message text, the
condition name, the FORTRAN-specific error code, the severity
the

error,

severity code

generated

and

when

the

type

included

error

recovery

part

condition

action.

the

error

detected.

All

code

that

FORTRAN-specific

errors have severity codes of either error (E) or
severe
error
(F).
As shown in Table 6-1, most FORTRAN-specific errors are severe.
If no
recovery action is specified for a
severe
error,
program
execution
terminates by default.
Two

types

include
and

the

recovery

ERR=
error

message,
and
ERR= transfer,

receive
ERR=

one

possible:

ERR=

However,

the

message

the

was

severe

RETURN.

specified

will

not

error

messages,

specified.

You

can

resulting
program termination,
error in your source program.

that

recovery

and

you

Most

I/0O

ERR=

type

RET

means

the

including

implication
the

an ERR=

recovery

and

error

specify an
Thus,
if
you

displayed.

circumvent

errors

transfer,

severity was F, your program terminates
with
an
exit status of severe error.
If you did

transfer

are

recovery.

that

error,

and

the

transfer

for

this

continuation

are

possible
only
if
you
establish a condition handler for that
User-written condition handlers are described in Section 6.3.

error.

The letter C in the "Sev" column means that you can continue execution
immediately
after
the
error if a user-written condition handler has
changed the severity code to error (E) or warning (W).
If there is no
letter

immediately after
execution will be
When

errors

occur

exits;

that

message

1is

is,

the

<column,

for

which

execution of

printed.

condition handler

"Sev"

the
error.
terminated.

recovery

cannot

attempt

type

the program

prevent

that performs

you

program

unwind

(see

continue

specified,

terminated,

execution

so,

the

and

termination,

Section 6.3).

program

an error

include

ERROR PROCESSING AND CONDITION HANDLERS

Table

6-1

Summary of FORTRAN Run-Time Errors

FORTRAN

Condition
Symbol

Err
#

|Sev

Rec.
Type

Message Text

FOR$_NOTFORSPE | 1 | F
FOR$_REWERR
20|
F
FORS DUPFILSPE | 21|
F

ERR=

NOT A FORTRAN-SPECIFIC ERROR
| REWIND ERROR

FOR$_FILNOTFOU | 29|

ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=

ERR= | FILE NOT FOUND

FORS$ INPRECTOO | 22|
FORS BACERR
23|
FOR$S ENDDURREA | 24|
FOR$ RECNUMOUT | 25|
FOR$ OPEDEFREQ | 26|
FORS MORONEREC | 27|
28|
FORS CLOERR
FOR$_OPEFAI
FOR$_MIXFILACC
FORS_INVLOGUNI
FOR$_ENDFILERR
FOR$_UNIALROPE
FOR$_SEGRECFOR
FOR$_ATTREANON
FOR$_INCRECLEN
FOR$_ERRDURWRI
FOR$_ERRDURREA

|
|
|
|
|
|
|
|
|

F
F
F
F
F
F
F

30|
F
31|
F
32|
F
33|
F
34|
F
35|
F
36 | F
37|
F
38 | F
39|
F

40|
F
41|
F
42|
F
43|
F
44|
F
45|
F
46|
F
47|
F
48|
F
59|F,C |
60|
F
61 |F,C |
62|
F

FORS_OUTCONERR | 63
FORS_INPCONERR | 64

[E,C |
|F,C |

FORS OUTSTAOVE | 66|
F
F
FORSINPSTAREQ | 67|
FOR$_VFEVALERR | 68 |F,C |

ERR= | DUPLICATE FILE SPECIFICATIONS

ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=
ERR=

ERR= | OUTPUT CONVERSION ERROR
ERR= | INPUT CONVERSION ERROR

SS$ INTOVF
SS$_INTDIV
SS$_FLTOVF
SS$ FLTDIV
SS$_FLTUND
SS$ SUBRNG

70 |F,C |
71 |F,C |
72 |F,C |

ERR= | OUTPUT STATEMENT OVERFLOWS RECORD
ERR= | INPUT STATEMENT REQUIRES TOO MUCH DATA
ERR= | VARIABLE FORMAT EXPRESSION VALUE ERROR
RET
INTEGER OVERFLOW
RET
INTEGER ZERO DIVIDE
RET
FLOATING OVERFLOW

73|F,C|

RET

MTH$S INVARGMAT |
MTHS UNDEXP
MTH$ LOGZERNEG. |
MTH$ SQUROONEG |
MTHS SINCOSSIG |

81|
F
82 |F,C|
83 [F,C |
84 [F,C |
87 |F,C |

RET
RET
RET
RET

74 |F,C |
77 |F,C|
MTHS_WRONUMARG | 80|
F

MTH$_FLOOVEMAT | 88 (F,C |
MTH$ FLOUNDMAT | 89 |F,C |
FOR$_ADJARRDIM | 93|
F

RET
RET

RET

FLOATING

ZERO DIVIDE

FLOATING UNDERFLOW
SUBSCRIPT OUT OF RANGE
WRONG NUMBER OF ARGUMENTS

INVALID ARGUMENT TO MATH LIBRARY
UNDEFINED EXPONENTIATION
LOGARITHM OF ZERO OR NEGATIVE VALUE
SQUARE ROOT OF NEGATIVE VALUE
SINE OR COSINE SIGNIFICANCE LOST

FLOATING OVERFLOW IN MATH LIBRARY

FLOATING UNDERFLOW IN MATH LIBRARY
ADJUSTABLE ARRAY DIMENSION ERROR

ERROR PROCESSING AND CONDITION HANDLERS

NOTES:

taken

after

completion

the

1I/0

The ERR= transfer

If no ERR= address has been defined for error 63, the program
The entire
message is printed.
error
the
after
continues
the
indicate
to
overflowed field is filled with asterisks,

statement for continuable errors numbered 59, 61, 63, 64, and
68;
the resulting file status and record
position
are
the
However, other I/0 errors
as if no error had occurred.
same
detected;
1is
error
the
as
take the ERR= transfer as soon
thus, file status and record position are undefined.

error

the output

87,
84,
83,
Function return values for errors numbered 82,
88, and 89 can be modified by means of user-written condition
handlers.

record.

See Section

Error number 1

detected

6.4.

(FOR$_NOTFORSPE)

indicates that an

that was not a FORTRAN-specific error;

error

that is,

was

table.
the
1in
was not reportable through any other message
you call ERRSNS, an error of this kind returns a value of
If
that
system condition value
unique
the
To obtain
1.
-error, use the fifth argument of the call to
the
identifies
ERRSNS

6.1.3

(condval).

See Section C.4.

Using the ERRSNS Subroutine

ERRSNS
errors.
You can use the ERRSNS system subroutine to process
the exact nature of the error and take action
lets you determine
error
integer
FORTRAN-specific
a
contains
Table 6-1
accordingly.
This error number can be
for each of the errors in the table.
number
situations
In many
obtained by calling ERRSNS after an error occurs.
code to react to specific I/0 errors, once you know
can provide
you
the ERRSNS
wuse
information,
this
To get
occurred.
which error
routine, as shown in the following example.
10

CHARACTER*40 FILN
ACCEPT *, FILN

R=100
,NAME=FILN)
OPEN (UNIT=INFILE,TYPE='OLD',ER
(process

.
1060

CALL ERRSNS (IERR)
IF (IERR .EQ. 29) THEN

TYPE *,

ELSE

TYPE *,

ELSE

TYPE

'FILE:',

(IERR

.EQ.

input

FILN,

43)

'FILENAME:',

file)

'DOES NOT EXIST, ENTER NEW FILENAME'

THEN

FILN,

'WAS BAD, ENTER NEW FILENAME'

'FAILURE ON INPUT FILE,

I/O ERROR CODE=',

IERR

STOP
ENDIF

GO TO 10
END

As shown, the OPEN statement contains an ERR=100
keyword,
causing
a
branch to the ERRSNS subroutine if an error occurs during execution of
the
The ERRSNS subroutine returns an error number value in
the OPEN.
The program then uses the value of IERR either
integer variable IERR.
to print a message indicating the nature of the error and continue, or
a message indicating that the error was too severe to allow
to print
the program to continue.

ERROR PROCESSING AND

See Appendix C

6.2

for

information

OVERVIEW OF THE VAX-11

A condition handler
condition
occurs.
software
memory

related.
access

CONDITION

about

the

HANDLERS

ERRSNS

subroutine.

CONDITION HANDLING FACILITY

a procedure that is invoked when
an
exception
The exception condition may be either hardware or
Hardware
exceptions
include
floating
overflows,

violations,

and

the

use

reserved

operands.

exceptions include output conversion errors,
end-of-file
and invalid arguments to mathematical procedures.
When the VAX/VMS system creates
a
system-defined
condition
handler

Software

conditions,

user
that

process,
it
establishes
a
will,
in the absence of any
user-written condition
handler,
process
errors
that
occur
during
execution
of
the user image.
Thus, by default, run-time errors will
cause this condition handler
to
print
one
of
the
standard
error

messages

and

terminate

severity

code

associated

with

This default
when
an

the

condition handler

error.

continue

is
no

undertaken
lower-level

The

default

the

appropriate

message

file,

SYSSOUTPUT
present.

the

and

condition

execution
continues.
termination;
and the
program exit
You

can

and

relying

during

the

Definitions

Before

reading

to
one

establish

the

established

message

user.

Messages

that

file,

system

further,

you

Condition value - an
particular exception

Procedure

subroutine,

an
or

Procedure
executes.

This

permits

reached
to

handler,

1list
to

of
the

output
are

these

the

respond

routine,

sent

files

are

continuation,

error

application

For

the

both

program

handlers

example,

describe
program,

messages.

become

INTEGER*4
condition.

executable
function.

condition

specifically

should

own

familiar

with

the

program

unit;

main

you

can
conditions
instead

following

value
that
identifies
See Section 6.3.4.

program,

activation

- the environment in
which
a
procedure
environment includes a unique stack frame on
stack;
the stack frame contains the address of a

the run-time
condition
handler

procedure

applications.

that

terms:

procedure
called;

handler

system's

the

your

execution

standard

6.2.1

has

the SYSSERROR

accommodate
the
needs
of
create and display
messages
encountered

condition

have
resignaled
(see
the
condition
handler
responds

calls

was

depending

Otherwise,
the default handler forces program
condition value
(see
Table
6-1)
becomes
the

status.

create

last
a

procedure

error.

handler

the

find

if
all
lower-level
handlers
definitions below), the default
send

execution,

error.

for
the
procedure
activation.
A
activation is created
every
time
a
procedure
the
procedure
activation
1is
deleted
when
returns.

Condition handler - a function
specifies
as
the
procedure
condition occurs.

that
to
be

new
is
the

a
procedure
activation
called when an exception

ERROR PROCESSING AND CONDITION HANDLERS

Establish - the process of placing the address of a
condition
handler
in
the
stack
frame
for
the
current
procedure
activation.
A condition handler established for
a
procedure
is
called
automatically
when an exception condition occurs.
In FORTRAN, condition handlers are established by means of the
LIBSESTABLISH procedure.
See Section 6.4.2.

Signal -

the means by which the

LIB$SSIGNAL execution;

LIBSSTOP -~
execution,
See

exception

signal

signal
unless

Section

a condition and possibly continue

program

a severe error and do not
continue
program
a condition handler performs an unwind.

6.2.2.

Resignal - the means by which a
condition
handler
indicates
that
the
signal
procedure
1is
to
continue searching for a
condition handler to process a previously signaled error.
To
resignal,
a condition handler returns the SS$_RESIGNAL value.
See

occurrence

condition is made known;
signals are initiated by a call to a
signal procedure, which then calls a condition handler.
There
are two signal procedures:

Section

6.3.3.

Condition symbol
in

the

symbol

used

specify a condition value,

form:

fac$_symbol

where fac is a facility name prefix and
specific condition.
Table

6.2.2

6~1

lists

FORTRAN-related

symbol

identifies

condition symbols.

Unwind - the
act
of
returning
control
to
a
particular
procedure
activation,
bypassing
any
intermediate procedure
activations.
For example, if X calls Y and Y calls Z;
and
1Z
detects
an error;
then a condition handler associated with Z
can
unwind
to
X,
bypassing
Y.
Control
returns
to
X
immediately following the point at which X called Y.

Condition Signals

A
condition
signal
consists
of
system-supplied signal procedures,

CALL LIBSSIGNAL
CALL LIBSSTOP

a
call
to
in either of

(condition-value,

one
of
the
two
the following forms:

arg,...)

You can use a condition signal when
you
do
not
want
to
handle
a
condition
in
the
routine in which it was detected, but wish to pass
the information to a higher-level routine.
If the
current
procedure
can
continue
after
the
signal
1is
made,
call
LIBSSIGNAL.
A
higher~-level procedure can then determine whether program execution is
continued.
If
the condition will not allow the current procedure to
continue, call LIBSSTOP.

ERROR PROCESSING AND CONDITION HANDLERS

To pass the condition value, you
must
use
the
$VAL
argument
1list
built-in function (see Section 5.2.3.1).
Condition values are usually

expressed

condition

CALL LIB$SIGNAL
Additional
information

symbols.

For

example:

(3%VAL (MTHS_FLOOVEMAT))

arguments
about the

may
be
error.

included

When called, a signal procedure searches
examining
the
preceding
stack
frames
procedure that handles the condition, or
is reached.

6.2.3

Handler

provide

supplementary

for a
condition
handler
by
in
order,
until it finds a
the default condition handler

Responses

A condition handler responds to an exception by taking action in three
major
areas:
1) condition correction, 2) condition reporting, and 3)
execution control.
First, the handler may determine whether it can correct the condition.
If it can, the handler will take the appropriate action, and execution
will continue.
If it cannot correct the condition,
the
handler
may
resignal
the
condition.
That
1is,
it
may
request
that
another
condition handler process the exception.
Condition reporting
following actions:

performed

count of

handlers

Maintaining
execution

Resignaling the same condition
to your terminal or log file

Changing
the
severity
field
resignaling the condition

Signaling a different
message oriented to a

Execution can be

affected

Continuing from the
a call to LIBSSTOP,

Unwinding

resulted

procedure's
®

the
the

exceptions

can

involve

encountered

send

the

number

ways.

condition

Among

signal.
If the signal was
the program will exit.

establisher
exception.

function value

during

the point

The

handler

returned.

the

program

appropriate message

condition, for
example,
specific application
a

one

wvalue

and

produce

them are:

issued

the

specifies

through

call
the

that
called

Unwinding to the
establisher's
caller
(the
procedure
that
called
the
procedure
that
established
the
handler).
The
handler specifies the called procedure's function value to
be
returned.

ERROR PROCESSING AND CONDITION HANDLERS

6.3

USER-WRITTEN CONDITION HANDLERS

The following sections describe how to code and establish a
condition
handler,
and
provide
some
simple
examples.
See Appendix C of the

VAX-11/780 Architecture
Handbook,
and
the
VAX-11]
Common
Run-Time
Procedure
Library
Reference
Manual
for
more
details on condition
handlers.

6.3.1

Establishing and Removing

Handlers

When
a
procedure
1is
called,
no
condition
handler
1is
1initially
established.
If
you
want
to
use
a
condition
handler, you must
"establish it by calling the Run-Time Library
procedure
LIBSESTABLISH
and including in the call such code as the following:
EXTERNAL

HANDLER

CALL LIBSESTABLISH

(HANDLER)

To remove an established handler,

LIBSREVERT,

follows:

call

the Run-Time Library

procedure

CALL LIBSREVERT

As a result of this call, the condition
handler
established
in
the
current
procedure
activation
1is removed.
When a procedure returns,
the condition handler established by the
procedure
1is
automatically
removed.

6.3.2

FORTRAN Condition Handlers

A FORTRAN condition handler is an INTEGER*4 function
that
when
an
exception
condition
occurs.
You
must
define
arguments for a condition handler:

1is
called
two dummy

An integer array to refer to the argument list from the
call
to
the
signal procedure (the "signal arguments").
That is,
the list of arguments included in
CALL
LIBS$SIGNAL
or
CALL
LIBSSTOP (see Section 6.2.2).

An integer array
to
refer
to
information
concerning
the
procedure
activation
that established the condition handler
(the "mechanism arguments").

For example,

you can define a condition handler

INTEGER*4
INTEGER*4

as follows:

FUNCTION
HANDLER (SIGARGS,MECHARGS)
SIGARGS
(6) ,MECHARGS (5)

ERROR PROCESSING AND CONDITION HANDLERS

The array SIGARGS is
wused
procedure.
The values from

to
the

obtain
information
signal procedure are

Value

Meaning

SIGARGS (1)

Indicates

how many

this vector
SIGARGS (2)

array

handler.

The

MECHARGS

activation

are

being

passed

count).

Indicates optional arguments as specified by the
call
to
LIBSSIGNAL or LIBSSTOP;
note that the
dimension bounds for the
SIGARGS
array
should
specify as many entries as necessary to refer to
the optional arguments

status

used

MECHARGS

the
a

obtain

5-element
4

FRAME

MECHARGS (3)
MECHARGS (4)
MECHARGS (5)

DEPTH
RO
R1

(MECHARGS (1))

that

array

MECHARGS (1)

element

information

procedure

MECHARGS (2)

first

arguments

(argument

Indicates
the
condition
being
signaled
(condition
value).
See
Section
6.3.4
for a
discussion of condition values.

SIGARGS
(3 to

The

from
the
signal
listed below.

the

about

established

the

procedure

the

condition

form:

argument

count of

this

vector

(4).
FRAME

(MECHARGS (2))

stack

frame

DEPTH

(MECHARGS (3))
made from the

been

contains

the

address

that established

the

handler.

the

procedure

contains the depth (number
procedure activation, up to

the

activation

calls)
that
have
point at which the

exception occurred.
MECHARGS
(4)

and

the

6.3.3

time

Handler

MECHARGS
(5)

the

contain

the

values

registers

and

Function Return Values

Condition
handlers
specify
function
return
subsequent
execution.
Function
return values
defined in Table 6-2.

values
to
control
and their effects are

Table 6-2
Condition Handler Function Return Values

Value

SS$_CONTINUE

Effect

Continue execution
from
signal
was
1issued
by

the
signal.
a
«call
to
the program exits.

however,

SS$_RESIGNAL

signal.

Resignal,
handler

continue

process

the

condition.

for

If
the
LIBSSTOP,

condition

ERROR PROCESSING AND CONDITION HANDLERS

A condition handler can
also
request
a
stack
unwind,
by
calling
SYSSUNWIND
before
returning.
If SYSSUNWIND is called, the function
return value is ignored.
The handler modifies the saved registers
RO
and

the mechanism arguments to specify the called procedure's

function value.

A stack

unwind can be made

to one of

Unwind to the establisher,
resulted in the exception.

two places:

at
the
point
Specify:

the

<call

that

CALL SYSSUNWIND
(3VAL (MECHARGS (3)) ,%VAL (0))

Unwind to the procedure that called the establisher.

Specify:

CALL SYSSUNWIND
($VAL (0) ,%VAL(O0))

6.3.4

Condition Values and Symbols

Condition values
are
used
by
VAX-1l1l
to
indicate
that
a
called
procedure
has
either
executed successfully or failed, and to report
exception
conditions..
Condition
values
are
INTEGER*4
values,
consisting
of
fields
that indicate which system component generated
the value;
the reason the value was generated;
and the
severity
of
the condition.
31

A condition value has

28 27

the

form:

16 15

FACILITY

C field

(31:28)

Message

(15:3)

describes the condition that
occurred.
Bit
15 = 1 to indicate the message is specific to

severity code,

indicate

follows:

- warning
-

success

N WNEEO

(2:0)

SEV

identifies
the
software
component
that
generated the condition value.
Bit 27 =1 to
indicate a customer facility.
Bit 27 = 0
to
indicate a DIGITAL facility.

a single facility.
Bit 15 =
system wide message.

Sev

control bits

(27:16)

Facility

MESSAGE

error

information

severe

reserved

error

A warning severity code (0) indicates that output
was
produced,
but
the
results
might
not be what you expected.
An error severity code
(2) indicates that output
was
produced
even
though
an
error
was
detected.
Execution
can
continue,
but the results will not all be
correct.
A severe error code (4) indicates that the error was of such
severity
that output was not produced.
One of the functions that can
be performed by a condition handler is altering the severity code of a

condition

value,

to allow execution to continue or to force an exit,

depending on the circumstances.

6-11

ERROR PROCESSING AND CONDITION HANDLERS

The condition value is passed as
the
second
element
of
the
array
SIGARGS.
There
are
times
when
you
may require that a particular
condition be identified by an exact match.
That is, each bit
of
the
condition
value
(31:0)
must
match
the
specified
condition.
For
example,

you

may want

process

floating

overflow condition

only

its
severity
code
1is
still
4 (that is, if no previous handler has
changed the
severity
code).
As
noted
above,
a
typical
handler
response is to change the severity code and resignal.
In many cases, however, you
may
want
to
respond
to
a
condition,
regardless
of the value of the severity code.
To ignore the severity
and control fields
of
a
condition
value,
use
the
LIB$MATCH_COND
function, as follows:

index

= LIBSMATCH_COND

(SIGARGS(2),con-1,con-2,...con-n)

This function compares bits 27:3 of the value in SIGARGS(2) with
each
of
the
condition
values
that
follow (con-1 through con-n).
If it
finds a match, the index value is assigned according to
the
position
of the matching condition value in the list.
That is, if the match is
with the third condition value following SIGARGS(2), then index
=
3.
If
no
match is found, index = 0.
The value of the index can then be
used to transfer control, as in the following example:
INTEGER*4

INCLUDE

INTEGER*4

FUNCTION HANDL (SIGARGS,MECHARGS)

'SYSS$LIBRARY:FORDEF'
SIGARGS
(6) ,MECHARGS (5)

INDEX=LIB$MATCH_COND
(SIGARGS (2) ,FOR$_FILNOTFOU,
FOR$_NO_SUCDEV,FOR$_FILNAMSPE ,FOR$_OPEFAI)

GO T0 (100,200,300,400) ,INDEX
HANDL=SS$_RESIGNAL
RETURN

If no match is found between the condition value in SIGARGS
(2) and any
of
the values in the list, INDEX=0, and control is transferred to the
next executable statement after the computed GO TO.
A match with
any
of
the
values
1in
the
1list
transfers control to the corresponding
statement in the GO TO list.
Thus, if SIGARGS(2) matches FORS$_OPEFAI,
control
is
transferred
to statement 400.
Note the use of condition
symbols

condition

represent

condition

symbols and

their

values..

Table

6-1

1lists

the

FORTRAN

values.

The
system
provides
parameter
files
comprising
condition
symbol
definitions.
When you write a condition handler, you must specify one
of the following
parameter
files,
as
appropriate,
1in
an
INCLUDE
statement:

SYSSLIBRARY
: FORDEF.FOR
This file contains definitions for
FORTRAN-specific Run-Time Library.

FOR$_xyz
For

example:

FOR$_INPCONERR

6-12

all condition symbols from the
These symbols have the form:

ERROR PROCESSING AND CONDITION HANDLERS

SYSSLIBRARY :MTHDEF.FOR

This file contains definitions for all condition symbols from the
These symbols have the form:
mathematical procedures library.
MTHS _xyz
For

example:

MTH$ SQUROONEG

SYSSLIBRARY :SIGDEF.FOR

contains

This file

miscellaneous

FORTRAN condition handlers.

definitions

symbol

These symbols have the form:

wused

SS$_xyz
For

example:

SS$_FLTOVF

You can obtain a listing of the condition symbols for any facility and
their corresponding hexadecimal values by means of the following
procedure:

Create a file that contains the lines

SxxDEF

GLOBAL

+END

To simplify the procedure, make sure the file type is MAR.

Specify a value for xx according
symbols you're concerned with.
$SSDEF

the

set

For example:

condition

GLOBAL

.END

the command:

Enter

$ MACRO

file-spec

If
where file-spec specifies the file you created in step 1.
you specified a file type of MAR, you need to enter only the
file name with this command.
the command:

Enter

$ LINK/MAP/FULL/NOEXE

file~-spec

Following execution of this command, you will have a file
the symbols defined in
that contains all
(file-name MAP)
xxDEF,

in numeric and alphabetic order.

Example:

$ EDIT
(create
$

MACRO

DEF.MAR
file)
DEF

$ LINK/MAP/FULL/NOEXE

DEF

6-13

ERROR PROCESSING AND

6.3.5

Floating Overflow,

CONDITION HANDLERS

Zero Divide

Exceptions

Some

conditions involving floating point operations require
that
you
take
special
action
if
vyou wish to continue execution.
Operations
that involve floating overflow, divide by 0, computing the square root
of
a negative number, etc.
store a unique result known as a floating

reserved

operand.

this

subsequent

floating

point

operation

accesses

result,
a
hardware
reserved
operand
fault
is generated
signaled.
This may continue indefinitely if the reserved
operand

not

changed

and

accessed

computation.
To
by

each

subsequent

reserved

operand

precision.
For
Common Run-Time

will

changed

it, the r acorvad
cperand
include the argument, and
This
ensures
that
the
et

correctly

hd N

regardless

information
on
LIBSFIXUP FLT,
Procedure Library Reference Manual.

see

its

the

VAX-11

that

apply

CONDITION HANDLER EXAMPLES
examples

typical

Example

this

FORTRAN

section

illustrate

procedures.

condition

handlers

The following example shows a
matrix
inversion
procedure, using
logical function INVE
to indicate
RT success or failure. That is, if
matrix can be inverted, INVERT returns the logical
value
.TRUE.
the
matrix
is singular, INVERT returns the logical value .FALSE.
the

1is
the

LIBSFIXUP_FLT (SIGARGS,MECHARGS,1.7D38)

The third argument is optional.
If you omit
is
changed
to
+0.0.
However, you should
specify it as’'
a
double
precision
value.

perform

example:

CALL

The

allow computation to continue, you must change the reserved operand
calling
the
Run-Time
Library routine LIBSFIXUP_FLT, as shown in

this

6.4

attempt

and

matrix

inversion

procedure

being

executed,

or divide-by-zero exception may occur.
A
provided to recover from these exceptions
to

the

INTEGER*4

calling

program.

function.

Note

that

the

floating

the
the
If
As

overflow

condition handler (HANDL) is
and return the value .FALSE.
condition handler is defined as

ERROR PROCESSING AND CONDITION HANDLERS

LOGICAL FUNCTION
DIMENSION A(N,N)

INVERT

(A,N)

HANDL

EXTERNAL

CALL LIBSESTABLISH
INVERT = .TRUE.
.

(HANDL)

!
!

ESTABLISH HANDLER
ASSUME SUCCESS

(INVERT THE MATRIX)

RETURN
END

INTEGER*4

FUNCTION

(SIGARGS, MECHARGS)

HANDL

(5)
SIGARGS (3), MECHARGS
INTEGER*4
INCLUDE 'SYSSLIBRARY:SIGDEF'
! ASSUME RESIGNAL
HANDL = SS$_RESIGNAL

(SIGARGS(2)

.EQ.

SIGARGS
(2) .EQ.
MECHARGS
(4) =

CALL

SS$ FLTOVF

SS$_FLTDIV)

.OR.

THEN

.FALSE.

SYSSUNWIND (%VAL(0),%VAL(0))

ENDIF

RETURN
END

the condition
If an exception occurs during the execution of INVERT,
The handler must first determine whether
is called.
(HANDL)
handler
the condition
Thus,
it can deal with the condition being signaled.
If the condition is
the condition value (SIGARGS(2)).
tests
handler
floating overflow or floating division by
0,
the
condition
handler
causes a return from INVERT with the value

.FALSE.

The condition handler uses the unwind procedure to force a

return

1is
The 1logical value .FALSE.,
that called INVERT.
the procedure
stored in the saved RO element of the mechanism vector
(MECHARGS(4)).

This value 1is used as the function value for INVERT when the unwind
the condition
occurs. The handler calls SYSSUNWIND and returns;
handling facility then gets control and actually performs the unwind
operation. Note that the function value from the wuser condition
handler

(HANDL = SS$_RESIGNAL)

is ignored if SYSSUNWIND is called.

division by
If the exception condition is not a floating overflow or
condition handler returns a value of SS$_RESIGNAL, indicating
the
0,
The
condition.
the
directly with
to deal
able
that it is not
immediately preceding procedure activation will then be checked for a
condition handler;
continuing until an established condition
handler
or the default condition handler

is reached.

ERROR PROCESSING AND

Example

CONDITION

HANDLERS

The following
example illustrates a condition
handler that processes
the Run-Time Library conditions MTH$ FLOOVEMAT and MTHS$ FLOUNDMAT. The
purpose

the

condition

the Run-Time Library
0.0, and to suppress

MAIN

handler

modify

the

value

returned

-0.0)

from the default value (reserved operand
the printing of an error message.

PROGRAM

EXTERNAL

HDLR

CALL LIBSESTABLISH

(HDLR)

X =.EXP(Y)
END

INTEGER*4
INTEGER*4

INCLUDE
INCLUDE
IF

FUNCTION HDLR (SIGARGS,MECHARGS)
SIGARGS(2), MECHARGS
(5)

'SYSSLIBRARY:SIGDEF'
'SYSSLIBRARY:MTHDEF'

(SIGARGS(2)

.EQ.

MTH$_FLOOVEMAT

SIGARGS(2)

.EQ.

MTHS$_FLOUNDMAT)

MECHARGS
(4) =

MECHARGS
(5)

HDLR =

.OR.

THEN

SS$_CONTINUE

ELSE

HDLR =

SS$_RESIGNAL

ENDIF
RETURN
END

When an exception condition occurs, HDLR is called, and
the
contents
of
the
second
element
of
the
signal
vector
are
compared
with
MTH$ FLOOVEMAT
and
MTH$ FLOUNDMAT.
If
either
of
the
specified

conditions is detected,

07s are placed in the saved RO and Rl elements

of array MECHARGS, the value SS§$_CONTINUE is
returned
in
HDLR,
and
execution
continues
in
the
math
1library
and
no error message is
printed.
RO and Rl are returned.
If the
condition
is
not
one
of
those specified, SS$_RESIGNAL is returned, and the preceding procedure

activations

are

searched

for

an established

6-16

condition handler.

CHAPTER 7
FORTRAN SYSTEM ENVIRONMENT

This chapter discusses aspects
of
the
relationship
between
VAX-11
FORTRAN
IV-PLUS
and
the
VAX-1ll
system.
The purpose is to provide
insights that will permit you to use VAX-11 FORTRAN IV-PLUS in
a
way
that
makes
best
use
of
its
features.
The following subjects are
discussed:

7.1

Program sections

Storage

FORTRAN-supplied

ENTRY

Floating point data representation

allocation
functions

loops
statement

arguments

PROGRAM SECTION USAGE

The storage required by a
VAX-1l1
FORTRAN
IV-PLUS
program
unit
is
allocated
1in
contiguous areas called program sections (PSECTs).
The
VAX~-11] FORTRAN IV-PLUS
compiler
implicitly
declares
three
PSECTs,
named:

SCODE

- This program section contains all executable code.

SPDATA

- This program section
constants and FORMAT

SLOCAL

-~ This program section contains read/write data local
the program unit.

In addition,
PSECT
with
block PSECT

contains
read-only
statements).

data

(e.g.,

each COMMON block you
declare
causes
allocation
of
a
the
same
name as the COMMON block.
(The unnamed COMMON

is named

$BLANK.)

The linker controls memory allocation and
attributes of each PSECT.
See Table 7-1.

sharing

according

Each module comprised by your program is named according to

the

name

specified
in
the
PROGRAM,
BLOCK
DATA,
FUNCTION,
or
SUBROUTINE
statement used in creating the module.
You can use the module name to
qualify
the
PSECT
name
specified
1in
LINK commands.
Refer to the
VAX-11l Linker Reference Manual.

FORTRAN

Defaults

are

applied

SYSTEM

PROGRAM and

PROGRAM default
DATA default

BLOCK

ENVIRONMENT

BLOCK DATA

statements.

They are:

source-file-name$MAIN
source-file-nameS$DATA

PSECT

Table 7-1
Names and Attributes

PSECT
Name

Use

Attributes

SCODE

Executable

SPDATA

Read-only data:
literals, read-only
FORMAT

SLOCAL

code.

PIC,CON,REL,LCL,SHR,NOEXE,RD,NOWRT,LONG

statements

Read/write data
to

PIC,CON,REL,LCL,SHR,EXE,RD,NOWRT,
LONG

the program

local

PIC,CON,REL,LCL,NOSHR,NOEXE ,RD ,WRT,LONG*

unit:

user local symbols,
compiler temporary
symbols, argument
lists, and descriptors

$BLANK

Blank

COMMON

block.

PIC,OVR,REL,GBL,SHR,NOEXE,RD,WRT
,LONG

name (s)

Named

COMMON

block(s).

PIC,OVR,REL,GBL,SHR,NOEXE,RD,WRT
,LONG

*
Program section SLOCAL is aligned
on
quadword
contains any double precision or complex data.

Table 7-2 describes the meanings
VAX-11 Linker Reference Manual.

the

boundaries

attributes.

Refer

also

the

When the VAX/VMS linker constructs an executable image, it divides the
executable
image
into
sections.
Each image section contains PSECTs
that have the same attributes.
By arranging image sections
according
to
PSECT attributes, the linker is able to control memory allocation.
The linker allows you to allocate memory to your own specification, by
means

commands

linker.

The

Manual.

)

7.2

you

options

include

file

options

is described

file

that

the VAX-1l1l

input

Linker

the

Reference

STORAGE ALLOCATION AND FIXED-POINT DATA TYPES

The default storage unit for VAX-11l FORTRAN IV-PLUS
is
the
(four
bytes).
A storage unit is the amount.
of memory.- needed
a real, logical, or
integer
value.
Double
precision
and
values
are
stored
1in
two successive storage units.
These
sizes must be taken into
account
when
you
associate
two

longword
to store
complex
relative
or
nmore

variables
through
association.

argument

an EQUIVALENCE or

COMMON

statement,

You can, however, declare integer
and
logical
variables
as
2-byte
values
to save space or to be compatible with PDP-11 FORTRAN.
Either
specify the /NOI4 qualifier in
the
FORTRAN
command,
or
explicitly
declare

variable

INTEGER*2

take advantage of VAX-1l's
32-bit data efficiently.

ability

LOGICAL*2
to

type.

manipulate

This

both

allows

16-bit

you

data

to
and

FORTRAN SYSTEM ENVIRONMENT

Table 7-2
PSECT Attributes

Attribute

Meaning

PIC/NOPIC

Position-independent or position~dependent.

CON/OVR

Concatenated or

REL/ABS

Relocatable or

GBL/LCL

Global

SHR/NOSHR

Shareable or

EXE/NOEXE

Executable or non-executable

RD/NORD

Readable or

WRT/NOWRT

Writeable or non-writeable

LONG/QUAD

Longword or quadword alignment

7.2.1

overlaid
absolute

local

scope

non-shareable

non-readable

1Integer Data Types Supported

VAX-11 FORTRAN IV-PLUS

supports INTEGER*2

and

INTEGER*4

data

types,

which
occupy
two and four bytes of storage respectively.
Both types
can be mixed
in
computations;
such
mixed
type
computations
are
carried out to 32 bits of significance, and produce INTEGER*4 results.
If you do not override the default storage allocation,
are allocated for integer values.

then four bytes

7.2.1.1
Relationship of INTEGER*2
and
INTEGER*4
Values - INTEGER*2
values
are
stored
as
two's
complement
signed binary numbers, and
occupy two bytes of storage.
INTEGER*4
values
are
also
stored
as
two's
complement
signed
binary
numbers,
but
occupy four bytes of
storage.
The lower addressed word of an INTEGER*4 value contains
the
low-order

part of

the value.

INTEGER*2 values are a
subset
of
INTEGER*4
values.
That
1is,
an
INTEGER*4
value
1in
the
range
-32768 to 32767 can be treated as an
INTEGER*2 value.
Conversion
from
INTEGER*4
to
INTEGER*2
(without
checks
for
overflow)
consists
of simply ignoring the high-order 16
bits of the INTEGER*4 value.
This
type
of
conversion
provides
an
important
FORTRAN
usage,
as
1illustrated
in the following example.
Given:

CALL SUB(2)

should the argument (2)
be
treated
as
an
INTEGER*2
value
or
as
INTEGER*4?
By providing an INTEGER*4 constant as the actual argument,
SUB executes correctly even if its dummy argument is typed INTEGER*2.

FORTRAN

SYSTEM ENVIRONMENT

7.2.1.2
1Integer Constant
Typing - Integer
constants
are
generally
typed
according
to the magnitude of the constant.
In most contexts,
INTEGER*2 and INTEGER*4 variables and integer constants can be
freely
mixed.
You
are
responsible,
however,
for
ensuring
that integer
overflow conditions do not occur, as happens in the following example.
INTEGER*2
INTEGER*4
I
32767
J
I+ 3

I
J

In this example, I and 3 are INTEGER*2 values, and an INTEGER*2 result
will
be
computed.
The
16-bit addition, however, will overflow the
valid INTEGER*2 range and be treated as -32766.
This
value
will
be
converted

detected

and

INTEGER*4

reported

type

and

the

assigned

program

default /CHECK=OVERFLOW qualifier

unit

The

was

specified.

overflow will

compiled

with

the
preceding
example
with
the
following
uivalent program, which produces different results.
qu

Contrast

the

apparently

PARAMETER I
INTEGER*4 J

32767

J=1IH++3

this

the

case

the compiler

parameter

performs

constant

32767,

The compiler recognizes this
assigned the value 32770.

7.2.1.3

1Integer-Valued

the

addition of

producing

INTEGER*4

Processor-Defined

the

constant

value.

constant
result

Thus,

Functions

and

32770.

will

- A

number

the

integer-valued

processor-defined
functions
provided
by
FORTRAN
(see Section 7.3)
produce integer results from real or double precision arguments;
for
example,
INT.
In
order
to
support
such
functions
in
a manner
compatible with both INTEGER*2 and INTEGER*4 modes,
two
versions
of
these
integer-valued
processor-defined
functions are supplied.
The
compiler chooses the version that matches the compiler
/I4
qualifier
setting (/I4 or /NOI4).
This is similar to generic function selection
(described in Section 7.3.1), except that the selection
is
based
on
the mode of the compiler, rather than on the argument data type.
In

some

cases,

you

may

need

use

the

version

processor-defined
function
that
1is
the
opposite
of
the compiler
qualifier
setting.
For
this
reason,
a
pair
of
additional
processor-defined
function
names
are
provided
for
each
standard

integer-valued processor-defined function.
The names of the INTEGER*2
versions
are prefixed with I, and the names of the INTEGER*4 versions
with J (for example, IIABS and JIABS).
See Appendix B of
the
VAX-11
FORTRAN

IV-PLUS

processor—-defined

7.2.2
The

Byte

FORTRAN

Language

(LOGICAL*1l)
IV-PLUS

In general, when
the

Manual

for

complete

list

Data Type

byte

data

processing
capabilities
signed integer data type,
Hollerith data.

operation,

Reference

functions.

data

type

lets

you

take

advantage

of VAX-1l.
Actually, BYTE
and is useful for storing

different

lower-ranked

type

1is

types

are

converted

used

the

byte

or LOGICAL*1l is a
and
manipulating

binary

the higher-ranked

FORTRAN SYSTEM ENVIRONMENT

discussed in the VAX-11l
type prior to computation. (Data type rank) isHowever
, in the case of a
Manual.
ce
Referen
e
FORTRAN IV-PLUS Languag
representable as a
byte variable and an integer constant in the range
t is treated as a byte
byte variable (-128 to 127), the integer constantype.
constant;

7.3

and the result is also of byte data

FUNCTIONS SUPPLIED WITH VAX-11l FORTRAN

processor—defined
of
VAX-11 FORTRAN IV-PLUS includes a number
N
These are listed in Appendix B of the VAX-11 FORTRA
functions.
thms
algori
the
of
ptions
descri
For
Manual.
IV-PLUS Language Reference

used

to generate processor-defined

functions,

Common Run-Time Procedure Library Reference Manual.

refer to the VAX-1ll

es provided by VAX-1l1
Processor-defined functions include routin
atical functions (such
mathem
d
ly-use
common
m
perfor
to
FORTRAN IV-PLUS
The
and TIME).
DATE
as
as COS and EXP) and utility services (such
of
use
the
by
cted
unaffe
is
on
functi
fined
sor-de
data type of a proces
of these

the IMPLICIT statement to change the default data type. Some
that is, you can
processor-defined functions are generic functions;
common name.
refer to a set of similar routines by one

7.3.1

Generic Functions

FORTRAN IV-PLUS are generic
Many of the functions supplied with VAX-11
them by a common name, to
to
refer
you
functions, which means that
, there are three
example
For
n.
functio
same
the
much
perform
, all of which are
cosines
ate
calcul
processor-defined functions that
names are COS,
their
COS;
name
c
generi
the
by
to
d
referre
(or can be)
precision,
single
return
they
They differ in that
DCOS, and CCOS.
request
you
If
ively.
respect
values,
x
comple
or
double precision,
to it
refer
only
need
that the cosine function be invoked, youthe appropriate routin
e
selects
er
compil
generically, as COS. The
single
is
nt
argume
the
If
.
specify
you
nts
argume
the
on
based
if it is double precision, DCOS is
precision, COS is selected;
selected;

and if complex, CCOS is selected.

to a particular routine
Note, however, that you can explicitly refer precis
ion cosine function,
double
Thus, to obtain the
if you wish.
the generic name.
you could specify DCOS, rather than using

The compiler provides the
selects

names of processor-defined

functions

by listing them by their internal names in the "FUNCTIONS AND

SUBROUTINES REFERENCED" section of the listing.

7.3.2

Use of the EXTERNAL Statement

in conjunction with
The EXTERNAL statement has special applicability
ram by means of
subprog
a
to
refer
you
If
ns.
functio
processor-defined
assumes that
er
compil
the
names,
one of the processor-defined function
if one
Thus,
used.
be
n
functio
fined
sor—de
proces
you intend that the
ed
-suppli
system—
a
as
name
of your own subprograms has the same
from
name
e's
routin
your
guish
distin
must
you
ine,
function or subrout
the processor-defined function name.

To do so, specify the name in an EXTERNAL statement as follows:
EXTERNAL *name

FORTRAN

The

compiler

is prefixed

interprets

SYSTEM

any name

by an asterisk,

ENVIRONMENT

listed

the name

The EXTERNAL statement is described
Language Reference Manual.

7.4

ITERATION COUNT MODEL

VAX-11

with

FORTRAN

the

IV-PLUS

following

The control variable
precision variable.

The
or

double

The

wvalue,

can be

of
and

parameter

7.4.1

the

size,

and

the

1loop
at

not

loop.

VAX-11l

practices
FORTRAN systems

FORTRAN

IV-PLUS.

Assigning

the

1loop

cause

associated
may

not
For

final

the

executed

(the

the

negative

7.4.2

1Iteration

Given

the
DO

(where

following

label,

ml,

computed

count=

This

Count

m2,

zero

the

successive

times

the

real,

iteration

during

number

changing

double

control

the

values

loop
of

is
the

statement.

Interchange
the

the

use

intended

statements

effects

when

used

the control variable within the
greater than the final value, in

termination

long loop that is
within the loop.

integer,

the

loop.

Similarly modifying either a step size
variable
variable to either modify the loop behavio
r
the loop.

Using

IV-PLUS

statement,

real,

value

with

have

the

value

that

subprogram.

initialization

example:

value to
that
is

early

statement,

FORTRAN

integer,

the

Thus,

used

form

re-evaluated

affected

variables

common

VAX-11

that produces

result,

Cautions Concerning Program

other

EXTERNAL

user-supplied

the

extended

expression

1is

not

determined

statement,

executions

can

times

executed

step

any

precision

number

count)

Some

provides

features:

initial

FOR DO LOOPS

variable

step

terminated

size

cause

conditional

or
or

on
with

body
order

of
to

a
final
terminate

arbitrarily

control

transfer

Computation

sample DO

statement

V=ml,m2,m3

and m3

are

any

follows:

expressions),

the

iteration

count

executed

MAX(1,INT(((m2-ml)/m3)+1))

computation:
e

Provides

least

that

once

the

body

the

loop

always

FORTRAN SYSTEM ENVIRONMENT

Permits the step size (m3) to be negative or positive, but not

zero

for
value
count
Gives a well-defined and predictable
allowed
the
of
ion
combinat
any
from
g
resultin
ons
expressi
ff
result types. Note, however, that the effects of round-othe
cause
may
tion
computa
point
g
floatin
any
in
t
error inheren
count to be greater or less than desired when real or double

precision values are used.

compute the iteration
Under certain conditions it is not necessary to paramet
ers are of type
the
of
all
if
,
example
For
tly.
count explici
the loop, then
in
ed
modifi
not
are
integer and if the parameter values
of iterations
number
the
s
control
code
ed
generat
IV-PLUS
the FORTRAN
the final
with
ly
direct
le
variab
of the loop by comparing the control
value.

7.5

ENTRY STATEMENT ARGUMENTS

ed in Chapter
The association of actual and dummy arguments is describ
Manual. In
ce
Referen
Language
IV-PLUS
FORTRAN
VAX-11]
the
6 of
However, the

general, that description suffices for most cases. ents varies from
implementation of argument association in ENTRY statem
the way this is done on some other FORTRAN systems.

5), VAX-11l FORTRAN
As described previously in this manual (Chapterriptor
methods to pass
y-desc
call-b
and
rence
y-refe
call-b
uses the
er arguments,
charact
and
arguments to called procedures (for numeric
the
use
s
ntation
impleme
FORTRAN
other
Some
respectively).
call-by-value/result method. This distinction becomes crucial when
reference is made to dummy arguments in ENTRY statements.

arguments in
While standard FORTRAN allows you to use the same dummy only
to those
refer
to
you
permits
it
nts,
stateme
different ENTRY
dummy arguments that are defined for the ENTRY point being called.
For

example:

SUBROUTINE SUBI1 (X,Y,Z)

ENTRY ENTI1(X,A)

ENTRY ENT2(B,Z,Y)

Given this, you can make the following references:
CALL
SUB1
ENT1

ENT2

Valid References
X
X

Y
A

FORTRAN

implementations

SYSTEM ENVIRONMENT

that

use

the

call-by-value/result

however,
permit
you to refer to dummy arguments that
in the ENTRY statement being called.
For example:
SUBROUTINE
ENTRY

method,

not defined

INIT(A,B,C)

RETURN

are

CALC(Y,X)

(A*X+B)/C

END
You

can

use

this

non-standard

device

call-by-value/result

implementations, because a separate internal variabl
e is allocated
each dummy argument in the called procedure.
When
the
procedure
called,
each
scalar
actual
argument
value
is
assigned
to
corresponding

whenever

internal

there

variable,

and

called

procedure.

return

from

arguments

the

are

copied

back

variables.

When an entry point is
with
the
values
of
referenced

advised
on

VAX-11l

not

other

and

systems

VAX-11l

attempt

FORTRAN.

VAX-11l,

Such

use

calls

to do

will

that

the

variables

dummy

are

then

argument

procedure,

corresponding

used

within

modified
actual

the
dummy

argument

referenced, all its dummy arguments are defined
the
corresponding actual arguments, and may be

subsequent

these

reference

for
is
the

the

this

references
produce

the

subprogram.

in programs
will

programs

that

However,

you

are

executed

are

not

have

the

intended

that

are

not

transportable

call-by-reference

(descriptor)

effect

creates

associations between
dummy
and
actual
arguments
the
address
of
each actual argument, or descriptor, to
called procedure.
Each reference to a
dummy
argument
generates

passing

indirect

address reference

control

returns

actual

and

from

dummy

the

through

called

arguments

the

actual

procedure,

ends.

The

argument

the

dummy

address.

association

arguments

For

SUBROUTINE
A = Al
B = Bl

the
an

When

between

not

their values, and therefore cannot be referenced on subseque
nt
Thus,
to
perform
the
sort
of non-standard references shown
previous example, the subprogram must copy the
values
of
the
arguments.

method.

example:

retain

calls.
in the
dummy

INIT(Al,Bl1,Cl)

)

=2C1l

RETURN

ENTRY

CALC(Y,X)

(A*X+B)
/C

END

This

will

work

call-by-value/result

avoided, because
depends
on
the
they will retain

7.6
The

VAX-11,
method.

and

However,

systems
this

it is also non-standard.
The
storage for A, B, and C being
their values from one call to

that

method

use

should

also

the
be

success of this example
statically allocated so
the next.

FLOATING POINT DATA REPRESENTATION
following

internally.

sections describe

how

floating point data

represented

FORTRAN SYSTEM ENVIRONMENT

7.6.1

8Single Precision Floating Point Data

four
A single precision floating point value is represented by 31.
through
0
right
the
from
d
numbere
are
bits
The
bytes.
ous
contigu
0

15 14

FRACTION

EXPONENT

FRACTION

A,
A single precision floating point value is specified by itstheaddress
value is

the address of the byte containing bit 0. The form of excess 128
sign magnitude with bit 15 the sign bit, bits 14:7 an
fraction
binary exponent, and bits 6:0 and 31:16 a normalizednot24-bit
represented.
with the redundant most significant fraction bit cance
go from 16
Within the fraction, bits of increasing signifi
through 31 and 0 through 6.

An
The 8-bit exponent field encodes the values 0 through s 255.
the
that
indicate
0,
of
bit
sign
a
with
exponent value of 0 together
of 1 through
floating point value has a value of 0. Exponent values
exponent value
An
+127.
through
-127
of
255 indicate binary exponents
operand.
reserved
a
as
taken
is
1,
of
bit
sign
a
with
together
of 0,
take a
operand
reserved
a
ng
Floating point instructions processi
reserved operand fault (see Section 7.6.3.1).

range of
The value of a floating point value is in the approxiamate
floating point
The precision of
.29*10**-38 through 1.7*10**38.
value is approximately one part in 2*%*23, i.e., typically 7 decimal
digits.

7.6.2

Double Precision Floating Point Data

by eight
A double precision floating point value is represented
63.
through
0
right
the
from
d
contiguous bytes. The bits are numbere
7

15 14

EXPONENT

FRACTION

FORTRAN

SYSTEM

ENVIRONMENT

A double

precision floating point value is specified by its
address A,
address
of
the
byte
containing
bit
0.
The form of a double
precision floating point value is
identical
to
a
single
precision
floating
point
value
except
for
an additional 32 low significance
fraction bits.
Within the fraction, bits of
increasing
significance
go 48 through 63, 32 through 47, 16 through 31,
and 0 through 6.

the

The

exponent

for

double

conventions and approximate
precision
floating
point

range

values

are
single

the
same
precision
floating point values.
The precision of a double
precision
floating
point
value
is
approximately
one part in 2*%*55, i.e., typically 16
decimal digits.

7.6.3

Floating

Certain
not

FORTRAN

permitted

values

Point Data Characteristics
programming
under

the

the expected behavior
described below.

when

practices
rules

for

attempted

that are commonly
wused,
standard FORTRAN, may not

VAX-11

FORTRAN.

though
produce

These

are

7.6.3.1
Reserved
Operand
Faults - Floating
point
variables
that
contain invalid floating point values (-0.0), indicate
d by an exponent
field of 0 and a sign bit of 1, cause a reserved operand
fault in
the

VAX-11

hardware.

terminates.
There

are
®

three ways

The

VAX-1l1l

result

the

and

create

hardware

overflow

error

reported,

and,

reserved

operand

stores

floating

reserved

point

zero

default,

invalid.

For

program

values:
operand

arithmetic

value

exceptions,

the

floating

divide.

The mathematical function library returns a
value if the function is called incorrectly

your

example,

reserved
or if the

operand
argument

SQRT (-1.0)

These return values can be
handler (see Chapter 6).
®

Integer
operand

modified

and logical operations can create
reserved
patterns
in floating point variables and arrays
with integers.
Associations of this kind can occur

EQUIVALENCE

COMMON,

32768

X+1.0

case

other

two

can
FORTRAN

cases

argument

association.

For

(X,I)

Adding 1.0 to X will cause
the integer value 32768 is
as a floating point value.
first

condition

bit

example:

last

arithmetic

associated
through
EQUIVALENCE,

The

by providing

a
a

occur

reserved
reserved

when invalid
inadvertently in a
systems.

occur

7-10

operand
operand

fault,
because
when interpreted

programs

data

are

program,

used.

and

The

not

detected

FORTRAN SYSTEM ENVIRONMENT

7.6.3.2 Representation of 0.0 - The VAX-1l hardware defines 0.0 as
any bit pattern that has an exponent field and sign bit of 0,
regardless of the value of the fraction. When a bit pattern that is
defined as 0.0 is used in a floating point operation, the VAX-1l1
hardware sets the fraction field to 0. One possible effect is that
non-zero integers that are equivalenced to floating point values may
be

interpreted as 0.

Logical operations
following example:

EQUIVALENCE
I = 64
IF (X .EQ.

can

have

similar

effect,

shown

the

(X,I)

GO TO 10

The branch will always be taken because the bit pattern that
represents the integer value is equivalent to 0 when interpreted as a
floating point value.

7.6.3.3 Sign Bit Tests - The bit used as the sign bit of a floating
point value is not the same bit as the sign bit of an equivalenced
INTEGER*4 value. Consequently, you must test the sign of a value by
testing the correct data type.
EQUIVALENCE
I = 40000

.GT.

For example:

(X,I)

GO TO 10

The branch will not be taken, because the bit pattern that

the

integer

represents

value 40000 is negative (bit 15 is set) when interpreted

as a floating point value.

7-11

CHAPTER 8

PROGRAMMING CONSIDERATIONS

This chapter discusses methods you can use to write
programs.

Topics covered are:

Creating efficient source programs

Compiler optimization

I/O system characteristics

source

programs

CREATING EFFICIENT SOURCE PROGRAMS

8.1

You can reduce the time and memory required for your

efficient

taking

advantage

features

language, as described below.

8.1.1

included

the FORTRAN IV-PLUS

PARAMETER Statement

The PARAMETER statement allows you to assign
constant.

For example:

symbolic

name

PARAMETER PI=3.1415927

to represent the constant
(PI)
You can then use the symbolic name
(3.1415927) anywhere a constant is valid (such as in expressions).
Using the PARAMETER statement promotes more efficient object code by
allowing constants to be used instead of variables, while permitting
easy program modification. Constants can generally be compiled into
(See Section 8.2.2.) Therefore, parametric
more efficient code.
variables should be defined by means of PARAMETER statements, rather
than by means of DATA or assignment statements. The source code
illustrated in Example A will produce more efficient code than either
Example B or Example C.
Example A

PARAMETER M = 50, N = 100

DIMENSION X (M),
DO 5, I =1, M
po 5, J=1, N

X(I)

= X(I)*Y(J)

Y(N)

+ X(M)*Y(N)

PROGRAMMING CONSIDERATIONS

Example

DIMENSION

DATA M,

X (50),

=1,

X(I)

Example

Y(100)

N/50,100/

N
X(I)*Y(J)

X(M)*Y(N)

DIMENSION

=50

X (50),

Y(100)

;

100

DO
DO

5, I = 1, M
5, J = 1, N
X(I) = X(I)*Y(J)
(

8.1.2
The

INCLUDE

program.

Statement

statement
This

programs.

code,

such

can

For

units.

create

INCLUDE

allows

source

program

+ X(M)*Y(N)

to
to

example,

COMMON

Rather

separate

statement

used
you

file

avoid

there

may

that

consists

unit

into

your

redundant

1lines

that

appear

each

program

several

specification,
repeat this code in

program

file

duplicating

block

than

each

incorporate

that

the

code;

requires

then

the

source
code

source

several
unit, you

specify

code.

Example

In this example, a COMMON block specification is required in a
program
as
well
as
in a subroutine called by the program.
The COMMON block
specification
is
put
into
a
file
(COMMON.FOR),
and
an
INCLUDE
statement is used in both the program and the subroutine, to
reference
the code.
The file, COMMON.FOR, consists of the following text:
PARAMETER M

COMMON

Main

INCLUDE

'COMMON.FOR'

CALL

CUBE

DO 5
Z(I)

I
=

INCLUDE

Z (M)

=1,M

X(I)

SUBROUTINE

100

Y (M)

Program

DIMENSION

X (M),

SQRT(Y(I))

CUBE

'COMMON.FOR'

DO 10 I = 1,M
X(I) = Y(I)**3
RETURN

END

The

file COMMON.FOR defines the size of the COMMON block
of
the
arrays
X, Y, and Z.
Any changes to the COMMON
reflected automatically after recompilation.

and the sizes
block will be

PROGRAMMING CONSIDERATIONS

Allocating Variables in Common Blocks

8.1.3

in
so
When you allocate variables in a common block, you should do
One
memory.
in
boundaries
natural
on
aligned
are
they
that
way
a
such
in order,
simple method to accomplish this is to allocate variables
First allocate INTEGER*4, LOGICAL*4, REAL,
to data type.
according
then INTEGER*2, LOGICAL*2 variables;
and DOUBLE PRECISION variables;
and finally BYTE and CHARACTER variables.

8.1.4

Conditional Branching

You will
generally produce
more
efficient
programs
if
vyou
use
IF...THEN...ELSE
statements
to
control program flow than if you use
IF...

8.2

GO TO

statements.

COMPILER OPTIMIZATIONS

Optimization refers to techniques used to produce the greatest

amount

The aim is to
of processing with the least amount of time and memory.
Optimum
create programs that are efficient in speed and size.
and
programs,
written
and
designed
carefully
from
results
efficiency
from compilation techniques that take advantage of the machine
You can produce optimum source programs by being aware
architecture.
of, and using, certain features provided by the VAX-11 FORTRAN IV-PLUS
the VAX-11 FORTRAN IV-PLUS compiler produces efficient code
language;
by deriving maximum benefit from the VAX-1ll hardware.

The primary goal of VAX-11 FORTRAN IV-PLUS optimizations is to produce
an object program that executes faster than an unoptimized version of

the same source program.

the

A secondary goal is to reduce

size

the object program.

The language elements you use in the source program directly affect
Therefore, you
the compiler's ability to optimize the object program.
in which you can assist compiler
the ways
should be aware of
the
performs
compiler
IV-PLUS
FORTRAN
The VAX-11
optimization.
following optimizations:

Evaluation at
compile
time
of
integer,
real,
and
double
precision
constant
expressions
involving
addition,
subtraction, multiplication, or division
(this
technique
is
called

"constant folding").

Compile-time constant conversion.

Compile-time evaluation of constant subscript

the
If two subprogram references have
Argument-list merging.
is
1list
argument
the
single copy of
a
arguments,
same

expressions

array element references.

generated.

Branch instruction optimizations for arithmetic,
block

1logical

and

statements.

Elimination of unreachable code.
An optional warning
message
unreachable statements in the source
indicate
to
issued
is
program.

PROGRAMMING CONSIDERATIONS

Recognition

Removal

Local register
are
retained

memory

and

replacement of

invariant

common

computations

assignment.
(if possible)
references needed.

from DO

loops.

Frequently-referenced
in registers to reduce

Assignment of frequently-used
registers across DO loops.

Assignment of
references to

Constant pooling.
Storage is allocated for only one
constant
in
the
compiled
program.
Constants

base
COMMON

variables

registers,
blocks.

immediate-mode
operands
are
not
includes most numeric constants.

and

provide

shorter

allocated

Fast calling sequences
for
the
real
and
double
versions of some processor-defined functions.

Reordering the evaluation
number of temporary values

Delaying unary minus and .NOT.
negation/complement operations.

Partial evaluation of Boolean expressions.
in
the
following expression has the value
evaluated:

expansion

Optimization of

Peephole

for

control

optimization

examining
code
operations that
operations.

Characteristics

some

processor-defined

of
expressions,
required.
operations

address

copy

of a
used
as
storage.
This

In-line

code

variables
the number

expressions

IF (el.AND.e2)

8.2.1

subexpressions.

functions.
precision

minimize

eliminate

the

wunary

For example, if el
.FALSE., e2 is not

transfers.
of

instruction

sequences;

i.e.,

on an instruction-by-instruction basis to find
can be replaced by shorter, faster
equivalent

of Optimized

Programs

Optimized programs produce results and
run-time
diagnostic
messages
identical
to those produced by an equivalent unoptimized program.
An
optimized

program may produce fewer run-time diagnostics, however,
the diagnostics may occur at different source program statements.

example:

Unoptimized

Program

Optimized

A = X/Y
B = X/Y
DO 10, I =1,10
C(I) = C(I)*(X/Y)

Program

t = X/Y
A=t
B

Do 10, 1 = 1,10
C(I) = C(I)*t

and
For

PROGRAMMING CONSIDERATIONS

In the example above, if Y has the value 0.0, the unoptimized program
produces 12 zero-divide errors at run time; the optimized program
(Note that t is a temporary variable created by the
produces 1.

compiler.) Eliminating redundant calculations and moving invariant
calculations out of loops can affect detection of such arithmetic
errors, and should be kept in mind when you include error-detecting
routines

in your program.

Compile-Time Operations on Constants

8.2.2

The compiler performs the following computations on expressions
involving constants (including PARAMETER constants) at compile time.

Negation of Constants:

Constants

signs are negated at compile time.

For

pfeceded

wunary

minus

example:
X = -10.0

is compiled as a single move instruction.

Type Conversion of

converted

compile time.

For

the

Constants:

data

Lower

ranked

constants

are

type of the higher ranked operand at

example:

X = 10*Y

is compiled

X = 10,0*Y

Arithmetic on Integer, Real, and Double Precision Constants:
Expressions involving +, -, *, or / operators are evaluated at
compile time.
For

example:

PARAMETER NN = 27
I = 2*NN+J
is compiled as
I

54+4J

Array subscript calculations involving
compile time wherever possible.

constants

example:

For

DIMENSION I(10,10)
I(1,2) = I(4,5)

is compiled as a single move instruction.

are

simplified

PROGRAMMING CONSIDERATIONS

8.2.3

Source Program Blocks

Some optimizations are performed within the confines of a single block
of
the source program, where a block consists of a sequence of one or

FORTRAN

source

defined

transfer

from

statements.

labeled
another

The

statement

(GO

start

that

1is

TO,

block

the

target

arithmetic

1IF,

generally

or
by
an
ENTRY statement.
The following kinds of statement
however, do NOT generally define the start of a new block:
®

unreferenced

a label terminating a DO loop,
label occur in DO statements

labels
of
FORMAT
statements;
labeled,
but
control
cannot

statement

labels

with

single

reference

example:
IF

(I.EQ.0)

labels,

labels
if

the

only

references

FORMAT
be

statement
®

control

ERR=option);

that

statements
transferred
to

precedes

the

must
a

be
FORMAT

1label.

labels

for

which

IF (A)
X =1

the

only

reference

logical

statement.

immediately

For

example:

10,20,20

block can contain one or more DO loops,
as
1long
as
none
of
labels
within
the
1loops defines the start of a new block.
Thus
following examples are considered single blocks and are
optimized

Example

For

CONTINUE

preceding arithmetic

complete

the

units:

the
the
as

B*C

10, 1=1,N
A(I) = A(I)/(B*C)

po 20,
Y(J) =

Example

10
20

J=1,N

Y(J)+B*C

20,

1=1,N

20,

J=1,N

SUM = 0.0
DO 10, K=1,N
SUM = SUM+A(I,K)*B(K,J)
C(I,J) = SUM

A more thoroughly optimized object program is produced if
the
number
of
separate
blocks is minimized.
Common subexpression, code motion,
and register allocation
optimizations
are
performed
within
single
blocks.

PROGRAMMING CONSIDERATIONS

8.2.4

Eliminating Common Subexpressions

The same

subexpression often appears

a program.
B*C+E*F

A+G-B*C

within

For

1in

than

one

computation

example:

((B*C)-H)10,20,30

In this code sequence, the
common
subexpression
B*C
appears
three
times.
If
the
values
of the operands of this subexpression do not
change between computations, its value can be computed
once
and
the
result
can be used in place of the subexpression.
Thus, the sequence
shown above is compiled as follows:
t
A

=
=

B*C
t+E*F

A+G-t

((t)-H)10,20,30

As you can see,

two computations of B*C have been eliminated.

Of course, you could have
optimized
the
source
program
itself
to
preclude
the
redundant
calculation
of
B*C.
The following example
shows
a
more
significant
application
of
this
kind
of
compiler
optimization, in which you could not reasonably modify the source code
to achieve the same effect.
Without optimization, the statements
DIMENSION A(25,25),
A(I,J) = B(I1,J)

can be

B(25,25)

compiled as

tl = J*25+1
t2 = J*25+1
A(tl)=B(t2)

Variables tl and t2 represent equivalent
expressions.
The
compiler
eliminates this redundancy by producing the following optimization:
t

= J*25+1

A(t)=B(t)

PROGRAMMING CONSIDERATIONS

8.2.5

Removing Invariant Computations

Execution speed is enhanced
loops.
For example, if the
DO

F =

10,

from Loops

by taking invariant
computations
out
of
compiler detected the following sequence

=1,100

3.0%Q*A(I)+F

it would recognize
each time the loop

that the subexpression 3.0*Q
has
the
is executed.
Thus, it would change the

same
value
sequence to

t = 3.0%Q
DO 10, I = 1,100
F = t*A(I)+F

This moves the calculation
multiply operations.

3.0*Q

out of

8.2.6

Compiler Optimization Example

Figure

8-1

illustrates many of

the

optimization

1loop,

and

techniques

saves

used

the

VAX-11 FORTRAN IV-PLUS compiler.
The
complete VAX-11l FORTRAN subroutine, a

first part
relaxation

in
engineering
function used to

subroutine
1is
a 2-dimensional
a variable at coordinates
on
a

surface;

for

applications.
This
obtain the values of

example,

temperatures

distributed

(Figure 8-la) shows a
function
often
used

across

a metal

plate.

The second part (Figure 8-1b) shows the VAX-11l machine code
generated
by
the
FORTRAN
compiler.
Several
compiler
optimizations
are
illustrated, as noted by circled numbers next to
the
generated
code
lines.
These are described in the notes that follow the figure.
0001

SUBROUTINE

0002
0003
0004

PARAMETER M=40, N=60
DIMENSION X (0:M,0:N)
COMMON X

0005

LOGICAL

DONE

.TRUE.

0006
0007
0008
0009
0010
0011

Do
DO

RELAX2 (EPS)

10 J = 1,N-1
10 I = 1,M-1
XNEW = ( X(I-1,J)+X(I+1,J)+X(I,J3-1)+X(1,J+1) ) /
IF ( ABS (XNEW-X(I,J))
.GT. EPS ) DONE = ,FALSE.
X(I,J) = XNEW

0012

(.NOT.

0013
0014

RETURN
END

Figure

DONE)

8-la

RELAX

Source

Program

PROGRAMMING

0000

{

.TITLE

RELAX2

.IDENT

.PSECT

S$BLANK

.PSECT

$CODE

0000

0000
0000
0000

RELAX2::

0002
0009

.WORD

"M<1V,R5,R6 ,R7,R8,R9,R10,R11>

MOVAL

S$SLOCAL,

R1l

0009

MNEGL

#1,

000C

MOVL

#1, R7

MOVAL

O00F
00le

L$IANE:

0016
0019
001D

MOVL
MULL3
LSIAGG:

001D

ADDL3

0021

ADDF3
ADDF2
ADDF2

0029
002F
0035

S$BLANK, R5
#1,

0006

;

0007

;

0008

;

0009

;

0010

;

0011

;

0012

#41,

R7,

R6‘,

R9,

R6, R10
X+4(R5) [R10],

X-4(R5) [R10],
X-164(R5) [R10], RO
X+164
(R5) [R10], RO

MULF3

;
DONE (R11)

#"X3F80, RO, R8 @

003D
0042

SUBF3

0047
004B

CMPF

BLEQ

LSIAPI

004D

CLRL

DONE (R11)

004F

CONSIDERATIONS

X(R5) [R10],

BICW2

#°X8000,

RO,

R8,

(7}

Q@EPS (AP)

LS$SIAPI:

004F

MOVL

R8,

X(R5)[R10]

0053

AOBLEQ

#39,

0057
005B

AOBLEQ

#59,

MOVL

R7,

005F
0063

MOVL

R8,

LSIAGG
LSIANE
J(R11l)
XNEW(R1l1l)

MOVL

R9,

I(R1ll)

0067
006A

BLBC

DONE (R11),

RY9,

R7,

RET

.END

Figure

Notes

for

Figure

8-1b

RELAX Machine Code

8-1

assignment

for

assignment

for

assignment

for

base

Invariant computation (J*41)
assigned to a register
Common

(Optimized)

subexpression

assignment;

uses

removed

evaluation

are made

the

for

from

and

value

blank

the

COMMON

inner

local

loop

and

PROGRAMMING

Notes

for

Figure 8-1

(Cont.)

Peephole optimization;
multiply by 0.25.
Inline

ABS

CONSIDERATIONS

divide

FORTRAN

4.0

1is

replaced

function

DO loop control using the Add One
Equal (AOBLEQ) instruction

8.3

and

Branch

Less

Than

programs
FORTRAN

by
I/0

I/0 SYSTEM CHARACTERISTICS

You can often reduce
use
of
the
making

the execution time of your
FORTRAN
following
facts
relevant
to
the

subsystem.

Unformatted I/0 is substantially faster and more accurate than
formatted
1I/0.
The
unformatted data representation usually
occupies less file storage space as
well.
Thus
unformatted
I/0
should
be
used
for
storing
intermediate
results
on
secondary

storage.

Specifying an array name
using
an
equivalent

in an I/O
implied

list is more efficient
DO
1list.
A
single

transmission call passes an entire array,

list

can pass

only a

single

while an implied

array element per

than
1I/0

I/0 call.

The
implementation
of
the
BACKSPACE
statement
involves
repositioning
the
file
and
scanning
previously
processed
records.
If a reread
capability
is
required,
it
1is
more
efficient to read the record into a temporary array and DECODE
the array several times than to read and backspace the record.

To obtain minimum I/0 processing, the record length
access
sequential
organization
files should be a
multiple of the device block
size
of
512
bytes
bytes,
64 bytes, etc.).
For relative organization
adds one overhead byte for
fixed
1length
records
overhead bytes

for

variable length records.

of
direct
divisor or
(e.g.,
32
files, RMS
and
three

If the approximate size of the
file
1is
known,
it
1is
more
efficient to allocate disk space when a file is opened than to
incrementally extend the file as records are written.

The use of run-time formats should be minimized.
The compiler
preprocesses
FORMAT
statements
into
an
efficient internal
form.
Run-time formats must be converted into
this
internal
form
at
run
time.
In
many
cases,
using variable format
expressions will allow the format to vary at run time.
The BLOCKSIZE keyword can be used
in
an
OPEN
statement
to
obtain large amounts of data with each physical I/0 operation,
thereby
greatly
improving
the
processing
of
sequentially
accessed
files.
For example, specifying a BLOCKSIZE of 4096
bytes results in the transfer of 8 disk blocks
for
each
1I/0
operation.

PROGRAMMING CONSIDERATIONS

The OPEN and

files.

and

devices

provide

statements

explicit

control

over

1I/0

Using these statements in the proper manner can

help create efficient source programs, as illustrated in the following
examples.

OPEN

(UNIT=1, TYPE='NEW',

INITIALSIZE=200)

is
file
This statement allocates space for a file when the
opened, which is more efficient than extending the size of the
file dynamically.

OPEN

(UNIT=3, TYPE='NEW', BLOCKSIZE=8192)

1I/0
for
factor
This statement specifies a large blocking
If the file is on magnetic tape, the physical tape
transfers.
disk
16
if the file is on disk,
blocks are 8192 bytes long;
transferred by each I1/0 operation, thus enhancing
are
blocks
though requiring more memory.

I/0 performance,

OPEN

(UNIT=J,

(IERR)

TYPE='NEW',

CLOSE (UNIT=J,

DISP='DELETE')

DISP='SAVE')

(UNIT=J,

...)

A file is created.
However, if an error occurs that makes the
file invalid or useless, it is deleted.
e

OPEN

(UNIT=2,

FORM='FORMATTED',

CARRIAGECONTROL='LIST')

This statement creates a file with implicit carriage
control.
The
first
character
of each record is NOT used for carriage
control, thus it can contain actual data.
®

CHARACTER*64

TYPE

FORMAT ('SINPUT FILE?')

100

101

102

FILNAM

100

ACCEPT 101 ,FILNAM
FORMAT (A)

OPEN

(UNIT=3, NAME=FILNAM,

TYPE

102,

FORMAT
GO TO

TYPE='OLD',

ERR=9)

FILNAM

ERROR OPENING FILE

',A)

This program reads a file

specification

into

the

character

1is then qpened for processing.
file
The
variable FILNAM.
This permits you to specify the input file name at run time.

APPENDIX

FORTRAN DATA REPRESENTATION

A.1

INTEGER*2 FORMAT

15 14
BINARY NUMBER

S(sign)

= 0(+),

1(-)

INTEGER*2
Integers are stored in a two's complement representation.
values 1lie in the range -32768 to +32767, and are stored in two
For example:
contiguous bytes aligned on an arbitrary byte boundary.
+22

-7

A.2

0016 (hex)
FFF9 (hex)

INTEGER*4 FORMAT

31 30
BINARY NUMBER

S(sign)=0(+),

1(-)

INTEGER*4 values are
INTEGER*4 values 1lie
value is stored in four
Note that
boundary.
i.e., -32768 to
value;
as an INTEGER*2 value.

A.3

stored in two's complement representation.
in the range -2147483648 to 2147483647. The
contiguous bytes, aligned on an arbitrary byte
if the value is in the range of an INTEGER*2
+32767, then the first word can be referenced

FLOATING-POINT FORMATS

The exponent for both floating-point formats is stored in excess 128
Binary exponents from -128 to +127 are represented by the
notation.
Fractions are represented 1in
binary equivalents of 0 through 255.
sign-magnitude notation with the binary radix point to the left.
the most
therefore,
Numbers are assumed to be normalized and,
significant bit is not stored, because it is redundant (this is called
"hidden bit normalization").

FORTRAN DATA REPRESENTATION

This
to

bit

assumed

2**-128)

represented

example,

in
by

+1.0

which

exponent

would

case

unlegs

the

exponent

assumed

field

sign

represented

and

hexadecimal

(corresponding
The

bit

value
of

0.0

is
For

by:

4080
0

the

4-byte

format,

or:

forsmat.

The decimal

4080
0
0
0

the

8-byte

number

-5.0

is:

COAOQ
V]

the

4-byte

format,

or:

COAOQ
0
0
0

the

A.3.1

8-byte

Real

format.

Format

(4-Byte Floating

Point)

A single precision real number is four contiguous
arbitrary
byte
boundary.
Bits are labeled from
31.
15 14

EXPONENT

bytes starting on an
the right, 0 through

FRACTION

S(sign)=0(+),
The

form of
the sign

and

31:16

1(-)

a single precision real number is sign magnitude, with
bit, bits 14:7 an excess 128 binary exponent, and bits
normalized

24-bit

fraction

with

the

redundant

bit

6:0
most

significant
fraction
bit
not
represented.
The
value of a single
precision real
number
is
1in
the
approximate
range:
.29%1(0**-38
through 1.7*10**38.
The precision is approximately one part in 2**23,

i.e.,

typically

decimal

digits.

FORTRAN DATA REPRESENTATION

A.3.2

Double Precision Format

(8-Byte Floating Point)

A double precision real number is eight contiguous bytes, starting
on
an
arbitrary
byte
boundary.
Bits
are
1labeled
from the right, 0
through

63.

15 14

FRACTION

EXPONENT

FRACTION

S(sign)=0(+), 1(-)

The form of a double precision real number is identical
to
a
single
precision real
number
except
for an additional 32 low significance
of
range
approximate
and
The exponent conventions,
fraction bits.
is
The precision
single precision.
as
same
the
are
values
approximately one part in 2**55, i.e., typically 16 decimal digits.

A.3.3

Complex

Format

A complex number is eight contiguous bytes, aligned
on
an
arbitrary
byte
boundary.
The
1low-order four bytes contain a single precision
real number that represents the real part of the complex number.
The
high-order
four
bytes
contain
a
single precision real number that
represents the imaginary part of the complex number.

15 14

S | BINARY EXCESS 128 EXP

0
FRACTION
REAL PART

FRACTION

S |BINARY EXCESS 128 EXP

FRACTION
IMAGINARY PART

FRACTION
63

FORTRAN DATA REPRESENTATION

A.4

LOGICAL*1

FORMAT

DATA ITEM

The range
format.

A.5

numbers

from +127

-128

can be

represented

in LOGICAL*1l

CHARACTER FORMAT
i

Character data is stored as one character
if necessary, to fill the data item.

A.6

per

byte,

padded

with

blanks

CHAR 2

CHAR 1

CHAR 4

CHAR 3

BLANK =40

CHAR N (N<255)

HOLLERITH FORMAT

CHAR 2

CHAR 1

CHAR 4

CHAR 3

BLANK =40

CHAR N (N<255)

Hollerith constants are stored
internally
one
character
Hollerith
values
are
padded
on
the
right with blanks
associated data item if necessary.

per
byte.
to fill the

FORTRAN DATA REPRESENTATION

A.7

LOGICAL FORMAT

Logical values are stored in two or four contiguous bytes, starting on
an
arbitary
byte boundary.
The low-order bit (bit 0) determines the
value.
If bit 0 is set, the value is .TRUE..
If bit 0 is clear,
the
value is .FALSE..
LOGICAL*2

TRUE:

UNDEFINED BITS

FALSE:

UNDEFINED BITS

LOGICAL*4

TRUE:

UNDEFINED

BITS
31

FALSE:

UNDEFINED

BITS

APPENDIX

DIAGNOSTIC MESSAGES

B.1

DIAGNOSTIC MESSAGES

Diagnostic messages
come

from

compiler
unmatched

OVERVIEW

a VAX-1l

compiler,

the

1linker,

detects

syntax

errors

the

illegal
characters,
parameters.
The Run-Time

DIAGNOSTIC MESSAGES

FROM THE

1IV-PLUS

program

can

Run-Time

Library.

The

source

parentheses,

missing or illegal
that
occur
during execution.
(Linker
VAX-11l Linker Reference Manual.)

B.2

FORTRAN

the

program,

misspelled

such

keywords,

Library
reports
messages are summarized

and
errors

the

COMPILER

The diagnostic messages issued by the compiler describe the error that
has
been
detected,
and
in
some cases contain an indication of the
action taken by the compiler in response to the error.
Besides reporting
errors
detected
in
source
program
compiler
will
issue
messages
1indicating
errors
that
compiler itself, such as I/0 errors, stack overflow, etc.

B.2.1

syntax,
involve

the
the

Source Program Diagnostic Messages

Three classes of source
severity of the error.
F

Fatal;

must

program errors are recognized,
based
on
the
These are (from greatest to least severity):
before

the

compiled.
No object file is
detected during compilation.

produced

Error;

not

fatal

corrected

but

should

program

corrected.

F-class

can

error

object

is
produced
in
spite of the presence of an E-class
but the program will probably not execute properly.
W

Warning;

issued

non-standard

for

syntax,

statements
and

for

that

use

statements

file

error,

acceptable,

but

corrected

the

compiler.
An object file is produced but you
should
check
the
statements
to
which
a
W-class
diagnostic
message
applies, to make sure the program will produce
the
correct
result.
Note that W-class messages are produced unless the
/NOWARNINGS qualifier is set in the FORTRAN command.

DIAGNOSTIC

Typing mistakes are a likely cause of
compiler
to
generate
misleading
especially of the following:

Missing comma or parenthesis
FORMAT

MESSAGES

syntax errors, and can cause the
diagnostic
messages.
Beware

in a complicated

expression

statement.

Misspelled variable names.
The compiler may not detect
error, so execution can be affected.

Inadvertent
1line
continuation
mark.
This
diagnostic message for the preceding line.

Confusion between
letter
O,
which
identical to you,

can

the
digit
=zero
(0)
and
the
can
result
in variable names
but not to the compiler.

cause

this
a

wupper-case
that appear

Another source of diagnostic messages
1is
the
inclusion
of
invalid
ASCII
characters
in
the
source program.
With the exception of the
tab, space,
and
form-feed
characters,
non-printing
ASCII
control
characters
are
not valid in a FORTRAN source program.
As the source
program is scanned, such invalid characters are replaced by a question
mark
(?).
However, because ?
cannot occur in a FORTRAN statement, a
syntax error

usually

results.

Thus, because the message indicates
only
the
immediate
should always check the entire source statement carefully.

cause,

you

Figure B-1 shows the form of source
program
diagnostic
messages
as
they
are displayed at your terminal, in interactive mode.
Figure B-2
shows how these messages appear in listings.

$FORT-W-ERROR 83, Extra comma in format list
[FORMAT (I3,)] in module MORTGAGE at line 9
$FORT-F-ERROR 69, Undefined statement label
[ 66] in module MORTGAGE at line 14
$FORT-I-ERRSUM,

Figure B-1

2 Errors MORT.FOR.10

Sample Diagnostic Messages

(Terminal Format)

" DIAGNOSTIC MESSAGES
PROGRAM TO CALCULATE MONTHLY MORTGAGE PAYMENTS

PROGRAM MORTGAGE

0001
0002

0003
0004
0005
0006

0007
0008
0009

TYPE

10
20

(' ENTER AMOUNT OF MORTGAGE

ACCEPT 20,IPV
FORMAT (I6)
TYPE

FORMAT

FORMAT (' ENTER LENGTH OF MORTGAGE IN MONTHS ')

ACCEPT 40,IMON
FORMAT (I3,)

$FORT-W-ERROR 83, Extra comma in format list
[FORMAT (I3,)] in module MORTGAGE at line 9
0010
0011
0012
0013
0014
0015
0016
0017
0018
0019
0020
0021

0022

0023

TYPE

50
60
65

FORMAT (' ENTER ANNUAL INTEREST RATE

ACCEPT 60,YINT
FORMAT (F6.4)
GO TO 66

YI = YI
FIMON =
FIMON =
FMNTHLY
TYPE

IGET MONTHLY RATE

YI = YINT/12
IMON =
FIPV =

0024

STOP

0025

END

-IMON
IPV*YI

+1
YI ** IMON
1-FIMON
= FIPV/FIMON

70,FMNTHLY

FORMAT

)

(' MONTHLY PAYMENT EQUALS

', F7.3 )

$FORT-F~ERROR 69, Undefined statement label
[ 66]

Figure B-2

in module MORTGAGE at line 14

Sample Diagnostic Messages

(Listing Format)

DIAGNOSTIC MESSAGES

Table
Source

B-1

Program Diagnostic

Messages

Message

Number

Text/Meaning

Statement
A

too

statement

subdivided
Compiler

complex

too

into

complex

two

or more

expression

stack

compiled.

complex
or
a subprogram
arguments
can
be

expression,

there
are
reference.

many

named

COMMON

Reduce

the

number

Source

line

too

source

line

Too
Up
by

many

COMMON

compilation

than
compiler

line.

continuation

blocks.

terminated

88 characters was encountered.
examines
only
the
first
72

lines,

remainder

ignored

to 99 continuation lines are permitted, as determined
the /CONTINUATIONS=n qualifier (default, 19).

Syntax

The

error

file

invalid

name

failure

The
an

specified
incorrect

unmounted

INCLUDE
Reduce

number

file
lines.

INCLUDE

string

qualifier,

Open

the

file

not

specification

acceptable

undefined

INCLUDE

device,

(invalid

file

nested

protection

too

violation.

deeply

level

of
INCLUDE
nesting
or
continuation
1lines permitted.
requires
space
equivalent
to
two

statement

improperly

letters)

has

initial

line.

syntax,

etc.).

file could not be opened, possibly due
to
file
specification,
non-existent
file,

volume,

files

Invalid
An

named

long,

You
can
number of

blocks

longer

NOTE:

The
characters on

too
many
A maximum

compiled.
reduce the

arguments.
Too

must

overflow

An expression is too
actual
arguments
in

of
60
actual
subdivide
a complex

statements.

label

continuation

ignored

formed

statement

been

detected

The

increase
the
Each INCLUDE

statement

label
(e.g.,
in
columns
1

label

containing
5
of an

ignored.

(continued

page)

DIAGNOSTIC MESSAGES

Table B-1 (Cont.)
Source Program Diagnostic Messages

Message

Number

Text/Meaning

Redundant continuation mark

ignored

A continuation mark was detected where
is required.
The continuation mark is
11

Extra characters
Superfluous
correct

following

text was

statement.

Overflow
while
expression

found
Check

valid

the

for

Missing

exponent

after

A floating point
notation, but the
14

Missing
A

of
or

syntactically

syntax

too

must

constant

1large

too

in character

constant

errors.

constant
was
specified
exponent was omitted.

apostrophe

character

end

constant

The specified value of a constant
small to be represented.

1line

statement

typing

converting

an
initial
ignored.

constant

enclosed

apostrophes.

Zero-length string

The
length
hexadecimal,

specified
octal
or

for
a
character,
Radix-50
constant

Hollerith,
must not be

zero.

String

constant

A character

characters.

truncated
Hollerith

Radix-50

to maximum length
constant

can

contain

constant can contain up to

255

characters.
17

Count

Hollerith

Radix50

constant

too

large,

reduced

The
is

value specified by
greater
than
the

the

source

the integer preceding the H or
number of characters remaining

R
in

statement.

(continued

next page)

DIAGNOSTIC MESSAGES

Table B-1 (Cont.)
Source Program Diagnostic Messages

Message

Number

Text/Meaning

Invalid

character

used

constant

An invalid character was detected in a constant.
characters

Hexadecimal:
Octal:
Radix-50:

For

Radix-50,

character.
constant 1s
20

space

For
set to

a 8,

name

for

the

octal,

subprogram

space

1invalid

the

name

entire

was

not

constant

required

Missing

period,

and

A required
found.

variable

substituted

hexadecimal
zero.

subprogram name

A
22

0 - 9, A - F,
0 - 7
A-12%2,0-9,

Missing variable or

Missing

Valid

are:

constant was

variable

An expression, or a
omitted.
Examples:

not

found.

constant
term

expression,

has

been

WRITE ( )
DIST = *TIME
23

Missing

operator

delimiter

symbol

Two terms of an
expression
are
not
separated
by
an
operator,
or
a
punctuation mark (such as a comma) has
been omitted.
Examples:
CIRCUM =
IF
(I
24

Missing
A

statement label

required

Missing

3.14
DIAM
10,20,30

statement

label

Non-integer

as
TO,
10 TO I.

expression where

An expression
data

omitted.

keyword

A required keyword, such
statement such as ASSIGN
26

reference was

that must be of

was

integer

type

omitted

value

INTEGER

from

required

was

another

type.

(continued on next page)

‘

DIAGNOSTIC MESSAGES

Table B-1 (Cont.)
Source Program Diagnostic Messages

Message

Number

Text/Meaning

Non-logical expression where logical value required
An expression
data

type LOGICAL

was

Arithmetic expression where character value
An expression
data

that must be

another

type.

that must

required

type CHARACTER was

another

type.

Character

expression where arithmetic value

required

An expression that must be
arithmetic
(INTEGER,
REAL,
LOGICAL,
DOUBLE
PRECISION,
or
COMPLEX)
was
of type
CHARACTER.
30

Variable name,

constant,

or expression invalid

this

context

A quantity has been used incorrectly;
for example,
the
name
of
a
subprogram
was
used
where
an arithmetic
expression is required.
31

Operation not permissible on

these data

An invalid operation was specified,
two real variables.
32

Arguments
supplied

incompatible

with

such

function,

types

.AND.

assumed

user

A function reference was made, using a processor-defined
function
name,
but the argument list does not agree in
order,
number,
or
type
with
the
processor-defined
function
requirements.
The
function is assumed to be
supplied by you as an EXTERNAL function.
33

Subscripted

reference to non-array variable

A variable that is not defined
with subscripts.
34

Substring reference used

as an array cannot appear

invalid context

A substring reference has been detected to a variable or
array that is not of type CHARACTER.
Example:
REAL X (10)
Y

Number of
More
or
declared

X(J:K)

subscripts does not match array declaration
fewer
dimensions
for the array.

are

referenced

than

were

(continued on next page)

DIAGNOSTIC

MESSAGES

’

Table B-1 (Cont.)
Program Diagnostic Messages

Source

Message

Number

Text/Meaning

More
An

than

array

dimensions
can

specified,

defined

remainder
having

ignored

seven

dimensions.
37

Non-constant

subscript where

Subscript and
EQUIVALENCE

Adjustable

substring

statements

array

constant

expressions
must

bounds

must

required

used

DATA

and

constants.

dummy

arquments

COMMON

Variables

must

specified

either

in dimension declarator

expressions

subprogram dummy arguments

appear

COMMON.

Adjustable
A

reference

where
40

such

Passed
A

array

was made

length

reference

array

1in

Lower
bound
declaration
The
to

upper
or

Character

greater

bound

greater

than

substring

adjustable

not

name
a

array

used

passed

invalid

length

such

the specified

outside

the

than

upper

lower

limits

context

character

reference

out

array

bounds,

bound

must

not

bounds

dimensions,

specified

dimension declarator

the

context

allowed.

expression value

beyond

character
substring
been referenced.
42

context

context where

substring

element

invalid

character

was made

Subscript or
An

reference

or
variable
allowed.
41

used

has

array

equal

bound.

out of

The first character position of a
is greater than the last character

order
substring
position.

expression
Example:

C(5:3)
(continued

next page)

DIAGNOSTIC

MESSAGES

Table B-1 (Cont.)
Program Diagnostic Messages

Source

Message

Text/Meaning

Number

Invalid

use of

A CALL

statement

used
as
included
that was

FUNCTION

name

referred

in CALL

statement

subprogram

name

that

was

a character function;
or a CALL statement that
alternate return specifiers referred to a
name
defined as a data type other than INTEGER*4.

Examples:

IMPLICIT CHARACTER*10
(C)
CSCAL = CFUNC (X)
CALL CFUNC (X)

REAL*4

CALL
45

$VAL,

TCB

TCB(Y,&100)

%REF,

$%$DESCR

used

invalid

context

The
argument
1list
built-in
functions
$DESCR)
cannot be used outside an actual
Example:
X

Invalid

argument

%VAL,

$REF,

$DESCR,

the

$VAL

(3.5D0)

argument
cannot
be
character, or complex.

$LOC

(X+Y)

argument

F Alternate
An

$REF,
list.

%REF(Y)

The argument specified for one
is not valid.
For example:

(%VAL,
argument

return

alternate

label

return

must

used

not

argument

$%LOC

built-in

functions

double

precision,

invalid
was

expression.

context

used

function

characters.

reference.

W Name

A
51

longer

symbolic

F Multiple
A

name

than

name

has

characters

been

declaration of

appears

two

truncated

name
more

inconsistent

declaration

statements.

E Multiple
used

declaration

A
variable
declaration
used.

appears
1in
statement.

data

type

more
The

for

than
first

variable,

first

one
data
type
type declaration is

(continued

next page)

DIAGNOSTIC

MESSAGES

Table B-1 (Cont.)
Program Diagnostic Messages

Source

Message

Number

Text/Meaning

Syntax error

in IMPLICIT

Improper syntax
was
Refer
to
the VAX-11]
Manual for the syntax
54

Letter mentioned

statement

used
in
an
IMPLICIT
statement.
FORTRAN IV-PLUS Language Reference
rules.

twice

IMPLICIT

statement,

A letter has been given an implicit data
once.
The last data type given is used.
‘55

Incorrect

length modifier

type more

used
than

in declaration

An unacceptable length has been specified in
declaration
(see
"Type
Declaration"
in

a data type
the
VAX-1l1l

FORTRAN

Example:

IV-PLUS

INTEGER

last

Left side
element

Language

Reference

Manual).

PIPES*8

assignment

must

variable

array

The symbolic name to which the value of an expression is
assigned must be a variable, array element, or character
substring reference.
57

Length
specified
declaration
The

length
character

must

match

character

function

specifications

for
all
ENTRY
names
subprogram
must
be
the

function

in
a
same.

Example:

CHARACTER*15
CHARACTER*20
ENTRY

FUNCTION
G

Inconsistent

function data

types

All entry names in a function subprogram must
character or numeric.
Example:
CHARACTER*15
REAL
ENTRY

FUNCTION

either

X
X

Undimensioned

array

function definition out of order

Either a statement function definition
has
been
found
among
executable statements, or an assignment statement
has
been
detected
1involving
an
array
for
which
dimensions have not been given.
(continued

page)

DIAGNOSTIC MESSAGES

Table B-1

(Cont.)

Source Program Diagnostic Messages
Message

Text/Meaning

Number

Statement out of order, statement ignored

The order of statements was not as specified in
VAX-11 FORTRAN IV-PLUS Language Reference Manual.
*

statement found out of order is 1ignored.

Statement not valid

this

unit,

program

the
The

statement

ignored

;
A program unit contains a statement that is not allowedan
ns
contai
gram
subpro
DATA
example, a BLOCK
for
executable statement.

Statement cannot appear in logical IF statement
A logical IF statement must not contain a DO statement
or another logical IF, IF THEN, ELSEIF, ELSE, ENDIF, or
END statement.

No path to this statement

Program

control

cannot

statement is deleted.

reach

statement.

this

The

Example:

I=1I+1
GO TO 10
STOP

Missing END statement, END is assumed

An END statement was missing at

the

input file, and has been inserted.

end

the

Statement cannot be labeled, label ignored

A label was placed on a statement that does
labels.

The label is ignored.

not

last

permit

Inconsistent usage of statement label

Labels of executable statements have been confused

with

labels of FORMAT statements.

Example:

10
68

GO TO 10

FORMAT

(I5)

Multiple definition of a statement label, second ignored
The
The same label appears on more than one statement.
first occurrence of the label is used.

(continued on next page)

Source

DIAGNOSTIC

MESSAGES

Table

(Cont.)

B-1

Program Diagnostic

Messages

Message

Number

Text/Meaning

Undefined

statement

label

Reference has been made to a
defined in the program unit.

and

loops and
levels.

statements

block

nested

statement

too

label

that

not

deeply

statements

cannot

nested

beyond

71
One

A statement label specified
been used previously.

the

following

conditions

was

found:

statement

has

1loop

Example:

I=I+1
J=J+1
DO

A DO loop
block.

K=1,10

contains

incomplete

Examples:

1=1,10

J=J+1
DO 20
10

The

start

IF,

ELSEIF

K=1,10

J=J+K
CONTINUE

the

incomplete

ELSE

block

statement.

can

block

page)

1=1,10

J=J+1I

(J.GT.20)

THEN

J=J-1
ELSE

J=J+1
CONTINUE
ENDIF

(continued

DIAGNOSTIC MESSAGES
(Cont.)

Table B-1

Source Program Diagnostic Messages
Message

Text/Meaning

Number

F Invalid control structure using ELSE IF, ELSE, or ENDIF

The order

incorrect.

ELSE,

ELSEIF,

statements

ENDIF

1is

ELSEIF, ELSE, and ENDIF statements cannot stand alone.
ELSEIF and ELSE must be preceded by either a block IF

statement
preceded

statement.

or
by

statement.
IF,

an ELSEIF
either a

block

ENDIF

ELSEIF,

be
must
or ELSE

Examples:

po 10

1=1,10

J=J+1

ELSEIF

Error:

(J.LE.K)

THEN

ELSEIF preceded by a DO statement.
IF

(J.LT.K)

THEN

J=1I+J
ELSE

J=I-J

ELSEIF

(J.EQ.K)

THEN

ENDIF

ELSEIF preceded by an ELSE statement.

Error:
73

Unclosed DO loop or IF block

The terminal statement of a DO loop
statement of an IF block was not found.

the

ENDIF

Example:

Do 20

1=1,10

X=Y
END

Statement cannot terminate a DO loop

The terminal statement of a DO loop cannot be a GO TO,
arithmetic IF, RETURN, block IF, ELSE, ELSEIF, ENDIF, DO
or

END

statement.

ENTRY within DO loop or IF block, statement ignored

An ENTRY statement is not allowed within the range of
DO loop or

IF block.

Assignment to DO variable within loop

The control variable of

loop

within the range of the DO statement.

has

been

altered

(continued on next page)

DIAGNOSTIC MESSAGES

‘Table B~1 (Cont.)
Program Diagnostic Messages

Source

Message
Number

Text/Meaning

F Alternate

return

subprogram

specifier

invalid

FUNCTION

The

argument list of a FUNCTION
declaration
contains
an
asterisk,
or
a RETURN statement in a function subprogram
specifies an alternate return.

Examples:

INTEGER FUNCTION TCB (ARG, * ,X)

FUNCTION

RETURN

IMAX

I+J

END

E Alternate

return

omitted

in -SUBROUTINE

ENTRY

statement

asterisk
is
missing
in
the
argument
1list
of
subroutine for which an alternate return is specified.

For

example:

SUBROUTINE

RETURN

XYZ (A,B)

Or:
ENTRY

MIN(Q,R)

RET&RN I+4
80

Format

groups

nested

too

Format

groups

cannot

Unbalanced

The

of
82

parentheses

deeply
nested

beyond

format

separator

between

format

A required separator character
fields in a FORMAT statement.

levels.

list

number of right parentheses does
left parentheses.

E Missing

eight

not match

the

number

items

has

been

omitted

(continued

between

on next page)

DIAGNOSTIC MESSAGES
Table B-1

(Cont.)

Source Program Diagnostic Messages

Message

Text/Meaning

Number

Extra comma in format list
FORMAT

_Example:
84

(I14,)

Constant in format item out of range

value in a FORMAT statement exceeds the
Refer to the VAX-11 FORTRAN IV-PLUS
range.

A numeric
allowable

Language Reference Manual.

Format item contains meaningless character

An invalid character or a syntax error was detected in a
FORMAT

statement.

Format item cannot be signed

A signed constant is valid only with the P format code.
87

Missing number in format list
FORMAT

Example:
88

Extra number in format list
FORMAT

Example:
89

(F6.)

(I4,3)

Invalid I/0 specification for this type of I/O statement

I/0

The format specifier in an I/O statement is invalid.

A syntax error was

found

the

portion

statement that precedes the I/O list.

Examples:

TYPE (6),J
WRITE 100,J
90

Format specifier

in error

must be one of the following:
e

label of a FORMAT statement

e
91

(list directed)

a run-time format

specifier:

variable,

element, or character substring reference.

array

END= or ERR= specification given twice, first used

Two instances of either END= or ERR= were found.
Control 1is transferred to the location specified in the
first

occurrence.

(continued on next page)

Source

DIAGNOSTIC

MESSAGES

Table

(Cont.)

B-1

Program Diagnostic

Messages

Message

Number

Text/Meaning

Syntax

error

Improper

Missing
An

I/0

I/O

list

syntax

was

detected

I/0

not present where

list

element

for

required.

input

statement

An input statement I/O list contains
such as an expression or a constant.
95

Duplicated
Each
can

keyword

in OPEN/CLOSE

keyword subparameter in
be specified only once.

UNIT=

keyword

OPEN

missing
CLOSE

Incorrect
A

keyword

statement

in an OPEN
100

Number

The

number
match

contains

exceeds

constants

the

number

must

number

Number

o0f

values

exceeds

that

variables

be initialized.
The remaining
elements are not initialized.
101

statement

include

keyword

specified
of

statement

must

element,

the

UNIT=

valid

only

statement

statement.
names

invalid

statement

OPEN/CLOSE

statement

an OPEN

subparameter.
97

list.

list

list was

Invalid

values

a
or

DATA
array

variables

number

DATA

statement
elements

and/or

names

statement

array

DATA

The
number
of
variables
or
array
elements
to
be
initialized must match the number of constants specified
in a DATA statement.
The remaining constant values
are
ignored.
102

Invalid

repeat

count

DATA

The repeat count in a DATA
nonzero integer constant.
103

Constant

size

A constant in
corresponding

exceeds

statement,

statement is not
It is ignored.

variable

a
DATA
variable.

count

size

statement

in DATA
1is

unsigned,

statement

larger

(continued

ignored

than

its

page)

DIAGNOSTIC

MESSAGES

Table B-1 (Cont.)
Source Program Diagnostic Messages
Message

Number

104

Text/Meaning

Character

name

incorrectly

1initialized

with

numeric

value
Character
data
with
initialized
with
a

a
1length
greater
than
1
was
numeric value in a DATA statement.

Example:

CHARACTER*4 A
DATA
A/l4/

105

F Program storage

requirements

exceed

addressable memory

The storage space allocated to the variables and
arrays
of
the program unit exceeds the addressing range of the
machine.
106

F Variable

inconsistently equivalenced

EQUIVALENCE
relationships

itself

statements
specify
between
variables
or

inconsistent
array
elements.

Example:

EQUIVALENCE

107

Invalid

equivalence

Variables
other.
108

(A(1l) ,A(2))

EQUIVALENCE

COMMON

two

variables

cannot

statement

in COMMON

equivalenced

incorrectly expands

A COMMON block cannot be extended beyond
by an EQUIVALENCE statement.
109

W Mixed

numeric

A
COMMON
character

110

W Mixed

and

block
data.

numeric

and

character
must

elements

not

contain

character

elements

Numeric and character variable
be equivalenced to each other.

111

Invalid

initialization of

An aftempt was made,
initialize

B.2.2

a variable

and

variable

that

BLOCK

not

each

COMMON

block

its

beginning

COMMON

both

numeric

and

in EQUIVALENCE

array

not

DATA

elements

cannot

in COMMON

subprogram,

common

block.

Compiler-Fatal Diagnostic Messages

Conditions can be encountered of such severity that
compilation
must
be
terminated
at
once.
These
conditions
are
caused by hardware
errors, software errors, and errors that require changing
the
source
program.
Table
B-2
1lists
the
diagnostic messages that report the
occurrence of such compiler-fatal errors.

B-17

DIAGNOSTIC MESSAGES

Table B-2
Compiler-Fatal Diagnostic Messages

Message

Number

Text/Meaning

201

Open error

on work

file

on WKO:

202

Open error

file

on WKO:

temp

FORTRAN IV-PLUS creates a temporary work file and
zero,
one,
or
two
temporary
scratch
files
during
the
compilation process.
The compiler was
unable
to
open
these
required
files.
Possibly
the
volume
was not
mounted, space was not available on
the
volume,
or
a
protection violation occurred.
203

I/0 error

on work

file

on WKO:

204

I/0

error

temp

file

on WKO:

205

I/0 error

source

file

206

I/0 error

on object

file

207

I/0 error

listing

file

I/0
errors
report
either
hardware
I/0
errors,
software error conditions such as an attempt to write
a write-protected volume.

208

Compiler

dynamic memory overflow

Reduce the
reduce

209

the

Compiler

number

INCLUDE

work

file

continuation

nesting depth.

Compiler

internal

lines

allowed

overflow

A single program unit
1is
too
Specify /WORK_FILES=3 or divide
units.
210

or
on

large
to
be
compiled.
the program into smaller

consistency check

An internal consistency check has
failed.
This
error
should
be reported to DIGITAL in a Software Performance
Report.

211

Compiler

control

stack

overflow

The compiler's control
stack
overflowed.
Simplifying
the FORTRAN source program will correct the problem.
212

Branch displacement out

range

A label referenced by a GO TO statement
1is
beyond
the
displacement
1limits
of a branch in the generated code.
Modify the source program to reduce the distance between
the
GO
TO
and its reference.
(See "Compiler Limits,"
Section B.2.3.)

DIAGNOSTIC MESSAGES

Compiler Limits

B.2.3

There are limits to the size and complexity of a single VAX-11 FORTRAN

There are also limits on the complexity of
IV-PLUS program unit.
FORTRAN statements. 1In some cases, the limits are readily described;
see Table B-3. In other cases, however, the limits are not so easily
defined.

For example, the compiler uses external work files to store the symbol
table and a compressed representation of the source program. The
the
number of work files is controlled by the /WORK_FILES qualifier:

maximum is 3, which provides space for approximately 2000 or more
If you run
lines of source code in a typical FORTRAN program unit.
out of work file space, error 209 occurs.

In some cases, the limits can be adjusted by re-linking
and

modifying

and

the

1limits

suit your needs.

the

compiler

Table B-3 shows two

values for such limits, in the form m(n), where m is the default limit
is

the maximum.

Limits for which only one value is shown are

not adjustable.

Table B-3
Compiler Limits

Limit

Language Element

DO and block-IF statement nesting

Actual arguments per CALL or function

(many)

(combined)

reference

Named COMMON blocks

Format group nesting

(250)

Labels in computed or assigned

250

Parentheses nesting in expressions

INCLUDE file nesting

Continuation lines

FORTRAN source line length

88 characters

Symbolic name length

15 characters

GOTO

list

Constants

Character, Hollerith
Radix-50

Array dimensions

(many)

255 characters
12 characters
7

DIAGNOSTIC

Error
too

212
far

occurs

apart.

The

branch

instruction

compiler

within a single program
displacements.
Thus, a

MESSAGES

and

generates

its

target

references

unit by means of
8-bit
jump must not cross more

and
than

statement

are

executable

code

16-bit
relative
32K bytes.

You will not reach this limit if a program unit
results
than
32K
bytes
of generated code.
You are more likely

work

The

file

limit

amount of

executable

address

B.3

first.

data

storage,

programs

space

are

available,

the

size

limited

only

arrays,

determined

the

and

amount

in
to

no
more
reach the

the

total

process

size

virtual

system generation.

RUN-TIME DIAGNOSTIC MESSAGES

Errors

that

reported

occur

messages

during

diagnostic

may

result

errors

detected

between

the

e xecution

messages

from:

hardware

by
RMS;
program and an

errors
internal

overflow

from

your
the

FORTRAN

Run-Time

file
system
errors;
that occur during transfer of data
record;
computations
that
cause
calls to the Run-Time Library;
and

Table

detected
by
the
Run-Time Procedure

Run-Time Library Diagnostic Message Presentation

Run-Time
terminal

B.3.2

are
These

conditions;

or underflow;
incorrect
problems
in
array
descriptions
and
conditions
operating
system.
Refer
to
the
VAX-11
Common
Library Reference Manual for more information.

B.3.1

program

Library.

Library diagnostic messages are usually sent to
(interactive mode) or the log file (batch mode).

either

your

Run-Time Library Diagnostic Messages
B-4

lists

the

Run-Time

library messages.

There is a HELP file available, which contains
these
error
messages
and descriptive text.
Thus you can print the descriptive text at your
terminal, on-line.
For example, to print the text for a
file
system
error

(such

HELP

rewind

error),

type

the

following:

ERROR FOR$_REWERR

The

condition

Table

the

VAX-11l

6-1.

symbols

corresponding

For more

Common

the

information about

Run-Time

Procedure

B-20

message

the

Library

numbers

are

run-time HELP

Reference

1listed

file,

Manual.

see

DIAGNOSTIC MESSAGES

Table B-4

Run-Time Diagnostic Messages

Messages 20 through 48 indicate errors related to the
(No message

numbers from 2 to 19

are used.)

Message

system.

Text/Meaning

Number
1

file

NOT A FORTRAN-SPECIFIC ERROR

Library
Run-Time
the
that
indicates
This message
a
by
caused
not
that was
error
an
encountered
a
was
it
is,
That
condition peculiar to FORTRAN.
condition not described by any other message in this
table.

REWIND ERROR

an
during
RMS
by
detected
An error condition was
RMSSREWIND
operation
used to position a file at its
beginning.

DUPLICATE FILE SPECIFICATIONS

file
specify
to
made
attempts were
Multiple
intervening close operation;
attributes without an
i.e., DEFINEFILE followed by

DEFINEFILE

followed

statement

OPEN

by DEFINEFILE.

INPUT RECORD TOO LONG

the
than
Your program tried to read a record longer
To read the file, use an OPEN
size.
record
maximum
appropriate
statement with a RECORDSIZE value of the
size.

BACKSPACE ERROR

One of the following has occurred:

BACKSPACE was attempted

opened

for

during
condition
error
an
detected
RMS
RMSSREWIND operation used to rewind the file

the

3.
24

file

appending

RMS detected an
forward

error

condition

to the desired record.

while

reading

the

ENDFILE

END-OF-FILE DURING READ

Either an end-file record

produced

was
condition
end-of-file
RMS
the
or
statement
END=
no
and
statement
READ
a
encountered during
transfer specification was provided.
(continued on next page)

DIAGNOSTIC MESSAGES

Table

B-4

(Cont.)

Run-Time Diagnostic Messages

Messages 20
(No message

through
numbers

48 indicate errors related
from 2 to 19 are used.)

the

file

system.

Message

Number

Text/Meaning

RECORD NUMBER OUTSIDE

RANGE

A
direct
access
READ,
WRITE,
or
FIND
statement
specified a record number outside the range specified
in a DEFINEFILE statement or in the MAXREC keyword of
the OPEN statement.
26

OPEN

OR DEFINEFILE

REQUIRED

SPECIFY DIRECT

ACCESS

A direct access READ, WRITE
or
FIND
operation
was
attempted
before
an
OPEN
statement
specifying
ACCESS='DIRECT',
or
DEFINEFILE
statement
was
performed.

THAN

ONE

RECORD

An attempt was made
single
record
in
statement,

CLOSE
An

error

attempted
FILE

to
a

I/O

read
or
write
direct
access

ENCODE

was

DECODE

NOT

detected