1 @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002, 2003
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter C Implementation-defined behavior
8 @cindex implementation-defined behavior, C language
10 A conforming implementation of ISO C is required to document its
11 choice of behavior in each of the areas that are designated
12 ``implementation defined.'' The following lists all such areas,
13 along with the section number from the ISO/IEC 9899:1999 standard.
16 * Translation implementation::
17 * Environment implementation::
18 * Identifiers implementation::
19 * Characters implementation::
20 * Integers implementation::
21 * Floating point implementation::
22 * Arrays and pointers implementation::
23 * Hints implementation::
24 * Structures unions enumerations and bit-fields implementation::
25 * Qualifiers implementation::
26 * Preprocessing directives implementation::
27 * Library functions implementation::
28 * Architecture implementation::
29 * Locale-specific behavior implementation::
32 @node Translation implementation
37 @cite{How a diagnostic is identified (3.10, 5.1.1.3).}
39 Diagnostics consist of all the output sent to stderr by GCC.
42 @cite{Whether each nonempty sequence of white-space characters other than
43 new-line is retained or replaced by one space character in translation
47 @node Environment implementation
50 The behavior of these points are dependent on the implementation
51 of the C library, and are not defined by GCC itself.
53 @node Identifiers implementation
58 @cite{Which additional multibyte characters may appear in identifiers
59 and their correspondence to universal character names (6.4.2).}
62 @cite{The number of significant initial characters in an identifier
65 For internal names, all characters are significant. For external names,
66 the number of significant characters are defined by the linker; for
67 almost all targets, all characters are significant.
71 @node Characters implementation
76 @cite{The number of bits in a byte (3.6).}
79 @cite{The values of the members of the execution character set (5.2.1).}
82 @cite{The unique value of the member of the execution character set produced
83 for each of the standard alphabetic escape sequences (5.2.2).}
86 @cite{The value of a @code{char} object into which has been stored any
87 character other than a member of the basic execution character set (6.2.5).}
90 @cite{Which of @code{signed char} or @code{unsigned char} has the same range,
91 representation, and behavior as ``plain'' @code{char} (6.2.5, 6.3.1.1).}
94 @cite{The mapping of members of the source character set (in character
95 constants and string literals) to members of the execution character
96 set (6.4.4.4, 5.1.1.2).}
99 @cite{The value of an integer character constant containing more than one
100 character or containing a character or escape sequence that does not map
101 to a single-byte execution character (6.4.4.4).}
104 @cite{The value of a wide character constant containing more than one
105 multibyte character, or containing a multibyte character or escape
106 sequence not represented in the extended execution character set (6.4.4.4).}
109 @cite{The current locale used to convert a wide character constant consisting
110 of a single multibyte character that maps to a member of the extended
111 execution character set into a corresponding wide character code (6.4.4.4).}
114 @cite{The current locale used to convert a wide string literal into
115 corresponding wide character codes (6.4.5).}
118 @cite{The value of a string literal containing a multibyte character or escape
119 sequence not represented in the execution character set (6.4.5).}
122 @node Integers implementation
127 @cite{Any extended integer types that exist in the implementation (6.2.5).}
130 @cite{Whether signed integer types are represented using sign and magnitude,
131 two's complement, or one's complement, and whether the extraordinary value
132 is a trap representation or an ordinary value (6.2.6.2).}
134 GCC supports only two's complement integer types, and all bit patterns
138 @cite{The rank of any extended integer type relative to another extended
139 integer type with the same precision (6.3.1.1).}
142 @cite{The result of, or the signal raised by, converting an integer to a
143 signed integer type when the value cannot be represented in an object of
144 that type (6.3.1.3).}
147 @cite{The results of some bitwise operations on signed integers (6.5).}
150 @node Floating point implementation
151 @section Floating point
155 @cite{The accuracy of the floating-point operations and of the library
156 functions in @code{<math.h>} and @code{<complex.h>} that return floating-point
157 results (5.2.4.2.2).}
160 @cite{The rounding behaviors characterized by non-standard values
161 of @code{FLT_ROUNDS} @gol
165 @cite{The evaluation methods characterized by non-standard negative
166 values of @code{FLT_EVAL_METHOD} (5.2.4.2.2).}
169 @cite{The direction of rounding when an integer is converted to a
170 floating-point number that cannot exactly represent the original
174 @cite{The direction of rounding when a floating-point number is
175 converted to a narrower floating-point number (6.3.1.5).}
178 @cite{How the nearest representable value or the larger or smaller
179 representable value immediately adjacent to the nearest representable
180 value is chosen for certain floating constants (6.4.4.2).}
183 @cite{Whether and how floating expressions are contracted when not
184 disallowed by the @code{FP_CONTRACT} pragma (6.5).}
187 @cite{The default state for the @code{FENV_ACCESS} pragma (7.6.1).}
190 @cite{Additional floating-point exceptions, rounding modes, environments,
191 and classifications, and their macro names (7.6, 7.12).}
194 @cite{The default state for the @code{FP_CONTRACT} pragma (7.12.2).}
197 @cite{Whether the ``inexact'' floating-point exception can be raised
198 when the rounded result actually does equal the mathematical result
199 in an IEC 60559 conformant implementation (F.9).}
202 @cite{Whether the ``underflow'' (and ``inexact'') floating-point
203 exception can be raised when a result is tiny but not inexact in an
204 IEC 60559 conformant implementation (F.9).}
208 @node Arrays and pointers implementation
209 @section Arrays and pointers
213 @cite{The result of converting a pointer to an integer or
214 vice versa (6.3.2.3).}
216 A cast from pointer to integer discards most-significant bits if the
217 pointer representation is larger than the integer type,
218 sign-extends@footnote{Future versions of GCC may zero-extend, or use
219 a target-defined @code{ptr_extend} pattern. Do not rely on sign extension.}
220 if the pointer representation is smaller than the integer type, otherwise
221 the bits are unchanged.
222 @c ??? We've always claimed that pointers were unsigned entities.
223 @c Shouldn't we therefore be doing zero-extension? If so, the bug
224 @c is in convert_to_integer, where we call type_for_size and request
225 @c a signed integral type. On the other hand, it might be most useful
226 @c for the target if we extend according to POINTERS_EXTEND_UNSIGNED.
228 A cast from integer to pointer discards most-significant bits if the
229 pointer representation is smaller than the integer type, extends according
230 to the signedness of the integer type if the pointer representation
231 is larger than the integer type, otherwise the bits are unchanged.
233 When casting from pointer to integer and back again, the resulting
234 pointer must reference the same object as the original pointer, otherwise
235 the behavior is undefined. That is, one may not use integer arithmetic to
236 avoid the undefined behavior of pointer arithmetic as proscribed in 6.5.6/8.
239 @cite{The size of the result of subtracting two pointers to elements
240 of the same array (6.5.6).}
244 @node Hints implementation
249 @cite{The extent to which suggestions made by using the @code{register}
250 storage-class specifier are effective (6.7.1).}
252 The @code{register} specifier affects code generation only in these ways:
256 When used as part of the register variable extension, see
257 @ref{Explicit Reg Vars}.
260 When @option{-O0} is in use, the compiler allocates distinct stack
261 memory for all variables that do not have the @code{register}
262 storage-class specifier; if @code{register} is specified, the variable
263 may have a shorter lifespan than the code would indicate and may never
267 On some rare x86 targets, @code{setjmp} doesn't save the registers in
268 all circumstances. In those cases, GCC doesn't allocate any variables
269 in registers unless they are marked @code{register}.
274 @cite{The extent to which suggestions made by using the inline function
275 specifier are effective (6.7.4).}
277 GCC will not inline any functions if the @option{-fno-inline} option is
278 used or if @option{-O0} is used. Otherwise, GCC may still be unable to
279 inline a function for many reasons; the @option{-Winline} option may be
280 used to determine if a function has not been inlined and why not.
284 @node Structures unions enumerations and bit-fields implementation
285 @section Structures, unions, enumerations, and bit-fields
289 @cite{Whether a ``plain'' int bit-field is treated as a @code{signed int}
290 bit-field or as an @code{unsigned int} bit-field (6.7.2, 6.7.2.1).}
293 @cite{Allowable bit-field types other than @code{_Bool}, @code{signed int},
294 and @code{unsigned int} (6.7.2.1).}
297 @cite{Whether a bit-field can straddle a storage-unit boundary (6.7.2.1).}
300 @cite{The order of allocation of bit-fields within a unit (6.7.2.1).}
303 @cite{The alignment of non-bit-field members of structures (6.7.2.1).}
306 @cite{The integer type compatible with each enumerated type (6.7.2.2).}
310 @node Qualifiers implementation
315 @cite{What constitutes an access to an object that has volatile-qualified
320 @node Preprocessing directives implementation
321 @section Preprocessing directives
325 @cite{How sequences in both forms of header names are mapped to headers
326 or external source file names (6.4.7).}
329 @cite{Whether the value of a character constant in a constant expression
330 that controls conditional inclusion matches the value of the same character
331 constant in the execution character set (6.10.1).}
334 @cite{Whether the value of a single-character character constant in a
335 constant expression that controls conditional inclusion may have a
336 negative value (6.10.1).}
339 @cite{The places that are searched for an included @samp{<>} delimited
340 header, and how the places are specified or the header is
341 identified (6.10.2).}
344 @cite{How the named source file is searched for in an included @samp{""}
345 delimited header (6.10.2).}
348 @cite{The method by which preprocessing tokens (possibly resulting from
349 macro expansion) in a @code{#include} directive are combined into a header
353 @cite{The nesting limit for @code{#include} processing (6.10.2).}
355 GCC imposes a limit of 200 nested @code{#include}s.
358 @cite{Whether the @samp{#} operator inserts a @samp{\} character before
359 the @samp{\} character that begins a universal character name in a
360 character constant or string literal (6.10.3.2).}
363 @cite{The behavior on each recognized non-@code{STDC #pragma}
367 @cite{The definitions for @code{__DATE__} and @code{__TIME__} when
368 respectively, the date and time of translation are not available (6.10.8).}
370 If the date and time are not available, @code{__DATE__} expands to
371 @code{@w{"??? ?? ????"}} and @code{__TIME__} expands to
376 @node Library functions implementation
377 @section Library functions
379 The behavior of these points are dependent on the implementation
380 of the C library, and are not defined by GCC itself.
382 @node Architecture implementation
383 @section Architecture
387 @cite{The values or expressions assigned to the macros specified in the
388 headers @code{<float.h>}, @code{<limits.h>}, and @code{<stdint.h>}
389 (5.2.4.2, 7.18.2, 7.18.3).}
392 @cite{The number, order, and encoding of bytes in any object
393 (when not explicitly specified in this International Standard) (6.2.6.1).}
396 @cite{The value of the result of the sizeof operator (6.5.3.4).}
400 @node Locale-specific behavior implementation
401 @section Locale-specific behavior
403 The behavior of these points are dependent on the implementation
404 of the C library, and are not defined by GCC itself.
407 @chapter Extensions to the C Language Family
408 @cindex extensions, C language
409 @cindex C language extensions
412 GNU C provides several language features not found in ISO standard C@.
413 (The @option{-pedantic} option directs GCC to print a warning message if
414 any of these features is used.) To test for the availability of these
415 features in conditional compilation, check for a predefined macro
416 @code{__GNUC__}, which is always defined under GCC@.
418 These extensions are available in C and Objective-C@. Most of them are
419 also available in C++. @xref{C++ Extensions,,Extensions to the
420 C++ Language}, for extensions that apply @emph{only} to C++.
422 Some features that are in ISO C99 but not C89 or C++ are also, as
423 extensions, accepted by GCC in C89 mode and in C++.
426 * Statement Exprs:: Putting statements and declarations inside expressions.
427 * Local Labels:: Labels local to a statement-expression.
428 * Labels as Values:: Getting pointers to labels, and computed gotos.
429 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
430 * Constructing Calls:: Dispatching a call to another function.
431 * Typeof:: @code{typeof}: referring to the type of an expression.
432 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
433 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
434 * Long Long:: Double-word integers---@code{long long int}.
435 * Complex:: Data types for complex numbers.
436 * Hex Floats:: Hexadecimal floating-point constants.
437 * Zero Length:: Zero-length arrays.
438 * Variable Length:: Arrays whose length is computed at run time.
439 * Empty Structures:: Structures with no members.
440 * Variadic Macros:: Macros with a variable number of arguments.
441 * Escaped Newlines:: Slightly looser rules for escaped newlines.
442 * Multi-line Strings:: String literals with embedded newlines.
443 * Subscripting:: Any array can be subscripted, even if not an lvalue.
444 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
445 * Initializers:: Non-constant initializers.
446 * Compound Literals:: Compound literals give structures, unions
448 * Designated Inits:: Labeling elements of initializers.
449 * Cast to Union:: Casting to union type from any member of the union.
450 * Case Ranges:: `case 1 ... 9' and such.
451 * Mixed Declarations:: Mixing declarations and code.
452 * Function Attributes:: Declaring that functions have no side effects,
453 or that they can never return.
454 * Attribute Syntax:: Formal syntax for attributes.
455 * Function Prototypes:: Prototype declarations and old-style definitions.
456 * C++ Comments:: C++ comments are recognized.
457 * Dollar Signs:: Dollar sign is allowed in identifiers.
458 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
459 * Variable Attributes:: Specifying attributes of variables.
460 * Type Attributes:: Specifying attributes of types.
461 * Alignment:: Inquiring about the alignment of a type or variable.
462 * Inline:: Defining inline functions (as fast as macros).
463 * Extended Asm:: Assembler instructions with C expressions as operands.
464 (With them you can define ``built-in'' functions.)
465 * Constraints:: Constraints for asm operands
466 * Asm Labels:: Specifying the assembler name to use for a C symbol.
467 * Explicit Reg Vars:: Defining variables residing in specified registers.
468 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
469 * Incomplete Enums:: @code{enum foo;}, with details to follow.
470 * Function Names:: Printable strings which are the name of the current
472 * Return Address:: Getting the return or frame address of a function.
473 * Vector Extensions:: Using vector instructions through built-in functions.
474 * Other Builtins:: Other built-in functions.
475 * Target Builtins:: Built-in functions specific to particular targets.
476 * Pragmas:: Pragmas accepted by GCC.
477 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
478 * Thread-Local:: Per-thread variables.
481 @node Statement Exprs
482 @section Statements and Declarations in Expressions
483 @cindex statements inside expressions
484 @cindex declarations inside expressions
485 @cindex expressions containing statements
486 @cindex macros, statements in expressions
488 @c the above section title wrapped and causes an underfull hbox.. i
489 @c changed it from "within" to "in". --mew 4feb93
490 A compound statement enclosed in parentheses may appear as an expression
491 in GNU C@. This allows you to use loops, switches, and local variables
492 within an expression.
494 Recall that a compound statement is a sequence of statements surrounded
495 by braces; in this construct, parentheses go around the braces. For
499 (@{ int y = foo (); int z;
506 is a valid (though slightly more complex than necessary) expression
507 for the absolute value of @code{foo ()}.
509 The last thing in the compound statement should be an expression
510 followed by a semicolon; the value of this subexpression serves as the
511 value of the entire construct. (If you use some other kind of statement
512 last within the braces, the construct has type @code{void}, and thus
513 effectively no value.)
515 This feature is especially useful in making macro definitions ``safe'' (so
516 that they evaluate each operand exactly once). For example, the
517 ``maximum'' function is commonly defined as a macro in standard C as
521 #define max(a,b) ((a) > (b) ? (a) : (b))
525 @cindex side effects, macro argument
526 But this definition computes either @var{a} or @var{b} twice, with bad
527 results if the operand has side effects. In GNU C, if you know the
528 type of the operands (here let's assume @code{int}), you can define
529 the macro safely as follows:
532 #define maxint(a,b) \
533 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
536 Embedded statements are not allowed in constant expressions, such as
537 the value of an enumeration constant, the width of a bit-field, or
538 the initial value of a static variable.
540 If you don't know the type of the operand, you can still do this, but you
541 must use @code{typeof} (@pxref{Typeof}).
543 Statement expressions are not supported fully in G++, and their fate
544 there is unclear. (It is possible that they will become fully supported
545 at some point, or that they will be deprecated, or that the bugs that
546 are present will continue to exist indefinitely.) Presently, statement
547 expressions do not work well as default arguments.
549 In addition, there are semantic issues with statement-expressions in
550 C++. If you try to use statement-expressions instead of inline
551 functions in C++, you may be surprised at the way object destruction is
552 handled. For example:
555 #define foo(a) (@{int b = (a); b + 3; @})
559 does not work the same way as:
562 inline int foo(int a) @{ int b = a; return b + 3; @}
566 In particular, if the expression passed into @code{foo} involves the
567 creation of temporaries, the destructors for those temporaries will be
568 run earlier in the case of the macro than in the case of the function.
570 These considerations mean that it is probably a bad idea to use
571 statement-expressions of this form in header files that are designed to
572 work with C++. (Note that some versions of the GNU C Library contained
573 header files using statement-expression that lead to precisely this
577 @section Locally Declared Labels
579 @cindex macros, local labels
581 Each statement expression is a scope in which @dfn{local labels} can be
582 declared. A local label is simply an identifier; you can jump to it
583 with an ordinary @code{goto} statement, but only from within the
584 statement expression it belongs to.
586 A local label declaration looks like this:
589 __label__ @var{label};
596 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
599 Local label declarations must come at the beginning of the statement
600 expression, right after the @samp{(@{}, before any ordinary
603 The label declaration defines the label @emph{name}, but does not define
604 the label itself. You must do this in the usual way, with
605 @code{@var{label}:}, within the statements of the statement expression.
607 The local label feature is useful because statement expressions are
608 often used in macros. If the macro contains nested loops, a @code{goto}
609 can be useful for breaking out of them. However, an ordinary label
610 whose scope is the whole function cannot be used: if the macro can be
611 expanded several times in one function, the label will be multiply
612 defined in that function. A local label avoids this problem. For
616 #define SEARCH(array, target) \
619 typeof (target) _SEARCH_target = (target); \
620 typeof (*(array)) *_SEARCH_array = (array); \
623 for (i = 0; i < max; i++) \
624 for (j = 0; j < max; j++) \
625 if (_SEARCH_array[i][j] == _SEARCH_target) \
626 @{ value = i; goto found; @} \
633 @node Labels as Values
634 @section Labels as Values
635 @cindex labels as values
636 @cindex computed gotos
637 @cindex goto with computed label
638 @cindex address of a label
640 You can get the address of a label defined in the current function
641 (or a containing function) with the unary operator @samp{&&}. The
642 value has type @code{void *}. This value is a constant and can be used
643 wherever a constant of that type is valid. For example:
651 To use these values, you need to be able to jump to one. This is done
652 with the computed goto statement@footnote{The analogous feature in
653 Fortran is called an assigned goto, but that name seems inappropriate in
654 C, where one can do more than simply store label addresses in label
655 variables.}, @code{goto *@var{exp};}. For example,
662 Any expression of type @code{void *} is allowed.
664 One way of using these constants is in initializing a static array that
665 will serve as a jump table:
668 static void *array[] = @{ &&foo, &&bar, &&hack @};
671 Then you can select a label with indexing, like this:
678 Note that this does not check whether the subscript is in bounds---array
679 indexing in C never does that.
681 Such an array of label values serves a purpose much like that of the
682 @code{switch} statement. The @code{switch} statement is cleaner, so
683 use that rather than an array unless the problem does not fit a
684 @code{switch} statement very well.
686 Another use of label values is in an interpreter for threaded code.
687 The labels within the interpreter function can be stored in the
688 threaded code for super-fast dispatching.
690 You may not use this mechanism to jump to code in a different function.
691 If you do that, totally unpredictable things will happen. The best way to
692 avoid this is to store the label address only in automatic variables and
693 never pass it as an argument.
695 An alternate way to write the above example is
698 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
700 goto *(&&foo + array[i]);
704 This is more friendly to code living in shared libraries, as it reduces
705 the number of dynamic relocations that are needed, and by consequence,
706 allows the data to be read-only.
708 @node Nested Functions
709 @section Nested Functions
710 @cindex nested functions
711 @cindex downward funargs
714 A @dfn{nested function} is a function defined inside another function.
715 (Nested functions are not supported for GNU C++.) The nested function's
716 name is local to the block where it is defined. For example, here we
717 define a nested function named @code{square}, and call it twice:
721 foo (double a, double b)
723 double square (double z) @{ return z * z; @}
725 return square (a) + square (b);
730 The nested function can access all the variables of the containing
731 function that are visible at the point of its definition. This is
732 called @dfn{lexical scoping}. For example, here we show a nested
733 function which uses an inherited variable named @code{offset}:
737 bar (int *array, int offset, int size)
739 int access (int *array, int index)
740 @{ return array[index + offset]; @}
743 for (i = 0; i < size; i++)
744 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
749 Nested function definitions are permitted within functions in the places
750 where variable definitions are allowed; that is, in any block, before
751 the first statement in the block.
753 It is possible to call the nested function from outside the scope of its
754 name by storing its address or passing the address to another function:
757 hack (int *array, int size)
759 void store (int index, int value)
760 @{ array[index] = value; @}
762 intermediate (store, size);
766 Here, the function @code{intermediate} receives the address of
767 @code{store} as an argument. If @code{intermediate} calls @code{store},
768 the arguments given to @code{store} are used to store into @code{array}.
769 But this technique works only so long as the containing function
770 (@code{hack}, in this example) does not exit.
772 If you try to call the nested function through its address after the
773 containing function has exited, all hell will break loose. If you try
774 to call it after a containing scope level has exited, and if it refers
775 to some of the variables that are no longer in scope, you may be lucky,
776 but it's not wise to take the risk. If, however, the nested function
777 does not refer to anything that has gone out of scope, you should be
780 GCC implements taking the address of a nested function using a technique
781 called @dfn{trampolines}. A paper describing them is available as
784 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
786 A nested function can jump to a label inherited from a containing
787 function, provided the label was explicitly declared in the containing
788 function (@pxref{Local Labels}). Such a jump returns instantly to the
789 containing function, exiting the nested function which did the
790 @code{goto} and any intermediate functions as well. Here is an example:
794 bar (int *array, int offset, int size)
797 int access (int *array, int index)
801 return array[index + offset];
805 for (i = 0; i < size; i++)
806 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
810 /* @r{Control comes here from @code{access}
811 if it detects an error.} */
818 A nested function always has internal linkage. Declaring one with
819 @code{extern} is erroneous. If you need to declare the nested function
820 before its definition, use @code{auto} (which is otherwise meaningless
821 for function declarations).
824 bar (int *array, int offset, int size)
827 auto int access (int *, int);
829 int access (int *array, int index)
833 return array[index + offset];
839 @node Constructing Calls
840 @section Constructing Function Calls
841 @cindex constructing calls
842 @cindex forwarding calls
844 Using the built-in functions described below, you can record
845 the arguments a function received, and call another function
846 with the same arguments, without knowing the number or types
849 You can also record the return value of that function call,
850 and later return that value, without knowing what data type
851 the function tried to return (as long as your caller expects
854 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
855 This built-in function returns a pointer to data
856 describing how to perform a call with the same arguments as were passed
857 to the current function.
859 The function saves the arg pointer register, structure value address,
860 and all registers that might be used to pass arguments to a function
861 into a block of memory allocated on the stack. Then it returns the
862 address of that block.
865 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
866 This built-in function invokes @var{function}
867 with a copy of the parameters described by @var{arguments}
870 The value of @var{arguments} should be the value returned by
871 @code{__builtin_apply_args}. The argument @var{size} specifies the size
872 of the stack argument data, in bytes.
874 This function returns a pointer to data describing
875 how to return whatever value was returned by @var{function}. The data
876 is saved in a block of memory allocated on the stack.
878 It is not always simple to compute the proper value for @var{size}. The
879 value is used by @code{__builtin_apply} to compute the amount of data
880 that should be pushed on the stack and copied from the incoming argument
884 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
885 This built-in function returns the value described by @var{result} from
886 the containing function. You should specify, for @var{result}, a value
887 returned by @code{__builtin_apply}.
891 @section Referring to a Type with @code{typeof}
894 @cindex macros, types of arguments
896 Another way to refer to the type of an expression is with @code{typeof}.
897 The syntax of using of this keyword looks like @code{sizeof}, but the
898 construct acts semantically like a type name defined with @code{typedef}.
900 There are two ways of writing the argument to @code{typeof}: with an
901 expression or with a type. Here is an example with an expression:
908 This assumes that @code{x} is an array of pointers to functions;
909 the type described is that of the values of the functions.
911 Here is an example with a typename as the argument:
918 Here the type described is that of pointers to @code{int}.
920 If you are writing a header file that must work when included in ISO C
921 programs, write @code{__typeof__} instead of @code{typeof}.
922 @xref{Alternate Keywords}.
924 A @code{typeof}-construct can be used anywhere a typedef name could be
925 used. For example, you can use it in a declaration, in a cast, or inside
926 of @code{sizeof} or @code{typeof}.
928 @code{typeof} is often useful in conjunction with the
929 statements-within-expressions feature. Here is how the two together can
930 be used to define a safe ``maximum'' macro that operates on any
931 arithmetic type and evaluates each of its arguments exactly once:
935 (@{ typeof (a) _a = (a); \
936 typeof (b) _b = (b); \
937 _a > _b ? _a : _b; @})
940 @cindex underscores in variables in macros
941 @cindex @samp{_} in variables in macros
942 @cindex local variables in macros
943 @cindex variables, local, in macros
944 @cindex macros, local variables in
946 The reason for using names that start with underscores for the local
947 variables is to avoid conflicts with variable names that occur within the
948 expressions that are substituted for @code{a} and @code{b}. Eventually we
949 hope to design a new form of declaration syntax that allows you to declare
950 variables whose scopes start only after their initializers; this will be a
951 more reliable way to prevent such conflicts.
954 Some more examples of the use of @code{typeof}:
958 This declares @code{y} with the type of what @code{x} points to.
965 This declares @code{y} as an array of such values.
972 This declares @code{y} as an array of pointers to characters:
975 typeof (typeof (char *)[4]) y;
979 It is equivalent to the following traditional C declaration:
985 To see the meaning of the declaration using @code{typeof}, and why it
986 might be a useful way to write, let's rewrite it with these macros:
989 #define pointer(T) typeof(T *)
990 #define array(T, N) typeof(T [N])
994 Now the declaration can be rewritten this way:
997 array (pointer (char), 4) y;
1001 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1002 pointers to @code{char}.
1005 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1006 a more limited extension which permitted one to write
1009 typedef @var{T} = @var{expr};
1013 with the effect of declaring @var{T} to have the type of the expression
1014 @var{expr}. This extension does not work with GCC 3 (versions between
1015 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1016 relies on it should be rewritten to use @code{typeof}:
1019 typedef typeof(@var{expr}) @var{T};
1023 This will work with all versions of GCC@.
1026 @section Generalized Lvalues
1027 @cindex compound expressions as lvalues
1028 @cindex expressions, compound, as lvalues
1029 @cindex conditional expressions as lvalues
1030 @cindex expressions, conditional, as lvalues
1031 @cindex casts as lvalues
1032 @cindex generalized lvalues
1033 @cindex lvalues, generalized
1034 @cindex extensions, @code{?:}
1035 @cindex @code{?:} extensions
1037 Compound expressions, conditional expressions and casts are allowed as
1038 lvalues provided their operands are lvalues. This means that you can take
1039 their addresses or store values into them.
1041 Standard C++ allows compound expressions and conditional expressions as
1042 lvalues, and permits casts to reference type, so use of this extension
1043 is deprecated for C++ code.
1045 For example, a compound expression can be assigned, provided the last
1046 expression in the sequence is an lvalue. These two expressions are
1054 Similarly, the address of the compound expression can be taken. These two
1055 expressions are equivalent:
1062 A conditional expression is a valid lvalue if its type is not void and the
1063 true and false branches are both valid lvalues. For example, these two
1064 expressions are equivalent:
1068 (a ? b = 5 : (c = 5))
1071 A cast is a valid lvalue if its operand is an lvalue. A simple
1072 assignment whose left-hand side is a cast works by converting the
1073 right-hand side first to the specified type, then to the type of the
1074 inner left-hand side expression. After this is stored, the value is
1075 converted back to the specified type to become the value of the
1076 assignment. Thus, if @code{a} has type @code{char *}, the following two
1077 expressions are equivalent:
1081 (int)(a = (char *)(int)5)
1084 An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
1085 performs the arithmetic using the type resulting from the cast, and then
1086 continues as in the previous case. Therefore, these two expressions are
1091 (int)(a = (char *)(int) ((int)a + 5))
1094 You cannot take the address of an lvalue cast, because the use of its
1095 address would not work out coherently. Suppose that @code{&(int)f} were
1096 permitted, where @code{f} has type @code{float}. Then the following
1097 statement would try to store an integer bit-pattern where a floating
1098 point number belongs:
1104 This is quite different from what @code{(int)f = 1} would do---that
1105 would convert 1 to floating point and store it. Rather than cause this
1106 inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
1108 If you really do want an @code{int *} pointer with the address of
1109 @code{f}, you can simply write @code{(int *)&f}.
1112 @section Conditionals with Omitted Operands
1113 @cindex conditional expressions, extensions
1114 @cindex omitted middle-operands
1115 @cindex middle-operands, omitted
1116 @cindex extensions, @code{?:}
1117 @cindex @code{?:} extensions
1119 The middle operand in a conditional expression may be omitted. Then
1120 if the first operand is nonzero, its value is the value of the conditional
1123 Therefore, the expression
1130 has the value of @code{x} if that is nonzero; otherwise, the value of
1133 This example is perfectly equivalent to
1139 @cindex side effect in ?:
1140 @cindex ?: side effect
1142 In this simple case, the ability to omit the middle operand is not
1143 especially useful. When it becomes useful is when the first operand does,
1144 or may (if it is a macro argument), contain a side effect. Then repeating
1145 the operand in the middle would perform the side effect twice. Omitting
1146 the middle operand uses the value already computed without the undesirable
1147 effects of recomputing it.
1150 @section Double-Word Integers
1151 @cindex @code{long long} data types
1152 @cindex double-word arithmetic
1153 @cindex multiprecision arithmetic
1154 @cindex @code{LL} integer suffix
1155 @cindex @code{ULL} integer suffix
1157 ISO C99 supports data types for integers that are at least 64 bits wide,
1158 and as an extension GCC supports them in C89 mode and in C++.
1159 Simply write @code{long long int} for a signed integer, or
1160 @code{unsigned long long int} for an unsigned integer. To make an
1161 integer constant of type @code{long long int}, add the suffix @samp{LL}
1162 to the integer. To make an integer constant of type @code{unsigned long
1163 long int}, add the suffix @samp{ULL} to the integer.
1165 You can use these types in arithmetic like any other integer types.
1166 Addition, subtraction, and bitwise boolean operations on these types
1167 are open-coded on all types of machines. Multiplication is open-coded
1168 if the machine supports fullword-to-doubleword a widening multiply
1169 instruction. Division and shifts are open-coded only on machines that
1170 provide special support. The operations that are not open-coded use
1171 special library routines that come with GCC@.
1173 There may be pitfalls when you use @code{long long} types for function
1174 arguments, unless you declare function prototypes. If a function
1175 expects type @code{int} for its argument, and you pass a value of type
1176 @code{long long int}, confusion will result because the caller and the
1177 subroutine will disagree about the number of bytes for the argument.
1178 Likewise, if the function expects @code{long long int} and you pass
1179 @code{int}. The best way to avoid such problems is to use prototypes.
1182 @section Complex Numbers
1183 @cindex complex numbers
1184 @cindex @code{_Complex} keyword
1185 @cindex @code{__complex__} keyword
1187 ISO C99 supports complex floating data types, and as an extension GCC
1188 supports them in C89 mode and in C++, and supports complex integer data
1189 types which are not part of ISO C99. You can declare complex types
1190 using the keyword @code{_Complex}. As an extension, the older GNU
1191 keyword @code{__complex__} is also supported.
1193 For example, @samp{_Complex double x;} declares @code{x} as a
1194 variable whose real part and imaginary part are both of type
1195 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1196 have real and imaginary parts of type @code{short int}; this is not
1197 likely to be useful, but it shows that the set of complex types is
1200 To write a constant with a complex data type, use the suffix @samp{i} or
1201 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1202 has type @code{_Complex float} and @code{3i} has type
1203 @code{_Complex int}. Such a constant always has a pure imaginary
1204 value, but you can form any complex value you like by adding one to a
1205 real constant. This is a GNU extension; if you have an ISO C99
1206 conforming C library (such as GNU libc), and want to construct complex
1207 constants of floating type, you should include @code{<complex.h>} and
1208 use the macros @code{I} or @code{_Complex_I} instead.
1210 @cindex @code{__real__} keyword
1211 @cindex @code{__imag__} keyword
1212 To extract the real part of a complex-valued expression @var{exp}, write
1213 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1214 extract the imaginary part. This is a GNU extension; for values of
1215 floating type, you should use the ISO C99 functions @code{crealf},
1216 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1217 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1218 built-in functions by GCC@.
1220 @cindex complex conjugation
1221 The operator @samp{~} performs complex conjugation when used on a value
1222 with a complex type. This is a GNU extension; for values of
1223 floating type, you should use the ISO C99 functions @code{conjf},
1224 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1225 provided as built-in functions by GCC@.
1227 GCC can allocate complex automatic variables in a noncontiguous
1228 fashion; it's even possible for the real part to be in a register while
1229 the imaginary part is on the stack (or vice-versa). Only the DWARF2
1230 debug info format can represent this, so use of DWARF2 is recommended.
1231 If you are using the stabs debug info format, GCC describes a noncontiguous
1232 complex variable as if it were two separate variables of noncomplex type.
1233 If the variable's actual name is @code{foo}, the two fictitious
1234 variables are named @code{foo$real} and @code{foo$imag}. You can
1235 examine and set these two fictitious variables with your debugger.
1241 ISO C99 supports floating-point numbers written not only in the usual
1242 decimal notation, such as @code{1.55e1}, but also numbers such as
1243 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1244 supports this in C89 mode (except in some cases when strictly
1245 conforming) and in C++. In that format the
1246 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1247 mandatory. The exponent is a decimal number that indicates the power of
1248 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1255 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1256 is the same as @code{1.55e1}.
1258 Unlike for floating-point numbers in the decimal notation the exponent
1259 is always required in the hexadecimal notation. Otherwise the compiler
1260 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1261 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1262 extension for floating-point constants of type @code{float}.
1265 @section Arrays of Length Zero
1266 @cindex arrays of length zero
1267 @cindex zero-length arrays
1268 @cindex length-zero arrays
1269 @cindex flexible array members
1271 Zero-length arrays are allowed in GNU C@. They are very useful as the
1272 last element of a structure which is really a header for a variable-length
1281 struct line *thisline = (struct line *)
1282 malloc (sizeof (struct line) + this_length);
1283 thisline->length = this_length;
1286 In ISO C90, you would have to give @code{contents} a length of 1, which
1287 means either you waste space or complicate the argument to @code{malloc}.
1289 In ISO C99, you would use a @dfn{flexible array member}, which is
1290 slightly different in syntax and semantics:
1294 Flexible array members are written as @code{contents[]} without
1298 Flexible array members have incomplete type, and so the @code{sizeof}
1299 operator may not be applied. As a quirk of the original implementation
1300 of zero-length arrays, @code{sizeof} evaluates to zero.
1303 Flexible array members may only appear as the last member of a
1304 @code{struct} that is otherwise non-empty.
1307 A structure containing a flexible array member, or a union containing
1308 such a structure (possibly recursively), may not be a member of a
1309 structure or an element of an array. (However, these uses are
1310 permitted by GCC as extensions.)
1313 GCC versions before 3.0 allowed zero-length arrays to be statically
1314 initialized, as if they were flexible arrays. In addition to those
1315 cases that were useful, it also allowed initializations in situations
1316 that would corrupt later data. Non-empty initialization of zero-length
1317 arrays is now treated like any case where there are more initializer
1318 elements than the array holds, in that a suitable warning about "excess
1319 elements in array" is given, and the excess elements (all of them, in
1320 this case) are ignored.
1322 Instead GCC allows static initialization of flexible array members.
1323 This is equivalent to defining a new structure containing the original
1324 structure followed by an array of sufficient size to contain the data.
1325 I.e.@: in the following, @code{f1} is constructed as if it were declared
1331 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1334 struct f1 f1; int data[3];
1335 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1339 The convenience of this extension is that @code{f1} has the desired
1340 type, eliminating the need to consistently refer to @code{f2.f1}.
1342 This has symmetry with normal static arrays, in that an array of
1343 unknown size is also written with @code{[]}.
1345 Of course, this extension only makes sense if the extra data comes at
1346 the end of a top-level object, as otherwise we would be overwriting
1347 data at subsequent offsets. To avoid undue complication and confusion
1348 with initialization of deeply nested arrays, we simply disallow any
1349 non-empty initialization except when the structure is the top-level
1350 object. For example:
1353 struct foo @{ int x; int y[]; @};
1354 struct bar @{ struct foo z; @};
1356 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1357 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1358 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1359 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1362 @node Empty Structures
1363 @section Structures With No Members
1364 @cindex empty structures
1365 @cindex zero-size structures
1367 GCC permits a C structure to have no members:
1374 The structure will have size zero. In C++, empty structures are part
1375 of the language. G++ treats empty structures as if they had a single
1376 member of type @code{char}.
1378 @node Variable Length
1379 @section Arrays of Variable Length
1380 @cindex variable-length arrays
1381 @cindex arrays of variable length
1384 Variable-length automatic arrays are allowed in ISO C99, and as an
1385 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1386 implementation of variable-length arrays does not yet conform in detail
1387 to the ISO C99 standard.) These arrays are
1388 declared like any other automatic arrays, but with a length that is not
1389 a constant expression. The storage is allocated at the point of
1390 declaration and deallocated when the brace-level is exited. For
1395 concat_fopen (char *s1, char *s2, char *mode)
1397 char str[strlen (s1) + strlen (s2) + 1];
1400 return fopen (str, mode);
1404 @cindex scope of a variable length array
1405 @cindex variable-length array scope
1406 @cindex deallocating variable length arrays
1407 Jumping or breaking out of the scope of the array name deallocates the
1408 storage. Jumping into the scope is not allowed; you get an error
1411 @cindex @code{alloca} vs variable-length arrays
1412 You can use the function @code{alloca} to get an effect much like
1413 variable-length arrays. The function @code{alloca} is available in
1414 many other C implementations (but not in all). On the other hand,
1415 variable-length arrays are more elegant.
1417 There are other differences between these two methods. Space allocated
1418 with @code{alloca} exists until the containing @emph{function} returns.
1419 The space for a variable-length array is deallocated as soon as the array
1420 name's scope ends. (If you use both variable-length arrays and
1421 @code{alloca} in the same function, deallocation of a variable-length array
1422 will also deallocate anything more recently allocated with @code{alloca}.)
1424 You can also use variable-length arrays as arguments to functions:
1428 tester (int len, char data[len][len])
1434 The length of an array is computed once when the storage is allocated
1435 and is remembered for the scope of the array in case you access it with
1438 If you want to pass the array first and the length afterward, you can
1439 use a forward declaration in the parameter list---another GNU extension.
1443 tester (int len; char data[len][len], int len)
1449 @cindex parameter forward declaration
1450 The @samp{int len} before the semicolon is a @dfn{parameter forward
1451 declaration}, and it serves the purpose of making the name @code{len}
1452 known when the declaration of @code{data} is parsed.
1454 You can write any number of such parameter forward declarations in the
1455 parameter list. They can be separated by commas or semicolons, but the
1456 last one must end with a semicolon, which is followed by the ``real''
1457 parameter declarations. Each forward declaration must match a ``real''
1458 declaration in parameter name and data type. ISO C99 does not support
1459 parameter forward declarations.
1461 @node Variadic Macros
1462 @section Macros with a Variable Number of Arguments.
1463 @cindex variable number of arguments
1464 @cindex macro with variable arguments
1465 @cindex rest argument (in macro)
1466 @cindex variadic macros
1468 In the ISO C standard of 1999, a macro can be declared to accept a
1469 variable number of arguments much as a function can. The syntax for
1470 defining the macro is similar to that of a function. Here is an
1474 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1477 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1478 such a macro, it represents the zero or more tokens until the closing
1479 parenthesis that ends the invocation, including any commas. This set of
1480 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1481 wherever it appears. See the CPP manual for more information.
1483 GCC has long supported variadic macros, and used a different syntax that
1484 allowed you to give a name to the variable arguments just like any other
1485 argument. Here is an example:
1488 #define debug(format, args...) fprintf (stderr, format, args)
1491 This is in all ways equivalent to the ISO C example above, but arguably
1492 more readable and descriptive.
1494 GNU CPP has two further variadic macro extensions, and permits them to
1495 be used with either of the above forms of macro definition.
1497 In standard C, you are not allowed to leave the variable argument out
1498 entirely; but you are allowed to pass an empty argument. For example,
1499 this invocation is invalid in ISO C, because there is no comma after
1506 GNU CPP permits you to completely omit the variable arguments in this
1507 way. In the above examples, the compiler would complain, though since
1508 the expansion of the macro still has the extra comma after the format
1511 To help solve this problem, CPP behaves specially for variable arguments
1512 used with the token paste operator, @samp{##}. If instead you write
1515 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1518 and if the variable arguments are omitted or empty, the @samp{##}
1519 operator causes the preprocessor to remove the comma before it. If you
1520 do provide some variable arguments in your macro invocation, GNU CPP
1521 does not complain about the paste operation and instead places the
1522 variable arguments after the comma. Just like any other pasted macro
1523 argument, these arguments are not macro expanded.
1525 @node Escaped Newlines
1526 @section Slightly Looser Rules for Escaped Newlines
1527 @cindex escaped newlines
1528 @cindex newlines (escaped)
1530 Recently, the preprocessor has relaxed its treatment of escaped
1531 newlines. Previously, the newline had to immediately follow a
1532 backslash. The current implementation allows whitespace in the form of
1533 spaces, horizontal and vertical tabs, and form feeds between the
1534 backslash and the subsequent newline. The preprocessor issues a
1535 warning, but treats it as a valid escaped newline and combines the two
1536 lines to form a single logical line. This works within comments and
1537 tokens, including multi-line strings, as well as between tokens.
1538 Comments are @emph{not} treated as whitespace for the purposes of this
1539 relaxation, since they have not yet been replaced with spaces.
1541 @node Multi-line Strings
1542 @section String Literals with Embedded Newlines
1543 @cindex multi-line string literals
1545 As an extension, GNU CPP permits string literals to cross multiple lines
1546 without escaping the embedded newlines. Each embedded newline is
1547 replaced with a single @samp{\n} character in the resulting string
1548 literal, regardless of what form the newline took originally.
1550 CPP currently allows such strings in directives as well (other than the
1551 @samp{#include} family). This is deprecated and will eventually be
1555 @section Non-Lvalue Arrays May Have Subscripts
1556 @cindex subscripting
1557 @cindex arrays, non-lvalue
1559 @cindex subscripting and function values
1560 In ISO C99, arrays that are not lvalues still decay to pointers, and
1561 may be subscripted, although they may not be modified or used after
1562 the next sequence point and the unary @samp{&} operator may not be
1563 applied to them. As an extension, GCC allows such arrays to be
1564 subscripted in C89 mode, though otherwise they do not decay to
1565 pointers outside C99 mode. For example,
1566 this is valid in GNU C though not valid in C89:
1570 struct foo @{int a[4];@};
1576 return f().a[index];
1582 @section Arithmetic on @code{void}- and Function-Pointers
1583 @cindex void pointers, arithmetic
1584 @cindex void, size of pointer to
1585 @cindex function pointers, arithmetic
1586 @cindex function, size of pointer to
1588 In GNU C, addition and subtraction operations are supported on pointers to
1589 @code{void} and on pointers to functions. This is done by treating the
1590 size of a @code{void} or of a function as 1.
1592 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1593 and on function types, and returns 1.
1595 @opindex Wpointer-arith
1596 The option @option{-Wpointer-arith} requests a warning if these extensions
1600 @section Non-Constant Initializers
1601 @cindex initializers, non-constant
1602 @cindex non-constant initializers
1604 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1605 automatic variable are not required to be constant expressions in GNU C@.
1606 Here is an example of an initializer with run-time varying elements:
1609 foo (float f, float g)
1611 float beat_freqs[2] = @{ f-g, f+g @};
1616 @node Compound Literals
1617 @section Compound Literals
1618 @cindex constructor expressions
1619 @cindex initializations in expressions
1620 @cindex structures, constructor expression
1621 @cindex expressions, constructor
1622 @cindex compound literals
1623 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1625 ISO C99 supports compound literals. A compound literal looks like
1626 a cast containing an initializer. Its value is an object of the
1627 type specified in the cast, containing the elements specified in
1628 the initializer; it is an lvalue. As an extension, GCC supports
1629 compound literals in C89 mode and in C++.
1631 Usually, the specified type is a structure. Assume that
1632 @code{struct foo} and @code{structure} are declared as shown:
1635 struct foo @{int a; char b[2];@} structure;
1639 Here is an example of constructing a @code{struct foo} with a compound literal:
1642 structure = ((struct foo) @{x + y, 'a', 0@});
1646 This is equivalent to writing the following:
1650 struct foo temp = @{x + y, 'a', 0@};
1655 You can also construct an array. If all the elements of the compound literal
1656 are (made up of) simple constant expressions, suitable for use in
1657 initializers of objects of static storage duration, then the compound
1658 literal can be coerced to a pointer to its first element and used in
1659 such an initializer, as shown here:
1662 char **foo = (char *[]) @{ "x", "y", "z" @};
1665 Compound literals for scalar types and union types are is
1666 also allowed, but then the compound literal is equivalent
1669 As a GNU extension, GCC allows initialization of objects with static storage
1670 duration by compound literals (which is not possible in ISO C99, because
1671 the initializer is not a constant).
1672 It is handled as if the object was initialized only with the bracket
1673 enclosed list if compound literal's and object types match.
1674 The initializer list of the compound literal must be constant.
1675 If the object being initialized has array type of unknown size, the size is
1676 determined by compound literal size.
1679 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1680 static int y[] = (int []) @{1, 2, 3@};
1681 static int z[] = (int [3]) @{1@};
1685 The above lines are equivalent to the following:
1687 static struct foo x = @{1, 'a', 'b'@};
1688 static int y[] = @{1, 2, 3@};
1689 static int z[] = @{1, 0, 0@};
1692 @node Designated Inits
1693 @section Designated Initializers
1694 @cindex initializers with labeled elements
1695 @cindex labeled elements in initializers
1696 @cindex case labels in initializers
1697 @cindex designated initializers
1699 Standard C89 requires the elements of an initializer to appear in a fixed
1700 order, the same as the order of the elements in the array or structure
1703 In ISO C99 you can give the elements in any order, specifying the array
1704 indices or structure field names they apply to, and GNU C allows this as
1705 an extension in C89 mode as well. This extension is not
1706 implemented in GNU C++.
1708 To specify an array index, write
1709 @samp{[@var{index}] =} before the element value. For example,
1712 int a[6] = @{ [4] = 29, [2] = 15 @};
1719 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1723 The index values must be constant expressions, even if the array being
1724 initialized is automatic.
1726 An alternative syntax for this which has been obsolete since GCC 2.5 but
1727 GCC still accepts is to write @samp{[@var{index}]} before the element
1728 value, with no @samp{=}.
1730 To initialize a range of elements to the same value, write
1731 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1732 extension. For example,
1735 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1739 If the value in it has side-effects, the side-effects will happen only once,
1740 not for each initialized field by the range initializer.
1743 Note that the length of the array is the highest value specified
1746 In a structure initializer, specify the name of a field to initialize
1747 with @samp{.@var{fieldname} =} before the element value. For example,
1748 given the following structure,
1751 struct point @{ int x, y; @};
1755 the following initialization
1758 struct point p = @{ .y = yvalue, .x = xvalue @};
1765 struct point p = @{ xvalue, yvalue @};
1768 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1769 @samp{@var{fieldname}:}, as shown here:
1772 struct point p = @{ y: yvalue, x: xvalue @};
1776 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1777 @dfn{designator}. You can also use a designator (or the obsolete colon
1778 syntax) when initializing a union, to specify which element of the union
1779 should be used. For example,
1782 union foo @{ int i; double d; @};
1784 union foo f = @{ .d = 4 @};
1788 will convert 4 to a @code{double} to store it in the union using
1789 the second element. By contrast, casting 4 to type @code{union foo}
1790 would store it into the union as the integer @code{i}, since it is
1791 an integer. (@xref{Cast to Union}.)
1793 You can combine this technique of naming elements with ordinary C
1794 initialization of successive elements. Each initializer element that
1795 does not have a designator applies to the next consecutive element of the
1796 array or structure. For example,
1799 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1806 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1809 Labeling the elements of an array initializer is especially useful
1810 when the indices are characters or belong to an @code{enum} type.
1815 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1816 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1819 @cindex designator lists
1820 You can also write a series of @samp{.@var{fieldname}} and
1821 @samp{[@var{index}]} designators before an @samp{=} to specify a
1822 nested subobject to initialize; the list is taken relative to the
1823 subobject corresponding to the closest surrounding brace pair. For
1824 example, with the @samp{struct point} declaration above:
1827 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1831 If the same field is initialized multiple times, it will have value from
1832 the last initialization. If any such overridden initialization has
1833 side-effect, it is unspecified whether the side-effect happens or not.
1834 Currently, gcc will discard them and issue a warning.
1837 @section Case Ranges
1839 @cindex ranges in case statements
1841 You can specify a range of consecutive values in a single @code{case} label,
1845 case @var{low} ... @var{high}:
1849 This has the same effect as the proper number of individual @code{case}
1850 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1852 This feature is especially useful for ranges of ASCII character codes:
1858 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1859 it may be parsed wrong when you use it with integer values. For example,
1874 @section Cast to a Union Type
1875 @cindex cast to a union
1876 @cindex union, casting to a
1878 A cast to union type is similar to other casts, except that the type
1879 specified is a union type. You can specify the type either with
1880 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1881 a constructor though, not a cast, and hence does not yield an lvalue like
1882 normal casts. (@xref{Compound Literals}.)
1884 The types that may be cast to the union type are those of the members
1885 of the union. Thus, given the following union and variables:
1888 union foo @{ int i; double d; @};
1894 both @code{x} and @code{y} can be cast to type @code{union foo}.
1896 Using the cast as the right-hand side of an assignment to a variable of
1897 union type is equivalent to storing in a member of the union:
1902 u = (union foo) x @equiv{} u.i = x
1903 u = (union foo) y @equiv{} u.d = y
1906 You can also use the union cast as a function argument:
1909 void hack (union foo);
1911 hack ((union foo) x);
1914 @node Mixed Declarations
1915 @section Mixed Declarations and Code
1916 @cindex mixed declarations and code
1917 @cindex declarations, mixed with code
1918 @cindex code, mixed with declarations
1920 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1921 within compound statements. As an extension, GCC also allows this in
1922 C89 mode. For example, you could do:
1931 Each identifier is visible from where it is declared until the end of
1932 the enclosing block.
1934 @node Function Attributes
1935 @section Declaring Attributes of Functions
1936 @cindex function attributes
1937 @cindex declaring attributes of functions
1938 @cindex functions that never return
1939 @cindex functions that have no side effects
1940 @cindex functions in arbitrary sections
1941 @cindex functions that behave like malloc
1942 @cindex @code{volatile} applied to function
1943 @cindex @code{const} applied to function
1944 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1945 @cindex functions with non-null pointer arguments
1946 @cindex functions that are passed arguments in registers on the 386
1947 @cindex functions that pop the argument stack on the 386
1948 @cindex functions that do not pop the argument stack on the 386
1950 In GNU C, you declare certain things about functions called in your program
1951 which help the compiler optimize function calls and check your code more
1954 The keyword @code{__attribute__} allows you to specify special
1955 attributes when making a declaration. This keyword is followed by an
1956 attribute specification inside double parentheses. The following
1957 attributes are currently defined for functions on all targets:
1958 @code{noreturn}, @code{noinline}, @code{always_inline},
1959 @code{pure}, @code{const}, @code{nothrow},
1960 @code{format}, @code{format_arg}, @code{no_instrument_function},
1961 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1962 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1963 @code{alias}, and @code{nonnull}. Several other attributes are defined
1964 for functions on particular target systems. Other attributes, including
1965 @code{section} are supported for variables declarations
1966 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1968 You may also specify attributes with @samp{__} preceding and following
1969 each keyword. This allows you to use them in header files without
1970 being concerned about a possible macro of the same name. For example,
1971 you may use @code{__noreturn__} instead of @code{noreturn}.
1973 @xref{Attribute Syntax}, for details of the exact syntax for using
1977 @cindex @code{noreturn} function attribute
1979 A few standard library functions, such as @code{abort} and @code{exit},
1980 cannot return. GCC knows this automatically. Some programs define
1981 their own functions that never return. You can declare them
1982 @code{noreturn} to tell the compiler this fact. For example,
1986 void fatal () __attribute__ ((noreturn));
1989 fatal (/* @r{@dots{}} */)
1991 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
1997 The @code{noreturn} keyword tells the compiler to assume that
1998 @code{fatal} cannot return. It can then optimize without regard to what
1999 would happen if @code{fatal} ever did return. This makes slightly
2000 better code. More importantly, it helps avoid spurious warnings of
2001 uninitialized variables.
2003 Do not assume that registers saved by the calling function are
2004 restored before calling the @code{noreturn} function.
2006 It does not make sense for a @code{noreturn} function to have a return
2007 type other than @code{void}.
2009 The attribute @code{noreturn} is not implemented in GCC versions
2010 earlier than 2.5. An alternative way to declare that a function does
2011 not return, which works in the current version and in some older
2012 versions, is as follows:
2015 typedef void voidfn ();
2017 volatile voidfn fatal;
2020 @cindex @code{noinline} function attribute
2022 This function attribute prevents a function from being considered for
2025 @cindex @code{always_inline} function attribute
2027 Generally, functions are not inlined unless optimization is specified.
2028 For functions declared inline, this attribute inlines the function even
2029 if no optimization level was specified.
2031 @cindex @code{pure} function attribute
2033 Many functions have no effects except the return value and their
2034 return value depends only on the parameters and/or global variables.
2035 Such a function can be subject
2036 to common subexpression elimination and loop optimization just as an
2037 arithmetic operator would be. These functions should be declared
2038 with the attribute @code{pure}. For example,
2041 int square (int) __attribute__ ((pure));
2045 says that the hypothetical function @code{square} is safe to call
2046 fewer times than the program says.
2048 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2049 Interesting non-pure functions are functions with infinite loops or those
2050 depending on volatile memory or other system resource, that may change between
2051 two consecutive calls (such as @code{feof} in a multithreading environment).
2053 The attribute @code{pure} is not implemented in GCC versions earlier
2055 @cindex @code{const} function attribute
2057 Many functions do not examine any values except their arguments, and
2058 have no effects except the return value. Basically this is just slightly
2059 more strict class than the @code{pure} attribute above, since function is not
2060 allowed to read global memory.
2062 @cindex pointer arguments
2063 Note that a function that has pointer arguments and examines the data
2064 pointed to must @emph{not} be declared @code{const}. Likewise, a
2065 function that calls a non-@code{const} function usually must not be
2066 @code{const}. It does not make sense for a @code{const} function to
2069 The attribute @code{const} is not implemented in GCC versions earlier
2070 than 2.5. An alternative way to declare that a function has no side
2071 effects, which works in the current version and in some older versions,
2075 typedef int intfn ();
2077 extern const intfn square;
2080 This approach does not work in GNU C++ from 2.6.0 on, since the language
2081 specifies that the @samp{const} must be attached to the return value.
2083 @cindex @code{nothrow} function attribute
2085 The @code{nothrow} attribute is used to inform the compiler that a
2086 function cannot throw an exception. For example, most functions in
2087 the standard C library can be guaranteed not to throw an exception
2088 with the notable exceptions of @code{qsort} and @code{bsearch} that
2089 take function pointer arguments. The @code{nothrow} attribute is not
2090 implemented in GCC versions earlier than 3.2.
2092 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2093 @cindex @code{format} function attribute
2095 The @code{format} attribute specifies that a function takes @code{printf},
2096 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2097 should be type-checked against a format string. For example, the
2102 my_printf (void *my_object, const char *my_format, ...)
2103 __attribute__ ((format (printf, 2, 3)));
2107 causes the compiler to check the arguments in calls to @code{my_printf}
2108 for consistency with the @code{printf} style format string argument
2111 The parameter @var{archetype} determines how the format string is
2112 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2113 or @code{strfmon}. (You can also use @code{__printf__},
2114 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2115 parameter @var{string-index} specifies which argument is the format
2116 string argument (starting from 1), while @var{first-to-check} is the
2117 number of the first argument to check against the format string. For
2118 functions where the arguments are not available to be checked (such as
2119 @code{vprintf}), specify the third parameter as zero. In this case the
2120 compiler only checks the format string for consistency. For
2121 @code{strftime} formats, the third parameter is required to be zero.
2122 Since non-static C++ methods have an implicit @code{this} argument, the
2123 arguments of such methods should be counted from two, not one, when
2124 giving values for @var{string-index} and @var{first-to-check}.
2126 In the example above, the format string (@code{my_format}) is the second
2127 argument of the function @code{my_print}, and the arguments to check
2128 start with the third argument, so the correct parameters for the format
2129 attribute are 2 and 3.
2131 @opindex ffreestanding
2132 The @code{format} attribute allows you to identify your own functions
2133 which take format strings as arguments, so that GCC can check the
2134 calls to these functions for errors. The compiler always (unless
2135 @option{-ffreestanding} is used) checks formats
2136 for the standard library functions @code{printf}, @code{fprintf},
2137 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2138 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2139 warnings are requested (using @option{-Wformat}), so there is no need to
2140 modify the header file @file{stdio.h}. In C99 mode, the functions
2141 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2142 @code{vsscanf} are also checked. Except in strictly conforming C
2143 standard modes, the X/Open function @code{strfmon} is also checked as
2144 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2145 @xref{C Dialect Options,,Options Controlling C Dialect}.
2147 @item format_arg (@var{string-index})
2148 @cindex @code{format_arg} function attribute
2149 @opindex Wformat-nonliteral
2150 The @code{format_arg} attribute specifies that a function takes a format
2151 string for a @code{printf}, @code{scanf}, @code{strftime} or
2152 @code{strfmon} style function and modifies it (for example, to translate
2153 it into another language), so the result can be passed to a
2154 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2155 function (with the remaining arguments to the format function the same
2156 as they would have been for the unmodified string). For example, the
2161 my_dgettext (char *my_domain, const char *my_format)
2162 __attribute__ ((format_arg (2)));
2166 causes the compiler to check the arguments in calls to a @code{printf},
2167 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2168 format string argument is a call to the @code{my_dgettext} function, for
2169 consistency with the format string argument @code{my_format}. If the
2170 @code{format_arg} attribute had not been specified, all the compiler
2171 could tell in such calls to format functions would be that the format
2172 string argument is not constant; this would generate a warning when
2173 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2174 without the attribute.
2176 The parameter @var{string-index} specifies which argument is the format
2177 string argument (starting from one). Since non-static C++ methods have
2178 an implicit @code{this} argument, the arguments of such methods should
2179 be counted from two.
2181 The @code{format-arg} attribute allows you to identify your own
2182 functions which modify format strings, so that GCC can check the
2183 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2184 type function whose operands are a call to one of your own function.
2185 The compiler always treats @code{gettext}, @code{dgettext}, and
2186 @code{dcgettext} in this manner except when strict ISO C support is
2187 requested by @option{-ansi} or an appropriate @option{-std} option, or
2188 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2189 Controlling C Dialect}.
2191 @item nonnull (@var{arg-index}, @dots{})
2192 @cindex @code{nonnull} function attribute
2193 The @code{nonnull} attribute specifies that some function parameters should
2194 be non-null pointers. For instance, the declaration:
2198 my_memcpy (void *dest, const void *src, size_t len)
2199 __attribute__((nonnull (1, 2)));
2203 causes the compiler to check that, in calls to @code{my_memcpy},
2204 arguments @var{dest} and @var{src} are non-null. If the compiler
2205 determines that a null pointer is passed in an argument slot marked
2206 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2207 is issued. The compiler may also choose to make optimizations based
2208 on the knowledge that certain function arguments will not be null.
2210 If no argument index list is given to the @code{nonnull} attribute,
2211 all pointer arguments are marked as non-null. To illustrate, the
2212 following declaration is equivalent to the previous example:
2216 my_memcpy (void *dest, const void *src, size_t len)
2217 __attribute__((nonnull));
2220 @item no_instrument_function
2221 @cindex @code{no_instrument_function} function attribute
2222 @opindex finstrument-functions
2223 If @option{-finstrument-functions} is given, profiling function calls will
2224 be generated at entry and exit of most user-compiled functions.
2225 Functions with this attribute will not be so instrumented.
2227 @item section ("@var{section-name}")
2228 @cindex @code{section} function attribute
2229 Normally, the compiler places the code it generates in the @code{text} section.
2230 Sometimes, however, you need additional sections, or you need certain
2231 particular functions to appear in special sections. The @code{section}
2232 attribute specifies that a function lives in a particular section.
2233 For example, the declaration:
2236 extern void foobar (void) __attribute__ ((section ("bar")));
2240 puts the function @code{foobar} in the @code{bar} section.
2242 Some file formats do not support arbitrary sections so the @code{section}
2243 attribute is not available on all platforms.
2244 If you need to map the entire contents of a module to a particular
2245 section, consider using the facilities of the linker instead.
2249 @cindex @code{constructor} function attribute
2250 @cindex @code{destructor} function attribute
2251 The @code{constructor} attribute causes the function to be called
2252 automatically before execution enters @code{main ()}. Similarly, the
2253 @code{destructor} attribute causes the function to be called
2254 automatically after @code{main ()} has completed or @code{exit ()} has
2255 been called. Functions with these attributes are useful for
2256 initializing data that will be used implicitly during the execution of
2259 These attributes are not currently implemented for Objective-C@.
2261 @cindex @code{unused} attribute.
2263 This attribute, attached to a function, means that the function is meant
2264 to be possibly unused. GCC will not produce a warning for this
2265 function. GNU C++ does not currently support this attribute as
2266 definitions without parameters are valid in C++.
2268 @cindex @code{used} attribute.
2270 This attribute, attached to a function, means that code must be emitted
2271 for the function even if it appears that the function is not referenced.
2272 This is useful, for example, when the function is referenced only in
2275 @cindex @code{deprecated} attribute.
2277 The @code{deprecated} attribute results in a warning if the function
2278 is used anywhere in the source file. This is useful when identifying
2279 functions that are expected to be removed in a future version of a
2280 program. The warning also includes the location of the declaration
2281 of the deprecated function, to enable users to easily find further
2282 information about why the function is deprecated, or what they should
2283 do instead. Note that the warnings only occurs for uses:
2286 int old_fn () __attribute__ ((deprecated));
2288 int (*fn_ptr)() = old_fn;
2291 results in a warning on line 3 but not line 2.
2293 The @code{deprecated} attribute can also be used for variables and
2294 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2297 @cindex @code{weak} attribute
2298 The @code{weak} attribute causes the declaration to be emitted as a weak
2299 symbol rather than a global. This is primarily useful in defining
2300 library functions which can be overridden in user code, though it can
2301 also be used with non-function declarations. Weak symbols are supported
2302 for ELF targets, and also for a.out targets when using the GNU assembler
2306 @cindex @code{malloc} attribute
2307 The @code{malloc} attribute is used to tell the compiler that a function
2308 may be treated as if it were the malloc function. The compiler assumes
2309 that calls to malloc result in a pointers that cannot alias anything.
2310 This will often improve optimization.
2312 @item alias ("@var{target}")
2313 @cindex @code{alias} attribute
2314 The @code{alias} attribute causes the declaration to be emitted as an
2315 alias for another symbol, which must be specified. For instance,
2318 void __f () @{ /* @r{Do something.} */; @}
2319 void f () __attribute__ ((weak, alias ("__f")));
2322 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2323 mangled name for the target must be used.
2325 Not all target machines support this attribute.
2327 @item visibility ("@var{visibility_type}")
2328 @cindex @code{visibility} attribute
2329 The @code{visibility} attribute on ELF targets causes the declaration
2330 to be emitted with default, hidden, protected or internal visibility.
2333 void __attribute__ ((visibility ("protected")))
2334 f () @{ /* @r{Do something.} */; @}
2335 int i __attribute__ ((visibility ("hidden")));
2338 See the ELF gABI for complete details, but the short story is
2342 Default visibility is the normal case for ELF. This value is
2343 available for the visibility attribute to override other options
2344 that may change the assumed visibility of symbols.
2347 Hidden visibility indicates that the symbol will not be placed into
2348 the dynamic symbol table, so no other @dfn{module} (executable or
2349 shared library) can reference it directly.
2352 Protected visibility indicates that the symbol will be placed in the
2353 dynamic symbol table, but that references within the defining module
2354 will bind to the local symbol. That is, the symbol cannot be overridden
2358 Internal visibility is like hidden visibility, but with additional
2359 processor specific semantics. Unless otherwise specified by the psABI,
2360 gcc defines internal visibility to mean that the function is @emph{never}
2361 called from another module. Note that hidden symbols, while then cannot
2362 be referenced directly by other modules, can be referenced indirectly via
2363 function pointers. By indicating that a symbol cannot be called from
2364 outside the module, gcc may for instance omit the load of a PIC register
2365 since it is known that the calling function loaded the correct value.
2368 Not all ELF targets support this attribute.
2370 @item regparm (@var{number})
2371 @cindex functions that are passed arguments in registers on the 386
2372 On the Intel 386, the @code{regparm} attribute causes the compiler to
2373 pass up to @var{number} integer arguments in registers EAX,
2374 EDX, and ECX instead of on the stack. Functions that take a
2375 variable number of arguments will continue to be passed all of their
2376 arguments on the stack.
2379 @cindex functions that pop the argument stack on the 386
2380 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2381 assume that the called function will pop off the stack space used to
2382 pass arguments, unless it takes a variable number of arguments.
2385 @cindex functions that pop the argument stack on the 386
2386 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2387 pass the first two arguments in the registers ECX and EDX. Subsequent
2388 arguments are passed on the stack. The called function will pop the
2389 arguments off the stack. If the number of arguments is variable all
2390 arguments are pushed on the stack.
2393 @cindex functions that do pop the argument stack on the 386
2395 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2396 assume that the calling function will pop off the stack space used to
2397 pass arguments. This is
2398 useful to override the effects of the @option{-mrtd} switch.
2400 @item longcall/shortcall
2401 @cindex functions called via pointer on the RS/6000 and PowerPC
2402 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2403 compiler to always call this function via a pointer, just as it would if
2404 the @option{-mlongcall} option had been specified. The @code{shortcall}
2405 attribute causes the compiler not to do this. These attributes override
2406 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2409 @xref{RS/6000 and PowerPC Options}, for more information on when long
2410 calls are and are not necessary.
2412 @item long_call/short_call
2413 @cindex indirect calls on ARM
2414 This attribute allows to specify how to call a particular function on
2415 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2416 command line switch and @code{#pragma long_calls} settings. The
2417 @code{long_call} attribute causes the compiler to always call the
2418 function by first loading its address into a register and then using the
2419 contents of that register. The @code{short_call} attribute always places
2420 the offset to the function from the call site into the @samp{BL}
2421 instruction directly.
2423 @item function_vector
2424 @cindex calling functions through the function vector on the H8/300 processors
2425 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2426 function should be called through the function vector. Calling a
2427 function through the function vector will reduce code size, however;
2428 the function vector has a limited size (maximum 128 entries on the H8/300
2429 and 64 entries on the H8/300H) and shares space with the interrupt vector.
2431 You must use GAS and GLD from GNU binutils version 2.7 or later for
2432 this attribute to work correctly.
2435 @cindex interrupt handler functions
2436 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2437 that the specified function is an interrupt handler. The compiler will
2438 generate function entry and exit sequences suitable for use in an
2439 interrupt handler when this attribute is present.
2441 Note, interrupt handlers for the H8/300, H8/300H and SH processors can
2442 be specified via the @code{interrupt_handler} attribute.
2444 Note, on the AVR interrupts will be enabled inside the function.
2446 Note, for the ARM you can specify the kind of interrupt to be handled by
2447 adding an optional parameter to the interrupt attribute like this:
2450 void f () __attribute__ ((interrupt ("IRQ")));
2453 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2455 @item interrupt_handler
2456 @cindex interrupt handler functions on the H8/300 and SH processors
2457 Use this attribute on the H8/300, H8/300H and SH to indicate that the
2458 specified function is an interrupt handler. The compiler will generate
2459 function entry and exit sequences suitable for use in an interrupt
2460 handler when this attribute is present.
2463 Use this attribute on the SH to indicate an @code{interrupt_handler}
2464 function should switch to an alternate stack. It expects a string
2465 argument that names a global variable holding the address of the
2470 void f () __attribute__ ((interrupt_handler,
2471 sp_switch ("alt_stack")));
2475 Use this attribute on the SH for an @code{interrupt_handle} to return using
2476 @code{trapa} instead of @code{rte}. This attribute expects an integer
2477 argument specifying the trap number to be used.
2480 @cindex eight bit data on the H8/300 and H8/300H
2481 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2482 variable should be placed into the eight bit data section.
2483 The compiler will generate more efficient code for certain operations
2484 on data in the eight bit data area. Note the eight bit data area is limited to
2487 You must use GAS and GLD from GNU binutils version 2.7 or later for
2488 this attribute to work correctly.
2491 @cindex tiny data section on the H8/300H
2492 Use this attribute on the H8/300H to indicate that the specified
2493 variable should be placed into the tiny data section.
2494 The compiler will generate more efficient code for loads and stores
2495 on data in the tiny data section. Note the tiny data area is limited to
2496 slightly under 32kbytes of data.
2499 @cindex signal handler functions on the AVR processors
2500 Use this attribute on the AVR to indicate that the specified
2501 function is an signal handler. The compiler will generate function
2502 entry and exit sequences suitable for use in an signal handler when this
2503 attribute is present. Interrupts will be disabled inside function.
2506 @cindex function without a prologue/epilogue code
2507 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2508 specified function do not need prologue/epilogue sequences generated by
2509 the compiler. It is up to the programmer to provide these sequences.
2511 @item model (@var{model-name})
2512 @cindex function addressability on the M32R/D
2513 Use this attribute on the M32R/D to set the addressability of an object,
2514 and the code generated for a function.
2515 The identifier @var{model-name} is one of @code{small}, @code{medium},
2516 or @code{large}, representing each of the code models.
2518 Small model objects live in the lower 16MB of memory (so that their
2519 addresses can be loaded with the @code{ld24} instruction), and are
2520 callable with the @code{bl} instruction.
2522 Medium model objects may live anywhere in the 32-bit address space (the
2523 compiler will generate @code{seth/add3} instructions to load their addresses),
2524 and are callable with the @code{bl} instruction.
2526 Large model objects may live anywhere in the 32-bit address space (the
2527 compiler will generate @code{seth/add3} instructions to load their addresses),
2528 and may not be reachable with the @code{bl} instruction (the compiler will
2529 generate the much slower @code{seth/add3/jl} instruction sequence).
2532 @cindex functions which handle memory bank switching
2533 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2534 use a calling convention that takes care of switching memory banks when
2535 entering and leaving a function. This calling convention is also the
2536 default when using the @option{-mlong-calls} option.
2538 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2539 to call and return from a function.
2541 On 68HC11 the compiler will generate a sequence of instructions
2542 to invoke a board-specific routine to switch the memory bank and call the
2543 real function. The board-specific routine simulates a @code{call}.
2544 At the end of a function, it will jump to a board-specific routine
2545 instead of using @code{rts}. The board-specific return routine simulates
2549 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2550 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2551 use the normal calling convention based on @code{jsr} and @code{rts}.
2552 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2557 You can specify multiple attributes in a declaration by separating them
2558 by commas within the double parentheses or by immediately following an
2559 attribute declaration with another attribute declaration.
2561 @cindex @code{#pragma}, reason for not using
2562 @cindex pragma, reason for not using
2563 Some people object to the @code{__attribute__} feature, suggesting that
2564 ISO C's @code{#pragma} should be used instead. At the time
2565 @code{__attribute__} was designed, there were two reasons for not doing
2570 It is impossible to generate @code{#pragma} commands from a macro.
2573 There is no telling what the same @code{#pragma} might mean in another
2577 These two reasons applied to almost any application that might have been
2578 proposed for @code{#pragma}. It was basically a mistake to use
2579 @code{#pragma} for @emph{anything}.
2581 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2582 to be generated from macros. In addition, a @code{#pragma GCC}
2583 namespace is now in use for GCC-specific pragmas. However, it has been
2584 found convenient to use @code{__attribute__} to achieve a natural
2585 attachment of attributes to their corresponding declarations, whereas
2586 @code{#pragma GCC} is of use for constructs that do not naturally form
2587 part of the grammar. @xref{Other Directives,,Miscellaneous
2588 Preprocessing Directives, cpp, The C Preprocessor}.
2590 @node Attribute Syntax
2591 @section Attribute Syntax
2592 @cindex attribute syntax
2594 This section describes the syntax with which @code{__attribute__} may be
2595 used, and the constructs to which attribute specifiers bind, for the C
2596 language. Some details may vary for C++ and Objective-C@. Because of
2597 infelicities in the grammar for attributes, some forms described here
2598 may not be successfully parsed in all cases.
2600 There are some problems with the semantics of attributes in C++. For
2601 example, there are no manglings for attributes, although they may affect
2602 code generation, so problems may arise when attributed types are used in
2603 conjunction with templates or overloading. Similarly, @code{typeid}
2604 does not distinguish between types with different attributes. Support
2605 for attributes in C++ may be restricted in future to attributes on
2606 declarations only, but not on nested declarators.
2608 @xref{Function Attributes}, for details of the semantics of attributes
2609 applying to functions. @xref{Variable Attributes}, for details of the
2610 semantics of attributes applying to variables. @xref{Type Attributes},
2611 for details of the semantics of attributes applying to structure, union
2612 and enumerated types.
2614 An @dfn{attribute specifier} is of the form
2615 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2616 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2617 each attribute is one of the following:
2621 Empty. Empty attributes are ignored.
2624 A word (which may be an identifier such as @code{unused}, or a reserved
2625 word such as @code{const}).
2628 A word, followed by, in parentheses, parameters for the attribute.
2629 These parameters take one of the following forms:
2633 An identifier. For example, @code{mode} attributes use this form.
2636 An identifier followed by a comma and a non-empty comma-separated list
2637 of expressions. For example, @code{format} attributes use this form.
2640 A possibly empty comma-separated list of expressions. For example,
2641 @code{format_arg} attributes use this form with the list being a single
2642 integer constant expression, and @code{alias} attributes use this form
2643 with the list being a single string constant.
2647 An @dfn{attribute specifier list} is a sequence of one or more attribute
2648 specifiers, not separated by any other tokens.
2650 An attribute specifier list may appear after the colon following a
2651 label, other than a @code{case} or @code{default} label. The only
2652 attribute it makes sense to use after a label is @code{unused}. This
2653 feature is intended for code generated by programs which contains labels
2654 that may be unused but which is compiled with @option{-Wall}. It would
2655 not normally be appropriate to use in it human-written code, though it
2656 could be useful in cases where the code that jumps to the label is
2657 contained within an @code{#ifdef} conditional.
2659 An attribute specifier list may appear as part of a @code{struct},
2660 @code{union} or @code{enum} specifier. It may go either immediately
2661 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2662 the closing brace. It is ignored if the content of the structure, union
2663 or enumerated type is not defined in the specifier in which the
2664 attribute specifier list is used---that is, in usages such as
2665 @code{struct __attribute__((foo)) bar} with no following opening brace.
2666 Where attribute specifiers follow the closing brace, they are considered
2667 to relate to the structure, union or enumerated type defined, not to any
2668 enclosing declaration the type specifier appears in, and the type
2669 defined is not complete until after the attribute specifiers.
2670 @c Otherwise, there would be the following problems: a shift/reduce
2671 @c conflict between attributes binding the struct/union/enum and
2672 @c binding to the list of specifiers/qualifiers; and "aligned"
2673 @c attributes could use sizeof for the structure, but the size could be
2674 @c changed later by "packed" attributes.
2676 Otherwise, an attribute specifier appears as part of a declaration,
2677 counting declarations of unnamed parameters and type names, and relates
2678 to that declaration (which may be nested in another declaration, for
2679 example in the case of a parameter declaration), or to a particular declarator
2680 within a declaration. Where an
2681 attribute specifier is applied to a parameter declared as a function or
2682 an array, it should apply to the function or array rather than the
2683 pointer to which the parameter is implicitly converted, but this is not
2684 yet correctly implemented.
2686 Any list of specifiers and qualifiers at the start of a declaration may
2687 contain attribute specifiers, whether or not such a list may in that
2688 context contain storage class specifiers. (Some attributes, however,
2689 are essentially in the nature of storage class specifiers, and only make
2690 sense where storage class specifiers may be used; for example,
2691 @code{section}.) There is one necessary limitation to this syntax: the
2692 first old-style parameter declaration in a function definition cannot
2693 begin with an attribute specifier, because such an attribute applies to
2694 the function instead by syntax described below (which, however, is not
2695 yet implemented in this case). In some other cases, attribute
2696 specifiers are permitted by this grammar but not yet supported by the
2697 compiler. All attribute specifiers in this place relate to the
2698 declaration as a whole. In the obsolescent usage where a type of
2699 @code{int} is implied by the absence of type specifiers, such a list of
2700 specifiers and qualifiers may be an attribute specifier list with no
2701 other specifiers or qualifiers.
2703 An attribute specifier list may appear immediately before a declarator
2704 (other than the first) in a comma-separated list of declarators in a
2705 declaration of more than one identifier using a single list of
2706 specifiers and qualifiers. Such attribute specifiers apply
2707 only to the identifier before whose declarator they appear. For
2711 __attribute__((noreturn)) void d0 (void),
2712 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2717 the @code{noreturn} attribute applies to all the functions
2718 declared; the @code{format} attribute only applies to @code{d1}.
2720 An attribute specifier list may appear immediately before the comma,
2721 @code{=} or semicolon terminating the declaration of an identifier other
2722 than a function definition. At present, such attribute specifiers apply
2723 to the declared object or function, but in future they may attach to the
2724 outermost adjacent declarator. In simple cases there is no difference,
2725 but, for example, in
2728 void (****f)(void) __attribute__((noreturn));
2732 at present the @code{noreturn} attribute applies to @code{f}, which
2733 causes a warning since @code{f} is not a function, but in future it may
2734 apply to the function @code{****f}. The precise semantics of what
2735 attributes in such cases will apply to are not yet specified. Where an
2736 assembler name for an object or function is specified (@pxref{Asm
2737 Labels}), at present the attribute must follow the @code{asm}
2738 specification; in future, attributes before the @code{asm} specification
2739 may apply to the adjacent declarator, and those after it to the declared
2742 An attribute specifier list may, in future, be permitted to appear after
2743 the declarator in a function definition (before any old-style parameter
2744 declarations or the function body).
2746 Attribute specifiers may be mixed with type qualifiers appearing inside
2747 the @code{[]} of a parameter array declarator, in the C99 construct by
2748 which such qualifiers are applied to the pointer to which the array is
2749 implicitly converted. Such attribute specifiers apply to the pointer,
2750 not to the array, but at present this is not implemented and they are
2753 An attribute specifier list may appear at the start of a nested
2754 declarator. At present, there are some limitations in this usage: the
2755 attributes correctly apply to the declarator, but for most individual
2756 attributes the semantics this implies are not implemented.
2757 When attribute specifiers follow the @code{*} of a pointer
2758 declarator, they may be mixed with any type qualifiers present.
2759 The following describes the formal semantics of this syntax. It will make the
2760 most sense if you are familiar with the formal specification of
2761 declarators in the ISO C standard.
2763 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2764 D1}, where @code{T} contains declaration specifiers that specify a type
2765 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2766 contains an identifier @var{ident}. The type specified for @var{ident}
2767 for derived declarators whose type does not include an attribute
2768 specifier is as in the ISO C standard.
2770 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2771 and the declaration @code{T D} specifies the type
2772 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2773 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2774 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2776 If @code{D1} has the form @code{*
2777 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2778 declaration @code{T D} specifies the type
2779 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2780 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2781 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2787 void (__attribute__((noreturn)) ****f) (void);
2791 specifies the type ``pointer to pointer to pointer to pointer to
2792 non-returning function returning @code{void}''. As another example,
2795 char *__attribute__((aligned(8))) *f;
2799 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2800 Note again that this does not work with most attributes; for example,
2801 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2802 is not yet supported.
2804 For compatibility with existing code written for compiler versions that
2805 did not implement attributes on nested declarators, some laxity is
2806 allowed in the placing of attributes. If an attribute that only applies
2807 to types is applied to a declaration, it will be treated as applying to
2808 the type of that declaration. If an attribute that only applies to
2809 declarations is applied to the type of a declaration, it will be treated
2810 as applying to that declaration; and, for compatibility with code
2811 placing the attributes immediately before the identifier declared, such
2812 an attribute applied to a function return type will be treated as
2813 applying to the function type, and such an attribute applied to an array
2814 element type will be treated as applying to the array type. If an
2815 attribute that only applies to function types is applied to a
2816 pointer-to-function type, it will be treated as applying to the pointer
2817 target type; if such an attribute is applied to a function return type
2818 that is not a pointer-to-function type, it will be treated as applying
2819 to the function type.
2821 @node Function Prototypes
2822 @section Prototypes and Old-Style Function Definitions
2823 @cindex function prototype declarations
2824 @cindex old-style function definitions
2825 @cindex promotion of formal parameters
2827 GNU C extends ISO C to allow a function prototype to override a later
2828 old-style non-prototype definition. Consider the following example:
2831 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2838 /* @r{Prototype function declaration.} */
2839 int isroot P((uid_t));
2841 /* @r{Old-style function definition.} */
2843 isroot (x) /* ??? lossage here ??? */
2850 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2851 not allow this example, because subword arguments in old-style
2852 non-prototype definitions are promoted. Therefore in this example the
2853 function definition's argument is really an @code{int}, which does not
2854 match the prototype argument type of @code{short}.
2856 This restriction of ISO C makes it hard to write code that is portable
2857 to traditional C compilers, because the programmer does not know
2858 whether the @code{uid_t} type is @code{short}, @code{int}, or
2859 @code{long}. Therefore, in cases like these GNU C allows a prototype
2860 to override a later old-style definition. More precisely, in GNU C, a
2861 function prototype argument type overrides the argument type specified
2862 by a later old-style definition if the former type is the same as the
2863 latter type before promotion. Thus in GNU C the above example is
2864 equivalent to the following:
2877 GNU C++ does not support old-style function definitions, so this
2878 extension is irrelevant.
2881 @section C++ Style Comments
2883 @cindex C++ comments
2884 @cindex comments, C++ style
2886 In GNU C, you may use C++ style comments, which start with @samp{//} and
2887 continue until the end of the line. Many other C implementations allow
2888 such comments, and they are included in the 1999 C standard. However,
2889 C++ style comments are not recognized if you specify an @option{-std}
2890 option specifying a version of ISO C before C99, or @option{-ansi}
2891 (equivalent to @option{-std=c89}).
2894 @section Dollar Signs in Identifier Names
2896 @cindex dollar signs in identifier names
2897 @cindex identifier names, dollar signs in
2899 In GNU C, you may normally use dollar signs in identifier names.
2900 This is because many traditional C implementations allow such identifiers.
2901 However, dollar signs in identifiers are not supported on a few target
2902 machines, typically because the target assembler does not allow them.
2904 @node Character Escapes
2905 @section The Character @key{ESC} in Constants
2907 You can use the sequence @samp{\e} in a string or character constant to
2908 stand for the ASCII character @key{ESC}.
2911 @section Inquiring on Alignment of Types or Variables
2913 @cindex type alignment
2914 @cindex variable alignment
2916 The keyword @code{__alignof__} allows you to inquire about how an object
2917 is aligned, or the minimum alignment usually required by a type. Its
2918 syntax is just like @code{sizeof}.
2920 For example, if the target machine requires a @code{double} value to be
2921 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2922 This is true on many RISC machines. On more traditional machine
2923 designs, @code{__alignof__ (double)} is 4 or even 2.
2925 Some machines never actually require alignment; they allow reference to any
2926 data type even at an odd addresses. For these machines, @code{__alignof__}
2927 reports the @emph{recommended} alignment of a type.
2929 If the operand of @code{__alignof__} is an lvalue rather than a type,
2930 its value is the required alignment for its type, taking into account
2931 any minimum alignment specified with GCC's @code{__attribute__}
2932 extension (@pxref{Variable Attributes}). For example, after this
2936 struct foo @{ int x; char y; @} foo1;
2940 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2941 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2943 It is an error to ask for the alignment of an incomplete type.
2945 @node Variable Attributes
2946 @section Specifying Attributes of Variables
2947 @cindex attribute of variables
2948 @cindex variable attributes
2950 The keyword @code{__attribute__} allows you to specify special
2951 attributes of variables or structure fields. This keyword is followed
2952 by an attribute specification inside double parentheses. Some
2953 attributes are currently defined generically for variables.
2954 Other attributes are defined for variables on particular target
2955 systems. Other attributes are available for functions
2956 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2957 Other front ends might define more attributes
2958 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2960 You may also specify attributes with @samp{__} preceding and following
2961 each keyword. This allows you to use them in header files without
2962 being concerned about a possible macro of the same name. For example,
2963 you may use @code{__aligned__} instead of @code{aligned}.
2965 @xref{Attribute Syntax}, for details of the exact syntax for using
2969 @cindex @code{aligned} attribute
2970 @item aligned (@var{alignment})
2971 This attribute specifies a minimum alignment for the variable or
2972 structure field, measured in bytes. For example, the declaration:
2975 int x __attribute__ ((aligned (16))) = 0;
2979 causes the compiler to allocate the global variable @code{x} on a
2980 16-byte boundary. On a 68040, this could be used in conjunction with
2981 an @code{asm} expression to access the @code{move16} instruction which
2982 requires 16-byte aligned operands.
2984 You can also specify the alignment of structure fields. For example, to
2985 create a double-word aligned @code{int} pair, you could write:
2988 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2992 This is an alternative to creating a union with a @code{double} member
2993 that forces the union to be double-word aligned.
2995 As in the preceding examples, you can explicitly specify the alignment
2996 (in bytes) that you wish the compiler to use for a given variable or
2997 structure field. Alternatively, you can leave out the alignment factor
2998 and just ask the compiler to align a variable or field to the maximum
2999 useful alignment for the target machine you are compiling for. For
3000 example, you could write:
3003 short array[3] __attribute__ ((aligned));
3006 Whenever you leave out the alignment factor in an @code{aligned} attribute
3007 specification, the compiler automatically sets the alignment for the declared
3008 variable or field to the largest alignment which is ever used for any data
3009 type on the target machine you are compiling for. Doing this can often make
3010 copy operations more efficient, because the compiler can use whatever
3011 instructions copy the biggest chunks of memory when performing copies to
3012 or from the variables or fields that you have aligned this way.
3014 The @code{aligned} attribute can only increase the alignment; but you
3015 can decrease it by specifying @code{packed} as well. See below.
3017 Note that the effectiveness of @code{aligned} attributes may be limited
3018 by inherent limitations in your linker. On many systems, the linker is
3019 only able to arrange for variables to be aligned up to a certain maximum
3020 alignment. (For some linkers, the maximum supported alignment may
3021 be very very small.) If your linker is only able to align variables
3022 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3023 in an @code{__attribute__} will still only provide you with 8 byte
3024 alignment. See your linker documentation for further information.
3028 @cindex @code{common} attribute
3029 @cindex @code{nocommon} attribute
3032 The @code{common} attribute requests GCC to place a variable in
3033 ``common'' storage. The @code{nocommon} attribute requests the
3034 opposite -- to allocate space for it directly.
3036 These attributes override the default chosen by the
3037 @option{-fno-common} and @option{-fcommon} flags respectively.
3040 @cindex @code{deprecated} attribute
3041 The @code{deprecated} attribute results in a warning if the variable
3042 is used anywhere in the source file. This is useful when identifying
3043 variables that are expected to be removed in a future version of a
3044 program. The warning also includes the location of the declaration
3045 of the deprecated variable, to enable users to easily find further
3046 information about why the variable is deprecated, or what they should
3047 do instead. Note that the warnings only occurs for uses:
3050 extern int old_var __attribute__ ((deprecated));
3052 int new_fn () @{ return old_var; @}
3055 results in a warning on line 3 but not line 2.
3057 The @code{deprecated} attribute can also be used for functions and
3058 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3060 @item mode (@var{mode})
3061 @cindex @code{mode} attribute
3062 This attribute specifies the data type for the declaration---whichever
3063 type corresponds to the mode @var{mode}. This in effect lets you
3064 request an integer or floating point type according to its width.
3066 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3067 indicate the mode corresponding to a one-byte integer, @samp{word} or
3068 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3069 or @samp{__pointer__} for the mode used to represent pointers.
3072 @cindex @code{packed} attribute
3073 The @code{packed} attribute specifies that a variable or structure field
3074 should have the smallest possible alignment---one byte for a variable,
3075 and one bit for a field, unless you specify a larger value with the
3076 @code{aligned} attribute.
3078 Here is a structure in which the field @code{x} is packed, so that it
3079 immediately follows @code{a}:
3085 int x[2] __attribute__ ((packed));
3089 @item section ("@var{section-name}")
3090 @cindex @code{section} variable attribute
3091 Normally, the compiler places the objects it generates in sections like
3092 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3093 or you need certain particular variables to appear in special sections,
3094 for example to map to special hardware. The @code{section}
3095 attribute specifies that a variable (or function) lives in a particular
3096 section. For example, this small program uses several specific section names:
3099 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3100 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3101 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3102 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3106 /* Initialize stack pointer */
3107 init_sp (stack + sizeof (stack));
3109 /* Initialize initialized data */
3110 memcpy (&init_data, &data, &edata - &data);
3112 /* Turn on the serial ports */
3119 Use the @code{section} attribute with an @emph{initialized} definition
3120 of a @emph{global} variable, as shown in the example. GCC issues
3121 a warning and otherwise ignores the @code{section} attribute in
3122 uninitialized variable declarations.
3124 You may only use the @code{section} attribute with a fully initialized
3125 global definition because of the way linkers work. The linker requires
3126 each object be defined once, with the exception that uninitialized
3127 variables tentatively go in the @code{common} (or @code{bss}) section
3128 and can be multiply ``defined''. You can force a variable to be
3129 initialized with the @option{-fno-common} flag or the @code{nocommon}
3132 Some file formats do not support arbitrary sections so the @code{section}
3133 attribute is not available on all platforms.
3134 If you need to map the entire contents of a module to a particular
3135 section, consider using the facilities of the linker instead.
3138 @cindex @code{shared} variable attribute
3139 On Windows NT, in addition to putting variable definitions in a named
3140 section, the section can also be shared among all running copies of an
3141 executable or DLL@. For example, this small program defines shared data
3142 by putting it in a named section @code{shared} and marking the section
3146 int foo __attribute__((section ("shared"), shared)) = 0;
3151 /* Read and write foo. All running
3152 copies see the same value. */
3158 You may only use the @code{shared} attribute along with @code{section}
3159 attribute with a fully initialized global definition because of the way
3160 linkers work. See @code{section} attribute for more information.
3162 The @code{shared} attribute is only available on Windows NT@.
3164 @item tls_model ("@var{tls_model}")
3165 @cindex @code{tls_model} attribute
3166 The @code{tls_model} attribute sets thread-local storage model
3167 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3168 overriding @code{-ftls-model=} command line switch on a per-variable
3170 The @var{tls_model} argument should be one of @code{global-dynamic},
3171 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3173 Not all targets support this attribute.
3175 @item transparent_union
3176 This attribute, attached to a function parameter which is a union, means
3177 that the corresponding argument may have the type of any union member,
3178 but the argument is passed as if its type were that of the first union
3179 member. For more details see @xref{Type Attributes}. You can also use
3180 this attribute on a @code{typedef} for a union data type; then it
3181 applies to all function parameters with that type.
3184 This attribute, attached to a variable, means that the variable is meant
3185 to be possibly unused. GCC will not produce a warning for this
3188 @item vector_size (@var{bytes})
3189 This attribute specifies the vector size for the variable, measured in
3190 bytes. For example, the declaration:
3193 int foo __attribute__ ((vector_size (16)));
3197 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3198 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3199 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3201 This attribute is only applicable to integral and float scalars,
3202 although arrays, pointers, and function return values are allowed in
3203 conjunction with this construct.
3205 Aggregates with this attribute are invalid, even if they are of the same
3206 size as a corresponding scalar. For example, the declaration:
3209 struct S @{ int a; @};
3210 struct S __attribute__ ((vector_size (16))) foo;
3214 is invalid even if the size of the structure is the same as the size of
3218 The @code{weak} attribute is described in @xref{Function Attributes}.
3221 @subsection M32R/D Variable Attributes
3223 One attribute is currently defined for the M32R/D.
3226 @item model (@var{model-name})
3227 @cindex variable addressability on the M32R/D
3228 Use this attribute on the M32R/D to set the addressability of an object.
3229 The identifier @var{model-name} is one of @code{small}, @code{medium},
3230 or @code{large}, representing each of the code models.
3232 Small model objects live in the lower 16MB of memory (so that their
3233 addresses can be loaded with the @code{ld24} instruction).
3235 Medium and large model objects may live anywhere in the 32-bit address space
3236 (the compiler will generate @code{seth/add3} instructions to load their
3240 @subsection i386 Variable Attributes
3242 Two attributes are currently defined for i386 configurations:
3243 @code{ms_struct} and @code{gcc_struct}
3248 @cindex @code{ms_struct} attribute
3249 @cindex @code{gcc_struct} attribute
3251 If @code{packed} is used on a structure, or if bit-fields are used
3252 it may be that the Microsoft ABI packs them differently
3253 than GCC would normally pack them. Particularly when moving packed
3254 data between functions compiled with GCC and the native Microsoft compiler
3255 (either via function call or as data in a file), it may be necessary to access
3258 Currently @option{-m[no-]ms-bitfields} is provided for the Windows X86
3259 compilers to match the native Microsoft compiler.
3262 @node Type Attributes
3263 @section Specifying Attributes of Types
3264 @cindex attribute of types
3265 @cindex type attributes
3267 The keyword @code{__attribute__} allows you to specify special
3268 attributes of @code{struct} and @code{union} types when you define such
3269 types. This keyword is followed by an attribute specification inside
3270 double parentheses. Six attributes are currently defined for types:
3271 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3272 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3273 functions (@pxref{Function Attributes}) and for variables
3274 (@pxref{Variable Attributes}).
3276 You may also specify any one of these attributes with @samp{__}
3277 preceding and following its keyword. This allows you to use these
3278 attributes in header files without being concerned about a possible
3279 macro of the same name. For example, you may use @code{__aligned__}
3280 instead of @code{aligned}.
3282 You may specify the @code{aligned} and @code{transparent_union}
3283 attributes either in a @code{typedef} declaration or just past the
3284 closing curly brace of a complete enum, struct or union type
3285 @emph{definition} and the @code{packed} attribute only past the closing
3286 brace of a definition.
3288 You may also specify attributes between the enum, struct or union
3289 tag and the name of the type rather than after the closing brace.
3291 @xref{Attribute Syntax}, for details of the exact syntax for using
3295 @cindex @code{aligned} attribute
3296 @item aligned (@var{alignment})
3297 This attribute specifies a minimum alignment (in bytes) for variables
3298 of the specified type. For example, the declarations:
3301 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3302 typedef int more_aligned_int __attribute__ ((aligned (8)));
3306 force the compiler to insure (as far as it can) that each variable whose
3307 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3308 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3309 variables of type @code{struct S} aligned to 8-byte boundaries allows
3310 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3311 store) instructions when copying one variable of type @code{struct S} to
3312 another, thus improving run-time efficiency.
3314 Note that the alignment of any given @code{struct} or @code{union} type
3315 is required by the ISO C standard to be at least a perfect multiple of
3316 the lowest common multiple of the alignments of all of the members of
3317 the @code{struct} or @code{union} in question. This means that you @emph{can}
3318 effectively adjust the alignment of a @code{struct} or @code{union}
3319 type by attaching an @code{aligned} attribute to any one of the members
3320 of such a type, but the notation illustrated in the example above is a
3321 more obvious, intuitive, and readable way to request the compiler to
3322 adjust the alignment of an entire @code{struct} or @code{union} type.
3324 As in the preceding example, you can explicitly specify the alignment
3325 (in bytes) that you wish the compiler to use for a given @code{struct}
3326 or @code{union} type. Alternatively, you can leave out the alignment factor
3327 and just ask the compiler to align a type to the maximum
3328 useful alignment for the target machine you are compiling for. For
3329 example, you could write:
3332 struct S @{ short f[3]; @} __attribute__ ((aligned));
3335 Whenever you leave out the alignment factor in an @code{aligned}
3336 attribute specification, the compiler automatically sets the alignment
3337 for the type to the largest alignment which is ever used for any data
3338 type on the target machine you are compiling for. Doing this can often
3339 make copy operations more efficient, because the compiler can use
3340 whatever instructions copy the biggest chunks of memory when performing
3341 copies to or from the variables which have types that you have aligned
3344 In the example above, if the size of each @code{short} is 2 bytes, then
3345 the size of the entire @code{struct S} type is 6 bytes. The smallest
3346 power of two which is greater than or equal to that is 8, so the
3347 compiler sets the alignment for the entire @code{struct S} type to 8
3350 Note that although you can ask the compiler to select a time-efficient
3351 alignment for a given type and then declare only individual stand-alone
3352 objects of that type, the compiler's ability to select a time-efficient
3353 alignment is primarily useful only when you plan to create arrays of
3354 variables having the relevant (efficiently aligned) type. If you
3355 declare or use arrays of variables of an efficiently-aligned type, then
3356 it is likely that your program will also be doing pointer arithmetic (or
3357 subscripting, which amounts to the same thing) on pointers to the
3358 relevant type, and the code that the compiler generates for these
3359 pointer arithmetic operations will often be more efficient for
3360 efficiently-aligned types than for other types.
3362 The @code{aligned} attribute can only increase the alignment; but you
3363 can decrease it by specifying @code{packed} as well. See below.
3365 Note that the effectiveness of @code{aligned} attributes may be limited
3366 by inherent limitations in your linker. On many systems, the linker is
3367 only able to arrange for variables to be aligned up to a certain maximum
3368 alignment. (For some linkers, the maximum supported alignment may
3369 be very very small.) If your linker is only able to align variables
3370 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3371 in an @code{__attribute__} will still only provide you with 8 byte
3372 alignment. See your linker documentation for further information.
3375 This attribute, attached to an @code{enum}, @code{struct}, or
3376 @code{union} type definition, specified that the minimum required memory
3377 be used to represent the type.
3379 @opindex fshort-enums
3380 Specifying this attribute for @code{struct} and @code{union} types is
3381 equivalent to specifying the @code{packed} attribute on each of the
3382 structure or union members. Specifying the @option{-fshort-enums}
3383 flag on the line is equivalent to specifying the @code{packed}
3384 attribute on all @code{enum} definitions.
3386 You may only specify this attribute after a closing curly brace on an
3387 @code{enum} definition, not in a @code{typedef} declaration, unless that
3388 declaration also contains the definition of the @code{enum}.
3390 @item transparent_union
3391 This attribute, attached to a @code{union} type definition, indicates
3392 that any function parameter having that union type causes calls to that
3393 function to be treated in a special way.
3395 First, the argument corresponding to a transparent union type can be of
3396 any type in the union; no cast is required. Also, if the union contains
3397 a pointer type, the corresponding argument can be a null pointer
3398 constant or a void pointer expression; and if the union contains a void
3399 pointer type, the corresponding argument can be any pointer expression.
3400 If the union member type is a pointer, qualifiers like @code{const} on
3401 the referenced type must be respected, just as with normal pointer
3404 Second, the argument is passed to the function using the calling
3405 conventions of first member of the transparent union, not the calling
3406 conventions of the union itself. All members of the union must have the
3407 same machine representation; this is necessary for this argument passing
3410 Transparent unions are designed for library functions that have multiple
3411 interfaces for compatibility reasons. For example, suppose the
3412 @code{wait} function must accept either a value of type @code{int *} to
3413 comply with Posix, or a value of type @code{union wait *} to comply with
3414 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3415 @code{wait} would accept both kinds of arguments, but it would also
3416 accept any other pointer type and this would make argument type checking
3417 less useful. Instead, @code{<sys/wait.h>} might define the interface
3425 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3427 pid_t wait (wait_status_ptr_t);
3430 This interface allows either @code{int *} or @code{union wait *}
3431 arguments to be passed, using the @code{int *} calling convention.
3432 The program can call @code{wait} with arguments of either type:
3435 int w1 () @{ int w; return wait (&w); @}
3436 int w2 () @{ union wait w; return wait (&w); @}
3439 With this interface, @code{wait}'s implementation might look like this:
3442 pid_t wait (wait_status_ptr_t p)
3444 return waitpid (-1, p.__ip, 0);
3449 When attached to a type (including a @code{union} or a @code{struct}),
3450 this attribute means that variables of that type are meant to appear
3451 possibly unused. GCC will not produce a warning for any variables of
3452 that type, even if the variable appears to do nothing. This is often
3453 the case with lock or thread classes, which are usually defined and then
3454 not referenced, but contain constructors and destructors that have
3455 nontrivial bookkeeping functions.
3458 The @code{deprecated} attribute results in a warning if the type
3459 is used anywhere in the source file. This is useful when identifying
3460 types that are expected to be removed in a future version of a program.
3461 If possible, the warning also includes the location of the declaration
3462 of the deprecated type, to enable users to easily find further
3463 information about why the type is deprecated, or what they should do
3464 instead. Note that the warnings only occur for uses and then only
3465 if the type is being applied to an identifier that itself is not being
3466 declared as deprecated.
3469 typedef int T1 __attribute__ ((deprecated));
3473 typedef T1 T3 __attribute__ ((deprecated));
3474 T3 z __attribute__ ((deprecated));
3477 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3478 warning is issued for line 4 because T2 is not explicitly
3479 deprecated. Line 5 has no warning because T3 is explicitly
3480 deprecated. Similarly for line 6.
3482 The @code{deprecated} attribute can also be used for functions and
3483 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3486 Accesses to objects with types with this attribute are not subjected to
3487 type-based alias analysis, but are instead assumed to be able to alias
3488 any other type of objects, just like the @code{char} type. See
3489 @option{-fstrict-aliasing} for more information on aliasing issues.
3494 typedef short __attribute__((__may_alias__)) short_a;
3500 short_a *b = (short_a *) &a;
3504 if (a == 0x12345678)
3511 If you replaced @code{short_a} with @code{short} in the variable
3512 declaration, the above program would abort when compiled with
3513 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3514 above in recent GCC versions.
3516 @subsection i386 Type Attributes
3518 Two attributes are currently defined for i386 configurations:
3519 @code{ms_struct} and @code{gcc_struct}
3523 @cindex @code{ms_struct}
3524 @cindex @code{gcc_struct}
3526 If @code{packed} is used on a structure, or if bit-fields are used
3527 it may be that the Microsoft ABI packs them differently
3528 than GCC would normally pack them. Particularly when moving packed
3529 data between functions compiled with GCC and the native Microsoft compiler
3530 (either via function call or as data in a file), it may be necessary to access
3533 Currently @option{-m[no-]ms-bitfields} is provided for the Windows X86
3534 compilers to match the native Microsoft compiler.
3537 To specify multiple attributes, separate them by commas within the
3538 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3542 @section An Inline Function is As Fast As a Macro
3543 @cindex inline functions
3544 @cindex integrating function code
3546 @cindex macros, inline alternative
3548 By declaring a function @code{inline}, you can direct GCC to
3549 integrate that function's code into the code for its callers. This
3550 makes execution faster by eliminating the function-call overhead; in
3551 addition, if any of the actual argument values are constant, their known
3552 values may permit simplifications at compile time so that not all of the
3553 inline function's code needs to be included. The effect on code size is
3554 less predictable; object code may be larger or smaller with function
3555 inlining, depending on the particular case. Inlining of functions is an
3556 optimization and it really ``works'' only in optimizing compilation. If
3557 you don't use @option{-O}, no function is really inline.
3559 Inline functions are included in the ISO C99 standard, but there are
3560 currently substantial differences between what GCC implements and what
3561 the ISO C99 standard requires.
3563 To declare a function inline, use the @code{inline} keyword in its
3564 declaration, like this:
3574 (If you are writing a header file to be included in ISO C programs, write
3575 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3576 You can also make all ``simple enough'' functions inline with the option
3577 @option{-finline-functions}.
3580 Note that certain usages in a function definition can make it unsuitable
3581 for inline substitution. Among these usages are: use of varargs, use of
3582 alloca, use of variable sized data types (@pxref{Variable Length}),
3583 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3584 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3585 will warn when a function marked @code{inline} could not be substituted,
3586 and will give the reason for the failure.
3588 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3589 does not affect the linkage of the function.
3591 @cindex automatic @code{inline} for C++ member fns
3592 @cindex @code{inline} automatic for C++ member fns
3593 @cindex member fns, automatically @code{inline}
3594 @cindex C++ member fns, automatically @code{inline}
3595 @opindex fno-default-inline
3596 GCC automatically inlines member functions defined within the class
3597 body of C++ programs even if they are not explicitly declared
3598 @code{inline}. (You can override this with @option{-fno-default-inline};
3599 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3601 @cindex inline functions, omission of
3602 @opindex fkeep-inline-functions
3603 When a function is both inline and @code{static}, if all calls to the
3604 function are integrated into the caller, and the function's address is
3605 never used, then the function's own assembler code is never referenced.
3606 In this case, GCC does not actually output assembler code for the
3607 function, unless you specify the option @option{-fkeep-inline-functions}.
3608 Some calls cannot be integrated for various reasons (in particular,
3609 calls that precede the function's definition cannot be integrated, and
3610 neither can recursive calls within the definition). If there is a
3611 nonintegrated call, then the function is compiled to assembler code as
3612 usual. The function must also be compiled as usual if the program
3613 refers to its address, because that can't be inlined.
3615 @cindex non-static inline function
3616 When an inline function is not @code{static}, then the compiler must assume
3617 that there may be calls from other source files; since a global symbol can
3618 be defined only once in any program, the function must not be defined in
3619 the other source files, so the calls therein cannot be integrated.
3620 Therefore, a non-@code{static} inline function is always compiled on its
3621 own in the usual fashion.
3623 If you specify both @code{inline} and @code{extern} in the function
3624 definition, then the definition is used only for inlining. In no case
3625 is the function compiled on its own, not even if you refer to its
3626 address explicitly. Such an address becomes an external reference, as
3627 if you had only declared the function, and had not defined it.
3629 This combination of @code{inline} and @code{extern} has almost the
3630 effect of a macro. The way to use it is to put a function definition in
3631 a header file with these keywords, and put another copy of the
3632 definition (lacking @code{inline} and @code{extern}) in a library file.
3633 The definition in the header file will cause most calls to the function
3634 to be inlined. If any uses of the function remain, they will refer to
3635 the single copy in the library.
3637 For future compatibility with when GCC implements ISO C99 semantics for
3638 inline functions, it is best to use @code{static inline} only. (The
3639 existing semantics will remain available when @option{-std=gnu89} is
3640 specified, but eventually the default will be @option{-std=gnu99} and
3641 that will implement the C99 semantics, though it does not do so yet.)
3643 GCC does not inline any functions when not optimizing unless you specify
3644 the @samp{always_inline} attribute for the function, like this:
3648 inline void foo (const char) __attribute__((always_inline));
3652 @section Assembler Instructions with C Expression Operands
3653 @cindex extended @code{asm}
3654 @cindex @code{asm} expressions
3655 @cindex assembler instructions
3658 In an assembler instruction using @code{asm}, you can specify the
3659 operands of the instruction using C expressions. This means you need not
3660 guess which registers or memory locations will contain the data you want
3663 You must specify an assembler instruction template much like what
3664 appears in a machine description, plus an operand constraint string for
3667 For example, here is how to use the 68881's @code{fsinx} instruction:
3670 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3674 Here @code{angle} is the C expression for the input operand while
3675 @code{result} is that of the output operand. Each has @samp{"f"} as its
3676 operand constraint, saying that a floating point register is required.
3677 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3678 output operands' constraints must use @samp{=}. The constraints use the
3679 same language used in the machine description (@pxref{Constraints}).
3681 Each operand is described by an operand-constraint string followed by
3682 the C expression in parentheses. A colon separates the assembler
3683 template from the first output operand and another separates the last
3684 output operand from the first input, if any. Commas separate the
3685 operands within each group. The total number of operands is currently
3686 limited to 30; this limitation may be lifted in some future version of
3689 If there are no output operands but there are input operands, you must
3690 place two consecutive colons surrounding the place where the output
3693 As of GCC version 3.1, it is also possible to specify input and output
3694 operands using symbolic names which can be referenced within the
3695 assembler code. These names are specified inside square brackets
3696 preceding the constraint string, and can be referenced inside the
3697 assembler code using @code{%[@var{name}]} instead of a percentage sign
3698 followed by the operand number. Using named operands the above example
3702 asm ("fsinx %[angle],%[output]"
3703 : [output] "=f" (result)
3704 : [angle] "f" (angle));
3708 Note that the symbolic operand names have no relation whatsoever to
3709 other C identifiers. You may use any name you like, even those of
3710 existing C symbols, but must ensure that no two operands within the same
3711 assembler construct use the same symbolic name.
3713 Output operand expressions must be lvalues; the compiler can check this.
3714 The input operands need not be lvalues. The compiler cannot check
3715 whether the operands have data types that are reasonable for the
3716 instruction being executed. It does not parse the assembler instruction
3717 template and does not know what it means or even whether it is valid
3718 assembler input. The extended @code{asm} feature is most often used for
3719 machine instructions the compiler itself does not know exist. If
3720 the output expression cannot be directly addressed (for example, it is a
3721 bit-field), your constraint must allow a register. In that case, GCC
3722 will use the register as the output of the @code{asm}, and then store
3723 that register into the output.
3725 The ordinary output operands must be write-only; GCC will assume that
3726 the values in these operands before the instruction are dead and need
3727 not be generated. Extended asm supports input-output or read-write
3728 operands. Use the constraint character @samp{+} to indicate such an
3729 operand and list it with the output operands.
3731 When the constraints for the read-write operand (or the operand in which
3732 only some of the bits are to be changed) allows a register, you may, as
3733 an alternative, logically split its function into two separate operands,
3734 one input operand and one write-only output operand. The connection
3735 between them is expressed by constraints which say they need to be in
3736 the same location when the instruction executes. You can use the same C
3737 expression for both operands, or different expressions. For example,
3738 here we write the (fictitious) @samp{combine} instruction with
3739 @code{bar} as its read-only source operand and @code{foo} as its
3740 read-write destination:
3743 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3747 The constraint @samp{"0"} for operand 1 says that it must occupy the
3748 same location as operand 0. A number in constraint is allowed only in
3749 an input operand and it must refer to an output operand.
3751 Only a number in the constraint can guarantee that one operand will be in
3752 the same place as another. The mere fact that @code{foo} is the value
3753 of both operands is not enough to guarantee that they will be in the
3754 same place in the generated assembler code. The following would not
3758 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3761 Various optimizations or reloading could cause operands 0 and 1 to be in
3762 different registers; GCC knows no reason not to do so. For example, the
3763 compiler might find a copy of the value of @code{foo} in one register and
3764 use it for operand 1, but generate the output operand 0 in a different
3765 register (copying it afterward to @code{foo}'s own address). Of course,
3766 since the register for operand 1 is not even mentioned in the assembler
3767 code, the result will not work, but GCC can't tell that.
3769 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3770 the operand number for a matching constraint. For example:
3773 asm ("cmoveq %1,%2,%[result]"
3774 : [result] "=r"(result)
3775 : "r" (test), "r"(new), "[result]"(old));
3778 Some instructions clobber specific hard registers. To describe this,
3779 write a third colon after the input operands, followed by the names of
3780 the clobbered hard registers (given as strings). Here is a realistic
3781 example for the VAX:
3784 asm volatile ("movc3 %0,%1,%2"
3786 : "g" (from), "g" (to), "g" (count)
3787 : "r0", "r1", "r2", "r3", "r4", "r5");
3790 You may not write a clobber description in a way that overlaps with an
3791 input or output operand. For example, you may not have an operand
3792 describing a register class with one member if you mention that register
3793 in the clobber list. Variables declared to live in specific registers
3794 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3795 have no part mentioned in the clobber description.
3796 There is no way for you to specify that an input
3797 operand is modified without also specifying it as an output
3798 operand. Note that if all the output operands you specify are for this
3799 purpose (and hence unused), you will then also need to specify
3800 @code{volatile} for the @code{asm} construct, as described below, to
3801 prevent GCC from deleting the @code{asm} statement as unused.
3803 If you refer to a particular hardware register from the assembler code,
3804 you will probably have to list the register after the third colon to
3805 tell the compiler the register's value is modified. In some assemblers,
3806 the register names begin with @samp{%}; to produce one @samp{%} in the
3807 assembler code, you must write @samp{%%} in the input.
3809 If your assembler instruction can alter the condition code register, add
3810 @samp{cc} to the list of clobbered registers. GCC on some machines
3811 represents the condition codes as a specific hardware register;
3812 @samp{cc} serves to name this register. On other machines, the
3813 condition code is handled differently, and specifying @samp{cc} has no
3814 effect. But it is valid no matter what the machine.
3816 If your assembler instruction modifies memory in an unpredictable
3817 fashion, add @samp{memory} to the list of clobbered registers. This
3818 will cause GCC to not keep memory values cached in registers across
3819 the assembler instruction. You will also want to add the
3820 @code{volatile} keyword if the memory affected is not listed in the
3821 inputs or outputs of the @code{asm}, as the @samp{memory} clobber does
3822 not count as a side-effect of the @code{asm}.
3824 You can put multiple assembler instructions together in a single
3825 @code{asm} template, separated by the characters normally used in assembly
3826 code for the system. A combination that works in most places is a newline
3827 to break the line, plus a tab character to move to the instruction field
3828 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3829 assembler allows semicolons as a line-breaking character. Note that some
3830 assembler dialects use semicolons to start a comment.
3831 The input operands are guaranteed not to use any of the clobbered
3832 registers, and neither will the output operands' addresses, so you can
3833 read and write the clobbered registers as many times as you like. Here
3834 is an example of multiple instructions in a template; it assumes the
3835 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3838 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3840 : "g" (from), "g" (to)
3844 Unless an output operand has the @samp{&} constraint modifier, GCC
3845 may allocate it in the same register as an unrelated input operand, on
3846 the assumption the inputs are consumed before the outputs are produced.
3847 This assumption may be false if the assembler code actually consists of
3848 more than one instruction. In such a case, use @samp{&} for each output
3849 operand that may not overlap an input. @xref{Modifiers}.
3851 If you want to test the condition code produced by an assembler
3852 instruction, you must include a branch and a label in the @code{asm}
3853 construct, as follows:
3856 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3862 This assumes your assembler supports local labels, as the GNU assembler
3863 and most Unix assemblers do.
3865 Speaking of labels, jumps from one @code{asm} to another are not
3866 supported. The compiler's optimizers do not know about these jumps, and
3867 therefore they cannot take account of them when deciding how to
3870 @cindex macros containing @code{asm}
3871 Usually the most convenient way to use these @code{asm} instructions is to
3872 encapsulate them in macros that look like functions. For example,
3876 (@{ double __value, __arg = (x); \
3877 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
3882 Here the variable @code{__arg} is used to make sure that the instruction
3883 operates on a proper @code{double} value, and to accept only those
3884 arguments @code{x} which can convert automatically to a @code{double}.
3886 Another way to make sure the instruction operates on the correct data
3887 type is to use a cast in the @code{asm}. This is different from using a
3888 variable @code{__arg} in that it converts more different types. For
3889 example, if the desired type were @code{int}, casting the argument to
3890 @code{int} would accept a pointer with no complaint, while assigning the
3891 argument to an @code{int} variable named @code{__arg} would warn about
3892 using a pointer unless the caller explicitly casts it.
3894 If an @code{asm} has output operands, GCC assumes for optimization
3895 purposes the instruction has no side effects except to change the output
3896 operands. This does not mean instructions with a side effect cannot be
3897 used, but you must be careful, because the compiler may eliminate them
3898 if the output operands aren't used, or move them out of loops, or
3899 replace two with one if they constitute a common subexpression. Also,
3900 if your instruction does have a side effect on a variable that otherwise
3901 appears not to change, the old value of the variable may be reused later
3902 if it happens to be found in a register.
3904 You can prevent an @code{asm} instruction from being deleted, moved
3905 significantly, or combined, by writing the keyword @code{volatile} after
3906 the @code{asm}. For example:
3909 #define get_and_set_priority(new) \
3911 asm volatile ("get_and_set_priority %0, %1" \
3912 : "=g" (__old) : "g" (new)); \
3917 If you write an @code{asm} instruction with no outputs, GCC will know
3918 the instruction has side-effects and will not delete the instruction or
3919 move it outside of loops.
3921 The @code{volatile} keyword indicates that the instruction has
3922 important side-effects. GCC will not delete a volatile @code{asm} if
3923 it is reachable. (The instruction can still be deleted if GCC can
3924 prove that control-flow will never reach the location of the
3925 instruction.) In addition, GCC will not reschedule instructions
3926 across a volatile @code{asm} instruction. For example:
3929 *(volatile int *)addr = foo;
3930 asm volatile ("eieio" : : );
3934 Assume @code{addr} contains the address of a memory mapped device
3935 register. The PowerPC @code{eieio} instruction (Enforce In-order
3936 Execution of I/O) tells the CPU to make sure that the store to that
3937 device register happens before it issues any other I/O@.
3939 Note that even a volatile @code{asm} instruction can be moved in ways
3940 that appear insignificant to the compiler, such as across jump
3941 instructions. You can't expect a sequence of volatile @code{asm}
3942 instructions to remain perfectly consecutive. If you want consecutive
3943 output, use a single @code{asm}. Also, GCC will perform some
3944 optimizations across a volatile @code{asm} instruction; GCC does not
3945 ``forget everything'' when it encounters a volatile @code{asm}
3946 instruction the way some other compilers do.
3948 An @code{asm} instruction without any operands or clobbers (an ``old
3949 style'' @code{asm}) will be treated identically to a volatile
3950 @code{asm} instruction.
3952 It is a natural idea to look for a way to give access to the condition
3953 code left by the assembler instruction. However, when we attempted to
3954 implement this, we found no way to make it work reliably. The problem
3955 is that output operands might need reloading, which would result in
3956 additional following ``store'' instructions. On most machines, these
3957 instructions would alter the condition code before there was time to
3958 test it. This problem doesn't arise for ordinary ``test'' and
3959 ``compare'' instructions because they don't have any output operands.
3961 For reasons similar to those described above, it is not possible to give
3962 an assembler instruction access to the condition code left by previous
3965 If you are writing a header file that should be includable in ISO C
3966 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
3969 @subsection i386 floating point asm operands
3971 There are several rules on the usage of stack-like regs in
3972 asm_operands insns. These rules apply only to the operands that are
3977 Given a set of input regs that die in an asm_operands, it is
3978 necessary to know which are implicitly popped by the asm, and
3979 which must be explicitly popped by gcc.
3981 An input reg that is implicitly popped by the asm must be
3982 explicitly clobbered, unless it is constrained to match an
3986 For any input reg that is implicitly popped by an asm, it is
3987 necessary to know how to adjust the stack to compensate for the pop.
3988 If any non-popped input is closer to the top of the reg-stack than
3989 the implicitly popped reg, it would not be possible to know what the
3990 stack looked like---it's not clear how the rest of the stack ``slides
3993 All implicitly popped input regs must be closer to the top of
3994 the reg-stack than any input that is not implicitly popped.
3996 It is possible that if an input dies in an insn, reload might
3997 use the input reg for an output reload. Consider this example:
4000 asm ("foo" : "=t" (a) : "f" (b));
4003 This asm says that input B is not popped by the asm, and that
4004 the asm pushes a result onto the reg-stack, i.e., the stack is one
4005 deeper after the asm than it was before. But, it is possible that
4006 reload will think that it can use the same reg for both the input and
4007 the output, if input B dies in this insn.
4009 If any input operand uses the @code{f} constraint, all output reg
4010 constraints must use the @code{&} earlyclobber.
4012 The asm above would be written as
4015 asm ("foo" : "=&t" (a) : "f" (b));
4019 Some operands need to be in particular places on the stack. All
4020 output operands fall in this category---there is no other way to
4021 know which regs the outputs appear in unless the user indicates
4022 this in the constraints.
4024 Output operands must specifically indicate which reg an output
4025 appears in after an asm. @code{=f} is not allowed: the operand
4026 constraints must select a class with a single reg.
4029 Output operands may not be ``inserted'' between existing stack regs.
4030 Since no 387 opcode uses a read/write operand, all output operands
4031 are dead before the asm_operands, and are pushed by the asm_operands.
4032 It makes no sense to push anywhere but the top of the reg-stack.
4034 Output operands must start at the top of the reg-stack: output
4035 operands may not ``skip'' a reg.
4038 Some asm statements may need extra stack space for internal
4039 calculations. This can be guaranteed by clobbering stack registers
4040 unrelated to the inputs and outputs.
4044 Here are a couple of reasonable asms to want to write. This asm
4045 takes one input, which is internally popped, and produces two outputs.
4048 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4051 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4052 and replaces them with one output. The user must code the @code{st(1)}
4053 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4056 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4062 @section Controlling Names Used in Assembler Code
4063 @cindex assembler names for identifiers
4064 @cindex names used in assembler code
4065 @cindex identifiers, names in assembler code
4067 You can specify the name to be used in the assembler code for a C
4068 function or variable by writing the @code{asm} (or @code{__asm__})
4069 keyword after the declarator as follows:
4072 int foo asm ("myfoo") = 2;
4076 This specifies that the name to be used for the variable @code{foo} in
4077 the assembler code should be @samp{myfoo} rather than the usual
4080 On systems where an underscore is normally prepended to the name of a C
4081 function or variable, this feature allows you to define names for the
4082 linker that do not start with an underscore.
4084 It does not make sense to use this feature with a non-static local
4085 variable since such variables do not have assembler names. If you are
4086 trying to put the variable in a particular register, see @ref{Explicit
4087 Reg Vars}. GCC presently accepts such code with a warning, but will
4088 probably be changed to issue an error, rather than a warning, in the
4091 You cannot use @code{asm} in this way in a function @emph{definition}; but
4092 you can get the same effect by writing a declaration for the function
4093 before its definition and putting @code{asm} there, like this:
4096 extern func () asm ("FUNC");
4103 It is up to you to make sure that the assembler names you choose do not
4104 conflict with any other assembler symbols. Also, you must not use a
4105 register name; that would produce completely invalid assembler code. GCC
4106 does not as yet have the ability to store static variables in registers.
4107 Perhaps that will be added.
4109 @node Explicit Reg Vars
4110 @section Variables in Specified Registers
4111 @cindex explicit register variables
4112 @cindex variables in specified registers
4113 @cindex specified registers
4114 @cindex registers, global allocation
4116 GNU C allows you to put a few global variables into specified hardware
4117 registers. You can also specify the register in which an ordinary
4118 register variable should be allocated.
4122 Global register variables reserve registers throughout the program.
4123 This may be useful in programs such as programming language
4124 interpreters which have a couple of global variables that are accessed
4128 Local register variables in specific registers do not reserve the
4129 registers. The compiler's data flow analysis is capable of determining
4130 where the specified registers contain live values, and where they are
4131 available for other uses. Stores into local register variables may be deleted
4132 when they appear to be dead according to dataflow analysis. References
4133 to local register variables may be deleted or moved or simplified.
4135 These local variables are sometimes convenient for use with the extended
4136 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4137 output of the assembler instruction directly into a particular register.
4138 (This will work provided the register you specify fits the constraints
4139 specified for that operand in the @code{asm}.)
4147 @node Global Reg Vars
4148 @subsection Defining Global Register Variables
4149 @cindex global register variables
4150 @cindex registers, global variables in
4152 You can define a global register variable in GNU C like this:
4155 register int *foo asm ("a5");
4159 Here @code{a5} is the name of the register which should be used. Choose a
4160 register which is normally saved and restored by function calls on your
4161 machine, so that library routines will not clobber it.
4163 Naturally the register name is cpu-dependent, so you would need to
4164 conditionalize your program according to cpu type. The register
4165 @code{a5} would be a good choice on a 68000 for a variable of pointer
4166 type. On machines with register windows, be sure to choose a ``global''
4167 register that is not affected magically by the function call mechanism.
4169 In addition, operating systems on one type of cpu may differ in how they
4170 name the registers; then you would need additional conditionals. For
4171 example, some 68000 operating systems call this register @code{%a5}.
4173 Eventually there may be a way of asking the compiler to choose a register
4174 automatically, but first we need to figure out how it should choose and
4175 how to enable you to guide the choice. No solution is evident.
4177 Defining a global register variable in a certain register reserves that
4178 register entirely for this use, at least within the current compilation.
4179 The register will not be allocated for any other purpose in the functions
4180 in the current compilation. The register will not be saved and restored by
4181 these functions. Stores into this register are never deleted even if they
4182 would appear to be dead, but references may be deleted or moved or
4185 It is not safe to access the global register variables from signal
4186 handlers, or from more than one thread of control, because the system
4187 library routines may temporarily use the register for other things (unless
4188 you recompile them specially for the task at hand).
4190 @cindex @code{qsort}, and global register variables
4191 It is not safe for one function that uses a global register variable to
4192 call another such function @code{foo} by way of a third function
4193 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4194 different source file in which the variable wasn't declared). This is
4195 because @code{lose} might save the register and put some other value there.
4196 For example, you can't expect a global register variable to be available in
4197 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4198 might have put something else in that register. (If you are prepared to
4199 recompile @code{qsort} with the same global register variable, you can
4200 solve this problem.)
4202 If you want to recompile @code{qsort} or other source files which do not
4203 actually use your global register variable, so that they will not use that
4204 register for any other purpose, then it suffices to specify the compiler
4205 option @option{-ffixed-@var{reg}}. You need not actually add a global
4206 register declaration to their source code.
4208 A function which can alter the value of a global register variable cannot
4209 safely be called from a function compiled without this variable, because it
4210 could clobber the value the caller expects to find there on return.
4211 Therefore, the function which is the entry point into the part of the
4212 program that uses the global register variable must explicitly save and
4213 restore the value which belongs to its caller.
4215 @cindex register variable after @code{longjmp}
4216 @cindex global register after @code{longjmp}
4217 @cindex value after @code{longjmp}
4220 On most machines, @code{longjmp} will restore to each global register
4221 variable the value it had at the time of the @code{setjmp}. On some
4222 machines, however, @code{longjmp} will not change the value of global
4223 register variables. To be portable, the function that called @code{setjmp}
4224 should make other arrangements to save the values of the global register
4225 variables, and to restore them in a @code{longjmp}. This way, the same
4226 thing will happen regardless of what @code{longjmp} does.
4228 All global register variable declarations must precede all function
4229 definitions. If such a declaration could appear after function
4230 definitions, the declaration would be too late to prevent the register from
4231 being used for other purposes in the preceding functions.
4233 Global register variables may not have initial values, because an
4234 executable file has no means to supply initial contents for a register.
4236 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4237 registers, but certain library functions, such as @code{getwd}, as well
4238 as the subroutines for division and remainder, modify g3 and g4. g1 and
4239 g2 are local temporaries.
4241 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4242 Of course, it will not do to use more than a few of those.
4244 @node Local Reg Vars
4245 @subsection Specifying Registers for Local Variables
4246 @cindex local variables, specifying registers
4247 @cindex specifying registers for local variables
4248 @cindex registers for local variables
4250 You can define a local register variable with a specified register
4254 register int *foo asm ("a5");
4258 Here @code{a5} is the name of the register which should be used. Note
4259 that this is the same syntax used for defining global register
4260 variables, but for a local variable it would appear within a function.
4262 Naturally the register name is cpu-dependent, but this is not a
4263 problem, since specific registers are most often useful with explicit
4264 assembler instructions (@pxref{Extended Asm}). Both of these things
4265 generally require that you conditionalize your program according to
4268 In addition, operating systems on one type of cpu may differ in how they
4269 name the registers; then you would need additional conditionals. For
4270 example, some 68000 operating systems call this register @code{%a5}.
4272 Defining such a register variable does not reserve the register; it
4273 remains available for other uses in places where flow control determines
4274 the variable's value is not live. However, these registers are made
4275 unavailable for use in the reload pass; excessive use of this feature
4276 leaves the compiler too few available registers to compile certain
4279 This option does not guarantee that GCC will generate code that has
4280 this variable in the register you specify at all times. You may not
4281 code an explicit reference to this register in an @code{asm} statement
4282 and assume it will always refer to this variable.
4284 Stores into local register variables may be deleted when they appear to be dead
4285 according to dataflow analysis. References to local register variables may
4286 be deleted or moved or simplified.
4288 @node Alternate Keywords
4289 @section Alternate Keywords
4290 @cindex alternate keywords
4291 @cindex keywords, alternate
4293 @option{-ansi} and the various @option{-std} options disable certain
4294 keywords. This causes trouble when you want to use GNU C extensions, or
4295 a general-purpose header file that should be usable by all programs,
4296 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4297 @code{inline} are not available in programs compiled with
4298 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4299 program compiled with @option{-std=c99}). The ISO C99 keyword
4300 @code{restrict} is only available when @option{-std=gnu99} (which will
4301 eventually be the default) or @option{-std=c99} (or the equivalent
4302 @option{-std=iso9899:1999}) is used.
4304 The way to solve these problems is to put @samp{__} at the beginning and
4305 end of each problematical keyword. For example, use @code{__asm__}
4306 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4308 Other C compilers won't accept these alternative keywords; if you want to
4309 compile with another compiler, you can define the alternate keywords as
4310 macros to replace them with the customary keywords. It looks like this:
4318 @findex __extension__
4320 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4322 prevent such warnings within one expression by writing
4323 @code{__extension__} before the expression. @code{__extension__} has no
4324 effect aside from this.
4326 @node Incomplete Enums
4327 @section Incomplete @code{enum} Types
4329 You can define an @code{enum} tag without specifying its possible values.
4330 This results in an incomplete type, much like what you get if you write
4331 @code{struct foo} without describing the elements. A later declaration
4332 which does specify the possible values completes the type.
4334 You can't allocate variables or storage using the type while it is
4335 incomplete. However, you can work with pointers to that type.
4337 This extension may not be very useful, but it makes the handling of
4338 @code{enum} more consistent with the way @code{struct} and @code{union}
4341 This extension is not supported by GNU C++.
4343 @node Function Names
4344 @section Function Names as Strings
4345 @cindex @code{__FUNCTION__} identifier
4346 @cindex @code{__PRETTY_FUNCTION__} identifier
4347 @cindex @code{__func__} identifier
4349 GCC predefines two magic identifiers to hold the name of the current
4350 function. The identifier @code{__FUNCTION__} holds the name of the function
4351 as it appears in the source. The identifier @code{__PRETTY_FUNCTION__}
4352 holds the name of the function pretty printed in a language specific
4355 These names are always the same in a C function, but in a C++ function
4356 they may be different. For example, this program:
4360 extern int printf (char *, ...);
4367 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4368 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4386 __PRETTY_FUNCTION__ = int a::sub (int)
4389 The compiler automagically replaces the identifiers with a string
4390 literal containing the appropriate name. Thus, they are neither
4391 preprocessor macros, like @code{__FILE__} and @code{__LINE__}, nor
4392 variables. This means that they catenate with other string literals, and
4393 that they can be used to initialize char arrays. For example
4396 char here[] = "Function " __FUNCTION__ " in " __FILE__;
4399 On the other hand, @samp{#ifdef __FUNCTION__} does not have any special
4400 meaning inside a function, since the preprocessor does not do anything
4401 special with the identifier @code{__FUNCTION__}.
4403 Note that these semantics are deprecated, and that GCC 3.2 will handle
4404 @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} the same way as
4405 @code{__func__}. @code{__func__} is defined by the ISO standard C99:
4408 The identifier @code{__func__} is implicitly declared by the translator
4409 as if, immediately following the opening brace of each function
4410 definition, the declaration
4413 static const char __func__[] = "function-name";
4416 appeared, where function-name is the name of the lexically-enclosing
4417 function. This name is the unadorned name of the function.
4420 By this definition, @code{__func__} is a variable, not a string literal.
4421 In particular, @code{__func__} does not catenate with other string
4424 In @code{C++}, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} are
4425 variables, declared in the same way as @code{__func__}.
4427 @node Return Address
4428 @section Getting the Return or Frame Address of a Function
4430 These functions may be used to get information about the callers of a
4433 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4434 This function returns the return address of the current function, or of
4435 one of its callers. The @var{level} argument is number of frames to
4436 scan up the call stack. A value of @code{0} yields the return address
4437 of the current function, a value of @code{1} yields the return address
4438 of the caller of the current function, and so forth. When inlining
4439 the expected behavior is that the function will return the address of
4440 the function that will be returned to. To work around this behavior use
4441 the @code{noinline} function attribute.
4443 The @var{level} argument must be a constant integer.
4445 On some machines it may be impossible to determine the return address of
4446 any function other than the current one; in such cases, or when the top
4447 of the stack has been reached, this function will return @code{0} or a
4448 random value. In addition, @code{__builtin_frame_address} may be used
4449 to determine if the top of the stack has been reached.
4451 This function should only be used with a nonzero argument for debugging
4455 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4456 This function is similar to @code{__builtin_return_address}, but it
4457 returns the address of the function frame rather than the return address
4458 of the function. Calling @code{__builtin_frame_address} with a value of
4459 @code{0} yields the frame address of the current function, a value of
4460 @code{1} yields the frame address of the caller of the current function,
4463 The frame is the area on the stack which holds local variables and saved
4464 registers. The frame address is normally the address of the first word
4465 pushed on to the stack by the function. However, the exact definition
4466 depends upon the processor and the calling convention. If the processor
4467 has a dedicated frame pointer register, and the function has a frame,
4468 then @code{__builtin_frame_address} will return the value of the frame
4471 On some machines it may be impossible to determine the frame address of
4472 any function other than the current one; in such cases, or when the top
4473 of the stack has been reached, this function will return @code{0} if
4474 the first frame pointer is properly initialized by the startup code.
4476 This function should only be used with a nonzero argument for debugging
4480 @node Vector Extensions
4481 @section Using vector instructions through built-in functions
4483 On some targets, the instruction set contains SIMD vector instructions that
4484 operate on multiple values contained in one large register at the same time.
4485 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4488 The first step in using these extensions is to provide the necessary data
4489 types. This should be done using an appropriate @code{typedef}:
4492 typedef int v4si __attribute__ ((mode(V4SI)));
4495 The base type @code{int} is effectively ignored by the compiler, the
4496 actual properties of the new type @code{v4si} are defined by the
4497 @code{__attribute__}. It defines the machine mode to be used; for vector
4498 types these have the form @code{V@var{n}@var{B}}; @var{n} should be the
4499 number of elements in the vector, and @var{B} should be the base mode of the
4500 individual elements. The following can be used as base modes:
4504 An integer that is as wide as the smallest addressable unit, usually 8 bits.
4506 An integer, twice as wide as a QI mode integer, usually 16 bits.
4508 An integer, four times as wide as a QI mode integer, usually 32 bits.
4510 An integer, eight times as wide as a QI mode integer, usually 64 bits.
4512 A floating point value, as wide as a SI mode integer, usually 32 bits.
4514 A floating point value, as wide as a DI mode integer, usually 64 bits.
4517 Specifying a combination that is not valid for the current architecture
4518 will cause gcc to synthesize the instructions using a narrower mode.
4519 For example, if you specify a variable of type @code{V4SI} and your
4520 architecture does not allow for this specific SIMD type, gcc will
4521 produce code that uses 4 @code{SIs}.
4523 The types defined in this manner can be used with a subset of normal C
4524 operations. Currently, gcc will allow using the following operators on
4525 these types: @code{+, -, *, /, unary minus}@.
4527 The operations behave like C++ @code{valarrays}. Addition is defined as
4528 the addition of the corresponding elements of the operands. For
4529 example, in the code below, each of the 4 elements in @var{a} will be
4530 added to the corresponding 4 elements in @var{b} and the resulting
4531 vector will be stored in @var{c}.
4534 typedef int v4si __attribute__ ((mode(V4SI)));
4541 Subtraction, multiplication, and division operate in a similar manner.
4542 Likewise, the result of using the unary minus operator on a vector type
4543 is a vector whose elements are the negative value of the corresponding
4544 elements in the operand.
4546 You can declare variables and use them in function calls and returns, as
4547 well as in assignments and some casts. You can specify a vector type as
4548 a return type for a function. Vector types can also be used as function
4549 arguments. It is possible to cast from one vector type to another,
4550 provided they are of the same size (in fact, you can also cast vectors
4551 to and from other datatypes of the same size).
4553 You cannot operate between vectors of different lengths or different
4554 signedness without a cast.
4556 A port that supports hardware vector operations, usually provides a set
4557 of built-in functions that can be used to operate on vectors. For
4558 example, a function to add two vectors and multiply the result by a
4559 third could look like this:
4562 v4si f (v4si a, v4si b, v4si c)
4564 v4si tmp = __builtin_addv4si (a, b);
4565 return __builtin_mulv4si (tmp, c);
4570 @node Other Builtins
4571 @section Other built-in functions provided by GCC
4572 @cindex built-in functions
4573 @findex __builtin_isgreater
4574 @findex __builtin_isgreaterequal
4575 @findex __builtin_isless
4576 @findex __builtin_islessequal
4577 @findex __builtin_islessgreater
4578 @findex __builtin_isunordered
4620 @findex fprintf_unlocked
4622 @findex fputs_unlocked
4642 @findex printf_unlocked
4683 GCC provides a large number of built-in functions other than the ones
4684 mentioned above. Some of these are for internal use in the processing
4685 of exceptions or variable-length argument lists and will not be
4686 documented here because they may change from time to time; we do not
4687 recommend general use of these functions.
4689 The remaining functions are provided for optimization purposes.
4691 @opindex fno-builtin
4692 GCC includes built-in versions of many of the functions in the standard
4693 C library. The versions prefixed with @code{__builtin_} will always be
4694 treated as having the same meaning as the C library function even if you
4695 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
4696 Many of these functions are only optimized in certain cases; if they are
4697 not optimized in a particular case, a call to the library function will
4702 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
4703 @option{-std=c99}), the functions @code{alloca}, @code{bcmp},
4704 @code{bzero}, @code{_exit}, @code{ffs}, @code{fprintf_unlocked},
4705 @code{fputs_unlocked}, @code{index}, @code{mempcpy}, @code{printf_unlocked},
4706 @code{rindex}, @code{stpcpy} and @code{strdup}
4707 may be handled as built-in functions.
4708 All these functions have corresponding versions
4709 prefixed with @code{__builtin_}, which may be used even in strict C89
4712 The ISO C99 functions @code{conj}, @code{conjf}, @code{conjl}, @code{creal},
4713 @code{crealf}, @code{creall}, @code{cimag}, @code{cimagf}, @code{cimagl},
4714 @code{_Exit}, @code{imaxabs}, @code{llabs},
4715 @code{nearbyint}, @code{nearbyintf}, @code{nearbyintl},
4716 @code{round}, @code{roundf}, @code{roundl}, @code{snprintf},
4717 @code{trunc}, @code{truncf}, @code{truncl},
4718 @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
4719 are handled as built-in functions
4720 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
4722 There are also built-in versions of the ISO C99 functions @code{atan2f},
4723 @code{atan2l}, @code{ceilf}, @code{ceill}, @code{cosf}, @code{cosl},
4724 @code{expf}, @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf},
4725 @code{floorl}, @code{fmodf}, @code{fmodl},
4726 @code{logf}, @code{logl}, @code{powf}, @code{powl},
4727 @code{sinf}, @code{sinl}, @code{sqrtf} and @code{sqrtl}
4728 that are recognized in any mode since ISO C90 reserves these names for
4729 the purpose to which ISO C99 puts them. All these functions have
4730 corresponding versions prefixed with @code{__builtin_}.
4732 The ISO C90 functions @code{abort}, @code{abs}, @code{atan2},
4733 @code{calloc}, @code{ceil}, @code{cos}, @code{exit},
4734 @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
4735 @code{fprintf}, @code{fputs}, @code{labs}, @code{log}, @code{malloc},
4736 @code{memcmp}, @code{memcpy}, @code{memset}, @code{pow}, @code{printf},
4737 @code{putchar}, @code{puts}, @code{scanf}, @code{sin}, @code{snprintf},
4738 @code{sprintf}, @code{sqrt}, @code{sscanf},
4739 @code{strcat}, @code{strchr}, @code{strcmp},
4740 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat}, @code{strncmp},
4741 @code{strncpy}, @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
4742 @code{vprintf} and @code{vsprintf}
4743 are all recognized as built-in functions unless
4744 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
4745 is specified for an individual function). All of these functions have
4746 corresponding versions prefixed with @code{__builtin_}.
4748 GCC provides built-in versions of the ISO C99 floating point comparison
4749 macros that avoid raising exceptions for unordered operands. They have
4750 the same names as the standard macros ( @code{isgreater},
4751 @code{isgreaterequal}, @code{isless}, @code{islessequal},
4752 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
4753 prefixed. We intend for a library implementor to be able to simply
4754 @code{#define} each standard macro to its built-in equivalent.
4756 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
4758 You can use the built-in function @code{__builtin_types_compatible_p} to
4759 determine whether two types are the same.
4761 This built-in function returns 1 if the unqualified versions of the
4762 types @var{type1} and @var{type2} (which are types, not expressions) are
4763 compatible, 0 otherwise. The result of this built-in function can be
4764 used in integer constant expressions.
4766 This built-in function ignores top level qualifiers (e.g., @code{const},
4767 @code{volatile}). For example, @code{int} is equivalent to @code{const
4770 The type @code{int[]} and @code{int[5]} are compatible. On the other
4771 hand, @code{int} and @code{char *} are not compatible, even if the size
4772 of their types, on the particular architecture are the same. Also, the
4773 amount of pointer indirection is taken into account when determining
4774 similarity. Consequently, @code{short *} is not similar to
4775 @code{short **}. Furthermore, two types that are typedefed are
4776 considered compatible if their underlying types are compatible.
4778 An @code{enum} type is considered to be compatible with another
4779 @code{enum} type. For example, @code{enum @{foo, bar@}} is similar to
4780 @code{enum @{hot, dog@}}.
4782 You would typically use this function in code whose execution varies
4783 depending on the arguments' types. For example:
4789 if (__builtin_types_compatible_p (typeof (x), long double)) \
4790 tmp = foo_long_double (tmp); \
4791 else if (__builtin_types_compatible_p (typeof (x), double)) \
4792 tmp = foo_double (tmp); \
4793 else if (__builtin_types_compatible_p (typeof (x), float)) \
4794 tmp = foo_float (tmp); \
4801 @emph{Note:} This construct is only available for C.
4805 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
4807 You can use the built-in function @code{__builtin_choose_expr} to
4808 evaluate code depending on the value of a constant expression. This
4809 built-in function returns @var{exp1} if @var{const_exp}, which is a
4810 constant expression that must be able to be determined at compile time,
4811 is nonzero. Otherwise it returns 0.
4813 This built-in function is analogous to the @samp{? :} operator in C,
4814 except that the expression returned has its type unaltered by promotion
4815 rules. Also, the built-in function does not evaluate the expression
4816 that was not chosen. For example, if @var{const_exp} evaluates to true,
4817 @var{exp2} is not evaluated even if it has side-effects.
4819 This built-in function can return an lvalue if the chosen argument is an
4822 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
4823 type. Similarly, if @var{exp2} is returned, its return type is the same
4830 __builtin_choose_expr ( \
4831 __builtin_types_compatible_p (typeof (x), double), \
4833 __builtin_choose_expr ( \
4834 __builtin_types_compatible_p (typeof (x), float), \
4836 /* @r{The void expression results in a compile-time error} \
4837 @r{when assigning the result to something.} */ \
4841 @emph{Note:} This construct is only available for C. Furthermore, the
4842 unused expression (@var{exp1} or @var{exp2} depending on the value of
4843 @var{const_exp}) may still generate syntax errors. This may change in
4848 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
4849 You can use the built-in function @code{__builtin_constant_p} to
4850 determine if a value is known to be constant at compile-time and hence
4851 that GCC can perform constant-folding on expressions involving that
4852 value. The argument of the function is the value to test. The function
4853 returns the integer 1 if the argument is known to be a compile-time
4854 constant and 0 if it is not known to be a compile-time constant. A
4855 return of 0 does not indicate that the value is @emph{not} a constant,
4856 but merely that GCC cannot prove it is a constant with the specified
4857 value of the @option{-O} option.
4859 You would typically use this function in an embedded application where
4860 memory was a critical resource. If you have some complex calculation,
4861 you may want it to be folded if it involves constants, but need to call
4862 a function if it does not. For example:
4865 #define Scale_Value(X) \
4866 (__builtin_constant_p (X) \
4867 ? ((X) * SCALE + OFFSET) : Scale (X))
4870 You may use this built-in function in either a macro or an inline
4871 function. However, if you use it in an inlined function and pass an
4872 argument of the function as the argument to the built-in, GCC will
4873 never return 1 when you call the inline function with a string constant
4874 or compound literal (@pxref{Compound Literals}) and will not return 1
4875 when you pass a constant numeric value to the inline function unless you
4876 specify the @option{-O} option.
4878 You may also use @code{__builtin_constant_p} in initializers for static
4879 data. For instance, you can write
4882 static const int table[] = @{
4883 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
4889 This is an acceptable initializer even if @var{EXPRESSION} is not a
4890 constant expression. GCC must be more conservative about evaluating the
4891 built-in in this case, because it has no opportunity to perform
4894 Previous versions of GCC did not accept this built-in in data
4895 initializers. The earliest version where it is completely safe is
4899 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
4900 @opindex fprofile-arcs
4901 You may use @code{__builtin_expect} to provide the compiler with
4902 branch prediction information. In general, you should prefer to
4903 use actual profile feedback for this (@option{-fprofile-arcs}), as
4904 programmers are notoriously bad at predicting how their programs
4905 actually perform. However, there are applications in which this
4906 data is hard to collect.
4908 The return value is the value of @var{exp}, which should be an
4909 integral expression. The value of @var{c} must be a compile-time
4910 constant. The semantics of the built-in are that it is expected
4911 that @var{exp} == @var{c}. For example:
4914 if (__builtin_expect (x, 0))
4919 would indicate that we do not expect to call @code{foo}, since
4920 we expect @code{x} to be zero. Since you are limited to integral
4921 expressions for @var{exp}, you should use constructions such as
4924 if (__builtin_expect (ptr != NULL, 1))
4929 when testing pointer or floating-point values.
4932 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
4933 This function is used to minimize cache-miss latency by moving data into
4934 a cache before it is accessed.
4935 You can insert calls to @code{__builtin_prefetch} into code for which
4936 you know addresses of data in memory that is likely to be accessed soon.
4937 If the target supports them, data prefetch instructions will be generated.
4938 If the prefetch is done early enough before the access then the data will
4939 be in the cache by the time it is accessed.
4941 The value of @var{addr} is the address of the memory to prefetch.
4942 There are two optional arguments, @var{rw} and @var{locality}.
4943 The value of @var{rw} is a compile-time constant one or zero; one
4944 means that the prefetch is preparing for a write to the memory address
4945 and zero, the default, means that the prefetch is preparing for a read.
4946 The value @var{locality} must be a compile-time constant integer between
4947 zero and three. A value of zero means that the data has no temporal
4948 locality, so it need not be left in the cache after the access. A value
4949 of three means that the data has a high degree of temporal locality and
4950 should be left in all levels of cache possible. Values of one and two
4951 mean, respectively, a low or moderate degree of temporal locality. The
4955 for (i = 0; i < n; i++)
4958 __builtin_prefetch (&a[i+j], 1, 1);
4959 __builtin_prefetch (&b[i+j], 0, 1);
4964 Data prefetch does not generate faults if @var{addr} is invalid, but
4965 the address expression itself must be valid. For example, a prefetch
4966 of @code{p->next} will not fault if @code{p->next} is not a valid
4967 address, but evaluation will fault if @code{p} is not a valid address.
4969 If the target does not support data prefetch, the address expression
4970 is evaluated if it includes side effects but no other code is generated
4971 and GCC does not issue a warning.
4974 @deftypefn {Built-in Function} double __builtin_huge_val (void)
4975 Returns a positive infinity, if supported by the floating-point format,
4976 else @code{DBL_MAX}. This function is suitable for implementing the
4977 ISO C macro @code{HUGE_VAL}.
4980 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
4981 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
4984 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
4985 Similar to @code{__builtin_huge_val}, except the return
4986 type is @code{long double}.
4989 @deftypefn {Built-in Function} double __builtin_inf (void)
4990 Similar to @code{__builtin_huge_val}, except a warning is generated
4991 if the target floating-point format does not support infinities.
4992 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
4995 @deftypefn {Built-in Function} float __builtin_inff (void)
4996 Similar to @code{__builtin_inf}, except the return type is @code{float}.
4999 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5000 Similar to @code{__builtin_inf}, except the return
5001 type is @code{long double}.
5004 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5005 This is an implementation of the ISO C99 function @code{nan}.
5007 Since ISO C99 defines this function in terms of @code{strtod}, which we
5008 do not implement, a description of the parsing is in order. The string
5009 is parsed as by @code{strtol}; that is, the base is recognized by
5010 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5011 in the significand such that the least significant bit of the number
5012 is at the least significant bit of the significand. The number is
5013 truncated to fit the significand field provided. The significand is
5014 forced to be a quiet NaN.
5016 This function, if given a string literal, is evaluated early enough
5017 that it is considered a compile-time constant.
5020 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5021 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5024 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5025 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5028 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5029 Similar to @code{__builtin_nan}, except the significand is forced
5030 to be a signaling NaN. The @code{nans} function is proposed by
5031 @uref{http://std.dkuug.dk/JTC1/SC22/WG14/www/docs/n965.htm,,WG14 N965}.
5034 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5035 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5038 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5039 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5042 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5043 Returns one plus the index of the least significant 1-bit of @var{x}, or
5044 if @var{x} is zero, returns zero.
5047 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5048 Returns the number of leading 0-bits in @var{x}, starting at the most
5049 significant bit position. If @var{x} is 0, the result is undefined.
5052 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5053 Returns the number of trailing 0-bits in @var{x}, starting at the least
5054 significant bit position. If @var{x} is 0, the result is undefined.
5057 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5058 Returns the number of 1-bits in @var{x}.
5061 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5062 Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5066 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5067 Similar to @code{__builtin_ffs}, except the argument type is
5068 @code{unsigned long}.
5071 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5072 Similar to @code{__builtin_clz}, except the argument type is
5073 @code{unsigned long}.
5076 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5077 Similar to @code{__builtin_ctz}, except the argument type is
5078 @code{unsigned long}.
5081 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5082 Similar to @code{__builtin_popcount}, except the argument type is
5083 @code{unsigned long}.
5086 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5087 Similar to @code{__builtin_parity}, except the argument type is
5088 @code{unsigned long}.
5091 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5092 Similar to @code{__builtin_ffs}, except the argument type is
5093 @code{unsigned long long}.
5096 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5097 Similar to @code{__builtin_clz}, except the argument type is
5098 @code{unsigned long long}.
5101 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5102 Similar to @code{__builtin_ctz}, except the argument type is
5103 @code{unsigned long long}.
5106 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5107 Similar to @code{__builtin_popcount}, except the argument type is
5108 @code{unsigned long long}.
5111 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5112 Similar to @code{__builtin_parity}, except the argument type is
5113 @code{unsigned long long}.
5117 @node Target Builtins
5118 @section Built-in Functions Specific to Particular Target Machines
5120 On some target machines, GCC supports many built-in functions specific
5121 to those machines. Generally these generate calls to specific machine
5122 instructions, but allow the compiler to schedule those calls.
5125 * Alpha Built-in Functions::
5126 * X86 Built-in Functions::
5127 * PowerPC AltiVec Built-in Functions::
5130 @node Alpha Built-in Functions
5131 @subsection Alpha Built-in Functions
5133 These built-in functions are available for the Alpha family of
5134 processors, depending on the command-line switches used.
5136 The following built-in functions are always available. They
5137 all generate the machine instruction that is part of the name.
5140 long __builtin_alpha_implver (void)
5141 long __builtin_alpha_rpcc (void)
5142 long __builtin_alpha_amask (long)
5143 long __builtin_alpha_cmpbge (long, long)
5144 long __builtin_alpha_extbl (long, long)
5145 long __builtin_alpha_extwl (long, long)
5146 long __builtin_alpha_extll (long, long)
5147 long __builtin_alpha_extql (long, long)
5148 long __builtin_alpha_extwh (long, long)
5149 long __builtin_alpha_extlh (long, long)
5150 long __builtin_alpha_extqh (long, long)
5151 long __builtin_alpha_insbl (long, long)
5152 long __builtin_alpha_inswl (long, long)
5153 long __builtin_alpha_insll (long, long)
5154 long __builtin_alpha_insql (long, long)
5155 long __builtin_alpha_inswh (long, long)
5156 long __builtin_alpha_inslh (long, long)
5157 long __builtin_alpha_insqh (long, long)
5158 long __builtin_alpha_mskbl (long, long)
5159 long __builtin_alpha_mskwl (long, long)
5160 long __builtin_alpha_mskll (long, long)
5161 long __builtin_alpha_mskql (long, long)
5162 long __builtin_alpha_mskwh (long, long)
5163 long __builtin_alpha_msklh (long, long)
5164 long __builtin_alpha_mskqh (long, long)
5165 long __builtin_alpha_umulh (long, long)
5166 long __builtin_alpha_zap (long, long)
5167 long __builtin_alpha_zapnot (long, long)
5170 The following built-in functions are always with @option{-mmax}
5171 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5172 later. They all generate the machine instruction that is part
5176 long __builtin_alpha_pklb (long)
5177 long __builtin_alpha_pkwb (long)
5178 long __builtin_alpha_unpkbl (long)
5179 long __builtin_alpha_unpkbw (long)
5180 long __builtin_alpha_minub8 (long, long)
5181 long __builtin_alpha_minsb8 (long, long)
5182 long __builtin_alpha_minuw4 (long, long)
5183 long __builtin_alpha_minsw4 (long, long)
5184 long __builtin_alpha_maxub8 (long, long)
5185 long __builtin_alpha_maxsb8 (long, long)
5186 long __builtin_alpha_maxuw4 (long, long)
5187 long __builtin_alpha_maxsw4 (long, long)
5188 long __builtin_alpha_perr (long, long)
5191 The following built-in functions are always with @option{-mcix}
5192 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5193 later. They all generate the machine instruction that is part
5197 long __builtin_alpha_cttz (long)
5198 long __builtin_alpha_ctlz (long)
5199 long __builtin_alpha_ctpop (long)
5202 The following builtins are available on systems that use the OSF/1
5203 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5204 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5205 @code{rdval} and @code{wrval}.
5208 void *__builtin_thread_pointer (void)
5209 void __builtin_set_thread_pointer (void *)
5212 @node X86 Built-in Functions
5213 @subsection X86 Built-in Functions
5215 These built-in functions are available for the i386 and x86-64 family
5216 of computers, depending on the command-line switches used.
5218 The following machine modes are available for use with MMX built-in functions
5219 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5220 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5221 vector of eight 8-bit integers. Some of the built-in functions operate on
5222 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5224 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5225 of two 32-bit floating point values.
5227 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5228 floating point values. Some instructions use a vector of four 32-bit
5229 integers, these use @code{V4SI}. Finally, some instructions operate on an
5230 entire vector register, interpreting it as a 128-bit integer, these use mode
5233 The following built-in functions are made available by @option{-mmmx}.
5234 All of them generate the machine instruction that is part of the name.
5237 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5238 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5239 v2si __builtin_ia32_paddd (v2si, v2si)
5240 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5241 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5242 v2si __builtin_ia32_psubd (v2si, v2si)
5243 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5244 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5245 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5246 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5247 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5248 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5249 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5250 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5251 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5252 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5253 di __builtin_ia32_pand (di, di)
5254 di __builtin_ia32_pandn (di,di)
5255 di __builtin_ia32_por (di, di)
5256 di __builtin_ia32_pxor (di, di)
5257 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5258 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5259 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5260 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5261 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5262 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5263 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5264 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5265 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5266 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5267 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5268 v2si __builtin_ia32_punpckldq (v2si, v2si)
5269 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5270 v4hi __builtin_ia32_packssdw (v2si, v2si)
5271 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5274 The following built-in functions are made available either with
5275 @option{-msse}, or with a combination of @option{-m3dnow} and
5276 @option{-march=athlon}. All of them generate the machine
5277 instruction that is part of the name.
5280 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5281 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5282 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5283 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5284 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5285 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5286 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5287 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5288 int __builtin_ia32_pextrw (v4hi, int)
5289 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5290 int __builtin_ia32_pmovmskb (v8qi)
5291 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5292 void __builtin_ia32_movntq (di *, di)
5293 void __builtin_ia32_sfence (void)
5296 The following built-in functions are available when @option{-msse} is used.
5297 All of them generate the machine instruction that is part of the name.
5300 int __builtin_ia32_comieq (v4sf, v4sf)
5301 int __builtin_ia32_comineq (v4sf, v4sf)
5302 int __builtin_ia32_comilt (v4sf, v4sf)
5303 int __builtin_ia32_comile (v4sf, v4sf)
5304 int __builtin_ia32_comigt (v4sf, v4sf)
5305 int __builtin_ia32_comige (v4sf, v4sf)
5306 int __builtin_ia32_ucomieq (v4sf, v4sf)
5307 int __builtin_ia32_ucomineq (v4sf, v4sf)
5308 int __builtin_ia32_ucomilt (v4sf, v4sf)
5309 int __builtin_ia32_ucomile (v4sf, v4sf)
5310 int __builtin_ia32_ucomigt (v4sf, v4sf)
5311 int __builtin_ia32_ucomige (v4sf, v4sf)
5312 v4sf __builtin_ia32_addps (v4sf, v4sf)
5313 v4sf __builtin_ia32_subps (v4sf, v4sf)
5314 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5315 v4sf __builtin_ia32_divps (v4sf, v4sf)
5316 v4sf __builtin_ia32_addss (v4sf, v4sf)
5317 v4sf __builtin_ia32_subss (v4sf, v4sf)
5318 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5319 v4sf __builtin_ia32_divss (v4sf, v4sf)
5320 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5321 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5322 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5323 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5324 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5325 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5326 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5327 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5328 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5329 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5330 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5331 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5332 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5333 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5334 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5335 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5336 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5337 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
5338 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
5339 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
5340 v4sf __builtin_ia32_maxps (v4sf, v4sf)
5341 v4sf __builtin_ia32_maxss (v4sf, v4sf)
5342 v4sf __builtin_ia32_minps (v4sf, v4sf)
5343 v4sf __builtin_ia32_minss (v4sf, v4sf)
5344 v4sf __builtin_ia32_andps (v4sf, v4sf)
5345 v4sf __builtin_ia32_andnps (v4sf, v4sf)
5346 v4sf __builtin_ia32_orps (v4sf, v4sf)
5347 v4sf __builtin_ia32_xorps (v4sf, v4sf)
5348 v4sf __builtin_ia32_movss (v4sf, v4sf)
5349 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
5350 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
5351 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
5352 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
5353 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
5354 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
5355 v2si __builtin_ia32_cvtps2pi (v4sf)
5356 int __builtin_ia32_cvtss2si (v4sf)
5357 v2si __builtin_ia32_cvttps2pi (v4sf)
5358 int __builtin_ia32_cvttss2si (v4sf)
5359 v4sf __builtin_ia32_rcpps (v4sf)
5360 v4sf __builtin_ia32_rsqrtps (v4sf)
5361 v4sf __builtin_ia32_sqrtps (v4sf)
5362 v4sf __builtin_ia32_rcpss (v4sf)
5363 v4sf __builtin_ia32_rsqrtss (v4sf)
5364 v4sf __builtin_ia32_sqrtss (v4sf)
5365 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
5366 void __builtin_ia32_movntps (float *, v4sf)
5367 int __builtin_ia32_movmskps (v4sf)
5370 The following built-in functions are available when @option{-msse} is used.
5373 @item v4sf __builtin_ia32_loadaps (float *)
5374 Generates the @code{movaps} machine instruction as a load from memory.
5375 @item void __builtin_ia32_storeaps (float *, v4sf)
5376 Generates the @code{movaps} machine instruction as a store to memory.
5377 @item v4sf __builtin_ia32_loadups (float *)
5378 Generates the @code{movups} machine instruction as a load from memory.
5379 @item void __builtin_ia32_storeups (float *, v4sf)
5380 Generates the @code{movups} machine instruction as a store to memory.
5381 @item v4sf __builtin_ia32_loadsss (float *)
5382 Generates the @code{movss} machine instruction as a load from memory.
5383 @item void __builtin_ia32_storess (float *, v4sf)
5384 Generates the @code{movss} machine instruction as a store to memory.
5385 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
5386 Generates the @code{movhps} machine instruction as a load from memory.
5387 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
5388 Generates the @code{movlps} machine instruction as a load from memory
5389 @item void __builtin_ia32_storehps (v4sf, v2si *)
5390 Generates the @code{movhps} machine instruction as a store to memory.
5391 @item void __builtin_ia32_storelps (v4sf, v2si *)
5392 Generates the @code{movlps} machine instruction as a store to memory.
5395 The following built-in functions are available when @option{-m3dnow} is used.
5396 All of them generate the machine instruction that is part of the name.
5399 void __builtin_ia32_femms (void)
5400 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
5401 v2si __builtin_ia32_pf2id (v2sf)
5402 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
5403 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
5404 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
5405 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
5406 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
5407 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
5408 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
5409 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
5410 v2sf __builtin_ia32_pfrcp (v2sf)
5411 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
5412 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
5413 v2sf __builtin_ia32_pfrsqrt (v2sf)
5414 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
5415 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
5416 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
5417 v2sf __builtin_ia32_pi2fd (v2si)
5418 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
5421 The following built-in functions are available when both @option{-m3dnow}
5422 and @option{-march=athlon} are used. All of them generate the machine
5423 instruction that is part of the name.
5426 v2si __builtin_ia32_pf2iw (v2sf)
5427 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
5428 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
5429 v2sf __builtin_ia32_pi2fw (v2si)
5430 v2sf __builtin_ia32_pswapdsf (v2sf)
5431 v2si __builtin_ia32_pswapdsi (v2si)
5434 @node PowerPC AltiVec Built-in Functions
5435 @subsection PowerPC AltiVec Built-in Functions
5437 These built-in functions are available for the PowerPC family
5438 of computers, depending on the command-line switches used.
5440 The following machine modes are available for use with AltiVec built-in
5441 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
5442 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
5443 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
5444 @code{V16QI} for a vector of sixteen 8-bit integers.
5446 The following functions are made available by including
5447 @code{<altivec.h>} and using @option{-maltivec} and
5448 @option{-mabi=altivec}. The functions implement the functionality
5449 described in Motorola's AltiVec Programming Interface Manual.
5451 There are a few differences from Motorola's documentation and GCC's
5452 implementation. Vector constants are done with curly braces (not
5453 parentheses). Vector initializers require no casts if the vector
5454 constant is of the same type as the variable it is initializing. The
5455 @code{vector bool} type is deprecated and will be discontinued in
5456 further revisions. Use @code{vector signed} instead. If @code{signed}
5457 or @code{unsigned} is omitted, the vector type will default to
5458 @code{signed}. Lastly, all overloaded functions are implemented with macros
5459 for the C implementation. So code the following example will not work:
5462 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
5465 Since vec_add is a macro, the vector constant in the above example will
5466 be treated as four different arguments. Wrap the entire argument in
5467 parentheses for this to work. The C++ implementation does not use
5470 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
5471 Internally, GCC uses built-in functions to achieve the functionality in
5472 the aforementioned header file, but they are not supported and are
5473 subject to change without notice.
5476 vector signed char vec_abs (vector signed char, vector signed char);
5477 vector signed short vec_abs (vector signed short, vector signed short);
5478 vector signed int vec_abs (vector signed int, vector signed int);
5479 vector signed float vec_abs (vector signed float, vector signed float);
5481 vector signed char vec_abss (vector signed char, vector signed char);
5482 vector signed short vec_abss (vector signed short, vector signed short);
5484 vector signed char vec_add (vector signed char, vector signed char);
5485 vector unsigned char vec_add (vector signed char, vector unsigned char);
5487 vector unsigned char vec_add (vector unsigned char, vector signed char);
5489 vector unsigned char vec_add (vector unsigned char,
5490 vector unsigned char);
5491 vector signed short vec_add (vector signed short, vector signed short);
5492 vector unsigned short vec_add (vector signed short,
5493 vector unsigned short);
5494 vector unsigned short vec_add (vector unsigned short,
5495 vector signed short);
5496 vector unsigned short vec_add (vector unsigned short,
5497 vector unsigned short);
5498 vector signed int vec_add (vector signed int, vector signed int);
5499 vector unsigned int vec_add (vector signed int, vector unsigned int);
5500 vector unsigned int vec_add (vector unsigned int, vector signed int);
5501 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
5502 vector float vec_add (vector float, vector float);
5504 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
5506 vector unsigned char vec_adds (vector signed char,
5507 vector unsigned char);
5508 vector unsigned char vec_adds (vector unsigned char,
5509 vector signed char);
5510 vector unsigned char vec_adds (vector unsigned char,
5511 vector unsigned char);
5512 vector signed char vec_adds (vector signed char, vector signed char);
5513 vector unsigned short vec_adds (vector signed short,
5514 vector unsigned short);
5515 vector unsigned short vec_adds (vector unsigned short,
5516 vector signed short);
5517 vector unsigned short vec_adds (vector unsigned short,
5518 vector unsigned short);
5519 vector signed short vec_adds (vector signed short, vector signed short);
5521 vector unsigned int vec_adds (vector signed int, vector unsigned int);
5522 vector unsigned int vec_adds (vector unsigned int, vector signed int);
5523 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
5525 vector signed int vec_adds (vector signed int, vector signed int);
5527 vector float vec_and (vector float, vector float);
5528 vector float vec_and (vector float, vector signed int);
5529 vector float vec_and (vector signed int, vector float);
5530 vector signed int vec_and (vector signed int, vector signed int);
5531 vector unsigned int vec_and (vector signed int, vector unsigned int);
5532 vector unsigned int vec_and (vector unsigned int, vector signed int);
5533 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
5534 vector signed short vec_and (vector signed short, vector signed short);
5535 vector unsigned short vec_and (vector signed short,
5536 vector unsigned short);
5537 vector unsigned short vec_and (vector unsigned short,
5538 vector signed short);
5539 vector unsigned short vec_and (vector unsigned short,
5540 vector unsigned short);
5541 vector signed char vec_and (vector signed char, vector signed char);
5542 vector unsigned char vec_and (vector signed char, vector unsigned char);
5544 vector unsigned char vec_and (vector unsigned char, vector signed char);
5546 vector unsigned char vec_and (vector unsigned char,
5547 vector unsigned char);
5549 vector float vec_andc (vector float, vector float);
5550 vector float vec_andc (vector float, vector signed int);
5551 vector float vec_andc (vector signed int, vector float);
5552 vector signed int vec_andc (vector signed int, vector signed int);
5553 vector unsigned int vec_andc (vector signed int, vector unsigned int);
5554 vector unsigned int vec_andc (vector unsigned int, vector signed int);
5555 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
5557 vector signed short vec_andc (vector signed short, vector signed short);
5559 vector unsigned short vec_andc (vector signed short,
5560 vector unsigned short);
5561 vector unsigned short vec_andc (vector unsigned short,
5562 vector signed short);
5563 vector unsigned short vec_andc (vector unsigned short,
5564 vector unsigned short);
5565 vector signed char vec_andc (vector signed char, vector signed char);
5566 vector unsigned char vec_andc (vector signed char,
5567 vector unsigned char);
5568 vector unsigned char vec_andc (vector unsigned char,
5569 vector signed char);
5570 vector unsigned char vec_andc (vector unsigned char,
5571 vector unsigned char);
5573 vector unsigned char vec_avg (vector unsigned char,
5574 vector unsigned char);
5575 vector signed char vec_avg (vector signed char, vector signed char);
5576 vector unsigned short vec_avg (vector unsigned short,
5577 vector unsigned short);
5578 vector signed short vec_avg (vector signed short, vector signed short);
5579 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
5580 vector signed int vec_avg (vector signed int, vector signed int);
5582 vector float vec_ceil (vector float);
5584 vector signed int vec_cmpb (vector float, vector float);
5586 vector signed char vec_cmpeq (vector signed char, vector signed char);
5587 vector signed char vec_cmpeq (vector unsigned char,
5588 vector unsigned char);
5589 vector signed short vec_cmpeq (vector signed short,
5590 vector signed short);
5591 vector signed short vec_cmpeq (vector unsigned short,
5592 vector unsigned short);
5593 vector signed int vec_cmpeq (vector signed int, vector signed int);
5594 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
5595 vector signed int vec_cmpeq (vector float, vector float);
5597 vector signed int vec_cmpge (vector float, vector float);
5599 vector signed char vec_cmpgt (vector unsigned char,
5600 vector unsigned char);
5601 vector signed char vec_cmpgt (vector signed char, vector signed char);
5602 vector signed short vec_cmpgt (vector unsigned short,
5603 vector unsigned short);
5604 vector signed short vec_cmpgt (vector signed short,
5605 vector signed short);
5606 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
5607 vector signed int vec_cmpgt (vector signed int, vector signed int);
5608 vector signed int vec_cmpgt (vector float, vector float);
5610 vector signed int vec_cmple (vector float, vector float);
5612 vector signed char vec_cmplt (vector unsigned char,
5613 vector unsigned char);
5614 vector signed char vec_cmplt (vector signed char, vector signed char);
5615 vector signed short vec_cmplt (vector unsigned short,
5616 vector unsigned short);
5617 vector signed short vec_cmplt (vector signed short,
5618 vector signed short);
5619 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
5620 vector signed int vec_cmplt (vector signed int, vector signed int);
5621 vector signed int vec_cmplt (vector float, vector float);
5623 vector float vec_ctf (vector unsigned int, const char);
5624 vector float vec_ctf (vector signed int, const char);
5626 vector signed int vec_cts (vector float, const char);
5628 vector unsigned int vec_ctu (vector float, const char);
5630 void vec_dss (const char);
5632 void vec_dssall (void);
5634 void vec_dst (void *, int, const char);
5636 void vec_dstst (void *, int, const char);
5638 void vec_dststt (void *, int, const char);
5640 void vec_dstt (void *, int, const char);
5642 vector float vec_expte (vector float, vector float);
5644 vector float vec_floor (vector float, vector float);
5646 vector float vec_ld (int, vector float *);
5647 vector float vec_ld (int, float *):
5648 vector signed int vec_ld (int, int *);
5649 vector signed int vec_ld (int, vector signed int *);
5650 vector unsigned int vec_ld (int, vector unsigned int *);
5651 vector unsigned int vec_ld (int, unsigned int *);
5652 vector signed short vec_ld (int, short *, vector signed short *);
5653 vector unsigned short vec_ld (int, unsigned short *,
5654 vector unsigned short *);
5655 vector signed char vec_ld (int, signed char *);
5656 vector signed char vec_ld (int, vector signed char *);
5657 vector unsigned char vec_ld (int, unsigned char *);
5658 vector unsigned char vec_ld (int, vector unsigned char *);
5660 vector signed char vec_lde (int, signed char *);
5661 vector unsigned char vec_lde (int, unsigned char *);
5662 vector signed short vec_lde (int, short *);
5663 vector unsigned short vec_lde (int, unsigned short *);
5664 vector float vec_lde (int, float *);
5665 vector signed int vec_lde (int, int *);
5666 vector unsigned int vec_lde (int, unsigned int *);
5668 void float vec_ldl (int, float *);
5669 void float vec_ldl (int, vector float *);
5670 void signed int vec_ldl (int, vector signed int *);
5671 void signed int vec_ldl (int, int *);
5672 void unsigned int vec_ldl (int, unsigned int *);
5673 void unsigned int vec_ldl (int, vector unsigned int *);
5674 void signed short vec_ldl (int, vector signed short *);
5675 void signed short vec_ldl (int, short *);
5676 void unsigned short vec_ldl (int, vector unsigned short *);
5677 void unsigned short vec_ldl (int, unsigned short *);
5678 void signed char vec_ldl (int, vector signed char *);
5679 void signed char vec_ldl (int, signed char *);
5680 void unsigned char vec_ldl (int, vector unsigned char *);
5681 void unsigned char vec_ldl (int, unsigned char *);
5683 vector float vec_loge (vector float);
5685 vector unsigned char vec_lvsl (int, void *, int *);
5687 vector unsigned char vec_lvsr (int, void *, int *);
5689 vector float vec_madd (vector float, vector float, vector float);
5691 vector signed short vec_madds (vector signed short, vector signed short,
5692 vector signed short);
5694 vector unsigned char vec_max (vector signed char, vector unsigned char);
5696 vector unsigned char vec_max (vector unsigned char, vector signed char);
5698 vector unsigned char vec_max (vector unsigned char,
5699 vector unsigned char);
5700 vector signed char vec_max (vector signed char, vector signed char);
5701 vector unsigned short vec_max (vector signed short,
5702 vector unsigned short);
5703 vector unsigned short vec_max (vector unsigned short,
5704 vector signed short);
5705 vector unsigned short vec_max (vector unsigned short,
5706 vector unsigned short);
5707 vector signed short vec_max (vector signed short, vector signed short);
5708 vector unsigned int vec_max (vector signed int, vector unsigned int);
5709 vector unsigned int vec_max (vector unsigned int, vector signed int);
5710 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
5711 vector signed int vec_max (vector signed int, vector signed int);
5712 vector float vec_max (vector float, vector float);
5714 vector signed char vec_mergeh (vector signed char, vector signed char);
5715 vector unsigned char vec_mergeh (vector unsigned char,
5716 vector unsigned char);
5717 vector signed short vec_mergeh (vector signed short,
5718 vector signed short);
5719 vector unsigned short vec_mergeh (vector unsigned short,
5720 vector unsigned short);
5721 vector float vec_mergeh (vector float, vector float);
5722 vector signed int vec_mergeh (vector signed int, vector signed int);
5723 vector unsigned int vec_mergeh (vector unsigned int,
5724 vector unsigned int);
5726 vector signed char vec_mergel (vector signed char, vector signed char);
5727 vector unsigned char vec_mergel (vector unsigned char,
5728 vector unsigned char);
5729 vector signed short vec_mergel (vector signed short,
5730 vector signed short);
5731 vector unsigned short vec_mergel (vector unsigned short,
5732 vector unsigned short);
5733 vector float vec_mergel (vector float, vector float);
5734 vector signed int vec_mergel (vector signed int, vector signed int);
5735 vector unsigned int vec_mergel (vector unsigned int,
5736 vector unsigned int);
5738 vector unsigned short vec_mfvscr (void);
5740 vector unsigned char vec_min (vector signed char, vector unsigned char);
5742 vector unsigned char vec_min (vector unsigned char, vector signed char);
5744 vector unsigned char vec_min (vector unsigned char,
5745 vector unsigned char);
5746 vector signed char vec_min (vector signed char, vector signed char);
5747 vector unsigned short vec_min (vector signed short,
5748 vector unsigned short);
5749 vector unsigned short vec_min (vector unsigned short,
5750 vector signed short);
5751 vector unsigned short vec_min (vector unsigned short,
5752 vector unsigned short);
5753 vector signed short vec_min (vector signed short, vector signed short);
5754 vector unsigned int vec_min (vector signed int, vector unsigned int);
5755 vector unsigned int vec_min (vector unsigned int, vector signed int);
5756 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
5757 vector signed int vec_min (vector signed int, vector signed int);
5758 vector float vec_min (vector float, vector float);
5760 vector signed short vec_mladd (vector signed short, vector signed short,
5761 vector signed short);
5762 vector signed short vec_mladd (vector signed short,
5763 vector unsigned short,
5764 vector unsigned short);
5765 vector signed short vec_mladd (vector unsigned short,
5766 vector signed short,
5767 vector signed short);
5768 vector unsigned short vec_mladd (vector unsigned short,
5769 vector unsigned short,
5770 vector unsigned short);
5772 vector signed short vec_mradds (vector signed short,
5773 vector signed short,
5774 vector signed short);
5776 vector unsigned int vec_msum (vector unsigned char,
5777 vector unsigned char,
5778 vector unsigned int);
5779 vector signed int vec_msum (vector signed char, vector unsigned char,
5781 vector unsigned int vec_msum (vector unsigned short,
5782 vector unsigned short,
5783 vector unsigned int);
5784 vector signed int vec_msum (vector signed short, vector signed short,
5787 vector unsigned int vec_msums (vector unsigned short,
5788 vector unsigned short,
5789 vector unsigned int);
5790 vector signed int vec_msums (vector signed short, vector signed short,
5793 void vec_mtvscr (vector signed int);
5794 void vec_mtvscr (vector unsigned int);
5795 void vec_mtvscr (vector signed short);
5796 void vec_mtvscr (vector unsigned short);
5797 void vec_mtvscr (vector signed char);
5798 void vec_mtvscr (vector unsigned char);
5800 vector unsigned short vec_mule (vector unsigned char,
5801 vector unsigned char);
5802 vector signed short vec_mule (vector signed char, vector signed char);
5803 vector unsigned int vec_mule (vector unsigned short,
5804 vector unsigned short);
5805 vector signed int vec_mule (vector signed short, vector signed short);
5807 vector unsigned short vec_mulo (vector unsigned char,
5808 vector unsigned char);
5809 vector signed short vec_mulo (vector signed char, vector signed char);
5810 vector unsigned int vec_mulo (vector unsigned short,
5811 vector unsigned short);
5812 vector signed int vec_mulo (vector signed short, vector signed short);
5814 vector float vec_nmsub (vector float, vector float, vector float);
5816 vector float vec_nor (vector float, vector float);
5817 vector signed int vec_nor (vector signed int, vector signed int);
5818 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
5819 vector signed short vec_nor (vector signed short, vector signed short);
5820 vector unsigned short vec_nor (vector unsigned short,
5821 vector unsigned short);
5822 vector signed char vec_nor (vector signed char, vector signed char);
5823 vector unsigned char vec_nor (vector unsigned char,
5824 vector unsigned char);
5826 vector float vec_or (vector float, vector float);
5827 vector float vec_or (vector float, vector signed int);
5828 vector float vec_or (vector signed int, vector float);
5829 vector signed int vec_or (vector signed int, vector signed int);
5830 vector unsigned int vec_or (vector signed int, vector unsigned int);
5831 vector unsigned int vec_or (vector unsigned int, vector signed int);
5832 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
5833 vector signed short vec_or (vector signed short, vector signed short);
5834 vector unsigned short vec_or (vector signed short,
5835 vector unsigned short);
5836 vector unsigned short vec_or (vector unsigned short,
5837 vector signed short);
5838 vector unsigned short vec_or (vector unsigned short,
5839 vector unsigned short);
5840 vector signed char vec_or (vector signed char, vector signed char);
5841 vector unsigned char vec_or (vector signed char, vector unsigned char);
5842 vector unsigned char vec_or (vector unsigned char, vector signed char);
5843 vector unsigned char vec_or (vector unsigned char,
5844 vector unsigned char);
5846 vector signed char vec_pack (vector signed short, vector signed short);
5847 vector unsigned char vec_pack (vector unsigned short,
5848 vector unsigned short);
5849 vector signed short vec_pack (vector signed int, vector signed int);
5850 vector unsigned short vec_pack (vector unsigned int,
5851 vector unsigned int);
5853 vector signed short vec_packpx (vector unsigned int,
5854 vector unsigned int);
5856 vector unsigned char vec_packs (vector unsigned short,
5857 vector unsigned short);
5858 vector signed char vec_packs (vector signed short, vector signed short);
5860 vector unsigned short vec_packs (vector unsigned int,
5861 vector unsigned int);
5862 vector signed short vec_packs (vector signed int, vector signed int);
5864 vector unsigned char vec_packsu (vector unsigned short,
5865 vector unsigned short);
5866 vector unsigned char vec_packsu (vector signed short,
5867 vector signed short);
5868 vector unsigned short vec_packsu (vector unsigned int,
5869 vector unsigned int);
5870 vector unsigned short vec_packsu (vector signed int, vector signed int);
5872 vector float vec_perm (vector float, vector float,
5873 vector unsigned char);
5874 vector signed int vec_perm (vector signed int, vector signed int,
5875 vector unsigned char);
5876 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
5877 vector unsigned char);
5878 vector signed short vec_perm (vector signed short, vector signed short,
5879 vector unsigned char);
5880 vector unsigned short vec_perm (vector unsigned short,
5881 vector unsigned short,
5882 vector unsigned char);
5883 vector signed char vec_perm (vector signed char, vector signed char,
5884 vector unsigned char);
5885 vector unsigned char vec_perm (vector unsigned char,
5886 vector unsigned char,
5887 vector unsigned char);
5889 vector float vec_re (vector float);
5891 vector signed char vec_rl (vector signed char, vector unsigned char);
5892 vector unsigned char vec_rl (vector unsigned char,
5893 vector unsigned char);
5894 vector signed short vec_rl (vector signed short, vector unsigned short);
5896 vector unsigned short vec_rl (vector unsigned short,
5897 vector unsigned short);
5898 vector signed int vec_rl (vector signed int, vector unsigned int);
5899 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
5901 vector float vec_round (vector float);
5903 vector float vec_rsqrte (vector float);
5905 vector float vec_sel (vector float, vector float, vector signed int);
5906 vector float vec_sel (vector float, vector float, vector unsigned int);
5907 vector signed int vec_sel (vector signed int, vector signed int,
5909 vector signed int vec_sel (vector signed int, vector signed int,
5910 vector unsigned int);
5911 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
5913 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
5914 vector unsigned int);
5915 vector signed short vec_sel (vector signed short, vector signed short,
5916 vector signed short);
5917 vector signed short vec_sel (vector signed short, vector signed short,
5918 vector unsigned short);
5919 vector unsigned short vec_sel (vector unsigned short,
5920 vector unsigned short,
5921 vector signed short);
5922 vector unsigned short vec_sel (vector unsigned short,
5923 vector unsigned short,
5924 vector unsigned short);
5925 vector signed char vec_sel (vector signed char, vector signed char,
5926 vector signed char);
5927 vector signed char vec_sel (vector signed char, vector signed char,
5928 vector unsigned char);
5929 vector unsigned char vec_sel (vector unsigned char,
5930 vector unsigned char,
5931 vector signed char);
5932 vector unsigned char vec_sel (vector unsigned char,
5933 vector unsigned char,
5934 vector unsigned char);
5936 vector signed char vec_sl (vector signed char, vector unsigned char);
5937 vector unsigned char vec_sl (vector unsigned char,
5938 vector unsigned char);
5939 vector signed short vec_sl (vector signed short, vector unsigned short);
5941 vector unsigned short vec_sl (vector unsigned short,
5942 vector unsigned short);
5943 vector signed int vec_sl (vector signed int, vector unsigned int);
5944 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
5946 vector float vec_sld (vector float, vector float, const char);
5947 vector signed int vec_sld (vector signed int, vector signed int,
5949 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
5951 vector signed short vec_sld (vector signed short, vector signed short,
5953 vector unsigned short vec_sld (vector unsigned short,
5954 vector unsigned short, const char);
5955 vector signed char vec_sld (vector signed char, vector signed char,
5957 vector unsigned char vec_sld (vector unsigned char,
5958 vector unsigned char,
5961 vector signed int vec_sll (vector signed int, vector unsigned int);
5962 vector signed int vec_sll (vector signed int, vector unsigned short);
5963 vector signed int vec_sll (vector signed int, vector unsigned char);
5964 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
5965 vector unsigned int vec_sll (vector unsigned int,
5966 vector unsigned short);
5967 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
5969 vector signed short vec_sll (vector signed short, vector unsigned int);
5970 vector signed short vec_sll (vector signed short,
5971 vector unsigned short);
5972 vector signed short vec_sll (vector signed short, vector unsigned char);
5974 vector unsigned short vec_sll (vector unsigned short,
5975 vector unsigned int);
5976 vector unsigned short vec_sll (vector unsigned short,
5977 vector unsigned short);
5978 vector unsigned short vec_sll (vector unsigned short,
5979 vector unsigned char);
5980 vector signed char vec_sll (vector signed char, vector unsigned int);
5981 vector signed char vec_sll (vector signed char, vector unsigned short);
5982 vector signed char vec_sll (vector signed char, vector unsigned char);
5983 vector unsigned char vec_sll (vector unsigned char,
5984 vector unsigned int);
5985 vector unsigned char vec_sll (vector unsigned char,
5986 vector unsigned short);
5987 vector unsigned char vec_sll (vector unsigned char,
5988 vector unsigned char);
5990 vector float vec_slo (vector float, vector signed char);
5991 vector float vec_slo (vector float, vector unsigned char);
5992 vector signed int vec_slo (vector signed int, vector signed char);
5993 vector signed int vec_slo (vector signed int, vector unsigned char);
5994 vector unsigned int vec_slo (vector unsigned int, vector signed char);
5995 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
5997 vector signed short vec_slo (vector signed short, vector signed char);
5998 vector signed short vec_slo (vector signed short, vector unsigned char);
6000 vector unsigned short vec_slo (vector unsigned short,
6001 vector signed char);
6002 vector unsigned short vec_slo (vector unsigned short,
6003 vector unsigned char);
6004 vector signed char vec_slo (vector signed char, vector signed char);
6005 vector signed char vec_slo (vector signed char, vector unsigned char);
6006 vector unsigned char vec_slo (vector unsigned char, vector signed char);
6008 vector unsigned char vec_slo (vector unsigned char,
6009 vector unsigned char);
6011 vector signed char vec_splat (vector signed char, const char);
6012 vector unsigned char vec_splat (vector unsigned char, const char);
6013 vector signed short vec_splat (vector signed short, const char);
6014 vector unsigned short vec_splat (vector unsigned short, const char);
6015 vector float vec_splat (vector float, const char);
6016 vector signed int vec_splat (vector signed int, const char);
6017 vector unsigned int vec_splat (vector unsigned int, const char);
6019 vector signed char vec_splat_s8 (const char);
6021 vector signed short vec_splat_s16 (const char);
6023 vector signed int vec_splat_s32 (const char);
6025 vector unsigned char vec_splat_u8 (const char);
6027 vector unsigned short vec_splat_u16 (const char);
6029 vector unsigned int vec_splat_u32 (const char);
6031 vector signed char vec_sr (vector signed char, vector unsigned char);
6032 vector unsigned char vec_sr (vector unsigned char,
6033 vector unsigned char);
6034 vector signed short vec_sr (vector signed short, vector unsigned short);
6036 vector unsigned short vec_sr (vector unsigned short,
6037 vector unsigned short);
6038 vector signed int vec_sr (vector signed int, vector unsigned int);
6039 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
6041 vector signed char vec_sra (vector signed char, vector unsigned char);
6042 vector unsigned char vec_sra (vector unsigned char,
6043 vector unsigned char);
6044 vector signed short vec_sra (vector signed short,
6045 vector unsigned short);
6046 vector unsigned short vec_sra (vector unsigned short,
6047 vector unsigned short);
6048 vector signed int vec_sra (vector signed int, vector unsigned int);
6049 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
6051 vector signed int vec_srl (vector signed int, vector unsigned int);
6052 vector signed int vec_srl (vector signed int, vector unsigned short);
6053 vector signed int vec_srl (vector signed int, vector unsigned char);
6054 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
6055 vector unsigned int vec_srl (vector unsigned int,
6056 vector unsigned short);
6057 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
6059 vector signed short vec_srl (vector signed short, vector unsigned int);
6060 vector signed short vec_srl (vector signed short,
6061 vector unsigned short);
6062 vector signed short vec_srl (vector signed short, vector unsigned char);
6064 vector unsigned short vec_srl (vector unsigned short,
6065 vector unsigned int);
6066 vector unsigned short vec_srl (vector unsigned short,
6067 vector unsigned short);
6068 vector unsigned short vec_srl (vector unsigned short,
6069 vector unsigned char);
6070 vector signed char vec_srl (vector signed char, vector unsigned int);
6071 vector signed char vec_srl (vector signed char, vector unsigned short);
6072 vector signed char vec_srl (vector signed char, vector unsigned char);
6073 vector unsigned char vec_srl (vector unsigned char,
6074 vector unsigned int);
6075 vector unsigned char vec_srl (vector unsigned char,
6076 vector unsigned short);
6077 vector unsigned char vec_srl (vector unsigned char,
6078 vector unsigned char);
6080 vector float vec_sro (vector float, vector signed char);
6081 vector float vec_sro (vector float, vector unsigned char);
6082 vector signed int vec_sro (vector signed int, vector signed char);
6083 vector signed int vec_sro (vector signed int, vector unsigned char);
6084 vector unsigned int vec_sro (vector unsigned int, vector signed char);
6085 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
6087 vector signed short vec_sro (vector signed short, vector signed char);
6088 vector signed short vec_sro (vector signed short, vector unsigned char);
6090 vector unsigned short vec_sro (vector unsigned short,
6091 vector signed char);
6092 vector unsigned short vec_sro (vector unsigned short,
6093 vector unsigned char);
6094 vector signed char vec_sro (vector signed char, vector signed char);
6095 vector signed char vec_sro (vector signed char, vector unsigned char);
6096 vector unsigned char vec_sro (vector unsigned char, vector signed char);
6098 vector unsigned char vec_sro (vector unsigned char,
6099 vector unsigned char);
6101 void vec_st (vector float, int, float *);
6102 void vec_st (vector float, int, vector float *);
6103 void vec_st (vector signed int, int, int *);
6104 void vec_st (vector signed int, int, unsigned int *);
6105 void vec_st (vector unsigned int, int, unsigned int *);
6106 void vec_st (vector unsigned int, int, vector unsigned int *);
6107 void vec_st (vector signed short, int, short *);
6108 void vec_st (vector signed short, int, vector unsigned short *);
6109 void vec_st (vector signed short, int, vector signed short *);
6110 void vec_st (vector unsigned short, int, unsigned short *);
6111 void vec_st (vector unsigned short, int, vector unsigned short *);
6112 void vec_st (vector signed char, int, signed char *);
6113 void vec_st (vector signed char, int, unsigned char *);
6114 void vec_st (vector signed char, int, vector signed char *);
6115 void vec_st (vector unsigned char, int, unsigned char *);
6116 void vec_st (vector unsigned char, int, vector unsigned char *);
6118 void vec_ste (vector signed char, int, unsigned char *);
6119 void vec_ste (vector signed char, int, signed char *);
6120 void vec_ste (vector unsigned char, int, unsigned char *);
6121 void vec_ste (vector signed short, int, short *);
6122 void vec_ste (vector signed short, int, unsigned short *);
6123 void vec_ste (vector unsigned short, int, void *);
6124 void vec_ste (vector signed int, int, unsigned int *);
6125 void vec_ste (vector signed int, int, int *);
6126 void vec_ste (vector unsigned int, int, unsigned int *);
6127 void vec_ste (vector float, int, float *);
6129 void vec_stl (vector float, int, vector float *);
6130 void vec_stl (vector float, int, float *);
6131 void vec_stl (vector signed int, int, vector signed int *);
6132 void vec_stl (vector signed int, int, int *);
6133 void vec_stl (vector signed int, int, unsigned int *);
6134 void vec_stl (vector unsigned int, int, vector unsigned int *);
6135 void vec_stl (vector unsigned int, int, unsigned int *);
6136 void vec_stl (vector signed short, int, short *);
6137 void vec_stl (vector signed short, int, unsigned short *);
6138 void vec_stl (vector signed short, int, vector signed short *);
6139 void vec_stl (vector unsigned short, int, unsigned short *);
6140 void vec_stl (vector unsigned short, int, vector signed short *);
6141 void vec_stl (vector signed char, int, signed char *);
6142 void vec_stl (vector signed char, int, unsigned char *);
6143 void vec_stl (vector signed char, int, vector signed char *);
6144 void vec_stl (vector unsigned char, int, unsigned char *);
6145 void vec_stl (vector unsigned char, int, vector unsigned char *);
6147 vector signed char vec_sub (vector signed char, vector signed char);
6148 vector unsigned char vec_sub (vector signed char, vector unsigned char);
6150 vector unsigned char vec_sub (vector unsigned char, vector signed char);
6152 vector unsigned char vec_sub (vector unsigned char,
6153 vector unsigned char);
6154 vector signed short vec_sub (vector signed short, vector signed short);
6155 vector unsigned short vec_sub (vector signed short,
6156 vector unsigned short);
6157 vector unsigned short vec_sub (vector unsigned short,
6158 vector signed short);
6159 vector unsigned short vec_sub (vector unsigned short,
6160 vector unsigned short);
6161 vector signed int vec_sub (vector signed int, vector signed int);
6162 vector unsigned int vec_sub (vector signed int, vector unsigned int);
6163 vector unsigned int vec_sub (vector unsigned int, vector signed int);
6164 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
6165 vector float vec_sub (vector float, vector float);
6167 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
6169 vector unsigned char vec_subs (vector signed char,
6170 vector unsigned char);
6171 vector unsigned char vec_subs (vector unsigned char,
6172 vector signed char);
6173 vector unsigned char vec_subs (vector unsigned char,
6174 vector unsigned char);
6175 vector signed char vec_subs (vector signed char, vector signed char);
6176 vector unsigned short vec_subs (vector signed short,
6177 vector unsigned short);
6178 vector unsigned short vec_subs (vector unsigned short,
6179 vector signed short);
6180 vector unsigned short vec_subs (vector unsigned short,
6181 vector unsigned short);
6182 vector signed short vec_subs (vector signed short, vector signed short);
6184 vector unsigned int vec_subs (vector signed int, vector unsigned int);
6185 vector unsigned int vec_subs (vector unsigned int, vector signed int);
6186 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
6188 vector signed int vec_subs (vector signed int, vector signed int);
6190 vector unsigned int vec_sum4s (vector unsigned char,
6191 vector unsigned int);
6192 vector signed int vec_sum4s (vector signed char, vector signed int);
6193 vector signed int vec_sum4s (vector signed short, vector signed int);
6195 vector signed int vec_sum2s (vector signed int, vector signed int);
6197 vector signed int vec_sums (vector signed int, vector signed int);
6199 vector float vec_trunc (vector float);
6201 vector signed short vec_unpackh (vector signed char);
6202 vector unsigned int vec_unpackh (vector signed short);
6203 vector signed int vec_unpackh (vector signed short);
6205 vector signed short vec_unpackl (vector signed char);
6206 vector unsigned int vec_unpackl (vector signed short);
6207 vector signed int vec_unpackl (vector signed short);
6209 vector float vec_xor (vector float, vector float);
6210 vector float vec_xor (vector float, vector signed int);
6211 vector float vec_xor (vector signed int, vector float);
6212 vector signed int vec_xor (vector signed int, vector signed int);
6213 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6214 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6215 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6216 vector signed short vec_xor (vector signed short, vector signed short);
6217 vector unsigned short vec_xor (vector signed short,
6218 vector unsigned short);
6219 vector unsigned short vec_xor (vector unsigned short,
6220 vector signed short);
6221 vector unsigned short vec_xor (vector unsigned short,
6222 vector unsigned short);
6223 vector signed char vec_xor (vector signed char, vector signed char);
6224 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6226 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6228 vector unsigned char vec_xor (vector unsigned char,
6229 vector unsigned char);
6231 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6233 vector signed int vec_all_eq (vector signed char, vector signed char);
6234 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6236 vector signed int vec_all_eq (vector unsigned char,
6237 vector unsigned char);
6238 vector signed int vec_all_eq (vector signed short,
6239 vector unsigned short);
6240 vector signed int vec_all_eq (vector signed short, vector signed short);
6242 vector signed int vec_all_eq (vector unsigned short,
6243 vector signed short);
6244 vector signed int vec_all_eq (vector unsigned short,
6245 vector unsigned short);
6246 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6247 vector signed int vec_all_eq (vector signed int, vector signed int);
6248 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6249 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6251 vector signed int vec_all_eq (vector float, vector float);
6253 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6255 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6257 vector signed int vec_all_ge (vector unsigned char,
6258 vector unsigned char);
6259 vector signed int vec_all_ge (vector signed char, vector signed char);
6260 vector signed int vec_all_ge (vector signed short,
6261 vector unsigned short);
6262 vector signed int vec_all_ge (vector unsigned short,
6263 vector signed short);
6264 vector signed int vec_all_ge (vector unsigned short,
6265 vector unsigned short);
6266 vector signed int vec_all_ge (vector signed short, vector signed short);
6268 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6269 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6270 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6272 vector signed int vec_all_ge (vector signed int, vector signed int);
6273 vector signed int vec_all_ge (vector float, vector float);
6275 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6277 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6279 vector signed int vec_all_gt (vector unsigned char,
6280 vector unsigned char);
6281 vector signed int vec_all_gt (vector signed char, vector signed char);
6282 vector signed int vec_all_gt (vector signed short,
6283 vector unsigned short);
6284 vector signed int vec_all_gt (vector unsigned short,
6285 vector signed short);
6286 vector signed int vec_all_gt (vector unsigned short,
6287 vector unsigned short);
6288 vector signed int vec_all_gt (vector signed short, vector signed short);
6290 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6291 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6292 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6294 vector signed int vec_all_gt (vector signed int, vector signed int);
6295 vector signed int vec_all_gt (vector float, vector float);
6297 vector signed int vec_all_in (vector float, vector float);
6299 vector signed int vec_all_le (vector signed char, vector unsigned char);
6301 vector signed int vec_all_le (vector unsigned char, vector signed char);
6303 vector signed int vec_all_le (vector unsigned char,
6304 vector unsigned char);
6305 vector signed int vec_all_le (vector signed char, vector signed char);
6306 vector signed int vec_all_le (vector signed short,
6307 vector unsigned short);
6308 vector signed int vec_all_le (vector unsigned short,
6309 vector signed short);
6310 vector signed int vec_all_le (vector unsigned short,
6311 vector unsigned short);
6312 vector signed int vec_all_le (vector signed short, vector signed short);
6314 vector signed int vec_all_le (vector signed int, vector unsigned int);
6315 vector signed int vec_all_le (vector unsigned int, vector signed int);
6316 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
6318 vector signed int vec_all_le (vector signed int, vector signed int);
6319 vector signed int vec_all_le (vector float, vector float);
6321 vector signed int vec_all_lt (vector signed char, vector unsigned char);
6323 vector signed int vec_all_lt (vector unsigned char, vector signed char);
6325 vector signed int vec_all_lt (vector unsigned char,
6326 vector unsigned char);
6327 vector signed int vec_all_lt (vector signed char, vector signed char);
6328 vector signed int vec_all_lt (vector signed short,
6329 vector unsigned short);
6330 vector signed int vec_all_lt (vector unsigned short,
6331 vector signed short);
6332 vector signed int vec_all_lt (vector unsigned short,
6333 vector unsigned short);
6334 vector signed int vec_all_lt (vector signed short, vector signed short);
6336 vector signed int vec_all_lt (vector signed int, vector unsigned int);
6337 vector signed int vec_all_lt (vector unsigned int, vector signed int);
6338 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
6340 vector signed int vec_all_lt (vector signed int, vector signed int);
6341 vector signed int vec_all_lt (vector float, vector float);
6343 vector signed int vec_all_nan (vector float);
6345 vector signed int vec_all_ne (vector signed char, vector unsigned char);
6347 vector signed int vec_all_ne (vector signed char, vector signed char);
6348 vector signed int vec_all_ne (vector unsigned char, vector signed char);
6350 vector signed int vec_all_ne (vector unsigned char,
6351 vector unsigned char);
6352 vector signed int vec_all_ne (vector signed short,
6353 vector unsigned short);
6354 vector signed int vec_all_ne (vector signed short, vector signed short);
6356 vector signed int vec_all_ne (vector unsigned short,
6357 vector signed short);
6358 vector signed int vec_all_ne (vector unsigned short,
6359 vector unsigned short);
6360 vector signed int vec_all_ne (vector signed int, vector unsigned int);
6361 vector signed int vec_all_ne (vector signed int, vector signed int);
6362 vector signed int vec_all_ne (vector unsigned int, vector signed int);
6363 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
6365 vector signed int vec_all_ne (vector float, vector float);
6367 vector signed int vec_all_nge (vector float, vector float);
6369 vector signed int vec_all_ngt (vector float, vector float);
6371 vector signed int vec_all_nle (vector float, vector float);
6373 vector signed int vec_all_nlt (vector float, vector float);
6375 vector signed int vec_all_numeric (vector float);
6377 vector signed int vec_any_eq (vector signed char, vector unsigned char);
6379 vector signed int vec_any_eq (vector signed char, vector signed char);
6380 vector signed int vec_any_eq (vector unsigned char, vector signed char);
6382 vector signed int vec_any_eq (vector unsigned char,
6383 vector unsigned char);
6384 vector signed int vec_any_eq (vector signed short,
6385 vector unsigned short);
6386 vector signed int vec_any_eq (vector signed short, vector signed short);
6388 vector signed int vec_any_eq (vector unsigned short,
6389 vector signed short);
6390 vector signed int vec_any_eq (vector unsigned short,
6391 vector unsigned short);
6392 vector signed int vec_any_eq (vector signed int, vector unsigned int);
6393 vector signed int vec_any_eq (vector signed int, vector signed int);
6394 vector signed int vec_any_eq (vector unsigned int, vector signed int);
6395 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
6397 vector signed int vec_any_eq (vector float, vector float);
6399 vector signed int vec_any_ge (vector signed char, vector unsigned char);
6401 vector signed int vec_any_ge (vector unsigned char, vector signed char);
6403 vector signed int vec_any_ge (vector unsigned char,
6404 vector unsigned char);
6405 vector signed int vec_any_ge (vector signed char, vector signed char);
6406 vector signed int vec_any_ge (vector signed short,
6407 vector unsigned short);
6408 vector signed int vec_any_ge (vector unsigned short,
6409 vector signed short);
6410 vector signed int vec_any_ge (vector unsigned short,
6411 vector unsigned short);
6412 vector signed int vec_any_ge (vector signed short, vector signed short);
6414 vector signed int vec_any_ge (vector signed int, vector unsigned int);
6415 vector signed int vec_any_ge (vector unsigned int, vector signed int);
6416 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
6418 vector signed int vec_any_ge (vector signed int, vector signed int);
6419 vector signed int vec_any_ge (vector float, vector float);
6421 vector signed int vec_any_gt (vector signed char, vector unsigned char);
6423 vector signed int vec_any_gt (vector unsigned char, vector signed char);
6425 vector signed int vec_any_gt (vector unsigned char,
6426 vector unsigned char);
6427 vector signed int vec_any_gt (vector signed char, vector signed char);
6428 vector signed int vec_any_gt (vector signed short,
6429 vector unsigned short);
6430 vector signed int vec_any_gt (vector unsigned short,
6431 vector signed short);
6432 vector signed int vec_any_gt (vector unsigned short,
6433 vector unsigned short);
6434 vector signed int vec_any_gt (vector signed short, vector signed short);
6436 vector signed int vec_any_gt (vector signed int, vector unsigned int);
6437 vector signed int vec_any_gt (vector unsigned int, vector signed int);
6438 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
6440 vector signed int vec_any_gt (vector signed int, vector signed int);
6441 vector signed int vec_any_gt (vector float, vector float);
6443 vector signed int vec_any_le (vector signed char, vector unsigned char);
6445 vector signed int vec_any_le (vector unsigned char, vector signed char);
6447 vector signed int vec_any_le (vector unsigned char,
6448 vector unsigned char);
6449 vector signed int vec_any_le (vector signed char, vector signed char);
6450 vector signed int vec_any_le (vector signed short,
6451 vector unsigned short);
6452 vector signed int vec_any_le (vector unsigned short,
6453 vector signed short);
6454 vector signed int vec_any_le (vector unsigned short,
6455 vector unsigned short);
6456 vector signed int vec_any_le (vector signed short, vector signed short);
6458 vector signed int vec_any_le (vector signed int, vector unsigned int);
6459 vector signed int vec_any_le (vector unsigned int, vector signed int);
6460 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
6462 vector signed int vec_any_le (vector signed int, vector signed int);
6463 vector signed int vec_any_le (vector float, vector float);
6465 vector signed int vec_any_lt (vector signed char, vector unsigned char);
6467 vector signed int vec_any_lt (vector unsigned char, vector signed char);
6469 vector signed int vec_any_lt (vector unsigned char,
6470 vector unsigned char);
6471 vector signed int vec_any_lt (vector signed char, vector signed char);
6472 vector signed int vec_any_lt (vector signed short,
6473 vector unsigned short);
6474 vector signed int vec_any_lt (vector unsigned short,
6475 vector signed short);
6476 vector signed int vec_any_lt (vector unsigned short,
6477 vector unsigned short);
6478 vector signed int vec_any_lt (vector signed short, vector signed short);
6480 vector signed int vec_any_lt (vector signed int, vector unsigned int);
6481 vector signed int vec_any_lt (vector unsigned int, vector signed int);
6482 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
6484 vector signed int vec_any_lt (vector signed int, vector signed int);
6485 vector signed int vec_any_lt (vector float, vector float);
6487 vector signed int vec_any_nan (vector float);
6489 vector signed int vec_any_ne (vector signed char, vector unsigned char);
6491 vector signed int vec_any_ne (vector signed char, vector signed char);
6492 vector signed int vec_any_ne (vector unsigned char, vector signed char);
6494 vector signed int vec_any_ne (vector unsigned char,
6495 vector unsigned char);
6496 vector signed int vec_any_ne (vector signed short,
6497 vector unsigned short);
6498 vector signed int vec_any_ne (vector signed short, vector signed short);
6500 vector signed int vec_any_ne (vector unsigned short,
6501 vector signed short);
6502 vector signed int vec_any_ne (vector unsigned short,
6503 vector unsigned short);
6504 vector signed int vec_any_ne (vector signed int, vector unsigned int);
6505 vector signed int vec_any_ne (vector signed int, vector signed int);
6506 vector signed int vec_any_ne (vector unsigned int, vector signed int);
6507 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
6509 vector signed int vec_any_ne (vector float, vector float);
6511 vector signed int vec_any_nge (vector float, vector float);
6513 vector signed int vec_any_ngt (vector float, vector float);
6515 vector signed int vec_any_nle (vector float, vector float);
6517 vector signed int vec_any_nlt (vector float, vector float);
6519 vector signed int vec_any_numeric (vector float);
6521 vector signed int vec_any_out (vector float, vector float);
6525 @section Pragmas Accepted by GCC
6529 GCC supports several types of pragmas, primarily in order to compile
6530 code originally written for other compilers. Note that in general
6531 we do not recommend the use of pragmas; @xref{Function Attributes},
6532 for further explanation.
6536 * RS/6000 and PowerPC Pragmas::
6543 @subsection ARM Pragmas
6545 The ARM target defines pragmas for controlling the default addition of
6546 @code{long_call} and @code{short_call} attributes to functions.
6547 @xref{Function Attributes}, for information about the effects of these
6552 @cindex pragma, long_calls
6553 Set all subsequent functions to have the @code{long_call} attribute.
6556 @cindex pragma, no_long_calls
6557 Set all subsequent functions to have the @code{short_call} attribute.
6559 @item long_calls_off
6560 @cindex pragma, long_calls_off
6561 Do not affect the @code{long_call} or @code{short_call} attributes of
6562 subsequent functions.
6565 @node RS/6000 and PowerPC Pragmas
6566 @subsection RS/6000 and PowerPC Pragmas
6568 The RS/6000 and PowerPC targets define one pragma for controlling
6569 whether or not the @code{longcall} attribute is added to function
6570 declarations by default. This pragma overrides the @option{-mlongcall}
6571 option, but not the @code{longcall} and @code{shortcall} attributes.
6572 @xref{RS/6000 and PowerPC Options}, for more information about when long
6573 calls are and are not necessary.
6577 @cindex pragma, longcall
6578 Apply the @code{longcall} attribute to all subsequent function
6582 Do not apply the @code{longcall} attribute to subsequent function
6586 @c Describe c4x pragmas here.
6587 @c Describe h8300 pragmas here.
6588 @c Describe i370 pragmas here.
6589 @c Describe i960 pragmas here.
6590 @c Describe sh pragmas here.
6591 @c Describe v850 pragmas here.
6593 @node Darwin Pragmas
6594 @subsection Darwin Pragmas
6596 The following pragmas are available for all architectures running the
6597 Darwin operating system. These are useful for compatibility with other
6601 @item mark @var{tokens}@dots{}
6602 @cindex pragma, mark
6603 This pragma is accepted, but has no effect.
6605 @item options align=@var{alignment}
6606 @cindex pragma, options align
6607 This pragma sets the alignment of fields in structures. The values of
6608 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
6609 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
6610 properly; to restore the previous setting, use @code{reset} for the
6613 @item segment @var{tokens}@dots{}
6614 @cindex pragma, segment
6615 This pragma is accepted, but has no effect.
6617 @item unused (@var{var} [, @var{var}]@dots{})
6618 @cindex pragma, unused
6619 This pragma declares variables to be possibly unused. GCC will not
6620 produce warnings for the listed variables. The effect is similar to
6621 that of the @code{unused} attribute, except that this pragma may appear
6622 anywhere within the variables' scopes.
6625 @node Solaris Pragmas
6626 @subsection Solaris Pragmas
6628 For compatibility with the SunPRO compiler, the following pragma
6632 @item redefine_extname @var{oldname} @var{newname}
6633 @cindex pragma, redefine_extname
6635 This pragma gives the C function @var{oldname} the assembler label
6636 @var{newname}. The pragma must appear before the function declaration.
6637 This pragma is equivalent to the asm labels extension (@pxref{Asm
6638 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
6639 if the pragma is available.
6643 @subsection Tru64 Pragmas
6645 For compatibility with the Compaq C compiler, the following pragma
6649 @item extern_prefix @var{string}
6650 @cindex pragma, extern_prefix
6652 This pragma renames all subsequent function and variable declarations
6653 such that @var{string} is prepended to the name. This effect may be
6654 terminated by using another @code{extern_prefix} pragma with the
6657 This pragma is similar in intent to to the asm labels extension
6658 (@pxref{Asm Labels}) in that the system programmer wants to change
6659 the assembly-level ABI without changing the source-level API. The
6660 preprocessor defines @code{__PRAGMA_EXTERN_PREFIX} if the pragma is
6664 @node Unnamed Fields
6665 @section Unnamed struct/union fields within structs/unions.
6669 For compatibility with other compilers, GCC allows you to define
6670 a structure or union that contains, as fields, structures and unions
6671 without names. For example:
6684 In this example, the user would be able to access members of the unnamed
6685 union with code like @samp{foo.b}. Note that only unnamed structs and
6686 unions are allowed, you may not have, for example, an unnamed
6689 You must never create such structures that cause ambiguous field definitions.
6690 For example, this structure:
6701 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
6702 Such constructs are not supported and must be avoided. In the future,
6703 such constructs may be detected and treated as compilation errors.
6706 @section Thread-Local Storage
6707 @cindex Thread-Local Storage
6708 @cindex @acronym{TLS}
6711 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
6712 are allocated such that there is one instance of the variable per extant
6713 thread. The run-time model GCC uses to implement this originates
6714 in the IA-64 processor-specific ABI, but has since been migrated
6715 to other processors as well. It requires significant support from
6716 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
6717 system libraries (@file{libc.so} and @file{libpthread.so}), so it
6718 is not available everywhere.
6720 At the user level, the extension is visible with a new storage
6721 class keyword: @code{__thread}. For example:
6725 extern __thread struct state s;
6726 static __thread char *p;
6729 The @code{__thread} specifier may be used alone, with the @code{extern}
6730 or @code{static} specifiers, but with no other storage class specifier.
6731 When used with @code{extern} or @code{static}, @code{__thread} must appear
6732 immediately after the other storage class specifier.
6734 The @code{__thread} specifier may be applied to any global, file-scoped
6735 static, function-scoped static, or static data member of a class. It may
6736 not be applied to block-scoped automatic or non-static data member.
6738 When the address-of operator is applied to a thread-local variable, it is
6739 evaluated at run-time and returns the address of the current thread's
6740 instance of that variable. An address so obtained may be used by any
6741 thread. When a thread terminates, any pointers to thread-local variables
6742 in that thread become invalid.
6744 No static initialization may refer to the address of a thread-local variable.
6746 In C++, if an initializer is present for a thread-local variable, it must
6747 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
6750 See @uref{http://people.redhat.com/drepper/tls.pdf,
6751 ELF Handling For Thread-Local Storage} for a detailed explanation of
6752 the four thread-local storage addressing models, and how the run-time
6753 is expected to function.
6756 * C99 Thread-Local Edits::
6757 * C++98 Thread-Local Edits::
6760 @node C99 Thread-Local Edits
6761 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
6763 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
6764 that document the exact semantics of the language extension.
6768 @cite{5.1.2 Execution environments}
6770 Add new text after paragraph 1
6773 Within either execution environment, a @dfn{thread} is a flow of
6774 control within a program. It is implementation defined whether
6775 or not there may be more than one thread associated with a program.
6776 It is implementation defined how threads beyond the first are
6777 created, the name and type of the function called at thread
6778 startup, and how threads may be terminated. However, objects
6779 with thread storage duration shall be initialized before thread
6784 @cite{6.2.4 Storage durations of objects}
6786 Add new text before paragraph 3
6789 An object whose identifier is declared with the storage-class
6790 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
6791 Its lifetime is the entire execution of the thread, and its
6792 stored value is initialized only once, prior to thread startup.
6796 @cite{6.4.1 Keywords}
6798 Add @code{__thread}.
6801 @cite{6.7.1 Storage-class specifiers}
6803 Add @code{__thread} to the list of storage class specifiers in
6806 Change paragraph 2 to
6809 With the exception of @code{__thread}, at most one storage-class
6810 specifier may be given [@dots{}]. The @code{__thread} specifier may
6811 be used alone, or immediately following @code{extern} or
6815 Add new text after paragraph 6
6818 The declaration of an identifier for a variable that has
6819 block scope that specifies @code{__thread} shall also
6820 specify either @code{extern} or @code{static}.
6822 The @code{__thread} specifier shall be used only with
6827 @node C++98 Thread-Local Edits
6828 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
6830 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
6831 that document the exact semantics of the language extension.
6835 @b{[intro.execution]}
6837 New text after paragraph 4
6840 A @dfn{thread} is a flow of control within the abstract machine.
6841 It is implementation defined whether or not there may be more than
6845 New text after paragraph 7
6848 It is unspecified whether additional action must be taken to
6849 ensure when and whether side effects are visible to other threads.
6855 Add @code{__thread}.
6858 @b{[basic.start.main]}
6860 Add after paragraph 5
6863 The thread that begins execution at the @code{main} function is called
6864 the @dfn{main thread}. It is implementation defined how functions
6865 beginning threads other than the main thread are designated or typed.
6866 A function so designated, as well as the @code{main} function, is called
6867 a @dfn{thread startup function}. It is implementation defined what
6868 happens if a thread startup function returns. It is implementation
6869 defined what happens to other threads when any thread calls @code{exit}.
6873 @b{[basic.start.init]}
6875 Add after paragraph 4
6878 The storage for an object of thread storage duration shall be
6879 statically initialized before the first statement of the thread startup
6880 function. An object of thread storage duration shall not require
6881 dynamic initialization.
6885 @b{[basic.start.term]}
6887 Add after paragraph 3
6890 The type of an object with thread storage duration shall not have a
6891 non-trivial destructor, nor shall it be an array type whose elements
6892 (directly or indirectly) have non-trivial destructors.
6898 Add ``thread storage duration'' to the list in paragraph 1.
6903 Thread, static, and automatic storage durations are associated with
6904 objects introduced by declarations [@dots{}].
6907 Add @code{__thread} to the list of specifiers in paragraph 3.
6910 @b{[basic.stc.thread]}
6912 New section before @b{[basic.stc.static]}
6915 The keyword @code{__thread} applied to an non-local object gives the
6916 object thread storage duration.
6918 A local variable or class data member declared both @code{static}
6919 and @code{__thread} gives the variable or member thread storage
6924 @b{[basic.stc.static]}
6929 All objects which have neither thread storage duration, dynamic
6930 storage duration nor are local [@dots{}].
6936 Add @code{__thread} to the list in paragraph 1.
6941 With the exception of @code{__thread}, at most one
6942 @var{storage-class-specifier} shall appear in a given
6943 @var{decl-specifier-seq}. The @code{__thread} specifier may
6944 be used alone, or immediately following the @code{extern} or
6945 @code{static} specifiers. [@dots{}]
6948 Add after paragraph 5
6951 The @code{__thread} specifier can be applied only to the names of objects
6952 and to anonymous unions.
6958 Add after paragraph 6
6961 Non-@code{static} members shall not be @code{__thread}.
6965 @node C++ Extensions
6966 @chapter Extensions to the C++ Language
6967 @cindex extensions, C++ language
6968 @cindex C++ language extensions
6970 The GNU compiler provides these extensions to the C++ language (and you
6971 can also use most of the C language extensions in your C++ programs). If you
6972 want to write code that checks whether these features are available, you can
6973 test for the GNU compiler the same way as for C programs: check for a
6974 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
6975 test specifically for GNU C++ (@pxref{Standard Predefined,,Standard
6976 Predefined Macros,cpp.info,The C Preprocessor}).
6979 * Min and Max:: C++ Minimum and maximum operators.
6980 * Volatiles:: What constitutes an access to a volatile object.
6981 * Restricted Pointers:: C99 restricted pointers and references.
6982 * Vague Linkage:: Where G++ puts inlines, vtables and such.
6983 * C++ Interface:: You can use a single C++ header file for both
6984 declarations and definitions.
6985 * Template Instantiation:: Methods for ensuring that exactly one copy of
6986 each needed template instantiation is emitted.
6987 * Bound member functions:: You can extract a function pointer to the
6988 method denoted by a @samp{->*} or @samp{.*} expression.
6989 * C++ Attributes:: Variable, function, and type attributes for C++ only.
6990 * Java Exceptions:: Tweaking exception handling to work with Java.
6991 * Deprecated Features:: Things might disappear from g++.
6992 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
6996 @section Minimum and Maximum Operators in C++
6998 It is very convenient to have operators which return the ``minimum'' or the
6999 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
7002 @item @var{a} <? @var{b}
7004 @cindex minimum operator
7005 is the @dfn{minimum}, returning the smaller of the numeric values
7006 @var{a} and @var{b};
7008 @item @var{a} >? @var{b}
7010 @cindex maximum operator
7011 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
7015 These operations are not primitive in ordinary C++, since you can
7016 use a macro to return the minimum of two things in C++, as in the
7020 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
7024 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
7025 the minimum value of variables @var{i} and @var{j}.
7027 However, side effects in @code{X} or @code{Y} may cause unintended
7028 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
7029 the smaller counter twice. The GNU C @code{typeof} extension allows you
7030 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
7031 However, writing @code{MIN} and @code{MAX} as macros also forces you to
7032 use function-call notation for a fundamental arithmetic operation.
7033 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
7036 Since @code{<?} and @code{>?} are built into the compiler, they properly
7037 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
7041 @section When is a Volatile Object Accessed?
7042 @cindex accessing volatiles
7043 @cindex volatile read
7044 @cindex volatile write
7045 @cindex volatile access
7047 Both the C and C++ standard have the concept of volatile objects. These
7048 are normally accessed by pointers and used for accessing hardware. The
7049 standards encourage compilers to refrain from optimizations
7050 concerning accesses to volatile objects that it might perform on
7051 non-volatile objects. The C standard leaves it implementation defined
7052 as to what constitutes a volatile access. The C++ standard omits to
7053 specify this, except to say that C++ should behave in a similar manner
7054 to C with respect to volatiles, where possible. The minimum either
7055 standard specifies is that at a sequence point all previous accesses to
7056 volatile objects have stabilized and no subsequent accesses have
7057 occurred. Thus an implementation is free to reorder and combine
7058 volatile accesses which occur between sequence points, but cannot do so
7059 for accesses across a sequence point. The use of volatiles does not
7060 allow you to violate the restriction on updating objects multiple times
7061 within a sequence point.
7063 In most expressions, it is intuitively obvious what is a read and what is
7064 a write. For instance
7067 volatile int *dst = @var{somevalue};
7068 volatile int *src = @var{someothervalue};
7073 will cause a read of the volatile object pointed to by @var{src} and stores the
7074 value into the volatile object pointed to by @var{dst}. There is no
7075 guarantee that these reads and writes are atomic, especially for objects
7076 larger than @code{int}.
7078 Less obvious expressions are where something which looks like an access
7079 is used in a void context. An example would be,
7082 volatile int *src = @var{somevalue};
7086 With C, such expressions are rvalues, and as rvalues cause a read of
7087 the object, GCC interprets this as a read of the volatile being pointed
7088 to. The C++ standard specifies that such expressions do not undergo
7089 lvalue to rvalue conversion, and that the type of the dereferenced
7090 object may be incomplete. The C++ standard does not specify explicitly
7091 that it is this lvalue to rvalue conversion which is responsible for
7092 causing an access. However, there is reason to believe that it is,
7093 because otherwise certain simple expressions become undefined. However,
7094 because it would surprise most programmers, G++ treats dereferencing a
7095 pointer to volatile object of complete type in a void context as a read
7096 of the object. When the object has incomplete type, G++ issues a
7101 struct T @{int m;@};
7102 volatile S *ptr1 = @var{somevalue};
7103 volatile T *ptr2 = @var{somevalue};
7108 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
7109 causes a read of the object pointed to. If you wish to force an error on
7110 the first case, you must force a conversion to rvalue with, for instance
7111 a static cast, @code{static_cast<S>(*ptr1)}.
7113 When using a reference to volatile, G++ does not treat equivalent
7114 expressions as accesses to volatiles, but instead issues a warning that
7115 no volatile is accessed. The rationale for this is that otherwise it
7116 becomes difficult to determine where volatile access occur, and not
7117 possible to ignore the return value from functions returning volatile
7118 references. Again, if you wish to force a read, cast the reference to
7121 @node Restricted Pointers
7122 @section Restricting Pointer Aliasing
7123 @cindex restricted pointers
7124 @cindex restricted references
7125 @cindex restricted this pointer
7127 As with gcc, g++ understands the C99 feature of restricted pointers,
7128 specified with the @code{__restrict__}, or @code{__restrict} type
7129 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
7130 language flag, @code{restrict} is not a keyword in C++.
7132 In addition to allowing restricted pointers, you can specify restricted
7133 references, which indicate that the reference is not aliased in the local
7137 void fn (int *__restrict__ rptr, int &__restrict__ rref)
7144 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
7145 @var{rref} refers to a (different) unaliased integer.
7147 You may also specify whether a member function's @var{this} pointer is
7148 unaliased by using @code{__restrict__} as a member function qualifier.
7151 void T::fn () __restrict__
7158 Within the body of @code{T::fn}, @var{this} will have the effective
7159 definition @code{T *__restrict__ const this}. Notice that the
7160 interpretation of a @code{__restrict__} member function qualifier is
7161 different to that of @code{const} or @code{volatile} qualifier, in that it
7162 is applied to the pointer rather than the object. This is consistent with
7163 other compilers which implement restricted pointers.
7165 As with all outermost parameter qualifiers, @code{__restrict__} is
7166 ignored in function definition matching. This means you only need to
7167 specify @code{__restrict__} in a function definition, rather than
7168 in a function prototype as well.
7171 @section Vague Linkage
7172 @cindex vague linkage
7174 There are several constructs in C++ which require space in the object
7175 file but are not clearly tied to a single translation unit. We say that
7176 these constructs have ``vague linkage''. Typically such constructs are
7177 emitted wherever they are needed, though sometimes we can be more
7181 @item Inline Functions
7182 Inline functions are typically defined in a header file which can be
7183 included in many different compilations. Hopefully they can usually be
7184 inlined, but sometimes an out-of-line copy is necessary, if the address
7185 of the function is taken or if inlining fails. In general, we emit an
7186 out-of-line copy in all translation units where one is needed. As an
7187 exception, we only emit inline virtual functions with the vtable, since
7188 it will always require a copy.
7190 Local static variables and string constants used in an inline function
7191 are also considered to have vague linkage, since they must be shared
7192 between all inlined and out-of-line instances of the function.
7196 C++ virtual functions are implemented in most compilers using a lookup
7197 table, known as a vtable. The vtable contains pointers to the virtual
7198 functions provided by a class, and each object of the class contains a
7199 pointer to its vtable (or vtables, in some multiple-inheritance
7200 situations). If the class declares any non-inline, non-pure virtual
7201 functions, the first one is chosen as the ``key method'' for the class,
7202 and the vtable is only emitted in the translation unit where the key
7205 @emph{Note:} If the chosen key method is later defined as inline, the
7206 vtable will still be emitted in every translation unit which defines it.
7207 Make sure that any inline virtuals are declared inline in the class
7208 body, even if they are not defined there.
7210 @item type_info objects
7213 C++ requires information about types to be written out in order to
7214 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7215 For polymorphic classes (classes with virtual functions), the type_info
7216 object is written out along with the vtable so that @samp{dynamic_cast}
7217 can determine the dynamic type of a class object at runtime. For all
7218 other types, we write out the type_info object when it is used: when
7219 applying @samp{typeid} to an expression, throwing an object, or
7220 referring to a type in a catch clause or exception specification.
7222 @item Template Instantiations
7223 Most everything in this section also applies to template instantiations,
7224 but there are other options as well.
7225 @xref{Template Instantiation,,Where's the Template?}.
7229 When used with GNU ld version 2.8 or later on an ELF system such as
7230 Linux/GNU or Solaris 2, or on Microsoft Windows, duplicate copies of
7231 these constructs will be discarded at link time. This is known as
7234 On targets that don't support COMDAT, but do support weak symbols, GCC
7235 will use them. This way one copy will override all the others, but
7236 the unused copies will still take up space in the executable.
7238 For targets which do not support either COMDAT or weak symbols,
7239 most entities with vague linkage will be emitted as local symbols to
7240 avoid duplicate definition errors from the linker. This will not happen
7241 for local statics in inlines, however, as having multiple copies will
7242 almost certainly break things.
7244 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7245 another way to control placement of these constructs.
7248 @section Declarations and Definitions in One Header
7250 @cindex interface and implementation headers, C++
7251 @cindex C++ interface and implementation headers
7252 C++ object definitions can be quite complex. In principle, your source
7253 code will need two kinds of things for each object that you use across
7254 more than one source file. First, you need an @dfn{interface}
7255 specification, describing its structure with type declarations and
7256 function prototypes. Second, you need the @dfn{implementation} itself.
7257 It can be tedious to maintain a separate interface description in a
7258 header file, in parallel to the actual implementation. It is also
7259 dangerous, since separate interface and implementation definitions may
7260 not remain parallel.
7262 @cindex pragmas, interface and implementation
7263 With GNU C++, you can use a single header file for both purposes.
7266 @emph{Warning:} The mechanism to specify this is in transition. For the
7267 nonce, you must use one of two @code{#pragma} commands; in a future
7268 release of GNU C++, an alternative mechanism will make these
7269 @code{#pragma} commands unnecessary.
7272 The header file contains the full definitions, but is marked with
7273 @samp{#pragma interface} in the source code. This allows the compiler
7274 to use the header file only as an interface specification when ordinary
7275 source files incorporate it with @code{#include}. In the single source
7276 file where the full implementation belongs, you can use either a naming
7277 convention or @samp{#pragma implementation} to indicate this alternate
7278 use of the header file.
7281 @item #pragma interface
7282 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7283 @kindex #pragma interface
7284 Use this directive in @emph{header files} that define object classes, to save
7285 space in most of the object files that use those classes. Normally,
7286 local copies of certain information (backup copies of inline member
7287 functions, debugging information, and the internal tables that implement
7288 virtual functions) must be kept in each object file that includes class
7289 definitions. You can use this pragma to avoid such duplication. When a
7290 header file containing @samp{#pragma interface} is included in a
7291 compilation, this auxiliary information will not be generated (unless
7292 the main input source file itself uses @samp{#pragma implementation}).
7293 Instead, the object files will contain references to be resolved at link
7296 The second form of this directive is useful for the case where you have
7297 multiple headers with the same name in different directories. If you
7298 use this form, you must specify the same string to @samp{#pragma
7301 @item #pragma implementation
7302 @itemx #pragma implementation "@var{objects}.h"
7303 @kindex #pragma implementation
7304 Use this pragma in a @emph{main input file}, when you want full output from
7305 included header files to be generated (and made globally visible). The
7306 included header file, in turn, should use @samp{#pragma interface}.
7307 Backup copies of inline member functions, debugging information, and the
7308 internal tables used to implement virtual functions are all generated in
7309 implementation files.
7311 @cindex implied @code{#pragma implementation}
7312 @cindex @code{#pragma implementation}, implied
7313 @cindex naming convention, implementation headers
7314 If you use @samp{#pragma implementation} with no argument, it applies to
7315 an include file with the same basename@footnote{A file's @dfn{basename}
7316 was the name stripped of all leading path information and of trailing
7317 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
7318 file. For example, in @file{allclass.cc}, giving just
7319 @samp{#pragma implementation}
7320 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
7322 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
7323 an implementation file whenever you would include it from
7324 @file{allclass.cc} even if you never specified @samp{#pragma
7325 implementation}. This was deemed to be more trouble than it was worth,
7326 however, and disabled.
7328 If you use an explicit @samp{#pragma implementation}, it must appear in
7329 your source file @emph{before} you include the affected header files.
7331 Use the string argument if you want a single implementation file to
7332 include code from multiple header files. (You must also use
7333 @samp{#include} to include the header file; @samp{#pragma
7334 implementation} only specifies how to use the file---it doesn't actually
7337 There is no way to split up the contents of a single header file into
7338 multiple implementation files.
7341 @cindex inlining and C++ pragmas
7342 @cindex C++ pragmas, effect on inlining
7343 @cindex pragmas in C++, effect on inlining
7344 @samp{#pragma implementation} and @samp{#pragma interface} also have an
7345 effect on function inlining.
7347 If you define a class in a header file marked with @samp{#pragma
7348 interface}, the effect on a function defined in that class is similar to
7349 an explicit @code{extern} declaration---the compiler emits no code at
7350 all to define an independent version of the function. Its definition
7351 is used only for inlining with its callers.
7353 @opindex fno-implement-inlines
7354 Conversely, when you include the same header file in a main source file
7355 that declares it as @samp{#pragma implementation}, the compiler emits
7356 code for the function itself; this defines a version of the function
7357 that can be found via pointers (or by callers compiled without
7358 inlining). If all calls to the function can be inlined, you can avoid
7359 emitting the function by compiling with @option{-fno-implement-inlines}.
7360 If any calls were not inlined, you will get linker errors.
7362 @node Template Instantiation
7363 @section Where's the Template?
7364 @cindex template instantiation
7366 C++ templates are the first language feature to require more
7367 intelligence from the environment than one usually finds on a UNIX
7368 system. Somehow the compiler and linker have to make sure that each
7369 template instance occurs exactly once in the executable if it is needed,
7370 and not at all otherwise. There are two basic approaches to this
7371 problem, which I will refer to as the Borland model and the Cfront model.
7375 Borland C++ solved the template instantiation problem by adding the code
7376 equivalent of common blocks to their linker; the compiler emits template
7377 instances in each translation unit that uses them, and the linker
7378 collapses them together. The advantage of this model is that the linker
7379 only has to consider the object files themselves; there is no external
7380 complexity to worry about. This disadvantage is that compilation time
7381 is increased because the template code is being compiled repeatedly.
7382 Code written for this model tends to include definitions of all
7383 templates in the header file, since they must be seen to be
7387 The AT&T C++ translator, Cfront, solved the template instantiation
7388 problem by creating the notion of a template repository, an
7389 automatically maintained place where template instances are stored. A
7390 more modern version of the repository works as follows: As individual
7391 object files are built, the compiler places any template definitions and
7392 instantiations encountered in the repository. At link time, the link
7393 wrapper adds in the objects in the repository and compiles any needed
7394 instances that were not previously emitted. The advantages of this
7395 model are more optimal compilation speed and the ability to use the
7396 system linker; to implement the Borland model a compiler vendor also
7397 needs to replace the linker. The disadvantages are vastly increased
7398 complexity, and thus potential for error; for some code this can be
7399 just as transparent, but in practice it can been very difficult to build
7400 multiple programs in one directory and one program in multiple
7401 directories. Code written for this model tends to separate definitions
7402 of non-inline member templates into a separate file, which should be
7403 compiled separately.
7406 When used with GNU ld version 2.8 or later on an ELF system such as
7407 Linux/GNU or Solaris 2, or on Microsoft Windows, g++ supports the
7408 Borland model. On other systems, g++ implements neither automatic
7411 A future version of g++ will support a hybrid model whereby the compiler
7412 will emit any instantiations for which the template definition is
7413 included in the compile, and store template definitions and
7414 instantiation context information into the object file for the rest.
7415 The link wrapper will extract that information as necessary and invoke
7416 the compiler to produce the remaining instantiations. The linker will
7417 then combine duplicate instantiations.
7419 In the mean time, you have the following options for dealing with
7420 template instantiations:
7425 Compile your template-using code with @option{-frepo}. The compiler will
7426 generate files with the extension @samp{.rpo} listing all of the
7427 template instantiations used in the corresponding object files which
7428 could be instantiated there; the link wrapper, @samp{collect2}, will
7429 then update the @samp{.rpo} files to tell the compiler where to place
7430 those instantiations and rebuild any affected object files. The
7431 link-time overhead is negligible after the first pass, as the compiler
7432 will continue to place the instantiations in the same files.
7434 This is your best option for application code written for the Borland
7435 model, as it will just work. Code written for the Cfront model will
7436 need to be modified so that the template definitions are available at
7437 one or more points of instantiation; usually this is as simple as adding
7438 @code{#include <tmethods.cc>} to the end of each template header.
7440 For library code, if you want the library to provide all of the template
7441 instantiations it needs, just try to link all of its object files
7442 together; the link will fail, but cause the instantiations to be
7443 generated as a side effect. Be warned, however, that this may cause
7444 conflicts if multiple libraries try to provide the same instantiations.
7445 For greater control, use explicit instantiation as described in the next
7449 @opindex fno-implicit-templates
7450 Compile your code with @option{-fno-implicit-templates} to disable the
7451 implicit generation of template instances, and explicitly instantiate
7452 all the ones you use. This approach requires more knowledge of exactly
7453 which instances you need than do the others, but it's less
7454 mysterious and allows greater control. You can scatter the explicit
7455 instantiations throughout your program, perhaps putting them in the
7456 translation units where the instances are used or the translation units
7457 that define the templates themselves; you can put all of the explicit
7458 instantiations you need into one big file; or you can create small files
7465 template class Foo<int>;
7466 template ostream& operator <<
7467 (ostream&, const Foo<int>&);
7470 for each of the instances you need, and create a template instantiation
7473 If you are using Cfront-model code, you can probably get away with not
7474 using @option{-fno-implicit-templates} when compiling files that don't
7475 @samp{#include} the member template definitions.
7477 If you use one big file to do the instantiations, you may want to
7478 compile it without @option{-fno-implicit-templates} so you get all of the
7479 instances required by your explicit instantiations (but not by any
7480 other files) without having to specify them as well.
7482 g++ has extended the template instantiation syntax given in the ISO
7483 standard to allow forward declaration of explicit instantiations
7484 (with @code{extern}), instantiation of the compiler support data for a
7485 template class (i.e.@: the vtable) without instantiating any of its
7486 members (with @code{inline}), and instantiation of only the static data
7487 members of a template class, without the support data or member
7488 functions (with (@code{static}):
7491 extern template int max (int, int);
7492 inline template class Foo<int>;
7493 static template class Foo<int>;
7497 Do nothing. Pretend g++ does implement automatic instantiation
7498 management. Code written for the Borland model will work fine, but
7499 each translation unit will contain instances of each of the templates it
7500 uses. In a large program, this can lead to an unacceptable amount of code
7503 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7504 more discussion of these pragmas.
7507 @node Bound member functions
7508 @section Extracting the function pointer from a bound pointer to member function
7510 @cindex pointer to member function
7511 @cindex bound pointer to member function
7513 In C++, pointer to member functions (PMFs) are implemented using a wide
7514 pointer of sorts to handle all the possible call mechanisms; the PMF
7515 needs to store information about how to adjust the @samp{this} pointer,
7516 and if the function pointed to is virtual, where to find the vtable, and
7517 where in the vtable to look for the member function. If you are using
7518 PMFs in an inner loop, you should really reconsider that decision. If
7519 that is not an option, you can extract the pointer to the function that
7520 would be called for a given object/PMF pair and call it directly inside
7521 the inner loop, to save a bit of time.
7523 Note that you will still be paying the penalty for the call through a
7524 function pointer; on most modern architectures, such a call defeats the
7525 branch prediction features of the CPU@. This is also true of normal
7526 virtual function calls.
7528 The syntax for this extension is
7532 extern int (A::*fp)();
7533 typedef int (*fptr)(A *);
7535 fptr p = (fptr)(a.*fp);
7538 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
7539 no object is needed to obtain the address of the function. They can be
7540 converted to function pointers directly:
7543 fptr p1 = (fptr)(&A::foo);
7546 @opindex Wno-pmf-conversions
7547 You must specify @option{-Wno-pmf-conversions} to use this extension.
7549 @node C++ Attributes
7550 @section C++-Specific Variable, Function, and Type Attributes
7552 Some attributes only make sense for C++ programs.
7555 @item init_priority (@var{priority})
7556 @cindex init_priority attribute
7559 In Standard C++, objects defined at namespace scope are guaranteed to be
7560 initialized in an order in strict accordance with that of their definitions
7561 @emph{in a given translation unit}. No guarantee is made for initializations
7562 across translation units. However, GNU C++ allows users to control the
7563 order of initialization of objects defined at namespace scope with the
7564 @code{init_priority} attribute by specifying a relative @var{priority},
7565 a constant integral expression currently bounded between 101 and 65535
7566 inclusive. Lower numbers indicate a higher priority.
7568 In the following example, @code{A} would normally be created before
7569 @code{B}, but the @code{init_priority} attribute has reversed that order:
7572 Some_Class A __attribute__ ((init_priority (2000)));
7573 Some_Class B __attribute__ ((init_priority (543)));
7577 Note that the particular values of @var{priority} do not matter; only their
7580 @item java_interface
7581 @cindex java_interface attribute
7583 This type attribute informs C++ that the class is a Java interface. It may
7584 only be applied to classes declared within an @code{extern "Java"} block.
7585 Calls to methods declared in this interface will be dispatched using GCJ's
7586 interface table mechanism, instead of regular virtual table dispatch.
7590 @node Java Exceptions
7591 @section Java Exceptions
7593 The Java language uses a slightly different exception handling model
7594 from C++. Normally, GNU C++ will automatically detect when you are
7595 writing C++ code that uses Java exceptions, and handle them
7596 appropriately. However, if C++ code only needs to execute destructors
7597 when Java exceptions are thrown through it, GCC will guess incorrectly.
7598 Sample problematic code is:
7601 struct S @{ ~S(); @};
7602 extern void bar(); // is written in Java, and may throw exceptions
7611 The usual effect of an incorrect guess is a link failure, complaining of
7612 a missing routine called @samp{__gxx_personality_v0}.
7614 You can inform the compiler that Java exceptions are to be used in a
7615 translation unit, irrespective of what it might think, by writing
7616 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
7617 @samp{#pragma} must appear before any functions that throw or catch
7618 exceptions, or run destructors when exceptions are thrown through them.
7620 You cannot mix Java and C++ exceptions in the same translation unit. It
7621 is believed to be safe to throw a C++ exception from one file through
7622 another file compiled for the Java exception model, or vice versa, but
7623 there may be bugs in this area.
7625 @node Deprecated Features
7626 @section Deprecated Features
7628 In the past, the GNU C++ compiler was extended to experiment with new
7629 features, at a time when the C++ language was still evolving. Now that
7630 the C++ standard is complete, some of those features are superseded by
7631 superior alternatives. Using the old features might cause a warning in
7632 some cases that the feature will be dropped in the future. In other
7633 cases, the feature might be gone already.
7635 While the list below is not exhaustive, it documents some of the options
7636 that are now deprecated:
7639 @item -fexternal-templates
7640 @itemx -falt-external-templates
7641 These are two of the many ways for g++ to implement template
7642 instantiation. @xref{Template Instantiation}. The C++ standard clearly
7643 defines how template definitions have to be organized across
7644 implementation units. g++ has an implicit instantiation mechanism that
7645 should work just fine for standard-conforming code.
7647 @item -fstrict-prototype
7648 @itemx -fno-strict-prototype
7649 Previously it was possible to use an empty prototype parameter list to
7650 indicate an unspecified number of parameters (like C), rather than no
7651 parameters, as C++ demands. This feature has been removed, except where
7652 it is required for backwards compatibility @xref{Backwards Compatibility}.
7655 The named return value extension has been deprecated, and is now
7658 The use of initializer lists with new expressions has been deprecated,
7659 and is now removed from g++.
7661 Floating and complex non-type template parameters have been deprecated,
7662 and are now removed from g++.
7664 The implicit typename extension has been deprecated and will be removed
7665 from g++ at some point. In some cases g++ determines that a dependent
7666 type such as @code{TPL<T>::X} is a type without needing a
7667 @code{typename} keyword, contrary to the standard.
7669 @node Backwards Compatibility
7670 @section Backwards Compatibility
7671 @cindex Backwards Compatibility
7672 @cindex ARM [Annotated C++ Reference Manual]
7674 Now that there is a definitive ISO standard C++, G++ has a specification
7675 to adhere to. The C++ language evolved over time, and features that
7676 used to be acceptable in previous drafts of the standard, such as the ARM
7677 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
7678 compilation of C++ written to such drafts, G++ contains some backwards
7679 compatibilities. @emph{All such backwards compatibility features are
7680 liable to disappear in future versions of G++.} They should be considered
7681 deprecated @xref{Deprecated Features}.
7685 If a variable is declared at for scope, it used to remain in scope until
7686 the end of the scope which contained the for statement (rather than just
7687 within the for scope). G++ retains this, but issues a warning, if such a
7688 variable is accessed outside the for scope.
7690 @item Implicit C language
7691 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
7692 scope to set the language. On such systems, all header files are
7693 implicitly scoped inside a C language scope. Also, an empty prototype
7694 @code{()} will be treated as an unspecified number of arguments, rather
7695 than no arguments, as C++ demands.