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 * Other Builtins:: Other built-in functions.
473 * Target Builtins:: Built-in functions specific to particular targets.
474 * Pragmas:: Pragmas accepted by GCC.
475 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
476 * Thread-Local:: Per-thread variables.
479 @node Statement Exprs
480 @section Statements and Declarations in Expressions
481 @cindex statements inside expressions
482 @cindex declarations inside expressions
483 @cindex expressions containing statements
484 @cindex macros, statements in expressions
486 @c the above section title wrapped and causes an underfull hbox.. i
487 @c changed it from "within" to "in". --mew 4feb93
488 A compound statement enclosed in parentheses may appear as an expression
489 in GNU C@. This allows you to use loops, switches, and local variables
490 within an expression.
492 Recall that a compound statement is a sequence of statements surrounded
493 by braces; in this construct, parentheses go around the braces. For
497 (@{ int y = foo (); int z;
504 is a valid (though slightly more complex than necessary) expression
505 for the absolute value of @code{foo ()}.
507 The last thing in the compound statement should be an expression
508 followed by a semicolon; the value of this subexpression serves as the
509 value of the entire construct. (If you use some other kind of statement
510 last within the braces, the construct has type @code{void}, and thus
511 effectively no value.)
513 This feature is especially useful in making macro definitions ``safe'' (so
514 that they evaluate each operand exactly once). For example, the
515 ``maximum'' function is commonly defined as a macro in standard C as
519 #define max(a,b) ((a) > (b) ? (a) : (b))
523 @cindex side effects, macro argument
524 But this definition computes either @var{a} or @var{b} twice, with bad
525 results if the operand has side effects. In GNU C, if you know the
526 type of the operands (here let's assume @code{int}), you can define
527 the macro safely as follows:
530 #define maxint(a,b) \
531 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
534 Embedded statements are not allowed in constant expressions, such as
535 the value of an enumeration constant, the width of a bit-field, or
536 the initial value of a static variable.
538 If you don't know the type of the operand, you can still do this, but you
539 must use @code{typeof} (@pxref{Typeof}).
541 In G++, the result value of a statement expression undergoes array and
542 function pointer decay, and is returned by value to the enclosing
543 expression. For instance, if @code{A} is a class, then
552 will construct a temporary @code{A} object to hold the result of the
553 statement expression, and that will be used to invoke @code{Foo}.
554 Therefore the @code{this} pointer observed by @code{Foo} will not be the
557 Any temporaries created within a statement within a statement expression
558 will be destroyed at the statement's end. This makes statement
559 expressions inside macros slightly different from function calls. In
560 the latter case temporaries introduced during argument evaluation will
561 be destroyed at the end of the statement that includes the function
562 call. In the statement expression case they will be destroyed during
563 the statement expression. For instance,
566 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
567 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
577 will have different places where temporaries are destroyed. For the
578 @code{macro} case, the temporary @code{X} will be destroyed just after
579 the initialization of @code{b}. In the @code{function} case that
580 temporary will be destroyed when the function returns.
582 These considerations mean that it is probably a bad idea to use
583 statement-expressions of this form in header files that are designed to
584 work with C++. (Note that some versions of the GNU C Library contained
585 header files using statement-expression that lead to precisely this
589 @section Locally Declared Labels
591 @cindex macros, local labels
593 GCC allows you to declare @dfn{local labels} in any nested block
594 scope. A local label is just like an ordinary label, but you can
595 only reference it (with a @code{goto} statement, or by taking its
596 address) within the block in which it was declared.
598 A local label declaration looks like this:
601 __label__ @var{label};
608 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
611 Local label declarations must come at the beginning of the block,
612 before any ordinary declarations or statements.
614 The label declaration defines the label @emph{name}, but does not define
615 the label itself. You must do this in the usual way, with
616 @code{@var{label}:}, within the statements of the statement expression.
618 The local label feature is useful for complex macros. If a macro
619 contains nested loops, a @code{goto} can be useful for breaking out of
620 them. However, an ordinary label whose scope is the whole function
621 cannot be used: if the macro can be expanded several times in one
622 function, the label will be multiply defined in that function. A
623 local label avoids this problem. For example:
626 #define SEARCH(value, array, target) \
629 typeof (target) _SEARCH_target = (target); \
630 typeof (*(array)) *_SEARCH_array = (array); \
633 for (i = 0; i < max; i++) \
634 for (j = 0; j < max; j++) \
635 if (_SEARCH_array[i][j] == _SEARCH_target) \
636 @{ (value) = i; goto found; @} \
642 This could also be written using a statement-expression:
645 #define SEARCH(array, target) \
648 typeof (target) _SEARCH_target = (target); \
649 typeof (*(array)) *_SEARCH_array = (array); \
652 for (i = 0; i < max; i++) \
653 for (j = 0; j < max; j++) \
654 if (_SEARCH_array[i][j] == _SEARCH_target) \
655 @{ value = i; goto found; @} \
662 Local label declarations also make the labels they declare visible to
663 nested functions, if there are any. @xref{Nested Functions}, for details.
665 @node Labels as Values
666 @section Labels as Values
667 @cindex labels as values
668 @cindex computed gotos
669 @cindex goto with computed label
670 @cindex address of a label
672 You can get the address of a label defined in the current function
673 (or a containing function) with the unary operator @samp{&&}. The
674 value has type @code{void *}. This value is a constant and can be used
675 wherever a constant of that type is valid. For example:
683 To use these values, you need to be able to jump to one. This is done
684 with the computed goto statement@footnote{The analogous feature in
685 Fortran is called an assigned goto, but that name seems inappropriate in
686 C, where one can do more than simply store label addresses in label
687 variables.}, @code{goto *@var{exp};}. For example,
694 Any expression of type @code{void *} is allowed.
696 One way of using these constants is in initializing a static array that
697 will serve as a jump table:
700 static void *array[] = @{ &&foo, &&bar, &&hack @};
703 Then you can select a label with indexing, like this:
710 Note that this does not check whether the subscript is in bounds---array
711 indexing in C never does that.
713 Such an array of label values serves a purpose much like that of the
714 @code{switch} statement. The @code{switch} statement is cleaner, so
715 use that rather than an array unless the problem does not fit a
716 @code{switch} statement very well.
718 Another use of label values is in an interpreter for threaded code.
719 The labels within the interpreter function can be stored in the
720 threaded code for super-fast dispatching.
722 You may not use this mechanism to jump to code in a different function.
723 If you do that, totally unpredictable things will happen. The best way to
724 avoid this is to store the label address only in automatic variables and
725 never pass it as an argument.
727 An alternate way to write the above example is
730 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
732 goto *(&&foo + array[i]);
736 This is more friendly to code living in shared libraries, as it reduces
737 the number of dynamic relocations that are needed, and by consequence,
738 allows the data to be read-only.
740 @node Nested Functions
741 @section Nested Functions
742 @cindex nested functions
743 @cindex downward funargs
746 A @dfn{nested function} is a function defined inside another function.
747 (Nested functions are not supported for GNU C++.) The nested function's
748 name is local to the block where it is defined. For example, here we
749 define a nested function named @code{square}, and call it twice:
753 foo (double a, double b)
755 double square (double z) @{ return z * z; @}
757 return square (a) + square (b);
762 The nested function can access all the variables of the containing
763 function that are visible at the point of its definition. This is
764 called @dfn{lexical scoping}. For example, here we show a nested
765 function which uses an inherited variable named @code{offset}:
769 bar (int *array, int offset, int size)
771 int access (int *array, int index)
772 @{ return array[index + offset]; @}
775 for (i = 0; i < size; i++)
776 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
781 Nested function definitions are permitted within functions in the places
782 where variable definitions are allowed; that is, in any block, before
783 the first statement in the block.
785 It is possible to call the nested function from outside the scope of its
786 name by storing its address or passing the address to another function:
789 hack (int *array, int size)
791 void store (int index, int value)
792 @{ array[index] = value; @}
794 intermediate (store, size);
798 Here, the function @code{intermediate} receives the address of
799 @code{store} as an argument. If @code{intermediate} calls @code{store},
800 the arguments given to @code{store} are used to store into @code{array}.
801 But this technique works only so long as the containing function
802 (@code{hack}, in this example) does not exit.
804 If you try to call the nested function through its address after the
805 containing function has exited, all hell will break loose. If you try
806 to call it after a containing scope level has exited, and if it refers
807 to some of the variables that are no longer in scope, you may be lucky,
808 but it's not wise to take the risk. If, however, the nested function
809 does not refer to anything that has gone out of scope, you should be
812 GCC implements taking the address of a nested function using a technique
813 called @dfn{trampolines}. A paper describing them is available as
816 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
818 A nested function can jump to a label inherited from a containing
819 function, provided the label was explicitly declared in the containing
820 function (@pxref{Local Labels}). Such a jump returns instantly to the
821 containing function, exiting the nested function which did the
822 @code{goto} and any intermediate functions as well. Here is an example:
826 bar (int *array, int offset, int size)
829 int access (int *array, int index)
833 return array[index + offset];
837 for (i = 0; i < size; i++)
838 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
842 /* @r{Control comes here from @code{access}
843 if it detects an error.} */
850 A nested function always has internal linkage. Declaring one with
851 @code{extern} is erroneous. If you need to declare the nested function
852 before its definition, use @code{auto} (which is otherwise meaningless
853 for function declarations).
856 bar (int *array, int offset, int size)
859 auto int access (int *, int);
861 int access (int *array, int index)
865 return array[index + offset];
871 @node Constructing Calls
872 @section Constructing Function Calls
873 @cindex constructing calls
874 @cindex forwarding calls
876 Using the built-in functions described below, you can record
877 the arguments a function received, and call another function
878 with the same arguments, without knowing the number or types
881 You can also record the return value of that function call,
882 and later return that value, without knowing what data type
883 the function tried to return (as long as your caller expects
886 However, these built-in functions may interact badly with some
887 sophisticated features or other extensions of the language. It
888 is, therefore, not recommended to use them outside very simple
889 functions acting as mere forwarders for their arguments.
891 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
892 This built-in function returns a pointer to data
893 describing how to perform a call with the same arguments as were passed
894 to the current function.
896 The function saves the arg pointer register, structure value address,
897 and all registers that might be used to pass arguments to a function
898 into a block of memory allocated on the stack. Then it returns the
899 address of that block.
902 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
903 This built-in function invokes @var{function}
904 with a copy of the parameters described by @var{arguments}
907 The value of @var{arguments} should be the value returned by
908 @code{__builtin_apply_args}. The argument @var{size} specifies the size
909 of the stack argument data, in bytes.
911 This function returns a pointer to data describing
912 how to return whatever value was returned by @var{function}. The data
913 is saved in a block of memory allocated on the stack.
915 It is not always simple to compute the proper value for @var{size}. The
916 value is used by @code{__builtin_apply} to compute the amount of data
917 that should be pushed on the stack and copied from the incoming argument
921 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
922 This built-in function returns the value described by @var{result} from
923 the containing function. You should specify, for @var{result}, a value
924 returned by @code{__builtin_apply}.
928 @section Referring to a Type with @code{typeof}
931 @cindex macros, types of arguments
933 Another way to refer to the type of an expression is with @code{typeof}.
934 The syntax of using of this keyword looks like @code{sizeof}, but the
935 construct acts semantically like a type name defined with @code{typedef}.
937 There are two ways of writing the argument to @code{typeof}: with an
938 expression or with a type. Here is an example with an expression:
945 This assumes that @code{x} is an array of pointers to functions;
946 the type described is that of the values of the functions.
948 Here is an example with a typename as the argument:
955 Here the type described is that of pointers to @code{int}.
957 If you are writing a header file that must work when included in ISO C
958 programs, write @code{__typeof__} instead of @code{typeof}.
959 @xref{Alternate Keywords}.
961 A @code{typeof}-construct can be used anywhere a typedef name could be
962 used. For example, you can use it in a declaration, in a cast, or inside
963 of @code{sizeof} or @code{typeof}.
965 @code{typeof} is often useful in conjunction with the
966 statements-within-expressions feature. Here is how the two together can
967 be used to define a safe ``maximum'' macro that operates on any
968 arithmetic type and evaluates each of its arguments exactly once:
972 (@{ typeof (a) _a = (a); \
973 typeof (b) _b = (b); \
974 _a > _b ? _a : _b; @})
977 @cindex underscores in variables in macros
978 @cindex @samp{_} in variables in macros
979 @cindex local variables in macros
980 @cindex variables, local, in macros
981 @cindex macros, local variables in
983 The reason for using names that start with underscores for the local
984 variables is to avoid conflicts with variable names that occur within the
985 expressions that are substituted for @code{a} and @code{b}. Eventually we
986 hope to design a new form of declaration syntax that allows you to declare
987 variables whose scopes start only after their initializers; this will be a
988 more reliable way to prevent such conflicts.
991 Some more examples of the use of @code{typeof}:
995 This declares @code{y} with the type of what @code{x} points to.
1002 This declares @code{y} as an array of such values.
1009 This declares @code{y} as an array of pointers to characters:
1012 typeof (typeof (char *)[4]) y;
1016 It is equivalent to the following traditional C declaration:
1022 To see the meaning of the declaration using @code{typeof}, and why it
1023 might be a useful way to write, let's rewrite it with these macros:
1026 #define pointer(T) typeof(T *)
1027 #define array(T, N) typeof(T [N])
1031 Now the declaration can be rewritten this way:
1034 array (pointer (char), 4) y;
1038 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1039 pointers to @code{char}.
1042 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1043 a more limited extension which permitted one to write
1046 typedef @var{T} = @var{expr};
1050 with the effect of declaring @var{T} to have the type of the expression
1051 @var{expr}. This extension does not work with GCC 3 (versions between
1052 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1053 relies on it should be rewritten to use @code{typeof}:
1056 typedef typeof(@var{expr}) @var{T};
1060 This will work with all versions of GCC@.
1063 @section Conditionals with Omitted Operands
1064 @cindex conditional expressions, extensions
1065 @cindex omitted middle-operands
1066 @cindex middle-operands, omitted
1067 @cindex extensions, @code{?:}
1068 @cindex @code{?:} extensions
1070 The middle operand in a conditional expression may be omitted. Then
1071 if the first operand is nonzero, its value is the value of the conditional
1074 Therefore, the expression
1081 has the value of @code{x} if that is nonzero; otherwise, the value of
1084 This example is perfectly equivalent to
1090 @cindex side effect in ?:
1091 @cindex ?: side effect
1093 In this simple case, the ability to omit the middle operand is not
1094 especially useful. When it becomes useful is when the first operand does,
1095 or may (if it is a macro argument), contain a side effect. Then repeating
1096 the operand in the middle would perform the side effect twice. Omitting
1097 the middle operand uses the value already computed without the undesirable
1098 effects of recomputing it.
1101 @section Double-Word Integers
1102 @cindex @code{long long} data types
1103 @cindex double-word arithmetic
1104 @cindex multiprecision arithmetic
1105 @cindex @code{LL} integer suffix
1106 @cindex @code{ULL} integer suffix
1108 ISO C99 supports data types for integers that are at least 64 bits wide,
1109 and as an extension GCC supports them in C89 mode and in C++.
1110 Simply write @code{long long int} for a signed integer, or
1111 @code{unsigned long long int} for an unsigned integer. To make an
1112 integer constant of type @code{long long int}, add the suffix @samp{LL}
1113 to the integer. To make an integer constant of type @code{unsigned long
1114 long int}, add the suffix @samp{ULL} to the integer.
1116 You can use these types in arithmetic like any other integer types.
1117 Addition, subtraction, and bitwise boolean operations on these types
1118 are open-coded on all types of machines. Multiplication is open-coded
1119 if the machine supports fullword-to-doubleword a widening multiply
1120 instruction. Division and shifts are open-coded only on machines that
1121 provide special support. The operations that are not open-coded use
1122 special library routines that come with GCC@.
1124 There may be pitfalls when you use @code{long long} types for function
1125 arguments, unless you declare function prototypes. If a function
1126 expects type @code{int} for its argument, and you pass a value of type
1127 @code{long long int}, confusion will result because the caller and the
1128 subroutine will disagree about the number of bytes for the argument.
1129 Likewise, if the function expects @code{long long int} and you pass
1130 @code{int}. The best way to avoid such problems is to use prototypes.
1133 @section Complex Numbers
1134 @cindex complex numbers
1135 @cindex @code{_Complex} keyword
1136 @cindex @code{__complex__} keyword
1138 ISO C99 supports complex floating data types, and as an extension GCC
1139 supports them in C89 mode and in C++, and supports complex integer data
1140 types which are not part of ISO C99. You can declare complex types
1141 using the keyword @code{_Complex}. As an extension, the older GNU
1142 keyword @code{__complex__} is also supported.
1144 For example, @samp{_Complex double x;} declares @code{x} as a
1145 variable whose real part and imaginary part are both of type
1146 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1147 have real and imaginary parts of type @code{short int}; this is not
1148 likely to be useful, but it shows that the set of complex types is
1151 To write a constant with a complex data type, use the suffix @samp{i} or
1152 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1153 has type @code{_Complex float} and @code{3i} has type
1154 @code{_Complex int}. Such a constant always has a pure imaginary
1155 value, but you can form any complex value you like by adding one to a
1156 real constant. This is a GNU extension; if you have an ISO C99
1157 conforming C library (such as GNU libc), and want to construct complex
1158 constants of floating type, you should include @code{<complex.h>} and
1159 use the macros @code{I} or @code{_Complex_I} instead.
1161 @cindex @code{__real__} keyword
1162 @cindex @code{__imag__} keyword
1163 To extract the real part of a complex-valued expression @var{exp}, write
1164 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1165 extract the imaginary part. This is a GNU extension; for values of
1166 floating type, you should use the ISO C99 functions @code{crealf},
1167 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1168 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1169 built-in functions by GCC@.
1171 @cindex complex conjugation
1172 The operator @samp{~} performs complex conjugation when used on a value
1173 with a complex type. This is a GNU extension; for values of
1174 floating type, you should use the ISO C99 functions @code{conjf},
1175 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1176 provided as built-in functions by GCC@.
1178 GCC can allocate complex automatic variables in a noncontiguous
1179 fashion; it's even possible for the real part to be in a register while
1180 the imaginary part is on the stack (or vice-versa). Only the DWARF2
1181 debug info format can represent this, so use of DWARF2 is recommended.
1182 If you are using the stabs debug info format, GCC describes a noncontiguous
1183 complex variable as if it were two separate variables of noncomplex type.
1184 If the variable's actual name is @code{foo}, the two fictitious
1185 variables are named @code{foo$real} and @code{foo$imag}. You can
1186 examine and set these two fictitious variables with your debugger.
1192 ISO C99 supports floating-point numbers written not only in the usual
1193 decimal notation, such as @code{1.55e1}, but also numbers such as
1194 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1195 supports this in C89 mode (except in some cases when strictly
1196 conforming) and in C++. In that format the
1197 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1198 mandatory. The exponent is a decimal number that indicates the power of
1199 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1206 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1207 is the same as @code{1.55e1}.
1209 Unlike for floating-point numbers in the decimal notation the exponent
1210 is always required in the hexadecimal notation. Otherwise the compiler
1211 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1212 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1213 extension for floating-point constants of type @code{float}.
1216 @section Arrays of Length Zero
1217 @cindex arrays of length zero
1218 @cindex zero-length arrays
1219 @cindex length-zero arrays
1220 @cindex flexible array members
1222 Zero-length arrays are allowed in GNU C@. They are very useful as the
1223 last element of a structure which is really a header for a variable-length
1232 struct line *thisline = (struct line *)
1233 malloc (sizeof (struct line) + this_length);
1234 thisline->length = this_length;
1237 In ISO C90, you would have to give @code{contents} a length of 1, which
1238 means either you waste space or complicate the argument to @code{malloc}.
1240 In ISO C99, you would use a @dfn{flexible array member}, which is
1241 slightly different in syntax and semantics:
1245 Flexible array members are written as @code{contents[]} without
1249 Flexible array members have incomplete type, and so the @code{sizeof}
1250 operator may not be applied. As a quirk of the original implementation
1251 of zero-length arrays, @code{sizeof} evaluates to zero.
1254 Flexible array members may only appear as the last member of a
1255 @code{struct} that is otherwise non-empty.
1258 A structure containing a flexible array member, or a union containing
1259 such a structure (possibly recursively), may not be a member of a
1260 structure or an element of an array. (However, these uses are
1261 permitted by GCC as extensions.)
1264 GCC versions before 3.0 allowed zero-length arrays to be statically
1265 initialized, as if they were flexible arrays. In addition to those
1266 cases that were useful, it also allowed initializations in situations
1267 that would corrupt later data. Non-empty initialization of zero-length
1268 arrays is now treated like any case where there are more initializer
1269 elements than the array holds, in that a suitable warning about "excess
1270 elements in array" is given, and the excess elements (all of them, in
1271 this case) are ignored.
1273 Instead GCC allows static initialization of flexible array members.
1274 This is equivalent to defining a new structure containing the original
1275 structure followed by an array of sufficient size to contain the data.
1276 I.e.@: in the following, @code{f1} is constructed as if it were declared
1282 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1285 struct f1 f1; int data[3];
1286 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1290 The convenience of this extension is that @code{f1} has the desired
1291 type, eliminating the need to consistently refer to @code{f2.f1}.
1293 This has symmetry with normal static arrays, in that an array of
1294 unknown size is also written with @code{[]}.
1296 Of course, this extension only makes sense if the extra data comes at
1297 the end of a top-level object, as otherwise we would be overwriting
1298 data at subsequent offsets. To avoid undue complication and confusion
1299 with initialization of deeply nested arrays, we simply disallow any
1300 non-empty initialization except when the structure is the top-level
1301 object. For example:
1304 struct foo @{ int x; int y[]; @};
1305 struct bar @{ struct foo z; @};
1307 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1308 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1309 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1310 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1313 @node Empty Structures
1314 @section Structures With No Members
1315 @cindex empty structures
1316 @cindex zero-size structures
1318 GCC permits a C structure to have no members:
1325 The structure will have size zero. In C++, empty structures are part
1326 of the language. G++ treats empty structures as if they had a single
1327 member of type @code{char}.
1329 @node Variable Length
1330 @section Arrays of Variable Length
1331 @cindex variable-length arrays
1332 @cindex arrays of variable length
1335 Variable-length automatic arrays are allowed in ISO C99, and as an
1336 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1337 implementation of variable-length arrays does not yet conform in detail
1338 to the ISO C99 standard.) These arrays are
1339 declared like any other automatic arrays, but with a length that is not
1340 a constant expression. The storage is allocated at the point of
1341 declaration and deallocated when the brace-level is exited. For
1346 concat_fopen (char *s1, char *s2, char *mode)
1348 char str[strlen (s1) + strlen (s2) + 1];
1351 return fopen (str, mode);
1355 @cindex scope of a variable length array
1356 @cindex variable-length array scope
1357 @cindex deallocating variable length arrays
1358 Jumping or breaking out of the scope of the array name deallocates the
1359 storage. Jumping into the scope is not allowed; you get an error
1362 @cindex @code{alloca} vs variable-length arrays
1363 You can use the function @code{alloca} to get an effect much like
1364 variable-length arrays. The function @code{alloca} is available in
1365 many other C implementations (but not in all). On the other hand,
1366 variable-length arrays are more elegant.
1368 There are other differences between these two methods. Space allocated
1369 with @code{alloca} exists until the containing @emph{function} returns.
1370 The space for a variable-length array is deallocated as soon as the array
1371 name's scope ends. (If you use both variable-length arrays and
1372 @code{alloca} in the same function, deallocation of a variable-length array
1373 will also deallocate anything more recently allocated with @code{alloca}.)
1375 You can also use variable-length arrays as arguments to functions:
1379 tester (int len, char data[len][len])
1385 The length of an array is computed once when the storage is allocated
1386 and is remembered for the scope of the array in case you access it with
1389 If you want to pass the array first and the length afterward, you can
1390 use a forward declaration in the parameter list---another GNU extension.
1394 tester (int len; char data[len][len], int len)
1400 @cindex parameter forward declaration
1401 The @samp{int len} before the semicolon is a @dfn{parameter forward
1402 declaration}, and it serves the purpose of making the name @code{len}
1403 known when the declaration of @code{data} is parsed.
1405 You can write any number of such parameter forward declarations in the
1406 parameter list. They can be separated by commas or semicolons, but the
1407 last one must end with a semicolon, which is followed by the ``real''
1408 parameter declarations. Each forward declaration must match a ``real''
1409 declaration in parameter name and data type. ISO C99 does not support
1410 parameter forward declarations.
1412 @node Variadic Macros
1413 @section Macros with a Variable Number of Arguments.
1414 @cindex variable number of arguments
1415 @cindex macro with variable arguments
1416 @cindex rest argument (in macro)
1417 @cindex variadic macros
1419 In the ISO C standard of 1999, a macro can be declared to accept a
1420 variable number of arguments much as a function can. The syntax for
1421 defining the macro is similar to that of a function. Here is an
1425 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1428 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1429 such a macro, it represents the zero or more tokens until the closing
1430 parenthesis that ends the invocation, including any commas. This set of
1431 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1432 wherever it appears. See the CPP manual for more information.
1434 GCC has long supported variadic macros, and used a different syntax that
1435 allowed you to give a name to the variable arguments just like any other
1436 argument. Here is an example:
1439 #define debug(format, args...) fprintf (stderr, format, args)
1442 This is in all ways equivalent to the ISO C example above, but arguably
1443 more readable and descriptive.
1445 GNU CPP has two further variadic macro extensions, and permits them to
1446 be used with either of the above forms of macro definition.
1448 In standard C, you are not allowed to leave the variable argument out
1449 entirely; but you are allowed to pass an empty argument. For example,
1450 this invocation is invalid in ISO C, because there is no comma after
1457 GNU CPP permits you to completely omit the variable arguments in this
1458 way. In the above examples, the compiler would complain, though since
1459 the expansion of the macro still has the extra comma after the format
1462 To help solve this problem, CPP behaves specially for variable arguments
1463 used with the token paste operator, @samp{##}. If instead you write
1466 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1469 and if the variable arguments are omitted or empty, the @samp{##}
1470 operator causes the preprocessor to remove the comma before it. If you
1471 do provide some variable arguments in your macro invocation, GNU CPP
1472 does not complain about the paste operation and instead places the
1473 variable arguments after the comma. Just like any other pasted macro
1474 argument, these arguments are not macro expanded.
1476 @node Escaped Newlines
1477 @section Slightly Looser Rules for Escaped Newlines
1478 @cindex escaped newlines
1479 @cindex newlines (escaped)
1481 Recently, the preprocessor has relaxed its treatment of escaped
1482 newlines. Previously, the newline had to immediately follow a
1483 backslash. The current implementation allows whitespace in the form
1484 of spaces, horizontal and vertical tabs, and form feeds between the
1485 backslash and the subsequent newline. The preprocessor issues a
1486 warning, but treats it as a valid escaped newline and combines the two
1487 lines to form a single logical line. This works within comments and
1488 tokens, as well as between tokens. Comments are @emph{not} treated as
1489 whitespace for the purposes of this relaxation, since they have not
1490 yet been replaced with spaces.
1493 @section Non-Lvalue Arrays May Have Subscripts
1494 @cindex subscripting
1495 @cindex arrays, non-lvalue
1497 @cindex subscripting and function values
1498 In ISO C99, arrays that are not lvalues still decay to pointers, and
1499 may be subscripted, although they may not be modified or used after
1500 the next sequence point and the unary @samp{&} operator may not be
1501 applied to them. As an extension, GCC allows such arrays to be
1502 subscripted in C89 mode, though otherwise they do not decay to
1503 pointers outside C99 mode. For example,
1504 this is valid in GNU C though not valid in C89:
1508 struct foo @{int a[4];@};
1514 return f().a[index];
1520 @section Arithmetic on @code{void}- and Function-Pointers
1521 @cindex void pointers, arithmetic
1522 @cindex void, size of pointer to
1523 @cindex function pointers, arithmetic
1524 @cindex function, size of pointer to
1526 In GNU C, addition and subtraction operations are supported on pointers to
1527 @code{void} and on pointers to functions. This is done by treating the
1528 size of a @code{void} or of a function as 1.
1530 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1531 and on function types, and returns 1.
1533 @opindex Wpointer-arith
1534 The option @option{-Wpointer-arith} requests a warning if these extensions
1538 @section Non-Constant Initializers
1539 @cindex initializers, non-constant
1540 @cindex non-constant initializers
1542 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1543 automatic variable are not required to be constant expressions in GNU C@.
1544 Here is an example of an initializer with run-time varying elements:
1547 foo (float f, float g)
1549 float beat_freqs[2] = @{ f-g, f+g @};
1554 @node Compound Literals
1555 @section Compound Literals
1556 @cindex constructor expressions
1557 @cindex initializations in expressions
1558 @cindex structures, constructor expression
1559 @cindex expressions, constructor
1560 @cindex compound literals
1561 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1563 ISO C99 supports compound literals. A compound literal looks like
1564 a cast containing an initializer. Its value is an object of the
1565 type specified in the cast, containing the elements specified in
1566 the initializer; it is an lvalue. As an extension, GCC supports
1567 compound literals in C89 mode and in C++.
1569 Usually, the specified type is a structure. Assume that
1570 @code{struct foo} and @code{structure} are declared as shown:
1573 struct foo @{int a; char b[2];@} structure;
1577 Here is an example of constructing a @code{struct foo} with a compound literal:
1580 structure = ((struct foo) @{x + y, 'a', 0@});
1584 This is equivalent to writing the following:
1588 struct foo temp = @{x + y, 'a', 0@};
1593 You can also construct an array. If all the elements of the compound literal
1594 are (made up of) simple constant expressions, suitable for use in
1595 initializers of objects of static storage duration, then the compound
1596 literal can be coerced to a pointer to its first element and used in
1597 such an initializer, as shown here:
1600 char **foo = (char *[]) @{ "x", "y", "z" @};
1603 Compound literals for scalar types and union types are is
1604 also allowed, but then the compound literal is equivalent
1607 As a GNU extension, GCC allows initialization of objects with static storage
1608 duration by compound literals (which is not possible in ISO C99, because
1609 the initializer is not a constant).
1610 It is handled as if the object was initialized only with the bracket
1611 enclosed list if compound literal's and object types match.
1612 The initializer list of the compound literal must be constant.
1613 If the object being initialized has array type of unknown size, the size is
1614 determined by compound literal size.
1617 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1618 static int y[] = (int []) @{1, 2, 3@};
1619 static int z[] = (int [3]) @{1@};
1623 The above lines are equivalent to the following:
1625 static struct foo x = @{1, 'a', 'b'@};
1626 static int y[] = @{1, 2, 3@};
1627 static int z[] = @{1, 0, 0@};
1630 @node Designated Inits
1631 @section Designated Initializers
1632 @cindex initializers with labeled elements
1633 @cindex labeled elements in initializers
1634 @cindex case labels in initializers
1635 @cindex designated initializers
1637 Standard C89 requires the elements of an initializer to appear in a fixed
1638 order, the same as the order of the elements in the array or structure
1641 In ISO C99 you can give the elements in any order, specifying the array
1642 indices or structure field names they apply to, and GNU C allows this as
1643 an extension in C89 mode as well. This extension is not
1644 implemented in GNU C++.
1646 To specify an array index, write
1647 @samp{[@var{index}] =} before the element value. For example,
1650 int a[6] = @{ [4] = 29, [2] = 15 @};
1657 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1661 The index values must be constant expressions, even if the array being
1662 initialized is automatic.
1664 An alternative syntax for this which has been obsolete since GCC 2.5 but
1665 GCC still accepts is to write @samp{[@var{index}]} before the element
1666 value, with no @samp{=}.
1668 To initialize a range of elements to the same value, write
1669 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1670 extension. For example,
1673 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1677 If the value in it has side-effects, the side-effects will happen only once,
1678 not for each initialized field by the range initializer.
1681 Note that the length of the array is the highest value specified
1684 In a structure initializer, specify the name of a field to initialize
1685 with @samp{.@var{fieldname} =} before the element value. For example,
1686 given the following structure,
1689 struct point @{ int x, y; @};
1693 the following initialization
1696 struct point p = @{ .y = yvalue, .x = xvalue @};
1703 struct point p = @{ xvalue, yvalue @};
1706 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1707 @samp{@var{fieldname}:}, as shown here:
1710 struct point p = @{ y: yvalue, x: xvalue @};
1714 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1715 @dfn{designator}. You can also use a designator (or the obsolete colon
1716 syntax) when initializing a union, to specify which element of the union
1717 should be used. For example,
1720 union foo @{ int i; double d; @};
1722 union foo f = @{ .d = 4 @};
1726 will convert 4 to a @code{double} to store it in the union using
1727 the second element. By contrast, casting 4 to type @code{union foo}
1728 would store it into the union as the integer @code{i}, since it is
1729 an integer. (@xref{Cast to Union}.)
1731 You can combine this technique of naming elements with ordinary C
1732 initialization of successive elements. Each initializer element that
1733 does not have a designator applies to the next consecutive element of the
1734 array or structure. For example,
1737 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1744 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1747 Labeling the elements of an array initializer is especially useful
1748 when the indices are characters or belong to an @code{enum} type.
1753 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1754 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1757 @cindex designator lists
1758 You can also write a series of @samp{.@var{fieldname}} and
1759 @samp{[@var{index}]} designators before an @samp{=} to specify a
1760 nested subobject to initialize; the list is taken relative to the
1761 subobject corresponding to the closest surrounding brace pair. For
1762 example, with the @samp{struct point} declaration above:
1765 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1769 If the same field is initialized multiple times, it will have value from
1770 the last initialization. If any such overridden initialization has
1771 side-effect, it is unspecified whether the side-effect happens or not.
1772 Currently, GCC will discard them and issue a warning.
1775 @section Case Ranges
1777 @cindex ranges in case statements
1779 You can specify a range of consecutive values in a single @code{case} label,
1783 case @var{low} ... @var{high}:
1787 This has the same effect as the proper number of individual @code{case}
1788 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1790 This feature is especially useful for ranges of ASCII character codes:
1796 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1797 it may be parsed wrong when you use it with integer values. For example,
1812 @section Cast to a Union Type
1813 @cindex cast to a union
1814 @cindex union, casting to a
1816 A cast to union type is similar to other casts, except that the type
1817 specified is a union type. You can specify the type either with
1818 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1819 a constructor though, not a cast, and hence does not yield an lvalue like
1820 normal casts. (@xref{Compound Literals}.)
1822 The types that may be cast to the union type are those of the members
1823 of the union. Thus, given the following union and variables:
1826 union foo @{ int i; double d; @};
1832 both @code{x} and @code{y} can be cast to type @code{union foo}.
1834 Using the cast as the right-hand side of an assignment to a variable of
1835 union type is equivalent to storing in a member of the union:
1840 u = (union foo) x @equiv{} u.i = x
1841 u = (union foo) y @equiv{} u.d = y
1844 You can also use the union cast as a function argument:
1847 void hack (union foo);
1849 hack ((union foo) x);
1852 @node Mixed Declarations
1853 @section Mixed Declarations and Code
1854 @cindex mixed declarations and code
1855 @cindex declarations, mixed with code
1856 @cindex code, mixed with declarations
1858 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1859 within compound statements. As an extension, GCC also allows this in
1860 C89 mode. For example, you could do:
1869 Each identifier is visible from where it is declared until the end of
1870 the enclosing block.
1872 @node Function Attributes
1873 @section Declaring Attributes of Functions
1874 @cindex function attributes
1875 @cindex declaring attributes of functions
1876 @cindex functions that never return
1877 @cindex functions that have no side effects
1878 @cindex functions in arbitrary sections
1879 @cindex functions that behave like malloc
1880 @cindex @code{volatile} applied to function
1881 @cindex @code{const} applied to function
1882 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1883 @cindex functions with non-null pointer arguments
1884 @cindex functions that are passed arguments in registers on the 386
1885 @cindex functions that pop the argument stack on the 386
1886 @cindex functions that do not pop the argument stack on the 386
1888 In GNU C, you declare certain things about functions called in your program
1889 which help the compiler optimize function calls and check your code more
1892 The keyword @code{__attribute__} allows you to specify special
1893 attributes when making a declaration. This keyword is followed by an
1894 attribute specification inside double parentheses. The following
1895 attributes are currently defined for functions on all targets:
1896 @code{noreturn}, @code{noinline}, @code{always_inline},
1897 @code{pure}, @code{const}, @code{nothrow},
1898 @code{format}, @code{format_arg}, @code{no_instrument_function},
1899 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1900 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1901 @code{alias}, @code{warn_unused_result} and @code{nonnull}. Several other
1902 attributes are defined for functions on particular target systems. Other
1903 attributes, including @code{section} are supported for variables declarations
1904 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1906 You may also specify attributes with @samp{__} preceding and following
1907 each keyword. This allows you to use them in header files without
1908 being concerned about a possible macro of the same name. For example,
1909 you may use @code{__noreturn__} instead of @code{noreturn}.
1911 @xref{Attribute Syntax}, for details of the exact syntax for using
1915 @cindex @code{noreturn} function attribute
1917 A few standard library functions, such as @code{abort} and @code{exit},
1918 cannot return. GCC knows this automatically. Some programs define
1919 their own functions that never return. You can declare them
1920 @code{noreturn} to tell the compiler this fact. For example,
1924 void fatal () __attribute__ ((noreturn));
1927 fatal (/* @r{@dots{}} */)
1929 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
1935 The @code{noreturn} keyword tells the compiler to assume that
1936 @code{fatal} cannot return. It can then optimize without regard to what
1937 would happen if @code{fatal} ever did return. This makes slightly
1938 better code. More importantly, it helps avoid spurious warnings of
1939 uninitialized variables.
1941 The @code{noreturn} keyword does not affect the exceptional path when that
1942 applies: a @code{noreturn}-marked function may still return to the caller
1943 by throwing an exception.
1945 Do not assume that registers saved by the calling function are
1946 restored before calling the @code{noreturn} function.
1948 It does not make sense for a @code{noreturn} function to have a return
1949 type other than @code{void}.
1951 The attribute @code{noreturn} is not implemented in GCC versions
1952 earlier than 2.5. An alternative way to declare that a function does
1953 not return, which works in the current version and in some older
1954 versions, is as follows:
1957 typedef void voidfn ();
1959 volatile voidfn fatal;
1962 @cindex @code{noinline} function attribute
1964 This function attribute prevents a function from being considered for
1967 @cindex @code{always_inline} function attribute
1969 Generally, functions are not inlined unless optimization is specified.
1970 For functions declared inline, this attribute inlines the function even
1971 if no optimization level was specified.
1973 @cindex @code{pure} function attribute
1975 Many functions have no effects except the return value and their
1976 return value depends only on the parameters and/or global variables.
1977 Such a function can be subject
1978 to common subexpression elimination and loop optimization just as an
1979 arithmetic operator would be. These functions should be declared
1980 with the attribute @code{pure}. For example,
1983 int square (int) __attribute__ ((pure));
1987 says that the hypothetical function @code{square} is safe to call
1988 fewer times than the program says.
1990 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
1991 Interesting non-pure functions are functions with infinite loops or those
1992 depending on volatile memory or other system resource, that may change between
1993 two consecutive calls (such as @code{feof} in a multithreading environment).
1995 The attribute @code{pure} is not implemented in GCC versions earlier
1997 @cindex @code{const} function attribute
1999 Many functions do not examine any values except their arguments, and
2000 have no effects except the return value. Basically this is just slightly
2001 more strict class than the @code{pure} attribute above, since function is not
2002 allowed to read global memory.
2004 @cindex pointer arguments
2005 Note that a function that has pointer arguments and examines the data
2006 pointed to must @emph{not} be declared @code{const}. Likewise, a
2007 function that calls a non-@code{const} function usually must not be
2008 @code{const}. It does not make sense for a @code{const} function to
2011 The attribute @code{const} is not implemented in GCC versions earlier
2012 than 2.5. An alternative way to declare that a function has no side
2013 effects, which works in the current version and in some older versions,
2017 typedef int intfn ();
2019 extern const intfn square;
2022 This approach does not work in GNU C++ from 2.6.0 on, since the language
2023 specifies that the @samp{const} must be attached to the return value.
2025 @cindex @code{nothrow} function attribute
2027 The @code{nothrow} attribute is used to inform the compiler that a
2028 function cannot throw an exception. For example, most functions in
2029 the standard C library can be guaranteed not to throw an exception
2030 with the notable exceptions of @code{qsort} and @code{bsearch} that
2031 take function pointer arguments. The @code{nothrow} attribute is not
2032 implemented in GCC versions earlier than 3.2.
2034 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2035 @cindex @code{format} function attribute
2037 The @code{format} attribute specifies that a function takes @code{printf},
2038 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2039 should be type-checked against a format string. For example, the
2044 my_printf (void *my_object, const char *my_format, ...)
2045 __attribute__ ((format (printf, 2, 3)));
2049 causes the compiler to check the arguments in calls to @code{my_printf}
2050 for consistency with the @code{printf} style format string argument
2053 The parameter @var{archetype} determines how the format string is
2054 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2055 or @code{strfmon}. (You can also use @code{__printf__},
2056 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2057 parameter @var{string-index} specifies which argument is the format
2058 string argument (starting from 1), while @var{first-to-check} is the
2059 number of the first argument to check against the format string. For
2060 functions where the arguments are not available to be checked (such as
2061 @code{vprintf}), specify the third parameter as zero. In this case the
2062 compiler only checks the format string for consistency. For
2063 @code{strftime} formats, the third parameter is required to be zero.
2064 Since non-static C++ methods have an implicit @code{this} argument, the
2065 arguments of such methods should be counted from two, not one, when
2066 giving values for @var{string-index} and @var{first-to-check}.
2068 In the example above, the format string (@code{my_format}) is the second
2069 argument of the function @code{my_print}, and the arguments to check
2070 start with the third argument, so the correct parameters for the format
2071 attribute are 2 and 3.
2073 @opindex ffreestanding
2074 The @code{format} attribute allows you to identify your own functions
2075 which take format strings as arguments, so that GCC can check the
2076 calls to these functions for errors. The compiler always (unless
2077 @option{-ffreestanding} is used) checks formats
2078 for the standard library functions @code{printf}, @code{fprintf},
2079 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2080 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2081 warnings are requested (using @option{-Wformat}), so there is no need to
2082 modify the header file @file{stdio.h}. In C99 mode, the functions
2083 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2084 @code{vsscanf} are also checked. Except in strictly conforming C
2085 standard modes, the X/Open function @code{strfmon} is also checked as
2086 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2087 @xref{C Dialect Options,,Options Controlling C Dialect}.
2089 @item format_arg (@var{string-index})
2090 @cindex @code{format_arg} function attribute
2091 @opindex Wformat-nonliteral
2092 The @code{format_arg} attribute specifies that a function takes a format
2093 string for a @code{printf}, @code{scanf}, @code{strftime} or
2094 @code{strfmon} style function and modifies it (for example, to translate
2095 it into another language), so the result can be passed to a
2096 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2097 function (with the remaining arguments to the format function the same
2098 as they would have been for the unmodified string). For example, the
2103 my_dgettext (char *my_domain, const char *my_format)
2104 __attribute__ ((format_arg (2)));
2108 causes the compiler to check the arguments in calls to a @code{printf},
2109 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2110 format string argument is a call to the @code{my_dgettext} function, for
2111 consistency with the format string argument @code{my_format}. If the
2112 @code{format_arg} attribute had not been specified, all the compiler
2113 could tell in such calls to format functions would be that the format
2114 string argument is not constant; this would generate a warning when
2115 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2116 without the attribute.
2118 The parameter @var{string-index} specifies which argument is the format
2119 string argument (starting from one). Since non-static C++ methods have
2120 an implicit @code{this} argument, the arguments of such methods should
2121 be counted from two.
2123 The @code{format-arg} attribute allows you to identify your own
2124 functions which modify format strings, so that GCC can check the
2125 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2126 type function whose operands are a call to one of your own function.
2127 The compiler always treats @code{gettext}, @code{dgettext}, and
2128 @code{dcgettext} in this manner except when strict ISO C support is
2129 requested by @option{-ansi} or an appropriate @option{-std} option, or
2130 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2131 Controlling C Dialect}.
2133 @item nonnull (@var{arg-index}, @dots{})
2134 @cindex @code{nonnull} function attribute
2135 The @code{nonnull} attribute specifies that some function parameters should
2136 be non-null pointers. For instance, the declaration:
2140 my_memcpy (void *dest, const void *src, size_t len)
2141 __attribute__((nonnull (1, 2)));
2145 causes the compiler to check that, in calls to @code{my_memcpy},
2146 arguments @var{dest} and @var{src} are non-null. If the compiler
2147 determines that a null pointer is passed in an argument slot marked
2148 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2149 is issued. The compiler may also choose to make optimizations based
2150 on the knowledge that certain function arguments will not be null.
2152 If no argument index list is given to the @code{nonnull} attribute,
2153 all pointer arguments are marked as non-null. To illustrate, the
2154 following declaration is equivalent to the previous example:
2158 my_memcpy (void *dest, const void *src, size_t len)
2159 __attribute__((nonnull));
2162 @item no_instrument_function
2163 @cindex @code{no_instrument_function} function attribute
2164 @opindex finstrument-functions
2165 If @option{-finstrument-functions} is given, profiling function calls will
2166 be generated at entry and exit of most user-compiled functions.
2167 Functions with this attribute will not be so instrumented.
2169 @item section ("@var{section-name}")
2170 @cindex @code{section} function attribute
2171 Normally, the compiler places the code it generates in the @code{text} section.
2172 Sometimes, however, you need additional sections, or you need certain
2173 particular functions to appear in special sections. The @code{section}
2174 attribute specifies that a function lives in a particular section.
2175 For example, the declaration:
2178 extern void foobar (void) __attribute__ ((section ("bar")));
2182 puts the function @code{foobar} in the @code{bar} section.
2184 Some file formats do not support arbitrary sections so the @code{section}
2185 attribute is not available on all platforms.
2186 If you need to map the entire contents of a module to a particular
2187 section, consider using the facilities of the linker instead.
2191 @cindex @code{constructor} function attribute
2192 @cindex @code{destructor} function attribute
2193 The @code{constructor} attribute causes the function to be called
2194 automatically before execution enters @code{main ()}. Similarly, the
2195 @code{destructor} attribute causes the function to be called
2196 automatically after @code{main ()} has completed or @code{exit ()} has
2197 been called. Functions with these attributes are useful for
2198 initializing data that will be used implicitly during the execution of
2201 These attributes are not currently implemented for Objective-C@.
2203 @cindex @code{unused} attribute.
2205 This attribute, attached to a function, means that the function is meant
2206 to be possibly unused. GCC will not produce a warning for this
2209 @cindex @code{used} attribute.
2211 This attribute, attached to a function, means that code must be emitted
2212 for the function even if it appears that the function is not referenced.
2213 This is useful, for example, when the function is referenced only in
2216 @cindex @code{deprecated} attribute.
2218 The @code{deprecated} attribute results in a warning if the function
2219 is used anywhere in the source file. This is useful when identifying
2220 functions that are expected to be removed in a future version of a
2221 program. The warning also includes the location of the declaration
2222 of the deprecated function, to enable users to easily find further
2223 information about why the function is deprecated, or what they should
2224 do instead. Note that the warnings only occurs for uses:
2227 int old_fn () __attribute__ ((deprecated));
2229 int (*fn_ptr)() = old_fn;
2232 results in a warning on line 3 but not line 2.
2234 The @code{deprecated} attribute can also be used for variables and
2235 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2237 @item warn_unused_result
2238 @cindex @code{warn_unused_result} attribute
2239 The @code{warn_unused_result} attribute causes a warning to be emitted
2240 if a caller of the function with this attribute does not use its
2241 return value. This is useful for functions where not checking
2242 the result is either a security problem or always a bug, such as
2246 int fn () __attribute__ ((warn_unused_result));
2249 if (fn () < 0) return -1;
2255 results in warning on line 5.
2258 @cindex @code{weak} attribute
2259 The @code{weak} attribute causes the declaration to be emitted as a weak
2260 symbol rather than a global. This is primarily useful in defining
2261 library functions which can be overridden in user code, though it can
2262 also be used with non-function declarations. Weak symbols are supported
2263 for ELF targets, and also for a.out targets when using the GNU assembler
2267 @cindex @code{malloc} attribute
2268 The @code{malloc} attribute is used to tell the compiler that a function
2269 may be treated as if any non-@code{NULL} pointer it returns cannot
2270 alias any other pointer valid when the function returns.
2271 This will often improve optimization.
2272 Standard functions with this property include @code{malloc} and
2273 @code{calloc}. @code{realloc}-like functions have this property as
2274 long as the old pointer is never referred to (including comparing it
2275 to the new pointer) after the function returns a non-@code{NULL}
2278 @item alias ("@var{target}")
2279 @cindex @code{alias} attribute
2280 The @code{alias} attribute causes the declaration to be emitted as an
2281 alias for another symbol, which must be specified. For instance,
2284 void __f () @{ /* @r{Do something.} */; @}
2285 void f () __attribute__ ((weak, alias ("__f")));
2288 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2289 mangled name for the target must be used.
2291 Not all target machines support this attribute.
2293 @item visibility ("@var{visibility_type}")
2294 @cindex @code{visibility} attribute
2295 The @code{visibility} attribute on ELF targets causes the declaration
2296 to be emitted with default, hidden, protected or internal visibility.
2299 void __attribute__ ((visibility ("protected")))
2300 f () @{ /* @r{Do something.} */; @}
2301 int i __attribute__ ((visibility ("hidden")));
2304 See the ELF gABI for complete details, but the short story is:
2308 Default visibility is the normal case for ELF. This value is
2309 available for the visibility attribute to override other options
2310 that may change the assumed visibility of symbols.
2313 Hidden visibility indicates that the symbol will not be placed into
2314 the dynamic symbol table, so no other @dfn{module} (executable or
2315 shared library) can reference it directly.
2318 Protected visibility indicates that the symbol will be placed in the
2319 dynamic symbol table, but that references within the defining module
2320 will bind to the local symbol. That is, the symbol cannot be overridden
2324 Internal visibility is like hidden visibility, but with additional
2325 processor specific semantics. Unless otherwise specified by the psABI,
2326 GCC defines internal visibility to mean that the function is @emph{never}
2327 called from another module. Note that hidden symbols, while they cannot
2328 be referenced directly by other modules, can be referenced indirectly via
2329 function pointers. By indicating that a symbol cannot be called from
2330 outside the module, GCC may for instance omit the load of a PIC register
2331 since it is known that the calling function loaded the correct value.
2334 Not all ELF targets support this attribute.
2336 @item regparm (@var{number})
2337 @cindex @code{regparm} attribute
2338 @cindex functions that are passed arguments in registers on the 386
2339 On the Intel 386, the @code{regparm} attribute causes the compiler to
2340 pass up to @var{number} integer arguments in registers EAX,
2341 EDX, and ECX instead of on the stack. Functions that take a
2342 variable number of arguments will continue to be passed all of their
2343 arguments on the stack.
2345 Beware that on some ELF systems this attribute is unsuitable for
2346 global functions in shared libraries with lazy binding (which is the
2347 default). Lazy binding will send the first call via resolving code in
2348 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2349 per the standard calling conventions. Solaris 8 is affected by this.
2350 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2351 safe since the loaders there save all registers. (Lazy binding can be
2352 disabled with the linker or the loader if desired, to avoid the
2356 @cindex functions that pop the argument stack on the 386
2357 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2358 assume that the called function will pop off the stack space used to
2359 pass arguments, unless it takes a variable number of arguments.
2362 @cindex functions that pop the argument stack on the 386
2363 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2364 pass the first two arguments in the registers ECX and EDX. Subsequent
2365 arguments are passed on the stack. The called function will pop the
2366 arguments off the stack. If the number of arguments is variable all
2367 arguments are pushed on the stack.
2370 @cindex functions that do pop the argument stack on the 386
2372 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2373 assume that the calling function will pop off the stack space used to
2374 pass arguments. This is
2375 useful to override the effects of the @option{-mrtd} switch.
2377 @item longcall/shortcall
2378 @cindex functions called via pointer on the RS/6000 and PowerPC
2379 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2380 compiler to always call this function via a pointer, just as it would if
2381 the @option{-mlongcall} option had been specified. The @code{shortcall}
2382 attribute causes the compiler not to do this. These attributes override
2383 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2386 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2387 calls are necessary.
2389 @item long_call/short_call
2390 @cindex indirect calls on ARM
2391 This attribute specifies how a particular function is called on
2392 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2393 command line switch and @code{#pragma long_calls} settings. The
2394 @code{long_call} attribute causes the compiler to always call the
2395 function by first loading its address into a register and then using the
2396 contents of that register. The @code{short_call} attribute always places
2397 the offset to the function from the call site into the @samp{BL}
2398 instruction directly.
2400 @item function_vector
2401 @cindex calling functions through the function vector on the H8/300 processors
2402 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2403 function should be called through the function vector. Calling a
2404 function through the function vector will reduce code size, however;
2405 the function vector has a limited size (maximum 128 entries on the H8/300
2406 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2408 You must use GAS and GLD from GNU binutils version 2.7 or later for
2409 this attribute to work correctly.
2412 @cindex interrupt handler functions
2413 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2414 that the specified function is an interrupt handler. The compiler will
2415 generate function entry and exit sequences suitable for use in an
2416 interrupt handler when this attribute is present.
2418 Note, interrupt handlers for the m68k, H8/300, H8/300H, H8S, and SH processors
2419 can be specified via the @code{interrupt_handler} attribute.
2421 Note, on the AVR, interrupts will be enabled inside the function.
2423 Note, for the ARM, you can specify the kind of interrupt to be handled by
2424 adding an optional parameter to the interrupt attribute like this:
2427 void f () __attribute__ ((interrupt ("IRQ")));
2430 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2432 @item interrupt_handler
2433 @cindex interrupt handler functions on the m68k, H8/300 and SH processors
2434 Use this attribute on the m68k, H8/300, H8/300H, H8S, and SH to indicate that
2435 the specified function is an interrupt handler. The compiler will generate
2436 function entry and exit sequences suitable for use in an interrupt
2437 handler when this attribute is present.
2440 Use this attribute on the SH to indicate an @code{interrupt_handler}
2441 function should switch to an alternate stack. It expects a string
2442 argument that names a global variable holding the address of the
2447 void f () __attribute__ ((interrupt_handler,
2448 sp_switch ("alt_stack")));
2452 Use this attribute on the SH for an @code{interrupt_handler} to return using
2453 @code{trapa} instead of @code{rte}. This attribute expects an integer
2454 argument specifying the trap number to be used.
2457 @cindex eight bit data on the H8/300, H8/300H, and H8S
2458 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2459 variable should be placed into the eight bit data section.
2460 The compiler will generate more efficient code for certain operations
2461 on data in the eight bit data area. Note the eight bit data area is limited to
2464 You must use GAS and GLD from GNU binutils version 2.7 or later for
2465 this attribute to work correctly.
2468 @cindex tiny data section on the H8/300H and H8S
2469 Use this attribute on the H8/300H and H8S to indicate that the specified
2470 variable should be placed into the tiny data section.
2471 The compiler will generate more efficient code for loads and stores
2472 on data in the tiny data section. Note the tiny data area is limited to
2473 slightly under 32kbytes of data.
2476 @cindex save all registers on the H8/300, H8/300H, and H8S
2477 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
2478 all registers except the stack pointer should be saved in the prologue
2479 regardless of whether they are used or not.
2482 @cindex signal handler functions on the AVR processors
2483 Use this attribute on the AVR to indicate that the specified
2484 function is a signal handler. The compiler will generate function
2485 entry and exit sequences suitable for use in a signal handler when this
2486 attribute is present. Interrupts will be disabled inside the function.
2489 @cindex function without a prologue/epilogue code
2490 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2491 specified function does not need prologue/epilogue sequences generated by
2492 the compiler. It is up to the programmer to provide these sequences.
2494 @item model (@var{model-name})
2495 @cindex function addressability on the M32R/D
2496 @cindex variable addressability on the IA-64
2498 On the M32R/D, use this attribute to set the addressability of an
2499 object, and of the code generated for a function. The identifier
2500 @var{model-name} is one of @code{small}, @code{medium}, or
2501 @code{large}, representing each of the code models.
2503 Small model objects live in the lower 16MB of memory (so that their
2504 addresses can be loaded with the @code{ld24} instruction), and are
2505 callable with the @code{bl} instruction.
2507 Medium model objects may live anywhere in the 32-bit address space (the
2508 compiler will generate @code{seth/add3} instructions to load their addresses),
2509 and are callable with the @code{bl} instruction.
2511 Large model objects may live anywhere in the 32-bit address space (the
2512 compiler will generate @code{seth/add3} instructions to load their addresses),
2513 and may not be reachable with the @code{bl} instruction (the compiler will
2514 generate the much slower @code{seth/add3/jl} instruction sequence).
2516 On IA-64, use this attribute to set the addressability of an object.
2517 At present, the only supported identifier for @var{model-name} is
2518 @code{small}, indicating addressability via ``small'' (22-bit)
2519 addresses (so that their addresses can be loaded with the @code{addl}
2520 instruction). Caveat: such addressing is by definition not position
2521 independent and hence this attribute must not be used for objects
2522 defined by shared libraries.
2525 @cindex functions which handle memory bank switching
2526 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2527 use a calling convention that takes care of switching memory banks when
2528 entering and leaving a function. This calling convention is also the
2529 default when using the @option{-mlong-calls} option.
2531 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2532 to call and return from a function.
2534 On 68HC11 the compiler will generate a sequence of instructions
2535 to invoke a board-specific routine to switch the memory bank and call the
2536 real function. The board-specific routine simulates a @code{call}.
2537 At the end of a function, it will jump to a board-specific routine
2538 instead of using @code{rts}. The board-specific return routine simulates
2542 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2543 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2544 use the normal calling convention based on @code{jsr} and @code{rts}.
2545 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2549 @cindex @code{__declspec(dllimport)}
2550 On Microsoft Windows targets, the @code{dllimport} attribute causes the compiler
2551 to reference a function or variable via a global pointer to a pointer
2552 that is set up by the Microsoft Windows dll library. The pointer name is formed by
2553 combining @code{_imp__} and the function or variable name. The attribute
2554 implies @code{extern} storage.
2556 Currently, the attribute is ignored for inlined functions. If the
2557 attribute is applied to a symbol @emph{definition}, an error is reported.
2558 If a symbol previously declared @code{dllimport} is later defined, the
2559 attribute is ignored in subsequent references, and a warning is emitted.
2560 The attribute is also overridden by a subsequent declaration as
2563 When applied to C++ classes, the attribute marks non-inlined
2564 member functions and static data members as imports. However, the
2565 attribute is ignored for virtual methods to allow creation of vtables
2568 On cygwin, mingw and arm-pe targets, @code{__declspec(dllimport)} is
2569 recognized as a synonym for @code{__attribute__ ((dllimport))} for
2570 compatibility with other Microsoft Windows compilers.
2572 The use of the @code{dllimport} attribute on functions is not necessary,
2573 but provides a small performance benefit by eliminating a thunk in the
2574 dll. The use of the @code{dllimport} attribute on imported variables was
2575 required on older versions of GNU ld, but can now be avoided by passing
2576 the @option{--enable-auto-import} switch to ld. As with functions, using
2577 the attribute for a variable eliminates a thunk in the dll.
2579 One drawback to using this attribute is that a pointer to a function or
2580 variable marked as dllimport cannot be used as a constant address. The
2581 attribute can be disabled for functions by setting the
2582 @option{-mnop-fun-dllimport} flag.
2585 @cindex @code{__declspec(dllexport)}
2586 On Microsoft Windows targets the @code{dllexport} attribute causes the compiler to
2587 provide a global pointer to a pointer in a dll, so that it can be
2588 referenced with the @code{dllimport} attribute. The pointer name is
2589 formed by combining @code{_imp__} and the function or variable name.
2591 Currently, the @code{dllexport}attribute is ignored for inlined
2592 functions, but export can be forced by using the
2593 @option{-fkeep-inline-functions} flag. The attribute is also ignored for
2596 When applied to C++ classes. the attribute marks defined non-inlined
2597 member functions and static data members as exports. Static consts
2598 initialized in-class are not marked unless they are also defined
2601 On cygwin, mingw and arm-pe targets, @code{__declspec(dllexport)} is
2602 recognized as a synonym for @code{__attribute__ ((dllexport))} for
2603 compatibility with other Microsoft Windows compilers.
2605 Alternative methods for including the symbol in the dll's export table
2606 are to use a .def file with an @code{EXPORTS} section or, with GNU ld,
2607 using the @option{--export-all} linker flag.
2611 You can specify multiple attributes in a declaration by separating them
2612 by commas within the double parentheses or by immediately following an
2613 attribute declaration with another attribute declaration.
2615 @cindex @code{#pragma}, reason for not using
2616 @cindex pragma, reason for not using
2617 Some people object to the @code{__attribute__} feature, suggesting that
2618 ISO C's @code{#pragma} should be used instead. At the time
2619 @code{__attribute__} was designed, there were two reasons for not doing
2624 It is impossible to generate @code{#pragma} commands from a macro.
2627 There is no telling what the same @code{#pragma} might mean in another
2631 These two reasons applied to almost any application that might have been
2632 proposed for @code{#pragma}. It was basically a mistake to use
2633 @code{#pragma} for @emph{anything}.
2635 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2636 to be generated from macros. In addition, a @code{#pragma GCC}
2637 namespace is now in use for GCC-specific pragmas. However, it has been
2638 found convenient to use @code{__attribute__} to achieve a natural
2639 attachment of attributes to their corresponding declarations, whereas
2640 @code{#pragma GCC} is of use for constructs that do not naturally form
2641 part of the grammar. @xref{Other Directives,,Miscellaneous
2642 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2644 @node Attribute Syntax
2645 @section Attribute Syntax
2646 @cindex attribute syntax
2648 This section describes the syntax with which @code{__attribute__} may be
2649 used, and the constructs to which attribute specifiers bind, for the C
2650 language. Some details may vary for C++ and Objective-C@. Because of
2651 infelicities in the grammar for attributes, some forms described here
2652 may not be successfully parsed in all cases.
2654 There are some problems with the semantics of attributes in C++. For
2655 example, there are no manglings for attributes, although they may affect
2656 code generation, so problems may arise when attributed types are used in
2657 conjunction with templates or overloading. Similarly, @code{typeid}
2658 does not distinguish between types with different attributes. Support
2659 for attributes in C++ may be restricted in future to attributes on
2660 declarations only, but not on nested declarators.
2662 @xref{Function Attributes}, for details of the semantics of attributes
2663 applying to functions. @xref{Variable Attributes}, for details of the
2664 semantics of attributes applying to variables. @xref{Type Attributes},
2665 for details of the semantics of attributes applying to structure, union
2666 and enumerated types.
2668 An @dfn{attribute specifier} is of the form
2669 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2670 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2671 each attribute is one of the following:
2675 Empty. Empty attributes are ignored.
2678 A word (which may be an identifier such as @code{unused}, or a reserved
2679 word such as @code{const}).
2682 A word, followed by, in parentheses, parameters for the attribute.
2683 These parameters take one of the following forms:
2687 An identifier. For example, @code{mode} attributes use this form.
2690 An identifier followed by a comma and a non-empty comma-separated list
2691 of expressions. For example, @code{format} attributes use this form.
2694 A possibly empty comma-separated list of expressions. For example,
2695 @code{format_arg} attributes use this form with the list being a single
2696 integer constant expression, and @code{alias} attributes use this form
2697 with the list being a single string constant.
2701 An @dfn{attribute specifier list} is a sequence of one or more attribute
2702 specifiers, not separated by any other tokens.
2704 In GNU C, an attribute specifier list may appear after the colon following a
2705 label, other than a @code{case} or @code{default} label. The only
2706 attribute it makes sense to use after a label is @code{unused}. This
2707 feature is intended for code generated by programs which contains labels
2708 that may be unused but which is compiled with @option{-Wall}. It would
2709 not normally be appropriate to use in it human-written code, though it
2710 could be useful in cases where the code that jumps to the label is
2711 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2712 such placement of attribute lists, as it is permissible for a
2713 declaration, which could begin with an attribute list, to be labelled in
2714 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2715 does not arise there.
2717 An attribute specifier list may appear as part of a @code{struct},
2718 @code{union} or @code{enum} specifier. It may go either immediately
2719 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2720 the closing brace. It is ignored if the content of the structure, union
2721 or enumerated type is not defined in the specifier in which the
2722 attribute specifier list is used---that is, in usages such as
2723 @code{struct __attribute__((foo)) bar} with no following opening brace.
2724 Where attribute specifiers follow the closing brace, they are considered
2725 to relate to the structure, union or enumerated type defined, not to any
2726 enclosing declaration the type specifier appears in, and the type
2727 defined is not complete until after the attribute specifiers.
2728 @c Otherwise, there would be the following problems: a shift/reduce
2729 @c conflict between attributes binding the struct/union/enum and
2730 @c binding to the list of specifiers/qualifiers; and "aligned"
2731 @c attributes could use sizeof for the structure, but the size could be
2732 @c changed later by "packed" attributes.
2734 Otherwise, an attribute specifier appears as part of a declaration,
2735 counting declarations of unnamed parameters and type names, and relates
2736 to that declaration (which may be nested in another declaration, for
2737 example in the case of a parameter declaration), or to a particular declarator
2738 within a declaration. Where an
2739 attribute specifier is applied to a parameter declared as a function or
2740 an array, it should apply to the function or array rather than the
2741 pointer to which the parameter is implicitly converted, but this is not
2742 yet correctly implemented.
2744 Any list of specifiers and qualifiers at the start of a declaration may
2745 contain attribute specifiers, whether or not such a list may in that
2746 context contain storage class specifiers. (Some attributes, however,
2747 are essentially in the nature of storage class specifiers, and only make
2748 sense where storage class specifiers may be used; for example,
2749 @code{section}.) There is one necessary limitation to this syntax: the
2750 first old-style parameter declaration in a function definition cannot
2751 begin with an attribute specifier, because such an attribute applies to
2752 the function instead by syntax described below (which, however, is not
2753 yet implemented in this case). In some other cases, attribute
2754 specifiers are permitted by this grammar but not yet supported by the
2755 compiler. All attribute specifiers in this place relate to the
2756 declaration as a whole. In the obsolescent usage where a type of
2757 @code{int} is implied by the absence of type specifiers, such a list of
2758 specifiers and qualifiers may be an attribute specifier list with no
2759 other specifiers or qualifiers.
2761 An attribute specifier list may appear immediately before a declarator
2762 (other than the first) in a comma-separated list of declarators in a
2763 declaration of more than one identifier using a single list of
2764 specifiers and qualifiers. Such attribute specifiers apply
2765 only to the identifier before whose declarator they appear. For
2769 __attribute__((noreturn)) void d0 (void),
2770 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2775 the @code{noreturn} attribute applies to all the functions
2776 declared; the @code{format} attribute only applies to @code{d1}.
2778 An attribute specifier list may appear immediately before the comma,
2779 @code{=} or semicolon terminating the declaration of an identifier other
2780 than a function definition. At present, such attribute specifiers apply
2781 to the declared object or function, but in future they may attach to the
2782 outermost adjacent declarator. In simple cases there is no difference,
2783 but, for example, in
2786 void (****f)(void) __attribute__((noreturn));
2790 at present the @code{noreturn} attribute applies to @code{f}, which
2791 causes a warning since @code{f} is not a function, but in future it may
2792 apply to the function @code{****f}. The precise semantics of what
2793 attributes in such cases will apply to are not yet specified. Where an
2794 assembler name for an object or function is specified (@pxref{Asm
2795 Labels}), at present the attribute must follow the @code{asm}
2796 specification; in future, attributes before the @code{asm} specification
2797 may apply to the adjacent declarator, and those after it to the declared
2800 An attribute specifier list may, in future, be permitted to appear after
2801 the declarator in a function definition (before any old-style parameter
2802 declarations or the function body).
2804 Attribute specifiers may be mixed with type qualifiers appearing inside
2805 the @code{[]} of a parameter array declarator, in the C99 construct by
2806 which such qualifiers are applied to the pointer to which the array is
2807 implicitly converted. Such attribute specifiers apply to the pointer,
2808 not to the array, but at present this is not implemented and they are
2811 An attribute specifier list may appear at the start of a nested
2812 declarator. At present, there are some limitations in this usage: the
2813 attributes correctly apply to the declarator, but for most individual
2814 attributes the semantics this implies are not implemented.
2815 When attribute specifiers follow the @code{*} of a pointer
2816 declarator, they may be mixed with any type qualifiers present.
2817 The following describes the formal semantics of this syntax. It will make the
2818 most sense if you are familiar with the formal specification of
2819 declarators in the ISO C standard.
2821 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2822 D1}, where @code{T} contains declaration specifiers that specify a type
2823 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2824 contains an identifier @var{ident}. The type specified for @var{ident}
2825 for derived declarators whose type does not include an attribute
2826 specifier is as in the ISO C standard.
2828 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2829 and the declaration @code{T D} specifies the type
2830 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2831 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2832 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2834 If @code{D1} has the form @code{*
2835 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2836 declaration @code{T D} specifies the type
2837 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2838 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2839 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2845 void (__attribute__((noreturn)) ****f) (void);
2849 specifies the type ``pointer to pointer to pointer to pointer to
2850 non-returning function returning @code{void}''. As another example,
2853 char *__attribute__((aligned(8))) *f;
2857 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2858 Note again that this does not work with most attributes; for example,
2859 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2860 is not yet supported.
2862 For compatibility with existing code written for compiler versions that
2863 did not implement attributes on nested declarators, some laxity is
2864 allowed in the placing of attributes. If an attribute that only applies
2865 to types is applied to a declaration, it will be treated as applying to
2866 the type of that declaration. If an attribute that only applies to
2867 declarations is applied to the type of a declaration, it will be treated
2868 as applying to that declaration; and, for compatibility with code
2869 placing the attributes immediately before the identifier declared, such
2870 an attribute applied to a function return type will be treated as
2871 applying to the function type, and such an attribute applied to an array
2872 element type will be treated as applying to the array type. If an
2873 attribute that only applies to function types is applied to a
2874 pointer-to-function type, it will be treated as applying to the pointer
2875 target type; if such an attribute is applied to a function return type
2876 that is not a pointer-to-function type, it will be treated as applying
2877 to the function type.
2879 @node Function Prototypes
2880 @section Prototypes and Old-Style Function Definitions
2881 @cindex function prototype declarations
2882 @cindex old-style function definitions
2883 @cindex promotion of formal parameters
2885 GNU C extends ISO C to allow a function prototype to override a later
2886 old-style non-prototype definition. Consider the following example:
2889 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2896 /* @r{Prototype function declaration.} */
2897 int isroot P((uid_t));
2899 /* @r{Old-style function definition.} */
2901 isroot (x) /* ??? lossage here ??? */
2908 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2909 not allow this example, because subword arguments in old-style
2910 non-prototype definitions are promoted. Therefore in this example the
2911 function definition's argument is really an @code{int}, which does not
2912 match the prototype argument type of @code{short}.
2914 This restriction of ISO C makes it hard to write code that is portable
2915 to traditional C compilers, because the programmer does not know
2916 whether the @code{uid_t} type is @code{short}, @code{int}, or
2917 @code{long}. Therefore, in cases like these GNU C allows a prototype
2918 to override a later old-style definition. More precisely, in GNU C, a
2919 function prototype argument type overrides the argument type specified
2920 by a later old-style definition if the former type is the same as the
2921 latter type before promotion. Thus in GNU C the above example is
2922 equivalent to the following:
2935 GNU C++ does not support old-style function definitions, so this
2936 extension is irrelevant.
2939 @section C++ Style Comments
2941 @cindex C++ comments
2942 @cindex comments, C++ style
2944 In GNU C, you may use C++ style comments, which start with @samp{//} and
2945 continue until the end of the line. Many other C implementations allow
2946 such comments, and they are included in the 1999 C standard. However,
2947 C++ style comments are not recognized if you specify an @option{-std}
2948 option specifying a version of ISO C before C99, or @option{-ansi}
2949 (equivalent to @option{-std=c89}).
2952 @section Dollar Signs in Identifier Names
2954 @cindex dollar signs in identifier names
2955 @cindex identifier names, dollar signs in
2957 In GNU C, you may normally use dollar signs in identifier names.
2958 This is because many traditional C implementations allow such identifiers.
2959 However, dollar signs in identifiers are not supported on a few target
2960 machines, typically because the target assembler does not allow them.
2962 @node Character Escapes
2963 @section The Character @key{ESC} in Constants
2965 You can use the sequence @samp{\e} in a string or character constant to
2966 stand for the ASCII character @key{ESC}.
2969 @section Inquiring on Alignment of Types or Variables
2971 @cindex type alignment
2972 @cindex variable alignment
2974 The keyword @code{__alignof__} allows you to inquire about how an object
2975 is aligned, or the minimum alignment usually required by a type. Its
2976 syntax is just like @code{sizeof}.
2978 For example, if the target machine requires a @code{double} value to be
2979 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2980 This is true on many RISC machines. On more traditional machine
2981 designs, @code{__alignof__ (double)} is 4 or even 2.
2983 Some machines never actually require alignment; they allow reference to any
2984 data type even at an odd address. For these machines, @code{__alignof__}
2985 reports the @emph{recommended} alignment of a type.
2987 If the operand of @code{__alignof__} is an lvalue rather than a type,
2988 its value is the required alignment for its type, taking into account
2989 any minimum alignment specified with GCC's @code{__attribute__}
2990 extension (@pxref{Variable Attributes}). For example, after this
2994 struct foo @{ int x; char y; @} foo1;
2998 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2999 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3001 It is an error to ask for the alignment of an incomplete type.
3003 @node Variable Attributes
3004 @section Specifying Attributes of Variables
3005 @cindex attribute of variables
3006 @cindex variable attributes
3008 The keyword @code{__attribute__} allows you to specify special
3009 attributes of variables or structure fields. This keyword is followed
3010 by an attribute specification inside double parentheses. Some
3011 attributes are currently defined generically for variables.
3012 Other attributes are defined for variables on particular target
3013 systems. Other attributes are available for functions
3014 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3015 Other front ends might define more attributes
3016 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3018 You may also specify attributes with @samp{__} preceding and following
3019 each keyword. This allows you to use them in header files without
3020 being concerned about a possible macro of the same name. For example,
3021 you may use @code{__aligned__} instead of @code{aligned}.
3023 @xref{Attribute Syntax}, for details of the exact syntax for using
3027 @cindex @code{aligned} attribute
3028 @item aligned (@var{alignment})
3029 This attribute specifies a minimum alignment for the variable or
3030 structure field, measured in bytes. For example, the declaration:
3033 int x __attribute__ ((aligned (16))) = 0;
3037 causes the compiler to allocate the global variable @code{x} on a
3038 16-byte boundary. On a 68040, this could be used in conjunction with
3039 an @code{asm} expression to access the @code{move16} instruction which
3040 requires 16-byte aligned operands.
3042 You can also specify the alignment of structure fields. For example, to
3043 create a double-word aligned @code{int} pair, you could write:
3046 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3050 This is an alternative to creating a union with a @code{double} member
3051 that forces the union to be double-word aligned.
3053 As in the preceding examples, you can explicitly specify the alignment
3054 (in bytes) that you wish the compiler to use for a given variable or
3055 structure field. Alternatively, you can leave out the alignment factor
3056 and just ask the compiler to align a variable or field to the maximum
3057 useful alignment for the target machine you are compiling for. For
3058 example, you could write:
3061 short array[3] __attribute__ ((aligned));
3064 Whenever you leave out the alignment factor in an @code{aligned} attribute
3065 specification, the compiler automatically sets the alignment for the declared
3066 variable or field to the largest alignment which is ever used for any data
3067 type on the target machine you are compiling for. Doing this can often make
3068 copy operations more efficient, because the compiler can use whatever
3069 instructions copy the biggest chunks of memory when performing copies to
3070 or from the variables or fields that you have aligned this way.
3072 The @code{aligned} attribute can only increase the alignment; but you
3073 can decrease it by specifying @code{packed} as well. See below.
3075 Note that the effectiveness of @code{aligned} attributes may be limited
3076 by inherent limitations in your linker. On many systems, the linker is
3077 only able to arrange for variables to be aligned up to a certain maximum
3078 alignment. (For some linkers, the maximum supported alignment may
3079 be very very small.) If your linker is only able to align variables
3080 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3081 in an @code{__attribute__} will still only provide you with 8 byte
3082 alignment. See your linker documentation for further information.
3084 @item cleanup (@var{cleanup_function})
3085 @cindex @code{cleanup} attribute
3086 The @code{cleanup} attribute runs a function when the variable goes
3087 out of scope. This attribute can only be applied to auto function
3088 scope variables; it may not be applied to parameters or variables
3089 with static storage duration. The function must take one parameter,
3090 a pointer to a type compatible with the variable. The return value
3091 of the function (if any) is ignored.
3093 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3094 will be run during the stack unwinding that happens during the
3095 processing of the exception. Note that the @code{cleanup} attribute
3096 does not allow the exception to be caught, only to perform an action.
3097 It is undefined what happens if @var{cleanup_function} does not
3102 @cindex @code{common} attribute
3103 @cindex @code{nocommon} attribute
3106 The @code{common} attribute requests GCC to place a variable in
3107 ``common'' storage. The @code{nocommon} attribute requests the
3108 opposite -- to allocate space for it directly.
3110 These attributes override the default chosen by the
3111 @option{-fno-common} and @option{-fcommon} flags respectively.
3114 @cindex @code{deprecated} attribute
3115 The @code{deprecated} attribute results in a warning if the variable
3116 is used anywhere in the source file. This is useful when identifying
3117 variables that are expected to be removed in a future version of a
3118 program. The warning also includes the location of the declaration
3119 of the deprecated variable, to enable users to easily find further
3120 information about why the variable is deprecated, or what they should
3121 do instead. Note that the warning only occurs for uses:
3124 extern int old_var __attribute__ ((deprecated));
3126 int new_fn () @{ return old_var; @}
3129 results in a warning on line 3 but not line 2.
3131 The @code{deprecated} attribute can also be used for functions and
3132 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3134 @item mode (@var{mode})
3135 @cindex @code{mode} attribute
3136 This attribute specifies the data type for the declaration---whichever
3137 type corresponds to the mode @var{mode}. This in effect lets you
3138 request an integer or floating point type according to its width.
3140 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3141 indicate the mode corresponding to a one-byte integer, @samp{word} or
3142 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3143 or @samp{__pointer__} for the mode used to represent pointers.
3146 @cindex @code{packed} attribute
3147 The @code{packed} attribute specifies that a variable or structure field
3148 should have the smallest possible alignment---one byte for a variable,
3149 and one bit for a field, unless you specify a larger value with the
3150 @code{aligned} attribute.
3152 Here is a structure in which the field @code{x} is packed, so that it
3153 immediately follows @code{a}:
3159 int x[2] __attribute__ ((packed));
3163 @item section ("@var{section-name}")
3164 @cindex @code{section} variable attribute
3165 Normally, the compiler places the objects it generates in sections like
3166 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3167 or you need certain particular variables to appear in special sections,
3168 for example to map to special hardware. The @code{section}
3169 attribute specifies that a variable (or function) lives in a particular
3170 section. For example, this small program uses several specific section names:
3173 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3174 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3175 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3176 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3180 /* Initialize stack pointer */
3181 init_sp (stack + sizeof (stack));
3183 /* Initialize initialized data */
3184 memcpy (&init_data, &data, &edata - &data);
3186 /* Turn on the serial ports */
3193 Use the @code{section} attribute with an @emph{initialized} definition
3194 of a @emph{global} variable, as shown in the example. GCC issues
3195 a warning and otherwise ignores the @code{section} attribute in
3196 uninitialized variable declarations.
3198 You may only use the @code{section} attribute with a fully initialized
3199 global definition because of the way linkers work. The linker requires
3200 each object be defined once, with the exception that uninitialized
3201 variables tentatively go in the @code{common} (or @code{bss}) section
3202 and can be multiply ``defined''. You can force a variable to be
3203 initialized with the @option{-fno-common} flag or the @code{nocommon}
3206 Some file formats do not support arbitrary sections so the @code{section}
3207 attribute is not available on all platforms.
3208 If you need to map the entire contents of a module to a particular
3209 section, consider using the facilities of the linker instead.
3212 @cindex @code{shared} variable attribute
3213 On Microsoft Windows, in addition to putting variable definitions in a named
3214 section, the section can also be shared among all running copies of an
3215 executable or DLL@. For example, this small program defines shared data
3216 by putting it in a named section @code{shared} and marking the section
3220 int foo __attribute__((section ("shared"), shared)) = 0;
3225 /* Read and write foo. All running
3226 copies see the same value. */
3232 You may only use the @code{shared} attribute along with @code{section}
3233 attribute with a fully initialized global definition because of the way
3234 linkers work. See @code{section} attribute for more information.
3236 The @code{shared} attribute is only available on Microsoft Windows@.
3238 @item tls_model ("@var{tls_model}")
3239 @cindex @code{tls_model} attribute
3240 The @code{tls_model} attribute sets thread-local storage model
3241 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3242 overriding @code{-ftls-model=} command line switch on a per-variable
3244 The @var{tls_model} argument should be one of @code{global-dynamic},
3245 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3247 Not all targets support this attribute.
3249 @item transparent_union
3250 This attribute, attached to a function parameter which is a union, means
3251 that the corresponding argument may have the type of any union member,
3252 but the argument is passed as if its type were that of the first union
3253 member. For more details see @xref{Type Attributes}. You can also use
3254 this attribute on a @code{typedef} for a union data type; then it
3255 applies to all function parameters with that type.
3258 This attribute, attached to a variable, means that the variable is meant
3259 to be possibly unused. GCC will not produce a warning for this
3262 @item vector_size (@var{bytes})
3263 This attribute specifies the vector size for the variable, measured in
3264 bytes. For example, the declaration:
3267 int foo __attribute__ ((vector_size (16)));
3271 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3272 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3273 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3275 This attribute is only applicable to integral and float scalars,
3276 although arrays, pointers, and function return values are allowed in
3277 conjunction with this construct.
3279 Aggregates with this attribute are invalid, even if they are of the same
3280 size as a corresponding scalar. For example, the declaration:
3283 struct S @{ int a; @};
3284 struct S __attribute__ ((vector_size (16))) foo;
3288 is invalid even if the size of the structure is the same as the size of
3292 The @code{weak} attribute is described in @xref{Function Attributes}.
3295 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3298 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3302 @subsection M32R/D Variable Attributes
3304 One attribute is currently defined for the M32R/D.
3307 @item model (@var{model-name})
3308 @cindex variable addressability on the M32R/D
3309 Use this attribute on the M32R/D to set the addressability of an object.
3310 The identifier @var{model-name} is one of @code{small}, @code{medium},
3311 or @code{large}, representing each of the code models.
3313 Small model objects live in the lower 16MB of memory (so that their
3314 addresses can be loaded with the @code{ld24} instruction).
3316 Medium and large model objects may live anywhere in the 32-bit address space
3317 (the compiler will generate @code{seth/add3} instructions to load their
3321 @subsection i386 Variable Attributes
3323 Two attributes are currently defined for i386 configurations:
3324 @code{ms_struct} and @code{gcc_struct}
3329 @cindex @code{ms_struct} attribute
3330 @cindex @code{gcc_struct} attribute
3332 If @code{packed} is used on a structure, or if bit-fields are used
3333 it may be that the Microsoft ABI packs them differently
3334 than GCC would normally pack them. Particularly when moving packed
3335 data between functions compiled with GCC and the native Microsoft compiler
3336 (either via function call or as data in a file), it may be necessary to access
3339 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3340 compilers to match the native Microsoft compiler.
3343 @node Type Attributes
3344 @section Specifying Attributes of Types
3345 @cindex attribute of types
3346 @cindex type attributes
3348 The keyword @code{__attribute__} allows you to specify special
3349 attributes of @code{struct} and @code{union} types when you define such
3350 types. This keyword is followed by an attribute specification inside
3351 double parentheses. Six attributes are currently defined for types:
3352 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3353 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3354 functions (@pxref{Function Attributes}) and for variables
3355 (@pxref{Variable Attributes}).
3357 You may also specify any one of these attributes with @samp{__}
3358 preceding and following its keyword. This allows you to use these
3359 attributes in header files without being concerned about a possible
3360 macro of the same name. For example, you may use @code{__aligned__}
3361 instead of @code{aligned}.
3363 You may specify the @code{aligned} and @code{transparent_union}
3364 attributes either in a @code{typedef} declaration or just past the
3365 closing curly brace of a complete enum, struct or union type
3366 @emph{definition} and the @code{packed} attribute only past the closing
3367 brace of a definition.
3369 You may also specify attributes between the enum, struct or union
3370 tag and the name of the type rather than after the closing brace.
3372 @xref{Attribute Syntax}, for details of the exact syntax for using
3376 @cindex @code{aligned} attribute
3377 @item aligned (@var{alignment})
3378 This attribute specifies a minimum alignment (in bytes) for variables
3379 of the specified type. For example, the declarations:
3382 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3383 typedef int more_aligned_int __attribute__ ((aligned (8)));
3387 force the compiler to insure (as far as it can) that each variable whose
3388 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3389 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3390 variables of type @code{struct S} aligned to 8-byte boundaries allows
3391 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3392 store) instructions when copying one variable of type @code{struct S} to
3393 another, thus improving run-time efficiency.
3395 Note that the alignment of any given @code{struct} or @code{union} type
3396 is required by the ISO C standard to be at least a perfect multiple of
3397 the lowest common multiple of the alignments of all of the members of
3398 the @code{struct} or @code{union} in question. This means that you @emph{can}
3399 effectively adjust the alignment of a @code{struct} or @code{union}
3400 type by attaching an @code{aligned} attribute to any one of the members
3401 of such a type, but the notation illustrated in the example above is a
3402 more obvious, intuitive, and readable way to request the compiler to
3403 adjust the alignment of an entire @code{struct} or @code{union} type.
3405 As in the preceding example, you can explicitly specify the alignment
3406 (in bytes) that you wish the compiler to use for a given @code{struct}
3407 or @code{union} type. Alternatively, you can leave out the alignment factor
3408 and just ask the compiler to align a type to the maximum
3409 useful alignment for the target machine you are compiling for. For
3410 example, you could write:
3413 struct S @{ short f[3]; @} __attribute__ ((aligned));
3416 Whenever you leave out the alignment factor in an @code{aligned}
3417 attribute specification, the compiler automatically sets the alignment
3418 for the type to the largest alignment which is ever used for any data
3419 type on the target machine you are compiling for. Doing this can often
3420 make copy operations more efficient, because the compiler can use
3421 whatever instructions copy the biggest chunks of memory when performing
3422 copies to or from the variables which have types that you have aligned
3425 In the example above, if the size of each @code{short} is 2 bytes, then
3426 the size of the entire @code{struct S} type is 6 bytes. The smallest
3427 power of two which is greater than or equal to that is 8, so the
3428 compiler sets the alignment for the entire @code{struct S} type to 8
3431 Note that although you can ask the compiler to select a time-efficient
3432 alignment for a given type and then declare only individual stand-alone
3433 objects of that type, the compiler's ability to select a time-efficient
3434 alignment is primarily useful only when you plan to create arrays of
3435 variables having the relevant (efficiently aligned) type. If you
3436 declare or use arrays of variables of an efficiently-aligned type, then
3437 it is likely that your program will also be doing pointer arithmetic (or
3438 subscripting, which amounts to the same thing) on pointers to the
3439 relevant type, and the code that the compiler generates for these
3440 pointer arithmetic operations will often be more efficient for
3441 efficiently-aligned types than for other types.
3443 The @code{aligned} attribute can only increase the alignment; but you
3444 can decrease it by specifying @code{packed} as well. See below.
3446 Note that the effectiveness of @code{aligned} attributes may be limited
3447 by inherent limitations in your linker. On many systems, the linker is
3448 only able to arrange for variables to be aligned up to a certain maximum
3449 alignment. (For some linkers, the maximum supported alignment may
3450 be very very small.) If your linker is only able to align variables
3451 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3452 in an @code{__attribute__} will still only provide you with 8 byte
3453 alignment. See your linker documentation for further information.
3456 This attribute, attached to @code{struct} or @code{union} type
3457 definition, specifies that each member of the structure or union is
3458 placed to minimize the memory required. When attached to an @code{enum}
3459 definition, it indicates that the smallest integral type should be used.
3461 @opindex fshort-enums
3462 Specifying this attribute for @code{struct} and @code{union} types is
3463 equivalent to specifying the @code{packed} attribute on each of the
3464 structure or union members. Specifying the @option{-fshort-enums}
3465 flag on the line is equivalent to specifying the @code{packed}
3466 attribute on all @code{enum} definitions.
3468 In the following example @code{struct my_packed_struct}'s members are
3469 packed closely together, but the internal layout of its @code{s} member
3470 is not packed -- to do that, @code{struct my_unpacked_struct} would need to
3474 struct my_unpacked_struct
3480 struct my_packed_struct __attribute__ ((__packed__))
3484 struct my_unpacked_struct s;
3488 You may only specify this attribute on the definition of a @code{enum},
3489 @code{struct} or @code{union}, not on a @code{typedef} which does not
3490 also define the enumerated type, structure or union.
3492 @item transparent_union
3493 This attribute, attached to a @code{union} type definition, indicates
3494 that any function parameter having that union type causes calls to that
3495 function to be treated in a special way.
3497 First, the argument corresponding to a transparent union type can be of
3498 any type in the union; no cast is required. Also, if the union contains
3499 a pointer type, the corresponding argument can be a null pointer
3500 constant or a void pointer expression; and if the union contains a void
3501 pointer type, the corresponding argument can be any pointer expression.
3502 If the union member type is a pointer, qualifiers like @code{const} on
3503 the referenced type must be respected, just as with normal pointer
3506 Second, the argument is passed to the function using the calling
3507 conventions of the first member of the transparent union, not the calling
3508 conventions of the union itself. All members of the union must have the
3509 same machine representation; this is necessary for this argument passing
3512 Transparent unions are designed for library functions that have multiple
3513 interfaces for compatibility reasons. For example, suppose the
3514 @code{wait} function must accept either a value of type @code{int *} to
3515 comply with Posix, or a value of type @code{union wait *} to comply with
3516 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3517 @code{wait} would accept both kinds of arguments, but it would also
3518 accept any other pointer type and this would make argument type checking
3519 less useful. Instead, @code{<sys/wait.h>} might define the interface
3527 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3529 pid_t wait (wait_status_ptr_t);
3532 This interface allows either @code{int *} or @code{union wait *}
3533 arguments to be passed, using the @code{int *} calling convention.
3534 The program can call @code{wait} with arguments of either type:
3537 int w1 () @{ int w; return wait (&w); @}
3538 int w2 () @{ union wait w; return wait (&w); @}
3541 With this interface, @code{wait}'s implementation might look like this:
3544 pid_t wait (wait_status_ptr_t p)
3546 return waitpid (-1, p.__ip, 0);
3551 When attached to a type (including a @code{union} or a @code{struct}),
3552 this attribute means that variables of that type are meant to appear
3553 possibly unused. GCC will not produce a warning for any variables of
3554 that type, even if the variable appears to do nothing. This is often
3555 the case with lock or thread classes, which are usually defined and then
3556 not referenced, but contain constructors and destructors that have
3557 nontrivial bookkeeping functions.
3560 The @code{deprecated} attribute results in a warning if the type
3561 is used anywhere in the source file. This is useful when identifying
3562 types that are expected to be removed in a future version of a program.
3563 If possible, the warning also includes the location of the declaration
3564 of the deprecated type, to enable users to easily find further
3565 information about why the type is deprecated, or what they should do
3566 instead. Note that the warnings only occur for uses and then only
3567 if the type is being applied to an identifier that itself is not being
3568 declared as deprecated.
3571 typedef int T1 __attribute__ ((deprecated));
3575 typedef T1 T3 __attribute__ ((deprecated));
3576 T3 z __attribute__ ((deprecated));
3579 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3580 warning is issued for line 4 because T2 is not explicitly
3581 deprecated. Line 5 has no warning because T3 is explicitly
3582 deprecated. Similarly for line 6.
3584 The @code{deprecated} attribute can also be used for functions and
3585 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3588 Accesses to objects with types with this attribute are not subjected to
3589 type-based alias analysis, but are instead assumed to be able to alias
3590 any other type of objects, just like the @code{char} type. See
3591 @option{-fstrict-aliasing} for more information on aliasing issues.
3596 typedef short __attribute__((__may_alias__)) short_a;
3602 short_a *b = (short_a *) &a;
3606 if (a == 0x12345678)
3613 If you replaced @code{short_a} with @code{short} in the variable
3614 declaration, the above program would abort when compiled with
3615 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3616 above in recent GCC versions.
3618 @subsection i386 Type Attributes
3620 Two attributes are currently defined for i386 configurations:
3621 @code{ms_struct} and @code{gcc_struct}
3625 @cindex @code{ms_struct}
3626 @cindex @code{gcc_struct}
3628 If @code{packed} is used on a structure, or if bit-fields are used
3629 it may be that the Microsoft ABI packs them differently
3630 than GCC would normally pack them. Particularly when moving packed
3631 data between functions compiled with GCC and the native Microsoft compiler
3632 (either via function call or as data in a file), it may be necessary to access
3635 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3636 compilers to match the native Microsoft compiler.
3639 To specify multiple attributes, separate them by commas within the
3640 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3644 @section An Inline Function is As Fast As a Macro
3645 @cindex inline functions
3646 @cindex integrating function code
3648 @cindex macros, inline alternative
3650 By declaring a function @code{inline}, you can direct GCC to
3651 integrate that function's code into the code for its callers. This
3652 makes execution faster by eliminating the function-call overhead; in
3653 addition, if any of the actual argument values are constant, their known
3654 values may permit simplifications at compile time so that not all of the
3655 inline function's code needs to be included. The effect on code size is
3656 less predictable; object code may be larger or smaller with function
3657 inlining, depending on the particular case. Inlining of functions is an
3658 optimization and it really ``works'' only in optimizing compilation. If
3659 you don't use @option{-O}, no function is really inline.
3661 Inline functions are included in the ISO C99 standard, but there are
3662 currently substantial differences between what GCC implements and what
3663 the ISO C99 standard requires.
3665 To declare a function inline, use the @code{inline} keyword in its
3666 declaration, like this:
3676 (If you are writing a header file to be included in ISO C programs, write
3677 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3678 You can also make all ``simple enough'' functions inline with the option
3679 @option{-finline-functions}.
3682 Note that certain usages in a function definition can make it unsuitable
3683 for inline substitution. Among these usages are: use of varargs, use of
3684 alloca, use of variable sized data types (@pxref{Variable Length}),
3685 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3686 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3687 will warn when a function marked @code{inline} could not be substituted,
3688 and will give the reason for the failure.
3690 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3691 does not affect the linkage of the function.
3693 @cindex automatic @code{inline} for C++ member fns
3694 @cindex @code{inline} automatic for C++ member fns
3695 @cindex member fns, automatically @code{inline}
3696 @cindex C++ member fns, automatically @code{inline}
3697 @opindex fno-default-inline
3698 GCC automatically inlines member functions defined within the class
3699 body of C++ programs even if they are not explicitly declared
3700 @code{inline}. (You can override this with @option{-fno-default-inline};
3701 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3703 @cindex inline functions, omission of
3704 @opindex fkeep-inline-functions
3705 When a function is both inline and @code{static}, if all calls to the
3706 function are integrated into the caller, and the function's address is
3707 never used, then the function's own assembler code is never referenced.
3708 In this case, GCC does not actually output assembler code for the
3709 function, unless you specify the option @option{-fkeep-inline-functions}.
3710 Some calls cannot be integrated for various reasons (in particular,
3711 calls that precede the function's definition cannot be integrated, and
3712 neither can recursive calls within the definition). If there is a
3713 nonintegrated call, then the function is compiled to assembler code as
3714 usual. The function must also be compiled as usual if the program
3715 refers to its address, because that can't be inlined.
3717 @cindex non-static inline function
3718 When an inline function is not @code{static}, then the compiler must assume
3719 that there may be calls from other source files; since a global symbol can
3720 be defined only once in any program, the function must not be defined in
3721 the other source files, so the calls therein cannot be integrated.
3722 Therefore, a non-@code{static} inline function is always compiled on its
3723 own in the usual fashion.
3725 If you specify both @code{inline} and @code{extern} in the function
3726 definition, then the definition is used only for inlining. In no case
3727 is the function compiled on its own, not even if you refer to its
3728 address explicitly. Such an address becomes an external reference, as
3729 if you had only declared the function, and had not defined it.
3731 This combination of @code{inline} and @code{extern} has almost the
3732 effect of a macro. The way to use it is to put a function definition in
3733 a header file with these keywords, and put another copy of the
3734 definition (lacking @code{inline} and @code{extern}) in a library file.
3735 The definition in the header file will cause most calls to the function
3736 to be inlined. If any uses of the function remain, they will refer to
3737 the single copy in the library.
3739 Since GCC eventually will implement ISO C99 semantics for
3740 inline functions, it is best to use @code{static inline} only
3741 to guarantee compatibility. (The
3742 existing semantics will remain available when @option{-std=gnu89} is
3743 specified, but eventually the default will be @option{-std=gnu99} and
3744 that will implement the C99 semantics, though it does not do so yet.)
3746 GCC does not inline any functions when not optimizing unless you specify
3747 the @samp{always_inline} attribute for the function, like this:
3751 inline void foo (const char) __attribute__((always_inline));
3755 @section Assembler Instructions with C Expression Operands
3756 @cindex extended @code{asm}
3757 @cindex @code{asm} expressions
3758 @cindex assembler instructions
3761 In an assembler instruction using @code{asm}, you can specify the
3762 operands of the instruction using C expressions. This means you need not
3763 guess which registers or memory locations will contain the data you want
3766 You must specify an assembler instruction template much like what
3767 appears in a machine description, plus an operand constraint string for
3770 For example, here is how to use the 68881's @code{fsinx} instruction:
3773 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3777 Here @code{angle} is the C expression for the input operand while
3778 @code{result} is that of the output operand. Each has @samp{"f"} as its
3779 operand constraint, saying that a floating point register is required.
3780 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3781 output operands' constraints must use @samp{=}. The constraints use the
3782 same language used in the machine description (@pxref{Constraints}).
3784 Each operand is described by an operand-constraint string followed by
3785 the C expression in parentheses. A colon separates the assembler
3786 template from the first output operand and another separates the last
3787 output operand from the first input, if any. Commas separate the
3788 operands within each group. The total number of operands is currently
3789 limited to 30; this limitation may be lifted in some future version of
3792 If there are no output operands but there are input operands, you must
3793 place two consecutive colons surrounding the place where the output
3796 As of GCC version 3.1, it is also possible to specify input and output
3797 operands using symbolic names which can be referenced within the
3798 assembler code. These names are specified inside square brackets
3799 preceding the constraint string, and can be referenced inside the
3800 assembler code using @code{%[@var{name}]} instead of a percentage sign
3801 followed by the operand number. Using named operands the above example
3805 asm ("fsinx %[angle],%[output]"
3806 : [output] "=f" (result)
3807 : [angle] "f" (angle));
3811 Note that the symbolic operand names have no relation whatsoever to
3812 other C identifiers. You may use any name you like, even those of
3813 existing C symbols, but you must ensure that no two operands within the same
3814 assembler construct use the same symbolic name.
3816 Output operand expressions must be lvalues; the compiler can check this.
3817 The input operands need not be lvalues. The compiler cannot check
3818 whether the operands have data types that are reasonable for the
3819 instruction being executed. It does not parse the assembler instruction
3820 template and does not know what it means or even whether it is valid
3821 assembler input. The extended @code{asm} feature is most often used for
3822 machine instructions the compiler itself does not know exist. If
3823 the output expression cannot be directly addressed (for example, it is a
3824 bit-field), your constraint must allow a register. In that case, GCC
3825 will use the register as the output of the @code{asm}, and then store
3826 that register into the output.
3828 The ordinary output operands must be write-only; GCC will assume that
3829 the values in these operands before the instruction are dead and need
3830 not be generated. Extended asm supports input-output or read-write
3831 operands. Use the constraint character @samp{+} to indicate such an
3832 operand and list it with the output operands. You should only use
3833 read-write operands when the constraints for the operand (or the
3834 operand in which only some of the bits are to be changed) allow a
3837 You may, as an alternative, logically split its function into two
3838 separate operands, one input operand and one write-only output
3839 operand. The connection between them is expressed by constraints
3840 which say they need to be in the same location when the instruction
3841 executes. You can use the same C expression for both operands, or
3842 different expressions. For example, here we write the (fictitious)
3843 @samp{combine} instruction with @code{bar} as its read-only source
3844 operand and @code{foo} as its read-write destination:
3847 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3851 The constraint @samp{"0"} for operand 1 says that it must occupy the
3852 same location as operand 0. A number in constraint is allowed only in
3853 an input operand and it must refer to an output operand.
3855 Only a number in the constraint can guarantee that one operand will be in
3856 the same place as another. The mere fact that @code{foo} is the value
3857 of both operands is not enough to guarantee that they will be in the
3858 same place in the generated assembler code. The following would not
3862 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3865 Various optimizations or reloading could cause operands 0 and 1 to be in
3866 different registers; GCC knows no reason not to do so. For example, the
3867 compiler might find a copy of the value of @code{foo} in one register and
3868 use it for operand 1, but generate the output operand 0 in a different
3869 register (copying it afterward to @code{foo}'s own address). Of course,
3870 since the register for operand 1 is not even mentioned in the assembler
3871 code, the result will not work, but GCC can't tell that.
3873 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3874 the operand number for a matching constraint. For example:
3877 asm ("cmoveq %1,%2,%[result]"
3878 : [result] "=r"(result)
3879 : "r" (test), "r"(new), "[result]"(old));
3882 Some instructions clobber specific hard registers. To describe this,
3883 write a third colon after the input operands, followed by the names of
3884 the clobbered hard registers (given as strings). Here is a realistic
3885 example for the VAX:
3888 asm volatile ("movc3 %0,%1,%2"
3890 : "g" (from), "g" (to), "g" (count)
3891 : "r0", "r1", "r2", "r3", "r4", "r5");
3894 You may not write a clobber description in a way that overlaps with an
3895 input or output operand. For example, you may not have an operand
3896 describing a register class with one member if you mention that register
3897 in the clobber list. Variables declared to live in specific registers
3898 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3899 have no part mentioned in the clobber description.
3900 There is no way for you to specify that an input
3901 operand is modified without also specifying it as an output
3902 operand. Note that if all the output operands you specify are for this
3903 purpose (and hence unused), you will then also need to specify
3904 @code{volatile} for the @code{asm} construct, as described below, to
3905 prevent GCC from deleting the @code{asm} statement as unused.
3907 If you refer to a particular hardware register from the assembler code,
3908 you will probably have to list the register after the third colon to
3909 tell the compiler the register's value is modified. In some assemblers,
3910 the register names begin with @samp{%}; to produce one @samp{%} in the
3911 assembler code, you must write @samp{%%} in the input.
3913 If your assembler instruction can alter the condition code register, add
3914 @samp{cc} to the list of clobbered registers. GCC on some machines
3915 represents the condition codes as a specific hardware register;
3916 @samp{cc} serves to name this register. On other machines, the
3917 condition code is handled differently, and specifying @samp{cc} has no
3918 effect. But it is valid no matter what the machine.
3920 If your assembler instructions access memory in an unpredictable
3921 fashion, add @samp{memory} to the list of clobbered registers. This
3922 will cause GCC to not keep memory values cached in registers across the
3923 assembler instruction and not optimize stores or loads to that memory.
3924 You will also want to add the @code{volatile} keyword if the memory
3925 affected is not listed in the inputs or outputs of the @code{asm}, as
3926 the @samp{memory} clobber does not count as a side-effect of the
3927 @code{asm}. If you know how large the accessed memory is, you can add
3928 it as input or output but if this is not known, you should add
3929 @samp{memory}. As an example, if you access ten bytes of a string, you
3930 can use a memory input like:
3933 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
3936 Note that in the following example the memory input is necessary,
3937 otherwise GCC might optimize the store to @code{x} away:
3944 asm ("magic stuff accessing an 'int' pointed to by '%1'"
3945 "=&d" (r) : "a" (y), "m" (*y));
3950 You can put multiple assembler instructions together in a single
3951 @code{asm} template, separated by the characters normally used in assembly
3952 code for the system. A combination that works in most places is a newline
3953 to break the line, plus a tab character to move to the instruction field
3954 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3955 assembler allows semicolons as a line-breaking character. Note that some
3956 assembler dialects use semicolons to start a comment.
3957 The input operands are guaranteed not to use any of the clobbered
3958 registers, and neither will the output operands' addresses, so you can
3959 read and write the clobbered registers as many times as you like. Here
3960 is an example of multiple instructions in a template; it assumes the
3961 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3964 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3966 : "g" (from), "g" (to)
3970 Unless an output operand has the @samp{&} constraint modifier, GCC
3971 may allocate it in the same register as an unrelated input operand, on
3972 the assumption the inputs are consumed before the outputs are produced.
3973 This assumption may be false if the assembler code actually consists of
3974 more than one instruction. In such a case, use @samp{&} for each output
3975 operand that may not overlap an input. @xref{Modifiers}.
3977 If you want to test the condition code produced by an assembler
3978 instruction, you must include a branch and a label in the @code{asm}
3979 construct, as follows:
3982 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3988 This assumes your assembler supports local labels, as the GNU assembler
3989 and most Unix assemblers do.
3991 Speaking of labels, jumps from one @code{asm} to another are not
3992 supported. The compiler's optimizers do not know about these jumps, and
3993 therefore they cannot take account of them when deciding how to
3996 @cindex macros containing @code{asm}
3997 Usually the most convenient way to use these @code{asm} instructions is to
3998 encapsulate them in macros that look like functions. For example,
4002 (@{ double __value, __arg = (x); \
4003 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4008 Here the variable @code{__arg} is used to make sure that the instruction
4009 operates on a proper @code{double} value, and to accept only those
4010 arguments @code{x} which can convert automatically to a @code{double}.
4012 Another way to make sure the instruction operates on the correct data
4013 type is to use a cast in the @code{asm}. This is different from using a
4014 variable @code{__arg} in that it converts more different types. For
4015 example, if the desired type were @code{int}, casting the argument to
4016 @code{int} would accept a pointer with no complaint, while assigning the
4017 argument to an @code{int} variable named @code{__arg} would warn about
4018 using a pointer unless the caller explicitly casts it.
4020 If an @code{asm} has output operands, GCC assumes for optimization
4021 purposes the instruction has no side effects except to change the output
4022 operands. This does not mean instructions with a side effect cannot be
4023 used, but you must be careful, because the compiler may eliminate them
4024 if the output operands aren't used, or move them out of loops, or
4025 replace two with one if they constitute a common subexpression. Also,
4026 if your instruction does have a side effect on a variable that otherwise
4027 appears not to change, the old value of the variable may be reused later
4028 if it happens to be found in a register.
4030 You can prevent an @code{asm} instruction from being deleted, moved
4031 significantly, or combined, by writing the keyword @code{volatile} after
4032 the @code{asm}. For example:
4035 #define get_and_set_priority(new) \
4037 asm volatile ("get_and_set_priority %0, %1" \
4038 : "=g" (__old) : "g" (new)); \
4043 If you write an @code{asm} instruction with no outputs, GCC will know
4044 the instruction has side-effects and will not delete the instruction or
4045 move it outside of loops.
4047 The @code{volatile} keyword indicates that the instruction has
4048 important side-effects. GCC will not delete a volatile @code{asm} if
4049 it is reachable. (The instruction can still be deleted if GCC can
4050 prove that control-flow will never reach the location of the
4051 instruction.) In addition, GCC will not reschedule instructions
4052 across a volatile @code{asm} instruction. For example:
4055 *(volatile int *)addr = foo;
4056 asm volatile ("eieio" : : );
4060 Assume @code{addr} contains the address of a memory mapped device
4061 register. The PowerPC @code{eieio} instruction (Enforce In-order
4062 Execution of I/O) tells the CPU to make sure that the store to that
4063 device register happens before it issues any other I/O@.
4065 Note that even a volatile @code{asm} instruction can be moved in ways
4066 that appear insignificant to the compiler, such as across jump
4067 instructions. You can't expect a sequence of volatile @code{asm}
4068 instructions to remain perfectly consecutive. If you want consecutive
4069 output, use a single @code{asm}. Also, GCC will perform some
4070 optimizations across a volatile @code{asm} instruction; GCC does not
4071 ``forget everything'' when it encounters a volatile @code{asm}
4072 instruction the way some other compilers do.
4074 An @code{asm} instruction without any operands or clobbers (an ``old
4075 style'' @code{asm}) will be treated identically to a volatile
4076 @code{asm} instruction.
4078 It is a natural idea to look for a way to give access to the condition
4079 code left by the assembler instruction. However, when we attempted to
4080 implement this, we found no way to make it work reliably. The problem
4081 is that output operands might need reloading, which would result in
4082 additional following ``store'' instructions. On most machines, these
4083 instructions would alter the condition code before there was time to
4084 test it. This problem doesn't arise for ordinary ``test'' and
4085 ``compare'' instructions because they don't have any output operands.
4087 For reasons similar to those described above, it is not possible to give
4088 an assembler instruction access to the condition code left by previous
4091 If you are writing a header file that should be includable in ISO C
4092 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4095 @subsection Size of an @code{asm}
4097 Some targets require that GCC track the size of each instruction used in
4098 order to generate correct code. Because the final length of an
4099 @code{asm} is only known by the assembler, GCC must make an estimate as
4100 to how big it will be. The estimate is formed by counting the number of
4101 statements in the pattern of the @code{asm} and multiplying that by the
4102 length of the longest instruction on that processor. Statements in the
4103 @code{asm} are identified by newline characters and whatever statement
4104 separator characters are supported by the assembler; on most processors
4105 this is the `@code{;}' character.
4107 Normally, GCC's estimate is perfectly adequate to ensure that correct
4108 code is generated, but it is possible to confuse the compiler if you use
4109 pseudo instructions or assembler macros that expand into multiple real
4110 instructions or if you use assembler directives that expand to more
4111 space in the object file than would be needed for a single instruction.
4112 If this happens then the assembler will produce a diagnostic saying that
4113 a label is unreachable.
4115 @subsection i386 floating point asm operands
4117 There are several rules on the usage of stack-like regs in
4118 asm_operands insns. These rules apply only to the operands that are
4123 Given a set of input regs that die in an asm_operands, it is
4124 necessary to know which are implicitly popped by the asm, and
4125 which must be explicitly popped by gcc.
4127 An input reg that is implicitly popped by the asm must be
4128 explicitly clobbered, unless it is constrained to match an
4132 For any input reg that is implicitly popped by an asm, it is
4133 necessary to know how to adjust the stack to compensate for the pop.
4134 If any non-popped input is closer to the top of the reg-stack than
4135 the implicitly popped reg, it would not be possible to know what the
4136 stack looked like---it's not clear how the rest of the stack ``slides
4139 All implicitly popped input regs must be closer to the top of
4140 the reg-stack than any input that is not implicitly popped.
4142 It is possible that if an input dies in an insn, reload might
4143 use the input reg for an output reload. Consider this example:
4146 asm ("foo" : "=t" (a) : "f" (b));
4149 This asm says that input B is not popped by the asm, and that
4150 the asm pushes a result onto the reg-stack, i.e., the stack is one
4151 deeper after the asm than it was before. But, it is possible that
4152 reload will think that it can use the same reg for both the input and
4153 the output, if input B dies in this insn.
4155 If any input operand uses the @code{f} constraint, all output reg
4156 constraints must use the @code{&} earlyclobber.
4158 The asm above would be written as
4161 asm ("foo" : "=&t" (a) : "f" (b));
4165 Some operands need to be in particular places on the stack. All
4166 output operands fall in this category---there is no other way to
4167 know which regs the outputs appear in unless the user indicates
4168 this in the constraints.
4170 Output operands must specifically indicate which reg an output
4171 appears in after an asm. @code{=f} is not allowed: the operand
4172 constraints must select a class with a single reg.
4175 Output operands may not be ``inserted'' between existing stack regs.
4176 Since no 387 opcode uses a read/write operand, all output operands
4177 are dead before the asm_operands, and are pushed by the asm_operands.
4178 It makes no sense to push anywhere but the top of the reg-stack.
4180 Output operands must start at the top of the reg-stack: output
4181 operands may not ``skip'' a reg.
4184 Some asm statements may need extra stack space for internal
4185 calculations. This can be guaranteed by clobbering stack registers
4186 unrelated to the inputs and outputs.
4190 Here are a couple of reasonable asms to want to write. This asm
4191 takes one input, which is internally popped, and produces two outputs.
4194 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4197 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4198 and replaces them with one output. The user must code the @code{st(1)}
4199 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4202 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4208 @section Controlling Names Used in Assembler Code
4209 @cindex assembler names for identifiers
4210 @cindex names used in assembler code
4211 @cindex identifiers, names in assembler code
4213 You can specify the name to be used in the assembler code for a C
4214 function or variable by writing the @code{asm} (or @code{__asm__})
4215 keyword after the declarator as follows:
4218 int foo asm ("myfoo") = 2;
4222 This specifies that the name to be used for the variable @code{foo} in
4223 the assembler code should be @samp{myfoo} rather than the usual
4226 On systems where an underscore is normally prepended to the name of a C
4227 function or variable, this feature allows you to define names for the
4228 linker that do not start with an underscore.
4230 It does not make sense to use this feature with a non-static local
4231 variable since such variables do not have assembler names. If you are
4232 trying to put the variable in a particular register, see @ref{Explicit
4233 Reg Vars}. GCC presently accepts such code with a warning, but will
4234 probably be changed to issue an error, rather than a warning, in the
4237 You cannot use @code{asm} in this way in a function @emph{definition}; but
4238 you can get the same effect by writing a declaration for the function
4239 before its definition and putting @code{asm} there, like this:
4242 extern func () asm ("FUNC");
4249 It is up to you to make sure that the assembler names you choose do not
4250 conflict with any other assembler symbols. Also, you must not use a
4251 register name; that would produce completely invalid assembler code. GCC
4252 does not as yet have the ability to store static variables in registers.
4253 Perhaps that will be added.
4255 @node Explicit Reg Vars
4256 @section Variables in Specified Registers
4257 @cindex explicit register variables
4258 @cindex variables in specified registers
4259 @cindex specified registers
4260 @cindex registers, global allocation
4262 GNU C allows you to put a few global variables into specified hardware
4263 registers. You can also specify the register in which an ordinary
4264 register variable should be allocated.
4268 Global register variables reserve registers throughout the program.
4269 This may be useful in programs such as programming language
4270 interpreters which have a couple of global variables that are accessed
4274 Local register variables in specific registers do not reserve the
4275 registers. The compiler's data flow analysis is capable of determining
4276 where the specified registers contain live values, and where they are
4277 available for other uses. Stores into local register variables may be deleted
4278 when they appear to be dead according to dataflow analysis. References
4279 to local register variables may be deleted or moved or simplified.
4281 These local variables are sometimes convenient for use with the extended
4282 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4283 output of the assembler instruction directly into a particular register.
4284 (This will work provided the register you specify fits the constraints
4285 specified for that operand in the @code{asm}.)
4293 @node Global Reg Vars
4294 @subsection Defining Global Register Variables
4295 @cindex global register variables
4296 @cindex registers, global variables in
4298 You can define a global register variable in GNU C like this:
4301 register int *foo asm ("a5");
4305 Here @code{a5} is the name of the register which should be used. Choose a
4306 register which is normally saved and restored by function calls on your
4307 machine, so that library routines will not clobber it.
4309 Naturally the register name is cpu-dependent, so you would need to
4310 conditionalize your program according to cpu type. The register
4311 @code{a5} would be a good choice on a 68000 for a variable of pointer
4312 type. On machines with register windows, be sure to choose a ``global''
4313 register that is not affected magically by the function call mechanism.
4315 In addition, operating systems on one type of cpu may differ in how they
4316 name the registers; then you would need additional conditionals. For
4317 example, some 68000 operating systems call this register @code{%a5}.
4319 Eventually there may be a way of asking the compiler to choose a register
4320 automatically, but first we need to figure out how it should choose and
4321 how to enable you to guide the choice. No solution is evident.
4323 Defining a global register variable in a certain register reserves that
4324 register entirely for this use, at least within the current compilation.
4325 The register will not be allocated for any other purpose in the functions
4326 in the current compilation. The register will not be saved and restored by
4327 these functions. Stores into this register are never deleted even if they
4328 would appear to be dead, but references may be deleted or moved or
4331 It is not safe to access the global register variables from signal
4332 handlers, or from more than one thread of control, because the system
4333 library routines may temporarily use the register for other things (unless
4334 you recompile them specially for the task at hand).
4336 @cindex @code{qsort}, and global register variables
4337 It is not safe for one function that uses a global register variable to
4338 call another such function @code{foo} by way of a third function
4339 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4340 different source file in which the variable wasn't declared). This is
4341 because @code{lose} might save the register and put some other value there.
4342 For example, you can't expect a global register variable to be available in
4343 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4344 might have put something else in that register. (If you are prepared to
4345 recompile @code{qsort} with the same global register variable, you can
4346 solve this problem.)
4348 If you want to recompile @code{qsort} or other source files which do not
4349 actually use your global register variable, so that they will not use that
4350 register for any other purpose, then it suffices to specify the compiler
4351 option @option{-ffixed-@var{reg}}. You need not actually add a global
4352 register declaration to their source code.
4354 A function which can alter the value of a global register variable cannot
4355 safely be called from a function compiled without this variable, because it
4356 could clobber the value the caller expects to find there on return.
4357 Therefore, the function which is the entry point into the part of the
4358 program that uses the global register variable must explicitly save and
4359 restore the value which belongs to its caller.
4361 @cindex register variable after @code{longjmp}
4362 @cindex global register after @code{longjmp}
4363 @cindex value after @code{longjmp}
4366 On most machines, @code{longjmp} will restore to each global register
4367 variable the value it had at the time of the @code{setjmp}. On some
4368 machines, however, @code{longjmp} will not change the value of global
4369 register variables. To be portable, the function that called @code{setjmp}
4370 should make other arrangements to save the values of the global register
4371 variables, and to restore them in a @code{longjmp}. This way, the same
4372 thing will happen regardless of what @code{longjmp} does.
4374 All global register variable declarations must precede all function
4375 definitions. If such a declaration could appear after function
4376 definitions, the declaration would be too late to prevent the register from
4377 being used for other purposes in the preceding functions.
4379 Global register variables may not have initial values, because an
4380 executable file has no means to supply initial contents for a register.
4382 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4383 registers, but certain library functions, such as @code{getwd}, as well
4384 as the subroutines for division and remainder, modify g3 and g4. g1 and
4385 g2 are local temporaries.
4387 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4388 Of course, it will not do to use more than a few of those.
4390 @node Local Reg Vars
4391 @subsection Specifying Registers for Local Variables
4392 @cindex local variables, specifying registers
4393 @cindex specifying registers for local variables
4394 @cindex registers for local variables
4396 You can define a local register variable with a specified register
4400 register int *foo asm ("a5");
4404 Here @code{a5} is the name of the register which should be used. Note
4405 that this is the same syntax used for defining global register
4406 variables, but for a local variable it would appear within a function.
4408 Naturally the register name is cpu-dependent, but this is not a
4409 problem, since specific registers are most often useful with explicit
4410 assembler instructions (@pxref{Extended Asm}). Both of these things
4411 generally require that you conditionalize your program according to
4414 In addition, operating systems on one type of cpu may differ in how they
4415 name the registers; then you would need additional conditionals. For
4416 example, some 68000 operating systems call this register @code{%a5}.
4418 Defining such a register variable does not reserve the register; it
4419 remains available for other uses in places where flow control determines
4420 the variable's value is not live.
4422 This option does not guarantee that GCC will generate code that has
4423 this variable in the register you specify at all times. You may not
4424 code an explicit reference to this register in an @code{asm} statement
4425 and assume it will always refer to this variable.
4427 Stores into local register variables may be deleted when they appear to be dead
4428 according to dataflow analysis. References to local register variables may
4429 be deleted or moved or simplified.
4431 @node Alternate Keywords
4432 @section Alternate Keywords
4433 @cindex alternate keywords
4434 @cindex keywords, alternate
4436 @option{-ansi} and the various @option{-std} options disable certain
4437 keywords. This causes trouble when you want to use GNU C extensions, or
4438 a general-purpose header file that should be usable by all programs,
4439 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4440 @code{inline} are not available in programs compiled with
4441 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4442 program compiled with @option{-std=c99}). The ISO C99 keyword
4443 @code{restrict} is only available when @option{-std=gnu99} (which will
4444 eventually be the default) or @option{-std=c99} (or the equivalent
4445 @option{-std=iso9899:1999}) is used.
4447 The way to solve these problems is to put @samp{__} at the beginning and
4448 end of each problematical keyword. For example, use @code{__asm__}
4449 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4451 Other C compilers won't accept these alternative keywords; if you want to
4452 compile with another compiler, you can define the alternate keywords as
4453 macros to replace them with the customary keywords. It looks like this:
4461 @findex __extension__
4463 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4465 prevent such warnings within one expression by writing
4466 @code{__extension__} before the expression. @code{__extension__} has no
4467 effect aside from this.
4469 @node Incomplete Enums
4470 @section Incomplete @code{enum} Types
4472 You can define an @code{enum} tag without specifying its possible values.
4473 This results in an incomplete type, much like what you get if you write
4474 @code{struct foo} without describing the elements. A later declaration
4475 which does specify the possible values completes the type.
4477 You can't allocate variables or storage using the type while it is
4478 incomplete. However, you can work with pointers to that type.
4480 This extension may not be very useful, but it makes the handling of
4481 @code{enum} more consistent with the way @code{struct} and @code{union}
4484 This extension is not supported by GNU C++.
4486 @node Function Names
4487 @section Function Names as Strings
4488 @cindex @code{__func__} identifier
4489 @cindex @code{__FUNCTION__} identifier
4490 @cindex @code{__PRETTY_FUNCTION__} identifier
4492 GCC provides three magic variables which hold the name of the current
4493 function, as a string. The first of these is @code{__func__}, which
4494 is part of the C99 standard:
4497 The identifier @code{__func__} is implicitly declared by the translator
4498 as if, immediately following the opening brace of each function
4499 definition, the declaration
4502 static const char __func__[] = "function-name";
4505 appeared, where function-name is the name of the lexically-enclosing
4506 function. This name is the unadorned name of the function.
4509 @code{__FUNCTION__} is another name for @code{__func__}. Older
4510 versions of GCC recognize only this name. However, it is not
4511 standardized. For maximum portability, we recommend you use
4512 @code{__func__}, but provide a fallback definition with the
4516 #if __STDC_VERSION__ < 199901L
4518 # define __func__ __FUNCTION__
4520 # define __func__ "<unknown>"
4525 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4526 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4527 the type signature of the function as well as its bare name. For
4528 example, this program:
4532 extern int printf (char *, ...);
4539 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4540 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4558 __PRETTY_FUNCTION__ = void a::sub(int)
4561 These identifiers are not preprocessor macros. In GCC 3.3 and
4562 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4563 were treated as string literals; they could be used to initialize
4564 @code{char} arrays, and they could be concatenated with other string
4565 literals. GCC 3.4 and later treat them as variables, like
4566 @code{__func__}. In C++, @code{__FUNCTION__} and
4567 @code{__PRETTY_FUNCTION__} have always been variables.
4569 @node Return Address
4570 @section Getting the Return or Frame Address of a Function
4572 These functions may be used to get information about the callers of a
4575 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4576 This function returns the return address of the current function, or of
4577 one of its callers. The @var{level} argument is number of frames to
4578 scan up the call stack. A value of @code{0} yields the return address
4579 of the current function, a value of @code{1} yields the return address
4580 of the caller of the current function, and so forth. When inlining
4581 the expected behavior is that the function will return the address of
4582 the function that will be returned to. To work around this behavior use
4583 the @code{noinline} function attribute.
4585 The @var{level} argument must be a constant integer.
4587 On some machines it may be impossible to determine the return address of
4588 any function other than the current one; in such cases, or when the top
4589 of the stack has been reached, this function will return @code{0} or a
4590 random value. In addition, @code{__builtin_frame_address} may be used
4591 to determine if the top of the stack has been reached.
4593 This function should only be used with a nonzero argument for debugging
4597 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4598 This function is similar to @code{__builtin_return_address}, but it
4599 returns the address of the function frame rather than the return address
4600 of the function. Calling @code{__builtin_frame_address} with a value of
4601 @code{0} yields the frame address of the current function, a value of
4602 @code{1} yields the frame address of the caller of the current function,
4605 The frame is the area on the stack which holds local variables and saved
4606 registers. The frame address is normally the address of the first word
4607 pushed on to the stack by the function. However, the exact definition
4608 depends upon the processor and the calling convention. If the processor
4609 has a dedicated frame pointer register, and the function has a frame,
4610 then @code{__builtin_frame_address} will return the value of the frame
4613 On some machines it may be impossible to determine the frame address of
4614 any function other than the current one; in such cases, or when the top
4615 of the stack has been reached, this function will return @code{0} if
4616 the first frame pointer is properly initialized by the startup code.
4618 This function should only be used with a nonzero argument for debugging
4622 @node Vector Extensions
4623 @section Using vector instructions through built-in functions
4625 On some targets, the instruction set contains SIMD vector instructions that
4626 operate on multiple values contained in one large register at the same time.
4627 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4630 The first step in using these extensions is to provide the necessary data
4631 types. This should be done using an appropriate @code{typedef}:
4634 typedef int v4si __attribute__ ((vector_size (16)));
4637 The @code{int} type specifies the base type, while the attribute specifies
4638 the vector size for the variable, measured in bytes. For example, the
4639 declaration above causes the compiler to set the mode for the @code{v4si}
4640 type to be 16 bytes wide and divided into @code{int} sized units. For
4641 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4642 corresponding mode of @code{foo} will be @acronym{V4SI}.
4644 The @code{vector_size} attribute is only applicable to integral and
4645 float scalars, although arrays, pointers, and function return values
4646 are allowed in conjunction with this construct.
4648 All the basic integer types can be used as base types, both as signed
4649 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4650 @code{long long}. In addition, @code{float} and @code{double} can be
4651 used to build floating-point vector types.
4653 Specifying a combination that is not valid for the current architecture
4654 will cause GCC to synthesize the instructions using a narrower mode.
4655 For example, if you specify a variable of type @code{V4SI} and your
4656 architecture does not allow for this specific SIMD type, GCC will
4657 produce code that uses 4 @code{SIs}.
4659 The types defined in this manner can be used with a subset of normal C
4660 operations. Currently, GCC will allow using the following operators
4661 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4663 The operations behave like C++ @code{valarrays}. Addition is defined as
4664 the addition of the corresponding elements of the operands. For
4665 example, in the code below, each of the 4 elements in @var{a} will be
4666 added to the corresponding 4 elements in @var{b} and the resulting
4667 vector will be stored in @var{c}.
4670 typedef int v4si __attribute__ ((vector_size (16)));
4677 Subtraction, multiplication, division, and the logical operations
4678 operate in a similar manner. Likewise, the result of using the unary
4679 minus or complement operators on a vector type is a vector whose
4680 elements are the negative or complemented values of the corresponding
4681 elements in the operand.
4683 You can declare variables and use them in function calls and returns, as
4684 well as in assignments and some casts. You can specify a vector type as
4685 a return type for a function. Vector types can also be used as function
4686 arguments. It is possible to cast from one vector type to another,
4687 provided they are of the same size (in fact, you can also cast vectors
4688 to and from other datatypes of the same size).
4690 You cannot operate between vectors of different lengths or different
4691 signedness without a cast.
4693 A port that supports hardware vector operations, usually provides a set
4694 of built-in functions that can be used to operate on vectors. For
4695 example, a function to add two vectors and multiply the result by a
4696 third could look like this:
4699 v4si f (v4si a, v4si b, v4si c)
4701 v4si tmp = __builtin_addv4si (a, b);
4702 return __builtin_mulv4si (tmp, c);
4707 @node Other Builtins
4708 @section Other built-in functions provided by GCC
4709 @cindex built-in functions
4710 @findex __builtin_isgreater
4711 @findex __builtin_isgreaterequal
4712 @findex __builtin_isless
4713 @findex __builtin_islessequal
4714 @findex __builtin_islessgreater
4715 @findex __builtin_isunordered
4870 @findex fprintf_unlocked
4872 @findex fputs_unlocked
4982 @findex printf_unlocked
5011 @findex significandf
5012 @findex significandl
5079 GCC provides a large number of built-in functions other than the ones
5080 mentioned above. Some of these are for internal use in the processing
5081 of exceptions or variable-length argument lists and will not be
5082 documented here because they may change from time to time; we do not
5083 recommend general use of these functions.
5085 The remaining functions are provided for optimization purposes.
5087 @opindex fno-builtin
5088 GCC includes built-in versions of many of the functions in the standard
5089 C library. The versions prefixed with @code{__builtin_} will always be
5090 treated as having the same meaning as the C library function even if you
5091 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5092 Many of these functions are only optimized in certain cases; if they are
5093 not optimized in a particular case, a call to the library function will
5098 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5099 @option{-std=c99}), the functions
5100 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5101 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5102 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5103 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5104 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5105 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5106 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5107 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5108 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5109 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5110 @code{significandf}, @code{significandl}, @code{significand},
5111 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5112 @code{strdup}, @code{strfmon}, @code{toascii}, @code{y0f}, @code{y0l},
5113 @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
5115 may be handled as built-in functions.
5116 All these functions have corresponding versions
5117 prefixed with @code{__builtin_}, which may be used even in strict C89
5120 The ISO C99 functions
5121 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5122 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5123 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5124 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5125 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5126 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5127 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5128 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5129 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5130 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5131 @code{cimagl}, @code{cimag}, @code{conjf}, @code{conjl}, @code{conj},
5132 @code{copysignf}, @code{copysignl}, @code{copysign}, @code{cpowf},
5133 @code{cpowl}, @code{cpow}, @code{cprojf}, @code{cprojl}, @code{cproj},
5134 @code{crealf}, @code{creall}, @code{creal}, @code{csinf}, @code{csinhf},
5135 @code{csinhl}, @code{csinh}, @code{csinl}, @code{csin}, @code{csqrtf},
5136 @code{csqrtl}, @code{csqrt}, @code{ctanf}, @code{ctanhf}, @code{ctanhl},
5137 @code{ctanh}, @code{ctanl}, @code{ctan}, @code{erfcf}, @code{erfcl},
5138 @code{erfc}, @code{erff}, @code{erfl}, @code{erf}, @code{exp2f},
5139 @code{exp2l}, @code{exp2}, @code{expm1f}, @code{expm1l}, @code{expm1},
5140 @code{fdimf}, @code{fdiml}, @code{fdim}, @code{fmaf}, @code{fmal},
5141 @code{fmaxf}, @code{fmaxl}, @code{fmax}, @code{fma}, @code{fminf},
5142 @code{fminl}, @code{fmin}, @code{hypotf}, @code{hypotl}, @code{hypot},
5143 @code{ilogbf}, @code{ilogbl}, @code{ilogb}, @code{imaxabs},
5144 @code{isblank}, @code{iswblank}, @code{lgammaf}, @code{lgammal},
5145 @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5146 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5147 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5148 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5149 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5150 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5151 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5152 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5153 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5154 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5155 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5156 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5157 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5158 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5159 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5160 are handled as built-in functions
5161 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5163 There are also built-in versions of the ISO C99 functions
5164 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5165 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5166 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5167 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5168 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5169 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5170 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5171 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5172 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5173 that are recognized in any mode since ISO C90 reserves these names for
5174 the purpose to which ISO C99 puts them. All these functions have
5175 corresponding versions prefixed with @code{__builtin_}.
5177 The ISO C94 functions
5178 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5179 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5180 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5182 are handled as built-in functions
5183 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5185 The ISO C90 functions
5186 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5187 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5188 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5189 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5190 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5191 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5192 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5193 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5194 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5195 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5196 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5197 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5198 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5199 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5200 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5201 @code{vprintf} and @code{vsprintf}
5202 are all recognized as built-in functions unless
5203 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5204 is specified for an individual function). All of these functions have
5205 corresponding versions prefixed with @code{__builtin_}.
5207 GCC provides built-in versions of the ISO C99 floating point comparison
5208 macros that avoid raising exceptions for unordered operands. They have
5209 the same names as the standard macros ( @code{isgreater},
5210 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5211 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5212 prefixed. We intend for a library implementor to be able to simply
5213 @code{#define} each standard macro to its built-in equivalent.
5215 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5217 You can use the built-in function @code{__builtin_types_compatible_p} to
5218 determine whether two types are the same.
5220 This built-in function returns 1 if the unqualified versions of the
5221 types @var{type1} and @var{type2} (which are types, not expressions) are
5222 compatible, 0 otherwise. The result of this built-in function can be
5223 used in integer constant expressions.
5225 This built-in function ignores top level qualifiers (e.g., @code{const},
5226 @code{volatile}). For example, @code{int} is equivalent to @code{const
5229 The type @code{int[]} and @code{int[5]} are compatible. On the other
5230 hand, @code{int} and @code{char *} are not compatible, even if the size
5231 of their types, on the particular architecture are the same. Also, the
5232 amount of pointer indirection is taken into account when determining
5233 similarity. Consequently, @code{short *} is not similar to
5234 @code{short **}. Furthermore, two types that are typedefed are
5235 considered compatible if their underlying types are compatible.
5237 An @code{enum} type is not considered to be compatible with another
5238 @code{enum} type even if both are compatible with the same integer
5239 type; this is what the C standard specifies.
5240 For example, @code{enum @{foo, bar@}} is not similar to
5241 @code{enum @{hot, dog@}}.
5243 You would typically use this function in code whose execution varies
5244 depending on the arguments' types. For example:
5250 if (__builtin_types_compatible_p (typeof (x), long double)) \
5251 tmp = foo_long_double (tmp); \
5252 else if (__builtin_types_compatible_p (typeof (x), double)) \
5253 tmp = foo_double (tmp); \
5254 else if (__builtin_types_compatible_p (typeof (x), float)) \
5255 tmp = foo_float (tmp); \
5262 @emph{Note:} This construct is only available for C.
5266 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5268 You can use the built-in function @code{__builtin_choose_expr} to
5269 evaluate code depending on the value of a constant expression. This
5270 built-in function returns @var{exp1} if @var{const_exp}, which is a
5271 constant expression that must be able to be determined at compile time,
5272 is nonzero. Otherwise it returns 0.
5274 This built-in function is analogous to the @samp{? :} operator in C,
5275 except that the expression returned has its type unaltered by promotion
5276 rules. Also, the built-in function does not evaluate the expression
5277 that was not chosen. For example, if @var{const_exp} evaluates to true,
5278 @var{exp2} is not evaluated even if it has side-effects.
5280 This built-in function can return an lvalue if the chosen argument is an
5283 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5284 type. Similarly, if @var{exp2} is returned, its return type is the same
5291 __builtin_choose_expr ( \
5292 __builtin_types_compatible_p (typeof (x), double), \
5294 __builtin_choose_expr ( \
5295 __builtin_types_compatible_p (typeof (x), float), \
5297 /* @r{The void expression results in a compile-time error} \
5298 @r{when assigning the result to something.} */ \
5302 @emph{Note:} This construct is only available for C. Furthermore, the
5303 unused expression (@var{exp1} or @var{exp2} depending on the value of
5304 @var{const_exp}) may still generate syntax errors. This may change in
5309 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5310 You can use the built-in function @code{__builtin_constant_p} to
5311 determine if a value is known to be constant at compile-time and hence
5312 that GCC can perform constant-folding on expressions involving that
5313 value. The argument of the function is the value to test. The function
5314 returns the integer 1 if the argument is known to be a compile-time
5315 constant and 0 if it is not known to be a compile-time constant. A
5316 return of 0 does not indicate that the value is @emph{not} a constant,
5317 but merely that GCC cannot prove it is a constant with the specified
5318 value of the @option{-O} option.
5320 You would typically use this function in an embedded application where
5321 memory was a critical resource. If you have some complex calculation,
5322 you may want it to be folded if it involves constants, but need to call
5323 a function if it does not. For example:
5326 #define Scale_Value(X) \
5327 (__builtin_constant_p (X) \
5328 ? ((X) * SCALE + OFFSET) : Scale (X))
5331 You may use this built-in function in either a macro or an inline
5332 function. However, if you use it in an inlined function and pass an
5333 argument of the function as the argument to the built-in, GCC will
5334 never return 1 when you call the inline function with a string constant
5335 or compound literal (@pxref{Compound Literals}) and will not return 1
5336 when you pass a constant numeric value to the inline function unless you
5337 specify the @option{-O} option.
5339 You may also use @code{__builtin_constant_p} in initializers for static
5340 data. For instance, you can write
5343 static const int table[] = @{
5344 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5350 This is an acceptable initializer even if @var{EXPRESSION} is not a
5351 constant expression. GCC must be more conservative about evaluating the
5352 built-in in this case, because it has no opportunity to perform
5355 Previous versions of GCC did not accept this built-in in data
5356 initializers. The earliest version where it is completely safe is
5360 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5361 @opindex fprofile-arcs
5362 You may use @code{__builtin_expect} to provide the compiler with
5363 branch prediction information. In general, you should prefer to
5364 use actual profile feedback for this (@option{-fprofile-arcs}), as
5365 programmers are notoriously bad at predicting how their programs
5366 actually perform. However, there are applications in which this
5367 data is hard to collect.
5369 The return value is the value of @var{exp}, which should be an
5370 integral expression. The value of @var{c} must be a compile-time
5371 constant. The semantics of the built-in are that it is expected
5372 that @var{exp} == @var{c}. For example:
5375 if (__builtin_expect (x, 0))
5380 would indicate that we do not expect to call @code{foo}, since
5381 we expect @code{x} to be zero. Since you are limited to integral
5382 expressions for @var{exp}, you should use constructions such as
5385 if (__builtin_expect (ptr != NULL, 1))
5390 when testing pointer or floating-point values.
5393 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5394 This function is used to minimize cache-miss latency by moving data into
5395 a cache before it is accessed.
5396 You can insert calls to @code{__builtin_prefetch} into code for which
5397 you know addresses of data in memory that is likely to be accessed soon.
5398 If the target supports them, data prefetch instructions will be generated.
5399 If the prefetch is done early enough before the access then the data will
5400 be in the cache by the time it is accessed.
5402 The value of @var{addr} is the address of the memory to prefetch.
5403 There are two optional arguments, @var{rw} and @var{locality}.
5404 The value of @var{rw} is a compile-time constant one or zero; one
5405 means that the prefetch is preparing for a write to the memory address
5406 and zero, the default, means that the prefetch is preparing for a read.
5407 The value @var{locality} must be a compile-time constant integer between
5408 zero and three. A value of zero means that the data has no temporal
5409 locality, so it need not be left in the cache after the access. A value
5410 of three means that the data has a high degree of temporal locality and
5411 should be left in all levels of cache possible. Values of one and two
5412 mean, respectively, a low or moderate degree of temporal locality. The
5416 for (i = 0; i < n; i++)
5419 __builtin_prefetch (&a[i+j], 1, 1);
5420 __builtin_prefetch (&b[i+j], 0, 1);
5425 Data prefetch does not generate faults if @var{addr} is invalid, but
5426 the address expression itself must be valid. For example, a prefetch
5427 of @code{p->next} will not fault if @code{p->next} is not a valid
5428 address, but evaluation will fault if @code{p} is not a valid address.
5430 If the target does not support data prefetch, the address expression
5431 is evaluated if it includes side effects but no other code is generated
5432 and GCC does not issue a warning.
5435 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5436 Returns a positive infinity, if supported by the floating-point format,
5437 else @code{DBL_MAX}. This function is suitable for implementing the
5438 ISO C macro @code{HUGE_VAL}.
5441 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5442 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5445 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5446 Similar to @code{__builtin_huge_val}, except the return
5447 type is @code{long double}.
5450 @deftypefn {Built-in Function} double __builtin_inf (void)
5451 Similar to @code{__builtin_huge_val}, except a warning is generated
5452 if the target floating-point format does not support infinities.
5453 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5456 @deftypefn {Built-in Function} float __builtin_inff (void)
5457 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5460 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5461 Similar to @code{__builtin_inf}, except the return
5462 type is @code{long double}.
5465 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5466 This is an implementation of the ISO C99 function @code{nan}.
5468 Since ISO C99 defines this function in terms of @code{strtod}, which we
5469 do not implement, a description of the parsing is in order. The string
5470 is parsed as by @code{strtol}; that is, the base is recognized by
5471 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5472 in the significand such that the least significant bit of the number
5473 is at the least significant bit of the significand. The number is
5474 truncated to fit the significand field provided. The significand is
5475 forced to be a quiet NaN.
5477 This function, if given a string literal, is evaluated early enough
5478 that it is considered a compile-time constant.
5481 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5482 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5485 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5486 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5489 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5490 Similar to @code{__builtin_nan}, except the significand is forced
5491 to be a signaling NaN. The @code{nans} function is proposed by
5492 @uref{http://std.dkuug.dk/JTC1/SC22/WG14/www/docs/n965.htm,,WG14 N965}.
5495 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5496 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5499 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5500 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5503 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5504 Returns one plus the index of the least significant 1-bit of @var{x}, or
5505 if @var{x} is zero, returns zero.
5508 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5509 Returns the number of leading 0-bits in @var{x}, starting at the most
5510 significant bit position. If @var{x} is 0, the result is undefined.
5513 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5514 Returns the number of trailing 0-bits in @var{x}, starting at the least
5515 significant bit position. If @var{x} is 0, the result is undefined.
5518 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5519 Returns the number of 1-bits in @var{x}.
5522 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5523 Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5527 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5528 Similar to @code{__builtin_ffs}, except the argument type is
5529 @code{unsigned long}.
5532 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5533 Similar to @code{__builtin_clz}, except the argument type is
5534 @code{unsigned long}.
5537 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5538 Similar to @code{__builtin_ctz}, except the argument type is
5539 @code{unsigned long}.
5542 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5543 Similar to @code{__builtin_popcount}, except the argument type is
5544 @code{unsigned long}.
5547 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5548 Similar to @code{__builtin_parity}, except the argument type is
5549 @code{unsigned long}.
5552 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5553 Similar to @code{__builtin_ffs}, except the argument type is
5554 @code{unsigned long long}.
5557 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5558 Similar to @code{__builtin_clz}, except the argument type is
5559 @code{unsigned long long}.
5562 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5563 Similar to @code{__builtin_ctz}, except the argument type is
5564 @code{unsigned long long}.
5567 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5568 Similar to @code{__builtin_popcount}, except the argument type is
5569 @code{unsigned long long}.
5572 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5573 Similar to @code{__builtin_parity}, except the argument type is
5574 @code{unsigned long long}.
5578 @node Target Builtins
5579 @section Built-in Functions Specific to Particular Target Machines
5581 On some target machines, GCC supports many built-in functions specific
5582 to those machines. Generally these generate calls to specific machine
5583 instructions, but allow the compiler to schedule those calls.
5586 * Alpha Built-in Functions::
5587 * ARM Built-in Functions::
5588 * X86 Built-in Functions::
5589 * PowerPC AltiVec Built-in Functions::
5592 @node Alpha Built-in Functions
5593 @subsection Alpha Built-in Functions
5595 These built-in functions are available for the Alpha family of
5596 processors, depending on the command-line switches used.
5598 The following built-in functions are always available. They
5599 all generate the machine instruction that is part of the name.
5602 long __builtin_alpha_implver (void)
5603 long __builtin_alpha_rpcc (void)
5604 long __builtin_alpha_amask (long)
5605 long __builtin_alpha_cmpbge (long, long)
5606 long __builtin_alpha_extbl (long, long)
5607 long __builtin_alpha_extwl (long, long)
5608 long __builtin_alpha_extll (long, long)
5609 long __builtin_alpha_extql (long, long)
5610 long __builtin_alpha_extwh (long, long)
5611 long __builtin_alpha_extlh (long, long)
5612 long __builtin_alpha_extqh (long, long)
5613 long __builtin_alpha_insbl (long, long)
5614 long __builtin_alpha_inswl (long, long)
5615 long __builtin_alpha_insll (long, long)
5616 long __builtin_alpha_insql (long, long)
5617 long __builtin_alpha_inswh (long, long)
5618 long __builtin_alpha_inslh (long, long)
5619 long __builtin_alpha_insqh (long, long)
5620 long __builtin_alpha_mskbl (long, long)
5621 long __builtin_alpha_mskwl (long, long)
5622 long __builtin_alpha_mskll (long, long)
5623 long __builtin_alpha_mskql (long, long)
5624 long __builtin_alpha_mskwh (long, long)
5625 long __builtin_alpha_msklh (long, long)
5626 long __builtin_alpha_mskqh (long, long)
5627 long __builtin_alpha_umulh (long, long)
5628 long __builtin_alpha_zap (long, long)
5629 long __builtin_alpha_zapnot (long, long)
5632 The following built-in functions are always with @option{-mmax}
5633 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5634 later. They all generate the machine instruction that is part
5638 long __builtin_alpha_pklb (long)
5639 long __builtin_alpha_pkwb (long)
5640 long __builtin_alpha_unpkbl (long)
5641 long __builtin_alpha_unpkbw (long)
5642 long __builtin_alpha_minub8 (long, long)
5643 long __builtin_alpha_minsb8 (long, long)
5644 long __builtin_alpha_minuw4 (long, long)
5645 long __builtin_alpha_minsw4 (long, long)
5646 long __builtin_alpha_maxub8 (long, long)
5647 long __builtin_alpha_maxsb8 (long, long)
5648 long __builtin_alpha_maxuw4 (long, long)
5649 long __builtin_alpha_maxsw4 (long, long)
5650 long __builtin_alpha_perr (long, long)
5653 The following built-in functions are always with @option{-mcix}
5654 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5655 later. They all generate the machine instruction that is part
5659 long __builtin_alpha_cttz (long)
5660 long __builtin_alpha_ctlz (long)
5661 long __builtin_alpha_ctpop (long)
5664 The following builtins are available on systems that use the OSF/1
5665 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5666 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5667 @code{rdval} and @code{wrval}.
5670 void *__builtin_thread_pointer (void)
5671 void __builtin_set_thread_pointer (void *)
5674 @node ARM Built-in Functions
5675 @subsection ARM Built-in Functions
5677 These built-in functions are available for the ARM family of
5678 processors, when the @option{-mcpu=iwmmxt} switch is used:
5681 typedef int v2si __attribute__ ((vector_size (8)));
5682 typedef short v4hi __attribute__ ((vector_size (8)));
5683 typedef char v8qi __attribute__ ((vector_size (8)));
5685 int __builtin_arm_getwcx (int)
5686 void __builtin_arm_setwcx (int, int)
5687 int __builtin_arm_textrmsb (v8qi, int)
5688 int __builtin_arm_textrmsh (v4hi, int)
5689 int __builtin_arm_textrmsw (v2si, int)
5690 int __builtin_arm_textrmub (v8qi, int)
5691 int __builtin_arm_textrmuh (v4hi, int)
5692 int __builtin_arm_textrmuw (v2si, int)
5693 v8qi __builtin_arm_tinsrb (v8qi, int)
5694 v4hi __builtin_arm_tinsrh (v4hi, int)
5695 v2si __builtin_arm_tinsrw (v2si, int)
5696 long long __builtin_arm_tmia (long long, int, int)
5697 long long __builtin_arm_tmiabb (long long, int, int)
5698 long long __builtin_arm_tmiabt (long long, int, int)
5699 long long __builtin_arm_tmiaph (long long, int, int)
5700 long long __builtin_arm_tmiatb (long long, int, int)
5701 long long __builtin_arm_tmiatt (long long, int, int)
5702 int __builtin_arm_tmovmskb (v8qi)
5703 int __builtin_arm_tmovmskh (v4hi)
5704 int __builtin_arm_tmovmskw (v2si)
5705 long long __builtin_arm_waccb (v8qi)
5706 long long __builtin_arm_wacch (v4hi)
5707 long long __builtin_arm_waccw (v2si)
5708 v8qi __builtin_arm_waddb (v8qi, v8qi)
5709 v8qi __builtin_arm_waddbss (v8qi, v8qi)
5710 v8qi __builtin_arm_waddbus (v8qi, v8qi)
5711 v4hi __builtin_arm_waddh (v4hi, v4hi)
5712 v4hi __builtin_arm_waddhss (v4hi, v4hi)
5713 v4hi __builtin_arm_waddhus (v4hi, v4hi)
5714 v2si __builtin_arm_waddw (v2si, v2si)
5715 v2si __builtin_arm_waddwss (v2si, v2si)
5716 v2si __builtin_arm_waddwus (v2si, v2si)
5717 v8qi __builtin_arm_walign (v8qi, v8qi, int)
5718 long long __builtin_arm_wand(long long, long long)
5719 long long __builtin_arm_wandn (long long, long long)
5720 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5721 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5722 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5723 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5724 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5725 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5726 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5727 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5728 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5729 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5730 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5731 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5732 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5733 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5734 long long __builtin_arm_wmacsz (v4hi, v4hi)
5735 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5736 long long __builtin_arm_wmacuz (v4hi, v4hi)
5737 v4hi __builtin_arm_wmadds (v4hi, v4hi)
5738 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5739 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5740 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5741 v2si __builtin_arm_wmaxsw (v2si, v2si)
5742 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5743 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5744 v2si __builtin_arm_wmaxuw (v2si, v2si)
5745 v8qi __builtin_arm_wminsb (v8qi, v8qi)
5746 v4hi __builtin_arm_wminsh (v4hi, v4hi)
5747 v2si __builtin_arm_wminsw (v2si, v2si)
5748 v8qi __builtin_arm_wminub (v8qi, v8qi)
5749 v4hi __builtin_arm_wminuh (v4hi, v4hi)
5750 v2si __builtin_arm_wminuw (v2si, v2si)
5751 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5752 v4hi __builtin_arm_wmulul (v4hi, v4hi)
5753 v4hi __builtin_arm_wmulum (v4hi, v4hi)
5754 long long __builtin_arm_wor (long long, long long)
5755 v2si __builtin_arm_wpackdss (long long, long long)
5756 v2si __builtin_arm_wpackdus (long long, long long)
5757 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5758 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5759 v4hi __builtin_arm_wpackwss (v2si, v2si)
5760 v4hi __builtin_arm_wpackwus (v2si, v2si)
5761 long long __builtin_arm_wrord (long long, long long)
5762 long long __builtin_arm_wrordi (long long, int)
5763 v4hi __builtin_arm_wrorh (v4hi, long long)
5764 v4hi __builtin_arm_wrorhi (v4hi, int)
5765 v2si __builtin_arm_wrorw (v2si, long long)
5766 v2si __builtin_arm_wrorwi (v2si, int)
5767 v2si __builtin_arm_wsadb (v8qi, v8qi)
5768 v2si __builtin_arm_wsadbz (v8qi, v8qi)
5769 v2si __builtin_arm_wsadh (v4hi, v4hi)
5770 v2si __builtin_arm_wsadhz (v4hi, v4hi)
5771 v4hi __builtin_arm_wshufh (v4hi, int)
5772 long long __builtin_arm_wslld (long long, long long)
5773 long long __builtin_arm_wslldi (long long, int)
5774 v4hi __builtin_arm_wsllh (v4hi, long long)
5775 v4hi __builtin_arm_wsllhi (v4hi, int)
5776 v2si __builtin_arm_wsllw (v2si, long long)
5777 v2si __builtin_arm_wsllwi (v2si, int)
5778 long long __builtin_arm_wsrad (long long, long long)
5779 long long __builtin_arm_wsradi (long long, int)
5780 v4hi __builtin_arm_wsrah (v4hi, long long)
5781 v4hi __builtin_arm_wsrahi (v4hi, int)
5782 v2si __builtin_arm_wsraw (v2si, long long)
5783 v2si __builtin_arm_wsrawi (v2si, int)
5784 long long __builtin_arm_wsrld (long long, long long)
5785 long long __builtin_arm_wsrldi (long long, int)
5786 v4hi __builtin_arm_wsrlh (v4hi, long long)
5787 v4hi __builtin_arm_wsrlhi (v4hi, int)
5788 v2si __builtin_arm_wsrlw (v2si, long long)
5789 v2si __builtin_arm_wsrlwi (v2si, int)
5790 v8qi __builtin_arm_wsubb (v8qi, v8qi)
5791 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5792 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5793 v4hi __builtin_arm_wsubh (v4hi, v4hi)
5794 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5795 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5796 v2si __builtin_arm_wsubw (v2si, v2si)
5797 v2si __builtin_arm_wsubwss (v2si, v2si)
5798 v2si __builtin_arm_wsubwus (v2si, v2si)
5799 v4hi __builtin_arm_wunpckehsb (v8qi)
5800 v2si __builtin_arm_wunpckehsh (v4hi)
5801 long long __builtin_arm_wunpckehsw (v2si)
5802 v4hi __builtin_arm_wunpckehub (v8qi)
5803 v2si __builtin_arm_wunpckehuh (v4hi)
5804 long long __builtin_arm_wunpckehuw (v2si)
5805 v4hi __builtin_arm_wunpckelsb (v8qi)
5806 v2si __builtin_arm_wunpckelsh (v4hi)
5807 long long __builtin_arm_wunpckelsw (v2si)
5808 v4hi __builtin_arm_wunpckelub (v8qi)
5809 v2si __builtin_arm_wunpckeluh (v4hi)
5810 long long __builtin_arm_wunpckeluw (v2si)
5811 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5812 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5813 v2si __builtin_arm_wunpckihw (v2si, v2si)
5814 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5815 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5816 v2si __builtin_arm_wunpckilw (v2si, v2si)
5817 long long __builtin_arm_wxor (long long, long long)
5818 long long __builtin_arm_wzero ()
5821 @node X86 Built-in Functions
5822 @subsection X86 Built-in Functions
5824 These built-in functions are available for the i386 and x86-64 family
5825 of computers, depending on the command-line switches used.
5827 The following machine modes are available for use with MMX built-in functions
5828 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5829 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5830 vector of eight 8-bit integers. Some of the built-in functions operate on
5831 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5833 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5834 of two 32-bit floating point values.
5836 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5837 floating point values. Some instructions use a vector of four 32-bit
5838 integers, these use @code{V4SI}. Finally, some instructions operate on an
5839 entire vector register, interpreting it as a 128-bit integer, these use mode
5842 The following built-in functions are made available by @option{-mmmx}.
5843 All of them generate the machine instruction that is part of the name.
5846 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5847 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5848 v2si __builtin_ia32_paddd (v2si, v2si)
5849 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5850 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5851 v2si __builtin_ia32_psubd (v2si, v2si)
5852 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5853 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5854 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5855 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5856 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5857 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5858 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5859 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5860 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5861 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5862 di __builtin_ia32_pand (di, di)
5863 di __builtin_ia32_pandn (di,di)
5864 di __builtin_ia32_por (di, di)
5865 di __builtin_ia32_pxor (di, di)
5866 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5867 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5868 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5869 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5870 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5871 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5872 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5873 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5874 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5875 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5876 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5877 v2si __builtin_ia32_punpckldq (v2si, v2si)
5878 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5879 v4hi __builtin_ia32_packssdw (v2si, v2si)
5880 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5883 The following built-in functions are made available either with
5884 @option{-msse}, or with a combination of @option{-m3dnow} and
5885 @option{-march=athlon}. All of them generate the machine
5886 instruction that is part of the name.
5889 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5890 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5891 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5892 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5893 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5894 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5895 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5896 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5897 int __builtin_ia32_pextrw (v4hi, int)
5898 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5899 int __builtin_ia32_pmovmskb (v8qi)
5900 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5901 void __builtin_ia32_movntq (di *, di)
5902 void __builtin_ia32_sfence (void)
5905 The following built-in functions are available when @option{-msse} is used.
5906 All of them generate the machine instruction that is part of the name.
5909 int __builtin_ia32_comieq (v4sf, v4sf)
5910 int __builtin_ia32_comineq (v4sf, v4sf)
5911 int __builtin_ia32_comilt (v4sf, v4sf)
5912 int __builtin_ia32_comile (v4sf, v4sf)
5913 int __builtin_ia32_comigt (v4sf, v4sf)
5914 int __builtin_ia32_comige (v4sf, v4sf)
5915 int __builtin_ia32_ucomieq (v4sf, v4sf)
5916 int __builtin_ia32_ucomineq (v4sf, v4sf)
5917 int __builtin_ia32_ucomilt (v4sf, v4sf)
5918 int __builtin_ia32_ucomile (v4sf, v4sf)
5919 int __builtin_ia32_ucomigt (v4sf, v4sf)
5920 int __builtin_ia32_ucomige (v4sf, v4sf)
5921 v4sf __builtin_ia32_addps (v4sf, v4sf)
5922 v4sf __builtin_ia32_subps (v4sf, v4sf)
5923 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5924 v4sf __builtin_ia32_divps (v4sf, v4sf)
5925 v4sf __builtin_ia32_addss (v4sf, v4sf)
5926 v4sf __builtin_ia32_subss (v4sf, v4sf)
5927 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5928 v4sf __builtin_ia32_divss (v4sf, v4sf)
5929 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5930 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5931 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5932 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5933 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5934 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5935 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5936 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5937 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5938 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5939 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5940 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5941 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5942 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5943 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5944 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5945 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5946 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
5947 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
5948 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
5949 v4sf __builtin_ia32_maxps (v4sf, v4sf)
5950 v4sf __builtin_ia32_maxss (v4sf, v4sf)
5951 v4sf __builtin_ia32_minps (v4sf, v4sf)
5952 v4sf __builtin_ia32_minss (v4sf, v4sf)
5953 v4sf __builtin_ia32_andps (v4sf, v4sf)
5954 v4sf __builtin_ia32_andnps (v4sf, v4sf)
5955 v4sf __builtin_ia32_orps (v4sf, v4sf)
5956 v4sf __builtin_ia32_xorps (v4sf, v4sf)
5957 v4sf __builtin_ia32_movss (v4sf, v4sf)
5958 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
5959 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
5960 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
5961 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
5962 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
5963 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
5964 v2si __builtin_ia32_cvtps2pi (v4sf)
5965 int __builtin_ia32_cvtss2si (v4sf)
5966 v2si __builtin_ia32_cvttps2pi (v4sf)
5967 int __builtin_ia32_cvttss2si (v4sf)
5968 v4sf __builtin_ia32_rcpps (v4sf)
5969 v4sf __builtin_ia32_rsqrtps (v4sf)
5970 v4sf __builtin_ia32_sqrtps (v4sf)
5971 v4sf __builtin_ia32_rcpss (v4sf)
5972 v4sf __builtin_ia32_rsqrtss (v4sf)
5973 v4sf __builtin_ia32_sqrtss (v4sf)
5974 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
5975 void __builtin_ia32_movntps (float *, v4sf)
5976 int __builtin_ia32_movmskps (v4sf)
5979 The following built-in functions are available when @option{-msse} is used.
5982 @item v4sf __builtin_ia32_loadaps (float *)
5983 Generates the @code{movaps} machine instruction as a load from memory.
5984 @item void __builtin_ia32_storeaps (float *, v4sf)
5985 Generates the @code{movaps} machine instruction as a store to memory.
5986 @item v4sf __builtin_ia32_loadups (float *)
5987 Generates the @code{movups} machine instruction as a load from memory.
5988 @item void __builtin_ia32_storeups (float *, v4sf)
5989 Generates the @code{movups} machine instruction as a store to memory.
5990 @item v4sf __builtin_ia32_loadsss (float *)
5991 Generates the @code{movss} machine instruction as a load from memory.
5992 @item void __builtin_ia32_storess (float *, v4sf)
5993 Generates the @code{movss} machine instruction as a store to memory.
5994 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
5995 Generates the @code{movhps} machine instruction as a load from memory.
5996 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
5997 Generates the @code{movlps} machine instruction as a load from memory
5998 @item void __builtin_ia32_storehps (v4sf, v2si *)
5999 Generates the @code{movhps} machine instruction as a store to memory.
6000 @item void __builtin_ia32_storelps (v4sf, v2si *)
6001 Generates the @code{movlps} machine instruction as a store to memory.
6004 The following built-in functions are available when @option{-msse3} is used.
6005 All of them generate the machine instruction that is part of the name.
6008 v2df __builtin_ia32_addsubpd (v2df, v2df)
6009 v2df __builtin_ia32_addsubps (v2df, v2df)
6010 v2df __builtin_ia32_haddpd (v2df, v2df)
6011 v2df __builtin_ia32_haddps (v2df, v2df)
6012 v2df __builtin_ia32_hsubpd (v2df, v2df)
6013 v2df __builtin_ia32_hsubps (v2df, v2df)
6014 v16qi __builtin_ia32_lddqu (char const *)
6015 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6016 v2df __builtin_ia32_movddup (v2df)
6017 v4sf __builtin_ia32_movshdup (v4sf)
6018 v4sf __builtin_ia32_movsldup (v4sf)
6019 void __builtin_ia32_mwait (unsigned int, unsigned int)
6022 The following built-in functions are available when @option{-msse3} is used.
6025 @item v2df __builtin_ia32_loadddup (double const *)
6026 Generates the @code{movddup} machine instruction as a load from memory.
6029 The following built-in functions are available when @option{-m3dnow} is used.
6030 All of them generate the machine instruction that is part of the name.
6033 void __builtin_ia32_femms (void)
6034 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6035 v2si __builtin_ia32_pf2id (v2sf)
6036 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6037 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6038 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6039 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6040 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6041 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6042 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6043 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6044 v2sf __builtin_ia32_pfrcp (v2sf)
6045 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6046 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6047 v2sf __builtin_ia32_pfrsqrt (v2sf)
6048 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6049 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6050 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6051 v2sf __builtin_ia32_pi2fd (v2si)
6052 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6055 The following built-in functions are available when both @option{-m3dnow}
6056 and @option{-march=athlon} are used. All of them generate the machine
6057 instruction that is part of the name.
6060 v2si __builtin_ia32_pf2iw (v2sf)
6061 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6062 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6063 v2sf __builtin_ia32_pi2fw (v2si)
6064 v2sf __builtin_ia32_pswapdsf (v2sf)
6065 v2si __builtin_ia32_pswapdsi (v2si)
6068 @node PowerPC AltiVec Built-in Functions
6069 @subsection PowerPC AltiVec Built-in Functions
6071 These built-in functions are available for the PowerPC family
6072 of computers, depending on the command-line switches used.
6074 The following machine modes are available for use with AltiVec built-in
6075 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
6076 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
6077 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
6078 @code{V16QI} for a vector of sixteen 8-bit integers.
6080 The following functions are made available by including
6081 @code{<altivec.h>} and using @option{-maltivec} and
6082 @option{-mabi=altivec}. The functions implement the functionality
6083 described in Motorola's AltiVec Programming Interface Manual.
6085 There are a few differences from Motorola's documentation and GCC's
6086 implementation. Vector constants are done with curly braces (not
6087 parentheses). Vector initializers require no casts if the vector
6088 constant is of the same type as the variable it is initializing. The
6089 @code{vector bool} type is deprecated and will be discontinued in
6090 further revisions. Use @code{vector signed} instead. If @code{signed}
6091 or @code{unsigned} is omitted, the vector type will default to
6092 @code{signed}. Lastly, all overloaded functions are implemented with macros
6093 for the C implementation. So code the following example will not work:
6096 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6099 Since vec_add is a macro, the vector constant in the above example will
6100 be treated as four different arguments. Wrap the entire argument in
6101 parentheses for this to work. The C++ implementation does not use
6104 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
6105 Internally, GCC uses built-in functions to achieve the functionality in
6106 the aforementioned header file, but they are not supported and are
6107 subject to change without notice.
6110 vector signed char vec_abs (vector signed char, vector signed char);
6111 vector signed short vec_abs (vector signed short, vector signed short);
6112 vector signed int vec_abs (vector signed int, vector signed int);
6113 vector signed float vec_abs (vector signed float, vector signed float);
6115 vector signed char vec_abss (vector signed char, vector signed char);
6116 vector signed short vec_abss (vector signed short, vector signed short);
6118 vector signed char vec_add (vector signed char, vector signed char);
6119 vector unsigned char vec_add (vector signed char, vector unsigned char);
6121 vector unsigned char vec_add (vector unsigned char, vector signed char);
6123 vector unsigned char vec_add (vector unsigned char,
6124 vector unsigned char);
6125 vector signed short vec_add (vector signed short, vector signed short);
6126 vector unsigned short vec_add (vector signed short,
6127 vector unsigned short);
6128 vector unsigned short vec_add (vector unsigned short,
6129 vector signed short);
6130 vector unsigned short vec_add (vector unsigned short,
6131 vector unsigned short);
6132 vector signed int vec_add (vector signed int, vector signed int);
6133 vector unsigned int vec_add (vector signed int, vector unsigned int);
6134 vector unsigned int vec_add (vector unsigned int, vector signed int);
6135 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6136 vector float vec_add (vector float, vector float);
6138 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6140 vector unsigned char vec_adds (vector signed char,
6141 vector unsigned char);
6142 vector unsigned char vec_adds (vector unsigned char,
6143 vector signed char);
6144 vector unsigned char vec_adds (vector unsigned char,
6145 vector unsigned char);
6146 vector signed char vec_adds (vector signed char, vector signed char);
6147 vector unsigned short vec_adds (vector signed short,
6148 vector unsigned short);
6149 vector unsigned short vec_adds (vector unsigned short,
6150 vector signed short);
6151 vector unsigned short vec_adds (vector unsigned short,
6152 vector unsigned short);
6153 vector signed short vec_adds (vector signed short, vector signed short);
6155 vector unsigned int vec_adds (vector signed int, vector unsigned int);
6156 vector unsigned int vec_adds (vector unsigned int, vector signed int);
6157 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6159 vector signed int vec_adds (vector signed int, vector signed int);
6161 vector float vec_and (vector float, vector float);
6162 vector float vec_and (vector float, vector signed int);
6163 vector float vec_and (vector signed int, vector float);
6164 vector signed int vec_and (vector signed int, vector signed int);
6165 vector unsigned int vec_and (vector signed int, vector unsigned int);
6166 vector unsigned int vec_and (vector unsigned int, vector signed int);
6167 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
6168 vector signed short vec_and (vector signed short, vector signed short);
6169 vector unsigned short vec_and (vector signed short,
6170 vector unsigned short);
6171 vector unsigned short vec_and (vector unsigned short,
6172 vector signed short);
6173 vector unsigned short vec_and (vector unsigned short,
6174 vector unsigned short);
6175 vector signed char vec_and (vector signed char, vector signed char);
6176 vector unsigned char vec_and (vector signed char, vector unsigned char);
6178 vector unsigned char vec_and (vector unsigned char, vector signed char);
6180 vector unsigned char vec_and (vector unsigned char,
6181 vector unsigned char);
6183 vector float vec_andc (vector float, vector float);
6184 vector float vec_andc (vector float, vector signed int);
6185 vector float vec_andc (vector signed int, vector float);
6186 vector signed int vec_andc (vector signed int, vector signed int);
6187 vector unsigned int vec_andc (vector signed int, vector unsigned int);
6188 vector unsigned int vec_andc (vector unsigned int, vector signed int);
6189 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
6191 vector signed short vec_andc (vector signed short, vector signed short);
6193 vector unsigned short vec_andc (vector signed short,
6194 vector unsigned short);
6195 vector unsigned short vec_andc (vector unsigned short,
6196 vector signed short);
6197 vector unsigned short vec_andc (vector unsigned short,
6198 vector unsigned short);
6199 vector signed char vec_andc (vector signed char, vector signed char);
6200 vector unsigned char vec_andc (vector signed char,
6201 vector unsigned char);
6202 vector unsigned char vec_andc (vector unsigned char,
6203 vector signed char);
6204 vector unsigned char vec_andc (vector unsigned char,
6205 vector unsigned char);
6207 vector unsigned char vec_avg (vector unsigned char,
6208 vector unsigned char);
6209 vector signed char vec_avg (vector signed char, vector signed char);
6210 vector unsigned short vec_avg (vector unsigned short,
6211 vector unsigned short);
6212 vector signed short vec_avg (vector signed short, vector signed short);
6213 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
6214 vector signed int vec_avg (vector signed int, vector signed int);
6216 vector float vec_ceil (vector float);
6218 vector signed int vec_cmpb (vector float, vector float);
6220 vector signed char vec_cmpeq (vector signed char, vector signed char);
6221 vector signed char vec_cmpeq (vector unsigned char,
6222 vector unsigned char);
6223 vector signed short vec_cmpeq (vector signed short,
6224 vector signed short);
6225 vector signed short vec_cmpeq (vector unsigned short,
6226 vector unsigned short);
6227 vector signed int vec_cmpeq (vector signed int, vector signed int);
6228 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
6229 vector signed int vec_cmpeq (vector float, vector float);
6231 vector signed int vec_cmpge (vector float, vector float);
6233 vector signed char vec_cmpgt (vector unsigned char,
6234 vector unsigned char);
6235 vector signed char vec_cmpgt (vector signed char, vector signed char);
6236 vector signed short vec_cmpgt (vector unsigned short,
6237 vector unsigned short);
6238 vector signed short vec_cmpgt (vector signed short,
6239 vector signed short);
6240 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
6241 vector signed int vec_cmpgt (vector signed int, vector signed int);
6242 vector signed int vec_cmpgt (vector float, vector float);
6244 vector signed int vec_cmple (vector float, vector float);
6246 vector signed char vec_cmplt (vector unsigned char,
6247 vector unsigned char);
6248 vector signed char vec_cmplt (vector signed char, vector signed char);
6249 vector signed short vec_cmplt (vector unsigned short,
6250 vector unsigned short);
6251 vector signed short vec_cmplt (vector signed short,
6252 vector signed short);
6253 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
6254 vector signed int vec_cmplt (vector signed int, vector signed int);
6255 vector signed int vec_cmplt (vector float, vector float);
6257 vector float vec_ctf (vector unsigned int, const char);
6258 vector float vec_ctf (vector signed int, const char);
6260 vector signed int vec_cts (vector float, const char);
6262 vector unsigned int vec_ctu (vector float, const char);
6264 void vec_dss (const char);
6266 void vec_dssall (void);
6268 void vec_dst (void *, int, const char);
6270 void vec_dstst (void *, int, const char);
6272 void vec_dststt (void *, int, const char);
6274 void vec_dstt (void *, int, const char);
6276 vector float vec_expte (vector float, vector float);
6278 vector float vec_floor (vector float, vector float);
6280 vector float vec_ld (int, vector float *);
6281 vector float vec_ld (int, float *):
6282 vector signed int vec_ld (int, int *);
6283 vector signed int vec_ld (int, vector signed int *);
6284 vector unsigned int vec_ld (int, vector unsigned int *);
6285 vector unsigned int vec_ld (int, unsigned int *);
6286 vector signed short vec_ld (int, short *, vector signed short *);
6287 vector unsigned short vec_ld (int, unsigned short *,
6288 vector unsigned short *);
6289 vector signed char vec_ld (int, signed char *);
6290 vector signed char vec_ld (int, vector signed char *);
6291 vector unsigned char vec_ld (int, unsigned char *);
6292 vector unsigned char vec_ld (int, vector unsigned char *);
6294 vector signed char vec_lde (int, signed char *);
6295 vector unsigned char vec_lde (int, unsigned char *);
6296 vector signed short vec_lde (int, short *);
6297 vector unsigned short vec_lde (int, unsigned short *);
6298 vector float vec_lde (int, float *);
6299 vector signed int vec_lde (int, int *);
6300 vector unsigned int vec_lde (int, unsigned int *);
6302 void float vec_ldl (int, float *);
6303 void float vec_ldl (int, vector float *);
6304 void signed int vec_ldl (int, vector signed int *);
6305 void signed int vec_ldl (int, int *);
6306 void unsigned int vec_ldl (int, unsigned int *);
6307 void unsigned int vec_ldl (int, vector unsigned int *);
6308 void signed short vec_ldl (int, vector signed short *);
6309 void signed short vec_ldl (int, short *);
6310 void unsigned short vec_ldl (int, vector unsigned short *);
6311 void unsigned short vec_ldl (int, unsigned short *);
6312 void signed char vec_ldl (int, vector signed char *);
6313 void signed char vec_ldl (int, signed char *);
6314 void unsigned char vec_ldl (int, vector unsigned char *);
6315 void unsigned char vec_ldl (int, unsigned char *);
6317 vector float vec_loge (vector float);
6319 vector unsigned char vec_lvsl (int, void *, int *);
6321 vector unsigned char vec_lvsr (int, void *, int *);
6323 vector float vec_madd (vector float, vector float, vector float);
6325 vector signed short vec_madds (vector signed short, vector signed short,
6326 vector signed short);
6328 vector unsigned char vec_max (vector signed char, vector unsigned char);
6330 vector unsigned char vec_max (vector unsigned char, vector signed char);
6332 vector unsigned char vec_max (vector unsigned char,
6333 vector unsigned char);
6334 vector signed char vec_max (vector signed char, vector signed char);
6335 vector unsigned short vec_max (vector signed short,
6336 vector unsigned short);
6337 vector unsigned short vec_max (vector unsigned short,
6338 vector signed short);
6339 vector unsigned short vec_max (vector unsigned short,
6340 vector unsigned short);
6341 vector signed short vec_max (vector signed short, vector signed short);
6342 vector unsigned int vec_max (vector signed int, vector unsigned int);
6343 vector unsigned int vec_max (vector unsigned int, vector signed int);
6344 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
6345 vector signed int vec_max (vector signed int, vector signed int);
6346 vector float vec_max (vector float, vector float);
6348 vector signed char vec_mergeh (vector signed char, vector signed char);
6349 vector unsigned char vec_mergeh (vector unsigned char,
6350 vector unsigned char);
6351 vector signed short vec_mergeh (vector signed short,
6352 vector signed short);
6353 vector unsigned short vec_mergeh (vector unsigned short,
6354 vector unsigned short);
6355 vector float vec_mergeh (vector float, vector float);
6356 vector signed int vec_mergeh (vector signed int, vector signed int);
6357 vector unsigned int vec_mergeh (vector unsigned int,
6358 vector unsigned int);
6360 vector signed char vec_mergel (vector signed char, vector signed char);
6361 vector unsigned char vec_mergel (vector unsigned char,
6362 vector unsigned char);
6363 vector signed short vec_mergel (vector signed short,
6364 vector signed short);
6365 vector unsigned short vec_mergel (vector unsigned short,
6366 vector unsigned short);
6367 vector float vec_mergel (vector float, vector float);
6368 vector signed int vec_mergel (vector signed int, vector signed int);
6369 vector unsigned int vec_mergel (vector unsigned int,
6370 vector unsigned int);
6372 vector unsigned short vec_mfvscr (void);
6374 vector unsigned char vec_min (vector signed char, vector unsigned char);
6376 vector unsigned char vec_min (vector unsigned char, vector signed char);
6378 vector unsigned char vec_min (vector unsigned char,
6379 vector unsigned char);
6380 vector signed char vec_min (vector signed char, vector signed char);
6381 vector unsigned short vec_min (vector signed short,
6382 vector unsigned short);
6383 vector unsigned short vec_min (vector unsigned short,
6384 vector signed short);
6385 vector unsigned short vec_min (vector unsigned short,
6386 vector unsigned short);
6387 vector signed short vec_min (vector signed short, vector signed short);
6388 vector unsigned int vec_min (vector signed int, vector unsigned int);
6389 vector unsigned int vec_min (vector unsigned int, vector signed int);
6390 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
6391 vector signed int vec_min (vector signed int, vector signed int);
6392 vector float vec_min (vector float, vector float);
6394 vector signed short vec_mladd (vector signed short, vector signed short,
6395 vector signed short);
6396 vector signed short vec_mladd (vector signed short,
6397 vector unsigned short,
6398 vector unsigned short);
6399 vector signed short vec_mladd (vector unsigned short,
6400 vector signed short,
6401 vector signed short);
6402 vector unsigned short vec_mladd (vector unsigned short,
6403 vector unsigned short,
6404 vector unsigned short);
6406 vector signed short vec_mradds (vector signed short,
6407 vector signed short,
6408 vector signed short);
6410 vector unsigned int vec_msum (vector unsigned char,
6411 vector unsigned char,
6412 vector unsigned int);
6413 vector signed int vec_msum (vector signed char, vector unsigned char,
6415 vector unsigned int vec_msum (vector unsigned short,
6416 vector unsigned short,
6417 vector unsigned int);
6418 vector signed int vec_msum (vector signed short, vector signed short,
6421 vector unsigned int vec_msums (vector unsigned short,
6422 vector unsigned short,
6423 vector unsigned int);
6424 vector signed int vec_msums (vector signed short, vector signed short,
6427 void vec_mtvscr (vector signed int);
6428 void vec_mtvscr (vector unsigned int);
6429 void vec_mtvscr (vector signed short);
6430 void vec_mtvscr (vector unsigned short);
6431 void vec_mtvscr (vector signed char);
6432 void vec_mtvscr (vector unsigned char);
6434 vector unsigned short vec_mule (vector unsigned char,
6435 vector unsigned char);
6436 vector signed short vec_mule (vector signed char, vector signed char);
6437 vector unsigned int vec_mule (vector unsigned short,
6438 vector unsigned short);
6439 vector signed int vec_mule (vector signed short, vector signed short);
6441 vector unsigned short vec_mulo (vector unsigned char,
6442 vector unsigned char);
6443 vector signed short vec_mulo (vector signed char, vector signed char);
6444 vector unsigned int vec_mulo (vector unsigned short,
6445 vector unsigned short);
6446 vector signed int vec_mulo (vector signed short, vector signed short);
6448 vector float vec_nmsub (vector float, vector float, vector float);
6450 vector float vec_nor (vector float, vector float);
6451 vector signed int vec_nor (vector signed int, vector signed int);
6452 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
6453 vector signed short vec_nor (vector signed short, vector signed short);
6454 vector unsigned short vec_nor (vector unsigned short,
6455 vector unsigned short);
6456 vector signed char vec_nor (vector signed char, vector signed char);
6457 vector unsigned char vec_nor (vector unsigned char,
6458 vector unsigned char);
6460 vector float vec_or (vector float, vector float);
6461 vector float vec_or (vector float, vector signed int);
6462 vector float vec_or (vector signed int, vector float);
6463 vector signed int vec_or (vector signed int, vector signed int);
6464 vector unsigned int vec_or (vector signed int, vector unsigned int);
6465 vector unsigned int vec_or (vector unsigned int, vector signed int);
6466 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
6467 vector signed short vec_or (vector signed short, vector signed short);
6468 vector unsigned short vec_or (vector signed short,
6469 vector unsigned short);
6470 vector unsigned short vec_or (vector unsigned short,
6471 vector signed short);
6472 vector unsigned short vec_or (vector unsigned short,
6473 vector unsigned short);
6474 vector signed char vec_or (vector signed char, vector signed char);
6475 vector unsigned char vec_or (vector signed char, vector unsigned char);
6476 vector unsigned char vec_or (vector unsigned char, vector signed char);
6477 vector unsigned char vec_or (vector unsigned char,
6478 vector unsigned char);
6480 vector signed char vec_pack (vector signed short, vector signed short);
6481 vector unsigned char vec_pack (vector unsigned short,
6482 vector unsigned short);
6483 vector signed short vec_pack (vector signed int, vector signed int);
6484 vector unsigned short vec_pack (vector unsigned int,
6485 vector unsigned int);
6487 vector signed short vec_packpx (vector unsigned int,
6488 vector unsigned int);
6490 vector unsigned char vec_packs (vector unsigned short,
6491 vector unsigned short);
6492 vector signed char vec_packs (vector signed short, vector signed short);
6494 vector unsigned short vec_packs (vector unsigned int,
6495 vector unsigned int);
6496 vector signed short vec_packs (vector signed int, vector signed int);
6498 vector unsigned char vec_packsu (vector unsigned short,
6499 vector unsigned short);
6500 vector unsigned char vec_packsu (vector signed short,
6501 vector signed short);
6502 vector unsigned short vec_packsu (vector unsigned int,
6503 vector unsigned int);
6504 vector unsigned short vec_packsu (vector signed int, vector signed int);
6506 vector float vec_perm (vector float, vector float,
6507 vector unsigned char);
6508 vector signed int vec_perm (vector signed int, vector signed int,
6509 vector unsigned char);
6510 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
6511 vector unsigned char);
6512 vector signed short vec_perm (vector signed short, vector signed short,
6513 vector unsigned char);
6514 vector unsigned short vec_perm (vector unsigned short,
6515 vector unsigned short,
6516 vector unsigned char);
6517 vector signed char vec_perm (vector signed char, vector signed char,
6518 vector unsigned char);
6519 vector unsigned char vec_perm (vector unsigned char,
6520 vector unsigned char,
6521 vector unsigned char);
6523 vector float vec_re (vector float);
6525 vector signed char vec_rl (vector signed char, vector unsigned char);
6526 vector unsigned char vec_rl (vector unsigned char,
6527 vector unsigned char);
6528 vector signed short vec_rl (vector signed short, vector unsigned short);
6530 vector unsigned short vec_rl (vector unsigned short,
6531 vector unsigned short);
6532 vector signed int vec_rl (vector signed int, vector unsigned int);
6533 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
6535 vector float vec_round (vector float);
6537 vector float vec_rsqrte (vector float);
6539 vector float vec_sel (vector float, vector float, vector signed int);
6540 vector float vec_sel (vector float, vector float, vector unsigned int);
6541 vector signed int vec_sel (vector signed int, vector signed int,
6543 vector signed int vec_sel (vector signed int, vector signed int,
6544 vector unsigned int);
6545 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6547 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6548 vector unsigned int);
6549 vector signed short vec_sel (vector signed short, vector signed short,
6550 vector signed short);
6551 vector signed short vec_sel (vector signed short, vector signed short,
6552 vector unsigned short);
6553 vector unsigned short vec_sel (vector unsigned short,
6554 vector unsigned short,
6555 vector signed short);
6556 vector unsigned short vec_sel (vector unsigned short,
6557 vector unsigned short,
6558 vector unsigned short);
6559 vector signed char vec_sel (vector signed char, vector signed char,
6560 vector signed char);
6561 vector signed char vec_sel (vector signed char, vector signed char,
6562 vector unsigned char);
6563 vector unsigned char vec_sel (vector unsigned char,
6564 vector unsigned char,
6565 vector signed char);
6566 vector unsigned char vec_sel (vector unsigned char,
6567 vector unsigned char,
6568 vector unsigned char);
6570 vector signed char vec_sl (vector signed char, vector unsigned char);
6571 vector unsigned char vec_sl (vector unsigned char,
6572 vector unsigned char);
6573 vector signed short vec_sl (vector signed short, vector unsigned short);
6575 vector unsigned short vec_sl (vector unsigned short,
6576 vector unsigned short);
6577 vector signed int vec_sl (vector signed int, vector unsigned int);
6578 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
6580 vector float vec_sld (vector float, vector float, const char);
6581 vector signed int vec_sld (vector signed int, vector signed int,
6583 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
6585 vector signed short vec_sld (vector signed short, vector signed short,
6587 vector unsigned short vec_sld (vector unsigned short,
6588 vector unsigned short, const char);
6589 vector signed char vec_sld (vector signed char, vector signed char,
6591 vector unsigned char vec_sld (vector unsigned char,
6592 vector unsigned char,
6595 vector signed int vec_sll (vector signed int, vector unsigned int);
6596 vector signed int vec_sll (vector signed int, vector unsigned short);
6597 vector signed int vec_sll (vector signed int, vector unsigned char);
6598 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
6599 vector unsigned int vec_sll (vector unsigned int,
6600 vector unsigned short);
6601 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
6603 vector signed short vec_sll (vector signed short, vector unsigned int);
6604 vector signed short vec_sll (vector signed short,
6605 vector unsigned short);
6606 vector signed short vec_sll (vector signed short, vector unsigned char);
6608 vector unsigned short vec_sll (vector unsigned short,
6609 vector unsigned int);
6610 vector unsigned short vec_sll (vector unsigned short,
6611 vector unsigned short);
6612 vector unsigned short vec_sll (vector unsigned short,
6613 vector unsigned char);
6614 vector signed char vec_sll (vector signed char, vector unsigned int);
6615 vector signed char vec_sll (vector signed char, vector unsigned short);
6616 vector signed char vec_sll (vector signed char, vector unsigned char);
6617 vector unsigned char vec_sll (vector unsigned char,
6618 vector unsigned int);
6619 vector unsigned char vec_sll (vector unsigned char,
6620 vector unsigned short);
6621 vector unsigned char vec_sll (vector unsigned char,
6622 vector unsigned char);
6624 vector float vec_slo (vector float, vector signed char);
6625 vector float vec_slo (vector float, vector unsigned char);
6626 vector signed int vec_slo (vector signed int, vector signed char);
6627 vector signed int vec_slo (vector signed int, vector unsigned char);
6628 vector unsigned int vec_slo (vector unsigned int, vector signed char);
6629 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
6631 vector signed short vec_slo (vector signed short, vector signed char);
6632 vector signed short vec_slo (vector signed short, vector unsigned char);
6634 vector unsigned short vec_slo (vector unsigned short,
6635 vector signed char);
6636 vector unsigned short vec_slo (vector unsigned short,
6637 vector unsigned char);
6638 vector signed char vec_slo (vector signed char, vector signed char);
6639 vector signed char vec_slo (vector signed char, vector unsigned char);
6640 vector unsigned char vec_slo (vector unsigned char, vector signed char);
6642 vector unsigned char vec_slo (vector unsigned char,
6643 vector unsigned char);
6645 vector signed char vec_splat (vector signed char, const char);
6646 vector unsigned char vec_splat (vector unsigned char, const char);
6647 vector signed short vec_splat (vector signed short, const char);
6648 vector unsigned short vec_splat (vector unsigned short, const char);
6649 vector float vec_splat (vector float, const char);
6650 vector signed int vec_splat (vector signed int, const char);
6651 vector unsigned int vec_splat (vector unsigned int, const char);
6653 vector signed char vec_splat_s8 (const char);
6655 vector signed short vec_splat_s16 (const char);
6657 vector signed int vec_splat_s32 (const char);
6659 vector unsigned char vec_splat_u8 (const char);
6661 vector unsigned short vec_splat_u16 (const char);
6663 vector unsigned int vec_splat_u32 (const char);
6665 vector signed char vec_sr (vector signed char, vector unsigned char);
6666 vector unsigned char vec_sr (vector unsigned char,
6667 vector unsigned char);
6668 vector signed short vec_sr (vector signed short, vector unsigned short);
6670 vector unsigned short vec_sr (vector unsigned short,
6671 vector unsigned short);
6672 vector signed int vec_sr (vector signed int, vector unsigned int);
6673 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
6675 vector signed char vec_sra (vector signed char, vector unsigned char);
6676 vector unsigned char vec_sra (vector unsigned char,
6677 vector unsigned char);
6678 vector signed short vec_sra (vector signed short,
6679 vector unsigned short);
6680 vector unsigned short vec_sra (vector unsigned short,
6681 vector unsigned short);
6682 vector signed int vec_sra (vector signed int, vector unsigned int);
6683 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
6685 vector signed int vec_srl (vector signed int, vector unsigned int);
6686 vector signed int vec_srl (vector signed int, vector unsigned short);
6687 vector signed int vec_srl (vector signed int, vector unsigned char);
6688 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
6689 vector unsigned int vec_srl (vector unsigned int,
6690 vector unsigned short);
6691 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
6693 vector signed short vec_srl (vector signed short, vector unsigned int);
6694 vector signed short vec_srl (vector signed short,
6695 vector unsigned short);
6696 vector signed short vec_srl (vector signed short, vector unsigned char);
6698 vector unsigned short vec_srl (vector unsigned short,
6699 vector unsigned int);
6700 vector unsigned short vec_srl (vector unsigned short,
6701 vector unsigned short);
6702 vector unsigned short vec_srl (vector unsigned short,
6703 vector unsigned char);
6704 vector signed char vec_srl (vector signed char, vector unsigned int);
6705 vector signed char vec_srl (vector signed char, vector unsigned short);
6706 vector signed char vec_srl (vector signed char, vector unsigned char);
6707 vector unsigned char vec_srl (vector unsigned char,
6708 vector unsigned int);
6709 vector unsigned char vec_srl (vector unsigned char,
6710 vector unsigned short);
6711 vector unsigned char vec_srl (vector unsigned char,
6712 vector unsigned char);
6714 vector float vec_sro (vector float, vector signed char);
6715 vector float vec_sro (vector float, vector unsigned char);
6716 vector signed int vec_sro (vector signed int, vector signed char);
6717 vector signed int vec_sro (vector signed int, vector unsigned char);
6718 vector unsigned int vec_sro (vector unsigned int, vector signed char);
6719 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
6721 vector signed short vec_sro (vector signed short, vector signed char);
6722 vector signed short vec_sro (vector signed short, vector unsigned char);
6724 vector unsigned short vec_sro (vector unsigned short,
6725 vector signed char);
6726 vector unsigned short vec_sro (vector unsigned short,
6727 vector unsigned char);
6728 vector signed char vec_sro (vector signed char, vector signed char);
6729 vector signed char vec_sro (vector signed char, vector unsigned char);
6730 vector unsigned char vec_sro (vector unsigned char, vector signed char);
6732 vector unsigned char vec_sro (vector unsigned char,
6733 vector unsigned char);
6735 void vec_st (vector float, int, float *);
6736 void vec_st (vector float, int, vector float *);
6737 void vec_st (vector signed int, int, int *);
6738 void vec_st (vector signed int, int, unsigned int *);
6739 void vec_st (vector unsigned int, int, unsigned int *);
6740 void vec_st (vector unsigned int, int, vector unsigned int *);
6741 void vec_st (vector signed short, int, short *);
6742 void vec_st (vector signed short, int, vector unsigned short *);
6743 void vec_st (vector signed short, int, vector signed short *);
6744 void vec_st (vector unsigned short, int, unsigned short *);
6745 void vec_st (vector unsigned short, int, vector unsigned short *);
6746 void vec_st (vector signed char, int, signed char *);
6747 void vec_st (vector signed char, int, unsigned char *);
6748 void vec_st (vector signed char, int, vector signed char *);
6749 void vec_st (vector unsigned char, int, unsigned char *);
6750 void vec_st (vector unsigned char, int, vector unsigned char *);
6752 void vec_ste (vector signed char, int, unsigned char *);
6753 void vec_ste (vector signed char, int, signed char *);
6754 void vec_ste (vector unsigned char, int, unsigned char *);
6755 void vec_ste (vector signed short, int, short *);
6756 void vec_ste (vector signed short, int, unsigned short *);
6757 void vec_ste (vector unsigned short, int, void *);
6758 void vec_ste (vector signed int, int, unsigned int *);
6759 void vec_ste (vector signed int, int, int *);
6760 void vec_ste (vector unsigned int, int, unsigned int *);
6761 void vec_ste (vector float, int, float *);
6763 void vec_stl (vector float, int, vector float *);
6764 void vec_stl (vector float, int, float *);
6765 void vec_stl (vector signed int, int, vector signed int *);
6766 void vec_stl (vector signed int, int, int *);
6767 void vec_stl (vector signed int, int, unsigned int *);
6768 void vec_stl (vector unsigned int, int, vector unsigned int *);
6769 void vec_stl (vector unsigned int, int, unsigned int *);
6770 void vec_stl (vector signed short, int, short *);
6771 void vec_stl (vector signed short, int, unsigned short *);
6772 void vec_stl (vector signed short, int, vector signed short *);
6773 void vec_stl (vector unsigned short, int, unsigned short *);
6774 void vec_stl (vector unsigned short, int, vector signed short *);
6775 void vec_stl (vector signed char, int, signed char *);
6776 void vec_stl (vector signed char, int, unsigned char *);
6777 void vec_stl (vector signed char, int, vector signed char *);
6778 void vec_stl (vector unsigned char, int, unsigned char *);
6779 void vec_stl (vector unsigned char, int, vector unsigned char *);
6781 vector signed char vec_sub (vector signed char, vector signed char);
6782 vector unsigned char vec_sub (vector signed char, vector unsigned char);
6784 vector unsigned char vec_sub (vector unsigned char, vector signed char);
6786 vector unsigned char vec_sub (vector unsigned char,
6787 vector unsigned char);
6788 vector signed short vec_sub (vector signed short, vector signed short);
6789 vector unsigned short vec_sub (vector signed short,
6790 vector unsigned short);
6791 vector unsigned short vec_sub (vector unsigned short,
6792 vector signed short);
6793 vector unsigned short vec_sub (vector unsigned short,
6794 vector unsigned short);
6795 vector signed int vec_sub (vector signed int, vector signed int);
6796 vector unsigned int vec_sub (vector signed int, vector unsigned int);
6797 vector unsigned int vec_sub (vector unsigned int, vector signed int);
6798 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
6799 vector float vec_sub (vector float, vector float);
6801 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
6803 vector unsigned char vec_subs (vector signed char,
6804 vector unsigned char);
6805 vector unsigned char vec_subs (vector unsigned char,
6806 vector signed char);
6807 vector unsigned char vec_subs (vector unsigned char,
6808 vector unsigned char);
6809 vector signed char vec_subs (vector signed char, vector signed char);
6810 vector unsigned short vec_subs (vector signed short,
6811 vector unsigned short);
6812 vector unsigned short vec_subs (vector unsigned short,
6813 vector signed short);
6814 vector unsigned short vec_subs (vector unsigned short,
6815 vector unsigned short);
6816 vector signed short vec_subs (vector signed short, vector signed short);
6818 vector unsigned int vec_subs (vector signed int, vector unsigned int);
6819 vector unsigned int vec_subs (vector unsigned int, vector signed int);
6820 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
6822 vector signed int vec_subs (vector signed int, vector signed int);
6824 vector unsigned int vec_sum4s (vector unsigned char,
6825 vector unsigned int);
6826 vector signed int vec_sum4s (vector signed char, vector signed int);
6827 vector signed int vec_sum4s (vector signed short, vector signed int);
6829 vector signed int vec_sum2s (vector signed int, vector signed int);
6831 vector signed int vec_sums (vector signed int, vector signed int);
6833 vector float vec_trunc (vector float);
6835 vector signed short vec_unpackh (vector signed char);
6836 vector unsigned int vec_unpackh (vector signed short);
6837 vector signed int vec_unpackh (vector signed short);
6839 vector signed short vec_unpackl (vector signed char);
6840 vector unsigned int vec_unpackl (vector signed short);
6841 vector signed int vec_unpackl (vector signed short);
6843 vector float vec_xor (vector float, vector float);
6844 vector float vec_xor (vector float, vector signed int);
6845 vector float vec_xor (vector signed int, vector float);
6846 vector signed int vec_xor (vector signed int, vector signed int);
6847 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6848 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6849 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6850 vector signed short vec_xor (vector signed short, vector signed short);
6851 vector unsigned short vec_xor (vector signed short,
6852 vector unsigned short);
6853 vector unsigned short vec_xor (vector unsigned short,
6854 vector signed short);
6855 vector unsigned short vec_xor (vector unsigned short,
6856 vector unsigned short);
6857 vector signed char vec_xor (vector signed char, vector signed char);
6858 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6860 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6862 vector unsigned char vec_xor (vector unsigned char,
6863 vector unsigned char);
6865 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6867 vector signed int vec_all_eq (vector signed char, vector signed char);
6868 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6870 vector signed int vec_all_eq (vector unsigned char,
6871 vector unsigned char);
6872 vector signed int vec_all_eq (vector signed short,
6873 vector unsigned short);
6874 vector signed int vec_all_eq (vector signed short, vector signed short);
6876 vector signed int vec_all_eq (vector unsigned short,
6877 vector signed short);
6878 vector signed int vec_all_eq (vector unsigned short,
6879 vector unsigned short);
6880 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6881 vector signed int vec_all_eq (vector signed int, vector signed int);
6882 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6883 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6885 vector signed int vec_all_eq (vector float, vector float);
6887 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6889 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6891 vector signed int vec_all_ge (vector unsigned char,
6892 vector unsigned char);
6893 vector signed int vec_all_ge (vector signed char, vector signed char);
6894 vector signed int vec_all_ge (vector signed short,
6895 vector unsigned short);
6896 vector signed int vec_all_ge (vector unsigned short,
6897 vector signed short);
6898 vector signed int vec_all_ge (vector unsigned short,
6899 vector unsigned short);
6900 vector signed int vec_all_ge (vector signed short, vector signed short);
6902 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6903 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6904 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6906 vector signed int vec_all_ge (vector signed int, vector signed int);
6907 vector signed int vec_all_ge (vector float, vector float);
6909 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6911 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6913 vector signed int vec_all_gt (vector unsigned char,
6914 vector unsigned char);
6915 vector signed int vec_all_gt (vector signed char, vector signed char);
6916 vector signed int vec_all_gt (vector signed short,
6917 vector unsigned short);
6918 vector signed int vec_all_gt (vector unsigned short,
6919 vector signed short);
6920 vector signed int vec_all_gt (vector unsigned short,
6921 vector unsigned short);
6922 vector signed int vec_all_gt (vector signed short, vector signed short);
6924 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6925 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6926 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6928 vector signed int vec_all_gt (vector signed int, vector signed int);
6929 vector signed int vec_all_gt (vector float, vector float);
6931 vector signed int vec_all_in (vector float, vector float);
6933 vector signed int vec_all_le (vector signed char, vector unsigned char);
6935 vector signed int vec_all_le (vector unsigned char, vector signed char);
6937 vector signed int vec_all_le (vector unsigned char,
6938 vector unsigned char);
6939 vector signed int vec_all_le (vector signed char, vector signed char);
6940 vector signed int vec_all_le (vector signed short,
6941 vector unsigned short);
6942 vector signed int vec_all_le (vector unsigned short,
6943 vector signed short);
6944 vector signed int vec_all_le (vector unsigned short,
6945 vector unsigned short);
6946 vector signed int vec_all_le (vector signed short, vector signed short);
6948 vector signed int vec_all_le (vector signed int, vector unsigned int);
6949 vector signed int vec_all_le (vector unsigned int, vector signed int);
6950 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
6952 vector signed int vec_all_le (vector signed int, vector signed int);
6953 vector signed int vec_all_le (vector float, vector float);
6955 vector signed int vec_all_lt (vector signed char, vector unsigned char);
6957 vector signed int vec_all_lt (vector unsigned char, vector signed char);
6959 vector signed int vec_all_lt (vector unsigned char,
6960 vector unsigned char);
6961 vector signed int vec_all_lt (vector signed char, vector signed char);
6962 vector signed int vec_all_lt (vector signed short,
6963 vector unsigned short);
6964 vector signed int vec_all_lt (vector unsigned short,
6965 vector signed short);
6966 vector signed int vec_all_lt (vector unsigned short,
6967 vector unsigned short);
6968 vector signed int vec_all_lt (vector signed short, vector signed short);
6970 vector signed int vec_all_lt (vector signed int, vector unsigned int);
6971 vector signed int vec_all_lt (vector unsigned int, vector signed int);
6972 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
6974 vector signed int vec_all_lt (vector signed int, vector signed int);
6975 vector signed int vec_all_lt (vector float, vector float);
6977 vector signed int vec_all_nan (vector float);
6979 vector signed int vec_all_ne (vector signed char, vector unsigned char);
6981 vector signed int vec_all_ne (vector signed char, vector signed char);
6982 vector signed int vec_all_ne (vector unsigned char, vector signed char);
6984 vector signed int vec_all_ne (vector unsigned char,
6985 vector unsigned char);
6986 vector signed int vec_all_ne (vector signed short,
6987 vector unsigned short);
6988 vector signed int vec_all_ne (vector signed short, vector signed short);
6990 vector signed int vec_all_ne (vector unsigned short,
6991 vector signed short);
6992 vector signed int vec_all_ne (vector unsigned short,
6993 vector unsigned short);
6994 vector signed int vec_all_ne (vector signed int, vector unsigned int);
6995 vector signed int vec_all_ne (vector signed int, vector signed int);
6996 vector signed int vec_all_ne (vector unsigned int, vector signed int);
6997 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
6999 vector signed int vec_all_ne (vector float, vector float);
7001 vector signed int vec_all_nge (vector float, vector float);
7003 vector signed int vec_all_ngt (vector float, vector float);
7005 vector signed int vec_all_nle (vector float, vector float);
7007 vector signed int vec_all_nlt (vector float, vector float);
7009 vector signed int vec_all_numeric (vector float);
7011 vector signed int vec_any_eq (vector signed char, vector unsigned char);
7013 vector signed int vec_any_eq (vector signed char, vector signed char);
7014 vector signed int vec_any_eq (vector unsigned char, vector signed char);
7016 vector signed int vec_any_eq (vector unsigned char,
7017 vector unsigned char);
7018 vector signed int vec_any_eq (vector signed short,
7019 vector unsigned short);
7020 vector signed int vec_any_eq (vector signed short, vector signed short);
7022 vector signed int vec_any_eq (vector unsigned short,
7023 vector signed short);
7024 vector signed int vec_any_eq (vector unsigned short,
7025 vector unsigned short);
7026 vector signed int vec_any_eq (vector signed int, vector unsigned int);
7027 vector signed int vec_any_eq (vector signed int, vector signed int);
7028 vector signed int vec_any_eq (vector unsigned int, vector signed int);
7029 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
7031 vector signed int vec_any_eq (vector float, vector float);
7033 vector signed int vec_any_ge (vector signed char, vector unsigned char);
7035 vector signed int vec_any_ge (vector unsigned char, vector signed char);
7037 vector signed int vec_any_ge (vector unsigned char,
7038 vector unsigned char);
7039 vector signed int vec_any_ge (vector signed char, vector signed char);
7040 vector signed int vec_any_ge (vector signed short,
7041 vector unsigned short);
7042 vector signed int vec_any_ge (vector unsigned short,
7043 vector signed short);
7044 vector signed int vec_any_ge (vector unsigned short,
7045 vector unsigned short);
7046 vector signed int vec_any_ge (vector signed short, vector signed short);
7048 vector signed int vec_any_ge (vector signed int, vector unsigned int);
7049 vector signed int vec_any_ge (vector unsigned int, vector signed int);
7050 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
7052 vector signed int vec_any_ge (vector signed int, vector signed int);
7053 vector signed int vec_any_ge (vector float, vector float);
7055 vector signed int vec_any_gt (vector signed char, vector unsigned char);
7057 vector signed int vec_any_gt (vector unsigned char, vector signed char);
7059 vector signed int vec_any_gt (vector unsigned char,
7060 vector unsigned char);
7061 vector signed int vec_any_gt (vector signed char, vector signed char);
7062 vector signed int vec_any_gt (vector signed short,
7063 vector unsigned short);
7064 vector signed int vec_any_gt (vector unsigned short,
7065 vector signed short);
7066 vector signed int vec_any_gt (vector unsigned short,
7067 vector unsigned short);
7068 vector signed int vec_any_gt (vector signed short, vector signed short);
7070 vector signed int vec_any_gt (vector signed int, vector unsigned int);
7071 vector signed int vec_any_gt (vector unsigned int, vector signed int);
7072 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
7074 vector signed int vec_any_gt (vector signed int, vector signed int);
7075 vector signed int vec_any_gt (vector float, vector float);
7077 vector signed int vec_any_le (vector signed char, vector unsigned char);
7079 vector signed int vec_any_le (vector unsigned char, vector signed char);
7081 vector signed int vec_any_le (vector unsigned char,
7082 vector unsigned char);
7083 vector signed int vec_any_le (vector signed char, vector signed char);
7084 vector signed int vec_any_le (vector signed short,
7085 vector unsigned short);
7086 vector signed int vec_any_le (vector unsigned short,
7087 vector signed short);
7088 vector signed int vec_any_le (vector unsigned short,
7089 vector unsigned short);
7090 vector signed int vec_any_le (vector signed short, vector signed short);
7092 vector signed int vec_any_le (vector signed int, vector unsigned int);
7093 vector signed int vec_any_le (vector unsigned int, vector signed int);
7094 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
7096 vector signed int vec_any_le (vector signed int, vector signed int);
7097 vector signed int vec_any_le (vector float, vector float);
7099 vector signed int vec_any_lt (vector signed char, vector unsigned char);
7101 vector signed int vec_any_lt (vector unsigned char, vector signed char);
7103 vector signed int vec_any_lt (vector unsigned char,
7104 vector unsigned char);
7105 vector signed int vec_any_lt (vector signed char, vector signed char);
7106 vector signed int vec_any_lt (vector signed short,
7107 vector unsigned short);
7108 vector signed int vec_any_lt (vector unsigned short,
7109 vector signed short);
7110 vector signed int vec_any_lt (vector unsigned short,
7111 vector unsigned short);
7112 vector signed int vec_any_lt (vector signed short, vector signed short);
7114 vector signed int vec_any_lt (vector signed int, vector unsigned int);
7115 vector signed int vec_any_lt (vector unsigned int, vector signed int);
7116 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
7118 vector signed int vec_any_lt (vector signed int, vector signed int);
7119 vector signed int vec_any_lt (vector float, vector float);
7121 vector signed int vec_any_nan (vector float);
7123 vector signed int vec_any_ne (vector signed char, vector unsigned char);
7125 vector signed int vec_any_ne (vector signed char, vector signed char);
7126 vector signed int vec_any_ne (vector unsigned char, vector signed char);
7128 vector signed int vec_any_ne (vector unsigned char,
7129 vector unsigned char);
7130 vector signed int vec_any_ne (vector signed short,
7131 vector unsigned short);
7132 vector signed int vec_any_ne (vector signed short, vector signed short);
7134 vector signed int vec_any_ne (vector unsigned short,
7135 vector signed short);
7136 vector signed int vec_any_ne (vector unsigned short,
7137 vector unsigned short);
7138 vector signed int vec_any_ne (vector signed int, vector unsigned int);
7139 vector signed int vec_any_ne (vector signed int, vector signed int);
7140 vector signed int vec_any_ne (vector unsigned int, vector signed int);
7141 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
7143 vector signed int vec_any_ne (vector float, vector float);
7145 vector signed int vec_any_nge (vector float, vector float);
7147 vector signed int vec_any_ngt (vector float, vector float);
7149 vector signed int vec_any_nle (vector float, vector float);
7151 vector signed int vec_any_nlt (vector float, vector float);
7153 vector signed int vec_any_numeric (vector float);
7155 vector signed int vec_any_out (vector float, vector float);
7159 @section Pragmas Accepted by GCC
7163 GCC supports several types of pragmas, primarily in order to compile
7164 code originally written for other compilers. Note that in general
7165 we do not recommend the use of pragmas; @xref{Function Attributes},
7166 for further explanation.
7170 * RS/6000 and PowerPC Pragmas::
7177 @subsection ARM Pragmas
7179 The ARM target defines pragmas for controlling the default addition of
7180 @code{long_call} and @code{short_call} attributes to functions.
7181 @xref{Function Attributes}, for information about the effects of these
7186 @cindex pragma, long_calls
7187 Set all subsequent functions to have the @code{long_call} attribute.
7190 @cindex pragma, no_long_calls
7191 Set all subsequent functions to have the @code{short_call} attribute.
7193 @item long_calls_off
7194 @cindex pragma, long_calls_off
7195 Do not affect the @code{long_call} or @code{short_call} attributes of
7196 subsequent functions.
7199 @node RS/6000 and PowerPC Pragmas
7200 @subsection RS/6000 and PowerPC Pragmas
7202 The RS/6000 and PowerPC targets define one pragma for controlling
7203 whether or not the @code{longcall} attribute is added to function
7204 declarations by default. This pragma overrides the @option{-mlongcall}
7205 option, but not the @code{longcall} and @code{shortcall} attributes.
7206 @xref{RS/6000 and PowerPC Options}, for more information about when long
7207 calls are and are not necessary.
7211 @cindex pragma, longcall
7212 Apply the @code{longcall} attribute to all subsequent function
7216 Do not apply the @code{longcall} attribute to subsequent function
7220 @c Describe c4x pragmas here.
7221 @c Describe h8300 pragmas here.
7222 @c Describe sh pragmas here.
7223 @c Describe v850 pragmas here.
7225 @node Darwin Pragmas
7226 @subsection Darwin Pragmas
7228 The following pragmas are available for all architectures running the
7229 Darwin operating system. These are useful for compatibility with other
7233 @item mark @var{tokens}@dots{}
7234 @cindex pragma, mark
7235 This pragma is accepted, but has no effect.
7237 @item options align=@var{alignment}
7238 @cindex pragma, options align
7239 This pragma sets the alignment of fields in structures. The values of
7240 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
7241 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
7242 properly; to restore the previous setting, use @code{reset} for the
7245 @item segment @var{tokens}@dots{}
7246 @cindex pragma, segment
7247 This pragma is accepted, but has no effect.
7249 @item unused (@var{var} [, @var{var}]@dots{})
7250 @cindex pragma, unused
7251 This pragma declares variables to be possibly unused. GCC will not
7252 produce warnings for the listed variables. The effect is similar to
7253 that of the @code{unused} attribute, except that this pragma may appear
7254 anywhere within the variables' scopes.
7257 @node Solaris Pragmas
7258 @subsection Solaris Pragmas
7260 For compatibility with the SunPRO compiler, the following pragma
7264 @item redefine_extname @var{oldname} @var{newname}
7265 @cindex pragma, redefine_extname
7267 This pragma gives the C function @var{oldname} the assembler label
7268 @var{newname}. The pragma must appear before the function declaration.
7269 This pragma is equivalent to the asm labels extension (@pxref{Asm
7270 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
7271 if the pragma is available.
7275 @subsection Tru64 Pragmas
7277 For compatibility with the Compaq C compiler, the following pragma
7281 @item extern_prefix @var{string}
7282 @cindex pragma, extern_prefix
7284 This pragma renames all subsequent function and variable declarations
7285 such that @var{string} is prepended to the name. This effect may be
7286 terminated by using another @code{extern_prefix} pragma with the
7289 This pragma is similar in intent to to the asm labels extension
7290 (@pxref{Asm Labels}) in that the system programmer wants to change
7291 the assembly-level ABI without changing the source-level API. The
7292 preprocessor defines @code{__PRAGMA_EXTERN_PREFIX} if the pragma is
7296 @node Unnamed Fields
7297 @section Unnamed struct/union fields within structs/unions.
7301 For compatibility with other compilers, GCC allows you to define
7302 a structure or union that contains, as fields, structures and unions
7303 without names. For example:
7316 In this example, the user would be able to access members of the unnamed
7317 union with code like @samp{foo.b}. Note that only unnamed structs and
7318 unions are allowed, you may not have, for example, an unnamed
7321 You must never create such structures that cause ambiguous field definitions.
7322 For example, this structure:
7333 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
7334 Such constructs are not supported and must be avoided. In the future,
7335 such constructs may be detected and treated as compilation errors.
7338 @section Thread-Local Storage
7339 @cindex Thread-Local Storage
7340 @cindex @acronym{TLS}
7343 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
7344 are allocated such that there is one instance of the variable per extant
7345 thread. The run-time model GCC uses to implement this originates
7346 in the IA-64 processor-specific ABI, but has since been migrated
7347 to other processors as well. It requires significant support from
7348 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
7349 system libraries (@file{libc.so} and @file{libpthread.so}), so it
7350 is not available everywhere.
7352 At the user level, the extension is visible with a new storage
7353 class keyword: @code{__thread}. For example:
7357 extern __thread struct state s;
7358 static __thread char *p;
7361 The @code{__thread} specifier may be used alone, with the @code{extern}
7362 or @code{static} specifiers, but with no other storage class specifier.
7363 When used with @code{extern} or @code{static}, @code{__thread} must appear
7364 immediately after the other storage class specifier.
7366 The @code{__thread} specifier may be applied to any global, file-scoped
7367 static, function-scoped static, or static data member of a class. It may
7368 not be applied to block-scoped automatic or non-static data member.
7370 When the address-of operator is applied to a thread-local variable, it is
7371 evaluated at run-time and returns the address of the current thread's
7372 instance of that variable. An address so obtained may be used by any
7373 thread. When a thread terminates, any pointers to thread-local variables
7374 in that thread become invalid.
7376 No static initialization may refer to the address of a thread-local variable.
7378 In C++, if an initializer is present for a thread-local variable, it must
7379 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
7382 See @uref{http://people.redhat.com/drepper/tls.pdf,
7383 ELF Handling For Thread-Local Storage} for a detailed explanation of
7384 the four thread-local storage addressing models, and how the run-time
7385 is expected to function.
7388 * C99 Thread-Local Edits::
7389 * C++98 Thread-Local Edits::
7392 @node C99 Thread-Local Edits
7393 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
7395 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
7396 that document the exact semantics of the language extension.
7400 @cite{5.1.2 Execution environments}
7402 Add new text after paragraph 1
7405 Within either execution environment, a @dfn{thread} is a flow of
7406 control within a program. It is implementation defined whether
7407 or not there may be more than one thread associated with a program.
7408 It is implementation defined how threads beyond the first are
7409 created, the name and type of the function called at thread
7410 startup, and how threads may be terminated. However, objects
7411 with thread storage duration shall be initialized before thread
7416 @cite{6.2.4 Storage durations of objects}
7418 Add new text before paragraph 3
7421 An object whose identifier is declared with the storage-class
7422 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
7423 Its lifetime is the entire execution of the thread, and its
7424 stored value is initialized only once, prior to thread startup.
7428 @cite{6.4.1 Keywords}
7430 Add @code{__thread}.
7433 @cite{6.7.1 Storage-class specifiers}
7435 Add @code{__thread} to the list of storage class specifiers in
7438 Change paragraph 2 to
7441 With the exception of @code{__thread}, at most one storage-class
7442 specifier may be given [@dots{}]. The @code{__thread} specifier may
7443 be used alone, or immediately following @code{extern} or
7447 Add new text after paragraph 6
7450 The declaration of an identifier for a variable that has
7451 block scope that specifies @code{__thread} shall also
7452 specify either @code{extern} or @code{static}.
7454 The @code{__thread} specifier shall be used only with
7459 @node C++98 Thread-Local Edits
7460 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
7462 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
7463 that document the exact semantics of the language extension.
7467 @b{[intro.execution]}
7469 New text after paragraph 4
7472 A @dfn{thread} is a flow of control within the abstract machine.
7473 It is implementation defined whether or not there may be more than
7477 New text after paragraph 7
7480 It is unspecified whether additional action must be taken to
7481 ensure when and whether side effects are visible to other threads.
7487 Add @code{__thread}.
7490 @b{[basic.start.main]}
7492 Add after paragraph 5
7495 The thread that begins execution at the @code{main} function is called
7496 the @dfn{main thread}. It is implementation defined how functions
7497 beginning threads other than the main thread are designated or typed.
7498 A function so designated, as well as the @code{main} function, is called
7499 a @dfn{thread startup function}. It is implementation defined what
7500 happens if a thread startup function returns. It is implementation
7501 defined what happens to other threads when any thread calls @code{exit}.
7505 @b{[basic.start.init]}
7507 Add after paragraph 4
7510 The storage for an object of thread storage duration shall be
7511 statically initialized before the first statement of the thread startup
7512 function. An object of thread storage duration shall not require
7513 dynamic initialization.
7517 @b{[basic.start.term]}
7519 Add after paragraph 3
7522 The type of an object with thread storage duration shall not have a
7523 non-trivial destructor, nor shall it be an array type whose elements
7524 (directly or indirectly) have non-trivial destructors.
7530 Add ``thread storage duration'' to the list in paragraph 1.
7535 Thread, static, and automatic storage durations are associated with
7536 objects introduced by declarations [@dots{}].
7539 Add @code{__thread} to the list of specifiers in paragraph 3.
7542 @b{[basic.stc.thread]}
7544 New section before @b{[basic.stc.static]}
7547 The keyword @code{__thread} applied to a non-local object gives the
7548 object thread storage duration.
7550 A local variable or class data member declared both @code{static}
7551 and @code{__thread} gives the variable or member thread storage
7556 @b{[basic.stc.static]}
7561 All objects which have neither thread storage duration, dynamic
7562 storage duration nor are local [@dots{}].
7568 Add @code{__thread} to the list in paragraph 1.
7573 With the exception of @code{__thread}, at most one
7574 @var{storage-class-specifier} shall appear in a given
7575 @var{decl-specifier-seq}. The @code{__thread} specifier may
7576 be used alone, or immediately following the @code{extern} or
7577 @code{static} specifiers. [@dots{}]
7580 Add after paragraph 5
7583 The @code{__thread} specifier can be applied only to the names of objects
7584 and to anonymous unions.
7590 Add after paragraph 6
7593 Non-@code{static} members shall not be @code{__thread}.
7597 @node C++ Extensions
7598 @chapter Extensions to the C++ Language
7599 @cindex extensions, C++ language
7600 @cindex C++ language extensions
7602 The GNU compiler provides these extensions to the C++ language (and you
7603 can also use most of the C language extensions in your C++ programs). If you
7604 want to write code that checks whether these features are available, you can
7605 test for the GNU compiler the same way as for C programs: check for a
7606 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
7607 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
7608 Predefined Macros,cpp,The GNU C Preprocessor}).
7611 * Min and Max:: C++ Minimum and maximum operators.
7612 * Volatiles:: What constitutes an access to a volatile object.
7613 * Restricted Pointers:: C99 restricted pointers and references.
7614 * Vague Linkage:: Where G++ puts inlines, vtables and such.
7615 * C++ Interface:: You can use a single C++ header file for both
7616 declarations and definitions.
7617 * Template Instantiation:: Methods for ensuring that exactly one copy of
7618 each needed template instantiation is emitted.
7619 * Bound member functions:: You can extract a function pointer to the
7620 method denoted by a @samp{->*} or @samp{.*} expression.
7621 * C++ Attributes:: Variable, function, and type attributes for C++ only.
7622 * Strong Using:: Strong using-directives for namespace composition.
7623 * Offsetof:: Special syntax for implementing @code{offsetof}.
7624 * Java Exceptions:: Tweaking exception handling to work with Java.
7625 * Deprecated Features:: Things will disappear from g++.
7626 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
7630 @section Minimum and Maximum Operators in C++
7632 It is very convenient to have operators which return the ``minimum'' or the
7633 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
7636 @item @var{a} <? @var{b}
7638 @cindex minimum operator
7639 is the @dfn{minimum}, returning the smaller of the numeric values
7640 @var{a} and @var{b};
7642 @item @var{a} >? @var{b}
7644 @cindex maximum operator
7645 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
7649 These operations are not primitive in ordinary C++, since you can
7650 use a macro to return the minimum of two things in C++, as in the
7654 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
7658 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
7659 the minimum value of variables @var{i} and @var{j}.
7661 However, side effects in @code{X} or @code{Y} may cause unintended
7662 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
7663 the smaller counter twice. The GNU C @code{typeof} extension allows you
7664 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
7665 However, writing @code{MIN} and @code{MAX} as macros also forces you to
7666 use function-call notation for a fundamental arithmetic operation.
7667 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
7670 Since @code{<?} and @code{>?} are built into the compiler, they properly
7671 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
7675 @section When is a Volatile Object Accessed?
7676 @cindex accessing volatiles
7677 @cindex volatile read
7678 @cindex volatile write
7679 @cindex volatile access
7681 Both the C and C++ standard have the concept of volatile objects. These
7682 are normally accessed by pointers and used for accessing hardware. The
7683 standards encourage compilers to refrain from optimizations
7684 concerning accesses to volatile objects that it might perform on
7685 non-volatile objects. The C standard leaves it implementation defined
7686 as to what constitutes a volatile access. The C++ standard omits to
7687 specify this, except to say that C++ should behave in a similar manner
7688 to C with respect to volatiles, where possible. The minimum either
7689 standard specifies is that at a sequence point all previous accesses to
7690 volatile objects have stabilized and no subsequent accesses have
7691 occurred. Thus an implementation is free to reorder and combine
7692 volatile accesses which occur between sequence points, but cannot do so
7693 for accesses across a sequence point. The use of volatiles does not
7694 allow you to violate the restriction on updating objects multiple times
7695 within a sequence point.
7697 In most expressions, it is intuitively obvious what is a read and what is
7698 a write. For instance
7701 volatile int *dst = @var{somevalue};
7702 volatile int *src = @var{someothervalue};
7707 will cause a read of the volatile object pointed to by @var{src} and stores the
7708 value into the volatile object pointed to by @var{dst}. There is no
7709 guarantee that these reads and writes are atomic, especially for objects
7710 larger than @code{int}.
7712 Less obvious expressions are where something which looks like an access
7713 is used in a void context. An example would be,
7716 volatile int *src = @var{somevalue};
7720 With C, such expressions are rvalues, and as rvalues cause a read of
7721 the object, GCC interprets this as a read of the volatile being pointed
7722 to. The C++ standard specifies that such expressions do not undergo
7723 lvalue to rvalue conversion, and that the type of the dereferenced
7724 object may be incomplete. The C++ standard does not specify explicitly
7725 that it is this lvalue to rvalue conversion which is responsible for
7726 causing an access. However, there is reason to believe that it is,
7727 because otherwise certain simple expressions become undefined. However,
7728 because it would surprise most programmers, G++ treats dereferencing a
7729 pointer to volatile object of complete type in a void context as a read
7730 of the object. When the object has incomplete type, G++ issues a
7735 struct T @{int m;@};
7736 volatile S *ptr1 = @var{somevalue};
7737 volatile T *ptr2 = @var{somevalue};
7742 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
7743 causes a read of the object pointed to. If you wish to force an error on
7744 the first case, you must force a conversion to rvalue with, for instance
7745 a static cast, @code{static_cast<S>(*ptr1)}.
7747 When using a reference to volatile, G++ does not treat equivalent
7748 expressions as accesses to volatiles, but instead issues a warning that
7749 no volatile is accessed. The rationale for this is that otherwise it
7750 becomes difficult to determine where volatile access occur, and not
7751 possible to ignore the return value from functions returning volatile
7752 references. Again, if you wish to force a read, cast the reference to
7755 @node Restricted Pointers
7756 @section Restricting Pointer Aliasing
7757 @cindex restricted pointers
7758 @cindex restricted references
7759 @cindex restricted this pointer
7761 As with the C front end, G++ understands the C99 feature of restricted pointers,
7762 specified with the @code{__restrict__}, or @code{__restrict} type
7763 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
7764 language flag, @code{restrict} is not a keyword in C++.
7766 In addition to allowing restricted pointers, you can specify restricted
7767 references, which indicate that the reference is not aliased in the local
7771 void fn (int *__restrict__ rptr, int &__restrict__ rref)
7778 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
7779 @var{rref} refers to a (different) unaliased integer.
7781 You may also specify whether a member function's @var{this} pointer is
7782 unaliased by using @code{__restrict__} as a member function qualifier.
7785 void T::fn () __restrict__
7792 Within the body of @code{T::fn}, @var{this} will have the effective
7793 definition @code{T *__restrict__ const this}. Notice that the
7794 interpretation of a @code{__restrict__} member function qualifier is
7795 different to that of @code{const} or @code{volatile} qualifier, in that it
7796 is applied to the pointer rather than the object. This is consistent with
7797 other compilers which implement restricted pointers.
7799 As with all outermost parameter qualifiers, @code{__restrict__} is
7800 ignored in function definition matching. This means you only need to
7801 specify @code{__restrict__} in a function definition, rather than
7802 in a function prototype as well.
7805 @section Vague Linkage
7806 @cindex vague linkage
7808 There are several constructs in C++ which require space in the object
7809 file but are not clearly tied to a single translation unit. We say that
7810 these constructs have ``vague linkage''. Typically such constructs are
7811 emitted wherever they are needed, though sometimes we can be more
7815 @item Inline Functions
7816 Inline functions are typically defined in a header file which can be
7817 included in many different compilations. Hopefully they can usually be
7818 inlined, but sometimes an out-of-line copy is necessary, if the address
7819 of the function is taken or if inlining fails. In general, we emit an
7820 out-of-line copy in all translation units where one is needed. As an
7821 exception, we only emit inline virtual functions with the vtable, since
7822 it will always require a copy.
7824 Local static variables and string constants used in an inline function
7825 are also considered to have vague linkage, since they must be shared
7826 between all inlined and out-of-line instances of the function.
7830 C++ virtual functions are implemented in most compilers using a lookup
7831 table, known as a vtable. The vtable contains pointers to the virtual
7832 functions provided by a class, and each object of the class contains a
7833 pointer to its vtable (or vtables, in some multiple-inheritance
7834 situations). If the class declares any non-inline, non-pure virtual
7835 functions, the first one is chosen as the ``key method'' for the class,
7836 and the vtable is only emitted in the translation unit where the key
7839 @emph{Note:} If the chosen key method is later defined as inline, the
7840 vtable will still be emitted in every translation unit which defines it.
7841 Make sure that any inline virtuals are declared inline in the class
7842 body, even if they are not defined there.
7844 @item type_info objects
7847 C++ requires information about types to be written out in order to
7848 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7849 For polymorphic classes (classes with virtual functions), the type_info
7850 object is written out along with the vtable so that @samp{dynamic_cast}
7851 can determine the dynamic type of a class object at runtime. For all
7852 other types, we write out the type_info object when it is used: when
7853 applying @samp{typeid} to an expression, throwing an object, or
7854 referring to a type in a catch clause or exception specification.
7856 @item Template Instantiations
7857 Most everything in this section also applies to template instantiations,
7858 but there are other options as well.
7859 @xref{Template Instantiation,,Where's the Template?}.
7863 When used with GNU ld version 2.8 or later on an ELF system such as
7864 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
7865 these constructs will be discarded at link time. This is known as
7868 On targets that don't support COMDAT, but do support weak symbols, GCC
7869 will use them. This way one copy will override all the others, but
7870 the unused copies will still take up space in the executable.
7872 For targets which do not support either COMDAT or weak symbols,
7873 most entities with vague linkage will be emitted as local symbols to
7874 avoid duplicate definition errors from the linker. This will not happen
7875 for local statics in inlines, however, as having multiple copies will
7876 almost certainly break things.
7878 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7879 another way to control placement of these constructs.
7882 @section Declarations and Definitions in One Header
7884 @cindex interface and implementation headers, C++
7885 @cindex C++ interface and implementation headers
7886 C++ object definitions can be quite complex. In principle, your source
7887 code will need two kinds of things for each object that you use across
7888 more than one source file. First, you need an @dfn{interface}
7889 specification, describing its structure with type declarations and
7890 function prototypes. Second, you need the @dfn{implementation} itself.
7891 It can be tedious to maintain a separate interface description in a
7892 header file, in parallel to the actual implementation. It is also
7893 dangerous, since separate interface and implementation definitions may
7894 not remain parallel.
7896 @cindex pragmas, interface and implementation
7897 With GNU C++, you can use a single header file for both purposes.
7900 @emph{Warning:} The mechanism to specify this is in transition. For the
7901 nonce, you must use one of two @code{#pragma} commands; in a future
7902 release of GNU C++, an alternative mechanism will make these
7903 @code{#pragma} commands unnecessary.
7906 The header file contains the full definitions, but is marked with
7907 @samp{#pragma interface} in the source code. This allows the compiler
7908 to use the header file only as an interface specification when ordinary
7909 source files incorporate it with @code{#include}. In the single source
7910 file where the full implementation belongs, you can use either a naming
7911 convention or @samp{#pragma implementation} to indicate this alternate
7912 use of the header file.
7915 @item #pragma interface
7916 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7917 @kindex #pragma interface
7918 Use this directive in @emph{header files} that define object classes, to save
7919 space in most of the object files that use those classes. Normally,
7920 local copies of certain information (backup copies of inline member
7921 functions, debugging information, and the internal tables that implement
7922 virtual functions) must be kept in each object file that includes class
7923 definitions. You can use this pragma to avoid such duplication. When a
7924 header file containing @samp{#pragma interface} is included in a
7925 compilation, this auxiliary information will not be generated (unless
7926 the main input source file itself uses @samp{#pragma implementation}).
7927 Instead, the object files will contain references to be resolved at link
7930 The second form of this directive is useful for the case where you have
7931 multiple headers with the same name in different directories. If you
7932 use this form, you must specify the same string to @samp{#pragma
7935 @item #pragma implementation
7936 @itemx #pragma implementation "@var{objects}.h"
7937 @kindex #pragma implementation
7938 Use this pragma in a @emph{main input file}, when you want full output from
7939 included header files to be generated (and made globally visible). The
7940 included header file, in turn, should use @samp{#pragma interface}.
7941 Backup copies of inline member functions, debugging information, and the
7942 internal tables used to implement virtual functions are all generated in
7943 implementation files.
7945 @cindex implied @code{#pragma implementation}
7946 @cindex @code{#pragma implementation}, implied
7947 @cindex naming convention, implementation headers
7948 If you use @samp{#pragma implementation} with no argument, it applies to
7949 an include file with the same basename@footnote{A file's @dfn{basename}
7950 was the name stripped of all leading path information and of trailing
7951 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
7952 file. For example, in @file{allclass.cc}, giving just
7953 @samp{#pragma implementation}
7954 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
7956 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
7957 an implementation file whenever you would include it from
7958 @file{allclass.cc} even if you never specified @samp{#pragma
7959 implementation}. This was deemed to be more trouble than it was worth,
7960 however, and disabled.
7962 If you use an explicit @samp{#pragma implementation}, it must appear in
7963 your source file @emph{before} you include the affected header files.
7965 Use the string argument if you want a single implementation file to
7966 include code from multiple header files. (You must also use
7967 @samp{#include} to include the header file; @samp{#pragma
7968 implementation} only specifies how to use the file---it doesn't actually
7971 There is no way to split up the contents of a single header file into
7972 multiple implementation files.
7975 @cindex inlining and C++ pragmas
7976 @cindex C++ pragmas, effect on inlining
7977 @cindex pragmas in C++, effect on inlining
7978 @samp{#pragma implementation} and @samp{#pragma interface} also have an
7979 effect on function inlining.
7981 If you define a class in a header file marked with @samp{#pragma
7982 interface}, the effect on a function defined in that class is similar to
7983 an explicit @code{extern} declaration---the compiler emits no code at
7984 all to define an independent version of the function. Its definition
7985 is used only for inlining with its callers.
7987 @opindex fno-implement-inlines
7988 Conversely, when you include the same header file in a main source file
7989 that declares it as @samp{#pragma implementation}, the compiler emits
7990 code for the function itself; this defines a version of the function
7991 that can be found via pointers (or by callers compiled without
7992 inlining). If all calls to the function can be inlined, you can avoid
7993 emitting the function by compiling with @option{-fno-implement-inlines}.
7994 If any calls were not inlined, you will get linker errors.
7996 @node Template Instantiation
7997 @section Where's the Template?
7998 @cindex template instantiation
8000 C++ templates are the first language feature to require more
8001 intelligence from the environment than one usually finds on a UNIX
8002 system. Somehow the compiler and linker have to make sure that each
8003 template instance occurs exactly once in the executable if it is needed,
8004 and not at all otherwise. There are two basic approaches to this
8005 problem, which I will refer to as the Borland model and the Cfront model.
8009 Borland C++ solved the template instantiation problem by adding the code
8010 equivalent of common blocks to their linker; the compiler emits template
8011 instances in each translation unit that uses them, and the linker
8012 collapses them together. The advantage of this model is that the linker
8013 only has to consider the object files themselves; there is no external
8014 complexity to worry about. This disadvantage is that compilation time
8015 is increased because the template code is being compiled repeatedly.
8016 Code written for this model tends to include definitions of all
8017 templates in the header file, since they must be seen to be
8021 The AT&T C++ translator, Cfront, solved the template instantiation
8022 problem by creating the notion of a template repository, an
8023 automatically maintained place where template instances are stored. A
8024 more modern version of the repository works as follows: As individual
8025 object files are built, the compiler places any template definitions and
8026 instantiations encountered in the repository. At link time, the link
8027 wrapper adds in the objects in the repository and compiles any needed
8028 instances that were not previously emitted. The advantages of this
8029 model are more optimal compilation speed and the ability to use the
8030 system linker; to implement the Borland model a compiler vendor also
8031 needs to replace the linker. The disadvantages are vastly increased
8032 complexity, and thus potential for error; for some code this can be
8033 just as transparent, but in practice it can been very difficult to build
8034 multiple programs in one directory and one program in multiple
8035 directories. Code written for this model tends to separate definitions
8036 of non-inline member templates into a separate file, which should be
8037 compiled separately.
8040 When used with GNU ld version 2.8 or later on an ELF system such as
8041 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
8042 Borland model. On other systems, G++ implements neither automatic
8045 A future version of G++ will support a hybrid model whereby the compiler
8046 will emit any instantiations for which the template definition is
8047 included in the compile, and store template definitions and
8048 instantiation context information into the object file for the rest.
8049 The link wrapper will extract that information as necessary and invoke
8050 the compiler to produce the remaining instantiations. The linker will
8051 then combine duplicate instantiations.
8053 In the mean time, you have the following options for dealing with
8054 template instantiations:
8059 Compile your template-using code with @option{-frepo}. The compiler will
8060 generate files with the extension @samp{.rpo} listing all of the
8061 template instantiations used in the corresponding object files which
8062 could be instantiated there; the link wrapper, @samp{collect2}, will
8063 then update the @samp{.rpo} files to tell the compiler where to place
8064 those instantiations and rebuild any affected object files. The
8065 link-time overhead is negligible after the first pass, as the compiler
8066 will continue to place the instantiations in the same files.
8068 This is your best option for application code written for the Borland
8069 model, as it will just work. Code written for the Cfront model will
8070 need to be modified so that the template definitions are available at
8071 one or more points of instantiation; usually this is as simple as adding
8072 @code{#include <tmethods.cc>} to the end of each template header.
8074 For library code, if you want the library to provide all of the template
8075 instantiations it needs, just try to link all of its object files
8076 together; the link will fail, but cause the instantiations to be
8077 generated as a side effect. Be warned, however, that this may cause
8078 conflicts if multiple libraries try to provide the same instantiations.
8079 For greater control, use explicit instantiation as described in the next
8083 @opindex fno-implicit-templates
8084 Compile your code with @option{-fno-implicit-templates} to disable the
8085 implicit generation of template instances, and explicitly instantiate
8086 all the ones you use. This approach requires more knowledge of exactly
8087 which instances you need than do the others, but it's less
8088 mysterious and allows greater control. You can scatter the explicit
8089 instantiations throughout your program, perhaps putting them in the
8090 translation units where the instances are used or the translation units
8091 that define the templates themselves; you can put all of the explicit
8092 instantiations you need into one big file; or you can create small files
8099 template class Foo<int>;
8100 template ostream& operator <<
8101 (ostream&, const Foo<int>&);
8104 for each of the instances you need, and create a template instantiation
8107 If you are using Cfront-model code, you can probably get away with not
8108 using @option{-fno-implicit-templates} when compiling files that don't
8109 @samp{#include} the member template definitions.
8111 If you use one big file to do the instantiations, you may want to
8112 compile it without @option{-fno-implicit-templates} so you get all of the
8113 instances required by your explicit instantiations (but not by any
8114 other files) without having to specify them as well.
8116 G++ has extended the template instantiation syntax given in the ISO
8117 standard to allow forward declaration of explicit instantiations
8118 (with @code{extern}), instantiation of the compiler support data for a
8119 template class (i.e.@: the vtable) without instantiating any of its
8120 members (with @code{inline}), and instantiation of only the static data
8121 members of a template class, without the support data or member
8122 functions (with (@code{static}):
8125 extern template int max (int, int);
8126 inline template class Foo<int>;
8127 static template class Foo<int>;
8131 Do nothing. Pretend G++ does implement automatic instantiation
8132 management. Code written for the Borland model will work fine, but
8133 each translation unit will contain instances of each of the templates it
8134 uses. In a large program, this can lead to an unacceptable amount of code
8137 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
8138 more discussion of these pragmas.
8141 @node Bound member functions
8142 @section Extracting the function pointer from a bound pointer to member function
8144 @cindex pointer to member function
8145 @cindex bound pointer to member function
8147 In C++, pointer to member functions (PMFs) are implemented using a wide
8148 pointer of sorts to handle all the possible call mechanisms; the PMF
8149 needs to store information about how to adjust the @samp{this} pointer,
8150 and if the function pointed to is virtual, where to find the vtable, and
8151 where in the vtable to look for the member function. If you are using
8152 PMFs in an inner loop, you should really reconsider that decision. If
8153 that is not an option, you can extract the pointer to the function that
8154 would be called for a given object/PMF pair and call it directly inside
8155 the inner loop, to save a bit of time.
8157 Note that you will still be paying the penalty for the call through a
8158 function pointer; on most modern architectures, such a call defeats the
8159 branch prediction features of the CPU@. This is also true of normal
8160 virtual function calls.
8162 The syntax for this extension is
8166 extern int (A::*fp)();
8167 typedef int (*fptr)(A *);
8169 fptr p = (fptr)(a.*fp);
8172 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
8173 no object is needed to obtain the address of the function. They can be
8174 converted to function pointers directly:
8177 fptr p1 = (fptr)(&A::foo);
8180 @opindex Wno-pmf-conversions
8181 You must specify @option{-Wno-pmf-conversions} to use this extension.
8183 @node C++ Attributes
8184 @section C++-Specific Variable, Function, and Type Attributes
8186 Some attributes only make sense for C++ programs.
8189 @item init_priority (@var{priority})
8190 @cindex init_priority attribute
8193 In Standard C++, objects defined at namespace scope are guaranteed to be
8194 initialized in an order in strict accordance with that of their definitions
8195 @emph{in a given translation unit}. No guarantee is made for initializations
8196 across translation units. However, GNU C++ allows users to control the
8197 order of initialization of objects defined at namespace scope with the
8198 @code{init_priority} attribute by specifying a relative @var{priority},
8199 a constant integral expression currently bounded between 101 and 65535
8200 inclusive. Lower numbers indicate a higher priority.
8202 In the following example, @code{A} would normally be created before
8203 @code{B}, but the @code{init_priority} attribute has reversed that order:
8206 Some_Class A __attribute__ ((init_priority (2000)));
8207 Some_Class B __attribute__ ((init_priority (543)));
8211 Note that the particular values of @var{priority} do not matter; only their
8214 @item java_interface
8215 @cindex java_interface attribute
8217 This type attribute informs C++ that the class is a Java interface. It may
8218 only be applied to classes declared within an @code{extern "Java"} block.
8219 Calls to methods declared in this interface will be dispatched using GCJ's
8220 interface table mechanism, instead of regular virtual table dispatch.
8224 See also @xref{Strong Using}.
8227 @section Strong Using
8229 @strong{Caution:} The semantics of this extension are not fully
8230 defined. Users should refrain from using this extension as its
8231 semantics may change subtly over time. It is possible that this
8232 extension wil be removed in future versions of G++.
8234 A using-directive with @code{__attribute ((strong))} is stronger
8235 than a normal using-directive in two ways:
8239 Templates from the used namespace can be specialized as though they were members of the using namespace.
8242 The using namespace is considered an associated namespace of all
8243 templates in the used namespace for purposes of argument-dependent
8247 This is useful for composing a namespace transparently from
8248 implementation namespaces. For example:
8253 template <class T> struct A @{ @};
8255 using namespace debug __attribute ((__strong__));
8256 template <> struct A<int> @{ @}; // ok to specialize
8258 template <class T> void f (A<T>);
8263 f (std::A<float>()); // lookup finds std::f
8271 G++ uses a syntactic extension to implement the @code{offsetof} macro.
8276 __offsetof__ (expression)
8279 is equivalent to the parenthesized expression, except that the
8280 expression is considered an integral constant expression even if it
8281 contains certain operators that are not normally permitted in an
8282 integral constant expression. Users should never use
8283 @code{__offsetof__} directly; the only valid use of
8284 @code{__offsetof__} is to implement the @code{offsetof} macro in
8287 @node Java Exceptions
8288 @section Java Exceptions
8290 The Java language uses a slightly different exception handling model
8291 from C++. Normally, GNU C++ will automatically detect when you are
8292 writing C++ code that uses Java exceptions, and handle them
8293 appropriately. However, if C++ code only needs to execute destructors
8294 when Java exceptions are thrown through it, GCC will guess incorrectly.
8295 Sample problematic code is:
8298 struct S @{ ~S(); @};
8299 extern void bar(); // is written in Java, and may throw exceptions
8308 The usual effect of an incorrect guess is a link failure, complaining of
8309 a missing routine called @samp{__gxx_personality_v0}.
8311 You can inform the compiler that Java exceptions are to be used in a
8312 translation unit, irrespective of what it might think, by writing
8313 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
8314 @samp{#pragma} must appear before any functions that throw or catch
8315 exceptions, or run destructors when exceptions are thrown through them.
8317 You cannot mix Java and C++ exceptions in the same translation unit. It
8318 is believed to be safe to throw a C++ exception from one file through
8319 another file compiled for the Java exception model, or vice versa, but
8320 there may be bugs in this area.
8322 @node Deprecated Features
8323 @section Deprecated Features
8325 In the past, the GNU C++ compiler was extended to experiment with new
8326 features, at a time when the C++ language was still evolving. Now that
8327 the C++ standard is complete, some of those features are superseded by
8328 superior alternatives. Using the old features might cause a warning in
8329 some cases that the feature will be dropped in the future. In other
8330 cases, the feature might be gone already.
8332 While the list below is not exhaustive, it documents some of the options
8333 that are now deprecated:
8336 @item -fexternal-templates
8337 @itemx -falt-external-templates
8338 These are two of the many ways for G++ to implement template
8339 instantiation. @xref{Template Instantiation}. The C++ standard clearly
8340 defines how template definitions have to be organized across
8341 implementation units. G++ has an implicit instantiation mechanism that
8342 should work just fine for standard-conforming code.
8344 @item -fstrict-prototype
8345 @itemx -fno-strict-prototype
8346 Previously it was possible to use an empty prototype parameter list to
8347 indicate an unspecified number of parameters (like C), rather than no
8348 parameters, as C++ demands. This feature has been removed, except where
8349 it is required for backwards compatibility @xref{Backwards Compatibility}.
8352 The named return value extension has been deprecated, and is now
8355 The use of initializer lists with new expressions has been deprecated,
8356 and is now removed from G++.
8358 Floating and complex non-type template parameters have been deprecated,
8359 and are now removed from G++.
8361 The implicit typename extension has been deprecated and is now
8364 The use of default arguments in function pointers, function typedefs and
8365 and other places where they are not permitted by the standard is
8366 deprecated and will be removed from a future version of G++.
8368 @node Backwards Compatibility
8369 @section Backwards Compatibility
8370 @cindex Backwards Compatibility
8371 @cindex ARM [Annotated C++ Reference Manual]
8373 Now that there is a definitive ISO standard C++, G++ has a specification
8374 to adhere to. The C++ language evolved over time, and features that
8375 used to be acceptable in previous drafts of the standard, such as the ARM
8376 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
8377 compilation of C++ written to such drafts, G++ contains some backwards
8378 compatibilities. @emph{All such backwards compatibility features are
8379 liable to disappear in future versions of G++.} They should be considered
8380 deprecated @xref{Deprecated Features}.
8384 If a variable is declared at for scope, it used to remain in scope until
8385 the end of the scope which contained the for statement (rather than just
8386 within the for scope). G++ retains this, but issues a warning, if such a
8387 variable is accessed outside the for scope.
8389 @item Implicit C language
8390 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
8391 scope to set the language. On such systems, all header files are
8392 implicitly scoped inside a C language scope. Also, an empty prototype
8393 @code{()} will be treated as an unspecified number of arguments, rather
8394 than no arguments, as C++ demands.