Microsoft Programmer's Library 1.3/Microsoft_Programmers_Library.7z/MPL/fortran/frtlref.txt

Text File | 2013-11-08 | 1.0 MB | 28,584 lines

 Microsoft FORTRAN Language Reference

 VERSION 5.0

 FOR MS(R) OS/2 AND MS-DOS(R) OPERATING SYSTEMS

 MICROSOFT CORPORATION

 Information in this document is subject to change without notice and does
 not represent a commitment on the part of Microsoft Corporation. The
 software described in this document is furnished under a license agreement
 or nondisclosure agreement. The software may be used or copied only in
 accordance with the terms of the agreement. It is against the law to copy
 the software on any medium except as specifically allowed in the license or
 nondisclosure agreement. No part of this manual may be reproduced or
 transmitted in any form or by any means, electronic or mechanical, including
 photocopying and recording, for any purpose without the express written
 permission of Microsoft.

 (c)Copyright Microsoft Corporation, 1987-1989. All rights reserved.

 Printed and bound in the United States of America.

 Microsoft, MS, MS-DOS, and CodeView are registered trademarks of Microsoft
 Corporation.

 DEC is a registered trademark of Digital Equipment Corporation.

 IBM is a registered trademark of International Business Machines
 Corporation.

 Intel is a registered trademark of Intel Corporation.

 VAX is a registered trademark of Digital Equipment Corporation.

 Document No.  LN0818-500-R00-0289
 Part No. 06383
 10  9 8 7 6 5 4 3 2 1


 Table of Contents

 Introduction

 About This Manual
 Document Conventions
 Books about FORTRAN Programming
 Requesting Assistance

 Part 1  Language Reference

 Chapter 1  Elements of FORTRAN

 1.1  Characters
 1.2  Names
       1.2.1  Global and Local Names
       1.2.2  Undeclared Names
 1.3  Data Types
       1.3.1  Integer Data Types
       1.3.2  The Single-Precision IEEE Real Data Type
       1.3.3  The Double-Precision IEEE Real Data Type
       1.3.4  Complex Data Types
       1.3.5  Logical Data Types
       1.3.6  The Character Data Type
 1.4  «Records»
 1.5  Arrays
 1.6  «Attributes»
       1.6.1   «ALLOCATABLE»
       1.6.2   «ALIAS»
       1.6.3   «C»
       1.6.4   «EXTERN»
       1.6.5   «FAR»
       1.6.6   «HUGE»
       1.6.7   «LOADDS»
       1.6.8   «NEAR»
       1.6.9   «PASCAL»
       1.6.10  «REFERENCE»
       1.6.11  «VALUE»
       1.6.12  «VARYING»
 1.7  Expressions
       1.7.1  Arithmetic Expressions
       1.7.2  Character Expressions
       1.7.3  Relational Expressions
       1.7.4  Logical Expressions
       1.7.5  «Array Expressions»
       1.7.6  Precedence of Operators

 Chapter 2  Program Structure

 2.1   Lines
 2.2   Statement Labels
 2.3   «Free-Form Source Code»
 2.4   Order of Statements «and Metacommands»
 2.5   Arguments
 2.6   Program Units
 2.7   Main Programs
 2.8   Subroutines
 2.9   Block-Data Subprograms
 2.10  Functions
        2.10.1  External Functions
        2.10.2  Statement Functions

 Chapter 3  The Input/Output System

 3.1  The FORTRAN I/O System
 3.2  I/O Statements
       3.2.1   File Names (FILE=)
       3.2.2   Units (UNIT=)
       3.2.3   File Access Method (ACCESS=)
       3.2.4   «Input/Output Buffer Size (BLOCKSIZE=)»
       3.2.5   The Edit List
       3.2.6   Error and End-of-File Handling (IOSTAT=, ERR=, END=)
       3.2.7   Format Specifier (FMT=)
       3.2.8   «Namelist Specifier (NML=)»
       3.2.9   File Structure (FORM=)
       3.2.10  Input/Output List
       3.2.11  «File Sharing (MODE=, SHARE=)»
       3.2.12  Record Number (REC=)
 3.3  Choosing File Types
 3.4  File Position
 3.5  Internal Files
 3.6  Carriage Control
 3.7  Formatted I/O
       3.7.1  Nonrepeatable Edit Descriptors
       3.7.2  Repeatable Edit Descriptors
       3.7.3  Interaction between Format and I/O List
 3.8  List-Directed I/O
       3.8.1  List-Directed Input
       3.8.2  List-Directed Output
 3.9  «Namelist-Directed I/O»

 Chapter 4  Statements

 4.1  Categories of Statements
 4.2  Statement Directory
       «ALLOCATE»
       ASSIGN (Label Assignment)
       Assignment (Computational)
       «AUTOMATIC»
       BACKSPACE
       BLOCK DATA
       CALL
       «CASE»
       CHARACTER
       CLOSE
       COMMON
       COMPLEX
       CONTINUE
       «CYCLE»
       DATA
       «DEALLOCATE»
       DIMENSION
       DO
       «DO WHILE»
       «DOUBLE COMPLEX»
       DOUBLE PRECISION
       ELSE
       ELSE IF
       END
       «END DO»
       END IF
       ENDFILE
       ENTRY
       EQUIVALENCE
       «EXIT»
       EXTERNAL
       FORMAT
       FUNCTION (External)
       GOTO (Assigned GOTO)
       GOTO (Computed GOTO)
       GOTO (Unconditional GOTO)
       IF (Arithmetic IF)
       IF (Logical IF)
       IF THEN ELSE (Block IF)
       IMPLICIT
       «INCLUDE»
       INQUIRE
       INTEGER
       «INTERFACE TO»
       INTRINSIC
       «LOCKING»
       LOGICAL
       «MAP...END MAP»
       «NAMELIST»
       OPEN
       PARAMETER
       PAUSE
       PRINT
       PROGRAM
       READ
       REAL
       «RECORD»
       RETURN
       REWIND
       SAVE
       «SELECT CASE...END SELECT»
       Statement Function
       STOP
       «STRUCTURE...END STRUCTURE»
       SUBROUTINE
       Type
       «UNION...END UNION»
       WRITE

 Chapter 5  Intrinsic Functions and Additional Procedures

 5.1  Using Intrinsic Functions
       5.1.1   Data-Type Conversion
       5.1.2   «Data-Type Information»
       5.1.3   Truncating and Rounding
       5.1.4   Absolute Value and Sign Transfer
       5.1.5   Remainders
       5.1.6   Positive Differences
       5.1.7   Maximums and Minimums
       5.1.8   Double-Precision Products
       5.1.9   Complex Functions
       5.1.10  Square Roots
       5.1.11  Exponents and Logarithms
       5.1.12  Trigonometric Functions
       5.1.13  Character Functions
       5.1.14  «End-of-File Function»
       5.1.15  «Address Functions»
       5.1.16  «Bit-Manipulation Functions»
 5.2  Alphabetical Function List
 5.3  Additional Procedures
       5.3.1  Time and Date Procedures
       5.3.2  Run-Time-Error Procedures
       5.3.3  Command-Line-Argument Procedures
       5.3.4  Random Number Procedures

 Part 2  Compiling and Linking

 «Chapter 6»  «Metacommands»

 6.1  Using Conditional-Compilation Metacommands
 6.2  Metacommand Directory
       $DEBUG and $NODEBUG
       $DECLARE and $NODECLARE
       $DEFINE and $UNDEFINE
       $DO66
       $ELSE
       $ELSEIF
       $ENDIF
       $FLOATCALLS and $NOFLOATCALLS
       $FREEFORM and $NOFREEFORM
       $IF
       $INCLUDE
       $LARGE and $NOTLARGE
       $LINESIZE
       $LIST and $NOLIST
       $LOOPOPT
       $MESSAGE
       $PACK
       $PAGE
       $PAGESIZE
       $STORAGE
       $STRICT and $NOTSTRICT
       $SUBTITLE
       $TITLE
       $TRUNCATE and $NOTRUNATE

 Chapter 7  The FL Command

 7.1  The Basics: Compiling, Linking, and Running FORTRAN Files
       7.1.1  Compiling and Linking with FL
       7.1.2  Using FL Options
       7.1.3  The FL Environment Variable
       7.1.4  Specifying the Next Compiler Pass
       7.1.5  Stopping FL
       7.1.6  Using the FL Command (Specific Examples)
       7.1.7  Running Your FORTRAN Program
 7.2  Getting Help with FL Options (/HELP)
 7.3  Floating-Point Options (/FP)
 7.4  Memory-Model Options (/A, /M)
 7.5  OS/2 Library Options (/Lp, /Lr, /Lc)
 7.6  Data Threshold Option (/Gt)
 7.7  Naming and Organizing Segments (/ND, /NM, /NT)
 7.8  Creating Bound Program Files (/Fb)
 7.9  FORTRAN-Specific Options (/4Y, /4N)
       7.9.1  Controlling Optional Language Features (/4Ys, /4Yi, /4Yv)
       7.9.2  Controlling Source-File Syntax (/4Yf, /4Nf, /4Yt, /4Nt, /4Y6, /
       7.9.3  Automatic Variables
       7.9.4  Setting the Default Integer Size (/4I2, /4I4)
       7.9.5  Conditional Compilation (/4cc, /D)
 7.10  Specifying Source Files (/Tf, /Ta)
 7.11  Compiling without Linking (/c)
 7.12  Naming the Object File (/Fo)
 7.13  Naming the Executable File (/Fe)
 7.14  Creating Listing Files (/F)
 7.15  Special File Names
 7.16  Line Size (/Sl) and Page Size (/Sp)
 7.17  Titles (/St) and Subtitles (/Ss)
 7.18  Formats for Listings
 7.19  Searching for Include Files (/I, /X)
 7.20  Handling Warnings and Errors
        7.20.1  Understanding Error Messages
        7.20.2  The Warning-Level Option (/W)
 7.21  Syntax Errors (/Zs)
 7.22  Preparing for Debugging (/Zi, /Od, /Zd)
 7.23  Using an 80186, 80188, 80286, or 80386 Processor (/G0, /G1, /G2)
 7.24  Optimizing (/O and /Zp)
 7.25  Enabling and Disabling Stack Probes (/Ge, /Gs)
 7.26  Suppressing Automatic Library Selection (/Zl)
 7.27  Setting the Stack Size (/F)
 7.28  Restricting the Length of External Names (/H)
 7.29  Labeling the Object File (/V)
 7.30  Linking with Libraries
 7.31  Creating Overlays
 7.32  Using FL to Link without Compiling
 7.33  Specifying Assembler Options (/MA)

 Appendix A  ASCII Character Codes

 Appendix B  Differences from Previous Versions

 B.1  Changes from Version 4.1 to Version 5.0
       B.1.1  Alphabetical Summary
       B.1.2  New Microsoft FORTRAN Functions and Procedures
       B.1.3  Microsoft FORTRAN Language Extensions
       B.1.4  OS/2
       B.1.5  Graphics
 B.2  Changes from Version 4.0 to Version 4.1
       B.2.1  OS/2 Support
       B.2.2  Enhanced FL Utility
       B.2.3  Extended Control Over Default Libraries (Linker Options)
 B.3  Changes from Versions 3.2 and 3.3 to Version 4.0
       B.3.1   Changes for ANSI Full-Language Standard
       B.3.2   Source Compatibility
       B.3.3   Attributes in Array Declarations
       B.3.4   Blanks in Formatted Files
       B.3.5   MODE and STATUS Options in OPEN Statement
       B.3.6   Temporary Scratch-File Names
       B.3.7   Binary Direct Files
       B.3.8   Precision of Floating-Point Operations
       B.3.9   Exponentiation Exceptions
       B.3.10  List-Directed Output
       B.3.11  DO-Loop Ranges
       B.3.12  Object Compatibility
       B.3.13  Library Compatibility
       B.3.14  Mixing Version 4.0 and Version 3.3 Modules
       B.3.15  Mixing Version 4.0 and Version 3.2 Modules
 B.4  Changes for Version 4.0
       B.4.1  Enhancements and Additions to the Compiler and Linker
       B.4.2  Run-Time Library Changes
       B.4.3  Changes to the Language
       B.4.4  New Language Features

 Appendix C  Compiler and Linker Limits

 C.1  Compiler Limits
       C.1.1  Limits on Number of Names
       C.1.2  Limits on Complicated Expressions
       C.1.3  Limits on Character Expressions
 C.2  Linker Limits
 C.3  Run-Time Limits
       C.3.1  Increasing the Maximum Number of Open Files
       C.3.2  Using the Modified Files
       C.3.3  Multithread and Dynamic Link Applications

 Appendix D  Error Messages

 D.1  Command-Line Error Messages
 D.2  Compiler Error Messages
       D.2.1  Compiler Fatal Error Messages
       D.2.2  Compilation Error Messages

       D.2.3  Recoverable Error Messages
 D.2.4  Warning Error Messages
 D.3  Run-Time Error Messages
       D.3.1  Run-Time-Library Error Messages
       D.3.2  Other Run-Time Error Messages

 Glossary


 Figures and Tables

  Figure 2.1  Order of Statements «and Metacommands»

  Table 1.1   Memory Requirements
  Table 1.2   Integers
 «Table 1.3   C String Escape Sequences»
 «Table 1.4   Objects to Which Attributes Can Refer»
  Table 1.5   Arithmetic Operators
  Table 1.6   Arithmetic Type Conversion
  Table 1.7   Relational Operators
  Table 1.8   Logical Operators
  Table 1.9   Values of Logical Expressions

  Table 3.1   I/O Statements
  Table 3.2   I/O Options
  Table 3.3   Errors and End-of-File Records When Reading
 «Table 3.4   Mode and Share Values»
  Table 3.5   Carriage-Control Characters
  Table 3.6   Nonrepeatable Edit Descriptors
  Table 3.7   Forms of Exponents for the E Edit Descriptor
  Table 3.8   Interpretation of G Edit Descriptor
  Table 3.9   Interpretation of GE Edit Descriptor
  Table 3.10  Forms of Exponents for the D Edit Descriptor

  Table 4.1   Categories of FORTRAN Statements
  Table 4.2   Specification Statements
  Table 4.3   Control Statements
  Table 4.4   I/O Statements
  Table 4.5   Repeatable Edit Descriptors

  Table 5.1   Abbreviations Used to Describe Intrinsic Functions
  Table 5.2   Intrinsic Functions: Type Conversion
 «Table 5.3   Intrinsic Functions: Data-Type Information»
  Table 5.4   Intrinsic Functions: Truncation and Rounding
  Table 5.5   Intrinsic Functions: Absolute Values and Sign Transfer
  Table 5.6   Intrinsic Functions: Remainders
  Table 5.7   Intrinsic Functions: Positive Difference
  Table 5.8   Intrinsic Functions: Maximums and Minimums
  Table 5.9   Intrinsic Functions: Double-Precision Product
  Table 5.10  Intrinsic Functions: Complex Operators
  Table 5.11  Intrinsic Functions: Square Roots
  Table 5.12  Intrinsic Functions: Exponents and Logarithms
  Table 5.13  Intrinsic Functions: Trigonometric Functions
  Table 5.14  Restrictions on Arguments and Results
  Table 5.15  Intrinsic Functions: Character Functions
 «Table 5.16  Intrinsic Functions: End-of-File Function»
 «Table 5.17  Intrinsic Functions: Addresses»
 «Table 5.18  Intrinsic Functions: Bit Manipulation»
 «Table 5.19  Bit-Manipulation Examples»
  Table 5.20  Intrinsic Functions
  Table 5.21  Time and Date Procedures

  Table 6.1   Metacommands

  Table 7.1   FL Options and Default Libraries
  Table 7.2   Segment-Naming Conventions
  Table 7.3   Default File Names and Extensions
  Table 7.4   Arguments to Listing Options

  Table B.1   Negative INTEGER or REAL Raised to a REAL Power
  Table B.2   Zero Raised to a Negative Power
  Table B.3   COMPLEX Zero Raised to a COMPLEX Power
  Table B.4   Zero Raised to the Zero Power

  Table C.1   Limits Imposed by the Microsoft FORTRAN Compiler
  Table C.2   Limits Imposed by the Microsoft Segmented-Executable Linker


 Introduction
 ───────────────────────────────────────────────────────────────────────────

 Microsoft(R) FORTRAN Version 5.0 improves on the already popular Microsoft
 FORTRAN programming language by adding several important new features. The
 language is now fully compatible with the Systems Application Architecture
 (SAA) FORTRAN extensions and many of the VAX(R) extensions. It supports new
 constructs such as compound data types (structures) and SELECT CASE
 decision-making blocks. New compiler directives allow advanced features like
 conditional compilation of specific pieces of program code. Also, the FL
 compiling and linking command has several new options and improvements to
 existing options.

 This new version of Microsoft FORTRAN offers enhanced OS/2 systems support,
 including the use of dynamic-link libraries and multiple threads of
 execution. For complete flexibility, programs can be designed to run under
 DOS, OS/2, or both operating systems. In addition, an extensive graphics
 library lets data and figures become an integral part of any FORTRAN
 application.

 This chapter introduces the Microsoft FORTRAN Reference, describes the
 document conventions used in the manual, and gives additional sources of
 information about FORTRAN.

 For discussions of memory models, calling non-FORTRAN subroutines and
 functions from a Microsoft FORTRAN program (mixed-language programming),
 programming under OS/2, and the use of graphics, see  Microsoft FORTRAN
 Advanced Topics. To find out how to use the Microsoft CodeView(R)
 Window-Oriented Debugger to debug your programs, see the Microsoft CodeView
 and Utilities User's Guide.


 About This Manual

 The Microsoft FORTRAN Reference defines the FORTRAN language as implemented
 by the Microsoft FORTRAN Optimizing Compiler, Version 5.0. It is intended as
 a reference for programmers who have experience in the FORTRAN language.
 This manual does not teach you how to program in FORTRAN; for a list of
 texts on FORTRAN, see "Books about FORTRAN Programming" at the end of this
 introduction.

 Microsoft documentation uses the term "OS/2" to refer to the OS/2
 systems──Microsoft Operating System/2 (MS(R) OS/2) and IBM(R) OS/2.
 Similarly, the term "DOS" refers to both the MS-DOS(R) and IBM Personal
 Computer DOS operating systems. The name of a specific operating system is
 used when it is necessary to note features that are unique to the system.

 Microsoft FORTRAN conforms to the American National Standard Programming
 Language FORTRAN 77, as described in the American National Standards
 Institute (ANSI) X3.9-1978 standard.

 ───────────────────────────────────────────────────────────────────────────
 «NOTE»
    «The Microsoft FORTRAN language contains many extensions to the full
    ANSI standard language. In this manual, information on all Microsoft
    extensions is indentified by chevron characters ('«' and '»')»
 ───────────────────────────────────────────────────────────────────────────

 Chapter 1 discusses the elements of the FORTRAN programming language.
 Chapter 2 explains the structure of FORTRAN programs. Chapter 3 gives the
 details of FORTRAN's input/output (I/O) system. Chapter 4 is a detailed
 description of all FORTRAN statements. Chapter 5 explains all of FORTRAN's
 intrinsic functions. «Chapter 6 covers Microsoft FORTRAN's metacommands»,
 and Chapter 7 discusses the FL command. The following list shows where to
 look for information on specific topics:

 Topic                                      Location of Information

 Characters, names, data types,             Chapter 1, "Elements of FORTRAN"
 «attributes,» and expressions in
 FORTRAN

 Formatting lines in your source            Chapter 2, "Program Structure"
 program; subroutines, functions,
 and arguments; and structuring
 your FORTRAN programs

 Input and output in FORTRAN                Chapter 3, "The Input/Output
                                            System"

 FORTRAN statements, listed                 Chapter 4, "Statements"
 alphabetically

 FORTRAN intrinsic functions,               Chapter 5, "Intrinsic Functions"
 listed alphabetically

 «Compiler directives, called               «Chapter 6, "Metacommands"»
 metacommands, listed
 alphabetically»

 FL command                                 Chapter 7, "The FL Command"

 Table of the American Standard             Appendix A, "ASCII Character
 Code for Information Interchange           Codes"
 (ASCII) character set

 Selected terms used in this                "Glossary"
 documentation


 Document Conventions

 This manual uses the following typographic conventions. (Note that, in most
 cases, blanks are not significant in FORTRAN).

 Example                              Description
 of Convention                        of Convention

 «Extensions to the ANSI standard     «Text enclosed by the chevron charcters
 language»                            indicates features that are extensions t
                                      the ANSI FORTRAN 77 full-language
                                      standard. These extensions may or
                                      may not be implemented by other
                                      compilers that conform to the
                                      full-language standard.»

 OUT.TXT, ANOVA.EXE, COPY, LINK,      Uppercase (capital) letters
 FL                                   indicate file names and DOS-level
                                      commands. Uppercase is also used
                                      for command-line options (unless
                                      the application accepts only
                                      lowercase).

 C       Comment line                 This kind of type is used for
         WRITE (*,*) 'Hello           program examples, program output,
         +World                       DOS-level commands, and error
                                      messages within the text. A
                                      capital C marks the beginning of
                                      a comment in sample programs.
                                      Continuation lines are indicated
                                      by a plus sign (+) in column 6.

 AUTOMATIC, INTRINSIC, WRITE          Bold capital letters indicate
                                      language-specific keywords with
                                      special meaning to FORTRAN.
                                      Keywords are a required part of
                                      statement syntax, unless enclosed
                                      in double brackets as explained
                                      below. In programs you write,
                                      FORTRAN keywords are entered in
                                      all-uppercase (capital) letters,
                                      or any combination of uppercase
                                      «and lowercase» letters.

 other keywords                       Bold lowercase letters are used
                                      for keywords of other languages.

                                      In the sentence, «"The value that
                                      is returned by LOCNEAR is
                                      equivalent to a near function or
                                      data pointer in Microsoft C or an
                                      ADR type in Microsoft Pascal,"»
                                      the word «LOCNEAR» is a FORTRAN
                                      keyword, and the words near and
                                      ADR are keywords of Microsoft C
                                      and Microsoft Pascal,
                                      respectively.

 Apostrophes: '''                     In Microsoft FORTRAN, an
                                      apostrophe is entered as a single
                                      right quotation mark ('), not a
                                      single left quotation mark (`).
                                      Note that in the typeface used in
                                      examples, such as 'string',
                                      apostrophes look like this: '.

 expression                           Words in italics indicate
                                      placeholders for information that
                                      you must supply. A file name is
                                      an example of this kind of
                                      information. Italics are also
                                      occasionally used in the text for
                                      emphasis.

 «optional item»                      Items inside bolded chevrons
                                      ('«', '») brackets are optional.

 {choice1|choice2}                    Braces and a vertical bar
                                      indicate a choice among two or
                                      more items. You must choose one
                                      of the items unless all of the
                                      items are also enclosed in double
                                      square brackets.

 Repeating elements...                Three dots following an item
                                      indicate that more items having
                                      the same form may be entered.

 CALL GetNum (i, *1O)                 A column of three dots indicates
 .                                    that part of the example has
 .                                    intentionally been  omitted.
 .
 SUBROUTINE GetNum (i, *)

 KEY NAMES                            Capital letters are used
                                      for the names of keys and key
                                      sequences, such as ENTER and
                                      CTRL+C.

                                      A plus (+) indicates a
                                      combination of keys. For example,
                                      CTRL+E means to hold down the
                                      CTRL key while pressing the E
                                      key.

                                      The carriage-return key is referred to
                                      as ENTER.

                                      The cursor movement ("arrow")
                                      keys on the numeric keypad are
                                      called DIRECTION keys. Individual
                                      DIRECTION keys are referred to by
                                      the direction of the arrow on the
                                      key top (LEFT, RIGHT, UP, DOWN)
                                      or the name on the key top (PGUP,
                                      PGDN).

                                      The key names used in this manual
                                      correspond to the name on the IBM
                                      Personal Computer keys. Other
                                      machines may use different names.

 "defined term"                       Quotation marks usually indicate
                                      a new term defined in the text.

 Video Graphics Array (VGA)           Acronyms are usually spelled out
                                      the first time they are used.


 Example

 The following example shows how this manual's typographic conventions are
 used to indicate the syntax of the EXTERNAL statement:

      EXTERNAL name «« [attrs] »» «, name «« [attrs] »» »...

 This syntax listing shows that when using the EXTERNAL statement, you must
 first enter the word EXTERNAL followed by a name that you specify. «Then,
 you can optionally enter a left bracket ([), followed by attributes (attrs)
 that you specify, followed by a right bracket (]).» If you want to specify
 more names, «optionally followed by attributes (attrs),» you must enter a
 comma, followed by a name, «optionally followed by a left bracket,
 attributes, and a right bracket.» Because the «, name « « [attrs] »» »
 sequence is followed by three dots (...), you can enter as many of those
 sequences (a comma, followed by a name, «optionally followed by attributes
 in brackets») as you want.


 Books about FORTRAN Programming

 Agelhoff, Roy, and Richard Mojena. Applied FORTRAN 77 featuring Structured
 Programming. Belmont, CA: Wadsworth, 1981.

 Ashcroft, J., R. H. Eldridge, R. W. Paulson, and G. A. Wilson. Programming
 with FORTRAN 77. Dobbs Ferry, NY: Sheridan House, 1981.

 Friedman, Frank, and E. Koffman. Problem Solving and Structured Programming
 in FORTRAN. 2d ed. Reading, MA: Addison-Wesley, 1981.

 Kernighan, Brian W., and P. J. Plauger. The Elements of Programming Style.
 New York, NY: McGraw-Hill, 1978.

 Ledgard, Henry F., and L. Chmura. FORTRAN with Style. Rochelle Park, NJ:
 Hayden, 1978.

 Wagener, Jerrold L. FORTRAN 77: Principles of Programming. New York, NY:
 Wiley, 1980.

 These books are listed for your convenience. Microsoft Corporation does not
 endorse these books or recommend them over others on the same subject.


 Requesting Assistance

 If you need help or feel you have discovered a problem in the software,
 please provide the following information to help us locate the problem:

   ■  The version of DOS you are running (use the DOS VER command)

   ■  Your system configuration:

        1. Make and model of your computer

        2. Total memory and total free memory at compiler execution time (use
           the DOS CHKDSK command to obtain these values)

        3. Any other information you think might be useful

   ■  The compiler command line used (or the link command line if the problem
      occurred during linking)

   ■  Any object files or libraries you linked with if the problem occurred
      at link time

 If your program is large, please try to reduce it to the smallest possible
 program that still produces the problem.

 Use the Product Assistance Request form at the back of this manual to send
 this information to Microsoft.

 If you have comments or suggestions regarding any of the manuals
 accompanying this product, please indicate them on the Document Feedback
 card at the back of this manual.

 If you are not already a registered Microsoft FORTRAN owner, you should fill
 out and return the Registration Card. This enables Microsoft to keep you
 informed of updates and other information.


 Part 1  Language Reference

 This section describes Version 5.0 of the Microsoft FORTRAN lanuage──both
 its ANSI-standard features and the special Microsoft extensions, which are
 indentified by the chevron characters '«' and '»'. The first three chapters
 introduce the language by reviewing the elements of FORTRAN (Chapter 1),
 giving an overview of the FORTRAN program structure (Chapter 2), and
 discussing the FORTRAN input/ouptput system (Chapter 3).

 The remaining two chapters provide the information you will need to put
 these features to work. Chapter 4 is an alphabetical statement reference,
 with information on syntax and usage for each Microsoft FORTRAN statement.
 Chapter 5 provides information on intrinsic functions and additional
 procedures supplied with this package.


 Chapter 1  Elements of FORTRAN
 ───────────────────────────────────────────────────────────────────────────

 This chapter explains the building blocks of FORTRAN programs: special
 characters, the scope and naming of FORTRAN identifiers, the data types that
 are available, and the rules that govern their use. The basic arithmetic and
 logical functions available are also covered.


 1.1  Characters

 FORTRAN source files can contain any printable characters in the ASCII
 character set. When character constants or character variables are logically
 compared, the collating sequence for the FORTRAN character set is the ASCII
 sequence. The ASCII character set, listed in Appendix A, "ASCII Character
 Codes," includes the following:

   ■  The 52 uppercase and lowercase alphabetic characters (A through Z and a
      through z).

      «The Microsoft FORTRAN Compiler interprets lowercase letters as
      uppercase letters in all contexts except character constants and
      Hollerith fields.» In character constants and Hollerith fields, case is
      significant. For example, «the statements WRITE (*, *) and write (*, *)
      are identical, but» the character constants 'ijk' and 'IjK' are
      different.

      «There is one exception to case sensitivity in character constants.
      Character constants that are a part of the FORTRAN statements listed in
      Chapter 4, "Statements," are not case sensitive unless the $STRICT
      metacommand is specified. For more information, see Chapter 6,
      "Metacommands."»

      «When using the CLOSE statement, for example, you can enter a character
      constant to specify whether to keep or delete a file. The syntax of
      this option is «,STATUS=status», and the acceptable values of status
      are 'KEEP' and 'DELETE'. As long as the $STRICT metacommand is not set,
      setting STATUS equal to 'KEEP' is equivalent to setting STATUS equal to
      'keep' or 'KeEp'.»

   ■  The 10 digits (0 through 9). Digits can be included in user-defined
      names; however, a digit may not be the first character.

   ■  All other printable characters in the ASCII character set, the blank
      character, and the TAB character.

 «In Microsoft FORTRAN, the dollar sign ($) and the underscore (_) can be
 included in user-defined names. The underscore, however, cannot be the first
 character in a name.»

 The blank character has no significance in a FORTRAN source file (except as
 listed below) so you can insert blanks to make your programs easier to read.
 The exceptions are the following:

   ■  Blanks in character constants or Hollerith fields are significant.

   ■  A blank or 0 in column 6 indicates an initial line (see Section 2.1,
      "Lines," for an explanation of initial lines).

 «The tab character's interpretation depends on which column it is in:»

 «Column                                 Interpretation»

 «1-5»                                   «The character following the tab
                                         character in the source line is
                                         interpreted as being in column
                                         7.»

 «6-72»                                  «The tab character is interpreted
                                         as a single blank, unless it is
                                         in a character or Hollerith
                                         constant (described in Section
                                         3.7.1.2). A tab character in a
                                         character or Hollerith constant
                                         is interpreted as a tab
                                         character.»


 1.2  Names

 All variables, arrays, functions, programs, and subprograms are identified
 by names. A name is a sequence of alphanumeric characters and must follow
 these guidelines:

   ■  The first character in a name must be alphabetic; the rest of the
      characters must be alphanumeric. «Microsoft FORTRAN allows the dollar
      sign ($) as an alphabetic character that follows Z in the IMPLICIT
      collating sequence for names. The underscore may also appear in names,
      but it may not be the first character.»

   ■  Blanks are ignored. Variable names like low  volt   age  and lowvoltage
      are identical to the compiler.

   ■  «Names may be up to 31 characters long; all characters are
      significant.» Only the first six alphanumeric characters are
      significant and the rest are ignored. «(This limitation does not apply
      to Microsoft FORTRAN unless the $STRICT or $TRUNCATE metacommand is in
      effect.)» Blank characters do not count. For example, the names
      delicate and d eli cat e  are both interpreted as delica «if $TRUNCATE
      or $STRICT is set.»

   ■  Your operating system or linker may impose other limits on name
      lengths.

 FORTRAN keywords are not reserved as in other languages. The compiler
 recognizes keywords by their context. For example, a program can have an
 array named IF, read, or Goto. Using keyword names for variables, however,
 makes programs harder to read and understand. For readability, and to reduce
 the possibility of hard-to-find bugs, programmers should avoid using names
 that look like parts of FORTRAN statements.

 Another type of error occurs when a typographical error causes the compiler
 to interpret a FORTRAN keyword as part of a variable name. Consider the
 following two statements:

            DO 5 INC = 1,20
            DO 5 INC = 1.20

 The first statement is the beginning of a DO loop. The second statement
 assigns the value 1.20 to a variable named DO5INC. The only difference
 between the two statements is that the first contains a comma and the second
 contains a period. The compiler cannot catch this type of error. The
 terminating line for the DO loop (a subsequent line labeled 5) is no
 different from any other line in the program, so the compiler has no way to
 recognize that an intended DO statement is missing.

 «Because the compiler reserves the following three predefined names, they
 cannot be used to name a program:»

   1. «_main, which is the external name for main programs. (The use of
      "main" is permitted under certain conditions, but is not recommended.
      See Section 2.7, "Main Program," for more information.)»

   2. «COMMQQ, which is the system name for blank common blocks.»

   3. «BLKDQQ, which is the default system name for block-data subprograms.»

 «Also, all names beginning with two underscore characters (__) or ending
 with QQ, such as __main  or MAINQQ, are reserved by the compiler. If you
 must use a name beginning with two underscore characters or ending with QQ,
 use the ALIAS attribute to provide an acceptable external name for the
 variable.»


 1.2.1  Global and Local Names

 There are two basic types of names: global and local.

 Global names are recognized anywhere in a program, so they can have only one
 global definition anywhere within that program. All subroutine, function,
 program, and common-block names are global. For example, if you use a
 subroutine named Sort in one program, you cannot also have a function named
 Sort in that program.

 You can, however, use Sort as a local name (described below) in a different
 program unit, provided you do not reference the global name Sort within that
 unit. For example, a program containing a function named Sort can also
 contain a subroutine that declares a variable named Sort, as long as the
 subroutine does not call the function Sort.

 Common-block names are a special case of global names. You can use the same
 name for a common block and a local name in the same program. This is
 permitted because common-block names are always enclosed in slashes,
 distinguishing them from other names. For example, if your program includes
 a common block named /distance/, you can also name an array in that program
 distance (arrays have local names).

 Local names have meaning only in a single program unit. In another program
 unit of the same program, the same name can be used again to refer to the
 same program object or to a different object.

 All variables, arrays, arguments, and statement functions have local names.

 Arguments to statement functions are a special case of local names. These
 arguments have meaning only in the statement-function statement. If,
 however, the arguments' names are used outside of the statement-function
 statement, the local variables in the enclosing subprogram must have the
 same data type as the statement function arguments with the same name. See
 the Statement Function entry in Section 4.2 for more information.


 1.2.2  Undeclared Names

 If a name is not explicitly defined, the compiler classifies the name
 according to the context in which it is first encountered. «If IMPLICIT NONE
 has been declared, an error message is generated at the first use of any
 variable that has not been explicitly typed. If the $DECLARE metacommand is
 in effect, a warning message is generated at the first use of any variable
 that has not been given a value in a DATA, PARAMETER, or assignment
 statement.» The following list explains how undeclared names are classified:

 Use of Name                                Classification

 As a variable or in a function             The first letter of a variable or
 reference                                  function determines its type. By
                                            default, variable names starting
                                            with I, J, K, L, M, or
                                            (uppercase «or lowercase») are
                                            given the type INTEGER, while
                                            variable names starting with any
                                            other letter or with a dollar
                                            sign are given the type REAL. You
                                            can use the IMPLICIT statement to
                                            change the association between
                                            type and first alphabetic
                                            character «(including the dollar
                                            sign)».

 As the target of a CALL statement          The compiler assumes the name is
                                            the name of a subroutine. A
                                            subroutine does not return a
                                            value through its name, so no
                                            typing occurs.


 1.3  Data Types

 The six basic data types in FORTRAN are described in this section. They are

   1. Integer (INTEGER, «INTEGER*1, INTEGER*2, and INTEGER*4»)

   2. Real (REAL, DOUBLE PRECISION, «REAL*4, or REAL*8»)

   3. Complex (COMPLEX, «COMPLEX*8, DOUBLE COMPLEX, and COMPLEX*16»)

   4. Logical (LOGICAL, «LOGICAL*1, LOGICAL*2, and LOGICAL*4»)

   5. Character (CHARACTER[«*n», where 1 <= n <= 32,767)

   6. «Record (variables defined with STRUCTURE types)»

 The data type of a variable, array, symbolic constant, or function can be
 declared in a specification statement. If its type is not declared, the
 compiler determines a name's type by its first letter (as described in
 Section 1.2.2, "Undeclared Names"). A type statement can also dimension an
 array variable. «In Microsoft FORTRAN a type statement can initialize
 variables and arrays.» See Chapter 4, "Statements," for detailed
 descriptions of type statements.

 The following sections describe each data type individually. Memory
 requirements are shown in Table 1.1.

 Table 1.1  Memory Requirements

 Type                       Bytes             Notes

 INTEGER                    «2 or» 4          «Defaults to 4 bytes. The
                                              setting of the $STORAGE
                                              metacommand determines the size
                                              of INTEGER and LOGICAL values.»
 «INTEGER*1                 1
 «INTEGER*2                 2
 «INTEGER*4                 4

 REAL                       4                 «Same as REAL*4.»
 «REAL*4                    4»

 DOUBLE PRECISION           8                 «Same as REAL*8.»
 «REAL*8                    8»

 COMPLEX                    8                 «Same as COMPLEX*8.»
 «COMPLEX*8                 8»
 «DOUBLE COMPLEX            16»               «Same as COMPLEX*16.»
 «COMPLEX*16                16»

 LOGICAL                    «2 or» 4          «Defaults to 4 bytes. The
                                              setting of the $STORAGE
                                              metacommand determines the size
                                              of INTEGER and LOGICAL values.»
 «LOGICAL*1                 1»
 «LOGICAL*2                 2»
 «LOGICAL*4                 4»

 CHARACTER                  1                 CHARACTER and CHARACTER*1 are
                                              the same.

 CHARACTER*n                n                 Maximum n is 32,767.

 «RECORD»                   «Size of          «Maximum 65,535 bytes; affected
                            structure         by $PACK metacommand.»
                            type»


 1.3.1  Integer Data Types

 An integer value is an exact representation of the corresponding integer.
 Table 1.2 shows the different types of integers, how many bytes of memory
 each type occupies, and the range of each type. «Note that variables and
 functions declared «as INTEGER are allocated as INTEGER*4, unless the
 $STORAGE metacommand is used to specify a 2-byte memory allocation. The
 $STORAGE metacommand also determines the default storage size of integer
 constants. If the $STORAGE:2 metacommand is specified, for example, integer
 constants are two bytes long, by default. However, a constant outside the
 INTEGER*2 range is given 4 bytes of storage.»

 Table 1.2  Integers

 Data Type                  Bytes            Range

 «INTEGER*1»                1                «-128 to 127»
 «INTEGER*2»                2                «-32,768 to 32,767»

 «INTEGER*4»                4                «-2,147,483,648 to
                                             2,147,483,647»

 INTEGER                    «2 or» 4         «Depends on setting of $STORAGE»

 «Although an INTEGER*4 constant or variable can take any value in its full
 32-bit range, it cannot be assigned its smallest negative value
 (-2,147,483,648) when that value is expressed as a numeric constant. For
 example, the statement varname = -2147483648 causes a compiler overflow
 error. If you wish to give a four-byte integer this value, the right side of
 the assignment statement must be a mathematical expression.»


 Syntax

 Constants are interpreted in base 10. To specify a constant «that is not in
 base 10,» use the following syntax:

 «sign»««base»#»constant

 The optional sign is a plus or minus sign. «The base can be any integer from
 2 through 36. If base is omitted but # is specified, the integer is
 interpreted in base 16. If both base and # are omitted, the integer is
 interpreted in base 10. For bases 11 through 36, the letters A through Z
 represent numbers greater than 9. For base 36, for example, A represents 10,
 B represents 11, C represents 12, and so on, through Z, which represents 35.
 The case of the letters is not significant.»


 Example

 The following seven integers are all assigned a value equal to 3,994,575
 decimal:

      «I = 2#1111001111001111001111
      m = 7#45644664
      J = +8#17171717
      K = #3CF3CF
      n = +17#2DE110»
      L = 3994575
      «index = 36#2DM8F»

 A decimal point is not allowed in an integer constant.

 Integer constants must also be in the ranges specified above. «However, for
 numbers with a radix that is other than 10, the compiler reads out-of-range
 numbers up to 2^(32). They are interpreted as the negative numbers with the
 corresponding internal representation. For example, 16#FFFFFFFF results in
 an arithmetic value of -1. If the $DEBUG metacommand is in effect, a
 compile-time range error occurs instead.»


 1.3.2  The Single-Precision IEEE Real Data Type

 A single-precision real value (REAL «or REAL*4») is usually an approximation
 of the real number desired and occupies 4 bytes of memory. The precision of
 this data type is between six and seven decimal digits. You can specify more
 than six digits, but only the first six are significant. The range of
 single-precision real values includes the negative numbers from
 approximately -3.4028235E+38 to -1.1754944E-38, the number 0, and the
 positive numbers from approximately +1.1754944E-38 to +3.4028235E+38.


 Syntax

 A real constant has the following form:

      «sign»«integer» «.«fraction»»«Eexponent»

 Parameter                                  Value

 sign                                       A sign (+ or -).

 integer                                    An integer part, consisting of
                                            one or more decimal digits.
                                            Either integer or fraction may be
                                            omitted, but not both.

 .                                          A decimal point.

 fraction                                   A fraction part, consisting of
                                            one or more decimal digits.
                                            Either fraction or integer may be
                                            omitted, but not both.

 Eexponent                                  An exponent part, consisting of
                                            an optionally signed one- or
                                            two-digit integer constant. An
                                            exponent indicates that the value
                                            preceding the exponent is to be
                                            multiplied by 10 raised to the
                                            value exponent.


 Example

 The following real constants all represent the same real number (1.23):

      +1.2300E0      .012300E2      1.23E0          123E-2
      +1.2300        123.0E-2       .000123E+4      1230E-3


 1.3.3  The Double-Precision IEEE Real Data Type

 A double-precision real value (DOUBLE PRECISION «or REAL*8») is usually an
 approximation of the real number desired, and occupies 8 bytes of memory.
 The precision is between 15 and 16 decimal digits. You can specify more
 digits, but only the first 15 are significant. The range of double-precision
 real values includes the number 0 and the negative numbers from
 approximately -1.797693134862316D+308 to -2.225073858507201D-308. It also
 includes the positive numbers from approximately +2.225073858507201D-308 to
 +1.797693134862316D+308.

 A double-precision real constant has the same form as a single-precision
 real constant, except that the letter D is used for exponents instead of the
 letter E, and an exponent part is mandatory. If the exponent is omitted, the
 number is interpreted as a single-precision constant.


 Example

 The following double-precision real constants all represent fifty-two
 one-thousandths:

      5.2D-2    +.00052D+2    .052D0    52.000D-3    52D-3


 1.3.4  Complex Data Types

 The COMPLEX «or COMPLEX*8» data type is an ordered pair of single-precision
 real numbers. «The DOUBLE COMPLEX or COMPLEX*16 data type is an ordered pair
 of double-precision real numbers.» The first number in the pair is the real
 part of a complex number, and the second number in the pair is the imaginary
 part. Both the real and imaginary components of a COMPLEX «or COMPLEX*8»
 number are REAL*4 numbers, so COMPLEX «or COMPLEX*8» numbers occupy 8 bytes
 of memory. «Both the real and imaginary components of a DOUBLE COMPLEX or
 COMPLEX*16 number are REAL*8 numbers, so DOUBLE COMPLEX and COMPLEX*16
 numbers occupy 16 bytes of memory.»


 Syntax

      ««sign»»(real, imag)

 Parameter                                  Value

 «sign»                                     «A sign (+ or -). If specified,
                                            the sign applies to both real and
                                            imag.»

 real                                       A real number «or an integer»,
                                            representing the real part.

 imag                                       A real number «or an integer»,
                                            representing the imaginary part.

 For example, the ordered pair (7,3.2) represents the complex number
 7.0+3.2i. «The ordered pair -(-.11E2,#5F) represents the complex number
 11.0-95.0i.»


 1.3.5  Logical Data Types

 The logical data type consists of two logical values, .TRUE. and .FALSE. A
 LOGICAL variable occupies «2 or» 4 bytes of memory«, depending on the
 setting of the $STORAGE metacommand. The default is 4 bytes. The value (true
 or false) of a logical variable is unaffected by the $STORAGE metacommand.»

 «LOGICAL*1 values occupy a single byte, which is either 0 (.FALSE.) or 1
 (.TRUE.). LOGICAL*2 values occupy 2 bytes: the least significant (first)
 byte contains a LOGICAL*1 value and the most significant byte is undefined.
 LOGICAL*4 variables occupy two words: the least significant (first) word
 contains a LOGICAL*2 value; the most significant word is undefined.»


 1.3.6  The Character Data Type

 Character variables occupy one byte of memory for each character and are
 assigned to contiguous bytes, independent of word boundaries. «However, when
 character and noncharacter variables are allocated in the same common block,
 the compiler places noncharacter variables that follow character variables
 on word boundaries. See the COMMON entry in Section 4.2, "Statement
 Directory," for more information on character variables in common blocks.»

 The length of a character variable, character array element, structure
 element, character function, or character constant with a symbolic name must
 be between 1 and 32,767. The length can be specified by the following:

   ■  An integer constant in the range 1-32,767

   ■  An expression (in parentheses) that evaluates to an integer in the
      range 1-32,767

   ■  An asterisk, which indicates the length of the string may vary


 Examples

 The following examples show proper usage:

            CHARACTER*32 string
            CHARACTER string*32
            CHARACTER string*(const*5)
            CHARACTER string*(*)
            CHARACTER*(*) string

 See the CHARACTER entry in Section 4.2, "Statement Directory," for more
 information.

 A character constant is a sequence of one or more of the printable ASCII
 characters enclosed in a pair of apostrophes (') «or quotes (")». The
 apostrophes «or quotes» that delimit the string are not stored with the
 string. To represent an apostrophe within a string delimited by apostrophes,
 specify two consecutive apostrophes with no blanks between them. «To
 represent a quotation mark within a string delimited by quotation marks,
 specify two consecutive quotation marks with no blanks between them.»

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    An apostrophe is a single right quotation mark ('), not a single left
    quotation mark (`). In the typeface used in examples, such as 'string',
    apostrophes look like this: '.
 ───────────────────────────────────────────────────────────────────────────

 Blank characters and tab characters are permitted in character constants and
 are significant. The case of alphabetic characters is also significant. «You
 can use C strings (as described in Section 1.3.6.1) to define strings with
 nonprintable characters, or to specify the null string.»

 The length of a character constant is equal to the number of characters
 between the apostrophes «or quotation marks». (A pair of apostrophes in a
 string delimited by apostrophes counts as a single character. «A pair of
 quotation marks in a string delimited by quotation marks counts as a single
 character.»)

 Some sample character constants are listed below:

 String                                     Constant

 'String'                                   String

 '1234!@#$'                                 1234!@#$

 'Blanks count'                             Blanks count

 ''''''                                     ''

 'Case Is Significant'                      Case Is Significant

 «"''"»                                     «''»

 «"""Double"" quotes count as               «"Double" quotes count as one»
 one"»

 FORTRAN source lines are 72 characters long (characters in columns 73-80 are
 ignored by the compiler), and lines with less than 72 characters are padded
 with blanks. Therefore, when a character constant extends across a line
 boundary, its value includes any blanks added to the lines. Consider the
 following FORTRAN statement:

      C     Sample FORTRAN continuation line
            Heading (secondcolumn) =  'Acceleration of Particles
           +from Group A'

 That statement sets the array element heading (secondcolumn) to
 'Acceleration of Particles              from Group A'. There are 14 blanks
 between Particles and from because Particles ends in column 58, leaving
 14 blanks to be added to the string.


 1.3.6.1  «C Strings»

 «String values in the C language are terminated with null characters
 (CHAR(0)), and may contain nonprintable characters (such as newline and
 backspace). Nonprintable characters are specified using the backslash as an
 escape character, followed by a single character indicating the nonprintable
 character desired. This type of string is specified in Microsoft FORTRAN by
 using a standard string constant followed by the character C. The standard
 string constant is then interpreted as a C-language constant. Backslashes
 are treated as escapes, and a null character is automatically appended to
 the end of the string (even if the string already ends in a null character).
 Table 1.3 shows the valid escape sequences. If a string contains an escape
 sequence that isn't in this table (such as \z), the backslash is ignored.»

 «Table 1.3  C String Escape Sequences»

 «Sequence                  Character»

 «\a»                       «Bell»

 «\b»                       «Backspace»

 «\f»                       «Form feed»

 «\n»                       «Newline»

 «\r»                       «Carriage return»

 «\t»                       «Horizontal tab»

 «\v»                       «Vertical tab»

 «\xhh»                     «Hexadecimal bit pattern»

 «\'»                       «Single quote»

 «\"»                       «Double quote»

 «\\»                       «Backslash»

 «\ooo»                     «Octal bit pattern»

 «A C string must also be a valid FORTRAN string. Therefore, if the string is
 delimited by apostrophes, all apostrophes in the string itself must be
 represented by double apostrophes. The escape sequence \'a causes a compiler
 syntax error because FORTRAN interprets the quotation mark as the end of a
 string.

 «The correct form is \''a. (If the string is delimited with quotation marks,
 an apostrophe may be entered as a single character.) C strings and ordinary
 strings differ only in how you specify the value of the string. The compiler
 treates them identically.»

 «The sequences \ooo and \xhh allow any ASCII character to be given as a one-
 to three-digit octal or a one- to two-digit hexadecimal character code. The
 o digit must be in the range 0-7, and the h digit must be in the range 0-F.
 For example, the C strings '\010'C and '\x08'C both represent a backspace
 character followed by a null character.»

 «The C string '\\abcd'C is equivalent to the string  '\abcd'  with a null
 character appended. The string ''C represents the ASCII null character. Note
 that the character constant '' is illegal because it has a length of 0, but
 ''C is legal because it has a length of 1.»


 1.3.6.2  Character Substrings

 Character substrings are used to access a contiguous part of a character
 variable.


 Syntax

      variable («first»:«last»)

 Parameter                                  Description

 variable                                   A character variable or a
                                            character array element.

 first                                      An integer expression «or
                                            arithmetic expression» that
                                            defines the first (leftmost)
                                            character in the substring. The
                                            compiler truncates first to an
                                            integer value. The default for
                                            first is 1, so if first is
                                            unspecified, the substring starts
                                            with the first character in the
                                            string.

 last                                       An integer expression «or
                                            arithmetic expression» that
                                            defines the last (rightmost)
                                            character in the substring. The
                                            compiler truncates last to an
                                            integer value. The default for
                                            last is the length of the string,
                                            so if last is unspecified, the
                                            substring ends with the last
                                            character in the string.

 The length of the substring is last - first + 1. For example, if a 10-byte
 character variable name contains the string 'Jane Doe ', the following is
 true:

 Character Variable                         Equivalent String

 name(:5)                                   'Jane '

 name(6:)                                   'Doe '

 name(3:7)                                  'ne Do'

 name(:)                                    'Jane Doe '

 Note that name(:) is equivalent to name.

 If the character variable is of length characters, the following
 relationships must be true:

      1 <= first <= last <= length

 That is, both first and last must be greater than zero; last must be greater
 than or equal to first; and neither first nor last can be greater than
 length.

 For the 10-byte character variable name, the following substring
 specifications are illegal:

      name(0:4)
      name(6:5)
      name(11:12)
      name(0:)
      name(:11)

 «If the $DEBUG metacommand is in effect, a run-time error occurs if these
 relationships are not true. If $DEBUG is not in effect, the results are
 undefined.»


 Example

      C     This program writes the second half of
      C     the alphabet, followed by the first half.
            CHARACTER alpha*26
            alpha = 'abcdefghijklmnopqrstuvwxyz'
            WRITE (*, *) alpha(14:), alpha(:13)
            END


 Output

      nopqrstuvwxyzabcdefghijklm

 The use of a noninteger expression for the first and last parameters causes
 an error. «(This limitation does not apply to Microsoft FORTRAN unless the
 $STRICT metacommand is in effect.)»


 1.4  «Records»

 «A record is a "structure" variable. A structure is a user-defined compound
 data type that consists of variable type definitions and unions of maps as
 explained below. Each item within a structure is called a "structure
 element."»


 «Example»

 «The following example is a structure definition for a variable type that
 contains employee data:»

           «STRUCTURE /employee_data/
               CHARACTER*25  last_name
               CHARACTER*15  first_name
               CHARACTER*20  street_name
               INTEGER*2     street_number
               INTEGER*2     apt_number
               CHARACTER*20  city
               CHARACTER*2   state
               INTEGER*4     zip
               INTEGER*4     telephone
               INTEGER*2     date_of_birth
               INTEGER*2     date_of_hire
               INTEGER*2     social_security(3)
               LOGICAL*2     married
               INTEGER*2     dependents
            END STRUCTURE»

 «A structure specification is not a variable, but a variable type. Structure
 variables are defined with the RECORD statement. In a company with 200
 employees, the following statement would define an array of structure
 variables to hold employee data, using the previous structure definition:»

            «RECORD /employee_data/ employees(200)»

 «Structure elements are referenced by specifying the sequence of elements
 needed to reach the desired element. The elements are separated by a period.
 For example, the number of dependents of the ninety-ninth employee is
 specified by employees(99).dependents. The first letter of the name of the
 state where this employee lives is specified by employees(99).state(1:1).
 Note that the data type of a structure reference is the type of the element
 referenced. In the two examples given here, the first reference is of
 type INTEGER*2. The second reference is of type CHARACTER*2. Any intervening
 references are for identification only; they do not affect the data type.»

 «Because periods are used to delimit structure elements, an element may not
 have the name of a relational or logical operator (NOT, AND, OR, GE, EQ,
 etc.). If the name of a relational or logical operator is used, the compiler
 will try to interpret the name as a relational operator.»


 «Example»

 «A structure definition can contain structure variable declarations. For
 example, the employee's name and address could have been defined as
 individual structure types:»

           «STRUCTURE /employee_name/
               CHARACTER*25  last_name
               CHARACTER*15  first_name
            END STRUCTURE»

           «STRUCTURE /employee_addr/
               CHARACTER*20  street_name
               INTEGER*2     street_number
               INTEGER*2     apt_number
               CHARACTER*20  city
               CHARACTER*2   state
               INTEGER*4     zip
            END STRUCTURE»

 «These become structure variables within the employee_data  structure:»

           «STRUCTURE /employee_data/
               RECORD  /employee_name/  name
               RECORD  /employee_addr/  addr
               INTEGER*4  telephone
               INTEGER*2  date_of_birth
               INTEGER*2  date_of_hire
               INTEGER*2  social_security(3)
               LOGICAL*2  married
               INTEGER*2  dependents
            END STRUCTURE»

 «Note that the reference to the first letter of the ninety-ninth employee's
 state of residence would now be employees(99).addr.state(1:1).»

 «A structure definition can also contain unions of maps. A map specifies
 that one or more variables are positioned contiguously in memory. The
 variables may be of any type, including structures, as shown below:»

           «MAP
               INTEGER*4  manny, moe, jack
               CHARACTER*21  myname
            END MAP»

 «In the examples, the four-byte integers manny, moe, and jack  appear first
 in memory, followed immediately by the 21-character variable myname. The
 starting addresses of each variable (and whether or not any bytes are used)
 is determined by the setting of the $PACK metacommand or /Zp command-line
 option.» «Here is another example of a map:»

           «MAP
               CHARACTER*7  yourname
               REAL*4  meg, joe, amy, zelda
            END MAP»

 «Maps can only appear within a UNION statement. When maps are combined in a
 union, the variables overlay each other as they do in an EQUIVALENCE
 statement, as shown in the following example:»

           «UNION
               MAP
                  INTEGER*4  manny, moe, jack
                  CHARACTER*21  myname
               END MAP
               MAP
                  CHARACTER*7  yourname
                  REAL*4  meg, joe, amy, zelda
               END MAP
            END UNION»

 «Assuming that $PACK:1 is in effect, the seven bytes of yourname will
 overlay the four bytes of manny, and the first three bytes of moe. The first
 byte of meg will overlay the last byte of moe, and so forth. If $PACK:2 is
 in effect, yourname and manny plus moe will be overlaid as before, but the
 odd byte at the end of yourname will be left empty. Therefore, the first
 byte of meg will begin on the first byte of jack. For more information on
 the $PACK metacommand, see Chapter 6, "Metacommands."»

 «Although individual structure elements may be written to or read from a
 file, any attempt to write a structure variable as a whole causes a
 compile-time error.»

 «For more information, see the RECORD and STRUCTURE entries in Section 4.2,
 "Statement Directory."»


 1.5  Arrays

 «The number of elements in an array and the number of its dimensions are
 limited only by available memory. If the $STRICT metacommand is in effect,»
 a warning is generated if more than seven dimensions are specified. An array
 element is referenced with the following syntax:

      array (subscripts)

 Parameter                                  Value

 array                                      The name of the array. If the
                                            type of the array is not declared
                                            in a type statement, the array
                                            elements have the type indicated
                                            by the first letter of array.

 subscripts                                 Subscript expression(s). If there
                                            is more than one subscript
                                            expression, they must be
                                            separated by commas. The number
                                            of subscript expressions must
                                            equal the number of dimensions in
                                            the array declaration. For
                                            information on declaring arrays,
                                            see the DIMENSION entry in
                                            Section 4.2, "Statement
                                            Directory."

                                            Each subscript must be an
                                            arithmetic expression. The result
                                            of the expression is converted to
                                            an integer by truncation. Each
                                            subscript must be an integer
                                            expression. «(This limitation
                                            does not apply to Microsoft
                                            FORTRAN unless the $STRICT
                                            metacommand is in effect.)»
                                            Function, array-element, and
                                            statement-element references are
                                            allowed. A subscript value can be
                                            positive, negative, or 0.


 Examples

 The following are examples of array-element references:

            DIMENSION  a(3,2), b(0:2,0:3), c(4,5), d(5,6),
           +           v(10), q(3.2,4.5)
            EQUIVALENCE (x, v(1)), (y, v(2))
            d(i,j) = d(i+b(i,j),j) / pivot
            c(i+1,j) = c(i,j) + a(i**2,k) * b(k*j,j-24)
            READ (*, *) (v(n), n = 1, 10)

 Array elements are stored contiguously in memory so that array element n+1
 follows array element n, and so on. With multidimensional arrays, the
 elements are organized in a linear sequence because computer memory has only
 one dimension.

 In FORTRAN, array elements are arranged in "column-major order." This means
 the order of storage is determined by incrementing the leftmost subscript
 first, then the next subscript to the right, and so on. The elements of the
 variable array(3,4), are stored in the following sequence:

      array(1,1)    array(2,1)    array(3,1)
      array(1,2)    array(2,2)    array(3,2)
      array(1,3)    array(2,3)    array(3,3)
      array(1,4)    array(2,4)    array(3,4)

 In Microsoft FORTRAN, array storage space can be allocated dynamically. That
 is, the size of each dimension can be set at run-time, rather than during
 compilation. This is done with an "allocatable" array.

 «For an array to be allocatable, it must have been declared with the
 attribute  ALLOCATABLE, and the number of its dimensions specified. When the
 program runs, the ALLOCATE statement sets the size of each dimension. When
 the array is no longer needed, the DEALLOCATE statement returns the array's
 memory space to the common pool.»

 «The following example shows the correct usage of the ALLOCATABLE attribute
 and the ALLOCATE and DEALLOCATE statements:»

           «INTEGER  dataset[ALLOCATABLE](:,:,:)
            INTEGER  reactor, level, points, error
            DATA  reactor, level, points  / 10, 50, 100 /
            .
            .
            .
            ALLOCATE (dataset(reactor,level,points), STAT = error)
            .
            .
            .
            DEALLOCATE (dataset, STAT = error)»

 «A significant enhancement to Microsoft FORTRAN 5.0 is the ability to
 perform operations on arrays as if they were ordinary scalar quantities. See
 Section 1.7.1, "Arithmetic Expressions," for more information.»


 1.6  «Attributes»

 «Attributes specify additional information about a variable, variable type,
 subprogram, or subprogram formal argument. Attributes allow your FORTRAN
 program, for example, to use the calling conventions of Microsoft C or
 Pascal, pass arguments by value or by reference, use segmented or
 unsegmented addresses, specify that a formal argument can span more than one
 segment, or specify an external name for a subprogram or common block.
 Attributes can be used in subroutine and function definitions, in type
 declarations, and with the INTERFACE and ENTRY statements (see Section 4.2).
 Table 1.4 summarizes which attributes can be used with which objects. See
 Sections 1.6.1-1.6.12 for information on each attribute.»

 «Table 1.4   Objects to Which Attributes Can Refer»

                 «Subprogram
                  Variable and          Specification and
                  Array                 Common-               EXTERNAL
 Attribute        Declarations          Block Names           Statements»

 «ALIAS           Yes                   Yes                   Yes»
 «ALLOCATABLE     Yes (arrays only)     No                    No»
 «C               Yes                   Yes                   Yes»
 «EXTERN          Yes                   No                    No»
 «FAR             Yes                   Yes                   Yes»
 «HUGE            Yes                   No                    No»
 «LOADDS          No                    No                    Yes»
 «NEAR            Yes                   Yes                   Yes»
 «PASCAL          Yes                   Yes                   Yes»
 «REFERENCE       Yes                   No                    No»
 «VALUE           Yes                   No                    No»
 «VARYING         No                    No                    Yes»


 «Syntax»

      «object « [attrs] »»

 «Attributes follow the object they refer to. If more than one attribute is
 specified, they must be separated by commas.»


 «Examples»

 «In the following example, the integer x is passed by reference, using a
 short address (offset only):»

           «INTEGER intvar[REFERENCE, NEAR]»

 «In the following example, PasFunc is a Pascal function with arguments i,
 j, and k that are C integers. PasFunc also returns a C integer.»

           «INTERFACE TO INTEGER [C] FUNCTION PasFunc [PASCAL]
           +(i, j, k)
            INTEGER [C] i, j, k
            END»


 1.6.1  «ALLOCATABLE»

 «This attribute specifies that an array is "allocatable"; that is, the size
 of each dimension is established dynamically at run-time, not during
 compilation.»

 «The ALLOCATABLE attribute can appear either in a type declaration or in a
 DIMENSION statement. Each dimension of the allocatable array is indicated by
 a colon. If there is more than one dimension, the colons are separated by
 commas, as in this example which declares the allocatable array dynamic with
 three dimensions:»

           «REAL*8  dynamic [ALLOCATABLE] (:,:,:)»

 «The ALLOCATABLE attribute may not be applied to formal arguments, nor may
 it be used with the NEAR attribute. If you anticipate that an allocatable
 array's size will exceed 65,536 bytes, you must also declare the array with
 the HUGE attribute so the correct addressing will be generated.»

 «For information on the use of allocatable arrays, see the reference
 sections on the ALLOCATE and DEALLOCATE statements.»


 1.6.2  «ALIAS»

 «This attribute allows you to specify an external name for a subprogram or
 common block. The name may differ from the name used in the declaration.»


 «Syntax»

      «ALIAS: string»

 «Parameter                                 Description»

 «string»                                   «A character constant (can be a C
                                            string, as described in Section
                                            1.3.6.1).»

                                            «No transformations are performed
                                            on string. Lowercase letters, for
                                            example, are not converted to
                                            uppercase. This is useful when
                                            interfacing with case-sensitive
                                            languages, such as C.»

 «Within the source file, the subprogram can be referred to only by the name
 given in its declaration. Outside the source file, the subprogram can be
 referred to only by its ALIAS name.»

 «You can also use the ALIAS attribute on an INTERFACE statement to redefine
 the name of a subprogram in another source file you wish to call.»

 «The ALIAS attribute overrides the C attribute. If the C attribute is used
 on a subprogram along with the ALIAS attribute, the subprogram will be given
 the C calling convention, but not the C naming convention. It will instead
 receive the name given for the ALIAS, with no modifications.»

 «The ALIAS attribute cannot be applied to formal arguments.»


 «Example»

 «This SUBROUTINE statement gives the subroutine PasSub the name OtherName
 outside this source file.»

           «SUBROUTINE PasSub [ALIAS:'OtherName']»


 1.6.3  «C»

 «The C attribute can be applied to subprograms, common blocks, and types.
 When applied to a subprogram, the C attribute defines the subprogram as
 having the same calling conventions as a Microsoft C procedure. The
 following list describes the differences between FORTRAN calling conventions
 and C calling conventions:»

 «Difference                                Explanation»

 «Order in which parameters are             «Microsoft FORTRAN pushes
 pushed on the stack»                       parameters on the stack in the
                                            order in which they appear in the
                                            procedure declaration. Microsoft
                                            C, by default, pushes its
                                            parameters in the reverse order
                                            (to allow varying numbers of
                                            arguments).»

 «Location of code that restores            «In Microsoft FORTRAN this code
 the stack when a procedure                 is in the called procedure. This
 completes execution»                       generates less code than
                                            Microsoft C, in which this code
                                            follows the procedure call.»

 «Arguments to subprograms with the C attribute are passed by value unless
 the formal argument is specified with the REFERENCE attribute. (Note that
 the VARYING attribute can only be specified for subprograms that also have
 the C attribute.)»

 «The names of subprograms using the C attribute are modified automatically
 to make it easier to match naming conventions used in C. External names are
 changed to lowercase, and begin with an underscore (_). To use a name
 containing uppercase letters, use the ALIAS attribute (described in Section
 1.6.2).»

 «When an integer variable is given the C attribute, it becomes a C integer
 variable. The default size for C and FORTRAN integers depends on the
 microprocessor. For example, on the 8086, Microsoft FORTRAN assumes 32-bit
 integers by default, while C assumes 16-bit. On other machines, both
 languages may assume 32-bit integers. The C attribute allows you to control
 integer size for compatibility between your FORTRAN programs and C
 functions.»


 «Syntax»

 «The C attribute cannot be applied to formal arguments, except those of
 INTEGER type, as in the following syntax line:»

      «INTEGER[C]argument»

 «The ALIAS attribute overrides the C attribute.»


 1.6.4  «EXTERN»

 «The EXTERN attribute can be used in variable declarations. It indicates
 that the variable is allocated in another source file. EXTERN must be used
 when acces-sing variables declared in other languages, and must not be
 applied to formal arguments.»


 1.6.5  «FAR»

 «When used with formal arguments, the FAR attribute specifies that the
 argument is to be passed using a segmented address. When used with
 variables, it specifies that the variable is allocated in far data areas.»


 1.6.6  «HUGE»

 «The HUGE attribute is a convenient way to specify that a formal argument or
 an allocatable array may span more than one segment. (The $LARGE metacommand
 specifies the same thing.) The following two fragments behave identically:»

           «FUNCTION Func (a[HUGE])
            DIMENSION a(200)»

           «$LARGE: a
            FUNCTION Func (a)
            DIMENSION a(200)»

 «The compiler does not ensure that HUGE is specified for all arguments that
 span more than one segment. Versions 3.3 and earlier of Microsoft Pascal and
 Versions 3.0 and earlier of Microsoft C do not support parameters with HUGE
 attributes.»


 1.6.7  «LOADDS»

 «The LOADDS attribute is applied only to separately compiled subprograms and
 functions. It directs the compiler to create a separate data segment for the
 data within that procedure; the base address (DGROUP) of this new data
 segment is automatically loaded into DS when the procedure is called. The
 use of a separate data segment permits the procedure's data to be called
 with 16-bit NEAR references rather than 32-bit FAR references. This speeds
 up data access. The default data segment for the program is automatically
 reloaded when execution of the procedure terminates.»

 «The LOADDS attribute is applied primarily to user-written routines that are
 to be included in an OS/2 dynamic link library (DLL). It is not needed for
 procedures that run in DOS programs because the FL command-line option /ND
 (name data segment) automatically assures that the new data segment's base
 address is loaded. The following is an example of the LOADDS attribute:»

           «REAL*8 FUNCTION [LOADDS] GetNewData»


 1.6.8  «NEAR»

 «The NEAR attribute specifies that the actual argument is in the default
 data segment and that only its offset is passed to the subprogram. This
 attribute can also be used with common blocks. Common blocks having the NEAR
 attribute are mapped into the default data segment.»


 «Syntax»

      «COMMON «/«name» [NEAR] /»...»

 «The parameter name is the name of the common block. When no name is
 specified, all blank common blocks are put in the default data segment. NEAR
 must be specified for at least the first definition of the common block in
 the source file. You can, however, specify NEAR for any of the COMMON
 statements in a subprogram.»

 «To make a common block near, specifying NEAR for all definitions of the
 common block is good programming practice. If, however, you are modifying an
 existing program, it can be easier to add a subroutine at the beginning of
 your source file to make common blocks near in the remainder of the
 program.»

 «The advantage of putting common blocks in the default data segment is that
 you can specify addresses with offsets only. This generates smaller, more
 efficient code. If you do not specify NEAR, the compiler uses segmented
 addresses to refer to everything in common blocks.»

 «If a common block is specified as near in one compiland, but not in
 another, it will be mapped into the default data segment. The compiland
 which recognizes it as near will use short addresses, and the other
 compiland will use long addresses. While this practice is not recommended,
 it does provide compatibility with libraries compiled with Version 3.2 of
 the compiler.»

 «Actual arguments passed to a near formal argument must be in the default
 data segment. You cannot pass any of the following to a near argument:»

   ■  «Data in common blocks that are not specified using the NEAR attribute»

   ■  «Arrays specified using the HUGE attritute»

   ■  «Arrays defined while the $LARGE metacommand is in effect»

   ■  «Variables named in a $LARGE metacommand»


 1.6.9  «PASCAL»

 «The PASCAL attribute can be used with subprograms, common blocks, and
 formal argument type declarations (but not on formal arguments in the formal
 argument list). This attribute identifies an argument or subprogram as
 having the following characteristics of Microsoft Pascal:»

   ■  «The argument or the subprogram's arguments are passed by value (unless
      the REFERENCE attribute is specified)»

   ■  «Microsoft FORTRAN's calling conventions are still used»


 1.6.10  «REFERENCE»

 «The REFERENCE attribute can only be applied to formal arguments. It
 specifies that the argument's memory location is to be passed, rather than
 the argument's value. This is called "passing by reference."»


 1.6.11  «VALUE»

 «The VALUE attribute can only be applied to formal arguments. It specifies
 that the argument's value is to be passed, rather than the argument's memory
 location. This is called "passing by value."»

 «If either the C or PASCAL attribute is specified on the subprogram
 definition, all arguments are assumed to have the VALUE attribute, since C
 and Pascal normally pass by value. Character values, substrings,
 assumed-size arrays, and adjustable-size arrays cannot be passed by value.»

 «When a formal argument has the VALUE attribute, the actual argument passed
 to it can be of a different type. If type conversion is necessary, it is
 performed before the call, following the rules discussed in Section 1.7.1.2,
 "Type Conversion of Arithmetic Operands."»

 «When passing by value, values cannot be returned through the formal
 argument, since the formal argument now addresses a stack location
 containing the value of the actual argument, rather than addressing the
 actual argument itself.»

 «In C, arrays never pass by value. If you specify the C attribute and your
 subprogram has an array argument, the array will be passed as if it were a C
 data «structure (struct). To pass an array and have it treated as an array
 (instead of as a struct), you can do one of two things:»

   1. «Use the REFERENCE attribute on the formal argument.»

   2. «Pass the address returned by the LOC, LOCNEAR, or LOCFAR functions by
      value.»


 «Example»

 «Integer x is passed by value in the following example:»

           «SUBROUTINE Subr (x[VALUE])
            INTEGER x[VALUE]»


 1.6.12  «VARYING»

 «In FORTRAN, a formal argument must be defined for each actual argument.
 Other languages (such as C) allow actual arguments for which no formal
 arguments are defined. These actual arguments are assumed to be passed by
 value, with no automatic data-type conversion. When the C attribute is
 specified you can also apply the VARYING attribute, which permits the number
 of actual arguments to be different from the number of formal arguments.
 Actual arguments for which a formal argument is defined must still follow
 the type rules, however.»

 «When writing a FORTRAN procedure with the VARYING attribute, be sure the
 procedure only uses arguments you actually passed, or you will get undefined
 results.»

 «Note that the FORTRAN calling sequence cannot support varying numbers of
 arguments; the VARYING attribute has no effect unless you have also
 specified the C attribute on the subprogram.»


 1.7  Expressions

 An expression is a formula for computing a value. Expressions consist of
 operands and operators. The operands can be function references, variables,
 structure elements, constants, or other expressions. The operators specify
 the actions to be performed on the operands. In the following expression,
 for example, the slash (/) is an operator and chickens and coops are
 operands:

      chickens / coops

 There are four kinds of expressions in FORTRAN, as shown below:

 Expression                                 Result

 Arithmetic                                 Integer, real, or complex value

 Character                                  Character or string value

 Relational                                 Logical value

 Logical                                    Logical «or integer» value

 Expressions are components of statements. In the following example, the
 entire line is a statement, and the portion after the equal sign (=) is an
 expression:

            cost = 10.95 * chickens / coops

 Any variable, array, array element, «structure, structure element,» or
 function that is referred to in an expression must be defined at the time
 the reference is made, or the results are undefined. Integer variables must
 have a numerical value, rather than a statement-label value set by an ASSIGN
 statement.

 FORTRAN only guarantees that expressions generate correct values, not that
 all parts of expressions are evaluated. For example, if an expression
 multiplies (37.8 / scale**expo + factor) by zero, then (37.8 / scale**expo +
 factor) may not be evaluated. Similarly, if a false value and a logical
 expression are operands of the .AND. operator, the expression may not be
 evaluated. In the following example, the expression (SWITCH .EQ. on) may not
 be evaluated:

      ((3 .LE. 1) .AND. (SWITCH .EQ. on))

 «Microsoft FORTRAN permits expressions to use full arrays in the same way
 they would normally use single (scalar) arguments. For a complete
 explanation, see Section 1.7.5, "Array Expressions."»


 1.7.1  Arithmetic Expressions

 An arithmetic expression produces a value that is an integer, a real or
 complex number, or an array of those types. The basic operands used in
 arithmetic expressions are

   ■  Arithmetic constants

   ■  Symbolic names for arithmetic constants

   ■  Variable references

   ■  Array-element references

   ■  Function references

   ■  Array references

   ■  «Structure-element references»

 Other arithmetic expressions are built from the basic operands in the
 preceding list, cusing parentheses and the arithmetic operators shown in
 Table 1.5.

 Table 1.5  Arithmetic Operators

 Operator          Operation                  Precedence

  **               Exponentiation             1 (highest)

  /                Division                   2

  *                Multiplication             2

  -                Subtraction (binary)       3
                   or negation (unary)

  +                Addition (binary) or       3
                   identity (unary)

 All of the arithmetic operators are binary operators, appearing between two
 operands. The plus and minus operators can also be used as unary operators,
 which precede a single operand.

 When consecutive operations are of equal precedence, the leftmost operation
 is performed first. For example, the expression first/second*third  is
 equivalent to (first/second)*third. There is one exception to this rule:
 exponentiation. When there are two consecutive exponentiation operations,
 the rightmost operation is performed first. For example, the following
 expressions are equivalent:

      first**second**third
      first**(second**third)

 FORTRAN does not allow two arithmetic operators to appear consecutively. For
 example, FORTRAN prohibits first**-second, but permits first** (-second).

 The following list shows examples of the precedence of arithmetic operators:

 Expression                          Equivalent Expression

 3 * 7 + 5                           (3 * 7) + 5

 -one**two                           -(one**two)

 + x / y                             +(x / y)

 area / g - qu**2**fact              (area / g) - (qu**(2**fact))

 The following arithmetic operations are prohibited:

   ■  Division by zero

   ■  Raising a zero-value operand to a negative power

   ■  Raising a negative-value operand to a nonintegral real power


 1.7.1.1  Integer Division

 When one integer is divided by another, the truncated quotient of the two
 operands is returned. Thus, 7/3 evaluates to 2, (-7)/3 evaluates to -2, and
 both 9/10 and 9/(-10) evaluate to zero.

 For example, look at the following assignment statement:

            X = 1/4 + 1/4 + 1/4 + 1/4

 First, note that division has higher precedence than addition, so the
 expression is equivalent to (1/4) + (1/4) + (1/4) + (1/4). Then, take the
 quotient of 1/4 and truncate it. The result is 0. The assignment, therefore,
 sets X equal to zero.


 1.7.1.2  Type Conversion of Arithmetic Operands

 When all operands of an arithmetic expression are of the same data type, the
 value returned by the expression is also of that type. When the operands are
 of different data types, the type of the value returned by the expression is
 the type of the highest-ranked operand. The exception to the rule is
 operations involving both REAL*8 numbers and COMPLEX*8 numbers, which yield
 COMPLEX*16 results.

 The ranking of arithmetic operands is as follows:

   1. «DOUBLE COMPLEX or COMPLEX*16 (highest rank)»

   2. COMPLEX««*8»»

   3. DOUBLE PRECISION «or REAL*8»

   4. REAL««*4»

   5. INTEGER«*4»

   6. «INTEGER*2»

   7. «INTEGER*1 (lowest rank)»

 For example, when an operation is performed on an INTEGER«*2» operand and a
 REAL«*4» operand, the INTEGER«*2» operand is first converted to REAL«*4».
 The result of the operation is also a value of data type REAL«*4».
 Similarly, in an operation on a real number and a complex number, the real
 number is first converted to a complex number, and the result of the
 operation is also complex.

 The following examples show how expressions are interpreted. The variables
 i1 and i2 are integers, r1 and r2 are single-precision real numbers, and c1
 and c2 are COMPLEX«*8» numbers.

 Statement                                  Interpretation

 r2 = i1/i2*r1                              First, integer division (as
                                            described in Section 1.7.1.1) is
                                            performed on i1 and i2. The
                                            quotient is converted to a real
                                            number and real multiplication is
                                            performed on the resulting
                                            operand and r1.

 c1=c2+i1                                   The integer i1 is first converted
                                            to type COMPLEX«*8», then added
                                            to c2.

 «For expressions with integer operands, the type of the result is controlled
 by the types of the operands and the setting of the $STORAGE metacommand,
 the result is of whichever type is larger. For example, if you declare  j
 and k as INTEGER*2 variables, and use the $STORAGE:4 metacommand, the result
 of the expression j + k is of type INTEGER*4. (The same result occurs if the
 $STORAGE metacommand is omitted, since the default for $STORAGE:n is 4.)»

 «Note that the compiler usually removes higher-precision arithmetic when
 optimizing if it will not affect the result and if $DEBUG is not set. For
 example, the intermediate value obtained when adding j to k is probably
 16-bit, even if $STORAGE:4 were specified as in the following example:»

           «INTEGER*2  i, j, k
            i = j + k»

 «Using $STORAGE:4 does not affect INTEGER*2 expressions which have only the
 plus (+), minus (-), or multiplication (*) operators in them. The results
 are the same because conversion to an INTEGER*4 intermediate result does not
 prevent overflows in INTEGER*2 arithmetic. (Note that such overflows are not
 reported unless the $DEBUG metacommand is in effect.)»

 Table 1.6 shows how arithmetic operands are converted from one data type to
 another.

 Table 1.6  Arithmetic Type Conversion

                            Converting to the          Converting to the

                            Next-Highest-Ranked        Next-Lowest-Ranked

 Data Type                  Data Type                  Data Type

 «DOUBLE COMPLEX or         «(Highest rank.)»          «Convert imaginary and
 COMPLEX*16»                                           real parts,
                                                       individually, from
                                                       REAL*8 to REAL*4».

 COMPLEX«*8»                Convert imaginary and      Delete imaginary part.
                            real parts,
                            individually, from
                            REAL*4 to REAL*8.

 DOUBLE PRECISION           Convert to REAL*4, and     Round off
 «or REAL*8»                add a 0.0 imaginary        least-significant
                            part.                      part.

 REAL«*4»                   Store in the DOUBLE        Truncate.
                            PRECISION format.

 INTEGER«*4»                Add zero fractional        Use least-significant
                            part.                      part.

 «INTEGER*2»                «Use as                    «Use least-significant
                            least-significant          part.»
                            part, and set sign bit
                            in most-significant
                            part.»

 «INTEGER*1»                «Use as                    «(Lowest rank.)»
                            least-significant
                            part, and set sign bit
                            in most-significant
                            part.»


 1.7.2  Character Expressions

 A character expression produces a value that is of type character. There are
 six basic operands used in character expressions:

   1. Character constants

   2. Character variable references

   3. Character array-element references

   4. Character function references

   5. Character substrings

   6. «Character structure-element references»

 The concatenation operator (//) is the only character operator. It is used
 as follows:

      first//second

 This produces a character string consisting of the value of first
 concatenated on the right with the value of second and whose length is the
 sum of the lengths of first and second. For example, the following
 expression produces the string 'AB  CDE':

      'AB  ' // 'CDE'

 When two or more string variables are concatenated, the resulting string is
 as long as the declared lengths of the string variables. Leading and
 trailing blanks are not discarded. For example:

            CHARACTER*10  first
            CHARACTER*6  second
            first = 'heaven'
            second = ' sent'
            WRITE (*, *) first//second

 The result is the 16-character string 'heaven     sent '. Note that
 there is a total of five spaces between 'heaven' and 'sent'.

 «If you concatenate C strings, remember a null character (\0) is
 automatically appended to each C string. For example, the expression»

      «'hello 'C // 'world'C»

 «is equivalent to the following C string:»

      «'hello \0world'C»


 1.7.3  Relational Expressions

 Relational expressions compare the values of two arithmetic or character
 expressions. You cannot compare an arithmetic variable with a character
 variable.

 «In Microsoft FORTRAN, an arithmetic expression can be compared with a
 character expression. The arithmetic expression is treated as if it were a
 character expression (that is, a sequence of byte values). The two
 expressions must be identical on a byte-by-byte basis, or they are not
 equal.»

 «For example, if 'A' were assigned to a four-byte integer, the ASCII value
 of the letter A (hex 41) would be the variable's least-significant byte, and
 the other bytes would be zeros. If 'A' were assigned to a character variable
 four characters long, the ASCII value of the letter A (hex 41) would be the
 variable's most-significant byte because character variables are
 left-justified. Therefore, the two variables would not be equal, even though
 they held the same nominal value.»

 A relational expression produces a result of type LOGICAL (.TRUE. or
 .FALSE.). Relational expressions can use any of the operators shown in Table
 1.7 to compare values.

 Table 1.7  Relational Operators

 Operator      Operation
 .LT.          Less than
 .LE.          Less than or equal to
 .EQ.          Equal to
 .NE.          Not equal to
 .GT.          Greater than
 .GE.          Greater than or equal to

 All relational operators are binary operators and appear between their
 operands. A relational expression cannot contain another relational
 expression, so there is no relative precedence or associativity among the
 relational operands. The following program fragment is therefore invalid:

            REAL*4  a, b, c, d
            IF ((a .LT. b) .NE. c)  d = 12.0

 Assume that a is less than b. After the first part of the expression is
 evaluated, the expression is

      .TRUE. .NE. c

 However, c is an arithmetic expression, and you cannot compare an
 arithmetic expression to .TRUE.. To compare relational expressions and
 logical values, use the logical operators (as discussed in Section 1.7.4).

 Relational expressions with arithmetic operands may have one operand that is
 an integer and one that is a real number. In this case, the integer operand
 is converted to a real number before the relational expression is evaluated.
 You can also have a complex operand, in which case the other operand is
 first converted to complex. However, you can use only the .NE. and .EQ.
 operators with complex operands.

 Relational expressions with character operands compare the position of their
 operands in the ASCII collating sequence. One operand is less than another
 if it appears earlier in the collating sequence. For example, the expression
 ('apple'.LT.'banana') returns the value .TRUE., and the expression
 ('Keith' .GE. 'Susan') returns the value .FALSE.. If operands of unequal
 length are compared, the shorter operand is extended to the length of the
 longer operand by the addition of blanks on the right.


 1.7.4  Logical Expressions

 A logical expression produces a logical value. There are seven basic
 operands used in logical expressions:

   1. Logical constants

   2. Logical variable references

   3. Logical array-element references

   4. Logical function references

   5. Relational expressions

   6. Integer constants or variables

   7. «Logical structure-element references»

 Other logical expressions are constructed from the basic operands in the
 preceding list by using parentheses and the logical operators of Table 1.8.

 Table 1.8  Logical Operators

 Operator      Operation                   Precedence

 .NOT.         Negation                    1 (highest)
 .AND.         Conjunction                 2
 .OR.          Inclusive disjunction       3
 «.XOR.»       «Exclusive disjunction»     «4»
 .EQV.         Equivalence                 4
 .NEQV.        Nonequivalence              4

 The .AND., .OR., «.XOR.,» .EQV., and .NEQV. operators are binary operators
 and appear between their logical expression operands. The .NOT. operator is
 unary and precedes its operand. If switch is .TRUE., then (.NOT. switch) is
 .FALSE..

 Logical operators allow only arguments of the LOGICAL type. «Microsoft
 FORTRAN also permits integer arguments, which may be integer constants,
 integer variables, integer structure elements, or integer expressions.
 Operations are "bitwise." For example, the expression k .XOR. m performs an
 "exclusive-or" comparison on matching bits in the operands, and sets or
 clears the corresponding bit in the integer value it returns. If both
 operands are not of the same integer type, the lower-precision operand is
 converted to the higher-precision type.»

 «Note that the result of comparing two integer expressions with a logical
 operator is of INTEGER type, not LOGICAL.»

 When two consecutive operations are of equal precedence, the leftmost
 operation is performed first.

 Two .NOT. operators cannot be adjacent, but the .NOT. operator can appear
 next to any of the other logical operators. The following statement, for
 example, is allowed:

            logvar = a .AND. .NOT. b

 Logical operators have the same meaning as in standard mathematical
 semantics; the .OR. operator is nonexclusive. For example,

            .TRUE. .OR. .TRUE.

 evaluates to .TRUE.. Table 1.9 shows the values of logical expressions.

 Table 1.9  Values of Logical Expressions

               Then These Expressions Evaluate as Follows:

 If Operands                                                 «a .XOR. b» or
 a and b Are:      a .AND. b      a .OR. b     a .EQV. b      a .NEQV. b

 Both true         True           True         True           False

 One true,
 one false         False          True         False          True

 Both false        False          False        True           False


 Examples

 The following example demonstrates precedence in logical expressions:

            LOGICAL  stop, go, wait, a, b, c, d, e

      C     The following two statements are equivalent:
            stop =  a .AND. b  .AND. c
            stop = (a .AND. b) .AND. c

      C     The following two statements are equivalent:
            go =   .NOT. a  .OR. b  .AND. c
            go = ((.NOT. a) .OR. b) .AND. c

      C     The following two statements are equivalent:
            wait =   .NOT. a  .EQV.  b .OR. c   .NEQV.  d .AND. e
            wait = ((.NOT. a) .EQV. (b .OR. c)) .NEQV. (d .AND. e)

 «The following example demonstrates the use of integers in logical
 expressions to perform byte masking:»

           «INTEGER*2  lowerbyte, dataval, mask
            mask = #00FF       ! mask the most-significant byte
            dataval = #1234
            lowerbyte = (dataval .AND. mask)
            WRITE (*, '('' '', 2Z4)') dataval, lowerbyte»

 «The  output is as follows:»

      «1234  34»


 1.7.5  «Array Expressions»

 «Microsoft FORTRAN permits operations on full arrays that would normally
 only work with single arguments (scalars). For example, two arrays may be
 added element-by-element using only the addition operator (+). Every element
 in an array can be divided by a constant value. Two arrays of LOGICAL or
 INTEGER variables can be compared with logical operators such as .AND. or
 .GE., and so on. Arrays can also be passed to functions (both intrinsic and
 external), with the function operating on each element and returning an
 array of the results.»

 «When two or more array operands appear in an assignment statement or
 expression, they all must "conform." That is, they must have the same number
 of dimensions, and corresponding dimensions must be the same size and have
 the same upper and lower boundaries. For example, the arrays first(6,3) and
 second(3,2,3) do not conform. Although they take up the same amount of
 memory, their dimensions do not match. The arrays third(3,4) and
 fourth(-1:1,5:8) do not conform either. They have the same number of
 dimensions, and each is of the same size, but the dimension boundaries do
 not match. Any attempt to combine non-conforming arrays in an arithmetic
 expression, or assign one to another, results in a compile-time error. Note
 that a scalar quantity conforms to any array.»

 «If arrays are to conform, the sizes of their dimensions must be fully
 specified at compile time. Therefore, adjustable and allocatable arrays do
 not conform; they cannot be combined in expressions either with themselves
 or fixed arrays, even if their dimensions match.»

 «All operations that are permitted for scalar arguments are permitted for
 array arguments. An array expression may be the right-hand term of an
 assignment statement whose left-hand term is an array. Array expressions may
 only appear in assignment statements and as function arguments. For example,
 if array1, array2, array3, and logarray are conforming arrays, then all the
 following statements are legal:»

      «array1 = 42                    !Assign constant to each element.
      array3 = array 1 / array2       !Divide corresponding element.
      array2 = - array1               !Negate each array1 element.
      array1 = array2 - 3             !Subtract constant from each element.
      array1 = array2 - array3(7)     !Subtract value of array3(7) from each e
      array2 = array3 ** array1       !Raise each element to power of
                                      !corresponding element.
      array3 = MyFunc (array2)        !Apply MyFunc to each element.
      logarray = array1 .LT. array2   !Logically compare corresponding element
      logarray = .NOT. array2         !Invert bits of each array2 element.
      chararray2 = chararray2(5:8)//chararray2(1:4) !Rearrange character
                                                    !string.»


 «All intrinsic functions that take scalar arguments may also take array
 arguments. If you wish to pass array arguments to external functions that
 take scalar arguments, you must declare the external function in an
 INTERFACE TO statement. None of the function's formal arguments may be
 arrays.»

 ───────────────────────────────────────────────────────────────────────────
 «NOTE»
    «In processing array expressions, the compiler may generate a less
    efficient sequence of machine instructions than it would if the arrays
    were processed in a conventional DO loop. If execution speed is critical,
    it may be more efficient to handle arrays element-by-element.»
 ───────────────────────────────────────────────────────────────────────────


 1.7.6  Precedence of Operators

 When arithmetic, character, relational, and logical operators appear in the
 same expression, precedence is as follows:

   1. Arithmetic operators have the highest precedence.

   2. Character operators are evaluated next.

   3. Relational operators are evaluated next.

   4. Logical operators have the lowest precedence.


 Chapter 2  Program Structure
 ───────────────────────────────────────────────────────────────────────────

 This chapter explains the structure of FORTRAN programs: what can be
 included and in what sequence. Information on line types and statement
 labels is provided, and restrictions on the order of statements and
 metacommands are discussed. This chapter also includes information on
 arguments and on the program units processed by the FORTRAN compiler: main
 programs, subroutines, functions, and block-data subprograms.


 2.1  Lines

 The position of characters within FORTRAN lines is significant. The
 following list shows how column position determines character
 interpretation:

 Column                                  Character Interpretation

 1-5                                     Statement label. «A dollar sign
                                         ($) in column 1 indicates a
                                         metacommand.» An asterisk or an
                                         uppercase «or lowercase» C in
                                         column 1 indicates a comment
                                         line.

 6                                       Continuation character.

 7-72                                    FORTRAN statement.

 73 and above                            Ignored.

 Lines shorter than 72 characters are padded with blanks.

 There are five kinds of lines in Microsoft FORTRAN:

 Type                                    Description

 «Metacommand lines»                     «A metacommand line has a dollar
                                         sign ($) in column 1.»

                                         «Metacommands control operation
                                         of the Microsoft FORTRAN
                                         Compiler. (See Chapter 6 for more
                                         information.) The following lines
                                         are metacommands:»

                                        «$DEBUG:'pdq'
                                         $DECLARE
                                         $LINESIZE:132»

 Initial lines                           The first (or only) line of a
                                         FORTRAN statement is called the
                                         "initial line." An initial line
                                         has either a blank or a zero in
                                         column 6, and has either blanks
                                         or a statement label in columns 1
                                         through 5. (The only exception to
                                         this rule is the statement which
                                         follows a logical IF statement.)
                                         The following are initial lines:

                                            GOTO 100
                                         0002  CHARACTER*10 name
                                         100  0CONTINUE
                                           1000STOP ' '

 Continuation lines                      A continuation line has blanks in
                                         columns 1 through 5 and a
                                         character (other than a blank or
                                         a zero) in column 6. A
                                         continuation line increases the
                                         room available to write a
                                         statement. «A statement may be
                                         extended to include as many
                                         continuation lines as memory
                                         allows.» The compiler issues a
                                         warning if a statement has more
                                         than 19 continuation lines. The
                                         second line below is a
                                         continuation line «(this
                                         limitation does not apply to
                                         Microsoft FORTRAN unless the
                                         $STRICT metacommand is in
                                         effect)»:

                                         C   Sample Continuation line
                                             INTEGER*4  count, popu, local,
                                            +       ovrflo, incrs, provnc

 Comment lines                           A comment line generally has an
                                         uppercase «or lowercase» C or an
                                         asterisk (*) in column 1 or is
                                         entirely blank. «An exclamation
                                         point appearing as the first
                                         nonblank character also indicates
                                         a comment line in Microsoft
                                         FORTRAN. (A line with an
                                         exclamation point in column 6 is
                                         interpreted as a continuation
                                         line.)»

                                         Comment lines do not affect the
                                         execution of the FORTRAN program
                                         in any way. Comment lines can
                                         appear within statements that
                                         have continuation lines. «Debug
                                         lines (described below) are
                                         sometimes treated as comment
                                         lines. For more information, see
                                         the $DEBUG and $NODEBUG entry in
                                         Section 6.2, "Metacommand
                                         Directory." The following are
                                         examples of comment lines:»

                                         C     This is a comment line,
                                         *            and so is this,
                                              «!    as is this line.»

                                         «Comments can also be added to
                                         program lines, following the
                                         statement. In this case, the
                                         comment must begin with an
                                         exclamation point (!).»

                                     «hyp = SQR (a**2 + b**2) ! hypotenuse»

 «Debug lines»                           «In Microsoft FORTRAN, any line
                                         with an uppercase or lowercase
                                         letter in the first column is
                                         considered a comment line. When a
                                         letter (except C or c) is
                                         specified in a $DEBUG
                                         metacommand, that letter is
                                         removed from all succeeding lines
                                         that include it in column 1, and
                                         the remainder of the line is
                                         compiled. The case of the letter
                                         (either in the $DEBUG metacommand
                                         or the program) is not
                                         significant. See the $DEBUG and
                                         $NODEBUG entry in Section 6.2 for
                                         more information on debug lines.
                                         The following lines are examples
                                         of debug lines:»

                                        «B    RETURN 1
                                         z    WRITE(*, *) count»


 2.2  Statement Labels

 Any statement can start with a label; however, only the labels of executable
 or FORMAT statements can be referenced. A statement label is a sequence of
 one to five digits, at least one of which must be nonzero. A label may be
 placed anywhere in columns 1 through 5 of an initial line, and blanks are
 ignored. (For example, the labels 23 4 and 234 are identical.) A label may
 not be repeated.


 2.3  «Free-Form Source Code»

 «The $FREEFORM metacommand lets you enter source code without most of the
 restrictions imposed by the conventional FORTRAN format. Most of the rules
 in Section 2.1, "Lines," do not apply to the free-form format. See the
 $FREEFORM and $NOFREEFORM entry in Section 6.2, "Metacommand Directory," for
 more information on free-form format.»

 «The following rules define the free-form format:»

   ■  «A double quotation mark ( " ) in column 1, or an exclamation point as
      the first nonblank character, indicates a comment line.»

   ■  «Initial lines may start in any column.»

   ■  «An initial line may begin with a statement label of from one to five
      digits (blanks are ignored). Statement labels need not be followed by
      blanks to separate them from the statement; the first nonblank,
      nonnumeric character is the beginning of the statement.»  p

   ■  «If the last nonblank character of a line is a minus sign, it is
      discarded and the next line is considered a continuation line. The
      continuation line may start in any column.»

   ■  «Alphabetic characters and asterisks are not allowed as comment markers
      in column 1.»


 2.4  Order of Statements «and Metacommands»

 Statements describe, specify, and classify the elements of your program, as
 well as the actions your program will take. Chapter 4, "Statements," defines
 each Microsoft FORTRAN statement.

 The ANSI standard for FORTRAN prescribes the ordering of the statements and
 lines in a FORTRAN program unit. «Some of Microsoft FORTRAN's extensions
 have additional requirements.» Figure 2.1 shows which statements «and
 metacommands» must precede and which must follow any specific statement «or
 metacommand».

 Suppose your program contains program elements a and b. In
 Figure 2.1, if the box containing element a is above the box containing
 element b, then a must appear before b in your program. An IMPLICIT
 statement, for example, must appear before a COMMON, DATA, or END statement,
 and so on.

 If, in Figure 2.1, a is in a box that is to the left or right of b, then a
 and b can appear in any order relative to each other. FORMAT, for example,
 «is to the left of the box containing most of the metacommands, and» is to
 the right of the boxes with DATA, PARAMETER, IMPLICIT, COMMON, and
 statement-function statements, and so on. Any of those elements can appear
 before or after a FORMAT statement.

 The following rules summarize the required order of statements «and
 metacommands» shown in Figure 2.1:

   ■  Every program unit must have an END statement as its last line.

   ■  Comment lines can appear anywhere, except after the last END statement
      in a source file.

   ■  The BLOCK DATA, FUNCTION, «INTERFACE TO,» PROGRAM, and SUBROUTINE
      statements must precede all other statements. They do not have to
      precede metacommands.

   ■  All specification statements must precede all DATA statements,
      statement-function statements, and executable statements. See Section
      4.1, "Categories of Statements," for listings of specification
      statements and executable statements.

   ■  IMPLICIT statements must precede other specification statements, with
      the exception of the PARAMETER statement.

   ■  Statement-function statements must precede executable statements.

   ■  When a specification statement defines the type of a constant to be
      used in the PARAMETER statement, the PARAMETER statement must follow
      that specification statement. The PARAMETER statement must precede all
      other specification statements that use the symbolic constants it
      defines.

   ■  «INTERFACE TO statements must precede references to the subprograms
      they define.»

   ■  «The $DO66, $«NO»FLOATCALLS, $«NO»FREEFORM, and $STORAGE
      metacommands, if present, must appear before anything else. $LARGE and
      $NOTLARGE, when used without arguments, cannot appear within the
      executable-statement section. $LARGE and $NOTLARGE, when used with
      arguments, must appear in the declarations section. Other metacommands
      can appear anywhere.»

   ■  Block-data subprograms may not contain statement function statements,
      FORMAT statements, or executable statements.


 Figure 2.1  Order of Statements «and Metacommands»

 ┌────────────────────────────────────────────────────────┬────────────────┐
 │  «$DO66, $«NO»FLOATCALLS, $«NO»FREEFORM, $STORAGE»                      │
 ├────────────────────────────────────────────────────────┤                │
 │ BLOCK DATA, FUNCTION, «INTERFACE,» PROGRAM, SUBROUTINE                  │
 ├────────────────┬───────────────────┬──────────┬────────┤                │
 │                      IMPLICIT                                           │
 │                ├───────────────────┤          │        │                │
 │                 COMMON,                                 «$«NO»DEBUG,»   │
 │                │DIMENSION,         │          │        │«$«NO»DECLARE,» │
 │                 EQUIVALENCE,                            «$INCLUDE,»     │
 │«$[NOT]LARGE,   │EXTERNAL,          │PARAMETER │        │«$LINESIZE,»    │
 │when used        INTRINSIC,                              «$«NO»LIST,»    │
 │without         │SAVE; type         │          │        │«$MESSAGE,»     │
 │arguments»       statements; «also                       «$PAGE,»        │
 │                │$[NOT]LARGE,       │          │        │«$PAGESIZE,»    │
 │                 when used with                 ENTRY,   «$«NOT»STRICT,» │
 │                │arguments          │          │FORMAT  │«$SUBTITLE,»    │
 │                 ─────────────────── ──────────          «$TITLE,»       │
 │                │Statement-function │          │        │«$«NO»TRUNCATE» │
 │                 statments                                               │
 ├────────────────┴───────────────────┤  DATA    │        │                │
 │     Executable statements                              │                │
 ├────────────────────────────────────┴──────────┴────────┴────────────────┤
 │                                   END                                   │
 └─────────────────────────────────────────────────────────────────────────┘


 2.5  Arguments

 "Arguments" are the values passed to and from functions and subroutines. A
 "formal argument" is the name by which an argument is known within a
 function or subroutine. The "actual argument" is the specific variable,
 expression, array, function name, or other item passed to a subroutine or
 function when it is called.

 The number of actual arguments must be the same as the number of formal
 arguments «(unless the VARYING attribute is specified)», and the
 corresponding types must agree. If a procedure is called more than once in a
 program unit, the compiler checks that the number and types of actual
 arguments are the same in each call (consistency). If a procedure was
 defined prior to its first use, «or in an INTERFACE TO statement,» the
 compiler also checks that the number and types of the actual arguments match
 the number and types of the formal arguments (validity).

 Arguments normally pass values into and out of subroutines or functions by
 reference (they pass the memory address of the argument). «In Microsoft
 «FORTRAN, this is the default. You can use the VALUE attribute (described in
 Section 1.6.11) to pass arguments by value.»

 Upon entry to a subroutine or function, the actual arguments are associated
 with the formal arguments. This association remains in effect until the
 subroutine or function terminates execution. If the actual argument has been
 passed by reference «(the default)», assigning a value to a formal argument
 during execution of a subroutine or function alters the value of the
 corresponding actual argument.

 If an actual argument is a constant, a function reference, or an expression
 other than a single variable, assigning a value to the corresponding formal
 argument is not permitted and has unpredictable results. In the following
 program, for example, the actual argument header is a constant and
 corresponds to the formal argument title. In the subroutine report, a value
 is assigned to title.

      C     This program is incorrect and has unpredictable
      C     results:
            CHARACTER*20  header
            REAL*4  grav
            PARAMETER  (header = 'Specific Gravity')
            DATA  grav / 2.8327 /
            WRITE (*, *) header, grav
      C     Header is an actual argument:
            CALL Report (header, grav)
            WRITE (*, *) header, grav
            END
      C     The formal argument corresponding to header is title:
            SUBROUTINE  Report (title, data)
            CHARACTER*20  title
            REAL*4  data
      C     The following statement is illegal; it assigns a value
      C     to a formal argument that corresponds to a constant:
            title = 'Density (kg/cubic m)'
            WRITE (*, *) title, data
            END

 The output of the above program is unpredictable. To change the value of
 title in the subroutine, header should have been a variable.

 If an actual argument is an expression, it is evaluated before the
 association of formal and actual arguments. If an actual argument is an
 array element, its subscript expressions are evaluated before the
 association. The subscript expressions remain constant throughout the
 execution of the subroutine or function, even if they contain variables that
 receive new values during the execution of the subroutine or function. The
 following list shows how actual and formal arguments can be associated:

 Actual Argument                            Formal Argument

 A variable, an array element,              Variable name.
 «a structure element,» or an
 expression.

 An array or an array element. The          Array name.
 number and size of dimensions in
 a formal argument may be
 different from those of the
 actual argument, but any
 reference to the formal array
 must be within the limits of the
 memory sequence in the actual
 array. A reference to an
 out-of-bounds element is not
 detected as an error, and has
 unpredictable results.

 «An array. Each element of the             «A variable. Its procedure must
 array is passed to the procedure,          be declared in an INTERFACE TO
 one element at a time, and the             statement, or it must be an
 procedure is executed once for             intrinsic function, if arrays are
 each element.»                             to be passed to a scalar formal
                                            argument.»

 An alternate-return specifier              An asterisk (*) .
 (*label) in the CALL statement.
 The same label can be used in
 more than one alternate-return
 specifier.

 The name of an external                    Any unique name which is used in
 subroutine or function, or                 a subroutine call or function
 intrinsic function. The actual             reference within the procedure.
 argument must be an external
 subroutine or function, declared
 with the EXTERNAL statement, or
 an intrinsic function permitted
 to be associated with a formal
 subroutine argument or function
 argument. The intrinsic function
 must first be declared with an
 INTRINSIC statement in the
 program unit where it is used as
 an actual argument.

 The following intrinsic functions may not be associated with formal
 subroutine arguments or function arguments:

     «ALLOCATED»
      AMAX0
      AMAX1
      AMIN0
      AMIN1
      CHAR
      CMPLX
      DBLE
     «DCMPLX»
      DFLOAT
      DMAX1
      DMIN1
     «DREAL
      EOF»
     «EPSILON»
      FLOAT
     «HFIX»
     «HUGE»
      ICHAR
      IDINT
      IFIX
     «IMAG
      INT1
      INT2
      INT4
      INTC»
      INT
     «JFIX»
     «LEN_TRIM»
      LGE
      LGT
      LLE
      LLT
     «LOC
      LOCFAR
      LOCNEAR»
      LOG
      LOG10
      MAX
      MAX0
      MAX1
     «MAXEXPONENT»
      MIN
      MIN0
      MIN1
     «MINEXPONENT»
     «NEAREST»
     «PRECISION»
      REAL
     «SCAN»
      SNGL
     «TINY»
     «VERIFY»

 «When passing integer arguments by reference, an INTEGER*2 variable cannot
 be passed to an INTEGER*4 formal argument, and an INTEGER*4 variable cannot
 be passed to an INTEGER*2 formal argument. You must convert the data type
 using the intrinsic functions INT4 or INT2 (described in Chapter 5,
 "Intrinsic Functions and Additional Procedures"). Note that the result of
 the conversion is stored at a temporary memory location, so the subroutine
 can no longer assign a new value to the actual argument. Also note that when
 $STORAGE:4 (the default) is in effect, even expressions with only INTEGER*2
 arguments have a result with type INTEGER*4. The following program, for
 example, results in an error:»

      «C     This is incorrect and produces an error:
      $STORAGE:4
            INTEGER*2 j, k
            CALL Subr(j + k)

            SUBROUTINE Subr (n)
            INTEGER*2 n
            .
            .
            .
            END»

 «An error occurs because the result of the expression j+k is of the type
 INTEGER*4. You must write the subroutine call as»

           «CALL Subr (INT2(j + k))»

 «Integer arguments passed by value are not subject to the same restrictions.
 The conversion rules for value arguments are the same as the conversion
 rules described in Table 1.6, "Arithmetic Type Conversion." For example, you
 can pass a real value to an integer argument because the conversion is
 performed automatically.»


 2.6  Program Units

 The FORTRAN compiler processes program units. A program unit can be a main
 program, a subroutine, a function, or a block-data subprogram. You can
 compile any of these units separately and link them together later. It is
 not necessary to compile or recompile them as a whole. The following list
 summarizes the four types of program units (discussed in Sections 2.7
 through 2.10):

 Program Unit                               Description

 Main program                               Any program unit that does not
                                            have a SUBROUTINE, FUNCTION, or
                                            BLOCK DATA statement as its first
                                            statement. A main program can
                                            have a PROGRAM statement as its
                                            first statement, but this is not
                                            required.

 Subroutine                                 A program unit that is called
                                            from other program units with a
                                            CALL statement.

 Block-data subprogram                      A program unit that provides
                                            initial values for variables in
                                            named common blocks.

 Function                                   A program unit that can be
                                            referred to in an expression.

 The PROGRAM, SUBROUTINE, BLOCK DATA, FUNCTION, and Statement Function
 statements are described in detail in Section 4.2, "Statement Directory."
 Related information is provided in the entries for the CALL and RETURN
 statements.

 Subprograms make it easier to develop large, well-structured programs,
 especially in the following situations:

 Situation                                  Result

 You have a large program                   You can more easily develop,
                                            test, maintain, and compile a
                                            large program when it is divided
                                            into parts.

 You intend to include certain              You can create object files that
 subprograms in more than one               contain these subprograms and
 program                                    link them to the programs in
                                            which they are used.

 You anticipate altering a                  You can place the procedure in
 procedure's implementation                 its own file and compile it
                                            separately. You can change the
                                            procedure «or even rewrite it in
                                            assembly language, Microsoft
                                            Pascal, or Microsoft C». The rest
                                            of your program does not need to
                                            change.


 2.7  Main Programs

 A main program is any program unit that does not have a FUNCTION,
 SUBROUTINE, or BLOCK DATA statement as its first statement. The first
 statement of a main program may be a PROGRAM statement. Main programs are
 always assigned the global name _main. The name _main should not be used for
 anything else in a program. The name "main" is actually permitted as a
 common-block name, as a local variable in a subprogram outside the main
 program, or as a subprogram name in a module that does not contain a PROGRAM
 statement and is not referenced by a module that contains the main program.
 However, the use of "main" is likely to cause confusion and is not
 recommended. If the main program has a PROGRAM statement, the name specified
 in the PROGRAM statement is assigned in addition to the name _main.

 Program execution always begins with the first executable statement in the
 main program, so there must be exactly one main program unit in every
 executable program.

 For further information about programs, see the PROGRAM entry in Section
 4.2, "Statement Directory."


 2.8  Subroutines

 A subroutine is a program unit that can be called from other program units
 with a CALL statement. When invoked, a subroutine performs the set of
 actions defined by its executable statements. The subroutine then returns
 control to the statement immediately following the one that called it, or to
 a statement specified as an alternate return. See the CALL entry in Section
 4.2, "Statement Directory," for more information.

 A subroutine does not directly return a value. However, values can be passed
 back to the calling program unit through arguments or common variables.

 See the SUBROUTINE entry in Section 4.2, "Statement Directory," for further
 information.


 2.9  Block-Data Subprograms

 A block-data subprogram is a program unit that defines initial values for
 variables in named common blocks.

 Variables are normally initialized with DATA statements. Variables in
 named-common blocks can be initialized only in block-data subprograms.
 Variables in the blank-common block cannot be initialized in block-data
 subprograms. See the BLOCK DATA entry in Section 4.2, "Statement Directory,"
 for more information.


 2.10  Functions

 A function is referred to in an expression and returns a value that is used
 in the computation of that expression. Functions can also return values
 through arguments and common variables. There are three kinds of functions:

   1. External functions

   2. Statement functions

   3. Intrinsic functions

 External and statement functions are described in more detail below.
 Intrinsic functions are described in Chapter 5, "Intrinsic Functions and
 Additional Procedures".

 An arithmetic, character, or logical expression may reference a function.
 Executing the function reference evaluates the function, and the returned
 value is used as an operand in the expression with the function reference.
 Note that the length of a character function must be specified by an
 integer-constant expression.

 The syntax of a function reference is as follows:

      fname («argument-list»)

 Parameter                                  Value

 fname                                      Name of an external, intrinsic,
                                            or statement function.

 argument-list                              Actual arguments. If more than
                                            one argument is given, they must
                                            be separated by commas.

 Function arguments follow the same rules as those for subroutines (except
 that alternate returns are not allowed); these are described in the CALL
 entry in Section 4.2, "Statement Directory." Some additional restrictions
 specific to statement functions and intrinsic functions are described in
 Section 2.10.2, "Statement Functions," and Chapter 5, "Intrinsic Functions
 and Additional Procedures," respectively.


 2.10.1  External Functions

 External functions are user-defined functions (as opposed to FORTRAN's
 intrinsic functions). They may be included with the main program source code
 or in a separately compiled unit. They begin with a FUNCTION statement and
 conclude with an END statement.


 Example

      C     The following external function calculates
      C     the Simpson approximation of a definite integral:

            INTEGER FUNCTION  Simpson (delx, steps, y)
            INTEGER  delx, y(100), sum, steps, factor
            sum = 0
            DO 100  i = 0, steps
               IF ((i .EQ. 0) .OR. (i .EQ. steps)) THEN
                  factor = 1
               ELSE IF (MOD (i, 2) .EQ. 0) THEN
                  factor = 2
               ELSE
                  factor = 4
               END IF
               sum = factor * y(i) + sum
      100   CONTINUE
            Simpson = INT ((REAL (delx) /3.0) * REAL (sum))
            END


 2.10.2  Statement Functions

 A statement function is defined by a single statement and is similar in form
 to an assignment statement. The body of a statement function defines the
 statement function.

 A statement function must be the first nondeclaration statement in a program
 unit, but it is not executed at that point; it is a nonexecutable statement.
 It is executed, like other functions, by referencing it in an expression.

 For information on the syntax and use of a statement function statement, see
 the Statement Function entry in Section 4.2, "Statement Directory."


 Chapter 3  The Input/Output System
 ───────────────────────────────────────────────────────────────────────────

 This chapter explains FORTRAN's input/output (I/O) system. The chapter
 introduces FORTRAN's file types and summarizes the I/O statements available
 to manipulate data within them. Additional sections describe options
 available in FORTRAN I/O statements and the use of  control characters,
 formatted I/O, and edit descriptors. List-directed and namelist-directed I/O
 are also discussed.


 3.1  The FORTRAN I/O System

 In FORTRAN's I/O system, data is stored in files and can be transferred
 between files. There are two basic types of files:

 Type                                       Description

 External files                             A device, such as the screen, the
                                            keyboard, or a printer, or a file
                                            that is stored on a device, such
                                            as a file on a disk.

 Internal files                             A character variable, character
                                            array, character array element,
                                            «character structure element,»
                                            character substring, «or
                                            noncharacter array». For
                                            information on internal files,
                                            see Section 3.5.

 All files consist of records, which are sequences of characters or values.
 See Microsoft FORTRAN Advanced Topics for information on the format of
 records.

 Input is the transfer of data from a file to internal storage. Output is the
 transfer of data from internal storage to a file. You input data by reading
 from a file, and output data by writing to a file.


 3.2  I/O Statements

 I/O statements transfer data, manipulate files, or determine the properties
 of the connections to files. Table 3.1 lists the I/O statements.

 Table 3.1  I/O Statements

 Statement                  Function

 BACKSPACE                  Positions a file back one record
 CLOSE                      Disconnects a unit
 ENDFILE                    Writes an end-of-file record

 INQUIRE                    Determines properties of a unit or named file

 «LOCKING»                  «Controls access to information in specific
                            direct-access files and/or records»

 OPEN                       Associates a unit number with a file

 PRINT                      Outputs data to the asterisk (*) unit
 READ                       Inputs data

 REWIND                     Repositions a file to the beginning
 WRITE                      Outputs data

 ───────────────────────────────────────────────────────────────────────────
 «NOTE»
    «In addition to the I/O statements, Microsoft FORTRAN includes an I/O
    intrinsic function, EOF (unitspec), which is described in Section 5.1.14,
    "End-of-File Function." EOF returns a logical value that indicates
    whether there is data remaining in the file after the current position.»
 ───────────────────────────────────────────────────────────────────────────

 In I/O statements, you can specify a series of options. The following, for
 example, is the syntax of the CLOSE statement:

      CLOSE («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»
      «, STATUS=status»)

 The three options of the CLOSE statement are «, ERR=errlabel», «, IOSTAT=
 iocheck», and «, STATUS=status». (Although the UNIT= descriptor is
 optional, the unitspec itself is not.)

 The ERR= and IOSTAT= options control error handling, and the STATUS= option
 specifies whether to keep or delete the file after disconnecting. The
 following statement closes the file connected to unit 2, transfers execution
 to the statement labeled 100 if there is an I/O error, and uses the default
 value of the STATUS= option:

            CLOSE (UNIT = 2, ERR = 100)

 Options used in only one statement are described in Section 4.2, "Statement
 Directory." Fourteen of the options in I/O statements, however, are used in
 more than one I/O statement. Table 3.2 lists these options, the I/O
 statements they are used in, and the section that describes each.

 Table 3.2  I/O Options

 Option                   I/O Statements                  Section

 ACCESS=access            INQUIRE, OPEN                   3.2.3, "File
                                                          Access Method"

 «BLOCKSIZE=blocksize»    «INQUIRE, OPEN»                 «3.2.4,
                                                          "Input/Output
                                                          Buffer Size"»

 editlist                 FORMAT, PRINT, READ, WRITE      3.2.5, "The Edit
                                                          List"

 ERR=errlabel             All except PRINT                3.2.6, "Error
                                                          and End-of-File
                                                          Handling"

 FILE=file                INQUIRE, OPEN                   3.2.1, "File
                                                          Names"

 «FMT=»formatspec         PRINT, READ, WRITE              3.2.7, "Format
                                                          Specifier"

 FORM=form                INQUIRE, OPEN                   3.2.9, "File
                                                          Structure"

 iolist                   PRINT, READ, WRITE              3.2.10,
                                                          "Input/Output
                                                          List"

 IOSTAT=iocheck           All except PRINT                3.2.6, "Error
                                                          and End-of-File
                                                          Handling"

 «MODE=mode»              «INQUIRE, OPEN»                 «3.2.11, "File
                                                          Sharing"»

 ««NML=»nmlspec»          «PRINT, READ, WRITE»            «3.2.8,
                                                          "Namelist
                                                          Specifier"»

 REC=rec                  LOCKING, READ, WRITE            3.2.12, "Record
                                                          Number"

 «SHARE=share»            «INQUIRE, OPEN»                 «3.2.11, "File
                                                          Sharing"»

 «UNIT=»unitspec          All except PRINT                3.2.2, "Units"

 The following list briefly summarizes the options listed in Table 3.2.
 Sections 3.2.1-3.2.12 provide a complete discussion of these options. The
 FILE=, UNIT=, and ACCESS= options are presented in the order you must use
 them, followed by the other options listed alphabetically.

 Option                                     Description

 ACCESS                                     Specifies the method of file
                                            access, which can be
                                            'SEQUENTIAL', 'DIRECT', «or
                                            'APPEND'». Returns the method of
                                            file access, which can be
                                            'SEQUENTIAL' or 'DIRECT'.

 «BLOCKSIZE»                                «Specifies or returns the
                                            internal buffer size used in
                                            I/O.»

 editlist                                   Lists edit descriptors. Used in
                                            FORMAT statements and format
                                            specifiers (the FMT=formatspec
                                            option) to describe the format of
                                            data.

 END                                        Transfers control to the
                                            statement with the label in the
                                            END= option when end-of-file is
                                            encountered.

 ERR                                        Controls I/O error handling.
                                            Specifies the label of an
                                            executable statement where
                                            execution is transferred after an
                                            error.

 IOSTAT                                     Controls I/O error handling.
                                            Specifies a variable whose value
                                            indicates whether an error has
                                            occurred.

 FILE                                       Specifies the name of a file.

 FMT                                        Specifies an editlist to use to
                                            format data.

 FORM                                       Specifies a file's format, which
                                            can be either  'FORMATTED',
                                            'UNFORMATTED'«, or 'BINARY'».

 iolist                                     Specifies items to be input or
                                            output.

 «MODE»                                     «Controls how other processes can
                                            access a file on networked
                                            systems. The MODE= option can be
                                            set to 'READWRITE', 'READ', or
                                            'WRITE'.»

 «NML»                                      «Specifies a namelist group to be
                                            input or output.»

 REC                                        Specifies the first (or only)
                                            record of a file to be locked,
                                            read from, or written to.

 «SHARE»                                    «Controls how other processes can
                                            simultaneously access a file on
                                            networked systems. The SHARE=
                                            option can be set to 'COMPAT',
                                            'DENYNONE', 'DENYWR', 'DENYRD',
                                            or 'DENYRW'.»

 UNIT                                       Specifies the unit to which a
                                            file is connected.


 3.2.1  File Names (FILE=)

 File names must follow the rules listed in Section 1.2, "Names." The name of
 an internal file is the name of the character variable, character array,
 character array element, «character structure element,» character substring,
 «or noncharacter array» that makes up the file. The name of an external file
 must be a character string that the operating system recognizes as a file
 name. The operating system assumes the file is in the current working
 directory if you do not specify a path. External file names must follow the
 file-naming conventions of your operating system, as well as the rules
 listed in Section 1.2, "Names." Wild-card specifications are not permitted.

 An external file can be connected to a FORTRAN unit number by any one of the
 following methods:

   ■  If the file is opened with an OPEN statement, the name can be specified
      in the OPEN statement.

   ■  If the file is opened with an OPEN statement for a unit other than one
      of the preconnected units (0, 5, or 6) and no name is specified in the
      OPEN statement, the file is considered a scratch or temporary file, and
      a default name is used.

   ■  «If the file is opened with an OPEN statement and the name is specified
      as all blanks, the name can be read from the command line or can be
      entered by the user, as described in the OPEN entry in Section 4.2,
      "Statement Directory."»

   ■  «If the file is referred to in a READ or WRITE statement before it has
      been opened, an implicit open operation is performed, as described in
      the READ and WRITE entries in Section 4.2, "Statement Directory." The
      implicit open operation is equivalent to executing an OPEN statement
      with a name specified as all blanks. Therefore the name is read from
      the command line or can be input by the user, as described in the OPEN
      entry in Section 4.2, "Statement Directory ."»


 3.2.2  Units (UNIT=)

 For most I/O operations, a file must be identified by a unit specifier. The
 unit specifier for an internal file is the name of that internal file (see
 Section 3.5 for information on internal files). For an external file, a file
 is connected to a unit specifier with the OPEN statement. Some external unit
 specifiers, listed below, are preconnected to certain devices and do not
 have to be opened. External units that you connect are disconnected when
 program execution terminates or when the unit is closed by a CLOSE
 statement.

 The unit specifier is required for all I/O statements except PRINT (which
 always writes to standard output), a READ statement that contains only an
 I/O list and format specifier (which always reads from standard input), and
 an INQUIRE by file (which specifies the file name, rather than the unit to
 which the file is connected).

 An external unit specifier must be either an integer expression or an
 asterisk (*). The integer expression must be in the range -32,767 to 32,767.
 The following example connects the external file UNDAMP to unit 10 and
 writes to it:

            OPEN (UNIT = 10, FILE = 'undamp')
            WRITE (10, *) ' Undamped Motion:'

 The asterisk (*) unit specifier is preconnected and cannot be connected by
 an OPEN statement. It is the only unit specifier that cannot be reconnected
 to another file. The asterisk unit specifier specifies the keyboard when
 reading and the screen when writing. The following example uses the asterisk
 unit specifier to write to the screen:

            WRITE (*, '(1X, ''Begin output.'')')

 Microsoft FORTRAN has four external units:

 External Unit                              Description

 Asterisk (*)                               Always represents the keyboard
                                            and screen

 0                                          Initially represents the keyboard
                                            and screen

 5                                          Initially represents the keyboard

 6                                          Initially represents the screen

 The asterisk (*) unit cannot be connected to any other file, and attempting
 to close this unit causes a compile-time error. Units 0, 5, and 6, however,
 can be connected to any file with the OPEN statement. «If you close unit 0,
 5, or 6, it is automatically reconnected to the keyboard and screen, the
 keyboard, or the screen, respectively.»

 If you read or write to a unit that has been closed, the file is opened
 implicitly, as described in the READ and WRITE entries in Section 4.2,
 "Statement Directory."


 Examples

 In the following program, the character variable fname is an internal file:

      C     The output of this program is FM004.DAT
            CHARACTER*14 fname
            ifile = 4
            WRITE (fname, 100) ifile
      100   FORMAT ('FM', I3.3, '.DAT')
            WRITE (*, *) fname
            END

 «The following example writes to the preconnected unit 6 (the screen), then
 reconnects unit 6 to an external file and writes to it, and finally
 reconnects unit 6 to the screen and writes to it.»

           «REAL a, b
      C     Write to the screen (preconnected unit 6):
            WRITE(6, '('' This is unit 6'')')
      C     Use the OPEN statement to connect unit 6
      C     to an external file named 'COSINES':
            OPEN (UNIT = 2*3, FILE = 'COSINES', STATUS = 'NEW')
            DO 200 a = 0.1, 6.3, 0.1
               b = COS (a)»

      «C     Write to the file 'COSINES':
               WRITE  (6, 100) a, b
      100      FORMAT (F3.1, F5.2)
      200   CONTINUE
      C     Reconnect unit 6 to the screen, by closing it:
            CLOSE (6, STATUS = 'KEEP')
      C     Write to the screen:
            WRITE(6,' ('' Cosines completed'')')
            END»


 3.2.3  File Access Method (ACCESS=)

 The following sections describe the three methods of file access:
 sequential, direct, and appended. Sequential files contain data recorded in
 the order in which it was written to the file; direct files are
 random-access files.


 3.2.3.1  Sequential File Access

 The records in a sequential file appear in the order in which they were
 written to the file.

 All internal files use sequential access. You must also use sequential
 access for files associated with "sequential devices." A sequential device
 is a device that does not allow explicit motion (other than reading or
 writing). The keyboard, screen, and printer are all sequential devices.

 It is not possible to access a particular sequential record directly because
 all the preceding records must be read through first. Direct operations to
 files opened for sequential access are not allowed. «An existing sequential
 file can be opened in 'APPEND' mode. The file is positioned immediately
 after the last record and before the end-of-file mark.»


 3.2.3.2  Direct File Access

 Direct files are random-access files whose records can be read or written to
 in any order. Direct-access files must reside on disk. You cannot associate
 a direct-access file with a sequential device, such as the keyboard, screen,
 or printer.

 Records are numbered sequentially, with the first record numbered 1. All
 records have the same length, specified by the RECL=option in the OPEN
 statement. One record is written for each unformatted READ or WRITE
 statement. A formatted READ or WRITE statement can transfer more than one
 record using the slash (/) edit descriptor.

 «Except for binary files,» the number of bytes written to a record must be
 less than or equal to the record length. «For binary files, a single READ or
 WRITE statement reads or writes as many records as needed to accommodate the
 number of bytes being transferred.» On output, incomplete formatted records
 are padded with spaces. Incomplete unformatted «and binary records» are
 padded with undefined bytes (zeros).

 In a direct-access file, it is possible to write records out of order (for
 example, 9, 5, and 11, in that order) without writing the records in
 between. It is not possible to delete a record once written; however, a
 record can be overwritten with a new value.

 Positioning a direct-access file past the end-of-file marker and attempting
 to read from it can cause a run-time error. Reading an unwritten record from
 a direct-access file is also illegal and can cause a run-time error. If a
 record is written beyond the old terminating file boundary, the operating
 system attempts to extend direct-access files. This works only if there is
 room available on the physical device.

 Each READ or WRITE operation on a direct-access file must explicitly specify
 the record to be accessed. «Microsoft FORTRAN also permits sequential
 operations on files opened for direct access; the operation takes place on
 the next record.»

 «The following statements read the third and fourth records of the file
 xxx:»

           «OPEN (1, FILE = 'xxx', ACCESS = 'DIRECT', RECL = 15
           +         FORM = 'FORMATTED')
            READ (1, '(3I5)', REC = 3) i, j, k
            READ (1, '(3I5)') l, m, n»

 «A file created as a direct-access file and opened in 'APPEND' mode defaults
 to sequential access.»


 3.2.4  «Input/Output Buffer Size (BLOCKSIZE=)»

 «Many programs are "I/O bound," meaning their speed of execution is largely
 determined by the speed of I/O operations. I/O speed, in turn, is influenced
 by the size of the I/O buffer. In general, the larger the buffer, the faster
 the I/O, since a large buffer reduces the total number of reads and writes
 needed to transfer a given amount of data.»

 «The value of BLOCKSIZE is an integer expression specifying the internal
 buffer size for use in I/O. See Microsoft FORTRAN Advanced Topics for
 specific information about block sizes and using the BLOCKSIZE= option.»


 3.2.5  The Edit List

 Edit lists describe the format of data. They are used in FORMAT statements
 and format specifiers.

 The edit list (including the outer parentheses) is a character constant, and
 is enclosed in apostrophes when it appears in a READ or WRITE statement. «In
 Microsoft FORTRAN, the edit list may be enclosed in quotes.» There are no
 apostrophes around the edit list when it appears in a FORMAT statement.

 The edit list is a series of formatting descriptors separated by commas.
 «You may omit the comma between two items if the resulting edit list is not
 ambiguous. For example, you can usually omit the comma after a P edit
 descriptor or before or after the slash (/) edit descriptor without any
 resulting ambiguity.»

 There are three types of editing descriptors:

   1. Nonrepeatable edit descriptors (nonrepeatable)

   2. Repeatable edit descriptors optionally preceded by a repeat
      specification (n-repeatable)

   3. An edit list, enclosed in parentheses, optionally preceded by a repeat
      specification (n(editlist))

 A repeat specification is a nonzero, unsigned integer constant. Edit
 descriptors are described in Sections 3.7.1, "Nonrepeatable Edit
 Descriptors," and 3.7.2, "Repeatable Edit Descriptors." The following list
 gives examples of each type of editing descriptor:

 Editing Descriptor              Examples

 nonrepeatable                   'Total = '
                                 3Hyes
                                 SP
                                 BN
                                 1X

 n-repeatable                    I5
                                 2I5
                                 10F8.2
                                 3A10
                                 E12.7E2

 n(editlist)                     2(1X, 2I5)
                                 (1X, 'Total = ', E12.7E2)
                                 (3A10, 10F8.2)
                                 920(10F8.2) 2(13(2I5), SP, 'Total =', F8.2)

 Up to 16 levels of nested parentheses are permitted within the outermost
 level of parentheses in an editlist.


 Examples

 The following program contains two examples of editlist:

            INTEGER a, b
            REAL x, y
            DATA a /52/, b /9/, x /5832.67/, y /1.02781/
            WRITE (*, 100) a, b
      C     The editlist in the next FORMAT statement is below the
      C     hyphens:   -----------------------
      100   FORMAT (' A = ', I5, 'B = ', I5)
            WRITE (*, 200) x, y
      C     The editlist in the next FORMAT statement is below the
      C     hyphens:   -----------------
      200   FORMAT (' ', 2(F8.3, 1Hm))
            END

 The program above produces the following output:

      A =    52B =     9
      5832.670m   1.028m

 Note that each formatted WRITE statement writes an initial blank to the
 terminal device as a carriage-control character.


 3.2.6  Error and End-of-File Handling (IOSTAT=, ERR=, END=)

 If an error occurs or the end-of-file record is encountered during an I/O
 operation, the action taken depends on the presence and definition of the
 ERR= errlabel, IOSTAT=iocheck, and END=endlabel options. Note that the READ
 statement is the only I/O statement that does not consider reaching an
 end-of-file record as an error.

 Since the PRINT statement does not allow any of these options to be
 specified, an error during execution of a PRINT statement always causes a
 run-time error.

 Table 3.3 indicates what action is taken when an error or end-of-file record
 is encountered by a READ statement. Note that any time an error occurs
 during a READ statement, all items in iolist become undefined.

 Table 3.3  Errors and End-of-File Records When Reading

 IOSTAT       END      ERR      End-of-File               Error, or Error and
 Set          Set      Set      Occurs                    End-of-File, Occurs

 No           No       No       Run-time error occurs.    Run-time error
                                                          occurs.

 No           No       Yes      Program terminates.       Go to errlabel.
                                Go to errlabel.

 No           Yes      No       Go to endlabel.           Run-time error
                                                          occurs.

 No           Yes      Yes      Go to endlabel.           Go to errlabel.

 Yes          No       No       Set iocheck = -1 and      Set iocheck =
                                continue.                 run-time error
                                                          number and
                                                          continue.

 Yes          No       Yes      Set iocheck = -1 and      Set iocheck =
                                continue.                 run-time error
                                                          number and go to
                                                          errlabel.

 Yes          Yes      No       Set iocheck = -1 and      Set iocheck =
                                go to endlabel.           run-time error
                                                          number and
                                                          continue.

 Yes          Yes      Yes      Set iocheck = -1 and      Set iocheck =
                                go to endlabel.           run-time error
                                                          number and go to
                                                          errlabel.

 The following list shows what happens when an error (including encountering
 an end-of-file record) occurs during any I/O statement other than READ or
 PRINT:

 Situation                                  Result

 Neither errlabel nor iocheck is            The program is terminated, and a
 present                                    run-time error message is given.

 Only errlabel is present                   Control is transferred to the
                                            statement at errlabel.

 Only iocheck is present                    The value of iocheck is set to
                                            the run-time error number and
                                            control is returned as if the
                                            statement had executed without
                                            error.

 Both errlabel and iocheck are              The value of iocheck is set to
 present                                    the run-time error number and
                                            control is transferred to the
                                            statement at errlabel.

 If an I/O statement terminates without encountering either an error or
 end-of-file record, and if iocheck is specified, iocheck is set to zero.


 Examples

 In the following program, none of the available options (ERR=, IOSTAT=, or
 END=) are set. Therefore, if an invalid value is entered for i, a run-time
 error is produced:

            INTEGER i
            WRITE (*, *) 'Please enter i'
            READ  (*, *) i
            WRITE (*, *) 'This is i:', i
            END

 The following program uses the ERR= option to prompt the user to enter a
 valid number:

            INTEGER i
            WRITE (*, *) 'Please enter i:'
       50   READ  (*, *, ERR = 100) i
            WRITE (*, *) 'This is i:', i
            STOP ' '
      100   WRITE (*, *) 'Invalid value. Please enter new i:'
            GOTO 50
            END

 The following program uses both the ERR= and IOSTAT= options to handle
 invalid input:

            INTEGER i, j
            WRITE (*, *) 'Please enter i:'
      50    READ  (*, *, ERR = 100, IOSTAT = j) i
            WRITE (*, *) 'This is i:', i, ' iostat = ', j
            STOP ' '
      100   WRITE (*, *) 'iostat = ', j, ' Please enter new i:'
            GOTO 50
            END


 3.2.7  Format Specifier (FMT=)

 Format specifiers either contain an edit list or specify the statement label
 of the format to be used. Format specifiers are used in PRINT, READ, and
 WRITE statements.

 The following sections show the seven acceptable types of format specifiers
 and provide an example of how each format specifier can be used.


 3.2.7.1  FORMAT Statement Label

 A format specifier can be the label of a FORMAT statement. The edit list in
 the FORMAT statement becomes the format for the data. The following syntax
 shows how a FORMAT statement label can specify an edit list for a WRITE
 statement:

            WRITE (*, label) iolist
      label FORMAT (editlist)


 3.2.7.2  Integer-Variable Name

 An ASSIGN statement can be used to associate an integer variable with the
 label of a FORMAT statement. The integer variable can then be used as a
 format specifier, as follows:

            ASSIGN label TO var
      label FORMAT (editlist)
            WRITE (*, var) iolist

 In the WRITE statement, the integer-variable name var refers to the FORMAT
 statement label, as assigned just before the FORMAT statement. For further
 information, see the ASSIGN entry in Section 4.2, "Statement Directory."


 3.2.7.3  Character Expression

 An edit list can be written as a character expression, and that character
 expression can be used as a format specifier, as follows:

      WRITE (*, '(editlist)')iolist

 The character expression can be a character constant. The expression cannot
 include concatenation of an operand whose length specifier is an asterisk in
 parentheses, unless that operand is the symbolic name of a constant.


 3.2.7.4  Character Variable

 An edit list can be written as a character expression, and that expression
 can be assigned to a character variable. The character variable is then used
 as the format specifier, as follows:

      CHARACTER*n var
      var = '(editlist)'
      WRITE (*, var) iolist


 3.2.7.5  Asterisk ( *)

 When an asterisk (*) is used as a format specifier, list-directed I/O is
 performed, as shown in the following syntax line:

      WRITE (6, *) iolist

 In this statement, the asterisk indicates a list-directed transfer. For more
 information, see Section 3.8, "List-Directed I/O."


 3.2.7.6  Character «or Noncharacter» Array Name

 An edit list can be written as a character expression, and that expression
 can be assigned to an array. The array is then used as the format specifier,
 as follows:

      CHARACTER*bytes array(dim)
      DATA array /'(editlist)'/...
      WRITE (*, array) iolist

 The array is interpreted as all the elements of the array concatenated in
 column-major order (see the DIMENSION entry in Section 4.2, "Statement
 Directory,"  for information on order of array elements).

 «A noncharacter array can also be specified where the elements of the array
 are treated as equivalent character variables of the same length.»

 If a Hollerith constant is used to specify an edit list, the edit list
 cannot contain an apostrophe edit descriptor or another Hollerith constant.


 Example

 The following example uses a character array, char, «and two real arrays,
 real1 and real2,» to print a header ( The results are:) and the values of
 two integer arrays, three times:

      «$NOTSTRICT»
            INTEGER  array1(6), array2(4)
            CHARACTER*8  char(4)
            «REAL*8  real1(4), real2(4)»
       C    format is   ==>   (" The results are: ", 6I6, 4I6)
            DATA char  /  "(' The r", "esults a", "re: ', 6",
           +              "I6, 4I6)" /
           «DATA real1 /  "(' The r", "esults a", "re: ', 6",
           +              "I6, 4I6)" /
            DATA real2 / 8H(' The r, 8Hesults a, 8Hre: ', 6,
           +             8HI6, 4I6)  /»
            WRITE (*, char ) array1, array2
           «WRITE (*, real1) array1, array2
            WRITE (*, real2) array1, array2»


 3.2.7.7  Character Array Element

 An edit list can be written as a character expression and that expression
 assigned to a character array element. The character array element is then
 used as the format specifier, as follows:

      CHARACTER*bytes array(dim)
      array(n) = '(editlist)'
      WRITE (*, array(n)) iolist

 In this example, the WRITE statement uses the character array element
 array(n) as the format specifier for data transfer.


 3.2.8  «Namelist Specifier (NML=)»

 «Within a WRITE statement, a namelist specifier causes the names and values
 of all the variables in the namelist to be written to the specified unit.
 Within a READ statement, a namelist specifier prompts the user to enter a
 value for one or more of the variables in the namelist. For more
 information, see Section 3.9, "Namelist-Directed I/O."»


 3.2.9  File Structure (FORM=)

 The structure of a file is determined by the file's access, set by the
 ACCESS= option described in Section 3.2.3, and the form of the data the file
 contains.

 A file is structured in one of three ways:

 Form         Structure

 Formatted    A file is a sequence of formatted records. Formatted records
              are a series of ASCII characters terminated by an end-of-record
              mark (a carriage-return and line-feed sequence). The records in
              a formatted sequential file can have varying lengths. All the
              records in a formatted direct file must be the same length. All
              internal files must be formatted.

 Unformatted  A file is a sequence of unformatted records. Unformatted
              records are a sequence of values. Unformatted direct files
              contain only this data, and each record is padded to a fixed
              length with undefined bytes. Unformatted sequential files also
              contain information that indicates the boundaries of each
              record.

 «Binary»     «Binary sequential files are sequences of bytes with no
              internal structure. There are no records. The file contains
              only the information specified as I/O list items in WRITE
              statements referring to the file.»

              «Binary direct files have very little structure. A record
              length is assigned by the RECL= option of the OPEN statement.
              This establishes record boundaries, which are used only for
              repositioning and padding before and after read and write
              operations and during BACKSPACE operations. These record
              boundaries do not, however, restrict the number of bytes that
              can be transferred during a read or write operation. If an I/O
              operation attempts to read or write more values than are
              contained in a record, the read or write operation is continued
              on the next record. I/O operations that can be performed on
              unformatted direct files produce the same results when applied
              to binary direct files.»

 See Microsoft FORTRAN Advanced Topics for information on how records are
 represented on your system.


 3.2.10  Input/Output List

 The input/output list, iolist, specifies the items to be input or output.
 See Section 3.7.3 for an explanation of how the iolist and editlist
 interact.

 The following items can be in an iolist:

   ■  No entry

      An iolist can be empty. The resulting record is either of zero length
      or contains only padding characters.

      C     An empty iolist:
            WRITE (unit, '(10I8)')

   ■  A variable name, an array-element name, «a structure-element name,» or
      a character-substring name

      These elements specify that the variable, array element, «structure
      element,» or character substring should be input or output.

      C     A variable and array element in iolist:
            READ (*, 500) n, bahb(n)

   ■  An array name

      An unsubscripted array name specifies in column-major order all the
      elements of the array. See the DIMENSION  entry in Section 4.2,
      "Statement Directory," for an explanation of how arrays are arranged in
      memory.

      C     An array in the iolist:
            INTEGER  handle(5)
            DATA  handle / 5*0 /
            WRITE (*, 99) handle
      99    FORMAT (1X, 5I5)

   ■  Any expression

      Output lists in WRITE and PRINT statements can contain arbitrary
      expressions, either numeric, logical, or character.

   ■  An implied-DO list

      Implied-DO lists have the following form:

      (iolist, dovar = start, stop «, inc»)

      Here, iolist is an input/output list (which can contain any of the
      items in this list, including another implied-DO list). The other
      variables are as described for implied-DO lists in the DATA entry in
      Section 4.2, "Statement Directory."

      An implied-DO list is analogous to an ordinary DO loop. The start,
      stop, and inc parameters determine the number of iterations, and dovar
      (where appropriate) can be used as an array element specifier.

      In a READ statement, the DO variable dovar (or other variables
      associated with dovar) must not appear in the iolist in the implied-DO
      list. The variable dovar can, however, be read in the same READ
      statement before the implied-DO list.

      C     Input and output with an implied-do list in iolist:
            INTEGER c, handle(10), n
            WRITE  (*, *) ' Enter c
      (<=10) and'
            WRITE  (*, *) ' handle(1)
      to handle(c)'
            READ   (*, 66) c,
      (handle(n), n = 1, c)
      66    FORMAT (I5, 10(:, /, I5))
            WRITE  (*, 99) c,
      (handle(n), n = 1, c)
      99    FORMAT (1X, 2I5)
            END


 3.2.11  «File Sharing (MODE=, SHARE=)»

 «In systems that use networking or allow multitasking, more than one program
 can attempt to access the same file at the same time. Two options (MODE= and
 SHARE=) in the OPEN statement control file access. These options are also
 available in the INQUIRE statement so you can determine the access status of
 a file.»

 «Option                                    Specification»

 «MODE=mode»                                «How the first process to open a
                                            file can use that file.»

                                            «The file can be opened for
                                            reading ('READ'), writing
                                            ('WRITE'), or both
                                            ('READWRITE').»

 «SHARE=share»                              «How subsequent processes are
                                            allowed to access the file (while
                                            that file is still open).»

                                            «Subsequent processes can be
                                            allowed to read ('DENYWR'), to
                                            write ('DENYRD'), to both read
                                            and write ('DENYNONE'), or to
                                            neither read nor write
                                            ('DENYRW'). In addition, all
                                            processes (except the process
                                            that originally opened the file)
                                            can be prohibited from opening
                                            the file ('COMPAT').»

 «Table 3.4, below, indicates the restrictions on opening a file that has
 already been opened with a particular value of mode and share.»

 «Table 3.4  Mode and Share Values»

 «Original           Original          Concurrent           Concurrent
 Process             Process           Processes Can        Processes Can
 Opened with         Opened with       Be Opened with       Be Opened with
 MODE=               SHARE=            MODE=                SHARE=»

 «'READWRITE' or     «'COMPAT'»        «'READWRITE'         «'COMPAT' by
 'READ' or 'WRITE'»                    'READ' or            original process
                                       'WRITE'»             only»

 «'READWRITE' or     «'DENYRW'»         «Cannot be opened
 'READ' or 'WRITE'                      concurrently»
 »
 «'READWRITE'        'DENYWR'          'READ'               'DENYNONE'»

 «'READ'»            «'DENYWR'»        «'READ'»             «'DENYNONE' or
                                                            'DENYWR'»

 «'WRITE'»           «'DENYWR'»        «'READ'»             «'DENYNONE' or
                                                            'DENYRD'»
 «'READWRITE'        'DENYRD'          'WRITE'              'DENYNONE'»

 «'READ'»            «'DENYRD'»        «'WRITE'»            «'DENYNONE' or
                                                            'DENYWR'»

 «'WRITE'»           «'DENYRD'»        «'WRITE'»            «'DENYNONE' or
                                                            'DENYRD'»

 «'READWRITE'»       «'DENYNONE'»      «'READWRITE'         «'DENYNONE'»
                                       or 'READ' or
                                       'WRITE'»

 «'READ'»            «'DENYNONE'»      «'READWRITE'         «'DENYNONE' or
                                       or 'READ' or         'DENYWR'»
                                       'WRITE'»

 «'WRITE'»           «'DENYNONE'»      «'READWRITE'         «'DENYNONE' or
                                       or 'READ' or         'DENYRD'»
                                       'WRITE'»

 «If, for example, a file is opened with MODE='READ' and SHARE='DENYRD', that
 file can also be opened with MODE='WRITE' and SHARE='DENYNONE' or
 SHARE='DENYWR'.»

 «Suppose, for example, you want several processes to read a file and you
 want to ensure that no process updates the file while those processes are
 reading it. First, determine what type of access to the file you want to
 allow the original process. In the example above, you want the original
 process to read the file only. Therefore, the original process should open
 the file with MODE='READ'. Next, determine what type of access the original
 process should allow other processes: in this case, other processes should
 only be able to read the file. The first process should open the file with
 SHARE='DENYWR'. Now, as indicated in Table 3.4, other processes can also
 open the same file with MODE='READ' and SHARE='DENYWR'.»


 3.2.12  Record Number (REC=)

 The REC=recnum option specifies a record number. «In the LOCKING statement,
 recnum specifies the first record to be locked or unlocked.» In the READ and
 WRITE statements, recnum specifies the first record to be read or written.
 The first record in a file is record number 1.


 3.3  Choosing File Types

 The available file properties can be combined to create many kinds of files.
 Two common file types are listed below:

   1. Sequential, formatted files associated with the asterisk (*) unit
      (which represents the keyboard and screen).

      When reading from the asterisk unit (the keyboard), lines must be
      terminated by pressing ENTER. To correct typing mistakes, follow the
      conventions of your operating system.

   2. A named, sequential, formatted external file.


 Example

 The following example uses the two types of files described above:

      C     Copy a file with three columns of integers, each 7
      C     columns wide, from a file whose name is entered by
      C     the user, to another file named OUT.TXT, reversing
      C     the positions of the first and second columns.

            PROGRAM  ColSwp
            CHARACTER*64  fname

      C     prompt for and read the file name:
            WRITE  (*, 900)
      900   FORMAT (' enter FILE NAME - ' \)
            READ   (*, '(A)') fname

      C     use unit 3 for input from external file:
            OPEN   (3, FILE = fname)

      C     use unit 4 for output to second external file:
            OPEN   (4, FILE = 'OUT.TXT')

      C     read and write until end of file:
      100   READ   (3, 920, END = 200) i, j, k
            WRITE  (4, 920) j, i, k
      920   FORMAT (3I7)
            GOTO 100
      200   WRITE  (*, '(A)') 'Done'
            END

 The file type you choose depends on your application. The following is a
 list of situations when file types other than formatted/sequential are used:

 Situation                                  File-Type Consideration

 You need random access I/O                 Use direct-access files. A common
                                            example of this type of
                                            application is a data base. The
                                            form can be «binary,» formatted,
                                            or unformatted.

 Your data is accessed only by              Accessing «binary and»
 Microsoft FORTRAN, and speed is            unformatted files is faster than
 important                                  accessing formatted files.

 Data must be transferred without           «Binary or» unformatted I/O is
 any system interpretation                  most practical.

 All 256 possible byte values               «Binary or» unformatted I/O is
 (ASCII 0-255) are to be                    necessary.
 transferred

 You are controlling a device with          «Binary or» unformatted I/O is
 a one-byte (binary) interface              necessary. In this situation,
                                            formatted I/O would interpret
                                            certain characters (such as the
                                            ASCII representation for
                                            carriage-return/line-feed)
                                            instead of passing them through
                                            to the program unaltered. «The
                                            binary format is preferable,
                                            since the file contains no
                                            record-structure information.»

 Data must be transferred without           «Binary format is recommended.
 any system interpretation, and             Unformatted files are blocked
 will be read by non-FORTRAN                internally, so the non-FORTRAN
 programs                                   program must be compatible with
                                            this format to interpret the data
                                            correctly. Binary files contain
                                            only the data written to them.»

 You are reading a file that was            «Binary I/O is recommended.
 not created by a Microsoft                 Non-FORTRAN files usually have a
 FORTRAN program                            different structure than FORTRAN
                                            files. Opening a file as
                                            FORM=BINARY and ACCESS=DIRECT
                                            with RECL=1 lets you move to any
                                            position and read an arbitrary
                                            sequence of values. Incomplete
                                            records don't cause undefined
                                            values because the record size is
                                            1.»


 3.4  File Position

 Opening a sequential file positions the file at its beginning «(unless the
 file was opened with ACCESS = 'APPEND')». If you write to the file, all
 records after the current record are discarded. The file position after
 sequential WRITE statements is at, but not beyond, the end-of-file record.

 Executing the ENDFILE statement inserts an end-of-file marker at the file's
 current position, then positions the file after the end-of-file marker. A
 BACKSPACE statement must be executed to position the file in front of the
 end-of-file marker. Any data past this position is lost.

 Executing a READ statement at the end of the file positions the file beyond
 the end-of-file record and produces an error unless you have specified the
 END=, ERR=, or IOSTAT= option in a READ statement.


 3.5  Internal Files

 An external file is a physical device, such as a printer or a screen, or is
 a file that is known to the operating system. An internal file is a
 character variable, character array, character array element, «character
 structure element,» character substring«, or noncharacter array».

 There are two basic types of internal files:

 Type                                       Properties

 Character variable, character              The file has exactly one record,
 array element, «character                  which is the same length as the
 structure element,» character              variable, array element,
 substring, «or noncharacter                substring«, or noncharacter
 array»                                     array. Noncharacter arrays are
                                            allowed for compatibility with
                                            older versions of FORTRAN.»

 Character array                            The file is a sequence of
                                            character array elements, each of
                                            which is a record. The order of
                                            records is the same as the order
                                            of array elements. All records
                                            are the same length: the length
                                            of array elements.

 You must follow these rules when using internal files:

   1. Use only formatted, sequential I/O.

   2. Use only the READ and WRITE statements.

   3. Don't use list-directed formatting.

 Before an I/O statement is executed, internal files are positioned at the
 beginning of the file. With internal files, you can use the formatting
 capabilities of the I/O system to convert values between external character
 representations and FORTRAN internal memory representations. That is,
 reading from an internal file converts the ASCII values into numeric,
 logical, or character values, and writing to an internal file converts
 values into their ASCII representations.

 If less than an entire record is written to an internal file, the rest of
 the record is filled with blanks.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    The FORTRAN 66 DECODE statement has been replaced by the internal READ
    function. The ENCODE statement has been replaced by the internal WRITE
    function.
 ───────────────────────────────────────────────────────────────────────────


 Example

 The following example prompts for a 3-digit file identification and uses an
 internal file to create a file name:

            CHARACTER fname*64
            fname = 'FILE   .DAT'
            WRITE (*, *) 'Enter 1-3 digit file identifier'
            READ  (*, *) id
            WRITE (fname(5:7), '(I3.3)') id
            OPEN  (1, FILE = fname, STATUS = 'NEW')
            .
            .
            .
            END


 3.6  Carriage Control

 When formatted I/O is used to transfer a record to a terminal device, such
 as the screen or a printer, the first character of that record is
 interpreted as a carriage-control character, and is not printed. The
 characters 0, 1, +, and the blank character, have the effects indicated in
 Table 3.5 below. Any other character is treated like the blank character. If
 the first character of your record does not print, make sure it is not being
 interpreted as the carriage-control character. In the following program
 fragment, for example, the number 2  is interpreted as a carriage-control
 character, and is treated like the space character:

            WRITE  (*, 100)
      100   FORMAT ('25 years')

 The example above produces the following output:

      5 years

 In list-directed I/O, the first character is not treated as a
 carriage-control character. In the following example of unformatted I/O, the
 full string is displayed:

            WRITE (*, *) '25 years'

 The following output is produced:

      25 years

 The effects of the characters 0, 1, +, and the blank character are listed in
 Table 3.5.

 Table 3.5  Carriage-Control Characters

 Character         Effect

 Blank             Advances one line.

 0                 Advances two lines.

 1                 Advances to top of next page. The screen behaves as if
                   this carriage-control character were ignored.

 +                 Does not advance. Permits overprinting.

 When writing to the screen or printer, the end-of-record mark (a
 carriage-return and line-feed sequence) is normally not sent until the next
 record is written. However, if a write to the screen is followed by a read
 from the keyboard, a new-line character is automatically emitted and the
 input line is positioned below the output line.

 «To suppress the new-line character and display the user's input on the same
 line as the previous output, add the backslash (\) edit descriptor at the
 end of the WRITE statement's edit list. The input then appears at the end of
 the last line written.» Since input lines are always terminated by the user
 with a new-line character, the next write operation always begins on a new
 line. Therefore, if the next operation to the console is a write operation,
 carriage control is adjusted to write one less end-of-record mark.

 Note that the plus (+) carriage-control character has no effect if the
 previous console operation was a read operation.


 3.7  Formatted I/O

 If a READ or WRITE statement includes a format specifier (other than an
 asterisk), the I/O statement is called a formatted I/O statement. The
 remainder of this section discusses the elements of format specifiers and
 the interaction between format specifiers and the I/O list. See Section
 3.2.7 for information on format specifiers.


 3.7.1  Nonrepeatable Edit Descriptors

 Table 3.6 summarizes the nonrepeatable edit descriptors. A discussion of
 edit descriptors follows.

 Table 3.6  Nonrepeatable Edit Descriptors

                                                     Used for  Used for
 Form       Name            Use                      Input     Output

 string     Apostrophe      Transmits string to      No        Yes
            editing         output unit

 nH         Hollerith       Transmits next n         No        Yes
            editing         characters to output
                            unit

 Tc, TLc,   Positional      Specifies position in    Yes       Yes
 TRc        editing         record

 nX         Positional      Specifies position in    Yes       Yes
            editing         record

 SP, SS, S  Optional-plus   Controls output of       No        Yes
            editing         plus signs

 /          Slash editing   Positions to next        Yes       Yes
                            record or writes
                            end-of-record mark

 «\»        «Backslash      «Continues same record»  «No»      «Yes (to
            editing»                                           terminal and
                                                               printer only)»

 :          Format          If no more items in      No        Yes
            control         iolist, terminates
            termination     statement

 kP         Scale-factor    Sets scale for           Yes       Yes
            editing         exponents in
                            subsequent F and E
                            (repeatable) edit
                            descriptors

 BN, BZ     Blank           Specifies                Yes       No
            interpretation  interpretation of
                            blanks in numeric
                            fields

 Sections 3.7.1.1-3.7.1.10 describe each nonrepeatable edit descriptor.


 3.7.1.1  Apostrophe Editing

 If a format specifier contains a character constant, string, that string is
 transmitted to the output unit. Embedded blanks are significant.

 Two adjacent apostrophes must be used to represent a single apostrophe
 within a character constant. Each additional level of nested apostrophes
 requires twice as many apostrophes as the previous level to resolve the
 ambiguity of the apostrophe's meaning. Note how in the second WRITE
 statement in the example below, the set of apostrophes that delimit the
 output string within the FORMAT statement are doubled, and four apostrophes
 are required within the output string itself to specify a single output
 apostrophe.


 Example

      C     These WRITE statements both output ABC'DEF
      C     (The leading blank is a carriage-control character)
            WRITE  (*, 970)
      970   FORMAT (' ABC''DEF')
            WRITE  (*, '('' ABC''''DEF'')')
      C     The following WRITE also outputs ABC'DEF. No carriage-
      C     control character is necessary for list-directed I/O
            WRITE (*, *) 'ABC''DEF'

 Apostrophe editing cannot be used with READ statements.


 3.7.1.2  Hollerith Editing (H)

 The nH edit descriptor transmits the next n characters, including blanks, to
 the output unit. Hollerith editing can be used in every context where
 character constants can be used.

 The n characters transmitted are called a "Hollerith constant."


 Example

      C     These WRITE statements both output ABC'DEF
      C     (The leading blank is a carriage-control character)
            WRITE (*, '(8H ABC''DEF)')
            WRITE (*, 960)
      960   FORMAT (8H ABC'DEF)


 3.7.1.3  Positional Editing: Tab, Tab Left, Tab Right (T, TL, TR)

 The T, TL, and TR edit descriptors specify the position in the record to
 which or from which the next character will be transmitted. The new position
 may be in either direction from the current position. This allows a record
 to be processed more than once on input. Note that moving the position
 backward more than 512 bytes (characters) is not recommended.

 The Tc edit descriptor specifies absolute tabbing; the transmission of the
 next character is to occur at the character position c. The TRc edit
 descriptor specifies relative tabbing to the right; the transmission of the
 next character is to occur c characters beyond the current position. The TLc
 edit descriptor specifies relative tabbing to the left; the transmission of
 the next character is to occur c characters prior to the current position.

 If TLc specifies a position before the first position of the current record,
 TLc editing causes transmission to or from position 1.

 Left tabbing is legal within records written to devices. However, if the
 record that is written is longer than the buffer associated with the device,
 you cannot left-tab to a position corresponding to the previous buffer.

 For example, the buffer associated with the console is 132 bytes. If a
 record of 140 bytes is written to the console, left tabbing is allowed for
 only eight bytes, since the first 132 bytes of the record have been sent to
 the device and are no longer accessible.

 If one of these edit descriptors is used to move to a position to the right
 of the last data item transmitted and another data item is then written, the
 space between the previous end of data in the record and the new position is
 filled with spaces. For example, consider the following:

            WRITE (*, '('' '', 3(''1234567890''))')
            WRITE (*, 100) 5, 9
        100 FORMAT (I5, 20X, TL10, I5)

 This example produces the following output:

      123456789012345678901234567890
          5              9

 «Be careful when using these edit descriptors if you read data from files
 that use commas as field delimiters. If you move backwards in a record using
 TLc or Tc (where c is less than the current position in the record), commas
 are disabled as field delimiters. If the format controller encounters a
 comma after you have moved backward in a record with TLc or Tc, a run-time
 error occurs. If you want to move backward in a record without disabling
 commas as field delimiters, advance to the end-of-record mark, then use the
 BACKSPACE statement to move to the beginning of the record.»


 3.7.1.4  Positional Editing (X)

 The nX edit descriptor advances the file position n characters. On output,
 if the nX edit descriptor moves past the end of data in the record, and if
 there are further items in the iolist, blanks are output, as described for
 the Tc and TRc edit descriptors.


 Example

      C     This writes 1   5   10   15 on the screen:
            WRITE  (*, 100)
      100   FORMAT (1X, '1', 3X, '5', 3X, '10', 3X, '15')

      C     This writes "zogoZOGozogo     !" on the screen:
            WRITE  (*, 200)
      200   FORMAT (1X, 'zogozogozogo', TL8, 'ZOG', 10X, '!')


 3.7.1.5  Optional-Plus Editing (SP, SS, S)

 The SP, SS, and S edit descriptors control optional-plus characters in
 numeric output fields. SP causes output of the plus sign in all subsequent
 positions that the processor recognizes as optional-plus fields. SS causes
 plus-sign suppression in all subsequent positions that the processor
 recognizes as optional-plus fields. S restores SS, the default.


 Example

      C     The following statements write: 251 +251  251 +251
      C     251
            INTEGER i
            i = 251
            WRITE  (*, 100) i, i, i, i, i
      100   FORMAT (1X, I5, SP, I5, SS, I5, SP, I5, S, I5)

      C     The following statements write:
      C     .673E+4+.673E+4 .673E+4+.673E+4 .673E+4
            REAL r
            r = 67.3E2
            WRITE  (*, 200) r, r, r, r, r
      200   FORMAT (1X, E8.3E1, SP, E8.3E1, SS, E8.3E1, SP,
           +            E8.3E1, S,  E8.3E1)


 3.7.1.6  Slash Editing (/)

 The slash indicates the end of data transfer on the current record. On
 input, the file is positioned to the beginning of the next record. On
 output, an end-of-record mark is written, and the file is positioned to
 write at the beginning of the next record.


 Example

      C     The following statements write a column and a row:
            WRITE  (*, 100)
      100   FORMAT (' c row', /, ' o', /, ' l', /, ' u',
           +                  /, ' m', /, ' n')

 The output from this example is shown below:

      c row
      o
      l
      u
      m
      n


 3.7.1.7  «Backslash Editing (\)»

 «The backslash edit descriptor is used only for formatted output to terminal
 devices, such as the screen or a printer. It is ignored in all other
 situations.»

 «When the format controller terminates a transmission to the terminal, it
 writes an end-of-record mark (a carriage-return and line-feed sequence). If
 the last edit descriptor encountered by the format controller is a backslash
 (\), no end-of-record mark is written, so the next I/O statement will
 continue writing on the same line.»

 «This mechanism can be used to write a prompt to the screen and then read a
 response from the same line, as in the following example:»

            «WRITE (*, '(A \)') 'enter an integer --> '
            READ  (*, '(BN, I6)') j»


 3.7.1.8  Terminating Format Control (:)

 The colon (:) edit descriptor terminates format control if there are no more
 items in the iolist. This feature is used to suppress output when some of
 the edit descriptors in the format do not have corresponding data in the
 iolist.


 Example

      C     The following example writes a= 3.20 b=  .99
            REAL  a, b, c, d
            DATA  a /3.2/, b /.9871515/
            WRITE  (*, 100) a, b
      100   FORMAT (' a=', F5.2, :, ' b=', F5.2, :,
           +        ' c=', F5.2, :, ' d=', F5.2)
            END


 3.7.1.9  Scale-Factor Editing (P)

 The kP edit descriptor sets the scale factor for all subsequent F and E edit
 descriptors (for information on F and E, see Section 3.7.2, "Repeatable Edit
 Descriptors") until another kP edit descriptor is encountered. At the start
 of each I/O statement, the scale factor is initialized to zero. The scale
 factor affects format editing in the following ways:

   ■  On input, with F and E editing, if there is no explicit exponent, the
      value read in is divided by 10^(k) before it is assigned to a variable.
      If there is an explicit exponent, the scale factor has no effect.

   ■  On output, with F editing, the value to be written out is multiplied by
      10^(k) before it is displayed.

   ■  On output, with E editing, the real part of the value to be displayed
      is multiplied by 10^(k), and its exponent is reduced by k. This alters
      the column position of the decimal point but not the value of the
      number.


 Examples

 The following fragment uses scale-factor editing when reading:

            READ (*, 100) a, b, c, d
       100  FORMAT (F10.6, 1P, F10.6, F10.6, -2P, F10.6)

            WRITE  (*, 200) a, b, c, d
       200  FORMAT (4F11.3)

 Assume the following data is entered:

      12340000  12340000  12340000  12340000
         12.34     12.34     12.34     12.34
       12.34e0   12.34e0   12.34e0   12.34e0
       12.34e3   12.34e3   12.34e3   12.34e3

 The program outputs the following:

          12.340      1.234      1.234   1234.000
          12.340      1.234      1.234   1234.000
          12.340     12.340     12.340     12.340
       12340.000  12340.000  12340.000  12340.000

 The following example uses scale-factor editing when writing:

            a = 12.34

            WRITE  (*, 100) a, a, a, a, a, a
      100   FORMAT (1X, F9.4, E11.4E2, 1P, F9.4, E11.4E2,
           +           -2P, F9.4, E11.4E2)

 This program outputs the following:

      12.3400  .1234E+02 123.4000 1.2340E+01    .1234  .0012E+04


 3.7.1.10  Blank Interpretation (BN, BZ)

 The edit descriptors BN and BZ control the interpretation of blanks in
 numeric input fields.

 The BN edit descriptor ignores blanks; it takes all the nonblank characters
 in the field and right-justifies them. For example, if an input field
 formatted for six integers contains ' 23 4 ', it is interpreted as '  234'.

 The BZ edit descriptor makes blanks identical to zeros. Trailing blanks in
 the field become zeros. Interspersed blanks also become zeros. In the
 previous example, the input field ' 23 4 ' would be interpreted as
 ' 23040'. If ' 23 4' were entered, the formatter would add one blank to pad
 the input to the six-integer format, but this extra space would be ignored,
 and the input would be interpreted as ' 2304 '. Note that the blanks
 following the E or D in real-number input are ignored, whatever form of
 blank interpretation is in effect.

 The default, BN, is set at the beginning of each I/O statement, unless the
 BLANK= option was specified in the OPEN statement. If you specify a BZ edit
 descriptor, BZ editing is in effect until the BN edit descriptor is
 specified.

 For example, look at the following program fragment:

            READ (*, 100) n
      100   FORMAT (BN, I6)

 If you enter any one of the following three records and terminate by
 pressing ENTER, the READ statement interprets that record as the value 123:

         123
      123
      123   456

 Because the repeatable edit descriptor associated with the I/O list item n
 is I6, only the first six characters of each record are read (three blanks
 followed by 123 for the first record, and 123 followed by three blanks for
 the last two records). Because blanks are ignored, all three records are
 interpreted as  123.

 The following example shows the effect of BN editing with an input record
 that has fewer characters than the number of characters specified by the
 edit descriptors and iolist. Suppose you enter 502, followed by ENTER, in
 response to the following READ statement:

            READ (*, '(I5)') n

 The I/O system is looking for five characters that it will try to interpret
 as an integer number. You have only entered three, so the first thing the
 I/O system does is to pad the record 502 on the right with two blanks. If BZ
 editing were in effect, those two blanks would be interpreted as zeros, and
 the record would be equal to 50200. However, with BN editing in effect (the
 default), the nonblank characters (502) are right-justified, so the record
 is equal to 502.


 3.7.2  Repeatable Edit Descriptors

 The I (integer), «Z (hexadecimal),» F (single-precision real), E (real with
 exponent), G (real with optional exponent), and D (double-precision real)
 edit descriptors are used for I/O of numeric data. The following rules apply
 to all of these numeric edit descriptors:

   ■  On input, fields that are all blanks are always interpreted as zero.
      The interpretation of trailing and interspersed blanks is controlled by
      the BN and BZ editing descriptors. Plus signs are optional. «The blanks
      supplied by the file system to pad a record to the required size are
      not significant.»

   ■  On input with F, E, G, and D editing, an explicit decimal point in the
      input field overrides any edit-descriptor specification of the
      decimal-point position.

   ■  On output, the characters generated are right-justified in the field
      and padded by leading blanks, if necessary.

   ■  On output, if the number of characters produced exceeds the field width
      or if the exponent exceeds its specified width, the entire field is
      filled with asterisks. If a real number contains more digits after the
      decimal point than are allowed in the field, the number is rounded.

   ■  «When reading with I, Z, F, E, G, D, or L edit descriptors, the input
      field may contain a comma that terminates the field. The next field
      starts at the character following the comma. The missing characters are
      not significant. For example, consider the following READ statement:»

          «READ (*, '(BZ 3I5)') i, j, k»

      «Entering the following data results in i = 1, j = 20, and k = 3:»

          «1, 2 ,  3,»

      «Do not use this feature when you use explicit positional editing (the
      T, TL, TR, or nX edit descriptors).»

   ■  Two successively interpreted edit descriptors of the types F, E, G, and
      D are required to format complex numbers. Two different descriptors may
      be used. The first edit descriptor specifies the real part of the
      complex number, and the second specifies the imaginary part.

   ■  Nonrepeatable edit descriptors may appear between repeatable edit
      descriptors.

 The following sections describe each repeatable edit descriptor.


 3.7.2.1  Integer Editing (I)


 Syntax

      Iw«.m»

 On input, any value entered that is associated with an I edit descriptor
 must have the form of an integer (it may not contain a decimal point or
 exponent), or a run-time error occurs. On output, the I/O list item
 associated with an I edit descriptor must have an integer value, or a
 run-time error occurs.

 The field is w characters wide. On input, an optional sign may appear in the
 field. If the optional unsigned integer m is specified, input is the same as
 Iw, but output is padded with leading zeros up to width m. For example,
 consider this statement:

            WRITE (*, '(1X, I5, I5.3)') 7, 7

 The following output is produced:

      7  007


 3.7.2.2  «Hexadecimal Editing (Z)»


 «Syntax»

      «Z«w»»

 «Hexadecimal editing converts between external data in hexadecimal form (the
 hexadecimal digits 0 through 9 and A through F) and internal four-bit binary
 data (0000 through 1111). Each byte of internal data corresponds to two
 four-bit hexadecimal characters. For example, the ASCII character "m," which
 is 01101101 in binary, is output as the hexadecimal characters 6D.
 Similarly, an INTEGER*4 value is output in its eight-hexadecimal-character
 form.»

 «The optional field width, w, specifies the number of hexadecimal characters
 to be read or written. If w is omitted, the field width defaults to 2*n,
 where n is the length of the iolist item in bytes. For example, an INTEGER*2
 value is represented by four hexadecimal characters.»

 «On output, character data types are written in the same order they appear
 in memory. For numeric and logical types, bytes are output in order of
 significance,  from the most significant on the left to the least
 significant on the right. The INTEGER*2 value 10, for example, will be
 output as 000A, although the order of bytes in memory on an 8086-based
 machine is actually 0A00.»

 «The following rules of truncation and padding apply. The value n is the
 length of the iolist item in bytes:»

 «Operation     Rule»

 «Output»       «If w > 2*n, the 2*n hexadecimal characters are right-
                justified and leading zeros are added to make the external
                field width equal to w.»

                «If w <= 2*n, the w rightmost hexadecimal characters are
                output.»

 «Input»        «If w >= 2*n, the rightmost 2*n hexadecimal characters are
                taken from the input field.»

                «If w < 2*n, the first w hexadecimal characters are read
                from the external field. Enough leading zeros are added to
                make the width equal to 2*n.»

 «Blanks in an input field are treated as zeros.»

 «Hexadecimal editing differs from conventional decimal editing in two
 significant ways. If there are more hex characters to be output than the
 field width can accommodate, the field is not filled with asterisks.
 Instead, w right-hand characters are displayed.»

 «The left-hand padding of short input fields does not take into account the
 sign bit of the entered value. For example, in a 8-character input field
 where only FFFF is entered, the entry is interpreted as 65,535, not -1.»

 «To edit complex numbers, two Z edit descriptors must be used. The first
 edit descriptor specifies the real part of the complex number, and the
 second specifies the imaginary part.»


 «Examples»

 «The following example demonstrates hexadecimal editing for output:»

            «CHARACTER*2 alpha
            INTEGER*2   num

            alpha = 'YZ'
            num   =  4096

            WRITE (*, '(1X, Z, 1X, Z2, 1X, Z6)') alpha, alpha,
           +alpha
            WRITE (*, '(1X, Z, 1X, Z2, 1X, Z6)') num,   num,   num»

 «This example produces the following  output:»

      «595A 5A 00595A
      1000 00 001000»

 «As an example of input, suppose the input record is 595A (hexadecimal
 characters), and the iolist item has CHARACTER*2 type. The record would be
 read as follows:»

 «Edit Descriptor                Value Read»

 «Z                               YZ»

 «Z2                              0Y»

 «Z6                              YZ»


 3.7.2.3  Real Editing without Exponents (F)


 Syntax

      Fw.d

 On output, the I/O list item associated with an F edit descriptor must be a
 single- or double-precision number, or a run-time error occurs. On input,
 the number entered may have any integer or real form, as long as its value
 is within the range of the associated variable.

 The field is w characters wide, with a fractional part d digits wide. The
 input field begins with an optional sign followed by a string of digits that
 may contain an optional decimal point. If the decimal point is present, it
 overrides the d specified in the edit descriptor; otherwise, the rightmost d
 digits of the string are interpreted as following the decimal point (with
 leading blanks converted to zeros, if necessary). Appearing after these
 digits is an optional exponent that must be one of the following:

   ■  + (plus) or - (minus) followed by an integer

   ■  E followed by zero or more blanks, followed by an optional sign,
      followed by an integer

 An example is the following READ statement:

            READ (*, '(F8.3)') xnum

 The above statement reads a given input record as follows:

 Input                            Number Read

 5                                .005

 2468                             2.468

 -24680                           -24.680

 -246801                          -246.801

 56789                            5.678

 -28E2                            -2.800

 The output field occupies w characters. One character is a decimal point,
 leaving w-1 characters available for digits. If the sign is negative, it
 must be included, leaving only w-2 characters available. Out of these w-1 or
 w-2 characters, d characters will be used for digits to the right of the
 decimal point. The remaining characters will be blanks or digits, as needed,
 in order to represent the digits to the left of the decimal point.

 The value output is controlled both by the iolist item and the current scale
 factor. The output value is rounded rather than truncated.


 Example

            REAL*4 g, h, e, r, k, i, n
            DATA  g /12345.678/, h /12345678./, e /-4.56E+1/, r
           +/-365./
            WRITE  (*, 100) g, h, e, r
      100   FORMAT (1X, F8.2)
            WRITE  (*, 200) g, h, e, r
      200   FORMAT (1X, 4F10.1)


 The program above produces the following output:

      12345.68
      ********
        -45.60
       -365.00
         12345.712345680.0     -45.6    -365.0


 3.7.2.4  Real Editing with Exponents (E)


 Syntax

      Ew.d«Ee»

 On output, the I/O list item associated with an E edit descriptor must be a
 single- or double-precision number, or a run-time error occurs. On input,
 the number entered may have any integer or real form, as long as its value
 is within the range of the associated variable.

 The field is w characters wide. The e parameter is ignored in input
 statements. The input field for the E edit descriptor is identical to that
 described by an F edit descriptor with the same w and d.

 The form of the output field depends on the scale factor (set by the P edit
 descriptor) in effect. For a scale factor of 0, the output field is a minus
 sign (if necessary), followed by a decimal point, followed by a string of
 digits, followed by an exponent field for exponent exp, having one of the
 forms shown in Table 3.7.

 Table 3.7  Forms of Exponents for the E Edit Descriptor

                            Absolute Value
 Edit Descriptor            of Exponent                Form of Exponent

 Ew.d                       |exp| <= 99                E followed by plus or
                                                       minus, followed by the
                                                       two-digit exponent

 Ew.d                       99 < |exp| <= 999          Plus or minus,
                                                       followed by the
                                                       three-digit exponent

 Ew.dEe                     |exp| <= (10^(e))-1        E followed by plus or
                                                       minus, followed by e
                                                       digits, which are the
                                                       exponent (with
                                                       possible leading
                                                       zeros)

 The form Ew.d must not be used if the absolute value of the exponent to be
 printed exceeds 999.

 The scale factor controls the decimal normalization of the printed E field.
 If the scale factor k is greater than -d and less than or equal to 0, then
 the output field contains exactly k leading zeros after the decimal point
 and d + k significant digits after this. If (0<k<d+2), the output field
 contains exactly k significant digits to the left of the decimal point and
 (d-k-1) places after the decimal point. Other values of k are errors.


 3.7.2.5  Real Editing-Wide Range (G)

 Syntax

      Gw.d«Ee»

 On output, the I/O list item associated with a G edit descriptor must be a
 single- or double-precision number, or a run-time error occurs. On input,
 the number entered may have any integer or real form, as long as its value
 is within the range of the associated variable.

 For either form, the input field is w characters wide, with a fractional
 part consisting of d digits. If the scale factor is greater than 1, the
 exponent part consists of e digits.

 G input editing is the same as F input editing. G output editing corresponds
 to either E or F editing, depending on the magnitude of the data. Tables 3.8
 and 3.9 show how the G edit descriptor is interpreted.

 Table 3.8  Interpretation of G Edit Descriptor

 Data Magnitude                                    Interpretation

 m < 0.1                                           Gw.d=Ew.d
 0.1 <= m < 1                                      Gw.d=F(w-4).d,4(' ')
 1 <= m < 10 (i.e., 10^(d-d) <= m < 10^(d-(d-1))   Gw.d=F(w-4).(d-1),4(' ')
 10^(d-2) <= m < 10^(d-1)                      Gw.d=F(w-4).1,4(' ')
 10^(d-2) <= m < 10^(d)                             Gw.d=F(w-4).0,4(' ')
 10^(d) <= m                                       Gw.d=Ew.d


 Table 3.9  Interpretation of GE Edit Descriptor

 Data Magnitude                                    Interpretation

 m < 0.1                                           Gw.dEe=Ew.d
 0.1 <= m < 1                                      Gw.dEe=F(w-e-2).d,(e+2)(' '
 1 <= m < 10 (i.e., 10^(d-d) <= m < 10^(d-(d-1))   Gw.dEe=F(w-e-2).(d-1),(e+2)









 10^(d-2) <= m < 10^(d-1)                          Gw.dEe=F(w-e-2).1,(e+2)(' '
 10^(d-1) <= m < 10^(d)                            Gw.dEe=F(w-e-2).0,(e+2)(' '
 10^(d) <= m                                       Gw.dEe=Ew.d


 3.7.2.6  Double-Precision Real Editing (D)

 Syntax

      Dw.d

 On output, the I/O list item associated with a D edit descriptor must be a
 single- or double-precision number, or a run-time error occurs. On input,
 the number entered may have any integer or real form, as long as its value
 is within the range of the associated variable. All parameters and rules for
 the E descriptor apply to the D descriptor.

 The field is w characters wide. The input field for the D edit descriptor is
 identical to that described by an F edit descriptor with the same w and d.

 The form of the output field depends on the scale factor (set by the P edit
 descriptor) in effect. For a scale factor of 0, the output field is a minus
 sign (if necessary), followed by a decimal point, followed by a string of
 digits, followed by an exponent field for exponent exp, in one of the forms
 shown in Table 3.10.

 Table 3.10  Forms of Exponents for the D Edit Descriptor

                            Absolute Value
 Edit Descriptor            of Exponent                Form of Exponent

 Dw.d                       |exp| <=  99               D followed by plus or
                                                       minus, followed by the
                                                       two-digit exponent

 Dw.d                       99 < |exp| <=  999         Plus or minus,
                                                       followed by the
                                                       three-digit exponent

 The form Dw.d must not be used if the absolute value of the exponent to be
 printed exceeds 999.

 The scale factor controls the decimal normalization of the printed D field.
 If the scale factor, k, is greater than -d and less than or equal to 0, then
 the output field contains exactly k leading zeros after the decimal point
 and d+k significant digits after this. If (0 < k < d+2), then the output
 field contains exactly k significant digits to the left of the decimal point
 and (d-k-1) places after the decimal point. Other values of k are errors.


 3.7.2.7  Logical Editing (L)


 Syntax

      Lw

 The field is w characters wide. On output, the iolist element associated
 with an L edit descriptor must be of type logical or a run-time error
 occurs. On input, the field consists of optional blanks, followed by an
 optional decimal point, followed by T (for true) or F (for false). Any
 further characters in the field are ignored, but accepted on input, so that
 .TRUE. and .FALSE. are also valid inputs. On output, w-1 blanks are followed
 by either T or F, as appropriate.


 3.7.2.8  Character Editing (A)


 Syntax

      A«w»

 If w is omitted, the field width defaults to the number of characters in the
 iolist associated item. The iolist item may be of any type. If it is not of
 the CHARACTER type, it is assumed to have one character per byte.

 «When the iolist item is of type INTEGER, REAL, or LOGICAL, Hollerith data
 types can be used. On input, the iolist item becomes defined with Hollerith
 data. On output, the iolist item must be defined with Hollerith data.»

 If the number of characters input is less than w, the input field is padded
 with blanks. If the number of characters input is greater than w, the input
 field is truncated on the right to the length of w. Only after these
 adjustments have been made will the input field be transmitted to the iolist
 item. For example, look at the following program fragment:

             CHARACTER*10 char
             READ (*, '(A15)') char

 Assume the following 13 characters are typed in at the keyboard:

      ABCDEFGHIJKLM

 The following two steps occur:

   1. Blanks are added to pad the input field to 15 characters:

      'ABCDEFGHIJKLM  '

   2. The rightmost 10 characters are transmitted to the iolist element char:

      'FGHIJKLM  '

 On output, if w exceeds the number of characters produced by the iolist
 item, leading blanks are provided. Otherwise, the leftmost w characters of
 the iolist item are output.


 3.7.3  Interaction between Format and I/O List

 If an iolist contains one or more items, at least one repeatable edit
 descriptor must exist in the format specification. The empty edit
 specification, ( ), can be used only if no items are specified in the
 iolist. A formatted WRITE statement with an empty edit specification writes
 a carriage return and line feed. A READ statement with an empty edit
 specification skips to the next record.

 «If you read a record in which the total number of characters is less than
 the number of characters specified by the edit descriptors and iolist, the
 following occurs:»

   1. «The record is padded with blanks on the right to the required length.»

   2. «Any blanks entered by the user are interpreted according to the
      blank-editing descriptor in effect (BN or BZ).»

 «For example, consider the following READ statement that uses BZ editing:»

            «READ (*, '(BZ, I5)') n»

 «Assume you enter the following in response:»

       «5»

 «The total number of characters in the input record is two (a blank followed
 by a 5). The record is padded on the right with three blanks, but these
 additional blanks added by the formatter are ignored. The input record is
 thus interpreted as 5, rather than 5000.»

 Each item in the iolist is associated with a repeatable edit descriptor
 during the I/O statement execution. Each COMPLEX item in the iolist requires
 two edit descriptors in the FORMAT statement or format descriptor.
 Nonrepeatable edit descriptors are not associated with items in the iolist.

 During the formatted I/O process, the format controller scans and processes
 the format items from left to right. Following is a list detailing
 situations the format controller may encounter and their explanations:

   ■  A repeatable edit descriptor is encountered, and a corresponding item
      appears in the iolist.

      The item and the edit descriptor are associated, and I/O of that item
      proceeds under the format control of the edit descriptor.

   ■  A repeatable edit descriptor is encountered, and no corresponding item
      appears in the iolist.

      The format controller terminates I/O. For the following statements, for
      example:

      i = 5
            WRITE (*, 100) i
      100   FORMAT (1X, 'I= ', I5, ' J= ',
           +I5, ' K= ', I5)

      the output would look like this:

      I=     5 J=

      The output terminates after J= because no corresponding item for the
      second I5 appears in the iolist.

   ■  The matching final right parenthesis of the format specification is
      encountered, and there are no further items in the iolist.

      The format controller terminates I/O.

   ■  A colon (:) edit descriptor is encountered, and there are no further
      items in the iolist.

      The format controller terminates I/O.

   ■  A colon (:) edit descriptor is encountered, but there are further items
      in the iolist.

      The colon edit descriptor is ignored.

   ■  The matching final right parenthesis of the format specification is
      encountered, and there are further items in the iolist.

      The file is positioned at the beginning of the next record and the
      format controller starts over at the beginning of the format
      specification terminated by the last preceding right parenthesis.

      If there is no such preceding right parenthesis, the format controller
      rescans the format from the beginning. Within the portion of the format
      rescanned, there must be at least one repeatable edit descriptor.

      If the rescan of the format specification begins with a repeated nested
      format specification, the repeat factor indicates the number of times
      to repeat that nested format specification. The rescan does not change
      the previously set scale factor or the BN or BZ blank control in
      effect.

 When the format controller terminates on input, any remaining characters of
 the record are ignored. When the format controller terminates on output, an
 end-of-record mark is written.

 For units connected to terminal devices, the end-of-record mark is not
 written until the next record is written to the unit. «If the device is the
 screen, you can use the backslash (/) edit descriptor to suppress the
 end-of-record mark.»


 3.8  List-Directed I/O

 A list-directed record is a sequence of values and value separators. Each
 value in a list-directed record must be one of the following:

   ■  A constant, optionally multiplied by an unsigned-nonzero-integer
      constant. For example, 5, or 2*5 (two successive fives) are both
      acceptable.

   ■  A null value, optionally multiplied by an unsigned-nonzero-integer
      constant. For example, 5* is five successive null values.

 Except in string constants, none of these may have embedded blanks.

 Each value separator in a list-directed record must be one of the following:

   ■  A comma (,).

   ■  A slash (/).

      A slash encountered as a value separator during execution of a
      list-directed input statement stops execution of that statement after
      assignment of the previous value. Any further items in the input list
      are treated as if they were null values.

   ■  One or more contiguous blanks between two constants or after the last
      constant.

      Blanks next to value separators are ignored. For example,  5  ,  6  /
      7  is equivalent to 5,6/7.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    List-directed I/O to or from internal files is prohibited by the ANSI
    standard.
 ───────────────────────────────────────────────────────────────────────────


 3.8.1  List-Directed Input

 In most cases, all the input forms available for formatted I/O are also
 available for list-directed formatting. This section describes all the
 exceptions to this rule.

 The following rules apply to list-directed input for all values:

   ■  The form of the input value must be acceptable for the type of input
      list item.

   ■  Blanks are always treated as separators and never as zeros.

   ■  Embedded blanks can only appear within character constants, as
      specified in the list below.

      Note that the end-of-record mark has the same effect as a blank, except
      when it appears within a character constant.

 In addition to the rules above, the following restrictions apply to the
 specified values:

 Type of Value        Restrictions

 Single- or           A real or double-precision constant must be a numeric
 double-precision     input field (a field suitable for F editing). It is
 real constants       assumed to have no fractional digits unless there is a
                      decimal point within the field.

 Complex constants    A complex constant is an ordered pair of real «or
                      integer» constants separated by a comma and surrounded
                      by opening and closing parentheses. The first constant
                      of the pair is the real part of the complex constant,
                      and the second is the imaginary part.

 Logical Constants    A logical constant must not include either slashes or
                      commas among the optional characters permitted for L
                      editing.

 Character Constants  A character constant is a nonempty string of characters
                      enclosed in apostrophes. Each apostrophe within a
                      character constant delimited by apostrophes must be
                      represented by two apostrophes, with no intervening
                      blanks.

                      Character constants may be continued from the end of
                      one record to the beginning of the next; the end of the
                      record doesn't cause a blank or other character to
                      become part of the constant. The constant may be
                      continued on as many records as needed and may include
                      the blank, comma, and slash characters.

                      If the length n of the list item is less than or equal
                      to the length m of the character constant, the leftmost
                      n characters of the latter are transmitted to the list
                      item.

                      If n is greater than m, the constant is transmitted to
                      the leftmost m characters of the list item. The
                      remaining n minus m characters of the list item are
                      filled with blanks. The effect is the same as if the
                      constant were assigned to the list item in a character
                      assignment statement.

 Null Values          You can specify a null value in one of three ways:

                      1. No characters between successive value separators

                      2. No characters preceding the first value separator in
                         the first record read by each execution of a list-
                         directed input statement

                      3. The r* form (for example, 10* is equivalent to 10
                         null values)

                      A null value has no effect on the current definition of
                      the corresponding input list item. If the input list
                      item is defined, it retains its previous value; if it
                      is undefined, it remains so.

                      A slash encountered as a value separator during
                      execution of a list-directed input statement stops
                      execution of that statement after the assignment of the
                      previous value. Any further items in the input list are
                      treated as if they were null values.

 Blanks               All blanks in a list-directed input record are
                      considered to be part of some value separator, except
                      for the following:

                      ■  Blanks embedded in a character constant

                      ■  Leading blanks in the first record read by each
                         execution of a list-directed input statement unless
                         immediately followed by a slash or comma


 Example

 The following example uses list-directed input and output:

            REAL    a
            INTEGER i
            COMPLEX c
            LOGICAL up, down
            DATA a /2358.2E-8/, i /91585/, c /(705.60,819.60)/
            DATA up /.TRUE./, down /.FALSE./
            OPEN   (UNIT = 9, FILE = 'listout', STATUS = 'NEW')
            WRITE  (9, *) a, i
            WRITE  (9, *) c, up, down
            REWIND (9)
            READ   (9, *) a, i
            READ   (9, *) c, up, down
            WRITE  (*, *) a, i
            WRITE  (*, *) c, up, down
            END

 This program produces the following output:

      2.358200E-005        91585
      (705.6000000,819.6000000) T F


 3.8.2  List-Directed Output

 The form of the values produced by list-directed output is the same as the
 form of values required for input, except as noted in this section. The
 list-directed line length is 79 columns.

 New records are created as necessary, but neither the end of a record nor
 blanks can occur within a constant (except in character constants). To
 provide carriage control when the record is printed, each output record
 automatically begins with a blank character.

 In addition, the following rules apply for the specified types of data:

 Type of Data             Characteristics

 Logical constants        Output as T for the value true and F for the value
                          false.

 Integer constants        Output in the format of an I11 edit descriptor.

 Single- and double-      Output in the format of either an F or E edit
 precision real           descriptor, depending on the value of the constant.
 constants
                          If:                     Then:

                          1 <= constant and       The constant
                          constant < 10^(7)       is output
                                                  using a 0PF15.6 edit
                                                  descriptor for
                                                  single-precision, or a
                                                  0PF24.15 edit descriptor
                                                  for double-precision.

                          constant < 1 or         The constant is
                          constant <= 10^(7)      output using
                                                  a 1PE15.6E2 edit descriptor
                                                  for single-precision, or a
                                                  1PE24.15E3 edit descriptor
                                                  for double-precision.

 Character constants      Not delimited by apostrophes. They are neither
                          preceded nor followed by a value separator.

                          Each internal apostrophe is represented by one
                          externally. A blank character is inserted at the
                          start of any record that begins with the
                          continuation of a character constant from the
                          preceding record.

 Slashes as value         Not produced by list-directed formatting.
 separators

 Null values              Not produced by list-directed formatting.


 Example

 The following example uses list-directed output:

            INTEGER i, j
            REAL    a, b
            LOGICAL on, off
            CHARACTER*20  c
            DATA i /123456/, j /500/, a /28.22/, b /.0015555/
            DATA on /.TRUE./, off/.FALSE./
            DATA c /'Here''s a string'/
            WRITE (*, *) i, j
            WRITE (*, *) a, b, on, off
            WRITE (*, *) c
            END

 The example above produces the following output:

             123456          500
              28.2200000  1.555500E-003 T F
      Here's a string


 3.9  «Namelist-Directed I/O»

 «Namelist-directed I/O is a powerful method for reading data in or writing
 data out to a file (or the terminal). By specifying one or more variables in
 a namelist group, you can read or write the values of all of them with a
 single I/O statement.»

 «A namelist-directed input statement scans the input file for the group
 name. Once found, the statement then scans for assignment statements that
 give values to one or more of the variables in the group. Namelist-directed
 input is terminated with a slash. A namelist-directed output statement
 writes the name of the namelist, followed by the name of each variable in
 the namelist, an equal sign, and the variable's current value.
 Namelist-directed output is terminated with a slash.»

 «A namelist group is created with the NAMELIST statement. It takes the
 form:»

 «NAMELIST / namelist / variablelist»

 «where namelist is an identifying name for the group, and variablelist is a
 list of variables and array names.»

 «The values of the namelist variables are written to a file or the screen
 with a WRITE statement in which namelist appears instead of a format
 specifier. Note that no iolist is needed or permitted.»

 «WRITE (*, NML = namelist)»

 «NML= is optional, and is only required if other keywords (such as END=) are
 used.»

 «The first output record is an ampersand (&), immediately followed by the
 namelist group name, in uppercase. Succeeding records list all variable
 names in the group and their values. Each output record begins with a blank
 character to provide carriage control if the record is printed. Values take
 the output format they would have in list-directed I/O, with one exception:
 character strings are delimited by apostrophes. This permits the file
 created to be read by a namelist-directed READ statement, which requires
 apostrophes as string delimiters. The last output record is a slash.»


 «Example»

 «The following example declares a number of variables which are placed in a
 namelist, initialized, then written to the screen with namelist I/O:»

           «INTEGER    int1*1, int2*2, int4*4, array(3)
            LOGICAL    log1*1, log2*2, log4*4
            REAL       real4*4, real8*8
            COMPLEX    z8*8, z16*16
            CHARACTER  char1*1, char10*10

            NAMELIST /example/ int1, int2, int4, log1, log2, log4,
           +          real4, real8, z8, z16, char1, char10, array

            int1     =  11
            int2     =  12
            int4     =  14
            log1     = .TRUE.
            log2     = .TRUE.
            log4     = .TRUE.
            real4    =  24.0
            real8    =  28.0d0
            z8       = (38.0,0.0)
            z16      = (316.0d0,0.0d0)
            char1    = 'A'
            char10   = '0123456789'
            array(1) = 41
            array(2) = 42
            array(3) = 43

            WRITE (*, example)»


 «Output»

      «&EXAMPLE
      INT1 =         11
      INT2 =         12
      INT4 =         14
      LOG1 = T
      LOG2 = T
      LOG4 = T
      REAL4 =     24.000000
      REAL8 =     28.000000000000000
      Z8 =           (38.000000,0.000000E+00)
      Z16 =         (316.000000000000000,0.000000000000000E+000)
      CHAR1 = 'A'
      CHAR10 = '0123456789'
      ARRAY =            41          42          43
      /»

 «The operation of a namelist-directed READ statement is almost the reverse
 of a WRITE operation. The statement first scans the file (either at the
 terminal or on disk) from its current position until it finds an ampersand
 immediately followed by the namelist group name, or until it reaches the end
 of the file. (Ampersands followed by other names are ignored.) There must be
 at least one blank or carriage return following the group name to separate
 it from the following value-assignment pairs.»

 «A value-assignment pair (optional) consists of a variable name, array
 element, or substring followed by an equal sign and one or more values and
 value separators. The equal sign may be preceded or followed by any number
 of blanks (including no blanks). A value separator is a single comma or one
 or more blanks. A comma that is not preceded by a value is treated as a null
 value, and the corresponding variable or array element is not altered.»

 «Variables may appear in any order. The same variable may appear in more
 than one assignment pair. Its final value is the value it received in its
 last assignment. All the variables in a namelist do not need to be assigned
 values; those which do not appear, or which are associated with null values,
 keep their current values. A variable name in the input file that is not in
 the namelist group causes a run-time error.»

 «If an array name appears without a qualifying subscript, the first value in
 the assignment statement is given to the first array element, the second to
 the second element, and so forth. Assignment to arrays is by row-major
 order.»

 «You may not assign more values than there are elements in an array. For
 example, you may not specify 101 values for a 100-element array. However, an
 array need not have values assigned to all its elements. Any missing values
 are treated as nulls, and the corresponding array elements are not altered.
 Individual values may also be assigned to subscripted array elements.»

 «A value may be repeated by placing a repeat factor and an asterisk in front
 of the value. For example, 7*'Hello'  assigns 'Hello'  to the next seven
 elements in an array or variable list. A repeat factor and asterisk without
 a value indicates multiple null values. The corresponding variables are not
 altered. Given the array matrix(0:101), the following statements assign 10
 to element 0, assign 25 to the elements 1 through 50, leave 51 through 100
 alone, assign -101 to element 101, then change the value of matrix(42) to
 63:»

            «matrix = 10, 50*25, 50*, -101
            matrix(42) = 63»

 «Character strings must be delimited by apostrophes.»

 «A namelist-directed READ statement is successfully terminated by a slash or
 by reaching the end of the file (in which case, an error results). Slashes
 must not be used as value separators unless it is desired to prematurely
 terminate the read.»

 «Suppose the preceding program wanted to read new values for some of the
 variables in namelist example. If a file connected to unit four contained
 the following namelist specifier and assignment statements:»

            «&example
            z8 = (99.0,0.0)
            INT1=99
            array(1)=99
            real4 = 99.0
            CHAR1='Z'
            char10(5:9) = 'GrUMp'
            log1=F
            /»

 «then the following namelist-directed READ statement would assign new values
 to the specified variables:»

            «READ (UNIT = 4, example)»

 «A second WRITE (*, example) statement would display their changed values as
 follows:»

      «&EXAMPLE
      INT1 =         99
      INT2 =         12
      INT4 =         14
      LOG1 = F
      LOG2 = T
      LOG4 = T
      REAL4 =     99.000000
      REAL8 =     28.000000000000000
      Z8 =           (99.000000,0.000000E+00)
      Z16 =         (316.000000000000000,0.000000000000000E+000)
      CHAR1 = 'Z'
      CHAR10 = '0123GrUMp9'
      ARRAY =            99          42          43
      /»


 Chapter 4  Statements
 ───────────────────────────────────────────────────────────────────────────

 The first part of this chapter describes the kinds of statements available
 in Microsoft FORTRAN. The second part of the chapter contains a directory of
 the Microsoft FORTRAN statements, listed alphabetically.

 A FORTRAN statement consists of an initial line, optionally followed by up
 to 19 continuation lines. «In Microsoft FORTRAN, the number of continuation
 lines is limited only by available memory.» Statements are written in
 columns 7 through 72. They perform actions such as computing, storing the
 results of computations, altering the flow of control, reading and writing
 files, and providing information for the compiler.


 4.1  Categories of Statements

 There are two basic types of statements in FORTRAN: executable and
 nonexecutable. An executable statement causes an action to be performed.
 Nonexecutable statements describe, classify, or specify the format of
 program elements, such as entry points, data, or program units.

 Table 4.1 summarizes the FORTRAN statements.

 Table 4.1  Categories of FORTRAN Statements

 Category                   Type                       Description

 Assignment statements      Executable                 Assign a value to a
                                                       variable or an array
                                                       element. See the
                                                       ASSIGN and
                                                       Assignment entries in
                                                       Section 4.2,
                                                       "Statement Directory,"
                                                       for more
                                                       information.

 BLOCK DATA, ENTRY,         Nonexecutable              Define the start of a
 FUNCTION,                                             program unit and
 «INTERFACE TO,»                                       specify its formal
 PROGRAM, and                                          arguments.
 SUBROUTINE
 statements

 Control statements         Executable                 Control the order of
                                                       execution of
                                                       statements. See Table
                                                       4.3.

 DATA statement             Nonexecutable              Assigns initial values
                                                       to variables.

 FORMAT statement           Nonexecutable              Provides data-editing
                                                       information.

 I/O statements             Executable                 Transfer data and
                                                       manipulate files and
                                                       records. See Table 4.4
                                                       and Chapter 3, "The
                                                       Input/Output
                                                       System."

 Specification              Nonexecutable              Define the attributes
 statements                                            of variables, arrays,
                                                       and subprograms. See
                                                       Table 4.2.

 Statement-function         Nonexecutable              Define simple, locally
 statements                                            used functions.

 Table 4.2 summarizes the specification statements.

 Table 4.2  Specification Statements

 Statement                           Purpose

 «AUTOMATIC»                         «Declares a variable on the stack,
                                     rather than at a static memory
                                     location»

 COMMON                              Shares variables between two or more
                                     program units

 DIMENSION                           Identifies a variable as an array and
                                     specifies the number of elements

 EQUIVALENCE                         Specifies that two or more variables or
                                     arrays share the same memory location

 EXTERNAL                            Allows a user-defined subroutine or
                                     function to be passed as an argument

 IMPLICIT                            Changes the default typing for real and
                                     integer variables and functions

 INTRINSIC                           Allows a predefined function to be
                                     passed as an argument

 «MAP...END MAP»                     «Within a UNION statement, delimits a
                                     group of variable type declarations that
                                     are to be ordered contiguously within
                                     memory»

 «NAMELIST»                          «Declares a group name for set of
                                     variables to be read or written in a
                                     single statement»

 PARAMETER                           Equates a constant expression with a
                                     name

 «RECORD»                            «Declares one or more variables of a
                                     user-
                                     defined structure type»

 SAVE                                Causes variables to retain their values
                                     between invocations of the procedure in
                                     which they are defined

 «STRUCTURE...END STRUCTURE»         «Defines a new variable type, composed
                                     of a collection of other variable types»

 Type:                               Specifies the type of user-defined names
 CHARACTER«*n»
 COMPLEX««*bytes»»
 «DOUBLE COMPLEX»
 DOUBLE PRECISION
 INTEGER««*bytes»»
 LOGICAL««*bytes»»
 REAL««*bytes»»
 «RECORD /struct-name/»
 «STRUCTURE /struct-name/»

 «UNION...END UNION»                 «Within a STRUCTURE statement, causes
                                     the variables in two or more maps to
                                     occupy the same memory location»

 Table 4.3 summarizes the control statements.

 Table 4.3  Control Statements

 Statement                  Purpose

 «ALLOCATE»                 «Dynamically establishes allocatable array
                            dimensions»

 CALL                       Executes a subroutine

 «CASE»                     «Within a SELECT CASE structure, marks a block of
                            statements that are executed if an associated
                            value matches the SELECT CASE expression»

 CONTINUE                   Has no effect; often used as target of GOTO or as
                            the terminal statement in a DO loop

 «CYCLE»                    «Advances control to the end statement of a DO
                            loop; the intervening loop statements are not
                            executed»

 «DEALLOCATE»               «Frees the storage space previously reserved in
                            an ALLOCATE statement»

 DO                         Evaluates statements in the DO loop, through and
                            including the ending statement, a specific number
                            of times

 «DO WHILE»                 «Evaluates statements in the DO WHILE loop,
                            through and including the ending statement, until
                            a logical condition becomes false»

 ELSE                       Introduces an ELSE block

 ELSE IF                    Introduces an ELSE IF block

 END                        Ends execution of a program unit

 «END DO»                   «Marks the end of a series of statements
                            following a DO or DO WHILE statement»

 END IF                     Marks the end of a series of statements following
                            a block IF statement

 «END SELECT»               «Marks the end of a SELECT CASE  statement»

 EXIT                       Leaves a DO loop; execution continues with the
                            first statement following

 GOTO                       Transfers control to another part of the program

 «INCLUDE»                  «Inserts contents of a specified file into the
                            source file»

 IF                         Controls conditional execution of other
                            statement(s)

 PAUSE                      Suspends program execution «and, optionally,
                            executes operating-system commands»

 RETURN                     Returns control to the program unit that called a
                            subroutine or function

 «SELECT CASE»              «Transfers program control to a selected block of
                            statements, based on value of a controlling
                            expression»

 STOP                       Terminates a program

 Table 4.4 summarizes the I/O statements.

 Table 4.4  I/O Statements

 Statement                  Purpose

 BACKSPACE                  Positions a file to the beginning of the previous
                            record

 CLOSE                      Disconnects the specified unit

 ENDFILE                    Writes an end-of-file record

 INQUIRE                    Returns values indicating the properties of a
                            file or unit

 «LOCKING»                  «Locks direct-access files and records»

 OPEN                       Associates a unit number with an external device
                            or file

 PRINT                      Displays data on the screen

 READ                       Transfers data from a file to the items in an I/O
                            list

 REWIND                     Repositions a file to its first record

 WRITE                      Transfers data from the items in an I/O list to a
                            file


 4.2  Statement Directory

 The rest of this chapter is an alphabetical listing of all Microsoft FORTRAN
 statements. Each statement is described using the following format:

 Heading                                    Information

 Action                                     Summary of what the statement
                                            does.

 Syntax                                     Statement syntax, and description
                                            of statement parameters.

 Remarks                                    Use of the statement.

 Example                                    Sample programs or program
                                            fragments that illustrate the use
                                            of the statement. This section
                                            does not appear with every
                                            reference entry.

 The syntax of statements that do not fit on one line is shown on more than
 one line, as in the following example:

      CLOSE («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»
      «, STATUS=status»)

 When you use these statements, you must still follow the formatting rules
 described in Section 2.1, "Lines." «If the $FREEFORM metacommand is
 specified, follow the formatting rules given in Section 2.3, "Free-Form
 Source Code."» The program fragment below, for example, is illegal:

            CLOSE (UNIT = 2,
            ERR = 100,
            IOSTAT = errvar)

 Either of the following two statements, however, is correct:

            CLOSE (UNIT = 2, ERR = 100, IOSTAT = errvar)

            CLOSE (UNIT = 2,
           +       ERR = 100,
           +       IOSTAT = errvar)


 ALLOCATE


 ■  «Action»

 «Dynamically sizes an array that has previously been declared with the
 ALLOCATABLE attribute»


 ■  «Syntax»

      «ALLOCATE (array(«l:»u«,«l:»u ...») «, STAT = ierr») ...»

 «Parameter                                 Description»

 «array»                                    «Name of allocatable array»

 «ierr»                                     «Integer variable that returns
                                            status of allocation attempt»

 «l»                                        «Integer expression that sets the
                                            lower bound of the array»

 «u»                                        «Integer expression that sets the
                                            upper bound of the array»


 ■  «Remarks»

 «Allocatable arrays may be dynamically allocated and deallocated at run
 time. An array must have previously been declared ALLOCATABLE, and the
 number of its dimensions declared (with colons only; no bounds may be
 specified). The ALLOCATE statement establishes the upper and lower bounds of
 each dimension and reserves sufficient memory.»

 «More than one allocatable array name may appear in an ALLOCATE statement,
 separated by commas. The STAT= parameter must appear last.»

 «Allocatable arrays may not have the NEAR attribute. If the array is to be
 larger than 65,536 bytes, you must specify the HUGE attribute so the array
 elements are correctly addressed. Allocatable arrays may not appear in
 AUTOMATIC, COMMON, DATA, or EQUIVALENCE statements.»

 «Attempting to reallocate a previously allocated array causes a run-time
 error.»

 «Any allocation failure causes a run-time error, unless the STAT= option is
 present. The ierr variable returns a value of zero if the allocation was
 successful, and the number of the run-time error if the allocation failed.»


 ■  «Example»

           «.
            .
            .
            INTEGER  dataset[ALLOCATABLE](:,:),
           +         results[ALLOCATABLE, HUGE](:,:,:)
            INTEGER  reactor, level, calcs, error
            DATA  reactor, level, calcs  / 10, 50, 100 /

            ALLOCATE (dataset(reactor,level),
           +          results(reactor,level,calcs), STAT = error)

            IF (error .NE. 0)
           +    STOP 'Not enough storage for data; aborting...'
            .
            .
            .»


 ■  «See also»

 «DEALLOCATE»


 ASSIGN (Label Assignment)


 ■  Action

 Assigns the value of a format or statement label to an integer variable


 ■  Syntax

      ASSIGN label TO variable

 Parameter                                  Description

 label                                      A format label or statement
                                            label. The label must appear in
                                            the same program unit.

 variable                                   An integer variable.


 ■  Remarks

 Variables with label values are used in the following situations:

 Situation                                  Use

 An assigned GOTO statement                 The assigned GOTO statement
                                            requires a variable with the
                                            value of the label of an
                                            executable statement.

 A format specifier                         Input/output statements can
                                            accept a variable which specifies
                                            the label of a FORMAT statement.

 The value of a label is not the same as the label number; the label is
 instead identified by a number assigned by the compiler. For example, the
 value of IVBL in the following example is not 400:

            ASSIGN 400 TO IVBL

 Therefore, variables used in ASSIGN statements are not defined as integers.
 If you want to use a variable defined by an ASSIGN statement in an
 arithmetic expression, you must first define the variable by a computational
 assignment statement or by a READ statement.

 «If you use INTEGER*1 variables for variable, note that INTEGER*1 variables
 can only be used for the first 128 ASSIGN statements in a subprogram.»


 ■  Example

      C     Assign statement label 100 to the integer variable ivar
            ASSIGN 100 TO ivar
      C     Use ivar as a FORMAT statement label
            WRITE (*, ivar)
      C     Assign statement label 200 to ivar
            ASSIGN 200 TO ivar
      C     Use ivar as the target label of an assigned GOTO statement
            GOTO ivar
            WRITE (*, *)' This is never written'
        200 CONTINUE
            WRITE (*, *)' This is written'
        100 FORMAT (' This is format 100')
            END


 Assignment (Computational)


 ■  Action

 Evaluates an expression and assigns the resulting value to the specified
 variable or array element


 ■  Syntax

      variable = expression

 Parameter                                  Description

 variable                                   A variable, array, array-element,
                                            or structure-element reference

 expression                                 Any expression


 ■  Remarks

 The variable, array, array-element, «or structure-element type» and the
 expression type must be compatible:

   ■  If expression is numeric, then variable must be numeric, and the
      statement is an arithmetic assignment statement. If the data types of
      expression and variable are not identical, expression is converted to
      the data type of variable.

      Section 1.7.1.2, "Type Conversion of Arithmetic Operands," explains how
      integer, real, and complex numbers are converted.

   ■  If expression is logical, then variable must be logical, and the
      statement is a logical assignment statement.

      «Logical expressions of any byte size can be assigned to logical
      variables of any byte size without changing the value of expression.»

      Note that integer and real expressions may not be assigned to logical
      variables, nor may logical expressions be assigned to integer or real
      variables.

   ■  If expression has the type CHARACTER, the statement is a character
      assignment statement. Both variable and expression must have type
      CHARACTER. «If the $NOTSTRICT metacommand (the default) is in effect,
      then a character expression can be assigned to a noncharacter variable,
      and a noncharacter variable or array element (but not an expression)
      can be assigned to a character variable.»

      For character assignment statements, if the length of expression does
      not match the size of variable, expression is adjusted as follows:

        ■  If variable is longer than expression, then expression is padded
           with blanks on the right.

        ■  If variable is shorter than expression, then characters on the
           right of expression are ignored.

        ■  If variable is an array, then array assignment occurs as described
           in Section 1.7.5, "Array Expressions."


 ■  Examples

 The following program demonstrates assignment statements:

            REAL         a, b, c
            LOGICAL      abigger
            CHARACTER*5  assertion

            c = .01
            a = SQRT (c)
            b = c**2

            assertion = 'a > b'
            abigger   = (a .GT. b)

            WRITE (*, 100) a, b
       100  FORMAT (' a =', F7.4, '    b =', F7.4)

            IF (abigger) THEN
                WRITE (*, *) assertion, ' is true.'
                ELSE
                WRITE (*, *) assertion, ' is false.'
            END IF
            END

 The program above has the following output:

      a =  .1000    b =  .0001
      a > b is true.

 The following fragment demonstrates legal and illegal assignment statements:

            INTEGER  i, int
            REAL  rone(4), rtwo(4), x, y
            COMPLEX  z      CHARACTER  char6*6, char8*8

            i       =  4
            x       =  2.0
            z       = (3.0, 4.0)
            rone(1) =  4.0
            rone(2) =  3.0
            rone(3) =  2.0
            rone(4) =  1.0
            char8   = 'Hello,'

      C     The following assignment statements are legal:
            i     = rone(2)
            int   = rone(i)
            int   = x
            y     = x
            y     = z
            y     = rone(3)
            rtwo  = rone
            rtwo  = 4.7
            char6 = char8

      C     The following assignment statements are illegal:
            char6 = x + 1.0
            int   = char8//'test'
            y     = rone


 AUTOMATIC


 ■  «Action»

 «Declares specified variables to be on the stack, rather than at a static
 memory location»


 ■  «Syntax»

 «AUTOMATIC «names»»

 «Parameter                                 Description»

 «names»                                    «A list of variables or array
                                            names to be made automatic. If
                                            there is more than one variable,
                                            they must be separated by
                                            commas.»


 ■  «Remarks»

 «In Microsoft FORTRAN, all variables are static by default. A variable
 declared as "automatic" has no fixed memory location; a section of stack
 memory is allocated for the variable as needed. Automatic variables within
 procedures are discarded when the procedure completes execution. Therefore,
 such variables cannot be guaranteed to have the same value on the next
 invocation of the procedure.»

 «If an AUTOMATIC statement contains no variable names, all the variables
 within that program unit (the main program, or an individual subprogram)
 that can legally be automatic are implicitly automatic.»

 «Common-block names and variables are not allowed in an AUTOMATIC statement.
 A variable cannot appear in both a SAVE statement and an AUTOMATIC
 statement.»

 «Variables with the ALLOCATABLE, EXTERNAL, FAR, or HUGE attribute cannot be
 automatic. A variable that has been explicitly declared automatic may not
 appear in a DATA statement. Variables that are implicitly automatic and
 appear in a DATA statement will be initialized and placed in static memory.
 A variable may appear in an AUTOMATIC statement only once. Formal arguments
 and procedure names may not appear in an AUTOMATIC statement.»

 «The ability to declare automatic variables has been added to support OS/2
 multithread applications.»


 ■  «Example»

     «C     In this example, all variables within the program unit
      C     are automatic, except for "clark" and "lois"; these are
      C     explicitly declared in a SAVE statement, and thus have
      C     static memory locations:

            INTEGER FUNCTION  Fibonacci (clark, lois)
            AUTOMATIC
            SAVE  clark, lois»


 BACKSPACE


 ■  Action

 Positions the file connected to the specified unit at the beginning of the
 preceding record


 ■  Syntax

      BACKSPACE {unitspec |
      («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»)}

 If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   An integer expression that
                                            specifies an external unit. If
                                            unitspec is not open, a run-time
                                            error occurs.

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit. If errlabel is specified,
                                            an I/O error causes transfer of
                                            control to the statement at
                                            errlabel. If errlabel is omitted,
                                            the effect of an I/O error is
                                            determined by the presence or
                                            absence of iocheck.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            no error occurs, or the number of
                                            the run-time error message if an
                                            error does occur. For more
                                            information on error handling,
                                            see Section 3.2.6, "Error and
                                            End-of-File Handling."


 ■  Remarks

 The BACKSPACE statement backs up by exactly one record, except in the
 following special cases:

 Special Case                               Result

 No preceding record                        The file position is not changed

 Preceding record is end-of-file            The file is positioned before the
 record                                     end-of-file record

 File position is in middle of              The file is positioned to the
 record                                     start of that record

 If a parameter of the BACKSPACE statement is an expression that calls a
 function, that function must not cause an I/O statement «or the EOF
 intrinsic function» to be executed, because the results are unpredictable.


 ■  Examples

            BACKSPACE 5
            BACKSPACE (5)
            BACKSPACE lunit
            BACKSPACE (UNIT = lunit, ERR = 30, IOSTAT = ios)


 BLOCK DATA


 ■  Action

 Identifies a block-data subprogram, where variables and array elements in
 named common blocks can be initialized


 ■  Syntax

      BLOCK DATA «blockdataname»

 Parameter                                  Description

 blockdataname                              A global symbolic name for the
                                            subprogram.

                                            This name must not be the same as
                                            any of the names for local
                                            variables or array elements
                                            defined in the subprogram labeled
                                            by blockdataname, and it must not
                                            be the same as any of the names
                                            given to the main program,
                                            external procedures, common
                                            blocks, or other block-data
                                            subprograms.


 ■  Remarks

 The BLOCK DATA statement must be the first statement in a block-data
 subprogram.

 Only one unnamed block-data subprogram may appear in an executable program.
 Otherwise, the default name will be defined twice, causing an error.

 The following restrictions apply to the use of block-data subprograms:

   ■  The only statements that may be used in a block-data subprogram are
      BLOCK DATA, COMMON, DATA, END, DIMENSION, EQUIVALENCE, IMPLICIT, «MAP,»
      PARAMETER, «RECORD,» SAVE, «STRUCTURE, UNION,» and type statements. No
      executable statements are permitted.

   ■  Only an entity defined in a named common block may be initially defined
      in a block-data subprogram.

   ■  All the constituents of a named common block must be specified in that
      block-data subprogram, even if not all of the constituents are
      initialized. This is because the length of the named common block must
      be the same in all subprograms.


 ■  Examples

      C     The following block-data subprogram initializes
      C     the named common block /greatlakes/:
      C
            BLOCK DATA Lakes
            COMMON /greatlakes/ erie, huron, michigan, ontario, superior
            DATA erie, huron, michigan, ontario, superior /1, 2, 3, 4, 5/
            END
      C
      C     Using the same common block, /greatlakes/, the
      C     following block-data subprogram is NOT allowed;
      C     not all the members of /greatlakes/ are specified.

            BLOCK DATA GrLaks
            COMMON /greatlakes/ erie, huron, ontario, superior
            DATA erie, huron, ontario, superior /1, 2, 4, 5/
            END


 CALL


 ■  Action

 Invokes a subroutine


 ■  Syntax

      CALL sub «(«actuals»)»

 Parameter                                  Description

 sub                                        The name of the subroutine to be
                                            executed.

 actuals                                    One or more actual arguments.

                                            If there is more than one
                                            argument, they are separated by
                                            commas. Each argument can be an
                                            alternate-return specifier
                                            (*label); a constant, variable,
                                            or expression; an array or array
                                            element; the name of a subroutine
                                            or external function; the name of
                                            an intrinsic function that can be
                                            passed as an argument«; or a
                                            Hollerith constant».


 ■  Remarks

 Execution of a CALL statement proceeds as follows:

   1. Arguments that are expressions are evaluated.

   2. Actual arguments are associated with their corresponding formal
      arguments.

   3. The body of the specified subroutine is executed.

   4. Control returns to the calling program or procedure, either to a
      statement specified by an alternate return or to the statement
      following the CALL statement.

 A subroutine can be called from any program unit.

 A CALL statement must contain as many actual arguments as there are formal
 arguments in the corresponding SUBROUTINE statement «(unless the C and
 VARYING attributes were used in declaring the subroutine)».

 If a SUBROUTINE statement contains no formal arguments, a CALL statement
 referencing that subroutine must not include any actual arguments. However,
 an empty pair of parentheses can follow sub.

 Formal arguments and their corresponding actual arguments must have the same
 data type (except for Hollerith constants). «When the actual argument is a
 Hollerith constant, the formal argument need not be the same type, as long
 as it is of type INTEGER, REAL, or LOGICAL.»

 For all arguments passed by reference «(see Section 1.6.11, "VALUE," for
 information on passing arguments by value)», the compiler assumes the type
 of the formal argument is the same as the type of the corresponding actual
 argument. If the type of the formal argument is known, it is used only to
 check that the arguments have the same data type.

 If a subroutine call appears more than once in a program unit, the compiler
 checks that the number and types of actual arguments passed are the same in
 each call (i.e., consistency).

 The compiler also checks that the actual arguments used in calls to a
 subroutine correspond in number and type to its formal arguments (i.e.,
 validity). In order to do this, the SUBROUTINE statement «(or an INTERFACE
 TO statement defining the subroutine)» must appear in the same source file
 as any calls to it, and it must precede those calls. The compiler will then
 ensure that all actual arguments agree with the formal arguments. If the
 actual arguments are not checked in this way, and if they do not agree with
 the formal arguments, the result of calling a subroutine is unpredictable.

 ───────────────────────────────────────────────────────────────────────────
 «NOTE»
    «When passing integer and logical arguments, you must pay attention to
    type agreement. The $STORAGE metacommand controls how integer and logical
    arguments are passed. When the default ($STORAGE:4) is in effect, all
    actual arguments that are integer or logical constants or expressions are
    assigned to INTEGER*4 or LOGICAL*4 temporary variables. When $STORAGE:2
    is in effect, all actual arguments that are integer or logical constants
    or expressions are assigned to temporary variables of type INTEGER*2 or
    LOGICAL*2.»

    «Therefore, if an integer or logical formal argument is of a different
    type than the default storage type, you must convert the actual argument
    to the same type. Use the INT2 and INT4 intrinsic functions within the
    actual argument list, as described in Section 5.1.1, "Data-Type
    Conversion."»
 ───────────────────────────────────────────────────────────────────────────

 The alternate-return feature lets you specify the statement to which a
 subroutine should return control. To use the alternate-return feature,
 follow these steps:

   1. Choose the statements in the calling routine to which you wish to
      return control. Enter the labels of these statements, preceded by
      asterisks, in the actual argument list of the CALL statement, as in
      this statement:

        CALL INVERT (row, column, *100, *200, *500)

   2. In the corresponding SUBROUTINE statement, enter asterisks for the
      formal arguments corresponding to the *label actual arguments in the
      CALL statement, as in the following SUBROUTINE statement:

        SUBROUTINE INVERT (r, c, *, *, *)

   3. In the subroutine, have at least one RETURN statement for each
      alternate return. As arguments for these RETURN statements, specify a 1
      for the RETURN statement that should return control to the first
      statement label in the CALL statement; a 2 for the RETURN statement
      that should return control to the second statement label in the CALL
      statement; and so on.

      For example, if the statement RETURN 1 is reached in the program
      containing the two statements in steps 1 and 2 above, control returns
      to the statement at label 100 in the calling routine; if RETURN 2 is
      reached, control returns to the statement at label 200; and if RETURN 3
      is reached, control returns to the statement at label 500. If a RETURN
      statement without any number is reached (RETURN), or a RETURN statement
      that has a number for which there is no corresponding return label is
      reached (such as RETURN 4, in this example), control returns to the
      statement following the CALL statement in the calling routine.

      A "recursive" subroutine is one that calls itself, or calls another
      subprogram which in turn calls the first subroutine before the first
      subroutine has completed execution. FORTRAN does not support recursive
      subroutine calls.


 ■  Examples

            .
            .
            .
            IF (ierr .NE. 0)  CALL Error (ierr)
            END
      C
            SUBROUTINE Error (ierrno)
            WRITE (*, 200) ierrno
        200 FORMAT (1X, 'error', I5, ' detected')
            END

      C     This example illustrates the alternate return feature:
          1 CALL Boomerang (count, *10, j, *20, *30)
            WRITE (*, *) 'normal return'
            GOTO 40
         10 WRITE (*, *) 'returned to 10'
            GOTO 40
         20 WRITE (*, *) 'returned to 20'
            GOTO 40
         30 WRITE (*, *) 'returned to 30'
         40 CONTINUE
            .
            .
            .
            SUBROUTINE Boomerang (i, *, j, *, *)
            IF (i .EQ. 10)  RETURN 1
            IF (i .EQ. 20)  RETURN 2
            IF (i .EQ. 30)  RETURN 3
            RETURN


 CASE


 ■  «Action»

 «Marks the beginning of a block of statements executed if an item in a list
 of expressions matches the test expression»


 ■  «Syntax»

      «CASE {DEFAULT | (expressionlist)}
                    statementblock»

 «Parameter                                 Description»

 «DEFAULT»                                  «The keyword indicating that the
                                            following statement block is to
                                            be executed if none of the
                                            expressions in any other CASE
                                            statements match the test
                                            expression in the SELECT CASE
                                            statement.»

 «expressionlist»                           «A list of values and ranges of
                                            values, which must be constant
                                            and must match the data type of
                                            the test expression in the SELECT
                                            CASE statement. The values must
                                            be of type INTEGER, LOGICAL, or
                                            CHARACTER*1. If testexpr matches
                                            one of the values, the following
                                            block of statements is executed.»

 «statementblock»                           «One or more executable
                                            statements. The block may also be
                                            empty.»


 ■  «Remarks»

 «The CASE statement may only appear within the SELECT CASE...END CASE
 construct.»

 «There are two ways to include values in the expressionlist. The first is to
 give a list of individual values, separated by commas. The second is to
 specify an inclusive range of values, separated by a colon, such as 5:10. If
 the lower bound is omitted (such as :10), then all values less than or equal
 to the upper bound match. If the upper bound is omitted (such as 5:), then
 all values greater than or equal to the lower bound match. Ranges of
 printable ASCII characters may be included, such as 'I':'N' or ' ':'/' .
 Both individual values and ranges of values may be included in the same
 expressionlist. You cannot specify a range of values when testexpr is of
 type LOGICAL. A given value (even when specified implicitly as part of a
 range) can only appear in one expressionlist.»

 «A statementblock need not contain executable statements. Empty blocks can
 be used to make it clear that no action is to be taken for a particular set
 of expression values.»

 «The CASE DEFAULT statement is optional. You can include only one CASE
 DEFAULT statement in a SELECT CASE block.»

 «If the value of testexpr does not match any value in any expressionlist,
 execution passes beyond the SELECT CASE construct to the next executable
 statement.»


 ■  «Example»

           «CHARACTER*1  cmdchar

            SELECT CASE (cmdchar)
            CASE ('0')
            WRITE (*, *) "Must retrieve one to nine files"
            CASE ('1':'9')
            CALL RetrieveNumFiles (cmdchar)
            CASE ('A', 'd')
            CALL AddEntry
            CASE ('D', 'd')
            CALL DeleteEntry
            CASE ('H', 'h')
            CALL Help
            CASE ('R':'T', 'r':'t')
            WRITE (*, *) "REDUCE, SPREAD and TRANSFER commands ",
           +                          "not yet supported"
            CASE DEFAULT
            WRITE (*, *) "Command not recognized; please re-enter"
            END SELECT»


 ■  «See also»

 «SELECT CASE...END SELECT»


 CHARACTER


 ■  Action

 Specifies the CHARACTER type for user-defined names


 ■  Syntax

      CHARACTER«*chars» vname ««[attrs]»»«*length»«(dim)»
      ««/values/»»«, vname ««[attrs]»»«*length»«(dim)»»
      ««/values/»»...

 The order of the dim and length parameters can be reversed.

 Parameter                                  Description

 chars                                      An unsigned integer constant in
                                            the range 1 through 32,767; an
                                            integer constant expression
                                            (evaluating to an integer between
                                            1 and 32,767) in parentheses; or
                                            an asterisk (*) (see REMARKS,
                                            below) in parentheses. The chars
                                            parameter specifies the length,
                                            in characters, of the items
                                            specified in the CHARACTER
                                            statement. This value can be
                                            overridden by the length
                                            parameter.

 vname                                      The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function, or
                                            intrinsic function; or a function
                                            subprogram or an array
                                            declarator. The parameter vname
                                            cannot be the name of a
                                            subroutine or main program.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            vname. The following attributes
                                            can be used with vname: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR, PASCAL, REFERENCE,
                                            VALUE.»

 length                                     An unsigned integer constant in
                                            the range 1 through 32,767; an
                                            integer constant expression
                                            (evaluating to an integer between
                                            1 and 32,767) in parentheses; or
                                            an asterisk (*) in parentheses.
                                            The length parameter specifies
                                            the length, in characters, of the
                                            vname immediately preceding it.
                                            This value, if specified,
                                            overrides the length indicated by
                                            chars.

 dim                                        A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant is written in
                                            the form n*constant, where n is a
                                            positive integer constant, and is
                                            equivalent to constant repeated n
                                            times. The /values/ option, if
                                            specified, initializes vname. The
                                            following statement declares word
                                            of type CHARACTER, and sets word
                                            equal to 'start':»

                                               «CHARACTER*5 word /'start'/»


 ■  Remarks

 A CHARACTER statement confirms or overrides the implicit type of vname. The
 name vname is defined for the entire program unit and cannot be defined by
 any other type statement in that program unit.

 An asterisk in parentheses ((*)) as a length specifier indicates that the
 length is specified elsewhere. An asterisk length specifier is allowed in
 the following cases:

   1. Character constants defined by PARAMETER statements. The actual length
      is determined by the length of the character string assigned to the
      parameter.

   2. Formal character arguments. The actual length is determined by the
      length of the actual argument.

   3. Character functions that are defined in one program unit, and
      referenced in another. The actual length is determined by a type
      declaration in the program unit that references the function.

 If neither length nor chars is specified, the length defaults to one.

 CHARACTER statements must precede all executable statements.


 ■  Examples

            CHARACTER  wt*10, city*80, ch

            CHARACTER  name(10)*20, eman*20(10)

            CHARACTER*20  tom, dick, harry*12, tarzan, jane*34


 CLOSE


 ■  Action

 Disconnects a specified unit


 ■  Syntax

      CLOSE («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»
      «, STATUS=status»)

 If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   An integer expression that
                                            specifies an external unit. No
                                            error occurs if the unit is not
                                            open.

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit. If errlabel is specified,
                                            an I/O error causes transfer of
                                            control to the statement at
                                            errlabel. If errlabel is omitted,
                                            the effect of an I/O error is
                                            determined by the presence or
                                            absence of iocheck.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            there is no error, or the number
                                            of the run-time error if an error
                                            occurs. For more information on
                                            error handling, see Section
                                            3.2.6, "Error and End-of-File
                                            Handling."

 status                                     A character expression that
                                            evaluates to either 'KEEP' or
                                            'DELETE'.

                                            Files opened without a file name
                                            are "scratch" files. For these
                                            files, the default for status is
                                            'DELETE'. Scratch files are
                                            temporary, and are always deleted
                                            upon normal program termination;
                                            specifying STATUS='KEEP' for
                                            scratch files causes a run-time
                                            error. The default for status for
                                            all other files is 'KEEP'.


 ■  Remarks

 Opened files do not have to be explicitly closed. Normal termination of a
 program will close each file according to its default status.

 «Closing unit 0 automatically reconnects unit 0 to the keyboard and screen.
 Closing units 5 and 6 automatically reconnects those units to the keyboard
 or screen, respectively.» Closing the asterisk (*) unit causes a
 compile-time error.

 If a parameter of the CLOSE statement is an expression that calls a
 function, that function must not cause an I/O operation «or the EOF
 intrinsic function» to be executed, because the results are unpredictable.


 ■  Example

      C     Close and discard file:
            CLOSE (7, STATUS = 'DELETE')


 COMMON


 ■  Action

 Allows two or more program units to directly share variables, without having
 to pass them as arguments


 ■  Syntax

      COMMON «/«cname»««[attrs]»»/» nlist ««,»/«cname»
      ««[attrs]»»/nlist»...

 Parameter                                  Description

 cname                                      A name for the common block. If
                                            none is given, the block is said
                                            to be "blank common." Omitting
                                            cname specifies that all elements
                                            in the nlist that follows are in
                                            the blank common block.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attributes
                                            describe cname. Only ALIAS, C,
                                            FAR, NEAR, and PASCAL can be used
                                            with common-block names.»

 nlist                                      A list of variable names, array
                                            names, and array declarators,
                                            separated by commas.

                                            Formal-argument names, function
                                            names, «automatic variables, and
                                            allocatable arrays» cannot appear
                                            in a COMMON statement. In each
                                            COMMON statement, all variables
                                            and arrays appearing in each
                                            nlist following a common-block
                                            name are declared to be in that
                                            common block.


 ■  Remarks

 Any common block can appear more than once in the same program unit. The
 list of elements in a particular common block is treated as a continuation
 of the list in the previous common block with the same name. For this
 reason, a given name can appear only once in all the common blocks of a
 program unit. Consider the following COMMON statements:

            COMMON /ralph/ ed, norton, trixie
            COMMON /     / fred, ethel, lucy
            COMMON /ralph/ audrey, meadows
            COMMON /jerry/ mortimer, tom, mickey
            COMMON         melvin, purvis

 They are equivalent to these COMMON statements:

            COMMON /ralph/ ed, norton, trixie, audrey, meadows
            COMMON         fred, ethel, lucy, melvin, purvis
            COMMON /jerry/ mortimer, tom, mickey

 All items in a common block must be only character or only noncharacter
 items. «Microsoft FORTRAN permits both character and noncharacter items in
 the same common block.»

 Microsoft FORTRAN normally starts all variables and arrays (including
 character variables and arrays) on even-byte addresses. If a single
 (non-array) character variable has an «»odd length, the next variable in
 memory (regardless of type) begins at the next even-byte address.«

 There are, however,  two exceptions. In an array of character variables,
 each element immediately follows the preceding element, without regard for
 even or odd addresses. Within a common block, a character variable that
 follows another character variable always begins on the next available byte.
 (Noncharacter variables still begin on an even byte.)

 «This can cause problems if a common block contains odd-length character
 variables interspersed with noncharacter variables. If the matching common
 block in another program unit contains only noncharacter variables (in
 effect, equivalencing the variables), there will be holes in the equivalence
 due to unused bytes in the block with the odd-length character variables.»

 The length of a common block equals the number of bytes of memory required
 to hold all elements in that common block. If several distinct program units
 refer to the same named common block, the common block must be the same
 length in each program unit. The blank common block, however, can have a
 different length in different program units. The blank common block is as
 long as the longest blank common block in any of the the program units.

 A variable that appears in a common block may not be initialized in a DATA
 statement unless that DATA statement is part of a block data subprogram. «In
 Microsoft FORTRAN, common-block variables may be initialized in DATA
 statements in any program unit.»

 «An item that appears in nlist cannot be initialized in a type statement.
 The following example causes a compile-time error:»

           «INTEGER i /1/
            COMMON i»


 ■  Example

            PROGRAM MyProg
            COMMON i, j, x, k(10)
            COMMON /mycom/ a(3)
            .
            .
            .
            END

            SUBROUTINE MySub
            COMMON pe, mn, z, idum(10)
            COMMON /mycom/ a(3)
            .
            .
            .
            END


 COMPLEX


 ■  Action

 Specifies the COMPLEX type for user-defined names


 ■  Syntax

      COMPLEX««*bytes»» vname««[attrs]»«*length»»«(dim)»
      ««/values/»»«,vname««[attrs]»«*length»»
      «(dim)»««/values/»»»...

 The order of the length and dim parameters can be reversed.

 Parameter                                  Description

 «bytes»                                    «Must be 8 or 16. The bytes
                                            parameter specifies the length,
                                            in bytes, of the items specified
                                            by the COMPLEX statement. This
                                            value can be overridden by the
                                            length parameter.»

 vname                                      The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function,
                                            intrinsic function, function
                                            subprogram, or array declarator.
                                            The vname parameter cannot be the
                                            name of a subroutine or a main
                                            program.

 «attrs»                                    «A list of attributes separated
                                            by commas. The attrs describe
                                            vname. The following attributes
                                            can be used with vname: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR, PASCAL, REFERENCE,
                                            VALUE.»

 «length»                                   «Must be 8 or 16. Specifies
                                            length of associated vname, in
                                            bytes. If length is given, it
                                            overrides the length specified by
                                            bytes.»

 dim                                        A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant is written in
                                            the form n*constant, where n is a
                                            positive integer constant, and is
                                            equivalent to the constant
                                            constant repeated n times. The
                                            /values/ option, if specified,
                                            initializes vname. The following
                                            statement declares that vector is
                                            of type complex, and sets vector
                                            equal to (32.0,10.0):»

                                             «COMPLEX vector / (32.0,10.0) /»


 ■  Remarks

 A COMPLEX statement confirms or overrides the implicit type of vname. The
 name vname is defined for the entire program unit and cannot be defined by
 any other type statement in that program unit.

 COMPLEX statements must precede all executable statements.

 «DOUBLE COMPLEX and COMPLEX*16 represent the same data type.»


 ■  Examples

            COMPLEX  ch, zdif«*8», xdif«*16»

            COMPLEX«*8»  zz

            «COMPLEX*16  ax, by»

            COMPLEX  «x*16», y(10)«*8», «z*16(10)»


 CONTINUE


 ■  Action

 Does not have any effect


 ■  Syntax

      CONTINUE


 ■  Remarks

 The CONTINUE statement is a convenient place for a statement label,
 especially as the terminal statement in a DO «or DO WHILE» loop.


 ■  Example

            DIMENSION narray(10)
            DO 100, n = 1, 10
               narray(n) = 120
        100 CONTINUE


 CYCLE


 ■  «Action»

 «Within a loop, advances control to the terminating statement of a DO or DO
 WHILE loop»


 ■  «Syntax»

 «CYCLE»


 ■  «Remarks»

 «The CYCLE statement skips over the remaining part of a DO or DO WHILE loop.
 By combining a CYCLE statement with a logical IF statement, you can control
 whether the subsequent code executes.»


 ■  «Example»

 «Suppose you wanted to print a table of relativistic time-dilation factors
 for every velocity from 0 to the speed of light, in steps of 100 km/second.
 Perhaps you do not want to calculate these factors for speeds less than 10
 percent of the speed of light. The following example computes the
 time-dilation factors accordingly, putting them in the array timedilation.
 You can use the WRITE statement to print out the array.»

 «Time-dilation factor:  1 / √(1 - (v/c)^(2))»

           «INTEGER sub        ! subscript for timedilation array
            REAL timedilation(0:300)
            speedolight = 300000e3        ! 300000 km per second
            speedstep = 100e3             ! 100 km per second

               sub = speedolight / speedstep
            DO velocity = 1, speedolight, speedstep
               timedilation(sub) = 1.0
               IF (velocity .LT. (0.1 * speedolight))   CYCLE
               timedilation(sub) =
           +   1.0 / SQRT (1.0 - (velocity / speedolight)**2)
            END DO»


 DATA


 ■  Action

 Assigns initial values to variables


 ■  Syntax

      DATA nlist /clist/ ««,» nlist /clist/»...

 Parameter                                  Description

 nlist                                      A list of variables, array
                                            elements, array names, substring
                                            names, «structure elements,» and
                                            implied-DO lists, separated by
                                            commas. Implied-DO lists are
                                            discussed in the REMARKS section
                                            below.

                                            Each subscript in nlist must be
                                            an integer constant expression,
                                            except for implied-DO variables.
                                            Each substring specifier in nlist
                                            must be an integer constant
                                            expression.

 clist                                      A list of constants and/or
                                            repeated constants «and/or
                                            Hollerith constants», separated
                                            by commas. A repeated constant is
                                            written in the form n*c, where
                                            the repeat factor n is a positive
                                            integer constant, and c is the
                                            constant to be repeated. The
                                            repeated constant 3*10, for
                                            example, is equivalent to the
                                            clist 10,10,10.

                                            «A Hollerith constant is written
                                            in the form nHdata, where n is a
                                            positive integer constant, and
                                            data is a string of n
                                            characters.»

                                            There must be the same number of
                                            values in each clist as there are
                                            variables or array elements in
                                            the corresponding nlist. The
                                            appearance of an array in an
                                            nlist is equivalent to a list of
                                            all elements in that array in
                                            column-major order. Array
                                            elements can be indexed only by
                                            constant subscripts.


 ■  Remarks

 «Structure variables and variables explicitly declared as automatic may not
 appear in DATA statements. However, structure elements may appear in DATA
 statements.»

 Each noncharacter element in clist is converted to the data type of the
 corresponding element in nlist, if necessary.

 If a character element in clist is shorter than its corresponding variable
 or array element, it is extended to the length of the variable by adding
 blank characters to the right. If a character element is longer than its
 variable or array element, it is truncated. A single character constant
 defines a single variable or array element. A repeat count can be used.

 «If the $STRICT metacommand is not specified, a character element in clist
 can initialize a variable of any type.»

 Only local variables, arrays, and array elements can appear in a DATA
 statement. Formal arguments, variables in blank common blocks, and function
 names cannot be assigned initial values in a DATA statement. Variables in
 named common blocks may not appear in a DATA statement unless the DATA
 statement is in a block-data subprogram. «In Microsoft FORTRAN, elements in
 named common blocks can be assigned initial values using a DATA statement in
 a main program, function, or subroutine; the DATA statement does not have to
 be in a block-data subprogram.»

 The form of an implied-DO list is as follows:

      (dolist, dovar = start, stop «, inc»)

 Parameter                                  Description

 dolist                                     A list of array-element names and
                                            implied-DO lists.

 dovar                                      The name of an integer variable
                                            called the implied-DO variable.

 start, stop, and inc                       The integer-constant expressions.
                                            Each expression can contain
                                            implied-DO variables (dovar) of
                                            other implied-DO lists that have
                                            this implied-DO list within their
                                            ranges.

 For example, the following are implied-DO lists:

      (count(i), i = 5, 15, 2)
      ((array(sub,low), low = 1, 12), sub = 1, 2)
      ((result(first,second), first = 1, max), second = 1, upper)

 The number of iterations and the values of the implied-DO variable are
 established from start, stop, and inc exactly as for a DO loop except that
 the iteration count must be positive. See the DO entry in this section for
 more information. When an implied-DO list appears in a DATA statement, the
 list items in dolist are initialized once for each iteration of the
 implied-DO  list. The range of an implied-DO list is dolist. If the program
 contains another variable with the same name as dovar, that variable is not
 affected by the use of dovar in a DATA statement.


 ■  Examples

            INTEGER n, order, alpha, list(100)
            REAL  coef(4), eps(2), pi(5), x(5,5)
            CHARACTER*12  help

            DATA  n /0/, order /3/
            DATA  alpha /'A'/
            DATA  coef /1.0, 2*3.0, 1.0/, eps(1) /.00001/

      C     The following example initializes diagonal and below in
      C     a 5x5 matrix:
            DATA  ((x(j,i), i=1,j), j=1,5) / 15*1.0 /
            DATA  pi / 5*3.14159 /
            DATA  list / 100*0 /
            DATA  help(1:4), help(5:8), help(9:12) /3*'HELP'/


 DEALLOCATE


 ■  «Action»

 «Frees the array storage space previously reserved in an ALLOCATE statement»


 ■  «Syntax»

      «DEALLOCATE (arraylist «, STAT = ierr»)»

 «Parameter                                 Description»

 «arraylist»                                «A list of allocatable array
                                            names; if more than one exists,
                                            they must be separated by
                                            commas.»

 «ierr»                                     «The integer variable that
                                            returns status of deallocation
                                            attempt.»


 ■  «Remarks»

 «The STAT= parameter must appear last.»

 «Attempting to deallocate an array that was not allocated causes a run-time
 error.»

 «Any deallocation failure causes a run-time error, unless the STAT=
 parameter is present. The ierr variable returns a value of zero if the
 deallocation was successful, and the number of the run-time error if the
 deallocation failed.»

 «If an allocatable array is referenced when it is not currently allocated,
 the results are unpredictable.»


 ■  «Example»

           «INTEGER  dataset[ALLOCATABLE](:,:,:)
            INTEGER  reactor, level, points, error
            DATA  reactor, level, points  / 10, 50, 10 /

            ALLOCATE (dataset(1:reactor,1:level,1:points), STAT = error)

            DEALLOCATE (dataset, STAT = error)»


 DIMENSION


 ■  Action

 Declares that a variable is an array, and specifies the number of dimensions
 and their bounds


 ■  Syntax

      DIMENSION array ««[attrs]»» («{»«lower:»upper «| :}»
      «,«{»«lower:»upper «| :}»...»)

 Parameter                                  Description

 array                                      The name of an array. More than
                                            one array may be declared in a
                                            single DIMENSION statement.
                                            Multiple names are separated by
                                            commas.

 «attrs»                                    «A list of attributes separated
                                            by commas. The attrs describe
                                            array. The following attributes
                                            can be used with array: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR, PASCAL, REFERENCE,
                                            VALUE.»

 lower                                      The lower dimension bound, which
                                            can be positive, negative, or
                                            zero. The default for lower is
                                            one.

 upper                                      The upper dimension bound, which
                                            can be positive, negative, zero,
                                            or an asterisk. It must be
                                            greater than or equal to lower.


 ■  Remarks

 The specification array(«{»«lower:» upper «| :}») is called an "array
 declarator." The «lower»upper specifier «or the : specifier» is called a
 "dimension declarator." An array has as many dimensions as it has dimension
 declarators. You may specify no more than seven dimensions. «In Microsoft
 FORTRAN, the number of dimensions and their sizes are limited only by
 available memory.»

 «When an array is specified as allocatable, the dimension declarator
 consists only of a colon for each dimension. The single-colon specifier may
 be used only when an array is specified as allocatable.»

 You are free to specify both the upper and lower dimension bounds. If, for
 example, one array contains data from experiments numbered 28 through 112,
 you could dimension the array as follows:

            DIMENSION exprmt(28:112)

 Then, to refer to the data from experiment 72, you would reference
 exprmt(72).

 You can use any of the following as dimension bounds:

 Bound                                      Description

 An arithmetic constant                     If all array dimensions are
                                            specified by arithmetic
                                            constants, the array has a
                                            constant size. The arithmetic
                                            value is truncated to an integer.

 A nonarray-integer formal                  The dimension size is the initial
 argument or a nonarray-integer             value of the variable upon entry
 variable in a common block in the          to the subprogram at execution
 same program unit as the                   time. If a dimension bound of
 DIMENSION statement                        array is an integer formal
                                            argument or an integer variable
                                            in a common block, the array is
                                            an "adjustable-size array." The
                                            variable must be given a value
                                            before the subprogram containing
                                            the adjustable-size array is
                                            called.

 An arithmetic expression                   Expressions cannot contain
                                            references to functions or array
                                            elements. Expressions can contain
                                            variables only in adjustable-size
                                            arrays. The result of the
                                            expression is truncated to an
                                            integer.

 An asterisk (*)                            Only upper can be an asterisk,
                                            and an asterisk can only be used
                                            for upper in the last dimension
                                            of array. If upper is an
                                            asterisk, then array is an
                                            "assumed-size array." For an
                                            assumed-size array, the
                                            subprogram array is defined at
                                            execution time to be the same
                                            size as the array in the calling
                                            program. The following DIMENSION
                                            statement defines an assumed-size
                                            array in a subprogram:

                                              DIMENSION data (19,*)

                                            At execution time, the array data
                                            is given the size of the
                                            corresponding array in the
                                            calling program.

 Within noncharacter arrays, all elements begin on even-byte (word)
 addresses. Within character arrays «(and arrays of INTEGER*1 or LOGICAL*1
 variables)», elements always begin at the next available byte (odd or even).

 All adjustable- and assumed-size arrays, as well as the bounds for
 adjustable-size arrays, must be formal arguments to the program unit in
 which they appear. «Allocatable arrays must not be formal arguments.»

 Array elements are stored in column-major order: the leftmost subscript is
 incremented first when the array is mapped into contiguous memory addresses.
 For example, look at the following statements:

            INTEGER*2 a(2, 0:2)
            DATA a /1, 2, 3, 4, 5, 6/

 If a is placed at location 1000 in memory, the preceding DATA statement
 produces the following mapping:


 Array Element              Address                    Value

 a(1,0)                      1000                      1

 a(2,0)                      1002                      2

 a(1,1)

 a(2,1)                      1006                      4

 a(1,2)                      1008                      5

 a(2,2)                      100A                      6


 ■  Examples

 The following program dimensions two arrays:

            DIMENSION a(2,3), v(10)
            CALL Subr (a, 2, v)
            .
            .
            .
            SUBROUTINE Subr (matrix, rows, vector)
            REAL MATRIX, VECTOR
            INTEGER ROWS
            DIMENSION MATRIX (ROWS,*), VECTOR (10),
           +  LOCAL (2,4,8)
            MATRIX (1,1) = VECTOR (5)
            .
            .
            .
            END

 The following program uses assumed- and adjustable-size arrays:

            REAL     magnitude, minimum
            INTEGER  vecs, space, vec

      C     Array data values are assigned in column-major order
            DIMENSION  vecs(3, 4)
            DATA       vecs /1,1,1,2, 1,0,3,4, 7,-2,2,1 /

      C     Find minimum magnitude
            minimum = 1E10
            DO 100 vec = 1, 4

      C     Call the function magnitude to calculate the magnitude of
      C     vector vec.
               minimum = AMIN1(minimum, magnitude(vecs, 3, vec))
       100  CONTINUE

            WRITE (*, 110) minimum
       110  FORMAT (' Vector closest to origin has a magnitude of',
           +          F12.6)
            END


      C     Function returns the magnitude of the j-th column vec in a
      C     matrix. Note that, because of the assumed-size array, the
      C     subroutine does not need to know the number of columns in
      C     the matrix. It only requires that the specified column
      C     vector be a valid column in the matrix. The number of rows
      C     must be passed so the function can do the sum.

            REAL FUNCTION Magnitude (matrix, rows, j)

            REAL       sum
            INTEGER    matrix, rows, i, j
            DIMENSION  matrix (rows,*)

            sum = 0.0
            DO 100 i = 1, rows
            sum = sum + matrix(i,j)**2
       100  CONTINUE
            magnitude = SQRT (sum)
            END


 DO


 ■  Action

 Repeatedly executes the statements following the DO statement through the
 statement which marks the end of the loop


 ■  Syntax

      DO «label «,» » dovar = start, stop «, inc»

 Parameter                                  Description

 label                                      The statement label of an
                                            executable statement.

 dovar                                      An integer, real, or
                                            double-precision variable, called
                                            the DO variable.

 start, stop                                The integer, real, or
                                            double-precision expressions.

 inc                                        An integer, real, or
                                            double-precision expression. The
                                            parameter inc cannot equal zero;
                                            inc defaults to one.


 ■  Remarks

 «The label is optional.» If label is used, the loop terminates with the
 labeled statement. «If label is not used, the loop is terminated with an END
 DO statement.»

 If you use a labeled terminating statement, it must follow the DO statement
 and be in the same program unit. This statement must not be an unconditional
 or assigned GOTO, a block or arithmetic IF, «CASE, CYCLE,» DO, ELSE, ELSE
 IF, END, END IF, «END SELECT CASE,» «EXIT,» RETURN, «SELECT CASE,» or STOP
 statement.

 The terminal statement may be a logical IF statement.

 The range of a DO loop begins with the statement immediately following the
 DO statement and includes the terminal statement of the DO loop.

 Execution of a CALL statement that is in the range of a DO loop does not
 terminate the DO loop unless an alternate-return specifier in a CALL
 statement returns control to a statement outside the DO loop.

 The following restrictions apply to DO loops:

   ■  If a DO loop appears within another DO «or DO WHILE» loop, its range
      must be entirely within the range of the enclosing loop.

   ■  If a DO statement appears within an IF, ELSE IF, or ELSE block, the DO
      loop must be contained within the block.

   ■  If a block IF statement appears within a DO loop, its associated END IF
      statement must also be within that DO loop.

   ■  The loop variable dovar may not be modified by statements within the
      loop.

   ■  Jumping into a DO loop from outside its range is not permitted.
      «However, a special feature added for compatibility with earlier
      versions of FORTRAN permits extended-range DO loops. See the $DO66
      entry in Section 6.2, "Metacommand Directory," for more information.»

   ■  «Two or more DO or DO WHILE loops may share one labeled terminal
      statement. An END DO statement may terminate only one DO or DO WHILE
      loop.»

   ■  «If a SELECT CASE statement appears within a DO loop, its associated
      END SELECT CASE statement must also appear within that DO loop.»

 Note that the number of iterations in a DO loop is limited to the maximum
 possible value of the loop variable. «For example, DO loops that use
 INTEGER*2 variables as the DO variable and bounds cannot be evaluated more
 than 32,767 times. When the $DEBUG metacommand is used, a run-time error
 occurs if a DO variable overflows. When $DEBUG is not used, the results of
 an overflow are unpredictable. In the following fragment the number of
 iterations is 64,001, and the DO variable i can only be evaluated 32,767
 times:»

      «$DEBUG
      C     This causes an error:
            INTEGER*2 i
            DO 10 i = -32000, 32000, 1
         10 CONTINUE»

 A DO statement is executed in the following sequence:

   1. The expressions start, stop, and inc are evaluated. If necessary, type
      conversion is performed. The DO variable dovar is set to the value of
      start.

   2. The iteration count for the loop is tested.

      The iteration count for the loop is calculated using the following
      formula:

      MAX(INT((stop - start + inc) / inc), 0)

      The iteration count is zero if either of the following is true:

        ■  start > stop and inc > 0

        ■  start < stop and inc < 0

   3. If the iteration count is greater than zero, the statements in the
      range of the DO loop are executed; if not, execution continues with the
      first statement following the DO loop.

      «If the $DO66 metacommand is in effect, the statements in the range of
      the DO loop are executed at least once, regardless of the value of the
      iteration count.» «

   4. After the execution of the terminal statement of the DO loop, the value
      of the DO variable dovar is increased by the value of inc that was
      computed when the DO statement was executed.

   5. The iteration count is decreased by one.

   6. The iteration count is tested. If the iteration count is greater than
      zero, the statements in the range of the DO loop are executed again.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    The iteration count is computed using two-byte precision (the default).
    If the iteration count overflows this precision, the results are
    unpredictable. The following example causes an iteration overflow:

      IMPLICIT INTEGER*2 (A-Z)
            stop = 32000
            step = 12000
            DO 100  n = 0, stop, step
            WRITE (*, *) 'N = ', n
        100 CONTINUE
 ───────────────────────────────────────────────────────────────────────────

 ■  Examples

 The following two program fragments are examples of DO statements:

      C     Initialize the even elements of a 20-element real array
            DIMENSION array(20)
            DO 100 j = 2, 20, 2
        100 array(j) = 12.0

      C     Perform a function 11 times
            DO 200,  k = -30, -60, -3
               int =  j / 3
               isb = -9 - k
               array(isb) = MyFunc (int)
        200 CONTINUE

 The following shows the final value of a DO variable (in this case 11):

            DO 200 j = 1, 10
        200 WRITE (*, '(I5)') j
            WRITE (*, '(I5)') j



 DO WHILE


 ■  «Action»

 «Executes a block of statements repeatedly while a logical condition remains
 true»


 ■  «Syntax»

 «DO «label «, » » WHILE (logicalexpr)»

 «Parameter                                 Description»

 «label»                                    «A label of executable statement
                                            or CONTINUE statement»

 «logicalexpr»                              «A test expression which
                                            evaluates to true or false»


 ■  «Remarks»

 «The label is optional. If label is used, the loop terminates with the
 labeled statement. If label is not used, the loop is terminated with an END
 DO statement.»

 «A terminating statement must follow the DO WHILE statement and be in the
 same program unit. This statement must not be an unconditional or assigned
 GOTO, a block or arithmetic IF, CASE, CYCLE, DO, ELSE, ELSE IF, END, END IF,
 END SELECT CASE, EXIT, RETURN, SELECT CASE, or STOP statement.»

 «The terminal statement may be a logical IF statement.»

 «The range of a DO WHILE loop begins with the statement immediately
 following the DO WHILE statement and includes the terminal statement of the
 DO WHILE loop.»

 «Execution of a CALL statement that is in the range of a DO WHILE loop does
 not terminate the DO WHILE loop unless an alternate-return specifier in a
 CALL statement returns control to a statement outside the DO WHILE loop.»

 «The following steps occur when a DO WHILE statement is executed:»

   1. «The logical expression is evaluated.»

   2. «If the expression is false, none of the statements within the range of
      the loop are executed. Execution jumps to the statement following the
      terminating statement.»

   3. «If the expression is true, the statements within the loop are
      executed, starting with the first statement following the DO WHILE
      statement.»

   4. «When the terminating statement is reached, execution returns to the DO
      WHILE statement. The logical expression is evaluated, and the cycle
      repeats.»

 «The following restrictions apply to DO WHILE loops:»

   ■  «If a DO WHILE loop appears within another DO or DO WHILE loop, its
      range must be entirely within the range of the enclosing loop.»

   ■  «Two or more DO or DO WHILE loops may share one labeled terminal
      statement. An END DO statement may terminate only one DO or DO WHILE
      loop.»

   ■  «If a DO WHILE statement appears within an IF, ELSE IF, or ELSE block,
      the range of the DO WHILE loop must be entirely within the block.»

   ■  «If a block IF statement appears within a DO WHILE loop, its associated
      END IF statement must also be within that DO WHILE loop.»

   ■  «If a SELECT CASE statement appears within the range of a DO WHILE
      loop, its associated END SELECT CASE statement must also be within that
      DO WHILE loop.»

   ■  «Jumping into the range of a DO WHILE loop is not permitted. However, a
      special feature added for compatibility with earlier versions of
      FORTRAN permits extended-range DO WHILE loops. See the $DO66 entry in
      Section 6.2, "Metacommand Directory," for more information.»


 ■  «Example»

            «CHARACTER*1 input

            input = ' '

            DO WHILE ((input .NE. 'n') .AND. (input .NE. 'y'))
            WRITE (*, '(A)') 'Enter y or n: '
            READ  (*, '(A)') input
            END DO»


 DOUBLE COMPLEX


 ■  «Action»

 «Specifies the DOUBLE COMPLEX type for user-defined names»


 ■  «Syntax»

      «DOUBLE COMPLEX vname « [attrs] » «(dim)» «/values/»
      «, vname« [attrs] » «(dim)» «/values/»»...»

 «Parameter                                 Description»

 «vname»                                    «The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function,
                                            intrinsic function, FUNCTION
                                            subprogram, or array declarator.
                                            The vname parameter cannot be the
                                            name of a subroutine or a main
                                            program.»

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            vname. The following attributes
                                            can be used with vname: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR, PASCAL, REFERENCE,
                                            VALUE.»

 «dim»                                      «A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.»

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant is written in
                                            the form n*constant, where n is a
                                            positive integer constant, and is
                                            equivalent to the constant
                                            constant repeated n times. The
                                            /values/ option, if specified,
                                            initializes vname. The following
                                            statement declares that vector is
                                            of type DOUBLE COMPLEX, and sets
                                            vector equal to (32.0,10.0):»

                                           «DOUBLE COMPLEX vector/(32.798d2,
                                            +10.985d3) /»


 ■  «Remarks»

 «A DOUBLE COMPLEX statement confirms or overrides the implicit type of
 vname. The name vname is defined for the entire program unit and cannot be
 defined by any other type of statement in that program unit.»

 «DOUBLE COMPLEX statements must precede all executable statements.»

 «DOUBLE COMPLEX and COMPLEX*16 are the same data type.»


 ■  «Examples

            «DOUBLE COMPLEX  vector, arrays(7,29)»

            «DOUBLE COMPLEX  pi, 2pi /3.141592654,6.283185308/»


 DOUBLE PRECISION


 ■  Action

 Specifies the DOUBLE PRECISION type for user-defined names


 ■  Syntax

      DOUBLE PRECISION vname «« [attrs] »» «(dim)» ««/values/»»
      «, vname«« [attrs] »» «(dim)» ««/values/»»»...

 Parameter                                  Description

 vname                                      The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function,
                                            intrinsic function, FUNCTION
                                            subprogram, or array declarator.
                                            The vname parameter cannot be the
                                            name of a subroutine or main
                                            program.

 «attrs»                                    «An optional list of attributes,
                                            separated by commas. The attrs
                                            describe vname. The following
                                            attributes can be used with
                                            vname: ALIAS, ALLOCATABLE, C,
                                            EXTERN, FAR, HUGE, NEAR, PASCAL,
                                            REFERENCE, VALUE.»

 dim                                        A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant is written in
                                            the form n*constant, where n is a
                                            positive integer constant, and is
                                            equivalent to constant repeated n
                                            times. The /values/ option, if
                                            specified, initializes vname. The
                                            following statement declares that
                                            pi  is of type DOUBLE PRECISION,
                                            and sets num equal to
                                            3.141592654:»

                                           «DOUBLE PRECISION pi/3.141592654/»


 ■  Remarks

 A DOUBLE PRECISION statement confirms or overrides the implicit type of
 vname. The name vname is defined for the entire program unit, and cannot be
 defined by any other type statement in that program unit.

 DOUBLE PRECISION statements must precede all executable statements.

 A DOUBLE PRESICION variable is accurate to 14 or 15 decimal digits.

 DOUBLE PRECISION and «REAL*8» are the same data type.


 ■  Example

            DOUBLE PRECISION  varnam


 ELSE


 ■  Action

 Marks the beginning of an ELSE block


 ■  Syntax

      ELSE


 ■  Remarks

 An ELSE block consists of any executable statements between the ELSE
 statement and the next END IF statement at the same IF level. The matching
 END IF statement must appear before any ELSE or ELSE IF statements of the
 same IF level.

 Control may not be transferred into an ELSE block from outside that block.


 ■  Example

            CHARACTER c
            .
            .
            .
            READ (*, '(A)') C
            IF (c .EQ. 'A') THEN
              CALL Asub
            ELSE
              CALL Other
            END IF
            .
            .
            .


 ■  See also

 ELSE IF
 END IF
 IF THEN ELSE (Block IF)


 ELSE IF


 ■  Action

 Causes execution of a block of statements if expression is true


 ■  Syntax

      ELSE IF (expression) THEN

 Parameter                                  Description

 expression                                 A logical expression


 ■  Remarks

 The associated ELSE IF block consists of any executable statements between
 the ELSE IF statement and the next ELSE IF, ELSE, or END IF statement at the
 same IF level.

 When the ELSE IF statement is executed, expression is evaluated, and the
 following steps are performed:

 If expression is:                          Then the next statement executed i

 True, and there is at least one            The first statement of the ELSE
 executable statement in the ELSE           IF block
 IF block

 True, and there are no executable          The next END IF statement at the
 statements in the ELSE IF block            same IF level as the ELSE IF
                                            statement

 False                                      The next ELSE IF, ELSE, or END IF
                                            statement that has the same IF
                                            level as the ELSE IF statement

 After the last statement in the ELSE IF block has been executed, the next
 statement executed is the next END IF statement at the same IF level as that
 ELSE IF statement.

 Control may not be transferred into an ELSE IF block from outside that
 block.


 ■  Example

            CHARACTER char
            READ (*, '(A)') char
            IF (char .EQ. 'L') THEN
              CALL Lsub
            ELSE IF (char .EQ. 'X') THEN
              CALL Xsub
            ELSE
              CALL Other
            END IF


 ■  See also

 ELSE
 END IF
 IF THEN ELSE (Block IF)


 END


 ■  Action

 Terminates execution of the main program, or returns control from a
 subprogram


 ■  Syntax

      END


 ■  Remarks

 The END statement marks the end of the program unit in which it appears, and
 it must be the last statement in every program unit. Comment lines may
 follow an END statement.

 An END statement must appear by itself on an unlabeled initial line (not a
 continuation line). No continuation lines may follow an END statement. No
 FORTRAN statement may have an initial line that appears to be an END
 statement (such as splitting an END IF statement over two lines).

 In a subprogram, the END statement has the same effect as a RETURN
 statement.


 ■  Example

      C     An END statement must be the last statement in a program
      C     unit:
            PROGRAM MyProg
            WRITE (*, '("Hello, world!")')
            END


 END DO


 ■  «Action»

 «Terminates a DO or DO WHILE loop»


 ■  «Syntax»

      «END DO»


 ■  «Remarks»

 «There must be a matching END DO statement for every DO or DO WHILE
 statement that does not contain a label reference.»

 «An END DO statement may terminate only one DO or DO WHILE statement.»


 ■  «Examples»

 «The following examples produce the same output.»

           «DO ivar = 1, 10»
              «PRINT ivar»
           «END DO»

           «ivar = 0»
           «DO WHILE (ivar .LT. 10)»
              «ivar = ivar + 1»
              «PRINT *«, ivar»
           «END DO»


 END IF


 ■  Action

 Terminates a block IF statement


 ■  Syntax

      END IF


 ■  Remarks

 There must be a matching END IF statement for every block IF statement in a
 program unit. Execution of an END IF statement has no effect on the program.


 ■  Example

            IF (n .LT. 0) THEN
              x = -n
              y = -n
            END IF


 ENDFILE


 ■  Action

 Writes an end-of-file record to the specified unit


 ■  Syntax

      ENDFILE {unitspec |
      («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»)}

 If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   An integer expression that
                                            specifies an external unit. If
                                            unitspec has not been opened, a
                                            run-time error occurs.

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit. If errlabel is specified,
                                            an I/O error causes transfer of
                                            control to the statement at
                                            errlabel. If errlabel is omitted,
                                            the effect of an I/O error is
                                            determined by the presence or
                                            absence of iocheck.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns the value zero if
                                            there is no error, or the error
                                            number if an error occurs. For
                                            more information on error
                                            handling, see Section 3.2.6.


 ■  Remarks

 After writing the end-of-file record, the ENDFILE statement positions the
 file after the end-of-file record. Further sequential data transfer is
 prohibited unless you execute either a BACKSPACE or REWIND statement.

 When ENDFILE operates on a direct-access file, all records beyond the new
 end-of-file record are erased.

 If a parameter of the ENDFILE statement is an expression that calls a
 function, that function must not cause an I/O statement «or the EOF
 intrinsic function» to be executed, because unpredictable results can occur.


 ■  Example

            WRITE (6, *) x
            ENDFILE 6
            REWIND 6
            READ (6, *) y


 ENTRY


 ■  Action

 Specifies an entry point to a subroutine or external function


 ■  Syntax

     ENTRY ename «« [eattrs] »»  «(«formal«« [attrs]»»
     «,formal« [attrs] »»...»)»

 Parameter                                  Description

 ename                                      The name of the entry point. If
                                            ename is an entry point for a
                                            user-defined function, ename must
                                            be given a data type by:

                                            ■  The default rules used in
                                               determining type; or

                                            ■  The type specified in an
                                               IMPLICIT
                                               statement; or

                                            ■  A declaration in the user-
                                               defined function's type-
                                               declaration section.

 «eattrs»                                   «A list of attributes, separated
                                            by commas. The eattrs describe
                                            ename. The following attributes
                                            can be used with ename: ALIAS, C,
                                            LOADDS, PASCAL, VARYING.»

 formal                                     A variable name, array name,
                                            «structure variable name,» or
                                            formal procedure name. If the
                                            ENTRY statement is in a
                                            subroutine, formal can be an
                                            asterisk.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            formal. The following attributes
                                            can be used with formal: FAR,
                                            HUGE, NEAR, REFERENCE, VALUE.»


 ■  Remarks

 To begin executing a subroutine or function at the first executable
 statement after an ENTRY statement, replace the name of the subprogram with
 the name of the entry point:

 Type of Call                               Form of Call

 Subroutine                                 CALL ename « («actual1«,actual2»..

 Function                                   ename («actual1«,actual2»...»

 Parentheses are required when calling a function entry point, even if the
 function has no formal arguments.

 Entry points cannot be called recursively. That is, a subprogram may not
 directly or indirectly call an entry point within that subprogram.

 There is no defined limit on the number of ENTRY statements you can use in a
 subprogram. The following restrictions apply to use of the ENTRY statement:

   ■  Within a subprogram, ename cannot have the same name as a formal
      argument in a FUNCTION, SUBROUTINE, ENTRY, or EXTERNAL statement.

   ■  Within a function, ename cannot appear in any statement other than a
      type statement until after ename has been defined in an ENTRY
      statement.

   ■  If one ename in a function is of character type, all the enames in that
      function that must be of character type, and all the enames must be the
      same length.

   ■  A formal argument cannot appear in an executable statement that occurs
      before the ENTRY statement containing the formal argument unless the
      formal argument also appears in a FUNCTION, SUBROUTINE, or ENTRY
      statement that precedes the executable statement.

   ■  An ENTRY statement cannot appear between a block IF statement and the
      corresponding END IF statement, or between a DO statement and the
      terminal statement of its  DO loop.


 ■  Example

      C     This fragment writes a message indicating
      C     whether num is positive or negative
            IF (num .GE. 0) THEN
                CALL Positive
            ELSE
                CALL Negative
            END IF
            .
            .
            .
            END

            SUBROUTINE Sign
            ENTRY Positive
            WRITE (*, *) 'It''s positive.'
            RETURN
            ENTRY Negative
            WRITE (*, *) 'It''s negative.'
            RETURN
            END


 EQUIVALENCE


 ■  Action

 Causes two or more variables or arrays to occupy the same memory location


 ■  Syntax

      EQUIVALENCE (nlist) «, (nlist)»...

 Parameter                                  Description

 nlist                                      A list of at least two variables,
                                            arrays, or array elements,
                                            separated by commas.

                                            The list may not include formal
                                            arguments «or allocatable
                                            arrays.» Subscripts must be
                                            integer constants and must be
                                            within the bounds of the array
                                            they index. An unsubscripted
                                            array name refers to the first
                                            element of the array.


 ■  Remarks

 An EQUIVALENCE statement causes all elements in nlist to have the same first
 memory location. Variable names are said to be associated if they refer to
 the same memory location.

 There is no automatic type conversion among the elements in nlist; they
 simply occupy the same memory space.

 Associated character entities may overlap, as in the following example:

            CHARACTER  a*4, b*4, c(2)*3
            EQUIVALENCE (a, c(1)), (b, c(2))

 The preceding example can be graphically illustrated as follows:

      |01|02|03|04|05|06|07|

      |-----a-----|
               |-----b-----|
      |--c(1)--|--c(2)--|

 The following rules restrict how you may associate elements:

   ■  A variable cannot be forced to occupy more than one memory location,
      nor can you force two or more elements of the same array to occupy the
      same memory location. The following statement would force r to occupy
      two different memory locations or s(1)  and s(2) to occupy the same
      memory location:

     C     this causes an error:
           REAL r, s(10)
           EQUIVALENCE (r, s(1))
           EQUIVALENCE (r, s(2))

   ■  Consecutive array elements must be stored in sequential order. The
      following is not permitted:

      C     this causes another error:
            REAL r(10), s(10)
            EQUIVALENCE (r(1), s(1))
            EQUIVALENCE (r(5), s(7))

      The compiler always aligns noncharacter entities on even-byte (word)
      boundaries. The following example causes a compile-time error, since
      variables a and b cannot be word-aligned simultaneously:

      CHARACTER*1 c1(10)
      REAL a, b
      EQUIVALENCE (a, c1(1))
      EQUIVALENCE (b, c1(2))

   ■  Character and noncharacter entities cannot be associated. «Microsoft
      FORTRAN permits character and noncharacter entities to be associated,
      but not in such a way that noncharacter entities start on an odd-byte
      boundary. If necessary, the compiler will adjust the storage location
      of the character entity so the noncharacter entity begins on an even
      byte. The following example causes a compile-time error because it is
      not possible to reposition the character array such that both
      noncharacter entities start on an even byte address:»

            «CHARACTER*«1  char1(10)»
            «REAL  reala, realb»
            «EQUIVALENCE  (reala, char1(1))»
            «EQUIVALENCE  (realb, char1(2))»

   ■  «An item that appears in nlist cannot be initialized in a type
      statement. The following example causes an error:»

      «INTEGER i /1/»
      «EQUIVALENCE (i, j)»

   ■  An EQUIVALENCE statement cannot share memory between two different
      common blocks.

   ■  An EQUIVALENCE statement can extend a common block by adding memory
      elements following the common block, as long as the EQUIVALENCE
      statement does not make a named common block's length different from
      the length of the same named common block in other program units.

   ■  An EQUIVALENCE statement cannot extend a common block by adding memory
      elements preceding the common block, as in the following example:

      C     This causes an error:
            COMMON /abcde/ r(10)
            REAL s(10)
            EQUIVALENCE (r(1), s(7))

 «Unless the $STRICT metacommand is in effect, only the first subscript of a
 multi-dimensional array is required in EQUIVALENCE statements.  This makes
 it easier to port FORTRAN 66 programs.»

 «For example, the array declaration var(3,3), var(4) could appear in an
 EQUIVALENCE statement. The reference is to the fourth element of the array
 (var(1,2)), not to the beginning of the fourth row or column.»


 ■  Example

            CHARACTER  name, first, middle, last
            DIMENSION  name(60), first(20), middle(20), last(20)
            EQUIVALENCE (name(1), first(1)), (name(21), middle(1))
            EQUIVALENCE (name(41), last(1))


 EXIT


 ■  «Action»

 «Transfers control from within a DO or DO WHILE loop to the first executable
 statement following the end of the loop»


 ■  «Syntax»

      «EXIT»


 ■  «Remarks»

 «Normally, a DO loop executes a fixed number of times. The EXIT statement
 lets you terminate the loop early if some specified condition warrants it.»

 «For example, you could set a loop to collect a maximum of 1,000 data
 points, while still being able to terminate the loop if there were fewer
 than 1,000 points.»


 ■  «Example»

      «C     Loop terminates early if one of the data points is zero:

            INTEGER numpoints, point
            REAL datarray(1000), sum

            sum = 0.0
            DO point = 1, 1000
               numpoints = point
               sum = sum + datarray(point)
               IF (datarray(point+1) .EQ. 0.0)  EXIT
            END DO»


 EXTERNAL


 ■  Action

 Identifies a user-defined name as an external subroutine or function


 ■  Syntax

      EXTERNAL name «« [attrs]»» «,name«« [attrs]»» »...

 Parameter                                  Description

 name                                       The name of an external
                                            subroutine or function.
                                            Statement-function (single-line
                                            function) names are not allowed.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            name. The following attributes
                                            can be used with name: ALIAS, C,
                                            FAR, LOADDS, NEAR, PASCAL,
                                            VARYING.»


 ■  Remarks

 The EXTERNAL statement is primarily used to specify that a particular
 user-defined name is a subroutine or function to be used as a procedural
 parameter. The EXTERNAL statement can also replace an intrinsic function
 with a user-defined function of the same name.

 If an intrinsic-function name appears in an EXTERNAL statement, that name
 becomes the name of an external procedure, and the corresponding intrinsic
 function can no longer be called from that program unit. A user name can
 only appear once in an EXTERNAL statement in any given program unit.

 FORTRAN assumes that the name of any subroutine or function referred to in a
 compilation unit (but not defined there) is defined externally.


 ■  Examples

            EXTERNAL MyFunc, MySub
      C     MyFunc and MySub are arguments to Calc
            CALL Calc (MyFunc, MySub)

      C     Example of a user-defined function replacing an
            intrinsic EXTERNAL SIN
            x = SIN (a, 4.2, 37)


 FORMAT


 ■  Action

 Sets the format in which data is written to or read from a file


 ■  Syntax

      FORMAT («editlist»)

 Parameter                                  Description

 editlist                                   A list of editing descriptions


 ■  Remarks

 FORMAT statements must be labeled.

 Invalid editlist strings generate error messages; some during compilation,
 others at run-time.

 Table 3.6 summarizes the nonrepeatable edit descriptors. Table 4.5
 summarizes the repeatable edit descriptors. See Section 3.7, "Formatted
 I/O," for further information on edit descriptors and formatted I/O.

 Table 4.5  Repeatable Edit Descriptors

 Descriptor       Use

 Iw«.m»            Integer values
 «Zw                 Hexadecimal values»
 Fw.d                Real values
 Ew.d«Ee»          Real values with exponents
 Gw.d«Ee»          Real values, extended range
 Dw.d                Double-precision real values
 Lw                  Logical values
 A«w»              Character values


 FUNCTION (External)


 ■  Action

 Identifies a program unit as a function, and supplies its name, data type,
 and optional formal parameters


 ■  Syntax

      «type»FUNCTION func ««[fattrs]»» («formal ««[attrs]»»»
      «,formal««[attrs]»»»...)

 Parameter                                  Description

 type                                       Declares the data type of the
                                            value returned by the function.
                                            The type parameter must be one of
                                            the following:

                                            CHARACTER
                                            CHARACTER*n
                                            COMPLEX
                                            «COMPLEX*8»
                                            «COMPLEX*16»
                                            «DOUBLE COMPLEX»
                                            DOUBLE PRECISION
                                            INTEGER
                                            «INTEGER*1»
                                            «INTEGER*2»
                                            «INTEGER*4»
                                            «INTEGER[C]»
                                            LOGICAL
                                            «LOGICAL*1»
                                            «LOGICAL*2»
                                            «LOGICAL*4»
                                            REAL
                                            «REAL*4»
                                            «REAL*8»

                                            If type is omitted, the
                                            function's type is determined by
                                            the applicable IMPLICIT
                                            statements. If there are none,
                                            the function's type is
                                            established by FORTRAN's default
                                            typing. «If IMPLICIT NONE or the
                                            $DECLARE metacommand is in
                                            effect, the function's type must
                                            be given in the function
                                            declaration or in a type
                                            declaration.»

                                            If type is specified, the
                                            function name cannot appear in a
                                            type statement.

 func                                       The name of the function. The
                                            name func cannot appear in
                                            «AUTOMATIC,» COMMON, DATA,
                                            EQUIVALENCE,  INTRINSIC,
                                            «NAMELIST,» or SAVE statements.

 «fattrs»                                   «A list of attributes, separated
                                            by commas. The fattrs describe
                                            func. The following attributes
                                            can be used with the func name:
                                            ALIAS, C, FAR, NEAR, PASCAL,
                                            VARYING.»

 formal                                     A formal-argument name. If more
                                            than one is specified, they must
                                            be separated by commas.
                                            Alternate-return specifiers are
                                            not allowed in FUNCTION
                                            statements.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            formal. The following attributes
                                            can be used with formal: FAR,
                                            HUGE, NEAR, REFERENCE, VALUE.»


 ■  Remarks

 A function begins with a FUNCTION statement and ends with the next END
 statement. A function can contain any statement except a BLOCK DATA,
 FUNCTION, «INTERFACE TO,» PROGRAM, or SUBROUTINE statement.

 The length specifier for the function type may follow the function name. For
 example, the following two declarations are equivalent:

             INTEGER*4  FUNCTION  FuncX   (var)
             INTEGER    FUNCTION  FuncX*4 (var)

 Within the calling program, func is global, and may not be used for any
 other variable or subprogram.

 The formal-argument list sets the number of arguments for that function. The
 argument types are set by any IMPLICIT, EXTERNAL, DIMENSION, or type
 statements within the function itself (or they default to implicit FORTRAN
 types). The argument types are not set by the calling program, even if the
 function appears in the calling program's source file. Formal-argument names
 cannot appear in «AUTOMATIC,» COMMON, DATA,  EQUIVALENCE, INTRINSIC, or SAVE
 statements.

 When a function is referenced, the actual arguments that are passed must
 agree with corresponding formal arguments in the FUNCTION statement in
 order, in number «(except when the C and VARYING attributes are specified)»,
 and in type or kind.

 The compiler will check for correspondence if the formal arguments are
 known. See the CALL entry in this section for more information on checking
 arguments for correspondence.

 The function name func acts as if it were a variable within the function
 definition. At some point in the execution of a function, a value must be
 assigned to the function name. It is this value that is returned to the
 calling program through the function's name. This value is usually assigned
 at the end of the function (but any number of assignments may be made), with
 RETURN statements terminating the function at the appropriate points.

 In addition to returning the value of the function, an external function can
 return values by assignment to any formal argument whose corresponding
 actual argument was passed by reference.

 A "recursive" function is one that calls itself, or calls another
 subprogram, which in turn calls the first function before the first function
 has completed execution. FORTRAN does not support recursive function calls.
 A function may otherwise be called from any unit.


 ■  Example

    C     GetNo is a function that
    reads a number from unit i
          i = 2
          10 IF (GetNo(i) .EQ. 0.0)
    GOTO 10
          END
    C
          FUNCTION GetNo (nounit)
          READ (nounit, '(F10.5)') r
          GetNo = r
          END


 GOTO (Assigned GOTO)


 ■  Action

 Transfers execution to the statement label assigned to variable


 ■  Syntax

      GOTO variable ««,» (labels)»

 Parameter                                  Description

 variable                                   An integer variable. The variable
                                            must have been assigned the label
                                            of an executable statement in the
                                            same program unit.

 labels                                     A list of one or more statement
                                            labels of executable statements
                                            in the same program unit. If more
                                            than one label is specified, the
                                            labels are separated by commas.

                                            A label may appear more than once
                                            in a list.


 ■  Remarks

 «If you specify the $DEBUG metacommand, a run-time error occurs if the label
 assigned to name is not one of the labels specified in the label list.»

 Jumping into a DO, IF, ELSE IF, or ELSE block from outside the block is not
 normally permitted. «A special feature, extended-range DO loops, does permit
 entry into a DO block. See the $DO66 entry in Section 6.2, "Metacommand
 Directory," for more information.»


 ■  Examples

            ASSIGN 10 TO n
            GOTO n
         10 CONTINUE

 The following example uses an assigned GOTO statement to check the value of
 clearance:

      «$DEBUG»

            INTEGER  view, clearance
      C     Assign an appropriate label to view:
            IF (clearance .EQ. 1) THEN
               ASSIGN 200 TO view
            ELSE IF (clearance .EQ. 2) THEN
               ASSIGN 400 TO view
            ELSE
               ASSIGN 100 TO view
            END IF

      C     Show user appropriate view of data depending on
      C     security clearance.
            GOTO view (100, 200, 400)

      «C     If view had not been assigned one of the valid labels»
      «C     (100, 200, or 400), a run-time error would have occurred.»

       100  CONTINUE
            .
            .
            .
       200  CONTINUE
            .
            .
            .
       400  CONTINUE
            .
            .
            .
            END


 GOTO (Computed GOTO)


 ■  Action

 Transfers control to the statement at the nth label in the list


 ■  Syntax

      GOTO (labels) «,» n

 Parameter                                  Description

 labels                                     One or more statement labels of
                                            executable statements in the same
                                            program unit. If there is more
                                            than one label, the labels are
                                            separated by commas.

                                            The same statement label may
                                            appear more than once in the
                                            list.

 n                                          An integer expression. Control is
                                            transferred to the nth label in
                                            the list.


 ■  Remarks

 If there are j labels in the list and n is out of range (that is, n > j or n
 < 1), the computed GOTO statement acts like a CONTINUE statement.

 Jumping into a DO, IF, ELSE IF, or ELSE block from outside the block is not
 normally permitted. «A special feature, extended-range DO loops, does permit
 entry into a DO block. See the $DO66 entry in Section 6.2,"Metacommand
 Directory," for more information.


 ■  Example

            next = 1
      C     The following statement transfers control to statement 10:
            GOTO (10, 20) next
            .
            .
            .
         10 CONTINUE
            .
            .
            .
         20 CONTINUE


 GOTO (Unconditional GOTO)


 ■  Action

 Transfers control to a labeled statement


 ■  Syntax

      GOTO label

 Parameter                                  Description

 label                                      The statement label of an
                                            executable statement in the same
                                            program unit


 ■  Remarks

 If the program unit does not contain a statement with the label specified in
 the GOTO statement, a compile-time error occurs.

 Jumping into a DO, IF, ELSE IF, or ELSE block from outside the block is not
 normally permitted. «A special feature, extended-range DO loops, does permit
 entry into a DO block. See the $DO66 entry in Section 6.2, "Metacommand
 Directory," for more information.»


 ■  Example

            GOTO 4077
            .
            .
            .
       4077 CONTINUE


 IF (Arithmetic IF)


 ■  Action

 Transfers control to one of three statement labels, depending on the value
 of expression


 ■  Syntax

      IF (expression) label1, label2, label3

 Parameter                                  Description

 expression                                 An integer expression or a
                                            single- or double-precision real
                                            expression.

 label1, label2, label3                     The statement labels of
                                            executable statements in the same
                                            program unit. The same statement
                                            label may appear more than once.


 ■  Remarks

 The arithmetic IF statement transfers control as follows:

 If expression is:                          The next statement executed is:

 < 0                                        The statement at label1.

 = 0                                        The statement at label2.

 > 0                                        The statement at label3.

 Jumping into a DO, IF, ELSE IF, or ELSE block from outside the block is not
 normally permitted. «A special feature, extended-range DO loops, does permit
 entry into a DO block. See the $DO66 entry in Section 6.2,"Metacommand
 Directory," for more information.»


 ■  Example

            DO 40 j = -1, 1
            n = j
      C     The following statement transfers control to statement 10
      C     for j = -1, to statement 20 for j = 0, and to statement 30
      C     for j = +1.
            IF (n) 10, 20, 30
         10 CONTINUE
            .
            .
            .
         20 CONTINUE
            .
            .
            .
         30 CONTINUE
            .
            .
            .
         40 CONTINUE


 IF (Logical IF)


 ■  Action

 If expression is true, statement is executed; if expression is false,
 program execution continues with the next executable statement


 ■  Syntax

      IF (expression) statement

 Parameter                                  Description

 expression                                 A logical expression.

 statement                                  Any executable statement except a
                                            «CASE,» DO, ELSE, ELSE IF, END,
                                            END IF, «END SELECT CASE, SELECT
                                            CASE,» block IF, or another
                                            logical IF statement. Note that
                                            the statement can be an
                                            arithmetic IF.


 ■  Examples

            IF (i .EQ. 0)  j = 2

            IF (x .GT. 2.3)  GOTO 100
            .
            .
            .
        100 CONTINUE


 IF THEN ELSE (Block IF)


 ■  Action

 If expression is true, statements in the IF block are executed; if
 expression is false, control is transferred to the next ELSE, ELSE IF, or
 END IF statement at the same IF level


 ■  Syntax

      IF (expression) THEN

 Parameter                                  Description

 expression                                 A logical expression


 ■  Remarks

 The associated IF block consists of all the executable statements (possibly
 none) between the IF statement and the next ELSE IF, ELSE, or END IF
 statement at the same IF level as this block IF statement. (IF levels are
 defined below.)

 The block IF statement transfers control as follows:

 If expression is:                          Then the next statement executed i

 True, and the IF block has no              The next END IF statement at the
 executable statements                      same IF level as the block IF
                                            statement

 True, and there is at least one            The first executable statement in
 executable statement in the IF             the IF block
 block

 False                                      The next ELSE IF, ELSE, or END IF
                                            statement at the same IF level as
                                            the block IF statement

 After execution of the last statement in the IF block, the next statement
 executed is the next END IF statement at the same IF level as the block IF
 statement.

 Transfer of control into an IF block from outside that block is not
 permitted.

 The IF level of a program statement is if minus endif, where if is the
 number of block IF statements from the beginning of the program unit in
 which the statement occurs, up to and including that statement, and endif is
 the number of END IF statements from the beginning of the program unit up
 to, but not including, that statement.

 Item                                       Required IF-Level Value

 Block IF, ELSE IF, ELSE, and END           Greater than 0 and less than
 IF                                         approximately 50

 END                                        0

 Other statements                           Greater than or equal to 0 and
                                            less than approximately 50


 ■  Examples

      C     Simple block IF that skips a group of statements
      C     if the expression is false:
            IF (i .LT. 10) THEN
      C     the next two statements are only executed if i is < 10
               j = i
               slice = TAN (angle)
            END IF

      C     Block IF with ELSE IF statements:

            IF (j .GT. 1000) THEN
      C     Statements here are executed only if J > 1000
            ELSE IF (j .GT. 100) THEN
      C     Statements here are executed only if J > 100 and j <= 1000
            ELSE IF (j .GT. 10) THEN
      C     Statements here are executed only if J > 10 and j <= 100
            ELSE
      C     Statements here are executed only if j <= 10
            END IF

      C     Nesting of constructs and use of an ELSE statement following
      C     a block IF without intervening ELSE IF statements:

            IF (i .LT. 100) THEN
      C     Statements here are executed only if i < 100
               IF (j .LT. 10) THEN
      C     Statements here are executed only if i < 100 and j < 10
               END IF
      C     Statements here are executed only if i < 100
            ELSE
      C     Statements here are executed only if i >= 100
               IF (j .LT. 10) THEN
      C     Statements here are executed only if i >= 100 and j < 10
               END IF
      C     Statements here are executed only if i >= 100
            END IF


 ■  See also

 ELSE
 ELSE IF
 END IF


 IMPLICIT


 ■  Action

 Defines the default type for user-declared names


 ■  Syntax

      IMPLICIT «{» type (letters) «, type (letters)»... «| NONE }»

 Parameter                                  Description

 type                                       One of the following types:

                                            CHARACTER
                                            CHARACTER*n
                                            COMPLEX
                                            «COMPLEX*8»
                                            «COMPLEX*16»
                                            «DOUBLE»
                                            «COMPLEX»
                                            DOUBLE PRECISION
                                            INTEGER
                                            «INTEGER*1»
                                            «INTEGER*2»
                                            «INTEGER*4»
                                            «INTEGER[C]»
                                            LOGICAL
                                            «LOGICAL*1»
                                            «LOGICAL*2»
                                            «LOGICAL*4»
                                            REAL
                                            «REAL*4»
                                            «REAL*8»

 letters                                    A list of single letters and
                                            ranges of letters. If more than
                                            one letter or range is listed,
                                            they must be separated by commas.
                                            Case is not significant.

                                            A range of letters is indicated
                                            by the first and last letters in
                                            the range, separated by a minus
                                            sign. The letters for a range
                                            must be in alphabetical order.
                                            «In this context, Microsoft
                                            FORTRAN allows the use of the
                                            dollar sign ($) as an alphabetic
                                            character that follows the letter
                                            Z.»


 ■  Remarks

 An IMPLICIT statement establishes the default type for all user-defined
 names that begin with any of the specified letters. An IMPLICIT statement
 applies only to the program unit in which it appears and does not change the
 type of any intrinsic function.

 «If IMPLICIT NONE is specified, then all user-defined names must be
 explicitly typed. An untyped name causes a compile-time error.» An explicit
 type statement can confirm or override the type established by an IMPLICIT
 statement. An explicit type in a FUNCTION statement also overrides the type
 indicated by an IMPLICIT statement. If the type in question is a character
 type, the length may be overridden by a later type definition.

 A program unit can have more than one IMPLICIT statement. However, all
 IMPLICIT statements must precede all other specification statements in that
 program unit. A particular letter cannot appear in more than one IMPLICIT
 statement in the same program unit.

 «A program unit can have only one IMPLICIT NONE statement. No other IMPLICIT
 statements may appear in a program unit that contains an IMPLICIT NONE
 statement.»


 ■  Example

            IMPLICIT INTEGER (a - b)
            IMPLICIT CHARACTER*10 (n)
      C     The following statement overrides the implicit
      C     INTEGER type for the variable anyname:
            CHARACTER*20 anyname
            age = 10
            name = 'PAUL'


 INCLUDE


 ■  «Action»

 «Inserts the contents of a specified text file at the location of the
 INCLUDE statement»


 ■  «Syntax»

      «INCLUDE 'filename'»

 «Parameter                                 Description»

 «filename»                                 «The name of the FORTRAN text
                                            file to include in the program,
                                            surrounded by apostrophes.


 ■  «Remarks»

 «The argument filename must be a valid text file specification for your
 operating system.»

 «The compiler considers the contents of the include file to be part of the
 program file and compiles them immediately. At the end of the include file,
 the compiler resumes processing the original source file at the line
 following the INCLUDE statement.»

 «Include files are primarily used for data or program units that appear in
 more than one program. Include files most often contain subroutines and
 functions, common block declarations, and EXTERNAL, INTERFACE, and INTRINSIC
 statements.»

 «Include files can also contain other INCLUDE statements and $INCLUDE
 metacommands; this is called "nesting" included files. The compiler allows
 you to nest any combination of up to ten INCLUDE statements or $INCLUDE
 metacommands. Your operating system may impose further restrictions.»


 ■  «Example»

 «This program implements a stack by declaring the common stack data in an
 include file. The contents of the file STKVARS.FOR (shown below in the
 following program) are inserted in the source code in place of every INCLUDE
 statement. This guarantees that all references to common storage for stack
 variables are consistent.»

          «INTEGER i
           REAL    x
           INCLUDE 'stkvars.for'

      C     read in five real numbers:
           DO 100 i = 1, 5
              READ (*, '(F10.5)') x
              CALL Push (x)
           100 CONTINUE

      C     write out the numbers in reverse order:
           WRITE (*, *) ' '
           DO 200 i = 1, 5
              CALL Pop (x)
              WRITE (*, *) x
           200 CONTINUE
           END
           SUBROUTINE Push (x)
      C     Pushes an element x onto the top of the stack.

           REAL x
           INCLUDE 'stkvars.for'

           top = top + 1
           IF (top .GT. stacksize)  STOP 'Stack overflow'
           stack(top) = x
           END

           SUBROUTINE pop(x)
      C     Pops an element from the top of the stack into x.

           REAL x
           INCLUDE 'stkvars.for'

           IF (top .LE. 0)  STOP 'Stack underflow'
              x   = stack(top)
              top = top - 1
           END»

 «The following is the text file STKVARS.FOR:»

     «C     This file contains the declaration of the common block
      C     for a stack implementation. Because this file contains an
      C     assignment statement, it must be included only after all
      C     other specification statements in each program unit.

           REAL     stack(500)
           INTEGER  top, stacksize

           COMMON  /stackbl/ stack, top

           stacksize = 500»


 INQUIRE


 ■  Action

 Returns the properties of a unit or external file


 ■  Syntax

      INQUIRE ({«UNIT=»unitspec | FILE=file}
      «, ACCESS=access»
     ««, BINARY=binary»»
      «, BLANK=blank»
     ««, BLOCKSIZE=blocksize»»
      «, DIRECT=direct»
      «, ERR=errlabel»
      «, EXIST=exist»
      «, FORM=form»
      «, FORMATTED=formatted»
      «, IOSTAT=iocheck»
     ««, MODE=mode»»
      «, NAME=name»
      «, NAMED=named»
      «, NEXTREC=nextrec»
      «, NUMBER=num»
      «, OPENED=opened»
      «, RECL=recl»
      «, SEQUENTIAL=seq»
     ««, SHARE=share»»
      «, UNFORMATTED=unformatted»)

 If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   The unitspec parameter can be
                                            either an integer expression or
                                            an asterisk (*). If UNIT = * is
                                            specified, you may not include
                                            the NUMBER= option, or a run-time
                                            error occurs.

                                            Exactly one unitspec or file must
                                            be specified, but not both. If
                                            unitspec is given, the inquiry is
                                            an "inquire-by-unit" operation.

 file                                       A character expression that the
                                            operating system recognizes as
                                            the name of the file you are
                                            interested in. Exactly one
                                            unitspec or file must be
                                            specified, but not both. If file
                                            is given, the inquiry is an
                                            "inquire-by-file" operation.

 access                                     A character variable, array
                                            element, «or structure element».
                                            Returns 'SEQUENTIAL' if the
                                            specified unit or file is open
                                            for appending data. Returns
                                            'SEQUENTIAL' if the specified
                                            unit or file is connected for
                                            sequential access. Returns
                                            'DIRECT' if the specified unit or
                                            file is connected for direct
                                            access. In an inquire-by-unit
                                            operation, if no file is
                                            connected to unitspec, access is
                                            undefined.

 «binary»                                   «A character variable, array
                                            element, or structure element.
                                            Returns 'YES' if binary is among
                                            the allowable forms for the
                                            specified file or unit. Returns
                                            'NO' or 'UNKNOWN' otherwise.»

 blank                                      A character variable, array
                                            element, «or structure element».
                                            Returns 'NULL' if the BN edit
                                            descriptor is in effect, and
                                            returns 'ZERO' if BZ is in
                                            effect.

 «blocksize»                                «An integer variable, array
                                            element, or structure element. If
                                            the unitspec or file is
                                            connected, blocksize returns the
                                            I/O buffer size. If unitspec or
                                            file is not connected, blocksize
                                            is undefined.»

 direct                                     A character variable, array
                                            element, «or structure element».
                                            Returns 'YES' if direct is among
                                            the allowable access modes for
                                            the specified file or unit.
                                            Returns 'NO' or 'UNKNOWN'
                                            otherwise.

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit. If errlabel is specified,
                                            an I/O error causes transfer of
                                            control to the statement at
                                            errlabel. If errlabel is omitted,
                                            the effect of an I/O error is
                                            determined by the presence or
                                            absence of iocheck.

 exist                                      A logical variable, array
                                            element, «or structure element».
                                            Returns .TRUE. if the specified
                                            unit or file exists; returns
                                            .FALSE. otherwise.

 form                                       A character variable or array
                                            element. Returns  'FORMATTED' if
                                            the specified unit or file is
                                            connected for formatted I/O;
                                            returns 'UNFORMATTED' for
                                            unformatted I/O; «returns
                                            'BINARY' for binary I/O».

 formatted                                  A character variable, array
                                            element, «or structure element».
                                            Returns 'YES' if formatted is
                                            among the allowable forms for the
                                            specified file or unit; returns
                                            'NO' or 'UNKNOWN' otherwise.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            there is no error, or the number
                                            of the error message if an error
                                            occurs. For more information on
                                            error handling, see Section
                                            3.2.6, "Error and End-of-File
                                            Handling."

 «mode»                                      «A character variable, array
                                            element, or structure element
                                            that returns the current mode
                                            status of the specified file or
                                            unit. The returned modes are the
                                            same as those specified in the
                                            OPEN statement: 'READ', 'WRITE',
                                            and 'READWRITE'. In an
                                            inquire-by-unit operation, if no
                                            file is connected to  unitspec,
                                            mode is undefined.»

 name                                       A character variable, array
                                            element, «or structure element».
                                            In an inquire-by-unit operation,
                                            returns the name of the file
                                            connected to unitspec. If no file
                                            is connected to unitspec, or if
                                            the file connected to unitspec
                                            does not have a name, name is
                                            undefined. In an inquire-by-file
                                            operation, returns the name
                                            specified for file.

 named                                      A logical variable, array
                                            element, «or structure element».
                                            Returns .FALSE. if the file
                                            specified by file or attached to
                                            unitspec is not open or if it is
                                            a scratch file; returns .TRUE.
                                            otherwise.

 nextrec                                    An integer variable, array
                                            element, «or structure element»
                                            that returns the record number of
                                            the next record in a
                                            direct-access file. (The first
                                            record in a direct-access file
                                            has record number 1.)

 num                                        An integer variable, array
                                            element, «or structure element».
                                            In an inquire-by-file operation,
                                            returns the number of the unit
                                            connected to file. If no unit is
                                            connected to file, num is
                                            undefined. In an inquire-by-unit
                                            operation, returns the number
                                            given in unitspec. If you specify
                                            UNIT = *, do not include the
                                            option NUMBER=, or a run-time
                                            error occurs.

 opened                                     A logical variable, array
                                            element, «or structure element».
                                            In an inquire-by-unit operation,
                                            returns .TRUE. if any file is
                                            currently connected to unitspec;
                                            returns .FALSE. otherwise. In an
                                            inquire-by-file operation,
                                            returns .TRUE. if file is
                                            currently connected to any unit;
                                            returns .FALSE. otherwise.

 recl                                       An integer variable, element
                                            name, «or structure element» that
                                            returns the length (in bytes) of
                                            each record in a direct-access
                                            file. If the file is connected
                                            for unformatted I/O, the value is
                                            in processor-dependent units.
                                            «(In Microsoft FORTRAN, this is
                                            also the number of bytes.)»

 seq                                        A character variable, array
                                            element, «or structure element».
                                            Returns 'YES' if sequential is
                                            among the allowable access modes
                                            for the specified file or unit;
                                            returns 'NO' or 'UNKNOWN'
                                            otherwise.

 «share»                                     «A character variable, array
                                            element, or structure element.
                                            Returns the current share status
                                            of the specified file or unit.
                                            The returned values are the same
                                            as those specified in the OPEN
                                            statement: 'COMPAT', 'DENYRW',
                                            'DENYWR', 'DENYRD', and
                                            'DENYNONE'. In an inquire-by-unit
                                            operation, if no file is
                                            connected to unitspec, share is
                                            undefined.»

 unformatted                                A character variable, array
                                            element, «or structure element».
                                            Returns 'YES' if unformatted is
                                            among the allowable forms for the
                                            specified file or unit; returns
                                            'NO' or 'UNKNOWN' otherwise.


 ■  Remarks

 The INQUIRE statement returns the values of the various attributes with
 which a file was opened. Note that the INQUIRE statement cannot return the
 properties of an unopened file, and it cannot distinguish between attributes
 that are specified by you and attributes that are set by default.

 You can execute the INQUIRE statement at any time. It returns the values
 that are current at the time of the call.

 If a parameter of the INQUIRE statement is an expression that calls a
 function, that function must not execute an I/O statement «or the EOF
 intrinsic function» because the results are unpredictable.


 ■  Example

      C     This program prompts for the name of a data file. The INQUIRE
      C     statement then determines whether or not the file exists.
      C     If it does not, the program prompts for another file name.

            CHARACTER*12  fname
            LOGICAL       exists

      C     Get the name of a file:
            100 WRITE (*, '(1X, A\)') 'Enter the file name: '
            READ  (*, '(A)') fname

      C     INQUIRE about file's existence:
            INQUIRE (FILE = fname, EXIST = exists)

            IF (.NOT. exists) THEN
               WRITE (*,'(2A/)') ' > Cannot find file ', fname
               GOTO 100
            END IF
            .
            .
            .
            END


 INTEGER


 ■  Action

 Specifies the INTEGER type for user-defined names


 ■  Syntax

     INTEGER«{« *bytes» |  « [C] »}» »vname ««[attrs] »
     « *length»»«*length»»«(dim)»««/values/»
     «, vname ««[attrs] »«*length»»«(dim)»«« /values/»»»...

 The order of the length and dim parameters can be reversed.

 Parameter                                  Description

 «bytes»                                    «Must be 1, 2, or 4. Specifies
                                            the length, in bytes, of the
                                            names in the INTEGER statement.
                                            This value can be overridden by
                                            the length parameter.»

 vname                                      The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function, or
                                            intrinsic function; or, a
                                            function subprogram or an array
                                            declarator. The vname parameter
                                            cannot be the name of a
                                            subroutine or main program.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            vname. These attributes can be
                                            used with vname: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR, PASCAL, REFERENCE,
                                            VALUE.»

 «length»                                   «Must be 1, 2, or 4. Gives vname
                                            the specified length. The length
                                            parameter overrides the length
                                            specified by bytes.»

 dim                                        A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant takes the form
                                            n*constant, where n is a positive
                                            integer constant. The /values/
                                            option initializes vname. The
                                            following statement declares that
                                            num is of type INTEGER, and sets
                                            num equal to 10:»

                                                 «INTEGER num / 10 /»


 ■  Remarks

 An INTEGER statement confirms or overrides the implicit type of vname. The
 name vname is defined for the entire program unit, and cannot be defined by
 any other type statement in that program unit.

 «If both the bytes and length parameters are omitted, the variable's length
 defaults to the value given in the $STORAGE metacommand.»

 INTEGER statements must precede all executable statements.


 ■  Examples

            INTEGER count, matrix(4, 4), «sum*2»
           «INTEGER*22 q, m12*24, ivec(10)*24, z*24(10)»


 INTERFACE TO


 ■  «Action»

 «Specifies the name of a subroutine or function, its attributes and formal
 argument types»


 ■  «Syntax»

      «INTERFACE TO {function-statement | subroutine-statement}
      «(Formal-parameter-declaration(s))»
      «END»

 «Parameter                                 Description»

 «function-statement»                       «A function declaration
                                            statement»

 «subroutine-statement»                     «A subroutine declaration
                                            statement»


 ■  «Remarks»

 «An interface has two parts. The first is the INTERFACE TO statement
 followed by a FUNCTION or SUBROUTINE declaration statement. The second is a
 list of declarations for all the procedure's formal arguments. Only type,
 DIMENSION, EXTERNAL,  INTRINSIC, and END statements can appear in the second
 part.»

 «The INTERFACE TO statement confirms the correctness of subprogram calls.
 The compiler verifies that the number and types of arguments in a subprogram
 call are consistent with those in the interface. (Note that the interface
 must appear prior to any subprogram references for the checking to work.)»

 «The attributes in an interface override the default definitions in the
 subprogram definition. However, if you use an attribute in the subprogram
 declaration or its arguments, the same attribute must also appear in the
 INTERFACE TO statement.»


 ■  «Example»

           «INTERFACE TO INTEGER FUNCTION Func (p, d, q)
            INTEGER*2 p
            REAL d[C]         ! required; appears in function definition
            REAL*8 q[FAR]     ! overrides default in function definition
            END»

 «The interface above describes the following function to the calling unit:»

           «INTEGER FUNCTION Func (r, s, t)
            INTEGER*2 r
            REAL s[C]
            REAL*8 t
            .
            .
            .
            END»


 INTRINSIC


 ■  Action

 Declares that a name is an intrinsic function


 ■  Syntax

      INTRINSIC names

 Parameter                                  Description

 names                                      One or more intrinsic-function
                                            names. If more than one name is
                                            specified, the names must be
                                            separated by commas.


 ■  Remarks

 You must specify the name of an intrinsic function in an INTRINSIC statement
 if you wish to pass that intrinsic function as an argument.

 A name may appear only once in an INTRINSIC statement. A name appearing in
 an  INTRINSIC statement cannot also appear in an EXTERNAL statement. All
 names used in an INTRINSIC statement must be system-defined intrinsic
 functions. For a list of these functions, see Chapter 5, "Intrinsic
 Functions and Additional Procedures."


 ■  Example

            INTRINSIC SIN, COS
      C     SIN and COS are arguments to Calc2:
            result = Calc2 (SIN, COS)


 LOCKING


 ■  «Action»

 «Locks direct-access files and records to prevent access by other users in a
 network environment»


 ■  «Syntax»

      «LOCKING («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»
      «, LOCKMODE=lockmode»
      «, REC=rec»
      «, RECORDS=recnum»)»

 «If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.»

 «Parameter                                 Description»

 «unitspec»                                 «An integer expression for the
                                            unit to be locked. The file
                                            attached to unitspec must have
                                            been opened for direct access.»

 «errlabel»                                 «The label of an executable
                                            statement in the same program
                                            unit. If errlabel is specified,
                                            an I/O error causes transfer of
                                            control to the statement at
                                            errlabel. If errlabel is omitted,
                                            the effect of an I/O error is
                                            determined by the presence or
                                            absence of iocheck.»

 «iocheck»                                  «An integer variable, array
                                            element, or structure element
                                            that returns the value zero if
                                            there is no error, or the number
                                            of the error message if an error
                                            occurs. For more information on
                                            error handling, see Section
                                            3.2.6, "Error and End-of-File
                                            Handling."»

 «lockmode»                                 «A character expression with one
                                            of the following values:»

                                            «Value    Description»

                                            «'NBLCK'  Locks the specified
                                                      region for reading and
                                                      writing. If any of the
                                                      records are already
                                                      locked by a different
                                                      process, a run-time
                                                      error occurs. This non-
                                                      blocking lock is the
                                                      default.»

                                            «'LOCK'   Locks the specified
                                                      region for reading and
                                                      writing. Waits for any
                                                      part of the region
                                                      locked by a different
                                                      process to become
                                                      available.»

                                            «'NBRLCK' Locks the specified
                                                      region for reading;
                                                      records may still be
                                                      written. This non-
                                                      blocking read lock is
                                                      otherwise the same as
                                                      'NBLCK'.»

                                            «'RLCK'   Locks the specified
                                                      region for reading;
                                                      records may still be
                                                      written. This read lock
                                                      is other-wise the same
                                                      as 'LOCK'.»

                                            «'UNLCK'  Unlocks the specified
                                                      region.»

 «rec»                                      «An integer expression that is
                                            the number of the first record in
                                            a group of records to be locked
                                            or unlocked. If rec is omitted,
                                            the next record (the one that
                                            would be read by a sequential
                                            read) is locked.»

 «recnum»                                   «An integer expression that is
                                            the number of records to be
                                            locked. It defaults to one.»


 ■  «Remarks»

 «The LOCKING statement has no effect when used with versions of MS-DOS
 before 3.0.»

 «If a parameter of the LOCKING statement is an expression that calls a
 function, that function must not cause an I/O statement or the EOF intrinsic
 function to be executed, because the results are unpredictable.»


 LOGICAL


 ■  Action

 Specifies the LOGICAL type for user-defined names


 ■  Syntax

 LOGICAL ««*bytes»»vname ««[attrs]»«*length»»«(dim)»
 ««/values/»»«, vname««[attrs]»«*length»»«(dim»
 ««/values»»»»»...

 Parameter                                  Description

 «bytes»                                    «Must be 1, 2, or 4. Specifies
                                            the length, in bytes, of the
                                            names in the LOGICAL statement.
                                            This value can be overridden by
                                            the length parameter.»

 vname                                      The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function, or
                                            intrinsic function; or, a
                                            function subprogram or an array
                                            declarator. The vname parameter
                                            cannot be the name of a
                                            subroutine or main program.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            vname. These attributes can be
                                            used with vname: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR,  PASCAL, REFERENCE,
                                            VALUE.»

 «length»                                   «Must be 1, 2, or 4. Gives vname
                                            the specified length. The length
                                            parameter overrides the length
                                            specified by bytes.»

 dim                                        A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant takes the form
                                            n*constant, where n is a
                                            positive-nonzero-integer
                                            constant, and is equivalent to
                                            constant repeated n times. The
                                            /values/ option initializes
                                            vname. The following statement
                                            declares that switch is of type
                                            LOGICAL, and sets switch  equal
                                            to .TRUE.:»

                                             «LOGICAL switch /.TRUE./»


 ■  Remarks

 A LOGICAL statement confirms or overrides the implicit type of vname. The
 name vname is defined for the entire program unit, and it cannot be defined
 by any other type statement in that program unit.

 «If both the bytes and length parameters are omitted, the variable's length
 defaults to the value given in the $STORAGE metacommand.»

 LOGICAL statements must precede all executable statements.


 ■  Example

            LOGICAL  switch
           «LOGICAL*1  flag»


 MAP...END MAP


 ■  «Action»

 «Delimits a group of variable type declarations within a STRUCTURE
 statement»


 ■  «Syntax»

      «MAP
             type-declaration(s)
             .
             .
             .
      END MAP»

 «Parameter                                 Description»

 «type-declaration»                         «A variable declaration; can be
                                            any conventional type (INTEGER,
                                            REAL*8, CHARACTER, etc.), or a
                                            RECORD declaration of a structure
                                            type


 ■  «Remarks»

 «A MAP...END MAP block must appear within a UNION...END UNION block.»

 «Any number of variables may appear within a map, and they may be a mixture
 of types. Variables are stored in memory contiguously, in the same order
 they appear in the MAP statement.»

 «There are always at least two maps within a union. The purpose of a union
 is to assign two or more maps to the same memory location. Unlike the
 FORTRAN EQUIVALENCE statement which overlays single variables or arrays, the
 MAP statement allows groups of variables to overlay other groups in the same
 union.»

 «For example, the two maps in the following union each contain three REAL*8
 variables. The variable manny shares the same memory space as meg, moe, and
 jo , and so forth.»

           «UNION
               MAP
                  REAL*8  meg, jo, amy
               END MAP
               MAP
                  REAL*8  manny, moe, jack
               END MAP
            END UNION»


 ■  «Examples»

 «Within a structure, maps can contain variables of intrinsic type:»

           «UNION
               MAP
                  CHARACTER*20 string
               END MAP
               MAP
            INTEGER*2 number(10)
               END MAP
            END UNION»

 «Or they can contain variables with types defined in previous STRUCTURE
 statements:»

           «UNION
               MAP
                  RECORD /Cartesian/ xcoord, ycoord
               END MAP
               MAP
                  RECORD /Polar/ length, angle
               END MAP
            END UNION»


 ■  «See also»

 «STRUCTURE...END STRUCTURE
 UNION...END UNION»


 NAMELIST


 ■  «Action»

 «Declares a group name for a set of variables so they may be read or written
 with a single namelist-directed READ or WRITE statement»


 ■  «Syntax»

      «NAMELIST /namlst/ varlist «/namlst/ varlist»»

 «Parameter                                 Description»

 «namlst»                                   «The name for a group of
                                            variables

 «varlist»                                  «A list of variable names and
                                            array names; may not include
                                            structure variables and formal
                                            arguments


 ■  «Remarks»

 «The same variable name may appear in more than one namelist.»

 «Any namelist can appear more than once in the same program unit. The list
 of elements in a particular namelist is treated as a continuation of the
 list in the previous namelist with the same name.»

 «If the namelist is used in a WRITE statement, the list of variable names
 followed by their values is written to a file or displayed.»

 «If the namelist is used in a READ statement, the file connected to unit or
 the user's input is scanned for the names in the namelist and their
 corresponding values. The order in which the names appear is not important,
 nor do all the names in the namelist have to appear.»

 «Although a namelist is limited to single variables and complete arrays, a
 file to be read may include the names of individual array elements and
 character variable substrings.»

 «Note that when you use a namelist in a» «READ» «statement, the program
 expects the variable names you enter to exactly match the names you gave in
 the NAMELIST statement. However, when the $TRUNCATE metacommand is used, the
 compiler truncates all variable names to six characters. In this
 circumstance, the program may not recognize the original, untruncated name;
 you must enter the shortened version of the name.»

 «For example, if one of your NAMELIST variables is impedance and you use the
 $TRUNCATE metacommand, a READ statement using the namelist expects impeda,
 not impedance.»


 ■  «Example»

           «INTEGER    i1*1, i2*2, i4*4, iarray(3)
            LOGICAL    11*1, l2*2, l4*4
            REAL       r4*4, r8*8
            COMPLEX    z8*8, z16*16
            CHARACTER  c1*1, c10*10

            NAMELIST /example/ i1, i2, i4, 11, l2, l4, r4, r8,
           +                   z8, z16, c1, c10, iarray

            i1  =  11
            i2  =  12
            i4  =  14
            11  = .TRUE.
            l2  = .TRUE.
            l4  = .TRUE.
            r4  =  24.0
            r8  =  28.0d0
            z8  = (38.0, 0.0)
            z16 = (316.0d0, 0.0d0)
            c1  = 'A'
            c10 = '0123456789'
            iarray(1) = 41
            iarray(2) = 42
            iarray(3) = 43

            WRITE (*, example)»

 «The following output is produced:»

      «&EXAMPLE
      I1 =           11
      I2 =           12
      I4 =           14
      L1 =  T
      L2 =  T
      L4 =  T
      R4 =        24.000000
      R8 =        28.000000000000000
      Z8 =           (38.000000, 0.000000E+00)
      Z16 =         (316.000000000000000, 0.000000000000000E+000)
      C1 = 'A'
      C10 = '0123456789'
      IARRAY =           41          42          43»

 «If a namelist-directed read is performed using READ (*, example) with the
 following input:»

      «&example
      I1=99
      L1=F
      R4=99.0
      Z8=(99.0, 0.0)
      C1='Z'
      IARRAY(1)=99
      /»

 «a second WRITE (*, example) statement produces the following output:»

      «&EXAMPLE
      I1 =           99
      I2 =           12
      I4 =           14
      L1 =  F
      L2 =  T
      L4 =  T
      R4 =        99.000000
      R8 =        28.000000000000000
      Z8 =           (99.000000, 0.000000E+00)
      Z16 =         (316.000000000000000, 0.000000000000000E+000)
      C1 = 'Z'
      C10 = '0123456789'
      IARRAY =           99          42          43
      /»


 OPEN


 ■  Action

 Associates a unit number with an external file or device


 ■  Syntax

      OPEN («UNIT=»unitspec
      «, ACCESS=access»
      «, BLANK=blanks»
     ««, BLOCKSIZE=blocksize»»
      «, ERR=errlabel»
      «, FILE=file»
      «, FORM=form»
      «, IOSTAT=iocheck»
     ««, MODE=mode»»
      «, RECL=recl»
     ««, SHARE=share»»
      «, STATUS=status»)

 If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   An integer expression that
                                            specifies an external unit.

 access                                     A character expression that
                                            evaluates to «'APPEND',»
                                            'DIRECT', or 'SEQUENTIAL' (the
                                            default).

 blanks                                     A character expression that
                                            evaluates to 'NULL' or 'ZERO'.
                                            Set to 'NULL' (the default) to
                                            ignore blanks (same as BN edit
                                            descriptor, unless overridden by
                                            an explicit BZ in the READ
                                            statement). Set to 'ZERO' to
                                            treat blanks as zeros (same as BZ
                                            edit descriptor, unless
                                            overridden by an explicit BN in
                                            the READ statement).

 «blocksize»                                «An integer expression specifying
                                            internal buffer size for the
                                            unit. See Section 3.2.4,
                                            "Input/Output Buffer Size."»

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit. If errlabel is specified,
                                            an I/O error causes transfer of
                                            control to the statement at
                                            errlabel. If errlabel is omitted,
                                            the effect of an I/O error is
                                            determined by the presence or
                                            absence of iocheck.

 file                                       A character expression. If file
                                            is omitted, the compiler creates
                                            a scratch file unique to the
                                            specified unit. The scratch file
                                            is deleted when it is either
                                            explicitly closed or the program
                                            terminates normally.

                                            «If the file name specified is
                                            blank (FILE=' '), the following
                                            steps are taken:»

                                            1. «The program reads a file name
                                               from the list of arguments (if
                                               any) in the command line that
                                               invoked the program. If the
                                               argument is a null string
                                               (''), the user is prompted for
                                               the corresponding file name.
                                               Each succeeding nameless OPEN
                                               statement reads a successive
                                               command-line argument.»

                                            2. «If there are more nameless
                                               OPEN statements than command-
                                               line arguments, the program
                                               prompts for additional file
                                               names.»

                                            «Assume the following command
                                            line has invoked the program
                                            MYPROG:»

                                            «myprog first.fil " " third.txt»

                                            «MYPROG contains four OPEN
                                            statements with blank file names,
                                            in the following order:»

                                            «OPEN (2,  FILE = ' ')
                                             OPEN (4,  FILE = ' ')
                                             OPEN (5,  FILE = ' ')
                                             OPEN (10, FILE = ' ')»

                                            «Unit 2 is associated with the
                                            file FIRST.FIL. Since a null
                                            argument was specified on the
                                            command line for the second file
                                            name, the OPEN statement for unit
                                            4 produces the following prompt:»

                                            «File name missing or blank
                                            -Please enter name UNIT 4?»

                                            «Unit 5 is associated with the
                                            file THIRD.TXT. As no fourth file
                                            was specified on the command
                                            line, the OPEN statement for unit
                                            10 produces the following
                                            prompt:»

                                            «File name missing or blank
                                            -Please enter name UNIT 10?»

 form                                       A character expression that
                                            evaluates to 'FORMATTED',
                                            'UNFORMATTED'«, or 'BINARY'».

                                            If access is sequential, the
                                            default for form is 'FORMATTED';
                                            if access is direct, the default
                                            is 'UNFORMATTED'.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            no error occurs, a negative
                                            integer if an end-of-file record
                                            is encountered, or the number of
                                            the error message if an error
                                            occurs. See Section 3.2.6, "Error
                                            and End-of-File Handling."

 «mode»                                     «A character expression that
                                            evaluates to 'READ' (the process
                                            can only read the file), 'WRITE'
                                            (the process can only write to
                                            the file), or 'READWRITE' (the
                                            process can both read from and
                                            write to the file).»

                                            «If mode is not specified, the
                                            FORTRAN run-time system attempts
                                            to open the file with a mode of
                                            'READWRITE'. If this open
                                            operation fails, the run-time
                                            system tries to open the file
                                            again, first using 'READ', then
                                            using 'WRITE'.»

                                            «Note that this action is not the
                                            same as specifying
                                            MODE='READWRITE'. If you specify
                                            the option MODE='READWRITE', and
                                            the file cannot be opened for
                                            both read and write access, the
                                            attempt to open the file fails.
                                            The default behavior (trying
                                            'READWRITE', then 'READ', then
                                            'WRITE') is more flexible (though
                                            you may have to use the INQUIRE
                                            statement to determine the actual
                                            access mode selected). The value
                                            of the STATUS= option does not
                                            affect mode.»

 recl                                       An integer expression that
                                            specifies the length of each
                                            record in bytes. This parameter
                                            is required for direct-access
                                            files, and is ignored for
                                            sequential files.

 «share»                                    «A character expression. The
                                            acceptable values of share are:»

                                            «Value      Description»

                                            «'COMPAT'   Compatibility mode
                                                        (the default mode in
                                                        DOS). While a file is
                                                        open in compatibility
                                                        mode, the original
                                                        user (the process
                                                        that opened the file)
                                                        may open the file in
                                                        compatibility mode
                                                        any number of times.
                                                        No other user may
                                                        open the file.»

                                                        A file that is
                                                        already open in a
                                                        mode other than
                                                        compatibility mode
                                                        cannot be opened in
                                                        compatibility mode.»

                                            «'DENYRW'   Deny-read/write mode.
                                                        While a file is open
                                                        in deny-read/write
                                                        mode, no other
                                                        process may open the
                                                        file.»

                                            «'DENYWR'   Deny-write mode.
                                                        While a file is open
                                                        in deny-write mode,
                                                        no process may open
                                                        the file with write
                                                        access.»

                                            «'DENYRD'   Deny-read mode. While
                                                        a file is open in
                                                        deny-read mode, no
                                                        process may open the
                                                        file with read
                                                        access.»

                                            «'DENYNONE' Deny-none mode (the
                                                        default mode in
                                                        OS/2). While a file
                                                        is open in deny-none
                                                        mode, any process may
                                                        open the file in any
                                                        mode (except
                                                        compatibility mode).»

                                                        See Section 3.2.11
                                                        for information on
                                                        file sharing.»

 status                                     A character expression that
                                            evaluates to 'OLD', 'NEW',
                                            'UNKNOWN' (default), or
                                            'SCRATCH'.

                                            The following list describes the
                                            four values of status:

                                            Value     Description

                                            'OLD'     The file must already
                                                      exist. If the file
                                                      exists, it is opened.
                                                      If it does not exist,
                                                      an I/O error occurs.

                                                      If you open an existing
                                                      sequential file and
                                                      write to it or close it
                                                      without first moving to
                                                      the end, you overwrite
                                                      the file. (Opening with
                                                      ACCESS = 'APPEND'
                                                      automatically positions
                                                      the file at the end.)

                                            'NEW'     The file must not
                                                      already exist. If the
                                                      file does not exist, it
                                                      is created. If it
                                                      exists, an I/O error
                                                      occurs.

                                                      If you open a file with
                                                      the option
                                                      STATUS='NEW', and close
                                                      it with STATUS='KEEP',
                                                      or if the program
                                                      terminates without
                                                      explicitly closing the
                                                      file, a permanent file
                                                      is created.

                                            'SCRATCH' If you omit the fname
                                                      parameter when opening
                                                      a file, the value of
                                                      status defaults to
                                                      'SCRATCH'. Scratch
                                                      files are temporary.
                                                      They are deleted when
                                                      explicitly closed or
                                                      when the program
                                                      terminates. You can
                                                      also open a named file
                                                      with the explicit
                                                      option STATUS=
                                                      'SCRATCH'. The file is
                                                      deleted at program
                                                      termination.

                                            'UNKNOWN' If you open a file with
                                                      the STATUS='UNKNOWN'
                                                      option, the FORTRAN
                                                      run-time system first
                                                      attempts to open the
                                                      file with status equal
                                                      to 'OLD', then with
                                                      status equal to 'NEW'.
                                                      Therefore, if the file
                                                      exists, it is opened;
                                                      if it does not exist,
                                                      it is created.

                                                      The 'UNKNOWN' status
                                                      avoids any run-time
                                                      errors associated with
                                                      opening an existing
                                                      file with STATUS='NEW'
                                                      or opening a non-
                                                      existent file with
                                                      STATUS='OLD'.

                                            Status values affect only disk
                                            files. When devices such as the
                                            keyboard or the printer are
                                            opened, the STATUS= option is
                                            ignored.


 ■  Remarks

 Opening a file for unit* has no effect, because unit* is permanently
 connected to the keyboard and screen. «You can, however, use the OPEN
 statement to connect the other preconnected units (0, 5, and 6) to a file.»

 If you open a file using a unit number that is already connected to another
 file, the already-open file is automatically closed first, then the new file
 name is connected to the unit; no error occurs. However, you cannot connect
 a single file to more than one unit at a time.

 «If you have not explicitly opened a file, and the first operation is a READ
 or WRITE, the FORTRAN run-time system attempts to open a file as if a blank
 name were specified, as described with the file parameter.»

 If a parameter of the OPEN statement is an expression that calls a function,
 that function must not cause an I/O statement «or the EOF intrinsic
 function» to be executed, because the results are unpredictable.

 «For more information on choosing values of share and mode when sharing
 files, see Section 3.2.11, "File Sharing."»


 ■  Examples

 The following example opens a new file:

      C     Prompt user for a file name and read it:
            CHARACTER*64 filename
            WRITE (*, '(A\)') ' enter file name '
            READ  (*, '(A)') filename
      C     Open the file for formatted sequential access as unit 7.
      C     Note that the specified access need not have been specified,
      C     since it is the default (as is "formatted").
            OPEN (7, FILE = filename, ACCESS = 'SEQUENTIAL',
           +      STATUS = 'NEW')

 The following example opens an existing file:

      C     Open a file created by an editor, "DATA3.TXT", as unit 3:
            OPEN (3, FILE = 'DATA3.TXT')


 PARAMETER


 ■  Action

 Gives a constant a symbolic name


 ■  Syntax

      PARAMETER (name=expression «, name=expression»...)

 Parameter                                  Description

 name                                       A symbolic name.

 expression                                 An expression. The expression can
                                            include symbolic names only if
                                            those names were defined in a
                                            previous PARAMETER statement in
                                            the program unit.

                                            If the expression is a character
                                            constant, it may contain no more
                                            than 1,957 bytes.


 ■  Remarks

 If expression is of a different numerical type than name, any needed
 conversion is performed automatically.

 Unlike the DATA statement (which declares an initial value for a variable),
 the PARAMETER statement sets fixed values for symbolic constants. Any
 attempt to change a symbolic constant with an assignment statement or a READ
 statement causes a compile-time error.

 A symbolic constant «cannot be of a structure type, and it» may not appear
 in a format specification.


 ■  Examples

      C     Example 1
            PARAMETER (nblocks = 10)

      C     Example 2
            REAL mass
            PARAMETER (mass = 47.3, pi = 3.14159)

      C     Example 3
            IMPLICIT REAL (L-M)
            PARAMETER (loads = 10.0, mass = 32.2)

      C     Example 4
            CHARACTER*(*) bigone
            PARAMETER (bigone = 'This constant is larger than 40
           +characters')


 PAUSE


 ■  Action

 Temporarily suspends program execution «and allows you to execute operating
 system commands during the suspension»


 ■  Syntax

      PAUSE «prompt»

 Parameter                                  Description

 prompt                                     Either a character constant or an
                                            integer from 0 to 99,999


 ■  Remarks

 «If prompt is not specified, the following default prompt appears:»

      «'Please enter a blank line (to continue) or a system command.'»

 If prompt is a number, the number appears on the screen preceded by PAUSE -.
 If prompt is a string, only the string is displayed.

 The following are possible responses to the prompt and their results:

 If the user enters:                        Then:

 A blank line                               Control returns to the program.

 «A command»                                «The command is executed and
                                            control returns to the program.»

 «The word "COMMAND" (uppercase or          «The user can carry out a
 lowercase)»                                sequence of commands. To return
                                            control to the program, enter
                                            EXIT (uppercase or lowercase).»


 ■  Examples

            SUBROUTINE SetDrive ( )
               PAUSE 'Please select default drive.'

 One of the following outputs is produced:

      Please select default drive.
      .
      .
      .

      PAUSE 999.
      .
      .
      .


      PAUSE- 999


 PRINT


 ■  Action

 Displays output on the screen


 ■  Syntax

      PRINT { *, | formatspec | «namelist» } «,iolist»

 Parameter                                  Description

 formatspec                                 A format specifier

 «namelist»                                 «A namelist specifier»

 iolist                                     A list specifying the data to be
                                            transferred


 ■  Remarks

 If *, replaces formatspec, output is list-directed.

 «If a namelist specifier is used, no iolist may appear.»

 The PRINT statement writes only to unit*.

 «The iolist may not contain structure variables, only structure elements.»

 If a parameter of the PRINT statement is an expression that calls a
 function, that function must not execute an I/O statement «or the EOF
 intrinsic function», because the results are unpredictable.


 ■  Example

      C     The two following statements are equivalent:
            PRINT     '(A11)', 'Abbottsford'
            WRITE (*, '(A11)') 'Abbottsford'


 PROGRAM


 ■  Action

 Identifies the program unit as a main program and gives it a name


 ■  Syntax

      PROGRAM program-name

 Parameter                                  Description

 program-name                               The name you have given to your
                                            main program.

                                            The program name is global.
                                            Therefore, it cannot be the name
                                            of a user-defined procedure or
                                            common block. It is also a local
                                            name and must not conflict with
                                            any name local to the main
                                            program.


 ■  Remarks

 The PROGRAM statement may only appear as the first statement of a main
 program.

 The main program is always assigned the system name of _main, in addition to
 any name specified by the PROGRAM statement.


 ■  Example

            PROGRAM Gauss
            .
            .
            .
            END


 READ


 ■  Action

 Transfers data from the file associated with unitspec to the items in the
 iolist until all read operations have been completed, the end-of-file is
 reached, or an error occurs


 ■  Syntax

      READ { «{» formatspec, «| nmlspec }» |
      («UNIT=»unitspec
      «, ««{»«FMT=»formatspec» «| «NML=»nmlspec}»»
      «, END=endlabel»
      «, ERR=errlabel»
      «, IOSTAT=iocheck»
      «, REC=rec»)}
      iolist

 If UNIT= is omitted, unitspec must be the first parameter. If FMT= «or NML=»
 is omitted, fmtspec «or nmlspec» must be the second parameter. The
 parameters can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   A unit specifier.

                                            When reading from an internal
                                            file, unitspec is a character
                                            substring, character variable,
                                            character array element,
                                            character array, «noncharacter
                                            array, or structure element».
                                            When reading from an external
                                            file, unitspec is an integer
                                            expression that specifies a
                                            unit.

                                            «If unitspec has not been
                                            explicitly associated with a
                                            file, an "implicit open," using
                                            the following syntax is
                                            performed:»

                                            «OPEN (unitspec, FILE = ' ',
                                            STATUS = 'OLD', ACCESS =
                                            'SEQUENTIAL', FORM = form)»

                                            «where form is 'FORMATTED' for a
                                            formatted read operation or
                                            'UNFORMATTED' for an unformatted
                                            read operation.»

                                            If you included a file name on
                                            the command line that executed
                                            this program, that name will be
                                            used for the file. If you did not
                                            include a file name, you will be
                                            prompted for one. For more
                                            information, see Chapter 3, "The
                                            Input/Output System."

 formatspec                                 A format specifier. It can be the
                                            label of a FORMAT statement, or
                                            the format specification itself.

                                            The formatspec argument is
                                            required for a formatted read,
                                            and must not appear in an
                                            unformatted read. If a READ
                                            statement omits the UNIT=, END=,
                                            ERR=, and REC= options, and
                                            specifies only a formatspec and
                                            iolist, that statement reads from
                                            the asterisk unit (the keyboard).

 «nmlspec»                                  «A namelist group specifier. If
                                            given, iolist must be omitted.
                                            Namelist-directed I/O may be
                                            performed only on files opened
                                            for sequential access.»

 endlabel                                   The label of a statement in the
                                            same program unit.

                                            If endlabel is present, reading
                                            the end-of-file record transfers
                                            control to the executable
                                            statement at endlabel. If
                                            endlabel is omitted, reading past
                                            the end-of-file record causes a
                                            run-time error, unless either
                                            errlabel or iocheck is specified.

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit.

                                            If errlabel is specified, an I/O
                                            error causes transfer of control
                                            to the statement at errlabel. If
                                            errlabel is omitted, the effect
                                            of an I/O error is determined by
                                            the presence or absence of
                                            iocheck.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            there is no error, -1 if
                                            end-of-file is encountered, or
                                            the number of the error message
                                            if an error occurs. For more
                                            information on error handling,
                                            see Section 3.2.6, "Error and
                                            End-of-File Handling."

 rec                                        An integer expression that
                                            evaluates to a positive number,
                                            called the record number. It is
                                            specified for direct-access files
                                            only. If rec is specified for a
                                            sequential-access or an internal
                                            file, an I/O error occurs.

                                            The file is positioned to record
                                            number rec before data is read
                                            from it. (The first record in the
                                            file is record number 1.) The
                                            default for rec is the current
                                            position in the file.

 iolist                                     A list of entities into which
                                            values are transferred from the
                                            file. An iolist may not contain
                                            structure variables, but it may
                                            contain structure elements.


 ■  Remarks

 «If the file were opened with MODE='READWRITE' (the default), you could read
 and write to the same file without having to reopen it to reset the access
 mode.»

 If you attempt to read an unwritten record that has not been written to from
 a direct file, the results are undefined.

 If a parameter of the READ statement is an expression that calls a function,
 that function must not execute an I/O statement «or the EOF intrinsic
 function», because the results are unpredictable.


 ■  Example

      C     Define a two-dimensional array:
            DIMENSION ia(10,20)
      C     Read in the bounds for the array. These bounds should
      C     be less than or equal to 10 and 20, respectively.
      C     Then read in the array in nested implied-DO lists
      C     with input format of 8 columns of width 5 each.
            READ   (3, 990) il, jl, ((ia(i,j), j = 1, jl), i =1, il)
        990 FORMAT (2I5, /, (8I5))


 REAL


 ■  Action

 Specifies the REAL type for user-defined names


 ■  Syntax

      REAL««*bytes»»
      vname««[attrs]»«*length»»«(dim)»««/values/»»
      «, vname««[attrs]»«*length»»«(dim)»««/values/»» »...

 The order of the length and dim parameters can be reversed.

 Parameter                                  Description

 «bytes»                                    «Must be 4 or 8. Specifies the
                                            length, in bytes, of the names in
                                            the REAL statement. This value
                                            can be overridden by the length
                                            parameter.»

 vname                                      The symbolic name of a constant,
                                            variable, array, external
                                            function, statement function,
                                            intrinsic function, FUNCTION
                                            subprogram or an array
                                            declarator. The vname parameter
                                            cannot be the name of a
                                            subroutine or main program.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            vname. These attributes can be
                                            used with vname: ALIAS,
                                            ALLOCATABLE, C, EXTERN, FAR,
                                            HUGE, NEAR, PASCAL, REFERENCE,
                                            VALUE.»

 «length»                                   «Must be 4 or 8. Gives vname the
                                            specified length. The length
                                            parameter overrides the length
                                            specified by bytes.»

 dim                                        A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.

 «values»                                   «A list of constants and repeated
                                            constants, separated by commas. A
                                            repeated constant takes the form
                                            n*constant, where n is a positive
                                            integer constant. The /values/
                                            option, if specified, initializes
                                            vname. The following statement
                                            declares that numis of type REAL,
                                            and sets num  equal to 10.0:»

                                             «REAL num /10.0/»


 ■  Remarks

 A REAL statement confirms or overrides the implicit type of vname. The name
 vname is defined for the entire program unit, and it cannot be defined by
 any other type statement in that program unit.

 «If both the bytes and length parameters are omitted, the default length is
 four bytes.»

 REAL statements must precede all executable statements.

 «REAL*8» and DOUBLE PRECISION are the same data type.


 ■  Examples

            REAL goof, abs
           «REAL*4 wx1, wx3*8, wx5, wx6*8»


 RECORD


 ■  «Action»

 «Specifies a user-defined structure type for user-defined names»


 ■  «Syntax»

      «RECORD /type_name/
      vname « [attrs] » «(dim)»
      «, vname « [attrs] » «(dim)» »...

 «Parameter                                 Description»

 «type_name»                                «The name of user-defined
                                            structure type.»

 «vname»                                    «A variable, array, or array
                                            declarator, which will be of this
                                            structure type.»

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describe
                                            vname.»

 «dim»                                      «A dimension declarator.
                                            Specifying dim declares vname as
                                            an array.»


 ■  «Remarks»

 «The STRUCTURE statement defines a new data type. The RECORD statement
 assigns this new data type to a variable. The type-name must be defined
 prior to its appearance in a RECORD statement.»

 «RECORD statements must precede all executable statements.»


 ■  «Example»

           «STRUCTURE /address/
              LOGICAL*2    house_or_apt
              INTEGER*2    apt
              INTEGER*2    housenumber
              CHARACTER*30 street
              CHARACTER*20 city
              CHARACTER*2  state
              INTEGER*4    zip
            END STRUCTURE

            RECORD /address/ mailing_addr[NEAR](20), shipping_addr(20)»


 ■  «See also»

 «STRUCTURE»


 RETURN


 ■  Action

 Returns control to the calling program unit


 ■  Syntax

      RETURN «ordinal»

 Parameter                                  Description

 ordinal                                    An integer constant. Specifies
                                            the position of an
                                            alternate-return label in the
                                            formal argument list for the
                                            subroutine. See the CALL entry in
                                            this section for information on
                                            alternate returns.


 ■  Remarks

 The RETURN statement terminates execution of the enclosing subroutine or
 function. If the RETURN statement is in a function, the function's value is
 equal to the current value of the function return variable. The END
 statement in a function or subroutine has the same effect as a RETURN
 statement.

 If the actual arguments of the CALL statement contain alternate-return
 specifiers, the  RETURN statement can return control to a specific statement
 via ordinal. The RETURN ordinal statement must appear only in a subroutine;
 it may not appear in a function.

 A RETURN statement in the main program is treated as a STOP statement with
 no  message parameter.


 ■  Examples

      C     This subroutine loops until you type "Y":
            SUBROUTINE Loop
            CHARACTER in

         10 READ (*, '(A)') in
            IF (in .EQ. 'Y') RETURN
            GOTO 10
      C     RETURN implied by the following statement:
            END

 The following example demonstrates the alternate-return feature:

         01 CALL AltRet (i, *10, j, *20, *30)
            WRITE (*, *) 'normal return'
            GOTO 40
         10 WRITE (*, *) 'I = 10'
            GOTO 40
         20 WRITE (*, *) 'I = 20'
            GOTO 40
         30 WRITE (*, *) 'I = 30'
         40 CONTINUE


            .
            .
            .
            SUBROUTINE AltRet (i, *, j, *, *)
            IF (i .EQ. 10)  RETURN 1
            IF (i .EQ. 20)  RETURN 2
            IF (i .EQ. 30)  RETURN 3
            RETURN

 In this example, RETURN 1 specifies the list's first alternate-return label,
 which is a symbol for the actual argument *10 in the CALL statement. RETURN
 2 specifies the second alternate-return label, and RETURN 3 specifies the
 third alternate-return label.


 REWIND


 ■  Action

 Repositions a file to its first record


 ■  Syntax

      REWIND { unitspec |
      («UNIT=»unitspec
      «, ERR=errlabel»
      «, IOSTAT=iocheck»)}

 If UNIT= is omitted, unitspec must be the first parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   An integer expression that
                                            specifies an external unit. If
                                            unitspec has not been opened,
                                            REWIND has no effect.

 errlabel                                   The label of an executable
                                            statement. If errlabel is
                                            specified, an I/O error causes
                                            transfer of control to the
                                            statement at  errlabel. If
                                            errlabel is omitted, the effect
                                            of an I/O error is determined by
                                            the presence or absence of
                                            iocheck.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            there is no I/O error, or the
                                            number of the error message if an
                                            error is encountered. See Section
                                            3.2.6, "Error and End-of-File
                                            Handling."


 ■  Remarks

 If a parameter of the REWIND statement is an expression that calls a
 function, that function must not execute an I/O statement «or the EOF
 intrinsic function», because the results are unpredictable.


 ■  Example

            INTEGER int(80)
            .
            .
            .
            WRITE (7, '(80I1)') int
            .
            .
            .
            REWIND 7
            .
            .
            .
            READ  (7, '(80I1)') int


 SAVE


 ■  Action

 Causes variables to retain their values between invocations of the procedure
 in which they are defined


 ■  Syntax

      SAVE «names»

 Parameter                                  Description

 names                                      One or more names of common
                                            blocks (enclosed in slashes),
                                            variables, or arrays. If more
                                            than one name is specified, they
                                            must be separated by commas. If
                                            no names are specified, all
                                            allowable items in the program
                                            unit are saved.


 ■  Remarks

 The SAVE statement does not allow the following:

   ■  Repeated appearances of an item

   ■  Formal argument names

   ■  Procedure names

   ■  Names of entities in a common block

   ■  «Names that appear in an AUTOMATIC statement»

 Within a program, a common block that is saved in one subprogram must be
 saved in every subprogram containing that common block.

 ───────────────────────────────────────────────────────────────────────────
 «NOTE»
    «Microsoft FORTRAN saves all variables by default. The SAVE statement
    simplifies porting code, and can specify variables that are not to be
    made automatic.»
 ───────────────────────────────────────────────────────────────────────────


 ■  Example

            SAVE /mycom/, myvar


 ■  See also

 «AUTOMATIC»


 SELECT CASE...END SELECT


 ■  «Action»

 «Transfers program control to a selected block of statements, determined by
 the value of a controlling expression»


 ■  «Syntax»

      «SELECT CASE (testexpr)

        CASE (expressionlist)
            «statementblock»

        CASE (expressionlist)
            «statementblock»
           .
           .
           .
        «CASE DEFAULT
        «statementblock»»

        END SELECT»

 «Parameter                                 Description»

 «testexpr»                                 «An INTEGER, LOGICAL, or
                                            CHARACTER*1 expression.»

 «expressionlist»                           «A list of values, which must be
                                            constant and must match the data
                                            type of testexpr. The values must
                                            be of type INTEGER, LOGICAL, or
                                            CHARACTER*1. If testexpr matches
                                            one of the values, the following
                                            statementblock is executed.»

 «statementblock»                           «One or more executable
                                            statements.»


 ■  «Remarks»

 «There are two ways to include values in the expressionlist. The first is to
 give a list of individual values, separated by commas. The second is to
 specify an inclusive range of values, separated by a colon, such as 5:10 or
 'I':'N'. The lower bound must be less than the upper bound. If the values
 are characters, the first character must appear before the second in the
 ASCII collating sequence.»

 «If the lower bound is omitted (such as :10), then all values less than or
 equal to the upper bound match. If the upper bound is omitted (such as 5:),
 then all values greater than or equal to the lower bound must match. Both
 individual values and ranges of values may be included in the same
 expressionlist. You cannot specify a range of values when»  «testexpr is of
 type LOGICAL. A value (even when specified implicitly as part of a colon
 range) can only appear in one expressionlist.»

 «A statementblock need not contain executable statements. Empty blocks can
 be used to make it clear that no action is to be taken for a particular set
 of expression values.»

 «The CASE DEFAULT statement is optional. You can include only one CASE
 DEFAULT statement in a SELECT CASE block.»

 «If the value of testexpr does not match any value in any expressionlist,
 execution passes beyond the SELECT CASE construct to the next executable
 statement.»

 «SELECT CASE blocks may be nested. Each block must have its own END SELECT
 statement; you may not terminate more than one block with a single END
 SELECT statement.»

 «If a SELECT CASE statement appears within a DO or DO WHILE loop, or within
 an IF, ELSE, or ELSE IF structure, the terminating END SELECT statement must
 appear within the range of that loop or structure.»

 «It is illegal to branch into a SELECT CASE block from outside, or to branch
 from one CASE section to another. Any attempt to do so causes a compile-time
 error.»


 ■  «Example»

     «CHARACTER*1  cmdchar
      .
      .
      .
      SELECT CASE (cmdchar)
        CASE ('0')
           WRITE (*, *) "Must retrieve one to nine files"
        CASE ('1':'9')
           CALL RetrieveNumFiles (cmdchar)
        CASE ('A', 'd')
             CALL AddEntry
        CASE ('D', 'd')
           CALL DeleteEntry
        CASE ('H', 'h')
           CALL Help
        CASE ('R':'T', 'r':'t')
           WRITE (*, *) "REDUCE, SPREAD and TRANSFER commands ",
          +    "not yet supported"
        CASE DEFAULT
           WRITE (*, *) "Command not recognized; please
          +re-enter"
      END SELECT»


 Statement Function


 ■  Action

 Defines a function in a single statement


 ■  Syntax

      fname («formals»)=expression

 Parameter                                  Description

 fname                                      The name of the statement
                                            function.

                                            The name fname is local to the
                                            enclosing program unit and must
                                            not be used otherwise, except as
                                            the name of a common block or as
                                            the name of a formal argument to
                                            another statement function. If
                                            fname is used as a formal
                                            argument to another statement
                                            function, fname must have the
                                            same data type every time it is
                                            used.

 formals                                    A list of formal arguments. If
                                            there is more than one name, the
                                            names must be separated by
                                            commas.

                                            The scope of formal-argument
                                            names is the statement function.
                                            Therefore, formal-argument names
                                            can be reused as other
                                            user-defined names in the rest of
                                            the program unit enclosing the
                                            statement-function definition.

                                            If formal is the same as another
                                            local name, a reference to formal
                                            within the statement function
                                            always refers to the formal
                                            argument, never to the other
                                            local name. The data type of
                                            formal is determined by the data
                                            type of the local variable,
                                            because the compiler has no other
                                            way to establish a type for it.

 expression                                 Any legal expression.

                                            References to variables, formal
                                            arguments, other functions, array
                                            elements, and constants are
                                            permitted within the  expression.
                                            Statement-function references,
                                            however, must refer to statement
                                            functions defined prior to the
                                            statement function in which they
                                            appear. Statement functions
                                            cannot be called recursively,
                                            either directly or indirectly.


 ■  Remarks

 Like a regular function, a statement function is executed by referring to it
 in an expression.

 The type of expression and the type of fname must be compatible in the same
 way that  expression and variable must be compatible in assignment
 statements. The expression is converted to the same data type as fname. The
 actual arguments to the statement function are converted to the same type as
 the formal arguments.   A statement function can only be referenced in the
 program unit in which it is defined. The name of a statement function cannot
 appear in any specification statement, except in a type statement (which may
 not be an array definition) and in a COMMON statement (as the name of a
 common block).


 ■  Example

            DIMENSION x(10)
            Add (a, b) = a + b

            DO 100, n = 1, 10
            x(n) = Add (y, z)
        100 CONTINUE


 STOP


 ■  Action

 Terminates program execution


 ■  Syntax

      STOP «message»

 Parameter                                  Description

 message                                    A character constant or an
                                            integer from 0 to 99,999


 ■  Remarks

 If no message is specified, the following default message is displayed:

      STOP - Program terminated.

 If message is a character constant, it is displayed, and the program returns
 zero to the operating system, for use by program units that retrieve status
 information.

 If message is a number:

   1. The words Return code, followed by the number, are displayed. For
      example, the statement STOP 0400 produces the output  Return code 0400.

   2. The program returns the least-significant byte of that integer value (a
      value from 0 to 255) to the operating system, for use by program units
      that wish to check status.


 ■  Example

            IF (ierror .EQ. 0) GOTO 200
            STOP 'ERROR DETECTED!'
        200 CONTINUE


 «STRUCTURE...END STRUCTURE»


 ■  «Action»

 «Defines a new, compound variable type constructed from other variable
 types»


 ■  «Syntax»

      «STRUCTURE /type_name/»
          element_declaration(s)
          .
          .
          .
      END STRUCTURE»

 «Parameter                                 Description»

 «type_name»                                «The name for the new data type;
                                            follows standard naming
                                            conventions. It may not duplicate
                                            the name of another variable or
                                            an intrinsic function, and it may
                                            not be the name of an intrinsic
                                            data type, such as COMPLEX. It is
                                            a local name.»

 «element_declarations»                     «Any combination of one or more
                                            variable-typing statements and
                                            UNION constructs. Can include
                                            RECORD statements that use
                                            previously defined structure
                                            types.»


 ■  «Remarks»

 «The STRUCTURE statement defines a new variable type, called a "structure";
 it is not a declaration of a specific program variable. To declare a
 variable of a particular structure type, use the RECORD statement.»

 «A structure is made of elements.  The simplest element is a conventional
 FORTRAN variable type with a dummy name, as in the following example:»

           «STRUCTURE /full_name/»
              «CHARACTER*15 first_name»
              «CHARACTER*20 last_name»
           «END STRUCTURE»

 «Elements cannot be declared with attributes. For example, INTEGER var[FAR]
 is not permitted.»

 «An element can also be a RECORD statement that references a previously
 defined structure type:»

          «STRUCTURE /full_address/
              RECORD /full_name/ personsname
              RECORD /address/   ship_to
              INTEGER*1          age
              INTEGER*4          phone
           END STRUCTURE»

 «An element can also be a union of several variable maps.»

 «Element names are local to the structure in which they appear. There is no
 conflict if the same element name appears in more than one structure. Nested
 structures may have elements with the same names.»

 «A particular element is specified by listing the sequence of elements
 required to reach the desired element, separated by a period (.). Suppose a
 structure variable, shippingaddress, were declared with the full_address
 structure declared in the previous example:»

            «RECORD /full_address/ shippingaddress»

 «The age element would then be specified by shippingaddress.age , the first
 name of the receiver by shippingaddress.personsname.first_name, and so on.
 An element is no different from any other variable having the same type.
 When a structure element appears in an expression, its type is the type of
 the element. When passed as an actual argument, it must match the formal
 argument in type, order, and dimension.»

 «A structure can be no longer than 65,536 bytes. This includes all the data,
 plus any padding bytes. The way structures are packed in memory is
 controlled by the $PACK metacommand and /Zp command-line option.»

 «Structures may look identical, but still be different. For two structures
 to be identical, they must have the same component types and names in the
 same sequence, and the packing must be the same.»


 ■  «See also»

 «MAP...END MAP»
 «RECORD»
 «UNION...END UNION»
 «$PACK»


 SUBROUTINE


 ■  Action

 Identifies a program unit as a subroutine, gives it a name, and identifies
 its formal arguments


 ■  Syntax

      SUBROUTINE subr ««sattrs]»»
      «(«formal ««[attrs]»»
      «,formal ««[attrs]»» »...»)»

 Parameter                                  Description

 subr                                       The global, external name of the
                                            subroutine. The name  subr cannot
                                            appear in «AUTOMATIC,» COMMON,
                                            DATA, EQUIVALENCE, INTRINSIC,
                                            «LOADDS,» SAVE, or type
                                            statements.

 «sattrs»                                   «A list of attributes, separated
                                            by commas. The sattrs describes
                                            subr. The following attributes
                                            can be used with subr: ALIAS, C,
                                            FAR, NEAR, PASCAL, VARYING.»

 formal                                     A formal-argument name. If more
                                            than one is specified, they must
                                            be separated by commas. A formal
                                            argument may be the name of a
                                            conventional variable, a
                                            structure variable, or an
                                            intrinsic function. A formal
                                            argument may also be an alternate
                                            return label (*).

                                            For an explanation of alternate
                                            return specifiers, see the CALL
                                            entry in this section.

 «attrs»                                    «A list of attributes, separated
                                            by commas. The attrs describes
                                            formal (except alternate-return
                                            specifiers). The following
                                            attributes can be used with
                                            formal: FAR, HUGE, NEAR,
                                            REFERENCE, VALUE.»


 ■  Remarks

 A subroutine begins with a SUBROUTINE statement and ends with the next END
 statement. A subroutine can contain any statement except a BLOCK DATA,
 FUNCTION, « INTERFACE TO,» PROGRAM, or SUBROUTINE statement.

 Within the calling program, subr is global, and may not be used for any
 other variable or subprogram.

 The formal-argument list sets the number of arguments for that subroutine.
 The argument types are set by any IMPLICIT, EXTERNAL, DIMENSION, or type
 statements within the subroutine itself (or they default to implicit FORTRAN
 types). The argument types are not set by the calling program, even if the
 subroutine appears in the calling program's source file. Formal-argument
 names cannot appear in «AUTOMATIC,» COMMON, DATA, EQUIVALENCE, INTRINSIC, or
 SAVE statements.

 In a CALL statement, the actual arguments that are passed must agree with
 corresponding formal arguments in the SUBROUTINE statement in order, in
 number «(except when the C and VARYING attributes are specified)», and in
 type or kind. The compiler checks for correspondence if the formal arguments
 are known.

 A "recursive" subroutine is one that calls itself, or calls another
 subprogram which in turn calls the first subroutine before the first
 subroutine has completed execution. FORTRAN does not support recursive
 subroutine calls. «When using Microsoft FORTRAN, any attempt at direct
 recursion results in a compile-time error. Indirect recursion, however, is
 not detected. The results of such indirect recursion are undefined and
 unpredictable.»


 ■  Example

            SUBROUTINE GetNum (num, unit)
            INTEGER num, unit
         10 READ (unit, '(I10)', ERR = 10) num
            END


 Type

 See the individual listings in this chapter for the CHARACTER, COMPLEX,
 «DOUBLE COMPLEX,» DOUBLE PRECISION, INTEGER, LOGICAL, REAL, «RECORD, and
 STRUCTURE...END STRUCTURE» statements.


 UNION...END UNION


 ■  «Action»

 «Causes two or more maps to occupy the same memory location»


 ■  «Syntax»

     «UNION»
      map-statement
      map-statement
      «map-statement»...
      END UNION»

 «Parameter                                 Description»

 «map-statement»                            «A map declaration. See the
                                            MAP..END MAP entry in this
                                            section for more information.»


 ■  «Remarks»

 «A UNION block may only appear within a STRUCTURE block. Each UNION block
 must be terminated with its own END UNION statement; a single END UNION
 statement may not be used to terminate more than one UNION block.»

 «UNION is similar to the EQUIVALENCE statement; both allocate the same
 memory area to more than one variable. However, maps can contain multiple,
 contiguous variables, which gives increased flexibility in assigning
 variables to the same memory location.»


 ■  «Example»

 «Note how the first 40 characters in the string2 array are assigned to
 four-byte integers, while the remaining 20 are assigned to two-byte
 integers:»

           «UNION»
              «MAP»
                 «CHARACTER*20 string1, CHARACTER*10 string2(6)»
              «END MAP»
              «MAP»
                 «INTEGER*2 number(10), INTEGER*4 var(10), INTEGER*2»
         «+       datum(10)»
              «END MAP»
           «END UNION»

 «The $PACK metacommand and /Zp command-line option control how variables in
 structures are assigned to beginning byte addresses. The particular packing
 option chosen may affect how particular variables are assigned to the same
 memory location. See the $PACK entry in Section 6.2, "Metacommand
 Directory," for more information.


 ■  «See also»

 «MAP...END MAP»
 «STRUCTURE...END STRUCTURE»
 «$PACK»


 WRITE


 ■  Action

 Transfers data from the iolist items to the file associated with unitspec


 ■  Syntax

      WRITE («UNIT=»unitspec
      «, ««{»«FMT=»formatspec»«|«NML=»nmlspec }»»
      «, ERR=errlabel»
      «, IOSTAT=iocheck»
      «, REC=rec»)
      iolist

 If UNIT= is omitted, unitspec must be the first parameter. If FMT= or NML=
 is omitted, fmtspec or nmlspec must be the second parameter. The parameters
 can otherwise appear in any order.

 Parameter                                  Description

 unitspec                                   When writing to an internal file,
                                            unitspec must be a character
                                            substring, variable, array, array
                                            element, «structure element, or
                                            noncharacter array». When writing
                                            to an external file, unitspec is
                                            an integer expression that
                                            specifies a unit.

                                            «If unitspec has not been
                                            explicitly associated with a
                                            file, an "implicit open," using
                                            the following syntax, is
                                            performed:»

                                            «OPEN (unitspec, FILE = ' ',»
                                            «STATUS = 'UNKNOWN',» «ACCESS =
                                            'SEQUENTIAL',» «FORM = form)»

                                            «where form is 'FORMATTED' for a
                                            formatted write operation and
                                            'UNFORMATTED' for an unformatted
                                            write operation.»

 formatspec                                 A format specifier. A format
                                            specifier is required for a
                                            formatted write; it must not
                                            appear in an unformatted write.

 «nmlspec»                                  «A namelist specifier. If
                                            specified, iolist must be
                                            omitted. Namelist-directed I/O
                                            may be performed only on files
                                            opened for sequential access.»

 errlabel                                   The label of an executable
                                            statement in the same program
                                            unit.

                                            If errlabel is specified, an I/O
                                            error causes transfer of control
                                            to the statement at errlabel. If
                                            errlabel is omitted, the effect
                                            of an I/O error is determined by
                                            the presence or absence of
                                            iocheck.

 iocheck                                    An integer variable, array
                                            element, «or structure element»
                                            that returns a value of zero if
                                            there is no error, or the number
                                            of the error message if an error
                                            occurs. See Section 3.2.6, "Error
                                            and End-of-File Handling."

 rec                                        A positive integer expression,
                                            called a record number, specified
                                            only for direct-access files
                                            (otherwise, an error results).

                                            The argument rec specifies the
                                            number of the record to be
                                            written. The first record in the
                                            file is record number 1. The
                                            default for rec is the current
                                            position in the file.

 iolist                                     A list of entities whose values
                                            are transferred by the WRITE
                                            statement. It may not contain
                                            structure variables, though
                                            structure elements are permitted.


 ■  Remarks

 «If the file were opened with MODE='READWRITE' (the default), you could
 alternately read and write to the same file without reopening it each time.»

 If you write to a sequential file, you delete any records that existed
 beyond the newly written record. Note that for sequential files, you are
 always effectively at the end of the file following a write operation, and
 you must backspace or rewind before the next read operation.

 If a parameter of the WRITE statement is an expression that calls a
 function, that function must not execute an I/O statement «or the EOF
 intrinsic function», because the results are unpredictable.


 ■  Example

      C     Generate a table of square and cube roots
      C     of the whole numbers from 1-100
            WRITE (*, 10) (n, SQRT(FLOAT(n)), FLOAT(n)**(1.0/3.0),
            +n = 1, 100)
            10 FORMAT (I5, F8.4, F8.5)



 Chapter 5  Intrinsic Functions and Additional Procedures

 ───────────────────────────────────────────────────────────────────────────

 Intrinsic functions are predefined by the Microsoft FORTRAN language. These
 functions carry out data-type conversions and return information about data
 types, perform operations on both numeric and character data, test for
 end-of-file, return addresses, and perform bit manipulation.

 The first part of this chapter is a detailed description of the intrinsic
 functions available in Microsoft FORTRAN. The second part of this chapter is
 an alphabetical tabular listing of all intrinsic functions. «The third part
 of this chapter describes other functions and subroutines supplied with
 Microsoft FORTRAN.»


 5.1  Using Intrinsic Functions

 An "intrinsic function" is a function that is part of the FORTRAN language.
 The compiler recognizes its name and knows that the object code for the
 function is in the language libraries. Intrinsic functions are automatically
 linked to the program without any additional effort on the programmer's
 part. (Functions written by the user are called "external" functions. If you
 wish to write your own function that has the same name as an intrinsic
 function, you must declare its name in an EXTERNAL statement.)

 Each function returns a value of integer, real, logical, or character, and
 therefore has a corresponding data type. For example, the ICHAR intrinsic
 function returns the ASCII value of a character string; it is therefore an
 integer function. Some intrinsic functions, such as INT, can take arguments
 of more than one type. Others, such as ABS, can return a value that has the
 same type as the argument.

 An IMPLICIT statement cannot change the type of an intrinsic function. For
 those generic intrinsic functions that allow several types of arguments, all
 arguments in a function call should be of the same type. The intrinsic
 function DIM, for example, takes two integer arguments and returns the
 positive difference. If you specify DIM (i,j), i and j must be of the same
 type.

 «If, however, two arguments are of different types, Microsoft FORTRAN first
 attempts to convert the arguments to the correct data type. For example, if
 i and j in the example above are real numbers, they are first converted to
 INTEGER type (see Table 1.6, "Arithmetic Type Conversion," for information
 on type conversion), and then the operation is performed. Or, if i  is of
 type INTEGER*2 and j is of type INTEGER*4, i is first converted to
 INTEGER*4, and then the operation is performed.»

 Intrinsic-function names, except those listed in Section 2.5, "Arguments,"
 can appear in an INTRINSIC statement. Intrinsic functions specified in
 INTRINSIC statements can be used as actual arguments in external procedure
 references. An intrinsic-function name can also appear in a type statement,
 but only if the type is the same as the standard type for that intrinsic
 function.

 Generic intrinsic functions, like ABS, let you use the same
 intrinsic-function name with more than one type of argument. Specific
 intrinsic functions, like IFIX, can only be used with one type of argument.

 An external procedure cannot have the same name as an intrinsic function,
 unless the EXTERNAL statement is used to tell the compiler the substitution
 has been made. (As a result, the intrinsic function of that name is no
 longer usable in the program unit where the EXTERNAL statement associated it
 with an external  procedure.) In the following statement, for example, sign,
 float, and index are the names of external procedures:

             EXTERNAL Sign, Float, Index
             SUBROUTINE Process (Sign, Float, Index)

 If you supply an argument that has no mathematically defined result or for
 which the result exceeds the numeric range of the processor, the result of
 the intrinsic function is undefined.

 Arguments must agree in order, number, and type with those specified in
 Tables 5.2-5.17. Arguments can also be expressions of the specified type.

 When logarithmic and trigonometric intrinsic functions act on a complex
 argument, they return the "principal value." The principal value of a
 complex number is the number whose argument (angle in radians) is less than
 or equal to p and greater than -p.

 «When the results of generic integer intrinsic functions are passed to
 subprograms, the $STORAGE setting determines the data type of the value to
 be passed.»

 ───────────────────────────────────────────────────────────────────────────
 «WARNING»
    «Microsoft FORTRAN contains several intrinsic functions whose names are
    longer than six characters (ALLOCATED, EPSILON, LEN_TRIM, LOCNEAR,
    MAXEXPONENT, MINEXPONENT, NEAREST, PRECISION). If the $TRUNCATE
    metacommand is enabled, the compiler considers any function name with the
    same first six characters to be one of these intrinsic functions.»
 ───────────────────────────────────────────────────────────────────────────

 Table 5.1 lists the abbreviations used in the tables of intrinsic functions
 in this chapter.

 Table 5.1  Abbreviations Used to Describe Intrinsic Functions

 Abbreviation      Data Type
 char              CHARACTER «*n»
 cmp               COMPLEX, «COMPLEX*8, DOUBLE COMPLEX, or COMPLEX*16»
 cmp8              COMPLEX«*8»
 cmp16             «DOUBLE COMPLEX or COMPLEX*16»
 dbl               DOUBLE PRECISION,«REAL*8»
 gen               More than one possible argument type; see "Argument Type"
                   column
 integer           INTEGER, «INTEGER*1, INTEGER*2, or INTEGER*4»
 «int1             INTEGER*1»
 «int2             INTEGER*2»
 int4              INTEGER«*4»
 log               LOGICAL, «LOGICAL*1, LOGICAL*2, or LOGICAL*4»
 log1              «LOGICAL*1»
 log2              «LOGICAL*2»
 log4              LOGICAL«*4»
 real              REAL, «REAL*4,» DOUBLE PRECISION«, or REAL*8»
 real4             REAL«*4»


 5.1.1  Data-Type Conversion

 This section describes the type-conversion intrinsic functions. Table 5.2
 summarizes the intrinsic functions that perform type conversion.

 Table 5.2  Intrinsic Functions: Type Conversion

                   Argument                   Function
 Name              Type                       Type

 INT (gen)         int, real, or cmp          int

 «INT1 (gen)»      «int, real, or cmp»        «INTEGER*1»

 «INT2 (gen)»      «int, real, or cmp»        «INTEGER*2»

 «INT4 (gen)»      «int, real, or cmp»        «INTEGER*4»

 «INTC (gen)»      «int, real, or cmp»        «INTEGER[C]»

 IFIX (real4)      REAL«*4»                   int

 «HFIX (gen)»      «int, real, or cmp»        «INTEGER*2»

 «JFIX (gen)»      «int, real, or cmp»        «INTEGER*4»

 IDINT (dbl)       DOUBLE PRECISION,          int
                   «REAL*8»

 REAL (gen)        int, real, or cmp          REAL«*4»

 «DREAL (cmp)»     «COMPLEX*16»               «REAL*8»

 FLOAT (int)       int                        REAL«*4»

 SNGL (dbl)        «REAL*8»                   REAL«*4»

 DBLE (gen)        int, real, or cmp          DOUBLE PRECISION

 DFLOAT (gen)      int, real, or cmp          DOUBLE PRECISION

 CMPLX (genA       int, real, or cmp          COMPLEX«*8»
 «, genB»)

 «DCMPLX (genA     «int, real, or cmp»        «COMPLEX*16»
 «, genB»)»

 ICHAR (char)      char                       int

 CHAR (int)        int                        char


 The INT intrinsic function converts arguments to integers. If the argument
 gen is an integer, then INT (gen) equals gen. If gen is real, then INT (gen)
 is the truncated value of gen. INT (1.9), for example, is equal to 1, and
 INT (-1.9) is equal to -1. If gen is complex, the real part of gen is taken,
 then truncated to an integer.

 «INT1 converts its arguments to INTEGER*1. INT2 and HFIX convert their
 arguments to INTEGER*2. INT4 and JFIX convert their arguments to INTEGER*4.
 They can be used to convert the data type of an expression or variable to an
 expression of the correct type for passing as an argument. INT2 can also be
 used to direct the compiler to use short arithmetic in expressions which
 would otherwise be long, and INT4 can specify long arithmetic in expressions
 which would otherwise be short.»

 «The INTC intrinsic function converts arguments to C integers. C integers
 are described in Section 1.6.3, "C."» The IFIX and IDINT intrinsic functions
 convert single- or double-precision arguments, respectively, to integers.

 The REAL intrinsic function converts numbers to the single-precision real
 data type. If gen is an integer, REAL(gen) is gen stored as a
 single-precision real number. If gen is a single-precision real number,
 REAL(gen) equals gen. If gen is complex, REAL(gen) equals the real part of
 gen. If gen is a double-precision number, REAL(gen) is the first six
 significant digits of gen.

 The DBLE intrinsic function converts numbers to the double-precision real
 data type. The FLOAT and SNGL intrinsic functions convert numbers to the
 single-precision real data type. They work like the REAL intrinsic function.
 «The DREAL intrinsic function converts COMPLEX*16 numbers to the
 double-precision real data type by deleting the imaginary part.»

 The CMPLX «and DCMPLX» intrinsic functions convert numbers to the complex
 data types. If only one argument, gen, is specified, gen can be an integer,
 a real, a double-precision, or a complex number. If gen is complex,
 CMPLX(gen) equals gen. If gen is an integer, a real, or a double-precision
 number, the real part of the result equals REAL(gen) and the imaginary part
 equals 0.0. If two arguments are specified, genA and genB must both have the
 same type, and they can be integers, real numbers, or double-precision
 numbers. In this case, the real part of the result equals REAL(genA), and
 the imaginary part of the result equals REAL(genB).

 The ICHAR intrinsic function translates ASCII characters into integers, and
 the CHAR intrinsic function translates integers into ASCII characters. (The
 ASCII character set is listed in Appendix A, "ASCII Character Codes.") Both
 the argument of the CHAR intrinsic function and the result of the ICHAR
 intrinsic function must be greater than or equal to 0, and less than or
 equal to 255. The argument of the ICHAR intrinsic function must be a single
 character.


 Example

 The following list shows examples of the type-conversion intrinsic
 functions:

 Function Reference                         Equivalent

 INT(-3.7)                                  -3

 INT(7.682)                                  7

 INT(0)                                      0

 INT((7.2,39.3))                             7


 5.1.2  «Data-Type Information»

 «This section describes the intrinsic functions that return information
 about data types. Except for the NEAREST function, the variables passed as
 arguments do not need to have been assigned a value; it is the data type of
 the argument that is significant. Table 5.3 summarizes these functions.»

 «Table 5.3  Intrinsic Functions: Data-Type Information»

 «Name                      Argument Type              Function Type»

 «ALLOCATED (gen)»          «array»                    «log»

 «EPSILON (real)»           «real»                     «real»

 «HUGE (gen)»               «int or real»              «Same as argument»

 «MAXEXPONENT (real)»       «real»                     «real»

 «MINEXPONENT (real)»       «real»                     «real»

 «NEAREST (real,            «real»                     «real»
  director)»

 «PRECISION (real)»         «real»                     «real»

 «TINY (real)»              «real»                     «real»


 «The ALLOCATED intrinsic function returns a logical value that is .TRUE. if
 memory is currently allocated to the array.»

 «The EPSILON intrinsic function returns the smallest increment that, when
 added to one, produces a number greater than one for the argument's data
 type. The increment is slightly larger than the best precision attainable
 for that data type. For example, a single-precision real can accurately
 represent some numbers to seven decimal places: EPSILON (real4) returns
 1.112093E-07.»

 «The EPSILON function makes it easy to select a delta for algorithms (such
 as root locators) that search until the calculation is within delta of an
 estimate. If delta is too small (smaller than the decimal resolution of the
 data type), the algorithm may never halt. By scaling the value returned by
 EPSILON to the estimate, you obtain a delta that ensures search
 termination.»

 «The HUGE intrinsic function returns the largest positive number that can be
 represented by the argument's data type.»

 «The MAXEXPONENT intrinsic function returns the largest positive decimal
 exponent that a number of the argument's data type can have. For example,
 MAXEXPONENT (real4) returns 38.»

 «The MINEXPONENT intrinsic function returns the largest negative decimal
 exponent that a number of the argument's data type can have. For example,
 MINEXPONENT (real8) returns -308.»

 «The NEAREST intrinsic function returns the nearest different number in the
 direction of director. (If director is positive, the value returned is
 larger than real. If director is negative, the value returned is smaller
 than real.) As with the EPSILON intrinsic function, NEAREST lets the program
 select an appropriate increment to guarantee a search terminates.»

 «The PRECISION intrinsic function returns the number of significant decimal
 digits for real's data type. It is useful in rounding off numbers.»

 «The TINY intrinsic function returns the smallest positive number that can
 be represented by the argument's data type.»


 5.1.3  Truncating and Rounding

 Table 5.4 summarizes the intrinsic functions that truncate and round.

 Table 5.4  Intrinsic Functions: Truncation and Rounding

                         Truncate          Argument          Function
 Name                    or Round          Type              Type

 AINT (real)             Truncate          real              Same as argument

 DINT (dbl)              Truncate          «REAL*8»          «REAL*8»

 ANINT (real)            Round             real              Same as argument

 DNINT (dbl)             Round             «REAL*8»          «REAL*8»

 NINT (real)             Round             real              int

 IDNINT (dbl)            Round             «REAL*8»          int


 The intrinsic functions AINT and DINT truncate their arguments. The
 intrinsic functions ANINT, DNINT, NINT, and IDNINT round their arguments and
 are evaluated as follows:

 Argument                                   Result

 Greater than zero                          INT(gen+0.5)

 Equal to zero                              Zero

 Less than zero                             INT(gen-0.5)


 Examples

 The following list shows examples of truncation and rounding:

 Function Reference                         Equivalent

 AINT(2.6)                                   2.0

 AINT(-2.6)                                 -2.0

 ANINT(2.6)                                  3.0

 ANINT(-2.6)                                -3.0

 The following example uses the ANINT intrinsic function to perform rounding:

      C     This program adds tax to a purchase amount
      C     and prints the total:

            REAL  amount, taxrate, tax, total
            taxrate = 0.081
            amount  = 12.99
      C     Calculate tax and round to nearest hundredth:
            tax = ANINT (amount * taxrate * 100.0) / 100.0
            total = amount + tax
            WRITE (*, 100) amount, tax, total
        100 FORMAT (1X, 'AMOUNT', F7.2 /
           +        1X, 'TAX   ', F7.2 /
           +        1X, 'TOTAL ', F7.2)
            END


 5.1.4  Absolute Value and Sign Transfer

 Table 5.5 summarizes the intrinsic functions that compute absolute values
 and perform sign transfers.

 The intrinsic functions ABS, IABS, DABS, CABS, and CDABS return the absolute
 value of their arguments. Note that for a complex number (real, imag), the
 absolute value equals the following:

     √(real^(2) + imag^(2)

 For two arguments (genA, genB), the intrinsic functions SIGN, ISIGN, and
 DSIGN return |genA| if genB is greater than or equal to zero, and -|genA| if
 genB is less than zero.

 Table 5.5  Intrinsic Functions: Absolute Values and Sign Transfer

                                                Argument                   Fun
 Name                Definition                 Type                       Typ

 ABS (gen)           Absolute value             int, real, or cmp          Fun
                                                                           sam
                                                                           typ
                                                                           whe
                                                                           arg
                                                                           com

 IABS (int)          Absolute value             int                        int

 DABS (dbl)          Absolute value             «REAL*8»                   «RE

 CABS (cmp)          Absolute value             cmp                        rea

 «CDABS (cmp16)»     «Absolute value»           «COMPLEX*16»               «RE

 SIGN (genA, genB)   Sign transfer              int or real                Sam
                                                                           arg

 ISIGN (intA, intB)  Sign transfer              int                        int

 DSIGN (dblA, dblB)  Sign transfer              «REAL*8»                   «RE


 Examples

 The following program uses a sign-transfer intrinsic function:

            a = 5.2
            b = -3.1

      C     The following statement transfers the sign of b to a
      C     and assigns the result to c.
            c = SIGN (a, b)

      C     The output is -5.2:
            WRITE (*, *) c
            END

 The following program uses sign-transfer and absolute-value intrinsic
 functions:

      C     This program takes the square root of a vector
      C     magnitude. Since the sign in a vector represents
      C     direction, the square root of a negative value is not
      C     meant to produce complex results. This routine removes
      C     the sign, takes the square root, and then restores
      C     the sign.
            REAL  mag, sgn

            WRITE (*, '(A)') ' ENTER A MAGNITUDE:  '
            READ  (*, '(F10.5)') mag
      C     Store the sign of mag by transferring its sign to 1
      C     and storing the result in sgn:
            sgn = SIGN (1.0, mag)

      C     Calculate the square root of the absolute value
      C     of the magnitude:
            result = SQRT (ABS (mag))

      C     Restore the sign by multiplying the result by -1 or
      C     +1:
            result = result * sgn
            WRITE (*, *) result
            END


 5.1.5  Remainders

 Table 5.6 summarizes the intrinsic functions that return remainders.

 Table 5.6  Intrinsic Functions: Remainders

                            Argument          Function
 Name                       Type              Type

 MOD (genA, genB)           int or real       Same as argument

 AMOD (real4A, real4B)      REAL«*4»          REAL«*4»

 DMOD (dblA, dblB)          «REAL*8»          «REAL*8»


 The intrinsic functions MOD, AMOD, and DMOD return remainders as follows
 (AMOD and DMOD use exactly the same formula as MOD):

      MOD (genA, genB) = genA-(INT (genA / genB) * genB)

 If genB is 0, the result is undefined.


 5.1.6  Positive Differences

 Table 5.7 summarizes the intrinsic functions that return the positive
 difference between two arguments.

 Table 5.7  Intrinsic Functions: Positive Difference

                            Argument                   Function
 Name                       Type                       Type

 DIM (genA, genB)           int or real                Same as argument

 IDIM (intA, intB)          int                        int

 DDIM (dblA, dblB)          «REAL*8»                   «REAL*8»


 The intrinsic functions DIM, IDIM, and DDIM return the positive difference
 of two arguments as follows:

 Situation                                  Result

 genA <= genB                               DIM (genA,  genB) = 0

 genA > genB                                DIM (genA,  genB) = genA-genB


 Examples

 The following list shows examples of positive differences:

 Function Reference                         Equivalent

 DIM(10,5)                                  5

 DIM(5,10)                                  0

 DIM(10,-5)                                 15


 5.1.7  Maximums and Minimums

 Table 5.8 summarizes the functions that return the maximum or minimum of two
 or more values.

 Table 5.8  Intrinsic Functions: Maximums and Minimums

                                                       Argument          Funct
 Name                                Definition        Type              Type

 MAX (genA,genB«,genC»...)           Maximum           int or real       Same
                                                                         argum

 MAX0 (intA,intB«,intC»...)          Maximum           int               int

 AMAX1 (real4A,real4B«,real4C»...)   Maximum           REAL«*4»          REAL«

 AMAX0 (intA,intB«,intC»...)         Maximum           int               REAL«

 MAX1 (real4A,real4B«,real4C»...)    Maximum           REAL«*4»          int

 DMAX1(dblA,dblB«,dblC»...)          Maximum           «REAL*8»          «REAL

 MIN (genA,genB«,genC»...)           Minimum           int or real       Same
                                                                         argum

 MIN0 (intA,intB«,intC»...)          Minimum           int               int

 AMIN1 (real4A,real4B«,real4C»...)   Minimum           REAL«*4»          REAL«

 AMIN0 (intA,intB«,intC»...)         Minimum           int               REAL«

 MIN1 (real4A,real4B«,real4C»...)    Minimum           REAL«*4»          int

 DMIN1 (dblA,dblB«,dblC»...)         Minimum           «REAL*8»          «REAL


 The intrinsic functions MAX, MAX0, AMAX1, and DMAX1 return the maximum value
 in the argument list. The intrinsic functions AMAX0 and MAX1 return the
 maximum and also perform type conversion. Similarly, MIN, MIN0, AMIN1, and
 DMIN1 return minimums, while AMIN0 and MIN1 return the minimum and also
 perform type conversion.


 Examples

 The following list shows examples of maximums and minimums:

 Function Reference                         Equivalent

 MAX(5,6,7,8)                               8

 MAX(-5.,-6.,-7.)                           -5.

 MIN(-5,-6,-7)                              -7

 MIN(.1E12,.1E14,.1E19)                     .1E12

 The following program uses the MIN and MAX intrinsic functions:

      C     This program uses the MAX intrinsic function to find
      C     the maximum and minimum elements in a vector x.

            INTEGER   i
            REAL      x(10), small, large

            DATA x /12.5, 2.7, -6.2, 14.1, -9.1, 17.5, 2.0, -6.3,
           +        2.5, -12.2/



      C     Initialize small and large with arbitrarily large and
      C     small values:
            small =  1e20
            large = -1e2

            DO 100, i = 1,10
               small = MIN (small, x(i))
               large = MAX (large, x(i))
        100 CONTINUE

            WRITE (*, 200) small, large
        200 FORMAT (' The smallest number was ', F6.1/
           +        ' The largest number was  ', F6.1)
            END


 Output

      The smallest number was  -12.2
      The largest number was    17.5


 5.1.8  Double-Precision Products

 Table 5.9 lists the intrinsic function that returns a double-precision
 product.

 Table 5.9  Intrinsic Functions: Double-Precision Product

                                     Argument          Function
 Name                                Type              Type

 DPROD (real4A, real4B)              REAL«*4»          «REAL*8»


 The intrinsic function DPROD returns the double-precision product of two
 single-precision real arguments.

 The following example uses the DPROD intrinsic function:

            REAL  a, b

            a = 3.72382
            b = 2.39265

            WRITE (*, *) a*b, DPROD (a,b)


 The following output is produced:

      8.9097980         8.90979744044290


 5.1.9  Complex Functions

 Table 5.10 lists the intrinsic operations that perform various other
 operations on complex numbers.

 Table 5.10  Intrinsic Functions: Complex Operators

                                                    Argument              Func
 Name                    Definition                 Type                  Type

 AIMAG (cmp8)            Imaginary part of          COMPLEX«*8»           REAL
                         COMPLEX«*8» number

 «IMAG (cmp)»            «Imaginary part of cmp     «cmp»                 «If
                         number»                                          COMP
                                                                          func
                                                                          REAL
                                                                          argu
                                                                          COMP
                                                                          func
                                                                          REAL

 «DIMAG (cmp16)»         «Imaginary part of         «COMPLEX*16»          «REA
                         COMPLEX*16 number»

 CONJG (cmp8)            Conjugate of               COMPLEX«*8»           COMP
                         COMPLEX«*8» number

 «DCONJG (cmp16)»        «Conjugate of              «COMPLEX*16»          «COM
                         COMPLEX*16 number»


 The intrinsic function«s» AIMAG, «IMAG, and DIMAG» return the imaginary part
 of complex numbers. The intrinsic function«s» CONJG «and DCONJG» return the
 complex conjugates of complex numbers. Therefore, for a complex number
 complex equal to (real, imag), AIMAG(complex) equals imag, and
 CONJG(complex) equals (real, - imag). Note that the REAL «and DBLE»
 intrinsic function«s», described in Section 5.1.1, can be used to return the
 real part of COMPLEX«*8» «and COMPLEX*16» numbers, respectively.


 Example

 The following program uses complex intrinsic functions:

      C     This program applies the quadratic formula to a
      C     polynomial and allows for complex results.

            REAL  a, b, c
            COMPLEX  ans1, ans2, desc
            WRITE  (*, 100)

       100  FORMAT (' Enter a, b, and c of the ',
           +        'polynomial ax^2 + bx + c: ' \)
            READ   (*, '(3F10.5)') a, b, c

            desc = CSQRT (CMPLX (b**2 - 4.0*a*c))
            ans1 = (-b + desc) / (2.0 * a)
            ans2 = (-b - desc) / (2.0 * a)

            WRITE  (*, 200)
       200  FORMAT (/ ' The roots are:' /)
            WRITE  (*, 300) REAL(ans1), AIMAG(ans1),
           +                REAL(ans2), AIMAG(ans2)
       300  FORMAT (' X = ', F10.5, '  +', F10.5, 'i')
            END


 5.1.10  Square Roots

 Table 5.11 summarizes the intrinsic functions that return square roots.

 Table 5.11  Intrinsic Functions: Square Roots

                            Argument                   Function
 Name                       Type                       Type

 SQRT (gen)                 real or cmp                Same as argument

 DSQRT (dbl)                «REAL*8»                   «REAL*8»

 CSQRT (cmp8)               COMPLEX«*8»                COMPLEX«*8»

 «CDSQRT (cmp16)»           «COMPLEX*16»               «COMPLEX*16»


 The intrinsic functions SQRT, DSQRT, CSQRT«, and CDSQRT» return the square
 root of their respective arguments. The arguments to SQRT and DSQRT must be
 greater than or equal to zero. For a complex argument, SQRT, CSQRT«, and
 CDSQRT» return a complex number whose magnitude is the square root of the
 magnitude of the argument and whose angle is one-half the angle of the
 argument.


 Example

 The following program uses the SQRT intrinsic function:

      C     This program calculates the length of the hypotenuse
      C     of a right triangle from the lengths of the other two
      C     sides.

            REAL  sidea, sideb, hyp

            sidea = 3.0
            sideb = 4.0

            hyp = SQRT (sidea**2 + sideb**2)

            WRITE (*, 100) hyp
        100 FORMAT (/ ' The hypotenuse is ', F10.3)

            END


 5.1.11  Exponents and Logarithms

 Table 5.12 lists the intrinsic functions that return exponents or
 logarithms.

 Table 5.12  Intrinsic Functions: Exponents and Logarithms

                                                    Argument           Functio
 Name                    Definition                 Type               Type

 EXP (gen)               Exponent                   real or cmp        Same as
                                                                       argumen

 DEXP (dbl)              Exponent                   «REAL*8»           «REAL*8

 CEXP (cmp8)             Exponent                   COMPLEX«*8»        COMPLEX

 «CDEXP (cmp16)»         «Exponent»                 «COMPLEX*16»       «COMPLE

 LOG (gen)               Natural logarithm          real or cmp        Same as
                                                                       argumen

 ALOG (real4)            Natural logarithm          REAL«*4»           REAL«*4

 DLOG (dbl)              Natural logarithm          «REAL*8»           «REAL*8

 CLOG (cmp8)             Natural logarithm          COMPLEX«*8»        COMPLEX

 «CDLOG (cmp16)»         «Natural logarithm»        «COMPLEX*16»       «COMPLE

 LOG10 (real)            Common logarithm           real               Same as
                                                                       argumen

 ALOG10 (real4)          Common logarithm           REAL«*4»           REAL«*4

 DLOG10 (dbl)            Common logarithm           «REAL*8»           «REAL*8


 The intrinsic functions EXP, DEXP, CEXP«, and CDEXP» return e**gen.

 The intrinsic functions LOG, ALOG, DLOG, CLOG«, and CDLOG» return the
 natural logarithm of their respective arguments. LOG10, ALOG10, and DLOG10
 return the base-10 logarithm of their arguments.

 For all intrinsic logarithmic functions, real arguments must be greater than
 zero. The CLOG and CDLOG functions return the logarithm of a complex number.
 The real component is the natural logarithm of the magnitude of the complex
 number. The imaginary component is the principal value of the angle of the
 complex number, in radians.


 Example

 The following program uses the EXP intrinsic function:

      C     Given the initial size and growth rate of a colony,
      C     this program computes the size of the colony at a
      C     specified time. The growth rate is assumed to be
      C     proportional to the colony's size.

            REAL  sizei, sizef, time, rate

            sizei = 10000.0
            time  = 40.5
            rate  = 0.0875

            sizef = sizei * EXP (rate * time)

            WRITE (*, 100) sizef
        100 FORMAT (' THE FINAL SIZE IS ', E12.6)
            END


 5.1.12  Trigonometric Functions

 Table 5.13 summarizes the trigonometric intrinsic functions.

 Table 5.13  Intrinsic Functions: Trigonometric Functions

                                                    Argument               Fun
 Name                    Definition                 Type                   Typ

 SIN (gen)               Sine                       real or cmp            Sam
                                                                           arg

 DSIN (dbl)              Sine                       «REAL*8»               «RE

 CSIN (cmp8)             Sine                       COMPLEX«*8»            COM

 «CDSIN (cmp16)»         «Sine»                     «COMPLEX*16»           «CO

 COS (gen)               Cosine                     real or cmp            Sam
                                                                           arg

 DCOS (dbl)              Cosine                     «REAL*8»               «RE

 CCOS (cmp8)             Cosine                     COMPLEX«*8»            COM

 «CDCOS (cmp16)»         «Cosine»                   «COMPLEX*16»           «CO

 TAN (real)              Tangent                    real                   Sam
                                                                           arg

 DTAN (dbl)              Tangent                    «REAL*8»               «RE

 ASIN (real)             Arc sine                   real                   Sam
                                                                           arg

 DASIN (dbl)             Arc sine                   «REAL*8»               «RE

 ACOS (real)             Arc cosine                 real                   Sam
                                                                           arg

 DACOS (dbl)             Arc cosine                 «REAL*8»               «RE

 ATAN (real)             Arc tangent                real                   Sam
                                                                           arg

 DATAN (dbl)             Arc tangent                «REAL*8»               «RE

 ATAN2 (realA,           Arc tangent (realA /       real                   Sam
 realB)                  realB)                                            arg

 DATAN2 (dblA,           Arc tangent (dblA /        «REAL*8»               «RE
 dblB)                   dblB)

 «COTAN (real)»          «Cotangent»                «real»                 «Sa
                                                                           arg

 «DCOTAN (dbl)»          «Cotangent»                «REAL*8»               «RE

 SINH (real)             Hyperbolic sine            real                   Sam
                                                                           arg

 DSINH (dbl)             Hyperbolic sine            «REAL*8»              «REA

 COSH (real)             Hyperbolic cosine          real                   Sam
                                                                           arg

 DCOSH (dbl)             Hyperbolic cosine          «REAL*8»               «RE

 TANH (real)             Hyperbolic tangent         real                   Sam
                                                                           arg

 DTANH (dbl)             Hyperbolic tangent         «REAL*8»               «RE


 All angles in trigonometric intrinsic functions are specified in radians.
 Table 5.14 indicates some restrictions on the arguments to and results of
 trigonometric intrinsic functions.

 Table 5.14  Restrictions on Arguments and Results

                            Restrictions               Range of
 Function                   on Arguments               Results

 SIN, DSIN, COS, DCOS,      None                       All real numbers
 TAN, DTAN

 ASIN, DASIN                |arg| <= 1                 -p/2 <= result <= p/2

 ACOS, DACOS                |arg| <= 1                 0 <= result <= p

 ATAN, DATAN                None                       -p/2 <= result <= p/2

 ATAN2, DATAN2              Arguments cannot both      -p <= result <= p
                            be zero

 «COTAN»                    «Argument cannot be        «All real numbers»
                            zero»

 The range of the results of the intrinsic functions ATAN2 and DATAN2 is as
 follows:

 Arguments                                  Result

 genA > 0                                   result > 0

 genA = 0 and genB > 0                      result = 0

 genA = 0 and genB < 0                      result = p

 genA < 0                                   result < 0

 genB = 0                                   |result| = p/2


 Example

 The following program uses trigonometric intrinsic functions:

      C     This program prompts for a polar coordinate
      C     and converts it to a rectangular coordinate.
      C
            REAL  theta, radius, x, y

            WRITE  (*, *) ' Enter polar coordinate (radius,
           +                angle)'
            READ   (*, '(2F10.5)') radius, theta

            x = radius * COS (theta)
            y = radius * SIN (theta)

            WRITE  (*, 100) x, y
       100  FORMAT (/ ' (X,Y) = (', F7.3, ',', F7.3,')' )
            END


 5.1.13  Character Functions

 Table 5.15 summarizes the intrinsic functions that operate on character
 constants and variables.

 Table 5.15  Intrinsic Functions: Character Functions

                                                    Argument              Func
 Name                    Definition                 Type                  Type

 LGE (charA, charB)      charA >= charB             char                  log

 LGT (charA, charB)      charA > charB              char                  log

 LLE (charA, charB)      charA <= charB             char                  log

 LLT (charA, charB)      charA < charB              char                  log

 LEN (char)              Length of string           char                  int

 INDEX (charA, charB     Position of substring      char, «log»           int
 ««, log»»)              charB in string charA

 «LEN_TRIM (char)»       «Length of string,         «char»                «int
                         less trailing blanks»

 «SCAN (char,            «Element of charset in     «char, log»           «int
 «charset«, log»)»        char»

 «VERIFY (char,          «Element of charset        «char, log»           «log
 charset«, log»)»        not in char»


 The intrinsic functions LGE, LGT, LLE, and LLT use the ASCII collating
 sequence to determine whether a character argument is less than (precedes in
 the ASCII collating sequence), greater than (follows in the ASCII collating
 sequence), or equal to (identical in the ASCII collating sequence) another
 character argument. If two character arguments are not of equal length, the
 shorter operand is padded to the length of the larger operand by adding
 blanks.

 The argument to the LEN intrinsic function does not have to be assigned a
 value.

 The INDEX intrinsic function returns an integer specifying the position of
 charB in charA. If the length of charA is less than the length of charB, or
 if charB does not occur in charA, the index equals zero. If charB occurs
 more than once in charA, the position of the first occurrence of charB is
 returned. «The log parameter, when .TRUE., starts the comparison at the end
 of the string and moves toward the beginning.»

 «The LEN_TRIM function returns the length of the string argument, less the
 number of trailing blanks.»

 «The SCAN and VERIFY functions both compare a string with the group of
 characters in charset. SCAN returns the position of the first string
 character that matches a character in charset, while VERIFY returns the
 first position that does not match a character in charset. If there is no
 match, or the string is of zero length, SCAN returns zero. If there is no
 mismatch, or the string is of zero length, VERIFY returns zero. The log
 parameter, when .TRUE., starts the comparison at the end of the string and
 moves toward the beginning.»


 Example

 The following list shows examples of the character intrinsic functions:

 Function Reference                         Equivalent

 LLE ('A', 'B')                             .TRUE.

 LLT ('A', 'a')                             .TRUE.

 LEN ('abcdef')                             6

 LEN_TRIM ('abc   ')                        3

 INDEX ('banana', 'an', «.TRUE.)»           4

 «SCAN ('banana', 'nbc')»                   «1»

 «VERIFY ('banana', 'nbc')»                 «2»


 5.1.14  «End-of-File Function»

 «Table 5.16 summarizes the end-of-file intrinsic function.»

 «Table 5.16  Intrinsic Functions: End-of-File Function

 «                                                  Argument              Func
 «Name                   Definition                 Type                  Type

 «EOF(int)»              «End-of-file»              «int»                 «log


 «If the unit specified by its argument is at or past the end-of-file record,
 the value .TRUE. is returned by the intrinsic function EOF (int). Otherwise,
 EOF returns the value .FALSE..The value of int must be the unit specifier
 corresponding to an open file. The value of int cannot be zero, unless you
 have reconnected unit zero to a unit other than the screen or keyboard.»


 «Example»

 «The following program uses the EOF intrinsic function:»

      «C     This program reads a file of integers»
      «C     and prints their average.»

           «CHARACTER*64  fname»
           «INTEGER       total, count, value»

           «WRITE (*, '(a  )') ' Enter file name:  '»
           «READ  (*, '(a  )') fname»

      «C     Open unit 1 for input (any unit except * is ok).»
           «OPEN (1, FILE = fname)»

           «total = 0»
           «count = 0»
      «100  IF (.NOT. EOF (1)) THEN»
              «count = count + 1»
              «read (1, '(I7)') value»
              «total = total + value»
              «GOTO 100»
           «ENDIF»
           «IF (count .GE. 0) THEN»
              «WRITE (*, *) 'Average is:', FLOAT (total) / count»
           «ELSE»
              «WRITE (*, *) 'Input file is empty'»
           «ENDIF»
           «END»


 5.1.15  «Address Functions»

 «Table 5.17 lists the intrinsic functions that return addresses.»

 «Table 5.17  Intrinsic Functions: Addresses»

 «                                                  Argument          Function
 «Name                   Definition                 Type              Type»

 «LOCNEAR (gen)»         «Unsegmented address»      «Any»             «INTEGER

 «LOCFAR (gen)»          «Segmented address»        «Any»             «INTEGER

 «LOC (gen)»             «Address»                  «Any»             «INTEGER
                                                                      INTEGER*


 «These three intrinsic functions return the machine address of the variable
 passed as an actual argument.»

 «The following list shows how the address is returned for different types of
 arguments:»

 «Argument                                  Return Value»

 «Expression, function call, or             «A temporary variable is created
 constant»                                  to hold the result of the
                                            expression, function call, or
                                            constant. The address of the
                                            temporary variable is then
                                            returned.»

 «All other arguments»                      «The machine address of the
                                            actual argument is returned.»

 «The value returned by the LOCNEAR intrinsic function is equivalent to a
 near procedure or data pointer in Microsoft C or an ADR type in Microsoft
 Pascal. Similarly, the value returned by the LOCFAR intrinsic function is
 equivalent to a far data or function pointer in Microsoft C, or an ads,
 adsfunc, or adsproc type in Microsoft Pascal.»

 «LOCNEAR can only be used with NEAR procedures and with objects in the
 default data segment, such as objects in NEAR common blocks and objects not
 named in $LARGE metacommands. For example, LOCNEAR will not usually return
 the correct address of an argument unless that argument is explicitly in the
 default data segment. LOC returns either a near or a far pointer, depending
 on the memory model used to compile.»


 5.1.16  «Bit-Manipulation Functions»

 «Table 5.18 summarizes the functions that perform bit manipulation.»

 «Table 5.18  Intrinsic Functions: Bit Manipulation»

 «                                                  Argument          Function
 «Name                   Definition                 Type              Type»

 «IOR (intA, intB)»      «Inclusive or»             «int»             «Same as
                                                                      argument

 «ISHL (intA,            «Logical shift»            «int»             «Same as
 intB)»                                                               argument

 «ISHFT (intA,           «Logical shift»            «int»             «Same as
 intB)»                                                               argument

 «ISHA (intA, intB)»     «Arithmetic shift»         «int»             «Same as
                                                                      argument

 «ISHC (intA, intB)»     «Rotate»                   «int»             «Same as
                                                                      argument

 «IEOR (intA, intB)»     «Exclusive or»             «int»             «Same as
                                                                      argument

 «IAND (intA, intB)»     «Logical product»          «int»             «Same as
                                                                      argument

 «NOT (intA)»            «Logical complement»       «int»             «Same as
                                                                      argument

 «IBCLR (intA,           «Bit clear»                «int»             «Same as
 intB)»                                                               argument

 «IBSET (intA,           «Bit set»                  «int»             «Same as
 intB)»                                                               argument

 «IBCHNG (intA,          «Bit change»               «int»             «Same as
 intB)»                                                               argument

 «BTEST (intA,           «Bit test»                 «int»             «log»
 intB)»


 «All bit-manipulation intrinsic functions can be passed as actual arguments.
 These intrinsic functions work as follows:»

 «Function                                  Operation»

 «Inclusive or»                             «If the nth bit of either the
                                            first or second argument is 1,
                                            then the nth bit of the result is
                                            set to 1.»

 «Logical shift»                            «If intB is greater than or equal
                                            to zero, shift intA logically
                                            left by intB bits. If intB is
                                            less than zero, shift intA
                                            logically right by intB bits.»

 «Arithmetic shift»                         «If intB is greater than or equal
                                            to zero, shift intA
                                            arithmetically left by intB bits.
                                            If intB is less than zero, shift
                                            intA arithmetically right by intB
                                            bits.»

 «Rotate»                                   «If intB is greater than or equal
                                            to zero, rotate intA left intB
                                            bits. If intB is less than zero,
                                            rotate intA right intB bits.»

 «Exclusive or»                             «If the nth bits of the first and
                                            second arguments are not equal to
                                            each other, then the nth bit of
                                            the result is set to 1.
                                            Otherwise, the nth bit of the
                                            result is set to 0.»

 «Logical product»                          «If the nth bits of both the
                                            first and second arguments are 1,
                                            then the nth bit of the result is
                                            set to 1. Otherwise, the nth bit
                                            of the result is set to 0.»

 «Logical complement»                       «If the nth bit of the argument
                                            is 1, then the nth bit of the
                                            result is set to 0. Otherwise,
                                            the nth bit of the result is set
                                            to 1.»

 «Bit clear»                                «Clear intB bit in intA.»

 «Bit set»                                  «Set intB bit in intA.»

 «Bit change»                               «Reverse value of intB bit in
                                            intA.»

 «Bit test»                                 «Return .TRUE. if bit intB in
                                            intA is set to 1. Return .FALSE.
                                            otherwise.»


 «Examples»

 «The following three examples show the results of three bit-manipulation
 intrinsic functions:»

 «Function                                       Binary Representation»

 «IOR(240,90) = 250»                                          «11110000»
                                            «IOR»             «01011010»
                                                              «────────»
                                                              «11111010»

 «IEOR(240,90) = 170»                                         «11110000»
                                            «IEOR»            «01011010»
                                                              «────────»
                                                              «10101010»

 «IAND(240,90) = 80»                                          «11110000»
                                            «IAND»            «01011010»
                                                              «────────»
                                                              «01010000»

 «Table 5.19 shows the results of other bit-manipulation intrinsic
 functions.»

 «Table 5.19    Bit-Manipulation Examples»

 «Function Reference      IntA                  Result»
 «ISHFT(IntA,2)           10010000 11000101     01000011 00010100»
 «ISHFT(IntA,-2)          10010000 11000101     00100100 00110001»
 «ISHA(IntA,3)            10000000 11011000     00000110 11000000»
 «ISHA(IntA,-3)           10000000 11011000     11110000 00011011»
 «ISHC(IntA,3)            01110000 00000100     10000000 00100011»
 «ISHC(IntA,-3)           01110000 00000100     10001110 00000000»
 «NOT(IntA)               00011100 01111000     11100011 10000111»
 «IBCLR(IntA,4)           00011100 01111000     00011100 01101000»
 «IBSET(IntA,14)          00011100 01111000     01011100 01111000»
 «IBCHNG(IntA,5)          00011100 01111000     00011100 01011000»
 «BTEST(IntA,2)           00011100 01111000     .FALSE.»
 «BTEST(IntA,3)           00011100 01111000     .TRUE.»


 5.2  Alphabetical Function List

 What follows in Table 5.20 is an alphabetical listing of all intrinsic
 functions in Microsoft FORTRAN. See Table 5.1 for a list of the
 abbreviations used for data types.

 Table 5.20  Intrinsic Functions

                                                         Argument          Fun
 Name                                  Definition        Type              Typ

 ABS (gen)                             Absolute          int, real, or     Sam
                                       value             cmp               typ
                                                                           arg
                                                                           cmp

 ACOS (real)                           Arc cosine        real              Sam
                                                                           arg

 AIMAG (cmp8)                          Imaginary         cmp8              rea
                                       part of cmp8
                                       number

 AINT (real)                           Truncate          real              Sam
                                                                           arg

 «ALLOCATED (array)»                   «Allocation       «Any»             «lo
                                       status of
                                       array»

 ALOG (real4)                          Natural           real4             rea
                                       logarithm

 ALOG10 (real4)                        Common            real4             rea
                                       logarithm

 AMAX0 (intA, intB«, intC»...)         Maximum           int               rea

 AMAX1 (real4A, real4B«, real4C»...)   Maximum           real4             rea

 AMIN0 (intA, intB«, intC»...)         Minimum           int               rea

 AMIN1 (real4A, real4B«, real4C»...)   Minimum           real4             rea

 AMOD (real4A, real4B)                 Remainder         real4             rea

 ANINT (real)                          Round             real              Sam
                                                                           arg

 ASIN (real)                           Arc sine          real              Sam
                                                                           arg

 ATAN (real)                           Arc tangent       real              Sam
                                                                           arg

 ATAN2 (realA, realB)                  Arc tangent       real              Sam
                                       (realA/realB)                       arg

 «BTEST (intA,intB)»                   «Bit test»        «int»             «lo

 CABS (cmp)                            Absolute          cmp               rea
                                       value

 CCOS (cmp8)                           Cosine            cmp8              cmp

 «CDABS (cmp16)»                       «Absolute         «cmp16»           «db
                                       value»

 «CDCOS (cmp16)»                       «Cosine»          «cmp16»           «cm

 «CDEXP (cmp16)»                       «Exponent»        «cmp16»           «cm

 «CDLOG (cmp16)»                       «Natural          «cmp16»           «cm
                                       logarithm»

 «CDSIN (cmp16)»                       «Sine»            «cmp16»           «cm

 «CDSQRT (cmp16)»                      «Square           «cmp16»           «cm
                                       root»

 CEXP (cmp8)                           Exponent          cmp8              cmp

 CHAR (int)                            Data-type         int               cha
                                       conversion

 CLOG (cmp8)                           Natural           cmp8              cmp
                                       logarithm

 CMPLX (genA«, genB»)                  Data-type         int, real, or     cmp
                                       conversion        cmp

 CONJG (cmp8)                          Conjugate of      cmp8              cmp
                                       cmp8 number

 COS (gen)                             Cosine            real or cmp       Sam
                                                                           arg

 COSH (real)                           Hyperbolic        real              Sam
                                       cosine                              arg

 «COTAN (real)»                        «Cotangent»       «real»            «Sa
                                                                           arg

 CSIN (cmp8)                           Sine              cmp8              cmp

 CSQRT (cmp8)                          Square root       cmp8              cmp

 DABS (dbl)                            Absolute          dbl               dbl
                                       value

 DACOS (dbl)                           Arc cosine        dbl               dbl

 DASIN (dbl)                           Arc sine          dbl               dbl

 DATAN (dbl)                           Arc tangent       dbl               dbl

 DATAN2 (dblA, dblB)                   Arc tangent       dbl               dbl
                                       (dblA/dblB)

 DBLE (gen)                            Data-type         int, real, or     dbl
                                       conversion        cmp

 «DCMPLX (genA«, genB»)»               «Data-type        «int, real,       «cm
                                       conversion»       or cmp»

 «DCONJG (cmp16)»                      «Conjugate of     «cmp16»           «cm
                                       cmp16
                                       number»

 DCOS (dbl)                            Cosine            dbl               dbl

 DCOSH (dbl)                           Hyperbolic        dbl               dbl
                                       cosine

 DCOTAN (dbl)                          Cotangent         dbl               dbl

 DDIM (dblA, dblB)                     Positive          dbl               dbl
                                       difference

 DEXP (dbl)                            Exponent          dbl               dbl

 «DFLOAT (gen)»                        «Data-type        «int, real,       «db
                                       conversion»       or cmp»

 DIM (genA, genB)                      Positive          int or real       Sam
                                       difference                          arg

 DIMAG (cmp16)                         Imaginary         cmp16             dbl
                                       part of cmp16
                                       number

 DINT (dbl)                            Truncate          dbl               dbl

 DLOG (dbl)                            Natural           dbl               dbl
                                       logarithm

 DLOG10 (dbl)                          Common            dbl               dbl
                                       logarithm

 DMAX1 (dblA, dblB«, dblC»...)         Maximum           dbl               dbl

 DMIN1 (dblA, dblB«, dblC»...)         Minimum           dbl               dbl

 DMOD (dblA, dblB)                     Remainder         dbl               dbl

 DNINT (dbl)                           Round             dbl               dbl

 DPROD (real4A, real4B)                Double-           real4             dbl
                                       precision
                                       product

 «DREAL (cmp16)»                       «Data-type        «cmp16»           «db
                                       conversion»

 DSIGN (dblA, dblB)                    Sign transfer     dbl               dbl

 DSIN (dbl)                            Sine              dbl               dbl

 DSINH (dbl)                           Hyperbolic        dbl               dbl
                                       sine

 DSQRT (dbl)                           Square root       dbl               dbl

 DTAN (dbl)                            Tangent           dbl               dbl

 DTANH (dbl)                           Hyperbolic        dbl               dbl
                                       tangent

 «EOF (int)»                           «End-of-file»     «int»             «lo

 «EPSILON (gen)»                       «Smallest         «real»            «re
                                       increment
                                       over 1»

 EXP (gen)                             Exponent          real or cmp       Sam
                                                                           arg

 FLOAT (int)                           Data-type         int               rea
                                       conversion

 «HFIX (gen)»                          «Data-type        «int, real,       «in
                                       conversion»       or cmp»

 «HUGE (gen)»                          «Largest          «int or real      «Sa
                                       positive                            arg
                                       number»

 IABS (int)                            Absolute          int               int
                                       value

 «IAND (intA, intB)»                   «Logical          «int»             «Sa
                                       product»                            arg

 «IBCHNG (intA, intB)»                 «Bit change»      «int»             «Sa
                                                                           arg

 «IBCLR (intA, intB)»                  «Bit clear»       «int»             «Sa
                                                                           arg

 «IBSET (intA, intB)»                  «Bit set»         «int»             «Sa
                                                                           arg

 ICHAR (char                           Data-type         char              int
                                       conversion

 IDIM (intA, intB)                     Positive          int               int
                                       difference

 IDINT (dbl)                           Data-type         dbl               int
                                       conversion

 IDNINT (dbl)                          Round             dbl               int

 «IEOR (intA, intB)»                   «Exclusive        «int»             «Sa
                                       or»                                 arg

 IFIX (real4)                          Data-type         real4             int
                                       conversion

 «IMAG (cmp)»                          «Imaginary        «cmp»             «re
                                       part of cmp
                                       number»

 INDEX (charA, charB ««, log»»)        Location of       char, «log»       int
                                       substring
                                       charB in
                                       string charA

 «INT (gen)»                           «Data-type        «int, real,       «in
                                       conversion»       or cmp»

 «INT1 (gen)»                          «Data-type        «int, real,       «in
                                       conversion»       or cmp»

 «INT2 (gen)»                          «Data-type        «int, real,       «in
                                       conversion»       or cmp»

 «INT4 (gen)»                          «Data-type        «int, real,       «in
                                       conversion»       or cmp»

 «INTC (gen)»                          «Data-type       «int, real,        «IN
                                       conversion»       or cmp»

 «IOR (intA, intB)»                    «Inclusive or»    «int»             «Sa
                                                                           arg

 «ISHA (intA, intB)»                   «Arithmetic       «int»             «Sa
                                       shift»                              arg

 «ISHC (intA, intB)»                   «Rotate»          «int»             «Sa
                                                                           arg

 «ISHFT (intA, intB)»                  «Logical          «int»             «Sa
                                       shift»                              arg

 «ISHL (intA, intB)»                   «Logical          «int»             «Sa
                                       shift»                              arg

 ISIGN (intA, intB)                    Sign transfer     int               int

 «JFIX (gen)»                          «Data-type        «int, real,       «in
                                       conversion»       or cmp»

 LEN (char)                            Length of         char              int
                                       string

 «LEN_TRIM (char)»                     «Length of        «char»            «in
                                       string,
                                       excluding
                                       trailing
                                       blanks»

 LGE (charA, charB)                    charA>=charB      char              log

 LGT (charA, charB)                    charA>charB       char              log

 LLE (charA, charB)                    charA<=charB      char              log

 LLT (charA, charB)                    charA<charB       char              log

 «LOC (gen)»                           «Address»         «Any»             «in

 «LOCFAR (gen)»                        «Segmented        «Any»             «in
                                       address»

 «LOCNEAR (gen)»                       «Unsegmented      «Any»             «in
                                       address»

 LOG (gen)                             Natural           real or cmp       Sam
                                       logarithm                           arg

 LOG10 (real)                          Common            real              Sam
                                       logarithm                           arg

 MAX (genA, genB«, genC»...)           Maximum           int or real       Sam
                                                                           arg

 MAX0 (intA, intB«, intC»...)          Maximum           int               int

 MAX1 (real4A, real4B«, real4C»...)    Maximum           real4             int

 «MAXEXPONENT (real)»                  «Largest          «real»            «re
                                       positive
                                       exponent of
                                       data type»

 MIN (genA, genB«, genC»...)           Minimum           int or real       Sam
                                                                           arg

 MIN0 (intA, intB«, intC»...)          Minimum           int               int

 MIN1 (real4A, real4B«, real4C»...)    Minimum           real4             int

 «MINEXPONENT (real)»                  «Largest          «real»            «re
                                       negative
                                       exponent of
                                       data type»

 MOD (genA, genB)                      Remainder         int or real       Sam
                                                                           arg

 «NEAREST (real, director)»            «Nearest          «real»            «re
                                       value in
                                       direction of
                                       sign of
                                       director»

 NINT (real)                           Round             real              int

 «NOT (intA)»                          «Logical          «int»             «Sa
                                       complement»                         arg

 «PRECISION (gen)»                     «Number of        «real»            «re
                                       significant
                                       digits for
                                       data type»

 REAL (gen)                            Data-type         int, real, or     rea
                                       conversion        cmp

 «SCAN (charA, charB «,log»)»         «Position of       «char and         «in
                                       first             log»
                                       occurrence of
                                       character
                                       from charB in
                                       charA»

 SIGN (genA, genB)                     Sign transfer     int or real       Sam
                                                                           arg

 SIN (gen)                             Sine              real or cmp       Sam
                                                                           arg

 SINH (real)                           Hyperbolic        real              Sam
                                       sine                                arg

 SNGL (dbl)                            Data-type         dbl               rea
                                       conversion

 SQRT (gen)                            Square root       real or cmp       Sam
                                                                           arg

 TAN (real)                            Tangent           real              Sam
                                                                           arg

 TANH (real)                           Hyperbolic        real              Sam
                                       tangent                             arg

 «TINY (real)»                         «Returns          «real»            «re
                                       small-
                                       est positive
                                       number > 0
                                       for data
                                       type»

 «VERIFY (charA, charB «,log»)»        «Position of      «char and         «in
                                       first             log»
                                       occurrence of
                                       character not
                                       from charB in
                                       charA»


 5.3  Additional Procedures

 Microsoft FORTRAN contains additional procedures that control and access
 system time and date, get and reset run-time error-code information, return
 command-line arguments, and generate pseudorandom numbers. The following
 sections describe these procedures.

 These procedures are included in the FORTRAN run-time library, and are
 automatically linked to your program if called. However, they are not
 intrinsic functions. You may write other functions that appropriate these
 functions' names without having to reference the names in an EXTERNAL
 statement. Your own functions will be linked instead, as long as LINK
 references their object code before calling the FORTRAN library.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Microsoft FORTRAN Advanced Topics discusses other functions in the
    FORTRAN libraries, including the C spawnlp and system functions (Chapter
    4), graphics and full-screen text functions (Chapters 8-11), and OS/2
    thread control functions (Chapter 3).
 ───────────────────────────────────────────────────────────────────────────


 5.3.1  Time and Date Procedures

 The functions SETTIM and SETDAT, and the subroutines GETTIM and GETDAT,
 allow you to use the system time and date in your programs. SETTIM and
 SETDAT set the system time and date; GETTIM and GETDAT return the time and
 date. Table 5.21 summarizes the time and date procedures.

 Table 5.21  Time and Date Procedures

                                                           Argument      Funct
 Name                                 Definition           Type          Type
 GETTIM (ihr, imin, isec, i100th)     Gets system time     INTEGER*2
 SETTIM (ihr, imin, isec, i100th)     Sets system time     INTEGER*2     LOGIC
 GETDAT (iyr, imon, iday)             Gets system date     INTEGER*2
 SETDAT (iyr, imon, iday)             Sets system date     INTEGER*2     LOGIC

 The arguments are defined as follows:

 Argument                                   Definition

 ihr                                        Hour (0-23)

 imin                                       Minute (0-59)

 isec                                       Second (0-59)

 i100th                                     Hundredth of a second (0-99)

 iyr                                        Year (xxxx AD)

 imon                                       Month (1-12)

 iday                                       Day of the month (1-31)

 Actual arguments used in calling GETTIM and GETDAT must be INTEGER*2
 variables, array elements, or structure elements. Because these subroutines
 redefine the values of their arguments, other kinds of expressions are
 prohibited.

 Actual arguments of the functions SETTIM and SETDAT can be any legal
 INTEGER*2 expression. SETTIM and SETDAT return .TRUE. if the system time or
 date is changed, or .FALSE. if no change is made.

 Refer to your operating system documentation for the range of permitted
 dates.


 Example

 The following program sets the date and time, then prints them on the
 screen:

      C     Warning: this program will reset
      C     your system date and time.

      $STORAGE:2
            CHARACTER*12  cdate, ctime
            LOGICAL  SETDAT, SETTIM
            DATA cdate / 'The date is ' /
            DATA ctime / 'The time is ' /
            IF (.NOT. (SETDAT (2001, 7, 4)))
           +     WRITE (*, *) 'SETDAT failed'
      C     sets the date to July 4th, 2001:
            IF (.NOT. (SETTIM (0, 0, 0, 0)))
           +     WRITE (*, *) 'SETTIM failed'
      C     sets the time to 00:00:00.00 (midnight)
            CALL GETDAT (iyr, imon, iday)
      C     gets the date from the system clock
            CALL GETTIM (ihr, imin, isec, i100th)
      C     gets the time of day from the system clock
            cdate, imon, iday, iyr
      C     writes the date
            WRITE (*, '(1X, A, I2.2, 1H:, I2.2, 1H:, I2.2, 1H.,
           +I2.2)') ctime, ihr, imin, isec, i100th
      C     writes the time in the format xx:xx:xx.xx
            END


 5.3.2  Run-Time-Error Procedures

 The IGETER function and the ICLRER subroutine are included for compatibility
 with previous versions of FORTRAN. Their functionality is provided in the
 current version by the IOSTAT= option. (See Section 3.2.6, "Error and
 End-of-File Handling," for more information about IOSTAT=.)

 IGETER is called after an I/O operation that includes the ERR= or IOSTAT=
 options. The significance of the value returned is explained in the
 following list:

 Return Value                               Description

 0                                          No error occurred.

 Negative value                             An end-of-file condition
                                            occurred, but no other error
                                            occurred.

 Positive value                             An error occurred. The return
                                            value is the error number.

 The IGETER calling interface has the following form:

            INTEGER*2 FUNCTION IGETER (iunit)
            INTEGER*2 iunit
            .
            .
            .
            END

 ICLRER resets the FORTRAN run-time-error code information after an error has
 been encountered and handled through ERR= and IOSTAT=. The ICLRER calling
 interface takes the following form:

            SUBROUTINE ICLRER (iunit)
            INTEGER*2 iunit
            .
            .
            .
            END


 5.3.3  Command-Line-Argument Procedures

 The NARGS function returns the total number of command-line arguments,
 including the command. The GETARG subroutine returns the nth command-line
 argument (where the command itself is argument number zero). The syntax of
 these procedures is shown below:

      numargs = NARGS ( )
      CALL GETARG (n, buffer, status)

 The NARGS function takes no arguments. It always returns an INTEGER*4 value,
 regardless of the $STORAGE setting.

 The GETARG subroutine takes three arguments. The first, of INTEGER*2 type,
 specifies the position of the desired argument. (The command itself is
 argument zero.)

 The buffer argument is a CHARACTER variable that returns the desired
 command-line argument. If the argument is shorter than buffer, GETARG pads
 buffer on the right with blanks. If the argument is longer than buffer,
 GETARG truncates the argument.

 The status argument is an INTEGER*2 type that returns a status value on
 completion. If there were no errors, status returns the number of characters
 in the command-line argument before truncation or blank-padding. (That is,
 status is the original number of characters in the command-line argument.)
 Errors include specifying an argument position less than zero or greater
 than the value returned by NARGS. For either of these errors, status returns
 a value of -1.

 The status argument is an INTEGER*2 type which returns a status value.

 If there were no errors, status contains the number of characters in the
 command-line argument before truncation or blank-padding. (That is, status
 is the original number of characters in the command-line argument.) Errors
 include passing a status less than one, or specifying an argument position
 less than zero or greater than the value returned by NARGS. For any of these
 errors, status returns a value of -1.


 Example

 Assume a command-line invocation of ANOVA -g -c -a, and that buffer is at
 least five characters long. The following GETARG statements return the
 corresponding arguments in the buffer:

 String                                     Length
 Statement                                  Returned          Returned

 CALL GETARG (0, buffer, status)            ANOVA             5

 CALL GETARG (1, buffer, status)            -g                2

 CALL GETARG (2, buffer, status)            -c                2

 CALL GETARG (3, buffer, status)            -a                2

 CALL GETARG (4, buffer, status)            undefined        -1


 5.3.4  Random Number Procedures

 The RANDOM subroutine returns a pseudorandom real value greater than or
 equal to zero and less than one. The SEED subroutine changes the starting
 point of the pseudorandom number generator. The syntax of these procedures
 is shown below:

      CALL RANDOM (ranval)
      CALL SEED (seedval)

 The RANDOM subroutine takes a single REAL*4 argument through which the
 random value is returned.

 The SEED subroutine takes a single INTEGER*2 argument. SEED uses this value
 to establish the starting point of the pseudorandom number generator. A
 given seed always produces the same sequence of values from RANDOM.

 If SEED is not called before the first call to RANDOM, RANDOM always begins
 with a seed value of one. If a program must have a different pseudorandom
 sequence each time it runs, use the GETTIM procedure to pick a seed value
 (the hundredth-of-a-second parameter is a good choice because it changes so
 rapidly).


 ───────────────────────────────────────────────────────────────────────────
 Part 2  Compiling and Linking

 The two chapters in this section provide information on using metacommands
 for compiler control and on using the FL command for compiling and linking.

 Chapter 6, which describes metacommands, includes overview material as well
 as an alphabetical metacommand directory. Chapter 7 shows how to use FL to
 compile and link FORTRAN programs in a single step. This chapter includes
 detailed information on available FL options.


 «Chapter 6  Metacommands»

 ───────────────────────────────────────────────────────────────────────────

 The first part of this chapter lists all Microsoft FORTRAN metacommands. The
 second part explains how the conditional-compilation metacommands are used
 to control which sections of your program are compiled. The third part is a
 directory of metacommands, arranged alphabetically.

 Metacommands tell the Microsoft FORTRAN Compiler how you want the source
 code compiled. Note that some command-line options, as described in
 Microsoft FORTRAN Advanced Topics, duplicate metacommand functions. If there
 is a conflict between a metacommand and a command-line switch, the
 metacommand takes precedence.

 Table 6.1 summarizes the Microsoft FORTRAN metacommands. Although some
 metacommands (such as $DO66 or $FREEFORM) can be used only once in a
 program, many can appear anywhere in the source code. This flexibility
 permits specific compilation features (such as loop optimization) to be
 enabled or disabled as desired.

 Table 6.1  Metacommands

 Metacommand             Instructions to Compiler                     Default

 $DEBUG«:string»         Turns on run-time checking for integer       $NODEBUG
                         arithmetic operations, assigned GOTO
                         values, subscript bounds, substrings,
                         and CASE selection. $NODEBUG turns off
                         checking. $DEBUG does not trigger or
                         suppress floating-point exceptions. $DEBUG
                         can also be used for conditional
                         compilation.

 $DECLARE                Generates warning messages for               $NODECLA
                         undeclared variables. $NODECLARE turns
                         off these messages.

 $DEFINE symbol-name     Creates (and optionally assigns a value      $UNDEFIN
 «= val»                 to) a variable whose existence may be
                         tested during conditional compilation.
                         $UNDEFINE removes a symbolic variable
                         name.

 $DO66                   Uses FORTRAN 66 semantics for DO             $DO66 no
                         statements.

 $ELSE                   Marks the beginning of a conditional         None
                         compilation block that is compiled if
                         the logical condition in the matching
                         $IF metacommand is .FALSE..

 $ELSEIF expr            Marks the beginning of a new conditional     None
                         compilation block that is compiled if
                         the logical condition in the matching
                         $IF metacommand is .FALSE. and expr is
                         .TRUE..

 $ENDIF                  Terminates $IF, $IF...$ELSE, or              None
                         $IF...$ELSEIF conditional compilation
                         blocks.

 $FLOATCALLS             Generates calls to subroutines in the        $NOFLOAT
                         emulator library. $NOFLOATCALLS causes
                         the compiler to generate in-line
                         interrupt
                         instructions.

 $FREEFORM               Uses free-form format for source code.       $NOFREEF
                         $NOFREEFORM uses fixed format.

 $IF expr                Marks the beginning of a conditional         None
                         compilation block. The succeeding
                         statements are compiled if the
                         conditional expression is .TRUE..

 $INCLUDE:'file'         Proceeds as if the contents of file were     None
                         inserted at this point in the current
                         source file.

 $LARGE«:name            Addresses the named array outside of the     None
 «, name»»...            DGROUP segment. $NOTLARGE disables
                         $LARGE for the named array. If name is
                         omitted, these metacommands affect all
                         arrays.

 $LINESIZE:n             Makes subsequent pages of listing n          $LINESIZ
                         columns wide. Minimum n equals 40;
                         maximum n equals 132.

 $LIST                   Begins generation of listing information     $LIST

                         that is sent to the listing file.
                         $NOLIST suppresses generation of listing
                         information.

 $LOOPOPT                $LOOPOPT optimizes loops in following        $LOOPOPT
                         code. $NOLOOPOPT disables loop
                         optimization.

 $MESSAGE:string         Sends a character string to the standard     None
                         output device.

 $PACK:n                 Sets number of bytes for packing width.      $PACK:2
                         n must be 1, 2, or 4.

 $PAGE                   Starts new page of listing.                  None

 $PAGESIZE:n             Makes subsequent pages of listing n          $PAGESIZ
                         lines long. Minimum n equals 15.

 $STORAGE:n              Allocates n bytes of memory (2 or 4) to      $STORAGE
                         all LOGICAL or INTEGER variables.

 $STRICT                 Disables Microsoft FORTRAN features          $NOTSTRI
                         not in 1977 full-language standard.
                         $NOTSTRICT enables them.

 $SUBTITLE:subtitle      Uses subtitle subtitle for subsequent        $SUBTITL
                         pages of listing.                            1

 $TITLE:title            Uses title title for subsequent pages of     $TITLE:'
                         listing.

 $TRUNCATE               Truncates variables to six characters.       $NOTRUNC
                         $NOTRUNCATE turns off truncation.


 Any line with a dollar sign ($) in column 1 is interpreted as a metacommand.
 A metacommand and its arguments (if any) must fit on a single source line;
 continuation lines are not permitted.


 6.1  Using Conditional-Compilation Metacommands

 In FORTRAN, the conditional-execution statements IF...THEN...ELSE, ELSE, and
 ELSE IF control the execution of statement blocks. In a similar fashion, the
 conditional-compilation metacommands control which sections of source code
 are compiled. These special metacommands make it easy to include or omit
 test code, customize code for specific applications by controlling which
 sections are included, or bypass incomplete code during development.

 Like all metacommands, conditional-compilation metacommands begin with a
 dollar sign and must start in column one. Otherwise, they look and work like
 their FORTRAN counterparts:

      $IF cond-expr
      $ELSE
      $ELSEIF cond-expr
      $ENDIF

 In conditional-compilation metacommands, conditional expressions
 (cond-exprs) can take several forms. The simplest is a FORTRAN logical
 value, .TRUE. or .FALSE..

      $IF .TRUE.
            WRITE (*, '('' This is compiled if .TRUE.'')')
      $ELSE
            WRITE (*, '('' This is compiled if .FALSE.'')')
      $ENDIF

 In the above example, the first block of code is compiled and the second is
 ignored. If .TRUE. were changed to .FALSE., the second block would be the
 only block compiled. However, the programmer must manually locate every
 .TRUE. or .FALSE. to be changed. An alternative is to create one or more
 symbolic names at the beginning of the program, using the $DEFINE
 metacommand:

 $DEFINE symbol-name

 The presence or absence of a particular definition is tested by using the
 DEFINED (symbol) conditional expression in the $IF metacommand. DEFINED
 (symbol) is .TRUE. if symbol appeared in a previous $DEFINE metacommand,
 without having been assigned a value. The advantage of this arrangement is
 the way the programmer can alter the meaning of a symbol throughout the
 program simply by adding or deleting a single metacommand. It is simpler and
 more reliable than changing every .TRUE. to .FALSE..

 For example, once the $DEFINE logicvar metacommand appears in a program, any
 subsequent $IF DEFINED (logicvar) metacommand evaluates to .TRUE.. If
 $DEFINE logicvar has not appeared, $IF DEFINED (logicvar) evaluates to
 .FALSE. as in the following examples:

      $DEFINE truthvar

      $IF DEFINED (truthvar)
            WRITE (*, '('' This is compiled if truthvar     DEFINED'')')
      $ELSE
            WRITE (*, '('' This is compiled if truthvar not DEFINED'')')
      $ENDIF

 Since the symbol truthvar appears in $DEFINE truthvar, DEFINED (truthvar)
 evaluates to .TRUE. and the first block of statements is compiled. If the
 $DEFINE truthvar metacommand were removed (or converted to a comment line),
 truthvar would not be defined, DEFINED (truthvar) would evaluate to .FALSE.,
 and the second block of statements would be compiled.

 The $UNDEFINE metacommand cancels a symbol definition. In the following
 example, the $UNDEFINE metacommand cancels truthvar, and DEFINED truthvar
 evaluates to .FALSE.. The compiler ignores the first statement block:

      $DEFINE truthvar
      $UNDEFINE truthvar

      $IF DEFINED (truthvar)
            WRITE (*, '('' This is compiled if truthvar     DEFINED'')')
      $ELSE
            WRITE (*, '('' This is compiled if truthvar not DEFINED'')')
      $ENDIF

 You may define as many symbolic names as you wish anywhere in the program.
 These symbols will not conflict with FORTRAN identifiers or variable names.
 You may use names like data, sin,  equivalence, or the name of an external
 procedure, if you wish.

 A $DEFINE metacommand can also give an integer value to a symbol name (any
 four-byte FORTRAN integer value is allowed):

      $DEFINE choice = 100000

      $IF choice .NE. 100000
            WRITE (*, '('' This is compiled if choice .NE. 100000'')')
      $ELSE
            WRITE (*, '('' This is compiled if choice .EQ. 100000'')')
      $ENDIF

 In the example above, choice .NE. 100000 evaluates to .FALSE., so the second
 statement block is compiled. (DEFINED (choice) would also evaluate to
 .FALSE., because choice was assigned a value.) This example also shows how a
 conditional expression can include any FORTRAN relational operator (.EQ.,
 .NE., .LT., .LE., .GT., or .GE.). The FORTRAN .NOT. operator can also be
 used:

      $DEFINE choice = 100000
      $DEFINE receiver = choice

      $IF .NOT. (receiver .NE. 100000)
            WRITE (*, '('' This is compiled if receiver .EQ. 100000'')')
      $ELSE
            WRITE (*, '('' This is compiled if receiver .NE. 100000'')')
      $ENDIF

 In the example above, the second $DEFINE metacommand gives the value of
 choice to receiver. (This assignment is not permitted unless choice has
 already been given an integer value.) The .NOT. operator reverses the
 logical value of receiver .NE. 100000, and the first statement block is
 compiled.

 The ability to assign integer variables to symbol names greatly increases
 the programmer's control over whether a statement block is compiled, as
 shown below:

      $DEFINE upper = 3
      $DEFINE lower = -1

      $IF upper .LT. 3
            WRITE (*, '('' Compiled if upper .LT. 3'')')
      $ELSEIF lower .GT. -2
            WRITE (*, '('' Compiled if upper .GE. 3 and lower .GT. -2'')')
      $ENDIF

 The FORTRAN conditional operators .AND. and .OR. can also be used in
 comparisons:

      $DEFINE upper
      $DEFINE lower = -1

      $IF (DEFINED (upper) .OR. (lower .LE. -2)) .AND. middle
            WRITE (*, '('' Compiled if all conditions met'')')
      $ENDIF

 In this example, upper is defined, but lower is not less than or equal to
 -2. However, the .OR. comparison is enclosed within parentheses, so the net
 evaluation of these two statements is .TRUE.. The symbol middle was not
 defined, so middle is .FALSE., making the full expression .FALSE. as well.
 The enclosed statement block is not compiled.

 Logical expressions may be as complex as desired, and may include
 parentheses wherever needed to clarify the logic or override precedence.
 (The rules of precedence and logical association are the same as in
 FORTRAN.) However, the expression must fit on one line.

 Comments may be added at the end of metacommand lines. Comments must begin
 with an exclamation point:

      $DEFINE test         ! controls compilation of test code

 Constantly editing a file to add or delete symbol declarations can be
 inconvenient. Therefore, you are also permitted to define symbols on the
 compiler command line using the /D command-line option. The option /Dsymbol
 defines symbol at the beginning of the compilation. (A subsequent $UNDEFINE
 metacommand within the program file can cancel the definition.) The option
 /Dsymbol=integer defines the symbol and gives it an integer value.

 Each /D option can define or assign only one symbol. However, several /D
 options can be included in one command line. The only real limit is the
 maximum length of a command line set by your operating system.


 6.2  Metacommand Directory

 The remainder of this chapter is an alphabetical directory of the Microsoft
 FORTRAN metacommands. Each metacommand is described using the following
 format:

 Heading                                    Information

 Action                                     Summary of what the metacommand
                                            does.

 Syntax                                     Correct syntax for the
                                            metacommand, and description of
                                            the metacommand's parameters.

 Remarks                                    Use of the metacommand.


 $DEBUG and $NODEBUG


 ■  Action

 $DEBUG directs the compiler to perform additional testing and expanded error
 handling, and can also be used for conditional compilation; $NODEBUG
 suppresses the additional testing and expanded error handling


 ■  Syntax

      $«NO»DEBUG«:string»


 ■  Remarks

 The default is $NODEBUG.

 These metacommands can appear anywhere in a program, enabling and disabling
 the debug features as desired.

 When $DEBUG is enabled, the compiler does the following:

   ■  Tests integer arithmetic for overflow.

   ■  Tests assigned GOTO values against the optional label list in an
      assigned GOTO statement.

   ■  Provides the run-time error-handling system with file names and line
      numbers. If run-time errors occur, the file name and line number are
      displayed on the console.

   ■  Checks range of subscripts and substrings.

   ■  Checks assignment range. This catches errors when larger integer
      variables are assigned to smaller integer variables, such as assigning
      an INTEGER*4 variable to an INTEGER*2 variable. If $DEBUG is not
      enabled, the variable is truncated, no error is reported, and the
      program returns unpredictable results. In the case of real numbers, an
      error is always reported.

   ■  Checks for a CASE DEFAULT statement. If there is no CASE DEFAULT in a
      SELECT CASE statement, and the value of the test expression does not
      match any value in any CASE expression list, program execution halts
      with an error message. (If you do not use $DEBUG, no error occurs, and
      execution passes beyond the SELECT CASE construct to the next
      executable statement.)

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    $DEBUG has no effect on floating-point exception handling. See Microsoft
    FORTRAN Advanced Topics for information on exception handling on your
    system.
 ───────────────────────────────────────────────────────────────────────────

 This metacommand should be placed in each source file to be compiled.

 If the optional string is specified, the characters in string specify that
 lines with those characters in column 1 are to be compiled into the program.
 Case is not significant. Note that the letter C always indicates a comment
 line; therefore, if string contains a C, the C is ignored. If more than one
 $DEBUG:string metacommand is specified, each string overrides the previous
 string. $DEBUG can be used for conditional compilation only if the $FREEFORM
 metacommand has not been specified. If the $DEBUG:string metacommand appears
 after $FREEFORM, the compiler emits a warning message.


 ■  Example

      C     If the $FREEFORM metacommand has been specified,
      C     the next line produces an error message.
      $DEBUG:'ABCD'
      A    I = 1
      E    I = 2
      B    I = I + I
      F    I = I * I
      C     This is always a comment. I equals 2,
      C     because only statements A and B are executed.


 $DECLARE and $NODECLARE


 ■  Action

 $DECLARE generates warnings for variables that have not appeared in
 specification statements, and $NODECLARE disables these warnings


 ■  Syntax

      $«NO»DECLARE


 ■  Remarks

 The default is $NODECLARE. When $DECLARE is enabled, a warning message is
 generated at the first use of any variable that has not been assigned a
 value in a specification statement. The $DECLARE metacommand is primarily a
 debugging tool that locates variables that have not been properly
 initialized, or that have been defined but never used.


 ■  Example

      $DECLARE
      C     Since the variable z has not been assigned a value,
      C     its use in the statement labeled 100 causes an error:

            REAL x, y, z
            y = 1.0
      100   x = y + z



 $DEFINE and $UNDEFINE


 ■  Action

 $DEFINE creates a symbolic variable whose existence or value can be tested
 during conditional compilation; $UNDEFINE removes the symbol


 ■  Syntax

      $DEFINE symbol-name «= val»
      $UNDEFINE symbol-name

 Parameter                                  Description

 symbol-name                                An alphanumeric identifier of up
                                            to 31 characters. It may include
                                            the dollar sign and underline; it
                                            may not begin with a numeral or
                                            the underline.

 val                                        Any positive or negative
                                            INTEGER*4 value.


 ■  Remarks

 The default is $UNDEFINE, since any symbol name not specified in a $DEFINE
 metacommand tests .FALSE..

 The existence of a symbol-name is tested by the DEFINED (symbol-name)
 conditional expression. The value of a symbol-name is tested with a
 FORTRAN-style conditional expression. If a symbol-name has been assigned a
 value, it tests .FALSE. in a DEFINED conditional expression.

 Symbol names are local to the metacommands, and may duplicate FORTRAN
 keywords, intrinsic functions, or user-defined names, without conflict.

 The $DEFINE and $UNDEFINE metacommands can appear anywhere in a program,
 enabling symbol definitions as desired.


 ■  Examples

      $DEFINE  testflag
      $DEFINE  testval = 2
      .
      .
      .
      $UNDEFINE  testflag


 $DO66


 ■  Action

 $DO66 causes DO statements to conform to FORTRAN 66 semantics


 ■  Syntax

      $DO66


 ■  Remarks

 You must obey the following rules when using $DO66:

   ■  $DO66 must precede the first declaration or executable statement of the
      source file in which it occurs.

   ■  $DO66 may only be preceded by a comment line or another metacommand.

   ■  $DO66 may only appear once in the source file.

 When $DO66 is enabled, the following FORTRAN 66 semantics are used:

   ■  Statements within a DO loop are always executed at least once.

   ■  Extended range is permitted; control may transfer into the syntactic
      body of a DO statement. The range of the DO statement is thereby
      extended to include, logically, any statement that may be executed
      between a DO statement and its terminal statement. However, the
      transfer of control into the range of a DO statement prior to the
      execution of the DO statement or following the final execution of its
      terminal statement is invalid.

 Note how this differs from the default (FORTRAN 77) semantics, which are as
 follows:

   ■  DO statements will not be executed if the value of the initial control
      variable exceeds that of the final control variable (or the
      corresponding condition for a DO statement with negative increment).


 $ELSE


 ■  Action

 Marks the beginning of a $ELSE metacommand block


 ■  Syntax

      $ELSE


 ■  Remarks

 If the logical expression in the matching $IF or $ELSEIF metacommand
 evaluates to .FALSE., a $ELSE block is compiled.

 A $ELSE block consists of any statements between the $ELSE metacommand and
 the next $ENDIF metacommand at the same $IF level. The matching $ENDIF must
 appear before any $ELSE or $ELSEIF at the same $IF level.


 ■  Example

      $DEFINE flag

      $IF DEFINED (flag)
            WRITE (*, '('' This is compiled if flag
           +               DEFINED'')')
      $ELSE
            WRITE (*, '('' This is compiled if flag
           +               not DEFINED'')')
      $ENDIF




 $ELSEIF


 ■  Action

 Causes compilation of a block of statements if expression is .TRUE. and the
 matching $IF metacommand is .FALSE.


 ■  Syntax

      $ELSEIF (expression)

 Parameter                                  Description

 expression                                 A logical expression


 ■  Remarks

 The logical expression in a $ELSEIF metacommand can take two forms. One is
 the conditional-compilation operator DEFINED, followed by a symbol name in
 parentheses. If the symbol name has previously appeared in a $DEFINE
 metacommand without having been assigned a value, DEFINED (symbol_name)
 evaluates to .TRUE.. If (symbol_name) did not appear in a preceding $DEFINE
 metacommand, or did appear and was assigned a value, DEFINED (symbol_name)
 evaluates to .FALSE..

 The second form is a logical comparison, where the value of the symbol is
 compared with an integer constant or the value of another symbol using the
 FORTRAN .EQ., .NE., .GT., .LT., .GE., or .LE. operators. The results of such
 comparisons may further be evaluated with the FORTRAN .AND., .OR., .XOR.,
 and .NOT. operators. The usual rules of precedence apply, and parentheses
 may be used to control the order of evaluation.

 The associated $ELSEIF block consists of any executable statements between
 the $ELSEIF metacommand and the next $ELSEIF, $ELSE, or $ENDIF metacommand
 at the same $IF level.


 ■  Example

      $DEFINE flag
      $DEFINE testval = 3

      $IF (.NOT. DEFINED (flag))
            WRITE (*, '('' Compiled if flag not DEFINED'')')
      $ELSEIF (testval .GT. 3)
            WRITE (*, '('' Compiled if flag DEFINED .AND. testval
           +.GT. 3'')')
      $ELSE
            WRITE (*, '('' Compiled if flag DEFINED .AND. testval
           +.LE. 3'')')
      $ENDIF
      $ENDIF


 $ENDIF


 ■  Action

 Terminates a $IF metacommand block


 ■  Syntax

      $ENDIF


 ■  Remarks

 There must be a matching $ENDIF metacommand for every $IF metacommand in a
 program unit.


 ■  Example

      $DEFINE  flag
      $DEFINE  testval = 3
      $IF DEFINED (flag)
            WRITE (*, '('' Compiled if all conditions met'')')
      $ENDIF
      .
      .
      .
      $IF (testval .LT. 1) .AND. .NOT. DEFINED (flag)
            WRITE (*, '('' This is compiled if flag .EQ. 3'')')
      $ELSE
            WRITE (*, '('' This is compiled if flag .NE. 3'')')
      $ENDIF


 $FLOATCALLS and $NOFLOATCALLS


 ■  Action

 $FLOATCALLS causes floating-point operations to be processed by calls to
 library subroutines; $NOFLOATCALLS causes floating-point operations to be
 processed by compiler-generated, in-line coprocessor instructions


 ■  Syntax

      $«NO»FLOATCALLS


 ■  Remarks

 $NOFLOATCALLS is the default.

 See Microsoft FORTRAN Advanced Topics for a discussion of the advantages and
 disadvantages of floating-point operations for handling each method.


 ■  Example

      $FLOATCALLS
            REAL  x, sine

            WRITE  (*, 100)
       100     FORMAT (1X, 'ENTER x: '\)
            READ   (*, '(F10.5)') x

            WRITE  (*, 200) x, SINE (x, .00001)
       200     FORMAT (1X, 'THE SINE OF ', F10.5, ' = ', F9.6)
            END

      C     The function calculates the sine of X using a power series.
      C     Successive terms are calculated until less than eps.

      C     Library calls are generated instead of in-line instructions,
      C     letting this routine run on machines without a coprocessor.

            REAL FUNCTION SINE (x, eps)

            REAL x, y, z, next, i, eps

            z    = AMOD (x, 6.2831853)
            y    = z
            i    = 4.0
            next = -z * z * z / 6.0
       100  IF (ABS (next) .GE. eps) THEN
               y    =  y + next
               next = -next * z * z / (i * (i + 1.0))
               i    =  i + 2.0
               GOTO 100
           END IF
            SINE = y
            END


 $FREEFORM and $NOFREEFORM


 ■  Action

 $NOFREEFORM specifies that a source file is in standard FORTRAN format;
 $FREEFORM specifies that a source file is in free-form format


 ■  Syntax

      $«NO»FREEFORM


 ■  Remarks

 If this metacommand appears, it must precede any FORTRAN statements. The
 default, $NOFREEFORM, tells the compiler that your source code is in the
 standard FORTRAN format: labels are in columns 1-5, continuation characters
 are in column 6, statements are in columns 7-72, and characters beyond
 column 73 are ignored. The standard FORTRAN format is described in Section
 2.1, "Lines." $FREEFORM tells the compiler that your source code is in the
 following format:

   ■  A double quotation mark (") in column 1 indicates a comment line. An
      exclamation point outside a character or Hollerith constant indicates
      the beginning of an in-line comment.

   ■  Initial lines may start in any column.

   ■  The first nonblank character of an initial line may be a digit (the
      first digit in a statement number). The statement number may be from
      one to five decimal digits: blanks and leading zeros are ignored.
      Blanks are not required to separate the statement number from the first
      character of the statement.

   ■  If the last nonblank, noncomment character of a line is a minus sign,
      it is discarded and the next line is taken to be a continuation line.
      The continuation line may start in any column.

   ■  Alphabetic characters and asterisks are not allowed as comment markers
      in column 1.


 ■  Example

      $FREEFORM

      "The sine of the number x is calculated using a power series.
      "Successive terms are calculated until one is less than epsi.

      REAL x, epsi, z, sine, next
      epsi = 0.0001

            WRITE  (*, 100)
       100  FORMAT (1X, 'ENTER x: ' \)
            READ   (*, '(F10.5)') x

            z    =  AMOD (x, 6.2831853)
            sine =  z
            i    =  4.0
            next = -z * z * z / 6.0

       200  IF (ABS (next) .GE. epsi) THEN
               sine =  sine + next
               next = -next * z * z / (i * (i + 1.0))
               i    =  i + 2.0
               GOTO 200
            ENDIF

            WRITE  (*, 300) x, sine
       300  FORMAT (1X, 'THE SINE OF ', F10.5,
           +        ' = ', F12.10)
            END




 $IF


 ■  Action

 If expression is .TRUE., statements in the $IF block are compiled; if
 expression is .FALSE., control is transferred to the next $ELSE, $ELSEIF, or
 $ENDIF metacommand at the same $IF level


 ■  Syntax

      $IF expression

 Parameter                                  Description

 expression                                 A logical expression


 ■  Remarks

 The logical expression in a $IF metacommand can take two forms. One is the
 metacommand keyword DEFINED, followed by a symbol name in parentheses. If
 the symbol name has previously appeared in a $DEFINE metacommand without
 having been assigned a value, DEFINED (symbol_name) evaluates to .TRUE.. If
 (symbol_name) did not appear in a preceding $DEFINE metacommand, or did
 appear and was assigned a value, DEFINED (symbol_name) evaluates to .FALSE..

 The second form is a logical comparison, where the value of the symbol is
 compared with an integer constant or the value of another symbol using the
 FORTRAN .EQ., .NE., .GT., .LT., .GE., or .LE. operators. The results of such
 comparisons may further be evaluated with the FORTRAN .AND., .OR., .XOR.,
 and .NOT. operators. The usual rules of precedence apply, and parentheses
 may be used to control the order of evaluation.


 ■  Example

      $DEFINE flag
      $DEFINE testval = 3
      $IF DEFINED (flag)
            WRITE (*, '('' This is compiled if flag DEFINED'')')
      $ENDIF
      .
      .
      .
      $IF (testval .LT. 1) .AND. .NOT. DEFINED (flag)
            WRITE (*, '('' Testval .LT. 1 .AND. flag .NOT. DEFINED'')')


      $ELSE
            WRITE (*, '('' Testval .GE. 1 .OR.  flag       DEFINED'')')
      $ENDIF


 $INCLUDE


 ■  Action

 Inserts the contents of a specified text file at the location of the
 $INCLUDE metacommand


 ■  Syntax

      $INCLUDE:'filename'

 Parameter                                  Description

 filename                                   Name of the FORTRAN text file to
                                            include in the program


 ■  Remarks

 The argument filename must be a valid file specification for your operating
 system.

 The compiler considers the contents of the include file to be part of the
 program file and compiles them immediately. At the end of the included file,
 the compiler resumes processing the original source file at the line
 following the $INCLUDE metacommand.

 Include files are primarily used for data or program units that appear in
 more than one program. Include files most often contain subroutines and
 functions, common block declarations, and EXTERNAL, INTERFACE TO, and
 INTRINSIC statements.

 Include files can also contain other $INCLUDE metacommands and INCLUDE
 statements (nested included files). The compiler allows you to nest any
 combination of up to ten $INCLUDE metacommands or INCLUDE statements. Your
 operating system may impose further restrictions.


 ■  Example

 This program implements a stack by declaring the common stack data in an
 include file. The contents of the file STKVARS.FOR (shown below the
 following program) are inserted in the source code in place of every
 $INCLUDE metacommand. This guarantees all references to common storage for
 stack variables are consistent.

            INTEGER i
            REAL    x
      $INCLUDE:'stkvars.for'

      C     read in five real numbers:
            DO 100 i = 1, 5
               READ (*, '(F10.5)') x
               CALL Push (x)
        100 CONTINUE

      C     write out the numbers in reverse order:
            WRITE (*, *) ' '
            DO 200 i = 1,5
               CALL Pop (x)
               WRITE (*, *) x
        200 CONTINUE
            END

            SUBROUTINE Push (x)
      C     Pushes an element x onto the top of the stack.

            REAL x
      $INCLUDE:'stkvars.for'

            top = top + 1
            IF (top .GT. stacksize)  STOP 'Stack overflow'
            stack(top) = x
            END

            SUBROUTINE pop(x)
      C     Pops an element from the top of the stack into x.

            REAL x
      $INCLUDE:'stkvars.for'

            IF (top .LE. 0)  STOP 'Stack underflow'
               x   = stack(top)
               top = top - 1
            END

 The following is the file STKVARS.FOR:

      C     This file contains the declaration of the common block
      C     for a stack implementation. Because this file contains an
      C     assignment statement, it must be included only after all
      C     other specification statements in each program unit.

            REAL     stack(500)
            INTEGER  top, stacksize

            COMMON  /stackbl/ stack, top

            stacksize = 500


 $LARGE and $NOTLARGE


 ■  Action

 $LARGE specifies that an actual argument can span more than one segment
 (64K); $NOTLARGE specifies that an actual argument cannot span more than one
 segment


 ■  Syntax

      $«NOT»LARGE«:names»

 Parameter                                  Value

 names                                      One or more names of array
                                            variables or formal array
                                            arguments. If more than one name
                                            is specified, they must be
                                            separated by commas.

                                            When names is specified in the
                                            $LARGE metacommand, it indicates
                                            that the array or formal array
                                            argument specified can span more
                                            than one segment (it is addressed
                                            outside of DGROUP). When names is
                                            specified in the $NOTLARGE
                                            metacommand, it excludes the
                                            specified items from the effects
                                            of a $LARGE metacommand that has
                                            no arguments.


 ■  Remarks

 $NOTLARGE is the default.

 If the optional names parameter is specified, the metacommand must appear in
 the declarative section of a subprogram.

 If names is omitted, the metacommand affects all arrays in all subsequent
 subprograms in the source file until the $NOTLARGE metacommand is specified
 without any arguments. This form without arguments may appear anywhere
 except in the executable section of a subprogram.

 Arrays with explicit dimensions indicating they are bigger than 64K are
 automatically allocated to multiple segments outside the default data
 segment. You do not need to specify $LARGE for these arrays.

 Only one $LARGE or one $NOTLARGE metacommand without arguments can occur in
 a single program unit. The following code fragment, for example, is illegal:

      C     This is illegal:
      $LARGE
            SUBROUTINE MySub
      $NOTLARGE
            a=1.0
            .
            .
            .

 Note that use of the $LARGE metacommand on the entire program corresponds to
 the huge memory model. You can also use the HUGE attribute to specify that
 an actual argument can span more than one segment.


 $LINESIZE


 ■  Action

 $LINESIZE formats subsequent pages of the listing to a width of n columns


 ■  Syntax

      $LINESIZE:n

 Parameter                                  Value

 n                                          An integer between 80 and 132.
                                            The default for n is 80.


 ■  Example

      $LINESIZE:100

      C     The compiler listing of this program will be one hundred
      C     columns wide:

            REAL x
            x = 20
            WRITE (*, 100) x, SQRT(x)
        100 FORMAT(' The square root of ',f5.2,' is ',f7.4)
            END


 $LIST and $NOLIST


 ■  Action

 $LIST sends subsequent listing information to the listing file specified
 when starting the compiler; $NOLIST directs that subsequent listing
 information be discarded, until there is a subsequent appearance of the
 $LIST metacommand


 ■  Syntax

      $«NO»LIST


 ■  Remarks

 The default is $LIST.

 If no listing file is specified in response to the compiler prompt, the
 metacommand has no effect.

 $LIST and $NOLIST can appear anywhere in a source file.


 $LOOPOPT


 ■  Action

 $LOOPOPT turns on compiler loop optimization; $NOLOOPOPT disables it


 ■  Syntax

      $«NO»LOOPOPT


 ■  Remarks

 $LOOPOPT is the default. The $LOOPOPT metacommand reenables loop
 optimization after it has been disabled, either by the $NOLOOPOPT
 metacommand or by the /Od command-line option.

 Loop optimization consists of placing invariant expressions outside the
 loop, and simplifying computations within the loop to speed up loop
 calculations.

 Loop optimization is not always desirable. In the following example, the
 calculation of upper divided by lower is invariant, so loop optimization
 repositions it ahead of the loop. This means the IF statement can no longer
 check the value of the divisor before the division; if it is zero, a
 run-time error will occur.

            upper = result
            lower = quantity

            DO 200 n = 1, 100

            IF (lower .EQ. 0)
            GOTO 300 C

      C     Loop optimization moves the following statement outside
      C     the loop:

            factor = upper / lower
       200  CONTINUE
       300  CONTINUE

 Due to unanticipated interactions (such as the one just described), it is
 dangerous to turn loop optimization on and off locally. Loop optimization
 should be applied to, or removed from, entire programs or procedures.


 $MESSAGE


 ■  Action

 $MESSAGE sends a character string to the standard output device during the
 first compiler pass


 ■  Syntax

      $MESSAGE:string

 Parameter                                  Description

 string                                     A character constant


 ■  Remarks

 The string must be enclosed in quotes or apostrophes.


 ■  Example

      $MESSAGE:'Compiling program'


 $PACK


 ■  Action

 Controls the starting addresses of variables in structures


 ■  Syntax

      $PACK«:{1 | 2 | 4}»


 ■  Remarks

 If there is no $PACK metacommand in the file, structures are packed
 according to the following default rules: INTEGER*1, LOGICAL*1, and all
 CHARACTER variables begin at the next available byte, whether odd or even;
 and all other variables begin at the next even byte. This arrangement wastes
 some memory space, but gives the quickest access to structure elements.

 If $PACK:1 is specified, all variables begin at the next available byte,
 whether odd or even. Although this slightly increases access time, no memory
 space is wasted.

 If $PACK:2 is specified, packing follows the default rules described above.

 If $PACK:4 is specified, INTEGER*1, LOGICAL*1, and all CHARACTER variables
 begin at the next available byte, whether odd or even; all other variables
 begin on four-byte boundaries.

 If $PACK is specified (no colon or number), packing reverts to whatever mode
 was specified in the command-line /Zp option.

 The $PACK metacommand may appear anywhere in a program to change structure
 packing as desired.


 ■  Example

      $PACK
      .
      .
      .
      $PACK:1


 $PAGE


 ■  Action

 $PAGE starts a new page in the source-listing file


 ■  Syntax

      $PAGE


 ■  Remarks

 If the first character of a line of source text is the ASCII form-feed
 character (hexadecimal number 0C), it is the same as having a $PAGE
 metacommand before that line.


 ■  Example

      C     This is page one. The following metacommand starts
      C     a new page in the source listing file:
      $PAGE
      C     This is page two. The following line starts with the ASCII
      C     form-feed character, so it will also start a new page in the
      C     source listing file:

      C     This is page 3.
            STOP ' '
            END


 $PAGESIZE


 ■  Action

 $PAGESIZE formats subsequent pages of the source listing to a length of n
 lines


 ■  Syntax

      $PAGESIZE:n


 ■  Remarks

 The argument n must be at least 15. The default page size is 63 lines.


 ■  Example

      $PAGESIZE:60


 $STORAGE


 ■  Action

 $STORAGE allocates n bytes of memory for all variables declared as integer
 or logical variables


 ■  Syntax

      $STORAGE:n


 ■  Remarks

 The argument n must be either 2 or 4. The default is 4.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    On many microprocessors, the code required to perform 16-bit arithmetic
    is faster and more compact than the code required to perform 32-bit
    arithmetic. Therefore, unless you set the Microsoft FORTRAN $STORAGE
    metacommand to a value of 2, programs will default to 32-bit arithmetic
    and may run slower than expected. Setting the $STORAGE metacommand to 2
    allows programs to run faster and to be smaller.
 ───────────────────────────────────────────────────────────────────────────

 The $STORAGE metacommand does not affect the allocation of memory for
 variables declared with an explicit length specification, such as INTEGER*2
 or LOGICAL*4.

 If several files of a source program are compiled and linked together, be
 careful that they are consistent in their allocation of memory for variables
 (such as actual and formal parameters) referred to in more than one module.

 The $STORAGE metacommand must precede the first declaration statement of the
 source file in which it appears.

 The default allocation for INTEGER, LOGICAL, and REAL variables is 4 bytes.
 This results in INTEGER, LOGICAL, and REAL variables being allocated the
 same amount of memory, as required by the FORTRAN 77 standard.

 For information on how the $STORAGE metacommand affects arithmetic
 expressions, see Section 1.7.1.2, "Type Conversion of Arithmetic Operands."
 For information on how the $STORAGE metacommand affects the passing of
 integer arguments, see Section 2.5, "Arguments."


 ■  Example

      $STORAGE:2

      C     b and c are declared without a byte length, so they default
      C     to the $STORAGE value of 2 bytes. a and d will be 4 bytes.

            INTEGER*4 a, d
            INTEGER   b, c

            a = 65537
            b = 1024

      C     Since c is 2 bytes, it is assigned only the lower 2 bytes
      C     of a+b.
            c = a + b
            d = a + b

      C     The following statement produces:  1025        66561
            WRITE (*, *) c, d
            END


 $STRICT and $NOTSTRICT


 ■  Action

 $STRICT disables the specific Microsoft FORTRAN features not found in the
 FORTRAN 77 full-language standard, and $NOTSTRICT enables these features


 ■  Syntax

      $«NOT»STRICT


 ■  Remarks

 The default is $NOTSTRICT. $NOTSTRICT and $STRICT can appear anywhere in a
 source file, enabling and disabling the extensions as desired.

 All Microsoft FORTRAN features that are not in the FORTRAN 77 full-language
 standard are printed in blue. The $STRICT metacommand disables them.


 ■  Example

      $STRICT
      C     The following statement produces an error, because
      C     INTEGER*2 is not part of the FORTRAN 77 standard:
            INTEGER*2  i

      C     The variable name (balance) will be truncated to six
      C     characters:
            REAL  balance(500)

      C     The following statement produces an error, because the
      C     MODE= option is not part of the FORTRAN 77 standard:
            OPEN (2, FILE = 'BALANCE.DAT', MODE = 'READ')

            DO 100 i = 1, 500
      C     The following statement produces an error, because the EOF
      C     intrinsic function is not part of the FORTRAN 77 standard;
      C     (EOF is treated as a REAL function):

               IF (EOF (2))  GOTO 200
               READ (2, '(F7.2)') balance(i)
       100  CONTINUE
       200  CONTINUE
            END


 $SUBTITLE


 ■  Action

 $SUBTITLE assigns the specified subtitle for subsequent pages of the source
 listing


 ■  Syntax

      $SUBTITLE:subtitle

 Parameter                                  Description

 subtitle                                   Any valid character constant


 ■  Remarks

 If a program contains no $SUBTITLE metacommand, the subtitle is a null
 string. The value of the subtitle string is printed in the upper-left corner
 of the source-listing file pages, below the title, if any appears. For a
 subtitle to appear on a specific page of the source-listing file, the
 $SUBTITLE metacommand must be the first statement on that page.


 ■  Example

 The following program produces a listing in which each page is titled GAUSS
 (the name of the program). Each subprogram begins on a new page of the
 listing, and the name of the subprogram appears as a subtitle.

      $TITLE:'GAUSS'

      C     main program here...
            END

      $SUBTITLE:'Row Division'
      $PAGE
            SUBROUTINE divide (row, matrix, pivot)
      C     subroutine body...
               RETURN
            END

      $SUBTITLE:'Back Substitution'
      $PAGE
            SUBROUTINE BackSub (matrix)
      C     subroutine body...
               RETURN
            END


 $TITLE


 ■  Action

 $TITLE assigns the specified title for subsequent pages of the source
 listing (until overridden by another $TITLE metacommand)


 ■  Syntax

      $TITLE:title

 Parameter                                  Description

 title                                      Any valid character constant


 ■  Remarks

 If a program contains no $TITLE metacommand, the title is a null string.


 ■  Example

 The following program produces a listing in which each page is titled GAUSS
 (the name of the program). Each subprogram begins on a new page of the
 listing, and the name of the subprogram appears as a subtitle.

      $TITLE:'GAUSS'

      C     main program here...
            END

      $SUBTITLE:'Row Division'
      $PAGE
            SUBROUTINE divide(row, matrix, pivot)
      C     subroutine body...
               RETURN
            END

      $SUBTITLE:'Back Substitution'
      $PAGE
            SUBROUTINE backsub(matrix)
      C     subroutine body...
               RETURN
            END


 $TRUNCATE and $NOTRUNATE


 ■  Action

 $TRUNCATE truncates all variable and program/subprogram names to six
 characters; $NOTRUNCATE disables the default or a previous $TRUNCATE
 metacommand


 ■  Syntax

      $«NO»TRUNCATE


 ■  Remarks

 The default is $NOTRUNCATE. When the default is in effect, the first 31
 characters in a name are significant. (Your operating system may also
 restrict the length of names.)

 When $TRUNCATE or $STRICT is in effect, names longer than six characters
 generate warning messages. This can make it easier to port your code to
 other systems.


 ■  Example

      C     This program produces the following output:
      C
      C     74 Las Vegas Street
      C     74 Las Vegas Street
      C
      C     Barry Floyd
      C     3 Prospect Drive


            IMPLICIT CHARACTER*20 (s)

      $TRUNCATE
            studentname    = 'Enrique Pieras'
            studentaddress = '74 Las Vegas Street'

            WRITE (*, 100) studentname, studentaddress

      $NOTRUNCATE
            studentname    = 'Barry Floyd'
            studentaddress = '3 Prospect Drive'

            WRITE (*, 100) studentname, studentaddress

        100 FORMAT (/ 1X, A20, / 1X, A20)

            END


 Chapter 7  The FL Command

 ───────────────────────────────────────────────────────────────────────────

 The FL command automatically compiles and links your FORTRAN program in a
 single step. FL executes the three compiler passes, then calls LINK (the
 Microsoft Segmented-Executable Linker) to link the object files. Except for
 a few unusual situations, FL is the only command you need to compile and
 link your FORTRAN source files. This chapter explains how it is used.

 When used by itself, the FL command creates an .EXE file that runs on most
 PC-compatible computers. In addition, the FL command offers a wide range of
 options, with which you can do the following:

   1. Set the memory model

   2. Select how floating-point calculations are performed

   3. Control which FORTRAN language features are available

   4. Create overlay files

   5. Create programs that run under OS/2

   6. Optimize the program for minimum size or maximum speed

   7. Assemble and link machine-language routines automatically

   8. Establish the search path for files

   9. Switch compiler warning messages on and off

  10. Select the page format and headers for output listings

 This chapter explains what each option does and how it is used.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    This chapter assumes you have a basic knowledge of FORTRAN and know how
    to create and edit program files. For questions relating to the
    definition of the FORTRAN language, see  specific chapters in this
    manual. The Microsoft CodeView and Utilities User's Guide explains how to
    use the symbolic debugger provided with this package.
    For a quick introduction to running the compiler and linker with FL, see
    the Microsoft FORTRAN Getting Started manual.
 ───────────────────────────────────────────────────────────────────────────

 Section 7.1 explains how to use the FL command to produce an executable
 program from a FORTRAN source file. It also describes how to run the program
 and pass command-line arguments to it, if desired.

 The remainder of the chapter discusses commonly used FL options. The FL
 options that control floating-point operations and memory models are
 discussed in Microsoft FORTRAN Advanced Topics. A summary of the FL commands
 and all available options appears in the Microsoft FORTRAN Quick Reference
 Guide provided with this package.


 7.1  The Basics: Compiling, Linking, and Running FORTRAN Files

 This section explains how to use FL to compile and link FORTRAN files, and
 it discusses the rules and conventions that apply to the file names and
 options used with FL. It also explains how to run the executable program
 created by FL.


 7.1.1  Compiling and Linking with FL

 The FL command takes the following form:

      FL «option...» «filespec...» «option...» «filespec...»
      «/link «libfield» «linkoptions»» «/MA option »

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Syntax examples too long to fit on one line are continued on two or more
    lines.
 ───────────────────────────────────────────────────────────────────────────

 Each option is one of the command-line options described in this manual, and
 each filespec names a file to be processed. The FL command automatically
 specifies the appropriate library to be used during linking; however, you
 can use the /link option with the optional libfield and linkoptions to
 specify additional libraries and options to be used during linking. See
 Section 7.32, "Using FL to Link without Compiling," for more information.

 You can give any number of options and file names on the command line,
 provided the command line does not exceed 128 characters.

 The FL command can process source files, object files, or a combination of
 source and object files. It uses the file-name extension (the period plus
 any letters following it) to determine what kind of processing the file
 needs, as shown below:

   ■  If the file has a .FOR extension, FL compiles the file.

   ■  If the file has an .OBJ extension, FL invokes the linker.

   ■  If the extension is omitted or is anything other than .FOR or .OBJ, FL
      assumes the file is an object file unless the file name is part of a
      /Tf option. If the file name appears in a /Tf option, FL assumes the
      file is a FORTRAN source file. See Section 7.10 for a description of
      the /Tf option.

 FL also accepts files with the .LIB and .DEF extensions and passes them to
 the linker in the correct order. Files that end with the .ASM extension or
 that appear in a /Ta option are passed to MASM.

 You can use the DOS wild-card characters (* and ?) to process all files
 meeting the wild-card specification, as long as the files have the required
 extensions. See the DOS manual for more information on wild-card characters.

 Any filespec on the FL command line can include a full or partial path
 specification, allowing you to process files in different directories or on
 different drives. A full path specification starts with the drive name. A
 partial path specification gives one or more directory names before the name
 of the file, but does not give a drive name. If a full path is not given, FL
 assumes the path starts from the current working directory.

 File names can contain any combination of uppercase or lowercase letters.
 For example, the following three file names are equivalent:

      abcde.for
      ABCDE.FOR
      aBcDe.foR

 When FL compiles source files, it creates object files. By default, these
 object files have the same base names as the corresponding source files, but
 with the extension .OBJ instead of .FOR. (The base name of a file is the
 portion of the name preceding the period, but excluding the path and drive
 name, if any.) You can use the /Fo option to assign a different name to an
 object file.

 These object files, along with any .OBJ files given on the command line, are
 linked to form an executable program file. The executable file has the base
 name of the first file (source or object) on the command line, plus an .EXE
 extension. If only .OBJ files are given on the command line, the compilation
 stage is skipped altogether, and the files are simply linked.

 You can tell whether FL is compiling or linking by the messages on the
 screen. When FL invokes the compiler, a message similar to the following
 appears:

      Microsoft (R) FORTRAN Optimizing Compiler Version 5.00
      Copyright (C) Microsoft Corp 1989.  All rights reserved.

 As each source file on the command line is compiled, its name appears on the
 screen. When all source files have been compiled and the linker is invoked,
 a message similar to the following appears:

      Microsoft (R) Segmented-Executable Linker  Version 5.03
      Copyright (C) Microsoft Corp 1984-1989. All rights reserved.

 This message is followed by several lines showing Microsoft LINK prompts and
 the responses provided by FL. The FL command uses the response-file method
 of invoking Microsoft LINK.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    FL ignores case in the linking stage. For example, the linker regards
    Global and GLOBAL as the same symbol. If you do not wish to ignore case,
    you must include the option, /link  /NOI (NOIGNORECASE). See Chapter 13,
    "Linking Object Files with LINK," in the Microsoft CodeView and Utilities
    User's Guide for more information.
 ───────────────────────────────────────────────────────────────────────────


 7.1.2  Using FL Options

 The FL command offers many command-line options to control and modify the
 compiler's operation. Options begin with a slash (/) and contain one or more
 letters. You can use a dash (-) instead of a slash if you prefer. For
 example, /I and -I are both acceptable forms of the I option.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Although file names can be either uppercase or lowercase, options must be
    given exactly as shown. For example, /Zd is a valid option, but /ZD and
    /zd are not.
 ───────────────────────────────────────────────────────────────────────────

 Options can appear anywhere on the FL command line. In general, an option
 applies to all files following it on the command line, and does not affect
 files preceding it. However, not all options obey this rule (see the
 discussion of a particular option for information on its behavior). Most FL
 options apply only to the compilation process. Unless otherwise noted,
 options do not affect object files on the command line.

 Some options take arguments, such as file names, strings, or numbers. In
 some cases, spaces are permitted between the option and its argument; check
 the syntax of each option to confirm this.

 Some options consist of more than one letter. When this is the case, no
 spaces are allowed between the letters of an option. For example, the /Sp
 option cannot appear as /S p or a command-line error occurs.

 Chapters 1 and 2 of Microsoft FORTRAN Advanced Topics describe FL options
 for floating-point operations and memory models, respectively. Additional
 linking options are explained in Chapter 13 of the Microsoft CodeView and
 Utilities User's Guide.


 7.1.3  The FL Environment Variable

 The FL environment variable is a convenient way to specify frequently used
 options and files without having to enter them manually. If the FL
 environment variable exists, the options specified in the variable are
 automatically added to the command line.

 Use the SET command to set the FL environment variable. It is usually easier
 to add the SET command to the AUTOEXEC.BAT file than to enter it manually
 when you reboot. For example, the following SET command sets FL so programs
 are compiled and linked for use with the Microsoft CodeView debugger:

      SET FL = /Zi /Od

 All options in the FL variable are processed before the options on the
 command line, with the exception of /link options. After all other options
 are processed, any /link options in the FL variable are processed, followed
 by the /link options on the command line. For example, if the FL environment
 variable is set to /Zi /Od /link /I, the following two commands are
 equivalent:

      FL         MAINPGM.FOR MODULE1.FOR /link    /NOP
      FL /Zi /Od MAINPGM.FOR MODULE1.FOR /link /I /NOP

 When conflicting options appear on the same command line, the last option
 usually takes precedence. Therefore, since options specified in the FL
 variable are processed first, they will be overridden by conflicting options
 on the com-mand line.


 7.1.4  Specifying the Next Compiler Pass

 If FL cannot locate the next compiler pass it needs (there are three), it
 prompts you for the complete path name. (This occurs when compiling from
 floppy disks, or if you have changed the file name of one of the passes.) If
 the pass is on another disk, insert it in any available drive. Type the full
 path name, then press ENTER.


 7.1.5  Stopping FL

 If you need to abort the compiling and linking session, press CTRL+C or
 CTRL+BREAK. You will be returned to the DOS command level, where you can
 restart FL.


 7.1.6  Using the FL Command (Specific Examples)

 The command line below compiles the files A.FOR and B.FOR, creating object
 files named A.OBJ and B.OBJ:

      FL  A.FOR  B.FOR  C.OBJ  D

 These object files are then linked with C.OBJ and D.OBJ to form an
 executable file named A.EXE (since the base name of the first file on the
 command line is A.) Note that the extension .OBJ is assumed for D since no
 extension is given on the command line.

 The command line below compiles all source files with the default extension
 (.FOR) in the current working directory:

      FL  *.FOR

 The resulting object files are linked to form an executable file whose base
 name is the same as the base name of the first file compiled.

 The command below links all object files with the default extension (.OBJ)
 in the current working directory, creating an executable file whose base
 name is the same as the base name of the first object file found.

      FL  *.OBJ

 This will usually be the first file listed in the disk directory, and not
 necessarily the first file alphabetically.


 7.1.7  Running Your FORTRAN Program

 Compiling and linking a program produces an executable file with the
 extension .EXE. This file can be run from the operating system. The PATH
 environment variable is used to find executable files. You can execute your
 program from any directory, as long as the executable program file is either
 in your current working directory or in one of the directories given in the
 PATH environment variable.

 The command line that executes the program can also include additional file
 names. These names are used to satisfy OPEN statements in your program that
 leave the file-name field blank. The first file name on the command line is
 used for the first such OPEN statement executed, the second file name is
 used for the second OPEN statement, and so on.

 These command-line file names and any other arguments are also available by
 calling the GETARG procedure. You can use the command-line file names in
 OPEN statements without filenames, and still retrieve them with GETARG. See
 Chapter 5, "Intrinsic Functions and Additional Procedures," for more
 information. There is no interaction between these uses, and none of the
 names are changed or deleted.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    If your program executes a READ or WRITE statement specifying a file that
    has not been opened, the effect is the same as that of an OPEN statement
    with a blank file name. Default values are assigned to the parameters
    normally given in the OPEN statement. This operation is called an
    "implicit OPEN."
 ───────────────────────────────────────────────────────────────────────────

 If the file names on the command line outnumber the OPEN statements with
 blank file names, the extra file names are ignored. However, the remaining
 arguments can still be accessed with the GETARG procedure.

 If more OPEN statements with blank file names are executed than there are
 file names on the command line, you will be prompted to enter a file name
 for each additional OPEN statement. You are also prompted if you give a null
 file name (see the example below).

 Each file name on the command line must be separated from the names around
 it by one or more spaces or tab characters. Each name can be enclosed in
 quotation marks ("filename") if desired, but this is not required. A null
 argument consists of an empty set of double quotation marks, with no file
 name enclosed ("").

 The following example runs the program MYPROG.EXE:

      MYPROG  ""  OUTPUT.DAT

 Since the first file-name argument is null, the first OPEN statement with a
 blank file-name field produces the following message:

      File name missing or blank - please enter file name
      UNIT number ?

 The number is the unit number specified in the OPEN statement. The file name
 OUTPUT.DAT is used for the second such OPEN statement executed. If
 additional OPEN statements with blank file-name fields are executed, you
 will be prompted for more file names.


 7.2  Getting Help with FL Options (/HELP)


 Option

      /HELP
      /help

 The /HELP option displays a list of the most commonly used FORTRAN options.
 (See the Microsoft FORTRAN Quick Reference Guide for a complete alphabetical
 listing of FL options.) For this option to work, the file containing the
 FORTRAN options, FL.HLP, must be in the current directory or in the path
 given in the PATH environment variable. If FL cannot find this file, it
 displays the following error message:

      cannot open help file, 'fl.hlp'

 This option is not case sensitive. Any combination of uppercase and
 lowercase letters will work.

 When the /HELP option appears on the FL command line, FL displays the list
 of options but does not take any other action, regardless of what other
 options appear on the command line. For example, if you give a source-file
 name along with the /HELP option, FL does not compile the source file.

 The help screen prompts you to press any key before returning to the
 operating system prompt. This keeps the top lines of the help screen in
 view. Once you press a key and return to the operating system prompt, the
 top lines scroll out of view.


 Examples

 The following examples show how you can save the help screen for future
 reference by sending it to a file or printer:

      FL  /HELP > HELP.DOC

      FL  /HELP >PRN

 The first example above saves the help screen in a file named HELP.DOC. In
 the second example, the screen output is sent directly to the printer
 device, PRN. (See Section 7.15, "Special File Names," or your operating
 system documentation for a list of device names that can be used in
 redirection.)

 Note that you may have to press ENTER several times to make sure that all
 help messages are saved or printed. Since the messages may be displayed on
 several separate screens, FL waits for you to enter a keystroke before
 displaying the next screenful of messages. Also, you must press an
 additional key (any key can be used, including ENTER) after giving the FL
 command, since the help screen requires you to press a key before returning
 to the operating system prompt.


 7.3  Floating-Point Options (/FP)

 The following options control the way your program handles floating-point
 operations:

 Option                          Effect

 /FPi                            Generates in-line instructions
                                 and selects the emulator math
                                 package

 /FPi87                          Generates in-line instructions
                                 and selects 8087/287/387 commands

 /FPc                            Generates floating-point calls
                                 and selects the emulator math
                                 package

 /FPc87                          Generates floating-point calls
                                 and selects 8087/80287 commands

 /FPa                            Generates floating-point calls
                                 and selects the alternate math
                                 package

 See Chapter 1, "Controlling Floating-Point Operations," in Microsoft FORTRAN
 Advanced Topics for a complete description of these options.

 By default, programs use the 8087/80287 coprocessor. If your computer does
 not have a coprocessor, you must specify a floating-point option that uses
 either the emulator or the alternate-math package. (The emulator package
 exactly duplicates the operations of the 8087/80287 coprocessor in software.
 The alternate-math package is not as accurate as the 8087/80287 emulation,
 nor does it adhere to the IEEE floating-point standards, but it does execute
 faster.)

 The Microsoft FORTRAN Compiler does not generate true in-line 8087/80287
 code when you use the /FPi87 option. Instead, the compiler inserts software
 interrupts to library code, which fixes each interrupt to use either the
 emulator or coprocessor, as appropriate. If you have force-linked the
 emulator code to your program, these interrupts will call the emulator if a
 math chip is not available.

 If your program is designed to run only with a coprocessor, you can speed up
 the program by assembling and linking the code below, which  eliminates the
 fix-up processing. In addition, all divide-by-zero errors will be masked.

      PUBLIC  FIARQQ, FICRQQ, FIDRQQ, FIERQQ, FISRQQ, FIWRQQ
      PUBLIC  FJARQQ, FJCRQQ, FJSRQQ

      FIARQQ    EQU  0
      FICRQQ    EQU  0
      FIDRQQ    EQU  0
      FIERQQ    EQU  0
      FISRQQ    EQU  0
      FIWRQQ    EQU  0
      FJARQQ    EQU  0
      FJCRQQ    EQU  0
      FJSRQQ    EQU  0

      extrn   __fpmath:for
      extrn   __fpsignal:for
      extrn   __fptaskdata:for

      CDATA   segment  word  common 'DATA'
              dw   0
              dw   __fpmath
              dw   __fpsignal
              dw   __fptaskdata
      CDATA   ends

      end


 7.4  Memory-Model Options (/A, /M)

 The /A option specifies the program's memory model. The memory model defines
 the rules the compiler uses to organize the program's code and data segments
 in memory. The following standard memory models are available:

 Option                          Effect

 /AL                             Chooses the large memory model
                                 (default)

 /AM                             Chooses the medium memory model

 /AH                             Chooses the huge memory model

 See Chapter 2, "Selecting a Memory Model," in Microsoft FORTRAN Advanced
 Topics for a more complete description of these options and the memory
 models they specify.

 The floating-point and memory-model options you choose determine the
 standard library name that FL places in the object file it creates. This
 library is the one the linker searches for by default. Table 7.1 shows each
 combination of memory-model and floating-point options and the corresponding
 library name that FL places in the object file.

 Table 7.1  FL Options and Default Libraries

 Floating-Point       Memory-Model     Default
 Option               Option           Library

 /FPi87 or /FPc87     /AL or /AH       LLIBFOR7.LIB
                      /AM              MLIBFOR7.LIB
 /FPi or /FPc         /AL or /AH       LLIBFORE.LIB
                      /AM              MLIBFORE.LIB
 /FPa                 /AL or /AH       LLIBFORA.LIB
                      /AM              MLIBFORA.LIB


 The /M option supports the use of dynamic-link and multithread applications
 in OS/2. These switches cannot be used with the /Am option since threads and
 dynamic-link libraries are not supported in medium model.

 Option                          Use

 /MT                             Multithread applications

 /MD                             Dynamic-link library and
                                 multithread applications

 When the /MT option is used, the program is automatically linked with
 LLIBFMT.LIB, regardless of the memory model. When the /MD option is used,
 all default libraries are overridden and you must create your own library,
 which is then specified on the command line (or when using LINK).

 Both the /MT and /MD option imply the /G2 (286 instruction set) option,
 since multithread and dynamic-link library programs run only under OS/2.
 They also imply the /FPi87 (math coprocessor) option. (This default may be
 overridden by any other floating-point option except /FPa, which, when used,
 causes a command-line error.) In addition, the compiler assumes that SS is
 not equal to DS (the stack may be in a segment other than the default data
 segment).

 Both the /MT and /MD options are incompatible with /Fb (bound) and /Lr or
 /Lc (real mode), since threaded and dynamic-link library applications are
 assumed to be unbound programs running in protected mode.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    If you renamed any of the libraries created while running SETUP, the
    library name embedded in the object file might not match the renamed
    library (as, for example, when you use the new default names). In these
    cases, you must explicitly specify the new library name to the linker.
    See Section 7.30, "Linking with Libraries," for more information.
 ───────────────────────────────────────────────────────────────────────────


 7.5  OS/2 Library Options (/Lp, /Lr, /Lc)

 If you set up the FORTRAN compiler to use DOS, or real-mode or
 protected-mode OS/2 (but not both modes), you do not need to specify the
 run-time library on the FL command line.

 If, however, you created FORTRAN libraries for both real- and protected-mode
 OS/2, you need to tell the FL command whether to use the real-mode or the
 protected-mode library.

 The following options are used to specify OS/2 libraries:

 Option                          Effect

 /Lp                             Specifies protected-mode
                                 libraries

 /Lr                             Specifies real-mode libraries

 /Lc                             Same as /Lr, but used by IBM C
                                 compiler; provided for
                                 compatibility

 The example below compiles and links the source file QSORT.FOR to create a
 real-mode program that can run under both DOS and the OS/2 compatibility
 box:

      FL  /Lr  QSORT.FOR

 (The /Lc option could have been used instead, with the same result.)

 The next example creates a protected-mode .EXE file that runs only under
 OS/2 protected mode:

      FL  /Lp  QSORT.FOR

 When you set up for only one operating system or system mode, SETUP creates
 only one library. The library has the same name, regardless of whether you
 set up for MS-DOS, OS/2 real mode, or OS/2 protected mode. The name has the
 following form:

 memory LIBFOR arith .LIB

 The memory argument is a single character indicating the memory model: L for
 large model, M for medium model, or H for huge model. The arith argument is
 a single character indicting the math option: 7 for math coprocessor
 arithmetic, A for arithmetic using the alternate-math package, or E for
 emulated coprocessor arithmetic. For example, the library LLIBFORA.LIB uses
 the large memory model and the alternate math package.

 When you set up for both OS/2 modes, SETUP creates two libraries, with names
 that have the following form:

 memory LIBF arith system .LIB

 The memory and arith arguments are the same as in the single operating
 system names. The system argument is a single character indicating the
 operating system mode R for OS/2 real-mode (real-mode libraries can also be
 used under DOS) and P for OS/2 protected-mode. A library for protected-mode
 OS/2 using the large memory model and the alternate-math package would have
 the name LLIBFAP.LIB.

 Remember that when you are set up for both real- and protected-mode OS/2,
 you must specify one of the /L options to be sure that the correct library
 is linked.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    There are other situations (besides being set up for dual-mode OS/2) when
    you must specify a link library other than the default library. Section
    7.30, "Linking with Libraries," explains how you can override the default
    library selection and specify another library.
 ───────────────────────────────────────────────────────────────────────────


 7.6  Data Threshold Option (/Gt)


 Option

 /Gt«number»

 The /Gt option sets the data threshold, a cutoff value the compiler uses
 when allocating data.

 In the medium memory model, the compiler allocates all static and global
 data items within the default data segment. In the large and huge models,
 the compiler allocates only initialized static and global data items to the
 default data segment.

 The /Gt option can be used only with large- and huge-model programs, since
 medium-model programs have only one data segment. This option is
 particularly useful with programs that have more than 64K of initialized
 static and global data as small data items.

 The /Gt option causes all data items whose size is greater than or equal to
 number bytes to be allocated to a new data segment outside the default data
 segment that is accessed with a far address. When the /Gt option is omitted,
 the default threshold value is 32,767 bytes. When the /Gt option is given,
 but no number is specified, the default threshold value is 256 bytes. When
 number is specified, it must immediately follow /Gt, with no intervening
 spaces.


 7.7  Naming and Organizing Segments (/ND, /NM, /NT)


 Options

 /NT textsegment /ND datasegment /NM textsegment

 An object-code file is sometimes referred to as a "module." A module can
 contain a main program, one or more procedures, or any combination of a main
 program and procedures.

 A "segment" is a contiguous block of code or data produced by the compiler.
 Every module has at least two segments: a text segment containing the
 program code (machine instructions), and a data segment containing the
 program data.

 All segments have names. Text segment and data segment names are normally
 created by the compiler. These default names depend on the memory model
 chosen. For example, in medium-model programs, the text segment from each is
 placed in a separate segment with a distinct name, formed from the module
 base name and the suffix _TEXT. The single data segment is named _DATA.

 In large- and huge-model programs, the text and data from each module are
 loaded into separate segments with distinct names. Each text segment is
 given the name of the module plus the suffix _TEXT. The data from each
 segment is placed in a private segment with a unique name (except for
 initialized global and static data that are placed in the default data
 segment). The naming conventions for text and data segments are summarized
 in Table 7.2.

 Table 7.2  Segment-Naming Conventions

 Model      Text            Data         Module
 Medium     module_TEXT     _DATA        filename
 Large      module_TEXT     _DATA       filename
 Huge       module_TEXT     _DATA       filename


 The linker uses segment names to define the order in which the segments of
 the program appear in memory when loaded for execution. (The segments in the
 group named DGROUP are an exception.) Segments with the same name (both
 module and suffix) are loaded into the same physical memory segment.

 The /NT (nametext) and /ND (namedata) options override the default names
 supplied by the compiler. This allows you, for example, to give two
 different modules the same text-segment name, ensuring that both are loaded
 into the same memory segment. This is useful when you wish to give a
 procedure the NEAR attribute, because it means the contents of the code
 segment (CS) register need not be changed when the procedure is called.

 If you change the name of the default data segment using /ND, there is no
 guarantee the data segment (DS) register will have the same value as the
 stack segment (SS) register. In this case, the compiler automatically
 generates code to load the DS register with the correct data-segment value
 on entry to the code and restores the previous value upon exit.

 The textsegment and datasegment arguments can be any combination of letters
 and digits.

 In previous versions of FL, the /NM option changed the name of a module. In
 the version supplied with FORTRAN 5.0, /NM is identical to the /NT option.


 7.8  Creating Bound Program Files (/Fb)

 Option                          Effect

 /Fb                             Creates a bound .EXE file with
                                 the source file base name

 /Fbfilename                     Creates a bound file named
                                 filename.EXE and a protected-mode
                                 .EXE file with the source file
                                 base name

 To create a bound program that can run under OS/2 or DOS, use a command of
 the following type:

      FL  /Lp  /Fb  QSORT.FOR

 To create both a protected-mode .EXE file and a bound .EXE file, specify a
 file name that is different from the base name of the source file:

      FL  /Lp  /FbQSORTB  QSORT.FOR

 In the preceding example, the protected-mode program QSORT.EXE is created
 first, then the bound program QSORTB.EXE is created. If the name given in
 the /Fb option is the same as the base name of the source file, the bound
 program file overwrites the protected-mode program that was created first.

 In order for the /Fb option to work, the files DOSCALLS.LIB, APILMR.OBJ, and
 API.LIB must be in the current directory or in one of the directories
 specified in the LIB environment variable.

 The file APILMR.OBJ is automatically bound with the program. If your
 medium-model program does not use allocatable arrays, and if you do not wish
 to reserve a full 64K DGROUP, run the BIND command manually, specifying the
 /n switch and omitting APILMR.OBJ. See the Microsoft CodeView and Utilities
 User's Guide for information about the BIND command.

 If you use the /Fm (map) option with the /Fb option, the resulting map file
 contains the map from BIND, rather than the linker map.

 When you bind mixed-language executable files that have non-Family API
 calls, you must execute the BIND command directly and specify the /n switch.
 See Chapter 4, "Mixed-Language Programming," of Microsoft FORTRAN Advanced
 Topics for more information.


 7.9  FORTRAN-Specific Options (/4Y, /4N)

 The following options begin with the numeral 4, indicating they are specific
 to FORTRAN 77. The letter Y (yes) indicates the option is enabled, and the
 letter N (no) indicates it is disabled.

 More than one /4 option may appear on the command line. An option applies to
 all the files succeeding it, until a conflicting or canceling option removes
 its effect for any remaining files. A contradictory metacommand within a
 file overrides the effect of a command-line option for that file only.

 The /4 options may be combined. For example, /4Ysfb disables all Microsoft
 FORTRAN extensions, permits free-form entry, and enables extended error
 handling. Note that groups of Y and N options must appear as separate
 command-line options; they cannot be mixed in a single option.


 7.9.1  Controlling Optional Language Features (/4Ys, /4Yi, /4Yv)

 Option                          Effect

 /4{Y | N}s                      Disables (Y) or enables (N) all
                                 Microsoft FORTRAN extensions

 /4{Y | N}i                      Enables (Y) only SAA extensions;
                                 disables (N) full Microsoft
                                 FORTRAN extensions

 /4{Y | N}v                      Enables (Y) only VAX extensions;
                                 disables (N) full Microsoft
                                 FORTRAN extensions

 Microsoft FORTRAN includes many optional features that are not part of
 ANSI-standard FORTRAN. All these extensions are normally available by
 default. To disable all of them, use the /4Ys command-line switch. Any
 language feature which is not part of the strict FORTRAN 77 definition is
 then tagged by the compiler as an error. (The /4Ys option is equivalent to
 the $STRICT metacommand at the beginning of the file.)

 The /4Ys option applies to all source-code files following it on the command
 line, unless the /4Ns option disables it for any of the remaining files. The
 $NOTSTRICT metacommand within a file overrides the /4Ys command option for
 that file only. The $STRICT metacommand within a file overrides the /4Ns
 command option for that file only.

 Many Microsoft FORTRAN extensions fall into one of two categories: IBM SAA
 (Systems Application Architecture) extensions, and DEC(R) VAX extensions.

 To simplify porting code from an SAA or VAX environment to a personal
 computer, you can disable all Microsoft FORTRAN extensions except the SAA
 extensions, or disable all Microsoft FORTRAN extensions except the VAX
 extensions. (The compiler recognizes only the strict FORTRAN 77 language
 plus either the SAA or the VAX extensions, but not both.)

 This is done with the /4Yi and /4Yv options, respectively. Either option
 applies to all files following it on the command line, until the /4Ni or
 /4Nv option reverts to using only Microsoft FORTRAN extensions.

 Microsoft FORTRAN includes all IBM SAA extensions, which are listed below:

   ■  31-character names

   ■  Bit-manipulation intrinsic

   ■  Case-insensitive source

   ■  COMMON allows character and noncharacter in same block

   ■  CONJG, HFIX, and IMAG intrinsic functions

   ■  Data initialization in type statements

   ■  EQUIVALENCE allows association of character and noncharacter

   ■  Functions IOR, IAND, NOT, IEOR, ISHFT, BTEST, IBSET, and IBCLR

   ■  IMPLICIT NONE

   ■  INCLUDE compiler directive

   ■  INTEGER*2, COMPLEX*16, and LOGICAL*1 data types

   ■  Optional length specifications in INTEGER, REAL, COMPLEX, and LOGICAL
      type statements

   ■  Use of underscore (_) in names

   ■  Z edit descriptor

 Microsoft FORTRAN includes many (but not all) VAX extensions, which are
 listed below:

   ■  31-character names

   ■  ACCESS selector 'APPEND' in the OPEN statement

   ■  Allowing integer arrays to contain FORMAT statements

   ■  BLOCKSIZE and NML I/O keywords

   ■  Debug comment lines

   ■  DO statements without specified labels

   ■  DO WHILE statement

   ■  END DO statement

   ■  EQUIVALENCE of character and noncharacter items

   ■  EQUIVALENCE to a multi-dimensioned array with only one subscript

   ■  Exponentiation of REAL and COMPLEX statements

   ■  IMPLICIT NONE

   ■  INCLUDE compiler directive

   ■  Initialization on the declaration line

   ■  In-line comments

   ■  Length specification within the FUNCTION statement

   ■  Length specifications within type declarations

   ■  Mixing of character and noncharacter items in COMMON statements

   ■  Noninteger alternate return values

   ■  Noninteger array subscripts

   ■  Numeric operands for .AND., .OR., etc.

   ■  Specified common-block variables in DATA statements outside a BLOCK
      DATA subroutine

   ■  STRUCTURE, UNION, MAP statements

   ■  Up to 99 continuation lines

   ■  Use of dollar sign ($) in names

   ■  .XOR. operator


 7.9.2  Controlling Source-File Syntax (/4Yf, /4Nf, /4Yt, /4Nt, /4Y6, /4N6)

 Option                          Effect

 /4{Y | N}f                      Enables (Y) or disables (N)
                                 free-form format

 /4{Y | N}t                      Enables (Y) or disables (N)
                                 truncation of variable names

 /4{Y | N}6                      Enables (Y) or disables (N)
                                 FORTRAN 66-style DO statements

 These options control the structure of a FORTRAN source file and two
 elements of FORTRAN syntax. They correspond to FORTRAN metacommands which
 are described in detail in Chapter 6, "Metacommands." The following list
 gives the metacommand corresponding to each option and identifies the
 option's default:

 Option                          Metacommand

 /4Yf                            $FREEFORM

 /4Nf (default)                  $NOFREEFORM

 /4Yt                            $TRUNCATE

 /4Nt (default)                  $NOTRUNCATE

 /4Y6                            $DO66

 /4N6 (default)                  None

 Each option has the same effect as placing the corresponding metacommand at
 the beginning of the source file. A conflicting metacommand in the source
 file overrides the command-line option for that file only. No error occurs
 if an option is used with a file that already contains the corresponding
 metacommand.


 Examples

      FL  /c  /4Yds  TEST.FOR  /4Nd  STABLE.FOR

 The command line above causes FL to compile TEST.FOR using truncation
 (disallowing all Microsoft extensions). The declare option is also enabled,
 so use of undeclared variables produces warning messages. When the second
 file, STABLE.FOR, is compiled, the truncation is still in effect, but the
 declare option is disabled.

      FL  /4Yf  /4Nt  *.FOR

 The command line above enables free-form format and disables truncation of
 variable names when compiling and linking all source files in the current
 working directory.


 7.9.2.1  The Debug Option


 Option

 /4{Y | N}b

 The debug option controls extended error handling at run time. It provides
 information to be used by the error-handling system in the program file. See
 the discussion of the $DEBUG metacommand in Chapter 6 for a description of
 the types of errors that are detected in extended error handling. When the
 debug option is enabled, loop optimization in the program is disabled.

 Debugging is enabled by giving the /4Yb option, and disabled with /4Nb. By
 default, debugging is disabled.

 The debug option has the same effect as a $DEBUG or $NODEBUG metacommand at
 the beginning of the source file being compiled. If a $DEBUG or $NODEBUG
 metacommand appears later in the file being compiled, debugging for that
 file is enabled or disabled, as appropriate.

 The /4Yb option does not accept a string argument for conditional
 compilation. Use the /4cc option, described in Section 7.9.5, instead.


 Examples

      FL  MAIN.FOR  /4Yb  /Fs  TEST.FOR

 The example above compiles and links two files. Debugging is enabled for
 TEST.FOR, and a source listing named TEST.LST is created. Neither the
 debugging option nor the source-listing option applies to MAIN.FOR.

      FL  /c  /4Ybd  ONE.FOR  /4Nd  TWO.FOR

 The example above compiles ONE.FOR with both the debug and declare options
 enabled. (The following section describes the declare option.) The declare
 option is disabled when compiling TWO.FOR, but the debug option is still in
 effect.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    The /4Yb and /Zi options are incompatible. See Section 7.22 for more
    information on /Zi .
 ───────────────────────────────────────────────────────────────────────────


 7.9.2.2  The Declare Option


 Option

 /4{Y | N}d

 The declare option controls warnings about undeclared variables. When the
 declare option is enabled, the compiler generates a warning message at the
 first use of any variable which has not been declared in a type statement.

 The /4Yd option enables the declare option and /4Nd disables it. The declare
 option is disabled by default (unless a $DECLARE metacommand is in the
 source file).

 The declare compiler option has the same effect as a $DECLARE metacom-mand
 or a $NODECLARE metacommand at the beginning of each source file. If
 $DECLARE or $NODECLARE metacommands appear later in the file, warnings are
 enabled or disabled, as appropriate. Note that if the source file contains a
 $DECLARE or $NODECLARE metacommand at the beginning, the /4Yd or /4Nd option
 has no effect.


 Examples

      FL  /4Ybd  *.FOR > DECLARE

 The example above compiles and links all source files with the default
 extension (.FOR) in the current working directory. The debug and declare
 options are both enabled. All messages (including warnings about undeclared
 variables) are redirected to the file DECLARE.

      FL  /4Yb  ONE.FOR  /4Yd  TWO.FOR

 The example above turns on debugging for both ONE.FOR and TWO.FOR; the
 declare option is also enabled for TWO.FOR.


 7.9.3  Automatic Variables


 Option

 /4{Y | N}a

 The /4Ya option causes all eligible variables in the succeeding files to be
 automatic. They are declared on the stack. (In Microsoft FORTRAN, the
 default is that all variables are static. They have defined memory
 locations.) The /4Na option disables automatic variables (except as declared
 within the program) for succeeding files. For more information, see the
 AUTOMATIC entry in Section 4.2, "Statement Directory."


 7.9.4  Setting the Default Integer Size (/4I2, /4I4)


 Option

 /4I{ 2 | 4}

 The /4I option allocates either 2 or 4 bytes of memory for all variables
 declared in the source file as INTEGER or LOGICAL variables. The default
 allocation is 4 bytes. The /4I option applies to the remainder of the
 command line or until another /4I option appears.


 Example

      FL  /4I2  /FeTESTPROG  *.FOR

 This example allocates 2 bytes of memory for INTEGER and LOGICAL variables
 when compiling and linking all source files in the current working
 directory. The executable file is named TESTPROG.EXE.

 The /4I option has the same effect as a $STORAGE metacommand at the top of
 each file that is being compiled. If a $STORAGE metacommand already appears
 in the file being compiled, the size given by the metacommand is used. The
 $STORAGE metacommand in a particular file affects only that file and does
 not change the effects of /4I on any other files on the command line. See
 Chapter 6,"Metacommands," for more information on the $STORAGE metacommand.


 7.9.5  Conditional Compilation (/4cc, /D)


 Options

 /4ccstring /Dsymbol«=val»

 The /4cc option permits conditional compilation of a source file. The string
 is a set of alphabetic characters controlling which lines in the source file
 are to be compiled.

 Any line with a letter in column 1 found in string is compiled; lines
 beginning with other letters are treated as comments. All lines not
 beginning with letters are compiled normally. Case is not significant. The
 letter must appear in column 1 of the source-file line.

 The string can be enclosed in quotation marks ("") if desired, but the
 quotation marks are not required.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Program lines with the character C or c in column 1 are always treated as
    comments.
 ───────────────────────────────────────────────────────────────────────────


 Example

      FL  /c  /4ccXYZ  PRELIM.FOR

 In this example, all lines in the file PRELIM.FOR beginning with  X, Y, or Z
 are compiled.

 The /D option defines a symbolic name used by the conditional compilation
 metacommands. It is equivalent to the $DEFINE metacommand within the program
 file. The name must immediately follow /D, with no intervening space.

 An equal sign and integer value may be added. This assigns a specific value
 to the symbol that the conditional-compilation metacommands can check.


 Examples

      FL /Dtestprocs foo.for
      FL /Dwhichpart=4 foo.for

 In the first example, the symbol testprocs is defined. The compiler will
 then evaluate any DEFINED  (testprocs)  expression in the
 conditional-compilation metacommands as true (unless an  $UNDEFINE testprocs
 metacommand in the program file cancels the definition).

 In the second example, the symbol whichpart is defined and given the integer
 value four.


 7.10  Specifying Source Files (/Tf, /Ta)


 Options

 /Tf« »sourcefile /Ta« »sourcefile

 The FL command assumes that any filename ending in .FOR is a FORTRAN source
 file, and invokes the compiler. Likewise, the FL command assumes that any
 filename ending in .ASM is an assembly source file, and tries to invoke the
 Microsoft assembler. If your files do not have these identifying extensions,
 you can use the /Tf and /Ta options to specify that the following filename
 is a FORTRAN or assembly file, respectively.

 If you have to specify more than one source file with an extension other
 than .FOR or .ASM, it is safest to list each source file in a separate /Tf
 or /Ta option. Although a sourcefile with a wild-card character is legal,
 this use of wild-card characters may cause problems. If a sourcefile with a
 wild-card character represents a single file, then FL behaves as expected:
 it considers that single file to be a FORTRAN (or assembly-language) source
 file. However, if a sourcefile with a wild-card character represents more
 than one file, FL treats only the first file as a FORTRAN (or
 assembly-language) source file. It treats any other files that sourcefile
 represents as object files. Masm must be in the path where the FL command
 can find it.


 Examples

      FL  MAIN.FOR  /TfTEST.PRG  /TfCOLLATE.PRG  PRINT.PRG

 In the example above, the FL command compiles the three source files
 MAIN.FOR, TEST.PRG, and COLLATE.PRG. Since the file PRINT.PRG is given
 without a /Tf option, FL treats it as an object file. Thus, after compiling
 the three source files, FL links the object files MAIN.OBJ, TEST.OBJ,
 COLLATE.OBJ, and PRINT.PRG.

      FL  /TfTEST?.F

 For the example above, assume the files TEST1.F, TEST2.F, and TEST3.F all
 exist in the current directory. In this example, the FL command would
 compile TEST1.F as a FORTRAN program and then try to treat TEST2.F and
 TEST3.F as object files. The FL command shown above would have the same
 effect as the following FL command:

      FL  /TfTEST1.F  TEST2.F  TEST3.F

 If your program is compiled using the /Tf option, you must explicitly select
 the FORTRAN expression evaluator when using the CodeView debugger. CodeView
 only uses the FORTRAN evaluator as the default when the source file has the
 .FOR extension.

 The /Ta option is used with the /MA option. The /MA option specifies that
 the following file is a Microsoft Assembler (MASM) file, and the file is to
 be assembled and linked with the program. If the file name does not end in
 .ASM, you must use the /Ta option to specify the full file name. The space
 between /Ta and sourcefile is optional.


 7.11  Compiling without Linking (/c)


 Option

 /c

 The /c (compile-only) option suppresses linking. Source files are compiled,
 but the resulting object files are not linked, no executable file is
 created, and any object files specified are ignored. This option is useful
 when you are compiling individual source files that do not make up a
 complete program.

 The /c option applies to the entire FL command line, regardless of the
 option's position on the command line.


 Example

      FL  /c *.FOR

 This command line compiles, but does not link, all files with the extension
 .FOR in the current working directory.


 7.12  Naming the Object File (/Fo)


 Option

 /Foobjfile

 By default, FL gives each object file the same base name as the
 corresponding source file, plus the extension .OBJ. The /Fo option gives an
 object file a different name or creates it in a different directory.

 The objfile argument must appear immediately after the option, with no
 intervening spaces. The objfile argument can be a file specification, a
 drive name, or a path specification.

 If objfile is a file specification, the /Fo option applies only to the
 source file immediately following the option on the command line. The object
 file created by compiling that source file has the name given by objfile.

 If objfile is a drive name or path specification, the FL command creates
 object files in the given location for every source file following the /Fo
 option on the command line. The object files take their default names (each
 object file has the base name of the corresponding source file).

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    When you give only a path specification, the objfile argument must end
    with a backslash ( \ ) so that FL can distinguish between it and an
    ordinary file name.
 ───────────────────────────────────────────────────────────────────────────

 You can use any name or extension you like for objfile. However, it is
 recommended that you use the conventional .OBJ extension because the FL
 command, as well as the LINK and LIB utilities, expects .OBJ when processing
 object files. If you choose an object-file name that lacks an extension, FL
 automatically adds the .OBJ extension. However, if you pick a name with a
 blank extension (a name ending in a period), FL does not add an extension.


 Examples

      FL  /c  /FoSUB\THAT  THIS.FOR

 The example above compiles the file THIS.FOR and creates an object file
 named THAT.OBJ in the subdirectory SUB. Note that FL automatically appends
 the .OBJ extension. Linking is suppressed because the /c option is given.

      FL  /FoB:\OBJECT\  *.FOR

 The example above compiles and links all source files with the extension
 .FOR in the current working directory. The option /FoB:\OBJECT\  tells FL to
 create all the object files in the directory named OBJECT on drive B. Each
 object file has the base name of the corresponding source file, plus the
 exten-sion .OBJ.


 7.13  Naming the Executable File (/Fe)


 Option

 /Feexefile

 By default, the executable file produced by the FL command is given the base
 name of the first file (source or object) on the command line, plus the
 extension .EXE. The /Fe option gives the executable file a different name or
 creates it in a different directory.

 Since only one executable file is created, it does not matter where the /Fe
 option appears on the command line. If the /Fe option appears more than
 once, the last name prevails.

 /Fe applies only during linking. If /c suppresses linking, /Fe has no
 effect.

 The exefile argument must appear immediately after the option, with no
 intervening spaces. The exefile argument can be a file specification, a
 drive name, or a path specification.

 If exefile is a file specification, the executable file is given the
 specified name. If exefile is a drive name or path specification, the FL
 command creates the executable file in the given location, using the default
 name (the base name of the first file plus .EXE).

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    When you give a path specification as the exefile argument, the path
    specification must end with a backslash ( \ ) so that FL can distinguish
    it from an ordinary file name.
 ───────────────────────────────────────────────────────────────────────────

 You can give any name or extension you like for the exefile. If you give a
 file name without an extension, FL automatically appends the .EXE extension.
 If you pick a name with a blank extension (a name ending in a period), FL
 does not add an extension.


 Examples

      FL  /FeC:\BIN\PROCESS  *.FOR

 The example above compiles and links all source files in the current working
 directory with the extension .FOR. The resulting executable file is named
 PROCESS.EXE and is created in the directory C:\BIN.

      FL  /FeC:\BIN\  *.FOR

 The example above is similar to the first example except that the executable
 file, instead of being named PROCESS.EXE, is given the same base name as the
 first file compiled. The executable file is created in the directory C:\BIN.


 7.14  Creating Listing Files (/F)

 The FL command offers a number of listing options. You can create a source
 listing, a map listing, or one of several kinds of object listings. You can
 also set the title and subtitle of the source listing and control the length
 of source-listing lines and pages.

 The FL command optimizes by default, so object listings reflect the
 optimized code. Since optimization may involve rearranging the code, the
 correspondence between your source file and the machine instructions may not
 be clear, especially when you use the /Fc option to mingle the source and
 assembly code. To suppress optimization, use the /Od option, discussed in
 Section 7.22.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Listings may contain names that begin with one or more underscores (for
    example, __chkstk) or that end with the suffix QQ. Names that use these
    conventions are reserved for internal use by the compiler.
 ───────────────────────────────────────────────────────────────────────────

 The following is a list of options that produce listings and control their
 appearances:

 Option                  Effect

 /Fs«listfile»           Produces source listing

 /Fl«listfile»           Produces object listing

 /Fa«listfile»           Produces assembly listing

 /Fc«listfile»           Produces combined source and
                         object listing

 /Fm«mapfile»            Produces map file that lists
                         segments in order

 This section describes how to use command-line options to create listings.
 For an example of each type of listing and a description of the information
 it contains, see Section 7.18, "Formats for Listings."

 The options in this section require the listfile or mapfile argument (if
 given) to follow the option immediately, with no intervening spaces. The
 listfile can be a file specification, a drive name, or a path specification.
 It can also be omitted.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    When you give only a path specification as the listfile argument, the
    path specification must end with a backslash ( \ ) so that FL can
    distinguish it from an ordinary file name.
 ───────────────────────────────────────────────────────────────────────────

 When you give a drive name or path specification as the argument to a
 listing option, or if you omit the argument altogether, FL uses the base
 name of the source file, plus the default extensions, to create a default
 file name. Table 7.3 lists the default file names and default extensions,
 which are used when you give a file-name argument that lacks an extension.

 Table 7.3  Default File Names and Extensions

                   Default           Default
 Option            Listing Type      File Name                 Extension

 /Fs               Source            Base name of source        .LST
                                     file plus .LST

 /Fl               Object            Base name of source        .COD
                                     file plus .COD

 /Fa               Assembly          Base name of source        .ASM
                                     file plus .ASM

 /Fc               Combined          Base name of source        .COD
                   source-object     file plus .COD

 /Fm               Map               Base name of first         .MAP
                                     source or object file
                                     on the command line
                                     plus .MAP


 Since you can process more than one file at a time with the FL command, the
 order in which you give listing options and the kind of argument you give
 for each option (file specification, path specification, or drive name) are
 significant. Table 7.4 summarizes the effects of each option with each type
 of argument.

 Table 7.4  Arguments to Listing Options

                Drive-Name
                File-Name                  or Path                    No
 Option         Argument                   Argument                  Argument

 /Fa, /Fc,      Creates a listing for      Creates listings in        Creates
 /Fl, /Fs       next source file on        the given location for     in the c
                command line; uses         every source file          director
                default extension if       listed after the           every so
                no extension is            option on the command      file lis
                supplied                   line; uses default         after th
                                           names                      on the c
                                                                      line; us
                                                                      default

 /Fm            Uses given file name       Creates map file in        Uses def
                for the map file; uses     the given directory;       name
                default extension if       uses default name
                no extension is
                supplied


 If a source file includes one or more $NOLIST metacommands, the portion of
 that source file between each $NOLIST metacommand and the following $LIST
 metacommand (if any) is omitted from the listing.

 Only one variation of the object or assembly listing can be produced for
 each source file. The /Fc option overrides the /Fa and /Fl options; whenever
 you use /Fc, a combined listing is produced. If you apply both the /Fa and
 the /Fl options to one source file, only the last listing specified is
 produced.

 The map file is produced during the linking stage. If linking is suppressed
 with the /c option, the /Fm option has no effect.


 7.15  Special File Names

 You can use the DOS device names listed below as file-name arguments to the
 listing options. These special names send listing files to your terminal or
 printer.

 Name                          Device

 AUX                           An auxiliary device.

 CON                           The console (terminal).

 PRN                           The printer.

 NUL                           Specifies a null (nonexistent)
                               file. Giving NUL as a file name
                               means that no file is created.

 Even if you add device designations or file-name extensions to these special
 file names, they remain associated with the devices listed above. For
 example, A:CON.tuv still refers to the console and is not the name of a disk
 file.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    When using device names, do not add a colon. The Microsoft FORTRAN
    Compiler does not recognize the colon. For example, use CON or PRN, not
    CON: or PRN:.
 ───────────────────────────────────────────────────────────────────────────


 7.16  Line Size (/Sl) and Page Size (/Sp)


 Options

 /Sl« »linesize /Sp« »pagesize

 The /Sl and /Sp options control the line size and page size of source
 listings, respectively. The default line size for source listings is 79
 columns, and the page size is 63 lines. These options are useful for
 preparing source listings that will be printed on a printer that uses
 nonstandard page sizes. They are valid only if you also specify the /Fs
 option on the FL command line. The space between /Sl and linesize, or /Sp
 and pagesize, is optional.

 The linesize argument gives the width of the listing line in columns (on
 line printers, columns usually correspond to characters). The number given
 must be a positive integer between 79 and 132, inclusive; any number outside
 this range produces an error message. Any line that exceeds the listing
 width is truncated.

 The pagesize argument gives the number of lines on each page of the listing.
 The minimum number is 15; if a smaller number is given, an error message
 appears and the default page size is used.

 The /Sl or /Sp option applies to the remainder of the command line or until
 the next occurrence of /Sl or /Sp on the command line. These options do not
 cause source listings to be created. They take effect only when the /Fs
 option is also given to create a source listing.

 You can use metacommands in the source file to override the /Sl and /Sp
 options. These options have the same effects as $LINESIZE and $PAGESIZE
 metacommands at the top of each file being compiled. If additional $LINESIZE
 or $PAGESIZE metacommands appear in the file being compiled, the line size
 or page size for that file is changed accordingly.

 The $LINESIZE or $PAGESIZE metacommands in a particular file affect only
 that file and do not change the effects of /Sl or /Sp on any other files on
 the command line.


 Examples

      FL  /c  /Fs  /Sl90  /Sp70  *.FOR

 The example above compiles all source files with the default extension
 (.FOR) in the current working directory, creating a source-listing file for
 each source file. Each page of the source listing is 90 columns wide and 70
 lines long.

      FL  /Fs  /Sp70  MAIN.FOR  /Sp63  SUB1.FOR  SUB2.FOR

 The example above compiles and links three source files, creating an
 executable file named MAIN.EXE. Three source listings are created: each page
 of MAIN.LST is 70 lines long, while each page of SUB1.LST and SUB2.LST is 63
 lines long.


 7.17  Titles (/St) and Subtitles (/Ss)


 Options

 /St« »"title"
 /Ss« »"subtitle"

 The /St and /Ss options set the title and subtitle, respectively, for source
 listings. The quotation marks around the title or subtitle argument can be
 omitted if the title or subtitle does not contain tabs or spaces. The space
 between /St and "title", or /Ss and "subtitle", is optional.

 The title appears in the upper-left corner of each page of the source
 listing. The subtitle appears below the title.

 The /St or /Ss option applies to the remainder of the command line or until
 the next occurrence of /St or /Ss on the command line. These options do not
 cause source listings to be created. They take effect only when the /Fs
 option is also used to create a source listing.

 Both the /St and /Ss options can be overridden by metacommands in the source
 file. These options have the same effect as $TITLE and $SUBTITLE
 metacommands at the beginning of the file being compiled. If additional
 $TITLE or $SUBTITLE metacommands appear in the file, the title or subtitle
 is changed accordingly.

 The $TITLE or $SUBTITLE metacommands in a particular file affect only that
 file and do not change the effects of /St or /Ss on any other files on the
 command line.


 Examples

      FL  /St"INCOME TAX"  /Ss15-APR  /Fs  TAX*.FOR

 The example above compiles and links all source files beginning with TAX and
 ending with the default extension (.FOR) in the current working directory.
 Each page of the source listing contains the title INCOME TAX in the
 upper-left corner. The subtitle 15-APR appears below the title on each page.

      FL  /c  /Fs  /St"CALC PROG"  /Ss"COUNT"  CT.FOR  /Ss"SORT"  SRT.FOR

 The example above compiles two source files and creates two source listings.
 Each source listing has a unique subtitle, but both listings have the title
 CALC PROG.


 7.18  Formats for Listings

 This section describes and gives examples of the five types of listings
 available with the FL command. See Section 7.14, "Creating Listing Files,"
 for information on how to create these listings.


 Source Listing

 Source listings are useful in debugging programs during development. These
 listings can also document the structure of a finished program.

 The source listing contains the numbered source-code lines of each procedure
 in the source file, along with expanded include files and any error
 messages. If the source file compiles with no errors more serious than
 warnings, the source listing also includes tables of local symbols, global
 symbols, and parameter symbols for each procedure. If the compiler is unable
 to finish compilation, it does not generate symbol tables.

 At the end of the source listing is a summary of segment sizes. This summary
 is useful when analyzing memory requirements.

 Error messages appear in the listing after the line that caused the error,
 as shown in the following example:

      9   hyp = SQRT ((sidea**2) + (sideb**2)
      ***** sqroot.for(9) : error F2115: syntax error

 The line number in the error message corresponds to the number of the source
 line immediately above the message in the source listing.

 The example below shows the source listing for a simple FORTRAN program:

      PAGE   1

      10-17-86

      18:20:36

        Line#  Source Line            Microsoft FORTRAN Compiler Version 5.00

            1   common a
            2   dimension a(10)
            3   real x
            4   complex c
            5   real *8 d
            6   complex *16 e
            7   character *50 f
            8   integer*2 j
            9   parameter (d=123456789.00056, e=-(.00000122, 1234354 e5))
           10   parameter (f='Note that character strings will be truncated')
           11   parameter (x=1.2345)
           12   parameter (c=(.12345, 123456.789), i = 123, j = 100)
           13   end

      main  Local Symbols

      Name                  Class   Type              Size   Offset

      A . . . . . . . . . . COMMQQ  REAL*4              40    0000


      Parameter Symbols        Type                Value

      X . . . . . . . . . . REAL*4            1.2345001E+000
      C . . . . . . . . . . COMPLEX*8       ( 1.2345000E-001,
                                              1.2345679E+005)
      D . . . . . . . . . . REAL*8            1.2345679E+008
      E . . . . . . . . . . COMPLEX*16      (-1.2199999E-006,-1.2343540E+011)
      F . . . . . . . . . . CHARACTER       Note that character
                                            strings will
      J . . . . . . . . . . INTEGER*2                     100
      I . . . . . . . . . . INTEGER*4                     123



      Global Symbols

      Name                      Class   Type              Size   Offset

      COMMQQ. . . . . . . . . . common                      40    0000
      main. . . . . . . . . . . FSUBRT  ***                ***    0000

      Code size = 0018 (24)
      Data size = 0000 (0)
      Bss size  = 0000 (0)

      No errors detected

 The Name column lists each global symbol, external symbol, and statically
 allocated variable declared in the source file. The  Parameter Symbols
 column lists each symbolic constant defined in a PARAMETER statement.

 For items other than functions and subroutines, the Class column contains
 either global, local, equiv, common, or extern, depending on how the symbol
 was defined. For functions and subroutines, the Class column contains the
 abbreviations in the following list:

 Type                                       Abbreviation

 Far function                               FFUNCT

 Near function                              NFUNCT

 Far subroutine                             FSUBRT

 Near subroutine                            NSUBRT

 The Type column lists a simplified version of the symbol's type as declared
 in the source file. The Type entry for functions is the type declared in the
 source file.

 The Size column is used only for variables. This column specifies the number
 of bytes allocated for the variable. Note that the allocation for an
 external array may be unknown, so its Size field may be undefined.

 The Offset column is used only for symbols with a global  or local entry in
 the Class field. For variables, the Offset column gives the variable's
 relative offset in the logical data segment for the program file. (Even if a
 program file contains more than one compilation unit, all data is allocated
 in the same data segment.) Since the linker generally combines several
 logical data segments into a physical segment, this number is useful only
 for determining a variable's relative storage position.

 The Value field appears only for parameter symbols. It lists the value of
 each symbolic constant. Character constants longer than 33 characters are
 truncated.

 The last table in the source listing shows the segments used and their size,
 as shown below:

      Code size = 0095 (149)
      Data size = 003c (60)
      Bss size  = 0000 (0)

 The byte size of each segment is given first in hexadecimal, and then in
 decimal (in parentheses). For information on specific segment usage by
 different memory models, see Chapter 2, "Selecting a Memory Model," in
 Microsoft FORTRAN Advanced Topics.


 Object-Listing File

 The object-listing file contains the machine instructions and assembly code
 for your program. The line numbers appear as comments, as shown below:

      SQRT_TEXT      SEGMENT
      ; Line 6
           PUBLIC    _main
      _main     PROC FAR
           *** 000000     55                  push bp
           *** 000001     8b ec               mov  bp,sp
           *** 000003     b8 02 00            mov  ax,2
           *** 000006     9a 00 00 00 00      call __chkstk
           *** 00000b     9b d9 06 00 00      fld  $T20002
           *** 000010     9b d9 1e 02 00      fstp $S14_SIDEA
      ; Line 7
           *** 000015     9b d9 06 04 00      fld  $T20003
           *** 00001a     9b d9 1e 06 00      fstp $S15_SIDEB
      ; Line 9
           *** 00001f     9b d9 06 08 00      fld  $T20004
           *** 000024     9a 00 00 00 00      call __FIsqrt
           *** 000029     9b d9 1e 0a 00      fstp $S16_HYP
           *** 00002e     90 9b               fwait


 Assembly-Listing File

 The assembly-listing file contains the assembly code corresponding to your
 program file, as shown below:

      SQRT_TEXT      SEGMENT
      ; Line 6
           PUBLIC    _main
      _main     PROC FAR
           push bp
           mov  bp,sp
           mov  ax,2
           call __chkstk
           fld  $T20002
           fstp $S14_SIDEA
      ; Line 7
           fld  $T20003
           fstp $S15_SIDEB
      ; Line 9
           fld  $T20004
           call __FIsqrt
           fstp $S16_HYP
           fwait
      .
      .
      .
      _main     ENDP
      SQRT_TEXT ENDS
      END

 Note that this sample shows the same code as the object listing sample,
 except the machine instructions are omitted. This ensures the listing is
 suitable as input for the Microsoft Macro Assembler (MASM).


 Combined Source and Object Listing

 The combined source and object listing shows one line of your source program
 followed by the corresponding line (or lines) of machine instructions, as in
 the following sample:

      SQRT_TEXT      SEGMENT
      ;|*** c   This program calculates the length of the hypotenuse of a
      ;|*** c   right triangle given the lengths of the other two sides.
      ;|***
      ;|***     real sidea, sideb, hyp
      ;|***
      ;|***     sidea = 3.
      ; Line 6
           PUBLIC    _main
      _main     PROC FAR
           *** 000000     55                  push bp
           *** 000001     8b ec               mov  bp,sp
           *** 000003     b8 02 00            mov  ax,2
           *** 000006     9a 00 00 00 00      call __chkstk
           *** 00000b     9b d9 06 00 00      fld  $T20002
           *** 000010     9b d9 1e 02 00      fstp $S14_SIDEA
      ;|***     sideb = 4.
      ; Line 7
           *** 000015     9b d9 06 04 00      fld  $T20003
           *** 00001a     9b d9 1e 06 00      fstp $S15_SIDEB
      ;|***
      ;|***     hyp = sqrt(sidea**2 + sideb**2)
      ; Line 9
           *** 00001f     9b d9 06 08 00      fld  $T20004
           *** 000024     9a 00 00 00 00      call __FIsqrt
           *** 000029     9b d9 1e 0a 00      fstp $S16_HYP
           *** 00002e     90 9b               fwait
      ;|***
      ;|***     write(*,100) hyp
      .
      .
      .
      _main     ENDP
      SQRT_TEXT ENDS
      END
      ;|***

 Note that this sample is like the object-listing sample, except the program
 source line is provided in addition to the line number.


 Map File

 The map file contains a list of segments in order of their appearance within
 the load module. An example is shown below:

      Start  Stop   Length Name                   Class
      00000H 00059H 0005AH SQRT_TEXT              CODE
      0005AH 018E1H 01888H _TEXT                  CODE
      .
      .
      .

 The information in the Start and Stop columns shows the 20-bit address (in
 hexadecimal) of each segment, relative to the beginning of the load module.
 The load module begins at location zero. The  Length column gives the length
 of the segment in bytes. The  Name column gives the name of the segment, and
 the Class column gives information about the segment type.

 The starting address and name of each group appear after the list of
 segments. A sample group listing is shown below:

      Origin   Group
      0643:0   DGROUP

 In the example above, DGROUP is the name of the data group. DGROUP is the
 only group used by most programs compiled with the Microsoft FORTRAN
 Compiler, Version 5.0. (Multithread and DLL applications use other groups.)

 The map file shown below contains two lists of global symbols: the first
 list is sorted by symbol address and the second is sorted alphabetically by
 symbol name. The notation Abs appears next to absolute symbol names (symbols
 containing 16-bit constant values that are not associated with program
 addresses).

 Many of the global symbols that appear in the map file are symbols used
 internally by the FORTRAN compiler. These symbols usually begin with one or
 two leading underscores or end with QQ.

      Address         Publics by Name

      0005:1594       $i8_output
      0005:1855       $i8_tpwr10
      0000:FE32  Abs  FIARQQ
      0000:0E32  Abs  FICRQQ
      0000:5C32  Abs  FIDRQQ
      0000:1632  Abs  FIERQQ
      0000:0632  Abs  FISRQQ
      0000:A23D  Abs  FIWRQQ
      0000:4000  Abs  FJARQQ
      0000:C000  Abs  FJCRQQ
      0000:8000  Abs  FJSRQQ
      018E:190B       ICLRER
      018E:1932       IGETER
      0643:058D       OFF_ARGPTR
      0643:058B       OFF_DESCPT
      0643:00F4       STKHQQ
      0005:0885       _access
      0005:106F       _brkctl
      0005:091A       _chsize
      .
      .
      .
      0643:00F0       __aaltstkovr
      0643:0278       __abrkp
      0643:0228       __abrktb
      0643:0278       __abrktbe
      .
      .
      .

 Finally, the map file gives the program entry point:

      Program entry point at 0005:03C9


 7.19  Searching for Include Files (/I, /X)


 Options

 /Idirectory «/Idirectory...»
 /X

 The /I and /X options temporarily override the environment variable INCLUDE.
 These options give special handling to a particular file or files, without
 changing the normal compiler environment.

 The /I (include) option causes the compiler to search the directory or
 directories you specify before it searches the standard places given by the
 INCLUDE environment variable.

 You can search more than one include-file directory by giving the /I option
 more than once. The directories are searched in order of their appearance on
 the command line. Each occurrence of an /I option applies only to source
 files following the option.

 The directories are searched only until the include file specified in the
 source file is found. If the file cannot be found, the compiler prints an
 error message and stops processing. When this occurs you must restart
 compilation with a corrected directory specification.

 The following list describes the compiler's search order for include files:

   1. All the directories in the "parent" file's path. The parent file is the
      file containing the INCLUDE statement or $INCLUDE metacommand. For
      example, if a file named FILE1 includes a file named FILE2, FILE1 is
      the parent file of FILE2.

      Include files can be nested; thus, in the preceding example, FILE2 can
      include another file named FILE3. In this case, FILE1 is said to be the
      "grandparent" file of FILE3. For nested include files, the search
      begins with the directories in the parent file's path, then proceeds
      through the directories of each of its grandparent files. (See the
      examples below for an illustration of this procedure.)

   2. The directories specified in each /I option.

   3. The places specified in the INCLUDE environment variable.

 The INCLUDE statement or $INCLUDE metacommand may give a full or partial
 path specification for the file. (A full path specification starts with the
 drive name; a partial path specification gives one or more directory names
 before the name of the file, but no drive name.) If a full path
 specification is given for the include file, the compiler uses the given
 path to find the file, and the INCLUDE environment variable and any /I
 options have no effect. If a partial path specification is given, the
 compiler attempts to find that path, starting from the parent file's
 directory, then from the grandparent file's directories, then from the
 directories given on the command line, and finally from the directories
 given by the INCLUDE environment variable.

 The /X (exclude) option stops the compiler from searching the standard
 places given by the INCLUDE environment variable. When /X is given, FL
 considers the list of standard places to be empty. The parent and
 grandparent directories are still searched, however.

 Like the /I option, /X applies only to source files following the option on
 the command line. The /X option can be followed by one or more /I options.
 This causes the compiler to search only the parent and grandparent
 directories and the directories given by the /I options, ignoring the
 standard places.


 Examples

      FL /IC:\TESTDIR /IC:\PREVIOUS *.FOR

 The example above assumes the INCLUDE environment variable is set to
 C:\FOR\INCLUDE. It compiles all source files with the default extension
 (.FOR) in the current working directory, searching for include files in the
 following order:

   1. The current working directory

   2. \TESTDIR,  the first directory on the command line

   3. \PREVIOUS,  the second directory on the command line

   4. \FOR\INCLUDE,  the directory given by the INCLUDE environment variable

 However, if the metacommand $INCLUDE:'\SUB\DEFS' is contained in one of the
 source files, the compiler adds the subdirectory  \SUB to the end of each
 path it searches. Thus, the search for the include file named DEFS proceeds
 in the following order:

   1. The current working directory (which contains any parent source files)

   2. The \SUB subdirectory of the current working directory

   3. \TESTDIR,  the first directory on the command line

   4. \PREVIOUS,  the second directory on the command line

   5. \FOR\INCLUDE,  the directory given by the INCLUDE environment variable

 The following example assumes that the INCLUDE environment variable is set
 to C:\FOR\INCLUDE:

      FL ..\TESTS\*.FOR

 It compiles all source files with the default extension (.FOR) in the
 directory named ..\TESTS, searching directories for include files in the
 following order:

   1. ..\TESTS,  the directory containing any possible parent files

   2. \FOR\INCLUDE,  the directory given by the INCLUDE environment variable

 However, if one of the source files in the directory ..\TESTS contains the
 metacommand $INCLUDE:'\SUB\DEFS', the compiler adds the subdirectory \SUB to
 the end of each path it searches. Thus, the search for the include file
 named DEFS proceeds in the following order:

   1. ..\TESTS\SUB,  adding the subdirectory \SUB to the directory ..\TESTS,
      where ..\TESTS is the directory containing the parent source file

   2. \FOR\INCLUDE\SUB,  adding the subdirectory \SUB to the directory
      \FOR\INCLUDE,  where \FOR\INCLUDE is the directory given by the INCLUDE
      environment variable

 If the file \SUB\DEFS contains the metacommand $INCLUDE:'COMS', the compiler
 searches directories for the nested include file named COMS in the following
 order:

   1. ..\TESTS\SUB,  the directory containing DEFS,  the parent file of the
      file named COM

   2. ..\TESTS,  the directory containing the grandparent source file of the
      file named COM

   3. ..\FOR\INCLUDE,  the directory given by the INCLUDE environment
      variable

 In this last case, since COMS is not specified as part of another
 subdirectory, no subdirectory is added to the end of the path specified in
 the INCLUDE environment variable.

 The search ends when the file is found.

 The following example uses a combination of the /I and /X options to control
 the search path:

      FL  MAIN.FOR  /X  /ITEST1  SUB1.FOR  /ITEST2  SUB2.FOR

 Since no /I option appears before MAIN.FOR on the command line, the compiler
 searches for any files included by MAIN.FOR in the standard places defined
 by the INCLUDE environment variable (after searching the parent file's
 directory). Since the /X option precedes the next file name, SUB1.FOR, the
 compiler does not search the standard places for any files SUB1.FOR includes
 (in this case, the environment variable is not used). Instead, only the
 directory of the parent source file SUB2.FOR and the directory TEST1  are
 searched. If the include file or files cannot be found in one of those
 places, an error occurs. The second /I option adds one more directory to be
 searched for any include files specified in the parent file SUB2.FOR. The
 TEST2 subdirectory is searched after the TEST1 subdirectory.


 7.20  Handling Warnings and Errors

 There are several kinds of errors that can occur when a program is compiled,
 linked, and run. Section 7.20.1 gives an overview of Microsoft FORTRAN error
 messages. Several options are available to control the types of warnings
 generated at compile time and to enable or disable expanded error handling
 at run time. See Section 7.20.2 for a description of these options.


 7.20.1  Understanding Error Messages

 Errors can occur at any stage of program development, as explained below:

   1. During compilation, the compiler generates a broad range of error and
      warning messages to help you locate errors and potential problems in
      your source files.

   2. While linking, the linker is responsible for generating error messages.

   3. When a program is executed, any error messages are run-time error
      messages. This category includes messages about floating-point
      exceptions, which are errors generated by an 8087, 80287, or 80387
      coprocessor.

 Other utilities included in this package, such as NMAKE and EXEMOD, generate
 their own error messages. You can also distinguish the type of a message by
 its format. See Appendix D, "Error Messages," for a description of
 error-message formats, a list of actual error messages, and explanations of
 the circumstances that cause them.

 When you are compiling and linking using the FL command, you may see both
 compiler and linker messages. The LINK program banner appears on the screen
 when linking begins. Compiler messages are any messages that appear before
 the LINK banner, and linker messages are those that appear after the banner.
 Compiler messages have numbers preceded by the letter F, and linker messages
 have numbers preceded by the letter L.

 Compiler error messages are sent to standard output, which is usually your
 terminal. You can redirect the messages to a file or printer by using one of
 the DOS redirection symbols: > or >>.

 Error redirection is useful in batch-file processing. For example, the
 following command redirects error messages to the printer device (designated
 by PRN):

      FL  /c  COUNT.FOR > PRN

 See Section 7.15, "Special File Names," or your DOS documentation for a list
 of device names, including PRN.

 In the following command, only output that ordinarily goes to the console
 screen is redirected.

      FL  COUNT.FOR > COUNT.ERR

 The FL control program returns an exit code indicating the compilation
 status. Exit codes are useful with the DOS batch command IF ERRORLEVEL and
 with the NMAKE utility. You can use them to test for the success or failure
 of the compilation before proceeding with other tasks. See Appendix B,
 "Using Exit Codes," in the Microsoft CodeView and Utilities User's Guide
 for more information.


 7.20.2  The Warning-Level Option (/W)


 Option

 /W{0 | 1| 2}

 You can suppress compiler warning messages with the /W (warning) option. Any
 message beginning with F4 is a compiler warning message. Warnings indicate
 potential problems (rather than actual errors) with statements that may not
 be compiled as you intend.

 /W1 (the default) causes the compiler to display warning messages. /W0 turns
 off warning messages. The /W0 option is useful when compiling programs that
 deliberately include questionable statements. /W2 suppresses the following
 error messages:

 F4998          variable used but not declared

 F4999          variable declared but not used

                /W0 and /W2 apply to the remainder of the command line or
                until the next occurrence of /W1 on the command line. These
                options have no effect on object files given on the command
                line.


 Example

      FL  /W0  CRUNCH.FOR  PRINT.FOR

 This example suppresses warning messages when the files CRUNCH.FOR and
 PRINT.FOR are compiled.


 7.21  Syntax Errors (/Zs)


 Option

 /Zs

 The /Zs option tells the compiler to perform a syntax check only. This is a
 quick way to find and correct syntax errors before compiling a source file.
 With /Zs, no code is generated and no object files or object listings are
 produced. However, you can specify the /Fs option on the same command line
 to generate a source listing.

 The /Zs option applies to all source files that follow the option on the
 command line but does not affect any source files preceding the option.


 Example

      FL  /Zs  TEST*.FOR

 This command causes the compiler to perform a syntax check on all source
 files in the current working directory that begin with TEST and end with the
 default extension (.FOR). The compiler displays messages for any errors
 found.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    The /4Yb and /4Yd options discussed in Section 7.9.2 above are also
    useful in identifying errors.
 ───────────────────────────────────────────────────────────────────────────


 7.22  Preparing for Debugging (/Zi, /Od, /Zd)

 Option                                     Effect

 /Zi                                        Prepares for debugging with the
                                            Microsoft CodeView debugger

 /Od                                        Disables optimization

 /Zd                                        Prepares for debugging with
                                            CodeView

 The /Zi option produces an object file containing full symbolic debugging
 information──including the symbol table and line numbers──for use with the
 Microsoft CodeView window-oriented debugger.

 When you use the FL command to compile and link, giving the /Zi option
 automatically causes the /CO option to be given at link time. If you link
 separately (whether using FL or the LINK command), instead of compiling and
 linking in one step, be sure to give the /CO option when you link.
 Otherwise, symbols and source-code lines will be missing when you run the
 CodeView debugger. See Section 1.3.5, "Preparing FORTRAN Programs," in the
 Microsoft CodeView and Utilities User's Guide for more information on /CO.

 The /Od option tells the compiler not to optimize. The default is to
 optimize. Using /Od is recommended whenever you use /Zi. It is also
 recommended while testing, since it can improve compilation speed by 30 to
 35 percent.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    If you use /Od when compiling, the F3S.EXE file must be in the current
    search path.
 ───────────────────────────────────────────────────────────────────────────

 Since optimization may rearrange instructions and store values in machine
 registers, you may have trouble finding and fixing errors if you optimize
 before debugging.

 Note that turning off or restricting optimization of a program usually
 increases the size of the generated code. If your program contains a module
 that is close to the 64K limit on compiled code, turning off optimization
 may cause the module to exceed the limit.

 See Section 7.24, "Optimizing," for a discussion of additional optimization
 options.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    When the debug option (/4Yb) is enabled, loop optimization is disabled.
    See Section 7.9.2.1 for a description of the debug option.
 ───────────────────────────────────────────────────────────────────────────

 The /Zd option produces an object file containing line-number records that
 correspond to the line numbers of the source file. The /Zd option is used
 when you want to pass an object file to the SYMDEB symbolic debugger,
 available with other Microsoft products (for instance, FORTRAN Version 3.0
 and earlier). The debugger can use the line numbers to refer to program
 locations. However, only global symbol-table information is available with
 SYMDEB (unlike the CodeView debugger, which also recognizes local symbols).

 When you use the FL command to compile and link, giving the /Zd option
 causes the /LI option to be given at link time. (See Chapter 13, "Linking
 Object Files with LINK," in the Microsoft Codeview and Utilities User's
 Guide for more information on /LI.) If you compile a source file with the
 /Zd option, and then link in a separate step using FL, be sure to give the
 /Zd option when you link. (If you link using the LINK command, give the /LI
 option.) Otherwise, your executable file will not contain line numbers.

 The /Zd option generates a map file, whether or not the /Fm option is given.
 If /Fm is not used to specify a file name or location for the file, the map
 file is created in the current working directory and given the default name,
 as described in Section 7.14.

 The /Zi, /Od, and /Zd options apply to any source files following the option
 on the command line, but do not affect source files preceding the option.
 The /Zi and /Od options have no effect on object files given on the command
 line. /Zd causes the /LI option to be given at link time.


 Example

      FL  /Zi  /Od  /Fs  P*.FOR  /FePROCESS  /FmPROCESS

 This command compiles all source files in the current working directory
 beginning with P and ending with the default extension (.FOR), creating
 object files that contain the symbolic information needed by the CodeView
 debugger. Optimization is disabled with /Od. The /Fs option creates a source
 listing for each source file. The executable file is named PROCESS.EXE, and
 a map file named PROCESS.MAP is also created.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    The /4Yb and /Zi options are incompatible. There is a conflict between
    the line information that each stores. If both are used and the code is
    brought up under CodeView, the source will come up in assembly mode with
    the statement, "No Symbolic Information." There is no need to turn on the
    debugging switch if you intend to debug in CodeView.
 ───────────────────────────────────────────────────────────────────────────


 7.23  Using an 80186, 80188, 80286, or 80386 Processor (/G0, /G1, /G2)

 Option                          Effect

 /G0                             8086/8088 instruction set
                                 (default)

 /G1                             80186/80188 instruction set

 /G2                             80286/80386 instruction set

 If you have an 80186/80188 or 80286/80386 processor, you can use the /G1 or
 /G2 option to enable the instruction set for your processor. Use /G1 for the
 80186 or 80188 processor; use /G2 for the 80286 or 80386. (80286 code runs
 on the 80386, but does not use any of the 80386's specialized instructions.)
 Although it is advantageous to use the appropriate instruction set, you are
 not required to do so. If you have an 80286 processor, for example, but want
 your code to also run on an 8086, do not use the /G1 or /G2 option.

 The /G0 option enables the instruction set for the 8086/8088 processor. You
 do not have to specify this option explicitly since the 8086/8088
 instruction set is used by default. Programs compiled this way also run on
 an 80186, 80188, 80286, or 80386 processor.

 Only one of these three options is allowed on the FL command line. If more
 than one appears, FL issues a warning and generates code using the last /G
 option on the line.


 Example

      FL  /G2  /FeFINAL  *.FOR

 The example above compiles and links all source files with the default
 extension (.FOR) in the current working directory, using the 80286
 instruction set. The resulting program, named FINAL.EXE, will run only on an
 80286 or 80386.


 7.24  Optimizing (/O and /Zp)

 The optimizing procedures performed by the Microsoft FORTRAN Compiler can
 reduce the storage space and execution time required for a compiled program
 by eliminating unnecessary instructions and rearranging code. The compiler
 performs some optimizations by default. You can use the /O options to
 exercise greater control over the optimization performed.


 Option

 /Oletters

 The /O (optimize) option controls optimization. The letters after /O alter
 the way the compiler optimizes your code. The letters are one or more of the
 following:

 Character                          Optimizing Procedure

 d                                  Disables optimization; leaves
                                    stack checking on

 l                                  Loop optimization

 p                                  Improves consistency of
                                    floating-point results

 s                                  Favors reduced code size

 t                                  Favors rapid execution

 x                                  Full optimization; equivalent to
                                    /Olt /Gs

 More than one /O option may appear on a command line. All of the options
 apply to all of the source-code files. FL processes the options in the order
 in which they appear, so their order is significant. For example, /Od
 followed by /Ol (or only /Odl) disables all optimization, then reinstates
 loop optimization only. On the other hand, /Ol followed by /Od (or just
 /Old) requests loop optimization, then disables all optimization. Hence, no
 optimization of any kind is performed.

 When you omit the /O option, or when you give an /O option but do not use
 the letter x, the compiler defaults to /Ox. Whenever the compiler has a
 choice between producing smaller (but slower) code and larger (but faster)
 code, the compiler chooses to generate the larger, faster code. To make the
 compiler produce smaller code, use the /Os option.

 The /Od option turns off optimization. This option is useful in the early
 stages of program development because it avoids optimizing code that is not
 in its final form, and it improves compilation speed by approximately 30 to
 35 percent. Because optimization may involve rearrangement of instructions,
 you may also want to specify the /Od option when you use a debugger other
 than the CodeView debugger with your program, or when you want to examine an
 object listing. (The /Zi option, which prepares a program for debugging with
 the CodeView debugger, automatically turns off loop optimization and
 optimization involving code rearrangement.) If you optimize before
 debugging, it can be difficult to recognize and correct your code.

 You may add options to the /Od option to reinstate specific optimizations.
 For example, /Odl performs only loop optimization, and /Odtp favors speed
 and floating-point consistency without optimizing loops. Any consistent
 combination is permitted. For example, /Odts would not be allowed since you
 cannot simultaneously optimize for size and speed.

 Note that turning off or restricting optimization usually increases the
 program's size. If your program contains a module that is close to the 64K
 limit on compiled code, turning off optimization may cause the module to
 exceed the limit.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    In performing optimizations on extremely complex code, the compiler may
    experience an internal error. It is sometimes possible to work around
    this problem by disabling the optimization pass with the /Od option.
    In all cases where you experience this type of compiler error, please
    contact Microsoft Corporation so that corrections can be made for
    subsequent releases.
 ───────────────────────────────────────────────────────────────────────────

 The /Op option is useful when floating-point results must be consistent.
 This option changes the compiler's default handling of floating-point
 values. Normally, the compiler tries to avoid assigning any value to a
 variable until all the calculations required for the variable's final value
 have been completed. It does this by storing intermediate values (wherever
 possible) in an 80-bit machine register.

 However, since floating-point types are allocated less than 80 bits of
 storage (32 bits for REAL*4 and 64 bits for REAL*8), a register value may
 actually be more precise than the same value stored in a floating-point
 variable. Over the course of many calculations, the value that results from
 the use of a machine register may be different from the value produced if
 the compiler assigned intermediate results to a variable. Furthermore,
 adding or deleting code may change the number of machine registers available
 to hold intermediate results, and thus alter the result of a particular
 calculation.

 Specifying the /Op option tells the compiler to place all intermediate
 results in variables rather than machine registers. Although this may give
 less precise results than using registers, and may increase program size, it
 guarantees consistent results in floating-point calculations.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    When the debug option (/4Yb) is enabled, loop optimization is disabled.
    See Section 7.9.2.1 for a description of the debug option.
 ───────────────────────────────────────────────────────────────────────────


 Examples

      FL  /c  /Os  FILE.FOR

 The command above favors code size over execution speed when compiling
 FILE.FOR.

      FL  /Od  *.FOR

 The command above compiles and links all FORTRAN source files with the
 default extension (.FOR) in the current directory and disables optimization.
 This command is useful during the early stages of program development, since
 it improves compilation speed.

      FL  /Op  /FeTESTRUN  *.FOR

 The command above causes floating-point assignments to variables to be
 carried out immediately (where specified) when compiling all source files
 with the default extension (.FOR) in the current working directory. By
 default, the optimization favors execution time. The resulting program is
 named TESTRUN.EXE.


 Option

 /Zp«{1 | 2 | 4}»

 The /Zp option controls the starting addresses (packing) of variables in
 structures. A tight pack saves memory space, at the expense of slightly
 slower access for noncharacter variables. Any $PACK:n metacommand in the
 file overrides a /Zp option, except for $PACK, which restores whatever
 command-line option was given.

 If no /Zp option is given, and if there is no $PACK metacommand in the file,
 structures are packed according to the following default rules: INTEGER*1,
 LOGICAL*1, and all CHARACTER variables begin at the next available byte,
 whether odd or even; all other variables begin at the next even byte. This
 arrangement wastes some memory space, but gives the quickest access.

 If /Zp1 is specified, all variables begin at the next available byte,
 whether odd or even. Although this slightly increases access time, no memory
 space is wasted.

 If /Zp2 is specified, INTEGER*1, LOGICAL*1, and all CHARACTER variables
 begin at the next available byte, whether odd or even. All other variables
 start on the next available even byte. This is equivalent to the default
 packing, described above.

 If /Zp4 is specified, INTEGER*1, LOGICAL*1, and all CHARACTER variables
 begin at the next available byte, whether odd or even. All other variables
 start on the next four-byte boundary.


 7.25  Enabling and Disabling Stack Probes (/Ge, /Gs)

 Option                          Effect

 /Ge                             Enables stack probes

 /Gs                             Disables stack probes (default)

 A stack probe is a short subroutine called on entry to a procedure to verify
 that the program stack has enough space for any automatic local variables.
 When stack probes are enabled, the stack-probe subroutine is automatically
 called at every entry point. The stack-probe subroutine generates a message
 and ends the program if it determines that the required stack space is not
 available.

 By default, all Microsoft FORTRAN variables are static, not automatic, so
 only procedures containing automatic variables require stack space for
 variable allocation. However, any procedure call consumes stack space for
 the return address and the addresses or values of the arguments passed. If
 your program has many nested procedure calls, or passes a lot of variables
 to procedures, the stack may overflow. Although enabling a stack probe does
 not prevent an overflow, it will tell you when one occurs.

 Note, however, that programs with stack checking are slightly larger and may
 perform less efficiently due to the calls to the probe routine. Stack probes
 can be used during development, then removed from the final version of the
 program.

 The /Ge option applies to all source files following the option on the
 command line. The /Gs option disables stack checking for all source files
 that follow it on the command line.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Although the default option, which disables stack probes, reduces program
    size, it means that no compiler error message is displayed if a stack
    overflow occurs. You may want to use the /Ge option when testing to make
    sure the program does not cause a stack overflow.
 ───────────────────────────────────────────────────────────────────────────


 Example

      FL  /c  /Ge  /Ot  FILE.FOR

 This example enables stack probes and favors execution time when compiling
 FILE.FOR.


 7.26  Suppressing Automatic Library Selection (/Zl)


 Option

 / Zl

 The compiler ordinarily places in the object file the name of the FORTRAN
 library corresponding to the floating-point and memory-model options you
 choose. The linker uses the library name to link the program automatically
 with the corresponding library. Thus, you do not need to specify a library
 name to the linker, provided that the appropriate library exists for the
 floating-point and memory-model options you are using.

 The /Zl option suppresses the insertion of library names in object files.
 When you specify /Zl, the compiler does not place a library name in the
 object file. As a result, the object file is slightly smaller.

 The /Zl option produces a significant space savings when building a library
 containing many object modules. When you link a library created using the
 /Zl option with a program file compiled without the /Zl option, the program
 file supplies the name of the library.

 The /Zl option applies to the remainder of the source files on the command
 line.


 Examples

      FL  ONE.FOR  /Zl  TWO.FOR

 The example above creates an object file named ONE.OBJ. Since no
 floating-point or memory-model options are specified on the FL command line,
 this object file contains the name of the FORTRAN library that corresponds
 to the default floating-point and memory-model options (LLIBFOR7.LIB). The
 example also creates an object file named TWO.OBJ without any library
 information, since the /Zl option appears before the file name on the
 command line. When ONE.OBJ and TWO.OBJ are linked to create an executable
 file, the library information in ONE.OBJ causes LLIBFOR7.LIB to be searched
 for any unresolved references in either ONE.OBJ or TWO.OBJ.

      FL  /c  /Zl  *.FOR

 The example above compiles all source files with the default extension
 (.FOR) in the current working directory. None of the resulting object files
 contain library information.


 7.27  Setting the Stack Size (/F)


 Option

 /F hexnum

 The /F option sets the size of the program stack. A space must separate /F
 and hexnum.

 The hexnum is a hexadecimal number representing the stack size in bytes. The
 number must be positive and cannot exceed 10,000 hexadecimal (65,536
 decimal). The default stack size is 2K.

 If many variables are declared as automatic, or if many parameters are
 passed to procedures, a larger stack may be needed to accommodate them.
 Stack overflow does not always cause a program to crash; it may have no
 effect at all, or it may cause erratic operation or incorrect results. If
 there is any doubt whether the stack is large enough, use the /Ge option to
 enable stack probes. Stack overflow will then automatically halt the
 program.

 This option affects object files only. It has no effect on source files.

 Using the /F option with the FL command has the same effect as using the
 /STACK option with the LINK program. See Chapter 13, "Linking Object Files
 with LINK," in the Microsoft CodeView and Utilities User's Guide for more
 information on the /STACK option.


 Example

      FL /F C00 *.OBJ

 This example sets the stack size to C00 hexadecimal (3K decimal) for the
 program created by linking all object files in the current working
 directory.


 7.28  Restricting the Length of External Names (/H)


 Option

 /Hnumber

 The /H option restricts the length of external names in the object file. The
 number is an integer specifying the maximum number of significant characters
 in external names.

 The /H option has no effect during compilation. The first 31 characters of a
 name are significant if $NOTRUNCATE (the default) is in effect, and only the
 first six if $TRUNCATE is in effect. It is only when the object file is
 created that any external names longer than number characters are truncated.

 This option has no effect on local names.


 7.29  Labeling the Object File (/V)


 Option

 /V"string"

 Use the /V (version) option to embed a given text string into an object
 file. The quotation marks may be omitted if the string does not contain
 blanks or tabs. A common use of the /V option is to label an object file
 with a version number or copyright notice.

 The /V option applies to all source files following the option on the
 com-mand line.


 Example

      FL /V"Microsoft FORTRAN Compiler Version 5.0" MAIN.FOR

 This command places the string Microsoft FORTRAN Compiler Version 5.0 in the
 object file MAIN.OBJ.


 7.30  Linking with Libraries

 When the FL command compiles a source file, it places a FORTRAN library name
 in the object file. The library name corresponds to the memory-model and
 floating-point options you chose on the FL command line, or the defaults for
 those options you did not explicitly select. See Table 7.1 in Section 7.4,
 "Memory-Model Options," for the library names FL includes in the object file
 for each combination of memory-model and floating-point options.

 The linker looks for a library matching the name embedded in the object
 file. If it finds a library with that name, it automatically links the
 library with the object file.

 The result is that you do not need to specify library names on the FL
 command line unless you want to link with standard libraries renamed during
 SETUP with libraries other than the default library for the floating-point
 and memory-model options you have chosen, or with user-created libraries.

 Setting up under OS/2 for both real- and protected-mode programming creates
 separate real- and protected-mode libraries, with different names. Use the
 /L option to specify the correct link library. See Section 7.5, "OS/2
 Library Options," for more information.

 If you want to link with other libraries, you must either use the /link
 option on the FL command line and include the new library names, or run the
 linker and specify the library names separately. In either case, the linker
 tries to resolve external references by searching the library you specified
 before it searches the library whose name is embedded in the object file. If
 you want the linker to ignore the library whose name is embedded in the
 object file, you must also include the /NOD (NODEFAULTLIBRARYSEARCH) linker
 option, either as part of the /link option on the FL command line, or as an
 option on the LINK command line.

 See Section 7.32, "Using FL to Link without Compiling," for information
 about the /link option of the FL command. See Chapter 13, "Linking Object
 Files with LINK," in the Microsoft Codeview and Utilities User's Guide for
 information about specifying library names to the linker.

 See Section 7.4, "Memory-Model Options," for more information on
 dynamic-link library and multithread linking.


 7.31  Creating Overlays

 You can specify program overlays on the command line. Overlays let several
 program modules use the same memory area, one module at a time. When needed,
 a module is loaded from the disk. See Chapter 13, "Linking Object Files with
 LINK," in the Microsoft CodeView and Utilities User's Guide for more
 information about overlays.

 The modules to be overlaid are enclosed in parentheses. For example, the
 following command line instructs FL to overlay modules OVER1.FOR  and
 OVER2.FOR:

      FL  MAIN.FOR  (OVER1.FOR  OVER2.FOR)


 7.32  Using FL to Link without Compiling

 Just as the FL command can compile files without linking the resulting
 object code, you can use FL to link object files that were previously
 compiled. If all of the files you give FL have extensions other than .FOR,
 and if no /Tf options appear, FL skips the compiling stage and links your
 files. To link object files, use the following special form of the FL
 command:

 FL objfile«,objfile...» /link «libfield» « linkoptions»

 When FL links object files, it gives the resulting executable file the base
 name of the first object file on the command line, plus an .EXE extension,
 by default. Alternately, you can specify files with the .LTB extension
 without the /link option. The /link option is most useful for passing link
 options. (This is the same naming convention FL uses when it compiles source
 files first, then links the resulting object files.)

 Command options beginning with /F allow you to supply the file names and
 options that would otherwise appear on the LINK command line (or in response
 to LINK prompts). The following list shows each FL option for the linker and
 the corresponding LINK command-line field, prompt, or option:

 FL Option                            LINK Field/Prompt/Option

 /Feexefile                           The exefile field or "Run File"
                                      prompt

 /Fmmapfile                           The mapfile field or "List File"
                                      prompt

 /link libfile linkoptions            The libfiles field or "Libraries"
                                      prompt, and any of the LINK
                                      options described in Chapter 13
                                      of the Microsoft CodeView and
                                      Utilities User's Guide

 /Fhexnum                             The /STACK option

 See Section 7.13 for a description of the /Fe option, Section 7.14 for a
 description of the /Fm option, and Section 7.27 for a description of the /F
 option. Chapter 13 of the Microsoft CodeView and Utilities User's Guide
 describes the LINK command-line fields and the /STACK option.

 If you use the /link libfile linkoptions option with the FL command, it must
 be the last option on the command line.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    The FL command normally links without having to specify the /link option.
    However, you can include the /link option in an FL command line if you
    wish to modify the linking process, such as by specifying a different
    library.
 ───────────────────────────────────────────────────────────────────────────


 7.33  Specifying Assembler Options (/MA)


 Option

 /MA option

 The /MA option allows you to specify Microsoft Assembler options when
 including assembly-language filenames on the command line. You must use a
 separate /MA option for each assembly-language option. The option applies to
 all of the assembly-language files that follow it on the command line,
 unless another option overrides it.


 Example

      FL FORT1.FOR /MA/B40 ASM1.ASM

 In this example, the /B40 option is passed to the Microsoft Assembler,
 specifying that the source file buffer is to be 40K when ASM1.ASM is
 assembled.


 Appendix A  ASCII Character Codes

 ───────────────────────────────────────────────────────────────────────────

 ASCII Character Codes

 Ctl D  H  O  C Code  D  H  O   C   D  H  O   C   D  H  O   C

 ^@   0  00  000     NUL│  32  20  040    │  64  40  100  @ │  96  60  140  `
 ^A   1  01  001  
  SOH│  33  21  041  ! │  65  41  101  A │  97  61  141  a
 ^B   2  02  002    STX│  34  22  042  " │  66  42  102  B │  98  62  142  b
 ^C   3  03  003    ETX│  35  23  043  # │  67  43  103  C │  99  63  143  c
 ^D   4  04  004    EOT│  36  24  044  $ │  68  44  104  D │ 100  64  144  d
 ^E   5  05  005    ENQ│  37  25  045  % │  69  45  105  E │ 101  65  145  e
 ^F   6  06  006    ACK│  38  26  046  & │  70  46  106  F │ 102  66  146  f
 ^G   7  07  007    BEL│  39  27  047  ' │  71  47  107  G │ 103  67  147  g
 ^H   8  08  010    BS │  40  28  050  ( │  72  48  110  H │ 104  68  150  h
 ^I   9  09  011        HT │  41  29  051  ) │  73  49  111  I │ 105  69  151  i
 ^J  10  0A  012     LF │  42  2A  052  * │  74  4A  112  J │ 106  6A  152  j
 ^K  11  0B  013    VT │  43  2B  053  + │  75  4B  113  K │ 107  6B  153  k
 ^L  12  0C  014    FF │  44  2C  054  , │  76  4C  114  L │ 108  6C  154  l
 ^M  13  0D  015     CR │  45  2D  055  - │  77  4D  115  M │ 109  6D  155  m
 ^N  14  0E  016    SO │  46  2E  056  . │  78  4E  116  N │ 110  6E  156  n
 ^O  15  0F  017    SI │  47  2F  057  / │  79  4F  117  O │ 111  6F  157  o
 ^P  16  10  020    DLE│  48  30  060  0 │  80  50  120  P │ 112  70  160  p
 ^Q  17  11  021    DC1│  49  31  061  1 │  81  51  121  Q │ 113  71  161  q
 ^R  18  12  022    DC2│  50  32  062  2 │  82  52  122  R │ 114  72  162  r
 ^S  19  13  023    DC3│  51  33  063  3 │  83  53  123  S │ 115  73  163  s
 ^T  20  14  024    DC4│  52  34  064  4 │  84  54  124  T │ 116  74  164  t
 ^U  21  15  025    NAK│  53  35  065  5 │  85  55  125  U │ 117  75  165  u
 ^V  22  16  026    SYN│  54  36  066  6 │  86  56  126  V │ 118  76  166  v
 ^W  23  17  027    ETB│  55  37  067  7 │  87  57  127  W │ 119  77  167  w
 ^X  24  18  030    CAN│  56  38  070  8 │  88  58  130  X │ 120  78  170  x
 ^Y  25  19  031    EM │  57  39  071  9 │  89  59  131  Y │ 121  79  171  y
 ^Z  26  1A  032     SUB│  58  3A  072  : │  90  5A  132  Z │ 122  7A  172  z
 ^[  27  1B  033    ESC│  59  3B  073  ; │  91  5B  133  [ │ 123  7B  173  {
 ^\  28  1C  034    FS │  60  3C  074  < │  92  5C  134  \ │ 124  7C  174  |
 ^]  29  1D  035    GS │  61  3D  075  = │  93  5D  135  ] │ 125  7D  175  }
 ^^  30  1E  036    RS │  62  3E  076  > │  94  5E  136  ^ │ 126  7E  176  ~
 ^_  31  1F  037    US │  63  3F  077  ? │  95  5F  137  _ │ 127  7F  177  

        D    H   O    C   D    H   O    C   D    H   O    C   D

        128  80  200  Ç │ 160  A0  240  á │ 192  C0  300  └ │ 224  E0  340  α
        129  81  201  ü │ 161  A1  241  í │ 193  C1  301  ┴ │ 225  E1  341  ß
        130  82  202  é │ 162  A2  242  ó │ 194  C2  302  ┬ │ 226  E2  342  Γ
        131  83  203  â │ 163  A3  243  ú │ 195  C3  303  ├ │ 227  E3  343  π
        132  84  204  ä │ 164  A4  244  ñ │ 196  C4  304  ─ │ 228  E4  344  Σ
        133  85  205  à │ 165  A5  245  Ñ │ 197  C5  305  ┼ │ 229  E5  345  σ
        134  86  206  å │ 166  A6  246  ª │ 198  C6  306  ╞ │ 230  E6  346  µ
        135  87  207  ç │ 167  A7  247  º │ 199  C7  307  ╟ │ 231  E7  347  τ
        136  88  210  ê │ 168  A8  250  ¿ │ 200  C8  310  ╚ │ 232  E8  350  Φ
        137  89  211  ë │ 169  A9  251  ⌐ │ 201  C9  311  ╔ │ 233  E9  351  Θ
        138  8A  212  è │ 170  AA  252  ¬ │ 202  CA  312  ╩ │ 234  EA  352  Ω
        139  8B  213  ï │ 171  AB  253  ½ │ 203  CB  313  ╦ │ 235  EB  353  δ
        140  8C  214  î │ 172  AC  254  ¼ │ 204  CC  314  ╠ │ 236  EC  354  ∞
        141  8D  215  ì │ 173  AD  255  ¡ │ 205  CD  315  ═ │ 237  ED  355  φ
        142  8E  216  Ä │ 174  AE  256  « │ 206  CE  316  ╬ │ 238  EE  356  ε
        143  8F  217  Å │ 175  AF  257  » │ 207  CF  317  ╧ │ 239  EF  357  ∩
        144  90  220  É │ 176  B0  260  ░ │ 208  D0  320  ╨ │ 240  F0  360  ≡
        145  91  221  æ │ 177  B1  261  ▒ │ 209  D1  321  ╤ │ 241  F1  361  ±
        146  92  222  Æ │ 178  B2  262  ▓ │ 210  D2  322  ╥ │ 242  F2  362  ≥
        147  93  223  ô │ 179  B3  263  │ │ 211  D3  323  ╙ │ 243  F3  363  ≤
        148  94  224  ö │ 180  B4  264  ┤ │ 212  D4  324  ╘ │ 244  F4  364  ⌠
        149  95  225  ò │ 181  B5  265  ╡ │ 213  D5  325  ╒ │ 245  F5  365  ⌡
        150  96  226  û │ 182  B6  266  ╢ │ 214  D6  326  ╓ │ 246  F6  366  ÷
        151  97  227  ù │ 183  B7  267  ╖ │ 215  D7  327  ╫ │ 247  F7  367  ≈
        152  98  230  ÿ │ 184  B8  270  ╕ │ 216  D8  330  ╪ │ 248  F8  370  °
        153  99  231  Ö │ 185  B9  271  ╣ │ 217  D9  331  ┘ │ 249  F9  371  ∙
        154  9A  232  Ü │ 186  BA  272  ║ │ 218  DA  332  ┌ │ 250  FA  372  ·
        155  9B  233  ¢ │ 187  BB  273  ╗ │ 219  DB  333  █ │ 251  FB  373  √
        156  9C  234  £ │ 188  BC  274  ╝ │ 220  DC  334  ▄ │ 252  FC  374  ⁿ
        157  9D  235  ¥ │ 189  BD  275  ╜ │ 221  DD  335  ▌ │ 253  FD  375  ²
        158  9E  236  ₧ │ 190  BE  276  ╛ │ 222  DE  336  ▐ │ 254  FE  376  ■
        159  9F  237  ƒ │ 191  BF  277  ┐ │ 223  DF  337  ▀ │ 255  FF  377


 Appendix B  Differences from Previous Versions

 ───────────────────────────────────────────────────────────────────────────

 This appendix documents the changes to Microsoft FORTRAN as the language has
 evolved from Versions 3.2 and 3.3 to Versions 4.0, 4.1, and 5.0. The
 appendix is in three sections, one for each update.

 The first section describes the current update, from Version 4.1 to Version
 5.0. The second section explains the changes that occurred when OS/2
 features were added to Version 4.0 to create Version 4.1. The third section
 describes the changes from Versions 3.2 and 3.3 to Version 4.0.

 The README.DOC file may also contain information unavailable when this
 update was printed.

 You may use the SETUP program to create libraries to link with Version 5.0
 programs. If you choose all the default responses for SETUP, the library
 that SETUP builds requires that you have an 8087 or 80287 coprocessor
 installed.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Since MS-DOS and PC-DOS are functionally the same operating system,
    Microsoft manuals and updates use the term DOS to include both systems,
    except when noting features that are unique to one or the other.
 ───────────────────────────────────────────────────────────────────────────


 B.1  Changes from Version 4.1 to Version 5.0

 This section summarizes the Microsoft FORTRAN features introduced in Version
 5.0.


 B.1.1  Alphabetical Summary

 Changes to individual statements, metacommands, and other features are
 listed alphabetically below.


 B.1.1.1  Allocatable Arrays

 Arrays may be declared as allocatable. Although the number of dimensions is
 fixed during compilation, the size and range of each dimension can be set at
 run time.


 B.1.1.2  APPEND Mode

 Files may be opened in ACCESS='APPEND' mode, which automatically positions
 the file pointer at the end of the file.


 B.1.1.3  Array Operations

 A complete array can now be an operand in an expression. For example, two
 arrays may be directly added together in a single statement, without having
 to use a DO loop. Similarly, a constant value may be subtracted from every
 element in an array, or two arrays of LOGICAL variables can be compared with
 a logical operator.

 An array may also be passed to a function (intrinsic or external). The
 function returns an identically-dimensioned array whose individual elements
 are the values obtained when the function is applied to the corresponding
 element of the original array. For more information, see Section 1.7.5,
 "Array Expressions."


 B.1.1.4  AUTOMATIC Variables

 Local variables may be declared "on the stack," rather than at fixed memory
 locations, with the AUTOMATIC statement.


 B.1.1.5  Compatibility with Versions 3.2 and 3.3

 Version 5.0 no longer supports compatibility with object files produced by
 Version 3.2 and 3.3 compilers. If you wish to use their code in new
 programs, they must be recompiled with the Version 5.0 compiler.


 B.1.1.6  Conditional Compilation

 The $DEFINE, $UNDEFINE, $IF, $ELSE, $ELSEIF, and $ENDIF metacommands control
 conditional compilation of FORTRAN source code. FL command-line arguments
 permit test values to be passed to these metacommands.


 B.1.1.7  CYCLE Statement

 The CYCLE statement transfers control back to the first line of a DO or DO
 WHILE loop.


 B.1.1.8  DO WHILE Statement

 The new DO WHILE statement permits a loop to be executed as long as a
 logical condition is .TRUE..


 B.1.1.9  END DO Statement

 The END DO statement can now terminate a DO or DO WHILE loop. Labeled
 terminating statements are no longer required (though they are still
 permitted).


 B.1.1.10  End-of-line Comments

 A comment line may now begin with an exclamation point. An exclamation point
 outside a character or Hollerith constant that is not in column six is
 considered the beginning of a comment, and it may appear anywhere on a line
 following the statement.


 B.1.1.11  Exclusive Or (.xor.) Logical Operator

 The .XOR. logical operator has been added. Although its function is
 identical to the existing .NEQV. operator, its meaning is clearer.


 B.1.1.12  EXIT Statement

 The EXIT statement transfers control to the first statement following a DO
 or DO WHILE loop.


 B.1.1.13  IMPLICIT NONE Statement

 The new IMPLICIT NONE statement removes implicit data typing. Any variable
 not explicitly typed produces a compile-time warning message.


 B.1.1.14  INCLUDE Statement

 The INCLUDE statement inserts the contents of a specified file at the point
 in the program where the INCLUDE statement appears. The compiler immediately
 processes these statements. The $INCLUDE metacommand, which has the same
 function, is retained.


 B.1.1.15  Loop Optimization

 Loop optimization may be enabled and disabled with the $LOOPOPT and
 $NOLOOPOPT metacommands.


 B.1.1.16  NAMELIST Statement

 The NAMELIST statement defines a group of variables that are written to or
 read from a formatted file with a single I/O statement. When written, the
 namelist variable values are labeled with their names. When read, a variable
 is assigned the value given in a varname = value statement. The namelist
 feature allows a programmer to display a large group of variables with only
 a single statement, or to read the values of a group of parameters from an
 ASCII file.


 B.1.1.17  Numeric Arguments for Logical Operators

 The .AND., .OR., .XOR., .EQV., .NEQV., and .NOT. logical operators now
 accept INTEGER arguments. The result of the operation is not a logical
 value, but an integer value determined by bitwise comparison of the
 operands.


 B.1.1.18  Quotes

 Character literals may now be delimited by quotes, as well as apostrophes.


 B.1.1.19  SELECT CASE Statement

 The SELECT CASE construct is similar to other languages' case-selection
 mechanisms. The test expression may be of INTEGER, LOGICAL, or  CHARACTER*1
 type. The case-list values may be single values, ranges of values, or any
 combination of the two. An optional CASE DEFAULT statement handles
 situations when none of the list values match the test expression.


 B.1.1.20  Structure Type

 A user-defined data type called a "structure" may be created. A structure
 type is a combination of other data types, including other structures. The
 RECORD statement creates variables of a specific structure type. Structure
 variables may be used both for internal data representation, and for reading
 and writing groups of data from unformatted files.


 B.1.1.21  Symbol Table Enlargement

 The Microsoft FORTRAN compiler has changed its internal allocation of the
 symbol table from near to far memory, allowing many more symbols in a given
 subprogram before the compiler runs out of memory.


 B.1.1.22  $TRUNCATE

 Microsoft FORTRAN no longer sets $TRUNCATE by default. Programs which
 contain variable names longer than six characters but with differences after
 the sixth character will need to explicitly set the $TRUNCATE metacommand.


 B.1.2  New Microsoft FORTRAN Functions and Procedures

 See Chapter 5 , "Intrinsic Functions and Additional Procedures," for full
 descriptions of the following functions and procedures.

 Function or Procedure                      Description

 EPSILON                                    Smallest number that is larger
                                            than one

 GETARG                                     Returns nth command-line argument

 HUGE                                       Largest positive number

 LEN_TRIM                                   Returns the length of a string
                                            (less trailing blanks)

 MAXEXPONENT                                Largest positive exponent for a
                                            data type

 MINEXPONENT                                Largest negative exponent for a
                                            data type

 NARGS                                      Returns number of command-line
                                            arguments

 NEAREST                                    Closest number (larger or
                                            smaller) to a given number

 SCAN                                       Searches a string for a character
                                            in a set

 TINY                                       Smallest positive number

 VERIFY                                     Searches a string for a character
                                            not in a set


 B.1.3  Microsoft FORTRAN Language Extensions

 Microsoft FORTRAN Version 5.0 supports all IBM SAA extensions. Microsoft
 FORTRAN Version 5.0 includes many (but not all) VAX extensions. Some of
 these features may be implemented in this manual differently than DEC
 implementation. Read the appropriate section to see if there are differences
 that are important to your application.

 For ease in transporting SAA or VAX applications to MS-DOS computers, there
 are two FL command-line options which disable all Microsoft FORTRAN
 extensions except for SAA extensions (/4Yi) or VAX extensions (/4Yv). See
 Chapter 7, "The FL Command," for more information on these options.


 B.1.4  OS/2

 Microsoft FORTRAN now supports a method for building dynamic-link and
 multithread libraries. For more information, see Microsoft FORTRAN Advanced
 Topics.


 B.1.5  Graphics

 Microsoft FORTRAN now provides a graphics library which supports text
 manipulation under both DOS and OS/2, and line- and shape-drawing functions
 under DOS. For more information , see Microsoft FORTRAN Advanced Topics.


 B.2  Changes from Version 4.0 to Version 4.1

 The principal change from Version 4.0 to Version 4.1 is the addition of OS/2
 support. The compiler, the linker, and the FL utility have been modified to
 correctly compile and link OS/2 real- or protected-mode programs.


 B.2.1  OS/2 Support

 You can write protected-mode programs that run under OS/2, real-mode
 programs that run under DOS Version 3.0 (or higher), or bound programs that
 run under both operating systems. For detailed information about developing
 programs for OS/2, see Chapter 3, "Programming Under OS/2," in Microsoft
 FORTRAN Advanced Topics.


 B.2.2  Enhanced FL Utility

 You can now use an environment variable to specify frequently used FL
 options. The new version of FL also lets you specify libraries anywhere on
 the command line and create overlays without a separate link step. The FL
 command also includes switches for compiling and linking under OS/2.


 B.2.2.1  Creating Overlays

 The FL utility now lets you specify program overlays on the command line, by
 enclosing the modules to be overlaid in parentheses.


 B.2.2.2  Specifying Libraries

 If you have set up the compiler for both real- and protected-mode OS/2, you
 must specify the correct library to use when linking. You can use the /Lp
 option to specify the protected-mode libraries, and the /Lr or /Lc option to
 specify the real-mode libraries.


 B.2.2.3  FL Environment Variable

 The FL environment variable can be used to specify a group of default
 options that are automatically added to the FL command line.


 B.2.2.4  Creating Bound Applications

 The /Fb option may be used to create a bound program that runs under either
 DOS or OS/2.


 B.2.2.5  Mixing .LIB and .DEF Files

 You can now mix .LIB and .DEF files on the command line with other options
 and files. FL sorts through the file names and sends them in the correct
 order to the linker.


 B.2.2.6  /NOI Option

 The /NOI (NOIGNORECASE) option is no longer the default. You must specify
 this option by including /link /NOI on the command line or in the
 environment variable.


 B.2.2.7  /Tf Option

 To specify a source file with an extension other than .FOR, use the /Tf
 option.


 B.2.3  Extended Control Over Default Libraries (Linker Options)

 An extension to the /NOD option lets you specify that a particular default
 library not be searched during linking.

 The version of the linker included with Microsoft FORTRAN Version 4.1
 includes an extension to the /NOD (ignore default libraries) option. In
 previous versions, /NOD made the linker avoid searching any library
 specified in the object file. You can now use the option to tell LINK not to
 search a specific default library by including the name after a colon (:).


 B.3  Changes from Versions 3.2 and 3.3 to Version 4.0

 This section describes features of the Microsoft FORTRAN Compiler, Version
 4.0, that are extensions of or changes to Version 3.3. It summarizes the
 changes made in Version 4.0 to support the ANSI full-language standard; it
 discusses compatibility between source and object files for Versions 3.2,
 3.3, and 4.0; and it describes changes and additions to the compiler and
 linker software, the run-time library system, and the language itself.


 B.3.1  Changes for ANSI Full-Language Standard

 Version 4.0 of the Microsoft FORTRAN Compiler is an implementation of the
 ANSI X3.9-1978 FORTRAN full-language standard; Version 3.3 implemented only
 the subset standard. The following list summarizes the new features in
 Version 4.0 that were required for the ANSI full-language standard.

 Language Construct                         Change for Version 4.0

 Concatenation operator (//)                Now supported.

 Asterisk length specifiers                 Can be used with character
                                            functions and character
                                            parameters.

 CHARACTER*n arguments                      Argument length passed with
                                            CHARACTER*n arguments to
                                            subprograms or functions. The
                                            maximum value of n is now 32,767
                                            instead of 127.

 Format specifiers                          Can be character arrays.

 Unit specifiers                            Unit specifiers that include the
                                            UNIT= keyword can appear at any
                                            position in the I/O control list.

 LEN intrinsic function                     Now supported.

 INDEX intrinsic function                   Now supported.

 Assignment statement                       Can include Hollerith constants.
 (computational)

 BACKSPACE, ENDFILE, and REWIND             Can transfer control to a label
 statements                                 after errors and use a variable
                                            to indicate error or end-of-file
                                            status.

 CLOSE and OPEN statements                  Can transfer control to a label
                                            after errors. The OPEN statement
                                            can specify how blanks are
                                            interpreted in numeric input.

 DATA statement                             Items that are assigned values
                                            can include substring names and
                                            implied-DO lists.

 CALL statement                             Can include Hollerith constants.

 DATA statement                             Can include Hollerith constants.

 DIMENSION statement                        Both upper and lower bounds
                                            allowed for dimension
                                            declarators.

 DO statement                               Loop indices can be of any REAL
                                            type.

 ENTRY statement                            Now supported.

 STATUS= in OPEN statements                 Opening existing files with the
                                            STATUS='NEW' option is illegal.
                                            STATUS='UNKNOWN' in Version 4.0
                                            behaves the same way as does
                                            STATUS='NEW' in Version 3.3.

 Constants in PARAMETER statements          Arithmetic, logical, and
                                            character constants fully
                                            supported.

 PRINT statement                            Now supported.

 READ statement                             READ statements without a control
                                            information list or without a
                                            unit specifier supported.


 B.3.2  Source Compatibility

 Version 4.0 of the Microsoft FORTRAN Compiler compiles any valid source
 program that you successfully compiled using an earlier version of the
 compiler, except where list-directed I/O and direct-access I/O are used
 together. However, source programs may behave differently when compiled with
 Version 4.0.


 B.3.3  Attributes in Array Declarations

 In array declarations in Version 4.0, attributes appear before the list of
 array bounds. In Version 3.3, attributes appear after the list of array
 bounds.

 For example, this declaration in a Version 3.3 source file

            DIMENSION x(10)[VALUE]

 should appear as shown below in a Version 4.0 source file:

            DIMENSION x[VALUE](10)


 B.3.4  Blanks in Formatted Files

 The ANSI full-language and subset standards treat blanks in formatted files
 differently. In the full-language standard, blanks are treated as null
 characters unless the BN and BZ format descriptors, or the BLANK= option in
 an OPEN statement, specify otherwise. In the subset standard, blanks are
 treated as zeros unless the BN and BZ format descriptors indicate otherwise.

 Version 4.0 supports the full-language treatment of blanks: it considers
 blanks to be null characters unless otherwise specified.

 If the files used by a program expect blanks to be treated as zeros by
 default, the program must include the BLANK='ZERO' option in the OPEN
 statements for those files.


 B.3.5  MODE and STATUS Options in OPEN Statement

 In Version 4.0, if the MODE=mode option does not appear in an OPEN
 statement, the FORTRAN run-time system tries to open the file with MODE
 values of 'READWRITE', 'READ', and 'WRITE', in that order. In Version 3.3,
 if the MODE=mode option does not appear in an OPEN statement, the FORTRAN
 run-time system tries to open the file with MODE values of 'READWRITE',
 'WRITE', and 'READ', in that order.

 In Version 4.0, when the STATUS='NEW' option appears in an OPEN statement,
 the file specified in the statement must not exist. If an existing file has
 the same path name as the file specified in the statement, an error results.
 In Version 3.3, when the STATUS='NEW' option appears in an OPEN statement,
 the file specified in the statement can exist at the time the statement is
 executed. Any file with the same path name as the file specified in the
 statement is overwritten. (This conflicts with a strict interpretation of
 the standard.)

 If you want programs compiled using Version 4.0 to behave in the same way as
 programs compiled using Version 3.3, substitute the STATUS='UNKNOWN' option
 for the STATUS='NEW' option in any OPEN statements that specify the path
 names of existing files.


 B.3.6  Temporary Scratch-File Names

 In Version 4.0, if no file name is specified in an OPEN statement, the
 FORTRAN run-time system creates a temporary scratch file with a file name in
 the following format:

      ZZprocessno

 In this file name, processno is an alphanumeric character followed by a
 5-digit process number. The alphanumeric character is "0" for the first
 temporary file opened, followed by the letters "a", "b", "c", and so on for
 each subsequent file name. For example, if you opened five files with no
 file names in a single program, the file names assigned to the temporary
 files would be the following (if "12345" is the process number):

 ZZ012345 (first file opened) ZZa12345 (second file opened) ZZb12345 (third
 file opened) ZZc12345 (fourth file opened) ZZd12345 (fifth file opened)

 In Version 3.3, if no file name is specified in an OPEN statement, the
 scratch file name has the following format:

      Tunitspec.TMP

 In this file name, unitspec is the unit number specified in the OPEN
 statement.


 B.3.7  Binary Direct Files

 In Version 4.0, binary files can be opened for direct access. In most cases,
 I/O operations performed on binary direct files produce the same results as
 the same operations performed on unformatted direct files. An exception is
 that the number of bytes transferred in a single binary direct read or write
 operation is no longer limited by the record length (although even multiples
 of the record length are still used in repositioning between successive READ
 and WRITE statements).

 See Chapter 6, "Record Structure: File Formats and Access," in Microsoft
 FORTRAN Advanced Topics.


 B.3.8  Precision of Floating-Point Operations

 Programs that use floating-point values may give slightly different results
 when compiled with Version 4.0 because Version 4.0 passes more information
 to the 8087/80287 coprocessor than Version 3.3. This has the effect of
 maintaining higher precision than if the values were truncated into double-
 or single-precision values.

 For example, in Version 4.0, arguments to transcendental functions are
 passed in the 8087/80287 registers. If these arguments are expressions,
 their values are in the 64-bit precision of the coprocessor. In Version 3.3,
 arguments to transcendental functions are passed in memory as either single-
 or double-precision values. Thus, these arguments are truncated to 23- or
 52-bit precision, respectively.

 See Chapter 1, "Controlling Floating-Point Operations," in Microsoft FORTRAN
 Advanced Topics for more information.


 B.3.9  Exponentiation Exceptions

 Versions 4.0 and 3.3 give different results for certain cases of
 exponentiation. These differences fall into four categories:

   1. Zero raised to a zero power

   2. Zero raised to a negative power

   3. COMPLEX zero raised to a COMPLEX power

   4. Negative INTEGER or REAL values raised to a REAL power

 Tables B.1-B.4 summarize these differences. The following abbreviations are
 used in the tables:

 Abbreviation                Meaning

 f                           Nonintegral real number (for
                             example, 1.5)

 -n                          Negative integer

 +r                          Positive real number

 -r                          Negative real number

 s                           Nonzero real number

 w                           Integral real number (for
                             example, 3.0)

 Table B.1  Negative INTEGER or REAL Raised to a REAL Power


                                                       Version     Version
 Base        Exponent                                  4.0         3.3
 Type        Type         Formula      Example         Returns     Returns

 INTEGER     REAL         (-n)w        (-3)^(3.0)      -27.0       Error
 INTEGER     REAL         (-n)f        (-1)^(1.5)      Error       Error
 REAL        REAL         (-r)^(w)     (-3.0)^(3.0)    -27.0       Error
 REAL        REAL         (-r)^(f)     (-1.0)^(1.5)    Error       Error


 Table B.2  Zero Raised to a Negative Power


                                                    Version       Version
 Base        Exponent                               4.0           3.3
 Type        Type         Formula                   Returns       Returns

 INTEGER     INTEGER      0^(-n)                    Error         Error
 REAL        INTEGER      0.0^(-n)                  Error         Error
 REAL        REAL         0.0^(-r)                  Error         Error
 COMPLEX     INTEGER      (0.0,0.0)^(-n)            Error         (0.0,0.0)
 COMPLEX     REAL         (0.0,0.0)^(-r)            Error         (0.0,0.0)
 COMPLEX     COMPLEX      (0.0,0.0)^(-(r,0))        Error         (0.0,0.0)


 Table B.3  COMPLEX Zero Raised to a COMPLEX Power

                                                    Version       Version
 Base        Exponent                               4.0           3.3
 Type        Type         Formula                   Returns       Returns

 COMPLEX     COMPLEX      (0.0, 0.0)^(+r,0.0)       (0.0,0.0)     (0.0,0.0)
 COMPLEX     COMPLEX      (0.0,0.0)^(0.0,0.0)       (1.0,0.0)     (0.0,0.0)
 COMPLEX     COMPLEX      (0.0,0.0)^(-r,0.0)        Error         (0.0,0.0)
 COMPLEX     COMPLEX      (0.0,0.0)^(+r,s)          (0.0,0.0)     (0.0,0.0)
 COMPLEX     COMPLEX      (0.0,0.0)^(0.0,s)         Error         (0.0,0.0)
 COMPLEX     COMPLEX      (0.0,0.0)^(-r,s)          Error         (0.0,0.0)

 Table B.4  Zero Raised to the Zero Power

                                                    Version        Version
 Base         Exponent                              4.0            3.3
 Type         Type         Formula                  Returns        Returns

 INTEGER      INTEGER      0^(0)                    1              1
 REAL         INTEGER      0.0^(0)                  1.0            1.0
 REAL         REAL         0.0^(0.0)                1.0            1.0
 COMPLEX      INTEGER      (0.0,0.0)^(0)            (1.0,0.0)      (0.0,0.0)
 COMPLEX      REAL         (0.0,0.0)^(0.0)          (1.0,0.0)      (0.0,0.0)
 COMPLEX      COMPLEX      (0.0,0.0)^(0.0,0.0)      (1.0,0.0)      (0.0,0.0)


 B.3.10  List-Directed Output

 In Version 4.0, the conventions for list-directed output have changed. The
 following conventions are used:

   1. Integer output constants are produced with the effect of an I11 edit
      descriptor. (Version 3.3 uses the I12 edit descriptor for this.)

   2. Real and double-precision constants are produced with the effect of
      either an F or an E edit descriptor, depending on the value of the
      constant c in the following range:

      1<=c<10^(7)

      a.   If c is within the range, the constant is produced by using
           0PF15.6 for single precision and 0PF24.15 for double precision.
           In Version 3.3, 0PF16.7 is used for single precision and
           0PF23.14 is used for double precision.

      b.   If c is outside the range, the constant is produced using
           1PE15.6E2 for single precision and 1PE24.15E3 for double
           precision. The value 0 is printed with this format. (In Version
           3.3, 1PE14.6E3 is used for single precision and 1PE21.13E3 is
           used for double precision.)

      The same field widths are used to force the constants in both cases to
      line up on a printed page.


 B.3.11  DO-Loop Ranges

 The code generated for DO loops in Version 4.0 uses the standard formula for
 determining the loop iteration count, which is, consequently, limited to the
 maximum allowable integer size. In Version 3.3, the code generated for DO
 loops allows more iterations than the maximum allowable integer value; for
 example, if the $STORAGE:2 metacommand is in effect, a DO loop of the
 following form loops 65,535 times in Version 3.3 but is illegal in Version
 4.0:

            DO 200 I = -32767,32767


 B.3.12  Object Compatibility

 Sections B.3.13-B.3.15 discuss compatibility between object files compiled
 with Versions 4.0, 3.3, and 3.2. If possible, you should recompile programs
 and subprograms to take advantage of the improved code generated by Version
 4.0. If you cannot do this (for example, if the source files are
 unavailable), you can continue to link object files generated by Version 3.3
 with those generated by Version 4.0. However, you should read the
 information in the following paragraphs to make sure that object files
 compiled under the two versions link correctly.


 B.3.13  Library Compatibility

 If your program mixes modules compiled with Version 4.0 and modules compiled
 with Version 3.3, you must link them with the FORTRAN.LIB library that comes
 with Version 4.0 in addition to a standard FORTRAN library built by the
 SETUP program. The SETUP program installs the Version 4.0 FORTRAN.LIB if you
 request compatibility with Version 3.3 or 3.2. This library is required
 because the standard Version 4.0 libraries are different internally from the
 standard Version 3.3 and Version 3.2 libraries, and the code generated by
 the Version 4.0 compiler accesses these libraries differently. Thus, special
 interfaces are required so that the code produced by the two versions can
 work together.

 The Version 4.0 FORTRAN.LIB library includes the interfaces required to work
 with Version 3.3 and Version 3.2 modules. It contains all the external
 interfaces supported by Version 3.3 and Version 3.2 FORTRAN.LIB. However,
 the interfaces in the Version 4.0 library generally use parts of the
 standard Version 4.0 library to perform their processing.

 FORTRAN.LIB is not required if all of the object files you are linking were
 compiled with Version 4.0. Also, since modules compiled with Versions 3.3
 and 3.2 have library search directives for FORTRAN.LIB embedded in them, you
 do not need to specify FORTRAN.LIB explicitly when you link. However, this
 library should be found in the standard place specified in the LIB
 environment variable.

 You can use Version 3.3 and Version 3.2 modules with Version 4.0 modules
 that are compiled with any /FP compiler option, subject to the restrictions
 that apply to the Version 4.0 modules: that is, you cannot link with an
 alternate math library (LLIBFORA.LIB) if any of the modules contains in-line
 instructions.

 However, you must still tell the SETUP program to include the compatibility
 math interfaces in the LLIBFORx.LIB library that it builds if you plan to
 use the library with Version 3.3 and Version 3.2 modules. The resulting
 program will not be affected, but the library that SETUP builds will be
 slightly larger. (The math interfaces are not included in FORTRAN.LIB since,
 unlike the standard FORTRAN libraries built by SETUP, FORTRAN.LIB is not
 typically associated with a particular /FP option.)


 B.3.14  Mixing Version 4.0 and Version 3.3 Modules

 Version 4.0 modules that are linked with Version 3.3 modules must be
 compiled using the large memory model. This model is the default for Version
 4.0 FORTRAN programs. (See Chapter 2 of Microsoft FORTRAN Advanced Topics
 for more information about memory models.)

 In most cases, the calling and argument-passing conventions are the same in
 Versions 3.3 and 4.0, so that routines compiled under either version can
 call each other freely. The only exception is the case of a Version 3.3
 routine calling a Version 4.0 routine and passing a CHARACTER*(*) argument.
 (This situation is most likely to arise when a Version 3.3 program passes a
 subprogram as an argument to another subprogram compiled with Version 4.0.)

 A routine compiled with Version 3.3 cannot call a Version 4.0 routine that
 has CHARACTER*(*) formal arguments. Version 4.0 expects the caller to
 specify the lengths of all such arguments in a special way. Since Version
 3.3 does not support arguments of this type, Version 3.3 programs cannot
 pass the argument length. Any such call gives undefined results at run time.
 (This change was made in order to support the more powerful feature of the
 full ANSI FORTRAN-77 standard.) This problem does not arise in calls from
 Version 4.0 routines to Version 3.3 routines. Version 4.0 routines pass the
 length of a CHARACTER*(*) argument in such a way that Version 3.3 routines
 can safely ignore it.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Certain additional rules apply if you are linking C modules with FORTRAN
    modules. Chapter 4 of Microsoft FORTRAN Advanced Topics explains these
    rules.
 ───────────────────────────────────────────────────────────────────────────

 If you compile a Version 3.3 source file that includes the STATUS='NEW'
 option and link the resulting object file with a Version 4.0 library that
 includes the Version 3.3 compatibility package, the STATUS='NEW' option is
 mapped to STATUS='UNKNOWN'. This results in behavior more similar to the
 Version 3.3 implementation of the STATUS='NEW' option.


 B.3.15  Mixing Version 4.0 and Version 3.2 Modules

 In general, programs can mix modules compiled with Versions 4.0 and 3.2 of
 Microsoft FORTRAN. However, the following considerations apply:

   ■  All considerations that apply to mixing Version 3.3 modules with
      Version 4.0 modules also apply to mixing Version 3.2 modules with
      Version 4.0 modules.

   ■  You must compile any Version 4.0 modules in these programs with the /Gr
      option to the FL command. This is because the code that Version 4.0
      generates by default preserves the SI and DI registers for the duration
      of a subprogram, while the code that Version 3.2 generates does not. If
      you specify /Gr, the Version 4.0 code does not expect the SI and DI
      registers to be preserved.


 B.4  Changes for Version 4.0

 Sections B.4.1-B.4.4 discuss changes and enhancements to the Microsoft
 FORTRAN Compiler for Version 4.0. These changes fall under the following
 categories:

   ■  Enhancements and additions to the compiler and linker

   ■  Run-time library changes

   ■  Language changes


 B.4.1  Enhancements and Additions to the Compiler and Linker

 Several features have been added to, or changed in, Version 4.0 of the
 Microsoft FORTRAN Compiler and the Microsoft Overlay Linker (LINK) to make
 them easier to use. These features should not affect your source code, but
 you may need to revise existing batch files or MAKE description files so
 that they work correctly with Version 4.0.


 B.4.1.1  The FL Command

 In Microsoft FORTRAN, a new command, FL, automatically executes the compiler
 and the linker. The options associated with this command give you
 considerable flexibility in controlling compilation and linking.

 You can specify the /c option with the FL command to compile without
 linking. You can invoke the linker separately after you compile, either
 through FL or through the LINK command.

 The FL command performs many of the same functions as any batch files that
 you may have created to compile and link your FORTRAN programs. It also
 allows you to specify on the command line all files you want to compile and
 link and all options for controlling the process. You can include wild-card
 characters in the files you specify so that you can easily compile and link
 more than one file. FL automatically prompts you if it cannot find a file
 that it needs at any point during compilation and linking. Note that you
 must give the entire source-file name, including the .FOR extension, to the
 FL command. If you do not include the .FOR extension, FL interprets the file
 name as an object-file name. If you wish to convert existing batch files so
 they compile and link correctly under Version 4.0, be sure that you
 substitute the appropriate FL command for any FOR1, PAS2, PAS3, and LINK
 commands that may have been in the original batch files.

 See Chapter 7, "The FL Command," for detailed instructions on using the FL
 command for program compilation and linking.


 B.4.1.2  Changes to the Linker

 Several linker options have been added for Version 4.0. You can specify
 these options either by using the /link option of the FL command, or by
 using the LINK command if you choose to invoke the linker separately.

 The following list gives the new linker options:

 Option                          Task

 /CO                             Prepares a special executable
                                 file for use with the Microsoft
                                 CodeView window-oriented debugger

 /DO                             Enforces the default
                                 segment-loading order for
                                 Microsoft language products
                                 (including Microsoft FORTRAN)

 /E                              Packs the executable file during
                                 linking

 /HE                             Lists all LINK command options on
                                 standard output

 /I                              Displays information about the
                                 effect of the linking process on
                                 standard error output


 B.4.1.3  Memory Models

 When you compile a program using Version 4.0 of the Microsoft FORTRAN
 Compiler, you can choose a memory model to be used for your program. The
 memory model you choose specifies how memory for the code and data in your
 program will be allocated. Three memory models are available: medium, large,
 and huge. You choose a memory model by specifying the /AL (large), /AM
 (medium), or /AH (huge) option with the FL command at compile time. The
 default is the large memory model. (See Chapter 2 of Microsoft FORTRAN
 Advanced Topics for information on the use of memory models.)

 All programs compiled with Version 3.3 of the Microsoft FORTRAN Compiler are
 large-model programs. The large model is the default memory model for
 Version 4.0.

 For programs that mix modules compiled under Versions 3.3 and 4.0, Version
 4.0 modules cannot be compiled using the medium memory model. If this model
 is used for the Version 4.0 modules, the program may produce undefined
 results, although it may appear to link correctly.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    Using the $LARGE metacommand on an entire program has the same effect as
    specifying the huge memory model, except that fixed-size arrays are
    implicitly declared with the HUGE attribute. The $LARGE metacommand is
    not associated with the large memory model.
 ───────────────────────────────────────────────────────────────────────────


 B.4.2  Run-Time Library Changes

 The following changes have been made to the libraries provided with Version
 4.0 of the Microsoft FORTRAN Compiler:

   ■  The auxiliary library DECMATH.LIB, which supported an alternative
      floating-point format in Version 3.3, is no longer provided.

   ■  The library structure for Version 4.0 is considerably different from
      the structure for Version 3.3. During installation, you can specify the
      memory model, the math package you wish to use, and various other
      options. Then the SETUP program builds a library according to your
      specifications.  The memory-model and floating-point options you
      specify on the FL command line allow your program to be linked with the
      library you build automatically.


 B.4.3  Changes to the Language

 This section lists the changes made to the Microsoft FORTRAN language for
 Version 4.0. For each difference, a reference to the appropriate section in
 the documentation for Version 4.0 or Version 3.3 is given.


 B.4.3.1  Underscore (_) as a Digit

 In Version 4.0, the underscore is classified as a digit, which can be used
 as any character of a name other than the first character. An underscore
 cannot be used in names if the /4Ys option is used in compiling (or the
 $STRICT metacommand is in effect). In Version 3.3, the underscore (_) is
 classified as a special character, which cannot be used in names.


 B.4.3.2  Dollar Sign ($) in Collating Sequence

 In Version 4.0, the dollar sign is classified as an alphanumeric character,
 which can be used in names and which appears after uppercase Z in the
 collating sequence.

 The dollar sign cannot be used as an alphanumeric character in names in the
 following cases:

   ■  If the /4Ys option is used in compiling (or the $STRICT metacommand is
      in effect)

   ■  If the name is declared using the C attribute

 In Version 3.3, the dollar sign ($) is classified as a special character,
 which appears as the first character in the FORTRAN collating sequence.


 B.4.3.3  Significant Characters in Names

 In Version 4.0, only the first six characters in a name are significant,
 unless the /4Nt option is used in compiling or the $NOTRUNCATE metacommand
 is in effect. In this case, the first 31 characters in a name are
 significant.

 In Version 3.3, only the first six characters in a name are significant
 under any circumstances.


 B.4.3.4  Column Restrictions for Source Files

 Version 4.0 allows source code to be in free-form format. The /4Yf option to
 the FL command (and the $FREEFORM metacommand) gives you this choice.

 In Version 3.3, statements in source programs are required to obey the
 standard FORTRAN column restrictions.


 B.4.3.5  Restrictions on Continuation Lines

 In Version 4.0, limits on the number of continuation lines have been
 removed, unless the /4Ys option is used in compiling (or the $STRICT
 metacommand is in effect). In these cases, the compiler  generates an error
 if a statement extends over more than 19 continuation lines or includes more
 than 1,320 characters.

 In Version 3.3, these restrictions are always in effect.


 B.4.3.6  Maximum Character-Value Length

 In Version 4.0, the maximum length of character values is 32,767 characters.
 Character constants are effectively limited to 1,958 characters.

 In Version 3.3, character values can have a maximum length of 127
 characters.


 B.4.3.7  Arithmetic Operations

 In Version 4.0, raising a negative-value operand to an integral real power
 is permitted.

 In Version 3.3, raising a negative-value operand to any real power produces
 an error.


 B.4.3.8  Character Editing and Hollerith Data Types

 In Version 4.0, Hollerith data types can be used with the A edit descriptor
 when an input/output list item is of type INTEGER, REAL, or LOGICAL.


 B.4.3.9  Expressions in Substring Specifications

 In Version 4.0, any type of arithmetic expression can be used to specify the
 first and last characters in a substring, unless the /4Ys is used in
 compiling (or the $STRICT metacommand is specified). In effect, noninteger
 substring expressions are truncated by an implicit use of the INT intrinsic
 function before substring operations are performed. If the /4Ys option (or
 $STRICT metacommand) appears, only integer expressions can be used to
 specify the first and last characters in a substring.

 In Version 3.3, these restrictions are always in effect.

 In Version 4.0 the compiler verifies the following relationships, where
 first is the arithmetic expression that defines the first character in the
 substring, last is the arithmetic expression that defines the last
 character, and length is the length of the character variable:

   ■  first<=last

   ■  1<=first<=length

   ■  1<=last<=length

 If either of these relationships is false and the /4Yb option is used in
 compiling (or the $DEBUG metacommand is in effect), the compiler generates
 an error message. If either of these relationships is false and the /4Yb
 option is not used (or the $DEBUG metacommand is not in effect), the
 substring is undefined.


 B.4.3.10  Array Subscripts

 In Version 4.0, array subscripts can be any arithmetic expression, unless
 the /4Ys option is used in compiling (or the $STRICT metacommand is
 specified). In effect, noninteger subscript expressions are truncated by an
 implicit use of the INT intrinsic function before subscripting operations
 are performed. If the /4Yb option is used in compiling (or the $DEBUG
 metacommand is specified), subscripts are checked on all arrays that are not
 formal arguments, and an error message is generated for invalid subscripts.

 In Version 3.3, array-element references must be integer expressions.


 B.4.3.11  Changes to the Input/Output System

 This section describes changes to the input/output system used in Version
 4.0 of Microsoft FORTRAN.


 Unit Specifiers

 In Version 4.0, unit specifiers can be used more flexibly. The optional
 UNIT= string can appear before the unit specifier in all I/O statements
 except PRINT,  INQUIRE with a FILE= option, and the EOF intrinsic function.
 If the optional UNIT= string appears in the unit specifier, the specifier
 can appear at any position in the I/O control list. This change was made to
 conform with the ANSI full-language standard for FORTRAN.

 In Version 3.3, the unit specifier must appear in the first position.

 In Version 4.0, the following external unit specifiers can be reconnected to
 another file:

 External Unit                     Description

 0                                 Initially represents the keyboard
                                   or the screen

 5                                 Initially represents the keyboard

 6                                 Initially represents the screen

 If you connect any of these specifiers to a different file using an OPEN
 statement and then close that file, the specifier resumes its preconnected
 status.


 Output Lists

 In Version 4.0, arbitrary expressions used in an output list can begin with
 a left parenthesis.

 In Version 3.3, arbitrary expressions used in an output list cannot begin
 with a left parenthesis because left parentheses are reserved for implied-DO
 lists.


 Format Specifiers

 In Version 4.0, statement labels, integer variables, character expressions,
 character variables, or character arrays can be used as format specifiers.
 If the /4Ys option is not used in compiling (and the $STRICT metacommand is
 not in effect), noncharacter arrays can also be used.

 In Version 3.3, only statement labels, integer variables, character
 expressions, or character variables can be used as format specifiers.


 Backslash (\) Edit Descriptor

 In Version 4.0, the backslash (\) edit descriptor is only recognized for
 files connected to terminal devices such as screens or printers. Otherwise,
 it is ignored.

 In Version 3.3, the backslash edit descriptor is recognized for all file
 types.


 B.4.3.12  Assignment Statement (Computational)

 In Version 4.0, the expression in a computational assignment statement can
 be a Hollerith constant. A Hollerith constant can be assigned to any type of
 variable. The normal rules for padding and truncation of character data
 types also apply to Hollerith constants.

 In Version 3.3, Hollerith constants cannot be used in assignments.


 B.4.3.13  CALL Statement

 In Version 4.0, the actuals parameter can include Hollerith constants.
 Hollerith constants cannot be passed to character formal arguments.

 In Version 3.3, Hollerith constants cannot be used in CALL statements.


 B.4.3.14  DATA Statement

 In Version 4.0, the nlist parameter in a DATA statement can include
 substring names and implied-DO lists, and the clist parameter can include
 Hollerith constants. The normal rules for padding and truncation of
 character data types also apply to Hollerith constants.

 In Version 3.3, these constructs are not allowed.


 B.4.3.15  BACKSPACE, ENDFILE, and REWIND Statements

 In Version 4.0, the BACKSPACE, ENDFILE, and REWIND statements can include an
 ERR= option to specify the flow of control after errors, and an  IOSTAT=
 option to specify a variable to be used to indicate error or end-of-file
 status.

 In Version 3.3, the only option allowed in the BACKSPACE, ENDFILE, and
 REWIND statements is a unit specifier, which specifies the unit location of
 the file that the command acts on.


 B.4.3.16  CLOSE and OPEN Statements

 In Version 4.0, the CLOSE and OPEN statements can include the ERR= option to
 specify the flow of control if an error occurs during statement execution.
 In addition, the OPEN statement can include the BLANK= option to indicate
 how blanks are interpreted in numeric input and the BLOCKSIZE= option to
 assign a new I/O-buffer size for the file being opened.


 B.4.3.17  DIMENSION Statement

 In Version 4.0, no restriction is placed on the number of array dimensions
 unless the /4Ys option is used in compiling, or the $STRICT metacommand is
 set. In that case, arrays are restricted to seven dimensions.

 Arrays in Version 3.3 are always restricted to seven dimensions.

 In Version 4.0, lower array-dimension bounds can be specified explicitly and
 can be positive, negative, or 0. If a lower dimension bound is not
 specified, it is 1 by default.

 The upper and lower bounds are checked according to the following rules:

   ■  If the upper and lower dimension bounds are constants, the compiler
      verifies that the upper dimension bound is greater than or equal to the
      lower dimension bound. If it is not, the compiler generates an error
      message.

   ■  If either the upper or the lower dimension bound is not a constant, the
      /4Yb compiler option must be used (or the $DEBUG metacommand must be in
      effect) if you want to verify that the upper bound is greater than or
      equal to the lower bound.

 ───────────────────────────────────────────────────────────────────────────
 NOTE
    If all of an array's dimensions are declared with no lower bounds and
    with upper bounds of 1, no bounds checking is performed, even if /4Yb or
    $DEBUG is used. In this case, the array is treated the same as an
    adjustable-size array, except that the declared size of the array is used
    to determined whether or not huge addressing is used.
 ───────────────────────────────────────────────────────────────────────────

 Dimension declarators in Version 3.3 do not include lower bounds; the lower
 bound is always 1.

 In Version 4.0, a dimension declarator can be an arithmetic expression,
 unless the /4Ys option is used in compiling (or the $STRICT metacommand is
 specified). The result of the expression is truncated to an integer by an
 implicit use of the INT intrinsic function. If an arithmetic expression is
 used as a dimension declarator, it cannot contain function or array-element
 references. If a dimension declarator with variables is used to declare an
 adjustable-size array, the variables either must be formal arguments to a
 routine or must exist in a common block. Also, the array itself must be a
 formal argument.


 B.4.3.18  DO Statement

 In Version 4.0, loop indices in a DO statement can be integer, real, or
 double-precision expressions.

 In Version 3.3, loop indices in a DO statement must be integer expressions.


 B.4.3.19  INQUIRE Statement

 In Version 4.0, the INQUIRE statement can include the BINARY= option to
 indicate whether the file (or the file connected to the unit) specified in
 the statement is in binary format. It can also include the BLOCKSIZE=
 option, which reports the I/O buffer size for the file.

 In Version 3.3., these options do not appear.


 B.4.3.20  PAUSE Statement

 In Version 4.0, the PAUSE statement allows the user to enter a blank line to
 return control to the program. It also allows the user to execute one or
 more DOS commands before returning control to the program. If this feature
 is used, the subdirectory containing COMMAND.COM should be part of the
 user's search path. While the program is suspended, the user can enter
 either of the following:

   ■  A DOS command. After the command is executed, control is automatically
      returned to the program.

   ■  The word COMMAND (uppercase or lowercase). After entering COMMAND, the
      user can enter as many DOS commands as desired, then type EXIT
      (uppercase or lowercase) to return control to the program.

 In Version 3.3, the PAUSE statement only allows the user to enter a blank
 line to return control to the program.


 B.4.3.21  READ and WRITE Statements

 In Version 4.0, the READ and WRITE statements can include the FMT=
 formatspec option, which can appear at any position in the I/O control list.
 However, the READ statement must include a unit specifier if the
 FMT=formatspec option is used.

 In Version 3.3, a format specifier must be the second argument in a
 formatted READ or WRITE statement.

 In Version 4.0, the unit specifier can be omitted in a READ statement of the
 following form:

 READ formatspec, iolist

 In this form of the READ statement, the unit is assumed to be the keyboard
 (*) unit.

 In Version 3.3, a unit specifier must be the first argument to a READ
 statement.


 B.4.3.22  STOP Statement

 In Version 4.0, if the message parameter in a STOP statement is an integer,
 the program displays this value on the screen and returns the
 least-significant byte of this value to the operating system. (This is a
 value between 0 and 255, inclusive.) If the message parameter is not an
 integer, the program displays this value on the screen and returns 0 to the
 operating system.

 In Version 3.3, if the message parameter in a STOP statement is an integer,
 the program displays the specified integer.


 B.4.3.23  Type Statements

 Type statements in Version 4.0 can be used to initialize the values of
 variables. However, variables that appear in COMMON and EQUIVALENCE
 statements cannot be initialized in this way.

 Also, length specifiers in type statements in Version 4.0 can appear either
 before or after dimension declarators.


 B.4.3.24  Conditional Compilation

 In Version 4.0, the /4cc option of the FL command (or the $DEBUG:string
 metacommand) can be used to specify conditional compilation. If one or more
 letters follows the /4cc option (or $DEBUG metacommand), lines in the source
 file that have one of those letters in column 1 are compiled into the
 program. Lines beginning with other characters are treated as comments.


 B.4.4  New Language Features

 Sections B.4.4.1-B.4.4.9 discuss new features for Version 4.0 of Microsoft
 FORTRAN and the changes you may have to make to source programs to take
 advantage of these features.


 B.4.4.1  INTEGER*1 and LOGICAL*1 Data Types

 Version 4.0 supports two new data types: INTEGER*1 and LOGICAL*1.

 An INTEGER*1 value occupies 1 byte and can be any number in the range -127
 to 127, inclusive. In an arithmetic expression, INTEGER*1 is the
 lowest-ranked operand. If an INTEGER*1 value is converted to an INTEGER*2
 value, the INTEGER*1 value is used as the least-significant part of the
 INTEGER*2 value, and the most-significant part is filled with copies of the
 sign bit (that is, it is sign extended). A new intrinsic function, INT1, is
 provided to convert values to type INTEGER*1. A LOGICAL*1 value occupies 1
 byte of storage. The value of this byte is either 0 (.FALSE.) or 1 (.TRUE.).


 B.4.4.2  C Strings

 The following new string escape sequences from the C language have been
 added for Version 4.0:

 Sequence                           Character

 \""                                Double quote

 \xhh                               Hexadecimal bit pattern (where h
                                    is between 0 and F, inclusive)

 \a                                 Bell

 See Chapter 2 in Microsoft FORTRAN Advanced Topics, for more information
 about C strings.


 B.4.4.3  Concatenation Operator

 Version 4.0 supports the use of the concatenation operator (//) in character
 expressions.


 B.4.4.4  New Intrinsic Functions

 New intrinsic functions that perform data-type conversion and bit
 manipulation have been added for Version 4.0.


 Data-Type Conversion

 The following list summarizes the new intrinsic functions that are used for
 data-type conversion:

 Function                          Operation

 INT1                              Converts arguments to type
                                   INTEGER*1

 HFIX                              Converts arguments to type
                                   INTEGER*2

 JFIX                              Converts arguments to type
                                   INTEGER*4

 See Chapter 5, "Intrinsic Functions and Additional Procedures," for more
 information about these functions.


 Bit Manipulation

 In Version 4.0, several new intrinsic functions can be used to perform
 bit-wise operations on variables. The list below summarizes these new
 intrinsic functions:

 Function                          Operation

 BTEST                             Bit test

 IAND                              Logical product

 IBCHNG                            Bit change

 IBCLR                             Bit clear

 IBSET                             Bit set

 IEOR                              Exclusive or

 IOR                               Inclusive or

 ISHA                              Arithmetic shift

 ISHC                              Rotate

 ISHFT                             Logical shift

 ISHL                              Logical shift

 NOT                               Logical complement

 All of these functions except NOT and BTEST accept two arguments of type
 INTEGER, INTEGER*1, INTEGER*2, or INTEGER*4 and return a result of the same
 type. If two arguments with different INTEGER types are given, the larger of
 the two types is returned (provided that it is also a legal type).

 NOT accepts one argument of one of these types and returns a result of the
 same type. BTEST accepts two arguments of one of these types and returns a
 LOGICAL result. All of these functions can be passed as actual arguments.


 B.4.4.5  New Time and Date Functions

 New subroutines and functions that get and set the date and time have been
 added for Version 4.0. The following list summarizes these functions:

 Function              Operation

 GETDAT                Gets the system date

 GETTIM                Gets the system time

 SETDAT                Sets the system date

 SETTIM                Sets the system time

 See Chapter 5, "Intrinsic Functions and Additional Procedures,"  for more
 information about these functions.


 B.4.4.6  Z Edit Descriptor

 The new Z repeatable edit descriptor allows you to specify hexadecimal
 editing in input/output lists. This edit descriptor has the form Zw, which
 specifies a field that is w characters wide. Hexadecimal digits A-F are
 output in uppercase. See Section 3.7.2 for rules for the use of this edit
 descriptor.


 B.4.4.7  ENTRY Statement

 ENTRY specifies an entry point for a subroutine or external function.


 B.4.4.8  PRINT Statement

 PRINT specifies output to the screen (unit *).


 B.4.4.9  $[NO]DECLARE, $[NO]FREEFORM, and $[NO]TRUNCATE Metacommands

 Six new metacommands have been added to Version 4.0 of Microsoft FORTRAN
 $DECLARE, $NODECLARE, $FREEFORM, $NOFREEFORM, $TRUNCATE, and $NOTRUNCATE.

 The $DECLARE metacommand causes the compiler to display warning messages for
 variables that are not declared in type statements. The $NODECLARE
 metacommand suppresses these warnings. The $NODECLARE metacommand is the
 default. Note that the /4Yd compiler option has the same effect as the
 $DECLARE metacommand, and the /4Nd compiler option has the same effect as
 the $NODECLARE metacommand.

 The $FREEFORM metacommand tells the compiler that the source program ignores
 the standard FORTRAN column restrictions (labels in columns 1-5,
 continuation characters in column 6, statements in columns 7-72, and any
 columns beyond 72 ignored). The $NOFREEFORM metacommand tells the compiler
 that the source program observes these column restrictions. $NOFREEFORM is
 the default. Note that the /4Yf compiler option has the same effect as the
 $FREEFORM metacommand, and the /4Nf compiler option has the same effect as
 the $NOFREEFORM metacommand.

 The $TRUNCATE metacommand tells the compiler to generate warning messages
 for any names longer than six characters. This option makes it easier to
 port your programs to other systems. The $NOTRUNCATE metacommand tells the
 compiler to treat the first 31 characters in a name as significant.
 $TRUNCATE is the default. Note that the /4Yt compiler option has the same
 effect as the $TRUNCATE metacommand, and the /4Nt compiler option has the
 same effect as the $NOTRUNCATE metacommand.

 See Chapter 6, "Metacommands," for more information on these metacommands.


 Appendix C  Compiler and Linker Limits

 ───────────────────────────────────────────────────────────────────────────

 This appendix discusses the limits imposed by the Microsoft FORTRAN Compiler
 and the Microsoft Segmented-Executable Linker (for example, the maximum
 length of an identifier) and suggests programming strategies for avoiding
 these limits.


 C.1  Compiler Limits

 To operate the Microsoft FORTRAN Compiler you must have sufficient disk
 space for the compiler to create temporary files used in processing. The
 required space is approximately two times the size of the source file.

 Table C.1 summarizes the limits imposed by the Microsoft FORTRAN Compiler.
 If your program exceeds one of these limits, an error message will inform
 you of the problem.

 Table C.1  Limits Imposed by the Microsoft FORTRAN Compiler

 Program Item               Maximum Limit

 Actual arguments           Number per subprogram: approximately 64

 Character constants        Length: approximately 1,900 bytes

 Names                      Length: 31 bytes (default) or 6 bytes (if /4Ys or
                            /4Yt is used in compiling or if $STRICT or
                            $TRUNCATE metacommand is in effect); in either
                            case, additional characters are discarded

                            Internal, length: 40 bytes per module: 20,000
                            names

 Simple variables           Number of simple variables per subprogram:
                            approximately 5,000 (depending on lengths of
                            variable names)

 Statements                 Levels of nesting: approximately 40 levels

 ENTRY statements           Number per subroutine: 32,000

 FORMAT statements          Number per module: 20,000

                            Format length: approximately 1,900 characters

                            Memory limitations: in medium-model programs, no
                            more than 64K internal formats in the default
                            data segment

                            Number of errors per statement: 10 errors

 GOTO statements            Number per subroutine: 64
 (assigned)

 INTEGER items              Size: 128 bytes for a string of digits

 Include files              Levels of nesting: 10 levels


 The compiler does not set explicit limits on the number and complexity of
 declarations, definitions, and statements in an individual function or in a
 program. If the compiler encounters a function or program that is too large
 or too complex to be processed, it produces an error message to that effect.

 During compilation, large programs are most often limited in the number of
 identifiers allowed in any one source file. They are also occasionally
 limited by the complexity of the program or one of its statements.


 C.1.1  Limits on Number of Names

 The Microsoft FORTRAN Compiler limits the number of names you can use in a
 source program. The compiler creates symbol-table entries for the names
 declared in source programs. Symbol-table entries are created for the
 following objects:

   ■  The program

   ■  Subroutines and functions declared or referenced in the program unit

   ■  Common blocks and variables

   ■  Statement functions

   ■  Formal parameters

   ■  Local variables

 Common variables, statement functions, formal parameters, and local
 variables are required only while the subroutine or function containing them
 is being compiled. These names are discarded at the end of the subroutine,
 and the space they used is made available for other names. Hence, you can
 create much bigger programs by splitting up your code into more subroutines
 and functions so that the space for local names can be shared. You can also
 place the subroutines and functions into their own files and compile them
 separately, since this usually reduces the number of names in groups being
 used per module.


 C.1.2  Limits on Complicated Expressions

 The compiler may run out of memory when it encounters any of the following:

   ■  A deeply nested statement or expression

   ■  A large number of error messages

   ■  A large block of specification statements (EQUIVALENCE statements in
      particular)

 Usually, if pass 1 runs successfully on a program without running out of
 memory, pass 2 will also run successfully, except for complicated basic
 blocks. A basic block is defined as follows:

   ■  A sequence of statements with no labels or other breaks

   ■  A sequence of statements containing long expressions or parameter lists
      (especially including I/O statements or character expressions)

 Pass 2 makes a smaller number of symbol-table entries than pass 1 (for
 example, for the program, subroutines, and functions declared or referenced
 in the program unit, for common blocks, and for many of the transcendental
 functions called in a program). If pass 2 runs out of memory, it displays a
 line-number reference and one of the following messages:

      out of heap space

      expression too complex, please simplify

 If a particularly long expression or parameter list appears near this line,
 break up the expression or parameter list by assigning parts of the
 expression to local variables or by using multiple WRITE statements. If this
 does not work, add labels to statements to break the basic block.


 C.1.3  Limits on Character Expressions

 Use the following programming strategies to avoid compiler limitations when
 initializing or assigning values to large character variables or array
 elements:

   ■  Use smaller pieces

   ■  Use substrings

   ■  Use EQUIVALENCE statements to assign values to a character array

 To avoid compiler limitations on character expressions, assign pieces of the
 character value to smaller variables or substrings. Just having nonconstants
 in the expression causes more of the expression to be evaluated at run time
 instead of at compile time, thus avoiding the 1,900-character compile-time
 limit on constants.


 C.2  Linker Limits

 Table C.2 summarizes the limits imposed by the linker. If you encounter one
 of these limits, you must adjust your program so that the linker can
 accommodate it.

 Table C.2  Limits Imposed by the Microsoft Segmented-Executable Linker

 Item                       Limit

 Symbol table               Only limited by available memory

 DOS load-time              Default is 32K. If /EXEPACK is used, the maximum
 relocations                is 512K.

 External symbols per       1,023
 module

 Groups                     Maximum number is 21, but the linker always
                            defines DGROUP so the effective maximum is 20.

 Overlays                   63

 Logical segments           128 by default; however, this maximum can be set
                            by using the /SEGMENTS option of the LINK
                            command.

 Libraries                  32

 Group definitions per      21
 module

 Physical segments per      255
 module

 Stack                      64K


 C.3  Run-Time Limits

 When running under MS-DOS or OS/2, a FORTRAN program cannot open more than
 20 files at one time. If an OS/2 program uses either the multithread library
 or a dynamically-linked FORTRAN run-time library, this limit is extended to
 40 files. In practice, the actual limit might be slightly less, depending on
 how the operating system uses available memory.

 Exceeding this limit will halt program execution and produce a run-time
 error message. The message varies, depending on whether you have exceeded
 the run-time limit set by FORTRAN, or the operating system limit on the
 number of open files.

 You can increase the maximum number of open files with the following
 procedures. In general, you should increase both the FORTRAN run-time limit
 and the operating system limit; if you have exceeded the FORTRAN run-time
 limit, increasing just the operating system limit will have no effect.


 C.3.1  Increasing the Maximum Number of Open Files

 FORTRAN 5.0 allows you to increase the maximum number of open files. To do
 this, you must be running under DOS 3.3 (or a later version), or under OS/2
 (any version).

 The following instructions refer to two .ASM files. These files are included
 on your Microsoft FORTRAN distribution disks in directories called STARTUP,
 STARTUP\DOS  and STARTUP\OS2, whose location is specified in the PACKING.LST
 file on the SETUP disk. The appropriate .ASM files for each operating system
 are in the corresponding subdirectory. Both multithread and
 dynamically-linked FORTRAN run-time libraries require special considerations
 (See Section C.3.3, below).


 C.3.1.1  Increasing File Handles

 Edit the startup source file CRT0DAT.ASM to increase the number of file
 handles. Change the number in the line

            _NFILE_ = 20

 to the maximum number of file handles desired. The limit is 256.


 C.3.1.2  Increasing Units

 The next step is to increase the size of the table which FORTRAN uses to
 manage units. Edit the source file UNIT.ASM so that the number in the line

            _NFILE_ = 20

 equals the same value you chose in CRT0DAT.ASM. Note that the number of
 handles must always be greater than or equal to the number of units.
 Therefore, if you increase the number of units in UNIT.ASM, there must be at
 least as many file handles specified in CRT0DAT.ASM.


 C.3.1.3  Increasing the System Limit

 To have more than 20 files open at one time, you must increase the file
 limit that the operating system imposes on your process. To do this, you
 must increase both the system-wide limit and the per-process limit.


 C.3.1.4  Increasing the System-Wide Limit

 You can increase the system limit on the number of open files by changing
 the FILES specification in the CONFIG.SYS file. If you wanted 100 files, you
 would place the following statement in your CONFIG.SYS file (or change the
 existing FILES statement):

            FILES=100


 C.3.1.5  Increasing the Per-Process Limit

 You must also increase the number of files the operating system makes
 available for your particular process. This is done by enabling the
 appropriate commented-out code in CRT0DAT.ASM.

 For example, the MS-DOS version of CRT0DAT.ASM contains the following
 commented-out code:

            ;     mov      ah, 67H
            ;     mov bx,  _NFILE_
            ;     callos

 In the OS/2 version of CRT0DAT.ASM, this code appears as a call to
 DOSSETMAXFH. Under OS/2, you must also enable the 'extern DOSSETMAXFH:far'
 declaration that appears near the beginning of the file.

 In either case, remove the semicolons to enable this code.


 C.3.2  Using the Modified Files

 After modifying CRT0DAT.ASM and UNIT.ASM, assemble the files, using the
 batch files and make files in the STARTUP  directory on the distribution
 disks. To use the new object code, either explicitly link your program with
 the new CRT0DAT.OBJ and UNIT.OBJ files, or replace the CRT0DAT.OBJ and
 UNIT.OBJ objects in the appropriate memory-model version of the FORTRAN
 run-time library.


 C.3.3  Multithread and Dynamic Link Applications

 To increase the default number of files that may be opened when linking with
 the multithread or dynamic link libraries from 40, you need only edit the
 UNIT.ASM file. Set the _NFILE_ constant defined between "else" and "endif"
 to the desired number of files, as described above. When assembling
 UNIT.ASM, be sure to define MTHREAD on the command line in the startup
 makefile to assure the correct conditional assembly:

            masm -DMTHREAD unit. asm

 You do not need to edit CRT0DAT.ASM. Instead, you must make an explicit call
 to DOSSETMAXFH in your application. The form of the call is shown in the
 OS/2 version of CRT0DAT.ASM. You must also increase the system-wide limit,
 as explained above.

 See the README.DOC file in the STARTUP  subdirectory of the distribution
 disks for more information on building memory-model-specific startup
 objects.


 Appendix D  Error Messages

 ───────────────────────────────────────────────────────────────────────────

 This appendix lists error messages you may encounter as you develop a
 program, and describes actions you can take to correct the errors. The list
 below indicates where to find error messages for various components of
 Microsoft FORTRAN:

 Component                                  Section

 Command line used to invoke the            Section D.1, "Command-Line Error
 Microsoft FORTRAN Compiler                 Messages"

 Microsoft FORTRAN Compiler                 Section D.2, "Compiler Error
                                            Messages"

 Microsoft FORTRAN run-time                 Section D.3, "Run-Time Error
 libraries and other run-time               Messages"
 situations


 D.1  Command-Line Error Messages

 Messages that indicate errors on the command line used to invoke the
 compiler have one of the following formats:

      command line fatal error D1xxx: messagetext
      command line error D2xxx: messagetext
      command line warning D4xxx: messagetext

 If possible, the compiler continues operation, printing error and warning
 messages. In some cases, command-line errors are fatal and the compiler
 terminates processing. The following messages indicate errors on the command
 line:

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Command-Line Error Message

 D1000          UNKNOWN COMMAND LINE FATAL ERROR Contact Microsoft
                Technical Support
 Number         Command-Line Error Message
               Technical Support

                An unknown error condition has been detected by the compiler.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 D2000          UNKNOWN COMMAND LINE ERROR Contact Microsoft Technical
                Support

                An unknown error condition has been detected by the compiler.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 D2001          too many symbols predefined with -D

                The maximum allowable number of -D switches is 20.

 D2002          a previously defined model specification has been
 Number         Command-Line Error Message
D2002          a previously defined model specification has been
                overridden

                Two different memory models were specified.

 D2003          missing source file name

                You must give the name of the source file to be compiled.

 D2008          too many option flags, 'string'

                Too many letters were given with a specific option (for
                example, with the /O option).

 D2011          only one floating point model allowed

                Only one /FP option can be given on the command line. For
                example, the following FL command would cause this error:

                FL /FPa test1.for /FPi test2.for
 Number         Command-Line Error Message
               FL /FPa test1.for /FPi test2.for

 D2012          too many linker flags on command line

                An attempt was made to pass more than 128 separate options
                and object files to the linker.

 D2016          switch and switch are incompatible

                The two switches you've chosen cannot be used together. The
                -AM and -MD switches, for example, are incompatible.

 D2018          cannot open linker cmd file

                The response file used to pass object-file names and options
                to the linker could not be opened.

                This error can occur if another read-only file has the same
                name as the re-sponse file.

 Number         Command-Line Error Message

 D2019          cannot overwrite the source/object file, 'filename'

                The source-file name was specified as an output-file name.
                The compiler does not allow compiler output files to
                overwrite the source file.

 D2021          invalid numerical argument 'string'

                An invalid number was read (for example, /Gt -1).

 D2022          cannot open help file, 'fl.hlp'

                The /HELP compiler option was used, but the file containing
                the help messages (FL.HLP) was not in the current directory
                or in any of the directories specified by the PATH
                environment variable.

 D2027          could not execute 'filename'

 Number         Command-Line Error Message

                The specified file containing a compiler pass or the linker
                was found but for some reason could not be executed.

                An illegal EXE file format is the most likely cause.

 D4000          UNKNOWN COMMAND LINE WARNING Contact Microsoft Technical
                Support

                An unknown error condition has been detected by the compiler.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 D4001          listing has precedence over assembly output

                Two different listing options were chosen (for example, /Fl
                and /Fa). The assembly-language listing was not created.

 D4002          ignoring unknown flag 'string'
 Number         Command-Line Error Message
D4002          ignoring unknown flag 'string'

                One of the options given on the command line was not
                recognized and was ignored.

 D4003          different processors selected for code generation

                Both the /G0 option and either the /G1 or the /G2 option were
                given; the last /G switch takes precedence.

 D4005          could not execute 'filename'; Please enter new file name
                (full path) or Ctrl+C to quit

                The compiler cannot find the indicated file either in the
                current working directory or in the directories specified by
                the PATH variable. Enter the correct path and file name to
                continue.

 D4008          non-standard model──defaulting to large model libraries

 Number         Command-Line Error Message

                A nonstandard memory model was specified with the /Astring
                option. The library-search records in the object module are
                set to use the large-model libraries.

 D4009          threshold only for far/huge data, ignored

                The /Gt option cannot be used in medium-model programs, which
                have near data pointers. It can be used only in large and
                huge models.

 D4013          combined listing has precedence over object listing

                When /Fc is specified along with either /Fl or /Fa, the
                combined listing (/Fc) is created.

 D4014          invalid value number for 'option'. Default number is used

                An out-of-range value was specified for an FL option. For
                example, the value for the /Sp or /Sl option may have been
 Number         Command-Line Error Message
               example, the value for the /Sp or /Sl option may have been
                too large or too small.

 D4018          .DEF files supported in protected mode only

                You used the /Lr (real-mode libraries) option and specified a
                .DEF file on the command line. .DEF files can only be used
                with protected-mode programs.

 D4019          String too long──truncated to 40 characters

                A string of more than 40 characters was specified for one of
                the following switches: /NM, /NT, /St, or /Ss. The string is
                truncated to 40 characters, and FL continues. This message is
                a warning.


 D.2  Compiler Error Messages

 The error messages produced by the Microsoft FORTRAN Compiler fall into the
 following four categories:

   1. Fatal error messages

   2. Compilation error messages

   3. Recoverable error messages

   4. Warning messages

 The messages for each category are listed in Sections D.2.1-D.2.4 in
 numerical order, with a brief explanation of each error. To look up an error
 message, first determine the message category, then find the error number.
 All messages give the file name and line number where the error occurs. The
 following paragraphs discuss error-message format.


 Fatal Error Messages

 Fatal error messages indicate a severe problem, one that prevents the
 compiler from processing your program any further. These messages have the
 following format:

 filename(line) : fatal error F1xxx: messagetext

 After the compiler displays a fatal error message, it terminates without
 producing an object file or checking for further errors.


 Compilation Error Messages

 Compilation error messages identify actual program errors. These messages
 appear in the following format:

 filename(line) : error F2xxx: messagetext

 The compiler does not produce an object file for a source file that has
 compilation errors in the program. When the compiler encounters such errors,
 it attempts to recover from the error. If possible, it continues to process
 the source file and produce error messages. If errors are too numerous or
 too severe, the compiler stops processing.


 Recoverable Error Messages

 Recoverable error messages are informational only; they do not prevent
 compiling and linking. These messages appear in the following format:

 filename(line) : error F3xxx: messagetext

 Recoverable error messages are similar to warning error messages (see
 below), except that you cannot suppress them using the /W option. (See
 Section 7.20.2 for a description of this option.)


 Warning Messages

 Warning messages are informational only; they do not prevent compilation and
 linking. These messages appear in the following format:

 filename(line) : warning F4xxx : messagetext

 You can use the /W option to control the level of warnings that the compiler
 generates.


 D.2.1  Compiler Fatal Error Messages

 The following messages identify fatal errors. The compiler cannot recover
 from a fatal error; it stops processing after printing the error message.

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Compiler Fatal Error Messages

 F1000          UNKNOWN FATAL ERROR Contact Microsoft Technical Support

                An unknown fatal error has occurred.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 F1001          Internal Compiler Error compiler file 'filename,' line
                'number' Contact Microsoft Technical Support

 Number         Compiler Fatal Error Messages

                An internal compiler error has occurred.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 F1002          out of heap space

                The compiler ran out of dynamic memory space. This usually
                means that your program has many complex expressions.

                Try breaking expressions into smaller subexpressions.

 F1003          error count exceeds number; stopping compilation

                The limit for compilation errors was exceeded.

 F1005          string too big for buffer

                A string in a compiler intermediate file overflowed a buffer.
 Number         Compiler Fatal Error Messages
               A string in a compiler intermediate file overflowed a buffer.

                This internal compiler error could result from initializing a
                very long character string with a DATA statement. Try
                decreasing the length of the character string.

 F1006          write error on compiler intermediate file

                A write error occurred on a compiler intermediate file.

                This could be caused by a faulty disk.

 F1027          DGROUP data allocation exceeds 64K

                Allocation of variables to the default data segment exceeded
                64K.

                For large- and huge-model programs, compile with the /Gt
                option to move items into separate segments.

 Number         Compiler Fatal Error Messages

 F1031          limit exceeded for nesting function calls

                Function calls were nested to more than 30 levels.

 F1032          cannot open object listing file name

                The compiler could not open the given object-listing file for
                writing.

                The file or disk may be write-protected, or the disk is full.

 F1033          cannot open assembly language output file name

                The compiler could not open the given assembly-listing file
                for writing.

                The file or disk may be write-protected, or the disk is full.

 F1035          expression too complex, please simplify
 Number         Compiler Fatal Error Messages
F1035          expression too complex, please simplify

                The compiler could not generate code for a complicated
                expression.

                Try breaking the expression into simpler subexpressions and
                recompiling. Please report this error to Microsoft
                Corporation using the Software Problem Report form at the
                back of this manual.

 F1036          cannot open source listing file name

                The compiler could not open the given source-listing file for
                writing.

                The file or disk may be write-protected, or the disk is full.

 F1037          cannot open object file name

                The compiler could not open the given object file for
 Number         Compiler Fatal Error Messages
               The compiler could not open the given object file for
                writing.

                The file or disk may be write-protected, or the disk is full.

 F1039          unrecoverable heap overflow in Pass 3

                The compiler ran out of dynamic memory space.

                A subroutine may have too many symbols; simplify the
                subroutine and make it smaller.

 F1041          cannot open compiler intermediate file - no more files

                The compiler was unable to open an intermediate file because
                too many files were already open.

 F1043          cannot open compiler intermediate file

                The compiler was unable to open an intermediate file. This
 Number         Compiler Fatal Error Messages
               The compiler was unable to open an intermediate file. This
                could occur if the environment variable TMP was set to a
                nonexistent directory.

                Try setting the environment variable TMP to an existing
                directory, or not setting TMP at all.

 F1044          out of disk space for compiler intermediate file

                The compiler ran out of disk space while writing to an
                intermediate file.

                Try making more disk space available and recompiling.

 F1045          floating-point overflow

                A compile-time evaluation of a floating or complex expression
                resulted in overflow, as shown in the following example:

                      real a,b,c
 Number         Compiler Fatal Error Messages
                     real a,b,c
                         a=10e30
                         b=10e30
                         c=a*b

 F1050          name : code segment too large

                The amount of object code in the named segment was larger
                than 64K.

 F1051          program too complex

                Your program caused the compiler to overflow one of its
                internal tables. For example, this error can occur if your
                program has too many labels.

                The /4Yb compiler option and the $DEBUG metacommand cause a
                large number of labels to be generated. If you encounter this
                message, try recompiling with the /4Nb option or changing the
                $DEBUG metacommand to $NODEBUG in your source file and
 Number         Compiler Fatal Error Messages
               $DEBUG metacommand to $NODEBUG in your source file and
                recompiling; or, if your file contains more than one
                procedure, try compiling the procedures in separate files.

 F1900          maximum memory-allocation size exceeded

                The program tried to allocate more than approximately 1900
                bytes at one time. This is the upper limit for the size of
                character constants. (See Appendix C, "Compiler and Linker
                Limits," for more information.)

 F1901          program too large for memory

                The combination of heap space and stack space overflowed the
                memory configurations of the machine.

 F1902          statement stack underflow

                This is an internal error. The compiler could not interpret
                the nesting of statements.
 Number         Compiler Fatal Error Messages
               the nesting of statements.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 F1903          statement-nesting limit exceeded

                Structured statements were nested too deeply.

                The maximum legal depth is about 40 statements and varies
                slightly depending on the type of statement.

 F1904          illegal command-line option

                This error should never occur.

                If it does, please report it to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 F1907          too many symbols
 Number         Compiler Fatal Error Messages
F1907          too many symbols

                The program overflowed the internal symbol counter.

                There is no set upper limit on the number of symbols allowed
                in a source file. However, in any case, no more than 20,000
                names are allowed in one module.

 F1908          ASSIGN : too many format labels

                The program overflowed the assigned format-label table.

                This error probably occurred because an INTEGER*1 variable,
                which has a limit of 127 labels, was specified. To avoid this
                error, use an INTEGER*2 variable instead. (See Appendix C,
                "Compiler and Linker Limits," for more information.)

 F1909          filename : include file nested too deeply

                More than 10 include files were active at the same time. (See
 Number         Compiler Fatal Error Messages
               More than 10 include files were active at the same time. (See
                Appendix C, "Compiler and Linker Limits," for more
                information.)

 F1910          name : unrecognized option

                This is an internal error; the compiler driver FL caught an
                illegal option.

 F1912          filename : cannot open file

                The specified file could not be opened.

 F1913          name : name too long

                The specified internal file name was more than 14 characters
                long. The compiler creates internal files in the directory
                specified by the TMP environment variable. If the combined
                length of the TMP environment variable and the unique
                internal name exceeds the name-buffer length, this message
 Number         Compiler Fatal Error Messages
               internal name exceeds the name-buffer length, this message
                appears (see Appendix C, "Compiler and Linker Limits," for
                more information).

                Create a smaller TMP environment variable. If no TMP variable
                is specified, this error should never occur.

 F1914          cannot open internal files

                Internal files could not be created.

 F1917          unknown primitive type

                An internal error has occurred.

                Please report this error to Microsoft Corporation using the
                Software Problem Report form at the back of this manual.

 F1918          missing symbol reference

 Number         Compiler Fatal Error Messages

                An internal error has occurred.

                Please report this error to Microsoft Corporation using the
                Software Problem Report form at the back of this manual.

 F1919          unknown constant type

                An internal error has occurred.

                Please report this error to Microsoft Corporation using the
                Software Problem Report form at the back of this manual.

 F1920          illegal -A option

                An invalid memory-model option was given for the FL command
                line.

 F1921          too many ENTRY statements

 Number         Compiler Fatal Error Messages

                More than 32,000 ENTRY statements were used in this
                subprogram. (This error is not likely to occur.)

 F1922          integer string too long

                An integer-constant string of digits overflowed an internal
                buffer. (This error should never occur in normal use.)

                Shorten the integer strings to legal value.

 F1923          CHARACTER constant too long

                A constant of type CHARACTER can have a maximum of
                approximately 1,900 characters. (See Appendix C, "Compiler
                and Linker Limits," for more information.)

 F1924          FORMAT string too long

                A FORMAT statement can have a maximum of approximately 1,900
 Number         Compiler Fatal Error Messages
               A FORMAT statement can have a maximum of approximately 1,900
                characters. (See Appendix C, "Compiler and Linker Limits,"
                for more information.)

 F1925          out of disk space for compiler internal file

                The disk became full while the compiler was writing to an
                internal file.

 F1926          write error on compiler internal file

                An error occurred while the compiler was writing to an
                internal file.

                Please report this error to Microsoft Corporation using the
                Software Problem Report form at the back of this manual.


 D.2.2  Compilation Error Messages

 The messages listed below indicate that your program has errors. When the
 compiler encounters any of the errors listed in this section, it continues
 compiling the program (if possible) and outputs additional error messages.
 However, no object file is produced.


╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Compiler Compilation Error Message

 F2000          UNKNOWN ERROR Contact Microsoft Technical Support

                An unknown compilation error has occurred.

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 F2001          INTEGER value overflow

                An INTEGER constant was too large to be of the specified
                type.

 Number         Compiler Compilation Error Message

 F2002          Hollerith not allowed

                Hollerith constants are not allowed when the /4Ys compiler
                option is used (or the $STRICT metacommand is in effect).

 F2003          illegal base value

                The specified base value was not between 2 and 36, inclusive.

 F2004          INTEGER constant must follow #

                No alphanumerics within the legal range for the base
                immediately (allowing for white space) followed the number
                sign (#).

 F2005          illegal REAL constant

                A floating-point constant was in an illegal form.

 Number         Compiler Compilation Error Message

 F2006          missing ] following attribute string

                The closing bracket for an attribute list was missing.

 F2007          opening quote missing

                The leading quote in a string value of an ALIAS attribute was
                missing.

 F2008          unrecognized attribute

                An item used as an attribute was not a legal FORTRAN
                attribute.

 F2009          character : illegal separator

                An attribute list did not end with a closing right bracket
                (]) and was not continued with a comma (,), or an illegal
                character was used in the list for the $NOTLARGE metacommand.
 Number         Compiler Compilation Error Message
               character was used in the list for the $NOTLARGE metacommand.

 F2010          name : name too long; truncated

                The specified name was more than 31 characters long. (The
                limit on the length of names may be less in some
                environments; see Appendix C, "Compiler and Linker Limits,"
                for more information.)

 F2011          octal value too big for byte

                An octal value was not within the range 8#0 to 8#377.

 F2012          name : already specified in $[NOT]LARGE

                The specified item appeared in the lists for both the $LARGE
                and the $NOTLARGE metacommands. (The message shows the
                metacommand that appears first in the source program.)

 F2013          too many continuation lines
 Number         Compiler Compilation Error Message
F2013          too many continuation lines

                Either the /4Ys compiler option was used in compiling or the
                $STRICT metacommand was in effect, and more than 19
                continuation lines were used.

 F2015          $DEBUG:'<debug list>' : string expected

                A quoted string was expected after a $DEBUG metacommand.

 F2016          $IF: no matching $ENDIF

                No matching ENDIF was found when an $IF was used.

 F2017          $INCLUDE:'<filename>' : string expected

                A quoted string specifying a file name was expected after an
                $INCLUDE metacommand.

 F2018          $LINESIZE  ($PACK or $PAGESIZE) : integer constant out of
 Number         Compiler Compilation Error Message
F2018          $LINESIZE  ($PACK or $PAGESIZE) : integer constant out of
                range

                An integer constant less than 80 or greater than 132 was
                specified in a $LINESIZE metacommand, or a lower bound of
                less than 15 was specified in the $PAGESIZE metacommand. An
                integer constant other than 1, 2, or 4, was specified for the
                $PACK metacommand.

 F2019          $LINESIZE (or $PAGESIZE) : integer constant expected

                An integer constant was expected after a $LINESIZE or
                $PAGESIZE metacommand.

 F2020          $[NOT]LARGE already set

                The $LARGE or $NOTLARGE metacommand appeared more than once
                in a procedure. (The metacommand appearing more than once is
                indicated in the message.)

 Number         Compiler Compilation Error Message

 F2021          $[NOT]LARGE illegal in executable statements

                The $LARGE or $NOTLARGE metacommand appeared between
                subprograms or within the specification section of a
                subprogram. (The metacommand that appeared between
                subprograms or within the specification section is indicated
                in the message.)

 F2022          $MESSAGE:'<message>' : string expected

                A quoted string containing a message was expected after a
                $MESSAGE metacommand.

 F2023          divide by 0

                The second operand in a division operation (/) evaluated to
                0, giving undefined results.

 F2024          mod by 0
 Number         Compiler Compilation Error Message
F2024          mod by 0

                The second operand in a MOD function evaluated to 0, giving
                undefined results.

 F2025          no matching $IF

                A $ELSE, $ELSEIF, or $ENDIF was seen without a previous $IF.

 F2026          metacommands are nonstandard

                The metacommands used are nonstandard.

 F2027          $STORAGE:<number> : 2 or 4 expected

                A number other than 2 or 4 followed the $STORAGE:
                metacommand.

 F2028          $SUBTITLE:'<subtitle>' : string expected

 Number         Compiler Compilation Error Message

                A quoted string was expected following a $SUBTITLE:
                metacommand.

 F2029          $TITLE:'<title>' : string expected

                A quoted string was expected following a $TITLE: metacommand.

 F2030          unrecognized metacommand

                An unrecognized string followed the dollar sign ($) in the
                source file.

 F2031          closing quote missing

                A quoted string did not end with a single quote (').

 F2032          zero-length CHARACTER constant

                An illegal CHARACTER constant of length 0 was used in the
 Number         Compiler Compilation Error Message
               An illegal CHARACTER constant of length 0 was used in the
                program.

 F2033          Hollerith constant exceeds 1313 characters

                A Hollerith constant exceeded the maximum legal length.

 F2034          zero-length Hollerith constant

                An illegal Hollerith constant of length 0 was used in the
                program.

 F2035          Hollerith constant : text length disagrees with given
                length

                A Hollerith constant was smaller than the size given in the
                length field of the Hollerith constant.

 F2036          character : non-FORTRAN character

 Number         Compiler Compilation Error Message

                A special character in the source file was not recognized.

 F2037          illegal label field

                A nondigit value was used in a label field.

 F2038          zero-value label field

                A label with the value 0 was used in the program.

                Labels must have values between 1 and 99,999, inclusive.

 F2039          free-form label too long

                A label was more than five digits long.

 F2040          label on continuation line

                A label was declared on a continuation line.
 Number         Compiler Compilation Error Message
               A label was declared on a continuation line.

 F2041          first statement line must have ' ' or '0' in column 6

                A continuation character was used on the first statement line
                in the program.

 F2042          label on blank line

                A label was used on a line with no statements.

 F2043          alternate bases illegal

                Alternate integer bases are not allowed if the /4Ys compiler
                option is used (or the $STRICT metacommand is in effect).

 F2044          " nonstandard character string delimiter

                The quotation mark is a nonstandard Microsoft extension
                character-string delimiter.
 Number         Compiler Compilation Error Message
               character-string delimiter.

 F2050          $DEFINE : name already defined

                A name may only be given a $DEFINE value once. To re-define,
                $UNDEF the variable first.

 F2051          invalid expression in metacommand

                The compiler could not interpret an expression in an $IF,
                $ELSE, $ELSEIF, $ENDIF, or $DEFINE metacommand.

 F2052          unmatched parenthesis

                A closing parenthesis was found without an opening
                parenthesis preceding it.

 F2053          Character : unrecognized character

                The compiler encountered an operator it did not recognize or
 Number         Compiler Compilation Error Message
               The compiler encountered an operator it did not recognize or
                a name beginning with a nonalphabetic character.

 F2054          name : not defined

                A name was referenced which had not been $DEFINEd.

 F2055          logical operator expected

                An .AND. or .OR. operator was expected: the other operators
                cannot appear consecutively.

 F2056          character: unexpected characters at end of expression

                There were additional characters on the line that could not
                be part of the expression.

 F2059          operand expected

                The operator requires an operand.
 Number         Compiler Compilation Error Message
               The operator requires an operand.

 F2060          invalid integer constant

                The constant is out of range for integer type or not in
                correct form for a constant.

 F2062          relational operator expected

                The operands supplied must have a relational operator.

 F2064          name: defined with no value

                If a name was $DEFINEd with no value, it can only be used as
                an argument to DEFINE( ).

 F2070          filename : cannot open include file

                The specified include file could not be opened because it was
                not found in the source directory or in other directories
 Number         Compiler Compilation Error Message
               not found in the source directory or in other directories
                specified by the include search paths given on the command
                line.

 F2071          INCLUDE : quoted string missing

                The argument for the INCLUDE statement or $INCLUDE
                metacommand was not of CHARACTER type.

 F2072          INCLUDE : argument must be character constant

                The argument must be enclosed in apostrophes or quotes.

 F2101          DO : too many expressions

                A DO statement had more than three items following the equal
                sign (=).

 F2102          I/O implied-DO list : list empty

 Number         Compiler Compilation Error Message

                No items appeared in an I/O implied-DO list.

 F2103          I/O implied-DO list : too many expressions

                More than three expressions appeared after the equal sign (=)
                in an I/O implied-DO list.

 F2104          I/O implied-DO list : illegal assignment

                Only one assignment is legal in an I/O implied-DO list.

 F2105          I/O implied-DO list : too few expressions

                Fewer than two expressions followed the equal sign (=) in an
                I/O implied-DO list.

 F2106          I/O implied-DO list : assignment missing

                No assignment appeared in an I/O implied-DO list, or more
 Number         Compiler Compilation Error Message
               No assignment appeared in an I/O implied-DO list, or more
                than two expressions in the list were embedded in
                parentheses.

 F2107          assignments in COMPLEX constant illegal

                An illegal embedded assignment appeared in a constant of type
                COMPLEX.

 F2108          illegal assignment in parenthesized expression

                An illegal embedded assignment appeared in an expression
                enclosed in parentheses.

 F2111          numeric constant expected

                A symbolic or numeric constant did not appear as part of a
                complex constant.

 F2112          name : not symbolic constant
 Number         Compiler Compilation Error Message
F2112          name : not symbolic constant

                The specified name was not a symbolic constant.

 F2113          component of COMPLEX number not INTEGER or REAL

                A component of a COMPLEX number was not of type INTEGER or
                REAL.

 F2114          parser stack overflow, statement too complex

                The statement being parsed was too large for the parser.

 F2115          syntax error

                The source file contained a syntax error at the specified
                line.

 F2124          CODE GENERATION ERROR Contact Microsoft Technical Support

 Number         Compiler Compilation Error Message

                The compiler could not generate code for an expression.
                Usually this error occurs with a complicated expression.

                Try rearranging the expression. Please report this error to
                Microsoft Corporation using the Software Problem Report form
                at the back of this manual.

 F2125          name : allocation exceeds 64K

                The specified item exceeded the limit of 64K.

                Huge arrays are the only items that are allowed to be larger
                than 64K.

 F2126          name : automatic allocation exceeds 32K

                The subroutine or function name has an exceedingly large
                amount of compiler-generated temporary variables that take up
                more than 32,767 bytes.
 Number         Compiler Compilation Error Message
               more than 32,767 bytes.

                Try splitting the subroutine or function into smaller pieces.

 F2127          parameter allocation exceeds 32K

                The storage space required for the arguments to a function
                exceeded the limit of 32K.

 F2128          name : huge array cannot be aligned to segment boundary

                The specified array violated one of the restrictions imposed
                on huge arrays.

 F2200          subprogram : formal argument name : CHARACTER*(*) cannot
                pass by value

                Arguments that are passed by value must have a length that
                can be determined at run time. CHARACTER*(*) lengths are
                determined at run time.
 Number         Compiler Compilation Error Message
               determined at run time.

 F2201          subprogram : type redefined

                The type given in the specified ENTRY, FUNCTION, or
                SUBROUTINE statement was redefined. It was defined with a
                different type in an earlier subprogram.

 F2202          subprogram : defined with different number of arguments

                The specified ENTRY, FUNCTION, or SUBROUTINE statement was
                defined or used earlier in the program with a different
                number of arguments.

 F2203          subprogram : formal argument name : symbol-class mismatch

                The specified formal argument was defined previously with a
                different class.

                An EXTERNAL statement that passes a function to a variable,
 Number         Compiler Compilation Error Message
               An EXTERNAL statement that passes a function to a variable,
                or a similar symbol-class mismatch, can cause this error.

 F2206          ENTRY seen before FUNCTION or SUBROUTINE

                An ENTRY statement appeared before any FUNCTION or SUBROUTINE
                statements in the program.

                An ENTRY statement can only appear in functions and
                subroutines.

 F2207          ENTRY not in function or subroutine

                An ENTRY statement appeared in a subprogram that was not a
                function or subroutine. It may have appeared in the main
                program.

 F2208          name : formal argument used as ENTRY

                The specified name was used as a formal argument in an
 Number         Compiler Compilation Error Message
               The specified name was used as a formal argument in an
                earlier ENTRY statement or in the subprogram header in the
                current subprogram.

 F2209          name : illegal as formal argument

                The symbol class of the formal argument was illegal.

                A formal argument can only be a variable, array, subroutine,
                function, or entry point.

 F2210          name : formal argument redefined

                The specified formal argument appeared in the argument list
                more than once.

 F2211          alternate RETURN only legal within subroutine

                An alternate RETURN statement was specified outside of a
                subroutine.
 Number         Compiler Compilation Error Message
               subroutine.

 F2212          subprogram : subprogram used or declared before INTERFACE

                The specified subprogram was used or declared before the
                corresponding INTERFACE statement appeared in the program.

 F2213          subprogram : already defined

                The specified subprogram was already defined in the current
                module.

 F2214          subprogram : already used or declared with different symbol
                class

                The specified subprogram was used earlier in the program with
                a different class. For example, a subprogram that was used
                earlier in the program as a function and then declared as a
                subroutine would cause this error.

 Number         Compiler Compilation Error Message

 F2215          subprogram : ENTRY : CHARACTER lengths differ

                In a subprogram, if an entry name of type CHARACTER is used,
                then all of the entry names in that subprogram must be of
                type CHARACTER. If one entry name is of type CHARACTER*(*),
                then all must be of that type.

 F2216          subprogram : CHARACTER and non-CHARACTER types mixed in
                ENTRY statements

                CHARACTER and non-CHARACTER types were mixed in a subprogram.

 F2217          too many PROGRAM statements

                More than one PROGRAM statement appeared in the source file.

                Only one PROGRAM statement is allowed per program.

 F2218          name : used or declared before ENTRY statement
 Number         Compiler Compilation Error Message
F2218          name : used or declared before ENTRY statement

                The name in an ENTRY statement was declared or used
                previously in the same subprogram. This caused a symbol-class
                conflict that prevented the name from being used in an ENTRY
                statement.

 F2219          subprogram : formal argument name : VALUE/REFERENCE
                mismatch

                An INTERFACE statement or prior call specified a different
                way of passing this argument than that specified in the
                current declaration.

 F2220          subprogram : length redefined

                The length of a function when called was different than when
                it was defined.

 F2221          subprogram : formal argument name : NEAR/FAR/HUGE mismatch
 Number         Compiler Compilation Error Message
F2221          subprogram : formal argument name : NEAR/FAR/HUGE mismatch

                The NEAR, FAR, or HUGE attributes were defined differently in
                the INTERFACE statement than in the subprogram definition or
                its arguments.

 F2222          name : formal argument previously initialized

                The formal argument to an ENTRY statement appeared previously
                in a DATA statement within the same subprogram.

 F2223          subprogram : formal argument name : subprogram passed by
                VALUE

                The formal argument had the VALUE attribute. Subprograms
                cannot be passed to items with the VALUE attribute.

 F2224          name : language attribute mismatch

                Language attributes were declared differently in the
 Number         Compiler Compilation Error Message
               Language attributes were declared differently in the
                subprogram declaration than in the INTERFACE statement.

 F2225          name : NEAR/FAR attribute mismatch

                The NEAR or FAR attribute was used differently in the
                INTERFACE statement than in the subprogram declaration.

 F2226          name : VARYING attribute mismatch

                The VARYING attribute was not used in both the INTERFACE
                statement and the subprogram declaration.

 F2227          subprogram : formal argument name : previously passed by
                value, now by reference

                A formal argument previously passed by value was passed by
                reference.

                The VALUE attribute should be specified for the formal
 Number         Compiler Compilation Error Message
               The VALUE attribute should be specified for the formal
                argument.

 F2228          subprogram : formal argument name : previously passed by
                reference, now by value

                A formal argument previously passed by reference was passed
                by value.

                The REFERENCE attribute should be specified for the formal
                argument.

 F2229          subprogram : formal argument name : previously passed with
                NEAR, now with FAR or HUGE

                An address-length mismatch occurred. This is because an
                INTERFACE statement specifying the FAR or HUGE attribute for
                the formal argument was not given.

 F2230          subprogram : formal argument name : previously passed with
 Number         Compiler Compilation Error Message
F2230          subprogram : formal argument name : previously passed with
                FAR or HUGE, now with NEAR

                An INTERFACE statement specifying the NEAR attribute for the
                formal argument was not given.

 F2231          name : PROGRAM : name redefined

                The program name already exists as a global entity.

 F2232          subprogram : formal argument name : Hollerith passed to
                CHARACTER formal argument.

                Hollerith constants may only be passed to formal arguments of
                type integer, logical, or real.

 F2233          name : previously called near

                A function that was previously declared or referenced with
                near addressing was used with a far call.
 Number         Compiler Compilation Error Message
               near addressing was used with a far call.

 F2234          name : previously called far

                A function that was previously declared or referenced with
                far addressing was used with a near call.

 F2299          name : EQUIVALENCE: structure components illegal

                A structure variable may appear in an EQUIVALENCE statement,
                but not in an individual field of a structure.

 F2301          name : EQUIVALENCE (or COMMON): formal argument illegal

                An item other than a local variable or array, or a variable
                or an array in a common block, appeared in an EQUIVALENCE or
                COMMON statement.

 F2302          name : EQUIVALENCE : not array

 Number         Compiler Compilation Error Message

                In an EQUIVALENCE  statement, an item that was not an array
                had an argument or subscript list attached to it.

 F2303          name : EQUIVALENCE : array subscripts missing

                A construct such as  x( ) was used to declare the specified
                array.

                If no bounds are required, delete the parentheses.

 F2304          name : EQUIVALENCE : nonconstant offset illegal

                A nonconstant offset was used for an array in an EQUIVALENCE
                statement.

 F2305          name : nonconstant lower substring expression illegal

                The lower bound of a substring expression was not a constant
                in an EQUIVALENCE statement.
 Number         Compiler Compilation Error Message
               in an EQUIVALENCE statement.

 F2306          name : EQUIVALENCE : enclosing class too big

                Arithmetic overflow occurred while the offset of an array
                expression in an EQUIVALENCE statement was being calculated.

 F2307          name : stmt : allocatable array illegal

                Allocatable arrays may not appear in COMMON or EQUIVALENCE
                lists.

 F2308          name : COMMON (or NAMELIST): length specification illegal

                It is illegal to specify the length of a type in a COMMON
                statement.

                Use a separate type statement to declare the length.

 F2309          name : COMMON (or NAMELIST): attributes on items illegal
 Number         Compiler Compilation Error Message
F2309          name : COMMON (or NAMELIST): attributes on items illegal

                STRUCTURE may appear as well as COMMON. Fields in a STRUCTURE
                may not have attributes.

 F2310          name : COMMON (or EQUIVALENCE) : SUBROUTINE (or FUNCTION)
                name illegal

                A function or subroutine name was included in a COMMON or
                EQUIVALENCE statement.

                Only local variables and arrays are legal.

 F2311          name : COMMON (or EQUIVALENCE or STRUCTURE) :
                preinitial-ization illegal

                Items in COMMON or EQUIVALENCE statements cannot be
                preinitialized in type-declaration statements.

                Use the standard notation for the DATA statement.
 Number         Compiler Compilation Error Message
               Use the standard notation for the DATA statement.

 F2312          name : COMMON (or EQUIVALENCE or NAMELIST) : formal
                argument illegal

                The specified formal argument was used in a COMMON or
                EQUIVALENCE statement.

 F2313          name : COMMON (or EQUIVALENCE or NAMELIST) : not an array
                or variable

                An item other than an array or variable was used in an
                EQUIVALENCE or COMMON statement.

 F2314          array : COMMON : too big

                Arithmetic overflow occurred while the size of a common block
                was being calculated. STRUCTURE can also appear. Common
                blocks can be very large; structures must be less than 64K
                bytes in size.
 Number         Compiler Compilation Error Message
               bytes in size.

 F2315          array : COMMON : array size nonconstant or zero

                A nonconstant or 0 value was used to dimension the array.

 F2316          name1, name2 : EQUIVALENCE : both in blank common block

                Two items specified in an EQUIVALENCE statement at different
                offsets were both in a blank common block. In the EQUIVALENCE
                statement, these items were specified to be at the same
                location in memory.

 F2317          name1, name2 : EQUIVALENCE : both in common block
                commonblock

                Two items specified in an EQUIVALENCE statement at different
                offsets were both in a named common block. These items were
                specified in the EQUIVALENCE statement to be at the same
                location in memory.
 Number         Compiler Compilation Error Message
               location in memory.

 F2318          name1, name2 : EQUIVALENCE : in different common blocks

                Two items in different common blocks were specified in an
                EQUIVALENCE statement.

 F2319          name : EQUIVALENCE : extends blank common block forward

                In an EQUIVALENCE statement, it is illegal to increase the
                size of a blank common block by adding memory elements before
                the beginning common block declared in the COMMON statement.

 F2320          name : EQUIVALENCE : extends common block commonblock
                forward

                In an EQUIVALENCE statement, it is illegal to increase the
                size of a named common block by adding memory elements before
                the beginning common block declared in the COMMON statement.

 Number         Compiler Compilation Error Message

 F2321          name1, name2 : EQUIVALENCE : conflicting offsets

                The processing of an EQUIVALENCE statement detected two items
                that should have had the same offsets but did not.
                Inconsistent use of EQUIVALENCE statements caused this
                problem.

 F2322          name : EQUIVALENCE : two different common blocks

                An EQUIVALENCE statement placed an item in two different
                common blocks.

 F2323          commonblock : COMMON : size changed

                The size of the specified common block differed from the size
                allocated in a prior subprogram.

 F2324          commonblock : COMMON : too big to be NEAR

 Number         Compiler Compilation Error Message

                The specified common block, declared with the NEAR attribute,
                is larger than a segment.

 F2325          name : COMMON : function or subroutine name

                The specified name was used as both a common-block name and a
                function or subroutine name.

 F2326          name : already in COMMON

                The specified name appeared in a COMMON statement elsewhere
                in this subprogram.

 F2327          name : EQUIVALENCE : needs at least two items

                An EQUIVALENCE statement had fewer than two items in a class.

 F2328          name : already typed

 Number         Compiler Compilation Error Message

                The specified item appeared in an earlier type statement in
                the same subprogram or derived-type definition.

 F2329          blank common cannot be HUGE

                In medium model, blank common items must be smaller than a
                single segment. Named common items do not have this
                restriction.

 F2330          name : already dimensioned

                Array bounds appeared for the specified item in an earlier
                specification statement in the same subprogram.

 F2331          name : types illegal on BLOCK
                DATA/COMMON/PROGRAM/SUBROUTINE

                The specified item was not one of the symbol classes that can
                be typed.
 Number         Compiler Compilation Error Message
               be typed.

 F2332          name : cannot initialize in type statements

                An attempt was made to initialize the specified item in a
                type statement while the /4Ys compiler option was used or the
                $STRICT metacommand was in effect.

 F2333          name : DIMENSION : not array

                The specified item in a DIMENSION statement (for example, an
                item already declared in an EXTERNAL or PARAMETER statement)
                was not an array.

 F2334          name : ALLOCATABLE: bounds must be omitted

                An allocatable array declaration's subscripts must be ':'
                only, with no upper or lower parameters.

 F2336          array : array bounds missing
 Number         Compiler Compilation Error Message
F2336          array : array bounds missing

                Both bound expressions were missing from the declaration of
                the specified array.

                At least an upper bound must be present.

 F2337          array : * : not last array bound

                An assumed-size array was declared with an asterisk (*) that
                did not occur in the last bound.

 F2338          array : bound size too small

                The bound size of the specified array was not a positive
                whole number.

                Bounds of adjustable-size arrays can be checked at run time.
                This compile-time error occurs only when the upper and
                (possibly implicit) lower bounds create a negative or zero
 Number         Compiler Compilation Error Message
               (possibly implicit) lower bounds create a negative or zero
                element count for an array bound.

 F2339          array : adjustable-size array not in subprogram

                The specified adjustable-size array was declared in a
                subprogram declared with a PROGRAM or BLOCK DATA statement.

                An adjustable-size array is legal only in an ENTRY, FUNCTION,
                or SUBROUTINE statement in a subprogram.

 F2340          IMPLICIT already seen

                An IMPLICIT statement has already been seen, so an IMPLICIT
                NONE statement is illegal.

 F2341          letters : IMPLICIT : only single letter allowed

                The upper or lower value of the range in an IMPLICIT
                statement was not a single character.
 Number         Compiler Compilation Error Message
               statement was not a single character.

 F2342          IMPLICIT NONE already seen

                There may only be one IMPLICIT NONE statement per subprogram.

 F2343          letter1, letter2 : IMPLICIT : lower limit exceeds upper
                limit

                The upper letter in the range in an IMPLICIT statement had a
                smaller value than the letter in the lower range.

 F2344          letter : already IMPLICIT

                The specified character already appeared in an IMPLICIT
                statement earlier in the same subprogram.

 F2345          name : illegal use of SAVE (or AUTOMATIC, EXTERNAL,
                INTRINSIC, PARAMETER)

 Number         Compiler Compilation Error Message

                The specified name appeared earlier in a conflicting type
                statement.

 F2346          name : INTRINSIC : unknown name

                The specified name is not the name of a supported intrinsic
                function.

 F2348          name : already declared SAVE (or AUTOMATIC, EXTERNAL,
                INTRINSIC, PARAMETER)

                The specified name was declared more than once with the same
                type of statement.

 F2349          name : PARAMETER : nonconstant expression

                The specified item was declared with a nonconstant value in a
                PARAMETER statement.

 Number         Compiler Compilation Error Message

 F2350          RECORD : structure type illegal in IMPLICIT statement

                Structure variables cannot be implicitly typed.

 F2351          name : repeated in formal-argument list

                The specified item was repeated in the formal-argument list
                to a statement function.

 F2352          name : formal argument not local variable

                Only local variables can be used as formal arguments to
                statement functions.

 F2354          name : statement function already declared

                The specified statement function was already declared in the
                current subprogram.

 Number         Compiler Compilation Error Message

 F2355          name : formal argument not a variable

                An argument list or substring operator for the specified item
                appeared in the formal-argument list to a statement function.

 F2356          name : statement function : too few actual arguments

                More formal arguments than actual arguments were declared for
                a statement function.

 F2357          name : statement function : too many actual arguments

                More actual arguments than formal arguments were declared for
                a statement function.

 F2359          type : illegal length

                An illegal length specifier for the given type was used in a
                declaration. For example,  REAL*13  would cause this error.
 Number         Compiler Compilation Error Message
               declaration. For example,  REAL*13  would cause this error.

 F2360          END STRUCTURE: no matching STRUCTURE statement

                An END STRUCTURE statement was seen without a preceding
                STRUCTURE statement.

 F2361          stmt : not a name

                The argument on the STRUCTURE or END STRUCTURE statement was
                not a symbol name.

 F2362          integer constant expression expected

                An integer value or integer constant expression was expected
                for an optional type-length specification.

 F2363          length value : illegal type length

                A zero or negative length specifier was used in a type
 Number         Compiler Compilation Error Message
               A zero or negative length specifier was used in a type
                statement, or the length specifier was larger than the
                largest allowed for all types.

 F2364          only C attribute legal on INTEGER type

                An attribute other than the C attribute appeared with an
                INTEGER type statement.

 F2365          attributes illegal on non-INTEGER types

                Attributes in type statements are illegal, other than the C
                attribute on the INTEGER statement. Attributes were not put
                on the variables themselves.

 F2366          DOUBLE PRECISION : length specifier illegal DOUBLE COMPLEX
                : length specifier illegal

                A DOUBLE PRECISION or DOUBLE COMPLEX statement included a
                length specifier. DOUBLE PRECISION is the same as REAL*8.
 Number         Compiler Compilation Error Message
               length specifier. DOUBLE PRECISION is the same as REAL*8.
                DOUBLE COMPLEX is the same as COMPLEX*8.

 F2367          value value : INTEGER : range error

                The specified constant was out of range for type conversion,
                or the type of an integer item was in conflict with the
                integer size specified in the /4I compiler option (or
                $STORAGE metacommand). For example, the following code
                induces this error:

                $STORAGE:2
                      INTEGER*4 i
                      i = 300000+30000
                      i =     10* 4000
                      = -30000-30000
                      END

                To correct this error, use the appropriate INT2 or INT4
                intrinsic function to make sure the appropriate (2- or
 Number         Compiler Compilation Error Message
               intrinsic function to make sure the appropriate (2- or
                4-byte) arithmetic is performed on the variable.

 F2368          name : truncated to 6 characters

                When the /4Ys compiler option is used or the $STRICT
                metacommand is in effect, only six characters can appear in
                identifier names.

 F2369          name : $ illegal in C name

                A character in the specified name was illegal for a C
                variable. C variables allow only underscores (_) and
                alphanumeric characters in names.

 F2370          length specification illegal

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), length specifications can only be
                used with CHARACTER type statements.
 Number         Compiler Compilation Error Message
               used with CHARACTER type statements.

 F2371          name1, name2 : EQUIVALENCE : character and noncharacter
                items mixed

                Character and noncharacter items were mixed in an EQUIVALENCE
                statement.

 F2372          name : more than 7 array bounds

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), an array cannot have more than
                seven bounds.

 F2373          name : REFERENCE or VALUE only legal on formal arguments

                A REFERENCE or VALUE attribute was used with an item that was
                not declared in the formal-argument list for the routine.

                If the item is used in a type statement, then the attributed
 Number         Compiler Compilation Error Message
               If the item is used in a type statement, then the attributed
                item must also appear in the formal-argument list of a
                subprogram. If the item appears in an ENTRY statement,
                include the attribute there instead.

 F2374          name : attributes illegal on array bounds

                No attributes are allowed on items that are used when
                dimensioning arrays.

 F2375          name : assumed-size array : cannot pass by value

                An assumed-size array was passed as an actual argument to a
                routine that had its formal argument declared with the VALUE
                attribute.

 F2376          name : adjustable-size array : cannot pass by value

                An adjustable-size array was passed as an actual argument to
                a routine that had its formal argument declared with the
 Number         Compiler Compilation Error Message
               a routine that had its formal argument declared with the
                VALUE attribute.

 F2377          name : NEAR common block has HUGE item

                A common block declared with the NEAR attribute included
                item(s) that required the common block to be huge.

 F2378          name : NEAR array bigger than segment

                An array declared with the NEAR attribute was larger than a
                segment.

 F2379          name : item in common block crosses segment

                An item or an array element in a common block crossed a
                segment boundary. Items or arrays must be evenly aligned to
                signal boundaries when a common block crosses a segment.

 F2380          name : VARYING illegal on symbol class
 Number         Compiler Compilation Error Message
F2380          name : VARYING illegal on symbol class

                The VARYING attribute was used on something other than a
                function or subroutine.

 F2381          commonblock : NEAR/FAR/HUGE attribute mismatches default

                An attribute declared for the given common block was
                different from the attribute implicitly applied to the common
                block in an earlier subprogram.

                In medium-model programs, the NEAR attribute is used
                implicitly, unless the size of the common block requires the
                common block to be huge. In large-model programs, the FAR or
                HUGE attribute is used implicitly.

 F2382          commonblock : attribute attribute mismatch with earlier
                NEAR/FAR/HUGE

                An attribute given in an earlier common-block declaration
 Number         Compiler Compilation Error Message
               An attribute given in an earlier common-block declaration
                (possibly in a different subprogram) was not the same as the
                current attribute.

 F2383          name : COMMON : character and noncharacter items mixed

                Character and noncharacter items cannot be mixed in a common
                block when the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect).

 F2384          name : attribute  variable cannot be AUTOMATIC

                The attribute can be one of the following four: HUGE, FAR,
                EXTERN, ALLOCATABLE.

 F2385          name : STRUCTURE has no elements

                A STRUCTURE was declared with no component fields.

 F2386          NAMELIST:  group name required
 Number         Compiler Compilation Error Message
F2386          NAMELIST:  group name required

                A name must appear between the slashes.

 F2387          name : STRUCTURE : intrinsic type name

                Derived type names may not be the same as the names of
                intrinsic types.

 F2388          name : NAMELIST : array bounds illegal

                Only variable and array names are allowed in a NAMELIST
                statement (the array must be declared someplace else).

 F2389          name : not a STRUCTURE name

                The argument in the STRUCTURE( ) statement has not previously
                been declared in a STRUCTURE statement.

 F2390          name : ALLOCATABLE : common block variable illegal
 Number         Compiler Compilation Error Message
F2390          name : ALLOCATABLE : common block variable illegal

                Common block variables may not be allocatable attribute.

 F2391          name : ALLOCATABLE : equivalenced variable illegal

                Equivalenced variables may not be given the allocatable
                attribute.

 F2392          MAP : no enclosing UNION statement

                A map can only appear within a UNION statement.

 F2393          name: element name conflicts with operator

 F2394          name: NAMELIST: structure illegal

                Structure variables cannot appear in NAMELIST groups.

 F2395          UNION : not in a STRUCTURE
 Number         Compiler Compilation Error Message
F2395          UNION : not in a STRUCTURE

                A union can only appear within a STRUCTURE statement.

 F2396          name : ALLOCATABLE : must be in array

                A variable was given the allocatable attribute but not
                dimensioned.

 F2401          name : DATA : illegal address expression

                An illegal expression was used for the offset in a DATA
                statement.

                Only constant offsets are legal for items in DATA statements.

 F2402          name : cannot initialize formal argument

                The item being initialized was a formal argument to a
                subprogram.
 Number         Compiler Compilation Error Message
               subprogram.

 F2403          name : cannot initialize item in blank common block

                An attempt was made to use a DATA statement to initialize the
                specified item in a blank common block.

 F2404          name : can only initialize common block in BLOCK DATA
                subprogram

                An attempt was made to initialize the specified item named in
                a common block in a DATA statement.

                Items in named common blocks can be initialized only in BLOCK
                DATA subprograms.

 F2405          name : DATA : not an array or variable

                Only arrays and variables can be initialized in DATA
                statements.
 Number         Compiler Compilation Error Message
               statements.

 F2406          name : repeat count not positive integer

                The repeat count for initialization of the specified item was
                not a positive integer value.

 F2407          name : DATA : nonconstant item in initializer list

                A nonconstant value was used to initialize the specified item
                in a DATA statement.

 F2408          name : DATA : too few constants to initialize item

                The DATA statement did not include enough constants to
                initialize the specified item.

 F2409          name : nonstatic address illegal in initialization

                During processing of an implied-DO list in a DATA statement,
 Number         Compiler Compilation Error Message
               During processing of an implied-DO list in a DATA statement,
                the specified item did not have a static address.

                Make sure the item has constant subscript specifiers.

 F2410          name : bound or increment not constant

                The specified item in implied-DO initialization in a DATA
                statement did not have constant bounds.

 F2411          name : DATA : zero increment

                In a DATA statement, the increment value in the implied-DO
                list for the specified item must be set so the loop exits.

 F2412          name : DATA : active implied-DO variable

                The specified implied-DO variable was used in nested DATA
                implied-DO initialization loops in a DATA statement.

 Number         Compiler Compilation Error Message

 F2413          name : DATA : implied-DO variable not INTEGER

                In DATA statements, only implied-DO variables of types
                INTEGER and  INTEGER*n  are legal.

 F2414          name : DATA : not array-element name

                Only array elements can be initialized in implied-DO
                initializations in DATA statements.

 F2415          DATA : too few constants to initialize names

                The constant list was exhausted before the initialization of
                the name list was complete.

 F2416          name: bound or increment not INTEGER

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), only items of type INTEGER are
 Number         Compiler Compilation Error Message
               metacommand is in effect), only items of type INTEGER are
                allowed for DATA implied-DO loop bounds and increments.
                Otherwise, any arithmetic type is allowed and is truncated to
                type INTEGER by an implicit use of the INT intrinsic
                function.

 F2417          DATA : iteration count not positive

                In the implied-DO list (...,dovar=start,stop,inc), if the
                increment inc is positive, then start must be greater than
                stop; if the increment inc is negative, then stop must be
                greater than start. If not, then the loop would execute zero
                times, which is not allowed.

 F2418          name  : variable was declared automatic

                A variable which is going to go on the stack (declared
                AUTOMATIC) cannot be given the HUGE attribute.

 F2419          name : cannot initialize allocatable array
 Number         Compiler Compilation Error Message
F2419          name : cannot initialize allocatable array

                An allocatable array cannot appear in a DATA statement.

 F2490          name : STAT= must be last parameter

                The STAT= parameter must be last in an ALLOCATE statement.

 F2491          name : stmt : not allocatable array

                The name in the allocate or deallocate statement is not an
                allocatable array. (The second parameter is ALLOCATE or
                DEALLOCATE.)

 F2492          stmt : STAT= variable must be scalar integer

                First parameter is ALLOCATE or DEALLOCATE.

 F2493          name : * :  illegal bound

 Number         Compiler Compilation Error Message

                There are no assumed-size dimensions in an allocatable array.

 F2500          array : adjustable-size array : used before definition

                An adjustable-size array was used before it was seen in an
                ENTRY statement.

 F2502          type : cannot convert to type

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), constants cannot be converted
                between CHARACTER and nonCHARACTER types.

 F2503          intrinsic : incorrect use of intrinsic function

                Invalid arguments were given for the specified intrinsic
                function.

 F2504          intrinsic : multiple arguments
 Number         Compiler Compilation Error Message
F2504          intrinsic : multiple arguments

                The specified intrinsic function had more than one argument;
                only one is legal.

 F2505          intrinsic : cannot convert FAR address to NEAR

                An item in the specified intrinsic function can only be
                referenced with addressing consistent with the FAR or HUGE
                attribute.

 F2506          cannot convert type to type

                An invalid type conversion to CHARACTER or LOGICAL type was
                attempted.

 F2508          array: array bound used array reference

                An expression having an array was used when declaring an
                adjustable-size array.
 Number         Compiler Compilation Error Message
               adjustable-size array.

                Only simple variables in common blocks on the current
                subprogram's formal argument list are allowed as variables in
                the bound expression.

 F2509          element : not an element of name

                The name given as a component of a derived type is undefined.

 F2510          name : symbolic constant : subscript illegal

                The specified symbolic constant had an array index or
                argument list.

 F2511          name : symbolic constant : substring illegal

                The specified symbolic constant had a substring operator.

 F2512          name : variable : argument list illegal
 Number         Compiler Compilation Error Message
F2512          name : variable : argument list illegal

                The specified simple variable included an argument list.

 F2513          name : not a variable

                The specified item was not a variable.

                A variable is expected in this context.

 F2514          concatenation with CHARACTER*(*)

                A CHARACTER*(*) item was used in a concatenation operation.

                Only items with specified lengths are legal in
                concatenations.

 F2515          left side of assignment illegal

                The left side of an assignment statement was illegal.
 Number         Compiler Compilation Error Message
               The left side of an assignment statement was illegal.

                Only variables, array elements, or function-return variables
                may appear on the left side of assignment statements.

 F2516          name : assignment using active DO variable illegal

                An active DO variable was used in an assignment statement.

 F2517          illegal implied-DO list in expression

                In this context, implied-DO statements are illegal in
                expressions.

 F2518          name : not a structure

                A variable which is not of derived type was specified with a
                component field.

 F2519          name : operation error with COMPLEX operands
 Number         Compiler Compilation Error Message
F2519          name : operation error with COMPLEX operands

                A constant-folding error occurred. The number created would
                probably overflow the allowed storage. Using smaller numbers
                will correct.

 F2520          name : operation error with REAL operands

                A constant-folding error occurred. The number created would
                probably overflow the allowed storage. Using smaller numbers
                will correct.

 F2521          negative exponent with zero base

                A negative exponent was used with a zero-value base.

 F2522          division by zero

                Division by zero occurred during constant folding.

 Number         Compiler Compilation Error Message

 F2523          only comparisons by .EQ. and .NE. allowed for complex items

                Only .EQ. and .NE. are legal as comparison operators for
                complex items.

 F2524          non-numeric operand

                A nonarithmetic operand was specified with an arithmetic
                operator.

 F2525          exponentiation of COMPLEX and DOUBLE PRECISION together
                illegal

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), exponentiation is illegal with
                bases of type COMPLEX having DOUBLE PRECISION exponents, or
                with bases of type DOUBLE PRECISION having COMPLEX exponents.

 F2526          concatenation of expressions illegal
 Number         Compiler Compilation Error Message
F2526          concatenation of expressions illegal

                An illegal concatenation operation occurred.

                If a noncharacter item is used in a concatenation, it must be
                a constant or it must be addressable.

 F2527          noncharacter operand

                When the /4Ys compiler option is used (or when the $STRICT
                metacommand is in effect), concatenation operators can be
                used only with character operands.

 F2528          nonlogical operand

                Logical operators (.AND., .OR., .NOT., .EQV., and .NEQV.)
                must be used with logical operands.

 F2529          operands of relation not numeric or character

 Number         Compiler Compilation Error Message

                Relational operators (.LT., .LE., .GT., .GE., .EQ., and .NE.)
                must be used with arithmetic or character operands.

 F2530          name: symbol class illegal here

                The class of the given symbol was illegal in this context.

 F2531          name: bound not integer

                The /4Ys compiler option was used in compiling (or the
                $STRICT metacommand was in effect), and a substring had a
                noninteger substring-bound expression.

                If the /4Ns compiler option is used in compiling (or the
                $NOTSTRICT metacommand is in effect), any arithmetic
                expression is legal and is truncated to integers through an
                implicit use of the INT intrinsic function.

 F2532          name : substring on noncharacter item
 Number         Compiler Compilation Error Message
F2532          name : substring on noncharacter item

                An attempt was made to take the substring from an item that
                was not a character variable or array item.

 F2533          name : lower substring bound exceeds upper bound

                The /4Yb compiler option was used (or the $DEBUG metacommand
                was in effect), and the value of the upper substring bound
                was less than the value of the lower substring bound.

 F2534          name : upper substring bound exceeds string length

                The /4Yb compiler option was used (or the $DEBUG metacommand
                was in effect), and the upper substring bound was greater
                than the length of the item from which the substring was
                taken.

                This error occurs only if the length of the item was not
                specified (that is, if it was declared as a CHARACTER*n
 Number         Compiler Compilation Error Message
               specified (that is, if it was declared as a CHARACTER*n
                item).

 F2535          name : lower substring bound not positive

                The /4Yb compiler option was used (or the $DEBUG metacommand
                was in effect), and the lower substring bound was less than
                or equal to 0.

                The minimum value for items of type CHARACTER is 1.

 F2536          array : subscript number out of range

                The /4Yb compiler option was used (or the $DEBUG metacommand
                was in effect), and a local array or an array in a common
                block had a bound out of range.

 F2537          array : array subscripts missing

                The specified array, which did not have array subscripts, was
 Number         Compiler Compilation Error Message
               The specified array, which did not have array subscripts, was
                used in an expression.

 F2538          array : subscript number : not integer

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), a subscripting expression used in
                the specified array must be of type INTEGER. Otherwise, it
                must be an arithmetic type that is truncated to  INTEGER by
                an implicit use of the INT intrinsic function.

 F2539          array : too few array subscripts

                Not enough subscripts were given when the specified array was
                used in an expression.

 F2540          array : too many array subscripts

                Too many subscripts were given when the array was used in an
                expression.
 Number         Compiler Compilation Error Message
               expression.

 F2541          cannot convert between CHARACTER and non CHARACTER
                constants

                If the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), constants cannot be converted
                between CHARACTER and nonCHARACTER types.

 F2542          one numeric, one character operand

                If the /4Ys compiler option is used in compiling (or the
                $STRICT metacommand is in effect), both operands used with
                relational operators must be character or both must be
                arithmetic. Operands of different types cannot be mixed.

 F2543          operand type must be logical or integer

                See F2528. When the $STRICT metacommand is off, integer
                operands are allowed also.
 Number         Compiler Compilation Error Message
               operands are allowed also.

 F2544          operand types do not match

                One argument is a derived type and the other argument is not.

 F2545          invalid operator for structure operands

                Only the .EQ. and .NE. operators are defined for structures.

 F2559          array : array bound used illegal variable

                Only variables in common blocks or in the formal argument
                list to the current subprogram are legal when declaring
                adjustable-size arrays.

 F2560          array : array bound used intrinsic call

                Only variables in common blocks or in the formal argument
                list to the current subprogram are legal when declaring
 Number         Compiler Compilation Error Message
               list to the current subprogram are legal when declaring
                adjustable-size arrays.

 F2561          array : array bound used function call

                Only variables in common blocks or in the formal argument
                list to the current subprogram are legal when declaring
                adjustable-size arrays.

 F2562          cannot pass CHARACTER*(*) by value

                The program tried to pass by value an item of type
                CHARACTER*(*). This is illegal because the length of such
                items is not known at compile time.

                Actual arguments with a length of n can be passed to
                CHARACTER*n items, and these items can be passed by value, if
                required.

 F2563          incompatible types for formal and actual arguments
 Number         Compiler Compilation Error Message
F2563          incompatible types for formal and actual arguments

                The types of the formal and actual arguments did not match.

                Formal and actual arguments must have the same types (except
                for arguments of type CHARACTER, where the lengths can
                differ).

 F2564          incompatible types in assignment

                The expressions on the left and right sides of an assignment
                statement were of different types. For example, a logical
                expression cannot be assigned to an integer variable.

 F2565          operation : COMPLEX : type conversion error

                An attempt was made to convert values of one type to types
                that hold a smaller range of values.

 F2566          operation : REAL : type conversion error
 Number         Compiler Compilation Error Message
F2566          operation : REAL : type conversion error

                An attempt was made to convert values of one type to types
                that hold a smaller range of values.

 F2567          LEN : illegal expression

                Only constants, symbols, concatenations, intrinsic type
                casts, and strings are allowed in the LEN intrinsic function.

 F2568          name : illegal bound type

                Only integer items are allowed as array bounds when the /4Ys
                compiler option is used (or the $STRICT metacommand is in
                effect). Otherwise, arithmetic types are allowed and are
                converted through an implicit use of the INT intrinsic
                function.

 F2569          name : Hollerith constant passed by value

 Number         Compiler Compilation Error Message

                A Hollerith constant must be passed by reference to a
                logical, real, or integer formal argument.

 F2570          consecutive arithmetic operators illegal

                Unary plus and minus cannot follow other arithmetic
                operators. For example,  I = I**-1  is illegal; I = I**(-1)
                must be used instead.

 F2571          consecutive relational operators illegal

                The .NOT. operator cannot follow another .NOT. operator.

 F2572          illegal use of Hollerith constant

                Hollerith constants are only allowed in assignments, DATA
                statements, and subprogram references.

 F2600          name : directly recursive
 Number         Compiler Compilation Error Message
F2600          name : directly recursive

                A subprogram is trying to call itself.

 F2601          intrinsic : intrinsic function illegal as actual argument

                This intrinsic function is illegal as an actual argument.
                (Some specific versions of the generic intrinsic functions
                can be passed as actual arguments.)

 F2602          name : formal argument name : character expressions cannot
                be passed by VALUE

                A character expression cannot be passed by value.

 F2604          subprogram : function : argument list missing

                The specified function was missing an argument list.

                At least an empty argument list (( )) must be present in
 Number         Compiler Compilation Error Message
               At least an empty argument list (( )) must be present in
                expressions.

 F2605          subprogram : function : substring operator illegal

                A substring operator was used illegally with the specified
                routine name.

                Substring operators can only be used with arrays and
                variables.

 F2606          subprogram : formal argument name : type mismatch

                The type of a formal argument was different from the type of
                the actual argument used in the subprogram call.

 F2607          subprogram : formal argument name : length mismatch

                The length of a formal argument was different from the length
                of the actual argument used in the subprogram call.
 Number         Compiler Compilation Error Message
               of the actual argument used in the subprogram call.

 F2608          subprogram : formal argument name : Hollerith illegal with
                CHARACTER

                Hollerith constants can only be used with items of type
                INTEGER, LOGICAL, and REAL in DATA statements and subprogram
                references.

 F2609          subprogram : formal argument * : actual not
                alternate-return label

                Because the specified formal argument was an alternate-return
                label, the current argument must also be an alternate-return
                label.

 F2610          subprogram : formal argument name : not alternate-return
                label

                Because the specified formal argument was not an
 Number         Compiler Compilation Error Message
               Because the specified formal argument was not an
                alternate-return label, the current argument must not be an
                alternate-return label.

 F2611          subprogram : formal argument name : actual not subprogram

                The formal argument used in a subprogram declaration was a
                subprogram, but the actual argument was not.

 F2612          subprogram : NEAR formal argument name : actual has FAR or
                HUGE address

                It is illegal to pass an item that must be addressed with far
                or huge addressing to a formal argument that must be
                addressed with near addressing.

 F2615          name : not function or subroutine

                The specified item was not a function or subroutine.

 Number         Compiler Compilation Error Message

                Check the item's use or declaration earlier in the program.

 F2616          subprogram : illegal use of function or subroutine

                The program tried to use a function as a subroutine or use a
                subroutine as a function.

 F2617          subprogram : adjustable-size array array : cannot pass by
                value

                An attempt was made to pass an adjustable-size array by
                value.

 F2618          subprogram : cannot use CHARACTER*(*) function

                CHARACTER*(*) functions cannot be directly referenced. They
                can only be passed as actual arguments.

 F2619          name : value argument bigger than segment
 Number         Compiler Compilation Error Message
F2619          name : value argument bigger than segment

                An argument with a VALUE attribute was too big to be passed
                onto the stack.

 F2620          subprogram : formal argument name : subprogram mismatch

                The type of the formal argument to the subprogram was not the
                same as the actual argument.

                Both the formal and the actual argument must be subroutines
                or functions.

 F2621          name : formal argument name : not subprogram

                The actual argument to the subprogram was a subprogram, but
                the formal argument was not a subprogram.

 F2622          assumed-size array array : cannot pass by value

 Number         Compiler Compilation Error Message

                An assumed-size array can only be passed by reference.

 F2623          name : nonconstant CHARACTER length : cannot pass by value

                If a substring is used when passing a CHARACTER*(n) or
                CHARACTER*(*) argument to a formal argument declared with the
                VALUE attribute, then the lower and upper substring values
                must be constant. Otherwise, the length cannot be determined.

 F2624          subprogram : too few actual arguments

                The number of actual and formal arguments for the given
                subprogram did not match.

                This practice is legal only when the C and VARYING attributes
                are specified for the subprogram.

 F2625          subprogram : too many actual arguments

 Number         Compiler Compilation Error Message

                The number of actual and formal arguments for the given
                subprogram did not match.

                This practice is legal only when the C and VARYING attributes
                are specified for the subprogram.

 F2626          formal argument <name> : cannot be SAVE or AUTOMATIC

                A formal argument cannot appear in either a SAVE or AUTOMATIC
                statement.

 F2650          name : array expression: cannot be adjustable-size array

                An adjustable-size array may not appear in an array
                expression.

 F2651          name : array expression: argument does not conform

                The array name is not the same size as other arrays in the
 Number         Compiler Compilation Error Message
               The array name is not the same size as other arrays in the
                expression.

 F2652          name : array expression: cannot be assumed-size array

                An assumed-size array may not appear in an array expression.

 F2653          <name> : array expression : cannot be allocatable array

                An allocatable array name cannot appear in an array
                expression, since its bounds are not known at compile time.

 F2702          iooption : array subscript missing

                In this context, the array in the specified I/O option cannot
                appear without subscripts.

 F2703          iooption : not type

                The specified I/O option required an item of a different
 Number         Compiler Compilation Error Message
               The specified I/O option required an item of a different
                type. For example, the REC=rec option requires an integer
                expression.

 F2704          iooption : not a variable or array element

                The specified I/O option required a variable or an array
                element, as opposed to an arbitrary expression.

 F2705          label number : not between 1 and 99999

                Statement labels are restricted to the range 1-99,999; they
                must be one to five digits, not all of which are 0.

 F2706          UNIT= * illegal for this statement

                The asterisk (*) unit specifier (console unit) cannot be
                specified for this I/O statement.

                Use the asterisk (*) unit specifier only with READ, WRITE, or
 Number         Compiler Compilation Error Message
               Use the asterisk (*) unit specifier only with READ, WRITE, or
                INQUIRE statements. For INQUIRE statements, the asterisk (*)
                unit specifier is allowed only when the /4Ns compiler option
                is used (or the $NOTSTRICT metacommand is in effect).

 F2707          illegal unit specifier

                The unit specifier in a UNIT= option was not an integer
                expression, asterisk (*), character variable, array element,
                array, or substring.

                A noncharacter array is a legal unit specifier if the /4Ns
                compiler option is used in compiling (or the $NOTSTRICT
                metacommand is in effect).

 F2708          illegal format specifier

                The format specifier in a FMT= option was not a statement
                label, integer variable, character expression, character
                array, noncharacter array, or asterisk (*).
 Number         Compiler Compilation Error Message
               array, noncharacter array, or asterisk (*).

 F2709          HUGE format illegal

                An array declared with a HUGE attribute that appeared in a
                $LARGE metacommand, or that spanned more than one segment,
                was used as a format specifier.

 F2710          UNIT=* : unformatted I/O illegal

                The keyboard or terminal is opened for sequential formatted
                I/O only.

 F2711          FAR format illegal in medium model

                Data allocated with the FAR attribute were used as a format
                specifier in a medium-model program.

 F2712          iooption : appears twice

 Number         Compiler Compilation Error Message

                The specified I/O option was used more than once in the same
                I/O statement.

 F2713          END= (or ERR=): illegal statement label

                An integer number in the range 1 to 99,999 must be specified
                for statement labels following the ERR= and END= input/output
                options. (Compile-time error.)

 F2714          I/O option number : <keyword=> missing

                The I/O option at position number in the option list appeared
                without a keyword.

                An I/O option without a keyword must not appear past the
                second position in the option list. Also, only UNIT= and FMT=
                options can appear without a keyword. If the UNIT= option
                appears without a keyword, it must be the first option in the
                option list. If the FMT= option appears without a keyword, it
 Number         Compiler Compilation Error Message
               option list. If the FMT= option appears without a keyword, it
                must follow a UNIT= option without a keyword.

                For example,

                      OPEN (2, 'F.DOT')

                would produce the message

                I/O option 2:  <keyword=> missing

                because the FILE= option is missing in the second option.

 F2715          iooption : option illegal for this statement

                The given I/O option could not be used with this I/O
                statement.

 F2716          INQUIRE : either UNIT= or FILE= needed

 Number         Compiler Compilation Error Message

                The INQUIRE statement must have either a UNIT= option or a
                FILE= option, but not both.

 F2717          UNIT= missing

                This I/O statement lacked a UNIT= option.

 F2718          illegal I/O formatting for internal unit

                Internal units do not allow the use of unformatted or
                list-directed I/O.

                A format specifier other than asterisk (*) must be used.

 F2719          REC= illegal for internal unit

                Direct-access I/O is illegal for internal units.

 F2720          FORMAT : label missing
 Number         Compiler Compilation Error Message
F2720          FORMAT : label missing

                A FORMAT statement lacked a statement label in the range 1 to
                99,999.

 F2721          no ASSIGN statements for FMT=<integer variable>

                The current format specifier had no corresponding ASSIGN
                statement to set the integer variable to a valid FORMAT
                statement label.

 F2722          UNIT= : not between -32767 and 32767

                An external unit number was out of range.

 F2723          iooption : unrecognized value in option

                An invalid or misspelled value was used with the given I/O
                option. For example, ACCESS='DIREKT'  and  ACCESS='RANDOM'
                are both illegal.
 Number         Compiler Compilation Error Message
               are both illegal.

 F2724          RECL= required to open direct-access file

                When opening a file for direct access, the RECL= option is
                required.

 F2725          illegal input list item

                An input list item was not a variable, array, array element,
                or substring.

 F2726          iooption: * illegal with this option

                The asterisk (*) unit specifier (console unit) cannot be used
                with the given I/O control specifier.

 F2727          array : assumed-size array illegal here

                An assumed-size array cannot be used in this context.
 Number         Compiler Compilation Error Message
               An assumed-size array cannot be used in this context.

 F2728          attributes are non-standard

                Specifying attributes is nonstandard (-4Ys or $STRICT has
                been specified).

 F2729          FAR or HUGE I/O item illegal in medium model

                Data items having the FAR or HUGE attribute cannot be used in
                I/O statements in medium-model programs.

 F2730          name : cannot modify active DO variable

                A DO variable cannot be modified within its range. For
                example, the following program fragments cause this error:

                      DO 100 I = 1,10
                      OPEN (33, IOSTAT = I)
                  100 CONTINUE
 Number         Compiler Compilation Error Message
                 100 CONTINUE

                      READ (*,*) (I, I = 1,10)

 F2731          iooption : noncharacter array nonstandard

                If the /4Ys compiler option is used in compiling (or the
                $STRICT metacommand is in effect), standard forms of the
                language must be used. In these cases, only character
                variables, arrays, array elements, and substrings are legal
                as I/O specifiers.

 F2732          stmt : nonstandard statement

                (IMPLICIT NONE, INCLUDE, SELECT, END DO, DO WHILE, NAMELIST,
                TYPE, END TYPE, TYPE ( ), etc.) The statement is non-standard
                and $STRICT has been specified. ALLOCATE, DEALLOCATE TYPE,
                and INTERFACE are also nonstandard features.

 F2733          iooption : option nonstandard
 Number         Compiler Compilation Error Message
F2733          iooption : option nonstandard

                The specified I/O option is not part of standard FORTRAN 77;
                it cannot be given if the /4Ys compiler option is used (or
                the $STRICT metacommand is in effect).

 F2734          END= : illegal when REC= present

                In READ statements, the REC= and END= options cannot both be
                present.

 F2735          REC= : illegal when FMT= *

                In READ and WRITE statements, the REC= option is illegal if
                list-directed I/O is in use.

 F2736          LOCKING : nonstandard

                If the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), the LOCKING statement is
 Number         Compiler Compilation Error Message
               metacommand is in effect), the LOCKING statement is
                prohibited.

 F2737          iooption : lowercase in string nonstandard

                If the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), the value of the specified I/O
                option must be given in uppercase. For example,
                ACCESS='DIRECT' is legal in this case, but  ACCESS='direct'
                is not.

 F2738          name : HUGE internal units illegal

                An array used as an internal unit cannot be declared with the
                HUGE attribute or used in a $LARGE metacommand. The array
                cannot be larger than one segment.

 F2739          name : record length too large for internal unit

                For a noncharacter array used as an internal unit, the
 Number         Compiler Compilation Error Message
               For a noncharacter array used as an internal unit, the
                element size multiplied by the element count (that is, the
                record length of the internal file) was too large.

 F2740          RECL= : out of range

                The value of the RECL= option was less than or equal to 0 or
                exceeded the maximum legal value.

 F2741          ACCESS= : nonstandard option value

                This option is not allowed for file access.

 F2742          format specification illegal when namelist specified

                The NML= and FMT=specifiers are mutually exclusive.

 F2743          name : NML= : not a namelist group name

                The name specified by the NML= spec was not declared in a
 Number         Compiler Compilation Error Message
               The name specified by the NML= spec was not declared in a
                NAMELIST statement.

 F2744          NML= : namelist group name missing

                The NML= spec requires a name as an argument.

 F2745          name : i/o of entire structures illegal

                Only structure variable elements can be written to or read
                from files using formatted I/O.

 F2800          name : CHARACTER*(*) type illegal

                An item was declared with CHARACTER*(*) type, but it was not
                in the formal-argument list in the current subprogram.

 F2801          no ASSIGN statements for assigned GOTO (or FMT=)

                The program unit had no ASSIGN statements for use with an
 Number         Compiler Compilation Error Message
               The program unit had no ASSIGN statements for use with an
                assigned GOTO statement or assigned FMT= specifier.

 F2803          name : ASSIGN : variable not INTEGER

                Only variables of type INTEGER*n or INTEGER are legal in
                ASSIGN statements.

 F2804          name : ASSIGN : too many INTEGER*1 variables

                Only the first 127 ASSIGN statements may use variables of
                type INTEGER*1 in a subprogram. This is caused by the storage
                limitations of INTEGER*1 items.

 F2805          label number : redefined in program unit

                The specified label appeared earlier in the subprogram.

                Labels may not be defined more than once within a single
                subprogram unit. This error may also occur if a DO loop
 Number         Compiler Compilation Error Message
               subprogram unit. This error may also occur if a DO loop
                references a previously defined label.

 F2806          DO-loop variable : not a variable

                A DO-loop variable was a symbolic constant, not an actual
                variable.

 F2807          name : illegal use of active DO-loop variable

                It is illegal to use an active DO-loop variable as another
                DO-loop variable in a nested DO statement. Values cannot be
                assigned to DO-loop variables within DO loops.

 F2808          DO-loop variable not INTEGER or REAL

                Only variables of type INTEGER*n and REAL*n are legal as
                DO-loop variables.

 F2809          DO-loop expression not INTEGER or REAL
 Number         Compiler Compilation Error Message
F2809          DO-loop expression not INTEGER or REAL

                Only expressions of type INTEGER*n and REAL*n are legal as
                DO-loop bounds.

 F2810          zero illegal as increment

                Only nonzero increments are legal as DO-loop increments.
                Otherwise, the loop would never exit.

 F2811          IF or ELSEIF missing

                No IF or ELSEIF statement matching an ELSE or ELSEIF
                statement appeared in the program.

 F2812          ENDIF missing

                Not all IF ENDIF blocks were exited before an END statement
                appeared.

 Number         Compiler Compilation Error Message

 F2813          DO-loop label number : not seen

                Not all DO loops were exited before an END statement
                appeared.

 F2814          IF, ELSEIF, or ELSE missing

                No IF, ELSEIF, or ELSE statement matching an ENDIF statement
                appeared in the program.

 F2815          assigned GOTO variable not INTEGER

                Only items of type INTEGER*n and INTEGER are legal for
                assigned GOTO variables.

 F2816          computed GOTO variable not INTEGER

                Only items of type INTEGER*n and INTEGER are legal for
                computed GOTO variables.
 Number         Compiler Compilation Error Message
               computed GOTO variables.

 F2817          expression type not LOGICAL

                Expression types for logical or block IF statements must be
                of type LOGICAL«*n».

 F2818          expression type not INTEGER or REAL

                Expression types for arithmetic IF statements must be of type
                INTEGER«*n» or REAL«*n».

 F2819          illegal statement after logical IF

                Only single-line statements can follow logical IF statements.
                All executable statements except DO, ELSE, ELSEIF, END,
                ENDIF, block IF, and logical IF can follow a logical IF
                statement.

 F2820          block label number : must not be referenced
 Number         Compiler Compilation Error Message
F2820          block label number : must not be referenced

                Labels that appear on ELSE and ELSEIF statements cannot be
                referenced.

 F2821          label number : previously used as executable label

                The specified label, previously referenced as an executable
                label, was used as a label for a FORMAT statement or
                specification statement.

 F2822          label number : previously used as FORMAT label

                The specified label, previously referenced as a label for a
                FORMAT statement, was used as an executable label or a label
                for a specification statement.

 F2823          DO-loop label number : out of order

                The specified termination label for a DO statement was out of
 Number         Compiler Compilation Error Message
               The specified termination label for a DO statement was out of
                order. DO-loop labels may have been reversed.

 F2824          assigned and unconditional GOTO illegal here

                Assigned and unconditional GOTO statements cannot terminate
                DO loops.

 F2825          block and arithmetic IF illegal here

                Block and arithmetic IF statements cannot terminate a DO
                loop.

 F2826          statement illegal as DO-loop termination

                An ELSE, ELSEIF, END, ENDIF, FORMAT, RETURN, or STOP
                statement cannot be used to terminate a DO loop.

 F2827          STOP (or PAUSE) : maximum of 5 digits

 Number         Compiler Compilation Error Message

                The STOP and PAUSE statements allow only numeric values
                between 0 and 99,999, inclusive.

 F2828          ASSIGN target not an INTEGER variable

                Only variables of type INTEGER are allowed as targets in
                ASSIGN statements.

 F2829          STOP (or PAUSE) : illegal expression

                Only integers or character constants are legal in STOP and
                PAUSE statements.

 F2830          END missing

                An END statement did not appear as the last statement in the
                module.

 F2831          label number : must not be referenced
 Number         Compiler Compilation Error Message
F2831          label number : must not be referenced

                The specified label appeared on a specification or DATA
                statement.

 F2832          statement illegal in INTERFACE

                Only specification statements are legal in INTERFACE
                statements.

 F2833          RETURN : integer or character expression required

                If the /4Ys compiler option is used for compiling (or the
                $STRICT metacommand is in effect), only integer or character
                expressions can follow the RETURN statement.

 F2834          name : alternate RETURN missing

                An alternate RETURN statement was given in the specified
                subprogram when none was given in the subprogram declaration.
 Number         Compiler Compilation Error Message
               subprogram when none was given in the subprogram declaration.

 F2835          statement out of order or END missing

                A specification statement was embedded in execution
                statements, another statement appeared out of the legal
                statement sequence, or an END statement did not terminate a
                previous subprogram.

 F2836          statement out of order

                A statement appeared out of the legal order of statements in
                the program. For example, a specification statement may have
                appeared with execution statements.

 F2837          label number : undefined

                The specified label, which was referenced in a subprogram,
                was not defined.

 Number         Compiler Compilation Error Message

 F2838          statement illegal in BLOCK DATA

                Only type-specification and DATA statements are legal in
                BLOCK DATA subprograms.

 F2839          only variables allowed in assigned GOTO statements

                Only variables are allowed in assigned GOTO statements.

 F2840          name : assumed-size array : not reference argument

                Assumed-size arrays must be passed by reference. They cannot
                be local entities to the subprogram.

 F2841          name : adjustable-size array : not reference argument

                Adjustable-size arrays must be passed by reference. They
                cannot be local entities to a subprogram.

 Number         Compiler Compilation Error Message

 F2842          too many assigned GOTO statements

                Only 255 assigned GOTO statements are allowed in a
                subprogram. (Compile-time error.)

 F2843          statement illegal with INTERFACE TO

                An interface must refer to a subroutine or function. The
                INTERFACE TO statement may not be used to declare calls to
                subprograms. For more information, see the entry for the
                INTERFACE TO statement in Section 4.2, "Statement
                Directory."

 F2844          no matching DO loop

                An EXIT, CYCLE, or END DO was seen without a matching DO
                loop.

 F2845          END SELECT missing
 Number         Compiler Compilation Error Message
F2845          END SELECT missing

                A SELECT construct was not closed by the end of the
                subprogram.

 F2846          DO-LOOP ENDDO : not seen

                A DO loop without a label was not closed by the end of the
                subprogram.

 F2847          statement illegal in STRUCTURE declaration

                Only STRUCTURE statements are allowed in a STRUCTURE
                declaration.

 F2860          expression must be integer, character, or logical

                SELECT CASE expressions must be integer, CHARACTER*1, or
                logical.

 Number         Compiler Compilation Error Message

 F2861          no matching SELECT CASE statement

                An END SELECT statement was seen without a previous SELECT
                CASE statement.

 F2862          only one CASE DEFAULT allowed

                No more than one CASE DEFAULT can appear in a SELECT CASE
                statement.

 F2863          CASE values must be constant expressions

                A CASE value cannot be a variable or an expression that
                contains variables.

 F2864          CASE value type does not match SELECT CASE expression type

                A CASE value type and SELECT CASE expression data types must
                match.
 Number         Compiler Compilation Error Message
               match.

 F2865          overlapping case values

                A given value must match only one CASE.

 F2866          The CASE statement must follow a SELECT CASE statement

                A CASE statement appeared without a previous SELECT CASE
                statement.

 F2867          LOGICAL case value ranges illegal

                A logical case value must not be expressed as a range.

 F2868          SELECT CASE : character expression must be of length 1

                Only CHARACTER*1 character expressions are allowed.

 F2869          lower value exceeds upper value in case value range
 Number         Compiler Compilation Error Message
F2869          lower value exceeds upper value in case value range

                The value to the left of the colon must be less than the
                value to the right of the colon.

 F2870          name : element is an array

                A field which is an array of structures is not being indexed.

 F2901          -4I2 or -4I4 expected

                Only 2- and 4-byte default integer and logical values are
                supported.

 F2902          -4Y and -4N : both options used for argument

                The $DO66 or $FREEFORM metacommand was specified in both the
                /4Y and the /4N compiler options.

 F2993          separator expected in format
 Number         Compiler Compilation Error Message
F2993          separator expected in format

                When the /4Ys compiler option is used (or the $STRICT
                metacommand is in effect), a comma (,), colon (:), right
                parenthesis ()), or slash (/) is expected to separate items
                in a format except in the following cases:

   1. Between a P edit descriptor and an immediately following F, D, E, or G
      edit descriptor

   2. Before or after a slash (/) edit descriptor

   3. Before or after a colon (:) edit descriptor

 F2994          \ or $ : nonstandard edit descriptor in format

                The \ and $ edit descriptors are not part of standard FORTRAN
                77 but are extensions to the language. This error occurs only
                if the /4Ys compiler option is used (or if the $STRICT
                metacommand is in effect).
 Number         Compiler Compilation Error Message
               metacommand is in effect).

 F2995          Z : nonstandard edit descriptor in format

                The Z edit descriptor is not part of standard FORTRAN 77 but
                is an extension to the language. This error occurs only if
                the /4Ys compiler option is used (or if the $STRICT
                metacommand is in effect).


 D.2.3  Recoverable Error Messages

 The messages listed below indicate potential problems but do not hinder
 compilation and linking. The /W compiler option has no effect on the output
 of these messages.

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Compiler Recoverable Error Message

 F3606          subprogram : formal argument name : type mismatch
 Number         Compiler Recoverable Error Message
F3606          subprogram : formal argument name : type mismatch

                The type of a formal argument was different from the type of
                the actual argument used in the subprogram call.

 F3607          subprogram : formal argument name : length mismatch

                The length of a formal argument was different from the length
                of the actual argument used in the subprogram call.


 D.2.4  Warning Error Messages

 The messages listed below indicate potential problems but do not hinder
 compilation and linking.

 F4000          UNKNOWN WARNING Contact Microsoft Technical Support

                An unknown warning has occurred.

 Number         Compiler Recoverable Error Message

                Please report this condition to Microsoft Corporation using
                the Software Problem Report form at the back of this manual.

 F4001          colon expected following ALIAS

                An ALIAS attribute had the wrong form.

                The correct form for ALIAS is the following:

      ALIAS:string

 F4002          $DEBUG:'<debug-list>' illegal with $FREEFORM

                This form of the $DEBUG metacommand was used when the
                $FREEFORM metacommand was in effect.

 F4003          $DECMATH not supported

                The $DECMATH metacommand is not supported in this version of
 Number         Compiler Recoverable Error Message
               The $DECMATH metacommand is not supported in this version of
                FORTRAN.

 F4006          metacommand already set

                A metacommand that may appear only once was reset.

 F4007          metacommand must come before all FORTRAN statements

                This metacommand must appear before all FORTRAN statements.

 F4008          characters following metacommand ignored

                Any characters that follow a fully processed metacommand are
                ignored.

 F4010          filename : error closing file

                A system error occurred while the specified source file was
                being closed.
 Number         Compiler Recoverable Error Message
               being closed.

 F4011          empty escape sequence

                A backslash (\) occurred at the end of a C string such as
                'abc\'. It is replaced by a zero.

                The backslash should be removed.

 F4014          character : nonalphabetic character in $DEBUG ignored

                A nonalphabetic character was included in the list for the
                $DEBUG metacommand.

                The characters a-z or A-Z are the only legal characters. Case
                is ignored.

 F4056          overflow in constant arithmetic

                The result of an operation exceeded #7FFFFFFF.
 Number         Compiler Recoverable Error Message
               The result of an operation exceeded #7FFFFFFF.

 F4057          overflow in constant multiplication

                The result of an operation exceeded #7FFFFFFF.

 F4058          address of frame variable taken, DS != SS

                The program was compiled with the default data segment (DS)
                not equal to the stack segment (SS), and the program tried to
                point to a frame variable with a near pointer.

 F4059          segment lost in conversion

                The conversion of a far pointer (a full segmented address) to
                a near pointer (a segmented offset) resulted in the loss of
                the segmented address.

 F4060          conversion of long address to short address

 Number         Compiler Recoverable Error Message

                The conversion of a long address (a 32-bit pointer) to a
                short address (a 16-bit pointer) resulted in the loss of the
                segmented address.

 F4061          long/short mismatch in argument : conversion supplied

                Actual and formal arguments of a function differed in base
                type. The type of the actual argument was converted to the
                type of the formal argument.

 F4062          near/far mismatch in argument : conversion supplied

                Actual and formal arguments of a function differed in pointer
                size. The size of the actual argument was converted to the
                size of the formal argument.

 F4063          name : function too large for post-optimizer

                The compiler tried to optimize a function but ran out of
 Number         Compiler Recoverable Error Message
               The compiler tried to optimize a function but ran out of
                memory while doing so. It flagged the warning, skipped the
                optimization, and continued the compilation.

                To avoid this problem, break the functions in the program
                into smaller functions.

 F4064          procedure too large, skipping optimization optimization and
                continuing

                The compiler tried to perform the given type of optimization
                on a function but ran out of memory while doing so. It
                flagged the warning, skipped the given part of the
                optimization, and continued the compilation.

                To avoid this problem, break the function into smaller
                functions.

 F4065          recoverable heap overflow in post-optimizer - some
                optimizations may be missed
 Number         Compiler Recoverable Error Message
               optimizations may be missed

                The compiler tried to optimize a function but ran out of
                memory while doing so. It flagged the warning, skipped the
                optimization, and continued the compilation.

                To avoid this problem, break the function into smaller
                functions.

 F4066          local symbol table overflow - some local symbols may be
                missing in listings

                The compiler ran out of memory when it tried to collect the
                local symbols for source listings. Not all of the symbols are
                listed.

 F4072          insufficient memory to process debugging information

                You specified the /Zi compiler option, but the compiler did
                not have enough memory to store all of the required debugging
 Number         Compiler Recoverable Error Message
               not have enough memory to store all of the required debugging
                information. (Compile-time warning.)

 F4186          string too long-truncated to 40 characters

                A string of more than 40 characters was used in a $TITLE or
                $SUBTITLE metacommand. The string is truncated to 40
                characters.

 F4201          ENTRY : formal argument name : ATTRIBUTE attribute :
                mismatch

                VALUE and REFERENCE attributes were mismatched in the
                declaration and use of an ENTRY statement.

 F4202          subprogram : formal argument name : never used

                If a formal argument is never referenced, the compiler must
                assume a variable was meant for this argument. In medium
                model, if a function is passed to the formal argument, the
 Number         Compiler Recoverable Error Message
               model, if a function is passed to the formal argument, the
                wrong amount of storage may be allocated. This message is
                suppressed by any previous compiler error message  (F2xxx).

 F4303          name : language attributes illegal on formal arguments

                A language attribute (C or PASCAL) was specified for a formal
                argument to the current routine. It has no effect.

 F4313          name : not previously declared

                While the /4Yd compiler option was used (or the $DECLARE
                metacommand was in effect), name was not declared in a type
                statement before it was used.

 F4314          intrinsic : declared with wrong type

                The specified name was declared with an incorrect type in an
                INTRINSIC statement. The incorrect type is ignored, and the
                correct type is used.
 Number         Compiler Recoverable Error Message
               correct type is used.

 F4315          name : attribute illegal with attributes specified in same
                list

                The specified attribute contradicts an earlier attribute for
                the item in the same attribute list.

 F4316          name : attribute illegal with attributes specified in
                earlier list

                The specified attribute contradicts an attribute in an
                earlier attribute list for the item.

 F4317          name : attribute attribute repeated

                The specified attribute was already used once in an earlier
                attribute list for the item, and it should only have appeared
                in one attribute list.

 Number         Compiler Recoverable Error Message

 F4318          name : attribute illegal on COMMON statements

                The specified attribute is illegal on common-block
                declarations.

 F4319          name : attribute illegal on formal arguments

                The specified attribute cannot be used on formal arguments.

 F4320          name : attribute illegal on ENTRY statements

                The specified attribute cannot be used on ENTRY statements.

 F4321          name : attribute illegal on subprogram statements

                The specified attribute cannot be used on SUBPROGRAM
                statements.

 F4322          name : attribute illegal on variable declarations
 Number         Compiler Recoverable Error Message
F4322          name : attribute illegal on variable declarations

                The specified attribute cannot be used on variable
                declarations.

 F4323          name : attribute illegal on type declarations

                The specified attribute cannot be used on type declarations.

 F4325          name : attribute illegal on NAMELIST declarations

                Attributes are illegal on NAMELIST declarations.

 F4326          name : EQUIVALENCE : nonconstant upper substring expression
                ignored

                The upper substring expression in an EQUIVALENCE statement
                was not a constant. Since the expression is not used in the
                addressing expression, it is ignored.

 Number         Compiler Recoverable Error Message

 F4327          name : INTERFACE : not formal argument

                A variable was declared that was not given in the
                formal-argument list to the subprogram specified in the
                INTERFACE statement.

 F4328          name : attribute illegal on STRUCTURE declarations

                An illegal attribute was specified on a STRUCTURE
                declaration.

 F4329          %fs : COMMON : size changed

                This is a warning level message which is the same as F2323.
                F2323 now only occurs if $STRICT (or -4Ys) is set. Otherwise,
                the warning occurs.

 F4400          DATA : more constants than names

 Number         Compiler Recoverable Error Message

                Extra constants appearing in a constant list of a DATA
                statement were ignored.

 F4501          array : subscript number out of range

                The /4Yb compiler option was used in compiling (or the $DEBUG
                metacommand was in effect), and an array passed as an
                argument had a bound out of range. (This practice is legal
                for formal arguments because it is common in FORTRAN to
                declare the last bound to be 1.)

 F4602          name : alternate RETURN statement missing

                The subprogram declaration where the specified name appeared
                had no alternate RETURN statement.

 F4605          name : FAR formal argument name : passed HUGE array

                An array declared with a HUGE attribute was passed to a
 Number         Compiler Recoverable Error Message
               An array declared with a HUGE attribute was passed to a
                formal argument declared with a FAR attribute.

 F4608          name : formal argument name : passed FAR/HUGE

                A variable which was declared NEAR is being passed FAR.

 F4801          label number : used across blocks

                An executable statement label was referenced across a
                statement block. This situation may arise in the following
                cases:

   ■  When a GOTO statement uses a statement label in a different arm of an
      IF...ELSE...ENDIF statement

   ■  When the program jumps into a DO loop

 F4802          no assigned GOTO or FMT= for ASSIGN statement

 Number         Compiler Recoverable Error Message

                An ASSIGN statement was used to assign a label to a variable
                in the subprogram, but the variable was not used.

 F4803          name : FUNCTION : return variable not set

                A return variable specified in a FUNCTION statement was not
                set at least once in the function.

 F4901          -4Y and -4N : both options used; -4Y assumed

                The $DEBUG, $DECLARE, $LIST, $STRICT, or $TRUNCATE
                metacommand was specified with both the /4Y and /4N compiler
                options. For example, $DEBUG was specified using both a /4Yb
                option and a /4Nb option.

 F4902          -Wnumber : illegal warning level ignored

                This is an internal check. Microsoft FORTRAN supports only
                warning levels 0 and 1.
 Number         Compiler Recoverable Error Message
               warning levels 0 and 1.

 F4903          -Zpnumber : illegal pack value ignored

                Only the structure packing values 1, 2, and 4 are valid.

 F4980          integer expected in format

                An edit descriptor lacked a required integer value.

 F4981          initial left parenthesis expected in format

                A format did not start with a left parenthesis (().

 F4982          positive integer expected in format

                An unexpected negative or 0 value was used in a format.

                Negative integer values can appear only with the P edit
                descriptor. Integer values of 0 can appear only in the d and
 Number         Compiler Recoverable Error Message
               descriptor. Integer values of 0 can appear only in the d and
                m fields of numeric edit descriptors.

 F4983          repeat count on nonrepeatable descriptor

                One or more BN, BZ, S, SP, SS, T, TL, TR, /, \, $, :, or
                apostrophe (') edit descriptors had repeat counts associated
                with them.

 F4984          integer expected preceding H, X, or P edit descriptor

                An integer did not precede a (nonrepeatable) H, X, or P edit
                descriptor.

                The correct formats for these edit descriptors are nH, nX,
                and kP, respectively, where n is a positive integer and k is
                an optionally signed integer.

 F4985          N or Z expected after B in format

 Number         Compiler Recoverable Error Message

                An illegal edit descriptor beginning with "B" was used.

                The only valid edit descriptors beginning with "B" are BN and
                BZ, used to specify the interpretation of blanks as nulls or
                zeros, respectively.

 F4986          format nesting limit exceeded

                More than 16 sets of parentheses were nested inside the main
                level of parentheses in a format.

 F4987          '.' : expected in format

                A period did not appear between the w and d fields of a D, E,
                F, or G edit descriptor.

 F4988          unexpected end of format

                An incomplete format was used.
 Number         Compiler Recoverable Error Message
               An incomplete format was used.

                Improperly matched parentheses, an unfinished Hollerith (H)
                descriptor, or another incomplete descriptor specification
                can cause this error.

 F4989          'character' : unexpected character in format

                A character that cannot be interpreted as a valid edit
                descriptor was used in a format.

 F4990          M field exceeds W field in I edit descriptor

                The length of the m field specified in an I edit descriptor
                exceeded the length of the w field.

 F4997          CHARACTER*(*) in multithread-thread may not work

                Character*(*) arguments and functions may not work when
                linked with the multithread-thread FORTRAN run-time library.
 Number         Compiler Recoverable Error Message
               linked with the multithread-thread FORTRAN run-time library.

 F4998          name :  variable used but not defined

                A variable was used in the subprogram but never given a
                value.

 F4999          name : variable declared but not used

                A variable was declared, but never referenced anywhere else
                in the subprogram. (This is one of only two warnings shut off
                by -W2.)


 D.3  Run-Time Error Messages

 Run-time error messages fall into two categories:

   1. Error messages generated by the run-time library to notify you of
      serious errors. These messages are listed and described in Section
      D.3.1.

   2. Floating-point exceptions generated by the 8087/80287 hardware or the
      emulator. These exceptions are listed and described in Section D.3.2,
      "Other Run-Time Error Messages."


 D.3.1  Run-Time-Library Error Messages

 The following messages may appear at run time when your program has serious
 errors. Run-time error-message numbers range from F6000 to F6999.

 A run-time error message takes the following general form:

      «sourcefile(line) :» run-time error F6nnn : operation«(filename)»
      - messagetext

 The sourcefile (line) information appears only when the $DEBUG metacommand
 is in effect.

 For operation, one of the following may appear: ALLOCATE, BACKSPACE,
 BEGINTHREAD, CLOSE, DEALLOCATE, ENDFILE, EOF, INQUIRE, LOCKING, OPEN, READ,
 REWIND, WRITE, or $DEBUG.

 The filename of the file affected by operation is shown except when
 operation is $DEBUG.

 The messagetext follows on the next line.

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Run-Time Error Messages

 F6096          array subscript expression out of range

                An expression used to index an array was smaller than the
                lower dimension bound or larger than the upper dimension
                bound. This message appears only if the /4Yb option is used
                in compiling (or the $DEBUG metacommand is in effect).

 F6097          CHARACTER substring expression out of range

                An expression used to index a character substring was
 Number         Run-Time Error Messages
               An expression used to index a character substring was
                illegal. This message appears only if the /4Yb option is used
                in compiling (or the $DEBUG metacommand is in effect).

 F6098          label not found in assigned GOTO list

                The label assigned to the integer-variable name was not
                specified in the label list of the assigned GOTO statement.
                This message appears only if the /4Yb option is used in
                compiling (or the $DEBUG metacommand is in effect).

 F6099          INTEGER overflow

                This error occurs whenever integer arithmetic results in
                overflow, or when assignment to an integer is out of range.
                This message appears only if the /4Yb option is used in
                compiling (or the $DEBUG metacommand is in effect).

 F6100          INTEGER overflow on input

 Number         Run-Time Error Messages

                An INTEGER*n item exceeded the legal size limits.

                An INTEGER*1 item must be in the range -127 to 127. An
                INTEGER*2 item must be in the range -32,767 to 32,767. An
                INTEGER*4 item must be in the range -2,147,483,647 to
                2,147,483,647.

 F6101          invalid INTEGER

                Either an illegal character appeared as part of an integer,
                or a numeric character larger than the radix was used in an
                alternate radix specifier.

 F6102          REAL indefinite (uninitialized or previous error)

                An illegal argument was specified for an intrinsic function
                (for example, SQRT(-1) or ASIN(2)). This error message does
                not always appear where the mistake was originally made. It
                may appear if the invalid value is used later in the program.
 Number         Run-Time Error Messages
               may appear if the invalid value is used later in the program.

 F6103          invalid REAL

                An illegal character appeared as part of a real number.

 F6104          REAL math overflow

                A real value was too large. Floating-point overflows in
                either direct or emulated mode generate NAN (Not-A-Number)
                exceptions, which appear in the output field as asterisks (*)
                or the letters  NAN.

 F6200          formatted I/O not consistent with OPEN options

                The program tried to perform formatted I/O on a unit opened
                with FORM='UNFORMATTED' or FORM='BINARY'.

 F6201          list-directed I/O not consistent with OPEN options

 Number         Run-Time Error Messages

                The program tried to perform list-directed I/O on a file that
                was not opened with FORM='FORMATTED' and ACCESS='SEQUENTIAL'.

 F6202          terminal I/O not consistent with OPEN options

                The ACCESS='SEQUENTIAL' option and either the
                FORM='FORMATTED' or the FORM='BINARY' option were not
                included in the OPEN statement for a special device name such
                as CON, LPT1, or PRN. These options are required because
                special device names are connected to devices that do not
                support direct access.

                When a unit is connected to a terminal device, an OPEN
                statement that has the options FORM='FORMATTED' and
                ACCESS='SEQUENTIAL' results in carriage control. If the
                FORM='BINARY' and ACCESS='SEQUENTIAL' options appear in an
                OPEN statement, binary data transfer takes place.

 F6203          direct I/O not consistent with OPEN options
 Number         Run-Time Error Messages
F6203          direct I/O not consistent with OPEN options

                A REC= option was included in a statement that transferred
                data to a file that was opened with the ACCESS='SEQUENTIAL'
                option.

 F6204          unformatted I/O not consistent with OPEN options

                If a file is opened with FORM='FORMATTED', unformatted or
                binary data transfer is prohibited.

 F6205          A edit descriptor expected for CHARACTER

                The A edit descriptor was not specified when a character data
                item was read or written using formatted I/O.

 F6206          E, F, D, or G edit descriptor expected for REAL

                The E, F, D, or G edit descriptor was not specified when a
                real data item was read or written using formatted I/O.
 Number         Run-Time Error Messages
               real data item was read or written using formatted I/O.

 F6207          I edit descriptor expected for INTEGER

                The I edit descriptor was not specified when an integer data
                item was read or written using formatted I/O.

 F6208          L edit descriptor expected for LOGICAL

                The L edit descriptor was not specified when a logical data
                item was read or written using formatted I/O.

 F6209          file already open : parameter mismatch

                An OPEN statement specified a connection between a unit and a
                file name that was already in effect. In this case, only the
                BLANK= option can have a different setting.

 F6210          namelist I/O not consistent with OPEN options

 Number         Run-Time Error Messages

                The program tried to perform namelist I/O on a file that was
                not opened with FORM='FORMATTED' and ACCESS='SEQUENTIAL'.

 F6300          KEEP illegal for scratch file

                STATUS='KEEP' was specified for a scratch file; this is
                illegal because scratch files are automatically deleted at
                program termination.

 F6301          SCRATCH illegal for named file

                STATUS='SCRATCH' should not be used in an OPEN statement that
                includes a file name.

 F6302          multiple radix specifiers

                More than one alternate radix for numeric I/O was specified.

 F6303          illegal radix specifier
 Number         Run-Time Error Messages
F6303          illegal radix specifier

                A radix specifier was not between 2 and 36, inclusive.

 F6304          illegal STATUS value

                An illegal value was used with the STATUS= option.

                STATUS= accepts the following values:

   1. 'KEEP' or 'DELETE' when used with CLOSE statements

   2. 'OLD', 'NEW', 'SCRATCH', or 'UNKNOWN' when used with OPEN statements

 F6305          illegal MODE value

                An illegal value was used with the MODE= option.

                MODE= accepts the values 'READ', 'WRITE', or 'READWRITE'.

 Number         Run-Time Error Messages

 F6306          illegal ACCESS value

                An illegal value was used with the ACCESS= option.

                ACCESS= accepts the values 'SEQUENTIAL' and 'DIRECT'.

 F6307          illegal BLANK value

                An illegal value was used with the BLANK= option.

                BLANK= accepts the values 'NULL' and 'ZERO'.

 F6308          illegal FORM value

                An illegal value was used with the FORM= option.

                FORM= accepts the following values: 'FORMATTED',
                'UNFORMATTED', and 'BINARY'.

 Number         Run-Time Error Messages

 F6309          illegal SHARE value

                An illegal value was used with the SHARE= option.

                SHARE= accepts the values 'COMPAT', 'DENYRW', 'DENYWR',
                'DENYRD', and 'DENYNONE'.

 F6310          illegal LOCKMODE value

                An illegal value was used with the LOCKMODE= option.

                LOCKMODE= accepts the values 'LOCK', 'NBLCK', 'NBRLCK',
                'RLCK', and 'UNLCK'.

 F6311          illegal record number

                An invalid number was specified as the record number for a
                direct-access file.

 Number         Run-Time Error Messages

                The first valid record number for direct-access files is 1.

 F6312          no unit number associated with *

                In an INQUIRE statement, the NUMBER= option was specified for
                the file associated with * (console).

 F6313          illegal RECORDS value

                The RECORDS= option in a LOCKING statement specified a
                negative number.

 F6314          illegal unit number

                An illegal unit number was specified.

                Legal unit numbers can range from -32,767 to 32,767,
                inclusive.

 Number         Run-Time Error Messages

 F6315          illegal RECL value

                A negative or zero record length was specified for a direct
                file.

                The smallest valid record length for direct files is 1.

 F6316          array already allocated

                The program attempted to reallocate an already allocated
                array.

 F6317          array size zero or negative

                The size specified for an array in an ALLOCATE statement must
                be greater than zero.

 F6318          non-HUGE array exceeds 64K

 Number         Run-Time Error Messages

                The memory space required for an array in an ALLOCATE
                statement exceeds 64K, but the HUGE attribute was not
                specified.

 F6319          array not allocated

                The program attempted to DEALLOCATE an array that was never
                allocated.

 F6400          BACKSPACE illegal on terminal device

                A BACKSPACE statement specified a unit connected to a
                terminal device such as a terminal or printer.

 F6401          EOF illegal on terminal device

                An EOF intrinsic function specified a unit connected to a
                terminal device such as a terminal or printer.

 Number         Run-Time Error Messages

 F6402          ENDFILE illegal on terminal device

                An ENDFILE statement specified a unit connected to a terminal
                device such as a terminal or printer.

 F6403          REWIND illegal on terminal device

                A REWIND statement specified a unit connected to a terminal
                device such as a terminal or printer.

 F6404          DELETE illegal for read-only file

                A CLOSE statement specified STATUS='DELETE' for a read-only
                file.

 F6405          external I/O illegal beyond end of file

                The program tried to access a file after executing an ENDFILE
                statement or after it encountered the end-of-file record
 Number         Run-Time Error Messages
               statement or after it encountered the end-of-file record
                during a read operation.

                A BACKSPACE, REWIND, or OPEN statement must be used to
                reposition the file before execution of any I/O statement
                that transfers data.

 F6406          truncation error : file closed

                This is a transient error. While the file was being
                truncated, it was temporarily closed.

                After a few minutes, the file should be run again. If this
                error message reappears, the file should be checked for
                characteristics, such as locking or permissions, that would
                prevent it from being accessed.

 F6407          terminal buffer overflow

                More than 131 characters were input to a record of a unit
 Number         Run-Time Error Messages
               More than 131 characters were input to a record of a unit
                connected to the terminal (keyboard). Note that the operating
                system may impose additional limits on the number of
                characters that can be input to the terminal in a single
                record.

 F6408          comma delimiter disabled after left repositioning

                A comma could not be used as a field delimiter. This is
                because the use of commas as input field delimiters is
                disabled if left tabbing leaves the file positioned in a
                previous buffer.

 F6409          LOCKING illegal on sequential file

                A LOCKING statement specified a unit that was not opened with
                ACCESS='DIRECT'.

 F6410          file already locked or unlocked

 Number         Run-Time Error Messages

                The program tried to lock a file that was already locked or
                tried to unlock a file that was already unlocked.

 F6411          file deadlocked

                A LOCKING statement that included the 'LOCK' or 'RLCK' value
                tried to lock a file, but the file could not be locked after
                10 attempts.

 F6412          SHARE not installed

                The SHARE.COM or SHARE.EXE file must be installed on your
                system before you can use the LOCKING statement, or the
                SHARE= option in an OPEN statement.

 F6413          file already connected to a different unit

                The program tried to connect an already connected file to a
                new unit.
 Number         Run-Time Error Messages
               new unit.

                A file can be connected to only one unit at a time.

 F6414          access not allowed

                This error is caused by one of the following occurrences:

   ■  The file name specified in an OPEN statement was a directory.

   ■  An OPEN statement tried to open a read-only file for writing.

   ■  The file's sharing mode does not allow the specified operations (DOS
      Versions 3.0 and later only).

 F6415          file already exists

                An OPEN statement specified STATUS='NEW' for a file that
                already exists.

 Number         Run-Time Error Messages

 F6416          file not found

                An OPEN statement specified STATUS='OLD' for a file that does
                not exist.

 F6417          too many open files

                The program exceeded the system limit on the number of open
                files allowed at one time.

                To fix this problem, change the FILES= command in the
                CONFIG.SYS file.

 F6418          too many units connected

                The program exceeded the limit on the number of open files
                per program.

                Close any unnecessary files. See the FILES= command in the
 Number         Run-Time Error Messages
               Close any unnecessary files. See the FILES= command in the
                Microsoft MS-DOS User's Guide and User's Reference for more
                information.

 F6419          illegal structure for unformatted file

                The file was opened with FORM='UNFORMATTED' and
                ACCESS='SEQUENTIAL', but its internal physical-record
                structure was incorrect or inconsistent.

 F6420          unknown unit number

                A statement such as BACKSPACE or ENDFILE specified a file
                that had not yet been opened. (The READ and WRITE statements
                do not cause this problem since, instead of generating this
                error, they prompt you for a file if the file has not been
                opened yet.)

 F6421          file read-only or locked against writing

 Number         Run-Time Error Messages

                The program tried to transfer data to a file that was opened
                in read-only mode or locked against writing.

 F6422          no space left on device

                The program tried to transfer data to a file residing on a
                device that was out of storage space.

 F6423          too many threads

                The program attempted to execute more threads than the
                FORTRAN run-time system can handle.

 F6424          invalid argument

                The system could not begin a thread of execution because an
                argument to the BEGINTHREAD routine is incorrect. This
                usually occurs when the stack argument does not start on a
                word boundary (even address), or the stack size argument is
 Number         Run-Time Error Messages
               word boundary (even address), or the stack size argument is
                odd or zero, or the stack crosses a segment boundary.

 F6500          file not open for reading or file locked

                The program tried to read from a file that was not opened for
                reading or was locked.

 F6501          end of file encountered

                The program tried to read more data than the file contains.

 F6502          positive integer expected in repeat field

                When the r*c form is used in list-directed input, the r must
                be a positive integer.

 F6503          multiple repeat field

                In list-directed input of the form r*c, an extra repeat field
 Number         Run-Time Error Messages
               In list-directed input of the form r*c, an extra repeat field
                was used. For example,

                      READ(*,*) I,J,K

                with input  2*1*3  returns this error. The  2*1  means send
                two values, each 1; the  *3  is an error.

 F6504          invalid number in  input

                Some of the values in a list-directed input record were not
                numeric. The following example would cause this error:

                123abc

 F6505          invalid string in input

                A string item was not enclosed in single quotation marks.

 F6506          comma missing in COMPLEX input
 Number         Run-Time Error Messages
F6506          comma missing in COMPLEX input

                When using list-directed input, the real and imaginary
                components of a complex number were not separated by a comma.

 F6507          T or F expected in LOGICAL read

                The wrong format was used for the input field for logical
                data.

                The input field for logical data consists of optional blanks,
                followed by an optional decimal point, followed by a T for
                true or F for false. The T or F may be followed by additional
                characters in the field, so that .TRUE. and .FALSE. are
                acceptable input forms.

 F6508          too many bytes read from unformatted record

                The program tried to read more data from an unformatted file
                than the current record contained. If the program was reading
 Number         Run-Time Error Messages
               than the current record contained. If the program was reading
                from an unformatted direct file, it tried to read more than
                the fixed record length as specified by the RECL= option. If
                the program was reading from an unformatted sequential file,
                it tried to read more data than was written to the record.

 F6509          H or apostrophe edit descriptor illegal on input

                Hollerith (H) or apostrophe edit descriptors were encountered
                in a format used by a READ statement.

 F6510          illegal character in hexadecimal input

                The input field contained a character that was not
                hexadecimal.

                Legal hexadecimal characters are 0-9 and A-F.

 F6511          variable name not found

 Number         Run-Time Error Messages

                A name encountered on input from a namelist record is not
                declared in the corresponding NAMELIST statement.

 F6512          invalid NAMELIST input format

                The input record is not in the correct form for namelist
                input.

 F6513          wrong number of array dimensions

                In namelist input, an array name was qualified with a
                different number of subscripts than its declaration, or a
                non-array name was qualified.

 F6514          array subscript exceeds allocated area

                A subscript was specified in namelist input which exceeded
                the declared dimensions of the array.

 Number         Run-Time Error Messages

 F6515          invalid subrange in namelist input

                A character item in namelist input was qualified with a
                subrange that did not meet the requirement that 1 <= e1 <= e2
                <= len (where 'len' is the length of the character item, 'e1'
                is the leftmost position of the substring, and 'e2' is the
                rightmost position of the substring).

 F6516          substring range specified on non-CHARACTER item

                A non-character item in namelist input was qualified with a
                substring range.

 F6600          internal file overflow

                The program either overflowed an internal-file record or
                tried to write to a record beyond the end of an internal
                file.

 Number         Run-Time Error Messages

 F6601          direct record overflow

                The program tried to write more than the number of bytes
                specified in the RECL= option to an individual record of a
                direct-access file.

 F6602          numeric field bigger than record size

                The program tried to write a noncharacter item across a
                record boundary in list-directed or namelist output. Only
                character constants can cross record boundaries.

 F6700          heap space limit exceeded

                The program tried to open too many files at once. A file
                control block (FCB) must be allocated from the heap for each
                file opened, but no more heap space was available.

 F6701          scratch file name limit exceeded
 Number         Run-Time Error Messages
F6701          scratch file name limit exceeded

                The program exhausted the template used to generate unique
                scratch-file names.

 F6980          integer expected in format

                An edit descriptor lacked a required integer value.

 F6981          initial left parenthesis expected in format

                A format did not begin with a left parenthesis (().

 F6982          positive integer expected in format

                A zero or negative integer value was used in a format.

                Negative integer values can appear only with the P edit
                descriptor. Integer values of 0 can appear only in the d and
                m fields of numeric edit descriptors.
 Number         Run-Time Error Messages
               m fields of numeric edit descriptors.

 F6983          repeat count on nonrepeatable descriptor

                One or more BN, BZ, S, SS, SP, T, TL, TR, /, \, $, :, or
                apostrophe (') edit descriptors had repeat counts associated
                with them.

 F6984          integer expected preceding H, X, or P edit descriptor

                An integer did not precede a (nonrepeatable) H, X, or P edit
                descriptor.

                The correct formats for these descriptors are nH, nX, and kP,
                respectively, where n is a positive integer and k is an
                optionally signed integer.

 F6985          N or Z expected after B in format

                An illegal edit descriptor beginning with "B" was used.
 Number         Run-Time Error Messages
               An illegal edit descriptor beginning with "B" was used.

                The only valid edit descriptors beginning with "B" are BN and
                BZ, used to specify the interpretation of blanks as nulls or
                zeros, respectively.

 F6986          format nesting limit exceeded

                More than 16 sets of parentheses were nested inside the main
                level of parentheses in a format.

 F6987          '.' expected in format

                No period appeared between the w and d fields of a D, E, F,
                or G edit descriptor.

 F6988          unexpected end of format

                An incomplete format was used.

 Number         Run-Time Error Messages

                Improperly matched parentheses, an unfinished Hollerith (H)
                descriptor, or another incomplete descriptor specification
                can cause this error.

 F6989          unexpected character in format

                A character that cannot be interpreted as part of a valid
                edit descriptor was used in a format.

 F6990          M field exceeds W field in I edit descriptor

                The value of the m field specified in an I edit descriptor
                exceeded the value of the w field.

 F6991          integer out of range in format

                An integer value specified in an edit descriptor was too
                large to represent as a 4-byte integer.

 Number         Run-Time Error Messages

 F6992          format not set by ASSIGN

                The format specifier in a READ, WRITE, or PRINT statement was
                an integer variable, but an ASSIGN statement did not properly
                assign it the statement label of a FORMAT statement in the
                same program unit.


 D.3.2  Other Run-Time Error Messages

 The following sections describe math run-time errors and general run-time
 errors. Math run-time errors are divided into low-level and function-level
 math errors.


 Low-Level Math Errors

 The error messages listed below correspond to exceptions generated by the
 8087/80287 hardware. Refer to the Intel documentation for your processor for
 a detailed discussion of hardware exceptions. These errors may also be
 detected by the floating-point emulator or alternate math library.

 Using FORTRAN's default 8087/80287 control-word settings, the following
 exceptions are masked and do not occur:

 Exception                                  Default Masked Action

 Denormal                                   Exception masked

 Underflow                                  Result goes to 0.0

 Inexact                                    Exception masked

 See Chapter 1, "Controlling Floating-Point Operations," in Microsoft FORTRAN
 Advanced Topics for information on how to change the floating-point control
 word.

 The following errors do not occur with code generated by the Microsoft
 FORTRAN Compiler or code provided in the standard Microsoft FORTRAN
 libraries:

      square root
      stack underflow
      unemulated

 The low-level math error messages, listed below, have the following format:

      «sourcefile(line) : » run-time error M61xx: MATH
      - floating-point error: messagetext

 The sourcefile and line where the error occurred appear only if the /4Yb
 option is used in compiling (or the $DEBUG metacommand is in effect).

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Low-Level Math Errors

 M6101          invalid

                An invalid operation occurred. This usually involves
                operating on NANs or infinities. This error terminates the
                program with exit code 129.
 Number         Low-Level Math Errors
               program with exit code 129.

 M6102          denormal

                A very small floating-point number was generated, which may
                no longer be valid due to loss of significance. Denormals are
                normally masked, causing them to be trapped and operated
                upon. This error terminates the program with exit code 130.

 M6103          divide by 0

                An attempt was made to divide by zero. This error terminates
                the program with exit code 131.

 M6104          overflow

                An overflow occurred in a floating-point operation. This
                error terminates the program with exit code 132.

 M6105          underflow
 Number         Low-Level Math Errors
M6105          underflow

                An underflow occurred in a floating-point operation. (An
                underflow is normally masked so that the underflowing value
                is replaced with 0.0.) This error terminates the program with
                exit code 133.

 M6106          inexact

                Loss of precision occurred in a floating-point operation.
                This exception is normally masked, since almost any
                floating-point operation can cause loss of precision. This
                error terminates the program with exit code 134.

 M6107          unemulated

                An attempt was made to execute an 8087/80287 instruction that
                is invalid or is not supported by the emulator. This error
                terminates the program with exit code 135.

 Number         Low-Level Math Errors

 M6108          square root

                The operand in a square-root operation was negative. The
                FORTRAN intrinsic function SQRT does not generate this
                message; instead, SQRT gives a function-level  DOMAIN  error.
                This error terminates the program with exit code 136 (see
                error message M6201 below).

 M6110          stack overflow

                A floating-point expression caused a stack overflow on the
                8087/80287 or emulator. Stack-overflow exceptions are trapped
                up to seven additional levels beyond the eight levels
                normally supported by the 8087/80287 processor. This error
                terminates the program with exit code 138.

 M6111          stack underflow

                A floating-point operation resulted in a stack underflow on
 Number         Low-Level Math Errors
               A floating-point operation resulted in a stack underflow on
                the 8087/80287 or emulator. This error terminates the program
                with exit code 139.


 Function-Level Math Errors

 The function-level math error messages, listed below, appear when there are
 errors in the use of intrinsic functions. The error messages have the
 following format:

      «sourcefile(line): » run-time error M62xx: MATH
      -  functionname: messagetext

 The sourcefile and the line where the error occurred appear only if the /4Yb
 compiler option is used (or the $DEBUG metacommand is in effect).

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         Function-Level Math Errors

 Number         Function-Level Math Errors

 M6201          functionname : DOMAIN error

                An argument to the given function was outside the domain of
                that function (that is, outside the legal set of input values
                for the function), as in the following examples:

                SQRT(-1.0)
                ACOS(-5.0)

 M6202          functionname : SING error

                This error refers to argument singularity. The given function
                was not properly defined for the value of its actual
                argument, although it may have been defined at nearby points,
                as in the following examples:

                LOG10(0.0)

                (0.0)**(-3.0)
 Number         Function-Level Math Errors
               (0.0)**(-3.0)

 M6203          functionname : OVERFLOW error

                The result of the given function or one of its intermediate
                computations was too large to be represented, as in the
                following example:

                EXP(25000.0)

 M6204          functionname : UNDERFLOW error

                The result of the given function or one of its intermediate
                computations was too small to be represented. (This error is
                not currently supported; instead, the underflowing value goes
                to 0.0.)

 M6205          functionname : TLOSS error

                A total loss of significance (precision) occurred, as in the
 Number         Function-Level Math Errors
               A total loss of significance (precision) occurred, as in the
                following example:

                COS(1.0E30)

 M6206          functionname : PLOSS error

                A partial loss of significance (precision) occurred. (This
                error is not currently supported; instead, the
                less-significant value is propagated to other computations or
                returned as the result.)


 General Run-Time Error Messages

 The following messages indicate general problems that may occur during
 program start-up, termination, or execution. These error messages have the
 following format:

      «sourcefile(line): » run-time error R6xxx
      - messagetext

 The sourcefile and the line where the error occurred appear only if the /4Yb
 compiler option is used to compile the program (or the $DEBUG metacommand is
 in effect). This additional information is not available for  R6002,  R6004,
 R6008,  and  R6009, which appear at start-up time.

╓┌─────────────┌─────────────────────────────────────────────────────────────╖
 Number         General Run-Time Error Messages

 R6000          stack overflow

                Your program ran out of stack space. This can occur when a
                program uses a large amount of space for local data or
                temporary files or uses subprogram calls that are nested too
                deeply. The program is terminated with exit code 255.

                To correct, relink using the linker /STACK option to allocate
                a large stack, or relink using the compiler option /Fhexnum.
                You can also compile your program with the /Ge compiler
                option to check for stack-overflow errors and modify the
 Number         General Run-Time Error Messages
               option to check for stack-overflow errors and modify the
                stack information in the executable-file header by using the
                EXEMOD program.

 R6001          null pointer assignment

                The contents of the NULL segment changed in the course of
                program execution. The NULL segment is a special low-memory
                location (starting at offset 0 in DGROUP) that is not
                normally used. If the contents of the NULL segment change
                during a program's execution, it means that the program has
                written to this area, usually by an inadvertent assignment
                through a null pointer (a memory address whose offset is 0 in
                the default data segment). Note that your program can contain
                null pointers without generating this message; the message
                appears only when you access a memory location through the
                null pointer.

                This error does not cause your program to terminate; the
                error message is printed following the normal termination of
 Number         General Run-Time Error Messages
               error message is printed following the normal termination of
                the program. This error yields a nonzero exit code.

                This message reflects a potentially serious error in your
                program. Although a program that produces this error may
                appear to operate correctly, it is likely to cause problems
                in the future and may fail to run in a different operating
                environment.

 R6002          floating point not loaded

                This error occurs when inadequate floating-point support has
                been loaded. The program terminates with exit code 255. Three
                situations can cause this error:

   1. The program was compiled or linked with an option (such as /FPi87) that
      required an 8087 or 80287 coprocessor, but the program was run on a
      machine that did not have a coprocessor installed.

      To fix this problem, recompile the program with the /FPi option, relink
 Number         General Run-Time Error Messages
     To fix this problem, recompile the program with the /FPi option, relink
      with an emulator library (LLIBFORE.LIB or MLIBFORE.LIB), or install a
      coprocessor. (See Chapter 1, "Controlling Floating-Point Operations,"
      in Microsoft FORTRAN Advanced Topics for more information about these
      options and libraries.)

   2. In a mixed-language program module that uses the C scanf or printf
      functions or their variants, a call to one of these functions included
      a floating-point format specification (such as f), but no
      floating-point values or variables appeared within the same C module.
      (The C compiler uses the presence of floating-point values and
      variables to determine whether or not to load floating-point conversion
      support.)

      To fix this problem, use a floating-point argument to correspond to the
      floating-point format specification in the scanf or printf call.

   3. In a mixed-language program that uses both C and FORTRAN modules, a C
      library (LLIBC.LIB or MLIBC.LIB) was specified before a FORTRAN library
      (LLIBFORx.LIB or MLIBFORx.LIB) in the linking stage.
 Number         General Run-Time Error Messages
     (LLIBFORx.LIB or MLIBFORx.LIB) in the linking stage.

      To fix this problem, relink and specify the libraries in reverse order:
      the FORTRAN library followed by the C library. (See Chapter 4,
      "Mixed-Language Programming," in Microsoft FORTRAN Advanced Topics for
      more information.)

 R6003          integer divide by 0

                An attempt was made to divide an integer by 0, giving an
                undefined result. Raising an integer 0 to a negative integer
                power may also cause this error. This error terminates the
                program with exit code 255.

 R6008          not enough space for arguments

                At start-up time, there was enough memory to load the
                program, but not enough room for the command-line arguments.

 R6009          not enough space for environment
 Number         General Run-Time Error Messages
R6009          not enough space for environment



 Glossary

 ───────────────────────────────────────────────────────────────────────────

 The definitions in this glossary are intended primarily for use with this
 manual and Microsoft FORTRAN Advanced Topics.

 8087, 80287, and 80387 coprocessors
   Intel(R) hardware products that provide very fast and precise number
   processing.

 Active page
   The area of memory that graphics instructions currently write to. This may
   or may not be the visual page.

 Actual argument
   The specific item (such as a variable, array, or expression) passed to a
   subroutine or function at a specific calling location.

 Alphanumeric
   A letter or a number.

 Argument
   A value passed to and from functions and subroutines.

 Array declarator
   The specifier array(«lower:»upper).

 Associated
   Referring to the same memory location.

 Attribute
   A keyword that specifies additional information about a variable, variable
   type, subprogram, or subprogram formal argument.

 Base name
   The portion of the file name that precedes the file-name extension. For
   example, samp  is the base name of the file  samp.for.

 Binary
   Base-2 notation.

 C string
   A character constant followed by the character C. The character constant
   is then interpreted as a C-language constant.

 Column-major order
   The order in which array elements are stored; the leftmost subscript is
   incremented first when the array is mapped into contiguous memory
   addresses.

 Compiland
   A file containing ASCII text to be compiled by the Microsoft FORTRAN
   Compiler. A compiland is also called a source file or program file.

 Compile time
   The time during which the compiler is executing, compiling a Microsoft
   FORTRAN source file, and creating a relocatable object file.

 Compiler
   A program that translates FORTRAN programs into code understood by the
   computer.

 Complex number
   A number with a real and an imaginary part.

 Constant folding
   The process of evaluating expressions that contain only constants, and
   then substituting the calculated constant for the expression. Constant
   folding is performed by the compiler during optimization. The expression
   9*3.1, for example, becomes  27.9.

 Dimension declarator
   The specifier «lower:»upper; an array has as many dimensions as it has
   dimension declarators. The number upper - lower + 1 is the size of a
   dimension.

 Domain
   The range of a function's valid input values. For example, in the
   expression  y = f(x), the domain of the function  f(x)  is the set of all
   values of  x  for which  f(x)  is defined. A DOMAIN error message is
   returned when an argument to a function is outside the domain of the
   function.

 Double-precision real number
   A real number that is allocated eight bytes of memory.

 Executable program
   A file containing executable program code. Such files usually end with the
   .EXE extension. When the name of the file is entered at the system prompt,
   the instructions in the program are performed.

 External
   In FORTRAN, user-defined subroutines and functions are said to be external
   (as opposed to the intrinsic procedures that are part of the language).
   The compiler assumes all references to non-intrinsic procedures are
   external, and will be satisfied during linkage.

 External reference
   A variable or routine in a given module that is referred to by a routine
   in another module.

 Far call
   An address that specifies the segment as well as the offset.

 FL
   A command used by Microsoft FORTRAN to compile and link programs.

 Formal argument
   The name by which a specific argument is known within a function or
   subroutine.

 Hexadecimal
   Base-16 notation.

 High-order bit
   The highest-numbered bit; the bit farthest to the left. It is also called
   the most-significant bit.

 Huge model
   A memory model that allows for more than one segment of code and more than
   one segment of data, and that allows individual data items to span more
   than one segment.

 IEEE
   Institute of Electrical and Electronics Engineers, Inc.

 Implicit open
   The file opening performed by a read or write operation when the file was
   not explicitly opened by an OPEN statement.

 In-line code
   Code that is in the main program, as opposed to code that is in a
   subroutine called by the main program. Using in-line code is faster, but
   it makes programs larger.

 Input/output list (I/O)
   A list of items to input or output. PRINT, READ, or WRITE statements can
   specify an I/O list.

 Intrinsic
   A subroutine or function that is part of the FORTRAN language. The
   compiler knows which procedures are intrinsic, and assumes their
   references will be linked with code from the FORTRAN libraries.

 Keyword
   A word with a special, predefined meaning for the compiler.

 Large model
   A memory model that allows for more than one segment of code and more than
   one segment of data.

 Large-model compiler
   A compiler that assumes a program has more than one segment of code and
   more than one segment of data.

 Least-significant byte
   The lowest-numbered byte; the first byte. It is also called the low-order
   byte.

 Library
   A file that stores modules of compiled code. These modules are used by the
   linker to create executable program files.

 Link time
   The time during which the linker is executing, that is, linking
   relocatable object files and library files.

 Linking
   The process by which the linker loads modules into memory, computes
   addresses for routines and variables in relocatable modules, and then
   resolves all external references by searching the run-time library. After
   loading and linking, the linker saves the modules it has loaded into
   memory as a single executable file.

 Long call
   An address that specifies the segment as well as the offset. It is also
   referred to as the long address.

 Low-order bit
   The lowest-numbered bit; the bit farthest to the right. It is also called
   the least-significant bit.

 Machine code
   Instructions that a microprocessor can execute.

 Mantissa
   The decimal part of a base-10 logarithm.

 Map
   An area of memory in which one or more variables are contiguous.

 Medium model
   A memory model that allows for more than one segment of code and only one
   segment of data.

 Memory map
   A representation of where in memory the compiler expects to find certain
   types of information.

 Most-significant byte
   The highest-numbered byte; the last byte. It is also called the high-order
   byte.

 NAN
   An abbreviation that stands for "Not A Number." NANs are generated when
   the result of an operation cannot be represented in the IEEE format. For
   example, if you try to add two positive numbers whose sum is larger than
   the maximum value permitted by the compiler, the processor will return a
   NAN instead of the sum.

 Near call
   A call to a routine in the same segment. The address of the called routine
   is specified with an offset.

 Object file
   A file that contains relocatable machine code.

 Offset
   The number of bytes from the beginning of a segment to a particular byte
   in that segment.

 Optimize
   To reduce the size of the executable file by eliminating unnecessary
   instructions or to increase the execution speed by using more efficient
   algorithms.

 Pass
   To transfer data between subroutines, functions, subprograms, and the main
   program. Data passes either by reference or by value. Individual readings
   of source code made by the compiler as it processes information. Each
   reading is called a pass.

 PLOSS
   Appears in an error message when the error caused a partial loss of
   accuracy in the significant digits of the result. For example, a PLOSS
   error on a single-precision result indicates that less than six decimal
   digits of the result are reliable.

 Principal value
   The angular representation of a complex number that falls between -112 and
   +112 radians.

 Program unit
   A main program, a subroutine, a function, or a block-data subprogram.

 Record
   A variable with a structure data type

 Relocatable
   Not containing absolute addresses.

 Row-major order
   The order in which array elements are stored: the rightmost subscript is
   incremented first when the array is mapped into contiguous memory
   addresses.

 Run time
   The time during which an executable file is running.

 Run-time library
   A file containing the routines needed to implement certain functions of
   the Microsoft FORTRAN language. A library module usually corresponds to a
   feature or subfeature of the Microsoft FORTRAN language.

 Scratch file
   A file created to hold temporary data, then discarded.

 Segment
   An area of memory, less than or equal to 64K long, containing program code
   or data.

 Short call
   A call to a routine in the same segment. The address of the called routine
   is specified with only an offset. It is also referred to as the short
   address.

 Sign extended
   The sign bit of a number is propagated through all the higher-order bits.
   In this way, the sign is preserved when the number is written into a
   larger format.

 Single-precision real number
   A real number that is allocated 4 bytes of memory.

 Source file
   A file containing the original ASCII text of a Microsoft FORTRAN program.

 Stack
   A dynamically shrinking and expanding area of memory in which data items
   are stored in consecutive order and removed on a last-in, first-out basis.

 String
   A character constant.

 Structure
   A data type compounded of other data types.

 Terminal I/O
   Any I/O done to a terminal device. Examples of a terminal device are the
   console, keyboard, and printer.

 TLOSS
   Appears in an error message when the error caused a total loss of accuracy
   in the significant digits of the result. For example, a TLOSS error on a
   single-precision result indicates that none of the six significant digits
   of the result are reliable.

 Truncate
   To convert a real number to an integer by discarding the fractional part;
   no rounding occurs.

 Two's complement
   A type of base-2 notation in which 1s and 0s are reversed (complemented),
   and 1 is added to the result.

 Type coercion
   The forcing of a variable to have a particular data type. For example,
   when integer values are used in expressions, if one operand of an
   expression containing the operators plus (+), minus (-), or multiplication
   (*) is of type REAL, the other operand is converted to a real number
   before the operation is performed.

 Undefined variable
   A variable that cannot be found, either in the routine being linked or,
   for an external reference, in a routine in another module.

 Union
   An overlaying of two or more maps; the variables in each map share the
   same memory locations.

 Unresolved reference
   A reference to a variable or a subprogram that cannot be found, either in
   the routine being linked or, for an external reference, in a routine in
   another module.

 Visual page
   The area of memory whose contents currently form the graphics display.


 Index

 Symbols
 ───────────────────────────────────────────────────────────────────────────

 $ (dollar sign), in user-defined names
 ' (single right quotation mark)
 (:) colon
 * (asterisk)
    alternate return
    formal argument
    multiplication operator
    unit specifier
       inquiring
       opening
       writing to
    upper dimension bound
 ** (asterisks), exponentiation operator
 + (plus sign)
    addition operator
    carriage-control character
 , (comma)
    edit list
    field delimiter
 ... (dots)
 / (forward slash), FL option character
 / (forward slash), division operator
 //(concatenation operator)
 { } (braces)
 | (bar)
 [ ] (brackets)
 \ (backslash)
    see also Backslash (\) editing
    character
    edit descriptor
    editing
 _ (underscore)
    FORTRAN 4.0 names, used in
    names using C attribute
 __ (double underscore),names
    in
 _main
 ` (single left quotation mark)

 1, carriage-control character
 16-bit arithmetic
    $DEBUG
    INT2
 2-byte arithmetic, see 16-bit arithmetic
 32-bit arithmetic
    $DEBUG
    INT4
 4-byte arithmetic, see 32-bit arithmetic
 /4I2 and /4I4 options (FL)
 /4Y6 and /4N6 options (FL)
 /4Yd and /4Nd options (FL)
 /4Yf and /4Nf options (FL)
 /4Yt and /4Nt options (FL)
 5 (unit specifier)
 6 (unit specifier)
 80186/80188 processor
 80286 processor
 8087/80287 coprocessor, defined

 A
 ───────────────────────────────────────────────────────────────────────────
 A editing
 Abbreviations, in intrinsic function tables
 ABS intrinsic function
 Absolute value
 Access
    described
    direct
    files, when networking
    internal file
    sequential
 ACCESS= option
 ACOS
 Action, metacommand heading
 Addresses
    common blocks
    even
    intrinsic function
    long
    odd
    offset
    segmented
    short
 Adjustable-size array
    defined
    passing by value
 /AH option (FL)
 Agreement, see Checking arguments
 AIMAG
 AINT
 /AL option (FL)
 Algorithms
    AIMAG
    AMOD
    ANINT
    CONJG
    DDIM
    DIM
    DMOD
    DNINT
    IDIM
    IDNINT
    MOD
    NINT
 ALIAS
 ALOG
 ALOG10
 Alphabetic characters
    character set
    names
 Alphanumeric
    characters, names
    defined
 Alternate return
    actual argument
    described
    formal argument
    function
    specifier
 /AM option (FL)
 AMAX0
 AMAX1
 American Standard Code for Information Interchange, see ASCII
 AMIN0
 AMIN1
 AMOD
 .AND. operator
 Angle, in trigonometric intrinsic function
 ANINT
 ANSI standard
    extensions
    identified
    $STRICT
    variables, size
 Apostrophe (')
    character string
    described
    editing
 Arc cosine intrinsic function
 Arc sine intrinsic function
 Arc tangent intrinsic function
 Arguments
    actual
       alternate-return specifier
       array
       array element
       associated
       corresponding formal argument
       default data segment
       defined
       EXTERNAL
       expression
       FAR
       function
       INTRINSIC
       multiple segments
       NEAR
       number
       subroutine
       usage
       variable
    agreement of data types
    CHAR
    checking
       integers
       INTERFACE, use of
       logical
       subroutine
    data type, subroutine
    data-type conversion
    defined
    different number of formal and actual
    FL options
    formal
       alternate return
       array
       assigning a value
       associated
       asterisk (*)
       C attribute
       corresponding actual argument (list)
       defined
       different number than actual arguments
       EXTERN
       EXTERNAL
       FAR
       function
       HUGE
       intrinsic function
       $LARGE
       number
       subroutine
       usage
       variable
    function
    ICHAR
    integer
       checking
       passing
    intrinsic function
       data type
       described
       LEN
       logarithm
       out of range
       square root
       undefined
    listing options (FL)
    name
    number
    spanning more than one segment
    statement function
    usage
    value, data-type conversion
 Arithmetic IF, described
 Arithmetic assignment statement, described
 Arithmetic expressions
 Arithmetic operand
    data-type conversion
    list
    rank
    type conversion
 Arithmetic shift intrinsic function
 Arithmetic
    16-bit (short)
       $DEBUG
       INT2
    32-bit (long)
       $DEBUG
       INT4
    high-precision, $DEBUG
    speed
    testing
 IArray bounds
 Array declarator
 Array element
    actual argument
    character
    expression, used in
    local
    referencing
    storage order
    syntax
    undefined
 Arrays
    actual argument
    adjustable size
    assumed size
    character
    default data type
    described
    DIMENSION statement
    dimensions
    EQUIVALENCE statement
    formal argument
    HUGE
    $LARGE
    names
    number of dimensions
    passing by value
    size
    storage order
 ASCII
    character set
    character values
    characters, representing
    collating sequence
       character set
       data-type conversion
       intrinsic functions
    values, input/output
 ASIN
 Assembly language
    accessing
    performance
 Assembly-listing files
    creating
    extensions
    format
 ASSIGN
    description
    format specifiers
    INTEGER*1 variables
 Assigned GOTO
 Assignment compatibility, statement functions
 Assignment statements
 Assignment, checking range
 Associated, defined
 Association
    address
    arguments, actual and formal
    common block
 Associativity, see Precedence
 Assumed-size array
    defined
    passing by value
 Asterisk (*)
    alternate return
    formal argument
    format specifier
    length specifier
    multiplication operator
    output
    unit specifier
       closing
       described
       inquiring
       opening
       writing to
    upper dimension bound
 Asterisks (**), exponentiation operator
 ATAN
 ATAN2
 Attributes
    ALIAS
    C
    defined
    described
    EXTERN
    FAR
    HUGE
    interface, used in
    NEAR
    PASCAL
    REFERENCE
    syntax
    (table)
    VALUE
    VARYING
 AUX

 B
 ───────────────────────────────────────────────────────────────────────────
 Backslash (\)
    character
    edit descriptor
    editing
 BACKSPACE
 Backspace character
 Bar (|)
 Base
 Base 10
    constants
    logarithm
 Base name, defined
 Batch files, FL command, converting for
 \bc*\ec (asterisk)
    unit specifier
       closing
 \bc196\ec (dash), FL option character
 \bc196\ec (minus sign), subtraction operator
 Bell character
 Big programs
 Binary files
    described
    reading
    record boundary
 Binary operators
    arithmetic
    logical
    relational
 Binary
    defined
    interface
 Bit change intrinsic function
 Bit clear intrinsic function
 Bit manipulation intrinsic function
 Bit set intrinsic function
 Bit test intrinsic function
 Bits
 Blank
    carriage-control character
    character constants
    column 6
    common block, initializing
    file name
    Hollerith fields
    input/output
    interpretation
    list-directed input
    names
    significance
 BLKDQQ
 BLOCK DATA
    also see Block-data subprogram
 Block ELSE IF, see ELSE IF block
 Block ELSE, see ELSE block
 Block IF, see IF block
 Block-data subprogram
     also see BLOCK DATA
 Block-data subprogram
    contents
    DATA statement
    described
    identifying
    named common blocks
    statements allowed
    summary
    unnamed
 BLOCKSIZE= option
 Blue type
 BN edit descriptor
 Bound, see Dimension bound
 Bounds, character substring
 Bounds, subscript, see Subscript
 Braces ({ })
 Brackets ([ ])
 BTEST
 Bytes
    also see Length
 BZ edit descriptor

 C
 ───────────────────────────────────────────────────────────────────────────
 C attribute
 C language
    also see Microsoft C
 C string
    defined
    described
    escape sequence, table
    nonprintable character
    null string
 /c option (FL)
 CABS
 Call
    long, defined
    short, defined
 CALL
    described
    DO loop, used with
    subroutine
 Calling conventions
 Calling subroutine
    CALL
    recursion
 Capital letter
    also see Case sensitivity
    notation
    small
 Card
 Carriage control
 Carriage-return character
 Case sensitivity
    ALIAS
    character constants
    described
    external names
    Hollerith fields
    keywords
 CCOS
 CDABS
 CDCOS
 CDEXP
 CDLOG
 CDSIN
 CDSQRT
 CEXP
 char
 CHAR
 CHARACTER
 Character array element
 Character assignment statement
 Character constants
    also see Strings
 Character data type
    common block
    list-directed input
    list-directed output
 Character editing
 Character expressions
 Character functions
 Character operand
 Character operator
 Character substring
    bounds
    checking
    described
    internal file
    length
    syntax
 Character variable
    common block
    internal file
    length
 Characters
    alphabetic
    ASCII
       using
    bell
    blank
    carriage control
    carriage return
    conversion to
    default length
    described
    digits
    form feed
    function
    hexadecimal bit pattern
    horizontal tab
    intrinsic function
    (list)
    lowercase
    names
    new line
    nonprintable
    octal bit pattern
    printable
    tab
    uppercase
    vertical tab
 Checking
    arguments
       described
       integers
       INTERFACE
       logical
       subroutines
    ranges
 CLOG
 CLOSE
    asterisk (*) unit
    described
    discarding files
    disconnecting units
    units 0, 5, 6, *
 Closing
    files
    units 0, 5, 6, *
 cmp
 cmp16
 cmp8
 CMPLX
 Code
    in-line
    machine
    size, optimizing
    source
 Coercion, see Conversion
 Collating sequence
    character set
    data-type conversion
    intrinsic functions
 Colon(:), nonrepeatable edit descriptor
 Column, statement
 Column-major order
 Comma (,)
    edit list
    field delimiter
 Command line
    error messages
    file names, entering
    FL
    switches
 Commands
    FL, defined
    MSF
    operating system
 Comment line
    described
    free form
    order
 COMMON
    also see Common block
 Common block
    blank, initializing
    character data type
    character variable
    COMMON
    EQUIVALENCE statement
    external name
    name
    named
       block-data subprogram
       DATA statement
       initializing
       length
    NEAR
 COMMQQ
 'COMPAT'
 Compatibility with old libraries
 Compatible data type
 Compiland, defined
 Compilation error messages
 Compilation unit
 Compilation, conditional
 Compile time, defined
 Compiler
    defined
    error messages
       categories
       compilation
       fatal
       identifying
       recoverable
       redirecting
       warning
    exit codes
    large model, defined
    limits
    options. See FL options
 Complement, logical
 Complex
    absolute value
    conjugate
    converting to
    described
    intrinsic function
       (table)
       described
       result
       square root
    number
    number, defined
    relational expression
    syntax
 COMPLEX*16
 COMPLEX*8
 Computed GOTO
 CON
 Concatenation operator (//)
 Conditional compilation
 CONJG
 Conjugate, complex
 Conjunction operator
 Consecutive operator
    arithmetic
    logical
 Constant
    base 10
    folding
    hexadecimal
    Hollerith
    integer
    naming
    specifying base
    specifying radix
 Continuation line
 CONTINUE
 Control statement (table)
 Conversion
    arithmetic operand
    character
    complex
    data type
       assignment
       DATA statement
       intrinsic functions
       subprogram argument
       (table)
       value arguments
    integer
    intrinsic functions
    real
 Coprocessor, defined
 Correctable error messages
 Correspondence, see Checking arguments
 COS
 COSH
 Cosine
   see arc, intrinsic function
    hyperbolic, intrinsic function
    intrinsic function
 COTAN
 Cotangent intrinsic function
 Count, iteration
 CSIN
 CSQRT

 D
 ───────────────────────────────────────────────────────────────────────────
 D editing
 D, real exponent
 DABS
 DACOS
 DASIN
 DATA
    also see Input/output
    described
    editing
    formatting
    reading
    writing
 Data segment, default
    LOCNEAR
    NEAR
 Data types
    abbreviation (table)
    arguments
       agreement
       subroutine
    arithmetic operand
    array
    character
       common block
       described
       list-directed input, output
       specifying
    compatible
    complex
    conversion
       arithmetic operand
       CHAR
       CMPLX
       DATA statement
       DBLE
       ICHAR
       intrinsic functions
       REAL
       subprogram argument
       value arguments
    declaring
    default
    expression
    func