Microsoft KB Archive/73256

= FORTRAN 5.1 README.DOC: Additions to Reference Manual =

Article ID: 73256

Article Last Modified on 10/23/2003

-

APPLIES TO


 * Microsoft FORTRAN Compiler 5.1
 * Microsoft FORTRAN Compiler 5.1

-



This article was previously published under Q73256



SUMMARY
The following is the &quot;Additions to Microsoft Reference Manual&quot; section from the FORTRAN version 5.1 README.DOC:



MORE INFORMATION
 README.DOC File Release notes for the Microsoft(R) FORTRAN Professional Development System, Version 5.1

This document contains release notes for version 5.1 of the Microsoft FORTRAN Professional Development System for MS-DOS(R) and the Microsoft Operating System/2 (MS(R) OS/2). The information in this document is more up-to-date than that in the manuals and the Microsoft Advisor (online help).

Microsoft improves its languages documentation at the time of reprinting, so some of the information in this file may already be in your manuals.

=
< Additions to the Microsoft FORTRAN Reference >============

/Gt option --

The /Gt, or data threshold, compiler option can remedy a number of link-time and run-time errors in FORTRAN 5.1. The /Gt option sets a threshold value that determines where data items are stored in memory. Data items that are greater in size than the threshold value, and that would normally be stored in the default data segment, or DGROUP, are moved to a far segment of data. This frees up more room in DGROUP for variables less than the threshold value and for other uses, such as the program STACK and NEAR HEAP.

Using the /Gt switch can often help eliminate the LINK errors:

L1070: 'segment name':Segment exceeds 64K

where segment can be _BSS or _DATA

L2041: Stack Plus Data exceeds 64K

and the run-time error:

F6700: Heap Space Limit Exceeded

In each of these cases, too much data is being stored in DGROUP, and using the /Gt switch may free up the required space. See Section 7.6 of the FORTRAN 5.1 Reference for more information.

/Oc Option --

When you use the /Oc (default common subexpression optimization), the compiler examines only short sections of code for common subexpressions. You can disable default common subexpression optimization with the /Od option.

Special Device Names

The list of special device names in the FORTRAN Reference Manual is incomplete. The complete list of special device names is:

Device Name             Comments ===========             ========     CON                      stdout, stdin USER                    stdout, stdin COM1                    COM1 port AUX                     COM1 alias LINE                    COM1 alias LPT1                    stdprn PRN                     stdprn LPT2 LPT3 LPT4 ERR                     stderr NUL                     null device

Note that COM2, COM3, and COM4 are not special device names in FORTRAN 5.1.

To access these devices, use an OPEN statement of the form:

OPEN (10, FILE='lpt1')

For connections to LPT2 or COM2, you are also required to specify STATUS='old'. These two devices are not treated as standard devices like LPT1 and COM1. Output to these devices is of the same form as to a file, i.e. the output is buffered and carriage control characters in column 1 are ignored.

are treated as special device names by FORTRAN 5.10 and do not require STATUS='OLD', nor do they ignore carriage control. The above information still applies to COM2-4, however.
 * NOTE ** The above section of the README.DOC is incorrect. LPT2-4

/ND Option --

When using the /ND option in a module, you must also set the loadds attribute on all subprograms called from outside the module.

Restrictions on String Assignment -

The right and left sides of a string assignment may not reference the same memory location. For example, the following may fail because the generated code overlays the input as it produces the output:

character *10 string/'0123456789' string = '#' // string write (*,*) 'string = ' string end

IBM VS Extensions in Microsoft FORTRAN -

Microsoft FORTRAN supports only the following IBM VS FORTRAN extensions:

- 31-character names - Allowing integer arrays to contain FORMAT specifications - DO statements without specified labels - DO WHILE statement - END DO statement - Equivalence of character and non character items - Mixing of character and non-character items in COMMON - NAMELIST - Specification of common block items in DATA outside block data subprograms - Initialization on the declaration line - IMPLICIT NONE - INCLUDE - Length specification within declarations - Noninteger array subscripts - Z edit descriptor - ! comments - $ and _ in identifiers

$NOTRUNCATE Metacommand ---

$NOTRUNCATE is now the default. As a result, it is possible to create variable names, common block names, and function or subroutine names that are longer than the 26-character space allotted in a .LST file. If these identifiers do not differ in the first 26 characters, the .LST file shows them with the same name.

New NINT Functions --

Three NINT intrinsic functions have been added: NINT1, NINT2, and NINT4. These return integer values of one, two, and four bytes respectively.

The /4I2 Switch and the INT and NINT Functions --

The length of the value returned by the INT and NINT intrinsic functions is determined by the integer storage switch: two bytes if /4I2 is used, four bytes if /4I4 (the default) is used.

If you want these functions to return a value with a length different from the length set by the integer storage switch, use INT1, INT2, or INT4 instead of INT, and use NINT1, NINT2, or NINT4 instead of NINT.

80387 Coprocessor Instructions --

The run-time libraries recognize the presence of an 80387 coprocessor. If you are not using the alternate math package, the library uses the 80387 SIN and COS instructions, rather than deriving sines and cosines from the FPTAN instruction.

Array Conformity in Functions and Subroutines -

A formal array argument in a subroutine or function does not have to have the same number of dimensions or total size as the actual array argument passed to it. This permits the procedure to contain a formal array that is as large as you think you will ever need; the calling program can then pass any array of the same size or smaller.

Note, however, that if the formal and actual dimensions do not match in quantity and size, the formal array elements do not point to the same memory locations as the actual array elements with the same subscripts. In this case, your program has to calculate the correct subscripts.

&quot;Out of Memory&quot; Error Message -

This compiler error does not have an error number. It occurs when the FL program does not have sufficient memory, as in the following cases:

1. You used wildcard characters to specify the files to be compiled (for example, FL *.FOR) and there were too many files. The exact number of files that can be handled depends on the amount of system memory.

2. The computer did not have enough free memory. This may occur when other processes are running under Microsoft Windows or OS/2.

Additional Compiler Error Messages --

F2399: : COMMON (or EQUIVALENCE) : automatic variable illegal.

A variable previously declared as AUTOMATIC cannot appear in a COMMON block or an EQUIVALENCE statement.

F4187: Debugging information exceeds 64K - extra ignored.

There was more debugging information than could fit in the allotted 64K space. The additional information (which is usually symbols) was discarded.

Revised Compiler Error Message --

The in error message F2384 can be &quot;common block&quot; or &quot;equivalenced,&quot; in addition to ALLOCATABLE, EXTERN, FAR, or HUGE.

Linker Path Specification -

The /link option does not allow you to specify the linker's path. FL assumes the linker is in the same directory as the compiler.

FL Switch Default Values

If no switches appear in the FL command line, the source code is compiled with the following defaults:

1. The Large memory model is used 2. The 8086/8088 instruction set is used 3. 8087/287/387 in-line instructions are used for floating-point calculations 4. Variable, subroutine, or function names are not truncated 5. All integer and single-precision floating-point numbers are four bytes 6. All Microsoft extensions to FORTRAN are enabled 7. FORTRAN 66-style DO statements are disabled 8. Full optimization is performed 9. Debugging (extended run-time error handling) is disabled

Any metacommands in a source code file override these defaults.

Compiling Extremely Large Programs --

This release contains the file F1L.EXE, which is an alternate form of the Compiler Pass 1 disk if you need it.

This compiler pass is used to compile programs that receive error message &quot;F1901: program too large for memory.&quot; The F1L.EXE pass is invoked by adding the /B1 option to the FL command line, as follows:

FL /B1  F1L.EXE  .FOR

where is the path (including drive and directory) where F1L.EXE resides, and is the name of the FORTRAN source file you are compiling. Alternatively, you can rename F1L.EXE to F1.EXE to make the high-capacity version the default.

Compiler Errors and Warnings Generated by $DEBUG and /4Yb -

In addition to the expanded run-time error handling and testing that $DEBUG and /4Yb enable, there are a number of compiler errors and warnings that appear if $DEBUG or /4Yb are used. They are:

Error F2367: INTEGER : range error Error F2533: lower substring bound exceeds upper bound Error F2534: upper substring bound exceeds string length Error F2535: lower substring bound not positive Error F2536: subscript out of range Warning F4501: subscript out of range (array argument)

Additional query words: 5.10

Keywords: KB73256

-

[mailto:TECHNET@MICROSOFT.COM Send feedback to Microsoft]

© Microsoft Corporation. All rights reserved.