This document contains information about Compaq Fortran for Compaq® Tru64^tm UNIX® (formerly DIGITAL® UNIX) Alpha Systems.

Software Version: Compaq Fortran
Version 5.4A

Compaq Computer Corporation
Houston, Texas

Compaq Computer Corporation makes no representations that the use of its products in the manner described in this publication will not infringe on existing or future patent rights, nor do the descriptions contained in this publication imply the granting of licenses to make, use, or sell equipment or software in accordance with the description.

Possession, use, or copying of the software described in this publication is authorized only pursuant to a valid written license from Compaq or an authorized sublicensor.

Compaq shall not be liable for technical or editorial errors or omissions contained herein. The information in this document is subject to change without notice.

© 2001 Compaq Computer Corporation. All Rights Reserved.

COMPAQ, the Compaq Logo, DEC, DECnet, DIGITAL, VAX, and VMS are registered in the U.S. Patent and Trademark Office.

AlphaGeneration, AlphaServer, Bookreader, DECchip, DEC Fortran, DECwindows, DEC FUSE, DECwindows, OpenVMS, Tru64, ULTRIX, and VAX FORTRAN are trademarks of Compaq Computer Corporation.

Microsoft, PowerStation, NT, and Windows are either trademarks or registered trademarks of Microsoft Corporation.

CRAY is a trademark of Cray Research, Inc.
IEEE is a registered trademark of the Institute of Electrical and Electronics Engineers, Inc.
Linux is a trademark of Linus Torvalds.
OpenMP is a trademark of the OpenMP Architecture Review Board.
PostScript is a trademark of Adobe Systems, Inc.
Motif, OSF, and OSF/1 are trademarks of the Open Software Foundation, Inc.
UNIX is a trademark in the United States and other countries of The Open Group.

All other trademarks and registered trademarks are the property of their respective holders.

This document is available on CD-ROM.

Contents

This chapter contains release notes for Compaq Fortran Version 5.4A (changes for this release are described in Section 1.5).

This chapter contains the following sections:

• Section 1.1 (Product Terminology Changes)
• Section 1.2 (Overview)
• Section 1.3 (Installation and Minimum Operating Systems Version)
• Section 1.4 (Software Versions)
• Section 1.5 (New Features, Corrections, and Known Problems in Version 5.4A)
• Section 1.6 (Known Limitations)
• Section 1.7 (New Features, Corrections, and Known Problems in Version 5.3)
• Section 1.8 (New Features, Corrections, and Known Problems in Version 5.2)
• Section 1.9 (High Performance Fortran (HPF) Support in Version 5.2)
• Section 1.10 (New Features and Corrections in Version 5.1)
• Section 1.11 (New Features and Corrections in Version 5.0)
• Section 1.12 (Additional Information)

This document uses the following new or changed product names:

• "Compaq Fortran" was previously called "DIGITAL Fortran 90".
• "Compaq Fortran 77" was previously called "DIGITAL Fortran 77".
• "Compaq Fortran" refers to the combined packaging of the Compaq Fortran 90 and Compaq Fortran 77 products. It also refers to the Compaq Fortran language.
• Compaq Fortran 90 and Compaq Fortran 77 product version numbers are now the same.
• The operating system formerly known as "DEC OSF/1" and "DIGITAL UNIX" is now called "Compaq Tru64 UNIX".
• The Compaq Extended Math Libraries {CXML} was previously called the DIGITAL Extended Math Libraries {DXML}.

Compaq Fortran conforms to the Fortran 95 Standard, Fortran 90 Standard, and previous Fortran standards. It also includes support for High Performance Fortran (HPF), and contains many but not all of Compaq Fortran 77's extensions to the FORTRAN-77 standard. Except in rare instances, a valid FORTRAN-77 program is also a valid Fortran 95 program.

Compaq Fortran fully supports the Fortran 95 Standard (ISO/IEC 1539-1:1997(E)) and the multi-vendor OpenMP^tm Fortran Specification, including support for directed parallel processing using OpenMP directives on shared memory multiprocessor systems.

The Compaq Fortran User Manual for Tru64 UNIX and Linux Alpha Systems contains a detailed description of Compaq Fortran 77 source language compatibility. Provided the types and shapes of the actual and formal arguments agree, routines compiled with Compaq Fortran 77 can call (or be called by) routines compiled with Compaq Fortran.

1.3 Installation and Minimum Operating Systems Version

Compaq Fortran Version 5.4A requires Version 4.0 (or later) of the Compaq Tru64 UNIX operating system. Use of non-NUMA parallel features requires Version 4.0D (or later) of the Compaq Tru64 UNIX system. Use of NUMA (Non-Uniform Memory Access) parallel features requires Version 5.1 (or later) of the Compaq Tru64 UNIX operating system.

For a detailed description of the installation procedure, see the Compaq Fortran Installation Guide for Tru64 UNIX Systems.

You may also send comments, questions and suggestions about the Compaq Fortran product to the following mail address:

fortran@compaq.com

Please note that this address is for informational inquiries and is not a formal support channel.

If you have Internet access and a World Wide Web (WWW) browser, you are welcome to view the Compaq Fortran home page, located at:

http://www.compaq.com/fortran

This section describes the contents of Compaq Fortran Version 5.4.

1.4.1 Contents of the Version 5.4A kit

The Compaq Fortran Version 5.4A kit consists of the following setld sets:

• DFABASE541---Compaq Fortran 90 and 77 v5.4A for Compaq Tru64 UNIX Alpha Systems
• DFADOC541 ---Compaq Fortran v5.4A Release Notes and Man Page
• DFACOM541 ---Compaq Fortran v5.4A Tools & their Man Pages
• DFARTL394 ---Compaq Fortran RTL #394 for Compaq Tru64 UNIX Alpha Systems
• OTABASE212---Compaq Compiled Code Support Library #212
• HPFLIBS185---Compaq Fortran V5.4 High Performance Fortran Runtime Libraries
• XMDCOM500 ---the CXML common subset files
• XMDLIB4500---the CXML archive and shared libraries (serial and parallel) for EV4 systems
• XMDLIB5500---the CXML archive and shared libraries (serial and parallel) for EV5 systems
• XMDLIB6500---the CXML archive and shared libraries (serial and parallel) for EV6 systems
• XMDSCI500 ---the SCIPORT Cray compatibility library and manpages
• XMDMAN500 ---the CXML manpages
• XMDHTM500 ---the CXML manpages in HTML format

1.4.2 Contents of High Performance Fortran in Version 5.4A

The following new Compaq Fortran features are now supported for HPF programs:

• HPF Programs with MPI
  HPF programs can execute with the support of Message Passing Interface (MPI) software. Before Version 5.3 ECO 01 of Compaq Fortran, HPF programs executed with the support of the HPF component of Version 1.6 of the Parallel Software Environment (PSE).
  The HPF/MPI connection enables development of programs that use both HPF and MPI together in the same application. In addition, MPI supports additional hardware configurations, some of which are faster than any supported by PSE. Finally, there are many tools available for debugging, profiling, and visualizing the execution of MPI programs.
  The Parallel Software Environment (PSE) runtime system is no longer available. Consequently, all HPF programs must now be run with the new HPF runtime system that uses MPI, instead of PSE, to perform its message passing. Section 1.7.3 (Version 5.3 ECO 01 HPF New Features) explains how to use the new runtime system.

The HPFLIBS subset contains High Performance Fortran (HPF) support and contains:

• Both the nonparallel (serial) and parallel versions of the High Performance Fortran (HPF) run-time library
• Associated reference pages for HPF_LIBRARY and HPF_LOCAL_LIBRARY, including intro(3hpf)

In order to use the HPF Scalar libraries, the HPFLIBS subset must be installed.

1.4.3 Miscellaneous Information about the Version 5.4A Kit

If you need to delete an older version of the Compaq Fortran Run-Time Library, delete the older subset before you install a new version. If you have installed multiple versions of the Run-Time Library and you delete one, you must reinstall the desired Run-Time Library version before you can correctly link Fortran programs.

The Compaq Fortran kit again includes the OTABASE subset (Compaq Compiled Code Support Library). The new OTABASE subset supports the use of parallel directives.

The following changes occurred to the OSFCMPLRS operating system subset starting with Compaq Tru64 UNIX (formerly DIGITAL UNIX) Version 4.0:

• Beginning with DIGITAL UNIX Version 4.0, the OSFCMPLRS subset now consists of multiple subsets: OSFCMPLRS (required for program development), OSFSDE (profiling tools), OSFLIBA (archive libraries), OSFINCLUDE (include files), OSFATOM ( atom tools).

1.5 New Features, Corrections, and Known Problems in Version 5.4A

For the Version 5.4A subset numbers, see Section 1.4.

The following topics are discussed:

• Section 1.5.1 (Version 5.4A New Features)
• Section 1.5.2 (Version 5.4 New Features)
• Section 1.5.3 (Version 5.4 Important Information)
• Section 1.5.4 (Version 5.4 Corrections)
• Section 1.5.5 (Version 5.4 Known Problems )

1.5.1 Version 5.4A New Features

The following new Compaq Fortran features are now supported:

• The following new features are now supported:
  • -pipeline is now the default at optimization level -o4 (the default).
  • Nested parallel regions are now supported by -omp .
  • Parallel programs running on Non-Uniform Memory Access (NUMA) machines have initial support using the following directives, compiler options, and environment variables:
    • !DEC$ MIGRATE_NEXT_TOUCH
    • !DEC$ MIGRATE_NEXT_TOUCH_NOPRESERVE
    • !DEC$ OMP NUMA
    • !DEC$ DISTRIBUTE - same as !HPF$ DISTRIBUTE
    • !DEC$ ALIGN - same as !HPF$ ALIGN
    
    

    -numa Fortran 90 only. Enables NUMA parallel processing and the NUMA command line options -numa_memories and -numa_tpm. NUMA paral- lel processing is indicated by inserting certain !DEC$ direc- tives in your source code. Option -omp must also be specified. The default is -nonuma. -numa_memories num Fortran 90 only. Specifies the number of memories (RADs) to be used for NUMA parallel processing. Option -numa must also be specified. If -numa_memories does not appear on the command line or if num is 0, the number of memories will be chosen at run-time either from the NUMA_MEMORIES environment variable (if it is set) or by the system. -numa_tpm num Fortran 90 only. Specifies the number of threads per memory to be used for NUMA parallel processing. Option -numa must also be specified. num is the number of threads per physical memory that will execute NUMA parallel features in the program. If -numa_tpm does not appear on the command line or if num is 0, the number of threads per memory will be chosen at run-time either from the NUMA_TPM environment variable (if it is set) or by the system.

• The following new f90 command options are now supported. See the f90(1) man page for details.
  • -arch ev67
  • -ccdefault fortran | list | none | default
  • -annotations
  • -assume noprotect_constants
  • -check arg_temp_created
• The following new run-time features are now supported:
  • The Fortran RTL has been changed so that it now processes the [.m] (the minimum number of digits) portion of the edit descriptor when w (the width field) is Zero for I, B, O, and Z editing.
  • An end-of-file or end-of-record status is no longer treated as an error status. This change was done to adhere to the Fortran 90 language standard. Prior to this change, if there was an ERR= specified and no END=|EOR= for a READ operation and the READ encountered an end-of-file or end-of-record situation, the ERR= path would be followed. With this change, this situation would now result in a fatal message being generated.
  • The runtime library now contains support for a new environment variable, FORT_CONVERT_ext that allows a user to associate a foreign data conversion option with files of a particular file extension. The values of FORT_CONVERT_ext are the same as FOR_CONVERTn.
  • The runtime library has been changed to perform more thorough edit checking on list directed input. Previously, the RTL was liberal in what it accepted for input to integer and real values. In accordance with the F95 Standard, the RTL no longer accepts "+", "-", ".", "D", "E", or "Q" without expressing at least 1 digit. For example, the RTL used to allow a single "+" to convert to a 0, but now the RTL will return a FOR$IOS_LISIO_SYN error. In addition, ambiguous expressions such as "+-" and "--" will be rejected.
  • Support for the -fpscomp options are now implemented in the Fortran RTL on UNIX:
    • -fpscomp all
    • -fpscomp general
    • -fpscomp ioformat
    • -fpscomp logicals
    • -fpscomp filesfromcmd
    
    This support includes the six Microsoft PowerStation compatible file types as described in the Visual Fortran Programmer's Guide:
    • MS Unformatted Sequential Access
    • MS Unformatted Direct Access
    • MS Formatted Sequential Access
    • MS Formatted Direct Access
    • MS Binary Sequential Access
    • MS Binary Direct Access
    
    This includes all fpscomp behavior that does not involve Windows-specific features. There is no QuickWin-like support. The 'Filenames from Command Line' option looks in the command line arguments for unspecified filenames in an OPEN(...,FILE='',...) statement and will also prompt for filenames at the terminal, but does not implement the equivalent of a Windows Dialog box, like a QuickWin application would in the PC Windows environment. Windows-specific physical device names used as filenames are given no special consideration or handling.
  • Support for reading nondelimited character strings as input for character NAMELIST items has been added.
    A character string does not need delimiting apostrophes or quotation marks if the corresponding NAMELIST item is of type default character, and the following is true:
    • The character string does not contain a blank, comma, slash, exclamation point (!), ampersand (&), dollar sign ($), left parenthesis, equal sign (=), percent sign (%), or period (.).
    • The character string is not continued across a record boundary.
    • The first nonblank character in the string is not an apostrophe or a quotation mark.
    • The leading character is not a string of digits followed by an asterisk.
    
    A nondelimited character string is terminated by the first blank, comma, slash, end-of-record, exclamation point, ampersand, or dollar sign encountered. Apostrophes and quotation marks within nondelimited character strings are transferred as is.
    Should an equal sign, percent sign, or period be encountered while scanning for a nondelimited character string, the string will be treated as a variable name (or part thereof) and not as a nondelimited character string.
    Be forewarned that nondelimited character strings that are written out by using a NAMELIST write may not be read in as expected by a corresponding NAMELIST read.
    Given the following example code:

    NAMELIST/TEST/ CHARR CHARACTER*3 CHARR(4) DATA CHARR/'AAA', 'BBB', 'CCC', 'DDD'/ OPEN (UNIT=1, FILE='NMLTEST.DAT') WRITE (1, NML=TEST) END

    
    The output file NMLTEST.DAT will contain:

    &TEST CHARR = AAABBBCCCDDD /

    
    Should an attempt be made to read the data in NMLTEST.DAT back in with a NAMELIST read using nondelimited character strings:

    NAMELIST/TEST/ CHARR CHARACTER*3 CHARR(4) DATA CHARR/4*' '/ OPEN (UNIT=1, FILE='NMLTEST.DAT') READ (1, NML=TEST) PRINT *, 'CHARR read in >', CHARR(1),'< >',CHARR(2),'< >', 1 CHARR(3), '< >', CHARR(4), '<' END

    
    This will result in:

    CHARR read in >AAA< > < > < > <

  • New run-time jacket routines provided for Fortran 3f lgamma, erf, erfc, short and long functions in libUfor. See the appropriate 3f man page for details.
  • The run-time now supports unlimited record sizes for writes, reads, backspaces of variable length unformatted files. Records greater than 2.1 giga-bytes use a new scheme that may not be portable to another vendor's Fortran.
  • When the DATE, TIME, or ZONE arguments to the DATE_AND_TIME intrinsic routine are not large enough to hold the required information, a fatal run-time error will now be generated.
  • The run-time support was enhanced to allow a REWIND operation to be performed on a direct access file. This is allowed without having to specify any command line options.
  • New bits were defined in the for_set_fpe routine interface to allow the compiler to specify counting and messages for floating point inexact traps (fpe:6). The runtime handler is updated to provide this service.
  • New run-time jacket routines provided for Fortran 3f lgamma, erf, erfc, short and long functions in libUfor. New entry point interfaces are:

    float lgamma_ ( float *x ) ; double dlgamma_ ( double *x ) ; float erf_ ( float *x ) ; double derf_ ( double *x ) ; float erfc_ ( float *x ) ; double derfc_ ( double *x ) ; short short_ ( int *x ) ; int long_ ( short *x ) ;

  • Support for generating traceback information has been added to the run-time support. Traceback will be produced for all severe errors. Traceback output can be disabled by setting the environment variable FOR_DISABLE_STACK_TRACE to true. All diagnostic output, which includes traceback, can be redirected to a file defined by the FORT0 environment variable, which was already supported.
  • Namelist input was not handling slices and strides of arrays, array segments with zero and negative positions, and character substrings of arrays. Note: this implementation adheres to the F90 Standard, so nested array slices are illegal.
  • Support for the environment variable FORT_BUFFERED was added. When it is set to TRUE, the run-time library will assume that buffered I/O will be used for output to all I/O units, except those whose output is to the terminal. This provides a run-time mechanism to support the behavior enabled by the -assume buffered_io option.
• See the new features listed in Section 1.7.2 (Version 5.3 ECO 01 New Features)
• The Compaq Extended Math Library (CXML) routines are updated in the Compaq Fortran kit. See the CXML release notes in:
  /usr/opt/XMDCOM410/docs/XMD410_release_note.txt

Some important information to note about this release:

• As of Compaq Fortran V5.3, the f77 command executes the Compaq Fortran 90 compiler instead of the Compaq Fortran 77 compiler. Use f77 -old_f77 to execute the Compaq Fortran 77 compiler.
• Object files created by Fortran 95 contain information about when they were compiled and by what version of the compiler. Use the command

  strings -a xxx.o | grep "@(.)"

  
  where xxx.o is the name of the object file. The strings generated by the compiler are

  • "(#)tttt" is a what(1) string with the text from a cDEC$ IDENT directive if present.
  • "@(c)Compaq Fortran Vn.n-eeee" is the version number of the Fortran 95 compiler that generated this object file.
  • "@(m)xxx" is the name of the first program unit in the source used to produce this object file.
  • "@(d)dd-mmm-yyyy hh:mm:tt" is the date and time this object file was created.

From DFA531 ECO 01 final X5.3-1120 -44A7B to DFA540 FT1 T5.4-1130-46A7R, the following corrections have been made:

• Eliminate internal compiler error for character function whose length expression includes LEN_TRIM of a dummy argument.
• Eliminate internal compiler error for SIZE of a derived type component pointer array.
• Software pipelining is now enabled at the default optimization level (-O4)
• Eliminate more unnecessary copies of contiguous array arguments.
• Correct parsing error for certain record component references using dots.
• Eliminate internal compiler error for particular use of RESHAPE.
• Do not give an unused variable warning where the only reference to a variable is as the argument to LOC.
• Eliminate internal compiler error for particular use of nested STRUCTURE.
• Disallow illegal ALLOCATED(A(:))
• Allow, as an extension for compatibility with our Fortran 77 compilers, a LOGICAL value as an argument to intrinsics which accept INTEGER, such as ABS.
• Diagnose ambiguous generic routine reference.
• When an integer constant is assigned to a variable whose KIND is smaller, don't change the KIND of the constant for future references.
• Allow IMPLICIT REAL with $.
• Eliminate internal compiler error for RESHAPE of array constructor

From DFA540 FT1 T5.4-1130-46A7R to DFA540 FT2 T5.4-1170-46A97, the following corrections have been made:

• Speed up processing of EQUIVALENCE groups in COMMON blocks.
• Properly handle SIZE(A(:)(I:J))
• Implement INT8 intrinsic (already documented)
• Implement !DEC$ PSECT NOWRT for COMMON blocks on UNIX.
• Allow omitted OPTIONAL arguments to be passed as optional arguments to intrinsics.
• If "too many errors", suppress informational/warning messages as well.
• Allow keyword specification of arguments A10, A20, etc. to MAX/MIN.
• Eliminate internal compiler error for case of ADJUSTL with array argument.
• Implement nested LASTPRIVATE for COMMON.
• Where DATA initializations initialize the same array element (not allowed according to the standard, but supported by Compaq Fortran), preserve the order in which initializations were specified, so that the last one takes precedence.
• Fix parsing error for typed function declaration with MS-style argument passing specification in argument list.
• Resolve incorrect generic resolution between subroutine with no arguments (but () specified as argument list) and routine with one required argument.
• Allow named constant to be used in specification expression later in same statement where it was defined.
• Give error instead of compiler abort for IF (expr) SELECT CASE
• For F90 standards checking, warn about I0, etc. in FORMAT.
• Warn about variables in BLOCK DATA subprogram that are not in a COMMON.
• Fix problem with continuation in free-form source when OpenMP conditional compilation is used.
• Eliminate internal compiler failure for incorrect ". AND." in free-form source.
• Fix a case where two NaNs sometimes compared as equal.
• Fix assertions caused by instruction scheduling at -fpe1.

From DFA540 FT2 T5.4-1170-46A97 to DF540 FT3 T5.4-1195-46AAC, the following corrections have been made:

• Support the full range of format width specifiers as documented.
• Correct problem with page-alignment for POINTER objects.
• Improve detection of contiguous array slices.
• Allow EOSHIFT to work on REAL*16 and COMPLEX*32 arrays.
• Correct support for SIZE with argument being array with vector subscripts.
• Eliminate internal compiler failure for incorrect ". AND." in free-form source.
• Correct problem with fixed form continuation with -assume cc_omp.
• Improve handling of continued C-string escape sequences.
• Preserve order of data initializations where the same location is initialized multiple times. While this is not allowed by the standard, we do support it.
• Eliminate spurious warnings for use of some intrinsics when -double_size 128 is used.
• Correct problem with module visibility and ONLY.
• Eliminate internal compiler error for certain use of POINTER and nested structures.
• Eliminate inappropriate error about multiple matching procedures when generic operators are used.
• Correct parsing error of certain variable names in an assignment following a logical IF.
• Eliminate inappropriate shape mismatch for certain use of RESHAPE in a PARAMETER initialization expression.

From DFA540 FT3 T5.4-1195-46AAC to DFA540 V5.4-1265-46ABA, the following corrections have been made:

• Add support for -mixed_str_len_arg switch on UNIX, which puts character argument lengths next to their addresses in the argument list (this is the default Windows method). Add support for [NO]MIXED_STR_LEN_ARG attribute in !DEC$ ATTRIBUTES.
• Eliminate spurious error for declaration of RECORD array in a COMMON block in a MODULE.
• Correct handling of generics where several routines have an optional argument at the same position.
• In C strings, allow octals of the form \ooo to have 1, 2 or 3 octal digits and hex objects of the form \Xxx to have 1 or 2 hexadecimal characters.
• Add support for -check arg_temp_created.
• Annotations now are displayed on a per-routine basis.
• Put out type information to the debugger for module functions

From DFA540 V5.4-1265-46ABA to DFA540 V5.4-1283-46ABA, the following corrections have been made:

• Allow Sequence/Nosequence directives on fields of derived types with -hpf, whether or not numa switch is set.
• Add diagmostic "The NOSEQUENCE directive may not be specified for a component of a derived type which has the SEQUENCE attribute.".
• Issue diagnostic if NOSEQUENCE directives are applied to fields of 'sequence' derived types.
• When calculating the size of an array to be allocated, check the size computation for integer overflow.
• Fix obscure bug with OPTIONAL mask for intrinsics MAXLOC et al. The mask has an expression for the dimension bound(s).
• When a psect becomes too big, give a better error message: "Psect xxx is too large. Reduce array sizes or make arrays AUTOMATIC or ALLOCATABLE."

The following known problems exist with Compaq Fortran Version 5.4:

• The following is a list of known problems for -omp and -mp parallel support in Version 5.4:
  • Global variables referenced by statement functions inside parallel regions should not reference local instances of those variable names. The following example should print 42.0 {10 times} instead of 0.0:

    real x,y(10) statement(a) = a + x x = 41.0 !$par parallel local(x) x = -1.0 !$par pdo local(i) do i=1,10 y(i) = statement(1.0) end do !$par end parallel type *,y end

• Please note that -warn decl gives an error level message, not a warning level message.
• When using Ladebug with certain versions of the UNIX operating system, be aware that a trailing underscore may be needed to display module variables. For example, to display variable X in module MOD, if typing print $mod$x$ does not display the variable's value, type:

  print $MOD$X$_

The following limitations exist in Version 5.4A:

• When using the -omp or -mp options, if you declare a parallel DO loop which accesses an INTEGER*1 array (one element at a time), the results may not be correct unless you also specify -granularity byte . In general, if a parallel loop tries to store things of "small" size into packed spaces, the granularity needs to be set to that size. However, while this "fixes" the program to do the right thing, it is a dreadful performance slowdown. It would be better to execute such code serially.
• If you specify the -hpf (or -wsf ) option to request parallel processing, the following Compaq Fortran language features are not supported:
  • REAL (KIND=16) (same as REAL*16) data type
  • Compaq Fortran 77 implementation of CRAY-style pointers
  • Variable format expressions (VFEs). For example:

    FORMAT(<N+1>I4)

  • Initialization of Compaq Fortran 77 structures. For example:

    STRUCTURE /S/ INTEGER I / 100 / END STRUCTURE

Version 5.3 is a minor release that includes corrections to problems discovered since Version 5.2 was released and certain new features.

The following topics are discussed:

• Section 1.7.1 (Version 5.3 ECO 02 New Features)
• Section 1.7.2 (Version 5.3 ECO 01 New Features)
• Section 1.7.3 (Version 5.3 ECO 01 HPF New Features)
• Section 1.7.4 (Version 5.3 New Features)
• Section 1.7.5 (Version 5.3 Important Information)
• Section 1.7.6 (Version 5.3 Corrections)
• Section 1.7.7 (HPF in Compaq Fortran Version 5.3 )
• Section 1.7.8 (Version 5.3 Known Problems )

1.7.1 Version 5.3 ECO 02 New Features

• COMMON blocks can be marked read-only using the NOWRT PSECT attribute.

From DFA531 ECO 01 X5.3-1120-44A7B to DFA532 ECO 02 X5.3-1155-44A8I, the following corrections have been made:

• Eliminate internal compiler error for SIZE of a derived type component pointer array.
• Eliminate more unnecessary copies of contiguous array arguments.
• Speed up processing of EQUIVALENCE groups in COMMON blocks.
• Properly handle SIZE(A(:)(I:J))
• Implement INT8 intrinsic (already documented) - UNIX)
• Implement !DEC$ PSECT NOWRT for COMMON blocks on UNIX.
• Allow omitted OPTIONAL arguments to be passed as optional arguments to intrinsics.
• If "too many errors", suppress informational/warning messages as well.
• Allow keyword specification of arguments A10, A20, etc. to MAX/MIN.
• Eliminate internal compiler error for case of ADJUSTL with array argument.
• Eliminate internal compiler error for particular use of nested STRUCTURE.
• Disallow illegal ALLOCATED(A(:))
• Allow, as an extension for compatibility with our Fortran 77 compilers, a LOGICAL value as an argument to intrinsics which accept INTEGER, such as ABS.
• Diagnose ambiguous generic routine reference.
• When an integer constant is assigned to a variable whose KIND is smaller, don't change the KIND of the constant for future references.
• Allow IMPLICIT REAL with $.
• Eliminate internal compiler error for RESHAPE of array constructor.
• Fix parsing error for typed function declaration with MS-style argument passing specification in argument list.
• Resolve incorrect generic resolution between subroutine with no arguments (but () specified as argument list) and routine with one required argument.
• Allow named constant to be used in specification expression later in same statement where it was defined.
• Give error instead of compiler abort for IF (expr) SELECT CASE
• For F90 standards checking, warn about I0, etc. in FORMAT.
• Warn about variables in BLOCK DATA subprogram that are not in a COMMON.
• Fix several compiler assertions including one from instruction scheduling at -fpe1.

The following new Compaq Fortran features are now supported:

• The new INT_PTR_KIND() intrinsic returns the kind of an integer pointer (ie, 8 for Tru64 UNIX/Alpha).
• An optional KIND argument is now allowed on the intrinsics LEN, SHAPE, SIZE, UBOUND, LBOUND, MAXLOC, MINLOC, INDEX, LEN_TRIM, SCAN, and VERIFY. This allows these intrinsics to return a result that is other than default integer kind.

The following new f90 command options are now supported:

• -wsf_target for use with -wsf .
  Notice the change in spelling of both -wsf_target and the environment variable DECF90_WSF_TARGET .
• -assume [no]cc_omp [dis]allows the use of OpenMP conditional compilation separate from -omp .
• Compiling -omp -non_shared now adds -qlpset_r -lpset to the link line.

Some important information to note about this release:

• An end-of-file condition on READ no longer triggers an ERR= branch - this is to conform with clearer wording in the recent standard. If an EOF condition occurs and there is no END= or IOSTAT=, an error is signalled.
• Add a NUL to the end of non-C character literals. This will not be reflected in the "length" of the constant.
• %VAL/%REF now overrides any mechanism specified in an explicit interface.
• The Compaq Extended Math Library (CXML) routines are not updated in this ECO 01 Compaq Fortran kit.

From released version V5.3-915-449BB to ECO 01 FT1 X5.3-953-44A17, the following corrections have been made:

• Improve compile time performance when USEing a module which contains a large COMMON block with many (thousands) of EQUIVALENCEd variables.
• Allow general CHARACTER expressions for the MOLD argument of the TRANSFER intrinsic.
• Correct problem that prevented scalar numeric variables and array elements which appeared in the iolist of an output statement (WRITE, etc.) from participating in uninitialized variable analysis.
• Add support for -ccdefault switch.
• In OPEN statement, if STATUS is not specified, default to UNKNOWN unless -f66 is specified, in which case default to NEW.
• Eliminate internal compiler error for case involving EQUIVALENCEd POINTER variables not in COMMON.
• Missing !DEC$ ENDIF no longer causes compiler abort.
• Correct rules for when SAVE in a PURE procedure is allowed or not.
• Correct parsing of assignment to field of RECORD whose name begins with TYPE.
• Eliminate E-level error when a BLOCK DATA subprogram name is the same as that of a COMMON block. A future revision will cause an appropriate diagnostic to appear.
• Issue clearer error message when a module name is repeated in a USE statement.
• Eliminate problem where UBOUND gave wrong value in certain cases.
• Allow substrings in left hand side of FORALL.
• Give proper error when a PARAMETER constant is CALLed.
• Give proper error when a variable is CALLed.
• In assignment statements, make sure that real constants are evaluated using the precision appropriate for their syntax (single/double/quad).

From DFA531 ECO 01 FT1 X5.3-953-44A17 to FT2 T5.3-1034-44A3V, the following corrections have been made:

• Fix problem which could cause incorrect code to be generated for certain uses of PACK, RESHAPE and UNPACK, primarily with CHARACTER arguments.
• Correct f77 driver problems, such as invalid syntax passed to cpp when a .F file was compiled.
• Speed up compilation of initialized nested structures.
• Fix problem with incorrect code generated for use of nested SPREAD intrinsic.
• Driver now understands shorter "-ieee" switch name.
• Eliminate internal compiler errors for use of defined assignment in FORALL.
• Speed up compilation where very large MODULEs are repeatedly used. Further work in this area remains.
• Correct problem with PACK intrinsic where array is array of structures with an array field with constant subscript/substring values.
• Improve generated code for SPREAD intrinsic.
• Improve generated code for array reductions.
• Improve generated code for UBOUND, LBOUND, SELECTED_INT_KIND, and SELECTED_REAL_KIND when the argument(s) are constants.
• Improve generated code for SIZEOF when bounds are constants.
• Fix UNIX driver to pass ld's position-specific qualifier within lib_list rather than cc_list.
• Eliminate internal compiler error for certain cases of integer (not F90) POINTER as a module variable.
• Reduce severity of "variable has not been used" diagnostic to "informational".
• Improve generated code for MINVAL/MAXVAL.
• Improve generated code for SIZE(A(:)).
• Improve generated code for SIZE with array argument that has vector subscripts.
• Add new INT_PTR_KIND() intrinsic which returns the kind of an integer pointer (either 4 or 8).
• Eliminate internal compiler error for use of allocatable array reference in a variable format expression.
• Improve generated code for ILEN.
• An end-of-file condition on READ no longer triggers an ERR= branch - this is to conform with clearer wording in the recent standard. If an EOF condition occurs and there is no END= or IOSTAT=, an error is signalled.
• Add a NUL to the end of non-C character literals. This will not be reflected in the "length" of the constant. This matches the Compaq Fortran 77 behavior, which was undocumented.
• %VAL/%REF now overrides any mechanism specified in an explicit interface.
• Generate much better code for certain array constructors (such as (/0,I=1,500000/) and allow lowerbound, upperbound and stride to have different KINDs.
• If -warn declarations is specified, do not issue diagnostic for use of IMPLICIT NONE.
• Eliminate internal compiler error for EOSHIFT with constant array argument.
• Eliminate internal compiler error for program fragment that is "ifdef-ed out".
• Report error for kind mismatch where an ac-value in an array constructor is an expression of a different kind than the other literal ac-values.
• Report error for assumed-size array passed as actual to deferred-shape array.
• Eliminate compiletime error for use of AIMAG in initialization expression when -real_size 128 is specified.
• Eliminate internal compiler error for format in a PRINT statement being an expression with concatenation.
• DFOR$PREFETCH, DFOR$PREFETCH_MODIFY, and DFOR$PREFETCH_EVICT_NEXT intrinsics now supported for Alpha processors.
• Generate correct values for a PARAMETER array whose element values are computed in an implied DO loop involving indexing into another PARAMETER array.
• Correct bad parse of .NE. as record field name.
• Allow a dummy argument that has an INTENT attribute specified to be specified in a NAMELIST list.
• Give standards warning for kind mismatch between source and pad arguments in RESHAPE.
• Long source lines are now correctly compiled when standards checking is requested (warning is still given).
• Correct problem with incorrect error given for a particular complex module renaming case.
• Allow as an extension (as does Compaq Fortran 77) a LOGICAL argument to an intrinsic which expects INTEGER.
• Correctly parse format which contains ">" characters and a variable format expression.
• Eliminate internal compiler error for a particularly complex and deeply nested structure reference in an IF.
• Don't give error for passing section of an assumed size array to a deferred shape array.
• Improve compilation speed for certain large DATA statements.
• Eliminate internal compiler error for a certain complicated FORALL.
• Correct problem with PUBLIC/PRIVATE attributes of a COMMON block in a MODULE.
• Allow (A .EQ. .NOT. B) (an extension).
• Correct problem with some COMPLEX*32 initialization expressions.
• Eliminate spurious unused variable diagnostic for variable used in pointer assignment.
• Correct problem where asking for standards checking disabled -D (/define)
• Fix a case where two NaNs sometimes compared as equal.

From DFA531 ECO 01 FT2 T5.3-1034-44A3V to final X5.3-1120 -44A7B, the following corrections have been made:

• Eliminate many unnecessary copies of assumed-shape arrays, when passed as arguments to explicit shape arrays, by keeping track of whether or not the array is known to be contiguous.
• Automatically force the alignment of a COMMON block to be at least as large as that required by the widest variable in the COMMON block.
• Fix a number of problems with WHERE inside of FORALL.
• Make defined assignment in FORALL work properly.
• Generate correct code for CSHIFT of a non-contiguous array slice.
• Allow user-defined types to be named BYTE and DOUBLECOMPLEX.
• Improve generated code for mixed COMPLEX-REAL multiplication and division.
• In array constructors with only one implied-DO, and nothing else, avoid creating an unnecessary temp.
• Allow SIZEOF(allocatable-array)
• Improve generated code for SIZEOF(array)
• Allow directory names specified with -module to be longer than 23 characters.
• Prevent incorrect collapsing of implied-DO loop in an I/O statement where there are nested loops and the implied-DO variable of one loop is used as a bound of an inner loop.
• When the error limit has been exceeded (default 30), simply suppress further messages rather than exiting the compiler. This means that the object file will get deleted appropriately, the listing file created, and subsequent source files on the command line will be processed.
• Re-allow complex constants to be passed by value. The real and imaginary parts are passed as separate arguments.
• Allow character array valued function as a format specifier.
• Allow debugging of a character function whose length is computed based on the length of a passed-length character argument.
• Solve a set of problems where the length of a character function is computed using a dummy array argument.
• The use of an INTENT(OUT) argument with LOC is now considered a "definition" for the purpose of uninitialized variable checking. Also, the use of LOC(ARRAY(N)) is now considered a "use" of ARRAY for unused variable checking.
• Eliminate internal compiler error for structure with %FILL component in module procedure.
• When standards checking is requested, do not give a warning for fixed-form source line exactly 72 columns long.
• Eliminate internal compiler error for assignment statement with variable whose name starts with an underscore. (Such names are not allowed, but a reasonable error should have been given.)
• Correct the problem where if a program unit contains two internal subroutines which both use host-association to access the same variable, the second one gets an inappropriate error.
• Eliminate internal compiler error for declaration of a derived type array with an initializer that is an implied-DO array constructor.
• Eliminate inappropriate error message for initialization expression of an implied-DO array constructor of length zero.
• When standards checking is enabled, give one warning, not three, for a !DEC$ ATTRIBUTES directive.
• Generate correct code for certain cases involving overloading of the .AND. and .OR. operators.
• Allow Variable Format Expression in a character literal when the I/O list has a subscripted array element.
• Eliminate compiler abort with incorrect program that names the enclosing program unit in an ONLY clause.
• Allow the extension syntax '101'B for a binary literal.
• Correctly handle whitespace in -omp conditional compilation lines.
• Fix problem where $INTEGER directive was not being properly handled.
• Add support for KIND= keyword in MINLOC/MAXLOC.
• Add suppprt for KIND= keyword in various string intrinsics.
• Prevent compiler abort for incorrect attempt to pass unsupported datatypes by value.
• Properly report invalid declaration EXTERNAL, INTEGER and recover so that remainder of program is properly parsed.
• Give standards warning for non-standard placement of NAMELIST.
• Eliminate internal compiler error for particular type of concatenation of substrings when Fortran 95 standards checking is requested.
• When converting a negative REAL value to COMPLEX, use +0.0 as the imaginary part rather than -0.0.
• Allow PARAMETER constant in ALIGN= specification of !DEC$ PSECT.
• Don't give shape mismatch for correct-shape RESHAPE in initialization expression.
• Don't give inappropriate alignment warnings for certain convoluted EQUIVALENCE/COMMON combinations.
• Eliminate internal compiler error for initialization expression which contains a constant expression in a structure constructor.
• Allow EXTERNAL :: FOO

The following information pertains to HPF using MPI.

Overview of HPF and MPI

The Compaq Fortran compiler now generates code that uses MPI as its message-passing library instead of PSE's HPF-specific support. The compiler provides a choice of three different variants of MPI: one for Compaq's SC supercomputer systems, one that supports shared-memory and Memory Channel^tm interconnects, and public domain MPI for other interconnects that include Ethernet and FDDI.

It is now possible to write HPF programs that also call or use MPI (such as distributed-memory libraries that invoke MPI). The compiler's MPI runtime library uses its own private MPI "communicator" so it won't interfere with other MPI code. A new example program, /usr/examples/hpf/call_mpi.f90, illustrates this.

You enable the new MPI-based runtime library, that supports Compaq Fortran's HPF directives, by adding the -wsf_target option. This option, which requires an argument, belongs in the compilation and link commands.

Compiling HPF Programs for MPI

You must now specify which variant of MPI support you wish to use for HPF programs by including the option -wsf_target with an MPI selection (argument target ) in the command to the f90 compiler. An example is next that selects Compaq MPI.

% f90 -wsf 2 -wsf_target cmpi -c lu.f90

An expansion of this example is next that invokes both the compiler and linker.

% f90 -wsf 2 -wsf_target cmpi -o lu lu.f90

The values of target in the option -wsf_target target appear next with their explanations.

target	Explanation
smpi	SC (Quadrics) MPI This MPI comes installed on SC-series systems. It works with the SC's RMS software that provides a set of commands for launching MPI jobs, scheduling these jobs on SC clusters, and performing other miscellaneous tasks.
cmpi	Compaq MPI This MPI is a version that is specifically tuned for Alpha systems. It is distributed as a Compaq layered product. Compaq MPI supports only Memory Channel clusters and shared-memory (SMP) machines.
gmpi	Generic MPI This target is for use with MPICH V1.2.0 or other compatible libraries. MPICH is a public domain implementation of the MPI specification that is available for many platforms. You can obtain this implementation from http://www-unix.mcs.anl.gov/mpi/mpich/. MPICH V1.2.0 supports many interconnection networks including Ethernet, FDDI, and other hardware. Using Compaq Fortran and HPF with this MPI is, officially, not supported. Compaq does not guarantee support of problems caused by specifying -wsf_target gmpi . However, Compaq remains quite interested in receiving problem reports and will attempt to respond to them.

If the command to the f90 compiler includes -wsf_target target , then the command must also include -wsf .

Another way of specifying the version of MPI to the compiler, instead of using the option -wsf_target , is to set the environment variable DECF90_WSF_TARGET to a value in the first column of the previous table. For example, the command

% f90 -wsf 2 -wsf_target cmpi -c lu.f90

is equivalent to the commands

% setenv DECF90_WSF_TARGET cmpi % f90 -wsf 2 -c lu.f90

If an f90 command contains -wsf_target with a value (such as cmpi) and environment variable DECF90_WSF_TARGET is set to a different value, then the value in the f90 command overrides the value of the environment variable.

Using the environment variable to select the desired MPI variant is the recommended method. This will require the fewest changes to existing scripts for building HPF programs, and will allow users generating code for more than one MPI variant to do so more easily. Compaq additionally recommends setting the environment variable in your shell initialization file (e.g. .cshrc if you use 'csh'), particularly if you usually use only one MPI variant.

A table, showing all changes to HPF-related compiler options between Fortran V5.3 and V5.3 ECO 01, is next.

Fortran V5.3	Fortran V5.3 ECO 01
-assume bigarrays	No change
-assume nozsize	No change
-hpf_matmul	Deleted
-nearest_neighbor	No change
-nowsf_main	No change (but currently does not work)
-pprof	Use only with -wsf_target pse
-show hpf*	No change
-show wsfinfo	No change
-wsf	No change
---	-wsf_target target

Linking HPF Programs with MPI

You must now specify which variant of MPI support you wish to use for HPF programs by including the option -wsf_target with an MPI selection (argument target ) in the link command. An example is next.

% f90 -wsf 2 -wsf_target cmpi -o lu lu.o

The values of target come from the table in the section "Compiling HPF Programs for MPI".

If you specified generic MPI at compilation time, either by including the -wsf_target gmpi option or by setting the environment variable DECF90_WSF_TARGET to gmpi, you must specify a path to the desired generic MPI library during linking. Do this in one of these ways:

• Set the environment variable DECF90_GMPILIB to the path of the desired generic MPI library to link with
• In the link command line, include -l (possibly along with -l ) with the path of the desired generic MPI library to link with. Or, explicitly add the library to the link command line.

An example of a link command for a generic MPI library is next.

% f90 -wsf 2 -wsf_target gmpi -o lu lu.o /usr/users/me/libmpich.a

In addition, you must have the Developer's Tool Kit software installed on your computer to link properly with the option -wsf_target gmpi .

Finally, programs linked with -wsf_target and an MPI target must be linked with -call_shared (which is the default); the -non_shared option does not link correctly.

Running HPF Programs Linked with MPI

The dmpirun command executes program files created with the -wsf_target cmpi option. Include the -n n option in the command line where n is the same value of -wsf n in the compilation command line. Or, if no value was given with the -wsf option, then set n to the desired number of peers. Also include the name of the program file.

An example is next where the compilation command line included -wsf 4 and the name of the program file is heat8.

% dmpirun -n 4 heat8

If your AlphaServer SC system is running with Revision A of the Quadrics switch, your boot log will contain the message:

elan0: Rev A Elite network detected - disabling adaptive routing (1)

To make MPI programs (including HPF programs generated with the "-wsf_target smpi" option) run properly with Revision A hardware, you need to set the LIBELAN_GROUP_HWBCAST environment variable to DISABLE; for example, from csh:

% setenv LIBELAN_GROUP_HWBCAST DISABLE

The manpage dmpirun contains a full description of this command.

The prun command executes program files created with the -wsf_target smpi option. Include the -n n option in the command line where n is the same value of -wsf n in the compilation command line. Or, if no value was given with the -wsf option, then set n to the desired number of peers. Also include the name of the program file.

An example is next where the compilation command line included -wsf 4 and the name of the program file is heat8.

% prun -n 4 -N 4 heat8

The mpirun command executes program files created with the -wsf_target gmpi option. Include the -np n option in the command line where n is the same value of -wsf n in the compilation command line. Also include the name of the program file. The mpirun command varies according to where you installed the generic MPI.

An example is next where the compilation command line included -wsf 4 and the name of the program file is heat8.

% /usr/users/me/mpirun -np 4 heat8

In the /usr/examples/hpf directory, there is a sample script that will launch an HPF program for any variant of MPI. This script, called "hpfrun", will even determine the number of processors a source program was compiled for (if that was specified at compile time), and invoke the proper MPI run command with the number of processors specified. Portions of the script, or the entire script, may be useful for users automating the building and running of HPF programs.

Cleaning up After Running HPF Programs Linked with MPI

Execution of the dmpirun command (but not the prun and mpirun commands) may leave various system resources allocated after the program has completed. To free them, give the mpiclean with no arguments. An example is next.

% mpiclean

Changing HPF Programs for MPI

There two changes you should make to Fortran source files before compiling them for MPI. If a module contains an EXTRINSIC (HPF_LOCAL) statement and it executes on a system different from peer 0, then its output intended for stdout may (depending on the variant of MPI used) go instead to /dev/null. Change such modules or your execution commands to have the extrinsic subroutine do input/output only from peer 0.