1 @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002,2003,2004
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 block.
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 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
433 * Long Long:: Double-word integers---@code{long long int}.
434 * Complex:: Data types for complex numbers.
435 * Hex Floats:: Hexadecimal floating-point constants.
436 * Zero Length:: Zero-length arrays.
437 * Variable Length:: Arrays whose length is computed at run time.
438 * Empty Structures:: Structures with no members.
439 * Variadic Macros:: Macros with a variable number of arguments.
440 * Escaped Newlines:: Slightly looser rules for escaped newlines.
441 * Subscripting:: Any array can be subscripted, even if not an lvalue.
442 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
443 * Initializers:: Non-constant initializers.
444 * Compound Literals:: Compound literals give structures, unions
446 * Designated Inits:: Labeling elements of initializers.
447 * Cast to Union:: Casting to union type from any member of the union.
448 * Case Ranges:: `case 1 ... 9' and such.
449 * Mixed Declarations:: Mixing declarations and code.
450 * Function Attributes:: Declaring that functions have no side effects,
451 or that they can never return.
452 * Attribute Syntax:: Formal syntax for attributes.
453 * Function Prototypes:: Prototype declarations and old-style definitions.
454 * C++ Comments:: C++ comments are recognized.
455 * Dollar Signs:: Dollar sign is allowed in identifiers.
456 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
457 * Variable Attributes:: Specifying attributes of variables.
458 * Type Attributes:: Specifying attributes of types.
459 * Alignment:: Inquiring about the alignment of a type or variable.
460 * Inline:: Defining inline functions (as fast as macros).
461 * Extended Asm:: Assembler instructions with C expressions as operands.
462 (With them you can define ``built-in'' functions.)
463 * Constraints:: Constraints for asm operands
464 * Asm Labels:: Specifying the assembler name to use for a C symbol.
465 * Explicit Reg Vars:: Defining variables residing in specified registers.
466 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
467 * Incomplete Enums:: @code{enum foo;}, with details to follow.
468 * Function Names:: Printable strings which are the name of the current
470 * Return Address:: Getting the return or frame address of a function.
471 * Vector Extensions:: Using vector instructions through built-in functions.
472 * Offsetof:: Special syntax for implementing @code{offsetof}.
473 * Other Builtins:: Other built-in functions.
474 * Target Builtins:: Built-in functions specific to particular targets.
475 * Pragmas:: Pragmas accepted by GCC.
476 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
477 * Thread-Local:: Per-thread variables.
480 @node Statement Exprs
481 @section Statements and Declarations in Expressions
482 @cindex statements inside expressions
483 @cindex declarations inside expressions
484 @cindex expressions containing statements
485 @cindex macros, statements in expressions
487 @c the above section title wrapped and causes an underfull hbox.. i
488 @c changed it from "within" to "in". --mew 4feb93
489 A compound statement enclosed in parentheses may appear as an expression
490 in GNU C@. This allows you to use loops, switches, and local variables
491 within an expression.
493 Recall that a compound statement is a sequence of statements surrounded
494 by braces; in this construct, parentheses go around the braces. For
498 (@{ int y = foo (); int z;
505 is a valid (though slightly more complex than necessary) expression
506 for the absolute value of @code{foo ()}.
508 The last thing in the compound statement should be an expression
509 followed by a semicolon; the value of this subexpression serves as the
510 value of the entire construct. (If you use some other kind of statement
511 last within the braces, the construct has type @code{void}, and thus
512 effectively no value.)
514 This feature is especially useful in making macro definitions ``safe'' (so
515 that they evaluate each operand exactly once). For example, the
516 ``maximum'' function is commonly defined as a macro in standard C as
520 #define max(a,b) ((a) > (b) ? (a) : (b))
524 @cindex side effects, macro argument
525 But this definition computes either @var{a} or @var{b} twice, with bad
526 results if the operand has side effects. In GNU C, if you know the
527 type of the operands (here let's assume @code{int}), you can define
528 the macro safely as follows:
531 #define maxint(a,b) \
532 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
535 Embedded statements are not allowed in constant expressions, such as
536 the value of an enumeration constant, the width of a bit-field, or
537 the initial value of a static variable.
539 If you don't know the type of the operand, you can still do this, but you
540 must use @code{typeof} (@pxref{Typeof}).
542 In G++, the result value of a statement expression undergoes array and
543 function pointer decay, and is returned by value to the enclosing
544 expression. For instance, if @code{A} is a class, then
553 will construct a temporary @code{A} object to hold the result of the
554 statement expression, and that will be used to invoke @code{Foo}.
555 Therefore the @code{this} pointer observed by @code{Foo} will not be the
558 Any temporaries created within a statement within a statement expression
559 will be destroyed at the statement's end. This makes statement
560 expressions inside macros slightly different from function calls. In
561 the latter case temporaries introduced during argument evaluation will
562 be destroyed at the end of the statement that includes the function
563 call. In the statement expression case they will be destroyed during
564 the statement expression. For instance,
567 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
568 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
578 will have different places where temporaries are destroyed. For the
579 @code{macro} case, the temporary @code{X} will be destroyed just after
580 the initialization of @code{b}. In the @code{function} case that
581 temporary will be destroyed when the function returns.
583 These considerations mean that it is probably a bad idea to use
584 statement-expressions of this form in header files that are designed to
585 work with C++. (Note that some versions of the GNU C Library contained
586 header files using statement-expression that lead to precisely this
590 @section Locally Declared Labels
592 @cindex macros, local labels
594 GCC allows you to declare @dfn{local labels} in any nested block
595 scope. A local label is just like an ordinary label, but you can
596 only reference it (with a @code{goto} statement, or by taking its
597 address) within the block in which it was declared.
599 A local label declaration looks like this:
602 __label__ @var{label};
609 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
612 Local label declarations must come at the beginning of the block,
613 before any ordinary declarations or statements.
615 The label declaration defines the label @emph{name}, but does not define
616 the label itself. You must do this in the usual way, with
617 @code{@var{label}:}, within the statements of the statement expression.
619 The local label feature is useful for complex macros. If a macro
620 contains nested loops, a @code{goto} can be useful for breaking out of
621 them. However, an ordinary label whose scope is the whole function
622 cannot be used: if the macro can be expanded several times in one
623 function, the label will be multiply defined in that function. A
624 local label avoids this problem. For example:
627 #define SEARCH(value, array, target) \
630 typeof (target) _SEARCH_target = (target); \
631 typeof (*(array)) *_SEARCH_array = (array); \
634 for (i = 0; i < max; i++) \
635 for (j = 0; j < max; j++) \
636 if (_SEARCH_array[i][j] == _SEARCH_target) \
637 @{ (value) = i; goto found; @} \
643 This could also be written using a statement-expression:
646 #define SEARCH(array, target) \
649 typeof (target) _SEARCH_target = (target); \
650 typeof (*(array)) *_SEARCH_array = (array); \
653 for (i = 0; i < max; i++) \
654 for (j = 0; j < max; j++) \
655 if (_SEARCH_array[i][j] == _SEARCH_target) \
656 @{ value = i; goto found; @} \
663 Local label declarations also make the labels they declare visible to
664 nested functions, if there are any. @xref{Nested Functions}, for details.
666 @node Labels as Values
667 @section Labels as Values
668 @cindex labels as values
669 @cindex computed gotos
670 @cindex goto with computed label
671 @cindex address of a label
673 You can get the address of a label defined in the current function
674 (or a containing function) with the unary operator @samp{&&}. The
675 value has type @code{void *}. This value is a constant and can be used
676 wherever a constant of that type is valid. For example:
684 To use these values, you need to be able to jump to one. This is done
685 with the computed goto statement@footnote{The analogous feature in
686 Fortran is called an assigned goto, but that name seems inappropriate in
687 C, where one can do more than simply store label addresses in label
688 variables.}, @code{goto *@var{exp};}. For example,
695 Any expression of type @code{void *} is allowed.
697 One way of using these constants is in initializing a static array that
698 will serve as a jump table:
701 static void *array[] = @{ &&foo, &&bar, &&hack @};
704 Then you can select a label with indexing, like this:
711 Note that this does not check whether the subscript is in bounds---array
712 indexing in C never does that.
714 Such an array of label values serves a purpose much like that of the
715 @code{switch} statement. The @code{switch} statement is cleaner, so
716 use that rather than an array unless the problem does not fit a
717 @code{switch} statement very well.
719 Another use of label values is in an interpreter for threaded code.
720 The labels within the interpreter function can be stored in the
721 threaded code for super-fast dispatching.
723 You may not use this mechanism to jump to code in a different function.
724 If you do that, totally unpredictable things will happen. The best way to
725 avoid this is to store the label address only in automatic variables and
726 never pass it as an argument.
728 An alternate way to write the above example is
731 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
733 goto *(&&foo + array[i]);
737 This is more friendly to code living in shared libraries, as it reduces
738 the number of dynamic relocations that are needed, and by consequence,
739 allows the data to be read-only.
741 @node Nested Functions
742 @section Nested Functions
743 @cindex nested functions
744 @cindex downward funargs
747 A @dfn{nested function} is a function defined inside another function.
748 (Nested functions are not supported for GNU C++.) The nested function's
749 name is local to the block where it is defined. For example, here we
750 define a nested function named @code{square}, and call it twice:
754 foo (double a, double b)
756 double square (double z) @{ return z * z; @}
758 return square (a) + square (b);
763 The nested function can access all the variables of the containing
764 function that are visible at the point of its definition. This is
765 called @dfn{lexical scoping}. For example, here we show a nested
766 function which uses an inherited variable named @code{offset}:
770 bar (int *array, int offset, int size)
772 int access (int *array, int index)
773 @{ return array[index + offset]; @}
776 for (i = 0; i < size; i++)
777 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
782 Nested function definitions are permitted within functions in the places
783 where variable definitions are allowed; that is, in any block, before
784 the first statement in the block.
786 It is possible to call the nested function from outside the scope of its
787 name by storing its address or passing the address to another function:
790 hack (int *array, int size)
792 void store (int index, int value)
793 @{ array[index] = value; @}
795 intermediate (store, size);
799 Here, the function @code{intermediate} receives the address of
800 @code{store} as an argument. If @code{intermediate} calls @code{store},
801 the arguments given to @code{store} are used to store into @code{array}.
802 But this technique works only so long as the containing function
803 (@code{hack}, in this example) does not exit.
805 If you try to call the nested function through its address after the
806 containing function has exited, all hell will break loose. If you try
807 to call it after a containing scope level has exited, and if it refers
808 to some of the variables that are no longer in scope, you may be lucky,
809 but it's not wise to take the risk. If, however, the nested function
810 does not refer to anything that has gone out of scope, you should be
813 GCC implements taking the address of a nested function using a technique
814 called @dfn{trampolines}. A paper describing them is available as
817 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
819 A nested function can jump to a label inherited from a containing
820 function, provided the label was explicitly declared in the containing
821 function (@pxref{Local Labels}). Such a jump returns instantly to the
822 containing function, exiting the nested function which did the
823 @code{goto} and any intermediate functions as well. Here is an example:
827 bar (int *array, int offset, int size)
830 int access (int *array, int index)
834 return array[index + offset];
838 for (i = 0; i < size; i++)
839 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
843 /* @r{Control comes here from @code{access}
844 if it detects an error.} */
851 A nested function always has internal linkage. Declaring one with
852 @code{extern} is erroneous. If you need to declare the nested function
853 before its definition, use @code{auto} (which is otherwise meaningless
854 for function declarations).
857 bar (int *array, int offset, int size)
860 auto int access (int *, int);
862 int access (int *array, int index)
866 return array[index + offset];
872 @node Constructing Calls
873 @section Constructing Function Calls
874 @cindex constructing calls
875 @cindex forwarding calls
877 Using the built-in functions described below, you can record
878 the arguments a function received, and call another function
879 with the same arguments, without knowing the number or types
882 You can also record the return value of that function call,
883 and later return that value, without knowing what data type
884 the function tried to return (as long as your caller expects
887 However, these built-in functions may interact badly with some
888 sophisticated features or other extensions of the language. It
889 is, therefore, not recommended to use them outside very simple
890 functions acting as mere forwarders for their arguments.
892 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
893 This built-in function returns a pointer to data
894 describing how to perform a call with the same arguments as were passed
895 to the current function.
897 The function saves the arg pointer register, structure value address,
898 and all registers that might be used to pass arguments to a function
899 into a block of memory allocated on the stack. Then it returns the
900 address of that block.
903 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
904 This built-in function invokes @var{function}
905 with a copy of the parameters described by @var{arguments}
908 The value of @var{arguments} should be the value returned by
909 @code{__builtin_apply_args}. The argument @var{size} specifies the size
910 of the stack argument data, in bytes.
912 This function returns a pointer to data describing
913 how to return whatever value was returned by @var{function}. The data
914 is saved in a block of memory allocated on the stack.
916 It is not always simple to compute the proper value for @var{size}. The
917 value is used by @code{__builtin_apply} to compute the amount of data
918 that should be pushed on the stack and copied from the incoming argument
922 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
923 This built-in function returns the value described by @var{result} from
924 the containing function. You should specify, for @var{result}, a value
925 returned by @code{__builtin_apply}.
929 @section Referring to a Type with @code{typeof}
932 @cindex macros, types of arguments
934 Another way to refer to the type of an expression is with @code{typeof}.
935 The syntax of using of this keyword looks like @code{sizeof}, but the
936 construct acts semantically like a type name defined with @code{typedef}.
938 There are two ways of writing the argument to @code{typeof}: with an
939 expression or with a type. Here is an example with an expression:
946 This assumes that @code{x} is an array of pointers to functions;
947 the type described is that of the values of the functions.
949 Here is an example with a typename as the argument:
956 Here the type described is that of pointers to @code{int}.
958 If you are writing a header file that must work when included in ISO C
959 programs, write @code{__typeof__} instead of @code{typeof}.
960 @xref{Alternate Keywords}.
962 A @code{typeof}-construct can be used anywhere a typedef name could be
963 used. For example, you can use it in a declaration, in a cast, or inside
964 of @code{sizeof} or @code{typeof}.
966 @code{typeof} is often useful in conjunction with the
967 statements-within-expressions feature. Here is how the two together can
968 be used to define a safe ``maximum'' macro that operates on any
969 arithmetic type and evaluates each of its arguments exactly once:
973 (@{ typeof (a) _a = (a); \
974 typeof (b) _b = (b); \
975 _a > _b ? _a : _b; @})
978 @cindex underscores in variables in macros
979 @cindex @samp{_} in variables in macros
980 @cindex local variables in macros
981 @cindex variables, local, in macros
982 @cindex macros, local variables in
984 The reason for using names that start with underscores for the local
985 variables is to avoid conflicts with variable names that occur within the
986 expressions that are substituted for @code{a} and @code{b}. Eventually we
987 hope to design a new form of declaration syntax that allows you to declare
988 variables whose scopes start only after their initializers; this will be a
989 more reliable way to prevent such conflicts.
992 Some more examples of the use of @code{typeof}:
996 This declares @code{y} with the type of what @code{x} points to.
1003 This declares @code{y} as an array of such values.
1010 This declares @code{y} as an array of pointers to characters:
1013 typeof (typeof (char *)[4]) y;
1017 It is equivalent to the following traditional C declaration:
1023 To see the meaning of the declaration using @code{typeof}, and why it
1024 might be a useful way to write, let's rewrite it with these macros:
1027 #define pointer(T) typeof(T *)
1028 #define array(T, N) typeof(T [N])
1032 Now the declaration can be rewritten this way:
1035 array (pointer (char), 4) y;
1039 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1040 pointers to @code{char}.
1043 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1044 a more limited extension which permitted one to write
1047 typedef @var{T} = @var{expr};
1051 with the effect of declaring @var{T} to have the type of the expression
1052 @var{expr}. This extension does not work with GCC 3 (versions between
1053 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1054 relies on it should be rewritten to use @code{typeof}:
1057 typedef typeof(@var{expr}) @var{T};
1061 This will work with all versions of GCC@.
1064 @section Conditionals with Omitted Operands
1065 @cindex conditional expressions, extensions
1066 @cindex omitted middle-operands
1067 @cindex middle-operands, omitted
1068 @cindex extensions, @code{?:}
1069 @cindex @code{?:} extensions
1071 The middle operand in a conditional expression may be omitted. Then
1072 if the first operand is nonzero, its value is the value of the conditional
1075 Therefore, the expression
1082 has the value of @code{x} if that is nonzero; otherwise, the value of
1085 This example is perfectly equivalent to
1091 @cindex side effect in ?:
1092 @cindex ?: side effect
1094 In this simple case, the ability to omit the middle operand is not
1095 especially useful. When it becomes useful is when the first operand does,
1096 or may (if it is a macro argument), contain a side effect. Then repeating
1097 the operand in the middle would perform the side effect twice. Omitting
1098 the middle operand uses the value already computed without the undesirable
1099 effects of recomputing it.
1102 @section Double-Word Integers
1103 @cindex @code{long long} data types
1104 @cindex double-word arithmetic
1105 @cindex multiprecision arithmetic
1106 @cindex @code{LL} integer suffix
1107 @cindex @code{ULL} integer suffix
1109 ISO C99 supports data types for integers that are at least 64 bits wide,
1110 and as an extension GCC supports them in C89 mode and in C++.
1111 Simply write @code{long long int} for a signed integer, or
1112 @code{unsigned long long int} for an unsigned integer. To make an
1113 integer constant of type @code{long long int}, add the suffix @samp{LL}
1114 to the integer. To make an integer constant of type @code{unsigned long
1115 long int}, add the suffix @samp{ULL} to the integer.
1117 You can use these types in arithmetic like any other integer types.
1118 Addition, subtraction, and bitwise boolean operations on these types
1119 are open-coded on all types of machines. Multiplication is open-coded
1120 if the machine supports fullword-to-doubleword a widening multiply
1121 instruction. Division and shifts are open-coded only on machines that
1122 provide special support. The operations that are not open-coded use
1123 special library routines that come with GCC@.
1125 There may be pitfalls when you use @code{long long} types for function
1126 arguments, unless you declare function prototypes. If a function
1127 expects type @code{int} for its argument, and you pass a value of type
1128 @code{long long int}, confusion will result because the caller and the
1129 subroutine will disagree about the number of bytes for the argument.
1130 Likewise, if the function expects @code{long long int} and you pass
1131 @code{int}. The best way to avoid such problems is to use prototypes.
1134 @section Complex Numbers
1135 @cindex complex numbers
1136 @cindex @code{_Complex} keyword
1137 @cindex @code{__complex__} keyword
1139 ISO C99 supports complex floating data types, and as an extension GCC
1140 supports them in C89 mode and in C++, and supports complex integer data
1141 types which are not part of ISO C99. You can declare complex types
1142 using the keyword @code{_Complex}. As an extension, the older GNU
1143 keyword @code{__complex__} is also supported.
1145 For example, @samp{_Complex double x;} declares @code{x} as a
1146 variable whose real part and imaginary part are both of type
1147 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1148 have real and imaginary parts of type @code{short int}; this is not
1149 likely to be useful, but it shows that the set of complex types is
1152 To write a constant with a complex data type, use the suffix @samp{i} or
1153 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1154 has type @code{_Complex float} and @code{3i} has type
1155 @code{_Complex int}. Such a constant always has a pure imaginary
1156 value, but you can form any complex value you like by adding one to a
1157 real constant. This is a GNU extension; if you have an ISO C99
1158 conforming C library (such as GNU libc), and want to construct complex
1159 constants of floating type, you should include @code{<complex.h>} and
1160 use the macros @code{I} or @code{_Complex_I} instead.
1162 @cindex @code{__real__} keyword
1163 @cindex @code{__imag__} keyword
1164 To extract the real part of a complex-valued expression @var{exp}, write
1165 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1166 extract the imaginary part. This is a GNU extension; for values of
1167 floating type, you should use the ISO C99 functions @code{crealf},
1168 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1169 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1170 built-in functions by GCC@.
1172 @cindex complex conjugation
1173 The operator @samp{~} performs complex conjugation when used on a value
1174 with a complex type. This is a GNU extension; for values of
1175 floating type, you should use the ISO C99 functions @code{conjf},
1176 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1177 provided as built-in functions by GCC@.
1179 GCC can allocate complex automatic variables in a noncontiguous
1180 fashion; it's even possible for the real part to be in a register while
1181 the imaginary part is on the stack (or vice-versa). Only the DWARF2
1182 debug info format can represent this, so use of DWARF2 is recommended.
1183 If you are using the stabs debug info format, GCC describes a noncontiguous
1184 complex variable as if it were two separate variables of noncomplex type.
1185 If the variable's actual name is @code{foo}, the two fictitious
1186 variables are named @code{foo$real} and @code{foo$imag}. You can
1187 examine and set these two fictitious variables with your debugger.
1193 ISO C99 supports floating-point numbers written not only in the usual
1194 decimal notation, such as @code{1.55e1}, but also numbers such as
1195 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1196 supports this in C89 mode (except in some cases when strictly
1197 conforming) and in C++. In that format the
1198 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1199 mandatory. The exponent is a decimal number that indicates the power of
1200 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1207 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1208 is the same as @code{1.55e1}.
1210 Unlike for floating-point numbers in the decimal notation the exponent
1211 is always required in the hexadecimal notation. Otherwise the compiler
1212 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1213 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1214 extension for floating-point constants of type @code{float}.
1217 @section Arrays of Length Zero
1218 @cindex arrays of length zero
1219 @cindex zero-length arrays
1220 @cindex length-zero arrays
1221 @cindex flexible array members
1223 Zero-length arrays are allowed in GNU C@. They are very useful as the
1224 last element of a structure which is really a header for a variable-length
1233 struct line *thisline = (struct line *)
1234 malloc (sizeof (struct line) + this_length);
1235 thisline->length = this_length;
1238 In ISO C90, you would have to give @code{contents} a length of 1, which
1239 means either you waste space or complicate the argument to @code{malloc}.
1241 In ISO C99, you would use a @dfn{flexible array member}, which is
1242 slightly different in syntax and semantics:
1246 Flexible array members are written as @code{contents[]} without
1250 Flexible array members have incomplete type, and so the @code{sizeof}
1251 operator may not be applied. As a quirk of the original implementation
1252 of zero-length arrays, @code{sizeof} evaluates to zero.
1255 Flexible array members may only appear as the last member of a
1256 @code{struct} that is otherwise non-empty.
1259 A structure containing a flexible array member, or a union containing
1260 such a structure (possibly recursively), may not be a member of a
1261 structure or an element of an array. (However, these uses are
1262 permitted by GCC as extensions.)
1265 GCC versions before 3.0 allowed zero-length arrays to be statically
1266 initialized, as if they were flexible arrays. In addition to those
1267 cases that were useful, it also allowed initializations in situations
1268 that would corrupt later data. Non-empty initialization of zero-length
1269 arrays is now treated like any case where there are more initializer
1270 elements than the array holds, in that a suitable warning about "excess
1271 elements in array" is given, and the excess elements (all of them, in
1272 this case) are ignored.
1274 Instead GCC allows static initialization of flexible array members.
1275 This is equivalent to defining a new structure containing the original
1276 structure followed by an array of sufficient size to contain the data.
1277 I.e.@: in the following, @code{f1} is constructed as if it were declared
1283 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1286 struct f1 f1; int data[3];
1287 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1291 The convenience of this extension is that @code{f1} has the desired
1292 type, eliminating the need to consistently refer to @code{f2.f1}.
1294 This has symmetry with normal static arrays, in that an array of
1295 unknown size is also written with @code{[]}.
1297 Of course, this extension only makes sense if the extra data comes at
1298 the end of a top-level object, as otherwise we would be overwriting
1299 data at subsequent offsets. To avoid undue complication and confusion
1300 with initialization of deeply nested arrays, we simply disallow any
1301 non-empty initialization except when the structure is the top-level
1302 object. For example:
1305 struct foo @{ int x; int y[]; @};
1306 struct bar @{ struct foo z; @};
1308 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1309 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1310 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1311 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1314 @node Empty Structures
1315 @section Structures With No Members
1316 @cindex empty structures
1317 @cindex zero-size structures
1319 GCC permits a C structure to have no members:
1326 The structure will have size zero. In C++, empty structures are part
1327 of the language. G++ treats empty structures as if they had a single
1328 member of type @code{char}.
1330 @node Variable Length
1331 @section Arrays of Variable Length
1332 @cindex variable-length arrays
1333 @cindex arrays of variable length
1336 Variable-length automatic arrays are allowed in ISO C99, and as an
1337 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1338 implementation of variable-length arrays does not yet conform in detail
1339 to the ISO C99 standard.) These arrays are
1340 declared like any other automatic arrays, but with a length that is not
1341 a constant expression. The storage is allocated at the point of
1342 declaration and deallocated when the brace-level is exited. For
1347 concat_fopen (char *s1, char *s2, char *mode)
1349 char str[strlen (s1) + strlen (s2) + 1];
1352 return fopen (str, mode);
1356 @cindex scope of a variable length array
1357 @cindex variable-length array scope
1358 @cindex deallocating variable length arrays
1359 Jumping or breaking out of the scope of the array name deallocates the
1360 storage. Jumping into the scope is not allowed; you get an error
1363 @cindex @code{alloca} vs variable-length arrays
1364 You can use the function @code{alloca} to get an effect much like
1365 variable-length arrays. The function @code{alloca} is available in
1366 many other C implementations (but not in all). On the other hand,
1367 variable-length arrays are more elegant.
1369 There are other differences between these two methods. Space allocated
1370 with @code{alloca} exists until the containing @emph{function} returns.
1371 The space for a variable-length array is deallocated as soon as the array
1372 name's scope ends. (If you use both variable-length arrays and
1373 @code{alloca} in the same function, deallocation of a variable-length array
1374 will also deallocate anything more recently allocated with @code{alloca}.)
1376 You can also use variable-length arrays as arguments to functions:
1380 tester (int len, char data[len][len])
1386 The length of an array is computed once when the storage is allocated
1387 and is remembered for the scope of the array in case you access it with
1390 If you want to pass the array first and the length afterward, you can
1391 use a forward declaration in the parameter list---another GNU extension.
1395 tester (int len; char data[len][len], int len)
1401 @cindex parameter forward declaration
1402 The @samp{int len} before the semicolon is a @dfn{parameter forward
1403 declaration}, and it serves the purpose of making the name @code{len}
1404 known when the declaration of @code{data} is parsed.
1406 You can write any number of such parameter forward declarations in the
1407 parameter list. They can be separated by commas or semicolons, but the
1408 last one must end with a semicolon, which is followed by the ``real''
1409 parameter declarations. Each forward declaration must match a ``real''
1410 declaration in parameter name and data type. ISO C99 does not support
1411 parameter forward declarations.
1413 @node Variadic Macros
1414 @section Macros with a Variable Number of Arguments.
1415 @cindex variable number of arguments
1416 @cindex macro with variable arguments
1417 @cindex rest argument (in macro)
1418 @cindex variadic macros
1420 In the ISO C standard of 1999, a macro can be declared to accept a
1421 variable number of arguments much as a function can. The syntax for
1422 defining the macro is similar to that of a function. Here is an
1426 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1429 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1430 such a macro, it represents the zero or more tokens until the closing
1431 parenthesis that ends the invocation, including any commas. This set of
1432 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1433 wherever it appears. See the CPP manual for more information.
1435 GCC has long supported variadic macros, and used a different syntax that
1436 allowed you to give a name to the variable arguments just like any other
1437 argument. Here is an example:
1440 #define debug(format, args...) fprintf (stderr, format, args)
1443 This is in all ways equivalent to the ISO C example above, but arguably
1444 more readable and descriptive.
1446 GNU CPP has two further variadic macro extensions, and permits them to
1447 be used with either of the above forms of macro definition.
1449 In standard C, you are not allowed to leave the variable argument out
1450 entirely; but you are allowed to pass an empty argument. For example,
1451 this invocation is invalid in ISO C, because there is no comma after
1458 GNU CPP permits you to completely omit the variable arguments in this
1459 way. In the above examples, the compiler would complain, though since
1460 the expansion of the macro still has the extra comma after the format
1463 To help solve this problem, CPP behaves specially for variable arguments
1464 used with the token paste operator, @samp{##}. If instead you write
1467 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1470 and if the variable arguments are omitted or empty, the @samp{##}
1471 operator causes the preprocessor to remove the comma before it. If you
1472 do provide some variable arguments in your macro invocation, GNU CPP
1473 does not complain about the paste operation and instead places the
1474 variable arguments after the comma. Just like any other pasted macro
1475 argument, these arguments are not macro expanded.
1477 @node Escaped Newlines
1478 @section Slightly Looser Rules for Escaped Newlines
1479 @cindex escaped newlines
1480 @cindex newlines (escaped)
1482 Recently, the preprocessor has relaxed its treatment of escaped
1483 newlines. Previously, the newline had to immediately follow a
1484 backslash. The current implementation allows whitespace in the form
1485 of spaces, horizontal and vertical tabs, and form feeds between the
1486 backslash and the subsequent newline. The preprocessor issues a
1487 warning, but treats it as a valid escaped newline and combines the two
1488 lines to form a single logical line. This works within comments and
1489 tokens, as well as between tokens. Comments are @emph{not} treated as
1490 whitespace for the purposes of this relaxation, since they have not
1491 yet been replaced with spaces.
1494 @section Non-Lvalue Arrays May Have Subscripts
1495 @cindex subscripting
1496 @cindex arrays, non-lvalue
1498 @cindex subscripting and function values
1499 In ISO C99, arrays that are not lvalues still decay to pointers, and
1500 may be subscripted, although they may not be modified or used after
1501 the next sequence point and the unary @samp{&} operator may not be
1502 applied to them. As an extension, GCC allows such arrays to be
1503 subscripted in C89 mode, though otherwise they do not decay to
1504 pointers outside C99 mode. For example,
1505 this is valid in GNU C though not valid in C89:
1509 struct foo @{int a[4];@};
1515 return f().a[index];
1521 @section Arithmetic on @code{void}- and Function-Pointers
1522 @cindex void pointers, arithmetic
1523 @cindex void, size of pointer to
1524 @cindex function pointers, arithmetic
1525 @cindex function, size of pointer to
1527 In GNU C, addition and subtraction operations are supported on pointers to
1528 @code{void} and on pointers to functions. This is done by treating the
1529 size of a @code{void} or of a function as 1.
1531 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1532 and on function types, and returns 1.
1534 @opindex Wpointer-arith
1535 The option @option{-Wpointer-arith} requests a warning if these extensions
1539 @section Non-Constant Initializers
1540 @cindex initializers, non-constant
1541 @cindex non-constant initializers
1543 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1544 automatic variable are not required to be constant expressions in GNU C@.
1545 Here is an example of an initializer with run-time varying elements:
1548 foo (float f, float g)
1550 float beat_freqs[2] = @{ f-g, f+g @};
1555 @node Compound Literals
1556 @section Compound Literals
1557 @cindex constructor expressions
1558 @cindex initializations in expressions
1559 @cindex structures, constructor expression
1560 @cindex expressions, constructor
1561 @cindex compound literals
1562 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1564 ISO C99 supports compound literals. A compound literal looks like
1565 a cast containing an initializer. Its value is an object of the
1566 type specified in the cast, containing the elements specified in
1567 the initializer; it is an lvalue. As an extension, GCC supports
1568 compound literals in C89 mode and in C++.
1570 Usually, the specified type is a structure. Assume that
1571 @code{struct foo} and @code{structure} are declared as shown:
1574 struct foo @{int a; char b[2];@} structure;
1578 Here is an example of constructing a @code{struct foo} with a compound literal:
1581 structure = ((struct foo) @{x + y, 'a', 0@});
1585 This is equivalent to writing the following:
1589 struct foo temp = @{x + y, 'a', 0@};
1594 You can also construct an array. If all the elements of the compound literal
1595 are (made up of) simple constant expressions, suitable for use in
1596 initializers of objects of static storage duration, then the compound
1597 literal can be coerced to a pointer to its first element and used in
1598 such an initializer, as shown here:
1601 char **foo = (char *[]) @{ "x", "y", "z" @};
1604 Compound literals for scalar types and union types are is
1605 also allowed, but then the compound literal is equivalent
1608 As a GNU extension, GCC allows initialization of objects with static storage
1609 duration by compound literals (which is not possible in ISO C99, because
1610 the initializer is not a constant).
1611 It is handled as if the object was initialized only with the bracket
1612 enclosed list if compound literal's and object types match.
1613 The initializer list of the compound literal must be constant.
1614 If the object being initialized has array type of unknown size, the size is
1615 determined by compound literal size.
1618 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1619 static int y[] = (int []) @{1, 2, 3@};
1620 static int z[] = (int [3]) @{1@};
1624 The above lines are equivalent to the following:
1626 static struct foo x = @{1, 'a', 'b'@};
1627 static int y[] = @{1, 2, 3@};
1628 static int z[] = @{1, 0, 0@};
1631 @node Designated Inits
1632 @section Designated Initializers
1633 @cindex initializers with labeled elements
1634 @cindex labeled elements in initializers
1635 @cindex case labels in initializers
1636 @cindex designated initializers
1638 Standard C89 requires the elements of an initializer to appear in a fixed
1639 order, the same as the order of the elements in the array or structure
1642 In ISO C99 you can give the elements in any order, specifying the array
1643 indices or structure field names they apply to, and GNU C allows this as
1644 an extension in C89 mode as well. This extension is not
1645 implemented in GNU C++.
1647 To specify an array index, write
1648 @samp{[@var{index}] =} before the element value. For example,
1651 int a[6] = @{ [4] = 29, [2] = 15 @};
1658 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1662 The index values must be constant expressions, even if the array being
1663 initialized is automatic.
1665 An alternative syntax for this which has been obsolete since GCC 2.5 but
1666 GCC still accepts is to write @samp{[@var{index}]} before the element
1667 value, with no @samp{=}.
1669 To initialize a range of elements to the same value, write
1670 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1671 extension. For example,
1674 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1678 If the value in it has side-effects, the side-effects will happen only once,
1679 not for each initialized field by the range initializer.
1682 Note that the length of the array is the highest value specified
1685 In a structure initializer, specify the name of a field to initialize
1686 with @samp{.@var{fieldname} =} before the element value. For example,
1687 given the following structure,
1690 struct point @{ int x, y; @};
1694 the following initialization
1697 struct point p = @{ .y = yvalue, .x = xvalue @};
1704 struct point p = @{ xvalue, yvalue @};
1707 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1708 @samp{@var{fieldname}:}, as shown here:
1711 struct point p = @{ y: yvalue, x: xvalue @};
1715 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1716 @dfn{designator}. You can also use a designator (or the obsolete colon
1717 syntax) when initializing a union, to specify which element of the union
1718 should be used. For example,
1721 union foo @{ int i; double d; @};
1723 union foo f = @{ .d = 4 @};
1727 will convert 4 to a @code{double} to store it in the union using
1728 the second element. By contrast, casting 4 to type @code{union foo}
1729 would store it into the union as the integer @code{i}, since it is
1730 an integer. (@xref{Cast to Union}.)
1732 You can combine this technique of naming elements with ordinary C
1733 initialization of successive elements. Each initializer element that
1734 does not have a designator applies to the next consecutive element of the
1735 array or structure. For example,
1738 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1745 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1748 Labeling the elements of an array initializer is especially useful
1749 when the indices are characters or belong to an @code{enum} type.
1754 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1755 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1758 @cindex designator lists
1759 You can also write a series of @samp{.@var{fieldname}} and
1760 @samp{[@var{index}]} designators before an @samp{=} to specify a
1761 nested subobject to initialize; the list is taken relative to the
1762 subobject corresponding to the closest surrounding brace pair. For
1763 example, with the @samp{struct point} declaration above:
1766 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1770 If the same field is initialized multiple times, it will have value from
1771 the last initialization. If any such overridden initialization has
1772 side-effect, it is unspecified whether the side-effect happens or not.
1773 Currently, GCC will discard them and issue a warning.
1776 @section Case Ranges
1778 @cindex ranges in case statements
1780 You can specify a range of consecutive values in a single @code{case} label,
1784 case @var{low} ... @var{high}:
1788 This has the same effect as the proper number of individual @code{case}
1789 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1791 This feature is especially useful for ranges of ASCII character codes:
1797 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1798 it may be parsed wrong when you use it with integer values. For example,
1813 @section Cast to a Union Type
1814 @cindex cast to a union
1815 @cindex union, casting to a
1817 A cast to union type is similar to other casts, except that the type
1818 specified is a union type. You can specify the type either with
1819 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1820 a constructor though, not a cast, and hence does not yield an lvalue like
1821 normal casts. (@xref{Compound Literals}.)
1823 The types that may be cast to the union type are those of the members
1824 of the union. Thus, given the following union and variables:
1827 union foo @{ int i; double d; @};
1833 both @code{x} and @code{y} can be cast to type @code{union foo}.
1835 Using the cast as the right-hand side of an assignment to a variable of
1836 union type is equivalent to storing in a member of the union:
1841 u = (union foo) x @equiv{} u.i = x
1842 u = (union foo) y @equiv{} u.d = y
1845 You can also use the union cast as a function argument:
1848 void hack (union foo);
1850 hack ((union foo) x);
1853 @node Mixed Declarations
1854 @section Mixed Declarations and Code
1855 @cindex mixed declarations and code
1856 @cindex declarations, mixed with code
1857 @cindex code, mixed with declarations
1859 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1860 within compound statements. As an extension, GCC also allows this in
1861 C89 mode. For example, you could do:
1870 Each identifier is visible from where it is declared until the end of
1871 the enclosing block.
1873 @node Function Attributes
1874 @section Declaring Attributes of Functions
1875 @cindex function attributes
1876 @cindex declaring attributes of functions
1877 @cindex functions that never return
1878 @cindex functions that have no side effects
1879 @cindex functions in arbitrary sections
1880 @cindex functions that behave like malloc
1881 @cindex @code{volatile} applied to function
1882 @cindex @code{const} applied to function
1883 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1884 @cindex functions with non-null pointer arguments
1885 @cindex functions that are passed arguments in registers on the 386
1886 @cindex functions that pop the argument stack on the 386
1887 @cindex functions that do not pop the argument stack on the 386
1889 In GNU C, you declare certain things about functions called in your program
1890 which help the compiler optimize function calls and check your code more
1893 The keyword @code{__attribute__} allows you to specify special
1894 attributes when making a declaration. This keyword is followed by an
1895 attribute specification inside double parentheses. The following
1896 attributes are currently defined for functions on all targets:
1897 @code{noreturn}, @code{noinline}, @code{always_inline},
1898 @code{pure}, @code{const}, @code{nothrow},
1899 @code{format}, @code{format_arg}, @code{no_instrument_function},
1900 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1901 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1902 @code{alias}, @code{warn_unused_result} and @code{nonnull}. Several other
1903 attributes are defined for functions on particular target systems. Other
1904 attributes, including @code{section} are supported for variables declarations
1905 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1907 You may also specify attributes with @samp{__} preceding and following
1908 each keyword. This allows you to use them in header files without
1909 being concerned about a possible macro of the same name. For example,
1910 you may use @code{__noreturn__} instead of @code{noreturn}.
1912 @xref{Attribute Syntax}, for details of the exact syntax for using
1916 @cindex @code{noreturn} function attribute
1918 A few standard library functions, such as @code{abort} and @code{exit},
1919 cannot return. GCC knows this automatically. Some programs define
1920 their own functions that never return. You can declare them
1921 @code{noreturn} to tell the compiler this fact. For example,
1925 void fatal () __attribute__ ((noreturn));
1928 fatal (/* @r{@dots{}} */)
1930 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
1936 The @code{noreturn} keyword tells the compiler to assume that
1937 @code{fatal} cannot return. It can then optimize without regard to what
1938 would happen if @code{fatal} ever did return. This makes slightly
1939 better code. More importantly, it helps avoid spurious warnings of
1940 uninitialized variables.
1942 The @code{noreturn} keyword does not affect the exceptional path when that
1943 applies: a @code{noreturn}-marked function may still return to the caller
1944 by throwing an exception.
1946 Do not assume that registers saved by the calling function are
1947 restored before calling the @code{noreturn} function.
1949 It does not make sense for a @code{noreturn} function to have a return
1950 type other than @code{void}.
1952 The attribute @code{noreturn} is not implemented in GCC versions
1953 earlier than 2.5. An alternative way to declare that a function does
1954 not return, which works in the current version and in some older
1955 versions, is as follows:
1958 typedef void voidfn ();
1960 volatile voidfn fatal;
1963 @cindex @code{noinline} function attribute
1965 This function attribute prevents a function from being considered for
1968 @cindex @code{always_inline} function attribute
1970 Generally, functions are not inlined unless optimization is specified.
1971 For functions declared inline, this attribute inlines the function even
1972 if no optimization level was specified.
1974 @cindex @code{pure} function attribute
1976 Many functions have no effects except the return value and their
1977 return value depends only on the parameters and/or global variables.
1978 Such a function can be subject
1979 to common subexpression elimination and loop optimization just as an
1980 arithmetic operator would be. These functions should be declared
1981 with the attribute @code{pure}. For example,
1984 int square (int) __attribute__ ((pure));
1988 says that the hypothetical function @code{square} is safe to call
1989 fewer times than the program says.
1991 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
1992 Interesting non-pure functions are functions with infinite loops or those
1993 depending on volatile memory or other system resource, that may change between
1994 two consecutive calls (such as @code{feof} in a multithreading environment).
1996 The attribute @code{pure} is not implemented in GCC versions earlier
1998 @cindex @code{const} function attribute
2000 Many functions do not examine any values except their arguments, and
2001 have no effects except the return value. Basically this is just slightly
2002 more strict class than the @code{pure} attribute above, since function is not
2003 allowed to read global memory.
2005 @cindex pointer arguments
2006 Note that a function that has pointer arguments and examines the data
2007 pointed to must @emph{not} be declared @code{const}. Likewise, a
2008 function that calls a non-@code{const} function usually must not be
2009 @code{const}. It does not make sense for a @code{const} function to
2012 The attribute @code{const} is not implemented in GCC versions earlier
2013 than 2.5. An alternative way to declare that a function has no side
2014 effects, which works in the current version and in some older versions,
2018 typedef int intfn ();
2020 extern const intfn square;
2023 This approach does not work in GNU C++ from 2.6.0 on, since the language
2024 specifies that the @samp{const} must be attached to the return value.
2026 @cindex @code{nothrow} function attribute
2028 The @code{nothrow} attribute is used to inform the compiler that a
2029 function cannot throw an exception. For example, most functions in
2030 the standard C library can be guaranteed not to throw an exception
2031 with the notable exceptions of @code{qsort} and @code{bsearch} that
2032 take function pointer arguments. The @code{nothrow} attribute is not
2033 implemented in GCC versions earlier than 3.2.
2035 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2036 @cindex @code{format} function attribute
2038 The @code{format} attribute specifies that a function takes @code{printf},
2039 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2040 should be type-checked against a format string. For example, the
2045 my_printf (void *my_object, const char *my_format, ...)
2046 __attribute__ ((format (printf, 2, 3)));
2050 causes the compiler to check the arguments in calls to @code{my_printf}
2051 for consistency with the @code{printf} style format string argument
2054 The parameter @var{archetype} determines how the format string is
2055 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2056 or @code{strfmon}. (You can also use @code{__printf__},
2057 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2058 parameter @var{string-index} specifies which argument is the format
2059 string argument (starting from 1), while @var{first-to-check} is the
2060 number of the first argument to check against the format string. For
2061 functions where the arguments are not available to be checked (such as
2062 @code{vprintf}), specify the third parameter as zero. In this case the
2063 compiler only checks the format string for consistency. For
2064 @code{strftime} formats, the third parameter is required to be zero.
2065 Since non-static C++ methods have an implicit @code{this} argument, the
2066 arguments of such methods should be counted from two, not one, when
2067 giving values for @var{string-index} and @var{first-to-check}.
2069 In the example above, the format string (@code{my_format}) is the second
2070 argument of the function @code{my_print}, and the arguments to check
2071 start with the third argument, so the correct parameters for the format
2072 attribute are 2 and 3.
2074 @opindex ffreestanding
2075 The @code{format} attribute allows you to identify your own functions
2076 which take format strings as arguments, so that GCC can check the
2077 calls to these functions for errors. The compiler always (unless
2078 @option{-ffreestanding} is used) checks formats
2079 for the standard library functions @code{printf}, @code{fprintf},
2080 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2081 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2082 warnings are requested (using @option{-Wformat}), so there is no need to
2083 modify the header file @file{stdio.h}. In C99 mode, the functions
2084 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2085 @code{vsscanf} are also checked. Except in strictly conforming C
2086 standard modes, the X/Open function @code{strfmon} is also checked as
2087 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2088 @xref{C Dialect Options,,Options Controlling C Dialect}.
2090 @item format_arg (@var{string-index})
2091 @cindex @code{format_arg} function attribute
2092 @opindex Wformat-nonliteral
2093 The @code{format_arg} attribute specifies that a function takes a format
2094 string for a @code{printf}, @code{scanf}, @code{strftime} or
2095 @code{strfmon} style function and modifies it (for example, to translate
2096 it into another language), so the result can be passed to a
2097 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2098 function (with the remaining arguments to the format function the same
2099 as they would have been for the unmodified string). For example, the
2104 my_dgettext (char *my_domain, const char *my_format)
2105 __attribute__ ((format_arg (2)));
2109 causes the compiler to check the arguments in calls to a @code{printf},
2110 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2111 format string argument is a call to the @code{my_dgettext} function, for
2112 consistency with the format string argument @code{my_format}. If the
2113 @code{format_arg} attribute had not been specified, all the compiler
2114 could tell in such calls to format functions would be that the format
2115 string argument is not constant; this would generate a warning when
2116 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2117 without the attribute.
2119 The parameter @var{string-index} specifies which argument is the format
2120 string argument (starting from one). Since non-static C++ methods have
2121 an implicit @code{this} argument, the arguments of such methods should
2122 be counted from two.
2124 The @code{format-arg} attribute allows you to identify your own
2125 functions which modify format strings, so that GCC can check the
2126 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2127 type function whose operands are a call to one of your own function.
2128 The compiler always treats @code{gettext}, @code{dgettext}, and
2129 @code{dcgettext} in this manner except when strict ISO C support is
2130 requested by @option{-ansi} or an appropriate @option{-std} option, or
2131 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2132 Controlling C Dialect}.
2134 @item nonnull (@var{arg-index}, @dots{})
2135 @cindex @code{nonnull} function attribute
2136 The @code{nonnull} attribute specifies that some function parameters should
2137 be non-null pointers. For instance, the declaration:
2141 my_memcpy (void *dest, const void *src, size_t len)
2142 __attribute__((nonnull (1, 2)));
2146 causes the compiler to check that, in calls to @code{my_memcpy},
2147 arguments @var{dest} and @var{src} are non-null. If the compiler
2148 determines that a null pointer is passed in an argument slot marked
2149 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2150 is issued. The compiler may also choose to make optimizations based
2151 on the knowledge that certain function arguments will not be null.
2153 If no argument index list is given to the @code{nonnull} attribute,
2154 all pointer arguments are marked as non-null. To illustrate, the
2155 following declaration is equivalent to the previous example:
2159 my_memcpy (void *dest, const void *src, size_t len)
2160 __attribute__((nonnull));
2163 @item no_instrument_function
2164 @cindex @code{no_instrument_function} function attribute
2165 @opindex finstrument-functions
2166 If @option{-finstrument-functions} is given, profiling function calls will
2167 be generated at entry and exit of most user-compiled functions.
2168 Functions with this attribute will not be so instrumented.
2170 @item section ("@var{section-name}")
2171 @cindex @code{section} function attribute
2172 Normally, the compiler places the code it generates in the @code{text} section.
2173 Sometimes, however, you need additional sections, or you need certain
2174 particular functions to appear in special sections. The @code{section}
2175 attribute specifies that a function lives in a particular section.
2176 For example, the declaration:
2179 extern void foobar (void) __attribute__ ((section ("bar")));
2183 puts the function @code{foobar} in the @code{bar} section.
2185 Some file formats do not support arbitrary sections so the @code{section}
2186 attribute is not available on all platforms.
2187 If you need to map the entire contents of a module to a particular
2188 section, consider using the facilities of the linker instead.
2192 @cindex @code{constructor} function attribute
2193 @cindex @code{destructor} function attribute
2194 The @code{constructor} attribute causes the function to be called
2195 automatically before execution enters @code{main ()}. Similarly, the
2196 @code{destructor} attribute causes the function to be called
2197 automatically after @code{main ()} has completed or @code{exit ()} has
2198 been called. Functions with these attributes are useful for
2199 initializing data that will be used implicitly during the execution of
2202 These attributes are not currently implemented for Objective-C@.
2204 @cindex @code{unused} attribute.
2206 This attribute, attached to a function, means that the function is meant
2207 to be possibly unused. GCC will not produce a warning for this
2210 @cindex @code{used} attribute.
2212 This attribute, attached to a function, means that code must be emitted
2213 for the function even if it appears that the function is not referenced.
2214 This is useful, for example, when the function is referenced only in
2217 @cindex @code{deprecated} attribute.
2219 The @code{deprecated} attribute results in a warning if the function
2220 is used anywhere in the source file. This is useful when identifying
2221 functions that are expected to be removed in a future version of a
2222 program. The warning also includes the location of the declaration
2223 of the deprecated function, to enable users to easily find further
2224 information about why the function is deprecated, or what they should
2225 do instead. Note that the warnings only occurs for uses:
2228 int old_fn () __attribute__ ((deprecated));
2230 int (*fn_ptr)() = old_fn;
2233 results in a warning on line 3 but not line 2.
2235 The @code{deprecated} attribute can also be used for variables and
2236 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2238 @item warn_unused_result
2239 @cindex @code{warn_unused_result} attribute
2240 The @code{warn_unused_result} attribute causes a warning to be emitted
2241 if a caller of the function with this attribute does not use its
2242 return value. This is useful for functions where not checking
2243 the result is either a security problem or always a bug, such as
2247 int fn () __attribute__ ((warn_unused_result));
2250 if (fn () < 0) return -1;
2256 results in warning on line 5.
2259 @cindex @code{weak} attribute
2260 The @code{weak} attribute causes the declaration to be emitted as a weak
2261 symbol rather than a global. This is primarily useful in defining
2262 library functions which can be overridden in user code, though it can
2263 also be used with non-function declarations. Weak symbols are supported
2264 for ELF targets, and also for a.out targets when using the GNU assembler
2268 @cindex @code{malloc} attribute
2269 The @code{malloc} attribute is used to tell the compiler that a function
2270 may be treated as if any non-@code{NULL} pointer it returns cannot
2271 alias any other pointer valid when the function returns.
2272 This will often improve optimization.
2273 Standard functions with this property include @code{malloc} and
2274 @code{calloc}. @code{realloc}-like functions have this property as
2275 long as the old pointer is never referred to (including comparing it
2276 to the new pointer) after the function returns a non-@code{NULL}
2279 @item alias ("@var{target}")
2280 @cindex @code{alias} attribute
2281 The @code{alias} attribute causes the declaration to be emitted as an
2282 alias for another symbol, which must be specified. For instance,
2285 void __f () @{ /* @r{Do something.} */; @}
2286 void f () __attribute__ ((weak, alias ("__f")));
2289 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2290 mangled name for the target must be used.
2292 Not all target machines support this attribute.
2294 @item visibility ("@var{visibility_type}")
2295 @cindex @code{visibility} attribute
2296 The @code{visibility} attribute on ELF targets causes the declaration
2297 to be emitted with default, hidden, protected or internal visibility.
2300 void __attribute__ ((visibility ("protected")))
2301 f () @{ /* @r{Do something.} */; @}
2302 int i __attribute__ ((visibility ("hidden")));
2305 See the ELF gABI for complete details, but the short story is:
2309 Default visibility is the normal case for ELF. This value is
2310 available for the visibility attribute to override other options
2311 that may change the assumed visibility of symbols.
2314 Hidden visibility indicates that the symbol will not be placed into
2315 the dynamic symbol table, so no other @dfn{module} (executable or
2316 shared library) can reference it directly.
2319 Protected visibility indicates that the symbol will be placed in the
2320 dynamic symbol table, but that references within the defining module
2321 will bind to the local symbol. That is, the symbol cannot be overridden
2325 Internal visibility is like hidden visibility, but with additional
2326 processor specific semantics. Unless otherwise specified by the psABI,
2327 GCC defines internal visibility to mean that the function is @emph{never}
2328 called from another module. Note that hidden symbols, while they cannot
2329 be referenced directly by other modules, can be referenced indirectly via
2330 function pointers. By indicating that a symbol cannot be called from
2331 outside the module, GCC may for instance omit the load of a PIC register
2332 since it is known that the calling function loaded the correct value.
2335 Not all ELF targets support this attribute.
2337 @item regparm (@var{number})
2338 @cindex @code{regparm} attribute
2339 @cindex functions that are passed arguments in registers on the 386
2340 On the Intel 386, the @code{regparm} attribute causes the compiler to
2341 pass up to @var{number} integer arguments in registers EAX,
2342 EDX, and ECX instead of on the stack. Functions that take a
2343 variable number of arguments will continue to be passed all of their
2344 arguments on the stack.
2346 Beware that on some ELF systems this attribute is unsuitable for
2347 global functions in shared libraries with lazy binding (which is the
2348 default). Lazy binding will send the first call via resolving code in
2349 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2350 per the standard calling conventions. Solaris 8 is affected by this.
2351 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2352 safe since the loaders there save all registers. (Lazy binding can be
2353 disabled with the linker or the loader if desired, to avoid the
2357 @cindex functions that pop the argument stack on the 386
2358 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2359 assume that the called function will pop off the stack space used to
2360 pass arguments, unless it takes a variable number of arguments.
2363 @cindex functions that pop the argument stack on the 386
2364 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2365 pass the first two arguments in the registers ECX and EDX. Subsequent
2366 arguments are passed on the stack. The called function will pop the
2367 arguments off the stack. If the number of arguments is variable all
2368 arguments are pushed on the stack.
2371 @cindex functions that do pop the argument stack on the 386
2373 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2374 assume that the calling function will pop off the stack space used to
2375 pass arguments. This is
2376 useful to override the effects of the @option{-mrtd} switch.
2378 @item longcall/shortcall
2379 @cindex functions called via pointer on the RS/6000 and PowerPC
2380 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2381 compiler to always call this function via a pointer, just as it would if
2382 the @option{-mlongcall} option had been specified. The @code{shortcall}
2383 attribute causes the compiler not to do this. These attributes override
2384 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2387 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2388 calls are necessary.
2390 @item long_call/short_call
2391 @cindex indirect calls on ARM
2392 This attribute specifies how a particular function is called on
2393 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2394 command line switch and @code{#pragma long_calls} settings. The
2395 @code{long_call} attribute causes the compiler to always call the
2396 function by first loading its address into a register and then using the
2397 contents of that register. The @code{short_call} attribute always places
2398 the offset to the function from the call site into the @samp{BL}
2399 instruction directly.
2401 @item function_vector
2402 @cindex calling functions through the function vector on the H8/300 processors
2403 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2404 function should be called through the function vector. Calling a
2405 function through the function vector will reduce code size, however;
2406 the function vector has a limited size (maximum 128 entries on the H8/300
2407 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2409 You must use GAS and GLD from GNU binutils version 2.7 or later for
2410 this attribute to work correctly.
2413 @cindex interrupt handler functions
2414 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2415 that the specified function is an interrupt handler. The compiler will
2416 generate function entry and exit sequences suitable for use in an
2417 interrupt handler when this attribute is present.
2419 Note, interrupt handlers for the m68k, H8/300, H8/300H, H8S, and SH processors
2420 can be specified via the @code{interrupt_handler} attribute.
2422 Note, on the AVR, interrupts will be enabled inside the function.
2424 Note, for the ARM, you can specify the kind of interrupt to be handled by
2425 adding an optional parameter to the interrupt attribute like this:
2428 void f () __attribute__ ((interrupt ("IRQ")));
2431 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2433 @item interrupt_handler
2434 @cindex interrupt handler functions on the m68k, H8/300 and SH processors
2435 Use this attribute on the m68k, H8/300, H8/300H, H8S, and SH to indicate that
2436 the specified function is an interrupt handler. The compiler will generate
2437 function entry and exit sequences suitable for use in an interrupt
2438 handler when this attribute is present.
2441 Use this attribute on the SH to indicate an @code{interrupt_handler}
2442 function should switch to an alternate stack. It expects a string
2443 argument that names a global variable holding the address of the
2448 void f () __attribute__ ((interrupt_handler,
2449 sp_switch ("alt_stack")));
2453 Use this attribute on the SH for an @code{interrupt_handler} to return using
2454 @code{trapa} instead of @code{rte}. This attribute expects an integer
2455 argument specifying the trap number to be used.
2458 @cindex eight bit data on the H8/300, H8/300H, and H8S
2459 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2460 variable should be placed into the eight bit data section.
2461 The compiler will generate more efficient code for certain operations
2462 on data in the eight bit data area. Note the eight bit data area is limited to
2465 You must use GAS and GLD from GNU binutils version 2.7 or later for
2466 this attribute to work correctly.
2469 @cindex tiny data section on the H8/300H and H8S
2470 Use this attribute on the H8/300H and H8S to indicate that the specified
2471 variable should be placed into the tiny data section.
2472 The compiler will generate more efficient code for loads and stores
2473 on data in the tiny data section. Note the tiny data area is limited to
2474 slightly under 32kbytes of data.
2477 @cindex save all registers on the H8/300, H8/300H, and H8S
2478 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
2479 all registers except the stack pointer should be saved in the prologue
2480 regardless of whether they are used or not.
2483 @cindex signal handler functions on the AVR processors
2484 Use this attribute on the AVR to indicate that the specified
2485 function is a signal handler. The compiler will generate function
2486 entry and exit sequences suitable for use in a signal handler when this
2487 attribute is present. Interrupts will be disabled inside the function.
2490 @cindex function without a prologue/epilogue code
2491 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2492 specified function does not need prologue/epilogue sequences generated by
2493 the compiler. It is up to the programmer to provide these sequences.
2495 @item model (@var{model-name})
2496 @cindex function addressability on the M32R/D
2497 @cindex variable addressability on the IA-64
2499 On the M32R/D, use this attribute to set the addressability of an
2500 object, and of the code generated for a function. The identifier
2501 @var{model-name} is one of @code{small}, @code{medium}, or
2502 @code{large}, representing each of the code models.
2504 Small model objects live in the lower 16MB of memory (so that their
2505 addresses can be loaded with the @code{ld24} instruction), and are
2506 callable with the @code{bl} instruction.
2508 Medium model objects may live anywhere in the 32-bit address space (the
2509 compiler will generate @code{seth/add3} instructions to load their addresses),
2510 and are callable with the @code{bl} instruction.
2512 Large model objects may live anywhere in the 32-bit address space (the
2513 compiler will generate @code{seth/add3} instructions to load their addresses),
2514 and may not be reachable with the @code{bl} instruction (the compiler will
2515 generate the much slower @code{seth/add3/jl} instruction sequence).
2517 On IA-64, use this attribute to set the addressability of an object.
2518 At present, the only supported identifier for @var{model-name} is
2519 @code{small}, indicating addressability via ``small'' (22-bit)
2520 addresses (so that their addresses can be loaded with the @code{addl}
2521 instruction). Caveat: such addressing is by definition not position
2522 independent and hence this attribute must not be used for objects
2523 defined by shared libraries.
2526 @cindex functions which handle memory bank switching
2527 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2528 use a calling convention that takes care of switching memory banks when
2529 entering and leaving a function. This calling convention is also the
2530 default when using the @option{-mlong-calls} option.
2532 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2533 to call and return from a function.
2535 On 68HC11 the compiler will generate a sequence of instructions
2536 to invoke a board-specific routine to switch the memory bank and call the
2537 real function. The board-specific routine simulates a @code{call}.
2538 At the end of a function, it will jump to a board-specific routine
2539 instead of using @code{rts}. The board-specific return routine simulates
2543 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2544 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2545 use the normal calling convention based on @code{jsr} and @code{rts}.
2546 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2550 @cindex @code{__declspec(dllimport)}
2551 On Microsoft Windows targets, the @code{dllimport} attribute causes the compiler
2552 to reference a function or variable via a global pointer to a pointer
2553 that is set up by the Microsoft Windows dll library. The pointer name is formed by
2554 combining @code{_imp__} and the function or variable name. The attribute
2555 implies @code{extern} storage.
2557 Currently, the attribute is ignored for inlined functions. If the
2558 attribute is applied to a symbol @emph{definition}, an error is reported.
2559 If a symbol previously declared @code{dllimport} is later defined, the
2560 attribute is ignored in subsequent references, and a warning is emitted.
2561 The attribute is also overridden by a subsequent declaration as
2564 When applied to C++ classes, the attribute marks non-inlined
2565 member functions and static data members as imports. However, the
2566 attribute is ignored for virtual methods to allow creation of vtables
2569 On cygwin, mingw and arm-pe targets, @code{__declspec(dllimport)} is
2570 recognized as a synonym for @code{__attribute__ ((dllimport))} for
2571 compatibility with other Microsoft Windows compilers.
2573 The use of the @code{dllimport} attribute on functions is not necessary,
2574 but provides a small performance benefit by eliminating a thunk in the
2575 dll. The use of the @code{dllimport} attribute on imported variables was
2576 required on older versions of GNU ld, but can now be avoided by passing
2577 the @option{--enable-auto-import} switch to ld. As with functions, using
2578 the attribute for a variable eliminates a thunk in the dll.
2580 One drawback to using this attribute is that a pointer to a function or
2581 variable marked as dllimport cannot be used as a constant address. The
2582 attribute can be disabled for functions by setting the
2583 @option{-mnop-fun-dllimport} flag.
2586 @cindex @code{__declspec(dllexport)}
2587 On Microsoft Windows targets the @code{dllexport} attribute causes the compiler to
2588 provide a global pointer to a pointer in a dll, so that it can be
2589 referenced with the @code{dllimport} attribute. The pointer name is
2590 formed by combining @code{_imp__} and the function or variable name.
2592 Currently, the @code{dllexport}attribute is ignored for inlined
2593 functions, but export can be forced by using the
2594 @option{-fkeep-inline-functions} flag. The attribute is also ignored for
2597 When applied to C++ classes. the attribute marks defined non-inlined
2598 member functions and static data members as exports. Static consts
2599 initialized in-class are not marked unless they are also defined
2602 On cygwin, mingw and arm-pe targets, @code{__declspec(dllexport)} is
2603 recognized as a synonym for @code{__attribute__ ((dllexport))} for
2604 compatibility with other Microsoft Windows compilers.
2606 Alternative methods for including the symbol in the dll's export table
2607 are to use a .def file with an @code{EXPORTS} section or, with GNU ld,
2608 using the @option{--export-all} linker flag.
2612 You can specify multiple attributes in a declaration by separating them
2613 by commas within the double parentheses or by immediately following an
2614 attribute declaration with another attribute declaration.
2616 @cindex @code{#pragma}, reason for not using
2617 @cindex pragma, reason for not using
2618 Some people object to the @code{__attribute__} feature, suggesting that
2619 ISO C's @code{#pragma} should be used instead. At the time
2620 @code{__attribute__} was designed, there were two reasons for not doing
2625 It is impossible to generate @code{#pragma} commands from a macro.
2628 There is no telling what the same @code{#pragma} might mean in another
2632 These two reasons applied to almost any application that might have been
2633 proposed for @code{#pragma}. It was basically a mistake to use
2634 @code{#pragma} for @emph{anything}.
2636 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2637 to be generated from macros. In addition, a @code{#pragma GCC}
2638 namespace is now in use for GCC-specific pragmas. However, it has been
2639 found convenient to use @code{__attribute__} to achieve a natural
2640 attachment of attributes to their corresponding declarations, whereas
2641 @code{#pragma GCC} is of use for constructs that do not naturally form
2642 part of the grammar. @xref{Other Directives,,Miscellaneous
2643 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2645 @node Attribute Syntax
2646 @section Attribute Syntax
2647 @cindex attribute syntax
2649 This section describes the syntax with which @code{__attribute__} may be
2650 used, and the constructs to which attribute specifiers bind, for the C
2651 language. Some details may vary for C++ and Objective-C@. Because of
2652 infelicities in the grammar for attributes, some forms described here
2653 may not be successfully parsed in all cases.
2655 There are some problems with the semantics of attributes in C++. For
2656 example, there are no manglings for attributes, although they may affect
2657 code generation, so problems may arise when attributed types are used in
2658 conjunction with templates or overloading. Similarly, @code{typeid}
2659 does not distinguish between types with different attributes. Support
2660 for attributes in C++ may be restricted in future to attributes on
2661 declarations only, but not on nested declarators.
2663 @xref{Function Attributes}, for details of the semantics of attributes
2664 applying to functions. @xref{Variable Attributes}, for details of the
2665 semantics of attributes applying to variables. @xref{Type Attributes},
2666 for details of the semantics of attributes applying to structure, union
2667 and enumerated types.
2669 An @dfn{attribute specifier} is of the form
2670 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2671 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2672 each attribute is one of the following:
2676 Empty. Empty attributes are ignored.
2679 A word (which may be an identifier such as @code{unused}, or a reserved
2680 word such as @code{const}).
2683 A word, followed by, in parentheses, parameters for the attribute.
2684 These parameters take one of the following forms:
2688 An identifier. For example, @code{mode} attributes use this form.
2691 An identifier followed by a comma and a non-empty comma-separated list
2692 of expressions. For example, @code{format} attributes use this form.
2695 A possibly empty comma-separated list of expressions. For example,
2696 @code{format_arg} attributes use this form with the list being a single
2697 integer constant expression, and @code{alias} attributes use this form
2698 with the list being a single string constant.
2702 An @dfn{attribute specifier list} is a sequence of one or more attribute
2703 specifiers, not separated by any other tokens.
2705 In GNU C, an attribute specifier list may appear after the colon following a
2706 label, other than a @code{case} or @code{default} label. The only
2707 attribute it makes sense to use after a label is @code{unused}. This
2708 feature is intended for code generated by programs which contains labels
2709 that may be unused but which is compiled with @option{-Wall}. It would
2710 not normally be appropriate to use in it human-written code, though it
2711 could be useful in cases where the code that jumps to the label is
2712 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2713 such placement of attribute lists, as it is permissible for a
2714 declaration, which could begin with an attribute list, to be labelled in
2715 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2716 does not arise there.
2718 An attribute specifier list may appear as part of a @code{struct},
2719 @code{union} or @code{enum} specifier. It may go either immediately
2720 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2721 the closing brace. It is ignored if the content of the structure, union
2722 or enumerated type is not defined in the specifier in which the
2723 attribute specifier list is used---that is, in usages such as
2724 @code{struct __attribute__((foo)) bar} with no following opening brace.
2725 Where attribute specifiers follow the closing brace, they are considered
2726 to relate to the structure, union or enumerated type defined, not to any
2727 enclosing declaration the type specifier appears in, and the type
2728 defined is not complete until after the attribute specifiers.
2729 @c Otherwise, there would be the following problems: a shift/reduce
2730 @c conflict between attributes binding the struct/union/enum and
2731 @c binding to the list of specifiers/qualifiers; and "aligned"
2732 @c attributes could use sizeof for the structure, but the size could be
2733 @c changed later by "packed" attributes.
2735 Otherwise, an attribute specifier appears as part of a declaration,
2736 counting declarations of unnamed parameters and type names, and relates
2737 to that declaration (which may be nested in another declaration, for
2738 example in the case of a parameter declaration), or to a particular declarator
2739 within a declaration. Where an
2740 attribute specifier is applied to a parameter declared as a function or
2741 an array, it should apply to the function or array rather than the
2742 pointer to which the parameter is implicitly converted, but this is not
2743 yet correctly implemented.
2745 Any list of specifiers and qualifiers at the start of a declaration may
2746 contain attribute specifiers, whether or not such a list may in that
2747 context contain storage class specifiers. (Some attributes, however,
2748 are essentially in the nature of storage class specifiers, and only make
2749 sense where storage class specifiers may be used; for example,
2750 @code{section}.) There is one necessary limitation to this syntax: the
2751 first old-style parameter declaration in a function definition cannot
2752 begin with an attribute specifier, because such an attribute applies to
2753 the function instead by syntax described below (which, however, is not
2754 yet implemented in this case). In some other cases, attribute
2755 specifiers are permitted by this grammar but not yet supported by the
2756 compiler. All attribute specifiers in this place relate to the
2757 declaration as a whole. In the obsolescent usage where a type of
2758 @code{int} is implied by the absence of type specifiers, such a list of
2759 specifiers and qualifiers may be an attribute specifier list with no
2760 other specifiers or qualifiers.
2762 An attribute specifier list may appear immediately before a declarator
2763 (other than the first) in a comma-separated list of declarators in a
2764 declaration of more than one identifier using a single list of
2765 specifiers and qualifiers. Such attribute specifiers apply
2766 only to the identifier before whose declarator they appear. For
2770 __attribute__((noreturn)) void d0 (void),
2771 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2776 the @code{noreturn} attribute applies to all the functions
2777 declared; the @code{format} attribute only applies to @code{d1}.
2779 An attribute specifier list may appear immediately before the comma,
2780 @code{=} or semicolon terminating the declaration of an identifier other
2781 than a function definition. At present, such attribute specifiers apply
2782 to the declared object or function, but in future they may attach to the
2783 outermost adjacent declarator. In simple cases there is no difference,
2784 but, for example, in
2787 void (****f)(void) __attribute__((noreturn));
2791 at present the @code{noreturn} attribute applies to @code{f}, which
2792 causes a warning since @code{f} is not a function, but in future it may
2793 apply to the function @code{****f}. The precise semantics of what
2794 attributes in such cases will apply to are not yet specified. Where an
2795 assembler name for an object or function is specified (@pxref{Asm
2796 Labels}), at present the attribute must follow the @code{asm}
2797 specification; in future, attributes before the @code{asm} specification
2798 may apply to the adjacent declarator, and those after it to the declared
2801 An attribute specifier list may, in future, be permitted to appear after
2802 the declarator in a function definition (before any old-style parameter
2803 declarations or the function body).
2805 Attribute specifiers may be mixed with type qualifiers appearing inside
2806 the @code{[]} of a parameter array declarator, in the C99 construct by
2807 which such qualifiers are applied to the pointer to which the array is
2808 implicitly converted. Such attribute specifiers apply to the pointer,
2809 not to the array, but at present this is not implemented and they are
2812 An attribute specifier list may appear at the start of a nested
2813 declarator. At present, there are some limitations in this usage: the
2814 attributes correctly apply to the declarator, but for most individual
2815 attributes the semantics this implies are not implemented.
2816 When attribute specifiers follow the @code{*} of a pointer
2817 declarator, they may be mixed with any type qualifiers present.
2818 The following describes the formal semantics of this syntax. It will make the
2819 most sense if you are familiar with the formal specification of
2820 declarators in the ISO C standard.
2822 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2823 D1}, where @code{T} contains declaration specifiers that specify a type
2824 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2825 contains an identifier @var{ident}. The type specified for @var{ident}
2826 for derived declarators whose type does not include an attribute
2827 specifier is as in the ISO C standard.
2829 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2830 and the declaration @code{T D} specifies the type
2831 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2832 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2833 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2835 If @code{D1} has the form @code{*
2836 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2837 declaration @code{T D} specifies the type
2838 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2839 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2840 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2846 void (__attribute__((noreturn)) ****f) (void);
2850 specifies the type ``pointer to pointer to pointer to pointer to
2851 non-returning function returning @code{void}''. As another example,
2854 char *__attribute__((aligned(8))) *f;
2858 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2859 Note again that this does not work with most attributes; for example,
2860 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2861 is not yet supported.
2863 For compatibility with existing code written for compiler versions that
2864 did not implement attributes on nested declarators, some laxity is
2865 allowed in the placing of attributes. If an attribute that only applies
2866 to types is applied to a declaration, it will be treated as applying to
2867 the type of that declaration. If an attribute that only applies to
2868 declarations is applied to the type of a declaration, it will be treated
2869 as applying to that declaration; and, for compatibility with code
2870 placing the attributes immediately before the identifier declared, such
2871 an attribute applied to a function return type will be treated as
2872 applying to the function type, and such an attribute applied to an array
2873 element type will be treated as applying to the array type. If an
2874 attribute that only applies to function types is applied to a
2875 pointer-to-function type, it will be treated as applying to the pointer
2876 target type; if such an attribute is applied to a function return type
2877 that is not a pointer-to-function type, it will be treated as applying
2878 to the function type.
2880 @node Function Prototypes
2881 @section Prototypes and Old-Style Function Definitions
2882 @cindex function prototype declarations
2883 @cindex old-style function definitions
2884 @cindex promotion of formal parameters
2886 GNU C extends ISO C to allow a function prototype to override a later
2887 old-style non-prototype definition. Consider the following example:
2890 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2897 /* @r{Prototype function declaration.} */
2898 int isroot P((uid_t));
2900 /* @r{Old-style function definition.} */
2902 isroot (x) /* ??? lossage here ??? */
2909 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2910 not allow this example, because subword arguments in old-style
2911 non-prototype definitions are promoted. Therefore in this example the
2912 function definition's argument is really an @code{int}, which does not
2913 match the prototype argument type of @code{short}.
2915 This restriction of ISO C makes it hard to write code that is portable
2916 to traditional C compilers, because the programmer does not know
2917 whether the @code{uid_t} type is @code{short}, @code{int}, or
2918 @code{long}. Therefore, in cases like these GNU C allows a prototype
2919 to override a later old-style definition. More precisely, in GNU C, a
2920 function prototype argument type overrides the argument type specified
2921 by a later old-style definition if the former type is the same as the
2922 latter type before promotion. Thus in GNU C the above example is
2923 equivalent to the following:
2936 GNU C++ does not support old-style function definitions, so this
2937 extension is irrelevant.
2940 @section C++ Style Comments
2942 @cindex C++ comments
2943 @cindex comments, C++ style
2945 In GNU C, you may use C++ style comments, which start with @samp{//} and
2946 continue until the end of the line. Many other C implementations allow
2947 such comments, and they are included in the 1999 C standard. However,
2948 C++ style comments are not recognized if you specify an @option{-std}
2949 option specifying a version of ISO C before C99, or @option{-ansi}
2950 (equivalent to @option{-std=c89}).
2953 @section Dollar Signs in Identifier Names
2955 @cindex dollar signs in identifier names
2956 @cindex identifier names, dollar signs in
2958 In GNU C, you may normally use dollar signs in identifier names.
2959 This is because many traditional C implementations allow such identifiers.
2960 However, dollar signs in identifiers are not supported on a few target
2961 machines, typically because the target assembler does not allow them.
2963 @node Character Escapes
2964 @section The Character @key{ESC} in Constants
2966 You can use the sequence @samp{\e} in a string or character constant to
2967 stand for the ASCII character @key{ESC}.
2970 @section Inquiring on Alignment of Types or Variables
2972 @cindex type alignment
2973 @cindex variable alignment
2975 The keyword @code{__alignof__} allows you to inquire about how an object
2976 is aligned, or the minimum alignment usually required by a type. Its
2977 syntax is just like @code{sizeof}.
2979 For example, if the target machine requires a @code{double} value to be
2980 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2981 This is true on many RISC machines. On more traditional machine
2982 designs, @code{__alignof__ (double)} is 4 or even 2.
2984 Some machines never actually require alignment; they allow reference to any
2985 data type even at an odd address. For these machines, @code{__alignof__}
2986 reports the @emph{recommended} alignment of a type.
2988 If the operand of @code{__alignof__} is an lvalue rather than a type,
2989 its value is the required alignment for its type, taking into account
2990 any minimum alignment specified with GCC's @code{__attribute__}
2991 extension (@pxref{Variable Attributes}). For example, after this
2995 struct foo @{ int x; char y; @} foo1;
2999 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3000 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3002 It is an error to ask for the alignment of an incomplete type.
3004 @node Variable Attributes
3005 @section Specifying Attributes of Variables
3006 @cindex attribute of variables
3007 @cindex variable attributes
3009 The keyword @code{__attribute__} allows you to specify special
3010 attributes of variables or structure fields. This keyword is followed
3011 by an attribute specification inside double parentheses. Some
3012 attributes are currently defined generically for variables.
3013 Other attributes are defined for variables on particular target
3014 systems. Other attributes are available for functions
3015 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3016 Other front ends might define more attributes
3017 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3019 You may also specify attributes with @samp{__} preceding and following
3020 each keyword. This allows you to use them in header files without
3021 being concerned about a possible macro of the same name. For example,
3022 you may use @code{__aligned__} instead of @code{aligned}.
3024 @xref{Attribute Syntax}, for details of the exact syntax for using
3028 @cindex @code{aligned} attribute
3029 @item aligned (@var{alignment})
3030 This attribute specifies a minimum alignment for the variable or
3031 structure field, measured in bytes. For example, the declaration:
3034 int x __attribute__ ((aligned (16))) = 0;
3038 causes the compiler to allocate the global variable @code{x} on a
3039 16-byte boundary. On a 68040, this could be used in conjunction with
3040 an @code{asm} expression to access the @code{move16} instruction which
3041 requires 16-byte aligned operands.
3043 You can also specify the alignment of structure fields. For example, to
3044 create a double-word aligned @code{int} pair, you could write:
3047 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3051 This is an alternative to creating a union with a @code{double} member
3052 that forces the union to be double-word aligned.
3054 As in the preceding examples, you can explicitly specify the alignment
3055 (in bytes) that you wish the compiler to use for a given variable or
3056 structure field. Alternatively, you can leave out the alignment factor
3057 and just ask the compiler to align a variable or field to the maximum
3058 useful alignment for the target machine you are compiling for. For
3059 example, you could write:
3062 short array[3] __attribute__ ((aligned));
3065 Whenever you leave out the alignment factor in an @code{aligned} attribute
3066 specification, the compiler automatically sets the alignment for the declared
3067 variable or field to the largest alignment which is ever used for any data
3068 type on the target machine you are compiling for. Doing this can often make
3069 copy operations more efficient, because the compiler can use whatever
3070 instructions copy the biggest chunks of memory when performing copies to
3071 or from the variables or fields that you have aligned this way.
3073 The @code{aligned} attribute can only increase the alignment; but you
3074 can decrease it by specifying @code{packed} as well. See below.
3076 Note that the effectiveness of @code{aligned} attributes may be limited
3077 by inherent limitations in your linker. On many systems, the linker is
3078 only able to arrange for variables to be aligned up to a certain maximum
3079 alignment. (For some linkers, the maximum supported alignment may
3080 be very very small.) If your linker is only able to align variables
3081 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3082 in an @code{__attribute__} will still only provide you with 8 byte
3083 alignment. See your linker documentation for further information.
3085 @item cleanup (@var{cleanup_function})
3086 @cindex @code{cleanup} attribute
3087 The @code{cleanup} attribute runs a function when the variable goes
3088 out of scope. This attribute can only be applied to auto function
3089 scope variables; it may not be applied to parameters or variables
3090 with static storage duration. The function must take one parameter,
3091 a pointer to a type compatible with the variable. The return value
3092 of the function (if any) is ignored.
3094 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3095 will be run during the stack unwinding that happens during the
3096 processing of the exception. Note that the @code{cleanup} attribute
3097 does not allow the exception to be caught, only to perform an action.
3098 It is undefined what happens if @var{cleanup_function} does not
3103 @cindex @code{common} attribute
3104 @cindex @code{nocommon} attribute
3107 The @code{common} attribute requests GCC to place a variable in
3108 ``common'' storage. The @code{nocommon} attribute requests the
3109 opposite -- to allocate space for it directly.
3111 These attributes override the default chosen by the
3112 @option{-fno-common} and @option{-fcommon} flags respectively.
3115 @cindex @code{deprecated} attribute
3116 The @code{deprecated} attribute results in a warning if the variable
3117 is used anywhere in the source file. This is useful when identifying
3118 variables that are expected to be removed in a future version of a
3119 program. The warning also includes the location of the declaration
3120 of the deprecated variable, to enable users to easily find further
3121 information about why the variable is deprecated, or what they should
3122 do instead. Note that the warning only occurs for uses:
3125 extern int old_var __attribute__ ((deprecated));
3127 int new_fn () @{ return old_var; @}
3130 results in a warning on line 3 but not line 2.
3132 The @code{deprecated} attribute can also be used for functions and
3133 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3135 @item mode (@var{mode})
3136 @cindex @code{mode} attribute
3137 This attribute specifies the data type for the declaration---whichever
3138 type corresponds to the mode @var{mode}. This in effect lets you
3139 request an integer or floating point type according to its width.
3141 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3142 indicate the mode corresponding to a one-byte integer, @samp{word} or
3143 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3144 or @samp{__pointer__} for the mode used to represent pointers.
3147 @cindex @code{packed} attribute
3148 The @code{packed} attribute specifies that a variable or structure field
3149 should have the smallest possible alignment---one byte for a variable,
3150 and one bit for a field, unless you specify a larger value with the
3151 @code{aligned} attribute.
3153 Here is a structure in which the field @code{x} is packed, so that it
3154 immediately follows @code{a}:
3160 int x[2] __attribute__ ((packed));
3164 @item section ("@var{section-name}")
3165 @cindex @code{section} variable attribute
3166 Normally, the compiler places the objects it generates in sections like
3167 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3168 or you need certain particular variables to appear in special sections,
3169 for example to map to special hardware. The @code{section}
3170 attribute specifies that a variable (or function) lives in a particular
3171 section. For example, this small program uses several specific section names:
3174 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3175 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3176 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3177 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3181 /* Initialize stack pointer */
3182 init_sp (stack + sizeof (stack));
3184 /* Initialize initialized data */
3185 memcpy (&init_data, &data, &edata - &data);
3187 /* Turn on the serial ports */
3194 Use the @code{section} attribute with an @emph{initialized} definition
3195 of a @emph{global} variable, as shown in the example. GCC issues
3196 a warning and otherwise ignores the @code{section} attribute in
3197 uninitialized variable declarations.
3199 You may only use the @code{section} attribute with a fully initialized
3200 global definition because of the way linkers work. The linker requires
3201 each object be defined once, with the exception that uninitialized
3202 variables tentatively go in the @code{common} (or @code{bss}) section
3203 and can be multiply ``defined''. You can force a variable to be
3204 initialized with the @option{-fno-common} flag or the @code{nocommon}
3207 Some file formats do not support arbitrary sections so the @code{section}
3208 attribute is not available on all platforms.
3209 If you need to map the entire contents of a module to a particular
3210 section, consider using the facilities of the linker instead.
3213 @cindex @code{shared} variable attribute
3214 On Microsoft Windows, in addition to putting variable definitions in a named
3215 section, the section can also be shared among all running copies of an
3216 executable or DLL@. For example, this small program defines shared data
3217 by putting it in a named section @code{shared} and marking the section
3221 int foo __attribute__((section ("shared"), shared)) = 0;
3226 /* Read and write foo. All running
3227 copies see the same value. */
3233 You may only use the @code{shared} attribute along with @code{section}
3234 attribute with a fully initialized global definition because of the way
3235 linkers work. See @code{section} attribute for more information.
3237 The @code{shared} attribute is only available on Microsoft Windows@.
3239 @item tls_model ("@var{tls_model}")
3240 @cindex @code{tls_model} attribute
3241 The @code{tls_model} attribute sets thread-local storage model
3242 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3243 overriding @code{-ftls-model=} command line switch on a per-variable
3245 The @var{tls_model} argument should be one of @code{global-dynamic},
3246 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3248 Not all targets support this attribute.
3250 @item transparent_union
3251 This attribute, attached to a function parameter which is a union, means
3252 that the corresponding argument may have the type of any union member,
3253 but the argument is passed as if its type were that of the first union
3254 member. For more details see @xref{Type Attributes}. You can also use
3255 this attribute on a @code{typedef} for a union data type; then it
3256 applies to all function parameters with that type.
3259 This attribute, attached to a variable, means that the variable is meant
3260 to be possibly unused. GCC will not produce a warning for this
3263 @item vector_size (@var{bytes})
3264 This attribute specifies the vector size for the variable, measured in
3265 bytes. For example, the declaration:
3268 int foo __attribute__ ((vector_size (16)));
3272 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3273 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3274 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3276 This attribute is only applicable to integral and float scalars,
3277 although arrays, pointers, and function return values are allowed in
3278 conjunction with this construct.
3280 Aggregates with this attribute are invalid, even if they are of the same
3281 size as a corresponding scalar. For example, the declaration:
3284 struct S @{ int a; @};
3285 struct S __attribute__ ((vector_size (16))) foo;
3289 is invalid even if the size of the structure is the same as the size of
3293 The @code{weak} attribute is described in @xref{Function Attributes}.
3296 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3299 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3303 @subsection M32R/D Variable Attributes
3305 One attribute is currently defined for the M32R/D.
3308 @item model (@var{model-name})
3309 @cindex variable addressability on the M32R/D
3310 Use this attribute on the M32R/D to set the addressability of an object.
3311 The identifier @var{model-name} is one of @code{small}, @code{medium},
3312 or @code{large}, representing each of the code models.
3314 Small model objects live in the lower 16MB of memory (so that their
3315 addresses can be loaded with the @code{ld24} instruction).
3317 Medium and large model objects may live anywhere in the 32-bit address space
3318 (the compiler will generate @code{seth/add3} instructions to load their
3322 @subsection i386 Variable Attributes
3324 Two attributes are currently defined for i386 configurations:
3325 @code{ms_struct} and @code{gcc_struct}
3330 @cindex @code{ms_struct} attribute
3331 @cindex @code{gcc_struct} attribute
3333 If @code{packed} is used on a structure, or if bit-fields are used
3334 it may be that the Microsoft ABI packs them differently
3335 than GCC would normally pack them. Particularly when moving packed
3336 data between functions compiled with GCC and the native Microsoft compiler
3337 (either via function call or as data in a file), it may be necessary to access
3340 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3341 compilers to match the native Microsoft compiler.
3344 @node Type Attributes
3345 @section Specifying Attributes of Types
3346 @cindex attribute of types
3347 @cindex type attributes
3349 The keyword @code{__attribute__} allows you to specify special
3350 attributes of @code{struct} and @code{union} types when you define such
3351 types. This keyword is followed by an attribute specification inside
3352 double parentheses. Six attributes are currently defined for types:
3353 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3354 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3355 functions (@pxref{Function Attributes}) and for variables
3356 (@pxref{Variable Attributes}).
3358 You may also specify any one of these attributes with @samp{__}
3359 preceding and following its keyword. This allows you to use these
3360 attributes in header files without being concerned about a possible
3361 macro of the same name. For example, you may use @code{__aligned__}
3362 instead of @code{aligned}.
3364 You may specify the @code{aligned} and @code{transparent_union}
3365 attributes either in a @code{typedef} declaration or just past the
3366 closing curly brace of a complete enum, struct or union type
3367 @emph{definition} and the @code{packed} attribute only past the closing
3368 brace of a definition.
3370 You may also specify attributes between the enum, struct or union
3371 tag and the name of the type rather than after the closing brace.
3373 @xref{Attribute Syntax}, for details of the exact syntax for using
3377 @cindex @code{aligned} attribute
3378 @item aligned (@var{alignment})
3379 This attribute specifies a minimum alignment (in bytes) for variables
3380 of the specified type. For example, the declarations:
3383 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3384 typedef int more_aligned_int __attribute__ ((aligned (8)));
3388 force the compiler to insure (as far as it can) that each variable whose
3389 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3390 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3391 variables of type @code{struct S} aligned to 8-byte boundaries allows
3392 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3393 store) instructions when copying one variable of type @code{struct S} to
3394 another, thus improving run-time efficiency.
3396 Note that the alignment of any given @code{struct} or @code{union} type
3397 is required by the ISO C standard to be at least a perfect multiple of
3398 the lowest common multiple of the alignments of all of the members of
3399 the @code{struct} or @code{union} in question. This means that you @emph{can}
3400 effectively adjust the alignment of a @code{struct} or @code{union}
3401 type by attaching an @code{aligned} attribute to any one of the members
3402 of such a type, but the notation illustrated in the example above is a
3403 more obvious, intuitive, and readable way to request the compiler to
3404 adjust the alignment of an entire @code{struct} or @code{union} type.
3406 As in the preceding example, you can explicitly specify the alignment
3407 (in bytes) that you wish the compiler to use for a given @code{struct}
3408 or @code{union} type. Alternatively, you can leave out the alignment factor
3409 and just ask the compiler to align a type to the maximum
3410 useful alignment for the target machine you are compiling for. For
3411 example, you could write:
3414 struct S @{ short f[3]; @} __attribute__ ((aligned));
3417 Whenever you leave out the alignment factor in an @code{aligned}
3418 attribute specification, the compiler automatically sets the alignment
3419 for the type to the largest alignment which is ever used for any data
3420 type on the target machine you are compiling for. Doing this can often
3421 make copy operations more efficient, because the compiler can use
3422 whatever instructions copy the biggest chunks of memory when performing
3423 copies to or from the variables which have types that you have aligned
3426 In the example above, if the size of each @code{short} is 2 bytes, then
3427 the size of the entire @code{struct S} type is 6 bytes. The smallest
3428 power of two which is greater than or equal to that is 8, so the
3429 compiler sets the alignment for the entire @code{struct S} type to 8
3432 Note that although you can ask the compiler to select a time-efficient
3433 alignment for a given type and then declare only individual stand-alone
3434 objects of that type, the compiler's ability to select a time-efficient
3435 alignment is primarily useful only when you plan to create arrays of
3436 variables having the relevant (efficiently aligned) type. If you
3437 declare or use arrays of variables of an efficiently-aligned type, then
3438 it is likely that your program will also be doing pointer arithmetic (or
3439 subscripting, which amounts to the same thing) on pointers to the
3440 relevant type, and the code that the compiler generates for these
3441 pointer arithmetic operations will often be more efficient for
3442 efficiently-aligned types than for other types.
3444 The @code{aligned} attribute can only increase the alignment; but you
3445 can decrease it by specifying @code{packed} as well. See below.
3447 Note that the effectiveness of @code{aligned} attributes may be limited
3448 by inherent limitations in your linker. On many systems, the linker is
3449 only able to arrange for variables to be aligned up to a certain maximum
3450 alignment. (For some linkers, the maximum supported alignment may
3451 be very very small.) If your linker is only able to align variables
3452 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3453 in an @code{__attribute__} will still only provide you with 8 byte
3454 alignment. See your linker documentation for further information.
3457 This attribute, attached to @code{struct} or @code{union} type
3458 definition, specifies that each member of the structure or union is
3459 placed to minimize the memory required. When attached to an @code{enum}
3460 definition, it indicates that the smallest integral type should be used.
3462 @opindex fshort-enums
3463 Specifying this attribute for @code{struct} and @code{union} types is
3464 equivalent to specifying the @code{packed} attribute on each of the
3465 structure or union members. Specifying the @option{-fshort-enums}
3466 flag on the line is equivalent to specifying the @code{packed}
3467 attribute on all @code{enum} definitions.
3469 In the following example @code{struct my_packed_struct}'s members are
3470 packed closely together, but the internal layout of its @code{s} member
3471 is not packed -- to do that, @code{struct my_unpacked_struct} would need to
3475 struct my_unpacked_struct
3481 struct my_packed_struct __attribute__ ((__packed__))
3485 struct my_unpacked_struct s;
3489 You may only specify this attribute on the definition of a @code{enum},
3490 @code{struct} or @code{union}, not on a @code{typedef} which does not
3491 also define the enumerated type, structure or union.
3493 @item transparent_union
3494 This attribute, attached to a @code{union} type definition, indicates
3495 that any function parameter having that union type causes calls to that
3496 function to be treated in a special way.
3498 First, the argument corresponding to a transparent union type can be of
3499 any type in the union; no cast is required. Also, if the union contains
3500 a pointer type, the corresponding argument can be a null pointer
3501 constant or a void pointer expression; and if the union contains a void
3502 pointer type, the corresponding argument can be any pointer expression.
3503 If the union member type is a pointer, qualifiers like @code{const} on
3504 the referenced type must be respected, just as with normal pointer
3507 Second, the argument is passed to the function using the calling
3508 conventions of the first member of the transparent union, not the calling
3509 conventions of the union itself. All members of the union must have the
3510 same machine representation; this is necessary for this argument passing
3513 Transparent unions are designed for library functions that have multiple
3514 interfaces for compatibility reasons. For example, suppose the
3515 @code{wait} function must accept either a value of type @code{int *} to
3516 comply with Posix, or a value of type @code{union wait *} to comply with
3517 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3518 @code{wait} would accept both kinds of arguments, but it would also
3519 accept any other pointer type and this would make argument type checking
3520 less useful. Instead, @code{<sys/wait.h>} might define the interface
3528 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3530 pid_t wait (wait_status_ptr_t);
3533 This interface allows either @code{int *} or @code{union wait *}
3534 arguments to be passed, using the @code{int *} calling convention.
3535 The program can call @code{wait} with arguments of either type:
3538 int w1 () @{ int w; return wait (&w); @}
3539 int w2 () @{ union wait w; return wait (&w); @}
3542 With this interface, @code{wait}'s implementation might look like this:
3545 pid_t wait (wait_status_ptr_t p)
3547 return waitpid (-1, p.__ip, 0);
3552 When attached to a type (including a @code{union} or a @code{struct}),
3553 this attribute means that variables of that type are meant to appear
3554 possibly unused. GCC will not produce a warning for any variables of
3555 that type, even if the variable appears to do nothing. This is often
3556 the case with lock or thread classes, which are usually defined and then
3557 not referenced, but contain constructors and destructors that have
3558 nontrivial bookkeeping functions.
3561 The @code{deprecated} attribute results in a warning if the type
3562 is used anywhere in the source file. This is useful when identifying
3563 types that are expected to be removed in a future version of a program.
3564 If possible, the warning also includes the location of the declaration
3565 of the deprecated type, to enable users to easily find further
3566 information about why the type is deprecated, or what they should do
3567 instead. Note that the warnings only occur for uses and then only
3568 if the type is being applied to an identifier that itself is not being
3569 declared as deprecated.
3572 typedef int T1 __attribute__ ((deprecated));
3576 typedef T1 T3 __attribute__ ((deprecated));
3577 T3 z __attribute__ ((deprecated));
3580 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3581 warning is issued for line 4 because T2 is not explicitly
3582 deprecated. Line 5 has no warning because T3 is explicitly
3583 deprecated. Similarly for line 6.
3585 The @code{deprecated} attribute can also be used for functions and
3586 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3589 Accesses to objects with types with this attribute are not subjected to
3590 type-based alias analysis, but are instead assumed to be able to alias
3591 any other type of objects, just like the @code{char} type. See
3592 @option{-fstrict-aliasing} for more information on aliasing issues.
3597 typedef short __attribute__((__may_alias__)) short_a;
3603 short_a *b = (short_a *) &a;
3607 if (a == 0x12345678)
3614 If you replaced @code{short_a} with @code{short} in the variable
3615 declaration, the above program would abort when compiled with
3616 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3617 above in recent GCC versions.
3619 @subsection i386 Type Attributes
3621 Two attributes are currently defined for i386 configurations:
3622 @code{ms_struct} and @code{gcc_struct}
3626 @cindex @code{ms_struct}
3627 @cindex @code{gcc_struct}
3629 If @code{packed} is used on a structure, or if bit-fields are used
3630 it may be that the Microsoft ABI packs them differently
3631 than GCC would normally pack them. Particularly when moving packed
3632 data between functions compiled with GCC and the native Microsoft compiler
3633 (either via function call or as data in a file), it may be necessary to access
3636 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3637 compilers to match the native Microsoft compiler.
3640 To specify multiple attributes, separate them by commas within the
3641 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3645 @section An Inline Function is As Fast As a Macro
3646 @cindex inline functions
3647 @cindex integrating function code
3649 @cindex macros, inline alternative
3651 By declaring a function @code{inline}, you can direct GCC to
3652 integrate that function's code into the code for its callers. This
3653 makes execution faster by eliminating the function-call overhead; in
3654 addition, if any of the actual argument values are constant, their known
3655 values may permit simplifications at compile time so that not all of the
3656 inline function's code needs to be included. The effect on code size is
3657 less predictable; object code may be larger or smaller with function
3658 inlining, depending on the particular case. Inlining of functions is an
3659 optimization and it really ``works'' only in optimizing compilation. If
3660 you don't use @option{-O}, no function is really inline.
3662 Inline functions are included in the ISO C99 standard, but there are
3663 currently substantial differences between what GCC implements and what
3664 the ISO C99 standard requires.
3666 To declare a function inline, use the @code{inline} keyword in its
3667 declaration, like this:
3677 (If you are writing a header file to be included in ISO C programs, write
3678 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3679 You can also make all ``simple enough'' functions inline with the option
3680 @option{-finline-functions}.
3683 Note that certain usages in a function definition can make it unsuitable
3684 for inline substitution. Among these usages are: use of varargs, use of
3685 alloca, use of variable sized data types (@pxref{Variable Length}),
3686 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3687 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3688 will warn when a function marked @code{inline} could not be substituted,
3689 and will give the reason for the failure.
3691 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3692 does not affect the linkage of the function.
3694 @cindex automatic @code{inline} for C++ member fns
3695 @cindex @code{inline} automatic for C++ member fns
3696 @cindex member fns, automatically @code{inline}
3697 @cindex C++ member fns, automatically @code{inline}
3698 @opindex fno-default-inline
3699 GCC automatically inlines member functions defined within the class
3700 body of C++ programs even if they are not explicitly declared
3701 @code{inline}. (You can override this with @option{-fno-default-inline};
3702 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3704 @cindex inline functions, omission of
3705 @opindex fkeep-inline-functions
3706 When a function is both inline and @code{static}, if all calls to the
3707 function are integrated into the caller, and the function's address is
3708 never used, then the function's own assembler code is never referenced.
3709 In this case, GCC does not actually output assembler code for the
3710 function, unless you specify the option @option{-fkeep-inline-functions}.
3711 Some calls cannot be integrated for various reasons (in particular,
3712 calls that precede the function's definition cannot be integrated, and
3713 neither can recursive calls within the definition). If there is a
3714 nonintegrated call, then the function is compiled to assembler code as
3715 usual. The function must also be compiled as usual if the program
3716 refers to its address, because that can't be inlined.
3718 @cindex non-static inline function
3719 When an inline function is not @code{static}, then the compiler must assume
3720 that there may be calls from other source files; since a global symbol can
3721 be defined only once in any program, the function must not be defined in
3722 the other source files, so the calls therein cannot be integrated.
3723 Therefore, a non-@code{static} inline function is always compiled on its
3724 own in the usual fashion.
3726 If you specify both @code{inline} and @code{extern} in the function
3727 definition, then the definition is used only for inlining. In no case
3728 is the function compiled on its own, not even if you refer to its
3729 address explicitly. Such an address becomes an external reference, as
3730 if you had only declared the function, and had not defined it.
3732 This combination of @code{inline} and @code{extern} has almost the
3733 effect of a macro. The way to use it is to put a function definition in
3734 a header file with these keywords, and put another copy of the
3735 definition (lacking @code{inline} and @code{extern}) in a library file.
3736 The definition in the header file will cause most calls to the function
3737 to be inlined. If any uses of the function remain, they will refer to
3738 the single copy in the library.
3740 Since GCC eventually will implement ISO C99 semantics for
3741 inline functions, it is best to use @code{static inline} only
3742 to guarantee compatibility. (The
3743 existing semantics will remain available when @option{-std=gnu89} is
3744 specified, but eventually the default will be @option{-std=gnu99} and
3745 that will implement the C99 semantics, though it does not do so yet.)
3747 GCC does not inline any functions when not optimizing unless you specify
3748 the @samp{always_inline} attribute for the function, like this:
3752 inline void foo (const char) __attribute__((always_inline));
3756 @section Assembler Instructions with C Expression Operands
3757 @cindex extended @code{asm}
3758 @cindex @code{asm} expressions
3759 @cindex assembler instructions
3762 In an assembler instruction using @code{asm}, you can specify the
3763 operands of the instruction using C expressions. This means you need not
3764 guess which registers or memory locations will contain the data you want
3767 You must specify an assembler instruction template much like what
3768 appears in a machine description, plus an operand constraint string for
3771 For example, here is how to use the 68881's @code{fsinx} instruction:
3774 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3778 Here @code{angle} is the C expression for the input operand while
3779 @code{result} is that of the output operand. Each has @samp{"f"} as its
3780 operand constraint, saying that a floating point register is required.
3781 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3782 output operands' constraints must use @samp{=}. The constraints use the
3783 same language used in the machine description (@pxref{Constraints}).
3785 Each operand is described by an operand-constraint string followed by
3786 the C expression in parentheses. A colon separates the assembler
3787 template from the first output operand and another separates the last
3788 output operand from the first input, if any. Commas separate the
3789 operands within each group. The total number of operands is currently
3790 limited to 30; this limitation may be lifted in some future version of
3793 If there are no output operands but there are input operands, you must
3794 place two consecutive colons surrounding the place where the output
3797 As of GCC version 3.1, it is also possible to specify input and output
3798 operands using symbolic names which can be referenced within the
3799 assembler code. These names are specified inside square brackets
3800 preceding the constraint string, and can be referenced inside the
3801 assembler code using @code{%[@var{name}]} instead of a percentage sign
3802 followed by the operand number. Using named operands the above example
3806 asm ("fsinx %[angle],%[output]"
3807 : [output] "=f" (result)
3808 : [angle] "f" (angle));
3812 Note that the symbolic operand names have no relation whatsoever to
3813 other C identifiers. You may use any name you like, even those of
3814 existing C symbols, but you must ensure that no two operands within the same
3815 assembler construct use the same symbolic name.
3817 Output operand expressions must be lvalues; the compiler can check this.
3818 The input operands need not be lvalues. The compiler cannot check
3819 whether the operands have data types that are reasonable for the
3820 instruction being executed. It does not parse the assembler instruction
3821 template and does not know what it means or even whether it is valid
3822 assembler input. The extended @code{asm} feature is most often used for
3823 machine instructions the compiler itself does not know exist. If
3824 the output expression cannot be directly addressed (for example, it is a
3825 bit-field), your constraint must allow a register. In that case, GCC
3826 will use the register as the output of the @code{asm}, and then store
3827 that register into the output.
3829 The ordinary output operands must be write-only; GCC will assume that
3830 the values in these operands before the instruction are dead and need
3831 not be generated. Extended asm supports input-output or read-write
3832 operands. Use the constraint character @samp{+} to indicate such an
3833 operand and list it with the output operands. You should only use
3834 read-write operands when the constraints for the operand (or the
3835 operand in which only some of the bits are to be changed) allow a
3838 You may, as an alternative, logically split its function into two
3839 separate operands, one input operand and one write-only output
3840 operand. The connection between them is expressed by constraints
3841 which say they need to be in the same location when the instruction
3842 executes. You can use the same C expression for both operands, or
3843 different expressions. For example, here we write the (fictitious)
3844 @samp{combine} instruction with @code{bar} as its read-only source
3845 operand and @code{foo} as its read-write destination:
3848 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3852 The constraint @samp{"0"} for operand 1 says that it must occupy the
3853 same location as operand 0. A number in constraint is allowed only in
3854 an input operand and it must refer to an output operand.
3856 Only a number in the constraint can guarantee that one operand will be in
3857 the same place as another. The mere fact that @code{foo} is the value
3858 of both operands is not enough to guarantee that they will be in the
3859 same place in the generated assembler code. The following would not
3863 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3866 Various optimizations or reloading could cause operands 0 and 1 to be in
3867 different registers; GCC knows no reason not to do so. For example, the
3868 compiler might find a copy of the value of @code{foo} in one register and
3869 use it for operand 1, but generate the output operand 0 in a different
3870 register (copying it afterward to @code{foo}'s own address). Of course,
3871 since the register for operand 1 is not even mentioned in the assembler
3872 code, the result will not work, but GCC can't tell that.
3874 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3875 the operand number for a matching constraint. For example:
3878 asm ("cmoveq %1,%2,%[result]"
3879 : [result] "=r"(result)
3880 : "r" (test), "r"(new), "[result]"(old));
3883 Some instructions clobber specific hard registers. To describe this,
3884 write a third colon after the input operands, followed by the names of
3885 the clobbered hard registers (given as strings). Here is a realistic
3886 example for the VAX:
3889 asm volatile ("movc3 %0,%1,%2"
3891 : "g" (from), "g" (to), "g" (count)
3892 : "r0", "r1", "r2", "r3", "r4", "r5");
3895 You may not write a clobber description in a way that overlaps with an
3896 input or output operand. For example, you may not have an operand
3897 describing a register class with one member if you mention that register
3898 in the clobber list. Variables declared to live in specific registers
3899 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3900 have no part mentioned in the clobber description.
3901 There is no way for you to specify that an input
3902 operand is modified without also specifying it as an output
3903 operand. Note that if all the output operands you specify are for this
3904 purpose (and hence unused), you will then also need to specify
3905 @code{volatile} for the @code{asm} construct, as described below, to
3906 prevent GCC from deleting the @code{asm} statement as unused.
3908 If you refer to a particular hardware register from the assembler code,
3909 you will probably have to list the register after the third colon to
3910 tell the compiler the register's value is modified. In some assemblers,
3911 the register names begin with @samp{%}; to produce one @samp{%} in the
3912 assembler code, you must write @samp{%%} in the input.
3914 If your assembler instruction can alter the condition code register, add
3915 @samp{cc} to the list of clobbered registers. GCC on some machines
3916 represents the condition codes as a specific hardware register;
3917 @samp{cc} serves to name this register. On other machines, the
3918 condition code is handled differently, and specifying @samp{cc} has no
3919 effect. But it is valid no matter what the machine.
3921 If your assembler instructions access memory in an unpredictable
3922 fashion, add @samp{memory} to the list of clobbered registers. This
3923 will cause GCC to not keep memory values cached in registers across the
3924 assembler instruction and not optimize stores or loads to that memory.
3925 You will also want to add the @code{volatile} keyword if the memory
3926 affected is not listed in the inputs or outputs of the @code{asm}, as
3927 the @samp{memory} clobber does not count as a side-effect of the
3928 @code{asm}. If you know how large the accessed memory is, you can add
3929 it as input or output but if this is not known, you should add
3930 @samp{memory}. As an example, if you access ten bytes of a string, you
3931 can use a memory input like:
3934 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
3937 Note that in the following example the memory input is necessary,
3938 otherwise GCC might optimize the store to @code{x} away:
3945 asm ("magic stuff accessing an 'int' pointed to by '%1'"
3946 "=&d" (r) : "a" (y), "m" (*y));
3951 You can put multiple assembler instructions together in a single
3952 @code{asm} template, separated by the characters normally used in assembly
3953 code for the system. A combination that works in most places is a newline
3954 to break the line, plus a tab character to move to the instruction field
3955 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3956 assembler allows semicolons as a line-breaking character. Note that some
3957 assembler dialects use semicolons to start a comment.
3958 The input operands are guaranteed not to use any of the clobbered
3959 registers, and neither will the output operands' addresses, so you can
3960 read and write the clobbered registers as many times as you like. Here
3961 is an example of multiple instructions in a template; it assumes the
3962 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3965 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3967 : "g" (from), "g" (to)
3971 Unless an output operand has the @samp{&} constraint modifier, GCC
3972 may allocate it in the same register as an unrelated input operand, on
3973 the assumption the inputs are consumed before the outputs are produced.
3974 This assumption may be false if the assembler code actually consists of
3975 more than one instruction. In such a case, use @samp{&} for each output
3976 operand that may not overlap an input. @xref{Modifiers}.
3978 If you want to test the condition code produced by an assembler
3979 instruction, you must include a branch and a label in the @code{asm}
3980 construct, as follows:
3983 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3989 This assumes your assembler supports local labels, as the GNU assembler
3990 and most Unix assemblers do.
3992 Speaking of labels, jumps from one @code{asm} to another are not
3993 supported. The compiler's optimizers do not know about these jumps, and
3994 therefore they cannot take account of them when deciding how to
3997 @cindex macros containing @code{asm}
3998 Usually the most convenient way to use these @code{asm} instructions is to
3999 encapsulate them in macros that look like functions. For example,
4003 (@{ double __value, __arg = (x); \
4004 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4009 Here the variable @code{__arg} is used to make sure that the instruction
4010 operates on a proper @code{double} value, and to accept only those
4011 arguments @code{x} which can convert automatically to a @code{double}.
4013 Another way to make sure the instruction operates on the correct data
4014 type is to use a cast in the @code{asm}. This is different from using a
4015 variable @code{__arg} in that it converts more different types. For
4016 example, if the desired type were @code{int}, casting the argument to
4017 @code{int} would accept a pointer with no complaint, while assigning the
4018 argument to an @code{int} variable named @code{__arg} would warn about
4019 using a pointer unless the caller explicitly casts it.
4021 If an @code{asm} has output operands, GCC assumes for optimization
4022 purposes the instruction has no side effects except to change the output
4023 operands. This does not mean instructions with a side effect cannot be
4024 used, but you must be careful, because the compiler may eliminate them
4025 if the output operands aren't used, or move them out of loops, or
4026 replace two with one if they constitute a common subexpression. Also,
4027 if your instruction does have a side effect on a variable that otherwise
4028 appears not to change, the old value of the variable may be reused later
4029 if it happens to be found in a register.
4031 You can prevent an @code{asm} instruction from being deleted, moved
4032 significantly, or combined, by writing the keyword @code{volatile} after
4033 the @code{asm}. For example:
4036 #define get_and_set_priority(new) \
4038 asm volatile ("get_and_set_priority %0, %1" \
4039 : "=g" (__old) : "g" (new)); \
4044 If you write an @code{asm} instruction with no outputs, GCC will know
4045 the instruction has side-effects and will not delete the instruction or
4046 move it outside of loops.
4048 The @code{volatile} keyword indicates that the instruction has
4049 important side-effects. GCC will not delete a volatile @code{asm} if
4050 it is reachable. (The instruction can still be deleted if GCC can
4051 prove that control-flow will never reach the location of the
4052 instruction.) In addition, GCC will not reschedule instructions
4053 across a volatile @code{asm} instruction. For example:
4056 *(volatile int *)addr = foo;
4057 asm volatile ("eieio" : : );
4061 Assume @code{addr} contains the address of a memory mapped device
4062 register. The PowerPC @code{eieio} instruction (Enforce In-order
4063 Execution of I/O) tells the CPU to make sure that the store to that
4064 device register happens before it issues any other I/O@.
4066 Note that even a volatile @code{asm} instruction can be moved in ways
4067 that appear insignificant to the compiler, such as across jump
4068 instructions. You can't expect a sequence of volatile @code{asm}
4069 instructions to remain perfectly consecutive. If you want consecutive
4070 output, use a single @code{asm}. Also, GCC will perform some
4071 optimizations across a volatile @code{asm} instruction; GCC does not
4072 ``forget everything'' when it encounters a volatile @code{asm}
4073 instruction the way some other compilers do.
4075 An @code{asm} instruction without any operands or clobbers (an ``old
4076 style'' @code{asm}) will be treated identically to a volatile
4077 @code{asm} instruction.
4079 It is a natural idea to look for a way to give access to the condition
4080 code left by the assembler instruction. However, when we attempted to
4081 implement this, we found no way to make it work reliably. The problem
4082 is that output operands might need reloading, which would result in
4083 additional following ``store'' instructions. On most machines, these
4084 instructions would alter the condition code before there was time to
4085 test it. This problem doesn't arise for ordinary ``test'' and
4086 ``compare'' instructions because they don't have any output operands.
4088 For reasons similar to those described above, it is not possible to give
4089 an assembler instruction access to the condition code left by previous
4092 If you are writing a header file that should be includable in ISO C
4093 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4096 @subsection Size of an @code{asm}
4098 Some targets require that GCC track the size of each instruction used in
4099 order to generate correct code. Because the final length of an
4100 @code{asm} is only known by the assembler, GCC must make an estimate as
4101 to how big it will be. The estimate is formed by counting the number of
4102 statements in the pattern of the @code{asm} and multiplying that by the
4103 length of the longest instruction on that processor. Statements in the
4104 @code{asm} are identified by newline characters and whatever statement
4105 separator characters are supported by the assembler; on most processors
4106 this is the `@code{;}' character.
4108 Normally, GCC's estimate is perfectly adequate to ensure that correct
4109 code is generated, but it is possible to confuse the compiler if you use
4110 pseudo instructions or assembler macros that expand into multiple real
4111 instructions or if you use assembler directives that expand to more
4112 space in the object file than would be needed for a single instruction.
4113 If this happens then the assembler will produce a diagnostic saying that
4114 a label is unreachable.
4116 @subsection i386 floating point asm operands
4118 There are several rules on the usage of stack-like regs in
4119 asm_operands insns. These rules apply only to the operands that are
4124 Given a set of input regs that die in an asm_operands, it is
4125 necessary to know which are implicitly popped by the asm, and
4126 which must be explicitly popped by gcc.
4128 An input reg that is implicitly popped by the asm must be
4129 explicitly clobbered, unless it is constrained to match an
4133 For any input reg that is implicitly popped by an asm, it is
4134 necessary to know how to adjust the stack to compensate for the pop.
4135 If any non-popped input is closer to the top of the reg-stack than
4136 the implicitly popped reg, it would not be possible to know what the
4137 stack looked like---it's not clear how the rest of the stack ``slides
4140 All implicitly popped input regs must be closer to the top of
4141 the reg-stack than any input that is not implicitly popped.
4143 It is possible that if an input dies in an insn, reload might
4144 use the input reg for an output reload. Consider this example:
4147 asm ("foo" : "=t" (a) : "f" (b));
4150 This asm says that input B is not popped by the asm, and that
4151 the asm pushes a result onto the reg-stack, i.e., the stack is one
4152 deeper after the asm than it was before. But, it is possible that
4153 reload will think that it can use the same reg for both the input and
4154 the output, if input B dies in this insn.
4156 If any input operand uses the @code{f} constraint, all output reg
4157 constraints must use the @code{&} earlyclobber.
4159 The asm above would be written as
4162 asm ("foo" : "=&t" (a) : "f" (b));
4166 Some operands need to be in particular places on the stack. All
4167 output operands fall in this category---there is no other way to
4168 know which regs the outputs appear in unless the user indicates
4169 this in the constraints.
4171 Output operands must specifically indicate which reg an output
4172 appears in after an asm. @code{=f} is not allowed: the operand
4173 constraints must select a class with a single reg.
4176 Output operands may not be ``inserted'' between existing stack regs.
4177 Since no 387 opcode uses a read/write operand, all output operands
4178 are dead before the asm_operands, and are pushed by the asm_operands.
4179 It makes no sense to push anywhere but the top of the reg-stack.
4181 Output operands must start at the top of the reg-stack: output
4182 operands may not ``skip'' a reg.
4185 Some asm statements may need extra stack space for internal
4186 calculations. This can be guaranteed by clobbering stack registers
4187 unrelated to the inputs and outputs.
4191 Here are a couple of reasonable asms to want to write. This asm
4192 takes one input, which is internally popped, and produces two outputs.
4195 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4198 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4199 and replaces them with one output. The user must code the @code{st(1)}
4200 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4203 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4209 @section Controlling Names Used in Assembler Code
4210 @cindex assembler names for identifiers
4211 @cindex names used in assembler code
4212 @cindex identifiers, names in assembler code
4214 You can specify the name to be used in the assembler code for a C
4215 function or variable by writing the @code{asm} (or @code{__asm__})
4216 keyword after the declarator as follows:
4219 int foo asm ("myfoo") = 2;
4223 This specifies that the name to be used for the variable @code{foo} in
4224 the assembler code should be @samp{myfoo} rather than the usual
4227 On systems where an underscore is normally prepended to the name of a C
4228 function or variable, this feature allows you to define names for the
4229 linker that do not start with an underscore.
4231 It does not make sense to use this feature with a non-static local
4232 variable since such variables do not have assembler names. If you are
4233 trying to put the variable in a particular register, see @ref{Explicit
4234 Reg Vars}. GCC presently accepts such code with a warning, but will
4235 probably be changed to issue an error, rather than a warning, in the
4238 You cannot use @code{asm} in this way in a function @emph{definition}; but
4239 you can get the same effect by writing a declaration for the function
4240 before its definition and putting @code{asm} there, like this:
4243 extern func () asm ("FUNC");
4250 It is up to you to make sure that the assembler names you choose do not
4251 conflict with any other assembler symbols. Also, you must not use a
4252 register name; that would produce completely invalid assembler code. GCC
4253 does not as yet have the ability to store static variables in registers.
4254 Perhaps that will be added.
4256 @node Explicit Reg Vars
4257 @section Variables in Specified Registers
4258 @cindex explicit register variables
4259 @cindex variables in specified registers
4260 @cindex specified registers
4261 @cindex registers, global allocation
4263 GNU C allows you to put a few global variables into specified hardware
4264 registers. You can also specify the register in which an ordinary
4265 register variable should be allocated.
4269 Global register variables reserve registers throughout the program.
4270 This may be useful in programs such as programming language
4271 interpreters which have a couple of global variables that are accessed
4275 Local register variables in specific registers do not reserve the
4276 registers. The compiler's data flow analysis is capable of determining
4277 where the specified registers contain live values, and where they are
4278 available for other uses. Stores into local register variables may be deleted
4279 when they appear to be dead according to dataflow analysis. References
4280 to local register variables may be deleted or moved or simplified.
4282 These local variables are sometimes convenient for use with the extended
4283 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4284 output of the assembler instruction directly into a particular register.
4285 (This will work provided the register you specify fits the constraints
4286 specified for that operand in the @code{asm}.)
4294 @node Global Reg Vars
4295 @subsection Defining Global Register Variables
4296 @cindex global register variables
4297 @cindex registers, global variables in
4299 You can define a global register variable in GNU C like this:
4302 register int *foo asm ("a5");
4306 Here @code{a5} is the name of the register which should be used. Choose a
4307 register which is normally saved and restored by function calls on your
4308 machine, so that library routines will not clobber it.
4310 Naturally the register name is cpu-dependent, so you would need to
4311 conditionalize your program according to cpu type. The register
4312 @code{a5} would be a good choice on a 68000 for a variable of pointer
4313 type. On machines with register windows, be sure to choose a ``global''
4314 register that is not affected magically by the function call mechanism.
4316 In addition, operating systems on one type of cpu may differ in how they
4317 name the registers; then you would need additional conditionals. For
4318 example, some 68000 operating systems call this register @code{%a5}.
4320 Eventually there may be a way of asking the compiler to choose a register
4321 automatically, but first we need to figure out how it should choose and
4322 how to enable you to guide the choice. No solution is evident.
4324 Defining a global register variable in a certain register reserves that
4325 register entirely for this use, at least within the current compilation.
4326 The register will not be allocated for any other purpose in the functions
4327 in the current compilation. The register will not be saved and restored by
4328 these functions. Stores into this register are never deleted even if they
4329 would appear to be dead, but references may be deleted or moved or
4332 It is not safe to access the global register variables from signal
4333 handlers, or from more than one thread of control, because the system
4334 library routines may temporarily use the register for other things (unless
4335 you recompile them specially for the task at hand).
4337 @cindex @code{qsort}, and global register variables
4338 It is not safe for one function that uses a global register variable to
4339 call another such function @code{foo} by way of a third function
4340 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4341 different source file in which the variable wasn't declared). This is
4342 because @code{lose} might save the register and put some other value there.
4343 For example, you can't expect a global register variable to be available in
4344 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4345 might have put something else in that register. (If you are prepared to
4346 recompile @code{qsort} with the same global register variable, you can
4347 solve this problem.)
4349 If you want to recompile @code{qsort} or other source files which do not
4350 actually use your global register variable, so that they will not use that
4351 register for any other purpose, then it suffices to specify the compiler
4352 option @option{-ffixed-@var{reg}}. You need not actually add a global
4353 register declaration to their source code.
4355 A function which can alter the value of a global register variable cannot
4356 safely be called from a function compiled without this variable, because it
4357 could clobber the value the caller expects to find there on return.
4358 Therefore, the function which is the entry point into the part of the
4359 program that uses the global register variable must explicitly save and
4360 restore the value which belongs to its caller.
4362 @cindex register variable after @code{longjmp}
4363 @cindex global register after @code{longjmp}
4364 @cindex value after @code{longjmp}
4367 On most machines, @code{longjmp} will restore to each global register
4368 variable the value it had at the time of the @code{setjmp}. On some
4369 machines, however, @code{longjmp} will not change the value of global
4370 register variables. To be portable, the function that called @code{setjmp}
4371 should make other arrangements to save the values of the global register
4372 variables, and to restore them in a @code{longjmp}. This way, the same
4373 thing will happen regardless of what @code{longjmp} does.
4375 All global register variable declarations must precede all function
4376 definitions. If such a declaration could appear after function
4377 definitions, the declaration would be too late to prevent the register from
4378 being used for other purposes in the preceding functions.
4380 Global register variables may not have initial values, because an
4381 executable file has no means to supply initial contents for a register.
4383 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4384 registers, but certain library functions, such as @code{getwd}, as well
4385 as the subroutines for division and remainder, modify g3 and g4. g1 and
4386 g2 are local temporaries.
4388 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4389 Of course, it will not do to use more than a few of those.
4391 @node Local Reg Vars
4392 @subsection Specifying Registers for Local Variables
4393 @cindex local variables, specifying registers
4394 @cindex specifying registers for local variables
4395 @cindex registers for local variables
4397 You can define a local register variable with a specified register
4401 register int *foo asm ("a5");
4405 Here @code{a5} is the name of the register which should be used. Note
4406 that this is the same syntax used for defining global register
4407 variables, but for a local variable it would appear within a function.
4409 Naturally the register name is cpu-dependent, but this is not a
4410 problem, since specific registers are most often useful with explicit
4411 assembler instructions (@pxref{Extended Asm}). Both of these things
4412 generally require that you conditionalize your program according to
4415 In addition, operating systems on one type of cpu may differ in how they
4416 name the registers; then you would need additional conditionals. For
4417 example, some 68000 operating systems call this register @code{%a5}.
4419 Defining such a register variable does not reserve the register; it
4420 remains available for other uses in places where flow control determines
4421 the variable's value is not live.
4423 This option does not guarantee that GCC will generate code that has
4424 this variable in the register you specify at all times. You may not
4425 code an explicit reference to this register in an @code{asm} statement
4426 and assume it will always refer to this variable.
4428 Stores into local register variables may be deleted when they appear to be dead
4429 according to dataflow analysis. References to local register variables may
4430 be deleted or moved or simplified.
4432 @node Alternate Keywords
4433 @section Alternate Keywords
4434 @cindex alternate keywords
4435 @cindex keywords, alternate
4437 @option{-ansi} and the various @option{-std} options disable certain
4438 keywords. This causes trouble when you want to use GNU C extensions, or
4439 a general-purpose header file that should be usable by all programs,
4440 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4441 @code{inline} are not available in programs compiled with
4442 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4443 program compiled with @option{-std=c99}). The ISO C99 keyword
4444 @code{restrict} is only available when @option{-std=gnu99} (which will
4445 eventually be the default) or @option{-std=c99} (or the equivalent
4446 @option{-std=iso9899:1999}) is used.
4448 The way to solve these problems is to put @samp{__} at the beginning and
4449 end of each problematical keyword. For example, use @code{__asm__}
4450 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4452 Other C compilers won't accept these alternative keywords; if you want to
4453 compile with another compiler, you can define the alternate keywords as
4454 macros to replace them with the customary keywords. It looks like this:
4462 @findex __extension__
4464 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4466 prevent such warnings within one expression by writing
4467 @code{__extension__} before the expression. @code{__extension__} has no
4468 effect aside from this.
4470 @node Incomplete Enums
4471 @section Incomplete @code{enum} Types
4473 You can define an @code{enum} tag without specifying its possible values.
4474 This results in an incomplete type, much like what you get if you write
4475 @code{struct foo} without describing the elements. A later declaration
4476 which does specify the possible values completes the type.
4478 You can't allocate variables or storage using the type while it is
4479 incomplete. However, you can work with pointers to that type.
4481 This extension may not be very useful, but it makes the handling of
4482 @code{enum} more consistent with the way @code{struct} and @code{union}
4485 This extension is not supported by GNU C++.
4487 @node Function Names
4488 @section Function Names as Strings
4489 @cindex @code{__func__} identifier
4490 @cindex @code{__FUNCTION__} identifier
4491 @cindex @code{__PRETTY_FUNCTION__} identifier
4493 GCC provides three magic variables which hold the name of the current
4494 function, as a string. The first of these is @code{__func__}, which
4495 is part of the C99 standard:
4498 The identifier @code{__func__} is implicitly declared by the translator
4499 as if, immediately following the opening brace of each function
4500 definition, the declaration
4503 static const char __func__[] = "function-name";
4506 appeared, where function-name is the name of the lexically-enclosing
4507 function. This name is the unadorned name of the function.
4510 @code{__FUNCTION__} is another name for @code{__func__}. Older
4511 versions of GCC recognize only this name. However, it is not
4512 standardized. For maximum portability, we recommend you use
4513 @code{__func__}, but provide a fallback definition with the
4517 #if __STDC_VERSION__ < 199901L
4519 # define __func__ __FUNCTION__
4521 # define __func__ "<unknown>"
4526 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4527 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4528 the type signature of the function as well as its bare name. For
4529 example, this program:
4533 extern int printf (char *, ...);
4540 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4541 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4559 __PRETTY_FUNCTION__ = void a::sub(int)
4562 These identifiers are not preprocessor macros. In GCC 3.3 and
4563 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4564 were treated as string literals; they could be used to initialize
4565 @code{char} arrays, and they could be concatenated with other string
4566 literals. GCC 3.4 and later treat them as variables, like
4567 @code{__func__}. In C++, @code{__FUNCTION__} and
4568 @code{__PRETTY_FUNCTION__} have always been variables.
4570 @node Return Address
4571 @section Getting the Return or Frame Address of a Function
4573 These functions may be used to get information about the callers of a
4576 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4577 This function returns the return address of the current function, or of
4578 one of its callers. The @var{level} argument is number of frames to
4579 scan up the call stack. A value of @code{0} yields the return address
4580 of the current function, a value of @code{1} yields the return address
4581 of the caller of the current function, and so forth. When inlining
4582 the expected behavior is that the function will return the address of
4583 the function that will be returned to. To work around this behavior use
4584 the @code{noinline} function attribute.
4586 The @var{level} argument must be a constant integer.
4588 On some machines it may be impossible to determine the return address of
4589 any function other than the current one; in such cases, or when the top
4590 of the stack has been reached, this function will return @code{0} or a
4591 random value. In addition, @code{__builtin_frame_address} may be used
4592 to determine if the top of the stack has been reached.
4594 This function should only be used with a nonzero argument for debugging
4598 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4599 This function is similar to @code{__builtin_return_address}, but it
4600 returns the address of the function frame rather than the return address
4601 of the function. Calling @code{__builtin_frame_address} with a value of
4602 @code{0} yields the frame address of the current function, a value of
4603 @code{1} yields the frame address of the caller of the current function,
4606 The frame is the area on the stack which holds local variables and saved
4607 registers. The frame address is normally the address of the first word
4608 pushed on to the stack by the function. However, the exact definition
4609 depends upon the processor and the calling convention. If the processor
4610 has a dedicated frame pointer register, and the function has a frame,
4611 then @code{__builtin_frame_address} will return the value of the frame
4614 On some machines it may be impossible to determine the frame address of
4615 any function other than the current one; in such cases, or when the top
4616 of the stack has been reached, this function will return @code{0} if
4617 the first frame pointer is properly initialized by the startup code.
4619 This function should only be used with a nonzero argument for debugging
4623 @node Vector Extensions
4624 @section Using vector instructions through built-in functions
4626 On some targets, the instruction set contains SIMD vector instructions that
4627 operate on multiple values contained in one large register at the same time.
4628 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4631 The first step in using these extensions is to provide the necessary data
4632 types. This should be done using an appropriate @code{typedef}:
4635 typedef int v4si __attribute__ ((vector_size (16)));
4638 The @code{int} type specifies the base type, while the attribute specifies
4639 the vector size for the variable, measured in bytes. For example, the
4640 declaration above causes the compiler to set the mode for the @code{v4si}
4641 type to be 16 bytes wide and divided into @code{int} sized units. For
4642 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4643 corresponding mode of @code{foo} will be @acronym{V4SI}.
4645 The @code{vector_size} attribute is only applicable to integral and
4646 float scalars, although arrays, pointers, and function return values
4647 are allowed in conjunction with this construct.
4649 All the basic integer types can be used as base types, both as signed
4650 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4651 @code{long long}. In addition, @code{float} and @code{double} can be
4652 used to build floating-point vector types.
4654 Specifying a combination that is not valid for the current architecture
4655 will cause GCC to synthesize the instructions using a narrower mode.
4656 For example, if you specify a variable of type @code{V4SI} and your
4657 architecture does not allow for this specific SIMD type, GCC will
4658 produce code that uses 4 @code{SIs}.
4660 The types defined in this manner can be used with a subset of normal C
4661 operations. Currently, GCC will allow using the following operators
4662 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4664 The operations behave like C++ @code{valarrays}. Addition is defined as
4665 the addition of the corresponding elements of the operands. For
4666 example, in the code below, each of the 4 elements in @var{a} will be
4667 added to the corresponding 4 elements in @var{b} and the resulting
4668 vector will be stored in @var{c}.
4671 typedef int v4si __attribute__ ((vector_size (16)));
4678 Subtraction, multiplication, division, and the logical operations
4679 operate in a similar manner. Likewise, the result of using the unary
4680 minus or complement operators on a vector type is a vector whose
4681 elements are the negative or complemented values of the corresponding
4682 elements in the operand.
4684 You can declare variables and use them in function calls and returns, as
4685 well as in assignments and some casts. You can specify a vector type as
4686 a return type for a function. Vector types can also be used as function
4687 arguments. It is possible to cast from one vector type to another,
4688 provided they are of the same size (in fact, you can also cast vectors
4689 to and from other datatypes of the same size).
4691 You cannot operate between vectors of different lengths or different
4692 signedness without a cast.
4694 A port that supports hardware vector operations, usually provides a set
4695 of built-in functions that can be used to operate on vectors. For
4696 example, a function to add two vectors and multiply the result by a
4697 third could look like this:
4700 v4si f (v4si a, v4si b, v4si c)
4702 v4si tmp = __builtin_addv4si (a, b);
4703 return __builtin_mulv4si (tmp, c);
4710 @findex __builtin_offsetof
4712 GCC implements for both C and C++ a syntactic extension to implement
4713 the @code{offsetof} macro.
4717 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4719 offsetof_member_designator:
4721 | offsetof_member_designator "." @code{identifier}
4722 | offsetof_member_designator "[" @code{expr} "]"
4725 This extension is sufficient such that
4728 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4731 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4732 may be dependent. In either case, @var{member} may consist of a single
4733 identifier, or a sequence of member accesses and array references.
4735 @node Other Builtins
4736 @section Other built-in functions provided by GCC
4737 @cindex built-in functions
4738 @findex __builtin_isgreater
4739 @findex __builtin_isgreaterequal
4740 @findex __builtin_isless
4741 @findex __builtin_islessequal
4742 @findex __builtin_islessgreater
4743 @findex __builtin_isunordered
4898 @findex fprintf_unlocked
4900 @findex fputs_unlocked
5010 @findex printf_unlocked
5039 @findex significandf
5040 @findex significandl
5107 GCC provides a large number of built-in functions other than the ones
5108 mentioned above. Some of these are for internal use in the processing
5109 of exceptions or variable-length argument lists and will not be
5110 documented here because they may change from time to time; we do not
5111 recommend general use of these functions.
5113 The remaining functions are provided for optimization purposes.
5115 @opindex fno-builtin
5116 GCC includes built-in versions of many of the functions in the standard
5117 C library. The versions prefixed with @code{__builtin_} will always be
5118 treated as having the same meaning as the C library function even if you
5119 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5120 Many of these functions are only optimized in certain cases; if they are
5121 not optimized in a particular case, a call to the library function will
5126 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5127 @option{-std=c99}), the functions
5128 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5129 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5130 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5131 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5132 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5133 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5134 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5135 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5136 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5137 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5138 @code{significandf}, @code{significandl}, @code{significand},
5139 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5140 @code{strdup}, @code{strfmon}, @code{toascii}, @code{y0f}, @code{y0l},
5141 @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
5143 may be handled as built-in functions.
5144 All these functions have corresponding versions
5145 prefixed with @code{__builtin_}, which may be used even in strict C89
5148 The ISO C99 functions
5149 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5150 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5151 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5152 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5153 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5154 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5155 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5156 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5157 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5158 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5159 @code{cimagl}, @code{cimag}, @code{conjf}, @code{conjl}, @code{conj},
5160 @code{copysignf}, @code{copysignl}, @code{copysign}, @code{cpowf},
5161 @code{cpowl}, @code{cpow}, @code{cprojf}, @code{cprojl}, @code{cproj},
5162 @code{crealf}, @code{creall}, @code{creal}, @code{csinf}, @code{csinhf},
5163 @code{csinhl}, @code{csinh}, @code{csinl}, @code{csin}, @code{csqrtf},
5164 @code{csqrtl}, @code{csqrt}, @code{ctanf}, @code{ctanhf}, @code{ctanhl},
5165 @code{ctanh}, @code{ctanl}, @code{ctan}, @code{erfcf}, @code{erfcl},
5166 @code{erfc}, @code{erff}, @code{erfl}, @code{erf}, @code{exp2f},
5167 @code{exp2l}, @code{exp2}, @code{expm1f}, @code{expm1l}, @code{expm1},
5168 @code{fdimf}, @code{fdiml}, @code{fdim}, @code{fmaf}, @code{fmal},
5169 @code{fmaxf}, @code{fmaxl}, @code{fmax}, @code{fma}, @code{fminf},
5170 @code{fminl}, @code{fmin}, @code{hypotf}, @code{hypotl}, @code{hypot},
5171 @code{ilogbf}, @code{ilogbl}, @code{ilogb}, @code{imaxabs},
5172 @code{isblank}, @code{iswblank}, @code{lgammaf}, @code{lgammal},
5173 @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5174 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5175 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5176 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5177 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5178 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5179 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5180 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5181 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5182 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5183 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5184 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5185 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5186 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5187 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5188 are handled as built-in functions
5189 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5191 There are also built-in versions of the ISO C99 functions
5192 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5193 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5194 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5195 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5196 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5197 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5198 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5199 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5200 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5201 that are recognized in any mode since ISO C90 reserves these names for
5202 the purpose to which ISO C99 puts them. All these functions have
5203 corresponding versions prefixed with @code{__builtin_}.
5205 The ISO C94 functions
5206 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5207 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5208 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5210 are handled as built-in functions
5211 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5213 The ISO C90 functions
5214 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5215 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5216 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5217 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5218 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5219 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5220 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5221 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5222 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5223 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5224 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5225 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5226 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5227 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5228 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5229 @code{vprintf} and @code{vsprintf}
5230 are all recognized as built-in functions unless
5231 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5232 is specified for an individual function). All of these functions have
5233 corresponding versions prefixed with @code{__builtin_}.
5235 GCC provides built-in versions of the ISO C99 floating point comparison
5236 macros that avoid raising exceptions for unordered operands. They have
5237 the same names as the standard macros ( @code{isgreater},
5238 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5239 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5240 prefixed. We intend for a library implementor to be able to simply
5241 @code{#define} each standard macro to its built-in equivalent.
5243 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5245 You can use the built-in function @code{__builtin_types_compatible_p} to
5246 determine whether two types are the same.
5248 This built-in function returns 1 if the unqualified versions of the
5249 types @var{type1} and @var{type2} (which are types, not expressions) are
5250 compatible, 0 otherwise. The result of this built-in function can be
5251 used in integer constant expressions.
5253 This built-in function ignores top level qualifiers (e.g., @code{const},
5254 @code{volatile}). For example, @code{int} is equivalent to @code{const
5257 The type @code{int[]} and @code{int[5]} are compatible. On the other
5258 hand, @code{int} and @code{char *} are not compatible, even if the size
5259 of their types, on the particular architecture are the same. Also, the
5260 amount of pointer indirection is taken into account when determining
5261 similarity. Consequently, @code{short *} is not similar to
5262 @code{short **}. Furthermore, two types that are typedefed are
5263 considered compatible if their underlying types are compatible.
5265 An @code{enum} type is not considered to be compatible with another
5266 @code{enum} type even if both are compatible with the same integer
5267 type; this is what the C standard specifies.
5268 For example, @code{enum @{foo, bar@}} is not similar to
5269 @code{enum @{hot, dog@}}.
5271 You would typically use this function in code whose execution varies
5272 depending on the arguments' types. For example:
5278 if (__builtin_types_compatible_p (typeof (x), long double)) \
5279 tmp = foo_long_double (tmp); \
5280 else if (__builtin_types_compatible_p (typeof (x), double)) \
5281 tmp = foo_double (tmp); \
5282 else if (__builtin_types_compatible_p (typeof (x), float)) \
5283 tmp = foo_float (tmp); \
5290 @emph{Note:} This construct is only available for C.
5294 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5296 You can use the built-in function @code{__builtin_choose_expr} to
5297 evaluate code depending on the value of a constant expression. This
5298 built-in function returns @var{exp1} if @var{const_exp}, which is a
5299 constant expression that must be able to be determined at compile time,
5300 is nonzero. Otherwise it returns 0.
5302 This built-in function is analogous to the @samp{? :} operator in C,
5303 except that the expression returned has its type unaltered by promotion
5304 rules. Also, the built-in function does not evaluate the expression
5305 that was not chosen. For example, if @var{const_exp} evaluates to true,
5306 @var{exp2} is not evaluated even if it has side-effects.
5308 This built-in function can return an lvalue if the chosen argument is an
5311 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5312 type. Similarly, if @var{exp2} is returned, its return type is the same
5319 __builtin_choose_expr ( \
5320 __builtin_types_compatible_p (typeof (x), double), \
5322 __builtin_choose_expr ( \
5323 __builtin_types_compatible_p (typeof (x), float), \
5325 /* @r{The void expression results in a compile-time error} \
5326 @r{when assigning the result to something.} */ \
5330 @emph{Note:} This construct is only available for C. Furthermore, the
5331 unused expression (@var{exp1} or @var{exp2} depending on the value of
5332 @var{const_exp}) may still generate syntax errors. This may change in
5337 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5338 You can use the built-in function @code{__builtin_constant_p} to
5339 determine if a value is known to be constant at compile-time and hence
5340 that GCC can perform constant-folding on expressions involving that
5341 value. The argument of the function is the value to test. The function
5342 returns the integer 1 if the argument is known to be a compile-time
5343 constant and 0 if it is not known to be a compile-time constant. A
5344 return of 0 does not indicate that the value is @emph{not} a constant,
5345 but merely that GCC cannot prove it is a constant with the specified
5346 value of the @option{-O} option.
5348 You would typically use this function in an embedded application where
5349 memory was a critical resource. If you have some complex calculation,
5350 you may want it to be folded if it involves constants, but need to call
5351 a function if it does not. For example:
5354 #define Scale_Value(X) \
5355 (__builtin_constant_p (X) \
5356 ? ((X) * SCALE + OFFSET) : Scale (X))
5359 You may use this built-in function in either a macro or an inline
5360 function. However, if you use it in an inlined function and pass an
5361 argument of the function as the argument to the built-in, GCC will
5362 never return 1 when you call the inline function with a string constant
5363 or compound literal (@pxref{Compound Literals}) and will not return 1
5364 when you pass a constant numeric value to the inline function unless you
5365 specify the @option{-O} option.
5367 You may also use @code{__builtin_constant_p} in initializers for static
5368 data. For instance, you can write
5371 static const int table[] = @{
5372 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5378 This is an acceptable initializer even if @var{EXPRESSION} is not a
5379 constant expression. GCC must be more conservative about evaluating the
5380 built-in in this case, because it has no opportunity to perform
5383 Previous versions of GCC did not accept this built-in in data
5384 initializers. The earliest version where it is completely safe is
5388 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5389 @opindex fprofile-arcs
5390 You may use @code{__builtin_expect} to provide the compiler with
5391 branch prediction information. In general, you should prefer to
5392 use actual profile feedback for this (@option{-fprofile-arcs}), as
5393 programmers are notoriously bad at predicting how their programs
5394 actually perform. However, there are applications in which this
5395 data is hard to collect.
5397 The return value is the value of @var{exp}, which should be an
5398 integral expression. The value of @var{c} must be a compile-time
5399 constant. The semantics of the built-in are that it is expected
5400 that @var{exp} == @var{c}. For example:
5403 if (__builtin_expect (x, 0))
5408 would indicate that we do not expect to call @code{foo}, since
5409 we expect @code{x} to be zero. Since you are limited to integral
5410 expressions for @var{exp}, you should use constructions such as
5413 if (__builtin_expect (ptr != NULL, 1))
5418 when testing pointer or floating-point values.
5421 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5422 This function is used to minimize cache-miss latency by moving data into
5423 a cache before it is accessed.
5424 You can insert calls to @code{__builtin_prefetch} into code for which
5425 you know addresses of data in memory that is likely to be accessed soon.
5426 If the target supports them, data prefetch instructions will be generated.
5427 If the prefetch is done early enough before the access then the data will
5428 be in the cache by the time it is accessed.
5430 The value of @var{addr} is the address of the memory to prefetch.
5431 There are two optional arguments, @var{rw} and @var{locality}.
5432 The value of @var{rw} is a compile-time constant one or zero; one
5433 means that the prefetch is preparing for a write to the memory address
5434 and zero, the default, means that the prefetch is preparing for a read.
5435 The value @var{locality} must be a compile-time constant integer between
5436 zero and three. A value of zero means that the data has no temporal
5437 locality, so it need not be left in the cache after the access. A value
5438 of three means that the data has a high degree of temporal locality and
5439 should be left in all levels of cache possible. Values of one and two
5440 mean, respectively, a low or moderate degree of temporal locality. The
5444 for (i = 0; i < n; i++)
5447 __builtin_prefetch (&a[i+j], 1, 1);
5448 __builtin_prefetch (&b[i+j], 0, 1);
5453 Data prefetch does not generate faults if @var{addr} is invalid, but
5454 the address expression itself must be valid. For example, a prefetch
5455 of @code{p->next} will not fault if @code{p->next} is not a valid
5456 address, but evaluation will fault if @code{p} is not a valid address.
5458 If the target does not support data prefetch, the address expression
5459 is evaluated if it includes side effects but no other code is generated
5460 and GCC does not issue a warning.
5463 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5464 Returns a positive infinity, if supported by the floating-point format,
5465 else @code{DBL_MAX}. This function is suitable for implementing the
5466 ISO C macro @code{HUGE_VAL}.
5469 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5470 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5473 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5474 Similar to @code{__builtin_huge_val}, except the return
5475 type is @code{long double}.
5478 @deftypefn {Built-in Function} double __builtin_inf (void)
5479 Similar to @code{__builtin_huge_val}, except a warning is generated
5480 if the target floating-point format does not support infinities.
5481 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5484 @deftypefn {Built-in Function} float __builtin_inff (void)
5485 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5488 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5489 Similar to @code{__builtin_inf}, except the return
5490 type is @code{long double}.
5493 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5494 This is an implementation of the ISO C99 function @code{nan}.
5496 Since ISO C99 defines this function in terms of @code{strtod}, which we
5497 do not implement, a description of the parsing is in order. The string
5498 is parsed as by @code{strtol}; that is, the base is recognized by
5499 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5500 in the significand such that the least significant bit of the number
5501 is at the least significant bit of the significand. The number is
5502 truncated to fit the significand field provided. The significand is
5503 forced to be a quiet NaN.
5505 This function, if given a string literal, is evaluated early enough
5506 that it is considered a compile-time constant.
5509 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5510 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5513 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5514 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5517 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5518 Similar to @code{__builtin_nan}, except the significand is forced
5519 to be a signaling NaN. The @code{nans} function is proposed by
5520 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
5523 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5524 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5527 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5528 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5531 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5532 Returns one plus the index of the least significant 1-bit of @var{x}, or
5533 if @var{x} is zero, returns zero.
5536 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5537 Returns the number of leading 0-bits in @var{x}, starting at the most
5538 significant bit position. If @var{x} is 0, the result is undefined.
5541 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5542 Returns the number of trailing 0-bits in @var{x}, starting at the least
5543 significant bit position. If @var{x} is 0, the result is undefined.
5546 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5547 Returns the number of 1-bits in @var{x}.
5550 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5551 Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5555 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5556 Similar to @code{__builtin_ffs}, except the argument type is
5557 @code{unsigned long}.
5560 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5561 Similar to @code{__builtin_clz}, except the argument type is
5562 @code{unsigned long}.
5565 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5566 Similar to @code{__builtin_ctz}, except the argument type is
5567 @code{unsigned long}.
5570 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5571 Similar to @code{__builtin_popcount}, except the argument type is
5572 @code{unsigned long}.
5575 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5576 Similar to @code{__builtin_parity}, except the argument type is
5577 @code{unsigned long}.
5580 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5581 Similar to @code{__builtin_ffs}, except the argument type is
5582 @code{unsigned long long}.
5585 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5586 Similar to @code{__builtin_clz}, except the argument type is
5587 @code{unsigned long long}.
5590 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5591 Similar to @code{__builtin_ctz}, except the argument type is
5592 @code{unsigned long long}.
5595 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5596 Similar to @code{__builtin_popcount}, except the argument type is
5597 @code{unsigned long long}.
5600 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5601 Similar to @code{__builtin_parity}, except the argument type is
5602 @code{unsigned long long}.
5606 @node Target Builtins
5607 @section Built-in Functions Specific to Particular Target Machines
5609 On some target machines, GCC supports many built-in functions specific
5610 to those machines. Generally these generate calls to specific machine
5611 instructions, but allow the compiler to schedule those calls.
5614 * Alpha Built-in Functions::
5615 * ARM Built-in Functions::
5616 * X86 Built-in Functions::
5617 * PowerPC AltiVec Built-in Functions::
5620 @node Alpha Built-in Functions
5621 @subsection Alpha Built-in Functions
5623 These built-in functions are available for the Alpha family of
5624 processors, depending on the command-line switches used.
5626 The following built-in functions are always available. They
5627 all generate the machine instruction that is part of the name.
5630 long __builtin_alpha_implver (void)
5631 long __builtin_alpha_rpcc (void)
5632 long __builtin_alpha_amask (long)
5633 long __builtin_alpha_cmpbge (long, long)
5634 long __builtin_alpha_extbl (long, long)
5635 long __builtin_alpha_extwl (long, long)
5636 long __builtin_alpha_extll (long, long)
5637 long __builtin_alpha_extql (long, long)
5638 long __builtin_alpha_extwh (long, long)
5639 long __builtin_alpha_extlh (long, long)
5640 long __builtin_alpha_extqh (long, long)
5641 long __builtin_alpha_insbl (long, long)
5642 long __builtin_alpha_inswl (long, long)
5643 long __builtin_alpha_insll (long, long)
5644 long __builtin_alpha_insql (long, long)
5645 long __builtin_alpha_inswh (long, long)
5646 long __builtin_alpha_inslh (long, long)
5647 long __builtin_alpha_insqh (long, long)
5648 long __builtin_alpha_mskbl (long, long)
5649 long __builtin_alpha_mskwl (long, long)
5650 long __builtin_alpha_mskll (long, long)
5651 long __builtin_alpha_mskql (long, long)
5652 long __builtin_alpha_mskwh (long, long)
5653 long __builtin_alpha_msklh (long, long)
5654 long __builtin_alpha_mskqh (long, long)
5655 long __builtin_alpha_umulh (long, long)
5656 long __builtin_alpha_zap (long, long)
5657 long __builtin_alpha_zapnot (long, long)
5660 The following built-in functions are always with @option{-mmax}
5661 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5662 later. They all generate the machine instruction that is part
5666 long __builtin_alpha_pklb (long)
5667 long __builtin_alpha_pkwb (long)
5668 long __builtin_alpha_unpkbl (long)
5669 long __builtin_alpha_unpkbw (long)
5670 long __builtin_alpha_minub8 (long, long)
5671 long __builtin_alpha_minsb8 (long, long)
5672 long __builtin_alpha_minuw4 (long, long)
5673 long __builtin_alpha_minsw4 (long, long)
5674 long __builtin_alpha_maxub8 (long, long)
5675 long __builtin_alpha_maxsb8 (long, long)
5676 long __builtin_alpha_maxuw4 (long, long)
5677 long __builtin_alpha_maxsw4 (long, long)
5678 long __builtin_alpha_perr (long, long)
5681 The following built-in functions are always with @option{-mcix}
5682 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5683 later. They all generate the machine instruction that is part
5687 long __builtin_alpha_cttz (long)
5688 long __builtin_alpha_ctlz (long)
5689 long __builtin_alpha_ctpop (long)
5692 The following builtins are available on systems that use the OSF/1
5693 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5694 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5695 @code{rdval} and @code{wrval}.
5698 void *__builtin_thread_pointer (void)
5699 void __builtin_set_thread_pointer (void *)
5702 @node ARM Built-in Functions
5703 @subsection ARM Built-in Functions
5705 These built-in functions are available for the ARM family of
5706 processors, when the @option{-mcpu=iwmmxt} switch is used:
5709 typedef int v2si __attribute__ ((vector_size (8)));
5710 typedef short v4hi __attribute__ ((vector_size (8)));
5711 typedef char v8qi __attribute__ ((vector_size (8)));
5713 int __builtin_arm_getwcx (int)
5714 void __builtin_arm_setwcx (int, int)
5715 int __builtin_arm_textrmsb (v8qi, int)
5716 int __builtin_arm_textrmsh (v4hi, int)
5717 int __builtin_arm_textrmsw (v2si, int)
5718 int __builtin_arm_textrmub (v8qi, int)
5719 int __builtin_arm_textrmuh (v4hi, int)
5720 int __builtin_arm_textrmuw (v2si, int)
5721 v8qi __builtin_arm_tinsrb (v8qi, int)
5722 v4hi __builtin_arm_tinsrh (v4hi, int)
5723 v2si __builtin_arm_tinsrw (v2si, int)
5724 long long __builtin_arm_tmia (long long, int, int)
5725 long long __builtin_arm_tmiabb (long long, int, int)
5726 long long __builtin_arm_tmiabt (long long, int, int)
5727 long long __builtin_arm_tmiaph (long long, int, int)
5728 long long __builtin_arm_tmiatb (long long, int, int)
5729 long long __builtin_arm_tmiatt (long long, int, int)
5730 int __builtin_arm_tmovmskb (v8qi)
5731 int __builtin_arm_tmovmskh (v4hi)
5732 int __builtin_arm_tmovmskw (v2si)
5733 long long __builtin_arm_waccb (v8qi)
5734 long long __builtin_arm_wacch (v4hi)
5735 long long __builtin_arm_waccw (v2si)
5736 v8qi __builtin_arm_waddb (v8qi, v8qi)
5737 v8qi __builtin_arm_waddbss (v8qi, v8qi)
5738 v8qi __builtin_arm_waddbus (v8qi, v8qi)
5739 v4hi __builtin_arm_waddh (v4hi, v4hi)
5740 v4hi __builtin_arm_waddhss (v4hi, v4hi)
5741 v4hi __builtin_arm_waddhus (v4hi, v4hi)
5742 v2si __builtin_arm_waddw (v2si, v2si)
5743 v2si __builtin_arm_waddwss (v2si, v2si)
5744 v2si __builtin_arm_waddwus (v2si, v2si)
5745 v8qi __builtin_arm_walign (v8qi, v8qi, int)
5746 long long __builtin_arm_wand(long long, long long)
5747 long long __builtin_arm_wandn (long long, long long)
5748 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5749 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5750 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5751 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5752 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5753 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5754 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5755 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5756 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5757 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5758 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5759 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5760 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5761 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5762 long long __builtin_arm_wmacsz (v4hi, v4hi)
5763 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5764 long long __builtin_arm_wmacuz (v4hi, v4hi)
5765 v4hi __builtin_arm_wmadds (v4hi, v4hi)
5766 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5767 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5768 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5769 v2si __builtin_arm_wmaxsw (v2si, v2si)
5770 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5771 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5772 v2si __builtin_arm_wmaxuw (v2si, v2si)
5773 v8qi __builtin_arm_wminsb (v8qi, v8qi)
5774 v4hi __builtin_arm_wminsh (v4hi, v4hi)
5775 v2si __builtin_arm_wminsw (v2si, v2si)
5776 v8qi __builtin_arm_wminub (v8qi, v8qi)
5777 v4hi __builtin_arm_wminuh (v4hi, v4hi)
5778 v2si __builtin_arm_wminuw (v2si, v2si)
5779 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5780 v4hi __builtin_arm_wmulul (v4hi, v4hi)
5781 v4hi __builtin_arm_wmulum (v4hi, v4hi)
5782 long long __builtin_arm_wor (long long, long long)
5783 v2si __builtin_arm_wpackdss (long long, long long)
5784 v2si __builtin_arm_wpackdus (long long, long long)
5785 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5786 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5787 v4hi __builtin_arm_wpackwss (v2si, v2si)
5788 v4hi __builtin_arm_wpackwus (v2si, v2si)
5789 long long __builtin_arm_wrord (long long, long long)
5790 long long __builtin_arm_wrordi (long long, int)
5791 v4hi __builtin_arm_wrorh (v4hi, long long)
5792 v4hi __builtin_arm_wrorhi (v4hi, int)
5793 v2si __builtin_arm_wrorw (v2si, long long)
5794 v2si __builtin_arm_wrorwi (v2si, int)
5795 v2si __builtin_arm_wsadb (v8qi, v8qi)
5796 v2si __builtin_arm_wsadbz (v8qi, v8qi)
5797 v2si __builtin_arm_wsadh (v4hi, v4hi)
5798 v2si __builtin_arm_wsadhz (v4hi, v4hi)
5799 v4hi __builtin_arm_wshufh (v4hi, int)
5800 long long __builtin_arm_wslld (long long, long long)
5801 long long __builtin_arm_wslldi (long long, int)
5802 v4hi __builtin_arm_wsllh (v4hi, long long)
5803 v4hi __builtin_arm_wsllhi (v4hi, int)
5804 v2si __builtin_arm_wsllw (v2si, long long)
5805 v2si __builtin_arm_wsllwi (v2si, int)
5806 long long __builtin_arm_wsrad (long long, long long)
5807 long long __builtin_arm_wsradi (long long, int)
5808 v4hi __builtin_arm_wsrah (v4hi, long long)
5809 v4hi __builtin_arm_wsrahi (v4hi, int)
5810 v2si __builtin_arm_wsraw (v2si, long long)
5811 v2si __builtin_arm_wsrawi (v2si, int)
5812 long long __builtin_arm_wsrld (long long, long long)
5813 long long __builtin_arm_wsrldi (long long, int)
5814 v4hi __builtin_arm_wsrlh (v4hi, long long)
5815 v4hi __builtin_arm_wsrlhi (v4hi, int)
5816 v2si __builtin_arm_wsrlw (v2si, long long)
5817 v2si __builtin_arm_wsrlwi (v2si, int)
5818 v8qi __builtin_arm_wsubb (v8qi, v8qi)
5819 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5820 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5821 v4hi __builtin_arm_wsubh (v4hi, v4hi)
5822 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5823 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5824 v2si __builtin_arm_wsubw (v2si, v2si)
5825 v2si __builtin_arm_wsubwss (v2si, v2si)
5826 v2si __builtin_arm_wsubwus (v2si, v2si)
5827 v4hi __builtin_arm_wunpckehsb (v8qi)
5828 v2si __builtin_arm_wunpckehsh (v4hi)
5829 long long __builtin_arm_wunpckehsw (v2si)
5830 v4hi __builtin_arm_wunpckehub (v8qi)
5831 v2si __builtin_arm_wunpckehuh (v4hi)
5832 long long __builtin_arm_wunpckehuw (v2si)
5833 v4hi __builtin_arm_wunpckelsb (v8qi)
5834 v2si __builtin_arm_wunpckelsh (v4hi)
5835 long long __builtin_arm_wunpckelsw (v2si)
5836 v4hi __builtin_arm_wunpckelub (v8qi)
5837 v2si __builtin_arm_wunpckeluh (v4hi)
5838 long long __builtin_arm_wunpckeluw (v2si)
5839 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5840 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5841 v2si __builtin_arm_wunpckihw (v2si, v2si)
5842 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5843 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5844 v2si __builtin_arm_wunpckilw (v2si, v2si)
5845 long long __builtin_arm_wxor (long long, long long)
5846 long long __builtin_arm_wzero ()
5849 @node X86 Built-in Functions
5850 @subsection X86 Built-in Functions
5852 These built-in functions are available for the i386 and x86-64 family
5853 of computers, depending on the command-line switches used.
5855 The following machine modes are available for use with MMX built-in functions
5856 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5857 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5858 vector of eight 8-bit integers. Some of the built-in functions operate on
5859 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5861 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5862 of two 32-bit floating point values.
5864 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5865 floating point values. Some instructions use a vector of four 32-bit
5866 integers, these use @code{V4SI}. Finally, some instructions operate on an
5867 entire vector register, interpreting it as a 128-bit integer, these use mode
5870 The following built-in functions are made available by @option{-mmmx}.
5871 All of them generate the machine instruction that is part of the name.
5874 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5875 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5876 v2si __builtin_ia32_paddd (v2si, v2si)
5877 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5878 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5879 v2si __builtin_ia32_psubd (v2si, v2si)
5880 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5881 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5882 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5883 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5884 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5885 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5886 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5887 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5888 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5889 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5890 di __builtin_ia32_pand (di, di)
5891 di __builtin_ia32_pandn (di,di)
5892 di __builtin_ia32_por (di, di)
5893 di __builtin_ia32_pxor (di, di)
5894 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5895 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5896 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5897 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5898 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5899 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5900 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5901 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5902 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5903 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5904 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5905 v2si __builtin_ia32_punpckldq (v2si, v2si)
5906 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5907 v4hi __builtin_ia32_packssdw (v2si, v2si)
5908 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5911 The following built-in functions are made available either with
5912 @option{-msse}, or with a combination of @option{-m3dnow} and
5913 @option{-march=athlon}. All of them generate the machine
5914 instruction that is part of the name.
5917 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5918 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5919 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5920 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5921 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5922 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5923 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5924 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5925 int __builtin_ia32_pextrw (v4hi, int)
5926 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5927 int __builtin_ia32_pmovmskb (v8qi)
5928 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5929 void __builtin_ia32_movntq (di *, di)
5930 void __builtin_ia32_sfence (void)
5933 The following built-in functions are available when @option{-msse} is used.
5934 All of them generate the machine instruction that is part of the name.
5937 int __builtin_ia32_comieq (v4sf, v4sf)
5938 int __builtin_ia32_comineq (v4sf, v4sf)
5939 int __builtin_ia32_comilt (v4sf, v4sf)
5940 int __builtin_ia32_comile (v4sf, v4sf)
5941 int __builtin_ia32_comigt (v4sf, v4sf)
5942 int __builtin_ia32_comige (v4sf, v4sf)
5943 int __builtin_ia32_ucomieq (v4sf, v4sf)
5944 int __builtin_ia32_ucomineq (v4sf, v4sf)
5945 int __builtin_ia32_ucomilt (v4sf, v4sf)
5946 int __builtin_ia32_ucomile (v4sf, v4sf)
5947 int __builtin_ia32_ucomigt (v4sf, v4sf)
5948 int __builtin_ia32_ucomige (v4sf, v4sf)
5949 v4sf __builtin_ia32_addps (v4sf, v4sf)
5950 v4sf __builtin_ia32_subps (v4sf, v4sf)
5951 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5952 v4sf __builtin_ia32_divps (v4sf, v4sf)
5953 v4sf __builtin_ia32_addss (v4sf, v4sf)
5954 v4sf __builtin_ia32_subss (v4sf, v4sf)
5955 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5956 v4sf __builtin_ia32_divss (v4sf, v4sf)
5957 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5958 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5959 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5960 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5961 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5962 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5963 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5964 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5965 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5966 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5967 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5968 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5969 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5970 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5971 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5972 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5973 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5974 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
5975 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
5976 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
5977 v4sf __builtin_ia32_maxps (v4sf, v4sf)
5978 v4sf __builtin_ia32_maxss (v4sf, v4sf)
5979 v4sf __builtin_ia32_minps (v4sf, v4sf)
5980 v4sf __builtin_ia32_minss (v4sf, v4sf)
5981 v4sf __builtin_ia32_andps (v4sf, v4sf)
5982 v4sf __builtin_ia32_andnps (v4sf, v4sf)
5983 v4sf __builtin_ia32_orps (v4sf, v4sf)
5984 v4sf __builtin_ia32_xorps (v4sf, v4sf)
5985 v4sf __builtin_ia32_movss (v4sf, v4sf)
5986 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
5987 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
5988 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
5989 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
5990 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
5991 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
5992 v2si __builtin_ia32_cvtps2pi (v4sf)
5993 int __builtin_ia32_cvtss2si (v4sf)
5994 v2si __builtin_ia32_cvttps2pi (v4sf)
5995 int __builtin_ia32_cvttss2si (v4sf)
5996 v4sf __builtin_ia32_rcpps (v4sf)
5997 v4sf __builtin_ia32_rsqrtps (v4sf)
5998 v4sf __builtin_ia32_sqrtps (v4sf)
5999 v4sf __builtin_ia32_rcpss (v4sf)
6000 v4sf __builtin_ia32_rsqrtss (v4sf)
6001 v4sf __builtin_ia32_sqrtss (v4sf)
6002 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6003 void __builtin_ia32_movntps (float *, v4sf)
6004 int __builtin_ia32_movmskps (v4sf)
6007 The following built-in functions are available when @option{-msse} is used.
6010 @item v4sf __builtin_ia32_loadaps (float *)
6011 Generates the @code{movaps} machine instruction as a load from memory.
6012 @item void __builtin_ia32_storeaps (float *, v4sf)
6013 Generates the @code{movaps} machine instruction as a store to memory.
6014 @item v4sf __builtin_ia32_loadups (float *)
6015 Generates the @code{movups} machine instruction as a load from memory.
6016 @item void __builtin_ia32_storeups (float *, v4sf)
6017 Generates the @code{movups} machine instruction as a store to memory.
6018 @item v4sf __builtin_ia32_loadsss (float *)
6019 Generates the @code{movss} machine instruction as a load from memory.
6020 @item void __builtin_ia32_storess (float *, v4sf)
6021 Generates the @code{movss} machine instruction as a store to memory.
6022 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6023 Generates the @code{movhps} machine instruction as a load from memory.
6024 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6025 Generates the @code{movlps} machine instruction as a load from memory
6026 @item void __builtin_ia32_storehps (v4sf, v2si *)
6027 Generates the @code{movhps} machine instruction as a store to memory.
6028 @item void __builtin_ia32_storelps (v4sf, v2si *)
6029 Generates the @code{movlps} machine instruction as a store to memory.
6032 The following built-in functions are available when @option{-msse3} is used.
6033 All of them generate the machine instruction that is part of the name.
6036 v2df __builtin_ia32_addsubpd (v2df, v2df)
6037 v2df __builtin_ia32_addsubps (v2df, v2df)
6038 v2df __builtin_ia32_haddpd (v2df, v2df)
6039 v2df __builtin_ia32_haddps (v2df, v2df)
6040 v2df __builtin_ia32_hsubpd (v2df, v2df)
6041 v2df __builtin_ia32_hsubps (v2df, v2df)
6042 v16qi __builtin_ia32_lddqu (char const *)
6043 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6044 v2df __builtin_ia32_movddup (v2df)
6045 v4sf __builtin_ia32_movshdup (v4sf)
6046 v4sf __builtin_ia32_movsldup (v4sf)
6047 void __builtin_ia32_mwait (unsigned int, unsigned int)
6050 The following built-in functions are available when @option{-msse3} is used.
6053 @item v2df __builtin_ia32_loadddup (double const *)
6054 Generates the @code{movddup} machine instruction as a load from memory.
6057 The following built-in functions are available when @option{-m3dnow} is used.
6058 All of them generate the machine instruction that is part of the name.
6061 void __builtin_ia32_femms (void)
6062 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6063 v2si __builtin_ia32_pf2id (v2sf)
6064 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6065 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6066 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6067 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6068 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6069 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6070 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6071 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6072 v2sf __builtin_ia32_pfrcp (v2sf)
6073 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6074 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6075 v2sf __builtin_ia32_pfrsqrt (v2sf)
6076 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6077 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6078 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6079 v2sf __builtin_ia32_pi2fd (v2si)
6080 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6083 The following built-in functions are available when both @option{-m3dnow}
6084 and @option{-march=athlon} are used. All of them generate the machine
6085 instruction that is part of the name.
6088 v2si __builtin_ia32_pf2iw (v2sf)
6089 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6090 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6091 v2sf __builtin_ia32_pi2fw (v2si)
6092 v2sf __builtin_ia32_pswapdsf (v2sf)
6093 v2si __builtin_ia32_pswapdsi (v2si)
6096 @node PowerPC AltiVec Built-in Functions
6097 @subsection PowerPC AltiVec Built-in Functions
6099 These built-in functions are available for the PowerPC family
6100 of computers, depending on the command-line switches used.
6102 The following machine modes are available for use with AltiVec built-in
6103 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
6104 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
6105 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
6106 @code{V16QI} for a vector of sixteen 8-bit integers.
6108 The following functions are made available by including
6109 @code{<altivec.h>} and using @option{-maltivec} and
6110 @option{-mabi=altivec}. The functions implement the functionality
6111 described in Motorola's AltiVec Programming Interface Manual.
6113 There are a few differences from Motorola's documentation and GCC's
6114 implementation. Vector constants are done with curly braces (not
6115 parentheses). Vector initializers require no casts if the vector
6116 constant is of the same type as the variable it is initializing. The
6117 @code{vector bool} type is deprecated and will be discontinued in
6118 further revisions. Use @code{vector signed} instead. If @code{signed}
6119 or @code{unsigned} is omitted, the vector type will default to
6120 @code{signed}. Lastly, all overloaded functions are implemented with macros
6121 for the C implementation. So code the following example will not work:
6124 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6127 Since vec_add is a macro, the vector constant in the above example will
6128 be treated as four different arguments. Wrap the entire argument in
6129 parentheses for this to work. The C++ implementation does not use
6132 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
6133 Internally, GCC uses built-in functions to achieve the functionality in
6134 the aforementioned header file, but they are not supported and are
6135 subject to change without notice.
6138 vector signed char vec_abs (vector signed char, vector signed char);
6139 vector signed short vec_abs (vector signed short, vector signed short);
6140 vector signed int vec_abs (vector signed int, vector signed int);
6141 vector signed float vec_abs (vector signed float, vector signed float);
6143 vector signed char vec_abss (vector signed char, vector signed char);
6144 vector signed short vec_abss (vector signed short, vector signed short);
6146 vector signed char vec_add (vector signed char, vector signed char);
6147 vector unsigned char vec_add (vector signed char, vector unsigned char);
6149 vector unsigned char vec_add (vector unsigned char, vector signed char);
6151 vector unsigned char vec_add (vector unsigned char,
6152 vector unsigned char);
6153 vector signed short vec_add (vector signed short, vector signed short);
6154 vector unsigned short vec_add (vector signed short,
6155 vector unsigned short);
6156 vector unsigned short vec_add (vector unsigned short,
6157 vector signed short);
6158 vector unsigned short vec_add (vector unsigned short,
6159 vector unsigned short);
6160 vector signed int vec_add (vector signed int, vector signed int);
6161 vector unsigned int vec_add (vector signed int, vector unsigned int);
6162 vector unsigned int vec_add (vector unsigned int, vector signed int);
6163 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6164 vector float vec_add (vector float, vector float);
6166 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6168 vector unsigned char vec_adds (vector signed char,
6169 vector unsigned char);
6170 vector unsigned char vec_adds (vector unsigned char,
6171 vector signed char);
6172 vector unsigned char vec_adds (vector unsigned char,
6173 vector unsigned char);
6174 vector signed char vec_adds (vector signed char, vector signed char);
6175 vector unsigned short vec_adds (vector signed short,
6176 vector unsigned short);
6177 vector unsigned short vec_adds (vector unsigned short,
6178 vector signed short);
6179 vector unsigned short vec_adds (vector unsigned short,
6180 vector unsigned short);
6181 vector signed short vec_adds (vector signed short, vector signed short);
6183 vector unsigned int vec_adds (vector signed int, vector unsigned int);
6184 vector unsigned int vec_adds (vector unsigned int, vector signed int);
6185 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6187 vector signed int vec_adds (vector signed int, vector signed int);
6189 vector float vec_and (vector float, vector float);
6190 vector float vec_and (vector float, vector signed int);
6191 vector float vec_and (vector signed int, vector float);
6192 vector signed int vec_and (vector signed int, vector signed int);
6193 vector unsigned int vec_and (vector signed int, vector unsigned int);
6194 vector unsigned int vec_and (vector unsigned int, vector signed int);
6195 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
6196 vector signed short vec_and (vector signed short, vector signed short);
6197 vector unsigned short vec_and (vector signed short,
6198 vector unsigned short);
6199 vector unsigned short vec_and (vector unsigned short,
6200 vector signed short);
6201 vector unsigned short vec_and (vector unsigned short,
6202 vector unsigned short);
6203 vector signed char vec_and (vector signed char, vector signed char);
6204 vector unsigned char vec_and (vector signed char, vector unsigned char);
6206 vector unsigned char vec_and (vector unsigned char, vector signed char);
6208 vector unsigned char vec_and (vector unsigned char,
6209 vector unsigned char);
6211 vector float vec_andc (vector float, vector float);
6212 vector float vec_andc (vector float, vector signed int);
6213 vector float vec_andc (vector signed int, vector float);
6214 vector signed int vec_andc (vector signed int, vector signed int);
6215 vector unsigned int vec_andc (vector signed int, vector unsigned int);
6216 vector unsigned int vec_andc (vector unsigned int, vector signed int);
6217 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
6219 vector signed short vec_andc (vector signed short, vector signed short);
6221 vector unsigned short vec_andc (vector signed short,
6222 vector unsigned short);
6223 vector unsigned short vec_andc (vector unsigned short,
6224 vector signed short);
6225 vector unsigned short vec_andc (vector unsigned short,
6226 vector unsigned short);
6227 vector signed char vec_andc (vector signed char, vector signed char);
6228 vector unsigned char vec_andc (vector signed char,
6229 vector unsigned char);
6230 vector unsigned char vec_andc (vector unsigned char,
6231 vector signed char);
6232 vector unsigned char vec_andc (vector unsigned char,
6233 vector unsigned char);
6235 vector unsigned char vec_avg (vector unsigned char,
6236 vector unsigned char);
6237 vector signed char vec_avg (vector signed char, vector signed char);
6238 vector unsigned short vec_avg (vector unsigned short,
6239 vector unsigned short);
6240 vector signed short vec_avg (vector signed short, vector signed short);
6241 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
6242 vector signed int vec_avg (vector signed int, vector signed int);
6244 vector float vec_ceil (vector float);
6246 vector signed int vec_cmpb (vector float, vector float);
6248 vector signed char vec_cmpeq (vector signed char, vector signed char);
6249 vector signed char vec_cmpeq (vector unsigned char,
6250 vector unsigned char);
6251 vector signed short vec_cmpeq (vector signed short,
6252 vector signed short);
6253 vector signed short vec_cmpeq (vector unsigned short,
6254 vector unsigned short);
6255 vector signed int vec_cmpeq (vector signed int, vector signed int);
6256 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
6257 vector signed int vec_cmpeq (vector float, vector float);
6259 vector signed int vec_cmpge (vector float, vector float);
6261 vector signed char vec_cmpgt (vector unsigned char,
6262 vector unsigned char);
6263 vector signed char vec_cmpgt (vector signed char, vector signed char);
6264 vector signed short vec_cmpgt (vector unsigned short,
6265 vector unsigned short);
6266 vector signed short vec_cmpgt (vector signed short,
6267 vector signed short);
6268 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
6269 vector signed int vec_cmpgt (vector signed int, vector signed int);
6270 vector signed int vec_cmpgt (vector float, vector float);
6272 vector signed int vec_cmple (vector float, vector float);
6274 vector signed char vec_cmplt (vector unsigned char,
6275 vector unsigned char);
6276 vector signed char vec_cmplt (vector signed char, vector signed char);
6277 vector signed short vec_cmplt (vector unsigned short,
6278 vector unsigned short);
6279 vector signed short vec_cmplt (vector signed short,
6280 vector signed short);
6281 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
6282 vector signed int vec_cmplt (vector signed int, vector signed int);
6283 vector signed int vec_cmplt (vector float, vector float);
6285 vector float vec_ctf (vector unsigned int, const char);
6286 vector float vec_ctf (vector signed int, const char);
6288 vector signed int vec_cts (vector float, const char);
6290 vector unsigned int vec_ctu (vector float, const char);
6292 void vec_dss (const char);
6294 void vec_dssall (void);
6296 void vec_dst (void *, int, const char);
6298 void vec_dstst (void *, int, const char);
6300 void vec_dststt (void *, int, const char);
6302 void vec_dstt (void *, int, const char);
6304 vector float vec_expte (vector float, vector float);
6306 vector float vec_floor (vector float, vector float);
6308 vector float vec_ld (int, vector float *);
6309 vector float vec_ld (int, float *):
6310 vector signed int vec_ld (int, int *);
6311 vector signed int vec_ld (int, vector signed int *);
6312 vector unsigned int vec_ld (int, vector unsigned int *);
6313 vector unsigned int vec_ld (int, unsigned int *);
6314 vector signed short vec_ld (int, short *, vector signed short *);
6315 vector unsigned short vec_ld (int, unsigned short *,
6316 vector unsigned short *);
6317 vector signed char vec_ld (int, signed char *);
6318 vector signed char vec_ld (int, vector signed char *);
6319 vector unsigned char vec_ld (int, unsigned char *);
6320 vector unsigned char vec_ld (int, vector unsigned char *);
6322 vector signed char vec_lde (int, signed char *);
6323 vector unsigned char vec_lde (int, unsigned char *);
6324 vector signed short vec_lde (int, short *);
6325 vector unsigned short vec_lde (int, unsigned short *);
6326 vector float vec_lde (int, float *);
6327 vector signed int vec_lde (int, int *);
6328 vector unsigned int vec_lde (int, unsigned int *);
6330 void float vec_ldl (int, float *);
6331 void float vec_ldl (int, vector float *);
6332 void signed int vec_ldl (int, vector signed int *);
6333 void signed int vec_ldl (int, int *);
6334 void unsigned int vec_ldl (int, unsigned int *);
6335 void unsigned int vec_ldl (int, vector unsigned int *);
6336 void signed short vec_ldl (int, vector signed short *);
6337 void signed short vec_ldl (int, short *);
6338 void unsigned short vec_ldl (int, vector unsigned short *);
6339 void unsigned short vec_ldl (int, unsigned short *);
6340 void signed char vec_ldl (int, vector signed char *);
6341 void signed char vec_ldl (int, signed char *);
6342 void unsigned char vec_ldl (int, vector unsigned char *);
6343 void unsigned char vec_ldl (int, unsigned char *);
6345 vector float vec_loge (vector float);
6347 vector unsigned char vec_lvsl (int, void *, int *);
6349 vector unsigned char vec_lvsr (int, void *, int *);
6351 vector float vec_madd (vector float, vector float, vector float);
6353 vector signed short vec_madds (vector signed short, vector signed short,
6354 vector signed short);
6356 vector unsigned char vec_max (vector signed char, vector unsigned char);
6358 vector unsigned char vec_max (vector unsigned char, vector signed char);
6360 vector unsigned char vec_max (vector unsigned char,
6361 vector unsigned char);
6362 vector signed char vec_max (vector signed char, vector signed char);
6363 vector unsigned short vec_max (vector signed short,
6364 vector unsigned short);
6365 vector unsigned short vec_max (vector unsigned short,
6366 vector signed short);
6367 vector unsigned short vec_max (vector unsigned short,
6368 vector unsigned short);
6369 vector signed short vec_max (vector signed short, vector signed short);
6370 vector unsigned int vec_max (vector signed int, vector unsigned int);
6371 vector unsigned int vec_max (vector unsigned int, vector signed int);
6372 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
6373 vector signed int vec_max (vector signed int, vector signed int);
6374 vector float vec_max (vector float, vector float);
6376 vector signed char vec_mergeh (vector signed char, vector signed char);
6377 vector unsigned char vec_mergeh (vector unsigned char,
6378 vector unsigned char);
6379 vector signed short vec_mergeh (vector signed short,
6380 vector signed short);
6381 vector unsigned short vec_mergeh (vector unsigned short,
6382 vector unsigned short);
6383 vector float vec_mergeh (vector float, vector float);
6384 vector signed int vec_mergeh (vector signed int, vector signed int);
6385 vector unsigned int vec_mergeh (vector unsigned int,
6386 vector unsigned int);
6388 vector signed char vec_mergel (vector signed char, vector signed char);
6389 vector unsigned char vec_mergel (vector unsigned char,
6390 vector unsigned char);
6391 vector signed short vec_mergel (vector signed short,
6392 vector signed short);
6393 vector unsigned short vec_mergel (vector unsigned short,
6394 vector unsigned short);
6395 vector float vec_mergel (vector float, vector float);
6396 vector signed int vec_mergel (vector signed int, vector signed int);
6397 vector unsigned int vec_mergel (vector unsigned int,
6398 vector unsigned int);
6400 vector unsigned short vec_mfvscr (void);
6402 vector unsigned char vec_min (vector signed char, vector unsigned char);
6404 vector unsigned char vec_min (vector unsigned char, vector signed char);
6406 vector unsigned char vec_min (vector unsigned char,
6407 vector unsigned char);
6408 vector signed char vec_min (vector signed char, vector signed char);
6409 vector unsigned short vec_min (vector signed short,
6410 vector unsigned short);
6411 vector unsigned short vec_min (vector unsigned short,
6412 vector signed short);
6413 vector unsigned short vec_min (vector unsigned short,
6414 vector unsigned short);
6415 vector signed short vec_min (vector signed short, vector signed short);
6416 vector unsigned int vec_min (vector signed int, vector unsigned int);
6417 vector unsigned int vec_min (vector unsigned int, vector signed int);
6418 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
6419 vector signed int vec_min (vector signed int, vector signed int);
6420 vector float vec_min (vector float, vector float);
6422 vector signed short vec_mladd (vector signed short, vector signed short,
6423 vector signed short);
6424 vector signed short vec_mladd (vector signed short,
6425 vector unsigned short,
6426 vector unsigned short);
6427 vector signed short vec_mladd (vector unsigned short,
6428 vector signed short,
6429 vector signed short);
6430 vector unsigned short vec_mladd (vector unsigned short,
6431 vector unsigned short,
6432 vector unsigned short);
6434 vector signed short vec_mradds (vector signed short,
6435 vector signed short,
6436 vector signed short);
6438 vector unsigned int vec_msum (vector unsigned char,
6439 vector unsigned char,
6440 vector unsigned int);
6441 vector signed int vec_msum (vector signed char, vector unsigned char,
6443 vector unsigned int vec_msum (vector unsigned short,
6444 vector unsigned short,
6445 vector unsigned int);
6446 vector signed int vec_msum (vector signed short, vector signed short,
6449 vector unsigned int vec_msums (vector unsigned short,
6450 vector unsigned short,
6451 vector unsigned int);
6452 vector signed int vec_msums (vector signed short, vector signed short,
6455 void vec_mtvscr (vector signed int);
6456 void vec_mtvscr (vector unsigned int);
6457 void vec_mtvscr (vector signed short);
6458 void vec_mtvscr (vector unsigned short);
6459 void vec_mtvscr (vector signed char);
6460 void vec_mtvscr (vector unsigned char);
6462 vector unsigned short vec_mule (vector unsigned char,
6463 vector unsigned char);
6464 vector signed short vec_mule (vector signed char, vector signed char);
6465 vector unsigned int vec_mule (vector unsigned short,
6466 vector unsigned short);
6467 vector signed int vec_mule (vector signed short, vector signed short);
6469 vector unsigned short vec_mulo (vector unsigned char,
6470 vector unsigned char);
6471 vector signed short vec_mulo (vector signed char, vector signed char);
6472 vector unsigned int vec_mulo (vector unsigned short,
6473 vector unsigned short);
6474 vector signed int vec_mulo (vector signed short, vector signed short);
6476 vector float vec_nmsub (vector float, vector float, vector float);
6478 vector float vec_nor (vector float, vector float);
6479 vector signed int vec_nor (vector signed int, vector signed int);
6480 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
6481 vector signed short vec_nor (vector signed short, vector signed short);
6482 vector unsigned short vec_nor (vector unsigned short,
6483 vector unsigned short);
6484 vector signed char vec_nor (vector signed char, vector signed char);
6485 vector unsigned char vec_nor (vector unsigned char,
6486 vector unsigned char);
6488 vector float vec_or (vector float, vector float);
6489 vector float vec_or (vector float, vector signed int);
6490 vector float vec_or (vector signed int, vector float);
6491 vector signed int vec_or (vector signed int, vector signed int);
6492 vector unsigned int vec_or (vector signed int, vector unsigned int);
6493 vector unsigned int vec_or (vector unsigned int, vector signed int);
6494 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
6495 vector signed short vec_or (vector signed short, vector signed short);
6496 vector unsigned short vec_or (vector signed short,
6497 vector unsigned short);
6498 vector unsigned short vec_or (vector unsigned short,
6499 vector signed short);
6500 vector unsigned short vec_or (vector unsigned short,
6501 vector unsigned short);
6502 vector signed char vec_or (vector signed char, vector signed char);
6503 vector unsigned char vec_or (vector signed char, vector unsigned char);
6504 vector unsigned char vec_or (vector unsigned char, vector signed char);
6505 vector unsigned char vec_or (vector unsigned char,
6506 vector unsigned char);
6508 vector signed char vec_pack (vector signed short, vector signed short);
6509 vector unsigned char vec_pack (vector unsigned short,
6510 vector unsigned short);
6511 vector signed short vec_pack (vector signed int, vector signed int);
6512 vector unsigned short vec_pack (vector unsigned int,
6513 vector unsigned int);
6515 vector signed short vec_packpx (vector unsigned int,
6516 vector unsigned int);
6518 vector unsigned char vec_packs (vector unsigned short,
6519 vector unsigned short);
6520 vector signed char vec_packs (vector signed short, vector signed short);
6522 vector unsigned short vec_packs (vector unsigned int,
6523 vector unsigned int);
6524 vector signed short vec_packs (vector signed int, vector signed int);
6526 vector unsigned char vec_packsu (vector unsigned short,
6527 vector unsigned short);
6528 vector unsigned char vec_packsu (vector signed short,
6529 vector signed short);
6530 vector unsigned short vec_packsu (vector unsigned int,
6531 vector unsigned int);
6532 vector unsigned short vec_packsu (vector signed int, vector signed int);
6534 vector float vec_perm (vector float, vector float,
6535 vector unsigned char);
6536 vector signed int vec_perm (vector signed int, vector signed int,
6537 vector unsigned char);
6538 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
6539 vector unsigned char);
6540 vector signed short vec_perm (vector signed short, vector signed short,
6541 vector unsigned char);
6542 vector unsigned short vec_perm (vector unsigned short,
6543 vector unsigned short,
6544 vector unsigned char);
6545 vector signed char vec_perm (vector signed char, vector signed char,
6546 vector unsigned char);
6547 vector unsigned char vec_perm (vector unsigned char,
6548 vector unsigned char,
6549 vector unsigned char);
6551 vector float vec_re (vector float);
6553 vector signed char vec_rl (vector signed char, vector unsigned char);
6554 vector unsigned char vec_rl (vector unsigned char,
6555 vector unsigned char);
6556 vector signed short vec_rl (vector signed short, vector unsigned short);
6558 vector unsigned short vec_rl (vector unsigned short,
6559 vector unsigned short);
6560 vector signed int vec_rl (vector signed int, vector unsigned int);
6561 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
6563 vector float vec_round (vector float);
6565 vector float vec_rsqrte (vector float);
6567 vector float vec_sel (vector float, vector float, vector signed int);
6568 vector float vec_sel (vector float, vector float, vector unsigned int);
6569 vector signed int vec_sel (vector signed int, vector signed int,
6571 vector signed int vec_sel (vector signed int, vector signed int,
6572 vector unsigned int);
6573 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6575 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6576 vector unsigned int);
6577 vector signed short vec_sel (vector signed short, vector signed short,
6578 vector signed short);
6579 vector signed short vec_sel (vector signed short, vector signed short,
6580 vector unsigned short);
6581 vector unsigned short vec_sel (vector unsigned short,
6582 vector unsigned short,
6583 vector signed short);
6584 vector unsigned short vec_sel (vector unsigned short,
6585 vector unsigned short,
6586 vector unsigned short);
6587 vector signed char vec_sel (vector signed char, vector signed char,
6588 vector signed char);
6589 vector signed char vec_sel (vector signed char, vector signed char,
6590 vector unsigned char);
6591 vector unsigned char vec_sel (vector unsigned char,
6592 vector unsigned char,
6593 vector signed char);
6594 vector unsigned char vec_sel (vector unsigned char,
6595 vector unsigned char,
6596 vector unsigned char);
6598 vector signed char vec_sl (vector signed char, vector unsigned char);
6599 vector unsigned char vec_sl (vector unsigned char,
6600 vector unsigned char);
6601 vector signed short vec_sl (vector signed short, vector unsigned short);
6603 vector unsigned short vec_sl (vector unsigned short,
6604 vector unsigned short);
6605 vector signed int vec_sl (vector signed int, vector unsigned int);
6606 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
6608 vector float vec_sld (vector float, vector float, const char);
6609 vector signed int vec_sld (vector signed int, vector signed int,
6611 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
6613 vector signed short vec_sld (vector signed short, vector signed short,
6615 vector unsigned short vec_sld (vector unsigned short,
6616 vector unsigned short, const char);
6617 vector signed char vec_sld (vector signed char, vector signed char,
6619 vector unsigned char vec_sld (vector unsigned char,
6620 vector unsigned char,
6623 vector signed int vec_sll (vector signed int, vector unsigned int);
6624 vector signed int vec_sll (vector signed int, vector unsigned short);
6625 vector signed int vec_sll (vector signed int, vector unsigned char);
6626 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
6627 vector unsigned int vec_sll (vector unsigned int,
6628 vector unsigned short);
6629 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
6631 vector signed short vec_sll (vector signed short, vector unsigned int);
6632 vector signed short vec_sll (vector signed short,
6633 vector unsigned short);
6634 vector signed short vec_sll (vector signed short, vector unsigned char);
6636 vector unsigned short vec_sll (vector unsigned short,
6637 vector unsigned int);
6638 vector unsigned short vec_sll (vector unsigned short,
6639 vector unsigned short);
6640 vector unsigned short vec_sll (vector unsigned short,
6641 vector unsigned char);
6642 vector signed char vec_sll (vector signed char, vector unsigned int);
6643 vector signed char vec_sll (vector signed char, vector unsigned short);
6644 vector signed char vec_sll (vector signed char, vector unsigned char);
6645 vector unsigned char vec_sll (vector unsigned char,
6646 vector unsigned int);
6647 vector unsigned char vec_sll (vector unsigned char,
6648 vector unsigned short);
6649 vector unsigned char vec_sll (vector unsigned char,
6650 vector unsigned char);
6652 vector float vec_slo (vector float, vector signed char);
6653 vector float vec_slo (vector float, vector unsigned char);
6654 vector signed int vec_slo (vector signed int, vector signed char);
6655 vector signed int vec_slo (vector signed int, vector unsigned char);
6656 vector unsigned int vec_slo (vector unsigned int, vector signed char);
6657 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
6659 vector signed short vec_slo (vector signed short, vector signed char);
6660 vector signed short vec_slo (vector signed short, vector unsigned char);
6662 vector unsigned short vec_slo (vector unsigned short,
6663 vector signed char);
6664 vector unsigned short vec_slo (vector unsigned short,
6665 vector unsigned char);
6666 vector signed char vec_slo (vector signed char, vector signed char);
6667 vector signed char vec_slo (vector signed char, vector unsigned char);
6668 vector unsigned char vec_slo (vector unsigned char, vector signed char);
6670 vector unsigned char vec_slo (vector unsigned char,
6671 vector unsigned char);
6673 vector signed char vec_splat (vector signed char, const char);
6674 vector unsigned char vec_splat (vector unsigned char, const char);
6675 vector signed short vec_splat (vector signed short, const char);
6676 vector unsigned short vec_splat (vector unsigned short, const char);
6677 vector float vec_splat (vector float, const char);
6678 vector signed int vec_splat (vector signed int, const char);
6679 vector unsigned int vec_splat (vector unsigned int, const char);
6681 vector signed char vec_splat_s8 (const char);
6683 vector signed short vec_splat_s16 (const char);
6685 vector signed int vec_splat_s32 (const char);
6687 vector unsigned char vec_splat_u8 (const char);
6689 vector unsigned short vec_splat_u16 (const char);
6691 vector unsigned int vec_splat_u32 (const char);
6693 vector signed char vec_sr (vector signed char, vector unsigned char);
6694 vector unsigned char vec_sr (vector unsigned char,
6695 vector unsigned char);
6696 vector signed short vec_sr (vector signed short, vector unsigned short);
6698 vector unsigned short vec_sr (vector unsigned short,
6699 vector unsigned short);
6700 vector signed int vec_sr (vector signed int, vector unsigned int);
6701 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
6703 vector signed char vec_sra (vector signed char, vector unsigned char);
6704 vector unsigned char vec_sra (vector unsigned char,
6705 vector unsigned char);
6706 vector signed short vec_sra (vector signed short,
6707 vector unsigned short);
6708 vector unsigned short vec_sra (vector unsigned short,
6709 vector unsigned short);
6710 vector signed int vec_sra (vector signed int, vector unsigned int);
6711 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
6713 vector signed int vec_srl (vector signed int, vector unsigned int);
6714 vector signed int vec_srl (vector signed int, vector unsigned short);
6715 vector signed int vec_srl (vector signed int, vector unsigned char);
6716 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
6717 vector unsigned int vec_srl (vector unsigned int,
6718 vector unsigned short);
6719 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
6721 vector signed short vec_srl (vector signed short, vector unsigned int);
6722 vector signed short vec_srl (vector signed short,
6723 vector unsigned short);
6724 vector signed short vec_srl (vector signed short, vector unsigned char);
6726 vector unsigned short vec_srl (vector unsigned short,
6727 vector unsigned int);
6728 vector unsigned short vec_srl (vector unsigned short,
6729 vector unsigned short);
6730 vector unsigned short vec_srl (vector unsigned short,
6731 vector unsigned char);
6732 vector signed char vec_srl (vector signed char, vector unsigned int);
6733 vector signed char vec_srl (vector signed char, vector unsigned short);
6734 vector signed char vec_srl (vector signed char, vector unsigned char);
6735 vector unsigned char vec_srl (vector unsigned char,
6736 vector unsigned int);
6737 vector unsigned char vec_srl (vector unsigned char,
6738 vector unsigned short);
6739 vector unsigned char vec_srl (vector unsigned char,
6740 vector unsigned char);
6742 vector float vec_sro (vector float, vector signed char);
6743 vector float vec_sro (vector float, vector unsigned char);
6744 vector signed int vec_sro (vector signed int, vector signed char);
6745 vector signed int vec_sro (vector signed int, vector unsigned char);
6746 vector unsigned int vec_sro (vector unsigned int, vector signed char);
6747 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
6749 vector signed short vec_sro (vector signed short, vector signed char);
6750 vector signed short vec_sro (vector signed short, vector unsigned char);
6752 vector unsigned short vec_sro (vector unsigned short,
6753 vector signed char);
6754 vector unsigned short vec_sro (vector unsigned short,
6755 vector unsigned char);
6756 vector signed char vec_sro (vector signed char, vector signed char);
6757 vector signed char vec_sro (vector signed char, vector unsigned char);
6758 vector unsigned char vec_sro (vector unsigned char, vector signed char);
6760 vector unsigned char vec_sro (vector unsigned char,
6761 vector unsigned char);
6763 void vec_st (vector float, int, float *);
6764 void vec_st (vector float, int, vector float *);
6765 void vec_st (vector signed int, int, int *);
6766 void vec_st (vector signed int, int, unsigned int *);
6767 void vec_st (vector unsigned int, int, unsigned int *);
6768 void vec_st (vector unsigned int, int, vector unsigned int *);
6769 void vec_st (vector signed short, int, short *);
6770 void vec_st (vector signed short, int, vector unsigned short *);
6771 void vec_st (vector signed short, int, vector signed short *);
6772 void vec_st (vector unsigned short, int, unsigned short *);
6773 void vec_st (vector unsigned short, int, vector unsigned short *);
6774 void vec_st (vector signed char, int, signed char *);
6775 void vec_st (vector signed char, int, unsigned char *);
6776 void vec_st (vector signed char, int, vector signed char *);
6777 void vec_st (vector unsigned char, int, unsigned char *);
6778 void vec_st (vector unsigned char, int, vector unsigned char *);
6780 void vec_ste (vector signed char, int, unsigned char *);
6781 void vec_ste (vector signed char, int, signed char *);
6782 void vec_ste (vector unsigned char, int, unsigned char *);
6783 void vec_ste (vector signed short, int, short *);
6784 void vec_ste (vector signed short, int, unsigned short *);
6785 void vec_ste (vector unsigned short, int, void *);
6786 void vec_ste (vector signed int, int, unsigned int *);
6787 void vec_ste (vector signed int, int, int *);
6788 void vec_ste (vector unsigned int, int, unsigned int *);
6789 void vec_ste (vector float, int, float *);
6791 void vec_stl (vector float, int, vector float *);
6792 void vec_stl (vector float, int, float *);
6793 void vec_stl (vector signed int, int, vector signed int *);
6794 void vec_stl (vector signed int, int, int *);
6795 void vec_stl (vector signed int, int, unsigned int *);
6796 void vec_stl (vector unsigned int, int, vector unsigned int *);
6797 void vec_stl (vector unsigned int, int, unsigned int *);
6798 void vec_stl (vector signed short, int, short *);
6799 void vec_stl (vector signed short, int, unsigned short *);
6800 void vec_stl (vector signed short, int, vector signed short *);
6801 void vec_stl (vector unsigned short, int, unsigned short *);
6802 void vec_stl (vector unsigned short, int, vector signed short *);
6803 void vec_stl (vector signed char, int, signed char *);
6804 void vec_stl (vector signed char, int, unsigned char *);
6805 void vec_stl (vector signed char, int, vector signed char *);
6806 void vec_stl (vector unsigned char, int, unsigned char *);
6807 void vec_stl (vector unsigned char, int, vector unsigned char *);
6809 vector signed char vec_sub (vector signed char, vector signed char);
6810 vector unsigned char vec_sub (vector signed char, vector unsigned char);
6812 vector unsigned char vec_sub (vector unsigned char, vector signed char);
6814 vector unsigned char vec_sub (vector unsigned char,
6815 vector unsigned char);
6816 vector signed short vec_sub (vector signed short, vector signed short);
6817 vector unsigned short vec_sub (vector signed short,
6818 vector unsigned short);
6819 vector unsigned short vec_sub (vector unsigned short,
6820 vector signed short);
6821 vector unsigned short vec_sub (vector unsigned short,
6822 vector unsigned short);
6823 vector signed int vec_sub (vector signed int, vector signed int);
6824 vector unsigned int vec_sub (vector signed int, vector unsigned int);
6825 vector unsigned int vec_sub (vector unsigned int, vector signed int);
6826 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
6827 vector float vec_sub (vector float, vector float);
6829 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
6831 vector unsigned char vec_subs (vector signed char,
6832 vector unsigned char);
6833 vector unsigned char vec_subs (vector unsigned char,
6834 vector signed char);
6835 vector unsigned char vec_subs (vector unsigned char,
6836 vector unsigned char);
6837 vector signed char vec_subs (vector signed char, vector signed char);
6838 vector unsigned short vec_subs (vector signed short,
6839 vector unsigned short);
6840 vector unsigned short vec_subs (vector unsigned short,
6841 vector signed short);
6842 vector unsigned short vec_subs (vector unsigned short,
6843 vector unsigned short);
6844 vector signed short vec_subs (vector signed short, vector signed short);
6846 vector unsigned int vec_subs (vector signed int, vector unsigned int);
6847 vector unsigned int vec_subs (vector unsigned int, vector signed int);
6848 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
6850 vector signed int vec_subs (vector signed int, vector signed int);
6852 vector unsigned int vec_sum4s (vector unsigned char,
6853 vector unsigned int);
6854 vector signed int vec_sum4s (vector signed char, vector signed int);
6855 vector signed int vec_sum4s (vector signed short, vector signed int);
6857 vector signed int vec_sum2s (vector signed int, vector signed int);
6859 vector signed int vec_sums (vector signed int, vector signed int);
6861 vector float vec_trunc (vector float);
6863 vector signed short vec_unpackh (vector signed char);
6864 vector unsigned int vec_unpackh (vector signed short);
6865 vector signed int vec_unpackh (vector signed short);
6867 vector signed short vec_unpackl (vector signed char);
6868 vector unsigned int vec_unpackl (vector signed short);
6869 vector signed int vec_unpackl (vector signed short);
6871 vector float vec_xor (vector float, vector float);
6872 vector float vec_xor (vector float, vector signed int);
6873 vector float vec_xor (vector signed int, vector float);
6874 vector signed int vec_xor (vector signed int, vector signed int);
6875 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6876 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6877 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6878 vector signed short vec_xor (vector signed short, vector signed short);
6879 vector unsigned short vec_xor (vector signed short,
6880 vector unsigned short);
6881 vector unsigned short vec_xor (vector unsigned short,
6882 vector signed short);
6883 vector unsigned short vec_xor (vector unsigned short,
6884 vector unsigned short);
6885 vector signed char vec_xor (vector signed char, vector signed char);
6886 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6888 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6890 vector unsigned char vec_xor (vector unsigned char,
6891 vector unsigned char);
6893 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6895 vector signed int vec_all_eq (vector signed char, vector signed char);
6896 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6898 vector signed int vec_all_eq (vector unsigned char,
6899 vector unsigned char);
6900 vector signed int vec_all_eq (vector signed short,
6901 vector unsigned short);
6902 vector signed int vec_all_eq (vector signed short, vector signed short);
6904 vector signed int vec_all_eq (vector unsigned short,
6905 vector signed short);
6906 vector signed int vec_all_eq (vector unsigned short,
6907 vector unsigned short);
6908 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6909 vector signed int vec_all_eq (vector signed int, vector signed int);
6910 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6911 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6913 vector signed int vec_all_eq (vector float, vector float);
6915 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6917 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6919 vector signed int vec_all_ge (vector unsigned char,
6920 vector unsigned char);
6921 vector signed int vec_all_ge (vector signed char, vector signed char);
6922 vector signed int vec_all_ge (vector signed short,
6923 vector unsigned short);
6924 vector signed int vec_all_ge (vector unsigned short,
6925 vector signed short);
6926 vector signed int vec_all_ge (vector unsigned short,
6927 vector unsigned short);
6928 vector signed int vec_all_ge (vector signed short, vector signed short);
6930 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6931 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6932 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6934 vector signed int vec_all_ge (vector signed int, vector signed int);
6935 vector signed int vec_all_ge (vector float, vector float);
6937 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6939 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6941 vector signed int vec_all_gt (vector unsigned char,
6942 vector unsigned char);
6943 vector signed int vec_all_gt (vector signed char, vector signed char);
6944 vector signed int vec_all_gt (vector signed short,
6945 vector unsigned short);
6946 vector signed int vec_all_gt (vector unsigned short,
6947 vector signed short);
6948 vector signed int vec_all_gt (vector unsigned short,
6949 vector unsigned short);
6950 vector signed int vec_all_gt (vector signed short, vector signed short);
6952 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6953 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6954 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6956 vector signed int vec_all_gt (vector signed int, vector signed int);
6957 vector signed int vec_all_gt (vector float, vector float);
6959 vector signed int vec_all_in (vector float, vector float);
6961 vector signed int vec_all_le (vector signed char, vector unsigned char);
6963 vector signed int vec_all_le (vector unsigned char, vector signed char);
6965 vector signed int vec_all_le (vector unsigned char,
6966 vector unsigned char);
6967 vector signed int vec_all_le (vector signed char, vector signed char);
6968 vector signed int vec_all_le (vector signed short,
6969 vector unsigned short);
6970 vector signed int vec_all_le (vector unsigned short,
6971 vector signed short);
6972 vector signed int vec_all_le (vector unsigned short,
6973 vector unsigned short);
6974 vector signed int vec_all_le (vector signed short, vector signed short);
6976 vector signed int vec_all_le (vector signed int, vector unsigned int);
6977 vector signed int vec_all_le (vector unsigned int, vector signed int);
6978 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
6980 vector signed int vec_all_le (vector signed int, vector signed int);
6981 vector signed int vec_all_le (vector float, vector float);
6983 vector signed int vec_all_lt (vector signed char, vector unsigned char);
6985 vector signed int vec_all_lt (vector unsigned char, vector signed char);
6987 vector signed int vec_all_lt (vector unsigned char,
6988 vector unsigned char);
6989 vector signed int vec_all_lt (vector signed char, vector signed char);
6990 vector signed int vec_all_lt (vector signed short,
6991 vector unsigned short);
6992 vector signed int vec_all_lt (vector unsigned short,
6993 vector signed short);
6994 vector signed int vec_all_lt (vector unsigned short,
6995 vector unsigned short);
6996 vector signed int vec_all_lt (vector signed short, vector signed short);
6998 vector signed int vec_all_lt (vector signed int, vector unsigned int);
6999 vector signed int vec_all_lt (vector unsigned int, vector signed int);
7000 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
7002 vector signed int vec_all_lt (vector signed int, vector signed int);
7003 vector signed int vec_all_lt (vector float, vector float);
7005 vector signed int vec_all_nan (vector float);
7007 vector signed int vec_all_ne (vector signed char, vector unsigned char);
7009 vector signed int vec_all_ne (vector signed char, vector signed char);
7010 vector signed int vec_all_ne (vector unsigned char, vector signed char);
7012 vector signed int vec_all_ne (vector unsigned char,
7013 vector unsigned char);
7014 vector signed int vec_all_ne (vector signed short,
7015 vector unsigned short);
7016 vector signed int vec_all_ne (vector signed short, vector signed short);
7018 vector signed int vec_all_ne (vector unsigned short,
7019 vector signed short);
7020 vector signed int vec_all_ne (vector unsigned short,
7021 vector unsigned short);
7022 vector signed int vec_all_ne (vector signed int, vector unsigned int);
7023 vector signed int vec_all_ne (vector signed int, vector signed int);
7024 vector signed int vec_all_ne (vector unsigned int, vector signed int);
7025 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
7027 vector signed int vec_all_ne (vector float, vector float);
7029 vector signed int vec_all_nge (vector float, vector float);
7031 vector signed int vec_all_ngt (vector float, vector float);
7033 vector signed int vec_all_nle (vector float, vector float);
7035 vector signed int vec_all_nlt (vector float, vector float);
7037 vector signed int vec_all_numeric (vector float);
7039 vector signed int vec_any_eq (vector signed char, vector unsigned char);
7041 vector signed int vec_any_eq (vector signed char, vector signed char);
7042 vector signed int vec_any_eq (vector unsigned char, vector signed char);
7044 vector signed int vec_any_eq (vector unsigned char,
7045 vector unsigned char);
7046 vector signed int vec_any_eq (vector signed short,
7047 vector unsigned short);
7048 vector signed int vec_any_eq (vector signed short, vector signed short);
7050 vector signed int vec_any_eq (vector unsigned short,
7051 vector signed short);
7052 vector signed int vec_any_eq (vector unsigned short,
7053 vector unsigned short);
7054 vector signed int vec_any_eq (vector signed int, vector unsigned int);
7055 vector signed int vec_any_eq (vector signed int, vector signed int);
7056 vector signed int vec_any_eq (vector unsigned int, vector signed int);
7057 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
7059 vector signed int vec_any_eq (vector float, vector float);
7061 vector signed int vec_any_ge (vector signed char, vector unsigned char);
7063 vector signed int vec_any_ge (vector unsigned char, vector signed char);
7065 vector signed int vec_any_ge (vector unsigned char,
7066 vector unsigned char);
7067 vector signed int vec_any_ge (vector signed char, vector signed char);
7068 vector signed int vec_any_ge (vector signed short,
7069 vector unsigned short);
7070 vector signed int vec_any_ge (vector unsigned short,
7071 vector signed short);
7072 vector signed int vec_any_ge (vector unsigned short,
7073 vector unsigned short);
7074 vector signed int vec_any_ge (vector signed short, vector signed short);
7076 vector signed int vec_any_ge (vector signed int, vector unsigned int);
7077 vector signed int vec_any_ge (vector unsigned int, vector signed int);
7078 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
7080 vector signed int vec_any_ge (vector signed int, vector signed int);
7081 vector signed int vec_any_ge (vector float, vector float);
7083 vector signed int vec_any_gt (vector signed char, vector unsigned char);
7085 vector signed int vec_any_gt (vector unsigned char, vector signed char);
7087 vector signed int vec_any_gt (vector unsigned char,
7088 vector unsigned char);
7089 vector signed int vec_any_gt (vector signed char, vector signed char);
7090 vector signed int vec_any_gt (vector signed short,
7091 vector unsigned short);
7092 vector signed int vec_any_gt (vector unsigned short,
7093 vector signed short);
7094 vector signed int vec_any_gt (vector unsigned short,
7095 vector unsigned short);
7096 vector signed int vec_any_gt (vector signed short, vector signed short);
7098 vector signed int vec_any_gt (vector signed int, vector unsigned int);
7099 vector signed int vec_any_gt (vector unsigned int, vector signed int);
7100 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
7102 vector signed int vec_any_gt (vector signed int, vector signed int);
7103 vector signed int vec_any_gt (vector float, vector float);
7105 vector signed int vec_any_le (vector signed char, vector unsigned char);
7107 vector signed int vec_any_le (vector unsigned char, vector signed char);
7109 vector signed int vec_any_le (vector unsigned char,
7110 vector unsigned char);
7111 vector signed int vec_any_le (vector signed char, vector signed char);
7112 vector signed int vec_any_le (vector signed short,
7113 vector unsigned short);
7114 vector signed int vec_any_le (vector unsigned short,
7115 vector signed short);
7116 vector signed int vec_any_le (vector unsigned short,
7117 vector unsigned short);
7118 vector signed int vec_any_le (vector signed short, vector signed short);
7120 vector signed int vec_any_le (vector signed int, vector unsigned int);
7121 vector signed int vec_any_le (vector unsigned int, vector signed int);
7122 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
7124 vector signed int vec_any_le (vector signed int, vector signed int);
7125 vector signed int vec_any_le (vector float, vector float);
7127 vector signed int vec_any_lt (vector signed char, vector unsigned char);
7129 vector signed int vec_any_lt (vector unsigned char, vector signed char);
7131 vector signed int vec_any_lt (vector unsigned char,
7132 vector unsigned char);
7133 vector signed int vec_any_lt (vector signed char, vector signed char);
7134 vector signed int vec_any_lt (vector signed short,
7135 vector unsigned short);
7136 vector signed int vec_any_lt (vector unsigned short,
7137 vector signed short);
7138 vector signed int vec_any_lt (vector unsigned short,
7139 vector unsigned short);
7140 vector signed int vec_any_lt (vector signed short, vector signed short);
7142 vector signed int vec_any_lt (vector signed int, vector unsigned int);
7143 vector signed int vec_any_lt (vector unsigned int, vector signed int);
7144 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
7146 vector signed int vec_any_lt (vector signed int, vector signed int);
7147 vector signed int vec_any_lt (vector float, vector float);
7149 vector signed int vec_any_nan (vector float);
7151 vector signed int vec_any_ne (vector signed char, vector unsigned char);
7153 vector signed int vec_any_ne (vector signed char, vector signed char);
7154 vector signed int vec_any_ne (vector unsigned char, vector signed char);
7156 vector signed int vec_any_ne (vector unsigned char,
7157 vector unsigned char);
7158 vector signed int vec_any_ne (vector signed short,
7159 vector unsigned short);
7160 vector signed int vec_any_ne (vector signed short, vector signed short);
7162 vector signed int vec_any_ne (vector unsigned short,
7163 vector signed short);
7164 vector signed int vec_any_ne (vector unsigned short,
7165 vector unsigned short);
7166 vector signed int vec_any_ne (vector signed int, vector unsigned int);
7167 vector signed int vec_any_ne (vector signed int, vector signed int);
7168 vector signed int vec_any_ne (vector unsigned int, vector signed int);
7169 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
7171 vector signed int vec_any_ne (vector float, vector float);
7173 vector signed int vec_any_nge (vector float, vector float);
7175 vector signed int vec_any_ngt (vector float, vector float);
7177 vector signed int vec_any_nle (vector float, vector float);
7179 vector signed int vec_any_nlt (vector float, vector float);
7181 vector signed int vec_any_numeric (vector float);
7183 vector signed int vec_any_out (vector float, vector float);
7187 @section Pragmas Accepted by GCC
7191 GCC supports several types of pragmas, primarily in order to compile
7192 code originally written for other compilers. Note that in general
7193 we do not recommend the use of pragmas; @xref{Function Attributes},
7194 for further explanation.
7198 * RS/6000 and PowerPC Pragmas::
7205 @subsection ARM Pragmas
7207 The ARM target defines pragmas for controlling the default addition of
7208 @code{long_call} and @code{short_call} attributes to functions.
7209 @xref{Function Attributes}, for information about the effects of these
7214 @cindex pragma, long_calls
7215 Set all subsequent functions to have the @code{long_call} attribute.
7218 @cindex pragma, no_long_calls
7219 Set all subsequent functions to have the @code{short_call} attribute.
7221 @item long_calls_off
7222 @cindex pragma, long_calls_off
7223 Do not affect the @code{long_call} or @code{short_call} attributes of
7224 subsequent functions.
7227 @node RS/6000 and PowerPC Pragmas
7228 @subsection RS/6000 and PowerPC Pragmas
7230 The RS/6000 and PowerPC targets define one pragma for controlling
7231 whether or not the @code{longcall} attribute is added to function
7232 declarations by default. This pragma overrides the @option{-mlongcall}
7233 option, but not the @code{longcall} and @code{shortcall} attributes.
7234 @xref{RS/6000 and PowerPC Options}, for more information about when long
7235 calls are and are not necessary.
7239 @cindex pragma, longcall
7240 Apply the @code{longcall} attribute to all subsequent function
7244 Do not apply the @code{longcall} attribute to subsequent function
7248 @c Describe c4x pragmas here.
7249 @c Describe h8300 pragmas here.
7250 @c Describe sh pragmas here.
7251 @c Describe v850 pragmas here.
7253 @node Darwin Pragmas
7254 @subsection Darwin Pragmas
7256 The following pragmas are available for all architectures running the
7257 Darwin operating system. These are useful for compatibility with other
7261 @item mark @var{tokens}@dots{}
7262 @cindex pragma, mark
7263 This pragma is accepted, but has no effect.
7265 @item options align=@var{alignment}
7266 @cindex pragma, options align
7267 This pragma sets the alignment of fields in structures. The values of
7268 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
7269 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
7270 properly; to restore the previous setting, use @code{reset} for the
7273 @item segment @var{tokens}@dots{}
7274 @cindex pragma, segment
7275 This pragma is accepted, but has no effect.
7277 @item unused (@var{var} [, @var{var}]@dots{})
7278 @cindex pragma, unused
7279 This pragma declares variables to be possibly unused. GCC will not
7280 produce warnings for the listed variables. The effect is similar to
7281 that of the @code{unused} attribute, except that this pragma may appear
7282 anywhere within the variables' scopes.
7285 @node Solaris Pragmas
7286 @subsection Solaris Pragmas
7288 For compatibility with the SunPRO compiler, the following pragma
7292 @item redefine_extname @var{oldname} @var{newname}
7293 @cindex pragma, redefine_extname
7295 This pragma gives the C function @var{oldname} the assembler label
7296 @var{newname}. The pragma must appear before the function declaration.
7297 This pragma is equivalent to the asm labels extension (@pxref{Asm
7298 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
7299 if the pragma is available.
7303 @subsection Tru64 Pragmas
7305 For compatibility with the Compaq C compiler, the following pragma
7309 @item extern_prefix @var{string}
7310 @cindex pragma, extern_prefix
7312 This pragma renames all subsequent function and variable declarations
7313 such that @var{string} is prepended to the name. This effect may be
7314 terminated by using another @code{extern_prefix} pragma with the
7317 This pragma is similar in intent to to the asm labels extension
7318 (@pxref{Asm Labels}) in that the system programmer wants to change
7319 the assembly-level ABI without changing the source-level API. The
7320 preprocessor defines @code{__PRAGMA_EXTERN_PREFIX} if the pragma is
7324 @node Unnamed Fields
7325 @section Unnamed struct/union fields within structs/unions.
7329 For compatibility with other compilers, GCC allows you to define
7330 a structure or union that contains, as fields, structures and unions
7331 without names. For example:
7344 In this example, the user would be able to access members of the unnamed
7345 union with code like @samp{foo.b}. Note that only unnamed structs and
7346 unions are allowed, you may not have, for example, an unnamed
7349 You must never create such structures that cause ambiguous field definitions.
7350 For example, this structure:
7361 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
7362 Such constructs are not supported and must be avoided. In the future,
7363 such constructs may be detected and treated as compilation errors.
7366 @section Thread-Local Storage
7367 @cindex Thread-Local Storage
7368 @cindex @acronym{TLS}
7371 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
7372 are allocated such that there is one instance of the variable per extant
7373 thread. The run-time model GCC uses to implement this originates
7374 in the IA-64 processor-specific ABI, but has since been migrated
7375 to other processors as well. It requires significant support from
7376 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
7377 system libraries (@file{libc.so} and @file{libpthread.so}), so it
7378 is not available everywhere.
7380 At the user level, the extension is visible with a new storage
7381 class keyword: @code{__thread}. For example:
7385 extern __thread struct state s;
7386 static __thread char *p;
7389 The @code{__thread} specifier may be used alone, with the @code{extern}
7390 or @code{static} specifiers, but with no other storage class specifier.
7391 When used with @code{extern} or @code{static}, @code{__thread} must appear
7392 immediately after the other storage class specifier.
7394 The @code{__thread} specifier may be applied to any global, file-scoped
7395 static, function-scoped static, or static data member of a class. It may
7396 not be applied to block-scoped automatic or non-static data member.
7398 When the address-of operator is applied to a thread-local variable, it is
7399 evaluated at run-time and returns the address of the current thread's
7400 instance of that variable. An address so obtained may be used by any
7401 thread. When a thread terminates, any pointers to thread-local variables
7402 in that thread become invalid.
7404 No static initialization may refer to the address of a thread-local variable.
7406 In C++, if an initializer is present for a thread-local variable, it must
7407 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
7410 See @uref{http://people.redhat.com/drepper/tls.pdf,
7411 ELF Handling For Thread-Local Storage} for a detailed explanation of
7412 the four thread-local storage addressing models, and how the run-time
7413 is expected to function.
7416 * C99 Thread-Local Edits::
7417 * C++98 Thread-Local Edits::
7420 @node C99 Thread-Local Edits
7421 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
7423 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
7424 that document the exact semantics of the language extension.
7428 @cite{5.1.2 Execution environments}
7430 Add new text after paragraph 1
7433 Within either execution environment, a @dfn{thread} is a flow of
7434 control within a program. It is implementation defined whether
7435 or not there may be more than one thread associated with a program.
7436 It is implementation defined how threads beyond the first are
7437 created, the name and type of the function called at thread
7438 startup, and how threads may be terminated. However, objects
7439 with thread storage duration shall be initialized before thread
7444 @cite{6.2.4 Storage durations of objects}
7446 Add new text before paragraph 3
7449 An object whose identifier is declared with the storage-class
7450 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
7451 Its lifetime is the entire execution of the thread, and its
7452 stored value is initialized only once, prior to thread startup.
7456 @cite{6.4.1 Keywords}
7458 Add @code{__thread}.
7461 @cite{6.7.1 Storage-class specifiers}
7463 Add @code{__thread} to the list of storage class specifiers in
7466 Change paragraph 2 to
7469 With the exception of @code{__thread}, at most one storage-class
7470 specifier may be given [@dots{}]. The @code{__thread} specifier may
7471 be used alone, or immediately following @code{extern} or
7475 Add new text after paragraph 6
7478 The declaration of an identifier for a variable that has
7479 block scope that specifies @code{__thread} shall also
7480 specify either @code{extern} or @code{static}.
7482 The @code{__thread} specifier shall be used only with
7487 @node C++98 Thread-Local Edits
7488 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
7490 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
7491 that document the exact semantics of the language extension.
7495 @b{[intro.execution]}
7497 New text after paragraph 4
7500 A @dfn{thread} is a flow of control within the abstract machine.
7501 It is implementation defined whether or not there may be more than
7505 New text after paragraph 7
7508 It is unspecified whether additional action must be taken to
7509 ensure when and whether side effects are visible to other threads.
7515 Add @code{__thread}.
7518 @b{[basic.start.main]}
7520 Add after paragraph 5
7523 The thread that begins execution at the @code{main} function is called
7524 the @dfn{main thread}. It is implementation defined how functions
7525 beginning threads other than the main thread are designated or typed.
7526 A function so designated, as well as the @code{main} function, is called
7527 a @dfn{thread startup function}. It is implementation defined what
7528 happens if a thread startup function returns. It is implementation
7529 defined what happens to other threads when any thread calls @code{exit}.
7533 @b{[basic.start.init]}
7535 Add after paragraph 4
7538 The storage for an object of thread storage duration shall be
7539 statically initialized before the first statement of the thread startup
7540 function. An object of thread storage duration shall not require
7541 dynamic initialization.
7545 @b{[basic.start.term]}
7547 Add after paragraph 3
7550 The type of an object with thread storage duration shall not have a
7551 non-trivial destructor, nor shall it be an array type whose elements
7552 (directly or indirectly) have non-trivial destructors.
7558 Add ``thread storage duration'' to the list in paragraph 1.
7563 Thread, static, and automatic storage durations are associated with
7564 objects introduced by declarations [@dots{}].
7567 Add @code{__thread} to the list of specifiers in paragraph 3.
7570 @b{[basic.stc.thread]}
7572 New section before @b{[basic.stc.static]}
7575 The keyword @code{__thread} applied to a non-local object gives the
7576 object thread storage duration.
7578 A local variable or class data member declared both @code{static}
7579 and @code{__thread} gives the variable or member thread storage
7584 @b{[basic.stc.static]}
7589 All objects which have neither thread storage duration, dynamic
7590 storage duration nor are local [@dots{}].
7596 Add @code{__thread} to the list in paragraph 1.
7601 With the exception of @code{__thread}, at most one
7602 @var{storage-class-specifier} shall appear in a given
7603 @var{decl-specifier-seq}. The @code{__thread} specifier may
7604 be used alone, or immediately following the @code{extern} or
7605 @code{static} specifiers. [@dots{}]
7608 Add after paragraph 5
7611 The @code{__thread} specifier can be applied only to the names of objects
7612 and to anonymous unions.
7618 Add after paragraph 6
7621 Non-@code{static} members shall not be @code{__thread}.
7625 @node C++ Extensions
7626 @chapter Extensions to the C++ Language
7627 @cindex extensions, C++ language
7628 @cindex C++ language extensions
7630 The GNU compiler provides these extensions to the C++ language (and you
7631 can also use most of the C language extensions in your C++ programs). If you
7632 want to write code that checks whether these features are available, you can
7633 test for the GNU compiler the same way as for C programs: check for a
7634 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
7635 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
7636 Predefined Macros,cpp,The GNU C Preprocessor}).
7639 * Min and Max:: C++ Minimum and maximum operators.
7640 * Volatiles:: What constitutes an access to a volatile object.
7641 * Restricted Pointers:: C99 restricted pointers and references.
7642 * Vague Linkage:: Where G++ puts inlines, vtables and such.
7643 * C++ Interface:: You can use a single C++ header file for both
7644 declarations and definitions.
7645 * Template Instantiation:: Methods for ensuring that exactly one copy of
7646 each needed template instantiation is emitted.
7647 * Bound member functions:: You can extract a function pointer to the
7648 method denoted by a @samp{->*} or @samp{.*} expression.
7649 * C++ Attributes:: Variable, function, and type attributes for C++ only.
7650 * Strong Using:: Strong using-directives for namespace composition.
7651 * Java Exceptions:: Tweaking exception handling to work with Java.
7652 * Deprecated Features:: Things will disappear from g++.
7653 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
7657 @section Minimum and Maximum Operators in C++
7659 It is very convenient to have operators which return the ``minimum'' or the
7660 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
7663 @item @var{a} <? @var{b}
7665 @cindex minimum operator
7666 is the @dfn{minimum}, returning the smaller of the numeric values
7667 @var{a} and @var{b};
7669 @item @var{a} >? @var{b}
7671 @cindex maximum operator
7672 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
7676 These operations are not primitive in ordinary C++, since you can
7677 use a macro to return the minimum of two things in C++, as in the
7681 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
7685 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
7686 the minimum value of variables @var{i} and @var{j}.
7688 However, side effects in @code{X} or @code{Y} may cause unintended
7689 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
7690 the smaller counter twice. The GNU C @code{typeof} extension allows you
7691 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
7692 However, writing @code{MIN} and @code{MAX} as macros also forces you to
7693 use function-call notation for a fundamental arithmetic operation.
7694 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
7697 Since @code{<?} and @code{>?} are built into the compiler, they properly
7698 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
7702 @section When is a Volatile Object Accessed?
7703 @cindex accessing volatiles
7704 @cindex volatile read
7705 @cindex volatile write
7706 @cindex volatile access
7708 Both the C and C++ standard have the concept of volatile objects. These
7709 are normally accessed by pointers and used for accessing hardware. The
7710 standards encourage compilers to refrain from optimizations
7711 concerning accesses to volatile objects that it might perform on
7712 non-volatile objects. The C standard leaves it implementation defined
7713 as to what constitutes a volatile access. The C++ standard omits to
7714 specify this, except to say that C++ should behave in a similar manner
7715 to C with respect to volatiles, where possible. The minimum either
7716 standard specifies is that at a sequence point all previous accesses to
7717 volatile objects have stabilized and no subsequent accesses have
7718 occurred. Thus an implementation is free to reorder and combine
7719 volatile accesses which occur between sequence points, but cannot do so
7720 for accesses across a sequence point. The use of volatiles does not
7721 allow you to violate the restriction on updating objects multiple times
7722 within a sequence point.
7724 In most expressions, it is intuitively obvious what is a read and what is
7725 a write. For instance
7728 volatile int *dst = @var{somevalue};
7729 volatile int *src = @var{someothervalue};
7734 will cause a read of the volatile object pointed to by @var{src} and stores the
7735 value into the volatile object pointed to by @var{dst}. There is no
7736 guarantee that these reads and writes are atomic, especially for objects
7737 larger than @code{int}.
7739 Less obvious expressions are where something which looks like an access
7740 is used in a void context. An example would be,
7743 volatile int *src = @var{somevalue};
7747 With C, such expressions are rvalues, and as rvalues cause a read of
7748 the object, GCC interprets this as a read of the volatile being pointed
7749 to. The C++ standard specifies that such expressions do not undergo
7750 lvalue to rvalue conversion, and that the type of the dereferenced
7751 object may be incomplete. The C++ standard does not specify explicitly
7752 that it is this lvalue to rvalue conversion which is responsible for
7753 causing an access. However, there is reason to believe that it is,
7754 because otherwise certain simple expressions become undefined. However,
7755 because it would surprise most programmers, G++ treats dereferencing a
7756 pointer to volatile object of complete type in a void context as a read
7757 of the object. When the object has incomplete type, G++ issues a
7762 struct T @{int m;@};
7763 volatile S *ptr1 = @var{somevalue};
7764 volatile T *ptr2 = @var{somevalue};
7769 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
7770 causes a read of the object pointed to. If you wish to force an error on
7771 the first case, you must force a conversion to rvalue with, for instance
7772 a static cast, @code{static_cast<S>(*ptr1)}.
7774 When using a reference to volatile, G++ does not treat equivalent
7775 expressions as accesses to volatiles, but instead issues a warning that
7776 no volatile is accessed. The rationale for this is that otherwise it
7777 becomes difficult to determine where volatile access occur, and not
7778 possible to ignore the return value from functions returning volatile
7779 references. Again, if you wish to force a read, cast the reference to
7782 @node Restricted Pointers
7783 @section Restricting Pointer Aliasing
7784 @cindex restricted pointers
7785 @cindex restricted references
7786 @cindex restricted this pointer
7788 As with the C front end, G++ understands the C99 feature of restricted pointers,
7789 specified with the @code{__restrict__}, or @code{__restrict} type
7790 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
7791 language flag, @code{restrict} is not a keyword in C++.
7793 In addition to allowing restricted pointers, you can specify restricted
7794 references, which indicate that the reference is not aliased in the local
7798 void fn (int *__restrict__ rptr, int &__restrict__ rref)
7805 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
7806 @var{rref} refers to a (different) unaliased integer.
7808 You may also specify whether a member function's @var{this} pointer is
7809 unaliased by using @code{__restrict__} as a member function qualifier.
7812 void T::fn () __restrict__
7819 Within the body of @code{T::fn}, @var{this} will have the effective
7820 definition @code{T *__restrict__ const this}. Notice that the
7821 interpretation of a @code{__restrict__} member function qualifier is
7822 different to that of @code{const} or @code{volatile} qualifier, in that it
7823 is applied to the pointer rather than the object. This is consistent with
7824 other compilers which implement restricted pointers.
7826 As with all outermost parameter qualifiers, @code{__restrict__} is
7827 ignored in function definition matching. This means you only need to
7828 specify @code{__restrict__} in a function definition, rather than
7829 in a function prototype as well.
7832 @section Vague Linkage
7833 @cindex vague linkage
7835 There are several constructs in C++ which require space in the object
7836 file but are not clearly tied to a single translation unit. We say that
7837 these constructs have ``vague linkage''. Typically such constructs are
7838 emitted wherever they are needed, though sometimes we can be more
7842 @item Inline Functions
7843 Inline functions are typically defined in a header file which can be
7844 included in many different compilations. Hopefully they can usually be
7845 inlined, but sometimes an out-of-line copy is necessary, if the address
7846 of the function is taken or if inlining fails. In general, we emit an
7847 out-of-line copy in all translation units where one is needed. As an
7848 exception, we only emit inline virtual functions with the vtable, since
7849 it will always require a copy.
7851 Local static variables and string constants used in an inline function
7852 are also considered to have vague linkage, since they must be shared
7853 between all inlined and out-of-line instances of the function.
7857 C++ virtual functions are implemented in most compilers using a lookup
7858 table, known as a vtable. The vtable contains pointers to the virtual
7859 functions provided by a class, and each object of the class contains a
7860 pointer to its vtable (or vtables, in some multiple-inheritance
7861 situations). If the class declares any non-inline, non-pure virtual
7862 functions, the first one is chosen as the ``key method'' for the class,
7863 and the vtable is only emitted in the translation unit where the key
7866 @emph{Note:} If the chosen key method is later defined as inline, the
7867 vtable will still be emitted in every translation unit which defines it.
7868 Make sure that any inline virtuals are declared inline in the class
7869 body, even if they are not defined there.
7871 @item type_info objects
7874 C++ requires information about types to be written out in order to
7875 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7876 For polymorphic classes (classes with virtual functions), the type_info
7877 object is written out along with the vtable so that @samp{dynamic_cast}
7878 can determine the dynamic type of a class object at runtime. For all
7879 other types, we write out the type_info object when it is used: when
7880 applying @samp{typeid} to an expression, throwing an object, or
7881 referring to a type in a catch clause or exception specification.
7883 @item Template Instantiations
7884 Most everything in this section also applies to template instantiations,
7885 but there are other options as well.
7886 @xref{Template Instantiation,,Where's the Template?}.
7890 When used with GNU ld version 2.8 or later on an ELF system such as
7891 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
7892 these constructs will be discarded at link time. This is known as
7895 On targets that don't support COMDAT, but do support weak symbols, GCC
7896 will use them. This way one copy will override all the others, but
7897 the unused copies will still take up space in the executable.
7899 For targets which do not support either COMDAT or weak symbols,
7900 most entities with vague linkage will be emitted as local symbols to
7901 avoid duplicate definition errors from the linker. This will not happen
7902 for local statics in inlines, however, as having multiple copies will
7903 almost certainly break things.
7905 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7906 another way to control placement of these constructs.
7909 @section Declarations and Definitions in One Header
7911 @cindex interface and implementation headers, C++
7912 @cindex C++ interface and implementation headers
7913 C++ object definitions can be quite complex. In principle, your source
7914 code will need two kinds of things for each object that you use across
7915 more than one source file. First, you need an @dfn{interface}
7916 specification, describing its structure with type declarations and
7917 function prototypes. Second, you need the @dfn{implementation} itself.
7918 It can be tedious to maintain a separate interface description in a
7919 header file, in parallel to the actual implementation. It is also
7920 dangerous, since separate interface and implementation definitions may
7921 not remain parallel.
7923 @cindex pragmas, interface and implementation
7924 With GNU C++, you can use a single header file for both purposes.
7927 @emph{Warning:} The mechanism to specify this is in transition. For the
7928 nonce, you must use one of two @code{#pragma} commands; in a future
7929 release of GNU C++, an alternative mechanism will make these
7930 @code{#pragma} commands unnecessary.
7933 The header file contains the full definitions, but is marked with
7934 @samp{#pragma interface} in the source code. This allows the compiler
7935 to use the header file only as an interface specification when ordinary
7936 source files incorporate it with @code{#include}. In the single source
7937 file where the full implementation belongs, you can use either a naming
7938 convention or @samp{#pragma implementation} to indicate this alternate
7939 use of the header file.
7942 @item #pragma interface
7943 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7944 @kindex #pragma interface
7945 Use this directive in @emph{header files} that define object classes, to save
7946 space in most of the object files that use those classes. Normally,
7947 local copies of certain information (backup copies of inline member
7948 functions, debugging information, and the internal tables that implement
7949 virtual functions) must be kept in each object file that includes class
7950 definitions. You can use this pragma to avoid such duplication. When a
7951 header file containing @samp{#pragma interface} is included in a
7952 compilation, this auxiliary information will not be generated (unless
7953 the main input source file itself uses @samp{#pragma implementation}).
7954 Instead, the object files will contain references to be resolved at link
7957 The second form of this directive is useful for the case where you have
7958 multiple headers with the same name in different directories. If you
7959 use this form, you must specify the same string to @samp{#pragma
7962 @item #pragma implementation
7963 @itemx #pragma implementation "@var{objects}.h"
7964 @kindex #pragma implementation
7965 Use this pragma in a @emph{main input file}, when you want full output from
7966 included header files to be generated (and made globally visible). The
7967 included header file, in turn, should use @samp{#pragma interface}.
7968 Backup copies of inline member functions, debugging information, and the
7969 internal tables used to implement virtual functions are all generated in
7970 implementation files.
7972 @cindex implied @code{#pragma implementation}
7973 @cindex @code{#pragma implementation}, implied
7974 @cindex naming convention, implementation headers
7975 If you use @samp{#pragma implementation} with no argument, it applies to
7976 an include file with the same basename@footnote{A file's @dfn{basename}
7977 was the name stripped of all leading path information and of trailing
7978 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
7979 file. For example, in @file{allclass.cc}, giving just
7980 @samp{#pragma implementation}
7981 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
7983 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
7984 an implementation file whenever you would include it from
7985 @file{allclass.cc} even if you never specified @samp{#pragma
7986 implementation}. This was deemed to be more trouble than it was worth,
7987 however, and disabled.
7989 If you use an explicit @samp{#pragma implementation}, it must appear in
7990 your source file @emph{before} you include the affected header files.
7992 Use the string argument if you want a single implementation file to
7993 include code from multiple header files. (You must also use
7994 @samp{#include} to include the header file; @samp{#pragma
7995 implementation} only specifies how to use the file---it doesn't actually
7998 There is no way to split up the contents of a single header file into
7999 multiple implementation files.
8002 @cindex inlining and C++ pragmas
8003 @cindex C++ pragmas, effect on inlining
8004 @cindex pragmas in C++, effect on inlining
8005 @samp{#pragma implementation} and @samp{#pragma interface} also have an
8006 effect on function inlining.
8008 If you define a class in a header file marked with @samp{#pragma
8009 interface}, the effect on a function defined in that class is similar to
8010 an explicit @code{extern} declaration---the compiler emits no code at
8011 all to define an independent version of the function. Its definition
8012 is used only for inlining with its callers.
8014 @opindex fno-implement-inlines
8015 Conversely, when you include the same header file in a main source file
8016 that declares it as @samp{#pragma implementation}, the compiler emits
8017 code for the function itself; this defines a version of the function
8018 that can be found via pointers (or by callers compiled without
8019 inlining). If all calls to the function can be inlined, you can avoid
8020 emitting the function by compiling with @option{-fno-implement-inlines}.
8021 If any calls were not inlined, you will get linker errors.
8023 @node Template Instantiation
8024 @section Where's the Template?
8025 @cindex template instantiation
8027 C++ templates are the first language feature to require more
8028 intelligence from the environment than one usually finds on a UNIX
8029 system. Somehow the compiler and linker have to make sure that each
8030 template instance occurs exactly once in the executable if it is needed,
8031 and not at all otherwise. There are two basic approaches to this
8032 problem, which I will refer to as the Borland model and the Cfront model.
8036 Borland C++ solved the template instantiation problem by adding the code
8037 equivalent of common blocks to their linker; the compiler emits template
8038 instances in each translation unit that uses them, and the linker
8039 collapses them together. The advantage of this model is that the linker
8040 only has to consider the object files themselves; there is no external
8041 complexity to worry about. This disadvantage is that compilation time
8042 is increased because the template code is being compiled repeatedly.
8043 Code written for this model tends to include definitions of all
8044 templates in the header file, since they must be seen to be
8048 The AT&T C++ translator, Cfront, solved the template instantiation
8049 problem by creating the notion of a template repository, an
8050 automatically maintained place where template instances are stored. A
8051 more modern version of the repository works as follows: As individual
8052 object files are built, the compiler places any template definitions and
8053 instantiations encountered in the repository. At link time, the link
8054 wrapper adds in the objects in the repository and compiles any needed
8055 instances that were not previously emitted. The advantages of this
8056 model are more optimal compilation speed and the ability to use the
8057 system linker; to implement the Borland model a compiler vendor also
8058 needs to replace the linker. The disadvantages are vastly increased
8059 complexity, and thus potential for error; for some code this can be
8060 just as transparent, but in practice it can been very difficult to build
8061 multiple programs in one directory and one program in multiple
8062 directories. Code written for this model tends to separate definitions
8063 of non-inline member templates into a separate file, which should be
8064 compiled separately.
8067 When used with GNU ld version 2.8 or later on an ELF system such as
8068 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
8069 Borland model. On other systems, G++ implements neither automatic
8072 A future version of G++ will support a hybrid model whereby the compiler
8073 will emit any instantiations for which the template definition is
8074 included in the compile, and store template definitions and
8075 instantiation context information into the object file for the rest.
8076 The link wrapper will extract that information as necessary and invoke
8077 the compiler to produce the remaining instantiations. The linker will
8078 then combine duplicate instantiations.
8080 In the mean time, you have the following options for dealing with
8081 template instantiations:
8086 Compile your template-using code with @option{-frepo}. The compiler will
8087 generate files with the extension @samp{.rpo} listing all of the
8088 template instantiations used in the corresponding object files which
8089 could be instantiated there; the link wrapper, @samp{collect2}, will
8090 then update the @samp{.rpo} files to tell the compiler where to place
8091 those instantiations and rebuild any affected object files. The
8092 link-time overhead is negligible after the first pass, as the compiler
8093 will continue to place the instantiations in the same files.
8095 This is your best option for application code written for the Borland
8096 model, as it will just work. Code written for the Cfront model will
8097 need to be modified so that the template definitions are available at
8098 one or more points of instantiation; usually this is as simple as adding
8099 @code{#include <tmethods.cc>} to the end of each template header.
8101 For library code, if you want the library to provide all of the template
8102 instantiations it needs, just try to link all of its object files
8103 together; the link will fail, but cause the instantiations to be
8104 generated as a side effect. Be warned, however, that this may cause
8105 conflicts if multiple libraries try to provide the same instantiations.
8106 For greater control, use explicit instantiation as described in the next
8110 @opindex fno-implicit-templates
8111 Compile your code with @option{-fno-implicit-templates} to disable the
8112 implicit generation of template instances, and explicitly instantiate
8113 all the ones you use. This approach requires more knowledge of exactly
8114 which instances you need than do the others, but it's less
8115 mysterious and allows greater control. You can scatter the explicit
8116 instantiations throughout your program, perhaps putting them in the
8117 translation units where the instances are used or the translation units
8118 that define the templates themselves; you can put all of the explicit
8119 instantiations you need into one big file; or you can create small files
8126 template class Foo<int>;
8127 template ostream& operator <<
8128 (ostream&, const Foo<int>&);
8131 for each of the instances you need, and create a template instantiation
8134 If you are using Cfront-model code, you can probably get away with not
8135 using @option{-fno-implicit-templates} when compiling files that don't
8136 @samp{#include} the member template definitions.
8138 If you use one big file to do the instantiations, you may want to
8139 compile it without @option{-fno-implicit-templates} so you get all of the
8140 instances required by your explicit instantiations (but not by any
8141 other files) without having to specify them as well.
8143 G++ has extended the template instantiation syntax given in the ISO
8144 standard to allow forward declaration of explicit instantiations
8145 (with @code{extern}), instantiation of the compiler support data for a
8146 template class (i.e.@: the vtable) without instantiating any of its
8147 members (with @code{inline}), and instantiation of only the static data
8148 members of a template class, without the support data or member
8149 functions (with (@code{static}):
8152 extern template int max (int, int);
8153 inline template class Foo<int>;
8154 static template class Foo<int>;
8158 Do nothing. Pretend G++ does implement automatic instantiation
8159 management. Code written for the Borland model will work fine, but
8160 each translation unit will contain instances of each of the templates it
8161 uses. In a large program, this can lead to an unacceptable amount of code
8164 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
8165 more discussion of these pragmas.
8168 @node Bound member functions
8169 @section Extracting the function pointer from a bound pointer to member function
8171 @cindex pointer to member function
8172 @cindex bound pointer to member function
8174 In C++, pointer to member functions (PMFs) are implemented using a wide
8175 pointer of sorts to handle all the possible call mechanisms; the PMF
8176 needs to store information about how to adjust the @samp{this} pointer,
8177 and if the function pointed to is virtual, where to find the vtable, and
8178 where in the vtable to look for the member function. If you are using
8179 PMFs in an inner loop, you should really reconsider that decision. If
8180 that is not an option, you can extract the pointer to the function that
8181 would be called for a given object/PMF pair and call it directly inside
8182 the inner loop, to save a bit of time.
8184 Note that you will still be paying the penalty for the call through a
8185 function pointer; on most modern architectures, such a call defeats the
8186 branch prediction features of the CPU@. This is also true of normal
8187 virtual function calls.
8189 The syntax for this extension is
8193 extern int (A::*fp)();
8194 typedef int (*fptr)(A *);
8196 fptr p = (fptr)(a.*fp);
8199 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
8200 no object is needed to obtain the address of the function. They can be
8201 converted to function pointers directly:
8204 fptr p1 = (fptr)(&A::foo);
8207 @opindex Wno-pmf-conversions
8208 You must specify @option{-Wno-pmf-conversions} to use this extension.
8210 @node C++ Attributes
8211 @section C++-Specific Variable, Function, and Type Attributes
8213 Some attributes only make sense for C++ programs.
8216 @item init_priority (@var{priority})
8217 @cindex init_priority attribute
8220 In Standard C++, objects defined at namespace scope are guaranteed to be
8221 initialized in an order in strict accordance with that of their definitions
8222 @emph{in a given translation unit}. No guarantee is made for initializations
8223 across translation units. However, GNU C++ allows users to control the
8224 order of initialization of objects defined at namespace scope with the
8225 @code{init_priority} attribute by specifying a relative @var{priority},
8226 a constant integral expression currently bounded between 101 and 65535
8227 inclusive. Lower numbers indicate a higher priority.
8229 In the following example, @code{A} would normally be created before
8230 @code{B}, but the @code{init_priority} attribute has reversed that order:
8233 Some_Class A __attribute__ ((init_priority (2000)));
8234 Some_Class B __attribute__ ((init_priority (543)));
8238 Note that the particular values of @var{priority} do not matter; only their
8241 @item java_interface
8242 @cindex java_interface attribute
8244 This type attribute informs C++ that the class is a Java interface. It may
8245 only be applied to classes declared within an @code{extern "Java"} block.
8246 Calls to methods declared in this interface will be dispatched using GCJ's
8247 interface table mechanism, instead of regular virtual table dispatch.
8251 See also @xref{Strong Using}.
8254 @section Strong Using
8256 @strong{Caution:} The semantics of this extension are not fully
8257 defined. Users should refrain from using this extension as its
8258 semantics may change subtly over time. It is possible that this
8259 extension wil be removed in future versions of G++.
8261 A using-directive with @code{__attribute ((strong))} is stronger
8262 than a normal using-directive in two ways:
8266 Templates from the used namespace can be specialized as though they were members of the using namespace.
8269 The using namespace is considered an associated namespace of all
8270 templates in the used namespace for purposes of argument-dependent
8274 This is useful for composing a namespace transparently from
8275 implementation namespaces. For example:
8280 template <class T> struct A @{ @};
8282 using namespace debug __attribute ((__strong__));
8283 template <> struct A<int> @{ @}; // ok to specialize
8285 template <class T> void f (A<T>);
8290 f (std::A<float>()); // lookup finds std::f
8295 @node Java Exceptions
8296 @section Java Exceptions
8298 The Java language uses a slightly different exception handling model
8299 from C++. Normally, GNU C++ will automatically detect when you are
8300 writing C++ code that uses Java exceptions, and handle them
8301 appropriately. However, if C++ code only needs to execute destructors
8302 when Java exceptions are thrown through it, GCC will guess incorrectly.
8303 Sample problematic code is:
8306 struct S @{ ~S(); @};
8307 extern void bar(); // is written in Java, and may throw exceptions
8316 The usual effect of an incorrect guess is a link failure, complaining of
8317 a missing routine called @samp{__gxx_personality_v0}.
8319 You can inform the compiler that Java exceptions are to be used in a
8320 translation unit, irrespective of what it might think, by writing
8321 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
8322 @samp{#pragma} must appear before any functions that throw or catch
8323 exceptions, or run destructors when exceptions are thrown through them.
8325 You cannot mix Java and C++ exceptions in the same translation unit. It
8326 is believed to be safe to throw a C++ exception from one file through
8327 another file compiled for the Java exception model, or vice versa, but
8328 there may be bugs in this area.
8330 @node Deprecated Features
8331 @section Deprecated Features
8333 In the past, the GNU C++ compiler was extended to experiment with new
8334 features, at a time when the C++ language was still evolving. Now that
8335 the C++ standard is complete, some of those features are superseded by
8336 superior alternatives. Using the old features might cause a warning in
8337 some cases that the feature will be dropped in the future. In other
8338 cases, the feature might be gone already.
8340 While the list below is not exhaustive, it documents some of the options
8341 that are now deprecated:
8344 @item -fexternal-templates
8345 @itemx -falt-external-templates
8346 These are two of the many ways for G++ to implement template
8347 instantiation. @xref{Template Instantiation}. The C++ standard clearly
8348 defines how template definitions have to be organized across
8349 implementation units. G++ has an implicit instantiation mechanism that
8350 should work just fine for standard-conforming code.
8352 @item -fstrict-prototype
8353 @itemx -fno-strict-prototype
8354 Previously it was possible to use an empty prototype parameter list to
8355 indicate an unspecified number of parameters (like C), rather than no
8356 parameters, as C++ demands. This feature has been removed, except where
8357 it is required for backwards compatibility @xref{Backwards Compatibility}.
8360 The named return value extension has been deprecated, and is now
8363 The use of initializer lists with new expressions has been deprecated,
8364 and is now removed from G++.
8366 Floating and complex non-type template parameters have been deprecated,
8367 and are now removed from G++.
8369 The implicit typename extension has been deprecated and is now
8372 The use of default arguments in function pointers, function typedefs and
8373 and other places where they are not permitted by the standard is
8374 deprecated and will be removed from a future version of G++.
8376 @node Backwards Compatibility
8377 @section Backwards Compatibility
8378 @cindex Backwards Compatibility
8379 @cindex ARM [Annotated C++ Reference Manual]
8381 Now that there is a definitive ISO standard C++, G++ has a specification
8382 to adhere to. The C++ language evolved over time, and features that
8383 used to be acceptable in previous drafts of the standard, such as the ARM
8384 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
8385 compilation of C++ written to such drafts, G++ contains some backwards
8386 compatibilities. @emph{All such backwards compatibility features are
8387 liable to disappear in future versions of G++.} They should be considered
8388 deprecated @xref{Deprecated Features}.
8392 If a variable is declared at for scope, it used to remain in scope until
8393 the end of the scope which contained the for statement (rather than just
8394 within the for scope). G++ retains this, but issues a warning, if such a
8395 variable is accessed outside the for scope.
8397 @item Implicit C language
8398 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
8399 scope to set the language. On such systems, all header files are
8400 implicitly scoped inside a C language scope. Also, an empty prototype
8401 @code{()} will be treated as an unspecified number of arguments, rather
8402 than no arguments, as C++ demands.