1 \input texinfo @c -*-texinfo-*-
3 @setfilename gfortran.info
4 @set copyrights-gfortran 1999-2006
6 @include gcc-common.texi
8 @settitle The GNU Fortran 95 Compiler
10 @c Create a separate index for command line options
12 @c Merge the standard indexes into a single one.
21 @c Use with @@smallbook.
23 @c %** start of document
25 @c Cause even numbered pages to be printed on the left hand side of
26 @c the page and odd numbered pages to be printed on the right hand
27 @c side of the page. Using this, you can print on both sides of a
28 @c sheet of paper and have the text on the same part of the sheet.
30 @c The text on right hand pages is pushed towards the right hand
31 @c margin and the text on left hand pages is pushed toward the left
33 @c (To provide the reverse effect, set bindingoffset to -0.75in.)
36 @c \global\bindingoffset=0.75in
37 @c \global\normaloffset =0.75in
41 Copyright @copyright{} @value{copyrights-gfortran} Free Software Foundation, Inc.
43 Permission is granted to copy, distribute and/or modify this document
44 under the terms of the GNU Free Documentation License, Version 1.1 or
45 any later version published by the Free Software Foundation; with the
46 Invariant Sections being ``GNU General Public License'' and ``Funding
47 Free Software'', the Front-Cover
48 texts being (a) (see below), and with the Back-Cover Texts being (b)
49 (see below). A copy of the license is included in the section entitled
50 ``GNU Free Documentation License''.
52 (a) The FSF's Front-Cover Text is:
56 (b) The FSF's Back-Cover Text is:
58 You have freedom to copy and modify this GNU Manual, like GNU
59 software. Copies published by the Free Software Foundation raise
60 funds for GNU development.
64 @dircategory Programming
66 * gfortran: (gfortran). The GNU Fortran 95 Compiler.
68 This file documents the use and the internals of
69 the GNU Fortran 95 compiler, (@command{gfortran}).
71 Published by the Free Software Foundation
72 51 Franklin Street, Fifth Floor
73 Boston, MA 02110-1301 USA
79 @setchapternewpage odd
81 @title Using GNU Fortran 95
83 @center The gfortran team
85 @vskip 0pt plus 1filll
86 For the @value{version-GCC} Version*
88 Published by the Free Software Foundation @*
89 51 Franklin Street, Fifth Floor@*
90 Boston, MA 02110-1301, USA@*
91 @c Last printed ??ber, 19??.@*
92 @c Printed copies are available for $? each.@*
105 This manual documents the use of @command{gfortran},
106 the GNU Fortran 95 compiler. You can find in this manual how to invoke
107 @command{gfortran}, as well as its features and incompatibilities.
110 @emph{Warning:} This document, and the compiler it describes, are still
111 under development. While efforts are made to keep it up-to-date, it might
112 not accurately reflect the status of the most recent @command{gfortran}.
116 @comment When you add a new menu item, please keep the right hand
117 @comment aligned to the same column. Do not use tabs. This provides
118 @comment better formatting.
121 * Getting Started:: What you should know about @command{gfortran}.
122 * GFORTRAN and GCC:: You can compile Fortran, C, or other programs.
123 * GFORTRAN and G77:: Why we chose to start from scratch.
124 * Invoking GFORTRAN:: Command options supported by @command{gfortran}.
125 * Project Status:: Status of @command{gfortran}, roadmap, proposed extensions.
126 * Contributing:: How you can help.
127 * Standards:: Standards supported by @command{gfortran}
128 * Runtime:: Influencing runtime behavior with environment variables.
129 * Extensions:: Language extensions implemented by @command{gfortran}
130 * Intrinsic Procedures:: Intrinsic procedures supported by @command{gfortran}
131 * Copying:: GNU General Public License says
132 how you can copy and share GNU Fortran.
133 * GNU Free Documentation License::
134 How you can copy and share this manual.
135 * Funding:: How to help assure continued work for free software.
136 * Index:: Index of this documentation.
141 @c ---------------------------------------------------------------------
143 @c ---------------------------------------------------------------------
145 @node Getting Started
146 @chapter Getting Started
148 Gfortran is the GNU Fortran 95 compiler front end,
149 designed initially as a free replacement for,
150 or alternative to, the unix @command{f95} command;
151 @command{gfortran} is the command you'll use to invoke the compiler.
153 Gfortran is still in an early state of development.
154 @command{gfortran} can generate code for most constructs and expressions,
155 but much work remains to be done.
157 When @command{gfortran} is finished,
158 it will do everything you expect from any decent compiler:
162 Read a user's program,
163 stored in a file and containing instructions written
164 in Fortran 77, Fortran 90 or Fortran 95.
165 This file contains @dfn{source code}.
168 Translate the user's program into instructions a computer
169 can carry out more quickly than it takes to translate the
170 instructions in the first
171 place. The result after compilation of a program is
173 code designed to be efficiently translated and processed
174 by a machine such as your computer.
175 Humans usually aren't as good writing machine code
176 as they are at writing Fortran (or C++, Ada, or Java),
177 because is easy to make tiny mistakes writing machine code.
180 Provide the user with information about the reasons why
181 the compiler is unable to create a binary from the source code.
182 Usually this will be the case if the source code is flawed.
183 When writing Fortran, it is easy to make big mistakes.
184 The Fortran 90 requires that the compiler can point out
185 mistakes to the user.
186 An incorrect usage of the language causes an @dfn{error message}.
188 The compiler will also attempt to diagnose cases where the
189 user's program contains a correct usage of the language,
190 but instructs the computer to do something questionable.
191 This kind of diagnostics message is called a @dfn{warning message}.
194 Provide optional information about the translation passes
195 from the source code to machine code.
196 This can help a user of the compiler to find the cause of
197 certain bugs which may not be obvious in the source code,
198 but may be more easily found at a lower level compiler output.
199 It also helps developers to find bugs in the compiler itself.
202 Provide information in the generated machine code that can
203 make it easier to find bugs in the program (using a debugging tool,
204 called a @dfn{debugger}, such as the GNU Debugger @command{gdb}).
207 Locate and gather machine code already generated to
208 perform actions requested by statements in the user's program.
209 This machine code is organized into @dfn{modules} and is located
210 and @dfn{linked} to the user program.
213 Gfortran consists of several components:
217 A version of the @command{gcc} command
218 (which also might be installed as the system's @command{cc} command)
219 that also understands and accepts Fortran source code.
220 The @command{gcc} command is the @dfn{driver} program for
221 all the languages in the GNU Compiler Collection (GCC);
223 you can compile the source code of any language for
224 which a front end is available in GCC.
227 The @command{gfortran} command itself,
228 which also might be installed as the
229 system's @command{f95} command.
230 @command{gfortran} is just another driver program,
231 but specifically for the Fortran 95 compiler only.
232 The difference with @command{gcc} is that @command{gfortran}
233 will automatically link the correct libraries to your program.
236 A collection of run-time libraries.
237 These libraries contain the machine code needed to support
238 capabilities of the Fortran language that are not directly
239 provided by the machine code generated by the
240 @command{gfortran} compilation phase,
241 such as intrinsic functions and subroutines,
242 and routines for interaction with files and the operating system.
243 @c and mechanisms to spawn,
244 @c unleash and pause threads in parallelized code.
247 The Fortran compiler itself, (@command{f951}).
248 This is the gfortran parser and code generator,
249 linked to and interfaced with the GCC backend library.
250 @command{f951} ``translates'' the source code to
251 assembler code. You would typically not use this
253 instead, the @command{gcc} or @command{gfortran} driver
254 programs will call it for you.
259 @c ---------------------------------------------------------------------
261 @c ---------------------------------------------------------------------
263 @node GFORTRAN and GCC
264 @chapter GFORTRAN and GCC
265 @cindex GNU Compiler Collection
267 GCC used to be the GNU ``C'' Compiler,
268 but is now known as the @dfn{GNU Compiler Collection}.
269 GCC provides the GNU system with a very versatile
270 compiler middle end (shared optimization passes),
271 and back ends (code generators) for many different
272 computer architectures and operating systems.
273 The code of the middle end and back end are shared by all
274 compiler front ends that are in the GNU Compiler Collection.
276 A GCC front end is essentially a source code parser
277 and an intermediate code generator. The code generator translates the
278 semantics of the source code into a language independent form called
281 The parser takes a source file written in a
282 particular computer language, reads and parses it,
283 and tries to make sure that the source code conforms to
285 Once the correctness of a program has been established,
286 the compiler will build a data structure known as the
287 @dfn{Abstract Syntax tree},
288 or just @dfn{AST} or ``tree'' for short.
289 This data structure represents the whole program
290 or a subroutine or a function.
291 The ``tree'' is passed to the GCC middle end,
292 which will perform optimization passes on it. The optimized AST is then
293 handed off too the back end which assembles the program unit.
295 Different phases in this translation process can be,
296 and in fact @emph{are} merged in many compiler front ends.
297 GNU Fortran 95 has a strict separation between the
298 parser and code generator.
300 The goal of the gfortran project is to build a new front end for GCC.
301 Specifically, a Fortran 95 front end.
302 In a non-gfortran installation,
303 @command{gcc} will not be able to compile Fortran 95 source code
304 (only the ``C'' front end has to be compiled if you want to build GCC,
305 all other languages are optional).
306 If you build GCC with gfortran, @command{gcc} will recognize
307 @file{.f/.f90/.f95} source files and accepts Fortran 95 specific
308 command line options.
312 @c ---------------------------------------------------------------------
314 @c ---------------------------------------------------------------------
316 @node GFORTRAN and G77
317 @chapter GFORTRAN and G77
321 Why do we write a compiler front end from scratch?
322 There's a fine Fortran 77 compiler in the
323 GNU Compiler Collection that accepts some features
324 of the Fortran 90 standard as extensions.
325 Why not start from there and revamp it?
327 One of the reasons is that Craig Burley, the author of G77,
328 has decided to stop working on the G77 front end.
329 On @uref{http://world.std.com/~burley/g77-why.html,
330 Craig explains the reasons for his decision to stop working on G77}
331 in one of the pages in his homepage.
332 Among the reasons is a lack of interest in improvements to
334 Users appear to be quite satisfied with @command{g77} as it is.
335 While @command{g77} is still being maintained (by Toon Moene),
336 it is unlikely that sufficient people will be willing
337 to completely rewrite the existing code.
339 But there are other reasons to start from scratch.
340 Many people, including Craig Burley,
341 no longer agreed with certain design decisions in the G77 front end.
342 Also, the interface of @command{g77} to the back end is written in
343 a style which is confusing and not up to date on recommended practice.
344 In fact, a full rewrite had already been planned for GCC 3.0.
346 When Craig decided to stop,
347 it just seemed to be a better idea to start a new project from scratch,
348 because it was expected to be easier to maintain code we
349 develop ourselves than to do a major overhaul of @command{g77} first,
350 and then build a Fortran 95 compiler out of it.
354 @c ---------------------------------------------------------------------
356 @c ---------------------------------------------------------------------
359 @chapter Project Status
362 As soon as gfortran can parse all of the statements correctly,
363 it will be in the ``larva'' state.
364 When we generate code, the ``puppa'' state.
365 When gfortran is done,
366 we'll see if it will be a beautiful butterfly,
367 or just a big bug....
369 --Andy Vaught, April 2000
372 The start of the GNU Fortran 95 project was announced on
373 the GCC homepage in March 18, 2000
374 (even though Andy had already been working on it for a while,
377 Gfortran is currently reaching the stage where is is able to compile real
378 world programs. However it is still under development and has many rough
384 * Proposed Extensions::
387 @node Compiler Status
388 @section Compiler Status
392 This is the part of gfortran which parses a source file, verifies that it
393 is valid Fortran 95, performs compile time replacement of constants
394 (PARAMETER variables) and reads and generate module files. This is
395 almost complete. Every Fortran 95 source should be accepted, and most
396 none-Fortran 95 source should be rejected. If you find a source file where
397 this is not true, please tell us. You can use the -fsyntax-only switch to
398 make gfortran quit after running the front end, effectively reducing it to
401 @item Middle end interface
402 These are the parts of gfortran that take the parse tree generated by the
403 front end and translate it to the GENERIC form required by the GCC back
404 end. Work is ongoing in these parts of gfortran, but a large part has
405 already been completed.
409 @section Library Status
411 Some intrinsic functions map directly to library functions, and in most
412 cases the name of the library function used depends on the type of the
413 arguments. For some intrinsics we generate inline code, and for others,
414 such as sin, cos and sqrt, we rely on the backend to use special
415 instructions in the floating point unit of the CPU if available, or to
416 fall back to a call to libm if these are not available.
418 Implementation of some non-elemental intrinsic functions (eg. DOT_PRODUCT,
419 AVERAGE) is not yet optimal. This is hard because we have to make decisions
420 whether to use inline code (good for small arrays as no function call
421 overhead occurs) or generate function calls (good for large arrays as it
422 allows use of hand-optimized assembly routines, SIMD instructions, etc.)
424 The IO library is in a mostly usable state. Unformatted I/O for
425 @code{REAL(KIND=10)} variables is currently not recommended.
427 Array intrinsics mostly work.
429 @node Proposed Extensions
430 @section Proposed Extensions
432 Here's a list of proposed extensions for @command{gfortran}, in no particular
433 order. Most of these are necessary to be fully compatible with
434 existing Fortran compilers, but they are not part of the official
435 J3 Fortran 95 standard.
437 @subsection Compiler extensions:
440 Flag for defining the kind number for default logicals.
443 User-specified alignment rules for structures.
445 Flag to generate @code{Makefile} info.
448 Automatically extend single precision constants to double.
451 Compile code that conserves memory by dynamically allocating common and
452 module storage either on stack or heap.
455 Flag to cause the compiler to distinguish between upper and lower case
456 names. The Fortran 95 standard does not distinguish them.
459 Compile flag to generate code for array conformance checking (suggest -CC).
462 User control of symbol names (underscores, etc).
465 Compile setting for maximum size of stack frame size before spilling
466 parts to static or heap.
469 Flag to force local variables into static space.
472 Flag to force local variables onto stack.
475 Flag to compile lines beginning with ``D''.
478 Flag to ignore lines beginning with ``D''.
481 Flag for maximum errors before ending compile.
484 Generate code to check for null pointer dereferences -- prints locus of
485 dereference instead of segfaulting. There was some discussion about this
486 option in the g95 development mailing list.
489 Allow setting the default unit number.
492 Option to initialize otherwise uninitialized integer and floating
496 Support for Fortran 200x. This includes several new features including
497 floating point exceptions, extended use of allocatable arrays, C
498 interoperability, Parameterizer data types and function pointers.
502 @subsection Environment Options
505 Pluggable library modules for random numbers, linear algebra.
506 LA should use BLAS calling conventions.
509 Environment variables controlling actions on arithmetic exceptions like
510 overflow, underflow, precision loss -- Generate NaN, abort, default.
514 Set precision for fp units that support it (i387).
517 Variable for setting fp rounding mode.
520 Variable to fill uninitialized variables with a user-defined bit
524 Environment variable controlling filename that is opened for that unit
528 Environment variable to clear/trash memory being freed.
531 Environment variable to control tracing of allocations and frees.
534 Environment variable to display allocated memory at normal program end.
537 Environment variable for filename for * IO-unit.
540 Environment variable for temporary file directory.
543 Environment variable forcing standard output to be line buffered (unix).
548 @chapter Runtime: Influencing runtime behavior with environment variables
551 The behaviour of the @command{gfortran} can be influenced by
552 environment variables.
554 Malformed environment variables are silently ignored.
557 * GFORTRAN_STDIN_UNIT:: Unit number for standard input
558 * GFORTRAN_STDOUT_UNIT:: Unit number for standard output
559 * GFORTRAN_STDERR_UNIT:: Unit number for standard error
560 * GFORTRAN_USE_STDERR:: Send library output to standard error
561 * GFORTRAN_TMPDIR:: Directory for scratch files
562 * GFORTRAN_UNBUFFERED_ALL:: Don't buffer output
563 * GFORTRAN_SHOW_LOCUS:: Show location for runtime errors
564 * GFORTRAN_OPTIONAL_PLUS:: Print leading + where permitted
565 * GFORTRAN_DEFAULT_RECL:: Default record length for new files
566 * GFORTRAN_LIST_SEPARATOR:: Separator for list output
567 * GFORTRAN_CONVERT_UNIT:: Set endianness for unformatted I/O
570 @node GFORTRAN_STDIN_UNIT
571 @section @env{GFORTRAN_STDIN_UNIT}---Unit number for standard input
573 This environment variable can be used to select the unit number
574 preconnected to standard input. This must be a positive integer.
575 The default value is 5.
577 @node GFORTRAN_STDOUT_UNIT
578 @section @env{GFORTRAN_STDOUT_UNIT}---Unit number for standard output
580 This environment variable can be used to select the unit number
581 preconnected to standard output. This must be a positive integer.
582 The default value is 6.
584 @node GFORTRAN_STDERR_UNIT
585 @section @env{GFORTRAN_STDERR_UNIT}---Unit number for standard error
587 This environment variable can be used to select the unit number
588 preconnected to standard error. This must be a positive integer.
589 The default value is 0.
591 @node GFORTRAN_USE_STDERR
592 @section @env{GFORTRAN_USE_STDERR}---Send library output to standard error
594 This environment variable controls where library output is sent.
595 If the first letter is @samp{y}, @samp{Y} or @samp{1}, standard
596 error is used. If the first letter is @samp{n}, @samp{N} or
597 @samp{0}, standard output is used.
599 @node GFORTRAN_TMPDIR
600 @section @env{GFORTRAN_TMPDIR}---Directory for scratch files
602 This environment variable controls where scratch files are
603 created. If this environment variable is missing,
604 gfortran searches for the environment variable @env{TMP}. If
605 this is also missing, the default is @file{/tmp}.
607 @node GFORTRAN_UNBUFFERED_ALL
608 @section @env{GFORTRAN_UNBUFFERED_ALL}---Don't buffer output
610 This environment variable controls wether all output is unbuffered.
611 If the first letter is @samp{y}, @samp{Y} or @samp{1}, all output is
612 unbuffered. This will slow down large writes. If the first letter is
613 @samp{n}, @samp{N} or @samp{0}, output is bufferred. This is the
616 @node GFORTRAN_SHOW_LOCUS
617 @section @env{GFORTRAN_SHOW_LOCUS}---Show location for runtime errors
619 If the first letter is @samp{y}, @samp{Y} or @samp{1}, filename and
620 line numbers for runtime errors are printed. If the first letter is
621 @samp{n}, @samp{N} or @samp{0}, don't print filename and line numbers
622 for runtime errors. The default is to print the location.
624 @node GFORTRAN_OPTIONAL_PLUS
625 @section @env{GFORTRAN_OPTIONAL_PLUS}---Print leading + where permitted
627 If the first letter is @samp{y}, @samp{Y} or @samp{1},
628 a plus sign is printed
629 where permitted by the Fortran standard. If the first lettter
630 is @samp{n}, @samp{N} or @samp{0}, a plus sign is not printed
631 in most cases. Default is not to print plus signs.
633 @node GFORTRAN_DEFAULT_RECL
634 @section @env{GFORTRAN_DEFAULT_RECL}---Default record length for new files
636 This environment variable specifies the default record length for
637 files which are opened without a @code{RECL} tag in the @code{OPEN}
638 statement. This must be a positive integer. The default value is
641 @node GFORTRAN_LIST_SEPARATOR
642 @section @env{GFORTRAN_LIST_SEPARATOR}---Separator for list output
644 This environment variable specifies the separator when writing
645 list-directed output. It may contain any number of spaces and
646 at most one comma. If you specify this on the command line,
647 be sure to quote spaces, as in
649 $ GFORTRAN_LIST_SEPARATOR=' , ' ./a.out
651 when @code{a.out} is the gfortran program that you want to run.
652 Default is a single space.
654 @node GFORTRAN_CONVERT_UNIT
655 @section @env{GFORTRAN_CONVERT_UNIT}---Set endianness for unformatted I/O
657 By setting the @env{GFORTRAN_CONVERT_UNIT} variable, it is possible
658 to change the representation of data for unformatted files.
659 The syntax for the @env{GFORTRAN_CONVERT_UNIT} variable is:
661 GFORTRAN_CONVERT_UNIT: mode | mode ';' exception ;
662 mode: 'native' | 'swap' | 'big_endian' | 'little_endian' ;
663 exception: mode ':' unit_list | unit_list ;
664 unit_list: unit_spec | unit_list unit_spec ;
665 unit_spec: INTEGER | INTEGER '-' INTEGER ;
667 The variable consists of an optional default mode, followed by
668 a list of optional exceptions, which are separated by semicolons
669 from the preceding default and each other. Each exception consists
670 of a format and a comma-separated list of units. Valid values for
671 the modes are the same as for the @code{CONVERT} specifier:
674 @item @code{NATIVE} Use the native format. This is the default.
675 @item @code{SWAP} Swap between little- and big-endian.
676 @item @code{LITTLE_ENDIAN} Use the little-endian format
677 for unformatted files.
678 @item @code{BIG_ENDIAN} Use the big-endian format for unformatted files.
680 A missing mode for an exception is taken to mean @code{BIG_ENDIAN}.
681 Examples of values for @code{GFORTRAN_CONVERT_UNIT} are:
683 @item @code{'big_endian'} Do all unformatted I/O in big_endian mode.
684 @item @code{'little_endian;native:10-20,25'} Do all unformatted I/O
685 in little_endian mode, except for units 10 to 20 and 25, which are in
687 @item @code{'10-20'} Units 10 to 20 are big-endian, the rest is native.
690 Setting the environment variables should be done on the command
691 line or via the @code{export}
692 command for @code{sh}-compatible shells and via @code{setenv}
693 for @code{csh}-compatible shells.
695 Example for @code{sh}:
698 $ GFORTRAN_CONVERT_UNIT='big_endian;native:10-20' ./a.out
701 Example code for @code{csh}:
704 % setenv GFORTRAN_CONVERT_UNIT 'big_endian;native:10-20'
708 Using anything but the native representation for unformatted data
709 carries a significant speed overhead. If speed in this area matters
710 to you, it is best if you use this only for data that needs to be
713 @xref{CONVERT specifier}, for an alternative way to specify the
714 data representation for unformatted files. @xref{Runtime Options}, for
715 setting a default data representation for the whole program. The
716 @code{CONVERT} specifier overrides the @code{-fconvert} compile options.
718 @c ---------------------------------------------------------------------
720 @c ---------------------------------------------------------------------
722 @c Maybe this chapter should be merged with the 'Standards' section,
723 @c whenever that is written :-)
729 @command{gfortran} implements a number of extensions over standard
730 Fortran. This chapter contains information on their syntax and
731 meaning. There are currently two categories of @command{gfortran}
732 extensions, those that provide functionality beyond that provided
733 by any standard, and those that are supported by @command{gfortran}
734 purely for backward compatibility with legacy compilers. By default,
735 @option{-std=gnu} allows the compiler to accept both types of
736 extensions, but to warn about the use of the latter. Specifying
737 either @option{-std=f95} or @option{-std=f2003} disables both types
738 of extensions, and @option{-std=legacy} allows both without warning.
741 * Old-style kind specifications::
742 * Old-style variable initialization::
743 * Extensions to namelist::
744 * X format descriptor::
745 * Commas in FORMAT specifications::
747 * Hexadecimal constants::
748 * Real array indices::
750 * Implicitly interconvert LOGICAL and INTEGER::
751 * Hollerith constants support::
753 * CONVERT specifier::
757 @node Old-style kind specifications
758 @section Old-style kind specifications
759 @cindex Kind specifications
761 @command{gfortran} allows old-style kind specifications in
762 declarations. These look like:
766 where @code{TYPESPEC} is a basic type, and where @code{k} is a valid kind
767 number for that type. The statement then declares @code{x}, @code{y}
768 and @code{z} to be of type @code{TYPESPEC} with kind @code{k}. In
769 other words, it is equivalent to the standard conforming declaration
774 @node Old-style variable initialization
775 @section Old-style variable initialization
776 @cindex Initialization
778 @command{gfortran} allows old-style initialization of variables of the
782 REAL*8 x(2,2) /3*0.,1./
784 These are only allowed in declarations without double colons
785 (@code{::}), as these were introduced in Fortran 90 which also
786 introduced a new syntax for variable initializations. The syntax for
787 the individual initializers is as for the @code{DATA} statement, but
788 unlike in a @code{DATA} statement, an initializer only applies to the
789 variable immediately preceding. In other words, something like
790 @code{INTEGER I,J/2,3/} is not valid.
792 Examples of standard conforming code equivalent to the above example, are:
795 INTEGER(4) :: i = 1, j = 2
796 REAL(8) :: x(2,2) = RESHAPE((/0.,0.,0.,1./),SHAPE(x))
799 DOUBLE PRECISION x(2,2)
800 DATA i,j,x /1,2,3*0.,1./
803 @node Extensions to namelist
804 @section Extensions to namelist
807 @command{gfortran} fully supports the Fortran 95 standard for namelist I/O
808 including array qualifiers, substrings and fully qualified derived types.
809 The output from a namelist write is compatible with namelist read. The
810 output has all names in upper case and indentation to column 1 after the
811 namelist name. Two extensions are permitted:
813 Old-style use of $ instead of &
816 X(:)%Y(2) = 1.0 2.0 3.0
821 It should be noticed that the default terminator is / rather than &END.
823 Querying of the namelist when inputting from stdin. After at least
824 one space, entering ? sends to stdout the namelist name and the names of
825 the variables in the namelist:
836 Entering =? outputs the namelist to stdout, as if WRITE (*,NML = mynml)
842 X(1)%Y= 0.000000 , 1.000000 , 0.000000 ,
843 X(2)%Y= 0.000000 , 2.000000 , 0.000000 ,
844 X(3)%Y= 0.000000 , 3.000000 , 0.000000 ,
848 To aid this dialog, when input is from stdin, errors send their
849 messages to stderr and execution continues, even if IOSTAT is set.
851 PRINT namelist is permitted. This causes an error if -std=f95 is used.
854 REAL, dimension (4) :: x = (/1.0, 2.0, 3.0, 4.0/)
857 END PROGRAM test_print
860 @node X format descriptor
861 @section X format descriptor
862 @cindex X format descriptor
864 To support legacy codes, @command{gfortran} permits the count field
865 of the X edit descriptor in FORMAT statements to be omitted. When
866 omitted, the count is implicitly assumed to be one.
870 10 FORMAT (I1, X, I1)
873 @node Commas in FORMAT specifications
874 @section Commas in FORMAT specifications
875 @cindex Commas in FORMAT specifications
877 To support legacy codes, @command{gfortran} allows the comma separator
878 to be omitted immediately before and after character string edit
879 descriptors in FORMAT statements.
883 10 FORMAT ('FOO='I1' BAR='I2)
887 @section I/O item lists
888 @cindex I/O item lists
890 To support legacy codes, @command{gfortran} allows the input item list
891 of the READ statement, and the output item lists of the WRITE and PRINT
892 statements to start with a comma.
894 @node Hexadecimal constants
895 @section Hexadecimal constants
896 @cindex Hexadecimal constants
898 As a GNU extension, @command{gfortran} allows hexadecimal constants to
899 be specified using the X prefix, in addition to the standard Z prefix.
901 @node Real array indices
902 @section Real array indices
903 @cindex Real array indices
905 As a GNU extension, @command{gfortran} allows arrays to be indexed using
906 real types, whose values are implicitly converted to integers.
908 @node Unary operators
909 @section Unary operators
910 @cindex Unary operators
912 As a GNU extension, @command{gfortran} allows unary plus and unary
913 minus operators to appear as the second operand of binary arithmetic
914 operators without the need for parenthesis.
920 @node Implicitly interconvert LOGICAL and INTEGER
921 @section Implicitly interconvert LOGICAL and INTEGER
922 @cindex Implicitly interconvert LOGICAL and INTEGER
924 As a GNU extension for backwards compatibility with other compilers,
925 @command{gfortran} allows the implicit conversion of LOGICALs to INTEGERs
926 and vice versa. When converting from a LOGICAL to an INTEGER, the numeric
927 value of @code{.FALSE.} is zero, and that of @code{.TRUE.} is one. When
928 converting from INTEGER to LOGICAL, the value zero is interpreted as
929 @code{.FALSE.} and any nonzero value is interpreted as @code{.TRUE.}.
936 @node Hollerith constants support
937 @section Hollerith constants support
938 @cindex Hollerith constants
940 A Hollerith constant is a string of characters preceded by the letter @samp{H}
941 or @samp{h}, and there must be an literal, unsigned, nonzero default integer
942 constant indicating the number of characters in the string. Hollerith constants
943 are stored as byte strings, one character per byte.
945 @command{gfortran} supports Hollerith constants. They can be used as the right
946 hands in the @code{DATA} statement and @code{ASSIGN} statement, also as the
947 arguments. The left hands can be of Integer, Real, Complex and Logical type.
948 The constant will be padded or truncated to fit the size of left hand.
950 Valid Hollerith constants examples:
953 data x /16Habcdefghijklmnop, 16Hqrstuvwxyz012345/
955 x(1) = 16Habcdefghijklmnop
958 Invalid Hollerith constants examples:
961 a = 8H12345678 ! The Hollerith constant is too long. It will be truncated.
962 a = 0H ! At least one character needed.
966 @section Cray pointers
967 @cindex Cray pointers
969 Cray pointers are part of a non-standard extension that provides a
970 C-like pointer in Fortran. This is accomplished through a pair of
971 variables: an integer "pointer" that holds a memory address, and a
972 "pointee" that is used to dereference the pointer.
974 Pointer/pointee pairs are declared in statements of the form:
976 pointer ( <pointer> , <pointee> )
980 pointer ( <pointer1> , <pointee1> ), ( <pointer2> , <pointee2> ), ...
982 The pointer is an integer that is intended to hold a memory address.
983 The pointee may be an array or scalar. A pointee can be an assumed
984 size array -- that is, the last dimension may be left unspecified by
985 using a '*' in place of a value -- but a pointee cannot be an assumed
986 shape array. No space is allocated for the pointee.
988 The pointee may have its type declared before or after the pointer
989 statement, and its array specification (if any) may be declared
990 before, during, or after the pointer statement. The pointer may be
991 declared as an integer prior to the pointer statement. However, some
992 machines have default integer sizes that are different than the size
993 of a pointer, and so the following code is not portable:
998 If a pointer is declared with a kind that is too small, the compiler
999 will issue a warning; the resulting binary will probably not work
1000 correctly, because the memory addresses stored in the pointers may be
1001 truncated. It is safer to omit the first line of the above example;
1002 if explicit declaration of ipt's type is omitted, then the compiler
1003 will ensure that ipt is an integer variable large enough to hold a
1006 Pointer arithmetic is valid with Cray pointers, but it is not the same
1007 as C pointer arithmetic. Cray pointers are just ordinary integers, so
1008 the user is responsible for determining how many bytes to add to a
1009 pointer in order to increment it. Consider the following example:
1013 pointer (ipt, pointee)
1017 The last statement does not set ipt to the address of
1018 @code{target(1)}, as one familiar with C pointer arithmetic might
1019 expect. Adding 1 to ipt just adds one byte to the address stored in
1022 Any expression involving the pointee will be translated to use the
1023 value stored in the pointer as the base address.
1025 To get the address of elements, this extension provides an intrinsic
1026 function loc(), loc() is essentially the C '&' operator, except the
1027 address is cast to an integer type:
1030 pointer(ipt, arpte(10))
1032 ipt = loc(ar) ! Makes arpte is an alias for ar
1033 arpte(1) = 1.0 ! Sets ar(1) to 1.0
1035 The pointer can also be set by a call to a malloc-type
1036 function. There is no malloc intrinsic implemented as part of the
1037 Cray pointer extension, but it might be a useful future addition to
1038 @command{gfortran}. Even without an intrinsic malloc function,
1039 dynamic memory allocation can be combined with Cray pointers by
1040 calling a short C function:
1044 void mymalloc_(void **ptr, int *nbytes)
1046 *ptr = malloc(*nbytes);
1055 pointer (ipdata, data(1024))
1056 call mymalloc(ipdata,4*1024)
1059 Cray pointees often are used to alias an existing variable. For
1067 As long as ipt remains unchanged, iarr is now an alias for target.
1068 The optimizer, however, will not detect this aliasing, so it is unsafe
1069 to use iarr and target simultaneously. Using a pointee in any way
1070 that violates the Fortran aliasing rules or assumptions is illegal.
1071 It is the user's responsibility to avoid doing this; the compiler
1072 works under the assumption that no such aliasing occurs.
1074 Cray pointers will work correctly when there is no aliasing (i.e.,
1075 when they're used to access a dynamically allocated block of memory),
1076 and also in any routine where a pointee is used, but any variable with
1077 which it shares storage is not used. Code that violates these rules
1078 may not run as the user intends. This is not a bug in the optimizer;
1079 any code that violates the aliasing rules is illegal. (Note that this
1080 is not unique to gfortran; any Fortran compiler that supports Cray
1081 pointers will ``incorrectly'' optimize code with illegal aliasing.)
1083 There are a number of restrictions on the attributes that can be
1084 applied to Cray pointers and pointees. Pointees may not have the
1085 attributes ALLOCATABLE, INTENT, OPTIONAL, DUMMY, TARGET, EXTERNAL,
1086 INTRINSIC, or POINTER. Pointers may not have the attributes
1087 DIMENSION, POINTER, TARGET, ALLOCATABLE, EXTERNAL, or INTRINSIC.
1088 Pointees may not occur in more than one pointer statement. A pointee
1089 cannot be a pointer. Pointees cannot occur in equivalence, common, or
1092 A pointer may be modified during the course of a program, and this
1093 will change the location to which the pointee refers. However, when
1094 pointees are passed as arguments, they are treated as ordinary
1095 variables in the invoked function. Subsequent changes to the pointer
1096 will not change the base address of the array that was passed.
1098 @node CONVERT specifier
1099 @section CONVERT specifier
1100 @cindex CONVERT specifier
1102 gfortran allows the conversion of unformatted data between little-
1103 and big-endian representation to facilitate moving of data
1104 between different systems. The conversion can be indicated with
1105 the @code{CONVERT} specifier on the @code{OPEN} statement.
1106 @xref{GFORTRAN_CONVERT_UNIT}, for an alternative way of specifying
1107 the data format via an environment variable.
1109 Valid values for @code{CONVERT} are:
1111 @item @code{CONVERT='NATIVE'} Use the native format. This is the default.
1112 @item @code{CONVERT='SWAP'} Swap between little- and big-endian.
1113 @item @code{CONVERT='LITTLE_ENDIAN'} Use the little-endian representation
1114 for unformatted files.
1115 @item @code{CONVERT='BIG_ENDIAN'} Use the big-endian representation for
1119 Using the option could look like this:
1121 open(file='big.dat',form='unformatted',access='sequential', &
1122 convert='big_endian')
1125 The value of the conversion can be queried by using
1126 @code{INQUIRE(CONVERT=ch)}. The values returned are
1127 @code{'BIG_ENDIAN'} and @code{'LITTLE_ENDIAN'}.
1129 @code{CONVERT} works between big- and little-endian for
1130 @code{INTEGER} values of all supported kinds and for @code{REAL}
1131 on IEEE systems of kinds 4 and 8. Conversion between different
1132 ``extended double'' types on different architectures such as
1133 m68k and x86_64, which gfortran
1134 supports as @code{REAL(KIND=10)} will probably not work.
1136 @emph{Note that the values specified via the GFORTRAN_CONVERT_UNIT
1137 environment variable will override the CONVERT specifier in the
1138 open statement}. This is to give control over data formats to
1139 a user who does not have the source code of his program available.
1141 Using anything but the native representation for unformatted data
1142 carries a significant speed overhead. If speed in this area matters
1143 to you, it is best if you use this only for data that needs to be
1150 gfortran attempts to be OpenMP Application Program Interface v2.5
1151 compatible when invoked with the @code{-fopenmp} option. gfortran
1152 then generates parallellized code according to the OpenMP directives
1153 used in the source. The OpenMP Fortran runtime library
1154 routines are provided both in a form of Fortran 90 module named
1155 @code{omp_lib} and in a form of a Fortran @code{include} file named
1158 For details refer to the actual
1159 @uref{http://www.openmp.org/drupal/mp-documents/spec25.pdf,
1160 OpenMP Application Program Interface v2.5} specification.
1162 @c ---------------------------------------------------------------------
1163 @include intrinsic.texi
1164 @c ---------------------------------------------------------------------
1166 @c ---------------------------------------------------------------------
1168 @c ---------------------------------------------------------------------
1171 @chapter Contributing
1172 @cindex Contributing
1174 Free software is only possible if people contribute to efforts
1176 We're always in need of more people helping out with ideas
1177 and comments, writing documentation and contributing code.
1179 If you want to contribute to GNU Fortran 95,
1180 have a look at the long lists of projects you can take on.
1181 Some of these projects are small,
1182 some of them are large;
1183 some are completely orthogonal to the rest of what is
1184 happening on @command{gfortran},
1185 but others are ``mainstream'' projects in need of enthusiastic hackers.
1186 All of these projects are important!
1187 We'll eventually get around to the things here,
1188 but they are also things doable by someone who is willing and able.
1197 @section Contributors to GNU Fortran 95
1198 @cindex Contributors
1202 Most of the parser was hand-crafted by @emph{Andy Vaught}, who is
1203 also the initiator of the whole project. Thanks Andy!
1204 Most of the interface with GCC was written by @emph{Paul Brook}.
1206 The following individuals have contributed code and/or
1207 ideas and significant help to the gfortran project
1208 (in no particular order):
1212 @item Katherine Holcomb
1213 @item Tobias Schlüter
1214 @item Steven Bosscher
1217 @item Niels Kristian Bech Jensen
1218 @item Steven Johnson
1223 @item François-Xavier Coudert
1226 @item Janne Blomqvist
1232 The following people have contributed bug reports,
1233 smaller or larger patches,
1234 and much needed feedback and encouragement for the
1235 @command{gfortran} project:
1238 @item Erik Schnetter
1243 Many other individuals have helped debug,
1244 test and improve @command{gfortran} over the past few years,
1245 and we welcome you to do the same!
1246 If you already have done so,
1247 and you would like to see your name listed in the
1248 list above, please contact us.
1256 @item Help build the test suite
1257 Solicit more code for donation to the test suite.
1258 We can keep code private on request.
1260 @item Bug hunting/squishing
1261 Find bugs and write more test cases!
1262 Test cases are especially very welcome,
1263 because it allows us to concentrate on fixing bugs
1264 instead of isolating them.
1266 @item Smaller projects (``bug'' fixes):
1268 @item Allow init exprs to be numbers raised to integer powers.
1269 @item Implement correct rounding.
1270 @item Implement F restrictions on Fortran 95 syntax.
1271 @item See about making Emacs-parsable error messages.
1275 If you wish to work on the runtime libraries,
1276 please contact a project maintainer.
1280 @c ---------------------------------------------------------------------
1282 @c ---------------------------------------------------------------------
1288 The GNU Fortran 95 Compiler aims to be a conforming implementation of
1289 ISO/IEC 1539:1997 (Fortran 95).
1291 In the future it may also support other variants of and extensions to
1292 the Fortran language. These include ANSI Fortran 77, ISO Fortran 90,
1293 ISO Fortran 2003 and OpenMP.
1296 * Fortran 2003 status::
1299 @node Fortran 2003 status
1300 @section Fortran 2003 status
1302 Although @command{gfortran} focuses on implementing the Fortran 95
1303 standard for the time being, a few Fortran 2003 features are currently
1308 Intrinsics @code{command_argument_count}, @code{get_command},
1309 @code{get_command_argument}, and @code{get_environment_variable}.
1312 @cindex Array constructors
1313 @cindex @code{[...]}
1314 Array constructors using square brackets. That is, @code{[...]} rather
1315 than @code{(/.../)}.
1318 @cindex @code{FLUSH} statement
1319 @code{FLUSH} statement.
1322 @cindex @code{IOMSG=} specifier
1323 @code{IOMSG=} specifier for I/O statements.
1326 @cindex @code{ENUM} statement
1327 @cindex @code{ENUMERATOR} statement
1328 @cindex @command{-fshort-enums}
1329 Support for the declaration of enumeration constants via the
1330 @code{ENUM} and @code{ENUMERATOR} statements. Interoperability with
1331 @command{gcc} is guaranteed also for the case where the
1332 @command{-fshort-enums} command line option is given.
1337 @c ---------------------------------------------------------------------
1338 @c GNU General Public License
1339 @c ---------------------------------------------------------------------
1345 @c ---------------------------------------------------------------------
1346 @c GNU Free Documentation License
1347 @c ---------------------------------------------------------------------
1353 @c ---------------------------------------------------------------------
1354 @c Funding Free Software
1355 @c ---------------------------------------------------------------------
1357 @include funding.texi
1359 @c ---------------------------------------------------------------------
1361 @c ---------------------------------------------------------------------