1\input texinfo @c -*-texinfo-*- 2@c %**start of header 3@setfilename gfortran.info 4@set copyrights-gfortran 1999-2015 5 6@include gcc-common.texi 7 8@settitle The GNU Fortran Compiler 9 10@c Create a separate index for command line options 11@defcodeindex op 12@c Merge the standard indexes into a single one. 13@syncodeindex fn cp 14@syncodeindex vr cp 15@syncodeindex ky cp 16@syncodeindex pg cp 17@syncodeindex tp cp 18 19@c TODO: The following "Part" definitions are included here temporarily 20@c until they are incorporated into the official Texinfo distribution. 21@c They borrow heavily from Texinfo's \unnchapentry definitions. 22 23@tex 24\gdef\part#1#2{% 25 \pchapsepmacro 26 \gdef\thischapter{} 27 \begingroup 28 \vglue\titlepagetopglue 29 \titlefonts \rm 30 \leftline{Part #1:@* #2} 31 \vskip4pt \hrule height 4pt width \hsize \vskip4pt 32 \endgroup 33 \writetocentry{part}{#2}{#1} 34} 35\gdef\blankpart{% 36 \writetocentry{blankpart}{}{} 37} 38% Part TOC-entry definition for summary contents. 39\gdef\dosmallpartentry#1#2#3#4{% 40 \vskip .5\baselineskip plus.2\baselineskip 41 \begingroup 42 \let\rm=\bf \rm 43 \tocentry{Part #2: #1}{\doshortpageno\bgroup#4\egroup} 44 \endgroup 45} 46\gdef\dosmallblankpartentry#1#2#3#4{% 47 \vskip .5\baselineskip plus.2\baselineskip 48} 49% Part TOC-entry definition for regular contents. This has to be 50% equated to an existing entry to not cause problems when the PDF 51% outline is created. 52\gdef\dopartentry#1#2#3#4{% 53 \unnchapentry{Part #2: #1}{}{#3}{#4} 54} 55\gdef\doblankpartentry#1#2#3#4{} 56@end tex 57 58@c %**end of header 59 60@c Use with @@smallbook. 61 62@c %** start of document 63 64@c Cause even numbered pages to be printed on the left hand side of 65@c the page and odd numbered pages to be printed on the right hand 66@c side of the page. Using this, you can print on both sides of a 67@c sheet of paper and have the text on the same part of the sheet. 68 69@c The text on right hand pages is pushed towards the right hand 70@c margin and the text on left hand pages is pushed toward the left 71@c hand margin. 72@c (To provide the reverse effect, set bindingoffset to -0.75in.) 73 74@c @tex 75@c \global\bindingoffset=0.75in 76@c \global\normaloffset =0.75in 77@c @end tex 78 79@copying 80Copyright @copyright{} @value{copyrights-gfortran} Free Software Foundation, Inc. 81 82Permission is granted to copy, distribute and/or modify this document 83under the terms of the GNU Free Documentation License, Version 1.3 or 84any later version published by the Free Software Foundation; with the 85Invariant Sections being ``Funding Free Software'', the Front-Cover 86Texts being (a) (see below), and with the Back-Cover Texts being (b) 87(see below). A copy of the license is included in the section entitled 88``GNU Free Documentation License''. 89 90(a) The FSF's Front-Cover Text is: 91 92 A GNU Manual 93 94(b) The FSF's Back-Cover Text is: 95 96 You have freedom to copy and modify this GNU Manual, like GNU 97 software. Copies published by the Free Software Foundation raise 98 funds for GNU development. 99@end copying 100 101@ifinfo 102@dircategory Software development 103@direntry 104* gfortran: (gfortran). The GNU Fortran Compiler. 105@end direntry 106This file documents the use and the internals of 107the GNU Fortran compiler, (@command{gfortran}). 108 109Published by the Free Software Foundation 11051 Franklin Street, Fifth Floor 111Boston, MA 02110-1301 USA 112 113@insertcopying 114@end ifinfo 115 116 117@setchapternewpage odd 118@titlepage 119@title Using GNU Fortran 120@versionsubtitle 121@author The @t{gfortran} team 122@page 123@vskip 0pt plus 1filll 124Published by the Free Software Foundation@* 12551 Franklin Street, Fifth Floor@* 126Boston, MA 02110-1301, USA@* 127@c Last printed ??ber, 19??.@* 128@c Printed copies are available for $? each.@* 129@c ISBN ??? 130@sp 1 131@insertcopying 132@end titlepage 133 134@c TODO: The following "Part" definitions are included here temporarily 135@c until they are incorporated into the official Texinfo distribution. 136 137@tex 138\global\let\partentry=\dosmallpartentry 139\global\let\blankpartentry=\dosmallblankpartentry 140@end tex 141@summarycontents 142 143@tex 144\global\let\partentry=\dopartentry 145\global\let\blankpartentry=\doblankpartentry 146@end tex 147@contents 148 149@page 150 151@c --------------------------------------------------------------------- 152@c TexInfo table of contents. 153@c --------------------------------------------------------------------- 154 155@ifnottex 156@node Top 157@top Introduction 158@cindex Introduction 159 160This manual documents the use of @command{gfortran}, 161the GNU Fortran compiler. You can find in this manual how to invoke 162@command{gfortran}, as well as its features and incompatibilities. 163 164@ifset DEVELOPMENT 165@emph{Warning:} This document, and the compiler it describes, are still 166under development. While efforts are made to keep it up-to-date, it might 167not accurately reflect the status of the most recent GNU Fortran compiler. 168@end ifset 169 170@comment 171@comment When you add a new menu item, please keep the right hand 172@comment aligned to the same column. Do not use tabs. This provides 173@comment better formatting. 174@comment 175@menu 176* Introduction:: 177 178Part I: Invoking GNU Fortran 179* Invoking GNU Fortran:: Command options supported by @command{gfortran}. 180* Runtime:: Influencing runtime behavior with environment variables. 181 182Part II: Language Reference 183* Fortran 2003 and 2008 status:: Fortran 2003 and 2008 features supported by GNU Fortran. 184* Compiler Characteristics:: User-visible implementation details. 185* Extensions:: Language extensions implemented by GNU Fortran. 186* Mixed-Language Programming:: Interoperability with C 187* Coarray Programming:: 188* Intrinsic Procedures:: Intrinsic procedures supported by GNU Fortran. 189* Intrinsic Modules:: Intrinsic modules supported by GNU Fortran. 190 191* Contributing:: How you can help. 192* Copying:: GNU General Public License says 193 how you can copy and share GNU Fortran. 194* GNU Free Documentation License:: 195 How you can copy and share this manual. 196* Funding:: How to help assure continued work for free software. 197* Option Index:: Index of command line options 198* Keyword Index:: Index of concepts 199@end menu 200@end ifnottex 201 202@c --------------------------------------------------------------------- 203@c Introduction 204@c --------------------------------------------------------------------- 205 206@node Introduction 207@chapter Introduction 208 209@c The following duplicates the text on the TexInfo table of contents. 210@iftex 211This manual documents the use of @command{gfortran}, the GNU Fortran 212compiler. You can find in this manual how to invoke @command{gfortran}, 213as well as its features and incompatibilities. 214 215@ifset DEVELOPMENT 216@emph{Warning:} This document, and the compiler it describes, are still 217under development. While efforts are made to keep it up-to-date, it 218might not accurately reflect the status of the most recent GNU Fortran 219compiler. 220@end ifset 221@end iftex 222 223The GNU Fortran compiler front end was 224designed initially as a free replacement for, 225or alternative to, the Unix @command{f95} command; 226@command{gfortran} is the command you will use to invoke the compiler. 227 228@menu 229* About GNU Fortran:: What you should know about the GNU Fortran compiler. 230* GNU Fortran and GCC:: You can compile Fortran, C, or other programs. 231* Preprocessing and conditional compilation:: The Fortran preprocessor 232* GNU Fortran and G77:: Why we chose to start from scratch. 233* Project Status:: Status of GNU Fortran, roadmap, proposed extensions. 234* Standards:: Standards supported by GNU Fortran. 235@end menu 236 237 238@c --------------------------------------------------------------------- 239@c About GNU Fortran 240@c --------------------------------------------------------------------- 241 242@node About GNU Fortran 243@section About GNU Fortran 244 245The GNU Fortran compiler supports the Fortran 77, 90 and 95 standards 246completely, parts of the Fortran 2003 and Fortran 2008 standards, and 247several vendor extensions. The development goal is to provide the 248following features: 249 250@itemize @bullet 251@item 252Read a user's program, 253stored in a file and containing instructions written 254in Fortran 77, Fortran 90, Fortran 95, Fortran 2003 or Fortran 2008. 255This file contains @dfn{source code}. 256 257@item 258Translate the user's program into instructions a computer 259can carry out more quickly than it takes to translate the 260instructions in the first 261place. The result after compilation of a program is 262@dfn{machine code}, 263code designed to be efficiently translated and processed 264by a machine such as your computer. 265Humans usually are not as good writing machine code 266as they are at writing Fortran (or C++, Ada, or Java), 267because it is easy to make tiny mistakes writing machine code. 268 269@item 270Provide the user with information about the reasons why 271the compiler is unable to create a binary from the source code. 272Usually this will be the case if the source code is flawed. 273The Fortran 90 standard requires that the compiler can point out 274mistakes to the user. 275An incorrect usage of the language causes an @dfn{error message}. 276 277The compiler will also attempt to diagnose cases where the 278user's program contains a correct usage of the language, 279but instructs the computer to do something questionable. 280This kind of diagnostics message is called a @dfn{warning message}. 281 282@item 283Provide optional information about the translation passes 284from the source code to machine code. 285This can help a user of the compiler to find the cause of 286certain bugs which may not be obvious in the source code, 287but may be more easily found at a lower level compiler output. 288It also helps developers to find bugs in the compiler itself. 289 290@item 291Provide information in the generated machine code that can 292make it easier to find bugs in the program (using a debugging tool, 293called a @dfn{debugger}, such as the GNU Debugger @command{gdb}). 294 295@item 296Locate and gather machine code already generated to 297perform actions requested by statements in the user's program. 298This machine code is organized into @dfn{modules} and is located 299and @dfn{linked} to the user program. 300@end itemize 301 302The GNU Fortran compiler consists of several components: 303 304@itemize @bullet 305@item 306A version of the @command{gcc} command 307(which also might be installed as the system's @command{cc} command) 308that also understands and accepts Fortran source code. 309The @command{gcc} command is the @dfn{driver} program for 310all the languages in the GNU Compiler Collection (GCC); 311With @command{gcc}, 312you can compile the source code of any language for 313which a front end is available in GCC. 314 315@item 316The @command{gfortran} command itself, 317which also might be installed as the 318system's @command{f95} command. 319@command{gfortran} is just another driver program, 320but specifically for the Fortran compiler only. 321The difference with @command{gcc} is that @command{gfortran} 322will automatically link the correct libraries to your program. 323 324@item 325A collection of run-time libraries. 326These libraries contain the machine code needed to support 327capabilities of the Fortran language that are not directly 328provided by the machine code generated by the 329@command{gfortran} compilation phase, 330such as intrinsic functions and subroutines, 331and routines for interaction with files and the operating system. 332@c and mechanisms to spawn, 333@c unleash and pause threads in parallelized code. 334 335@item 336The Fortran compiler itself, (@command{f951}). 337This is the GNU Fortran parser and code generator, 338linked to and interfaced with the GCC backend library. 339@command{f951} ``translates'' the source code to 340assembler code. You would typically not use this 341program directly; 342instead, the @command{gcc} or @command{gfortran} driver 343programs will call it for you. 344@end itemize 345 346 347@c --------------------------------------------------------------------- 348@c GNU Fortran and GCC 349@c --------------------------------------------------------------------- 350 351@node GNU Fortran and GCC 352@section GNU Fortran and GCC 353@cindex GNU Compiler Collection 354@cindex GCC 355 356GNU Fortran is a part of GCC, the @dfn{GNU Compiler Collection}. GCC 357consists of a collection of front ends for various languages, which 358translate the source code into a language-independent form called 359@dfn{GENERIC}. This is then processed by a common middle end which 360provides optimization, and then passed to one of a collection of back 361ends which generate code for different computer architectures and 362operating systems. 363 364Functionally, this is implemented with a driver program (@command{gcc}) 365which provides the command-line interface for the compiler. It calls 366the relevant compiler front-end program (e.g., @command{f951} for 367Fortran) for each file in the source code, and then calls the assembler 368and linker as appropriate to produce the compiled output. In a copy of 369GCC which has been compiled with Fortran language support enabled, 370@command{gcc} will recognize files with @file{.f}, @file{.for}, @file{.ftn}, 371@file{.f90}, @file{.f95}, @file{.f03} and @file{.f08} extensions as 372Fortran source code, and compile it accordingly. A @command{gfortran} 373driver program is also provided, which is identical to @command{gcc} 374except that it automatically links the Fortran runtime libraries into the 375compiled program. 376 377Source files with @file{.f}, @file{.for}, @file{.fpp}, @file{.ftn}, @file{.F}, 378@file{.FOR}, @file{.FPP}, and @file{.FTN} extensions are treated as fixed form. 379Source files with @file{.f90}, @file{.f95}, @file{.f03}, @file{.f08}, 380@file{.F90}, @file{.F95}, @file{.F03} and @file{.F08} extensions are 381treated as free form. The capitalized versions of either form are run 382through preprocessing. Source files with the lower case @file{.fpp} 383extension are also run through preprocessing. 384 385This manual specifically documents the Fortran front end, which handles 386the programming language's syntax and semantics. The aspects of GCC 387which relate to the optimization passes and the back-end code generation 388are documented in the GCC manual; see 389@ref{Top,,Introduction,gcc,Using the GNU Compiler Collection (GCC)}. 390The two manuals together provide a complete reference for the GNU 391Fortran compiler. 392 393 394@c --------------------------------------------------------------------- 395@c Preprocessing and conditional compilation 396@c --------------------------------------------------------------------- 397 398@node Preprocessing and conditional compilation 399@section Preprocessing and conditional compilation 400@cindex CPP 401@cindex FPP 402@cindex Conditional compilation 403@cindex Preprocessing 404@cindex preprocessor, include file handling 405 406Many Fortran compilers including GNU Fortran allow passing the source code 407through a C preprocessor (CPP; sometimes also called the Fortran preprocessor, 408FPP) to allow for conditional compilation. In the case of GNU Fortran, 409this is the GNU C Preprocessor in the traditional mode. On systems with 410case-preserving file names, the preprocessor is automatically invoked if the 411filename extension is @file{.F}, @file{.FOR}, @file{.FTN}, @file{.fpp}, 412@file{.FPP}, @file{.F90}, @file{.F95}, @file{.F03} or @file{.F08}. To manually 413invoke the preprocessor on any file, use @option{-cpp}, to disable 414preprocessing on files where the preprocessor is run automatically, use 415@option{-nocpp}. 416 417If a preprocessed file includes another file with the Fortran @code{INCLUDE} 418statement, the included file is not preprocessed. To preprocess included 419files, use the equivalent preprocessor statement @code{#include}. 420 421If GNU Fortran invokes the preprocessor, @code{__GFORTRAN__} 422is defined and @code{__GNUC__}, @code{__GNUC_MINOR__} and 423@code{__GNUC_PATCHLEVEL__} can be used to determine the version of the 424compiler. See @ref{Top,,Overview,cpp,The C Preprocessor} for details. 425 426While CPP is the de-facto standard for preprocessing Fortran code, 427Part 3 of the Fortran 95 standard (ISO/IEC 1539-3:1998) defines 428Conditional Compilation, which is not widely used and not directly 429supported by the GNU Fortran compiler. You can use the program coco 430to preprocess such files (@uref{http://www.daniellnagle.com/coco.html}). 431 432 433@c --------------------------------------------------------------------- 434@c GNU Fortran and G77 435@c --------------------------------------------------------------------- 436 437@node GNU Fortran and G77 438@section GNU Fortran and G77 439@cindex Fortran 77 440@cindex @command{g77} 441 442The GNU Fortran compiler is the successor to @command{g77}, the Fortran 44377 front end included in GCC prior to version 4. It is an entirely new 444program that has been designed to provide Fortran 95 support and 445extensibility for future Fortran language standards, as well as providing 446backwards compatibility for Fortran 77 and nearly all of the GNU language 447extensions supported by @command{g77}. 448 449 450@c --------------------------------------------------------------------- 451@c Project Status 452@c --------------------------------------------------------------------- 453 454@node Project Status 455@section Project Status 456 457@quotation 458As soon as @command{gfortran} can parse all of the statements correctly, 459it will be in the ``larva'' state. 460When we generate code, the ``puppa'' state. 461When @command{gfortran} is done, 462we'll see if it will be a beautiful butterfly, 463or just a big bug.... 464 465--Andy Vaught, April 2000 466@end quotation 467 468The start of the GNU Fortran 95 project was announced on 469the GCC homepage in March 18, 2000 470(even though Andy had already been working on it for a while, 471of course). 472 473The GNU Fortran compiler is able to compile nearly all 474standard-compliant Fortran 95, Fortran 90, and Fortran 77 programs, 475including a number of standard and non-standard extensions, and can be 476used on real-world programs. In particular, the supported extensions 477include OpenMP, Cray-style pointers, and several Fortran 2003 and Fortran 4782008 features, including TR 15581. However, it is still under 479development and has a few remaining rough edges. 480There also is initial support for OpenACC. 481Note that this is an experimental feature, incomplete, and subject to 482change in future versions of GCC. See 483@uref{https://gcc.gnu.org/wiki/OpenACC} for more information. 484 485At present, the GNU Fortran compiler passes the 486@uref{http://www.fortran-2000.com/ArnaudRecipes/fcvs21_f95.html, 487NIST Fortran 77 Test Suite}, and produces acceptable results on the 488@uref{http://www.netlib.org/lapack/faq.html#1.21, LAPACK Test Suite}. 489It also provides respectable performance on 490the @uref{http://www.polyhedron.com/fortran-compiler-comparisons/polyhedron-benchmark-suite, 491Polyhedron Fortran 492compiler benchmarks} and the 493@uref{http://www.netlib.org/benchmark/livermore, 494Livermore Fortran Kernels test}. It has been used to compile a number of 495large real-world programs, including 496@uref{http://hirlam.org/, the HARMONIE and HIRLAM weather forecasting code} and 497@uref{http://physical-chemistry.scb.uwa.edu.au/tonto/wiki/index.php/Main_Page, 498the Tonto quantum chemistry package}; see 499@url{https://gcc.gnu.org/@/wiki/@/GfortranApps} for an extended list. 500 501Among other things, the GNU Fortran compiler is intended as a replacement 502for G77. At this point, nearly all programs that could be compiled with 503G77 can be compiled with GNU Fortran, although there are a few minor known 504regressions. 505 506The primary work remaining to be done on GNU Fortran falls into three 507categories: bug fixing (primarily regarding the treatment of invalid code 508and providing useful error messages), improving the compiler optimizations 509and the performance of compiled code, and extending the compiler to support 510future standards---in particular, Fortran 2003 and Fortran 2008. 511 512 513@c --------------------------------------------------------------------- 514@c Standards 515@c --------------------------------------------------------------------- 516 517@node Standards 518@section Standards 519@cindex Standards 520 521@menu 522* Varying Length Character Strings:: 523@end menu 524 525The GNU Fortran compiler implements 526ISO/IEC 1539:1997 (Fortran 95). As such, it can also compile essentially all 527standard-compliant Fortran 90 and Fortran 77 programs. It also supports 528the ISO/IEC TR-15581 enhancements to allocatable arrays. 529 530GNU Fortran also have a partial support for ISO/IEC 1539-1:2004 (Fortran 5312003), ISO/IEC 1539-1:2010 (Fortran 2008), the Technical Specification 532@code{Further Interoperability of Fortran with C} (ISO/IEC TS 29113:2012). 533Full support of those standards and future Fortran standards is planned. 534The current status of the support is can be found in the 535@ref{Fortran 2003 status}, @ref{Fortran 2008 status} and 536@ref{TS 29113 status} sections of the documentation. 537 538Additionally, the GNU Fortran compilers supports the OpenMP specification 539(version 4.0, @url{http://openmp.org/@/wp/@/openmp-specifications/}). 540There also is initial support for the OpenACC specification (targeting 541version 2.0, @uref{http://www.openacc.org/}). 542Note that this is an experimental feature, incomplete, and subject to 543change in future versions of GCC. See 544@uref{https://gcc.gnu.org/wiki/OpenACC} for more information. 545 546@node Varying Length Character Strings 547@subsection Varying Length Character Strings 548@cindex Varying length character strings 549@cindex Varying length strings 550@cindex strings, varying length 551 552The Fortran 95 standard specifies in Part 2 (ISO/IEC 1539-2:2000) 553varying length character strings. While GNU Fortran currently does not 554support such strings directly, there exist two Fortran implementations 555for them, which work with GNU Fortran. They can be found at 556@uref{http://www.fortran.com/@/iso_varying_string.f95} and at 557@uref{ftp://ftp.nag.co.uk/@/sc22wg5/@/ISO_VARYING_STRING/}. 558 559Deferred-length character strings of Fortran 2003 supports part of 560the features of @code{ISO_VARYING_STRING} and should be considered as 561replacement. (Namely, allocatable or pointers of the type 562@code{character(len=:)}.) 563 564 565@c ===================================================================== 566@c PART I: INVOCATION REFERENCE 567@c ===================================================================== 568 569@tex 570\part{I}{Invoking GNU Fortran} 571@end tex 572 573@c --------------------------------------------------------------------- 574@c Compiler Options 575@c --------------------------------------------------------------------- 576 577@include invoke.texi 578 579 580@c --------------------------------------------------------------------- 581@c Runtime 582@c --------------------------------------------------------------------- 583 584@node Runtime 585@chapter Runtime: Influencing runtime behavior with environment variables 586@cindex environment variable 587 588The behavior of the @command{gfortran} can be influenced by 589environment variables. 590 591Malformed environment variables are silently ignored. 592 593@menu 594* TMPDIR:: Directory for scratch files 595* GFORTRAN_STDIN_UNIT:: Unit number for standard input 596* GFORTRAN_STDOUT_UNIT:: Unit number for standard output 597* GFORTRAN_STDERR_UNIT:: Unit number for standard error 598* GFORTRAN_UNBUFFERED_ALL:: Do not buffer I/O for all units. 599* GFORTRAN_UNBUFFERED_PRECONNECTED:: Do not buffer I/O for preconnected units. 600* GFORTRAN_SHOW_LOCUS:: Show location for runtime errors 601* GFORTRAN_OPTIONAL_PLUS:: Print leading + where permitted 602* GFORTRAN_DEFAULT_RECL:: Default record length for new files 603* GFORTRAN_LIST_SEPARATOR:: Separator for list output 604* GFORTRAN_CONVERT_UNIT:: Set endianness for unformatted I/O 605* GFORTRAN_ERROR_BACKTRACE:: Show backtrace on run-time errors 606@end menu 607 608@node TMPDIR 609@section @env{TMPDIR}---Directory for scratch files 610 611When opening a file with @code{STATUS='SCRATCH'}, GNU Fortran tries to 612create the file in one of the potential directories by testing each 613directory in the order below. 614 615@enumerate 616@item 617The environment variable @env{TMPDIR}, if it exists. 618 619@item 620On the MinGW target, the directory returned by the @code{GetTempPath} 621function. Alternatively, on the Cygwin target, the @env{TMP} and 622@env{TEMP} environment variables, if they exist, in that order. 623 624@item 625The @code{P_tmpdir} macro if it is defined, otherwise the directory 626@file{/tmp}. 627@end enumerate 628 629@node GFORTRAN_STDIN_UNIT 630@section @env{GFORTRAN_STDIN_UNIT}---Unit number for standard input 631 632This environment variable can be used to select the unit number 633preconnected to standard input. This must be a positive integer. 634The default value is 5. 635 636@node GFORTRAN_STDOUT_UNIT 637@section @env{GFORTRAN_STDOUT_UNIT}---Unit number for standard output 638 639This environment variable can be used to select the unit number 640preconnected to standard output. This must be a positive integer. 641The default value is 6. 642 643@node GFORTRAN_STDERR_UNIT 644@section @env{GFORTRAN_STDERR_UNIT}---Unit number for standard error 645 646This environment variable can be used to select the unit number 647preconnected to standard error. This must be a positive integer. 648The default value is 0. 649 650@node GFORTRAN_UNBUFFERED_ALL 651@section @env{GFORTRAN_UNBUFFERED_ALL}---Do not buffer I/O on all units 652 653This environment variable controls whether all I/O is unbuffered. If 654the first letter is @samp{y}, @samp{Y} or @samp{1}, all I/O is 655unbuffered. This will slow down small sequential reads and writes. If 656the first letter is @samp{n}, @samp{N} or @samp{0}, I/O is buffered. 657This is the default. 658 659@node GFORTRAN_UNBUFFERED_PRECONNECTED 660@section @env{GFORTRAN_UNBUFFERED_PRECONNECTED}---Do not buffer I/O on preconnected units 661 662The environment variable named @env{GFORTRAN_UNBUFFERED_PRECONNECTED} controls 663whether I/O on a preconnected unit (i.e.@: STDOUT or STDERR) is unbuffered. If 664the first letter is @samp{y}, @samp{Y} or @samp{1}, I/O is unbuffered. This 665will slow down small sequential reads and writes. If the first letter 666is @samp{n}, @samp{N} or @samp{0}, I/O is buffered. This is the default. 667 668@node GFORTRAN_SHOW_LOCUS 669@section @env{GFORTRAN_SHOW_LOCUS}---Show location for runtime errors 670 671If the first letter is @samp{y}, @samp{Y} or @samp{1}, filename and 672line numbers for runtime errors are printed. If the first letter is 673@samp{n}, @samp{N} or @samp{0}, do not print filename and line numbers 674for runtime errors. The default is to print the location. 675 676@node GFORTRAN_OPTIONAL_PLUS 677@section @env{GFORTRAN_OPTIONAL_PLUS}---Print leading + where permitted 678 679If the first letter is @samp{y}, @samp{Y} or @samp{1}, 680a plus sign is printed 681where permitted by the Fortran standard. If the first letter 682is @samp{n}, @samp{N} or @samp{0}, a plus sign is not printed 683in most cases. Default is not to print plus signs. 684 685@node GFORTRAN_DEFAULT_RECL 686@section @env{GFORTRAN_DEFAULT_RECL}---Default record length for new files 687 688This environment variable specifies the default record length, in 689bytes, for files which are opened without a @code{RECL} tag in the 690@code{OPEN} statement. This must be a positive integer. The 691default value is 1073741824 bytes (1 GB). 692 693@node GFORTRAN_LIST_SEPARATOR 694@section @env{GFORTRAN_LIST_SEPARATOR}---Separator for list output 695 696This environment variable specifies the separator when writing 697list-directed output. It may contain any number of spaces and 698at most one comma. If you specify this on the command line, 699be sure to quote spaces, as in 700@smallexample 701$ GFORTRAN_LIST_SEPARATOR=' , ' ./a.out 702@end smallexample 703when @command{a.out} is the compiled Fortran program that you want to run. 704Default is a single space. 705 706@node GFORTRAN_CONVERT_UNIT 707@section @env{GFORTRAN_CONVERT_UNIT}---Set endianness for unformatted I/O 708 709By setting the @env{GFORTRAN_CONVERT_UNIT} variable, it is possible 710to change the representation of data for unformatted files. 711The syntax for the @env{GFORTRAN_CONVERT_UNIT} variable is: 712@smallexample 713GFORTRAN_CONVERT_UNIT: mode | mode ';' exception | exception ; 714mode: 'native' | 'swap' | 'big_endian' | 'little_endian' ; 715exception: mode ':' unit_list | unit_list ; 716unit_list: unit_spec | unit_list unit_spec ; 717unit_spec: INTEGER | INTEGER '-' INTEGER ; 718@end smallexample 719The variable consists of an optional default mode, followed by 720a list of optional exceptions, which are separated by semicolons 721from the preceding default and each other. Each exception consists 722of a format and a comma-separated list of units. Valid values for 723the modes are the same as for the @code{CONVERT} specifier: 724 725@itemize @w{} 726@item @code{NATIVE} Use the native format. This is the default. 727@item @code{SWAP} Swap between little- and big-endian. 728@item @code{LITTLE_ENDIAN} Use the little-endian format 729for unformatted files. 730@item @code{BIG_ENDIAN} Use the big-endian format for unformatted files. 731@end itemize 732A missing mode for an exception is taken to mean @code{BIG_ENDIAN}. 733Examples of values for @env{GFORTRAN_CONVERT_UNIT} are: 734@itemize @w{} 735@item @code{'big_endian'} Do all unformatted I/O in big_endian mode. 736@item @code{'little_endian;native:10-20,25'} Do all unformatted I/O 737in little_endian mode, except for units 10 to 20 and 25, which are in 738native format. 739@item @code{'10-20'} Units 10 to 20 are big-endian, the rest is native. 740@end itemize 741 742Setting the environment variables should be done on the command 743line or via the @command{export} 744command for @command{sh}-compatible shells and via @command{setenv} 745for @command{csh}-compatible shells. 746 747Example for @command{sh}: 748@smallexample 749$ gfortran foo.f90 750$ GFORTRAN_CONVERT_UNIT='big_endian;native:10-20' ./a.out 751@end smallexample 752 753Example code for @command{csh}: 754@smallexample 755% gfortran foo.f90 756% setenv GFORTRAN_CONVERT_UNIT 'big_endian;native:10-20' 757% ./a.out 758@end smallexample 759 760Using anything but the native representation for unformatted data 761carries a significant speed overhead. If speed in this area matters 762to you, it is best if you use this only for data that needs to be 763portable. 764 765@xref{CONVERT specifier}, for an alternative way to specify the 766data representation for unformatted files. @xref{Runtime Options}, for 767setting a default data representation for the whole program. The 768@code{CONVERT} specifier overrides the @option{-fconvert} compile options. 769 770@emph{Note that the values specified via the GFORTRAN_CONVERT_UNIT 771environment variable will override the CONVERT specifier in the 772open statement}. This is to give control over data formats to 773users who do not have the source code of their program available. 774 775@node GFORTRAN_ERROR_BACKTRACE 776@section @env{GFORTRAN_ERROR_BACKTRACE}---Show backtrace on run-time errors 777 778If the @env{GFORTRAN_ERROR_BACKTRACE} variable is set to @samp{y}, 779@samp{Y} or @samp{1} (only the first letter is relevant) then a 780backtrace is printed when a serious run-time error occurs. To disable 781the backtracing, set the variable to @samp{n}, @samp{N}, @samp{0}. 782Default is to print a backtrace unless the @option{-fno-backtrace} 783compile option was used. 784 785@c ===================================================================== 786@c PART II: LANGUAGE REFERENCE 787@c ===================================================================== 788 789@tex 790\part{II}{Language Reference} 791@end tex 792 793@c --------------------------------------------------------------------- 794@c Fortran 2003 and 2008 Status 795@c --------------------------------------------------------------------- 796 797@node Fortran 2003 and 2008 status 798@chapter Fortran 2003 and 2008 Status 799 800@menu 801* Fortran 2003 status:: 802* Fortran 2008 status:: 803* TS 29113 status:: 804@end menu 805 806@node Fortran 2003 status 807@section Fortran 2003 status 808 809GNU Fortran supports several Fortran 2003 features; an incomplete 810list can be found below. See also the 811@uref{https://gcc.gnu.org/wiki/Fortran2003, wiki page} about Fortran 2003. 812 813@itemize 814@item Procedure pointers including procedure-pointer components with 815@code{PASS} attribute. 816 817@item Procedures which are bound to a derived type (type-bound procedures) 818including @code{PASS}, @code{PROCEDURE} and @code{GENERIC}, and 819operators bound to a type. 820 821@item Abstract interfaces and type extension with the possibility to 822override type-bound procedures or to have deferred binding. 823 824@item Polymorphic entities (``@code{CLASS}'') for derived types and unlimited 825polymorphism (``@code{CLASS(*)}'') -- including @code{SAME_TYPE_AS}, 826@code{EXTENDS_TYPE_OF} and @code{SELECT TYPE} for scalars and arrays and 827finalization. 828 829@item Generic interface names, which have the same name as derived types, 830are now supported. This allows one to write constructor functions. Note 831that Fortran does not support static constructor functions. For static 832variables, only default initialization or structure-constructor 833initialization are available. 834 835@item The @code{ASSOCIATE} construct. 836 837@item Interoperability with C including enumerations, 838 839@item In structure constructors the components with default values may be 840omitted. 841 842@item Extensions to the @code{ALLOCATE} statement, allowing for a 843type-specification with type parameter and for allocation and initialization 844from a @code{SOURCE=} expression; @code{ALLOCATE} and @code{DEALLOCATE} 845optionally return an error message string via @code{ERRMSG=}. 846 847@item Reallocation on assignment: If an intrinsic assignment is 848used, an allocatable variable on the left-hand side is automatically allocated 849(if unallocated) or reallocated (if the shape is different). Currently, scalar 850deferred character length left-hand sides are correctly handled but arrays 851are not yet fully implemented. 852 853@item Deferred-length character variables and scalar deferred-length character 854components of derived types are supported. (Note that array-valued compoents 855are not yet implemented.) 856 857@item Transferring of allocations via @code{MOVE_ALLOC}. 858 859@item The @code{PRIVATE} and @code{PUBLIC} attributes may be given individually 860to derived-type components. 861 862@item In pointer assignments, the lower bound may be specified and 863the remapping of elements is supported. 864 865@item For pointers an @code{INTENT} may be specified which affect the 866association status not the value of the pointer target. 867 868@item Intrinsics @code{command_argument_count}, @code{get_command}, 869@code{get_command_argument}, and @code{get_environment_variable}. 870 871@item Support for Unicode characters (ISO 10646) and UTF-8, including 872the @code{SELECTED_CHAR_KIND} and @code{NEW_LINE} intrinsic functions. 873 874@item Support for binary, octal and hexadecimal (BOZ) constants in the 875intrinsic functions @code{INT}, @code{REAL}, @code{CMPLX} and @code{DBLE}. 876 877@item Support for namelist variables with allocatable and pointer 878attribute and nonconstant length type parameter. 879 880@item 881@cindex array, constructors 882@cindex @code{[...]} 883Array constructors using square brackets. That is, @code{[...]} rather 884than @code{(/.../)}. Type-specification for array constructors like 885@code{(/ some-type :: ... /)}. 886 887@item Extensions to the specification and initialization expressions, 888including the support for intrinsics with real and complex arguments. 889 890@item Support for the asynchronous input/output syntax; however, the 891data transfer is currently always synchronously performed. 892 893@item 894@cindex @code{FLUSH} statement 895@cindex statement, @code{FLUSH} 896@code{FLUSH} statement. 897 898@item 899@cindex @code{IOMSG=} specifier 900@code{IOMSG=} specifier for I/O statements. 901 902@item 903@cindex @code{ENUM} statement 904@cindex @code{ENUMERATOR} statement 905@cindex statement, @code{ENUM} 906@cindex statement, @code{ENUMERATOR} 907@opindex @code{fshort-enums} 908Support for the declaration of enumeration constants via the 909@code{ENUM} and @code{ENUMERATOR} statements. Interoperability with 910@command{gcc} is guaranteed also for the case where the 911@command{-fshort-enums} command line option is given. 912 913@item 914@cindex TR 15581 915TR 15581: 916@itemize 917@item 918@cindex @code{ALLOCATABLE} dummy arguments 919@code{ALLOCATABLE} dummy arguments. 920@item 921@cindex @code{ALLOCATABLE} function results 922@code{ALLOCATABLE} function results 923@item 924@cindex @code{ALLOCATABLE} components of derived types 925@code{ALLOCATABLE} components of derived types 926@end itemize 927 928@item 929@cindex @code{STREAM} I/O 930@cindex @code{ACCESS='STREAM'} I/O 931The @code{OPEN} statement supports the @code{ACCESS='STREAM'} specifier, 932allowing I/O without any record structure. 933 934@item 935Namelist input/output for internal files. 936 937@item Minor I/O features: Rounding during formatted output, using of 938a decimal comma instead of a decimal point, setting whether a plus sign 939should appear for positive numbers. On systems where @code{strtod} honours 940the rounding mode, the rounding mode is also supported for input. 941 942@item 943@cindex @code{PROTECTED} statement 944@cindex statement, @code{PROTECTED} 945The @code{PROTECTED} statement and attribute. 946 947@item 948@cindex @code{VALUE} statement 949@cindex statement, @code{VALUE} 950The @code{VALUE} statement and attribute. 951 952@item 953@cindex @code{VOLATILE} statement 954@cindex statement, @code{VOLATILE} 955The @code{VOLATILE} statement and attribute. 956 957@item 958@cindex @code{IMPORT} statement 959@cindex statement, @code{IMPORT} 960The @code{IMPORT} statement, allowing to import 961host-associated derived types. 962 963@item The intrinsic modules @code{ISO_FORTRAN_ENVIRONMENT} is supported, 964which contains parameters of the I/O units, storage sizes. Additionally, 965procedures for C interoperability are available in the @code{ISO_C_BINDING} 966module. 967 968@item 969@cindex @code{USE, INTRINSIC} statement 970@cindex statement, @code{USE, INTRINSIC} 971@cindex @code{ISO_FORTRAN_ENV} statement 972@cindex statement, @code{ISO_FORTRAN_ENV} 973@code{USE} statement with @code{INTRINSIC} and @code{NON_INTRINSIC} 974attribute; supported intrinsic modules: @code{ISO_FORTRAN_ENV}, 975@code{ISO_C_BINDING}, @code{OMP_LIB} and @code{OMP_LIB_KINDS}, 976and @code{OPENACC}. 977 978@item 979Renaming of operators in the @code{USE} statement. 980 981@end itemize 982 983 984@node Fortran 2008 status 985@section Fortran 2008 status 986 987The latest version of the Fortran standard is ISO/IEC 1539-1:2010, informally 988known as Fortran 2008. The official version is available from International 989Organization for Standardization (ISO) or its national member organizations. 990The the final draft (FDIS) can be downloaded free of charge from 991@url{http://www.nag.co.uk/@/sc22wg5/@/links.html}. Fortran is developed by the 992Working Group 5 of Sub-Committee 22 of the Joint Technical Committee 1 of the 993International Organization for Standardization and the International 994Electrotechnical Commission (IEC). This group is known as 995@uref{http://www.nag.co.uk/sc22wg5/, WG5}. 996 997The GNU Fortran compiler supports several of the new features of Fortran 2008; 998the @uref{https://gcc.gnu.org/wiki/Fortran2008Status, wiki} has some information 999about the current Fortran 2008 implementation status. In particular, the 1000following is implemented. 1001 1002@itemize 1003@item The @option{-std=f2008} option and support for the file extensions 1004@file{.f08} and @file{.F08}. 1005 1006@item The @code{OPEN} statement now supports the @code{NEWUNIT=} option, 1007which returns a unique file unit, thus preventing inadvertent use of the 1008same unit in different parts of the program. 1009 1010@item The @code{g0} format descriptor and unlimited format items. 1011 1012@item The mathematical intrinsics @code{ASINH}, @code{ACOSH}, @code{ATANH}, 1013@code{ERF}, @code{ERFC}, @code{GAMMA}, @code{LOG_GAMMA}, @code{BESSEL_J0}, 1014@code{BESSEL_J1}, @code{BESSEL_JN}, @code{BESSEL_Y0}, @code{BESSEL_Y1}, 1015@code{BESSEL_YN}, @code{HYPOT}, @code{NORM2}, and @code{ERFC_SCALED}. 1016 1017@item Using complex arguments with @code{TAN}, @code{SINH}, @code{COSH}, 1018@code{TANH}, @code{ASIN}, @code{ACOS}, and @code{ATAN} is now possible; 1019@code{ATAN}(@var{Y},@var{X}) is now an alias for @code{ATAN2}(@var{Y},@var{X}). 1020 1021@item Support of the @code{PARITY} intrinsic functions. 1022 1023@item The following bit intrinsics: @code{LEADZ} and @code{TRAILZ} for 1024counting the number of leading and trailing zero bits, @code{POPCNT} and 1025@code{POPPAR} for counting the number of one bits and returning the parity; 1026@code{BGE}, @code{BGT}, @code{BLE}, and @code{BLT} for bitwise comparisons; 1027@code{DSHIFTL} and @code{DSHIFTR} for combined left and right shifts, 1028@code{MASKL} and @code{MASKR} for simple left and right justified masks, 1029@code{MERGE_BITS} for a bitwise merge using a mask, @code{SHIFTA}, 1030@code{SHIFTL} and @code{SHIFTR} for shift operations, and the 1031transformational bit intrinsics @code{IALL}, @code{IANY} and @code{IPARITY}. 1032 1033@item Support of the @code{EXECUTE_COMMAND_LINE} intrinsic subroutine. 1034 1035@item Support for the @code{STORAGE_SIZE} intrinsic inquiry function. 1036 1037@item The @code{INT@{8,16,32@}} and @code{REAL@{32,64,128@}} kind type 1038parameters and the array-valued named constants @code{INTEGER_KINDS}, 1039@code{LOGICAL_KINDS}, @code{REAL_KINDS} and @code{CHARACTER_KINDS} of 1040the intrinsic module @code{ISO_FORTRAN_ENV}. 1041 1042@item The module procedures @code{C_SIZEOF} of the intrinsic module 1043@code{ISO_C_BINDINGS} and @code{COMPILER_VERSION} and @code{COMPILER_OPTIONS} 1044of @code{ISO_FORTRAN_ENV}. 1045 1046@item Coarray support for serial programs with @option{-fcoarray=single} flag 1047and experimental support for multiple images with the @option{-fcoarray=lib} 1048flag. 1049 1050@item The @code{DO CONCURRENT} construct is supported. 1051 1052@item The @code{BLOCK} construct is supported. 1053 1054@item The @code{STOP} and the new @code{ERROR STOP} statements now 1055support all constant expressions. Both show the signals which were signaling 1056at termination. 1057 1058@item Support for the @code{CONTIGUOUS} attribute. 1059 1060@item Support for @code{ALLOCATE} with @code{MOLD}. 1061 1062@item Support for the @code{IMPURE} attribute for procedures, which 1063allows for @code{ELEMENTAL} procedures without the restrictions of 1064@code{PURE}. 1065 1066@item Null pointers (including @code{NULL()}) and not-allocated variables 1067can be used as actual argument to optional non-pointer, non-allocatable 1068dummy arguments, denoting an absent argument. 1069 1070@item Non-pointer variables with @code{TARGET} attribute can be used as 1071actual argument to @code{POINTER} dummies with @code{INTENT(IN)}. 1072 1073@item Pointers including procedure pointers and those in a derived 1074type (pointer components) can now be initialized by a target instead 1075of only by @code{NULL}. 1076 1077@item The @code{EXIT} statement (with construct-name) can be now be 1078used to leave not only the @code{DO} but also the @code{ASSOCIATE}, 1079@code{BLOCK}, @code{IF}, @code{SELECT CASE} and @code{SELECT TYPE} 1080constructs. 1081 1082@item Internal procedures can now be used as actual argument. 1083 1084@item Minor features: obsolesce diagnostics for @code{ENTRY} with 1085@option{-std=f2008}; a line may start with a semicolon; for internal 1086and module procedures @code{END} can be used instead of 1087@code{END SUBROUTINE} and @code{END FUNCTION}; @code{SELECTED_REAL_KIND} 1088now also takes a @code{RADIX} argument; intrinsic types are supported 1089for @code{TYPE}(@var{intrinsic-type-spec}); multiple type-bound procedures 1090can be declared in a single @code{PROCEDURE} statement; implied-shape 1091arrays are supported for named constants (@code{PARAMETER}). 1092@end itemize 1093 1094 1095 1096@node TS 29113 status 1097@section Technical Specification 29113 Status 1098 1099GNU Fortran supports some of the new features of the Technical 1100Specification (TS) 29113 on Further Interoperability of Fortran with C. 1101The @uref{https://gcc.gnu.org/wiki/TS29113Status, wiki} has some information 1102about the current TS 29113 implementation status. In particular, the 1103following is implemented. 1104 1105See also @ref{Further Interoperability of Fortran with C}. 1106 1107@itemize 1108@item The @option{-std=f2008ts} option. 1109 1110@item The @code{OPTIONAL} attribute is allowed for dummy arguments 1111of @code{BIND(C) procedures.} 1112 1113@item The @code{RANK} intrinsic is supported. 1114 1115@item GNU Fortran's implementation for variables with @code{ASYNCHRONOUS} 1116attribute is compatible with TS 29113. 1117 1118@item Assumed types (@code{TYPE(*)}. 1119 1120@item Assumed-rank (@code{DIMENSION(..)}). However, the array descriptor 1121of the TS is not yet supported. 1122@end itemize 1123 1124 1125 1126@c --------------------------------------------------------------------- 1127@c Compiler Characteristics 1128@c --------------------------------------------------------------------- 1129 1130@node Compiler Characteristics 1131@chapter Compiler Characteristics 1132 1133This chapter describes certain characteristics of the GNU Fortran 1134compiler, that are not specified by the Fortran standard, but which 1135might in some way or another become visible to the programmer. 1136 1137@menu 1138* KIND Type Parameters:: 1139* Internal representation of LOGICAL variables:: 1140* Thread-safety of the runtime library:: 1141* Data consistency and durability:: 1142* Files opened without an explicit ACTION= specifier:: 1143@end menu 1144 1145 1146@node KIND Type Parameters 1147@section KIND Type Parameters 1148@cindex kind 1149 1150The @code{KIND} type parameters supported by GNU Fortran for the primitive 1151data types are: 1152 1153@table @code 1154 1155@item INTEGER 11561, 2, 4, 8*, 16*, default: 4** 1157 1158@item LOGICAL 11591, 2, 4, 8*, 16*, default: 4** 1160 1161@item REAL 11624, 8, 10*, 16*, default: 4*** 1163 1164@item COMPLEX 11654, 8, 10*, 16*, default: 4*** 1166 1167@item DOUBLE PRECISION 11684, 8, 10*, 16*, default: 8*** 1169 1170@item CHARACTER 11711, 4, default: 1 1172 1173@end table 1174 1175@noindent 1176* not available on all systems @* 1177** unless @option{-fdefault-integer-8} is used @* 1178*** unless @option{-fdefault-real-8} is used (see @ref{Fortran Dialect Options}) 1179 1180@noindent 1181The @code{KIND} value matches the storage size in bytes, except for 1182@code{COMPLEX} where the storage size is twice as much (or both real and 1183imaginary part are a real value of the given size). It is recommended to use 1184the @ref{SELECTED_CHAR_KIND}, @ref{SELECTED_INT_KIND} and 1185@ref{SELECTED_REAL_KIND} intrinsics or the @code{INT8}, @code{INT16}, 1186@code{INT32}, @code{INT64}, @code{REAL32}, @code{REAL64}, and @code{REAL128} 1187parameters of the @code{ISO_FORTRAN_ENV} module instead of the concrete values. 1188The available kind parameters can be found in the constant arrays 1189@code{CHARACTER_KINDS}, @code{INTEGER_KINDS}, @code{LOGICAL_KINDS} and 1190@code{REAL_KINDS} in the @ref{ISO_FORTRAN_ENV} module. For C interoperability, 1191the kind parameters of the @ref{ISO_C_BINDING} module should be used. 1192 1193 1194@node Internal representation of LOGICAL variables 1195@section Internal representation of LOGICAL variables 1196@cindex logical, variable representation 1197 1198The Fortran standard does not specify how variables of @code{LOGICAL} 1199type are represented, beyond requiring that @code{LOGICAL} variables 1200of default kind have the same storage size as default @code{INTEGER} 1201and @code{REAL} variables. The GNU Fortran internal representation is 1202as follows. 1203 1204A @code{LOGICAL(KIND=N)} variable is represented as an 1205@code{INTEGER(KIND=N)} variable, however, with only two permissible 1206values: @code{1} for @code{.TRUE.} and @code{0} for 1207@code{.FALSE.}. Any other integer value results in undefined behavior. 1208 1209See also @ref{Argument passing conventions} and @ref{Interoperability with C}. 1210 1211 1212@node Thread-safety of the runtime library 1213@section Thread-safety of the runtime library 1214@cindex thread-safety, threads 1215 1216GNU Fortran can be used in programs with multiple threads, e.g.@: by 1217using OpenMP, by calling OS thread handling functions via the 1218@code{ISO_C_BINDING} facility, or by GNU Fortran compiled library code 1219being called from a multi-threaded program. 1220 1221The GNU Fortran runtime library, (@code{libgfortran}), supports being 1222called concurrently from multiple threads with the following 1223exceptions. 1224 1225During library initialization, the C @code{getenv} function is used, 1226which need not be thread-safe. Similarly, the @code{getenv} 1227function is used to implement the @code{GET_ENVIRONMENT_VARIABLE} and 1228@code{GETENV} intrinsics. It is the responsibility of the user to 1229ensure that the environment is not being updated concurrently when any 1230of these actions are taking place. 1231 1232The @code{EXECUTE_COMMAND_LINE} and @code{SYSTEM} intrinsics are 1233implemented with the @code{system} function, which need not be 1234thread-safe. It is the responsibility of the user to ensure that 1235@code{system} is not called concurrently. 1236 1237For platforms not supporting thread-safe POSIX functions, further 1238functionality might not be thread-safe. For details, please consult 1239the documentation for your operating system. 1240 1241The GNU Fortran runtime library uses various C library functions that 1242depend on the locale, such as @code{strtod} and @code{snprintf}. In 1243order to work correctly in locale-aware programs that set the locale 1244using @code{setlocale}, the locale is reset to the default ``C'' 1245locale while executing a formatted @code{READ} or @code{WRITE} 1246statement. On targets supporting the POSIX 2008 per-thread locale 1247functions (e.g. @code{newlocale}, @code{uselocale}, 1248@code{freelocale}), these are used and thus the global locale set 1249using @code{setlocale} or the per-thread locales in other threads are 1250not affected. However, on targets lacking this functionality, the 1251global LC_NUMERIC locale is set to ``C'' during the formatted I/O. 1252Thus, on such targets it's not safe to call @code{setlocale} 1253concurrently from another thread while a Fortran formatted I/O 1254operation is in progress. Also, other threads doing something 1255dependent on the LC_NUMERIC locale might not work correctly if a 1256formatted I/O operation is in progress in another thread. 1257 1258@node Data consistency and durability 1259@section Data consistency and durability 1260@cindex consistency, durability 1261 1262This section contains a brief overview of data and metadata 1263consistency and durability issues when doing I/O. 1264 1265With respect to durability, GNU Fortran makes no effort to ensure that 1266data is committed to stable storage. If this is required, the GNU 1267Fortran programmer can use the intrinsic @code{FNUM} to retrieve the 1268low level file descriptor corresponding to an open Fortran unit. Then, 1269using e.g. the @code{ISO_C_BINDING} feature, one can call the 1270underlying system call to flush dirty data to stable storage, such as 1271@code{fsync} on POSIX, @code{_commit} on MingW, or @code{fcntl(fd, 1272F_FULLSYNC, 0)} on Mac OS X. The following example shows how to call 1273fsync: 1274 1275@smallexample 1276 ! Declare the interface for POSIX fsync function 1277 interface 1278 function fsync (fd) bind(c,name="fsync") 1279 use iso_c_binding, only: c_int 1280 integer(c_int), value :: fd 1281 integer(c_int) :: fsync 1282 end function fsync 1283 end interface 1284 1285 ! Variable declaration 1286 integer :: ret 1287 1288 ! Opening unit 10 1289 open (10,file="foo") 1290 1291 ! ... 1292 ! Perform I/O on unit 10 1293 ! ... 1294 1295 ! Flush and sync 1296 flush(10) 1297 ret = fsync(fnum(10)) 1298 1299 ! Handle possible error 1300 if (ret /= 0) stop "Error calling FSYNC" 1301@end smallexample 1302 1303With respect to consistency, for regular files GNU Fortran uses 1304buffered I/O in order to improve performance. This buffer is flushed 1305automatically when full and in some other situations, e.g. when 1306closing a unit. It can also be explicitly flushed with the 1307@code{FLUSH} statement. Also, the buffering can be turned off with the 1308@code{GFORTRAN_UNBUFFERED_ALL} and 1309@code{GFORTRAN_UNBUFFERED_PRECONNECTED} environment variables. Special 1310files, such as terminals and pipes, are always unbuffered. Sometimes, 1311however, further things may need to be done in order to allow other 1312processes to see data that GNU Fortran has written, as follows. 1313 1314The Windows platform supports a relaxed metadata consistency model, 1315where file metadata is written to the directory lazily. This means 1316that, for instance, the @code{dir} command can show a stale size for a 1317file. One can force a directory metadata update by closing the unit, 1318or by calling @code{_commit} on the file descriptor. Note, though, 1319that @code{_commit} will force all dirty data to stable storage, which 1320is often a very slow operation. 1321 1322The Network File System (NFS) implements a relaxed consistency model 1323called open-to-close consistency. Closing a file forces dirty data and 1324metadata to be flushed to the server, and opening a file forces the 1325client to contact the server in order to revalidate cached 1326data. @code{fsync} will also force a flush of dirty data and metadata 1327to the server. Similar to @code{open} and @code{close}, acquiring and 1328releasing @code{fcntl} file locks, if the server supports them, will 1329also force cache validation and flushing dirty data and metadata. 1330 1331 1332@node Files opened without an explicit ACTION= specifier 1333@section Files opened without an explicit ACTION= specifier 1334@cindex open, action 1335 1336The Fortran standard says that if an @code{OPEN} statement is executed 1337without an explicit @code{ACTION=} specifier, the default value is 1338processor dependent. GNU Fortran behaves as follows: 1339 1340@enumerate 1341@item Attempt to open the file with @code{ACTION='READWRITE'} 1342@item If that fails, try to open with @code{ACTION='READ'} 1343@item If that fails, try to open with @code{ACTION='WRITE'} 1344@item If that fails, generate an error 1345@end enumerate 1346 1347 1348@c --------------------------------------------------------------------- 1349@c Extensions 1350@c --------------------------------------------------------------------- 1351 1352@c Maybe this chapter should be merged with the 'Standards' section, 1353@c whenever that is written :-) 1354 1355@node Extensions 1356@chapter Extensions 1357@cindex extensions 1358 1359The two sections below detail the extensions to standard Fortran that are 1360implemented in GNU Fortran, as well as some of the popular or 1361historically important extensions that are not (or not yet) implemented. 1362For the latter case, we explain the alternatives available to GNU Fortran 1363users, including replacement by standard-conforming code or GNU 1364extensions. 1365 1366@menu 1367* Extensions implemented in GNU Fortran:: 1368* Extensions not implemented in GNU Fortran:: 1369@end menu 1370 1371 1372@node Extensions implemented in GNU Fortran 1373@section Extensions implemented in GNU Fortran 1374@cindex extensions, implemented 1375 1376GNU Fortran implements a number of extensions over standard 1377Fortran. This chapter contains information on their syntax and 1378meaning. There are currently two categories of GNU Fortran 1379extensions, those that provide functionality beyond that provided 1380by any standard, and those that are supported by GNU Fortran 1381purely for backward compatibility with legacy compilers. By default, 1382@option{-std=gnu} allows the compiler to accept both types of 1383extensions, but to warn about the use of the latter. Specifying 1384either @option{-std=f95}, @option{-std=f2003} or @option{-std=f2008} 1385disables both types of extensions, and @option{-std=legacy} allows both 1386without warning. 1387 1388@menu 1389* Old-style kind specifications:: 1390* Old-style variable initialization:: 1391* Extensions to namelist:: 1392* X format descriptor without count field:: 1393* Commas in FORMAT specifications:: 1394* Missing period in FORMAT specifications:: 1395* I/O item lists:: 1396* @code{Q} exponent-letter:: 1397* BOZ literal constants:: 1398* Real array indices:: 1399* Unary operators:: 1400* Implicitly convert LOGICAL and INTEGER values:: 1401* Hollerith constants support:: 1402* Cray pointers:: 1403* CONVERT specifier:: 1404* OpenMP:: 1405* OpenACC:: 1406* Argument list functions:: 1407* Read/Write after EOF marker:: 1408@end menu 1409 1410@node Old-style kind specifications 1411@subsection Old-style kind specifications 1412@cindex kind, old-style 1413 1414GNU Fortran allows old-style kind specifications in declarations. These 1415look like: 1416@smallexample 1417 TYPESPEC*size x,y,z 1418@end smallexample 1419@noindent 1420where @code{TYPESPEC} is a basic type (@code{INTEGER}, @code{REAL}, 1421etc.), and where @code{size} is a byte count corresponding to the 1422storage size of a valid kind for that type. (For @code{COMPLEX} 1423variables, @code{size} is the total size of the real and imaginary 1424parts.) The statement then declares @code{x}, @code{y} and @code{z} to 1425be of type @code{TYPESPEC} with the appropriate kind. This is 1426equivalent to the standard-conforming declaration 1427@smallexample 1428 TYPESPEC(k) x,y,z 1429@end smallexample 1430@noindent 1431where @code{k} is the kind parameter suitable for the intended precision. As 1432kind parameters are implementation-dependent, use the @code{KIND}, 1433@code{SELECTED_INT_KIND} and @code{SELECTED_REAL_KIND} intrinsics to retrieve 1434the correct value, for instance @code{REAL*8 x} can be replaced by: 1435@smallexample 1436INTEGER, PARAMETER :: dbl = KIND(1.0d0) 1437REAL(KIND=dbl) :: x 1438@end smallexample 1439 1440@node Old-style variable initialization 1441@subsection Old-style variable initialization 1442 1443GNU Fortran allows old-style initialization of variables of the 1444form: 1445@smallexample 1446 INTEGER i/1/,j/2/ 1447 REAL x(2,2) /3*0.,1./ 1448@end smallexample 1449The syntax for the initializers is as for the @code{DATA} statement, but 1450unlike in a @code{DATA} statement, an initializer only applies to the 1451variable immediately preceding the initialization. In other words, 1452something like @code{INTEGER I,J/2,3/} is not valid. This style of 1453initialization is only allowed in declarations without double colons 1454(@code{::}); the double colons were introduced in Fortran 90, which also 1455introduced a standard syntax for initializing variables in type 1456declarations. 1457 1458Examples of standard-conforming code equivalent to the above example 1459are: 1460@smallexample 1461! Fortran 90 1462 INTEGER :: i = 1, j = 2 1463 REAL :: x(2,2) = RESHAPE((/0.,0.,0.,1./),SHAPE(x)) 1464! Fortran 77 1465 INTEGER i, j 1466 REAL x(2,2) 1467 DATA i/1/, j/2/, x/3*0.,1./ 1468@end smallexample 1469 1470Note that variables which are explicitly initialized in declarations 1471or in @code{DATA} statements automatically acquire the @code{SAVE} 1472attribute. 1473 1474@node Extensions to namelist 1475@subsection Extensions to namelist 1476@cindex Namelist 1477 1478GNU Fortran fully supports the Fortran 95 standard for namelist I/O 1479including array qualifiers, substrings and fully qualified derived types. 1480The output from a namelist write is compatible with namelist read. The 1481output has all names in upper case and indentation to column 1 after the 1482namelist name. Two extensions are permitted: 1483 1484Old-style use of @samp{$} instead of @samp{&} 1485@smallexample 1486$MYNML 1487 X(:)%Y(2) = 1.0 2.0 3.0 1488 CH(1:4) = "abcd" 1489$END 1490@end smallexample 1491 1492It should be noted that the default terminator is @samp{/} rather than 1493@samp{&END}. 1494 1495Querying of the namelist when inputting from stdin. After at least 1496one space, entering @samp{?} sends to stdout the namelist name and the names of 1497the variables in the namelist: 1498@smallexample 1499 ? 1500 1501&mynml 1502 x 1503 x%y 1504 ch 1505&end 1506@end smallexample 1507 1508Entering @samp{=?} outputs the namelist to stdout, as if 1509@code{WRITE(*,NML = mynml)} had been called: 1510@smallexample 1511=? 1512 1513&MYNML 1514 X(1)%Y= 0.000000 , 1.000000 , 0.000000 , 1515 X(2)%Y= 0.000000 , 2.000000 , 0.000000 , 1516 X(3)%Y= 0.000000 , 3.000000 , 0.000000 , 1517 CH=abcd, / 1518@end smallexample 1519 1520To aid this dialog, when input is from stdin, errors send their 1521messages to stderr and execution continues, even if @code{IOSTAT} is set. 1522 1523@code{PRINT} namelist is permitted. This causes an error if 1524@option{-std=f95} is used. 1525@smallexample 1526PROGRAM test_print 1527 REAL, dimension (4) :: x = (/1.0, 2.0, 3.0, 4.0/) 1528 NAMELIST /mynml/ x 1529 PRINT mynml 1530END PROGRAM test_print 1531@end smallexample 1532 1533Expanded namelist reads are permitted. This causes an error if 1534@option{-std=f95} is used. In the following example, the first element 1535of the array will be given the value 0.00 and the two succeeding 1536elements will be given the values 1.00 and 2.00. 1537@smallexample 1538&MYNML 1539 X(1,1) = 0.00 , 1.00 , 2.00 1540/ 1541@end smallexample 1542 1543When writing a namelist, if no @code{DELIM=} is specified, by default a 1544double quote is used to delimit character strings. If -std=F95, F2003, 1545or F2008, etc, the delim status is set to 'none'. Defaulting to 1546quotes ensures that namelists with character strings can be subsequently 1547read back in accurately. 1548 1549@node X format descriptor without count field 1550@subsection @code{X} format descriptor without count field 1551 1552To support legacy codes, GNU Fortran permits the count field of the 1553@code{X} edit descriptor in @code{FORMAT} statements to be omitted. 1554When omitted, the count is implicitly assumed to be one. 1555 1556@smallexample 1557 PRINT 10, 2, 3 155810 FORMAT (I1, X, I1) 1559@end smallexample 1560 1561@node Commas in FORMAT specifications 1562@subsection Commas in @code{FORMAT} specifications 1563 1564To support legacy codes, GNU Fortran allows the comma separator 1565to be omitted immediately before and after character string edit 1566descriptors in @code{FORMAT} statements. 1567 1568@smallexample 1569 PRINT 10, 2, 3 157010 FORMAT ('FOO='I1' BAR='I2) 1571@end smallexample 1572 1573 1574@node Missing period in FORMAT specifications 1575@subsection Missing period in @code{FORMAT} specifications 1576 1577To support legacy codes, GNU Fortran allows missing periods in format 1578specifications if and only if @option{-std=legacy} is given on the 1579command line. This is considered non-conforming code and is 1580discouraged. 1581 1582@smallexample 1583 REAL :: value 1584 READ(*,10) value 158510 FORMAT ('F4') 1586@end smallexample 1587 1588@node I/O item lists 1589@subsection I/O item lists 1590@cindex I/O item lists 1591 1592To support legacy codes, GNU Fortran allows the input item list 1593of the @code{READ} statement, and the output item lists of the 1594@code{WRITE} and @code{PRINT} statements, to start with a comma. 1595 1596@node @code{Q} exponent-letter 1597@subsection @code{Q} exponent-letter 1598@cindex @code{Q} exponent-letter 1599 1600GNU Fortran accepts real literal constants with an exponent-letter 1601of @code{Q}, for example, @code{1.23Q45}. The constant is interpreted 1602as a @code{REAL(16)} entity on targets that support this type. If 1603the target does not support @code{REAL(16)} but has a @code{REAL(10)} 1604type, then the real-literal-constant will be interpreted as a 1605@code{REAL(10)} entity. In the absence of @code{REAL(16)} and 1606@code{REAL(10)}, an error will occur. 1607 1608@node BOZ literal constants 1609@subsection BOZ literal constants 1610@cindex BOZ literal constants 1611 1612Besides decimal constants, Fortran also supports binary (@code{b}), 1613octal (@code{o}) and hexadecimal (@code{z}) integer constants. The 1614syntax is: @samp{prefix quote digits quote}, were the prefix is 1615either @code{b}, @code{o} or @code{z}, quote is either @code{'} or 1616@code{"} and the digits are for binary @code{0} or @code{1}, for 1617octal between @code{0} and @code{7}, and for hexadecimal between 1618@code{0} and @code{F}. (Example: @code{b'01011101'}.) 1619 1620Up to Fortran 95, BOZ literals were only allowed to initialize 1621integer variables in DATA statements. Since Fortran 2003 BOZ literals 1622are also allowed as argument of @code{REAL}, @code{DBLE}, @code{INT} 1623and @code{CMPLX}; the result is the same as if the integer BOZ 1624literal had been converted by @code{TRANSFER} to, respectively, 1625@code{real}, @code{double precision}, @code{integer} or @code{complex}. 1626As GNU Fortran extension the intrinsic procedures @code{FLOAT}, 1627@code{DFLOAT}, @code{COMPLEX} and @code{DCMPLX} are treated alike. 1628 1629As an extension, GNU Fortran allows hexadecimal BOZ literal constants to 1630be specified using the @code{X} prefix, in addition to the standard 1631@code{Z} prefix. The BOZ literal can also be specified by adding a 1632suffix to the string, for example, @code{Z'ABC'} and @code{'ABC'Z} are 1633equivalent. 1634 1635Furthermore, GNU Fortran allows using BOZ literal constants outside 1636DATA statements and the four intrinsic functions allowed by Fortran 2003. 1637In DATA statements, in direct assignments, where the right-hand side 1638only contains a BOZ literal constant, and for old-style initializers of 1639the form @code{integer i /o'0173'/}, the constant is transferred 1640as if @code{TRANSFER} had been used; for @code{COMPLEX} numbers, only 1641the real part is initialized unless @code{CMPLX} is used. In all other 1642cases, the BOZ literal constant is converted to an @code{INTEGER} value with 1643the largest decimal representation. This value is then converted 1644numerically to the type and kind of the variable in question. 1645(For instance, @code{real :: r = b'0000001' + 1} initializes @code{r} 1646with @code{2.0}.) As different compilers implement the extension 1647differently, one should be careful when doing bitwise initialization 1648of non-integer variables. 1649 1650Note that initializing an @code{INTEGER} variable with a statement such 1651as @code{DATA i/Z'FFFFFFFF'/} will give an integer overflow error rather 1652than the desired result of @math{-1} when @code{i} is a 32-bit integer 1653on a system that supports 64-bit integers. The @samp{-fno-range-check} 1654option can be used as a workaround for legacy code that initializes 1655integers in this manner. 1656 1657@node Real array indices 1658@subsection Real array indices 1659@cindex array, indices of type real 1660 1661As an extension, GNU Fortran allows the use of @code{REAL} expressions 1662or variables as array indices. 1663 1664@node Unary operators 1665@subsection Unary operators 1666@cindex operators, unary 1667 1668As an extension, GNU Fortran allows unary plus and unary minus operators 1669to appear as the second operand of binary arithmetic operators without 1670the need for parenthesis. 1671 1672@smallexample 1673 X = Y * -Z 1674@end smallexample 1675 1676@node Implicitly convert LOGICAL and INTEGER values 1677@subsection Implicitly convert @code{LOGICAL} and @code{INTEGER} values 1678@cindex conversion, to integer 1679@cindex conversion, to logical 1680 1681As an extension for backwards compatibility with other compilers, GNU 1682Fortran allows the implicit conversion of @code{LOGICAL} values to 1683@code{INTEGER} values and vice versa. When converting from a 1684@code{LOGICAL} to an @code{INTEGER}, @code{.FALSE.} is interpreted as 1685zero, and @code{.TRUE.} is interpreted as one. When converting from 1686@code{INTEGER} to @code{LOGICAL}, the value zero is interpreted as 1687@code{.FALSE.} and any nonzero value is interpreted as @code{.TRUE.}. 1688 1689@smallexample 1690 LOGICAL :: l 1691 l = 1 1692@end smallexample 1693@smallexample 1694 INTEGER :: i 1695 i = .TRUE. 1696@end smallexample 1697 1698However, there is no implicit conversion of @code{INTEGER} values in 1699@code{if}-statements, nor of @code{LOGICAL} or @code{INTEGER} values 1700in I/O operations. 1701 1702@node Hollerith constants support 1703@subsection Hollerith constants support 1704@cindex Hollerith constants 1705 1706GNU Fortran supports Hollerith constants in assignments, function 1707arguments, and @code{DATA} and @code{ASSIGN} statements. A Hollerith 1708constant is written as a string of characters preceded by an integer 1709constant indicating the character count, and the letter @code{H} or 1710@code{h}, and stored in bytewise fashion in a numeric (@code{INTEGER}, 1711@code{REAL}, or @code{complex}) or @code{LOGICAL} variable. The 1712constant will be padded or truncated to fit the size of the variable in 1713which it is stored. 1714 1715Examples of valid uses of Hollerith constants: 1716@smallexample 1717 complex*16 x(2) 1718 data x /16Habcdefghijklmnop, 16Hqrstuvwxyz012345/ 1719 x(1) = 16HABCDEFGHIJKLMNOP 1720 call foo (4h abc) 1721@end smallexample 1722 1723Invalid Hollerith constants examples: 1724@smallexample 1725 integer*4 a 1726 a = 8H12345678 ! Valid, but the Hollerith constant will be truncated. 1727 a = 0H ! At least one character is needed. 1728@end smallexample 1729 1730In general, Hollerith constants were used to provide a rudimentary 1731facility for handling character strings in early Fortran compilers, 1732prior to the introduction of @code{CHARACTER} variables in Fortran 77; 1733in those cases, the standard-compliant equivalent is to convert the 1734program to use proper character strings. On occasion, there may be a 1735case where the intent is specifically to initialize a numeric variable 1736with a given byte sequence. In these cases, the same result can be 1737obtained by using the @code{TRANSFER} statement, as in this example. 1738@smallexample 1739 INTEGER(KIND=4) :: a 1740 a = TRANSFER ("abcd", a) ! equivalent to: a = 4Habcd 1741@end smallexample 1742 1743 1744@node Cray pointers 1745@subsection Cray pointers 1746@cindex pointer, Cray 1747 1748Cray pointers are part of a non-standard extension that provides a 1749C-like pointer in Fortran. This is accomplished through a pair of 1750variables: an integer "pointer" that holds a memory address, and a 1751"pointee" that is used to dereference the pointer. 1752 1753Pointer/pointee pairs are declared in statements of the form: 1754@smallexample 1755 pointer ( <pointer> , <pointee> ) 1756@end smallexample 1757or, 1758@smallexample 1759 pointer ( <pointer1> , <pointee1> ), ( <pointer2> , <pointee2> ), ... 1760@end smallexample 1761The pointer is an integer that is intended to hold a memory address. 1762The pointee may be an array or scalar. A pointee can be an assumed 1763size array---that is, the last dimension may be left unspecified by 1764using a @code{*} in place of a value---but a pointee cannot be an 1765assumed shape array. No space is allocated for the pointee. 1766 1767The pointee may have its type declared before or after the pointer 1768statement, and its array specification (if any) may be declared 1769before, during, or after the pointer statement. The pointer may be 1770declared as an integer prior to the pointer statement. However, some 1771machines have default integer sizes that are different than the size 1772of a pointer, and so the following code is not portable: 1773@smallexample 1774 integer ipt 1775 pointer (ipt, iarr) 1776@end smallexample 1777If a pointer is declared with a kind that is too small, the compiler 1778will issue a warning; the resulting binary will probably not work 1779correctly, because the memory addresses stored in the pointers may be 1780truncated. It is safer to omit the first line of the above example; 1781if explicit declaration of ipt's type is omitted, then the compiler 1782will ensure that ipt is an integer variable large enough to hold a 1783pointer. 1784 1785Pointer arithmetic is valid with Cray pointers, but it is not the same 1786as C pointer arithmetic. Cray pointers are just ordinary integers, so 1787the user is responsible for determining how many bytes to add to a 1788pointer in order to increment it. Consider the following example: 1789@smallexample 1790 real target(10) 1791 real pointee(10) 1792 pointer (ipt, pointee) 1793 ipt = loc (target) 1794 ipt = ipt + 1 1795@end smallexample 1796The last statement does not set @code{ipt} to the address of 1797@code{target(1)}, as it would in C pointer arithmetic. Adding @code{1} 1798to @code{ipt} just adds one byte to the address stored in @code{ipt}. 1799 1800Any expression involving the pointee will be translated to use the 1801value stored in the pointer as the base address. 1802 1803To get the address of elements, this extension provides an intrinsic 1804function @code{LOC()}. The @code{LOC()} function is equivalent to the 1805@code{&} operator in C, except the address is cast to an integer type: 1806@smallexample 1807 real ar(10) 1808 pointer(ipt, arpte(10)) 1809 real arpte 1810 ipt = loc(ar) ! Makes arpte is an alias for ar 1811 arpte(1) = 1.0 ! Sets ar(1) to 1.0 1812@end smallexample 1813The pointer can also be set by a call to the @code{MALLOC} intrinsic 1814(see @ref{MALLOC}). 1815 1816Cray pointees often are used to alias an existing variable. For 1817example: 1818@smallexample 1819 integer target(10) 1820 integer iarr(10) 1821 pointer (ipt, iarr) 1822 ipt = loc(target) 1823@end smallexample 1824As long as @code{ipt} remains unchanged, @code{iarr} is now an alias for 1825@code{target}. The optimizer, however, will not detect this aliasing, so 1826it is unsafe to use @code{iarr} and @code{target} simultaneously. Using 1827a pointee in any way that violates the Fortran aliasing rules or 1828assumptions is illegal. It is the user's responsibility to avoid doing 1829this; the compiler works under the assumption that no such aliasing 1830occurs. 1831 1832Cray pointers will work correctly when there is no aliasing (i.e., when 1833they are used to access a dynamically allocated block of memory), and 1834also in any routine where a pointee is used, but any variable with which 1835it shares storage is not used. Code that violates these rules may not 1836run as the user intends. This is not a bug in the optimizer; any code 1837that violates the aliasing rules is illegal. (Note that this is not 1838unique to GNU Fortran; any Fortran compiler that supports Cray pointers 1839will ``incorrectly'' optimize code with illegal aliasing.) 1840 1841There are a number of restrictions on the attributes that can be applied 1842to Cray pointers and pointees. Pointees may not have the 1843@code{ALLOCATABLE}, @code{INTENT}, @code{OPTIONAL}, @code{DUMMY}, 1844@code{TARGET}, @code{INTRINSIC}, or @code{POINTER} attributes. Pointers 1845may not have the @code{DIMENSION}, @code{POINTER}, @code{TARGET}, 1846@code{ALLOCATABLE}, @code{EXTERNAL}, or @code{INTRINSIC} attributes, nor 1847may they be function results. Pointees may not occur in more than one 1848pointer statement. A pointee cannot be a pointer. Pointees cannot occur 1849in equivalence, common, or data statements. 1850 1851A Cray pointer may also point to a function or a subroutine. For 1852example, the following excerpt is valid: 1853@smallexample 1854 implicit none 1855 external sub 1856 pointer (subptr,subpte) 1857 external subpte 1858 subptr = loc(sub) 1859 call subpte() 1860 [...] 1861 subroutine sub 1862 [...] 1863 end subroutine sub 1864@end smallexample 1865 1866A pointer may be modified during the course of a program, and this 1867will change the location to which the pointee refers. However, when 1868pointees are passed as arguments, they are treated as ordinary 1869variables in the invoked function. Subsequent changes to the pointer 1870will not change the base address of the array that was passed. 1871 1872@node CONVERT specifier 1873@subsection @code{CONVERT} specifier 1874@cindex @code{CONVERT} specifier 1875 1876GNU Fortran allows the conversion of unformatted data between little- 1877and big-endian representation to facilitate moving of data 1878between different systems. The conversion can be indicated with 1879the @code{CONVERT} specifier on the @code{OPEN} statement. 1880@xref{GFORTRAN_CONVERT_UNIT}, for an alternative way of specifying 1881the data format via an environment variable. 1882 1883Valid values for @code{CONVERT} are: 1884@itemize @w{} 1885@item @code{CONVERT='NATIVE'} Use the native format. This is the default. 1886@item @code{CONVERT='SWAP'} Swap between little- and big-endian. 1887@item @code{CONVERT='LITTLE_ENDIAN'} Use the little-endian representation 1888for unformatted files. 1889@item @code{CONVERT='BIG_ENDIAN'} Use the big-endian representation for 1890unformatted files. 1891@end itemize 1892 1893Using the option could look like this: 1894@smallexample 1895 open(file='big.dat',form='unformatted',access='sequential', & 1896 convert='big_endian') 1897@end smallexample 1898 1899The value of the conversion can be queried by using 1900@code{INQUIRE(CONVERT=ch)}. The values returned are 1901@code{'BIG_ENDIAN'} and @code{'LITTLE_ENDIAN'}. 1902 1903@code{CONVERT} works between big- and little-endian for 1904@code{INTEGER} values of all supported kinds and for @code{REAL} 1905on IEEE systems of kinds 4 and 8. Conversion between different 1906``extended double'' types on different architectures such as 1907m68k and x86_64, which GNU Fortran 1908supports as @code{REAL(KIND=10)} and @code{REAL(KIND=16)}, will 1909probably not work. 1910 1911@emph{Note that the values specified via the GFORTRAN_CONVERT_UNIT 1912environment variable will override the CONVERT specifier in the 1913open statement}. This is to give control over data formats to 1914users who do not have the source code of their program available. 1915 1916Using anything but the native representation for unformatted data 1917carries a significant speed overhead. If speed in this area matters 1918to you, it is best if you use this only for data that needs to be 1919portable. 1920 1921@node OpenMP 1922@subsection OpenMP 1923@cindex OpenMP 1924 1925OpenMP (Open Multi-Processing) is an application programming 1926interface (API) that supports multi-platform shared memory 1927multiprocessing programming in C/C++ and Fortran on many 1928architectures, including Unix and Microsoft Windows platforms. 1929It consists of a set of compiler directives, library routines, 1930and environment variables that influence run-time behavior. 1931 1932GNU Fortran strives to be compatible to the 1933@uref{http://openmp.org/wp/openmp-specifications/, 1934OpenMP Application Program Interface v4.0}. 1935 1936To enable the processing of the OpenMP directive @code{!$omp} in 1937free-form source code; the @code{c$omp}, @code{*$omp} and @code{!$omp} 1938directives in fixed form; the @code{!$} conditional compilation sentinels 1939in free form; and the @code{c$}, @code{*$} and @code{!$} sentinels 1940in fixed form, @command{gfortran} needs to be invoked with the 1941@option{-fopenmp}. This also arranges for automatic linking of the 1942GNU Offloading and Multi Processing Runtime Library 1943@ref{Top,,libgomp,libgomp,GNU Offloading and Multi Processing Runtime 1944Library}. 1945 1946The OpenMP Fortran runtime library routines are provided both in a 1947form of a Fortran 90 module named @code{omp_lib} and in a form of 1948a Fortran @code{include} file named @file{omp_lib.h}. 1949 1950An example of a parallelized loop taken from Appendix A.1 of 1951the OpenMP Application Program Interface v2.5: 1952@smallexample 1953SUBROUTINE A1(N, A, B) 1954 INTEGER I, N 1955 REAL B(N), A(N) 1956!$OMP PARALLEL DO !I is private by default 1957 DO I=2,N 1958 B(I) = (A(I) + A(I-1)) / 2.0 1959 ENDDO 1960!$OMP END PARALLEL DO 1961END SUBROUTINE A1 1962@end smallexample 1963 1964Please note: 1965@itemize 1966@item 1967@option{-fopenmp} implies @option{-frecursive}, i.e., all local arrays 1968will be allocated on the stack. When porting existing code to OpenMP, 1969this may lead to surprising results, especially to segmentation faults 1970if the stacksize is limited. 1971 1972@item 1973On glibc-based systems, OpenMP enabled applications cannot be statically 1974linked due to limitations of the underlying pthreads-implementation. It 1975might be possible to get a working solution if 1976@command{-Wl,--whole-archive -lpthread -Wl,--no-whole-archive} is added 1977to the command line. However, this is not supported by @command{gcc} and 1978thus not recommended. 1979@end itemize 1980 1981@node OpenACC 1982@subsection OpenACC 1983@cindex OpenACC 1984 1985OpenACC is an application programming interface (API) that supports 1986offloading of code to accelerator devices. It consists of a set of 1987compiler directives, library routines, and environment variables that 1988influence run-time behavior. 1989 1990GNU Fortran strives to be compatible to the 1991@uref{http://www.openacc.org/, OpenACC Application Programming 1992Interface v2.0}. 1993 1994To enable the processing of the OpenACC directive @code{!$acc} in 1995free-form source code; the @code{c$acc}, @code{*$acc} and @code{!$acc} 1996directives in fixed form; the @code{!$} conditional compilation 1997sentinels in free form; and the @code{c$}, @code{*$} and @code{!$} 1998sentinels in fixed form, @command{gfortran} needs to be invoked with 1999the @option{-fopenacc}. This also arranges for automatic linking of 2000the GNU Offloading and Multi Processing Runtime Library 2001@ref{Top,,libgomp,libgomp,GNU Offloading and Multi Processing Runtime 2002Library}. 2003 2004The OpenACC Fortran runtime library routines are provided both in a 2005form of a Fortran 90 module named @code{openacc} and in a form of a 2006Fortran @code{include} file named @file{openacc_lib.h}. 2007 2008Note that this is an experimental feature, incomplete, and subject to 2009change in future versions of GCC. See 2010@uref{https://gcc.gnu.org/wiki/OpenACC} for more information. 2011 2012@node Argument list functions 2013@subsection Argument list functions @code{%VAL}, @code{%REF} and @code{%LOC} 2014@cindex argument list functions 2015@cindex @code{%VAL} 2016@cindex @code{%REF} 2017@cindex @code{%LOC} 2018 2019GNU Fortran supports argument list functions @code{%VAL}, @code{%REF} 2020and @code{%LOC} statements, for backward compatibility with g77. 2021It is recommended that these should be used only for code that is 2022accessing facilities outside of GNU Fortran, such as operating system 2023or windowing facilities. It is best to constrain such uses to isolated 2024portions of a program--portions that deal specifically and exclusively 2025with low-level, system-dependent facilities. Such portions might well 2026provide a portable interface for use by the program as a whole, but are 2027themselves not portable, and should be thoroughly tested each time they 2028are rebuilt using a new compiler or version of a compiler. 2029 2030@code{%VAL} passes a scalar argument by value, @code{%REF} passes it by 2031reference and @code{%LOC} passes its memory location. Since gfortran 2032already passes scalar arguments by reference, @code{%REF} is in effect 2033a do-nothing. @code{%LOC} has the same effect as a Fortran pointer. 2034 2035An example of passing an argument by value to a C subroutine foo.: 2036@smallexample 2037C 2038C prototype void foo_ (float x); 2039C 2040 external foo 2041 real*4 x 2042 x = 3.14159 2043 call foo (%VAL (x)) 2044 end 2045@end smallexample 2046 2047For details refer to the g77 manual 2048@uref{https://gcc.gnu.org/@/onlinedocs/@/gcc-3.4.6/@/g77/@/index.html#Top}. 2049 2050Also, @code{c_by_val.f} and its partner @code{c_by_val.c} of the 2051GNU Fortran testsuite are worth a look. 2052 2053@node Read/Write after EOF marker 2054@subsection Read/Write after EOF marker 2055@cindex @code{EOF} 2056@cindex @code{BACKSPACE} 2057@cindex @code{REWIND} 2058 2059Some legacy codes rely on allowing @code{READ} or @code{WRITE} after the 2060EOF file marker in order to find the end of a file. GNU Fortran normally 2061rejects these codes with a run-time error message and suggests the user 2062consider @code{BACKSPACE} or @code{REWIND} to properly position 2063the file before the EOF marker. As an extension, the run-time error may 2064be disabled using -std=legacy. 2065 2066@node Extensions not implemented in GNU Fortran 2067@section Extensions not implemented in GNU Fortran 2068@cindex extensions, not implemented 2069 2070The long history of the Fortran language, its wide use and broad 2071userbase, the large number of different compiler vendors and the lack of 2072some features crucial to users in the first standards have lead to the 2073existence of a number of important extensions to the language. While 2074some of the most useful or popular extensions are supported by the GNU 2075Fortran compiler, not all existing extensions are supported. This section 2076aims at listing these extensions and offering advice on how best make 2077code that uses them running with the GNU Fortran compiler. 2078 2079@c More can be found here: 2080@c -- https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/Missing-Features.html 2081@c -- the list of Fortran and libgfortran bugs closed as WONTFIX: 2082@c http://tinyurl.com/2u4h5y 2083 2084@menu 2085* STRUCTURE and RECORD:: 2086@c * UNION and MAP:: 2087* ENCODE and DECODE statements:: 2088* Variable FORMAT expressions:: 2089@c * Q edit descriptor:: 2090@c * AUTOMATIC statement:: 2091@c * TYPE and ACCEPT I/O Statements:: 2092@c * .XOR. operator:: 2093@c * CARRIAGECONTROL, DEFAULTFILE, DISPOSE and RECORDTYPE I/O specifiers:: 2094@c * Omitted arguments in procedure call:: 2095* Alternate complex function syntax:: 2096* Volatile COMMON blocks:: 2097@end menu 2098 2099 2100@node STRUCTURE and RECORD 2101@subsection @code{STRUCTURE} and @code{RECORD} 2102@cindex @code{STRUCTURE} 2103@cindex @code{RECORD} 2104 2105Record structures are a pre-Fortran-90 vendor extension to create 2106user-defined aggregate data types. GNU Fortran does not support 2107record structures, only Fortran 90's ``derived types'', which have 2108a different syntax. 2109 2110In many cases, record structures can easily be converted to derived types. 2111To convert, replace @code{STRUCTURE /}@var{structure-name}@code{/} 2112by @code{TYPE} @var{type-name}. Additionally, replace 2113@code{RECORD /}@var{structure-name}@code{/} by 2114@code{TYPE(}@var{type-name}@code{)}. Finally, in the component access, 2115replace the period (@code{.}) by the percent sign (@code{%}). 2116 2117Here is an example of code using the non portable record structure syntax: 2118 2119@example 2120! Declaring a structure named ``item'' and containing three fields: 2121! an integer ID, an description string and a floating-point price. 2122STRUCTURE /item/ 2123 INTEGER id 2124 CHARACTER(LEN=200) description 2125 REAL price 2126END STRUCTURE 2127 2128! Define two variables, an single record of type ``item'' 2129! named ``pear'', and an array of items named ``store_catalog'' 2130RECORD /item/ pear, store_catalog(100) 2131 2132! We can directly access the fields of both variables 2133pear.id = 92316 2134pear.description = "juicy D'Anjou pear" 2135pear.price = 0.15 2136store_catalog(7).id = 7831 2137store_catalog(7).description = "milk bottle" 2138store_catalog(7).price = 1.2 2139 2140! We can also manipulate the whole structure 2141store_catalog(12) = pear 2142print *, store_catalog(12) 2143@end example 2144 2145@noindent 2146This code can easily be rewritten in the Fortran 90 syntax as following: 2147 2148@example 2149! ``STRUCTURE /name/ ... END STRUCTURE'' becomes 2150! ``TYPE name ... END TYPE'' 2151TYPE item 2152 INTEGER id 2153 CHARACTER(LEN=200) description 2154 REAL price 2155END TYPE 2156 2157! ``RECORD /name/ variable'' becomes ``TYPE(name) variable'' 2158TYPE(item) pear, store_catalog(100) 2159 2160! Instead of using a dot (.) to access fields of a record, the 2161! standard syntax uses a percent sign (%) 2162pear%id = 92316 2163pear%description = "juicy D'Anjou pear" 2164pear%price = 0.15 2165store_catalog(7)%id = 7831 2166store_catalog(7)%description = "milk bottle" 2167store_catalog(7)%price = 1.2 2168 2169! Assignments of a whole variable do not change 2170store_catalog(12) = pear 2171print *, store_catalog(12) 2172@end example 2173 2174 2175@c @node UNION and MAP 2176@c @subsection @code{UNION} and @code{MAP} 2177@c @cindex @code{UNION} 2178@c @cindex @code{MAP} 2179@c 2180@c For help writing this one, see 2181@c http://www.eng.umd.edu/~nsw/ench250/fortran1.htm#UNION and 2182@c http://www.tacc.utexas.edu/services/userguides/pgi/pgiws_ug/pgi32u06.htm 2183 2184 2185@node ENCODE and DECODE statements 2186@subsection @code{ENCODE} and @code{DECODE} statements 2187@cindex @code{ENCODE} 2188@cindex @code{DECODE} 2189 2190GNU Fortran does not support the @code{ENCODE} and @code{DECODE} 2191statements. These statements are best replaced by @code{READ} and 2192@code{WRITE} statements involving internal files (@code{CHARACTER} 2193variables and arrays), which have been part of the Fortran standard since 2194Fortran 77. For example, replace a code fragment like 2195 2196@smallexample 2197 INTEGER*1 LINE(80) 2198 REAL A, B, C 2199c ... Code that sets LINE 2200 DECODE (80, 9000, LINE) A, B, C 2201 9000 FORMAT (1X, 3(F10.5)) 2202@end smallexample 2203 2204@noindent 2205with the following: 2206 2207@smallexample 2208 CHARACTER(LEN=80) LINE 2209 REAL A, B, C 2210c ... Code that sets LINE 2211 READ (UNIT=LINE, FMT=9000) A, B, C 2212 9000 FORMAT (1X, 3(F10.5)) 2213@end smallexample 2214 2215Similarly, replace a code fragment like 2216 2217@smallexample 2218 INTEGER*1 LINE(80) 2219 REAL A, B, C 2220c ... Code that sets A, B and C 2221 ENCODE (80, 9000, LINE) A, B, C 2222 9000 FORMAT (1X, 'OUTPUT IS ', 3(F10.5)) 2223@end smallexample 2224 2225@noindent 2226with the following: 2227 2228@smallexample 2229 CHARACTER(LEN=80) LINE 2230 REAL A, B, C 2231c ... Code that sets A, B and C 2232 WRITE (UNIT=LINE, FMT=9000) A, B, C 2233 9000 FORMAT (1X, 'OUTPUT IS ', 3(F10.5)) 2234@end smallexample 2235 2236 2237@node Variable FORMAT expressions 2238@subsection Variable @code{FORMAT} expressions 2239@cindex @code{FORMAT} 2240 2241A variable @code{FORMAT} expression is format statement which includes 2242angle brackets enclosing a Fortran expression: @code{FORMAT(I<N>)}. GNU 2243Fortran does not support this legacy extension. The effect of variable 2244format expressions can be reproduced by using the more powerful (and 2245standard) combination of internal output and string formats. For example, 2246replace a code fragment like this: 2247 2248@smallexample 2249 WRITE(6,20) INT1 2250 20 FORMAT(I<N+1>) 2251@end smallexample 2252 2253@noindent 2254with the following: 2255 2256@smallexample 2257c Variable declaration 2258 CHARACTER(LEN=20) FMT 2259c 2260c Other code here... 2261c 2262 WRITE(FMT,'("(I", I0, ")")') N+1 2263 WRITE(6,FMT) INT1 2264@end smallexample 2265 2266@noindent 2267or with: 2268 2269@smallexample 2270c Variable declaration 2271 CHARACTER(LEN=20) FMT 2272c 2273c Other code here... 2274c 2275 WRITE(FMT,*) N+1 2276 WRITE(6,"(I" // ADJUSTL(FMT) // ")") INT1 2277@end smallexample 2278 2279 2280@node Alternate complex function syntax 2281@subsection Alternate complex function syntax 2282@cindex Complex function 2283 2284Some Fortran compilers, including @command{g77}, let the user declare 2285complex functions with the syntax @code{COMPLEX FUNCTION name*16()}, as 2286well as @code{COMPLEX*16 FUNCTION name()}. Both are non-standard, legacy 2287extensions. @command{gfortran} accepts the latter form, which is more 2288common, but not the former. 2289 2290 2291@node Volatile COMMON blocks 2292@subsection Volatile @code{COMMON} blocks 2293@cindex @code{VOLATILE} 2294@cindex @code{COMMON} 2295 2296Some Fortran compilers, including @command{g77}, let the user declare 2297@code{COMMON} with the @code{VOLATILE} attribute. This is 2298invalid standard Fortran syntax and is not supported by 2299@command{gfortran}. Note that @command{gfortran} accepts 2300@code{VOLATILE} variables in @code{COMMON} blocks since revision 4.3. 2301 2302 2303 2304@c --------------------------------------------------------------------- 2305@c Mixed-Language Programming 2306@c --------------------------------------------------------------------- 2307 2308@node Mixed-Language Programming 2309@chapter Mixed-Language Programming 2310@cindex Interoperability 2311@cindex Mixed-language programming 2312 2313@menu 2314* Interoperability with C:: 2315* GNU Fortran Compiler Directives:: 2316* Non-Fortran Main Program:: 2317* Naming and argument-passing conventions:: 2318@end menu 2319 2320This chapter is about mixed-language interoperability, but also applies 2321if one links Fortran code compiled by different compilers. In most cases, 2322use of the C Binding features of the Fortran 2003 standard is sufficient, 2323and their use is highly recommended. 2324 2325 2326@node Interoperability with C 2327@section Interoperability with C 2328 2329@menu 2330* Intrinsic Types:: 2331* Derived Types and struct:: 2332* Interoperable Global Variables:: 2333* Interoperable Subroutines and Functions:: 2334* Working with Pointers:: 2335* Further Interoperability of Fortran with C:: 2336@end menu 2337 2338Since Fortran 2003 (ISO/IEC 1539-1:2004(E)) there is a 2339standardized way to generate procedure and derived-type 2340declarations and global variables which are interoperable with C 2341(ISO/IEC 9899:1999). The @code{bind(C)} attribute has been added 2342to inform the compiler that a symbol shall be interoperable with C; 2343also, some constraints are added. Note, however, that not 2344all C features have a Fortran equivalent or vice versa. For instance, 2345neither C's unsigned integers nor C's functions with variable number 2346of arguments have an equivalent in Fortran. 2347 2348Note that array dimensions are reversely ordered in C and that arrays in 2349C always start with index 0 while in Fortran they start by default with 23501. Thus, an array declaration @code{A(n,m)} in Fortran matches 2351@code{A[m][n]} in C and accessing the element @code{A(i,j)} matches 2352@code{A[j-1][i-1]}. The element following @code{A(i,j)} (C: @code{A[j-1][i-1]}; 2353assuming @math{i < n}) in memory is @code{A(i+1,j)} (C: @code{A[j-1][i]}). 2354 2355@node Intrinsic Types 2356@subsection Intrinsic Types 2357 2358In order to ensure that exactly the same variable type and kind is used 2359in C and Fortran, the named constants shall be used which are defined in the 2360@code{ISO_C_BINDING} intrinsic module. That module contains named constants 2361for kind parameters and character named constants for the escape sequences 2362in C. For a list of the constants, see @ref{ISO_C_BINDING}. 2363 2364For logical types, please note that the Fortran standard only guarantees 2365interoperability between C99's @code{_Bool} and Fortran's @code{C_Bool}-kind 2366logicals and C99 defines that @code{true} has the value 1 and @code{false} 2367the value 0. Using any other integer value with GNU Fortran's @code{LOGICAL} 2368(with any kind parameter) gives an undefined result. (Passing other integer 2369values than 0 and 1 to GCC's @code{_Bool} is also undefined, unless the 2370integer is explicitly or implicitly casted to @code{_Bool}.) 2371 2372 2373 2374@node Derived Types and struct 2375@subsection Derived Types and struct 2376 2377For compatibility of derived types with @code{struct}, one needs to use 2378the @code{BIND(C)} attribute in the type declaration. For instance, the 2379following type declaration 2380 2381@smallexample 2382 USE ISO_C_BINDING 2383 TYPE, BIND(C) :: myType 2384 INTEGER(C_INT) :: i1, i2 2385 INTEGER(C_SIGNED_CHAR) :: i3 2386 REAL(C_DOUBLE) :: d1 2387 COMPLEX(C_FLOAT_COMPLEX) :: c1 2388 CHARACTER(KIND=C_CHAR) :: str(5) 2389 END TYPE 2390@end smallexample 2391 2392matches the following @code{struct} declaration in C 2393 2394@smallexample 2395 struct @{ 2396 int i1, i2; 2397 /* Note: "char" might be signed or unsigned. */ 2398 signed char i3; 2399 double d1; 2400 float _Complex c1; 2401 char str[5]; 2402 @} myType; 2403@end smallexample 2404 2405Derived types with the C binding attribute shall not have the @code{sequence} 2406attribute, type parameters, the @code{extends} attribute, nor type-bound 2407procedures. Every component must be of interoperable type and kind and may not 2408have the @code{pointer} or @code{allocatable} attribute. The names of the 2409components are irrelevant for interoperability. 2410 2411As there exist no direct Fortran equivalents, neither unions nor structs 2412with bit field or variable-length array members are interoperable. 2413 2414@node Interoperable Global Variables 2415@subsection Interoperable Global Variables 2416 2417Variables can be made accessible from C using the C binding attribute, 2418optionally together with specifying a binding name. Those variables 2419have to be declared in the declaration part of a @code{MODULE}, 2420be of interoperable type, and have neither the @code{pointer} nor 2421the @code{allocatable} attribute. 2422 2423@smallexample 2424 MODULE m 2425 USE myType_module 2426 USE ISO_C_BINDING 2427 integer(C_INT), bind(C, name="_MyProject_flags") :: global_flag 2428 type(myType), bind(C) :: tp 2429 END MODULE 2430@end smallexample 2431 2432Here, @code{_MyProject_flags} is the case-sensitive name of the variable 2433as seen from C programs while @code{global_flag} is the case-insensitive 2434name as seen from Fortran. If no binding name is specified, as for 2435@var{tp}, the C binding name is the (lowercase) Fortran binding name. 2436If a binding name is specified, only a single variable may be after the 2437double colon. Note of warning: You cannot use a global variable to 2438access @var{errno} of the C library as the C standard allows it to be 2439a macro. Use the @code{IERRNO} intrinsic (GNU extension) instead. 2440 2441@node Interoperable Subroutines and Functions 2442@subsection Interoperable Subroutines and Functions 2443 2444Subroutines and functions have to have the @code{BIND(C)} attribute to 2445be compatible with C. The dummy argument declaration is relatively 2446straightforward. However, one needs to be careful because C uses 2447call-by-value by default while Fortran behaves usually similar to 2448call-by-reference. Furthermore, strings and pointers are handled 2449differently. Note that in Fortran 2003 and 2008 only explicit size 2450and assumed-size arrays are supported but not assumed-shape or 2451deferred-shape (i.e. allocatable or pointer) arrays. However, those 2452are allowed since the Technical Specification 29113, see 2453@ref{Further Interoperability of Fortran with C} 2454 2455To pass a variable by value, use the @code{VALUE} attribute. 2456Thus, the following C prototype 2457 2458@smallexample 2459@code{int func(int i, int *j)} 2460@end smallexample 2461 2462matches the Fortran declaration 2463 2464@smallexample 2465 integer(c_int) function func(i,j) 2466 use iso_c_binding, only: c_int 2467 integer(c_int), VALUE :: i 2468 integer(c_int) :: j 2469@end smallexample 2470 2471Note that pointer arguments also frequently need the @code{VALUE} attribute, 2472see @ref{Working with Pointers}. 2473 2474Strings are handled quite differently in C and Fortran. In C a string 2475is a @code{NUL}-terminated array of characters while in Fortran each string 2476has a length associated with it and is thus not terminated (by e.g. 2477@code{NUL}). For example, if one wants to use the following C function, 2478 2479@smallexample 2480 #include <stdio.h> 2481 void print_C(char *string) /* equivalent: char string[] */ 2482 @{ 2483 printf("%s\n", string); 2484 @} 2485@end smallexample 2486 2487to print ``Hello World'' from Fortran, one can call it using 2488 2489@smallexample 2490 use iso_c_binding, only: C_CHAR, C_NULL_CHAR 2491 interface 2492 subroutine print_c(string) bind(C, name="print_C") 2493 use iso_c_binding, only: c_char 2494 character(kind=c_char) :: string(*) 2495 end subroutine print_c 2496 end interface 2497 call print_c(C_CHAR_"Hello World"//C_NULL_CHAR) 2498@end smallexample 2499 2500As the example shows, one needs to ensure that the 2501string is @code{NUL} terminated. Additionally, the dummy argument 2502@var{string} of @code{print_C} is a length-one assumed-size 2503array; using @code{character(len=*)} is not allowed. The example 2504above uses @code{c_char_"Hello World"} to ensure the string 2505literal has the right type; typically the default character 2506kind and @code{c_char} are the same and thus @code{"Hello World"} 2507is equivalent. However, the standard does not guarantee this. 2508 2509The use of strings is now further illustrated using the C library 2510function @code{strncpy}, whose prototype is 2511 2512@smallexample 2513 char *strncpy(char *restrict s1, const char *restrict s2, size_t n); 2514@end smallexample 2515 2516The function @code{strncpy} copies at most @var{n} characters from 2517string @var{s2} to @var{s1} and returns @var{s1}. In the following 2518example, we ignore the return value: 2519 2520@smallexample 2521 use iso_c_binding 2522 implicit none 2523 character(len=30) :: str,str2 2524 interface 2525 ! Ignore the return value of strncpy -> subroutine 2526 ! "restrict" is always assumed if we do not pass a pointer 2527 subroutine strncpy(dest, src, n) bind(C) 2528 import 2529 character(kind=c_char), intent(out) :: dest(*) 2530 character(kind=c_char), intent(in) :: src(*) 2531 integer(c_size_t), value, intent(in) :: n 2532 end subroutine strncpy 2533 end interface 2534 str = repeat('X',30) ! Initialize whole string with 'X' 2535 call strncpy(str, c_char_"Hello World"//C_NULL_CHAR, & 2536 len(c_char_"Hello World",kind=c_size_t)) 2537 print '(a)', str ! prints: "Hello WorldXXXXXXXXXXXXXXXXXXX" 2538 end 2539@end smallexample 2540 2541The intrinsic procedures are described in @ref{Intrinsic Procedures}. 2542 2543@node Working with Pointers 2544@subsection Working with Pointers 2545 2546C pointers are represented in Fortran via the special opaque derived type 2547@code{type(c_ptr)} (with private components). Thus one needs to 2548use intrinsic conversion procedures to convert from or to C pointers. 2549 2550For some applications, using an assumed type (@code{TYPE(*)}) can be an 2551alternative to a C pointer; see 2552@ref{Further Interoperability of Fortran with C}. 2553 2554For example, 2555 2556@smallexample 2557 use iso_c_binding 2558 type(c_ptr) :: cptr1, cptr2 2559 integer, target :: array(7), scalar 2560 integer, pointer :: pa(:), ps 2561 cptr1 = c_loc(array(1)) ! The programmer needs to ensure that the 2562 ! array is contiguous if required by the C 2563 ! procedure 2564 cptr2 = c_loc(scalar) 2565 call c_f_pointer(cptr2, ps) 2566 call c_f_pointer(cptr2, pa, shape=[7]) 2567@end smallexample 2568 2569When converting C to Fortran arrays, the one-dimensional @code{SHAPE} argument 2570has to be passed. 2571 2572If a pointer is a dummy-argument of an interoperable procedure, it usually 2573has to be declared using the @code{VALUE} attribute. @code{void*} 2574matches @code{TYPE(C_PTR), VALUE}, while @code{TYPE(C_PTR)} alone 2575matches @code{void**}. 2576 2577Procedure pointers are handled analogously to pointers; the C type is 2578@code{TYPE(C_FUNPTR)} and the intrinsic conversion procedures are 2579@code{C_F_PROCPOINTER} and @code{C_FUNLOC}. 2580 2581Let us consider two examples of actually passing a procedure pointer from 2582C to Fortran and vice versa. Note that these examples are also very 2583similar to passing ordinary pointers between both languages. First, 2584consider this code in C: 2585 2586@smallexample 2587/* Procedure implemented in Fortran. */ 2588void get_values (void (*)(double)); 2589 2590/* Call-back routine we want called from Fortran. */ 2591void 2592print_it (double x) 2593@{ 2594 printf ("Number is %f.\n", x); 2595@} 2596 2597/* Call Fortran routine and pass call-back to it. */ 2598void 2599foobar () 2600@{ 2601 get_values (&print_it); 2602@} 2603@end smallexample 2604 2605A matching implementation for @code{get_values} in Fortran, that correctly 2606receives the procedure pointer from C and is able to call it, is given 2607in the following @code{MODULE}: 2608 2609@smallexample 2610MODULE m 2611 IMPLICIT NONE 2612 2613 ! Define interface of call-back routine. 2614 ABSTRACT INTERFACE 2615 SUBROUTINE callback (x) 2616 USE, INTRINSIC :: ISO_C_BINDING 2617 REAL(KIND=C_DOUBLE), INTENT(IN), VALUE :: x 2618 END SUBROUTINE callback 2619 END INTERFACE 2620 2621CONTAINS 2622 2623 ! Define C-bound procedure. 2624 SUBROUTINE get_values (cproc) BIND(C) 2625 USE, INTRINSIC :: ISO_C_BINDING 2626 TYPE(C_FUNPTR), INTENT(IN), VALUE :: cproc 2627 2628 PROCEDURE(callback), POINTER :: proc 2629 2630 ! Convert C to Fortran procedure pointer. 2631 CALL C_F_PROCPOINTER (cproc, proc) 2632 2633 ! Call it. 2634 CALL proc (1.0_C_DOUBLE) 2635 CALL proc (-42.0_C_DOUBLE) 2636 CALL proc (18.12_C_DOUBLE) 2637 END SUBROUTINE get_values 2638 2639END MODULE m 2640@end smallexample 2641 2642Next, we want to call a C routine that expects a procedure pointer argument 2643and pass it a Fortran procedure (which clearly must be interoperable!). 2644Again, the C function may be: 2645 2646@smallexample 2647int 2648call_it (int (*func)(int), int arg) 2649@{ 2650 return func (arg); 2651@} 2652@end smallexample 2653 2654It can be used as in the following Fortran code: 2655 2656@smallexample 2657MODULE m 2658 USE, INTRINSIC :: ISO_C_BINDING 2659 IMPLICIT NONE 2660 2661 ! Define interface of C function. 2662 INTERFACE 2663 INTEGER(KIND=C_INT) FUNCTION call_it (func, arg) BIND(C) 2664 USE, INTRINSIC :: ISO_C_BINDING 2665 TYPE(C_FUNPTR), INTENT(IN), VALUE :: func 2666 INTEGER(KIND=C_INT), INTENT(IN), VALUE :: arg 2667 END FUNCTION call_it 2668 END INTERFACE 2669 2670CONTAINS 2671 2672 ! Define procedure passed to C function. 2673 ! It must be interoperable! 2674 INTEGER(KIND=C_INT) FUNCTION double_it (arg) BIND(C) 2675 INTEGER(KIND=C_INT), INTENT(IN), VALUE :: arg 2676 double_it = arg + arg 2677 END FUNCTION double_it 2678 2679 ! Call C function. 2680 SUBROUTINE foobar () 2681 TYPE(C_FUNPTR) :: cproc 2682 INTEGER(KIND=C_INT) :: i 2683 2684 ! Get C procedure pointer. 2685 cproc = C_FUNLOC (double_it) 2686 2687 ! Use it. 2688 DO i = 1_C_INT, 10_C_INT 2689 PRINT *, call_it (cproc, i) 2690 END DO 2691 END SUBROUTINE foobar 2692 2693END MODULE m 2694@end smallexample 2695 2696@node Further Interoperability of Fortran with C 2697@subsection Further Interoperability of Fortran with C 2698 2699The Technical Specification ISO/IEC TS 29113:2012 on further 2700interoperability of Fortran with C extends the interoperability support 2701of Fortran 2003 and Fortran 2008. Besides removing some restrictions 2702and constraints, it adds assumed-type (@code{TYPE(*)}) and assumed-rank 2703(@code{dimension}) variables and allows for interoperability of 2704assumed-shape, assumed-rank and deferred-shape arrays, including 2705allocatables and pointers. 2706 2707Note: Currently, GNU Fortran does not support the array descriptor 2708(dope vector) as specified in the Technical Specification, but uses 2709an array descriptor with different fields. The Chasm Language 2710Interoperability Tools, @url{http://chasm-interop.sourceforge.net/}, 2711provide an interface to GNU Fortran's array descriptor. 2712 2713The Technical Specification adds the following new features, which 2714are supported by GNU Fortran: 2715 2716@itemize @bullet 2717 2718@item The @code{ASYNCHRONOUS} attribute has been clarified and 2719extended to allow its use with asynchronous communication in 2720user-provided libraries such as in implementations of the 2721Message Passing Interface specification. 2722 2723@item Many constraints have been relaxed, in particular for 2724the @code{C_LOC} and @code{C_F_POINTER} intrinsics. 2725 2726@item The @code{OPTIONAL} attribute is now allowed for dummy 2727arguments; an absent argument matches a @code{NULL} pointer. 2728 2729@item Assumed types (@code{TYPE(*)}) have been added, which may 2730only be used for dummy arguments. They are unlimited polymorphic 2731but contrary to @code{CLASS(*)} they do not contain any type 2732information, similar to C's @code{void *} pointers. Expressions 2733of any type and kind can be passed; thus, it can be used as 2734replacement for @code{TYPE(C_PTR)}, avoiding the use of 2735@code{C_LOC} in the caller. 2736 2737Note, however, that @code{TYPE(*)} only accepts scalar arguments, 2738unless the @code{DIMENSION} is explicitly specified. As 2739@code{DIMENSION(*)} only supports array (including array elements) but 2740no scalars, it is not a full replacement for @code{C_LOC}. On the 2741other hand, assumed-type assumed-rank dummy arguments 2742(@code{TYPE(*), DIMENSION(..)}) allow for both scalars and arrays, but 2743require special code on the callee side to handle the array descriptor. 2744 2745@item Assumed-rank arrays (@code{DIMENSION(..)}) as dummy argument 2746allow that scalars and arrays of any rank can be passed as actual 2747argument. As the Technical Specification does not provide for direct 2748means to operate with them, they have to be used either from the C side 2749or be converted using @code{C_LOC} and @code{C_F_POINTER} to scalars 2750or arrays of a specific rank. The rank can be determined using the 2751@code{RANK} intrinisic. 2752@end itemize 2753 2754 2755Currently unimplemented: 2756 2757@itemize @bullet 2758 2759@item GNU Fortran always uses an array descriptor, which does not 2760match the one of the Technical Specification. The 2761@code{ISO_Fortran_binding.h} header file and the C functions it 2762specifies are not available. 2763 2764@item Using assumed-shape, assumed-rank and deferred-shape arrays in 2765@code{BIND(C)} procedures is not fully supported. In particular, 2766C interoperable strings of other length than one are not supported 2767as this requires the new array descriptor. 2768@end itemize 2769 2770 2771@node GNU Fortran Compiler Directives 2772@section GNU Fortran Compiler Directives 2773 2774The Fortran standard describes how a conforming program shall 2775behave; however, the exact implementation is not standardized. In order 2776to allow the user to choose specific implementation details, compiler 2777directives can be used to set attributes of variables and procedures 2778which are not part of the standard. Whether a given attribute is 2779supported and its exact effects depend on both the operating system and 2780on the processor; see 2781@ref{Top,,C Extensions,gcc,Using the GNU Compiler Collection (GCC)} 2782for details. 2783 2784For procedures and procedure pointers, the following attributes can 2785be used to change the calling convention: 2786 2787@itemize 2788@item @code{CDECL} -- standard C calling convention 2789@item @code{STDCALL} -- convention where the called procedure pops the stack 2790@item @code{FASTCALL} -- part of the arguments are passed via registers 2791instead using the stack 2792@end itemize 2793 2794Besides changing the calling convention, the attributes also influence 2795the decoration of the symbol name, e.g., by a leading underscore or by 2796a trailing at-sign followed by the number of bytes on the stack. When 2797assigning a procedure to a procedure pointer, both should use the same 2798calling convention. 2799 2800On some systems, procedures and global variables (module variables and 2801@code{COMMON} blocks) need special handling to be accessible when they 2802are in a shared library. The following attributes are available: 2803 2804@itemize 2805@item @code{DLLEXPORT} -- provide a global pointer to a pointer in the DLL 2806@item @code{DLLIMPORT} -- reference the function or variable using a 2807global pointer 2808@end itemize 2809 2810For dummy arguments, the @code{NO_ARG_CHECK} attribute can be used; in 2811other compilers, it is also known as @code{IGNORE_TKR}. For dummy arguments 2812with this attribute actual arguments of any type and kind (similar to 2813@code{TYPE(*)}), scalars and arrays of any rank (no equivalent 2814in Fortran standard) are accepted. As with @code{TYPE(*)}, the argument 2815is unlimited polymorphic and no type information is available. 2816Additionally, the argument may only be passed to dummy arguments 2817with the @code{NO_ARG_CHECK} attribute and as argument to the 2818@code{PRESENT} intrinsic function and to @code{C_LOC} of the 2819@code{ISO_C_BINDING} module. 2820 2821Variables with @code{NO_ARG_CHECK} attribute shall be of assumed-type 2822(@code{TYPE(*)}; recommended) or of type @code{INTEGER}, @code{LOGICAL}, 2823@code{REAL} or @code{COMPLEX}. They shall not have the @code{ALLOCATE}, 2824@code{CODIMENSION}, @code{INTENT(OUT)}, @code{POINTER} or @code{VALUE} 2825attribute; furthermore, they shall be either scalar or of assumed-size 2826(@code{dimension(*)}). As @code{TYPE(*)}, the @code{NO_ARG_CHECK} attribute 2827requires an explicit interface. 2828 2829@itemize 2830@item @code{NO_ARG_CHECK} -- disable the type, kind and rank checking 2831@end itemize 2832 2833 2834The attributes are specified using the syntax 2835 2836@code{!GCC$ ATTRIBUTES} @var{attribute-list} @code{::} @var{variable-list} 2837 2838where in free-form source code only whitespace is allowed before @code{!GCC$} 2839and in fixed-form source code @code{!GCC$}, @code{cGCC$} or @code{*GCC$} shall 2840start in the first column. 2841 2842For procedures, the compiler directives shall be placed into the body 2843of the procedure; for variables and procedure pointers, they shall be in 2844the same declaration part as the variable or procedure pointer. 2845 2846 2847 2848@node Non-Fortran Main Program 2849@section Non-Fortran Main Program 2850 2851@menu 2852* _gfortran_set_args:: Save command-line arguments 2853* _gfortran_set_options:: Set library option flags 2854* _gfortran_set_convert:: Set endian conversion 2855* _gfortran_set_record_marker:: Set length of record markers 2856* _gfortran_set_fpe:: Set when a Floating Point Exception should be raised 2857* _gfortran_set_max_subrecord_length:: Set subrecord length 2858@end menu 2859 2860Even if you are doing mixed-language programming, it is very 2861likely that you do not need to know or use the information in this 2862section. Since it is about the internal structure of GNU Fortran, 2863it may also change in GCC minor releases. 2864 2865When you compile a @code{PROGRAM} with GNU Fortran, a function 2866with the name @code{main} (in the symbol table of the object file) 2867is generated, which initializes the libgfortran library and then 2868calls the actual program which uses the name @code{MAIN__}, for 2869historic reasons. If you link GNU Fortran compiled procedures 2870to, e.g., a C or C++ program or to a Fortran program compiled by 2871a different compiler, the libgfortran library is not initialized 2872and thus a few intrinsic procedures do not work properly, e.g. 2873those for obtaining the command-line arguments. 2874 2875Therefore, if your @code{PROGRAM} is not compiled with 2876GNU Fortran and the GNU Fortran compiled procedures require 2877intrinsics relying on the library initialization, you need to 2878initialize the library yourself. Using the default options, 2879gfortran calls @code{_gfortran_set_args} and 2880@code{_gfortran_set_options}. The initialization of the former 2881is needed if the called procedures access the command line 2882(and for backtracing); the latter sets some flags based on the 2883standard chosen or to enable backtracing. In typical programs, 2884it is not necessary to call any initialization function. 2885 2886If your @code{PROGRAM} is compiled with GNU Fortran, you shall 2887not call any of the following functions. The libgfortran 2888initialization functions are shown in C syntax but using C 2889bindings they are also accessible from Fortran. 2890 2891 2892@node _gfortran_set_args 2893@subsection @code{_gfortran_set_args} --- Save command-line arguments 2894@fnindex _gfortran_set_args 2895@cindex libgfortran initialization, set_args 2896 2897@table @asis 2898@item @emph{Description}: 2899@code{_gfortran_set_args} saves the command-line arguments; this 2900initialization is required if any of the command-line intrinsics 2901is called. Additionally, it shall be called if backtracing is 2902enabled (see @code{_gfortran_set_options}). 2903 2904@item @emph{Syntax}: 2905@code{void _gfortran_set_args (int argc, char *argv[])} 2906 2907@item @emph{Arguments}: 2908@multitable @columnfractions .15 .70 2909@item @var{argc} @tab number of command line argument strings 2910@item @var{argv} @tab the command-line argument strings; argv[0] 2911is the pathname of the executable itself. 2912@end multitable 2913 2914@item @emph{Example}: 2915@smallexample 2916int main (int argc, char *argv[]) 2917@{ 2918 /* Initialize libgfortran. */ 2919 _gfortran_set_args (argc, argv); 2920 return 0; 2921@} 2922@end smallexample 2923@end table 2924 2925 2926@node _gfortran_set_options 2927@subsection @code{_gfortran_set_options} --- Set library option flags 2928@fnindex _gfortran_set_options 2929@cindex libgfortran initialization, set_options 2930 2931@table @asis 2932@item @emph{Description}: 2933@code{_gfortran_set_options} sets several flags related to the Fortran 2934standard to be used, whether backtracing should be enabled 2935and whether range checks should be performed. The syntax allows for 2936upward compatibility since the number of passed flags is specified; for 2937non-passed flags, the default value is used. See also 2938@pxref{Code Gen Options}. Please note that not all flags are actually 2939used. 2940 2941@item @emph{Syntax}: 2942@code{void _gfortran_set_options (int num, int options[])} 2943 2944@item @emph{Arguments}: 2945@multitable @columnfractions .15 .70 2946@item @var{num} @tab number of options passed 2947@item @var{argv} @tab The list of flag values 2948@end multitable 2949 2950@item @emph{option flag list}: 2951@multitable @columnfractions .15 .70 2952@item @var{option}[0] @tab Allowed standard; can give run-time errors 2953if e.g. an input-output edit descriptor is invalid in a given standard. 2954Possible values are (bitwise or-ed) @code{GFC_STD_F77} (1), 2955@code{GFC_STD_F95_OBS} (2), @code{GFC_STD_F95_DEL} (4), @code{GFC_STD_F95} 2956(8), @code{GFC_STD_F2003} (16), @code{GFC_STD_GNU} (32), 2957@code{GFC_STD_LEGACY} (64), @code{GFC_STD_F2008} (128), 2958@code{GFC_STD_F2008_OBS} (256) and GFC_STD_F2008_TS (512). Default: 2959@code{GFC_STD_F95_OBS | GFC_STD_F95_DEL | GFC_STD_F95 | GFC_STD_F2003 2960| GFC_STD_F2008 | GFC_STD_F2008_TS | GFC_STD_F2008_OBS | GFC_STD_F77 2961| GFC_STD_GNU | GFC_STD_LEGACY}. 2962@item @var{option}[1] @tab Standard-warning flag; prints a warning to 2963standard error. Default: @code{GFC_STD_F95_DEL | GFC_STD_LEGACY}. 2964@item @var{option}[2] @tab If non zero, enable pedantic checking. 2965Default: off. 2966@item @var{option}[3] @tab Unused. 2967@item @var{option}[4] @tab If non zero, enable backtracing on run-time 2968errors. Default: off. (Default in the compiler: on.) 2969Note: Installs a signal handler and requires command-line 2970initialization using @code{_gfortran_set_args}. 2971@item @var{option}[5] @tab If non zero, supports signed zeros. 2972Default: enabled. 2973@item @var{option}[6] @tab Enables run-time checking. Possible values 2974are (bitwise or-ed): GFC_RTCHECK_BOUNDS (1), GFC_RTCHECK_ARRAY_TEMPS (2), 2975GFC_RTCHECK_RECURSION (4), GFC_RTCHECK_DO (16), GFC_RTCHECK_POINTER (32). 2976Default: disabled. 2977@item @var{option}[7] @tab Unused. 2978@item @var{option}[8] @tab Show a warning when invoking @code{STOP} and 2979@code{ERROR STOP} if a floating-point exception occurred. Possible values 2980are (bitwise or-ed) @code{GFC_FPE_INVALID} (1), @code{GFC_FPE_DENORMAL} (2), 2981@code{GFC_FPE_ZERO} (4), @code{GFC_FPE_OVERFLOW} (8), 2982@code{GFC_FPE_UNDERFLOW} (16), @code{GFC_FPE_INEXACT} (32). Default: None (0). 2983(Default in the compiler: @code{GFC_FPE_INVALID | GFC_FPE_DENORMAL | 2984GFC_FPE_ZERO | GFC_FPE_OVERFLOW | GFC_FPE_UNDERFLOW}.) 2985@end multitable 2986 2987@item @emph{Example}: 2988@smallexample 2989 /* Use gfortran 4.9 default options. */ 2990 static int options[] = @{68, 511, 0, 0, 1, 1, 0, 0, 31@}; 2991 _gfortran_set_options (9, &options); 2992@end smallexample 2993@end table 2994 2995 2996@node _gfortran_set_convert 2997@subsection @code{_gfortran_set_convert} --- Set endian conversion 2998@fnindex _gfortran_set_convert 2999@cindex libgfortran initialization, set_convert 3000 3001@table @asis 3002@item @emph{Description}: 3003@code{_gfortran_set_convert} set the representation of data for 3004unformatted files. 3005 3006@item @emph{Syntax}: 3007@code{void _gfortran_set_convert (int conv)} 3008 3009@item @emph{Arguments}: 3010@multitable @columnfractions .15 .70 3011@item @var{conv} @tab Endian conversion, possible values: 3012GFC_CONVERT_NATIVE (0, default), GFC_CONVERT_SWAP (1), 3013GFC_CONVERT_BIG (2), GFC_CONVERT_LITTLE (3). 3014@end multitable 3015 3016@item @emph{Example}: 3017@smallexample 3018int main (int argc, char *argv[]) 3019@{ 3020 /* Initialize libgfortran. */ 3021 _gfortran_set_args (argc, argv); 3022 _gfortran_set_convert (1); 3023 return 0; 3024@} 3025@end smallexample 3026@end table 3027 3028 3029@node _gfortran_set_record_marker 3030@subsection @code{_gfortran_set_record_marker} --- Set length of record markers 3031@fnindex _gfortran_set_record_marker 3032@cindex libgfortran initialization, set_record_marker 3033 3034@table @asis 3035@item @emph{Description}: 3036@code{_gfortran_set_record_marker} sets the length of record markers 3037for unformatted files. 3038 3039@item @emph{Syntax}: 3040@code{void _gfortran_set_record_marker (int val)} 3041 3042@item @emph{Arguments}: 3043@multitable @columnfractions .15 .70 3044@item @var{val} @tab Length of the record marker; valid values 3045are 4 and 8. Default is 4. 3046@end multitable 3047 3048@item @emph{Example}: 3049@smallexample 3050int main (int argc, char *argv[]) 3051@{ 3052 /* Initialize libgfortran. */ 3053 _gfortran_set_args (argc, argv); 3054 _gfortran_set_record_marker (8); 3055 return 0; 3056@} 3057@end smallexample 3058@end table 3059 3060 3061@node _gfortran_set_fpe 3062@subsection @code{_gfortran_set_fpe} --- Enable floating point exception traps 3063@fnindex _gfortran_set_fpe 3064@cindex libgfortran initialization, set_fpe 3065 3066@table @asis 3067@item @emph{Description}: 3068@code{_gfortran_set_fpe} enables floating point exception traps for 3069the specified exceptions. On most systems, this will result in a 3070SIGFPE signal being sent and the program being aborted. 3071 3072@item @emph{Syntax}: 3073@code{void _gfortran_set_fpe (int val)} 3074 3075@item @emph{Arguments}: 3076@multitable @columnfractions .15 .70 3077@item @var{option}[0] @tab IEEE exceptions. Possible values are 3078(bitwise or-ed) zero (0, default) no trapping, 3079@code{GFC_FPE_INVALID} (1), @code{GFC_FPE_DENORMAL} (2), 3080@code{GFC_FPE_ZERO} (4), @code{GFC_FPE_OVERFLOW} (8), 3081@code{GFC_FPE_UNDERFLOW} (16), and @code{GFC_FPE_INEXACT} (32). 3082@end multitable 3083 3084@item @emph{Example}: 3085@smallexample 3086int main (int argc, char *argv[]) 3087@{ 3088 /* Initialize libgfortran. */ 3089 _gfortran_set_args (argc, argv); 3090 /* FPE for invalid operations such as SQRT(-1.0). */ 3091 _gfortran_set_fpe (1); 3092 return 0; 3093@} 3094@end smallexample 3095@end table 3096 3097 3098@node _gfortran_set_max_subrecord_length 3099@subsection @code{_gfortran_set_max_subrecord_length} --- Set subrecord length 3100@fnindex _gfortran_set_max_subrecord_length 3101@cindex libgfortran initialization, set_max_subrecord_length 3102 3103@table @asis 3104@item @emph{Description}: 3105@code{_gfortran_set_max_subrecord_length} set the maximum length 3106for a subrecord. This option only makes sense for testing and 3107debugging of unformatted I/O. 3108 3109@item @emph{Syntax}: 3110@code{void _gfortran_set_max_subrecord_length (int val)} 3111 3112@item @emph{Arguments}: 3113@multitable @columnfractions .15 .70 3114@item @var{val} @tab the maximum length for a subrecord; 3115the maximum permitted value is 2147483639, which is also 3116the default. 3117@end multitable 3118 3119@item @emph{Example}: 3120@smallexample 3121int main (int argc, char *argv[]) 3122@{ 3123 /* Initialize libgfortran. */ 3124 _gfortran_set_args (argc, argv); 3125 _gfortran_set_max_subrecord_length (8); 3126 return 0; 3127@} 3128@end smallexample 3129@end table 3130 3131 3132@node Naming and argument-passing conventions 3133@section Naming and argument-passing conventions 3134 3135This section gives an overview about the naming convention of procedures 3136and global variables and about the argument passing conventions used by 3137GNU Fortran. If a C binding has been specified, the naming convention 3138and some of the argument-passing conventions change. If possible, 3139mixed-language and mixed-compiler projects should use the better defined 3140C binding for interoperability. See @pxref{Interoperability with C}. 3141 3142@menu 3143* Naming conventions:: 3144* Argument passing conventions:: 3145@end menu 3146 3147 3148@node Naming conventions 3149@subsection Naming conventions 3150 3151According the Fortran standard, valid Fortran names consist of a letter 3152between @code{A} to @code{Z}, @code{a} to @code{z}, digits @code{0}, 3153@code{1} to @code{9} and underscores (@code{_}) with the restriction 3154that names may only start with a letter. As vendor extension, the 3155dollar sign (@code{$}) is additionally permitted with the option 3156@option{-fdollar-ok}, but not as first character and only if the 3157target system supports it. 3158 3159By default, the procedure name is the lower-cased Fortran name with an 3160appended underscore (@code{_}); using @option{-fno-underscoring} no 3161underscore is appended while @code{-fsecond-underscore} appends two 3162underscores. Depending on the target system and the calling convention, 3163the procedure might be additionally dressed; for instance, on 32bit 3164Windows with @code{stdcall}, an at-sign @code{@@} followed by an integer 3165number is appended. For the changing the calling convention, see 3166@pxref{GNU Fortran Compiler Directives}. 3167 3168For common blocks, the same convention is used, i.e. by default an 3169underscore is appended to the lower-cased Fortran name. Blank commons 3170have the name @code{__BLNK__}. 3171 3172For procedures and variables declared in the specification space of a 3173module, the name is formed by @code{__}, followed by the lower-cased 3174module name, @code{_MOD_}, and the lower-cased Fortran name. Note that 3175no underscore is appended. 3176 3177 3178@node Argument passing conventions 3179@subsection Argument passing conventions 3180 3181Subroutines do not return a value (matching C99's @code{void}) while 3182functions either return a value as specified in the platform ABI or 3183the result variable is passed as hidden argument to the function and 3184no result is returned. A hidden result variable is used when the 3185result variable is an array or of type @code{CHARACTER}. 3186 3187Arguments are passed according to the platform ABI. In particular, 3188complex arguments might not be compatible to a struct with two real 3189components for the real and imaginary part. The argument passing 3190matches the one of C99's @code{_Complex}. Functions with scalar 3191complex result variables return their value and do not use a 3192by-reference argument. Note that with the @option{-ff2c} option, 3193the argument passing is modified and no longer completely matches 3194the platform ABI. Some other Fortran compilers use @code{f2c} 3195semantic by default; this might cause problems with 3196interoperablility. 3197 3198GNU Fortran passes most arguments by reference, i.e. by passing a 3199pointer to the data. Note that the compiler might use a temporary 3200variable into which the actual argument has been copied, if required 3201semantically (copy-in/copy-out). 3202 3203For arguments with @code{ALLOCATABLE} and @code{POINTER} 3204attribute (including procedure pointers), a pointer to the pointer 3205is passed such that the pointer address can be modified in the 3206procedure. 3207 3208For dummy arguments with the @code{VALUE} attribute: Scalar arguments 3209of the type @code{INTEGER}, @code{LOGICAL}, @code{REAL} and 3210@code{COMPLEX} are passed by value according to the platform ABI. 3211(As vendor extension and not recommended, using @code{%VAL()} in the 3212call to a procedure has the same effect.) For @code{TYPE(C_PTR)} and 3213procedure pointers, the pointer itself is passed such that it can be 3214modified without affecting the caller. 3215@c FIXME: Document how VALUE is handled for CHARACTER, TYPE, 3216@c CLASS and arrays, i.e. whether the copy-in is done in the caller 3217@c or in the callee. 3218 3219For Boolean (@code{LOGICAL}) arguments, please note that GCC expects 3220only the integer value 0 and 1. If a GNU Fortran @code{LOGICAL} 3221variable contains another integer value, the result is undefined. 3222As some other Fortran compilers use @math{-1} for @code{.TRUE.}, 3223extra care has to be taken -- such as passing the value as 3224@code{INTEGER}. (The same value restriction also applies to other 3225front ends of GCC, e.g. to GCC's C99 compiler for @code{_Bool} 3226or GCC's Ada compiler for @code{Boolean}.) 3227 3228For arguments of @code{CHARACTER} type, the character length is passed 3229as hidden argument. For deferred-length strings, the value is passed 3230by reference, otherwise by value. The character length has the type 3231@code{INTEGER(kind=4)}. Note with C binding, @code{CHARACTER(len=1)} 3232result variables are returned according to the platform ABI and no 3233hidden length argument is used for dummy arguments; with @code{VALUE}, 3234those variables are passed by value. 3235 3236For @code{OPTIONAL} dummy arguments, an absent argument is denoted 3237by a NULL pointer, except for scalar dummy arguments of type 3238@code{INTEGER}, @code{LOGICAL}, @code{REAL} and @code{COMPLEX} 3239which have the @code{VALUE} attribute. For those, a hidden Boolean 3240argument (@code{logical(kind=C_bool),value}) is used to indicate 3241whether the argument is present. 3242 3243Arguments which are assumed-shape, assumed-rank or deferred-rank 3244arrays or, with @option{-fcoarray=lib}, allocatable scalar coarrays use 3245an array descriptor. All other arrays pass the address of the 3246first element of the array. With @option{-fcoarray=lib}, the token 3247and the offset belonging to nonallocatable coarrays dummy arguments 3248are passed as hidden argument along the character length hidden 3249arguments. The token is an oparque pointer identifying the coarray 3250and the offset is a passed-by-value integer of kind @code{C_PTRDIFF_T}, 3251denoting the byte offset between the base address of the coarray and 3252the passed scalar or first element of the passed array. 3253 3254The arguments are passed in the following order 3255@itemize @bullet 3256@item Result variable, when the function result is passed by reference 3257@item Character length of the function result, if it is a of type 3258@code{CHARACTER} and no C binding is used 3259@item The arguments in the order in which they appear in the Fortran 3260declaration 3261@item The the present status for optional arguments with value attribute, 3262which are internally passed by value 3263@item The character length and/or coarray token and offset for the first 3264argument which is a @code{CHARACTER} or a nonallocatable coarray dummy 3265argument, followed by the hidden arguments of the next dummy argument 3266of such a type 3267@end itemize 3268 3269 3270@c --------------------------------------------------------------------- 3271@c Coarray Programming 3272@c --------------------------------------------------------------------- 3273 3274@node Coarray Programming 3275@chapter Coarray Programming 3276@cindex Coarrays 3277 3278@menu 3279* Type and enum ABI Documentation:: 3280* Function ABI Documentation:: 3281@end menu 3282 3283 3284@node Type and enum ABI Documentation 3285@section Type and enum ABI Documentation 3286 3287@menu 3288* caf_token_t:: 3289* caf_register_t:: 3290@end menu 3291 3292@node caf_token_t 3293@subsection @code{caf_token_t} 3294 3295Typedef of type @code{void *} on the compiler side. Can be any data 3296type on the library side. 3297 3298@node caf_register_t 3299@subsection @code{caf_register_t} 3300 3301Indicates which kind of coarray variable should be registered. 3302 3303@verbatim 3304typedef enum caf_register_t { 3305 CAF_REGTYPE_COARRAY_STATIC, 3306 CAF_REGTYPE_COARRAY_ALLOC, 3307 CAF_REGTYPE_LOCK_STATIC, 3308 CAF_REGTYPE_LOCK_ALLOC, 3309 CAF_REGTYPE_CRITICAL, 3310 CAF_REGTYPE_EVENT_STATIC, 3311 CAF_REGTYPE_EVENT_ALLOC 3312} 3313caf_register_t; 3314@end verbatim 3315 3316 3317@node Function ABI Documentation 3318@section Function ABI Documentation 3319 3320@menu 3321* _gfortran_caf_init:: Initialiation function 3322* _gfortran_caf_finish:: Finalization function 3323* _gfortran_caf_this_image:: Querying the image number 3324* _gfortran_caf_num_images:: Querying the maximal number of images 3325* _gfortran_caf_register:: Registering coarrays 3326* _gfortran_caf_deregister:: Deregistering coarrays 3327* _gfortran_caf_send:: Sending data from a local image to a remote image 3328* _gfortran_caf_get:: Getting data from a remote image 3329* _gfortran_caf_sendget:: Sending data between remote images 3330* _gfortran_caf_lock:: Locking a lock variable 3331* _gfortran_caf_unlock:: Unlocking a lock variable 3332* _gfortran_caf_event_post:: Post an event 3333* _gfortran_caf_event_wait:: Wait that an event occurred 3334* _gfortran_caf_event_query:: Query event count 3335* _gfortran_caf_sync_all:: All-image barrier 3336* _gfortran_caf_sync_images:: Barrier for selected images 3337* _gfortran_caf_sync_memory:: Wait for completion of segment-memory operations 3338* _gfortran_caf_error_stop:: Error termination with exit code 3339* _gfortran_caf_error_stop_str:: Error termination with string 3340* _gfortran_caf_atomic_define:: Atomic variable assignment 3341* _gfortran_caf_atomic_ref:: Atomic variable reference 3342* _gfortran_caf_atomic_cas:: Atomic compare and swap 3343* _gfortran_caf_atomic_op:: Atomic operation 3344* _gfortran_caf_co_broadcast:: Sending data to all images 3345* _gfortran_caf_co_max:: Collective maximum reduction 3346* _gfortran_caf_co_min:: Collective minimum reduction 3347* _gfortran_caf_co_sum:: Collective summing reduction 3348* _gfortran_caf_co_reduce:: Generic collective reduction 3349@end menu 3350 3351 3352@node _gfortran_caf_init 3353@subsection @code{_gfortran_caf_init} --- Initialiation function 3354@cindex Coarray, _gfortran_caf_init 3355 3356@table @asis 3357@item @emph{Description}: 3358This function is called at startup of the program before the Fortran main 3359program, if the latter has been compiled with @option{-fcoarray=lib}. 3360It takes as arguments the command-line arguments of the program. It is 3361permitted to pass to @code{NULL} pointers as argument; if non-@code{NULL}, 3362the library is permitted to modify the arguments. 3363 3364@item @emph{Syntax}: 3365@code{void _gfortran_caf_init (int *argc, char ***argv)} 3366 3367@item @emph{Arguments}: 3368@multitable @columnfractions .15 .70 3369@item @var{argc} @tab intent(inout) An integer pointer with the number of 3370arguments passed to the program or @code{NULL}. 3371@item @var{argv} @tab intent(inout) A pointer to an array of strings with the 3372command-line arguments or @code{NULL}. 3373@end multitable 3374 3375@item @emph{NOTES} 3376The function is modelled after the initialization function of the Message 3377Passing Interface (MPI) specification. Due to the way coarray registration 3378works, it might not be the first call to the libaray. If the main program is 3379not written in Fortran and only a library uses coarrays, it can happen that 3380this function is never called. Therefore, it is recommended that the library 3381does not rely on the passed arguments and whether the call has been done. 3382@end table 3383 3384 3385@node _gfortran_caf_finish 3386@subsection @code{_gfortran_caf_finish} --- Finalization function 3387@cindex Coarray, _gfortran_caf_finish 3388 3389@table @asis 3390@item @emph{Description}: 3391This function is called at the end of the Fortran main program, if it has 3392been compiled with the @option{-fcoarray=lib} option. 3393 3394@item @emph{Syntax}: 3395@code{void _gfortran_caf_finish (void)} 3396 3397@item @emph{NOTES} 3398For non-Fortran programs, it is recommended to call the function at the end 3399of the main program. To ensure that the shutdown is also performed for 3400programs where this function is not explicitly invoked, for instance 3401non-Fortran programs or calls to the system's exit() function, the library 3402can use a destructor function. Note that programs can also be terminated 3403using the STOP and ERROR STOP statements; those use different library calls. 3404@end table 3405 3406 3407@node _gfortran_caf_this_image 3408@subsection @code{_gfortran_caf_this_image} --- Querying the image number 3409@cindex Coarray, _gfortran_caf_this_image 3410 3411@table @asis 3412@item @emph{Description}: 3413This function returns the current image number, which is a positive number. 3414 3415@item @emph{Syntax}: 3416@code{int _gfortran_caf_this_image (int distance)} 3417 3418@item @emph{Arguments}: 3419@multitable @columnfractions .15 .70 3420@item @var{distance} @tab As specified for the @code{this_image} intrinsic 3421in TS18508. Shall be a nonnegative number. 3422@end multitable 3423 3424@item @emph{NOTES} 3425If the Fortran intrinsic @code{this_image} is invoked without an argument, which 3426is the only permitted form in Fortran 2008, GCC passes @code{0} as 3427first argument. 3428@end table 3429 3430 3431@node _gfortran_caf_num_images 3432@subsection @code{_gfortran_caf_num_images} --- Querying the maximal number of images 3433@cindex Coarray, _gfortran_caf_num_images 3434 3435@table @asis 3436@item @emph{Description}: 3437This function returns the number of images in the current team, if 3438@var{distance} is 0 or the number of images in the parent team at the specified 3439distance. If failed is -1, the function returns the number of all images at 3440the specified distance; if it is 0, the function returns the number of 3441nonfailed images, and if it is 1, it returns the number of failed images. 3442 3443@item @emph{Syntax}: 3444@code{int _gfortran_caf_num_images(int distance, int failed)} 3445 3446@item @emph{Arguments}: 3447@multitable @columnfractions .15 .70 3448@item @var{distance} @tab the distance from this image to the ancestor. 3449Shall be positive. 3450@item @var{failed} @tab shall be -1, 0, or 1 3451@end multitable 3452 3453@item @emph{NOTES} 3454This function follows TS18508. If the num_image intrinsic has no arguments, 3455the the compiler passes @code{distance=0} and @code{failed=-1} to the function. 3456@end table 3457 3458 3459@node _gfortran_caf_register 3460@subsection @code{_gfortran_caf_register} --- Registering coarrays 3461@cindex Coarray, _gfortran_caf_deregister 3462 3463@table @asis 3464@item @emph{Description}: 3465Allocates memory for a coarray and creates a token to identify the coarray. The 3466function is called for both coarrays with @code{SAVE} attribute and using an 3467explicit @code{ALLOCATE} statement. If an error occurs and @var{STAT} is a 3468@code{NULL} pointer, the function shall abort with printing an error message 3469and starting the error termination. If no error occurs and @var{STAT} is 3470present, it shall be set to zero. Otherwise, it shall be set to a positive 3471value and, if not-@code{NULL}, @var{ERRMSG} shall be set to a string describing 3472the failure. The function shall return a pointer to the requested memory 3473for the local image as a call to @code{malloc} would do. 3474 3475For @code{CAF_REGTYPE_COARRAY_STATIC} and @code{CAF_REGTYPE_COARRAY_ALLOC}, 3476the passed size is the byte size requested. For @code{CAF_REGTYPE_LOCK_STATIC}, 3477@code{CAF_REGTYPE_LOCK_ALLOC} and @code{CAF_REGTYPE_CRITICAL} it is the array 3478size or one for a scalar. 3479 3480 3481@item @emph{Syntax}: 3482@code{void *caf_register (size_t size, caf_register_t type, caf_token_t *token, 3483int *stat, char *errmsg, int errmsg_len)} 3484 3485@item @emph{Arguments}: 3486@multitable @columnfractions .15 .70 3487@item @var{size} @tab For normal coarrays, the byte size of the coarray to be 3488allocated; for lock types and event types, the number of elements. 3489@item @var{type} @tab one of the caf_register_t types. 3490@item @var{token} @tab intent(out) An opaque pointer identifying the coarray. 3491@item @var{stat} @tab intent(out) For allocatable coarrays, stores the STAT=; 3492may be NULL 3493@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3494an error message; may be NULL 3495@item @var{errmsg_len} @tab the buffer size of errmsg. 3496@end multitable 3497 3498@item @emph{NOTES} 3499Nonalloatable coarrays have to be registered prior use from remote images. 3500In order to guarantee this, they have to be registered before the main 3501program. This can be achieved by creating constructor functions. That is what 3502GCC does such that also nonallocatable coarrays the memory is allocated and no 3503static memory is used. The token permits to identify the coarray; to the 3504processor, the token is a nonaliasing pointer. The library can, for instance, 3505store the base address of the coarray in the token, some handle or a more 3506complicated struct. 3507 3508For normal coarrays, the returned pointer is used for accesses on the local 3509image. For lock types, the value shall only used for checking the allocation 3510status. Note that for critical blocks, the locking is only required on one 3511image; in the locking statement, the processor shall always pass always an 3512image index of one for critical-block lock variables 3513(@code{CAF_REGTYPE_CRITICAL}). For lock types and critical-block variables, 3514the initial value shall be unlocked (or, respecitively, not in critical 3515section) such as the value false; for event types, the initial state should 3516be no event, e.g. zero. 3517@end table 3518 3519 3520@node _gfortran_caf_deregister 3521@subsection @code{_gfortran_caf_deregister} --- Deregistering coarrays 3522@cindex Coarray, _gfortran_caf_deregister 3523 3524@table @asis 3525@item @emph{Description}: 3526Called to free the memory of a coarray; the processor calls this function for 3527automatic and explicit deallocation. In case of an error, this function shall 3528fail with an error message, unless the @var{STAT} variable is not null. 3529 3530@item @emph{Syntax}: 3531@code{void caf_deregister (const caf_token_t *token, int *stat, char *errmsg, 3532int errmsg_len)} 3533 3534@item @emph{Arguments}: 3535@multitable @columnfractions .15 .70 3536@item @var{stat} @tab intent(out) Stores the STAT=; may be NULL 3537@item @var{errmsg} @tab intent(out) When an error occurs, this will be set 3538to an error message; may be NULL 3539@item @var{errmsg_len} @tab the buffer size of errmsg. 3540@end multitable 3541 3542@item @emph{NOTES} 3543For nonalloatable coarrays this function is never called. If a cleanup is 3544required, it has to be handled via the finish, stop and error stop functions, 3545and via destructors. 3546@end table 3547 3548 3549@node _gfortran_caf_send 3550@subsection @code{_gfortran_caf_send} --- Sending data from a local image to a remote image 3551@cindex Coarray, _gfortran_caf_send 3552 3553@table @asis 3554@item @emph{Description}: 3555Called to send a scalar, an array section or whole array from a local 3556to a remote image identified by the image_index. 3557 3558@item @emph{Syntax}: 3559@code{void _gfortran_caf_send (caf_token_t token, size_t offset, 3560int image_index, gfc_descriptor_t *dest, caf_vector_t *dst_vector, 3561gfc_descriptor_t *src, int dst_kind, int src_kind, bool may_require_tmp)} 3562 3563@item @emph{Arguments}: 3564@multitable @columnfractions .15 .70 3565@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3566@item @var{offset} @tab By which amount of bytes the actual data is shifted 3567compared to the base address of the coarray. 3568@item @var{image_index} @tab The ID of the remote image; must be a positive 3569number. 3570@item @var{dest} @tab intent(in) Array descriptor for the remote image for the 3571bounds and the size. The base_addr shall not be accessed. 3572@item @var{dst_vector} @tab intent(int) If not NULL, it contains the vector 3573subscript of the destination array; the values are relative to the dimension 3574triplet of the dest argument. 3575@item @var{src} @tab intent(in) Array descriptor of the local array to be 3576transferred to the remote image 3577@item @var{dst_kind} @tab Kind of the destination argument 3578@item @var{src_kind} @tab Kind of the source argument 3579@item @var{may_require_tmp} @tab The variable is false it is known at compile 3580time that the @var{dest} and @var{src} either cannot overlap or overlap (fully 3581or partially) such that walking @var{src} and @var{dest} in element wise 3582element order (honoring the stride value) will not lead to wrong results. 3583Otherwise, the value is true. 3584@end multitable 3585 3586@item @emph{NOTES} 3587It is permitted to have image_id equal the current image; the memory of the 3588send-to and the send-from might (partially) overlap in that case. The 3589implementation has to take care that it handles this case, e.g. using 3590@code{memmove} which handles (partially) overlapping memory. If 3591@var{may_require_tmp} is true, the library might additionally create a 3592temporary variable, unless additional checks show that this is not required 3593(e.g. because walking backward is possible or because both arrays are 3594contiguous and @code{memmove} takes care of overlap issues). 3595 3596Note that the assignment of a scalar to an array is permitted. In addition, 3597the library has to handle numeric-type conversion and for strings, padding 3598and different character kinds. 3599@end table 3600 3601 3602@node _gfortran_caf_get 3603@subsection @code{_gfortran_caf_get} --- Getting data from a remote image 3604@cindex Coarray, _gfortran_caf_get 3605 3606@table @asis 3607@item @emph{Description}: 3608Called to get an array section or whole array from a a remote, 3609image identified by the image_index. 3610 3611@item @emph{Syntax}: 3612@code{void _gfortran_caf_get_desc (caf_token_t token, size_t offset, 3613int image_index, gfc_descriptor_t *src, caf_vector_t *src_vector, 3614gfc_descriptor_t *dest, int src_kind, int dst_kind, bool may_require_tmp)} 3615 3616@item @emph{Arguments}: 3617@multitable @columnfractions .15 .70 3618@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3619@item @var{offset} @tab By which amount of bytes the actual data is shifted 3620compared to the base address of the coarray. 3621@item @var{image_index} @tab The ID of the remote image; must be a positive 3622number. 3623@item @var{dest} @tab intent(in) Array descriptor of the local array to be 3624transferred to the remote image 3625@item @var{src} @tab intent(in) Array descriptor for the remote image for the 3626bounds and the size. The base_addr shall not be accessed. 3627@item @var{src_vector} @tab intent(int) If not NULL, it contains the vector 3628subscript of the destination array; the values are relative to the dimension 3629triplet of the dest argument. 3630@item @var{dst_kind} @tab Kind of the destination argument 3631@item @var{src_kind} @tab Kind of the source argument 3632@item @var{may_require_tmp} @tab The variable is false it is known at compile 3633time that the @var{dest} and @var{src} either cannot overlap or overlap (fully 3634or partially) such that walking @var{src} and @var{dest} in element wise 3635element order (honoring the stride value) will not lead to wrong results. 3636Otherwise, the value is true. 3637@end multitable 3638 3639@item @emph{NOTES} 3640It is permitted to have image_id equal the current image; the memory of the 3641send-to and the send-from might (partially) overlap in that case. The 3642implementation has to take care that it handles this case, e.g. using 3643@code{memmove} which handles (partially) overlapping memory. If 3644@var{may_require_tmp} is true, the library might additionally create a 3645temporary variable, unless additional checks show that this is not required 3646(e.g. because walking backward is possible or because both arrays are 3647contiguous and @code{memmove} takes care of overlap issues). 3648 3649Note that the library has to handle numeric-type conversion and for strings, 3650padding and different character kinds. 3651@end table 3652 3653 3654@node _gfortran_caf_sendget 3655@subsection @code{_gfortran_caf_sendget} --- Sending data between remote images 3656@cindex Coarray, _gfortran_caf_sendget 3657 3658@table @asis 3659@item @emph{Description}: 3660Called to send a scalar, an array section or whole array from a remote image 3661identified by the src_image_index to a remote image identified by the 3662dst_image_index. 3663 3664@item @emph{Syntax}: 3665@code{void _gfortran_caf_sendget (caf_token_t dst_token, size_t dst_offset, 3666int dst_image_index, gfc_descriptor_t *dest, caf_vector_t *dst_vector, 3667caf_token_t src_token, size_t src_offset, int src_image_index, 3668gfc_descriptor_t *src, caf_vector_t *src_vector, int dst_kind, int src_kind, 3669bool may_require_tmp)} 3670 3671@item @emph{Arguments}: 3672@multitable @columnfractions .15 .70 3673@item @var{dst_token} @tab intent(in) An opaque pointer identifying the 3674destination coarray. 3675@item @var{dst_offset} @tab By which amount of bytes the actual data is 3676shifted compared to the base address of the destination coarray. 3677@item @var{dst_image_index} @tab The ID of the destination remote image; must 3678be a positive number. 3679@item @var{dest} @tab intent(in) Array descriptor for the destination 3680remote image for the bounds and the size. The base_addr shall not be accessed. 3681@item @var{dst_vector} @tab intent(int) If not NULL, it contains the vector 3682subscript of the destination array; the values are relative to the dimension 3683triplet of the dest argument. 3684@item @var{src_token} @tab An opaque pointer identifying the source coarray. 3685@item @var{src_offset} @tab By which amount of bytes the actual data is shifted 3686compared to the base address of the source coarray. 3687@item @var{src_image_index} @tab The ID of the source remote image; must be a 3688positive number. 3689@item @var{src} @tab intent(in) Array descriptor of the local array to be 3690transferred to the remote image. 3691@item @var{src_vector} @tab intent(in) Array descriptor of the local array to 3692be transferred to the remote image 3693@item @var{dst_kind} @tab Kind of the destination argument 3694@item @var{src_kind} @tab Kind of the source argument 3695@item @var{may_require_tmp} @tab The variable is false it is known at compile 3696time that the @var{dest} and @var{src} either cannot overlap or overlap (fully 3697or partially) such that walking @var{src} and @var{dest} in element wise 3698element order (honoring the stride value) will not lead to wrong results. 3699Otherwise, the value is true. 3700@end multitable 3701 3702@item @emph{NOTES} 3703It is permitted to have image_ids equal; the memory of the send-to and the 3704send-from might (partially) overlap in that case. The implementation has to 3705take care that it handles this case, e.g. using @code{memmove} which handles 3706(partially) overlapping memory. If @var{may_require_tmp} is true, the library 3707might additionally create a temporary variable, unless additional checks show 3708that this is not required (e.g. because walking backward is possible or because 3709both arrays are contiguous and @code{memmove} takes care of overlap issues). 3710 3711Note that the assignment of a scalar to an array is permitted. In addition, 3712the library has to handle numeric-type conversion and for strings, padding and 3713different character kinds. 3714@end table 3715 3716 3717@node _gfortran_caf_lock 3718@subsection @code{_gfortran_caf_lock} --- Locking a lock variable 3719@cindex Coarray, _gfortran_caf_lock 3720 3721@table @asis 3722@item @emph{Description}: 3723Acquire a lock on the given image on a scalar locking variable or for the 3724given array element for an array-valued variable. If the @var{aquired_lock} 3725is @code{NULL}, the function return after having obtained the lock. If it is 3726nonnull, the result is is assigned the value true (one) when the lock could be 3727obtained and false (zero) otherwise. Locking a lock variable which has already 3728been locked by the same image is an error. 3729 3730@item @emph{Syntax}: 3731@code{void _gfortran_caf_lock (caf_token_t token, size_t index, int image_index, 3732int *aquired_lock, int *stat, char *errmsg, int errmsg_len)} 3733 3734@item @emph{Arguments}: 3735@multitable @columnfractions .15 .70 3736@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3737@item @var{index} @tab Array index; first array index is 0. For scalars, it is 3738always 0. 3739@item @var{image_index} @tab The ID of the remote image; must be a positive 3740number. 3741@item @var{aquired_lock} @tab intent(out) If not NULL, it returns whether lock 3742could be obtained 3743@item @var{stat} @tab intent(out) Stores the STAT=; may be NULL 3744@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3745an error message; may be NULL 3746@item @var{errmsg_len} @tab the buffer size of errmsg. 3747@end multitable 3748 3749@item @emph{NOTES} 3750This function is also called for critical blocks; for those, the array index 3751is always zero and the image index is one. Libraries are permitted to use other 3752images for critical-block locking variables. 3753@end table 3754 3755@node _gfortran_caf_unlock 3756@subsection @code{_gfortran_caf_lock} --- Unlocking a lock variable 3757@cindex Coarray, _gfortran_caf_unlock 3758 3759@table @asis 3760@item @emph{Description}: 3761Release a lock on the given image on a scalar locking variable or for the 3762given array element for an array-valued variable. Unlocking a lock variable 3763which is unlocked or has been locked by a different image is an error. 3764 3765@item @emph{Syntax}: 3766@code{void _gfortran_caf_unlock (caf_token_t token, size_t index, int image_index, 3767int *stat, char *errmsg, int errmsg_len)} 3768 3769@item @emph{Arguments}: 3770@multitable @columnfractions .15 .70 3771@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3772@item @var{index} @tab Array index; first array index is 0. For scalars, it is 3773always 0. 3774@item @var{image_index} @tab The ID of the remote image; must be a positive 3775number. 3776@item @var{stat} @tab intent(out) For allocatable coarrays, stores the STAT=; 3777may be NULL 3778@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3779an error message; may be NULL 3780@item @var{errmsg_len} @tab the buffer size of errmsg. 3781@end multitable 3782 3783@item @emph{NOTES} 3784This function is also called for critical block; for those, the array index 3785is always zero and the image index is one. Libraries are permitted to use other 3786images for critical-block locking variables. 3787@end table 3788 3789@node _gfortran_caf_event_post 3790@subsection @code{_gfortran_caf_event_post} --- Post an event 3791@cindex Coarray, _gfortran_caf_event_post 3792 3793@table @asis 3794@item @emph{Description}: 3795Increment the event count of the specified event variable. 3796 3797@item @emph{Syntax}: 3798@code{void _gfortran_caf_event_post (caf_token_t token, size_t index, 3799int image_index, int *stat, char *errmsg, int errmsg_len)} 3800 3801@item @emph{Arguments}: 3802@multitable @columnfractions .15 .70 3803@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3804@item @var{index} @tab Array index; first array index is 0. For scalars, it is 3805always 0. 3806@item @var{image_index} @tab The ID of the remote image; must be a positive 3807number; zero indicates the current image when accessed noncoindexed. 3808@item @var{stat} @tab intent(out) Stores the STAT=; may be NULL 3809@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3810an error message; may be NULL 3811@item @var{errmsg_len} @tab the buffer size of errmsg. 3812@end multitable 3813 3814@item @emph{NOTES} 3815This acts like an atomic add of one to the remote image's event variable. 3816The statement is an image-control statement but does not imply sync memory. 3817Still, all preceeding push communications of this image to the specified 3818remote image has to be completed before @code{event_wait} on the remote 3819image returns. 3820@end table 3821 3822 3823 3824@node _gfortran_caf_event_wait 3825@subsection @code{_gfortran_caf_event_wait} --- Wait that an event occurred 3826@cindex Coarray, _gfortran_caf_event_wait 3827 3828@table @asis 3829@item @emph{Description}: 3830Wait until the event count has reached at least the specified 3831@var{until_count}; if so, atomically decrement the event variable by this 3832amount and return. 3833 3834@item @emph{Syntax}: 3835@code{void _gfortran_caf_event_wait (caf_token_t token, size_t index, 3836int until_count, int *stat, char *errmsg, int errmsg_len)} 3837 3838@item @emph{Arguments}: 3839@multitable @columnfractions .15 .70 3840@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3841@item @var{index} @tab Array index; first array index is 0. For scalars, it is 3842always 0. 3843@item @var{until_count} @tab The number of events which have to be available 3844before the function returns. 3845@item @var{stat} @tab intent(out) Stores the STAT=; may be NULL 3846@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3847an error message; may be NULL 3848@item @var{errmsg_len} @tab the buffer size of errmsg. 3849@end multitable 3850 3851@item @emph{NOTES} 3852This function only operates on a local coarray. It acts like a loop checking 3853atomically the value of the event variable, breaking if the value is greater 3854or equal the requested number of counts. Before the function returns, the 3855event variable has to be decremented by the requested @var{until_count} value. 3856A possible implementation would be a busy loop for a certain number of spins 3857(possibly depending on the number of threads relative to the number of available 3858cores) followed by other waiting strategy such as a sleeping wait (possibly with 3859an increasing number of sleep time) or, if possible, a futex wait. 3860 3861The statement is an image-control statement but does not imply sync memory. 3862Still, all preceeding push communications to this image of images having 3863issued a @code{event_push} have to be completed before this function returns. 3864@end table 3865 3866 3867 3868@node _gfortran_caf_event_query 3869@subsection @code{_gfortran_caf_event_query} --- Query event count 3870@cindex Coarray, _gfortran_caf_event_query 3871 3872@table @asis 3873@item @emph{Description}: 3874Return the event count of the specified event count. 3875 3876@item @emph{Syntax}: 3877@code{void _gfortran_caf_event_query (caf_token_t token, size_t index, 3878int image_index, int *count, int *stat)} 3879 3880@item @emph{Arguments}: 3881@multitable @columnfractions .15 .70 3882@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 3883@item @var{index} @tab Array index; first array index is 0. For scalars, it is 3884always 0. 3885@item @var{image_index} @tab The ID of the remote image; must be a positive 3886number; zero indicates the current image when accessed noncoindexed. 3887@item @var{count} @tab intent(out) The number of events currently posted to 3888the event variable 3889@item @var{stat} @tab intent(out) Stores the STAT=; may be NULL 3890@end multitable 3891 3892@item @emph{NOTES} 3893The typical use is to check the local even variable to only call 3894@code{event_wait} when the data is available. However, a coindexed variable 3895is permitted; there is no ordering or synchronization implied. It acts like 3896an atomic fetch of the value of the event variable. 3897@end table 3898 3899@node _gfortran_caf_sync_all 3900@subsection @code{_gfortran_caf_sync_all} --- All-image barrier 3901@cindex Coarray, _gfortran_caf_sync_all 3902 3903@table @asis 3904@item @emph{Description}: 3905Synchronization of all images in the current team; the program only continues 3906on a given image after this function has been called on all images of the 3907current team. Additionally, it ensures that all pending data transfers of 3908previous segment have completed. 3909 3910@item @emph{Syntax}: 3911@code{void _gfortran_caf_sync_all (int *stat, char *errmsg, int errmsg_len)} 3912 3913@item @emph{Arguments}: 3914@multitable @columnfractions .15 .70 3915@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 3916@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3917an error message; may be NULL 3918@item @var{errmsg_len} @tab the buffer size of errmsg. 3919@end multitable 3920@end table 3921 3922 3923 3924@node _gfortran_caf_sync_images 3925@subsection @code{_gfortran_caf_sync_images} --- Barrier for selected images 3926@cindex Coarray, _gfortran_caf_sync_images 3927 3928@table @asis 3929@item @emph{Description}: 3930Synchronization between the specified images; the program only continues on a 3931given image after this function has been called on all images specified for 3932that image. Note that one image can wait for all other images in the current 3933team (e.g. via @code{sync images(*)}) while those only wait for that specific 3934image. Additionally, @code{sync images} it ensures that all pending data 3935transfers of previous segment have completed. 3936 3937@item @emph{Syntax}: 3938@code{void _gfortran_caf_sync_images (int count, int images[], int *stat, 3939char *errmsg, int errmsg_len)} 3940 3941@item @emph{Arguments}: 3942@multitable @columnfractions .15 .70 3943@item @var{count} @tab the number of images which are provided in the next 3944argument. For a zero-sized array, the value is zero. For @code{sync 3945images (*)}, the value is @math{-1}. 3946@item @var{images} @tab intent(in) an array with the images provided by the 3947user. If @var{count} is zero, a NULL pointer is passed. 3948@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 3949@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3950an error message; may be NULL 3951@item @var{errmsg_len} @tab the buffer size of errmsg. 3952@end multitable 3953@end table 3954 3955 3956 3957@node _gfortran_caf_sync_memory 3958@subsection @code{_gfortran_caf_sync_memory} --- Wait for completion of segment-memory operations 3959@cindex Coarray, _gfortran_caf_sync_memory 3960 3961@table @asis 3962@item @emph{Description}: 3963Acts as optimization barrier between different segments. It also ensures that 3964all pending memory operations of this image have been completed. 3965 3966@item @emph{Syntax}: 3967@code{void _gfortran_caf_sync_memory (int *stat, char *errmsg, int errmsg_len)} 3968 3969@item @emph{Arguments}: 3970@multitable @columnfractions .15 .70 3971@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 3972@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 3973an error message; may be NULL 3974@item @var{errmsg_len} @tab the buffer size of errmsg. 3975@end multitable 3976 3977@item @emph{NOTE} A simple implementation could be 3978@code{__asm__ __volatile__ ("":::"memory")} to prevent code movements. 3979@end table 3980 3981 3982 3983@node _gfortran_caf_error_stop 3984@subsection @code{_gfortran_caf_error_stop} --- Error termination with exit code 3985@cindex Coarray, _gfortran_caf_error_stop 3986 3987@table @asis 3988@item @emph{Description}: 3989Invoked for an @code{ERROR STOP} statement which has an integer argument. The 3990function should terminate the program with the specified exit code. 3991 3992 3993@item @emph{Syntax}: 3994@code{void _gfortran_caf_error_stop (int32_t error)} 3995 3996@item @emph{Arguments}: 3997@multitable @columnfractions .15 .70 3998@item @var{error} @tab the exit status to be used. 3999@end multitable 4000@end table 4001 4002 4003 4004@node _gfortran_caf_error_stop_str 4005@subsection @code{_gfortran_caf_error_stop_str} --- Error termination with string 4006@cindex Coarray, _gfortran_caf_error_stop_str 4007 4008@table @asis 4009@item @emph{Description}: 4010Invoked for an @code{ERROR STOP} statement which has a string as argument. The 4011function should terminate the program with a nonzero-exit code. 4012 4013@item @emph{Syntax}: 4014@code{void _gfortran_caf_error_stop (const char *string, int32_t len)} 4015 4016@item @emph{Arguments}: 4017@multitable @columnfractions .15 .70 4018@item @var{string} @tab the error message (not zero terminated) 4019@item @var{len} @tab the length of the string 4020@end multitable 4021@end table 4022 4023 4024 4025@node _gfortran_caf_atomic_define 4026@subsection @code{_gfortran_caf_atomic_define} --- Atomic variable assignment 4027@cindex Coarray, _gfortran_caf_atomic_define 4028 4029@table @asis 4030@item @emph{Description}: 4031Assign atomically a value to an integer or logical variable. 4032 4033@item @emph{Syntax}: 4034@code{void _gfortran_caf_atomic_define (caf_token_t token, size_t offset, 4035int image_index, void *value, int *stat, int type, int kind)} 4036 4037@item @emph{Arguments}: 4038@multitable @columnfractions .15 .70 4039@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 4040@item @var{offset} @tab By which amount of bytes the actual data is shifted 4041compared to the base address of the coarray. 4042@item @var{image_index} @tab The ID of the remote image; must be a positive 4043number; zero indicates the current image when used noncoindexed. 4044@item @var{value} @tab intent(in) the value to be assigned, passed by reference. 4045@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4046@item @var{type} @tab the data type, i.e. @code{BT_INTEGER} (1) or 4047@code{BT_LOGICAL} (2). 4048@item @var{kind} @tab The kind value (only 4; always @code{int}) 4049@end multitable 4050@end table 4051 4052 4053 4054@node _gfortran_caf_atomic_ref 4055@subsection @code{_gfortran_caf_atomic_ref} --- Atomic variable reference 4056@cindex Coarray, _gfortran_caf_atomic_ref 4057 4058@table @asis 4059@item @emph{Description}: 4060Reference atomically a value of a kind-4 integer or logical variable. 4061 4062@item @emph{Syntax}: 4063@code{void _gfortran_caf_atomic_ref (caf_token_t token, size_t offset, 4064int image_index, void *value, int *stat, int type, int kind)} 4065 4066@item @emph{Arguments}: 4067@item @emph{Arguments}: 4068@multitable @columnfractions .15 .70 4069@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 4070@item @var{offset} @tab By which amount of bytes the actual data is shifted 4071compared to the base address of the coarray. 4072@item @var{image_index} @tab The ID of the remote image; must be a positive 4073number; zero indicates the current image when used noncoindexed. 4074@item @var{value} @tab intent(out) The variable assigned the atomically 4075referenced variable. 4076@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4077@item @var{type} @tab the data type, i.e. @code{BT_INTEGER} (1) or 4078@code{BT_LOGICAL} (2). 4079@item @var{kind} @tab The kind value (only 4; always @code{int}) 4080@end multitable 4081@end table 4082 4083 4084 4085@node _gfortran_caf_atomic_cas 4086@subsection @code{_gfortran_caf_atomic_cas} --- Atomic compare and swap 4087@cindex Coarray, _gfortran_caf_atomic_cas 4088 4089@table @asis 4090@item @emph{Description}: 4091Atomic compare and swap of a kind-4 integer or logical variable. Assigns 4092atomically the specified value to the atomic variable, if the latter has 4093the value specified by the passed condition value. 4094 4095@item @emph{Syntax}: 4096@code{void _gfortran_caf_atomic_cas (caf_token_t token, size_t offset, 4097int image_index, void *old, void *compare, void *new_val, int *stat, 4098int type, int kind)} 4099 4100@item @emph{Arguments}: 4101@multitable @columnfractions .15 .70 4102@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 4103@item @var{offset} @tab By which amount of bytes the actual data is shifted 4104compared to the base address of the coarray. 4105@item @var{image_index} @tab The ID of the remote image; must be a positive 4106number; zero indicates the current image when used noncoindexed. 4107@item @var{old} @tab intent(out) the value which the atomic variable had 4108just before the cas operation. 4109@item @var{compare} @tab intent(in) The value used for comparision. 4110@item @var{new_val} @tab intent(in) The new value for the atomic variable, 4111assigned to the atomic variable, if @code{compare} equals the value of the 4112atomic variable. 4113@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4114@item @var{type} @tab the data type, i.e. @code{BT_INTEGER} (1) or 4115@code{BT_LOGICAL} (2). 4116@item @var{kind} @tab The kind value (only 4; always @code{int}) 4117@end multitable 4118@end table 4119 4120 4121 4122@node _gfortran_caf_atomic_op 4123@subsection @code{_gfortran_caf_atomic_op} --- Atomic operation 4124@cindex Coarray, _gfortran_caf_atomic_op 4125 4126@table @asis 4127@item @emph{Description}: 4128Apply an operation atomically to an atomic integer or logical variable. 4129After the operation, @var{old} contains the value just before the operation, 4130which, respectively, adds (GFC_CAF_ATOMIC_ADD) atomically the @code{value} to 4131the atomic integer variable or does a bitwise AND, OR or exclusive OR of the 4132between the atomic variable and @var{value}; the result is then stored in the 4133atomic variable. 4134 4135@item @emph{Syntax}: 4136@code{void _gfortran_caf_atomic_op (int op, caf_token_t token, size_t offset, 4137int image_index, void *value, void *old, int *stat, int type, int kind)} 4138 4139@item @emph{Arguments}: 4140@multitable @columnfractions .15 .70 4141@item @var{op} @tab the operation to be performed; possible values 4142@code{GFC_CAF_ATOMIC_ADD} (1), @code{GFC_CAF_ATOMIC_AND} (2), 4143@code{GFC_CAF_ATOMIC_OR} (3), @code{GFC_CAF_ATOMIC_XOR} (4). 4144@item @var{token} @tab intent(in) An opaque pointer identifying the coarray. 4145@item @var{offset} @tab By which amount of bytes the actual data is shifted 4146compared to the base address of the coarray. 4147@item @var{image_index} @tab The ID of the remote image; must be a positive 4148number; zero indicates the current image when used noncoindexed. 4149@item @var{old} @tab intent(out) the value which the atomic variable had 4150just before the atomic operation. 4151@item @var{val} @tab intent(in) The new value for the atomic variable, 4152assigned to the atomic variable, if @code{compare} equals the value of the 4153atomic variable. 4154@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4155@item @var{type} @tab the data type, i.e. @code{BT_INTEGER} (1) or 4156@code{BT_LOGICAL} (2). 4157@item @var{kind} @tab The kind value (only 4; always @code{int}) 4158@end multitable 4159@end table 4160 4161 4162 4163 4164@node _gfortran_caf_co_broadcast 4165@subsection @code{_gfortran_caf_co_broadcast} --- Sending data to all images 4166@cindex Coarray, _gfortran_caf_co_broadcast 4167 4168@table @asis 4169@item @emph{Description}: 4170Distribute a value from a given image to all other images in the team. Has to 4171be called collectively. 4172 4173@item @emph{Syntax}: 4174@code{void _gfortran_caf_co_broadcast (gfc_descriptor_t *a, 4175int source_image, int *stat, char *errmsg, int errmsg_len)} 4176 4177@item @emph{Arguments}: 4178@multitable @columnfractions .15 .70 4179@item @var{a} @tab intent(inout) And array descriptor with the data to be 4180breoadcasted (on @var{source_image}) or to be received (other images). 4181@item @var{source_image} @tab The ID of the image from which the data should 4182be taken. 4183@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4184@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 4185an error message; may be NULL 4186@item @var{errmsg_len} @tab the buffer size of errmsg. 4187@end multitable 4188@end table 4189 4190 4191 4192@node _gfortran_caf_co_max 4193@subsection @code{_gfortran_caf_co_max} --- Collective maximum reduction 4194@cindex Coarray, _gfortran_caf_co_max 4195 4196@table @asis 4197@item @emph{Description}: 4198Calculates the for the each array element of the variable @var{a} the maximum 4199value for that element in the current team; if @var{result_image} has the 4200value 0, the result shall be stored on all images, otherwise, only on the 4201specified image. This function operates on numeric values and character 4202strings. 4203 4204@item @emph{Syntax}: 4205@code{void _gfortran_caf_co_max (gfc_descriptor_t *a, int result_image, 4206int *stat, char *errmsg, int a_len, int errmsg_len)} 4207 4208@item @emph{Arguments}: 4209@multitable @columnfractions .15 .70 4210@item @var{a} @tab intent(inout) And array descriptor with the data to be 4211breoadcasted (on @var{source_image}) or to be received (other images). 4212@item @var{result_image} @tab The ID of the image to which the reduced 4213value should be copied to; if zero, it has to be copied to all images. 4214@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4215@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 4216an error message; may be NULL 4217@item @var{a_len} @tab The string length of argument @var{a}. 4218@item @var{errmsg_len} @tab the buffer size of errmsg. 4219@end multitable 4220 4221@item @emph{NOTES} 4222If @var{result_image} is nonzero, the value on all images except of the 4223specified one become undefined; hence, the library may make use of this. 4224@end table 4225 4226 4227 4228@node _gfortran_caf_co_min 4229@subsection @code{_gfortran_caf_co_min} --- Collective minimum reduction 4230@cindex Coarray, _gfortran_caf_co_min 4231 4232@table @asis 4233@item @emph{Description}: 4234Calculates the for the each array element of the variable @var{a} the minimum 4235value for that element in the current team; if @var{result_image} has the 4236value 0, the result shall be stored on all images, otherwise, only on the 4237specified image. This function operates on numeric values and character 4238strings. 4239 4240@item @emph{Syntax}: 4241@code{void _gfortran_caf_co_min (gfc_descriptor_t *a, int result_image, 4242int *stat, char *errmsg, int a_len, int errmsg_len)} 4243 4244@item @emph{Arguments}: 4245@multitable @columnfractions .15 .70 4246@item @var{a} @tab intent(inout) And array descriptor with the data to be 4247breoadcasted (on @var{source_image}) or to be received (other images). 4248@item @var{result_image} @tab The ID of the image to which the reduced 4249value should be copied to; if zero, it has to be copied to all images. 4250@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4251@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 4252an error message; may be NULL 4253@item @var{a_len} @tab The string length of argument @var{a}. 4254@item @var{errmsg_len} @tab the buffer size of errmsg. 4255@end multitable 4256 4257@item @emph{NOTES} 4258If @var{result_image} is nonzero, the value on all images except of the 4259specified one become undefined; hence, the library may make use of this. 4260@end table 4261 4262 4263 4264@node _gfortran_caf_co_sum 4265@subsection @code{_gfortran_caf_co_sum} --- Collective summing reduction 4266@cindex Coarray, _gfortran_caf_co_sum 4267 4268@table @asis 4269@item @emph{Description}: 4270Calculates the for the each array element of the variable @var{a} the sum 4271value for that element in the current team; if @var{result_image} has the 4272value 0, the result shall be stored on all images, otherwise, only on the 4273specified image. This function operates on numeric values. 4274 4275@item @emph{Syntax}: 4276@code{void _gfortran_caf_co_sum (gfc_descriptor_t *a, int result_image, 4277int *stat, char *errmsg, int errmsg_len)} 4278 4279@item @emph{Arguments}: 4280@multitable @columnfractions .15 .70 4281@item @var{a} @tab intent(inout) And array descriptor with the data to be 4282breoadcasted (on @var{source_image}) or to be received (other images). 4283@item @var{result_image} @tab The ID of the image to which the reduced 4284value should be copied to; if zero, it has to be copied to all images. 4285@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4286@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 4287an error message; may be NULL 4288@item @var{errmsg_len} @tab the buffer size of errmsg. 4289@end multitable 4290 4291@item @emph{NOTES} 4292If @var{result_image} is nonzero, the value on all images except of the 4293specified one become undefined; hence, the library may make use of this. 4294@end table 4295 4296 4297 4298@node _gfortran_caf_co_reduce 4299@subsection @code{_gfortran_caf_co_reduce} --- Generic collective reduction 4300@cindex Coarray, _gfortran_caf_co_reduce 4301 4302@table @asis 4303@item @emph{Description}: 4304Calculates the for the each array element of the variable @var{a} the reduction 4305value for that element in the current team; if @var{result_image} has the 4306value 0, the result shall be stored on all images, otherwise, only on the 4307specified image. The @var{opr} is a pure function doing a mathematically 4308commutative and associative operation. 4309 4310The @var{opr_flags} denote the following; the values are bitwise ored. 4311@code{GFC_CAF_BYREF} (1) if the result should be returned 4312by value; @code{GFC_CAF_HIDDENLEN} (2) whether the result and argument 4313string lengths shall be specified as hidden argument; 4314@code{GFC_CAF_ARG_VALUE} (4) whether the arguments shall be passed by value, 4315@code{GFC_CAF_ARG_DESC} (8) whether the arguments shall be passed by descriptor. 4316 4317 4318@item @emph{Syntax}: 4319@code{void _gfortran_caf_co_reduce (gfc_descriptor_t *a, 4320void * (*opr) (void *, void *), int opr_flags, int result_image, 4321int *stat, char *errmsg, int a_len, int errmsg_len)} 4322 4323@item @emph{Arguments}: 4324@multitable @columnfractions .15 .70 4325@item @var{opr} @tab Function pointer to the reduction function. 4326@item @var{opr_flags} @tab Flags regarding the reduction function 4327@item @var{a} @tab intent(inout) And array descriptor with the data to be 4328breoadcasted (on @var{source_image}) or to be received (other images). 4329@item @var{result_image} @tab The ID of the image to which the reduced 4330value should be copied to; if zero, it has to be copied to all images. 4331@item @var{stat} @tab intent(out) Stores the status STAT= and may be NULL. 4332@item @var{errmsg} @tab intent(out) When an error occurs, this will be set to 4333an error message; may be NULL 4334@item @var{a_len} @tab The string length of argument @var{a}. 4335@item @var{errmsg_len} @tab the buffer size of errmsg. 4336@end multitable 4337 4338@item @emph{NOTES} 4339If @var{result_image} is nonzero, the value on all images except of the 4340specified one become undefined; hence, the library may make use of this. 4341For character arguments, the result is passed as first argument, followed 4342by the result string length, next come the two string arguments, followed 4343by the two hidden arguments. With C binding, there are no hidden arguments 4344and by-reference passing and either only a single character is passed or 4345an array descriptor. 4346@end table 4347 4348 4349@c Intrinsic Procedures 4350@c --------------------------------------------------------------------- 4351 4352@include intrinsic.texi 4353 4354 4355@tex 4356\blankpart 4357@end tex 4358 4359@c --------------------------------------------------------------------- 4360@c Contributing 4361@c --------------------------------------------------------------------- 4362 4363@node Contributing 4364@unnumbered Contributing 4365@cindex Contributing 4366 4367Free software is only possible if people contribute to efforts 4368to create it. 4369We're always in need of more people helping out with ideas 4370and comments, writing documentation and contributing code. 4371 4372If you want to contribute to GNU Fortran, 4373have a look at the long lists of projects you can take on. 4374Some of these projects are small, 4375some of them are large; 4376some are completely orthogonal to the rest of what is 4377happening on GNU Fortran, 4378but others are ``mainstream'' projects in need of enthusiastic hackers. 4379All of these projects are important! 4380We will eventually get around to the things here, 4381but they are also things doable by someone who is willing and able. 4382 4383@menu 4384* Contributors:: 4385* Projects:: 4386* Proposed Extensions:: 4387@end menu 4388 4389 4390@node Contributors 4391@section Contributors to GNU Fortran 4392@cindex Contributors 4393@cindex Credits 4394@cindex Authors 4395 4396Most of the parser was hand-crafted by @emph{Andy Vaught}, who is 4397also the initiator of the whole project. Thanks Andy! 4398Most of the interface with GCC was written by @emph{Paul Brook}. 4399 4400The following individuals have contributed code and/or 4401ideas and significant help to the GNU Fortran project 4402(in alphabetical order): 4403 4404@itemize @minus 4405@item Janne Blomqvist 4406@item Steven Bosscher 4407@item Paul Brook 4408@item Tobias Burnus 4409@item Fran@,{c}ois-Xavier Coudert 4410@item Bud Davis 4411@item Jerry DeLisle 4412@item Erik Edelmann 4413@item Bernhard Fischer 4414@item Daniel Franke 4415@item Richard Guenther 4416@item Richard Henderson 4417@item Katherine Holcomb 4418@item Jakub Jelinek 4419@item Niels Kristian Bech Jensen 4420@item Steven Johnson 4421@item Steven G. Kargl 4422@item Thomas Koenig 4423@item Asher Langton 4424@item H. J. Lu 4425@item Toon Moene 4426@item Brooks Moses 4427@item Andrew Pinski 4428@item Tim Prince 4429@item Christopher D. Rickett 4430@item Richard Sandiford 4431@item Tobias Schl@"uter 4432@item Roger Sayle 4433@item Paul Thomas 4434@item Andy Vaught 4435@item Feng Wang 4436@item Janus Weil 4437@item Daniel Kraft 4438@end itemize 4439 4440The following people have contributed bug reports, 4441smaller or larger patches, 4442and much needed feedback and encouragement for the 4443GNU Fortran project: 4444 4445@itemize @minus 4446@item Bill Clodius 4447@item Dominique d'Humi@`eres 4448@item Kate Hedstrom 4449@item Erik Schnetter 4450@item Joost VandeVondele 4451@end itemize 4452 4453Many other individuals have helped debug, 4454test and improve the GNU Fortran compiler over the past few years, 4455and we welcome you to do the same! 4456If you already have done so, 4457and you would like to see your name listed in the 4458list above, please contact us. 4459 4460 4461@node Projects 4462@section Projects 4463 4464@table @emph 4465 4466@item Help build the test suite 4467Solicit more code for donation to the test suite: the more extensive the 4468testsuite, the smaller the risk of breaking things in the future! We can 4469keep code private on request. 4470 4471@item Bug hunting/squishing 4472Find bugs and write more test cases! Test cases are especially very 4473welcome, because it allows us to concentrate on fixing bugs instead of 4474isolating them. Going through the bugzilla database at 4475@url{https://gcc.gnu.org/@/bugzilla/} to reduce testcases posted there and 4476add more information (for example, for which version does the testcase 4477work, for which versions does it fail?) is also very helpful. 4478 4479@end table 4480 4481 4482@node Proposed Extensions 4483@section Proposed Extensions 4484 4485Here's a list of proposed extensions for the GNU Fortran compiler, in no particular 4486order. Most of these are necessary to be fully compatible with 4487existing Fortran compilers, but they are not part of the official 4488J3 Fortran 95 standard. 4489 4490@subsection Compiler extensions: 4491@itemize @bullet 4492@item 4493User-specified alignment rules for structures. 4494 4495@item 4496Automatically extend single precision constants to double. 4497 4498@item 4499Compile code that conserves memory by dynamically allocating common and 4500module storage either on stack or heap. 4501 4502@item 4503Compile flag to generate code for array conformance checking (suggest -CC). 4504 4505@item 4506User control of symbol names (underscores, etc). 4507 4508@item 4509Compile setting for maximum size of stack frame size before spilling 4510parts to static or heap. 4511 4512@item 4513Flag to force local variables into static space. 4514 4515@item 4516Flag to force local variables onto stack. 4517@end itemize 4518 4519 4520@subsection Environment Options 4521@itemize @bullet 4522@item 4523Pluggable library modules for random numbers, linear algebra. 4524LA should use BLAS calling conventions. 4525 4526@item 4527Environment variables controlling actions on arithmetic exceptions like 4528overflow, underflow, precision loss---Generate NaN, abort, default. 4529action. 4530 4531@item 4532Set precision for fp units that support it (i387). 4533 4534@item 4535Variable for setting fp rounding mode. 4536 4537@item 4538Variable to fill uninitialized variables with a user-defined bit 4539pattern. 4540 4541@item 4542Environment variable controlling filename that is opened for that unit 4543number. 4544 4545@item 4546Environment variable to clear/trash memory being freed. 4547 4548@item 4549Environment variable to control tracing of allocations and frees. 4550 4551@item 4552Environment variable to display allocated memory at normal program end. 4553 4554@item 4555Environment variable for filename for * IO-unit. 4556 4557@item 4558Environment variable for temporary file directory. 4559 4560@item 4561Environment variable forcing standard output to be line buffered (Unix). 4562 4563@end itemize 4564 4565 4566@c --------------------------------------------------------------------- 4567@c GNU General Public License 4568@c --------------------------------------------------------------------- 4569 4570@include gpl_v3.texi 4571 4572 4573 4574@c --------------------------------------------------------------------- 4575@c GNU Free Documentation License 4576@c --------------------------------------------------------------------- 4577 4578@include fdl.texi 4579 4580 4581 4582@c --------------------------------------------------------------------- 4583@c Funding Free Software 4584@c --------------------------------------------------------------------- 4585 4586@include funding.texi 4587 4588@c --------------------------------------------------------------------- 4589@c Indices 4590@c --------------------------------------------------------------------- 4591 4592@node Option Index 4593@unnumbered Option Index 4594@command{gfortran}'s command line options are indexed here without any 4595initial @samp{-} or @samp{--}. Where an option has both positive and 4596negative forms (such as -foption and -fno-option), relevant entries in 4597the manual are indexed under the most appropriate form; it may sometimes 4598be useful to look up both forms. 4599@printindex op 4600 4601@node Keyword Index 4602@unnumbered Keyword Index 4603@printindex cp 4604 4605@bye 4606