1 @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002, 2003
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter C Implementation-defined behavior
8 @cindex implementation-defined behavior, C language
10 A conforming implementation of ISO C is required to document its
11 choice of behavior in each of the areas that are designated
12 ``implementation defined.'' The following lists all such areas,
13 along with the section number from the ISO/IEC 9899:1999 standard.
16 * Translation implementation::
17 * Environment implementation::
18 * Identifiers implementation::
19 * Characters implementation::
20 * Integers implementation::
21 * Floating point implementation::
22 * Arrays and pointers implementation::
23 * Hints implementation::
24 * Structures unions enumerations and bit-fields implementation::
25 * Qualifiers implementation::
26 * Preprocessing directives implementation::
27 * Library functions implementation::
28 * Architecture implementation::
29 * Locale-specific behavior implementation::
32 @node Translation implementation
37 @cite{How a diagnostic is identified (3.10, 5.1.1.3).}
39 Diagnostics consist of all the output sent to stderr by GCC.
42 @cite{Whether each nonempty sequence of white-space characters other than
43 new-line is retained or replaced by one space character in translation
47 @node Environment implementation
50 The behavior of these points are dependent on the implementation
51 of the C library, and are not defined by GCC itself.
53 @node Identifiers implementation
58 @cite{Which additional multibyte characters may appear in identifiers
59 and their correspondence to universal character names (6.4.2).}
62 @cite{The number of significant initial characters in an identifier
65 For internal names, all characters are significant. For external names,
66 the number of significant characters are defined by the linker; for
67 almost all targets, all characters are significant.
71 @node Characters implementation
76 @cite{The number of bits in a byte (3.6).}
79 @cite{The values of the members of the execution character set (5.2.1).}
82 @cite{The unique value of the member of the execution character set produced
83 for each of the standard alphabetic escape sequences (5.2.2).}
86 @cite{The value of a @code{char} object into which has been stored any
87 character other than a member of the basic execution character set (6.2.5).}
90 @cite{Which of @code{signed char} or @code{unsigned char} has the same range,
91 representation, and behavior as ``plain'' @code{char} (6.2.5, 6.3.1.1).}
94 @cite{The mapping of members of the source character set (in character
95 constants and string literals) to members of the execution character
96 set (6.4.4.4, 5.1.1.2).}
99 @cite{The value of an integer character constant containing more than one
100 character or containing a character or escape sequence that does not map
101 to a single-byte execution character (6.4.4.4).}
104 @cite{The value of a wide character constant containing more than one
105 multibyte character, or containing a multibyte character or escape
106 sequence not represented in the extended execution character set (6.4.4.4).}
109 @cite{The current locale used to convert a wide character constant consisting
110 of a single multibyte character that maps to a member of the extended
111 execution character set into a corresponding wide character code (6.4.4.4).}
114 @cite{The current locale used to convert a wide string literal into
115 corresponding wide character codes (6.4.5).}
118 @cite{The value of a string literal containing a multibyte character or escape
119 sequence not represented in the execution character set (6.4.5).}
122 @node Integers implementation
127 @cite{Any extended integer types that exist in the implementation (6.2.5).}
130 @cite{Whether signed integer types are represented using sign and magnitude,
131 two's complement, or one's complement, and whether the extraordinary value
132 is a trap representation or an ordinary value (6.2.6.2).}
134 GCC supports only two's complement integer types, and all bit patterns
138 @cite{The rank of any extended integer type relative to another extended
139 integer type with the same precision (6.3.1.1).}
142 @cite{The result of, or the signal raised by, converting an integer to a
143 signed integer type when the value cannot be represented in an object of
144 that type (6.3.1.3).}
147 @cite{The results of some bitwise operations on signed integers (6.5).}
150 @node Floating point implementation
151 @section Floating point
155 @cite{The accuracy of the floating-point operations and of the library
156 functions in @code{<math.h>} and @code{<complex.h>} that return floating-point
157 results (5.2.4.2.2).}
160 @cite{The rounding behaviors characterized by non-standard values
161 of @code{FLT_ROUNDS} @gol
165 @cite{The evaluation methods characterized by non-standard negative
166 values of @code{FLT_EVAL_METHOD} (5.2.4.2.2).}
169 @cite{The direction of rounding when an integer is converted to a
170 floating-point number that cannot exactly represent the original
174 @cite{The direction of rounding when a floating-point number is
175 converted to a narrower floating-point number (6.3.1.5).}
178 @cite{How the nearest representable value or the larger or smaller
179 representable value immediately adjacent to the nearest representable
180 value is chosen for certain floating constants (6.4.4.2).}
183 @cite{Whether and how floating expressions are contracted when not
184 disallowed by the @code{FP_CONTRACT} pragma (6.5).}
187 @cite{The default state for the @code{FENV_ACCESS} pragma (7.6.1).}
190 @cite{Additional floating-point exceptions, rounding modes, environments,
191 and classifications, and their macro names (7.6, 7.12).}
194 @cite{The default state for the @code{FP_CONTRACT} pragma (7.12.2).}
197 @cite{Whether the ``inexact'' floating-point exception can be raised
198 when the rounded result actually does equal the mathematical result
199 in an IEC 60559 conformant implementation (F.9).}
202 @cite{Whether the ``underflow'' (and ``inexact'') floating-point
203 exception can be raised when a result is tiny but not inexact in an
204 IEC 60559 conformant implementation (F.9).}
208 @node Arrays and pointers implementation
209 @section Arrays and pointers
213 @cite{The result of converting a pointer to an integer or
214 vice versa (6.3.2.3).}
216 A cast from pointer to integer discards most-significant bits if the
217 pointer representation is larger than the integer type,
218 sign-extends@footnote{Future versions of GCC may zero-extend, or use
219 a target-defined @code{ptr_extend} pattern. Do not rely on sign extension.}
220 if the pointer representation is smaller than the integer type, otherwise
221 the bits are unchanged.
222 @c ??? We've always claimed that pointers were unsigned entities.
223 @c Shouldn't we therefore be doing zero-extension? If so, the bug
224 @c is in convert_to_integer, where we call type_for_size and request
225 @c a signed integral type. On the other hand, it might be most useful
226 @c for the target if we extend according to POINTERS_EXTEND_UNSIGNED.
228 A cast from integer to pointer discards most-significant bits if the
229 pointer representation is smaller than the integer type, extends according
230 to the signedness of the integer type if the pointer representation
231 is larger than the integer type, otherwise the bits are unchanged.
233 When casting from pointer to integer and back again, the resulting
234 pointer must reference the same object as the original pointer, otherwise
235 the behavior is undefined. That is, one may not use integer arithmetic to
236 avoid the undefined behavior of pointer arithmetic as proscribed in 6.5.6/8.
239 @cite{The size of the result of subtracting two pointers to elements
240 of the same array (6.5.6).}
244 @node Hints implementation
249 @cite{The extent to which suggestions made by using the @code{register}
250 storage-class specifier are effective (6.7.1).}
252 The @code{register} specifier affects code generation only in these ways:
256 When used as part of the register variable extension, see
257 @ref{Explicit Reg Vars}.
260 When @option{-O0} is in use, the compiler allocates distinct stack
261 memory for all variables that do not have the @code{register}
262 storage-class specifier; if @code{register} is specified, the variable
263 may have a shorter lifespan than the code would indicate and may never
267 On some rare x86 targets, @code{setjmp} doesn't save the registers in
268 all circumstances. In those cases, GCC doesn't allocate any variables
269 in registers unless they are marked @code{register}.
274 @cite{The extent to which suggestions made by using the inline function
275 specifier are effective (6.7.4).}
277 GCC will not inline any functions if the @option{-fno-inline} option is
278 used or if @option{-O0} is used. Otherwise, GCC may still be unable to
279 inline a function for many reasons; the @option{-Winline} option may be
280 used to determine if a function has not been inlined and why not.
284 @node Structures unions enumerations and bit-fields implementation
285 @section Structures, unions, enumerations, and bit-fields
289 @cite{Whether a ``plain'' int bit-field is treated as a @code{signed int}
290 bit-field or as an @code{unsigned int} bit-field (6.7.2, 6.7.2.1).}
293 @cite{Allowable bit-field types other than @code{_Bool}, @code{signed int},
294 and @code{unsigned int} (6.7.2.1).}
297 @cite{Whether a bit-field can straddle a storage-unit boundary (6.7.2.1).}
300 @cite{The order of allocation of bit-fields within a unit (6.7.2.1).}
303 @cite{The alignment of non-bit-field members of structures (6.7.2.1).}
306 @cite{The integer type compatible with each enumerated type (6.7.2.2).}
310 @node Qualifiers implementation
315 @cite{What constitutes an access to an object that has volatile-qualified
320 @node Preprocessing directives implementation
321 @section Preprocessing directives
325 @cite{How sequences in both forms of header names are mapped to headers
326 or external source file names (6.4.7).}
329 @cite{Whether the value of a character constant in a constant expression
330 that controls conditional inclusion matches the value of the same character
331 constant in the execution character set (6.10.1).}
334 @cite{Whether the value of a single-character character constant in a
335 constant expression that controls conditional inclusion may have a
336 negative value (6.10.1).}
339 @cite{The places that are searched for an included @samp{<>} delimited
340 header, and how the places are specified or the header is
341 identified (6.10.2).}
344 @cite{How the named source file is searched for in an included @samp{""}
345 delimited header (6.10.2).}
348 @cite{The method by which preprocessing tokens (possibly resulting from
349 macro expansion) in a @code{#include} directive are combined into a header
353 @cite{The nesting limit for @code{#include} processing (6.10.2).}
355 GCC imposes a limit of 200 nested @code{#include}s.
358 @cite{Whether the @samp{#} operator inserts a @samp{\} character before
359 the @samp{\} character that begins a universal character name in a
360 character constant or string literal (6.10.3.2).}
363 @cite{The behavior on each recognized non-@code{STDC #pragma}
367 @cite{The definitions for @code{__DATE__} and @code{__TIME__} when
368 respectively, the date and time of translation are not available (6.10.8).}
370 If the date and time are not available, @code{__DATE__} expands to
371 @code{@w{"??? ?? ????"}} and @code{__TIME__} expands to
376 @node Library functions implementation
377 @section Library functions
379 The behavior of these points are dependent on the implementation
380 of the C library, and are not defined by GCC itself.
382 @node Architecture implementation
383 @section Architecture
387 @cite{The values or expressions assigned to the macros specified in the
388 headers @code{<float.h>}, @code{<limits.h>}, and @code{<stdint.h>}
389 (5.2.4.2, 7.18.2, 7.18.3).}
392 @cite{The number, order, and encoding of bytes in any object
393 (when not explicitly specified in this International Standard) (6.2.6.1).}
396 @cite{The value of the result of the sizeof operator (6.5.3.4).}
400 @node Locale-specific behavior implementation
401 @section Locale-specific behavior
403 The behavior of these points are dependent on the implementation
404 of the C library, and are not defined by GCC itself.
407 @chapter Extensions to the C Language Family
408 @cindex extensions, C language
409 @cindex C language extensions
412 GNU C provides several language features not found in ISO standard C@.
413 (The @option{-pedantic} option directs GCC to print a warning message if
414 any of these features is used.) To test for the availability of these
415 features in conditional compilation, check for a predefined macro
416 @code{__GNUC__}, which is always defined under GCC@.
418 These extensions are available in C and Objective-C@. Most of them are
419 also available in C++. @xref{C++ Extensions,,Extensions to the
420 C++ Language}, for extensions that apply @emph{only} to C++.
422 Some features that are in ISO C99 but not C89 or C++ are also, as
423 extensions, accepted by GCC in C89 mode and in C++.
426 * Statement Exprs:: Putting statements and declarations inside expressions.
427 * Local Labels:: Labels local to a 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 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
433 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
434 * Long Long:: Double-word integers---@code{long long int}.
435 * Complex:: Data types for complex numbers.
436 * Hex Floats:: Hexadecimal floating-point constants.
437 * Zero Length:: Zero-length arrays.
438 * Variable Length:: Arrays whose length is computed at run time.
439 * Empty Structures:: Structures with no members.
440 * Variadic Macros:: Macros with a variable number of arguments.
441 * Escaped Newlines:: Slightly looser rules for escaped newlines.
442 * Subscripting:: Any array can be subscripted, even if not an lvalue.
443 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
444 * Initializers:: Non-constant initializers.
445 * Compound Literals:: Compound literals give structures, unions
447 * Designated Inits:: Labeling elements of initializers.
448 * Cast to Union:: Casting to union type from any member of the union.
449 * Case Ranges:: `case 1 ... 9' and such.
450 * Mixed Declarations:: Mixing declarations and code.
451 * Function Attributes:: Declaring that functions have no side effects,
452 or that they can never return.
453 * Attribute Syntax:: Formal syntax for attributes.
454 * Function Prototypes:: Prototype declarations and old-style definitions.
455 * C++ Comments:: C++ comments are recognized.
456 * Dollar Signs:: Dollar sign is allowed in identifiers.
457 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
458 * Variable Attributes:: Specifying attributes of variables.
459 * Type Attributes:: Specifying attributes of types.
460 * Alignment:: Inquiring about the alignment of a type or variable.
461 * Inline:: Defining inline functions (as fast as macros).
462 * Extended Asm:: Assembler instructions with C expressions as operands.
463 (With them you can define ``built-in'' functions.)
464 * Constraints:: Constraints for asm operands
465 * Asm Labels:: Specifying the assembler name to use for a C symbol.
466 * Explicit Reg Vars:: Defining variables residing in specified registers.
467 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
468 * Incomplete Enums:: @code{enum foo;}, with details to follow.
469 * Function Names:: Printable strings which are the name of the current
471 * Return Address:: Getting the return or frame address of a function.
472 * Vector Extensions:: Using vector instructions through built-in functions.
473 * Other Builtins:: Other built-in functions.
474 * Target Builtins:: Built-in functions specific to particular targets.
475 * Pragmas:: Pragmas accepted by GCC.
476 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
477 * Thread-Local:: Per-thread variables.
480 @node Statement Exprs
481 @section Statements and Declarations in Expressions
482 @cindex statements inside expressions
483 @cindex declarations inside expressions
484 @cindex expressions containing statements
485 @cindex macros, statements in expressions
487 @c the above section title wrapped and causes an underfull hbox.. i
488 @c changed it from "within" to "in". --mew 4feb93
489 A compound statement enclosed in parentheses may appear as an expression
490 in GNU C@. This allows you to use loops, switches, and local variables
491 within an expression.
493 Recall that a compound statement is a sequence of statements surrounded
494 by braces; in this construct, parentheses go around the braces. For
498 (@{ int y = foo (); int z;
505 is a valid (though slightly more complex than necessary) expression
506 for the absolute value of @code{foo ()}.
508 The last thing in the compound statement should be an expression
509 followed by a semicolon; the value of this subexpression serves as the
510 value of the entire construct. (If you use some other kind of statement
511 last within the braces, the construct has type @code{void}, and thus
512 effectively no value.)
514 This feature is especially useful in making macro definitions ``safe'' (so
515 that they evaluate each operand exactly once). For example, the
516 ``maximum'' function is commonly defined as a macro in standard C as
520 #define max(a,b) ((a) > (b) ? (a) : (b))
524 @cindex side effects, macro argument
525 But this definition computes either @var{a} or @var{b} twice, with bad
526 results if the operand has side effects. In GNU C, if you know the
527 type of the operands (here let's assume @code{int}), you can define
528 the macro safely as follows:
531 #define maxint(a,b) \
532 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
535 Embedded statements are not allowed in constant expressions, such as
536 the value of an enumeration constant, the width of a bit-field, or
537 the initial value of a static variable.
539 If you don't know the type of the operand, you can still do this, but you
540 must use @code{typeof} (@pxref{Typeof}).
542 In G++, the result value of a statement expression undergoes array and
543 function pointer decay, and is returned by value to the enclosing
544 expression. For instance, if @code{A} is a class, then
553 will construct a temporary @code{A} object to hold the result of the
554 statement expression, and that will be used to invoke @code{Foo}.
555 Therefore the @code{this} pointer observed by @code{Foo} will not be the
558 Any temporaries created within a statement within a statement expression
559 will be destroyed at the statement's end. This makes statement
560 expressions inside macros slightly different from function calls. In
561 the latter case temporaries introduced during argument evaluation will
562 be destroyed at the end of the statement that includes the function
563 call. In the statement expression case they will be destroyed during
564 the statement expression. For instance,
567 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
568 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
578 will have different places where temporaries are destroyed. For the
579 @code{macro} case, the temporary @code{X} will be destroyed just after
580 the initialization of @code{b}. In the @code{function} case that
581 temporary will be destroyed when the function returns.
583 These considerations mean that it is probably a bad idea to use
584 statement-expressions of this form in header files that are designed to
585 work with C++. (Note that some versions of the GNU C Library contained
586 header files using statement-expression that lead to precisely this
590 @section Locally Declared Labels
592 @cindex macros, local labels
594 GCC allows you to declare @dfn{local labels} in any nested block
595 scope. A local label is just like an ordinary label, but you can
596 only reference it (with a @code{goto} statement, or by taking its
597 address) within the block in which it was declared.
599 A local label declaration looks like this:
602 __label__ @var{label};
609 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
612 Local label declarations must come at the beginning of the block,
613 before any ordinary declarations or statements.
615 The label declaration defines the label @emph{name}, but does not define
616 the label itself. You must do this in the usual way, with
617 @code{@var{label}:}, within the statements of the statement expression.
619 The local label feature is useful for complex macros. If a macro
620 contains nested loops, a @code{goto} can be useful for breaking out of
621 them. However, an ordinary label whose scope is the whole function
622 cannot be used: if the macro can be expanded several times in one
623 function, the label will be multiply defined in that function. A
624 local label avoids this problem. For example:
627 #define SEARCH(value, array, target) \
630 typeof (target) _SEARCH_target = (target); \
631 typeof (*(array)) *_SEARCH_array = (array); \
634 for (i = 0; i < max; i++) \
635 for (j = 0; j < max; j++) \
636 if (_SEARCH_array[i][j] == _SEARCH_target) \
637 @{ (value) = i; goto found; @} \
643 This could also be written using a statement-expression:
646 #define SEARCH(array, target) \
649 typeof (target) _SEARCH_target = (target); \
650 typeof (*(array)) *_SEARCH_array = (array); \
653 for (i = 0; i < max; i++) \
654 for (j = 0; j < max; j++) \
655 if (_SEARCH_array[i][j] == _SEARCH_target) \
656 @{ value = i; goto found; @} \
663 Local label declarations also make the labels they declare visible to
664 nested functions, if there are any. @xref{Nested Functions}, for details.
666 @node Labels as Values
667 @section Labels as Values
668 @cindex labels as values
669 @cindex computed gotos
670 @cindex goto with computed label
671 @cindex address of a label
673 You can get the address of a label defined in the current function
674 (or a containing function) with the unary operator @samp{&&}. The
675 value has type @code{void *}. This value is a constant and can be used
676 wherever a constant of that type is valid. For example:
684 To use these values, you need to be able to jump to one. This is done
685 with the computed goto statement@footnote{The analogous feature in
686 Fortran is called an assigned goto, but that name seems inappropriate in
687 C, where one can do more than simply store label addresses in label
688 variables.}, @code{goto *@var{exp};}. For example,
695 Any expression of type @code{void *} is allowed.
697 One way of using these constants is in initializing a static array that
698 will serve as a jump table:
701 static void *array[] = @{ &&foo, &&bar, &&hack @};
704 Then you can select a label with indexing, like this:
711 Note that this does not check whether the subscript is in bounds---array
712 indexing in C never does that.
714 Such an array of label values serves a purpose much like that of the
715 @code{switch} statement. The @code{switch} statement is cleaner, so
716 use that rather than an array unless the problem does not fit a
717 @code{switch} statement very well.
719 Another use of label values is in an interpreter for threaded code.
720 The labels within the interpreter function can be stored in the
721 threaded code for super-fast dispatching.
723 You may not use this mechanism to jump to code in a different function.
724 If you do that, totally unpredictable things will happen. The best way to
725 avoid this is to store the label address only in automatic variables and
726 never pass it as an argument.
728 An alternate way to write the above example is
731 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
733 goto *(&&foo + array[i]);
737 This is more friendly to code living in shared libraries, as it reduces
738 the number of dynamic relocations that are needed, and by consequence,
739 allows the data to be read-only.
741 @node Nested Functions
742 @section Nested Functions
743 @cindex nested functions
744 @cindex downward funargs
747 A @dfn{nested function} is a function defined inside another function.
748 (Nested functions are not supported for GNU C++.) The nested function's
749 name is local to the block where it is defined. For example, here we
750 define a nested function named @code{square}, and call it twice:
754 foo (double a, double b)
756 double square (double z) @{ return z * z; @}
758 return square (a) + square (b);
763 The nested function can access all the variables of the containing
764 function that are visible at the point of its definition. This is
765 called @dfn{lexical scoping}. For example, here we show a nested
766 function which uses an inherited variable named @code{offset}:
770 bar (int *array, int offset, int size)
772 int access (int *array, int index)
773 @{ return array[index + offset]; @}
776 for (i = 0; i < size; i++)
777 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
782 Nested function definitions are permitted within functions in the places
783 where variable definitions are allowed; that is, in any block, before
784 the first statement in the block.
786 It is possible to call the nested function from outside the scope of its
787 name by storing its address or passing the address to another function:
790 hack (int *array, int size)
792 void store (int index, int value)
793 @{ array[index] = value; @}
795 intermediate (store, size);
799 Here, the function @code{intermediate} receives the address of
800 @code{store} as an argument. If @code{intermediate} calls @code{store},
801 the arguments given to @code{store} are used to store into @code{array}.
802 But this technique works only so long as the containing function
803 (@code{hack}, in this example) does not exit.
805 If you try to call the nested function through its address after the
806 containing function has exited, all hell will break loose. If you try
807 to call it after a containing scope level has exited, and if it refers
808 to some of the variables that are no longer in scope, you may be lucky,
809 but it's not wise to take the risk. If, however, the nested function
810 does not refer to anything that has gone out of scope, you should be
813 GCC implements taking the address of a nested function using a technique
814 called @dfn{trampolines}. A paper describing them is available as
817 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
819 A nested function can jump to a label inherited from a containing
820 function, provided the label was explicitly declared in the containing
821 function (@pxref{Local Labels}). Such a jump returns instantly to the
822 containing function, exiting the nested function which did the
823 @code{goto} and any intermediate functions as well. Here is an example:
827 bar (int *array, int offset, int size)
830 int access (int *array, int index)
834 return array[index + offset];
838 for (i = 0; i < size; i++)
839 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
843 /* @r{Control comes here from @code{access}
844 if it detects an error.} */
851 A nested function always has internal linkage. Declaring one with
852 @code{extern} is erroneous. If you need to declare the nested function
853 before its definition, use @code{auto} (which is otherwise meaningless
854 for function declarations).
857 bar (int *array, int offset, int size)
860 auto int access (int *, int);
862 int access (int *array, int index)
866 return array[index + offset];
872 @node Constructing Calls
873 @section Constructing Function Calls
874 @cindex constructing calls
875 @cindex forwarding calls
877 Using the built-in functions described below, you can record
878 the arguments a function received, and call another function
879 with the same arguments, without knowing the number or types
882 You can also record the return value of that function call,
883 and later return that value, without knowing what data type
884 the function tried to return (as long as your caller expects
887 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
888 This built-in function returns a pointer to data
889 describing how to perform a call with the same arguments as were passed
890 to the current function.
892 The function saves the arg pointer register, structure value address,
893 and all registers that might be used to pass arguments to a function
894 into a block of memory allocated on the stack. Then it returns the
895 address of that block.
898 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
899 This built-in function invokes @var{function}
900 with a copy of the parameters described by @var{arguments}
903 The value of @var{arguments} should be the value returned by
904 @code{__builtin_apply_args}. The argument @var{size} specifies the size
905 of the stack argument data, in bytes.
907 This function returns a pointer to data describing
908 how to return whatever value was returned by @var{function}. The data
909 is saved in a block of memory allocated on the stack.
911 It is not always simple to compute the proper value for @var{size}. The
912 value is used by @code{__builtin_apply} to compute the amount of data
913 that should be pushed on the stack and copied from the incoming argument
917 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
918 This built-in function returns the value described by @var{result} from
919 the containing function. You should specify, for @var{result}, a value
920 returned by @code{__builtin_apply}.
924 @section Referring to a Type with @code{typeof}
927 @cindex macros, types of arguments
929 Another way to refer to the type of an expression is with @code{typeof}.
930 The syntax of using of this keyword looks like @code{sizeof}, but the
931 construct acts semantically like a type name defined with @code{typedef}.
933 There are two ways of writing the argument to @code{typeof}: with an
934 expression or with a type. Here is an example with an expression:
941 This assumes that @code{x} is an array of pointers to functions;
942 the type described is that of the values of the functions.
944 Here is an example with a typename as the argument:
951 Here the type described is that of pointers to @code{int}.
953 If you are writing a header file that must work when included in ISO C
954 programs, write @code{__typeof__} instead of @code{typeof}.
955 @xref{Alternate Keywords}.
957 A @code{typeof}-construct can be used anywhere a typedef name could be
958 used. For example, you can use it in a declaration, in a cast, or inside
959 of @code{sizeof} or @code{typeof}.
961 @code{typeof} is often useful in conjunction with the
962 statements-within-expressions feature. Here is how the two together can
963 be used to define a safe ``maximum'' macro that operates on any
964 arithmetic type and evaluates each of its arguments exactly once:
968 (@{ typeof (a) _a = (a); \
969 typeof (b) _b = (b); \
970 _a > _b ? _a : _b; @})
973 @cindex underscores in variables in macros
974 @cindex @samp{_} in variables in macros
975 @cindex local variables in macros
976 @cindex variables, local, in macros
977 @cindex macros, local variables in
979 The reason for using names that start with underscores for the local
980 variables is to avoid conflicts with variable names that occur within the
981 expressions that are substituted for @code{a} and @code{b}. Eventually we
982 hope to design a new form of declaration syntax that allows you to declare
983 variables whose scopes start only after their initializers; this will be a
984 more reliable way to prevent such conflicts.
987 Some more examples of the use of @code{typeof}:
991 This declares @code{y} with the type of what @code{x} points to.
998 This declares @code{y} as an array of such values.
1005 This declares @code{y} as an array of pointers to characters:
1008 typeof (typeof (char *)[4]) y;
1012 It is equivalent to the following traditional C declaration:
1018 To see the meaning of the declaration using @code{typeof}, and why it
1019 might be a useful way to write, let's rewrite it with these macros:
1022 #define pointer(T) typeof(T *)
1023 #define array(T, N) typeof(T [N])
1027 Now the declaration can be rewritten this way:
1030 array (pointer (char), 4) y;
1034 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1035 pointers to @code{char}.
1038 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1039 a more limited extension which permitted one to write
1042 typedef @var{T} = @var{expr};
1046 with the effect of declaring @var{T} to have the type of the expression
1047 @var{expr}. This extension does not work with GCC 3 (versions between
1048 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1049 relies on it should be rewritten to use @code{typeof}:
1052 typedef typeof(@var{expr}) @var{T};
1056 This will work with all versions of GCC@.
1059 @section Generalized Lvalues
1060 @cindex compound expressions as lvalues
1061 @cindex expressions, compound, as lvalues
1062 @cindex conditional expressions as lvalues
1063 @cindex expressions, conditional, as lvalues
1064 @cindex casts as lvalues
1065 @cindex generalized lvalues
1066 @cindex lvalues, generalized
1067 @cindex extensions, @code{?:}
1068 @cindex @code{?:} extensions
1070 Compound expressions, conditional expressions and casts are allowed as
1071 lvalues provided their operands are lvalues. This means that you can take
1072 their addresses or store values into them.
1074 Standard C++ allows compound expressions and conditional expressions as
1075 lvalues, and permits casts to reference type, so use of this extension
1076 is deprecated for C++ code.
1078 For example, a compound expression can be assigned, provided the last
1079 expression in the sequence is an lvalue. These two expressions are
1087 Similarly, the address of the compound expression can be taken. These two
1088 expressions are equivalent:
1095 A conditional expression is a valid lvalue if its type is not void and the
1096 true and false branches are both valid lvalues. For example, these two
1097 expressions are equivalent:
1101 (a ? b = 5 : (c = 5))
1104 A cast is a valid lvalue if its operand is an lvalue. A simple
1105 assignment whose left-hand side is a cast works by converting the
1106 right-hand side first to the specified type, then to the type of the
1107 inner left-hand side expression. After this is stored, the value is
1108 converted back to the specified type to become the value of the
1109 assignment. Thus, if @code{a} has type @code{char *}, the following two
1110 expressions are equivalent:
1114 (int)(a = (char *)(int)5)
1117 An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
1118 performs the arithmetic using the type resulting from the cast, and then
1119 continues as in the previous case. Therefore, these two expressions are
1124 (int)(a = (char *)(int) ((int)a + 5))
1127 You cannot take the address of an lvalue cast, because the use of its
1128 address would not work out coherently. Suppose that @code{&(int)f} were
1129 permitted, where @code{f} has type @code{float}. Then the following
1130 statement would try to store an integer bit-pattern where a floating
1131 point number belongs:
1137 This is quite different from what @code{(int)f = 1} would do---that
1138 would convert 1 to floating point and store it. Rather than cause this
1139 inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
1141 If you really do want an @code{int *} pointer with the address of
1142 @code{f}, you can simply write @code{(int *)&f}.
1145 @section Conditionals with Omitted Operands
1146 @cindex conditional expressions, extensions
1147 @cindex omitted middle-operands
1148 @cindex middle-operands, omitted
1149 @cindex extensions, @code{?:}
1150 @cindex @code{?:} extensions
1152 The middle operand in a conditional expression may be omitted. Then
1153 if the first operand is nonzero, its value is the value of the conditional
1156 Therefore, the expression
1163 has the value of @code{x} if that is nonzero; otherwise, the value of
1166 This example is perfectly equivalent to
1172 @cindex side effect in ?:
1173 @cindex ?: side effect
1175 In this simple case, the ability to omit the middle operand is not
1176 especially useful. When it becomes useful is when the first operand does,
1177 or may (if it is a macro argument), contain a side effect. Then repeating
1178 the operand in the middle would perform the side effect twice. Omitting
1179 the middle operand uses the value already computed without the undesirable
1180 effects of recomputing it.
1183 @section Double-Word Integers
1184 @cindex @code{long long} data types
1185 @cindex double-word arithmetic
1186 @cindex multiprecision arithmetic
1187 @cindex @code{LL} integer suffix
1188 @cindex @code{ULL} integer suffix
1190 ISO C99 supports data types for integers that are at least 64 bits wide,
1191 and as an extension GCC supports them in C89 mode and in C++.
1192 Simply write @code{long long int} for a signed integer, or
1193 @code{unsigned long long int} for an unsigned integer. To make an
1194 integer constant of type @code{long long int}, add the suffix @samp{LL}
1195 to the integer. To make an integer constant of type @code{unsigned long
1196 long int}, add the suffix @samp{ULL} to the integer.
1198 You can use these types in arithmetic like any other integer types.
1199 Addition, subtraction, and bitwise boolean operations on these types
1200 are open-coded on all types of machines. Multiplication is open-coded
1201 if the machine supports fullword-to-doubleword a widening multiply
1202 instruction. Division and shifts are open-coded only on machines that
1203 provide special support. The operations that are not open-coded use
1204 special library routines that come with GCC@.
1206 There may be pitfalls when you use @code{long long} types for function
1207 arguments, unless you declare function prototypes. If a function
1208 expects type @code{int} for its argument, and you pass a value of type
1209 @code{long long int}, confusion will result because the caller and the
1210 subroutine will disagree about the number of bytes for the argument.
1211 Likewise, if the function expects @code{long long int} and you pass
1212 @code{int}. The best way to avoid such problems is to use prototypes.
1215 @section Complex Numbers
1216 @cindex complex numbers
1217 @cindex @code{_Complex} keyword
1218 @cindex @code{__complex__} keyword
1220 ISO C99 supports complex floating data types, and as an extension GCC
1221 supports them in C89 mode and in C++, and supports complex integer data
1222 types which are not part of ISO C99. You can declare complex types
1223 using the keyword @code{_Complex}. As an extension, the older GNU
1224 keyword @code{__complex__} is also supported.
1226 For example, @samp{_Complex double x;} declares @code{x} as a
1227 variable whose real part and imaginary part are both of type
1228 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1229 have real and imaginary parts of type @code{short int}; this is not
1230 likely to be useful, but it shows that the set of complex types is
1233 To write a constant with a complex data type, use the suffix @samp{i} or
1234 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1235 has type @code{_Complex float} and @code{3i} has type
1236 @code{_Complex int}. Such a constant always has a pure imaginary
1237 value, but you can form any complex value you like by adding one to a
1238 real constant. This is a GNU extension; if you have an ISO C99
1239 conforming C library (such as GNU libc), and want to construct complex
1240 constants of floating type, you should include @code{<complex.h>} and
1241 use the macros @code{I} or @code{_Complex_I} instead.
1243 @cindex @code{__real__} keyword
1244 @cindex @code{__imag__} keyword
1245 To extract the real part of a complex-valued expression @var{exp}, write
1246 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1247 extract the imaginary part. This is a GNU extension; for values of
1248 floating type, you should use the ISO C99 functions @code{crealf},
1249 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1250 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1251 built-in functions by GCC@.
1253 @cindex complex conjugation
1254 The operator @samp{~} performs complex conjugation when used on a value
1255 with a complex type. This is a GNU extension; for values of
1256 floating type, you should use the ISO C99 functions @code{conjf},
1257 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1258 provided as built-in functions by GCC@.
1260 GCC can allocate complex automatic variables in a noncontiguous
1261 fashion; it's even possible for the real part to be in a register while
1262 the imaginary part is on the stack (or vice-versa). Only the DWARF2
1263 debug info format can represent this, so use of DWARF2 is recommended.
1264 If you are using the stabs debug info format, GCC describes a noncontiguous
1265 complex variable as if it were two separate variables of noncomplex type.
1266 If the variable's actual name is @code{foo}, the two fictitious
1267 variables are named @code{foo$real} and @code{foo$imag}. You can
1268 examine and set these two fictitious variables with your debugger.
1274 ISO C99 supports floating-point numbers written not only in the usual
1275 decimal notation, such as @code{1.55e1}, but also numbers such as
1276 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1277 supports this in C89 mode (except in some cases when strictly
1278 conforming) and in C++. In that format the
1279 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1280 mandatory. The exponent is a decimal number that indicates the power of
1281 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1288 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1289 is the same as @code{1.55e1}.
1291 Unlike for floating-point numbers in the decimal notation the exponent
1292 is always required in the hexadecimal notation. Otherwise the compiler
1293 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1294 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1295 extension for floating-point constants of type @code{float}.
1298 @section Arrays of Length Zero
1299 @cindex arrays of length zero
1300 @cindex zero-length arrays
1301 @cindex length-zero arrays
1302 @cindex flexible array members
1304 Zero-length arrays are allowed in GNU C@. They are very useful as the
1305 last element of a structure which is really a header for a variable-length
1314 struct line *thisline = (struct line *)
1315 malloc (sizeof (struct line) + this_length);
1316 thisline->length = this_length;
1319 In ISO C90, you would have to give @code{contents} a length of 1, which
1320 means either you waste space or complicate the argument to @code{malloc}.
1322 In ISO C99, you would use a @dfn{flexible array member}, which is
1323 slightly different in syntax and semantics:
1327 Flexible array members are written as @code{contents[]} without
1331 Flexible array members have incomplete type, and so the @code{sizeof}
1332 operator may not be applied. As a quirk of the original implementation
1333 of zero-length arrays, @code{sizeof} evaluates to zero.
1336 Flexible array members may only appear as the last member of a
1337 @code{struct} that is otherwise non-empty.
1340 A structure containing a flexible array member, or a union containing
1341 such a structure (possibly recursively), may not be a member of a
1342 structure or an element of an array. (However, these uses are
1343 permitted by GCC as extensions.)
1346 GCC versions before 3.0 allowed zero-length arrays to be statically
1347 initialized, as if they were flexible arrays. In addition to those
1348 cases that were useful, it also allowed initializations in situations
1349 that would corrupt later data. Non-empty initialization of zero-length
1350 arrays is now treated like any case where there are more initializer
1351 elements than the array holds, in that a suitable warning about "excess
1352 elements in array" is given, and the excess elements (all of them, in
1353 this case) are ignored.
1355 Instead GCC allows static initialization of flexible array members.
1356 This is equivalent to defining a new structure containing the original
1357 structure followed by an array of sufficient size to contain the data.
1358 I.e.@: in the following, @code{f1} is constructed as if it were declared
1364 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1367 struct f1 f1; int data[3];
1368 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1372 The convenience of this extension is that @code{f1} has the desired
1373 type, eliminating the need to consistently refer to @code{f2.f1}.
1375 This has symmetry with normal static arrays, in that an array of
1376 unknown size is also written with @code{[]}.
1378 Of course, this extension only makes sense if the extra data comes at
1379 the end of a top-level object, as otherwise we would be overwriting
1380 data at subsequent offsets. To avoid undue complication and confusion
1381 with initialization of deeply nested arrays, we simply disallow any
1382 non-empty initialization except when the structure is the top-level
1383 object. For example:
1386 struct foo @{ int x; int y[]; @};
1387 struct bar @{ struct foo z; @};
1389 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1390 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1391 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1392 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1395 @node Empty Structures
1396 @section Structures With No Members
1397 @cindex empty structures
1398 @cindex zero-size structures
1400 GCC permits a C structure to have no members:
1407 The structure will have size zero. In C++, empty structures are part
1408 of the language. G++ treats empty structures as if they had a single
1409 member of type @code{char}.
1411 @node Variable Length
1412 @section Arrays of Variable Length
1413 @cindex variable-length arrays
1414 @cindex arrays of variable length
1417 Variable-length automatic arrays are allowed in ISO C99, and as an
1418 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1419 implementation of variable-length arrays does not yet conform in detail
1420 to the ISO C99 standard.) These arrays are
1421 declared like any other automatic arrays, but with a length that is not
1422 a constant expression. The storage is allocated at the point of
1423 declaration and deallocated when the brace-level is exited. For
1428 concat_fopen (char *s1, char *s2, char *mode)
1430 char str[strlen (s1) + strlen (s2) + 1];
1433 return fopen (str, mode);
1437 @cindex scope of a variable length array
1438 @cindex variable-length array scope
1439 @cindex deallocating variable length arrays
1440 Jumping or breaking out of the scope of the array name deallocates the
1441 storage. Jumping into the scope is not allowed; you get an error
1444 @cindex @code{alloca} vs variable-length arrays
1445 You can use the function @code{alloca} to get an effect much like
1446 variable-length arrays. The function @code{alloca} is available in
1447 many other C implementations (but not in all). On the other hand,
1448 variable-length arrays are more elegant.
1450 There are other differences between these two methods. Space allocated
1451 with @code{alloca} exists until the containing @emph{function} returns.
1452 The space for a variable-length array is deallocated as soon as the array
1453 name's scope ends. (If you use both variable-length arrays and
1454 @code{alloca} in the same function, deallocation of a variable-length array
1455 will also deallocate anything more recently allocated with @code{alloca}.)
1457 You can also use variable-length arrays as arguments to functions:
1461 tester (int len, char data[len][len])
1467 The length of an array is computed once when the storage is allocated
1468 and is remembered for the scope of the array in case you access it with
1471 If you want to pass the array first and the length afterward, you can
1472 use a forward declaration in the parameter list---another GNU extension.
1476 tester (int len; char data[len][len], int len)
1482 @cindex parameter forward declaration
1483 The @samp{int len} before the semicolon is a @dfn{parameter forward
1484 declaration}, and it serves the purpose of making the name @code{len}
1485 known when the declaration of @code{data} is parsed.
1487 You can write any number of such parameter forward declarations in the
1488 parameter list. They can be separated by commas or semicolons, but the
1489 last one must end with a semicolon, which is followed by the ``real''
1490 parameter declarations. Each forward declaration must match a ``real''
1491 declaration in parameter name and data type. ISO C99 does not support
1492 parameter forward declarations.
1494 @node Variadic Macros
1495 @section Macros with a Variable Number of Arguments.
1496 @cindex variable number of arguments
1497 @cindex macro with variable arguments
1498 @cindex rest argument (in macro)
1499 @cindex variadic macros
1501 In the ISO C standard of 1999, a macro can be declared to accept a
1502 variable number of arguments much as a function can. The syntax for
1503 defining the macro is similar to that of a function. Here is an
1507 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1510 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1511 such a macro, it represents the zero or more tokens until the closing
1512 parenthesis that ends the invocation, including any commas. This set of
1513 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1514 wherever it appears. See the CPP manual for more information.
1516 GCC has long supported variadic macros, and used a different syntax that
1517 allowed you to give a name to the variable arguments just like any other
1518 argument. Here is an example:
1521 #define debug(format, args...) fprintf (stderr, format, args)
1524 This is in all ways equivalent to the ISO C example above, but arguably
1525 more readable and descriptive.
1527 GNU CPP has two further variadic macro extensions, and permits them to
1528 be used with either of the above forms of macro definition.
1530 In standard C, you are not allowed to leave the variable argument out
1531 entirely; but you are allowed to pass an empty argument. For example,
1532 this invocation is invalid in ISO C, because there is no comma after
1539 GNU CPP permits you to completely omit the variable arguments in this
1540 way. In the above examples, the compiler would complain, though since
1541 the expansion of the macro still has the extra comma after the format
1544 To help solve this problem, CPP behaves specially for variable arguments
1545 used with the token paste operator, @samp{##}. If instead you write
1548 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1551 and if the variable arguments are omitted or empty, the @samp{##}
1552 operator causes the preprocessor to remove the comma before it. If you
1553 do provide some variable arguments in your macro invocation, GNU CPP
1554 does not complain about the paste operation and instead places the
1555 variable arguments after the comma. Just like any other pasted macro
1556 argument, these arguments are not macro expanded.
1558 @node Escaped Newlines
1559 @section Slightly Looser Rules for Escaped Newlines
1560 @cindex escaped newlines
1561 @cindex newlines (escaped)
1563 Recently, the preprocessor has relaxed its treatment of escaped
1564 newlines. Previously, the newline had to immediately follow a
1565 backslash. The current implementation allows whitespace in the form
1566 of spaces, horizontal and vertical tabs, and form feeds between the
1567 backslash and the subsequent newline. The preprocessor issues a
1568 warning, but treats it as a valid escaped newline and combines the two
1569 lines to form a single logical line. This works within comments and
1570 tokens, as well as between tokens. Comments are @emph{not} treated as
1571 whitespace for the purposes of this relaxation, since they have not
1572 yet been replaced with spaces.
1575 @section Non-Lvalue Arrays May Have Subscripts
1576 @cindex subscripting
1577 @cindex arrays, non-lvalue
1579 @cindex subscripting and function values
1580 In ISO C99, arrays that are not lvalues still decay to pointers, and
1581 may be subscripted, although they may not be modified or used after
1582 the next sequence point and the unary @samp{&} operator may not be
1583 applied to them. As an extension, GCC allows such arrays to be
1584 subscripted in C89 mode, though otherwise they do not decay to
1585 pointers outside C99 mode. For example,
1586 this is valid in GNU C though not valid in C89:
1590 struct foo @{int a[4];@};
1596 return f().a[index];
1602 @section Arithmetic on @code{void}- and Function-Pointers
1603 @cindex void pointers, arithmetic
1604 @cindex void, size of pointer to
1605 @cindex function pointers, arithmetic
1606 @cindex function, size of pointer to
1608 In GNU C, addition and subtraction operations are supported on pointers to
1609 @code{void} and on pointers to functions. This is done by treating the
1610 size of a @code{void} or of a function as 1.
1612 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1613 and on function types, and returns 1.
1615 @opindex Wpointer-arith
1616 The option @option{-Wpointer-arith} requests a warning if these extensions
1620 @section Non-Constant Initializers
1621 @cindex initializers, non-constant
1622 @cindex non-constant initializers
1624 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1625 automatic variable are not required to be constant expressions in GNU C@.
1626 Here is an example of an initializer with run-time varying elements:
1629 foo (float f, float g)
1631 float beat_freqs[2] = @{ f-g, f+g @};
1636 @node Compound Literals
1637 @section Compound Literals
1638 @cindex constructor expressions
1639 @cindex initializations in expressions
1640 @cindex structures, constructor expression
1641 @cindex expressions, constructor
1642 @cindex compound literals
1643 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1645 ISO C99 supports compound literals. A compound literal looks like
1646 a cast containing an initializer. Its value is an object of the
1647 type specified in the cast, containing the elements specified in
1648 the initializer; it is an lvalue. As an extension, GCC supports
1649 compound literals in C89 mode and in C++.
1651 Usually, the specified type is a structure. Assume that
1652 @code{struct foo} and @code{structure} are declared as shown:
1655 struct foo @{int a; char b[2];@} structure;
1659 Here is an example of constructing a @code{struct foo} with a compound literal:
1662 structure = ((struct foo) @{x + y, 'a', 0@});
1666 This is equivalent to writing the following:
1670 struct foo temp = @{x + y, 'a', 0@};
1675 You can also construct an array. If all the elements of the compound literal
1676 are (made up of) simple constant expressions, suitable for use in
1677 initializers of objects of static storage duration, then the compound
1678 literal can be coerced to a pointer to its first element and used in
1679 such an initializer, as shown here:
1682 char **foo = (char *[]) @{ "x", "y", "z" @};
1685 Compound literals for scalar types and union types are is
1686 also allowed, but then the compound literal is equivalent
1689 As a GNU extension, GCC allows initialization of objects with static storage
1690 duration by compound literals (which is not possible in ISO C99, because
1691 the initializer is not a constant).
1692 It is handled as if the object was initialized only with the bracket
1693 enclosed list if compound literal's and object types match.
1694 The initializer list of the compound literal must be constant.
1695 If the object being initialized has array type of unknown size, the size is
1696 determined by compound literal size.
1699 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1700 static int y[] = (int []) @{1, 2, 3@};
1701 static int z[] = (int [3]) @{1@};
1705 The above lines are equivalent to the following:
1707 static struct foo x = @{1, 'a', 'b'@};
1708 static int y[] = @{1, 2, 3@};
1709 static int z[] = @{1, 0, 0@};
1712 @node Designated Inits
1713 @section Designated Initializers
1714 @cindex initializers with labeled elements
1715 @cindex labeled elements in initializers
1716 @cindex case labels in initializers
1717 @cindex designated initializers
1719 Standard C89 requires the elements of an initializer to appear in a fixed
1720 order, the same as the order of the elements in the array or structure
1723 In ISO C99 you can give the elements in any order, specifying the array
1724 indices or structure field names they apply to, and GNU C allows this as
1725 an extension in C89 mode as well. This extension is not
1726 implemented in GNU C++.
1728 To specify an array index, write
1729 @samp{[@var{index}] =} before the element value. For example,
1732 int a[6] = @{ [4] = 29, [2] = 15 @};
1739 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1743 The index values must be constant expressions, even if the array being
1744 initialized is automatic.
1746 An alternative syntax for this which has been obsolete since GCC 2.5 but
1747 GCC still accepts is to write @samp{[@var{index}]} before the element
1748 value, with no @samp{=}.
1750 To initialize a range of elements to the same value, write
1751 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1752 extension. For example,
1755 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1759 If the value in it has side-effects, the side-effects will happen only once,
1760 not for each initialized field by the range initializer.
1763 Note that the length of the array is the highest value specified
1766 In a structure initializer, specify the name of a field to initialize
1767 with @samp{.@var{fieldname} =} before the element value. For example,
1768 given the following structure,
1771 struct point @{ int x, y; @};
1775 the following initialization
1778 struct point p = @{ .y = yvalue, .x = xvalue @};
1785 struct point p = @{ xvalue, yvalue @};
1788 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1789 @samp{@var{fieldname}:}, as shown here:
1792 struct point p = @{ y: yvalue, x: xvalue @};
1796 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1797 @dfn{designator}. You can also use a designator (or the obsolete colon
1798 syntax) when initializing a union, to specify which element of the union
1799 should be used. For example,
1802 union foo @{ int i; double d; @};
1804 union foo f = @{ .d = 4 @};
1808 will convert 4 to a @code{double} to store it in the union using
1809 the second element. By contrast, casting 4 to type @code{union foo}
1810 would store it into the union as the integer @code{i}, since it is
1811 an integer. (@xref{Cast to Union}.)
1813 You can combine this technique of naming elements with ordinary C
1814 initialization of successive elements. Each initializer element that
1815 does not have a designator applies to the next consecutive element of the
1816 array or structure. For example,
1819 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1826 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1829 Labeling the elements of an array initializer is especially useful
1830 when the indices are characters or belong to an @code{enum} type.
1835 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1836 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1839 @cindex designator lists
1840 You can also write a series of @samp{.@var{fieldname}} and
1841 @samp{[@var{index}]} designators before an @samp{=} to specify a
1842 nested subobject to initialize; the list is taken relative to the
1843 subobject corresponding to the closest surrounding brace pair. For
1844 example, with the @samp{struct point} declaration above:
1847 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1851 If the same field is initialized multiple times, it will have value from
1852 the last initialization. If any such overridden initialization has
1853 side-effect, it is unspecified whether the side-effect happens or not.
1854 Currently, gcc will discard them and issue a warning.
1857 @section Case Ranges
1859 @cindex ranges in case statements
1861 You can specify a range of consecutive values in a single @code{case} label,
1865 case @var{low} ... @var{high}:
1869 This has the same effect as the proper number of individual @code{case}
1870 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1872 This feature is especially useful for ranges of ASCII character codes:
1878 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1879 it may be parsed wrong when you use it with integer values. For example,
1894 @section Cast to a Union Type
1895 @cindex cast to a union
1896 @cindex union, casting to a
1898 A cast to union type is similar to other casts, except that the type
1899 specified is a union type. You can specify the type either with
1900 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1901 a constructor though, not a cast, and hence does not yield an lvalue like
1902 normal casts. (@xref{Compound Literals}.)
1904 The types that may be cast to the union type are those of the members
1905 of the union. Thus, given the following union and variables:
1908 union foo @{ int i; double d; @};
1914 both @code{x} and @code{y} can be cast to type @code{union foo}.
1916 Using the cast as the right-hand side of an assignment to a variable of
1917 union type is equivalent to storing in a member of the union:
1922 u = (union foo) x @equiv{} u.i = x
1923 u = (union foo) y @equiv{} u.d = y
1926 You can also use the union cast as a function argument:
1929 void hack (union foo);
1931 hack ((union foo) x);
1934 @node Mixed Declarations
1935 @section Mixed Declarations and Code
1936 @cindex mixed declarations and code
1937 @cindex declarations, mixed with code
1938 @cindex code, mixed with declarations
1940 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1941 within compound statements. As an extension, GCC also allows this in
1942 C89 mode. For example, you could do:
1951 Each identifier is visible from where it is declared until the end of
1952 the enclosing block.
1954 @node Function Attributes
1955 @section Declaring Attributes of Functions
1956 @cindex function attributes
1957 @cindex declaring attributes of functions
1958 @cindex functions that never return
1959 @cindex functions that have no side effects
1960 @cindex functions in arbitrary sections
1961 @cindex functions that behave like malloc
1962 @cindex @code{volatile} applied to function
1963 @cindex @code{const} applied to function
1964 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1965 @cindex functions with non-null pointer arguments
1966 @cindex functions that are passed arguments in registers on the 386
1967 @cindex functions that pop the argument stack on the 386
1968 @cindex functions that do not pop the argument stack on the 386
1970 In GNU C, you declare certain things about functions called in your program
1971 which help the compiler optimize function calls and check your code more
1974 The keyword @code{__attribute__} allows you to specify special
1975 attributes when making a declaration. This keyword is followed by an
1976 attribute specification inside double parentheses. The following
1977 attributes are currently defined for functions on all targets:
1978 @code{noreturn}, @code{noinline}, @code{always_inline},
1979 @code{pure}, @code{const}, @code{nothrow},
1980 @code{format}, @code{format_arg}, @code{no_instrument_function},
1981 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1982 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1983 @code{alias}, and @code{nonnull}. Several other attributes are defined
1984 for functions on particular target systems. Other attributes, including
1985 @code{section} are supported for variables declarations
1986 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1988 You may also specify attributes with @samp{__} preceding and following
1989 each keyword. This allows you to use them in header files without
1990 being concerned about a possible macro of the same name. For example,
1991 you may use @code{__noreturn__} instead of @code{noreturn}.
1993 @xref{Attribute Syntax}, for details of the exact syntax for using
1997 @cindex @code{noreturn} function attribute
1999 A few standard library functions, such as @code{abort} and @code{exit},
2000 cannot return. GCC knows this automatically. Some programs define
2001 their own functions that never return. You can declare them
2002 @code{noreturn} to tell the compiler this fact. For example,
2006 void fatal () __attribute__ ((noreturn));
2009 fatal (/* @r{@dots{}} */)
2011 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2017 The @code{noreturn} keyword tells the compiler to assume that
2018 @code{fatal} cannot return. It can then optimize without regard to what
2019 would happen if @code{fatal} ever did return. This makes slightly
2020 better code. More importantly, it helps avoid spurious warnings of
2021 uninitialized variables.
2023 Do not assume that registers saved by the calling function are
2024 restored before calling the @code{noreturn} function.
2026 It does not make sense for a @code{noreturn} function to have a return
2027 type other than @code{void}.
2029 The attribute @code{noreturn} is not implemented in GCC versions
2030 earlier than 2.5. An alternative way to declare that a function does
2031 not return, which works in the current version and in some older
2032 versions, is as follows:
2035 typedef void voidfn ();
2037 volatile voidfn fatal;
2040 @cindex @code{noinline} function attribute
2042 This function attribute prevents a function from being considered for
2045 @cindex @code{always_inline} function attribute
2047 Generally, functions are not inlined unless optimization is specified.
2048 For functions declared inline, this attribute inlines the function even
2049 if no optimization level was specified.
2051 @cindex @code{pure} function attribute
2053 Many functions have no effects except the return value and their
2054 return value depends only on the parameters and/or global variables.
2055 Such a function can be subject
2056 to common subexpression elimination and loop optimization just as an
2057 arithmetic operator would be. These functions should be declared
2058 with the attribute @code{pure}. For example,
2061 int square (int) __attribute__ ((pure));
2065 says that the hypothetical function @code{square} is safe to call
2066 fewer times than the program says.
2068 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2069 Interesting non-pure functions are functions with infinite loops or those
2070 depending on volatile memory or other system resource, that may change between
2071 two consecutive calls (such as @code{feof} in a multithreading environment).
2073 The attribute @code{pure} is not implemented in GCC versions earlier
2075 @cindex @code{const} function attribute
2077 Many functions do not examine any values except their arguments, and
2078 have no effects except the return value. Basically this is just slightly
2079 more strict class than the @code{pure} attribute above, since function is not
2080 allowed to read global memory.
2082 @cindex pointer arguments
2083 Note that a function that has pointer arguments and examines the data
2084 pointed to must @emph{not} be declared @code{const}. Likewise, a
2085 function that calls a non-@code{const} function usually must not be
2086 @code{const}. It does not make sense for a @code{const} function to
2089 The attribute @code{const} is not implemented in GCC versions earlier
2090 than 2.5. An alternative way to declare that a function has no side
2091 effects, which works in the current version and in some older versions,
2095 typedef int intfn ();
2097 extern const intfn square;
2100 This approach does not work in GNU C++ from 2.6.0 on, since the language
2101 specifies that the @samp{const} must be attached to the return value.
2103 @cindex @code{nothrow} function attribute
2105 The @code{nothrow} attribute is used to inform the compiler that a
2106 function cannot throw an exception. For example, most functions in
2107 the standard C library can be guaranteed not to throw an exception
2108 with the notable exceptions of @code{qsort} and @code{bsearch} that
2109 take function pointer arguments. The @code{nothrow} attribute is not
2110 implemented in GCC versions earlier than 3.2.
2112 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2113 @cindex @code{format} function attribute
2115 The @code{format} attribute specifies that a function takes @code{printf},
2116 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2117 should be type-checked against a format string. For example, the
2122 my_printf (void *my_object, const char *my_format, ...)
2123 __attribute__ ((format (printf, 2, 3)));
2127 causes the compiler to check the arguments in calls to @code{my_printf}
2128 for consistency with the @code{printf} style format string argument
2131 The parameter @var{archetype} determines how the format string is
2132 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2133 or @code{strfmon}. (You can also use @code{__printf__},
2134 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2135 parameter @var{string-index} specifies which argument is the format
2136 string argument (starting from 1), while @var{first-to-check} is the
2137 number of the first argument to check against the format string. For
2138 functions where the arguments are not available to be checked (such as
2139 @code{vprintf}), specify the third parameter as zero. In this case the
2140 compiler only checks the format string for consistency. For
2141 @code{strftime} formats, the third parameter is required to be zero.
2142 Since non-static C++ methods have an implicit @code{this} argument, the
2143 arguments of such methods should be counted from two, not one, when
2144 giving values for @var{string-index} and @var{first-to-check}.
2146 In the example above, the format string (@code{my_format}) is the second
2147 argument of the function @code{my_print}, and the arguments to check
2148 start with the third argument, so the correct parameters for the format
2149 attribute are 2 and 3.
2151 @opindex ffreestanding
2152 The @code{format} attribute allows you to identify your own functions
2153 which take format strings as arguments, so that GCC can check the
2154 calls to these functions for errors. The compiler always (unless
2155 @option{-ffreestanding} is used) checks formats
2156 for the standard library functions @code{printf}, @code{fprintf},
2157 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2158 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2159 warnings are requested (using @option{-Wformat}), so there is no need to
2160 modify the header file @file{stdio.h}. In C99 mode, the functions
2161 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2162 @code{vsscanf} are also checked. Except in strictly conforming C
2163 standard modes, the X/Open function @code{strfmon} is also checked as
2164 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2165 @xref{C Dialect Options,,Options Controlling C Dialect}.
2167 @item format_arg (@var{string-index})
2168 @cindex @code{format_arg} function attribute
2169 @opindex Wformat-nonliteral
2170 The @code{format_arg} attribute specifies that a function takes a format
2171 string for a @code{printf}, @code{scanf}, @code{strftime} or
2172 @code{strfmon} style function and modifies it (for example, to translate
2173 it into another language), so the result can be passed to a
2174 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2175 function (with the remaining arguments to the format function the same
2176 as they would have been for the unmodified string). For example, the
2181 my_dgettext (char *my_domain, const char *my_format)
2182 __attribute__ ((format_arg (2)));
2186 causes the compiler to check the arguments in calls to a @code{printf},
2187 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2188 format string argument is a call to the @code{my_dgettext} function, for
2189 consistency with the format string argument @code{my_format}. If the
2190 @code{format_arg} attribute had not been specified, all the compiler
2191 could tell in such calls to format functions would be that the format
2192 string argument is not constant; this would generate a warning when
2193 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2194 without the attribute.
2196 The parameter @var{string-index} specifies which argument is the format
2197 string argument (starting from one). Since non-static C++ methods have
2198 an implicit @code{this} argument, the arguments of such methods should
2199 be counted from two.
2201 The @code{format-arg} attribute allows you to identify your own
2202 functions which modify format strings, so that GCC can check the
2203 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2204 type function whose operands are a call to one of your own function.
2205 The compiler always treats @code{gettext}, @code{dgettext}, and
2206 @code{dcgettext} in this manner except when strict ISO C support is
2207 requested by @option{-ansi} or an appropriate @option{-std} option, or
2208 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2209 Controlling C Dialect}.
2211 @item nonnull (@var{arg-index}, @dots{})
2212 @cindex @code{nonnull} function attribute
2213 The @code{nonnull} attribute specifies that some function parameters should
2214 be non-null pointers. For instance, the declaration:
2218 my_memcpy (void *dest, const void *src, size_t len)
2219 __attribute__((nonnull (1, 2)));
2223 causes the compiler to check that, in calls to @code{my_memcpy},
2224 arguments @var{dest} and @var{src} are non-null. If the compiler
2225 determines that a null pointer is passed in an argument slot marked
2226 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2227 is issued. The compiler may also choose to make optimizations based
2228 on the knowledge that certain function arguments will not be null.
2230 If no argument index list is given to the @code{nonnull} attribute,
2231 all pointer arguments are marked as non-null. To illustrate, the
2232 following declaration is equivalent to the previous example:
2236 my_memcpy (void *dest, const void *src, size_t len)
2237 __attribute__((nonnull));
2240 @item no_instrument_function
2241 @cindex @code{no_instrument_function} function attribute
2242 @opindex finstrument-functions
2243 If @option{-finstrument-functions} is given, profiling function calls will
2244 be generated at entry and exit of most user-compiled functions.
2245 Functions with this attribute will not be so instrumented.
2247 @item section ("@var{section-name}")
2248 @cindex @code{section} function attribute
2249 Normally, the compiler places the code it generates in the @code{text} section.
2250 Sometimes, however, you need additional sections, or you need certain
2251 particular functions to appear in special sections. The @code{section}
2252 attribute specifies that a function lives in a particular section.
2253 For example, the declaration:
2256 extern void foobar (void) __attribute__ ((section ("bar")));
2260 puts the function @code{foobar} in the @code{bar} section.
2262 Some file formats do not support arbitrary sections so the @code{section}
2263 attribute is not available on all platforms.
2264 If you need to map the entire contents of a module to a particular
2265 section, consider using the facilities of the linker instead.
2269 @cindex @code{constructor} function attribute
2270 @cindex @code{destructor} function attribute
2271 The @code{constructor} attribute causes the function to be called
2272 automatically before execution enters @code{main ()}. Similarly, the
2273 @code{destructor} attribute causes the function to be called
2274 automatically after @code{main ()} has completed or @code{exit ()} has
2275 been called. Functions with these attributes are useful for
2276 initializing data that will be used implicitly during the execution of
2279 These attributes are not currently implemented for Objective-C@.
2281 @cindex @code{unused} attribute.
2283 This attribute, attached to a function, means that the function is meant
2284 to be possibly unused. GCC will not produce a warning for this
2287 @cindex @code{used} attribute.
2289 This attribute, attached to a function, means that code must be emitted
2290 for the function even if it appears that the function is not referenced.
2291 This is useful, for example, when the function is referenced only in
2294 @cindex @code{deprecated} attribute.
2296 The @code{deprecated} attribute results in a warning if the function
2297 is used anywhere in the source file. This is useful when identifying
2298 functions that are expected to be removed in a future version of a
2299 program. The warning also includes the location of the declaration
2300 of the deprecated function, to enable users to easily find further
2301 information about why the function is deprecated, or what they should
2302 do instead. Note that the warnings only occurs for uses:
2305 int old_fn () __attribute__ ((deprecated));
2307 int (*fn_ptr)() = old_fn;
2310 results in a warning on line 3 but not line 2.
2312 The @code{deprecated} attribute can also be used for variables and
2313 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2316 @cindex @code{weak} attribute
2317 The @code{weak} attribute causes the declaration to be emitted as a weak
2318 symbol rather than a global. This is primarily useful in defining
2319 library functions which can be overridden in user code, though it can
2320 also be used with non-function declarations. Weak symbols are supported
2321 for ELF targets, and also for a.out targets when using the GNU assembler
2325 @cindex @code{malloc} attribute
2326 The @code{malloc} attribute is used to tell the compiler that a function
2327 may be treated as if it were the malloc function. The compiler assumes
2328 that calls to malloc result in pointers that cannot alias anything.
2329 This will often improve optimization.
2331 @item alias ("@var{target}")
2332 @cindex @code{alias} attribute
2333 The @code{alias} attribute causes the declaration to be emitted as an
2334 alias for another symbol, which must be specified. For instance,
2337 void __f () @{ /* @r{Do something.} */; @}
2338 void f () __attribute__ ((weak, alias ("__f")));
2341 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2342 mangled name for the target must be used.
2344 Not all target machines support this attribute.
2346 @item visibility ("@var{visibility_type}")
2347 @cindex @code{visibility} attribute
2348 The @code{visibility} attribute on ELF targets causes the declaration
2349 to be emitted with default, hidden, protected or internal visibility.
2352 void __attribute__ ((visibility ("protected")))
2353 f () @{ /* @r{Do something.} */; @}
2354 int i __attribute__ ((visibility ("hidden")));
2357 See the ELF gABI for complete details, but the short story is:
2361 Default visibility is the normal case for ELF. This value is
2362 available for the visibility attribute to override other options
2363 that may change the assumed visibility of symbols.
2366 Hidden visibility indicates that the symbol will not be placed into
2367 the dynamic symbol table, so no other @dfn{module} (executable or
2368 shared library) can reference it directly.
2371 Protected visibility indicates that the symbol will be placed in the
2372 dynamic symbol table, but that references within the defining module
2373 will bind to the local symbol. That is, the symbol cannot be overridden
2377 Internal visibility is like hidden visibility, but with additional
2378 processor specific semantics. Unless otherwise specified by the psABI,
2379 gcc defines internal visibility to mean that the function is @emph{never}
2380 called from another module. Note that hidden symbols, while they cannot
2381 be referenced directly by other modules, can be referenced indirectly via
2382 function pointers. By indicating that a symbol cannot be called from
2383 outside the module, gcc may for instance omit the load of a PIC register
2384 since it is known that the calling function loaded the correct value.
2387 Not all ELF targets support this attribute.
2389 @item regparm (@var{number})
2390 @cindex @code{regparm} attribute
2391 @cindex functions that are passed arguments in registers on the 386
2392 On the Intel 386, the @code{regparm} attribute causes the compiler to
2393 pass up to @var{number} integer arguments in registers EAX,
2394 EDX, and ECX instead of on the stack. Functions that take a
2395 variable number of arguments will continue to be passed all of their
2396 arguments on the stack.
2398 Beware that on some ELF systems this attribute is unsuitable for
2399 global functions in shared libraries with lazy binding (which is the
2400 default). Lazy binding will send the first call via resolving code in
2401 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2402 per the standard calling conventions. Solaris 8 is affected by this.
2403 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2404 safe since the loaders there save all registers. (Lazy binding can be
2405 disabled with the linker or the loader if desired, to avoid the
2409 @cindex functions that pop the argument stack on the 386
2410 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2411 assume that the called function will pop off the stack space used to
2412 pass arguments, unless it takes a variable number of arguments.
2415 @cindex functions that pop the argument stack on the 386
2416 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2417 pass the first two arguments in the registers ECX and EDX. Subsequent
2418 arguments are passed on the stack. The called function will pop the
2419 arguments off the stack. If the number of arguments is variable all
2420 arguments are pushed on the stack.
2423 @cindex functions that do pop the argument stack on the 386
2425 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2426 assume that the calling function will pop off the stack space used to
2427 pass arguments. This is
2428 useful to override the effects of the @option{-mrtd} switch.
2430 @item longcall/shortcall
2431 @cindex functions called via pointer on the RS/6000 and PowerPC
2432 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2433 compiler to always call this function via a pointer, just as it would if
2434 the @option{-mlongcall} option had been specified. The @code{shortcall}
2435 attribute causes the compiler not to do this. These attributes override
2436 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2439 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2440 calls are necessary.
2442 @item long_call/short_call
2443 @cindex indirect calls on ARM
2444 This attribute specifies how a particular function is called on
2445 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2446 command line switch and @code{#pragma long_calls} settings. The
2447 @code{long_call} attribute causes the compiler to always call the
2448 function by first loading its address into a register and then using the
2449 contents of that register. The @code{short_call} attribute always places
2450 the offset to the function from the call site into the @samp{BL}
2451 instruction directly.
2453 @item function_vector
2454 @cindex calling functions through the function vector on the H8/300 processors
2455 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2456 function should be called through the function vector. Calling a
2457 function through the function vector will reduce code size, however;
2458 the function vector has a limited size (maximum 128 entries on the H8/300
2459 and 64 entries on the H8/300H) and shares space with the interrupt vector.
2461 You must use GAS and GLD from GNU binutils version 2.7 or later for
2462 this attribute to work correctly.
2465 @cindex interrupt handler functions
2466 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2467 that the specified function is an interrupt handler. The compiler will
2468 generate function entry and exit sequences suitable for use in an
2469 interrupt handler when this attribute is present.
2471 Note, interrupt handlers for the H8/300, H8/300H and SH processors can
2472 be specified via the @code{interrupt_handler} attribute.
2474 Note, on the AVR, interrupts will be enabled inside the function.
2476 Note, for the ARM, you can specify the kind of interrupt to be handled by
2477 adding an optional parameter to the interrupt attribute like this:
2480 void f () __attribute__ ((interrupt ("IRQ")));
2483 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2485 @item interrupt_handler
2486 @cindex interrupt handler functions on the H8/300 and SH processors
2487 Use this attribute on the H8/300, H8/300H and SH to indicate that the
2488 specified function is an interrupt handler. The compiler will generate
2489 function entry and exit sequences suitable for use in an interrupt
2490 handler when this attribute is present.
2493 Use this attribute on the SH to indicate an @code{interrupt_handler}
2494 function should switch to an alternate stack. It expects a string
2495 argument that names a global variable holding the address of the
2500 void f () __attribute__ ((interrupt_handler,
2501 sp_switch ("alt_stack")));
2505 Use this attribute on the SH for an @code{interrupt_handle} to return using
2506 @code{trapa} instead of @code{rte}. This attribute expects an integer
2507 argument specifying the trap number to be used.
2510 @cindex eight bit data on the H8/300 and H8/300H
2511 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2512 variable should be placed into the eight bit data section.
2513 The compiler will generate more efficient code for certain operations
2514 on data in the eight bit data area. Note the eight bit data area is limited to
2517 You must use GAS and GLD from GNU binutils version 2.7 or later for
2518 this attribute to work correctly.
2521 @cindex tiny data section on the H8/300H
2522 Use this attribute on the H8/300H to indicate that the specified
2523 variable should be placed into the tiny data section.
2524 The compiler will generate more efficient code for loads and stores
2525 on data in the tiny data section. Note the tiny data area is limited to
2526 slightly under 32kbytes of data.
2529 @cindex signal handler functions on the AVR processors
2530 Use this attribute on the AVR to indicate that the specified
2531 function is a signal handler. The compiler will generate function
2532 entry and exit sequences suitable for use in a signal handler when this
2533 attribute is present. Interrupts will be disabled inside the function.
2536 @cindex function without a prologue/epilogue code
2537 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2538 specified function does not need prologue/epilogue sequences generated by
2539 the compiler. It is up to the programmer to provide these sequences.
2541 @item model (@var{model-name})
2542 @cindex function addressability on the M32R/D
2543 @cindex variable addressability on the IA-64
2545 On the M32R/D, use this attribute to set the addressability of an
2546 object, and of the code generated for a function. The identifier
2547 @var{model-name} is one of @code{small}, @code{medium}, or
2548 @code{large}, representing each of the code models.
2550 Small model objects live in the lower 16MB of memory (so that their
2551 addresses can be loaded with the @code{ld24} instruction), and are
2552 callable with the @code{bl} instruction.
2554 Medium model objects may live anywhere in the 32-bit address space (the
2555 compiler will generate @code{seth/add3} instructions to load their addresses),
2556 and are callable with the @code{bl} instruction.
2558 Large model objects may live anywhere in the 32-bit address space (the
2559 compiler will generate @code{seth/add3} instructions to load their addresses),
2560 and may not be reachable with the @code{bl} instruction (the compiler will
2561 generate the much slower @code{seth/add3/jl} instruction sequence).
2563 On IA-64, use this attribute to set the addressability of an object.
2564 At present, the only supported identifier for @var{model-name} is
2565 @code{small}, indicating addressability via ``small'' (22-bit)
2566 addresses (so that their addresses can be loaded with the @code{addl}
2567 instruction). Caveat: such addressing is by definition not position
2568 independent and hence this attribute must not be used for objects
2569 defined by shared libraries.
2572 @cindex functions which handle memory bank switching
2573 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2574 use a calling convention that takes care of switching memory banks when
2575 entering and leaving a function. This calling convention is also the
2576 default when using the @option{-mlong-calls} option.
2578 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2579 to call and return from a function.
2581 On 68HC11 the compiler will generate a sequence of instructions
2582 to invoke a board-specific routine to switch the memory bank and call the
2583 real function. The board-specific routine simulates a @code{call}.
2584 At the end of a function, it will jump to a board-specific routine
2585 instead of using @code{rts}. The board-specific return routine simulates
2589 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2590 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2591 use the normal calling convention based on @code{jsr} and @code{rts}.
2592 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2596 @cindex @code{__declspec(dllimport)}
2597 On Windows targets, the @code{dllimport} attribute causes the compiler
2598 to reference a function or variable via a global pointer to a pointer
2599 that is set up by the Windows dll library. The pointer name is formed by
2600 combining @code{_imp__} and the function or variable name. The attribute
2601 implies @code{extern} storage.
2603 Currently, the attribute is ignored for inlined functions. If the
2604 attribute is applied to a symbol @emph{definition}, an error is reported.
2605 If a symbol previously declared @code{dllimport} is later defined, the
2606 attribute is ignored in subsequent references, and a warning is emitted.
2607 The attribute is also overriden by a subsequent declaration as
2610 When applied to C++ classes, the attribute marks non-inlined
2611 member functions and static data members as imports. However, the
2612 attribute is ignored for virtual methods to allow creation of vtables
2615 On cygwin, mingw and arm-pe targets, @code{__declspec(dllimport)} is
2616 recognized as a synonym for @code{__attribute__ ((dllimport))} for
2617 compatibility with other Windows compilers.
2619 The use of the @code{dllimport} attribute on functions is not necessary,
2620 but provides a small performance benefit by eliminating a thunk in the
2621 dll. The use of the @code{dllimport} attribute on imported variables was
2622 required on older versions of GNU ld, but can now be avoided by passing
2623 the @option{--enable-auto-import} switch to ld. As with functions, using
2624 the attribute for a variable eliminates a thunk in the dll.
2626 One drawback to using this attribute is that a pointer to a function or
2627 variable marked as dllimport cannot be used as a constant address. The
2628 attribute can be disabled for functions by setting the
2629 @option{-mnop-fun-dllimport} flag.
2632 @cindex @code{__declspec(dllexport)}
2633 On Windows targets the @code{dllexport} attribute causes the compiler to
2634 provide a global pointer to a pointer in a dll, so that it can be
2635 referenced with the @code{dllimport} attribute. The pointer name is
2636 formed by combining @code{_imp__} and the function or variable name.
2638 Currently, the @code{dllexport}attribute is ignored for inlined
2639 functions, but export can be forced by using the
2640 @option{-fkeep-inline-functions} flag. The attribute is also ignored for
2643 When applied to C++ classes. the attribute marks defined non-inlined
2644 member functions and static data members as exports. Static consts
2645 initialized in-class are not marked unless they are also defined
2648 On cygwin, mingw and arm-pe targets, @code{__declspec(dllexport)} is
2649 recognized as a synonym for @code{__attribute__ ((dllexport))} for
2650 compatibility with other Windows compilers.
2652 Alternative methods for including the symbol in the dll's export table
2653 are to use a .def file with an @code{EXPORTS} section or, with GNU ld,
2654 using the @option{--export-all} linker flag.
2658 You can specify multiple attributes in a declaration by separating them
2659 by commas within the double parentheses or by immediately following an
2660 attribute declaration with another attribute declaration.
2662 @cindex @code{#pragma}, reason for not using
2663 @cindex pragma, reason for not using
2664 Some people object to the @code{__attribute__} feature, suggesting that
2665 ISO C's @code{#pragma} should be used instead. At the time
2666 @code{__attribute__} was designed, there were two reasons for not doing
2671 It is impossible to generate @code{#pragma} commands from a macro.
2674 There is no telling what the same @code{#pragma} might mean in another
2678 These two reasons applied to almost any application that might have been
2679 proposed for @code{#pragma}. It was basically a mistake to use
2680 @code{#pragma} for @emph{anything}.
2682 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2683 to be generated from macros. In addition, a @code{#pragma GCC}
2684 namespace is now in use for GCC-specific pragmas. However, it has been
2685 found convenient to use @code{__attribute__} to achieve a natural
2686 attachment of attributes to their corresponding declarations, whereas
2687 @code{#pragma GCC} is of use for constructs that do not naturally form
2688 part of the grammar. @xref{Other Directives,,Miscellaneous
2689 Preprocessing Directives, cpp, The C Preprocessor}.
2691 @node Attribute Syntax
2692 @section Attribute Syntax
2693 @cindex attribute syntax
2695 This section describes the syntax with which @code{__attribute__} may be
2696 used, and the constructs to which attribute specifiers bind, for the C
2697 language. Some details may vary for C++ and Objective-C@. Because of
2698 infelicities in the grammar for attributes, some forms described here
2699 may not be successfully parsed in all cases.
2701 There are some problems with the semantics of attributes in C++. For
2702 example, there are no manglings for attributes, although they may affect
2703 code generation, so problems may arise when attributed types are used in
2704 conjunction with templates or overloading. Similarly, @code{typeid}
2705 does not distinguish between types with different attributes. Support
2706 for attributes in C++ may be restricted in future to attributes on
2707 declarations only, but not on nested declarators.
2709 @xref{Function Attributes}, for details of the semantics of attributes
2710 applying to functions. @xref{Variable Attributes}, for details of the
2711 semantics of attributes applying to variables. @xref{Type Attributes},
2712 for details of the semantics of attributes applying to structure, union
2713 and enumerated types.
2715 An @dfn{attribute specifier} is of the form
2716 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2717 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2718 each attribute is one of the following:
2722 Empty. Empty attributes are ignored.
2725 A word (which may be an identifier such as @code{unused}, or a reserved
2726 word such as @code{const}).
2729 A word, followed by, in parentheses, parameters for the attribute.
2730 These parameters take one of the following forms:
2734 An identifier. For example, @code{mode} attributes use this form.
2737 An identifier followed by a comma and a non-empty comma-separated list
2738 of expressions. For example, @code{format} attributes use this form.
2741 A possibly empty comma-separated list of expressions. For example,
2742 @code{format_arg} attributes use this form with the list being a single
2743 integer constant expression, and @code{alias} attributes use this form
2744 with the list being a single string constant.
2748 An @dfn{attribute specifier list} is a sequence of one or more attribute
2749 specifiers, not separated by any other tokens.
2751 In GNU C, an attribute specifier list may appear after the colon following a
2752 label, other than a @code{case} or @code{default} label. The only
2753 attribute it makes sense to use after a label is @code{unused}. This
2754 feature is intended for code generated by programs which contains labels
2755 that may be unused but which is compiled with @option{-Wall}. It would
2756 not normally be appropriate to use in it human-written code, though it
2757 could be useful in cases where the code that jumps to the label is
2758 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2759 such placement of attribute lists, as it is permissible for a
2760 declaration, which could begin with an attribute list, to be labelled in
2761 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2762 does not arise there.
2764 An attribute specifier list may appear as part of a @code{struct},
2765 @code{union} or @code{enum} specifier. It may go either immediately
2766 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2767 the closing brace. It is ignored if the content of the structure, union
2768 or enumerated type is not defined in the specifier in which the
2769 attribute specifier list is used---that is, in usages such as
2770 @code{struct __attribute__((foo)) bar} with no following opening brace.
2771 Where attribute specifiers follow the closing brace, they are considered
2772 to relate to the structure, union or enumerated type defined, not to any
2773 enclosing declaration the type specifier appears in, and the type
2774 defined is not complete until after the attribute specifiers.
2775 @c Otherwise, there would be the following problems: a shift/reduce
2776 @c conflict between attributes binding the struct/union/enum and
2777 @c binding to the list of specifiers/qualifiers; and "aligned"
2778 @c attributes could use sizeof for the structure, but the size could be
2779 @c changed later by "packed" attributes.
2781 Otherwise, an attribute specifier appears as part of a declaration,
2782 counting declarations of unnamed parameters and type names, and relates
2783 to that declaration (which may be nested in another declaration, for
2784 example in the case of a parameter declaration), or to a particular declarator
2785 within a declaration. Where an
2786 attribute specifier is applied to a parameter declared as a function or
2787 an array, it should apply to the function or array rather than the
2788 pointer to which the parameter is implicitly converted, but this is not
2789 yet correctly implemented.
2791 Any list of specifiers and qualifiers at the start of a declaration may
2792 contain attribute specifiers, whether or not such a list may in that
2793 context contain storage class specifiers. (Some attributes, however,
2794 are essentially in the nature of storage class specifiers, and only make
2795 sense where storage class specifiers may be used; for example,
2796 @code{section}.) There is one necessary limitation to this syntax: the
2797 first old-style parameter declaration in a function definition cannot
2798 begin with an attribute specifier, because such an attribute applies to
2799 the function instead by syntax described below (which, however, is not
2800 yet implemented in this case). In some other cases, attribute
2801 specifiers are permitted by this grammar but not yet supported by the
2802 compiler. All attribute specifiers in this place relate to the
2803 declaration as a whole. In the obsolescent usage where a type of
2804 @code{int} is implied by the absence of type specifiers, such a list of
2805 specifiers and qualifiers may be an attribute specifier list with no
2806 other specifiers or qualifiers.
2808 An attribute specifier list may appear immediately before a declarator
2809 (other than the first) in a comma-separated list of declarators in a
2810 declaration of more than one identifier using a single list of
2811 specifiers and qualifiers. Such attribute specifiers apply
2812 only to the identifier before whose declarator they appear. For
2816 __attribute__((noreturn)) void d0 (void),
2817 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2822 the @code{noreturn} attribute applies to all the functions
2823 declared; the @code{format} attribute only applies to @code{d1}.
2825 An attribute specifier list may appear immediately before the comma,
2826 @code{=} or semicolon terminating the declaration of an identifier other
2827 than a function definition. At present, such attribute specifiers apply
2828 to the declared object or function, but in future they may attach to the
2829 outermost adjacent declarator. In simple cases there is no difference,
2830 but, for example, in
2833 void (****f)(void) __attribute__((noreturn));
2837 at present the @code{noreturn} attribute applies to @code{f}, which
2838 causes a warning since @code{f} is not a function, but in future it may
2839 apply to the function @code{****f}. The precise semantics of what
2840 attributes in such cases will apply to are not yet specified. Where an
2841 assembler name for an object or function is specified (@pxref{Asm
2842 Labels}), at present the attribute must follow the @code{asm}
2843 specification; in future, attributes before the @code{asm} specification
2844 may apply to the adjacent declarator, and those after it to the declared
2847 An attribute specifier list may, in future, be permitted to appear after
2848 the declarator in a function definition (before any old-style parameter
2849 declarations or the function body).
2851 Attribute specifiers may be mixed with type qualifiers appearing inside
2852 the @code{[]} of a parameter array declarator, in the C99 construct by
2853 which such qualifiers are applied to the pointer to which the array is
2854 implicitly converted. Such attribute specifiers apply to the pointer,
2855 not to the array, but at present this is not implemented and they are
2858 An attribute specifier list may appear at the start of a nested
2859 declarator. At present, there are some limitations in this usage: the
2860 attributes correctly apply to the declarator, but for most individual
2861 attributes the semantics this implies are not implemented.
2862 When attribute specifiers follow the @code{*} of a pointer
2863 declarator, they may be mixed with any type qualifiers present.
2864 The following describes the formal semantics of this syntax. It will make the
2865 most sense if you are familiar with the formal specification of
2866 declarators in the ISO C standard.
2868 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2869 D1}, where @code{T} contains declaration specifiers that specify a type
2870 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2871 contains an identifier @var{ident}. The type specified for @var{ident}
2872 for derived declarators whose type does not include an attribute
2873 specifier is as in the ISO C standard.
2875 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2876 and the declaration @code{T D} specifies the type
2877 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2878 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2879 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2881 If @code{D1} has the form @code{*
2882 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2883 declaration @code{T D} specifies the type
2884 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2885 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2886 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2892 void (__attribute__((noreturn)) ****f) (void);
2896 specifies the type ``pointer to pointer to pointer to pointer to
2897 non-returning function returning @code{void}''. As another example,
2900 char *__attribute__((aligned(8))) *f;
2904 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2905 Note again that this does not work with most attributes; for example,
2906 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2907 is not yet supported.
2909 For compatibility with existing code written for compiler versions that
2910 did not implement attributes on nested declarators, some laxity is
2911 allowed in the placing of attributes. If an attribute that only applies
2912 to types is applied to a declaration, it will be treated as applying to
2913 the type of that declaration. If an attribute that only applies to
2914 declarations is applied to the type of a declaration, it will be treated
2915 as applying to that declaration; and, for compatibility with code
2916 placing the attributes immediately before the identifier declared, such
2917 an attribute applied to a function return type will be treated as
2918 applying to the function type, and such an attribute applied to an array
2919 element type will be treated as applying to the array type. If an
2920 attribute that only applies to function types is applied to a
2921 pointer-to-function type, it will be treated as applying to the pointer
2922 target type; if such an attribute is applied to a function return type
2923 that is not a pointer-to-function type, it will be treated as applying
2924 to the function type.
2926 @node Function Prototypes
2927 @section Prototypes and Old-Style Function Definitions
2928 @cindex function prototype declarations
2929 @cindex old-style function definitions
2930 @cindex promotion of formal parameters
2932 GNU C extends ISO C to allow a function prototype to override a later
2933 old-style non-prototype definition. Consider the following example:
2936 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2943 /* @r{Prototype function declaration.} */
2944 int isroot P((uid_t));
2946 /* @r{Old-style function definition.} */
2948 isroot (x) /* ??? lossage here ??? */
2955 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2956 not allow this example, because subword arguments in old-style
2957 non-prototype definitions are promoted. Therefore in this example the
2958 function definition's argument is really an @code{int}, which does not
2959 match the prototype argument type of @code{short}.
2961 This restriction of ISO C makes it hard to write code that is portable
2962 to traditional C compilers, because the programmer does not know
2963 whether the @code{uid_t} type is @code{short}, @code{int}, or
2964 @code{long}. Therefore, in cases like these GNU C allows a prototype
2965 to override a later old-style definition. More precisely, in GNU C, a
2966 function prototype argument type overrides the argument type specified
2967 by a later old-style definition if the former type is the same as the
2968 latter type before promotion. Thus in GNU C the above example is
2969 equivalent to the following:
2982 GNU C++ does not support old-style function definitions, so this
2983 extension is irrelevant.
2986 @section C++ Style Comments
2988 @cindex C++ comments
2989 @cindex comments, C++ style
2991 In GNU C, you may use C++ style comments, which start with @samp{//} and
2992 continue until the end of the line. Many other C implementations allow
2993 such comments, and they are included in the 1999 C standard. However,
2994 C++ style comments are not recognized if you specify an @option{-std}
2995 option specifying a version of ISO C before C99, or @option{-ansi}
2996 (equivalent to @option{-std=c89}).
2999 @section Dollar Signs in Identifier Names
3001 @cindex dollar signs in identifier names
3002 @cindex identifier names, dollar signs in
3004 In GNU C, you may normally use dollar signs in identifier names.
3005 This is because many traditional C implementations allow such identifiers.
3006 However, dollar signs in identifiers are not supported on a few target
3007 machines, typically because the target assembler does not allow them.
3009 @node Character Escapes
3010 @section The Character @key{ESC} in Constants
3012 You can use the sequence @samp{\e} in a string or character constant to
3013 stand for the ASCII character @key{ESC}.
3016 @section Inquiring on Alignment of Types or Variables
3018 @cindex type alignment
3019 @cindex variable alignment
3021 The keyword @code{__alignof__} allows you to inquire about how an object
3022 is aligned, or the minimum alignment usually required by a type. Its
3023 syntax is just like @code{sizeof}.
3025 For example, if the target machine requires a @code{double} value to be
3026 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3027 This is true on many RISC machines. On more traditional machine
3028 designs, @code{__alignof__ (double)} is 4 or even 2.
3030 Some machines never actually require alignment; they allow reference to any
3031 data type even at an odd address. For these machines, @code{__alignof__}
3032 reports the @emph{recommended} alignment of a type.
3034 If the operand of @code{__alignof__} is an lvalue rather than a type,
3035 its value is the required alignment for its type, taking into account
3036 any minimum alignment specified with GCC's @code{__attribute__}
3037 extension (@pxref{Variable Attributes}). For example, after this
3041 struct foo @{ int x; char y; @} foo1;
3045 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3046 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3048 It is an error to ask for the alignment of an incomplete type.
3050 @node Variable Attributes
3051 @section Specifying Attributes of Variables
3052 @cindex attribute of variables
3053 @cindex variable attributes
3055 The keyword @code{__attribute__} allows you to specify special
3056 attributes of variables or structure fields. This keyword is followed
3057 by an attribute specification inside double parentheses. Some
3058 attributes are currently defined generically for variables.
3059 Other attributes are defined for variables on particular target
3060 systems. Other attributes are available for functions
3061 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3062 Other front ends might define more attributes
3063 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3065 You may also specify attributes with @samp{__} preceding and following
3066 each keyword. This allows you to use them in header files without
3067 being concerned about a possible macro of the same name. For example,
3068 you may use @code{__aligned__} instead of @code{aligned}.
3070 @xref{Attribute Syntax}, for details of the exact syntax for using
3074 @cindex @code{aligned} attribute
3075 @item aligned (@var{alignment})
3076 This attribute specifies a minimum alignment for the variable or
3077 structure field, measured in bytes. For example, the declaration:
3080 int x __attribute__ ((aligned (16))) = 0;
3084 causes the compiler to allocate the global variable @code{x} on a
3085 16-byte boundary. On a 68040, this could be used in conjunction with
3086 an @code{asm} expression to access the @code{move16} instruction which
3087 requires 16-byte aligned operands.
3089 You can also specify the alignment of structure fields. For example, to
3090 create a double-word aligned @code{int} pair, you could write:
3093 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3097 This is an alternative to creating a union with a @code{double} member
3098 that forces the union to be double-word aligned.
3100 As in the preceding examples, you can explicitly specify the alignment
3101 (in bytes) that you wish the compiler to use for a given variable or
3102 structure field. Alternatively, you can leave out the alignment factor
3103 and just ask the compiler to align a variable or field to the maximum
3104 useful alignment for the target machine you are compiling for. For
3105 example, you could write:
3108 short array[3] __attribute__ ((aligned));
3111 Whenever you leave out the alignment factor in an @code{aligned} attribute
3112 specification, the compiler automatically sets the alignment for the declared
3113 variable or field to the largest alignment which is ever used for any data
3114 type on the target machine you are compiling for. Doing this can often make
3115 copy operations more efficient, because the compiler can use whatever
3116 instructions copy the biggest chunks of memory when performing copies to
3117 or from the variables or fields that you have aligned this way.
3119 The @code{aligned} attribute can only increase the alignment; but you
3120 can decrease it by specifying @code{packed} as well. See below.
3122 Note that the effectiveness of @code{aligned} attributes may be limited
3123 by inherent limitations in your linker. On many systems, the linker is
3124 only able to arrange for variables to be aligned up to a certain maximum
3125 alignment. (For some linkers, the maximum supported alignment may
3126 be very very small.) If your linker is only able to align variables
3127 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3128 in an @code{__attribute__} will still only provide you with 8 byte
3129 alignment. See your linker documentation for further information.
3131 @item cleanup (@var{cleanup_function})
3132 @cindex @code{cleanup} attribute
3133 The @code{cleanup} attribute runs a function when the variable goes
3134 out of scope. This attribute can only be applied to auto function
3135 scope variables; it may not be applied to parameters or variables
3136 with static storage duration. The function must take one parameter,
3137 a pointer to a type compatible with the variable. The return value
3138 of the function (if any) is ignored.
3140 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3141 will be run during the stack unwinding that happens during the
3142 processing of the exception. Note that the @code{cleanup} attribute
3143 does not allow the exception to be caught, only to perform an action.
3144 It is undefined what happens if @var{cleanup_function} does not
3149 @cindex @code{common} attribute
3150 @cindex @code{nocommon} attribute
3153 The @code{common} attribute requests GCC to place a variable in
3154 ``common'' storage. The @code{nocommon} attribute requests the
3155 opposite -- to allocate space for it directly.
3157 These attributes override the default chosen by the
3158 @option{-fno-common} and @option{-fcommon} flags respectively.
3161 @cindex @code{deprecated} attribute
3162 The @code{deprecated} attribute results in a warning if the variable
3163 is used anywhere in the source file. This is useful when identifying
3164 variables that are expected to be removed in a future version of a
3165 program. The warning also includes the location of the declaration
3166 of the deprecated variable, to enable users to easily find further
3167 information about why the variable is deprecated, or what they should
3168 do instead. Note that the warning only occurs for uses:
3171 extern int old_var __attribute__ ((deprecated));
3173 int new_fn () @{ return old_var; @}
3176 results in a warning on line 3 but not line 2.
3178 The @code{deprecated} attribute can also be used for functions and
3179 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3181 @item mode (@var{mode})
3182 @cindex @code{mode} attribute
3183 This attribute specifies the data type for the declaration---whichever
3184 type corresponds to the mode @var{mode}. This in effect lets you
3185 request an integer or floating point type according to its width.
3187 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3188 indicate the mode corresponding to a one-byte integer, @samp{word} or
3189 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3190 or @samp{__pointer__} for the mode used to represent pointers.
3193 @cindex @code{packed} attribute
3194 The @code{packed} attribute specifies that a variable or structure field
3195 should have the smallest possible alignment---one byte for a variable,
3196 and one bit for a field, unless you specify a larger value with the
3197 @code{aligned} attribute.
3199 Here is a structure in which the field @code{x} is packed, so that it
3200 immediately follows @code{a}:
3206 int x[2] __attribute__ ((packed));
3210 @item section ("@var{section-name}")
3211 @cindex @code{section} variable attribute
3212 Normally, the compiler places the objects it generates in sections like
3213 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3214 or you need certain particular variables to appear in special sections,
3215 for example to map to special hardware. The @code{section}
3216 attribute specifies that a variable (or function) lives in a particular
3217 section. For example, this small program uses several specific section names:
3220 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3221 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3222 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3223 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3227 /* Initialize stack pointer */
3228 init_sp (stack + sizeof (stack));
3230 /* Initialize initialized data */
3231 memcpy (&init_data, &data, &edata - &data);
3233 /* Turn on the serial ports */
3240 Use the @code{section} attribute with an @emph{initialized} definition
3241 of a @emph{global} variable, as shown in the example. GCC issues
3242 a warning and otherwise ignores the @code{section} attribute in
3243 uninitialized variable declarations.
3245 You may only use the @code{section} attribute with a fully initialized
3246 global definition because of the way linkers work. The linker requires
3247 each object be defined once, with the exception that uninitialized
3248 variables tentatively go in the @code{common} (or @code{bss}) section
3249 and can be multiply ``defined''. You can force a variable to be
3250 initialized with the @option{-fno-common} flag or the @code{nocommon}
3253 Some file formats do not support arbitrary sections so the @code{section}
3254 attribute is not available on all platforms.
3255 If you need to map the entire contents of a module to a particular
3256 section, consider using the facilities of the linker instead.
3259 @cindex @code{shared} variable attribute
3260 On Windows, in addition to putting variable definitions in a named
3261 section, the section can also be shared among all running copies of an
3262 executable or DLL@. For example, this small program defines shared data
3263 by putting it in a named section @code{shared} and marking the section
3267 int foo __attribute__((section ("shared"), shared)) = 0;
3272 /* Read and write foo. All running
3273 copies see the same value. */
3279 You may only use the @code{shared} attribute along with @code{section}
3280 attribute with a fully initialized global definition because of the way
3281 linkers work. See @code{section} attribute for more information.
3283 The @code{shared} attribute is only available on Windows@.
3285 @item tls_model ("@var{tls_model}")
3286 @cindex @code{tls_model} attribute
3287 The @code{tls_model} attribute sets thread-local storage model
3288 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3289 overriding @code{-ftls-model=} command line switch on a per-variable
3291 The @var{tls_model} argument should be one of @code{global-dynamic},
3292 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3294 Not all targets support this attribute.
3296 @item transparent_union
3297 This attribute, attached to a function parameter which is a union, means
3298 that the corresponding argument may have the type of any union member,
3299 but the argument is passed as if its type were that of the first union
3300 member. For more details see @xref{Type Attributes}. You can also use
3301 this attribute on a @code{typedef} for a union data type; then it
3302 applies to all function parameters with that type.
3305 This attribute, attached to a variable, means that the variable is meant
3306 to be possibly unused. GCC will not produce a warning for this
3309 @item vector_size (@var{bytes})
3310 This attribute specifies the vector size for the variable, measured in
3311 bytes. For example, the declaration:
3314 int foo __attribute__ ((vector_size (16)));
3318 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3319 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3320 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3322 This attribute is only applicable to integral and float scalars,
3323 although arrays, pointers, and function return values are allowed in
3324 conjunction with this construct.
3326 Aggregates with this attribute are invalid, even if they are of the same
3327 size as a corresponding scalar. For example, the declaration:
3330 struct S @{ int a; @};
3331 struct S __attribute__ ((vector_size (16))) foo;
3335 is invalid even if the size of the structure is the same as the size of
3339 The @code{weak} attribute is described in @xref{Function Attributes}.
3342 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3345 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3349 @subsection M32R/D Variable Attributes
3351 One attribute is currently defined for the M32R/D.
3354 @item model (@var{model-name})
3355 @cindex variable addressability on the M32R/D
3356 Use this attribute on the M32R/D to set the addressability of an object.
3357 The identifier @var{model-name} is one of @code{small}, @code{medium},
3358 or @code{large}, representing each of the code models.
3360 Small model objects live in the lower 16MB of memory (so that their
3361 addresses can be loaded with the @code{ld24} instruction).
3363 Medium and large model objects may live anywhere in the 32-bit address space
3364 (the compiler will generate @code{seth/add3} instructions to load their
3368 @subsection i386 Variable Attributes
3370 Two attributes are currently defined for i386 configurations:
3371 @code{ms_struct} and @code{gcc_struct}
3376 @cindex @code{ms_struct} attribute
3377 @cindex @code{gcc_struct} attribute
3379 If @code{packed} is used on a structure, or if bit-fields are used
3380 it may be that the Microsoft ABI packs them differently
3381 than GCC would normally pack them. Particularly when moving packed
3382 data between functions compiled with GCC and the native Microsoft compiler
3383 (either via function call or as data in a file), it may be necessary to access
3386 Currently @option{-m[no-]ms-bitfields} is provided for the Windows X86
3387 compilers to match the native Microsoft compiler.
3390 @node Type Attributes
3391 @section Specifying Attributes of Types
3392 @cindex attribute of types
3393 @cindex type attributes
3395 The keyword @code{__attribute__} allows you to specify special
3396 attributes of @code{struct} and @code{union} types when you define such
3397 types. This keyword is followed by an attribute specification inside
3398 double parentheses. Six attributes are currently defined for types:
3399 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3400 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3401 functions (@pxref{Function Attributes}) and for variables
3402 (@pxref{Variable Attributes}).
3404 You may also specify any one of these attributes with @samp{__}
3405 preceding and following its keyword. This allows you to use these
3406 attributes in header files without being concerned about a possible
3407 macro of the same name. For example, you may use @code{__aligned__}
3408 instead of @code{aligned}.
3410 You may specify the @code{aligned} and @code{transparent_union}
3411 attributes either in a @code{typedef} declaration or just past the
3412 closing curly brace of a complete enum, struct or union type
3413 @emph{definition} and the @code{packed} attribute only past the closing
3414 brace of a definition.
3416 You may also specify attributes between the enum, struct or union
3417 tag and the name of the type rather than after the closing brace.
3419 @xref{Attribute Syntax}, for details of the exact syntax for using
3423 @cindex @code{aligned} attribute
3424 @item aligned (@var{alignment})
3425 This attribute specifies a minimum alignment (in bytes) for variables
3426 of the specified type. For example, the declarations:
3429 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3430 typedef int more_aligned_int __attribute__ ((aligned (8)));
3434 force the compiler to insure (as far as it can) that each variable whose
3435 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3436 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3437 variables of type @code{struct S} aligned to 8-byte boundaries allows
3438 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3439 store) instructions when copying one variable of type @code{struct S} to
3440 another, thus improving run-time efficiency.
3442 Note that the alignment of any given @code{struct} or @code{union} type
3443 is required by the ISO C standard to be at least a perfect multiple of
3444 the lowest common multiple of the alignments of all of the members of
3445 the @code{struct} or @code{union} in question. This means that you @emph{can}
3446 effectively adjust the alignment of a @code{struct} or @code{union}
3447 type by attaching an @code{aligned} attribute to any one of the members
3448 of such a type, but the notation illustrated in the example above is a
3449 more obvious, intuitive, and readable way to request the compiler to
3450 adjust the alignment of an entire @code{struct} or @code{union} type.
3452 As in the preceding example, you can explicitly specify the alignment
3453 (in bytes) that you wish the compiler to use for a given @code{struct}
3454 or @code{union} type. Alternatively, you can leave out the alignment factor
3455 and just ask the compiler to align a type to the maximum
3456 useful alignment for the target machine you are compiling for. For
3457 example, you could write:
3460 struct S @{ short f[3]; @} __attribute__ ((aligned));
3463 Whenever you leave out the alignment factor in an @code{aligned}
3464 attribute specification, the compiler automatically sets the alignment
3465 for the type to the largest alignment which is ever used for any data
3466 type on the target machine you are compiling for. Doing this can often
3467 make copy operations more efficient, because the compiler can use
3468 whatever instructions copy the biggest chunks of memory when performing
3469 copies to or from the variables which have types that you have aligned
3472 In the example above, if the size of each @code{short} is 2 bytes, then
3473 the size of the entire @code{struct S} type is 6 bytes. The smallest
3474 power of two which is greater than or equal to that is 8, so the
3475 compiler sets the alignment for the entire @code{struct S} type to 8
3478 Note that although you can ask the compiler to select a time-efficient
3479 alignment for a given type and then declare only individual stand-alone
3480 objects of that type, the compiler's ability to select a time-efficient
3481 alignment is primarily useful only when you plan to create arrays of
3482 variables having the relevant (efficiently aligned) type. If you
3483 declare or use arrays of variables of an efficiently-aligned type, then
3484 it is likely that your program will also be doing pointer arithmetic (or
3485 subscripting, which amounts to the same thing) on pointers to the
3486 relevant type, and the code that the compiler generates for these
3487 pointer arithmetic operations will often be more efficient for
3488 efficiently-aligned types than for other types.
3490 The @code{aligned} attribute can only increase the alignment; but you
3491 can decrease it by specifying @code{packed} as well. See below.
3493 Note that the effectiveness of @code{aligned} attributes may be limited
3494 by inherent limitations in your linker. On many systems, the linker is
3495 only able to arrange for variables to be aligned up to a certain maximum
3496 alignment. (For some linkers, the maximum supported alignment may
3497 be very very small.) If your linker is only able to align variables
3498 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3499 in an @code{__attribute__} will still only provide you with 8 byte
3500 alignment. See your linker documentation for further information.
3503 This attribute, attached to @code{struct} or @code{union} type
3504 definition, specifies that each member of the structure or union is
3505 placed to minimize the memory required. When attached to an @code{enum}
3506 definition, it indicates that the smallest integral type should be used.
3508 @opindex fshort-enums
3509 Specifying this attribute for @code{struct} and @code{union} types is
3510 equivalent to specifying the @code{packed} attribute on each of the
3511 structure or union members. Specifying the @option{-fshort-enums}
3512 flag on the line is equivalent to specifying the @code{packed}
3513 attribute on all @code{enum} definitions.
3515 In the following example @code{struct my_packed_struct}'s members are
3516 packed closely together, but the internal layout of its @code{s} member
3517 is not packed -- to do that, @code{struct my_unpacked_struct} would need to
3521 struct my_unpacked_struct
3527 struct my_packed_struct __attribute__ ((__packed__))
3531 struct my_unpacked_struct s;
3535 You may only specify this attribute on the definition of a @code{enum},
3536 @code{struct} or @code{union}, not on a @code{typedef} which does not
3537 also define the enumerated type, structure or union.
3539 @item transparent_union
3540 This attribute, attached to a @code{union} type definition, indicates
3541 that any function parameter having that union type causes calls to that
3542 function to be treated in a special way.
3544 First, the argument corresponding to a transparent union type can be of
3545 any type in the union; no cast is required. Also, if the union contains
3546 a pointer type, the corresponding argument can be a null pointer
3547 constant or a void pointer expression; and if the union contains a void
3548 pointer type, the corresponding argument can be any pointer expression.
3549 If the union member type is a pointer, qualifiers like @code{const} on
3550 the referenced type must be respected, just as with normal pointer
3553 Second, the argument is passed to the function using the calling
3554 conventions of the first member of the transparent union, not the calling
3555 conventions of the union itself. All members of the union must have the
3556 same machine representation; this is necessary for this argument passing
3559 Transparent unions are designed for library functions that have multiple
3560 interfaces for compatibility reasons. For example, suppose the
3561 @code{wait} function must accept either a value of type @code{int *} to
3562 comply with Posix, or a value of type @code{union wait *} to comply with
3563 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3564 @code{wait} would accept both kinds of arguments, but it would also
3565 accept any other pointer type and this would make argument type checking
3566 less useful. Instead, @code{<sys/wait.h>} might define the interface
3574 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3576 pid_t wait (wait_status_ptr_t);
3579 This interface allows either @code{int *} or @code{union wait *}
3580 arguments to be passed, using the @code{int *} calling convention.
3581 The program can call @code{wait} with arguments of either type:
3584 int w1 () @{ int w; return wait (&w); @}
3585 int w2 () @{ union wait w; return wait (&w); @}
3588 With this interface, @code{wait}'s implementation might look like this:
3591 pid_t wait (wait_status_ptr_t p)
3593 return waitpid (-1, p.__ip, 0);
3598 When attached to a type (including a @code{union} or a @code{struct}),
3599 this attribute means that variables of that type are meant to appear
3600 possibly unused. GCC will not produce a warning for any variables of
3601 that type, even if the variable appears to do nothing. This is often
3602 the case with lock or thread classes, which are usually defined and then
3603 not referenced, but contain constructors and destructors that have
3604 nontrivial bookkeeping functions.
3607 The @code{deprecated} attribute results in a warning if the type
3608 is used anywhere in the source file. This is useful when identifying
3609 types that are expected to be removed in a future version of a program.
3610 If possible, the warning also includes the location of the declaration
3611 of the deprecated type, to enable users to easily find further
3612 information about why the type is deprecated, or what they should do
3613 instead. Note that the warnings only occur for uses and then only
3614 if the type is being applied to an identifier that itself is not being
3615 declared as deprecated.
3618 typedef int T1 __attribute__ ((deprecated));
3622 typedef T1 T3 __attribute__ ((deprecated));
3623 T3 z __attribute__ ((deprecated));
3626 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3627 warning is issued for line 4 because T2 is not explicitly
3628 deprecated. Line 5 has no warning because T3 is explicitly
3629 deprecated. Similarly for line 6.
3631 The @code{deprecated} attribute can also be used for functions and
3632 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3635 Accesses to objects with types with this attribute are not subjected to
3636 type-based alias analysis, but are instead assumed to be able to alias
3637 any other type of objects, just like the @code{char} type. See
3638 @option{-fstrict-aliasing} for more information on aliasing issues.
3643 typedef short __attribute__((__may_alias__)) short_a;
3649 short_a *b = (short_a *) &a;
3653 if (a == 0x12345678)
3660 If you replaced @code{short_a} with @code{short} in the variable
3661 declaration, the above program would abort when compiled with
3662 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3663 above in recent GCC versions.
3665 @subsection i386 Type Attributes
3667 Two attributes are currently defined for i386 configurations:
3668 @code{ms_struct} and @code{gcc_struct}
3672 @cindex @code{ms_struct}
3673 @cindex @code{gcc_struct}
3675 If @code{packed} is used on a structure, or if bit-fields are used
3676 it may be that the Microsoft ABI packs them differently
3677 than GCC would normally pack them. Particularly when moving packed
3678 data between functions compiled with GCC and the native Microsoft compiler
3679 (either via function call or as data in a file), it may be necessary to access
3682 Currently @option{-m[no-]ms-bitfields} is provided for the Windows X86
3683 compilers to match the native Microsoft compiler.
3686 To specify multiple attributes, separate them by commas within the
3687 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3691 @section An Inline Function is As Fast As a Macro
3692 @cindex inline functions
3693 @cindex integrating function code
3695 @cindex macros, inline alternative
3697 By declaring a function @code{inline}, you can direct GCC to
3698 integrate that function's code into the code for its callers. This
3699 makes execution faster by eliminating the function-call overhead; in
3700 addition, if any of the actual argument values are constant, their known
3701 values may permit simplifications at compile time so that not all of the
3702 inline function's code needs to be included. The effect on code size is
3703 less predictable; object code may be larger or smaller with function
3704 inlining, depending on the particular case. Inlining of functions is an
3705 optimization and it really ``works'' only in optimizing compilation. If
3706 you don't use @option{-O}, no function is really inline.
3708 Inline functions are included in the ISO C99 standard, but there are
3709 currently substantial differences between what GCC implements and what
3710 the ISO C99 standard requires.
3712 To declare a function inline, use the @code{inline} keyword in its
3713 declaration, like this:
3723 (If you are writing a header file to be included in ISO C programs, write
3724 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3725 You can also make all ``simple enough'' functions inline with the option
3726 @option{-finline-functions}.
3729 Note that certain usages in a function definition can make it unsuitable
3730 for inline substitution. Among these usages are: use of varargs, use of
3731 alloca, use of variable sized data types (@pxref{Variable Length}),
3732 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3733 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3734 will warn when a function marked @code{inline} could not be substituted,
3735 and will give the reason for the failure.
3737 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3738 does not affect the linkage of the function.
3740 @cindex automatic @code{inline} for C++ member fns
3741 @cindex @code{inline} automatic for C++ member fns
3742 @cindex member fns, automatically @code{inline}
3743 @cindex C++ member fns, automatically @code{inline}
3744 @opindex fno-default-inline
3745 GCC automatically inlines member functions defined within the class
3746 body of C++ programs even if they are not explicitly declared
3747 @code{inline}. (You can override this with @option{-fno-default-inline};
3748 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3750 @cindex inline functions, omission of
3751 @opindex fkeep-inline-functions
3752 When a function is both inline and @code{static}, if all calls to the
3753 function are integrated into the caller, and the function's address is
3754 never used, then the function's own assembler code is never referenced.
3755 In this case, GCC does not actually output assembler code for the
3756 function, unless you specify the option @option{-fkeep-inline-functions}.
3757 Some calls cannot be integrated for various reasons (in particular,
3758 calls that precede the function's definition cannot be integrated, and
3759 neither can recursive calls within the definition). If there is a
3760 nonintegrated call, then the function is compiled to assembler code as
3761 usual. The function must also be compiled as usual if the program
3762 refers to its address, because that can't be inlined.
3764 @cindex non-static inline function
3765 When an inline function is not @code{static}, then the compiler must assume
3766 that there may be calls from other source files; since a global symbol can
3767 be defined only once in any program, the function must not be defined in
3768 the other source files, so the calls therein cannot be integrated.
3769 Therefore, a non-@code{static} inline function is always compiled on its
3770 own in the usual fashion.
3772 If you specify both @code{inline} and @code{extern} in the function
3773 definition, then the definition is used only for inlining. In no case
3774 is the function compiled on its own, not even if you refer to its
3775 address explicitly. Such an address becomes an external reference, as
3776 if you had only declared the function, and had not defined it.
3778 This combination of @code{inline} and @code{extern} has almost the
3779 effect of a macro. The way to use it is to put a function definition in
3780 a header file with these keywords, and put another copy of the
3781 definition (lacking @code{inline} and @code{extern}) in a library file.
3782 The definition in the header file will cause most calls to the function
3783 to be inlined. If any uses of the function remain, they will refer to
3784 the single copy in the library.
3786 Since GCC eventually will implement ISO C99 semantics for
3787 inline functions, it is best to use @code{static inline} only
3788 to guarentee compatibility. (The
3789 existing semantics will remain available when @option{-std=gnu89} is
3790 specified, but eventually the default will be @option{-std=gnu99} and
3791 that will implement the C99 semantics, though it does not do so yet.)
3793 GCC does not inline any functions when not optimizing unless you specify
3794 the @samp{always_inline} attribute for the function, like this:
3798 inline void foo (const char) __attribute__((always_inline));
3802 @section Assembler Instructions with C Expression Operands
3803 @cindex extended @code{asm}
3804 @cindex @code{asm} expressions
3805 @cindex assembler instructions
3808 In an assembler instruction using @code{asm}, you can specify the
3809 operands of the instruction using C expressions. This means you need not
3810 guess which registers or memory locations will contain the data you want
3813 You must specify an assembler instruction template much like what
3814 appears in a machine description, plus an operand constraint string for
3817 For example, here is how to use the 68881's @code{fsinx} instruction:
3820 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3824 Here @code{angle} is the C expression for the input operand while
3825 @code{result} is that of the output operand. Each has @samp{"f"} as its
3826 operand constraint, saying that a floating point register is required.
3827 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3828 output operands' constraints must use @samp{=}. The constraints use the
3829 same language used in the machine description (@pxref{Constraints}).
3831 Each operand is described by an operand-constraint string followed by
3832 the C expression in parentheses. A colon separates the assembler
3833 template from the first output operand and another separates the last
3834 output operand from the first input, if any. Commas separate the
3835 operands within each group. The total number of operands is currently
3836 limited to 30; this limitation may be lifted in some future version of
3839 If there are no output operands but there are input operands, you must
3840 place two consecutive colons surrounding the place where the output
3843 As of GCC version 3.1, it is also possible to specify input and output
3844 operands using symbolic names which can be referenced within the
3845 assembler code. These names are specified inside square brackets
3846 preceding the constraint string, and can be referenced inside the
3847 assembler code using @code{%[@var{name}]} instead of a percentage sign
3848 followed by the operand number. Using named operands the above example
3852 asm ("fsinx %[angle],%[output]"
3853 : [output] "=f" (result)
3854 : [angle] "f" (angle));
3858 Note that the symbolic operand names have no relation whatsoever to
3859 other C identifiers. You may use any name you like, even those of
3860 existing C symbols, but you must ensure that no two operands within the same
3861 assembler construct use the same symbolic name.
3863 Output operand expressions must be lvalues; the compiler can check this.
3864 The input operands need not be lvalues. The compiler cannot check
3865 whether the operands have data types that are reasonable for the
3866 instruction being executed. It does not parse the assembler instruction
3867 template and does not know what it means or even whether it is valid
3868 assembler input. The extended @code{asm} feature is most often used for
3869 machine instructions the compiler itself does not know exist. If
3870 the output expression cannot be directly addressed (for example, it is a
3871 bit-field), your constraint must allow a register. In that case, GCC
3872 will use the register as the output of the @code{asm}, and then store
3873 that register into the output.
3875 The ordinary output operands must be write-only; GCC will assume that
3876 the values in these operands before the instruction are dead and need
3877 not be generated. Extended asm supports input-output or read-write
3878 operands. Use the constraint character @samp{+} to indicate such an
3879 operand and list it with the output operands.
3881 When the constraints for the read-write operand (or the operand in which
3882 only some of the bits are to be changed) allows a register, you may, as
3883 an alternative, logically split its function into two separate operands,
3884 one input operand and one write-only output operand. The connection
3885 between them is expressed by constraints which say they need to be in
3886 the same location when the instruction executes. You can use the same C
3887 expression for both operands, or different expressions. For example,
3888 here we write the (fictitious) @samp{combine} instruction with
3889 @code{bar} as its read-only source operand and @code{foo} as its
3890 read-write destination:
3893 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3897 The constraint @samp{"0"} for operand 1 says that it must occupy the
3898 same location as operand 0. A number in constraint is allowed only in
3899 an input operand and it must refer to an output operand.
3901 Only a number in the constraint can guarantee that one operand will be in
3902 the same place as another. The mere fact that @code{foo} is the value
3903 of both operands is not enough to guarantee that they will be in the
3904 same place in the generated assembler code. The following would not
3908 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3911 Various optimizations or reloading could cause operands 0 and 1 to be in
3912 different registers; GCC knows no reason not to do so. For example, the
3913 compiler might find a copy of the value of @code{foo} in one register and
3914 use it for operand 1, but generate the output operand 0 in a different
3915 register (copying it afterward to @code{foo}'s own address). Of course,
3916 since the register for operand 1 is not even mentioned in the assembler
3917 code, the result will not work, but GCC can't tell that.
3919 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3920 the operand number for a matching constraint. For example:
3923 asm ("cmoveq %1,%2,%[result]"
3924 : [result] "=r"(result)
3925 : "r" (test), "r"(new), "[result]"(old));
3928 Some instructions clobber specific hard registers. To describe this,
3929 write a third colon after the input operands, followed by the names of
3930 the clobbered hard registers (given as strings). Here is a realistic
3931 example for the VAX:
3934 asm volatile ("movc3 %0,%1,%2"
3936 : "g" (from), "g" (to), "g" (count)
3937 : "r0", "r1", "r2", "r3", "r4", "r5");
3940 You may not write a clobber description in a way that overlaps with an
3941 input or output operand. For example, you may not have an operand
3942 describing a register class with one member if you mention that register
3943 in the clobber list. Variables declared to live in specific registers
3944 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3945 have no part mentioned in the clobber description.
3946 There is no way for you to specify that an input
3947 operand is modified without also specifying it as an output
3948 operand. Note that if all the output operands you specify are for this
3949 purpose (and hence unused), you will then also need to specify
3950 @code{volatile} for the @code{asm} construct, as described below, to
3951 prevent GCC from deleting the @code{asm} statement as unused.
3953 If you refer to a particular hardware register from the assembler code,
3954 you will probably have to list the register after the third colon to
3955 tell the compiler the register's value is modified. In some assemblers,
3956 the register names begin with @samp{%}; to produce one @samp{%} in the
3957 assembler code, you must write @samp{%%} in the input.
3959 If your assembler instruction can alter the condition code register, add
3960 @samp{cc} to the list of clobbered registers. GCC on some machines
3961 represents the condition codes as a specific hardware register;
3962 @samp{cc} serves to name this register. On other machines, the
3963 condition code is handled differently, and specifying @samp{cc} has no
3964 effect. But it is valid no matter what the machine.
3966 If your assembler instruction modifies memory in an unpredictable
3967 fashion, add @samp{memory} to the list of clobbered registers. This
3968 will cause GCC to not keep memory values cached in registers across
3969 the assembler instruction. You will also want to add the
3970 @code{volatile} keyword if the memory affected is not listed in the
3971 inputs or outputs of the @code{asm}, as the @samp{memory} clobber does
3972 not count as a side-effect of the @code{asm}.
3974 You can put multiple assembler instructions together in a single
3975 @code{asm} template, separated by the characters normally used in assembly
3976 code for the system. A combination that works in most places is a newline
3977 to break the line, plus a tab character to move to the instruction field
3978 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3979 assembler allows semicolons as a line-breaking character. Note that some
3980 assembler dialects use semicolons to start a comment.
3981 The input operands are guaranteed not to use any of the clobbered
3982 registers, and neither will the output operands' addresses, so you can
3983 read and write the clobbered registers as many times as you like. Here
3984 is an example of multiple instructions in a template; it assumes the
3985 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3988 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3990 : "g" (from), "g" (to)
3994 Unless an output operand has the @samp{&} constraint modifier, GCC
3995 may allocate it in the same register as an unrelated input operand, on
3996 the assumption the inputs are consumed before the outputs are produced.
3997 This assumption may be false if the assembler code actually consists of
3998 more than one instruction. In such a case, use @samp{&} for each output
3999 operand that may not overlap an input. @xref{Modifiers}.
4001 If you want to test the condition code produced by an assembler
4002 instruction, you must include a branch and a label in the @code{asm}
4003 construct, as follows:
4006 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4012 This assumes your assembler supports local labels, as the GNU assembler
4013 and most Unix assemblers do.
4015 Speaking of labels, jumps from one @code{asm} to another are not
4016 supported. The compiler's optimizers do not know about these jumps, and
4017 therefore they cannot take account of them when deciding how to
4020 @cindex macros containing @code{asm}
4021 Usually the most convenient way to use these @code{asm} instructions is to
4022 encapsulate them in macros that look like functions. For example,
4026 (@{ double __value, __arg = (x); \
4027 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4032 Here the variable @code{__arg} is used to make sure that the instruction
4033 operates on a proper @code{double} value, and to accept only those
4034 arguments @code{x} which can convert automatically to a @code{double}.
4036 Another way to make sure the instruction operates on the correct data
4037 type is to use a cast in the @code{asm}. This is different from using a
4038 variable @code{__arg} in that it converts more different types. For
4039 example, if the desired type were @code{int}, casting the argument to
4040 @code{int} would accept a pointer with no complaint, while assigning the
4041 argument to an @code{int} variable named @code{__arg} would warn about
4042 using a pointer unless the caller explicitly casts it.
4044 If an @code{asm} has output operands, GCC assumes for optimization
4045 purposes the instruction has no side effects except to change the output
4046 operands. This does not mean instructions with a side effect cannot be
4047 used, but you must be careful, because the compiler may eliminate them
4048 if the output operands aren't used, or move them out of loops, or
4049 replace two with one if they constitute a common subexpression. Also,
4050 if your instruction does have a side effect on a variable that otherwise
4051 appears not to change, the old value of the variable may be reused later
4052 if it happens to be found in a register.
4054 You can prevent an @code{asm} instruction from being deleted, moved
4055 significantly, or combined, by writing the keyword @code{volatile} after
4056 the @code{asm}. For example:
4059 #define get_and_set_priority(new) \
4061 asm volatile ("get_and_set_priority %0, %1" \
4062 : "=g" (__old) : "g" (new)); \
4067 If you write an @code{asm} instruction with no outputs, GCC will know
4068 the instruction has side-effects and will not delete the instruction or
4069 move it outside of loops.
4071 The @code{volatile} keyword indicates that the instruction has
4072 important side-effects. GCC will not delete a volatile @code{asm} if
4073 it is reachable. (The instruction can still be deleted if GCC can
4074 prove that control-flow will never reach the location of the
4075 instruction.) In addition, GCC will not reschedule instructions
4076 across a volatile @code{asm} instruction. For example:
4079 *(volatile int *)addr = foo;
4080 asm volatile ("eieio" : : );
4084 Assume @code{addr} contains the address of a memory mapped device
4085 register. The PowerPC @code{eieio} instruction (Enforce In-order
4086 Execution of I/O) tells the CPU to make sure that the store to that
4087 device register happens before it issues any other I/O@.
4089 Note that even a volatile @code{asm} instruction can be moved in ways
4090 that appear insignificant to the compiler, such as across jump
4091 instructions. You can't expect a sequence of volatile @code{asm}
4092 instructions to remain perfectly consecutive. If you want consecutive
4093 output, use a single @code{asm}. Also, GCC will perform some
4094 optimizations across a volatile @code{asm} instruction; GCC does not
4095 ``forget everything'' when it encounters a volatile @code{asm}
4096 instruction the way some other compilers do.
4098 An @code{asm} instruction without any operands or clobbers (an ``old
4099 style'' @code{asm}) will be treated identically to a volatile
4100 @code{asm} instruction.
4102 It is a natural idea to look for a way to give access to the condition
4103 code left by the assembler instruction. However, when we attempted to
4104 implement this, we found no way to make it work reliably. The problem
4105 is that output operands might need reloading, which would result in
4106 additional following ``store'' instructions. On most machines, these
4107 instructions would alter the condition code before there was time to
4108 test it. This problem doesn't arise for ordinary ``test'' and
4109 ``compare'' instructions because they don't have any output operands.
4111 For reasons similar to those described above, it is not possible to give
4112 an assembler instruction access to the condition code left by previous
4115 If you are writing a header file that should be includable in ISO C
4116 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4119 @subsection i386 floating point asm operands
4121 There are several rules on the usage of stack-like regs in
4122 asm_operands insns. These rules apply only to the operands that are
4127 Given a set of input regs that die in an asm_operands, it is
4128 necessary to know which are implicitly popped by the asm, and
4129 which must be explicitly popped by gcc.
4131 An input reg that is implicitly popped by the asm must be
4132 explicitly clobbered, unless it is constrained to match an
4136 For any input reg that is implicitly popped by an asm, it is
4137 necessary to know how to adjust the stack to compensate for the pop.
4138 If any non-popped input is closer to the top of the reg-stack than
4139 the implicitly popped reg, it would not be possible to know what the
4140 stack looked like---it's not clear how the rest of the stack ``slides
4143 All implicitly popped input regs must be closer to the top of
4144 the reg-stack than any input that is not implicitly popped.
4146 It is possible that if an input dies in an insn, reload might
4147 use the input reg for an output reload. Consider this example:
4150 asm ("foo" : "=t" (a) : "f" (b));
4153 This asm says that input B is not popped by the asm, and that
4154 the asm pushes a result onto the reg-stack, i.e., the stack is one
4155 deeper after the asm than it was before. But, it is possible that
4156 reload will think that it can use the same reg for both the input and
4157 the output, if input B dies in this insn.
4159 If any input operand uses the @code{f} constraint, all output reg
4160 constraints must use the @code{&} earlyclobber.
4162 The asm above would be written as
4165 asm ("foo" : "=&t" (a) : "f" (b));
4169 Some operands need to be in particular places on the stack. All
4170 output operands fall in this category---there is no other way to
4171 know which regs the outputs appear in unless the user indicates
4172 this in the constraints.
4174 Output operands must specifically indicate which reg an output
4175 appears in after an asm. @code{=f} is not allowed: the operand
4176 constraints must select a class with a single reg.
4179 Output operands may not be ``inserted'' between existing stack regs.
4180 Since no 387 opcode uses a read/write operand, all output operands
4181 are dead before the asm_operands, and are pushed by the asm_operands.
4182 It makes no sense to push anywhere but the top of the reg-stack.
4184 Output operands must start at the top of the reg-stack: output
4185 operands may not ``skip'' a reg.
4188 Some asm statements may need extra stack space for internal
4189 calculations. This can be guaranteed by clobbering stack registers
4190 unrelated to the inputs and outputs.
4194 Here are a couple of reasonable asms to want to write. This asm
4195 takes one input, which is internally popped, and produces two outputs.
4198 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4201 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4202 and replaces them with one output. The user must code the @code{st(1)}
4203 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4206 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4212 @section Controlling Names Used in Assembler Code
4213 @cindex assembler names for identifiers
4214 @cindex names used in assembler code
4215 @cindex identifiers, names in assembler code
4217 You can specify the name to be used in the assembler code for a C
4218 function or variable by writing the @code{asm} (or @code{__asm__})
4219 keyword after the declarator as follows:
4222 int foo asm ("myfoo") = 2;
4226 This specifies that the name to be used for the variable @code{foo} in
4227 the assembler code should be @samp{myfoo} rather than the usual
4230 On systems where an underscore is normally prepended to the name of a C
4231 function or variable, this feature allows you to define names for the
4232 linker that do not start with an underscore.
4234 It does not make sense to use this feature with a non-static local
4235 variable since such variables do not have assembler names. If you are
4236 trying to put the variable in a particular register, see @ref{Explicit
4237 Reg Vars}. GCC presently accepts such code with a warning, but will
4238 probably be changed to issue an error, rather than a warning, in the
4241 You cannot use @code{asm} in this way in a function @emph{definition}; but
4242 you can get the same effect by writing a declaration for the function
4243 before its definition and putting @code{asm} there, like this:
4246 extern func () asm ("FUNC");
4253 It is up to you to make sure that the assembler names you choose do not
4254 conflict with any other assembler symbols. Also, you must not use a
4255 register name; that would produce completely invalid assembler code. GCC
4256 does not as yet have the ability to store static variables in registers.
4257 Perhaps that will be added.
4259 @node Explicit Reg Vars
4260 @section Variables in Specified Registers
4261 @cindex explicit register variables
4262 @cindex variables in specified registers
4263 @cindex specified registers
4264 @cindex registers, global allocation
4266 GNU C allows you to put a few global variables into specified hardware
4267 registers. You can also specify the register in which an ordinary
4268 register variable should be allocated.
4272 Global register variables reserve registers throughout the program.
4273 This may be useful in programs such as programming language
4274 interpreters which have a couple of global variables that are accessed
4278 Local register variables in specific registers do not reserve the
4279 registers. The compiler's data flow analysis is capable of determining
4280 where the specified registers contain live values, and where they are
4281 available for other uses. Stores into local register variables may be deleted
4282 when they appear to be dead according to dataflow analysis. References
4283 to local register variables may be deleted or moved or simplified.
4285 These local variables are sometimes convenient for use with the extended
4286 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4287 output of the assembler instruction directly into a particular register.
4288 (This will work provided the register you specify fits the constraints
4289 specified for that operand in the @code{asm}.)
4297 @node Global Reg Vars
4298 @subsection Defining Global Register Variables
4299 @cindex global register variables
4300 @cindex registers, global variables in
4302 You can define a global register variable in GNU C like this:
4305 register int *foo asm ("a5");
4309 Here @code{a5} is the name of the register which should be used. Choose a
4310 register which is normally saved and restored by function calls on your
4311 machine, so that library routines will not clobber it.
4313 Naturally the register name is cpu-dependent, so you would need to
4314 conditionalize your program according to cpu type. The register
4315 @code{a5} would be a good choice on a 68000 for a variable of pointer
4316 type. On machines with register windows, be sure to choose a ``global''
4317 register that is not affected magically by the function call mechanism.
4319 In addition, operating systems on one type of cpu may differ in how they
4320 name the registers; then you would need additional conditionals. For
4321 example, some 68000 operating systems call this register @code{%a5}.
4323 Eventually there may be a way of asking the compiler to choose a register
4324 automatically, but first we need to figure out how it should choose and
4325 how to enable you to guide the choice. No solution is evident.
4327 Defining a global register variable in a certain register reserves that
4328 register entirely for this use, at least within the current compilation.
4329 The register will not be allocated for any other purpose in the functions
4330 in the current compilation. The register will not be saved and restored by
4331 these functions. Stores into this register are never deleted even if they
4332 would appear to be dead, but references may be deleted or moved or
4335 It is not safe to access the global register variables from signal
4336 handlers, or from more than one thread of control, because the system
4337 library routines may temporarily use the register for other things (unless
4338 you recompile them specially for the task at hand).
4340 @cindex @code{qsort}, and global register variables
4341 It is not safe for one function that uses a global register variable to
4342 call another such function @code{foo} by way of a third function
4343 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4344 different source file in which the variable wasn't declared). This is
4345 because @code{lose} might save the register and put some other value there.
4346 For example, you can't expect a global register variable to be available in
4347 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4348 might have put something else in that register. (If you are prepared to
4349 recompile @code{qsort} with the same global register variable, you can
4350 solve this problem.)
4352 If you want to recompile @code{qsort} or other source files which do not
4353 actually use your global register variable, so that they will not use that
4354 register for any other purpose, then it suffices to specify the compiler
4355 option @option{-ffixed-@var{reg}}. You need not actually add a global
4356 register declaration to their source code.
4358 A function which can alter the value of a global register variable cannot
4359 safely be called from a function compiled without this variable, because it
4360 could clobber the value the caller expects to find there on return.
4361 Therefore, the function which is the entry point into the part of the
4362 program that uses the global register variable must explicitly save and
4363 restore the value which belongs to its caller.
4365 @cindex register variable after @code{longjmp}
4366 @cindex global register after @code{longjmp}
4367 @cindex value after @code{longjmp}
4370 On most machines, @code{longjmp} will restore to each global register
4371 variable the value it had at the time of the @code{setjmp}. On some
4372 machines, however, @code{longjmp} will not change the value of global
4373 register variables. To be portable, the function that called @code{setjmp}
4374 should make other arrangements to save the values of the global register
4375 variables, and to restore them in a @code{longjmp}. This way, the same
4376 thing will happen regardless of what @code{longjmp} does.
4378 All global register variable declarations must precede all function
4379 definitions. If such a declaration could appear after function
4380 definitions, the declaration would be too late to prevent the register from
4381 being used for other purposes in the preceding functions.
4383 Global register variables may not have initial values, because an
4384 executable file has no means to supply initial contents for a register.
4386 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4387 registers, but certain library functions, such as @code{getwd}, as well
4388 as the subroutines for division and remainder, modify g3 and g4. g1 and
4389 g2 are local temporaries.
4391 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4392 Of course, it will not do to use more than a few of those.
4394 @node Local Reg Vars
4395 @subsection Specifying Registers for Local Variables
4396 @cindex local variables, specifying registers
4397 @cindex specifying registers for local variables
4398 @cindex registers for local variables
4400 You can define a local register variable with a specified register
4404 register int *foo asm ("a5");
4408 Here @code{a5} is the name of the register which should be used. Note
4409 that this is the same syntax used for defining global register
4410 variables, but for a local variable it would appear within a function.
4412 Naturally the register name is cpu-dependent, but this is not a
4413 problem, since specific registers are most often useful with explicit
4414 assembler instructions (@pxref{Extended Asm}). Both of these things
4415 generally require that you conditionalize your program according to
4418 In addition, operating systems on one type of cpu may differ in how they
4419 name the registers; then you would need additional conditionals. For
4420 example, some 68000 operating systems call this register @code{%a5}.
4422 Defining such a register variable does not reserve the register; it
4423 remains available for other uses in places where flow control determines
4424 the variable's value is not live. However, these registers are made
4425 unavailable for use in the reload pass; excessive use of this feature
4426 leaves the compiler too few available registers to compile certain
4429 This option does not guarantee that GCC will generate code that has
4430 this variable in the register you specify at all times. You may not
4431 code an explicit reference to this register in an @code{asm} statement
4432 and assume it will always refer to this variable.
4434 Stores into local register variables may be deleted when they appear to be dead
4435 according to dataflow analysis. References to local register variables may
4436 be deleted or moved or simplified.
4438 @node Alternate Keywords
4439 @section Alternate Keywords
4440 @cindex alternate keywords
4441 @cindex keywords, alternate
4443 @option{-ansi} and the various @option{-std} options disable certain
4444 keywords. This causes trouble when you want to use GNU C extensions, or
4445 a general-purpose header file that should be usable by all programs,
4446 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4447 @code{inline} are not available in programs compiled with
4448 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4449 program compiled with @option{-std=c99}). The ISO C99 keyword
4450 @code{restrict} is only available when @option{-std=gnu99} (which will
4451 eventually be the default) or @option{-std=c99} (or the equivalent
4452 @option{-std=iso9899:1999}) is used.
4454 The way to solve these problems is to put @samp{__} at the beginning and
4455 end of each problematical keyword. For example, use @code{__asm__}
4456 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4458 Other C compilers won't accept these alternative keywords; if you want to
4459 compile with another compiler, you can define the alternate keywords as
4460 macros to replace them with the customary keywords. It looks like this:
4468 @findex __extension__
4470 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4472 prevent such warnings within one expression by writing
4473 @code{__extension__} before the expression. @code{__extension__} has no
4474 effect aside from this.
4476 @node Incomplete Enums
4477 @section Incomplete @code{enum} Types
4479 You can define an @code{enum} tag without specifying its possible values.
4480 This results in an incomplete type, much like what you get if you write
4481 @code{struct foo} without describing the elements. A later declaration
4482 which does specify the possible values completes the type.
4484 You can't allocate variables or storage using the type while it is
4485 incomplete. However, you can work with pointers to that type.
4487 This extension may not be very useful, but it makes the handling of
4488 @code{enum} more consistent with the way @code{struct} and @code{union}
4491 This extension is not supported by GNU C++.
4493 @node Function Names
4494 @section Function Names as Strings
4495 @cindex @code{__func__} identifier
4496 @cindex @code{__FUNCTION__} identifier
4497 @cindex @code{__PRETTY_FUNCTION__} identifier
4499 GCC provides three magic variables which hold the name of the current
4500 function, as a string. The first of these is @code{__func__}, which
4501 is part of the C99 standard:
4504 The identifier @code{__func__} is implicitly declared by the translator
4505 as if, immediately following the opening brace of each function
4506 definition, the declaration
4509 static const char __func__[] = "function-name";
4512 appeared, where function-name is the name of the lexically-enclosing
4513 function. This name is the unadorned name of the function.
4516 @code{__FUNCTION__} is another name for @code{__func__}. Older
4517 versions of GCC recognize only this name. However, it is not
4518 standardized. For maximum portability, we recommend you use
4519 @code{__func__}, but provide a fallback definition with the
4523 #if __STDC_VERSION__ < 199901L
4525 # define __func__ __FUNCTION__
4527 # define __func__ "<unknown>"
4532 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4533 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4534 the type signature of the function as well as its bare name. For
4535 example, this program:
4539 extern int printf (char *, ...);
4546 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4547 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4565 __PRETTY_FUNCTION__ = void a::sub(int)
4568 These identifiers are not preprocessor macros. In GCC 3.3 and
4569 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4570 were treated as string literals; they could be used to initialize
4571 @code{char} arrays, and they could be concatenated with other string
4572 literals. GCC 3.4 and later treat them as variables, like
4573 @code{__func__}. In C++, @code{__FUNCTION__} and
4574 @code{__PRETTY_FUNCTION__} have always been variables.
4576 @node Return Address
4577 @section Getting the Return or Frame Address of a Function
4579 These functions may be used to get information about the callers of a
4582 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4583 This function returns the return address of the current function, or of
4584 one of its callers. The @var{level} argument is number of frames to
4585 scan up the call stack. A value of @code{0} yields the return address
4586 of the current function, a value of @code{1} yields the return address
4587 of the caller of the current function, and so forth. When inlining
4588 the expected behavior is that the function will return the address of
4589 the function that will be returned to. To work around this behavior use
4590 the @code{noinline} function attribute.
4592 The @var{level} argument must be a constant integer.
4594 On some machines it may be impossible to determine the return address of
4595 any function other than the current one; in such cases, or when the top
4596 of the stack has been reached, this function will return @code{0} or a
4597 random value. In addition, @code{__builtin_frame_address} may be used
4598 to determine if the top of the stack has been reached.
4600 This function should only be used with a nonzero argument for debugging
4604 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4605 This function is similar to @code{__builtin_return_address}, but it
4606 returns the address of the function frame rather than the return address
4607 of the function. Calling @code{__builtin_frame_address} with a value of
4608 @code{0} yields the frame address of the current function, a value of
4609 @code{1} yields the frame address of the caller of the current function,
4612 The frame is the area on the stack which holds local variables and saved
4613 registers. The frame address is normally the address of the first word
4614 pushed on to the stack by the function. However, the exact definition
4615 depends upon the processor and the calling convention. If the processor
4616 has a dedicated frame pointer register, and the function has a frame,
4617 then @code{__builtin_frame_address} will return the value of the frame
4620 On some machines it may be impossible to determine the frame address of
4621 any function other than the current one; in such cases, or when the top
4622 of the stack has been reached, this function will return @code{0} if
4623 the first frame pointer is properly initialized by the startup code.
4625 This function should only be used with a nonzero argument for debugging
4629 @node Vector Extensions
4630 @section Using vector instructions through built-in functions
4632 On some targets, the instruction set contains SIMD vector instructions that
4633 operate on multiple values contained in one large register at the same time.
4634 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4637 The first step in using these extensions is to provide the necessary data
4638 types. This should be done using an appropriate @code{typedef}:
4641 typedef int v4si __attribute__ ((mode(V4SI)));
4644 The base type @code{int} is effectively ignored by the compiler, the
4645 actual properties of the new type @code{v4si} are defined by the
4646 @code{__attribute__}. It defines the machine mode to be used; for vector
4647 types these have the form @code{V@var{n}@var{B}}; @var{n} should be the
4648 number of elements in the vector, and @var{B} should be the base mode of the
4649 individual elements. The following can be used as base modes:
4653 An integer that is as wide as the smallest addressable unit, usually 8 bits.
4655 An integer, twice as wide as a QI mode integer, usually 16 bits.
4657 An integer, four times as wide as a QI mode integer, usually 32 bits.
4659 An integer, eight times as wide as a QI mode integer, usually 64 bits.
4661 A floating point value, as wide as a SI mode integer, usually 32 bits.
4663 A floating point value, as wide as a DI mode integer, usually 64 bits.
4666 Specifying a combination that is not valid for the current architecture
4667 will cause gcc to synthesize the instructions using a narrower mode.
4668 For example, if you specify a variable of type @code{V4SI} and your
4669 architecture does not allow for this specific SIMD type, gcc will
4670 produce code that uses 4 @code{SIs}.
4672 The types defined in this manner can be used with a subset of normal C
4673 operations. Currently, gcc will allow using the following operators on
4674 these types: @code{+, -, *, /, unary minus}@.
4676 The operations behave like C++ @code{valarrays}. Addition is defined as
4677 the addition of the corresponding elements of the operands. For
4678 example, in the code below, each of the 4 elements in @var{a} will be
4679 added to the corresponding 4 elements in @var{b} and the resulting
4680 vector will be stored in @var{c}.
4683 typedef int v4si __attribute__ ((mode(V4SI)));
4690 Subtraction, multiplication, and division operate in a similar manner.
4691 Likewise, the result of using the unary minus operator on a vector type
4692 is a vector whose elements are the negative value of the corresponding
4693 elements in the operand.
4695 You can declare variables and use them in function calls and returns, as
4696 well as in assignments and some casts. You can specify a vector type as
4697 a return type for a function. Vector types can also be used as function
4698 arguments. It is possible to cast from one vector type to another,
4699 provided they are of the same size (in fact, you can also cast vectors
4700 to and from other datatypes of the same size).
4702 You cannot operate between vectors of different lengths or different
4703 signedness without a cast.
4705 A port that supports hardware vector operations, usually provides a set
4706 of built-in functions that can be used to operate on vectors. For
4707 example, a function to add two vectors and multiply the result by a
4708 third could look like this:
4711 v4si f (v4si a, v4si b, v4si c)
4713 v4si tmp = __builtin_addv4si (a, b);
4714 return __builtin_mulv4si (tmp, c);
4719 @node Other Builtins
4720 @section Other built-in functions provided by GCC
4721 @cindex built-in functions
4722 @findex __builtin_isgreater
4723 @findex __builtin_isgreaterequal
4724 @findex __builtin_isless
4725 @findex __builtin_islessequal
4726 @findex __builtin_islessgreater
4727 @findex __builtin_isunordered
4777 @findex fprintf_unlocked
4779 @findex fputs_unlocked
4801 @findex printf_unlocked
4849 GCC provides a large number of built-in functions other than the ones
4850 mentioned above. Some of these are for internal use in the processing
4851 of exceptions or variable-length argument lists and will not be
4852 documented here because they may change from time to time; we do not
4853 recommend general use of these functions.
4855 The remaining functions are provided for optimization purposes.
4857 @opindex fno-builtin
4858 GCC includes built-in versions of many of the functions in the standard
4859 C library. The versions prefixed with @code{__builtin_} will always be
4860 treated as having the same meaning as the C library function even if you
4861 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
4862 Many of these functions are only optimized in certain cases; if they are
4863 not optimized in a particular case, a call to the library function will
4868 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
4869 @option{-std=c99}), the functions @code{alloca}, @code{bcmp},
4870 @code{bzero}, @code{dcgettext}, @code{dgettext}, @code{_exit},
4871 @code{ffs}, @code{fprintf_unlocked},
4872 @code{fputs_unlocked}, @code{gettext},
4873 @code{index}, @code{mempcpy}, @code{printf_unlocked},
4874 @code{rindex}, @code{stpcpy}, @code{strdup} and @code{strfmon}
4875 may be handled as built-in functions.
4876 All these functions have corresponding versions
4877 prefixed with @code{__builtin_}, which may be used even in strict C89
4880 The ISO C99 functions
4881 @code{cabs}, @code{cabsf}, @code{cabsl},
4882 @code{conj}, @code{conjf}, @code{conjl},
4883 @code{creal}, @code{crealf}, @code{creall},
4884 @code{cimag}, @code{cimagf}, @code{cimagl},
4885 @code{_Exit}, @code{imaxabs}, @code{llabs},
4886 @code{nearbyint}, @code{nearbyintf}, @code{nearbyintl},
4887 @code{round}, @code{roundf}, @code{roundl}, @code{snprintf},
4888 @code{trunc}, @code{truncf}, @code{truncl}, @code{vfscanf},
4889 @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
4890 are handled as built-in functions
4891 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
4893 There are also built-in versions of the ISO C99 functions @code{atan2f},
4894 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
4895 @code{cosf}, @code{cosl},
4896 @code{expf}, @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf},
4897 @code{floorl}, @code{fmodf}, @code{fmodl},
4898 @code{logf}, @code{logl}, @code{powf}, @code{powl},
4899 @code{sinf}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
4900 @code{tanf} and @code{tanl}
4901 that are recognized in any mode since ISO C90 reserves these names for
4902 the purpose to which ISO C99 puts them. All these functions have
4903 corresponding versions prefixed with @code{__builtin_}.
4905 The ISO C90 functions @code{abort}, @code{abs}, @code{atan}, @code{atan2},
4906 @code{calloc}, @code{ceil}, @code{cos}, @code{exit},
4907 @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
4908 @code{fprintf}, @code{fputs}, @code{fscanf},
4909 @code{labs}, @code{log}, @code{malloc},
4910 @code{memcmp}, @code{memcpy}, @code{memset}, @code{pow}, @code{printf},
4911 @code{putchar}, @code{puts}, @code{scanf}, @code{sin}, @code{snprintf},
4912 @code{sprintf}, @code{sqrt}, @code{sscanf},
4913 @code{strcat}, @code{strchr}, @code{strcmp},
4914 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat}, @code{strncmp},
4915 @code{strncpy}, @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
4916 @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
4917 are all recognized as built-in functions unless
4918 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
4919 is specified for an individual function). All of these functions have
4920 corresponding versions prefixed with @code{__builtin_}.
4922 GCC provides built-in versions of the ISO C99 floating point comparison
4923 macros that avoid raising exceptions for unordered operands. They have
4924 the same names as the standard macros ( @code{isgreater},
4925 @code{isgreaterequal}, @code{isless}, @code{islessequal},
4926 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
4927 prefixed. We intend for a library implementor to be able to simply
4928 @code{#define} each standard macro to its built-in equivalent.
4930 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
4932 You can use the built-in function @code{__builtin_types_compatible_p} to
4933 determine whether two types are the same.
4935 This built-in function returns 1 if the unqualified versions of the
4936 types @var{type1} and @var{type2} (which are types, not expressions) are
4937 compatible, 0 otherwise. The result of this built-in function can be
4938 used in integer constant expressions.
4940 This built-in function ignores top level qualifiers (e.g., @code{const},
4941 @code{volatile}). For example, @code{int} is equivalent to @code{const
4944 The type @code{int[]} and @code{int[5]} are compatible. On the other
4945 hand, @code{int} and @code{char *} are not compatible, even if the size
4946 of their types, on the particular architecture are the same. Also, the
4947 amount of pointer indirection is taken into account when determining
4948 similarity. Consequently, @code{short *} is not similar to
4949 @code{short **}. Furthermore, two types that are typedefed are
4950 considered compatible if their underlying types are compatible.
4952 An @code{enum} type is considered to be compatible with another
4953 @code{enum} type. For example, @code{enum @{foo, bar@}} is similar to
4954 @code{enum @{hot, dog@}}.
4956 You would typically use this function in code whose execution varies
4957 depending on the arguments' types. For example:
4963 if (__builtin_types_compatible_p (typeof (x), long double)) \
4964 tmp = foo_long_double (tmp); \
4965 else if (__builtin_types_compatible_p (typeof (x), double)) \
4966 tmp = foo_double (tmp); \
4967 else if (__builtin_types_compatible_p (typeof (x), float)) \
4968 tmp = foo_float (tmp); \
4975 @emph{Note:} This construct is only available for C.
4979 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
4981 You can use the built-in function @code{__builtin_choose_expr} to
4982 evaluate code depending on the value of a constant expression. This
4983 built-in function returns @var{exp1} if @var{const_exp}, which is a
4984 constant expression that must be able to be determined at compile time,
4985 is nonzero. Otherwise it returns 0.
4987 This built-in function is analogous to the @samp{? :} operator in C,
4988 except that the expression returned has its type unaltered by promotion
4989 rules. Also, the built-in function does not evaluate the expression
4990 that was not chosen. For example, if @var{const_exp} evaluates to true,
4991 @var{exp2} is not evaluated even if it has side-effects.
4993 This built-in function can return an lvalue if the chosen argument is an
4996 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
4997 type. Similarly, if @var{exp2} is returned, its return type is the same
5004 __builtin_choose_expr ( \
5005 __builtin_types_compatible_p (typeof (x), double), \
5007 __builtin_choose_expr ( \
5008 __builtin_types_compatible_p (typeof (x), float), \
5010 /* @r{The void expression results in a compile-time error} \
5011 @r{when assigning the result to something.} */ \
5015 @emph{Note:} This construct is only available for C. Furthermore, the
5016 unused expression (@var{exp1} or @var{exp2} depending on the value of
5017 @var{const_exp}) may still generate syntax errors. This may change in
5022 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5023 You can use the built-in function @code{__builtin_constant_p} to
5024 determine if a value is known to be constant at compile-time and hence
5025 that GCC can perform constant-folding on expressions involving that
5026 value. The argument of the function is the value to test. The function
5027 returns the integer 1 if the argument is known to be a compile-time
5028 constant and 0 if it is not known to be a compile-time constant. A
5029 return of 0 does not indicate that the value is @emph{not} a constant,
5030 but merely that GCC cannot prove it is a constant with the specified
5031 value of the @option{-O} option.
5033 You would typically use this function in an embedded application where
5034 memory was a critical resource. If you have some complex calculation,
5035 you may want it to be folded if it involves constants, but need to call
5036 a function if it does not. For example:
5039 #define Scale_Value(X) \
5040 (__builtin_constant_p (X) \
5041 ? ((X) * SCALE + OFFSET) : Scale (X))
5044 You may use this built-in function in either a macro or an inline
5045 function. However, if you use it in an inlined function and pass an
5046 argument of the function as the argument to the built-in, GCC will
5047 never return 1 when you call the inline function with a string constant
5048 or compound literal (@pxref{Compound Literals}) and will not return 1
5049 when you pass a constant numeric value to the inline function unless you
5050 specify the @option{-O} option.
5052 You may also use @code{__builtin_constant_p} in initializers for static
5053 data. For instance, you can write
5056 static const int table[] = @{
5057 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5063 This is an acceptable initializer even if @var{EXPRESSION} is not a
5064 constant expression. GCC must be more conservative about evaluating the
5065 built-in in this case, because it has no opportunity to perform
5068 Previous versions of GCC did not accept this built-in in data
5069 initializers. The earliest version where it is completely safe is
5073 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5074 @opindex fprofile-arcs
5075 You may use @code{__builtin_expect} to provide the compiler with
5076 branch prediction information. In general, you should prefer to
5077 use actual profile feedback for this (@option{-fprofile-arcs}), as
5078 programmers are notoriously bad at predicting how their programs
5079 actually perform. However, there are applications in which this
5080 data is hard to collect.
5082 The return value is the value of @var{exp}, which should be an
5083 integral expression. The value of @var{c} must be a compile-time
5084 constant. The semantics of the built-in are that it is expected
5085 that @var{exp} == @var{c}. For example:
5088 if (__builtin_expect (x, 0))
5093 would indicate that we do not expect to call @code{foo}, since
5094 we expect @code{x} to be zero. Since you are limited to integral
5095 expressions for @var{exp}, you should use constructions such as
5098 if (__builtin_expect (ptr != NULL, 1))
5103 when testing pointer or floating-point values.
5106 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5107 This function is used to minimize cache-miss latency by moving data into
5108 a cache before it is accessed.
5109 You can insert calls to @code{__builtin_prefetch} into code for which
5110 you know addresses of data in memory that is likely to be accessed soon.
5111 If the target supports them, data prefetch instructions will be generated.
5112 If the prefetch is done early enough before the access then the data will
5113 be in the cache by the time it is accessed.
5115 The value of @var{addr} is the address of the memory to prefetch.
5116 There are two optional arguments, @var{rw} and @var{locality}.
5117 The value of @var{rw} is a compile-time constant one or zero; one
5118 means that the prefetch is preparing for a write to the memory address
5119 and zero, the default, means that the prefetch is preparing for a read.
5120 The value @var{locality} must be a compile-time constant integer between
5121 zero and three. A value of zero means that the data has no temporal
5122 locality, so it need not be left in the cache after the access. A value
5123 of three means that the data has a high degree of temporal locality and
5124 should be left in all levels of cache possible. Values of one and two
5125 mean, respectively, a low or moderate degree of temporal locality. The
5129 for (i = 0; i < n; i++)
5132 __builtin_prefetch (&a[i+j], 1, 1);
5133 __builtin_prefetch (&b[i+j], 0, 1);
5138 Data prefetch does not generate faults if @var{addr} is invalid, but
5139 the address expression itself must be valid. For example, a prefetch
5140 of @code{p->next} will not fault if @code{p->next} is not a valid
5141 address, but evaluation will fault if @code{p} is not a valid address.
5143 If the target does not support data prefetch, the address expression
5144 is evaluated if it includes side effects but no other code is generated
5145 and GCC does not issue a warning.
5148 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5149 Returns a positive infinity, if supported by the floating-point format,
5150 else @code{DBL_MAX}. This function is suitable for implementing the
5151 ISO C macro @code{HUGE_VAL}.
5154 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5155 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5158 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5159 Similar to @code{__builtin_huge_val}, except the return
5160 type is @code{long double}.
5163 @deftypefn {Built-in Function} double __builtin_inf (void)
5164 Similar to @code{__builtin_huge_val}, except a warning is generated
5165 if the target floating-point format does not support infinities.
5166 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5169 @deftypefn {Built-in Function} float __builtin_inff (void)
5170 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5173 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5174 Similar to @code{__builtin_inf}, except the return
5175 type is @code{long double}.
5178 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5179 This is an implementation of the ISO C99 function @code{nan}.
5181 Since ISO C99 defines this function in terms of @code{strtod}, which we
5182 do not implement, a description of the parsing is in order. The string
5183 is parsed as by @code{strtol}; that is, the base is recognized by
5184 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5185 in the significand such that the least significant bit of the number
5186 is at the least significant bit of the significand. The number is
5187 truncated to fit the significand field provided. The significand is
5188 forced to be a quiet NaN.
5190 This function, if given a string literal, is evaluated early enough
5191 that it is considered a compile-time constant.
5194 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5195 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5198 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5199 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5202 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5203 Similar to @code{__builtin_nan}, except the significand is forced
5204 to be a signaling NaN. The @code{nans} function is proposed by
5205 @uref{http://std.dkuug.dk/JTC1/SC22/WG14/www/docs/n965.htm,,WG14 N965}.
5208 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5209 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5212 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5213 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5216 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5217 Returns one plus the index of the least significant 1-bit of @var{x}, or
5218 if @var{x} is zero, returns zero.
5221 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5222 Returns the number of leading 0-bits in @var{x}, starting at the most
5223 significant bit position. If @var{x} is 0, the result is undefined.
5226 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5227 Returns the number of trailing 0-bits in @var{x}, starting at the least
5228 significant bit position. If @var{x} is 0, the result is undefined.
5231 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5232 Returns the number of 1-bits in @var{x}.
5235 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5236 Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5240 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5241 Similar to @code{__builtin_ffs}, except the argument type is
5242 @code{unsigned long}.
5245 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5246 Similar to @code{__builtin_clz}, except the argument type is
5247 @code{unsigned long}.
5250 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5251 Similar to @code{__builtin_ctz}, except the argument type is
5252 @code{unsigned long}.
5255 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5256 Similar to @code{__builtin_popcount}, except the argument type is
5257 @code{unsigned long}.
5260 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5261 Similar to @code{__builtin_parity}, except the argument type is
5262 @code{unsigned long}.
5265 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5266 Similar to @code{__builtin_ffs}, except the argument type is
5267 @code{unsigned long long}.
5270 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5271 Similar to @code{__builtin_clz}, except the argument type is
5272 @code{unsigned long long}.
5275 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5276 Similar to @code{__builtin_ctz}, except the argument type is
5277 @code{unsigned long long}.
5280 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5281 Similar to @code{__builtin_popcount}, except the argument type is
5282 @code{unsigned long long}.
5285 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5286 Similar to @code{__builtin_parity}, except the argument type is
5287 @code{unsigned long long}.
5291 @node Target Builtins
5292 @section Built-in Functions Specific to Particular Target Machines
5294 On some target machines, GCC supports many built-in functions specific
5295 to those machines. Generally these generate calls to specific machine
5296 instructions, but allow the compiler to schedule those calls.
5299 * Alpha Built-in Functions::
5300 * ARM Built-in Functions::
5301 * X86 Built-in Functions::
5302 * PowerPC AltiVec Built-in Functions::
5305 @node Alpha Built-in Functions
5306 @subsection Alpha Built-in Functions
5308 These built-in functions are available for the Alpha family of
5309 processors, depending on the command-line switches used.
5311 The following built-in functions are always available. They
5312 all generate the machine instruction that is part of the name.
5315 long __builtin_alpha_implver (void)
5316 long __builtin_alpha_rpcc (void)
5317 long __builtin_alpha_amask (long)
5318 long __builtin_alpha_cmpbge (long, long)
5319 long __builtin_alpha_extbl (long, long)
5320 long __builtin_alpha_extwl (long, long)
5321 long __builtin_alpha_extll (long, long)
5322 long __builtin_alpha_extql (long, long)
5323 long __builtin_alpha_extwh (long, long)
5324 long __builtin_alpha_extlh (long, long)
5325 long __builtin_alpha_extqh (long, long)
5326 long __builtin_alpha_insbl (long, long)
5327 long __builtin_alpha_inswl (long, long)
5328 long __builtin_alpha_insll (long, long)
5329 long __builtin_alpha_insql (long, long)
5330 long __builtin_alpha_inswh (long, long)
5331 long __builtin_alpha_inslh (long, long)
5332 long __builtin_alpha_insqh (long, long)
5333 long __builtin_alpha_mskbl (long, long)
5334 long __builtin_alpha_mskwl (long, long)
5335 long __builtin_alpha_mskll (long, long)
5336 long __builtin_alpha_mskql (long, long)
5337 long __builtin_alpha_mskwh (long, long)
5338 long __builtin_alpha_msklh (long, long)
5339 long __builtin_alpha_mskqh (long, long)
5340 long __builtin_alpha_umulh (long, long)
5341 long __builtin_alpha_zap (long, long)
5342 long __builtin_alpha_zapnot (long, long)
5345 The following built-in functions are always with @option{-mmax}
5346 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5347 later. They all generate the machine instruction that is part
5351 long __builtin_alpha_pklb (long)
5352 long __builtin_alpha_pkwb (long)
5353 long __builtin_alpha_unpkbl (long)
5354 long __builtin_alpha_unpkbw (long)
5355 long __builtin_alpha_minub8 (long, long)
5356 long __builtin_alpha_minsb8 (long, long)
5357 long __builtin_alpha_minuw4 (long, long)
5358 long __builtin_alpha_minsw4 (long, long)
5359 long __builtin_alpha_maxub8 (long, long)
5360 long __builtin_alpha_maxsb8 (long, long)
5361 long __builtin_alpha_maxuw4 (long, long)
5362 long __builtin_alpha_maxsw4 (long, long)
5363 long __builtin_alpha_perr (long, long)
5366 The following built-in functions are always with @option{-mcix}
5367 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5368 later. They all generate the machine instruction that is part
5372 long __builtin_alpha_cttz (long)
5373 long __builtin_alpha_ctlz (long)
5374 long __builtin_alpha_ctpop (long)
5377 The following builtins are available on systems that use the OSF/1
5378 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5379 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5380 @code{rdval} and @code{wrval}.
5383 void *__builtin_thread_pointer (void)
5384 void __builtin_set_thread_pointer (void *)
5387 @node ARM Built-in Functions
5388 @subsection ARM Built-in Functions
5390 These built-in functions are available for the ARM family of
5391 processors, when the @option{-mcpu=iwmmxt} switch is used:
5394 typedef int __v2si __attribute__ ((__mode__ (__V2SI__)))
5396 v2si __builtin_arm_waddw (v2si, v2si)
5397 v2si __builtin_arm_waddw (v2si, v2si)
5398 v2si __builtin_arm_wsubw (v2si, v2si)
5399 v2si __builtin_arm_wsubw (v2si, v2si)
5400 v2si __builtin_arm_waddwss (v2si, v2si)
5401 v2si __builtin_arm_wsubwss (v2si, v2si)
5402 v2si __builtin_arm_wsubwss (v2si, v2si)
5403 v2si __builtin_arm_wsubwss (v2si, v2si)
5404 v2si __builtin_arm_wsubwss (v2si, v2si)
5405 v2si __builtin_arm_waddwus (v2si, v2si)
5406 v2si __builtin_arm_wsubwus (v2si, v2si)
5407 v2si __builtin_arm_wsubwus (v2si, v2si)
5408 v2si __builtin_arm_wmaxuw (v2si, v2si)
5409 v2si __builtin_arm_wmaxsw (v2si, v2si)
5410 v2si __builtin_arm_wavg2br (v2si, v2si)
5411 v2si __builtin_arm_wavg2hr (v2si, v2si)
5412 v2si __builtin_arm_wavg2b (v2si, v2si)
5413 v2si __builtin_arm_wavg2h (v2si, v2si)
5414 v2si __builtin_arm_waccb (v2si)
5415 v2si __builtin_arm_wacch (v2si)
5416 v2si __builtin_arm_waccw (v2si)
5417 v2si __builtin_arm_wmacs (v2si, v2si, v2si)
5418 v2si __builtin_arm_wmacsz (v2si, v2si, v2si)
5419 v2si __builtin_arm_wmacu (v2si, v2si, v2si)
5420 v2si __builtin_arm_wmacuz (v2si, v2si)
5421 v2si __builtin_arm_wsadb (v2si, v2si)
5422 v2si __builtin_arm_wsadbz (v2si, v2si)
5423 v2si __builtin_arm_wsadh (v2si, v2si)
5424 v2si __builtin_arm_wsadhz (v2si, v2si)
5425 v2si __builtin_arm_walign (v2si, v2si)
5426 v2si __builtin_arm_tmia (v2si, int, int)
5427 v2si __builtin_arm_tmiaph (v2si, int, int)
5428 v2si __builtin_arm_tmiabb (v2si, int, int)
5429 v2si __builtin_arm_tmiabt (v2si, int, int)
5430 v2si __builtin_arm_tmiatb (v2si, int, int)
5431 v2si __builtin_arm_tmiatt (v2si, int, int)
5432 int __builtin_arm_tmovmskb (v2si)
5433 int __builtin_arm_tmovmskh (v2si)
5434 int __builtin_arm_tmovmskw (v2si)
5435 v2si __builtin_arm_wmadds (v2si, v2si)
5436 v2si __builtin_arm_wmaddu (v2si, v2si)
5437 v2si __builtin_arm_wpackhss (v2si, v2si)
5438 v2si __builtin_arm_wpackwss (v2si, v2si)
5439 v2si __builtin_arm_wpackdss (v2si, v2si)
5440 v2si __builtin_arm_wpackhus (v2si, v2si)
5441 v2si __builtin_arm_wpackwus (v2si, v2si)
5442 v2si __builtin_arm_wpackdus (v2si, v2si)
5443 v2si __builtin_arm_waddb (v2si, v2si)
5444 v2si __builtin_arm_waddh (v2si, v2si)
5445 v2si __builtin_arm_waddw (v2si, v2si)
5446 v2si __builtin_arm_waddbss (v2si, v2si)
5447 v2si __builtin_arm_waddhss (v2si, v2si)
5448 v2si __builtin_arm_waddwss (v2si, v2si)
5449 v2si __builtin_arm_waddbus (v2si, v2si)
5450 v2si __builtin_arm_waddhus (v2si, v2si)
5451 v2si __builtin_arm_waddwus (v2si, v2si)
5452 v2si __builtin_arm_wsubb (v2si, v2si)
5453 v2si __builtin_arm_wsubh (v2si, v2si)
5454 v2si __builtin_arm_wsubw (v2si, v2si)
5455 v2si __builtin_arm_wsubbss (v2si, v2si)
5456 v2si __builtin_arm_wsubhss (v2si, v2si)
5457 v2si __builtin_arm_wsubwss (v2si, v2si)
5458 v2si __builtin_arm_wsubbus (v2si, v2si)
5459 v2si __builtin_arm_wsubhus (v2si, v2si)
5460 v2si __builtin_arm_wsubwus (v2si, v2si)
5461 v2si __builtin_arm_wand (v2si, v2si)
5462 v2si __builtin_arm_wandn (v2si, v2si)
5463 v2si __builtin_arm_wor (v2si, v2si)
5464 v2si __builtin_arm_wxor (v2si, v2si)
5465 v2si __builtin_arm_wcmpeqb (v2si, v2si)
5466 v2si __builtin_arm_wcmpeqh (v2si, v2si)
5467 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5468 v2si __builtin_arm_wcmpgtub (v2si, v2si)
5469 v2si __builtin_arm_wcmpgtuh (v2si, v2si)
5470 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5471 v2si __builtin_arm_wcmpgtsb (v2si, v2si)
5472 v2si __builtin_arm_wcmpgtsh (v2si, v2si)
5473 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5474 int __builtin_arm_textrmsb (v2si, int)
5475 int __builtin_arm_textrmsh (v2si, int)
5476 int __builtin_arm_textrmsw (v2si, int)
5477 int __builtin_arm_textrmub (v2si, int)
5478 int __builtin_arm_textrmuh (v2si, int)
5479 int __builtin_arm_textrmuw (v2si, int)
5480 v2si __builtin_arm_tinsrb (v2si, int, int)
5481 v2si __builtin_arm_tinsrh (v2si, int, int)
5482 v2si __builtin_arm_tinsrw (v2si, int, int)
5483 v2si __builtin_arm_wmaxsw (v2si, v2si)
5484 v2si __builtin_arm_wmaxsh (v2si, v2si)
5485 v2si __builtin_arm_wmaxsb (v2si, v2si)
5486 v2si __builtin_arm_wmaxuw (v2si, v2si)
5487 v2si __builtin_arm_wmaxuh (v2si, v2si)
5488 v2si __builtin_arm_wmaxub (v2si, v2si)
5489 v2si __builtin_arm_wminsw (v2si, v2si)
5490 v2si __builtin_arm_wminsh (v2si, v2si)
5491 v2si __builtin_arm_wminsb (v2si, v2si)
5492 v2si __builtin_arm_wminuw (v2si, v2si)
5493 v2si __builtin_arm_wminuh (v2si, v2si)
5494 v2si __builtin_arm_wminub (v2si, v2si)
5495 v2si __builtin_arm_wmuluh (v2si, v2si)
5496 v2si __builtin_arm_wmulsh (v2si, v2si)
5497 v2si __builtin_arm_wmulul (v2si, v2si)
5498 v2si __builtin_arm_wshufh (v2si, int)
5499 v2si __builtin_arm_wsllh (v2si, v2si)
5500 v2si __builtin_arm_wsllw (v2si, v2si)
5501 v2si __builtin_arm_wslld (v2si, v2si)
5502 v2si __builtin_arm_wsrah (v2si, v2si)
5503 v2si __builtin_arm_wsraw (v2si, v2si)
5504 v2si __builtin_arm_wsrad (v2si, v2si)
5505 v2si __builtin_arm_wsrlh (v2si, v2si)
5506 v2si __builtin_arm_wsrlw (v2si, v2si)
5507 v2si __builtin_arm_wsrld (v2si, v2si)
5508 v2si __builtin_arm_wrorh (v2si, v2si)
5509 v2si __builtin_arm_wrorw (v2si, v2si)
5510 v2si __builtin_arm_wrord (v2si, v2si)
5511 v2si __builtin_arm_wsllhi (v2si, int)
5512 v2si __builtin_arm_wsllwi (v2si, int)
5513 v2si __builtin_arm_wslldi (v2si, v2si)
5514 v2si __builtin_arm_wsrahi (v2si, int)
5515 v2si __builtin_arm_wsrawi (v2si, int)
5516 v2si __builtin_arm_wsradi (v2si, v2si)
5517 v2si __builtin_arm_wsrlwi (v2si, int)
5518 v2si __builtin_arm_wsrldi (v2si, int)
5519 v2si __builtin_arm_wrorhi (v2si, int)
5520 v2si __builtin_arm_wrorwi (v2si, int)
5521 v2si __builtin_arm_wrordi (v2si, int)
5522 v2si __builtin_arm_wunpckihb (v2si, v2si)
5523 v2si __builtin_arm_wunpckihh (v2si, v2si)
5524 v2si __builtin_arm_wunpckihw (v2si, v2si)
5525 v2si __builtin_arm_wunpckilb (v2si, v2si)
5526 v2si __builtin_arm_wunpckilh (v2si, v2si)
5527 v2si __builtin_arm_wunpckilw (v2si, v2si)
5528 v2si __builtin_arm_wunpckehsb (v2si)
5529 v2si __builtin_arm_wunpckehsh (v2si)
5530 v2si __builtin_arm_wunpckehsw (v2si)
5531 v2si __builtin_arm_wunpckehub (v2si)
5532 v2si __builtin_arm_wunpckehuh (v2si)
5533 v2si __builtin_arm_wunpckehuw (v2si)
5534 v2si __builtin_arm_wunpckelsb (v2si)
5535 v2si __builtin_arm_wunpckelsh (v2si)
5536 v2si __builtin_arm_wunpckelsw (v2si)
5537 v2si __builtin_arm_wunpckelub (v2si)
5538 v2si __builtin_arm_wunpckeluh (v2si)
5539 v2si __builtin_arm_wunpckeluw (v2si)
5540 v2si __builtin_arm_wsubwss (v2si, v2si)
5541 v2si __builtin_arm_wsraw (v2si, v2si)
5542 v2si __builtin_arm_wsrad (v2si, v2si)
5545 @node X86 Built-in Functions
5546 @subsection X86 Built-in Functions
5548 These built-in functions are available for the i386 and x86-64 family
5549 of computers, depending on the command-line switches used.
5551 The following machine modes are available for use with MMX built-in functions
5552 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5553 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5554 vector of eight 8-bit integers. Some of the built-in functions operate on
5555 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5557 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5558 of two 32-bit floating point values.
5560 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5561 floating point values. Some instructions use a vector of four 32-bit
5562 integers, these use @code{V4SI}. Finally, some instructions operate on an
5563 entire vector register, interpreting it as a 128-bit integer, these use mode
5566 The following built-in functions are made available by @option{-mmmx}.
5567 All of them generate the machine instruction that is part of the name.
5570 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5571 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5572 v2si __builtin_ia32_paddd (v2si, v2si)
5573 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5574 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5575 v2si __builtin_ia32_psubd (v2si, v2si)
5576 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5577 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5578 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5579 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5580 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5581 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5582 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5583 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5584 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5585 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5586 di __builtin_ia32_pand (di, di)
5587 di __builtin_ia32_pandn (di,di)
5588 di __builtin_ia32_por (di, di)
5589 di __builtin_ia32_pxor (di, di)
5590 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5591 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5592 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5593 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5594 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5595 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5596 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5597 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5598 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5599 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5600 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5601 v2si __builtin_ia32_punpckldq (v2si, v2si)
5602 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5603 v4hi __builtin_ia32_packssdw (v2si, v2si)
5604 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5607 The following built-in functions are made available either with
5608 @option{-msse}, or with a combination of @option{-m3dnow} and
5609 @option{-march=athlon}. All of them generate the machine
5610 instruction that is part of the name.
5613 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5614 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5615 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5616 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5617 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5618 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5619 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5620 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5621 int __builtin_ia32_pextrw (v4hi, int)
5622 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5623 int __builtin_ia32_pmovmskb (v8qi)
5624 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5625 void __builtin_ia32_movntq (di *, di)
5626 void __builtin_ia32_sfence (void)
5629 The following built-in functions are available when @option{-msse} is used.
5630 All of them generate the machine instruction that is part of the name.
5633 int __builtin_ia32_comieq (v4sf, v4sf)
5634 int __builtin_ia32_comineq (v4sf, v4sf)
5635 int __builtin_ia32_comilt (v4sf, v4sf)
5636 int __builtin_ia32_comile (v4sf, v4sf)
5637 int __builtin_ia32_comigt (v4sf, v4sf)
5638 int __builtin_ia32_comige (v4sf, v4sf)
5639 int __builtin_ia32_ucomieq (v4sf, v4sf)
5640 int __builtin_ia32_ucomineq (v4sf, v4sf)
5641 int __builtin_ia32_ucomilt (v4sf, v4sf)
5642 int __builtin_ia32_ucomile (v4sf, v4sf)
5643 int __builtin_ia32_ucomigt (v4sf, v4sf)
5644 int __builtin_ia32_ucomige (v4sf, v4sf)
5645 v4sf __builtin_ia32_addps (v4sf, v4sf)
5646 v4sf __builtin_ia32_subps (v4sf, v4sf)
5647 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5648 v4sf __builtin_ia32_divps (v4sf, v4sf)
5649 v4sf __builtin_ia32_addss (v4sf, v4sf)
5650 v4sf __builtin_ia32_subss (v4sf, v4sf)
5651 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5652 v4sf __builtin_ia32_divss (v4sf, v4sf)
5653 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5654 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5655 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5656 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5657 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5658 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5659 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5660 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5661 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5662 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5663 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5664 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5665 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5666 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5667 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5668 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5669 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5670 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
5671 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
5672 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
5673 v4sf __builtin_ia32_maxps (v4sf, v4sf)
5674 v4sf __builtin_ia32_maxss (v4sf, v4sf)
5675 v4sf __builtin_ia32_minps (v4sf, v4sf)
5676 v4sf __builtin_ia32_minss (v4sf, v4sf)
5677 v4sf __builtin_ia32_andps (v4sf, v4sf)
5678 v4sf __builtin_ia32_andnps (v4sf, v4sf)
5679 v4sf __builtin_ia32_orps (v4sf, v4sf)
5680 v4sf __builtin_ia32_xorps (v4sf, v4sf)
5681 v4sf __builtin_ia32_movss (v4sf, v4sf)
5682 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
5683 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
5684 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
5685 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
5686 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
5687 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
5688 v2si __builtin_ia32_cvtps2pi (v4sf)
5689 int __builtin_ia32_cvtss2si (v4sf)
5690 v2si __builtin_ia32_cvttps2pi (v4sf)
5691 int __builtin_ia32_cvttss2si (v4sf)
5692 v4sf __builtin_ia32_rcpps (v4sf)
5693 v4sf __builtin_ia32_rsqrtps (v4sf)
5694 v4sf __builtin_ia32_sqrtps (v4sf)
5695 v4sf __builtin_ia32_rcpss (v4sf)
5696 v4sf __builtin_ia32_rsqrtss (v4sf)
5697 v4sf __builtin_ia32_sqrtss (v4sf)
5698 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
5699 void __builtin_ia32_movntps (float *, v4sf)
5700 int __builtin_ia32_movmskps (v4sf)
5703 The following built-in functions are available when @option{-msse} is used.
5706 @item v4sf __builtin_ia32_loadaps (float *)
5707 Generates the @code{movaps} machine instruction as a load from memory.
5708 @item void __builtin_ia32_storeaps (float *, v4sf)
5709 Generates the @code{movaps} machine instruction as a store to memory.
5710 @item v4sf __builtin_ia32_loadups (float *)
5711 Generates the @code{movups} machine instruction as a load from memory.
5712 @item void __builtin_ia32_storeups (float *, v4sf)
5713 Generates the @code{movups} machine instruction as a store to memory.
5714 @item v4sf __builtin_ia32_loadsss (float *)
5715 Generates the @code{movss} machine instruction as a load from memory.
5716 @item void __builtin_ia32_storess (float *, v4sf)
5717 Generates the @code{movss} machine instruction as a store to memory.
5718 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
5719 Generates the @code{movhps} machine instruction as a load from memory.
5720 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
5721 Generates the @code{movlps} machine instruction as a load from memory
5722 @item void __builtin_ia32_storehps (v4sf, v2si *)
5723 Generates the @code{movhps} machine instruction as a store to memory.
5724 @item void __builtin_ia32_storelps (v4sf, v2si *)
5725 Generates the @code{movlps} machine instruction as a store to memory.
5728 The following built-in functions are available when @option{-mpni} is used.
5729 All of them generate the machine instruction that is part of the name.
5732 v2df __builtin_ia32_addsubpd (v2df, v2df)
5733 v2df __builtin_ia32_addsubps (v2df, v2df)
5734 v2df __builtin_ia32_haddpd (v2df, v2df)
5735 v2df __builtin_ia32_haddps (v2df, v2df)
5736 v2df __builtin_ia32_hsubpd (v2df, v2df)
5737 v2df __builtin_ia32_hsubps (v2df, v2df)
5738 v16qi __builtin_ia32_lddqu (char const *)
5739 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
5740 v2df __builtin_ia32_movddup (v2df)
5741 v4sf __builtin_ia32_movshdup (v4sf)
5742 v4sf __builtin_ia32_movsldup (v4sf)
5743 void __builtin_ia32_mwait (unsigned int, unsigned int)
5746 The following built-in functions are available when @option{-mpni} is used.
5749 @item v2df __builtin_ia32_loadddup (double const *)
5750 Generates the @code{movddup} machine instruction as a load from memory.
5753 The following built-in functions are available when @option{-m3dnow} is used.
5754 All of them generate the machine instruction that is part of the name.
5757 void __builtin_ia32_femms (void)
5758 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
5759 v2si __builtin_ia32_pf2id (v2sf)
5760 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
5761 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
5762 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
5763 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
5764 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
5765 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
5766 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
5767 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
5768 v2sf __builtin_ia32_pfrcp (v2sf)
5769 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
5770 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
5771 v2sf __builtin_ia32_pfrsqrt (v2sf)
5772 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
5773 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
5774 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
5775 v2sf __builtin_ia32_pi2fd (v2si)
5776 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
5779 The following built-in functions are available when both @option{-m3dnow}
5780 and @option{-march=athlon} are used. All of them generate the machine
5781 instruction that is part of the name.
5784 v2si __builtin_ia32_pf2iw (v2sf)
5785 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
5786 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
5787 v2sf __builtin_ia32_pi2fw (v2si)
5788 v2sf __builtin_ia32_pswapdsf (v2sf)
5789 v2si __builtin_ia32_pswapdsi (v2si)
5792 @node PowerPC AltiVec Built-in Functions
5793 @subsection PowerPC AltiVec Built-in Functions
5795 These built-in functions are available for the PowerPC family
5796 of computers, depending on the command-line switches used.
5798 The following machine modes are available for use with AltiVec built-in
5799 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
5800 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
5801 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
5802 @code{V16QI} for a vector of sixteen 8-bit integers.
5804 The following functions are made available by including
5805 @code{<altivec.h>} and using @option{-maltivec} and
5806 @option{-mabi=altivec}. The functions implement the functionality
5807 described in Motorola's AltiVec Programming Interface Manual.
5809 There are a few differences from Motorola's documentation and GCC's
5810 implementation. Vector constants are done with curly braces (not
5811 parentheses). Vector initializers require no casts if the vector
5812 constant is of the same type as the variable it is initializing. The
5813 @code{vector bool} type is deprecated and will be discontinued in
5814 further revisions. Use @code{vector signed} instead. If @code{signed}
5815 or @code{unsigned} is omitted, the vector type will default to
5816 @code{signed}. Lastly, all overloaded functions are implemented with macros
5817 for the C implementation. So code the following example will not work:
5820 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
5823 Since vec_add is a macro, the vector constant in the above example will
5824 be treated as four different arguments. Wrap the entire argument in
5825 parentheses for this to work. The C++ implementation does not use
5828 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
5829 Internally, GCC uses built-in functions to achieve the functionality in
5830 the aforementioned header file, but they are not supported and are
5831 subject to change without notice.
5834 vector signed char vec_abs (vector signed char, vector signed char);
5835 vector signed short vec_abs (vector signed short, vector signed short);
5836 vector signed int vec_abs (vector signed int, vector signed int);
5837 vector signed float vec_abs (vector signed float, vector signed float);
5839 vector signed char vec_abss (vector signed char, vector signed char);
5840 vector signed short vec_abss (vector signed short, vector signed short);
5842 vector signed char vec_add (vector signed char, vector signed char);
5843 vector unsigned char vec_add (vector signed char, vector unsigned char);
5845 vector unsigned char vec_add (vector unsigned char, vector signed char);
5847 vector unsigned char vec_add (vector unsigned char,
5848 vector unsigned char);
5849 vector signed short vec_add (vector signed short, vector signed short);
5850 vector unsigned short vec_add (vector signed short,
5851 vector unsigned short);
5852 vector unsigned short vec_add (vector unsigned short,
5853 vector signed short);
5854 vector unsigned short vec_add (vector unsigned short,
5855 vector unsigned short);
5856 vector signed int vec_add (vector signed int, vector signed int);
5857 vector unsigned int vec_add (vector signed int, vector unsigned int);
5858 vector unsigned int vec_add (vector unsigned int, vector signed int);
5859 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
5860 vector float vec_add (vector float, vector float);
5862 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
5864 vector unsigned char vec_adds (vector signed char,
5865 vector unsigned char);
5866 vector unsigned char vec_adds (vector unsigned char,
5867 vector signed char);
5868 vector unsigned char vec_adds (vector unsigned char,
5869 vector unsigned char);
5870 vector signed char vec_adds (vector signed char, vector signed char);
5871 vector unsigned short vec_adds (vector signed short,
5872 vector unsigned short);
5873 vector unsigned short vec_adds (vector unsigned short,
5874 vector signed short);
5875 vector unsigned short vec_adds (vector unsigned short,
5876 vector unsigned short);
5877 vector signed short vec_adds (vector signed short, vector signed short);
5879 vector unsigned int vec_adds (vector signed int, vector unsigned int);
5880 vector unsigned int vec_adds (vector unsigned int, vector signed int);
5881 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
5883 vector signed int vec_adds (vector signed int, vector signed int);
5885 vector float vec_and (vector float, vector float);
5886 vector float vec_and (vector float, vector signed int);
5887 vector float vec_and (vector signed int, vector float);
5888 vector signed int vec_and (vector signed int, vector signed int);
5889 vector unsigned int vec_and (vector signed int, vector unsigned int);
5890 vector unsigned int vec_and (vector unsigned int, vector signed int);
5891 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
5892 vector signed short vec_and (vector signed short, vector signed short);
5893 vector unsigned short vec_and (vector signed short,
5894 vector unsigned short);
5895 vector unsigned short vec_and (vector unsigned short,
5896 vector signed short);
5897 vector unsigned short vec_and (vector unsigned short,
5898 vector unsigned short);
5899 vector signed char vec_and (vector signed char, vector signed char);
5900 vector unsigned char vec_and (vector signed char, vector unsigned char);
5902 vector unsigned char vec_and (vector unsigned char, vector signed char);
5904 vector unsigned char vec_and (vector unsigned char,
5905 vector unsigned char);
5907 vector float vec_andc (vector float, vector float);
5908 vector float vec_andc (vector float, vector signed int);
5909 vector float vec_andc (vector signed int, vector float);
5910 vector signed int vec_andc (vector signed int, vector signed int);
5911 vector unsigned int vec_andc (vector signed int, vector unsigned int);
5912 vector unsigned int vec_andc (vector unsigned int, vector signed int);
5913 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
5915 vector signed short vec_andc (vector signed short, vector signed short);
5917 vector unsigned short vec_andc (vector signed short,
5918 vector unsigned short);
5919 vector unsigned short vec_andc (vector unsigned short,
5920 vector signed short);
5921 vector unsigned short vec_andc (vector unsigned short,
5922 vector unsigned short);
5923 vector signed char vec_andc (vector signed char, vector signed char);
5924 vector unsigned char vec_andc (vector signed char,
5925 vector unsigned char);
5926 vector unsigned char vec_andc (vector unsigned char,
5927 vector signed char);
5928 vector unsigned char vec_andc (vector unsigned char,
5929 vector unsigned char);
5931 vector unsigned char vec_avg (vector unsigned char,
5932 vector unsigned char);
5933 vector signed char vec_avg (vector signed char, vector signed char);
5934 vector unsigned short vec_avg (vector unsigned short,
5935 vector unsigned short);
5936 vector signed short vec_avg (vector signed short, vector signed short);
5937 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
5938 vector signed int vec_avg (vector signed int, vector signed int);
5940 vector float vec_ceil (vector float);
5942 vector signed int vec_cmpb (vector float, vector float);
5944 vector signed char vec_cmpeq (vector signed char, vector signed char);
5945 vector signed char vec_cmpeq (vector unsigned char,
5946 vector unsigned char);
5947 vector signed short vec_cmpeq (vector signed short,
5948 vector signed short);
5949 vector signed short vec_cmpeq (vector unsigned short,
5950 vector unsigned short);
5951 vector signed int vec_cmpeq (vector signed int, vector signed int);
5952 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
5953 vector signed int vec_cmpeq (vector float, vector float);
5955 vector signed int vec_cmpge (vector float, vector float);
5957 vector signed char vec_cmpgt (vector unsigned char,
5958 vector unsigned char);
5959 vector signed char vec_cmpgt (vector signed char, vector signed char);
5960 vector signed short vec_cmpgt (vector unsigned short,
5961 vector unsigned short);
5962 vector signed short vec_cmpgt (vector signed short,
5963 vector signed short);
5964 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
5965 vector signed int vec_cmpgt (vector signed int, vector signed int);
5966 vector signed int vec_cmpgt (vector float, vector float);
5968 vector signed int vec_cmple (vector float, vector float);
5970 vector signed char vec_cmplt (vector unsigned char,
5971 vector unsigned char);
5972 vector signed char vec_cmplt (vector signed char, vector signed char);
5973 vector signed short vec_cmplt (vector unsigned short,
5974 vector unsigned short);
5975 vector signed short vec_cmplt (vector signed short,
5976 vector signed short);
5977 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
5978 vector signed int vec_cmplt (vector signed int, vector signed int);
5979 vector signed int vec_cmplt (vector float, vector float);
5981 vector float vec_ctf (vector unsigned int, const char);
5982 vector float vec_ctf (vector signed int, const char);
5984 vector signed int vec_cts (vector float, const char);
5986 vector unsigned int vec_ctu (vector float, const char);
5988 void vec_dss (const char);
5990 void vec_dssall (void);
5992 void vec_dst (void *, int, const char);
5994 void vec_dstst (void *, int, const char);
5996 void vec_dststt (void *, int, const char);
5998 void vec_dstt (void *, int, const char);
6000 vector float vec_expte (vector float, vector float);
6002 vector float vec_floor (vector float, vector float);
6004 vector float vec_ld (int, vector float *);
6005 vector float vec_ld (int, float *):
6006 vector signed int vec_ld (int, int *);
6007 vector signed int vec_ld (int, vector signed int *);
6008 vector unsigned int vec_ld (int, vector unsigned int *);
6009 vector unsigned int vec_ld (int, unsigned int *);
6010 vector signed short vec_ld (int, short *, vector signed short *);
6011 vector unsigned short vec_ld (int, unsigned short *,
6012 vector unsigned short *);
6013 vector signed char vec_ld (int, signed char *);
6014 vector signed char vec_ld (int, vector signed char *);
6015 vector unsigned char vec_ld (int, unsigned char *);
6016 vector unsigned char vec_ld (int, vector unsigned char *);
6018 vector signed char vec_lde (int, signed char *);
6019 vector unsigned char vec_lde (int, unsigned char *);
6020 vector signed short vec_lde (int, short *);
6021 vector unsigned short vec_lde (int, unsigned short *);
6022 vector float vec_lde (int, float *);
6023 vector signed int vec_lde (int, int *);
6024 vector unsigned int vec_lde (int, unsigned int *);
6026 void float vec_ldl (int, float *);
6027 void float vec_ldl (int, vector float *);
6028 void signed int vec_ldl (int, vector signed int *);
6029 void signed int vec_ldl (int, int *);
6030 void unsigned int vec_ldl (int, unsigned int *);
6031 void unsigned int vec_ldl (int, vector unsigned int *);
6032 void signed short vec_ldl (int, vector signed short *);
6033 void signed short vec_ldl (int, short *);
6034 void unsigned short vec_ldl (int, vector unsigned short *);
6035 void unsigned short vec_ldl (int, unsigned short *);
6036 void signed char vec_ldl (int, vector signed char *);
6037 void signed char vec_ldl (int, signed char *);
6038 void unsigned char vec_ldl (int, vector unsigned char *);
6039 void unsigned char vec_ldl (int, unsigned char *);
6041 vector float vec_loge (vector float);
6043 vector unsigned char vec_lvsl (int, void *, int *);
6045 vector unsigned char vec_lvsr (int, void *, int *);
6047 vector float vec_madd (vector float, vector float, vector float);
6049 vector signed short vec_madds (vector signed short, vector signed short,
6050 vector signed short);
6052 vector unsigned char vec_max (vector signed char, vector unsigned char);
6054 vector unsigned char vec_max (vector unsigned char, vector signed char);
6056 vector unsigned char vec_max (vector unsigned char,
6057 vector unsigned char);
6058 vector signed char vec_max (vector signed char, vector signed char);
6059 vector unsigned short vec_max (vector signed short,
6060 vector unsigned short);
6061 vector unsigned short vec_max (vector unsigned short,
6062 vector signed short);
6063 vector unsigned short vec_max (vector unsigned short,
6064 vector unsigned short);
6065 vector signed short vec_max (vector signed short, vector signed short);
6066 vector unsigned int vec_max (vector signed int, vector unsigned int);
6067 vector unsigned int vec_max (vector unsigned int, vector signed int);
6068 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
6069 vector signed int vec_max (vector signed int, vector signed int);
6070 vector float vec_max (vector float, vector float);
6072 vector signed char vec_mergeh (vector signed char, vector signed char);
6073 vector unsigned char vec_mergeh (vector unsigned char,
6074 vector unsigned char);
6075 vector signed short vec_mergeh (vector signed short,
6076 vector signed short);
6077 vector unsigned short vec_mergeh (vector unsigned short,
6078 vector unsigned short);
6079 vector float vec_mergeh (vector float, vector float);
6080 vector signed int vec_mergeh (vector signed int, vector signed int);
6081 vector unsigned int vec_mergeh (vector unsigned int,
6082 vector unsigned int);
6084 vector signed char vec_mergel (vector signed char, vector signed char);
6085 vector unsigned char vec_mergel (vector unsigned char,
6086 vector unsigned char);
6087 vector signed short vec_mergel (vector signed short,
6088 vector signed short);
6089 vector unsigned short vec_mergel (vector unsigned short,
6090 vector unsigned short);
6091 vector float vec_mergel (vector float, vector float);
6092 vector signed int vec_mergel (vector signed int, vector signed int);
6093 vector unsigned int vec_mergel (vector unsigned int,
6094 vector unsigned int);
6096 vector unsigned short vec_mfvscr (void);
6098 vector unsigned char vec_min (vector signed char, vector unsigned char);
6100 vector unsigned char vec_min (vector unsigned char, vector signed char);
6102 vector unsigned char vec_min (vector unsigned char,
6103 vector unsigned char);
6104 vector signed char vec_min (vector signed char, vector signed char);
6105 vector unsigned short vec_min (vector signed short,
6106 vector unsigned short);
6107 vector unsigned short vec_min (vector unsigned short,
6108 vector signed short);
6109 vector unsigned short vec_min (vector unsigned short,
6110 vector unsigned short);
6111 vector signed short vec_min (vector signed short, vector signed short);
6112 vector unsigned int vec_min (vector signed int, vector unsigned int);
6113 vector unsigned int vec_min (vector unsigned int, vector signed int);
6114 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
6115 vector signed int vec_min (vector signed int, vector signed int);
6116 vector float vec_min (vector float, vector float);
6118 vector signed short vec_mladd (vector signed short, vector signed short,
6119 vector signed short);
6120 vector signed short vec_mladd (vector signed short,
6121 vector unsigned short,
6122 vector unsigned short);
6123 vector signed short vec_mladd (vector unsigned short,
6124 vector signed short,
6125 vector signed short);
6126 vector unsigned short vec_mladd (vector unsigned short,
6127 vector unsigned short,
6128 vector unsigned short);
6130 vector signed short vec_mradds (vector signed short,
6131 vector signed short,
6132 vector signed short);
6134 vector unsigned int vec_msum (vector unsigned char,
6135 vector unsigned char,
6136 vector unsigned int);
6137 vector signed int vec_msum (vector signed char, vector unsigned char,
6139 vector unsigned int vec_msum (vector unsigned short,
6140 vector unsigned short,
6141 vector unsigned int);
6142 vector signed int vec_msum (vector signed short, vector signed short,
6145 vector unsigned int vec_msums (vector unsigned short,
6146 vector unsigned short,
6147 vector unsigned int);
6148 vector signed int vec_msums (vector signed short, vector signed short,
6151 void vec_mtvscr (vector signed int);
6152 void vec_mtvscr (vector unsigned int);
6153 void vec_mtvscr (vector signed short);
6154 void vec_mtvscr (vector unsigned short);
6155 void vec_mtvscr (vector signed char);
6156 void vec_mtvscr (vector unsigned char);
6158 vector unsigned short vec_mule (vector unsigned char,
6159 vector unsigned char);
6160 vector signed short vec_mule (vector signed char, vector signed char);
6161 vector unsigned int vec_mule (vector unsigned short,
6162 vector unsigned short);
6163 vector signed int vec_mule (vector signed short, vector signed short);
6165 vector unsigned short vec_mulo (vector unsigned char,
6166 vector unsigned char);
6167 vector signed short vec_mulo (vector signed char, vector signed char);
6168 vector unsigned int vec_mulo (vector unsigned short,
6169 vector unsigned short);
6170 vector signed int vec_mulo (vector signed short, vector signed short);
6172 vector float vec_nmsub (vector float, vector float, vector float);
6174 vector float vec_nor (vector float, vector float);
6175 vector signed int vec_nor (vector signed int, vector signed int);
6176 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
6177 vector signed short vec_nor (vector signed short, vector signed short);
6178 vector unsigned short vec_nor (vector unsigned short,
6179 vector unsigned short);
6180 vector signed char vec_nor (vector signed char, vector signed char);
6181 vector unsigned char vec_nor (vector unsigned char,
6182 vector unsigned char);
6184 vector float vec_or (vector float, vector float);
6185 vector float vec_or (vector float, vector signed int);
6186 vector float vec_or (vector signed int, vector float);
6187 vector signed int vec_or (vector signed int, vector signed int);
6188 vector unsigned int vec_or (vector signed int, vector unsigned int);
6189 vector unsigned int vec_or (vector unsigned int, vector signed int);
6190 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
6191 vector signed short vec_or (vector signed short, vector signed short);
6192 vector unsigned short vec_or (vector signed short,
6193 vector unsigned short);
6194 vector unsigned short vec_or (vector unsigned short,
6195 vector signed short);
6196 vector unsigned short vec_or (vector unsigned short,
6197 vector unsigned short);
6198 vector signed char vec_or (vector signed char, vector signed char);
6199 vector unsigned char vec_or (vector signed char, vector unsigned char);
6200 vector unsigned char vec_or (vector unsigned char, vector signed char);
6201 vector unsigned char vec_or (vector unsigned char,
6202 vector unsigned char);
6204 vector signed char vec_pack (vector signed short, vector signed short);
6205 vector unsigned char vec_pack (vector unsigned short,
6206 vector unsigned short);
6207 vector signed short vec_pack (vector signed int, vector signed int);
6208 vector unsigned short vec_pack (vector unsigned int,
6209 vector unsigned int);
6211 vector signed short vec_packpx (vector unsigned int,
6212 vector unsigned int);
6214 vector unsigned char vec_packs (vector unsigned short,
6215 vector unsigned short);
6216 vector signed char vec_packs (vector signed short, vector signed short);
6218 vector unsigned short vec_packs (vector unsigned int,
6219 vector unsigned int);
6220 vector signed short vec_packs (vector signed int, vector signed int);
6222 vector unsigned char vec_packsu (vector unsigned short,
6223 vector unsigned short);
6224 vector unsigned char vec_packsu (vector signed short,
6225 vector signed short);
6226 vector unsigned short vec_packsu (vector unsigned int,
6227 vector unsigned int);
6228 vector unsigned short vec_packsu (vector signed int, vector signed int);
6230 vector float vec_perm (vector float, vector float,
6231 vector unsigned char);
6232 vector signed int vec_perm (vector signed int, vector signed int,
6233 vector unsigned char);
6234 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
6235 vector unsigned char);
6236 vector signed short vec_perm (vector signed short, vector signed short,
6237 vector unsigned char);
6238 vector unsigned short vec_perm (vector unsigned short,
6239 vector unsigned short,
6240 vector unsigned char);
6241 vector signed char vec_perm (vector signed char, vector signed char,
6242 vector unsigned char);
6243 vector unsigned char vec_perm (vector unsigned char,
6244 vector unsigned char,
6245 vector unsigned char);
6247 vector float vec_re (vector float);
6249 vector signed char vec_rl (vector signed char, vector unsigned char);
6250 vector unsigned char vec_rl (vector unsigned char,
6251 vector unsigned char);
6252 vector signed short vec_rl (vector signed short, vector unsigned short);
6254 vector unsigned short vec_rl (vector unsigned short,
6255 vector unsigned short);
6256 vector signed int vec_rl (vector signed int, vector unsigned int);
6257 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
6259 vector float vec_round (vector float);
6261 vector float vec_rsqrte (vector float);
6263 vector float vec_sel (vector float, vector float, vector signed int);
6264 vector float vec_sel (vector float, vector float, vector unsigned int);
6265 vector signed int vec_sel (vector signed int, vector signed int,
6267 vector signed int vec_sel (vector signed int, vector signed int,
6268 vector unsigned int);
6269 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6271 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6272 vector unsigned int);
6273 vector signed short vec_sel (vector signed short, vector signed short,
6274 vector signed short);
6275 vector signed short vec_sel (vector signed short, vector signed short,
6276 vector unsigned short);
6277 vector unsigned short vec_sel (vector unsigned short,
6278 vector unsigned short,
6279 vector signed short);
6280 vector unsigned short vec_sel (vector unsigned short,
6281 vector unsigned short,
6282 vector unsigned short);
6283 vector signed char vec_sel (vector signed char, vector signed char,
6284 vector signed char);
6285 vector signed char vec_sel (vector signed char, vector signed char,
6286 vector unsigned char);
6287 vector unsigned char vec_sel (vector unsigned char,
6288 vector unsigned char,
6289 vector signed char);
6290 vector unsigned char vec_sel (vector unsigned char,
6291 vector unsigned char,
6292 vector unsigned char);
6294 vector signed char vec_sl (vector signed char, vector unsigned char);
6295 vector unsigned char vec_sl (vector unsigned char,
6296 vector unsigned char);
6297 vector signed short vec_sl (vector signed short, vector unsigned short);
6299 vector unsigned short vec_sl (vector unsigned short,
6300 vector unsigned short);
6301 vector signed int vec_sl (vector signed int, vector unsigned int);
6302 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
6304 vector float vec_sld (vector float, vector float, const char);
6305 vector signed int vec_sld (vector signed int, vector signed int,
6307 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
6309 vector signed short vec_sld (vector signed short, vector signed short,
6311 vector unsigned short vec_sld (vector unsigned short,
6312 vector unsigned short, const char);
6313 vector signed char vec_sld (vector signed char, vector signed char,
6315 vector unsigned char vec_sld (vector unsigned char,
6316 vector unsigned char,
6319 vector signed int vec_sll (vector signed int, vector unsigned int);
6320 vector signed int vec_sll (vector signed int, vector unsigned short);
6321 vector signed int vec_sll (vector signed int, vector unsigned char);
6322 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
6323 vector unsigned int vec_sll (vector unsigned int,
6324 vector unsigned short);
6325 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
6327 vector signed short vec_sll (vector signed short, vector unsigned int);
6328 vector signed short vec_sll (vector signed short,
6329 vector unsigned short);
6330 vector signed short vec_sll (vector signed short, vector unsigned char);
6332 vector unsigned short vec_sll (vector unsigned short,
6333 vector unsigned int);
6334 vector unsigned short vec_sll (vector unsigned short,
6335 vector unsigned short);
6336 vector unsigned short vec_sll (vector unsigned short,
6337 vector unsigned char);
6338 vector signed char vec_sll (vector signed char, vector unsigned int);
6339 vector signed char vec_sll (vector signed char, vector unsigned short);
6340 vector signed char vec_sll (vector signed char, vector unsigned char);
6341 vector unsigned char vec_sll (vector unsigned char,
6342 vector unsigned int);
6343 vector unsigned char vec_sll (vector unsigned char,
6344 vector unsigned short);
6345 vector unsigned char vec_sll (vector unsigned char,
6346 vector unsigned char);
6348 vector float vec_slo (vector float, vector signed char);
6349 vector float vec_slo (vector float, vector unsigned char);
6350 vector signed int vec_slo (vector signed int, vector signed char);
6351 vector signed int vec_slo (vector signed int, vector unsigned char);
6352 vector unsigned int vec_slo (vector unsigned int, vector signed char);
6353 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
6355 vector signed short vec_slo (vector signed short, vector signed char);
6356 vector signed short vec_slo (vector signed short, vector unsigned char);
6358 vector unsigned short vec_slo (vector unsigned short,
6359 vector signed char);
6360 vector unsigned short vec_slo (vector unsigned short,
6361 vector unsigned char);
6362 vector signed char vec_slo (vector signed char, vector signed char);
6363 vector signed char vec_slo (vector signed char, vector unsigned char);
6364 vector unsigned char vec_slo (vector unsigned char, vector signed char);
6366 vector unsigned char vec_slo (vector unsigned char,
6367 vector unsigned char);
6369 vector signed char vec_splat (vector signed char, const char);
6370 vector unsigned char vec_splat (vector unsigned char, const char);
6371 vector signed short vec_splat (vector signed short, const char);
6372 vector unsigned short vec_splat (vector unsigned short, const char);
6373 vector float vec_splat (vector float, const char);
6374 vector signed int vec_splat (vector signed int, const char);
6375 vector unsigned int vec_splat (vector unsigned int, const char);
6377 vector signed char vec_splat_s8 (const char);
6379 vector signed short vec_splat_s16 (const char);
6381 vector signed int vec_splat_s32 (const char);
6383 vector unsigned char vec_splat_u8 (const char);
6385 vector unsigned short vec_splat_u16 (const char);
6387 vector unsigned int vec_splat_u32 (const char);
6389 vector signed char vec_sr (vector signed char, vector unsigned char);
6390 vector unsigned char vec_sr (vector unsigned char,
6391 vector unsigned char);
6392 vector signed short vec_sr (vector signed short, vector unsigned short);
6394 vector unsigned short vec_sr (vector unsigned short,
6395 vector unsigned short);
6396 vector signed int vec_sr (vector signed int, vector unsigned int);
6397 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
6399 vector signed char vec_sra (vector signed char, vector unsigned char);
6400 vector unsigned char vec_sra (vector unsigned char,
6401 vector unsigned char);
6402 vector signed short vec_sra (vector signed short,
6403 vector unsigned short);
6404 vector unsigned short vec_sra (vector unsigned short,
6405 vector unsigned short);
6406 vector signed int vec_sra (vector signed int, vector unsigned int);
6407 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
6409 vector signed int vec_srl (vector signed int, vector unsigned int);
6410 vector signed int vec_srl (vector signed int, vector unsigned short);
6411 vector signed int vec_srl (vector signed int, vector unsigned char);
6412 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
6413 vector unsigned int vec_srl (vector unsigned int,
6414 vector unsigned short);
6415 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
6417 vector signed short vec_srl (vector signed short, vector unsigned int);
6418 vector signed short vec_srl (vector signed short,
6419 vector unsigned short);
6420 vector signed short vec_srl (vector signed short, vector unsigned char);
6422 vector unsigned short vec_srl (vector unsigned short,
6423 vector unsigned int);
6424 vector unsigned short vec_srl (vector unsigned short,
6425 vector unsigned short);
6426 vector unsigned short vec_srl (vector unsigned short,
6427 vector unsigned char);
6428 vector signed char vec_srl (vector signed char, vector unsigned int);
6429 vector signed char vec_srl (vector signed char, vector unsigned short);
6430 vector signed char vec_srl (vector signed char, vector unsigned char);
6431 vector unsigned char vec_srl (vector unsigned char,
6432 vector unsigned int);
6433 vector unsigned char vec_srl (vector unsigned char,
6434 vector unsigned short);
6435 vector unsigned char vec_srl (vector unsigned char,
6436 vector unsigned char);
6438 vector float vec_sro (vector float, vector signed char);
6439 vector float vec_sro (vector float, vector unsigned char);
6440 vector signed int vec_sro (vector signed int, vector signed char);
6441 vector signed int vec_sro (vector signed int, vector unsigned char);
6442 vector unsigned int vec_sro (vector unsigned int, vector signed char);
6443 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
6445 vector signed short vec_sro (vector signed short, vector signed char);
6446 vector signed short vec_sro (vector signed short, vector unsigned char);
6448 vector unsigned short vec_sro (vector unsigned short,
6449 vector signed char);
6450 vector unsigned short vec_sro (vector unsigned short,
6451 vector unsigned char);
6452 vector signed char vec_sro (vector signed char, vector signed char);
6453 vector signed char vec_sro (vector signed char, vector unsigned char);
6454 vector unsigned char vec_sro (vector unsigned char, vector signed char);
6456 vector unsigned char vec_sro (vector unsigned char,
6457 vector unsigned char);
6459 void vec_st (vector float, int, float *);
6460 void vec_st (vector float, int, vector float *);
6461 void vec_st (vector signed int, int, int *);
6462 void vec_st (vector signed int, int, unsigned int *);
6463 void vec_st (vector unsigned int, int, unsigned int *);
6464 void vec_st (vector unsigned int, int, vector unsigned int *);
6465 void vec_st (vector signed short, int, short *);
6466 void vec_st (vector signed short, int, vector unsigned short *);
6467 void vec_st (vector signed short, int, vector signed short *);
6468 void vec_st (vector unsigned short, int, unsigned short *);
6469 void vec_st (vector unsigned short, int, vector unsigned short *);
6470 void vec_st (vector signed char, int, signed char *);
6471 void vec_st (vector signed char, int, unsigned char *);
6472 void vec_st (vector signed char, int, vector signed char *);
6473 void vec_st (vector unsigned char, int, unsigned char *);
6474 void vec_st (vector unsigned char, int, vector unsigned char *);
6476 void vec_ste (vector signed char, int, unsigned char *);
6477 void vec_ste (vector signed char, int, signed char *);
6478 void vec_ste (vector unsigned char, int, unsigned char *);
6479 void vec_ste (vector signed short, int, short *);
6480 void vec_ste (vector signed short, int, unsigned short *);
6481 void vec_ste (vector unsigned short, int, void *);
6482 void vec_ste (vector signed int, int, unsigned int *);
6483 void vec_ste (vector signed int, int, int *);
6484 void vec_ste (vector unsigned int, int, unsigned int *);
6485 void vec_ste (vector float, int, float *);
6487 void vec_stl (vector float, int, vector float *);
6488 void vec_stl (vector float, int, float *);
6489 void vec_stl (vector signed int, int, vector signed int *);
6490 void vec_stl (vector signed int, int, int *);
6491 void vec_stl (vector signed int, int, unsigned int *);
6492 void vec_stl (vector unsigned int, int, vector unsigned int *);
6493 void vec_stl (vector unsigned int, int, unsigned int *);
6494 void vec_stl (vector signed short, int, short *);
6495 void vec_stl (vector signed short, int, unsigned short *);
6496 void vec_stl (vector signed short, int, vector signed short *);
6497 void vec_stl (vector unsigned short, int, unsigned short *);
6498 void vec_stl (vector unsigned short, int, vector signed short *);
6499 void vec_stl (vector signed char, int, signed char *);
6500 void vec_stl (vector signed char, int, unsigned char *);
6501 void vec_stl (vector signed char, int, vector signed char *);
6502 void vec_stl (vector unsigned char, int, unsigned char *);
6503 void vec_stl (vector unsigned char, int, vector unsigned char *);
6505 vector signed char vec_sub (vector signed char, vector signed char);
6506 vector unsigned char vec_sub (vector signed char, vector unsigned char);
6508 vector unsigned char vec_sub (vector unsigned char, vector signed char);
6510 vector unsigned char vec_sub (vector unsigned char,
6511 vector unsigned char);
6512 vector signed short vec_sub (vector signed short, vector signed short);
6513 vector unsigned short vec_sub (vector signed short,
6514 vector unsigned short);
6515 vector unsigned short vec_sub (vector unsigned short,
6516 vector signed short);
6517 vector unsigned short vec_sub (vector unsigned short,
6518 vector unsigned short);
6519 vector signed int vec_sub (vector signed int, vector signed int);
6520 vector unsigned int vec_sub (vector signed int, vector unsigned int);
6521 vector unsigned int vec_sub (vector unsigned int, vector signed int);
6522 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
6523 vector float vec_sub (vector float, vector float);
6525 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
6527 vector unsigned char vec_subs (vector signed char,
6528 vector unsigned char);
6529 vector unsigned char vec_subs (vector unsigned char,
6530 vector signed char);
6531 vector unsigned char vec_subs (vector unsigned char,
6532 vector unsigned char);
6533 vector signed char vec_subs (vector signed char, vector signed char);
6534 vector unsigned short vec_subs (vector signed short,
6535 vector unsigned short);
6536 vector unsigned short vec_subs (vector unsigned short,
6537 vector signed short);
6538 vector unsigned short vec_subs (vector unsigned short,
6539 vector unsigned short);
6540 vector signed short vec_subs (vector signed short, vector signed short);
6542 vector unsigned int vec_subs (vector signed int, vector unsigned int);
6543 vector unsigned int vec_subs (vector unsigned int, vector signed int);
6544 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
6546 vector signed int vec_subs (vector signed int, vector signed int);
6548 vector unsigned int vec_sum4s (vector unsigned char,
6549 vector unsigned int);
6550 vector signed int vec_sum4s (vector signed char, vector signed int);
6551 vector signed int vec_sum4s (vector signed short, vector signed int);
6553 vector signed int vec_sum2s (vector signed int, vector signed int);
6555 vector signed int vec_sums (vector signed int, vector signed int);
6557 vector float vec_trunc (vector float);
6559 vector signed short vec_unpackh (vector signed char);
6560 vector unsigned int vec_unpackh (vector signed short);
6561 vector signed int vec_unpackh (vector signed short);
6563 vector signed short vec_unpackl (vector signed char);
6564 vector unsigned int vec_unpackl (vector signed short);
6565 vector signed int vec_unpackl (vector signed short);
6567 vector float vec_xor (vector float, vector float);
6568 vector float vec_xor (vector float, vector signed int);
6569 vector float vec_xor (vector signed int, vector float);
6570 vector signed int vec_xor (vector signed int, vector signed int);
6571 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6572 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6573 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6574 vector signed short vec_xor (vector signed short, vector signed short);
6575 vector unsigned short vec_xor (vector signed short,
6576 vector unsigned short);
6577 vector unsigned short vec_xor (vector unsigned short,
6578 vector signed short);
6579 vector unsigned short vec_xor (vector unsigned short,
6580 vector unsigned short);
6581 vector signed char vec_xor (vector signed char, vector signed char);
6582 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6584 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6586 vector unsigned char vec_xor (vector unsigned char,
6587 vector unsigned char);
6589 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6591 vector signed int vec_all_eq (vector signed char, vector signed char);
6592 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6594 vector signed int vec_all_eq (vector unsigned char,
6595 vector unsigned char);
6596 vector signed int vec_all_eq (vector signed short,
6597 vector unsigned short);
6598 vector signed int vec_all_eq (vector signed short, vector signed short);
6600 vector signed int vec_all_eq (vector unsigned short,
6601 vector signed short);
6602 vector signed int vec_all_eq (vector unsigned short,
6603 vector unsigned short);
6604 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6605 vector signed int vec_all_eq (vector signed int, vector signed int);
6606 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6607 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6609 vector signed int vec_all_eq (vector float, vector float);
6611 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6613 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6615 vector signed int vec_all_ge (vector unsigned char,
6616 vector unsigned char);
6617 vector signed int vec_all_ge (vector signed char, vector signed char);
6618 vector signed int vec_all_ge (vector signed short,
6619 vector unsigned short);
6620 vector signed int vec_all_ge (vector unsigned short,
6621 vector signed short);
6622 vector signed int vec_all_ge (vector unsigned short,
6623 vector unsigned short);
6624 vector signed int vec_all_ge (vector signed short, vector signed short);
6626 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6627 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6628 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6630 vector signed int vec_all_ge (vector signed int, vector signed int);
6631 vector signed int vec_all_ge (vector float, vector float);
6633 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6635 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6637 vector signed int vec_all_gt (vector unsigned char,
6638 vector unsigned char);
6639 vector signed int vec_all_gt (vector signed char, vector signed char);
6640 vector signed int vec_all_gt (vector signed short,
6641 vector unsigned short);
6642 vector signed int vec_all_gt (vector unsigned short,
6643 vector signed short);
6644 vector signed int vec_all_gt (vector unsigned short,
6645 vector unsigned short);
6646 vector signed int vec_all_gt (vector signed short, vector signed short);
6648 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6649 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6650 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6652 vector signed int vec_all_gt (vector signed int, vector signed int);
6653 vector signed int vec_all_gt (vector float, vector float);
6655 vector signed int vec_all_in (vector float, vector float);
6657 vector signed int vec_all_le (vector signed char, vector unsigned char);
6659 vector signed int vec_all_le (vector unsigned char, vector signed char);
6661 vector signed int vec_all_le (vector unsigned char,
6662 vector unsigned char);
6663 vector signed int vec_all_le (vector signed char, vector signed char);
6664 vector signed int vec_all_le (vector signed short,
6665 vector unsigned short);
6666 vector signed int vec_all_le (vector unsigned short,
6667 vector signed short);
6668 vector signed int vec_all_le (vector unsigned short,
6669 vector unsigned short);
6670 vector signed int vec_all_le (vector signed short, vector signed short);
6672 vector signed int vec_all_le (vector signed int, vector unsigned int);
6673 vector signed int vec_all_le (vector unsigned int, vector signed int);
6674 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
6676 vector signed int vec_all_le (vector signed int, vector signed int);
6677 vector signed int vec_all_le (vector float, vector float);
6679 vector signed int vec_all_lt (vector signed char, vector unsigned char);
6681 vector signed int vec_all_lt (vector unsigned char, vector signed char);
6683 vector signed int vec_all_lt (vector unsigned char,
6684 vector unsigned char);
6685 vector signed int vec_all_lt (vector signed char, vector signed char);
6686 vector signed int vec_all_lt (vector signed short,
6687 vector unsigned short);
6688 vector signed int vec_all_lt (vector unsigned short,
6689 vector signed short);
6690 vector signed int vec_all_lt (vector unsigned short,
6691 vector unsigned short);
6692 vector signed int vec_all_lt (vector signed short, vector signed short);
6694 vector signed int vec_all_lt (vector signed int, vector unsigned int);
6695 vector signed int vec_all_lt (vector unsigned int, vector signed int);
6696 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
6698 vector signed int vec_all_lt (vector signed int, vector signed int);
6699 vector signed int vec_all_lt (vector float, vector float);
6701 vector signed int vec_all_nan (vector float);
6703 vector signed int vec_all_ne (vector signed char, vector unsigned char);
6705 vector signed int vec_all_ne (vector signed char, vector signed char);
6706 vector signed int vec_all_ne (vector unsigned char, vector signed char);
6708 vector signed int vec_all_ne (vector unsigned char,
6709 vector unsigned char);
6710 vector signed int vec_all_ne (vector signed short,
6711 vector unsigned short);
6712 vector signed int vec_all_ne (vector signed short, vector signed short);
6714 vector signed int vec_all_ne (vector unsigned short,
6715 vector signed short);
6716 vector signed int vec_all_ne (vector unsigned short,
6717 vector unsigned short);
6718 vector signed int vec_all_ne (vector signed int, vector unsigned int);
6719 vector signed int vec_all_ne (vector signed int, vector signed int);
6720 vector signed int vec_all_ne (vector unsigned int, vector signed int);
6721 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
6723 vector signed int vec_all_ne (vector float, vector float);
6725 vector signed int vec_all_nge (vector float, vector float);
6727 vector signed int vec_all_ngt (vector float, vector float);
6729 vector signed int vec_all_nle (vector float, vector float);
6731 vector signed int vec_all_nlt (vector float, vector float);
6733 vector signed int vec_all_numeric (vector float);
6735 vector signed int vec_any_eq (vector signed char, vector unsigned char);
6737 vector signed int vec_any_eq (vector signed char, vector signed char);
6738 vector signed int vec_any_eq (vector unsigned char, vector signed char);
6740 vector signed int vec_any_eq (vector unsigned char,
6741 vector unsigned char);
6742 vector signed int vec_any_eq (vector signed short,
6743 vector unsigned short);
6744 vector signed int vec_any_eq (vector signed short, vector signed short);
6746 vector signed int vec_any_eq (vector unsigned short,
6747 vector signed short);
6748 vector signed int vec_any_eq (vector unsigned short,
6749 vector unsigned short);
6750 vector signed int vec_any_eq (vector signed int, vector unsigned int);
6751 vector signed int vec_any_eq (vector signed int, vector signed int);
6752 vector signed int vec_any_eq (vector unsigned int, vector signed int);
6753 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
6755 vector signed int vec_any_eq (vector float, vector float);
6757 vector signed int vec_any_ge (vector signed char, vector unsigned char);
6759 vector signed int vec_any_ge (vector unsigned char, vector signed char);
6761 vector signed int vec_any_ge (vector unsigned char,
6762 vector unsigned char);
6763 vector signed int vec_any_ge (vector signed char, vector signed char);
6764 vector signed int vec_any_ge (vector signed short,
6765 vector unsigned short);
6766 vector signed int vec_any_ge (vector unsigned short,
6767 vector signed short);
6768 vector signed int vec_any_ge (vector unsigned short,
6769 vector unsigned short);
6770 vector signed int vec_any_ge (vector signed short, vector signed short);
6772 vector signed int vec_any_ge (vector signed int, vector unsigned int);
6773 vector signed int vec_any_ge (vector unsigned int, vector signed int);
6774 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
6776 vector signed int vec_any_ge (vector signed int, vector signed int);
6777 vector signed int vec_any_ge (vector float, vector float);
6779 vector signed int vec_any_gt (vector signed char, vector unsigned char);
6781 vector signed int vec_any_gt (vector unsigned char, vector signed char);
6783 vector signed int vec_any_gt (vector unsigned char,
6784 vector unsigned char);
6785 vector signed int vec_any_gt (vector signed char, vector signed char);
6786 vector signed int vec_any_gt (vector signed short,
6787 vector unsigned short);
6788 vector signed int vec_any_gt (vector unsigned short,
6789 vector signed short);
6790 vector signed int vec_any_gt (vector unsigned short,
6791 vector unsigned short);
6792 vector signed int vec_any_gt (vector signed short, vector signed short);
6794 vector signed int vec_any_gt (vector signed int, vector unsigned int);
6795 vector signed int vec_any_gt (vector unsigned int, vector signed int);
6796 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
6798 vector signed int vec_any_gt (vector signed int, vector signed int);
6799 vector signed int vec_any_gt (vector float, vector float);
6801 vector signed int vec_any_le (vector signed char, vector unsigned char);
6803 vector signed int vec_any_le (vector unsigned char, vector signed char);
6805 vector signed int vec_any_le (vector unsigned char,
6806 vector unsigned char);
6807 vector signed int vec_any_le (vector signed char, vector signed char);
6808 vector signed int vec_any_le (vector signed short,
6809 vector unsigned short);
6810 vector signed int vec_any_le (vector unsigned short,
6811 vector signed short);
6812 vector signed int vec_any_le (vector unsigned short,
6813 vector unsigned short);
6814 vector signed int vec_any_le (vector signed short, vector signed short);
6816 vector signed int vec_any_le (vector signed int, vector unsigned int);
6817 vector signed int vec_any_le (vector unsigned int, vector signed int);
6818 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
6820 vector signed int vec_any_le (vector signed int, vector signed int);
6821 vector signed int vec_any_le (vector float, vector float);
6823 vector signed int vec_any_lt (vector signed char, vector unsigned char);
6825 vector signed int vec_any_lt (vector unsigned char, vector signed char);
6827 vector signed int vec_any_lt (vector unsigned char,
6828 vector unsigned char);
6829 vector signed int vec_any_lt (vector signed char, vector signed char);
6830 vector signed int vec_any_lt (vector signed short,
6831 vector unsigned short);
6832 vector signed int vec_any_lt (vector unsigned short,
6833 vector signed short);
6834 vector signed int vec_any_lt (vector unsigned short,
6835 vector unsigned short);
6836 vector signed int vec_any_lt (vector signed short, vector signed short);
6838 vector signed int vec_any_lt (vector signed int, vector unsigned int);
6839 vector signed int vec_any_lt (vector unsigned int, vector signed int);
6840 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
6842 vector signed int vec_any_lt (vector signed int, vector signed int);
6843 vector signed int vec_any_lt (vector float, vector float);
6845 vector signed int vec_any_nan (vector float);
6847 vector signed int vec_any_ne (vector signed char, vector unsigned char);
6849 vector signed int vec_any_ne (vector signed char, vector signed char);
6850 vector signed int vec_any_ne (vector unsigned char, vector signed char);
6852 vector signed int vec_any_ne (vector unsigned char,
6853 vector unsigned char);
6854 vector signed int vec_any_ne (vector signed short,
6855 vector unsigned short);
6856 vector signed int vec_any_ne (vector signed short, vector signed short);
6858 vector signed int vec_any_ne (vector unsigned short,
6859 vector signed short);
6860 vector signed int vec_any_ne (vector unsigned short,
6861 vector unsigned short);
6862 vector signed int vec_any_ne (vector signed int, vector unsigned int);
6863 vector signed int vec_any_ne (vector signed int, vector signed int);
6864 vector signed int vec_any_ne (vector unsigned int, vector signed int);
6865 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
6867 vector signed int vec_any_ne (vector float, vector float);
6869 vector signed int vec_any_nge (vector float, vector float);
6871 vector signed int vec_any_ngt (vector float, vector float);
6873 vector signed int vec_any_nle (vector float, vector float);
6875 vector signed int vec_any_nlt (vector float, vector float);
6877 vector signed int vec_any_numeric (vector float);
6879 vector signed int vec_any_out (vector float, vector float);
6883 @section Pragmas Accepted by GCC
6887 GCC supports several types of pragmas, primarily in order to compile
6888 code originally written for other compilers. Note that in general
6889 we do not recommend the use of pragmas; @xref{Function Attributes},
6890 for further explanation.
6894 * RS/6000 and PowerPC Pragmas::
6901 @subsection ARM Pragmas
6903 The ARM target defines pragmas for controlling the default addition of
6904 @code{long_call} and @code{short_call} attributes to functions.
6905 @xref{Function Attributes}, for information about the effects of these
6910 @cindex pragma, long_calls
6911 Set all subsequent functions to have the @code{long_call} attribute.
6914 @cindex pragma, no_long_calls
6915 Set all subsequent functions to have the @code{short_call} attribute.
6917 @item long_calls_off
6918 @cindex pragma, long_calls_off
6919 Do not affect the @code{long_call} or @code{short_call} attributes of
6920 subsequent functions.
6923 @node RS/6000 and PowerPC Pragmas
6924 @subsection RS/6000 and PowerPC Pragmas
6926 The RS/6000 and PowerPC targets define one pragma for controlling
6927 whether or not the @code{longcall} attribute is added to function
6928 declarations by default. This pragma overrides the @option{-mlongcall}
6929 option, but not the @code{longcall} and @code{shortcall} attributes.
6930 @xref{RS/6000 and PowerPC Options}, for more information about when long
6931 calls are and are not necessary.
6935 @cindex pragma, longcall
6936 Apply the @code{longcall} attribute to all subsequent function
6940 Do not apply the @code{longcall} attribute to subsequent function
6944 @c Describe c4x pragmas here.
6945 @c Describe h8300 pragmas here.
6946 @c Describe i370 pragmas here.
6947 @c Describe i960 pragmas here.
6948 @c Describe sh pragmas here.
6949 @c Describe v850 pragmas here.
6951 @node Darwin Pragmas
6952 @subsection Darwin Pragmas
6954 The following pragmas are available for all architectures running the
6955 Darwin operating system. These are useful for compatibility with other
6959 @item mark @var{tokens}@dots{}
6960 @cindex pragma, mark
6961 This pragma is accepted, but has no effect.
6963 @item options align=@var{alignment}
6964 @cindex pragma, options align
6965 This pragma sets the alignment of fields in structures. The values of
6966 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
6967 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
6968 properly; to restore the previous setting, use @code{reset} for the
6971 @item segment @var{tokens}@dots{}
6972 @cindex pragma, segment
6973 This pragma is accepted, but has no effect.
6975 @item unused (@var{var} [, @var{var}]@dots{})
6976 @cindex pragma, unused
6977 This pragma declares variables to be possibly unused. GCC will not
6978 produce warnings for the listed variables. The effect is similar to
6979 that of the @code{unused} attribute, except that this pragma may appear
6980 anywhere within the variables' scopes.
6983 @node Solaris Pragmas
6984 @subsection Solaris Pragmas
6986 For compatibility with the SunPRO compiler, the following pragma
6990 @item redefine_extname @var{oldname} @var{newname}
6991 @cindex pragma, redefine_extname
6993 This pragma gives the C function @var{oldname} the assembler label
6994 @var{newname}. The pragma must appear before the function declaration.
6995 This pragma is equivalent to the asm labels extension (@pxref{Asm
6996 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
6997 if the pragma is available.
7001 @subsection Tru64 Pragmas
7003 For compatibility with the Compaq C compiler, the following pragma
7007 @item extern_prefix @var{string}
7008 @cindex pragma, extern_prefix
7010 This pragma renames all subsequent function and variable declarations
7011 such that @var{string} is prepended to the name. This effect may be
7012 terminated by using another @code{extern_prefix} pragma with the
7015 This pragma is similar in intent to to the asm labels extension
7016 (@pxref{Asm Labels}) in that the system programmer wants to change
7017 the assembly-level ABI without changing the source-level API. The
7018 preprocessor defines @code{__PRAGMA_EXTERN_PREFIX} if the pragma is
7022 @node Unnamed Fields
7023 @section Unnamed struct/union fields within structs/unions.
7027 For compatibility with other compilers, GCC allows you to define
7028 a structure or union that contains, as fields, structures and unions
7029 without names. For example:
7042 In this example, the user would be able to access members of the unnamed
7043 union with code like @samp{foo.b}. Note that only unnamed structs and
7044 unions are allowed, you may not have, for example, an unnamed
7047 You must never create such structures that cause ambiguous field definitions.
7048 For example, this structure:
7059 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
7060 Such constructs are not supported and must be avoided. In the future,
7061 such constructs may be detected and treated as compilation errors.
7064 @section Thread-Local Storage
7065 @cindex Thread-Local Storage
7066 @cindex @acronym{TLS}
7069 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
7070 are allocated such that there is one instance of the variable per extant
7071 thread. The run-time model GCC uses to implement this originates
7072 in the IA-64 processor-specific ABI, but has since been migrated
7073 to other processors as well. It requires significant support from
7074 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
7075 system libraries (@file{libc.so} and @file{libpthread.so}), so it
7076 is not available everywhere.
7078 At the user level, the extension is visible with a new storage
7079 class keyword: @code{__thread}. For example:
7083 extern __thread struct state s;
7084 static __thread char *p;
7087 The @code{__thread} specifier may be used alone, with the @code{extern}
7088 or @code{static} specifiers, but with no other storage class specifier.
7089 When used with @code{extern} or @code{static}, @code{__thread} must appear
7090 immediately after the other storage class specifier.
7092 The @code{__thread} specifier may be applied to any global, file-scoped
7093 static, function-scoped static, or static data member of a class. It may
7094 not be applied to block-scoped automatic or non-static data member.
7096 When the address-of operator is applied to a thread-local variable, it is
7097 evaluated at run-time and returns the address of the current thread's
7098 instance of that variable. An address so obtained may be used by any
7099 thread. When a thread terminates, any pointers to thread-local variables
7100 in that thread become invalid.
7102 No static initialization may refer to the address of a thread-local variable.
7104 In C++, if an initializer is present for a thread-local variable, it must
7105 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
7108 See @uref{http://people.redhat.com/drepper/tls.pdf,
7109 ELF Handling For Thread-Local Storage} for a detailed explanation of
7110 the four thread-local storage addressing models, and how the run-time
7111 is expected to function.
7114 * C99 Thread-Local Edits::
7115 * C++98 Thread-Local Edits::
7118 @node C99 Thread-Local Edits
7119 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
7121 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
7122 that document the exact semantics of the language extension.
7126 @cite{5.1.2 Execution environments}
7128 Add new text after paragraph 1
7131 Within either execution environment, a @dfn{thread} is a flow of
7132 control within a program. It is implementation defined whether
7133 or not there may be more than one thread associated with a program.
7134 It is implementation defined how threads beyond the first are
7135 created, the name and type of the function called at thread
7136 startup, and how threads may be terminated. However, objects
7137 with thread storage duration shall be initialized before thread
7142 @cite{6.2.4 Storage durations of objects}
7144 Add new text before paragraph 3
7147 An object whose identifier is declared with the storage-class
7148 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
7149 Its lifetime is the entire execution of the thread, and its
7150 stored value is initialized only once, prior to thread startup.
7154 @cite{6.4.1 Keywords}
7156 Add @code{__thread}.
7159 @cite{6.7.1 Storage-class specifiers}
7161 Add @code{__thread} to the list of storage class specifiers in
7164 Change paragraph 2 to
7167 With the exception of @code{__thread}, at most one storage-class
7168 specifier may be given [@dots{}]. The @code{__thread} specifier may
7169 be used alone, or immediately following @code{extern} or
7173 Add new text after paragraph 6
7176 The declaration of an identifier for a variable that has
7177 block scope that specifies @code{__thread} shall also
7178 specify either @code{extern} or @code{static}.
7180 The @code{__thread} specifier shall be used only with
7185 @node C++98 Thread-Local Edits
7186 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
7188 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
7189 that document the exact semantics of the language extension.
7193 @b{[intro.execution]}
7195 New text after paragraph 4
7198 A @dfn{thread} is a flow of control within the abstract machine.
7199 It is implementation defined whether or not there may be more than
7203 New text after paragraph 7
7206 It is unspecified whether additional action must be taken to
7207 ensure when and whether side effects are visible to other threads.
7213 Add @code{__thread}.
7216 @b{[basic.start.main]}
7218 Add after paragraph 5
7221 The thread that begins execution at the @code{main} function is called
7222 the @dfn{main thread}. It is implementation defined how functions
7223 beginning threads other than the main thread are designated or typed.
7224 A function so designated, as well as the @code{main} function, is called
7225 a @dfn{thread startup function}. It is implementation defined what
7226 happens if a thread startup function returns. It is implementation
7227 defined what happens to other threads when any thread calls @code{exit}.
7231 @b{[basic.start.init]}
7233 Add after paragraph 4
7236 The storage for an object of thread storage duration shall be
7237 statically initialized before the first statement of the thread startup
7238 function. An object of thread storage duration shall not require
7239 dynamic initialization.
7243 @b{[basic.start.term]}
7245 Add after paragraph 3
7248 The type of an object with thread storage duration shall not have a
7249 non-trivial destructor, nor shall it be an array type whose elements
7250 (directly or indirectly) have non-trivial destructors.
7256 Add ``thread storage duration'' to the list in paragraph 1.
7261 Thread, static, and automatic storage durations are associated with
7262 objects introduced by declarations [@dots{}].
7265 Add @code{__thread} to the list of specifiers in paragraph 3.
7268 @b{[basic.stc.thread]}
7270 New section before @b{[basic.stc.static]}
7273 The keyword @code{__thread} applied to a non-local object gives the
7274 object thread storage duration.
7276 A local variable or class data member declared both @code{static}
7277 and @code{__thread} gives the variable or member thread storage
7282 @b{[basic.stc.static]}
7287 All objects which have neither thread storage duration, dynamic
7288 storage duration nor are local [@dots{}].
7294 Add @code{__thread} to the list in paragraph 1.
7299 With the exception of @code{__thread}, at most one
7300 @var{storage-class-specifier} shall appear in a given
7301 @var{decl-specifier-seq}. The @code{__thread} specifier may
7302 be used alone, or immediately following the @code{extern} or
7303 @code{static} specifiers. [@dots{}]
7306 Add after paragraph 5
7309 The @code{__thread} specifier can be applied only to the names of objects
7310 and to anonymous unions.
7316 Add after paragraph 6
7319 Non-@code{static} members shall not be @code{__thread}.
7323 @node C++ Extensions
7324 @chapter Extensions to the C++ Language
7325 @cindex extensions, C++ language
7326 @cindex C++ language extensions
7328 The GNU compiler provides these extensions to the C++ language (and you
7329 can also use most of the C language extensions in your C++ programs). If you
7330 want to write code that checks whether these features are available, you can
7331 test for the GNU compiler the same way as for C programs: check for a
7332 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
7333 test specifically for GNU C++ (@pxref{Standard Predefined,,Standard
7334 Predefined Macros,cpp.info,The C Preprocessor}).
7337 * Min and Max:: C++ Minimum and maximum operators.
7338 * Volatiles:: What constitutes an access to a volatile object.
7339 * Restricted Pointers:: C99 restricted pointers and references.
7340 * Vague Linkage:: Where G++ puts inlines, vtables and such.
7341 * C++ Interface:: You can use a single C++ header file for both
7342 declarations and definitions.
7343 * Template Instantiation:: Methods for ensuring that exactly one copy of
7344 each needed template instantiation is emitted.
7345 * Bound member functions:: You can extract a function pointer to the
7346 method denoted by a @samp{->*} or @samp{.*} expression.
7347 * C++ Attributes:: Variable, function, and type attributes for C++ only.
7348 * Java Exceptions:: Tweaking exception handling to work with Java.
7349 * Deprecated Features:: Things will disappear from g++.
7350 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
7354 @section Minimum and Maximum Operators in C++
7356 It is very convenient to have operators which return the ``minimum'' or the
7357 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
7360 @item @var{a} <? @var{b}
7362 @cindex minimum operator
7363 is the @dfn{minimum}, returning the smaller of the numeric values
7364 @var{a} and @var{b};
7366 @item @var{a} >? @var{b}
7368 @cindex maximum operator
7369 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
7373 These operations are not primitive in ordinary C++, since you can
7374 use a macro to return the minimum of two things in C++, as in the
7378 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
7382 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
7383 the minimum value of variables @var{i} and @var{j}.
7385 However, side effects in @code{X} or @code{Y} may cause unintended
7386 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
7387 the smaller counter twice. The GNU C @code{typeof} extension allows you
7388 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
7389 However, writing @code{MIN} and @code{MAX} as macros also forces you to
7390 use function-call notation for a fundamental arithmetic operation.
7391 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
7394 Since @code{<?} and @code{>?} are built into the compiler, they properly
7395 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
7399 @section When is a Volatile Object Accessed?
7400 @cindex accessing volatiles
7401 @cindex volatile read
7402 @cindex volatile write
7403 @cindex volatile access
7405 Both the C and C++ standard have the concept of volatile objects. These
7406 are normally accessed by pointers and used for accessing hardware. The
7407 standards encourage compilers to refrain from optimizations
7408 concerning accesses to volatile objects that it might perform on
7409 non-volatile objects. The C standard leaves it implementation defined
7410 as to what constitutes a volatile access. The C++ standard omits to
7411 specify this, except to say that C++ should behave in a similar manner
7412 to C with respect to volatiles, where possible. The minimum either
7413 standard specifies is that at a sequence point all previous accesses to
7414 volatile objects have stabilized and no subsequent accesses have
7415 occurred. Thus an implementation is free to reorder and combine
7416 volatile accesses which occur between sequence points, but cannot do so
7417 for accesses across a sequence point. The use of volatiles does not
7418 allow you to violate the restriction on updating objects multiple times
7419 within a sequence point.
7421 In most expressions, it is intuitively obvious what is a read and what is
7422 a write. For instance
7425 volatile int *dst = @var{somevalue};
7426 volatile int *src = @var{someothervalue};
7431 will cause a read of the volatile object pointed to by @var{src} and stores the
7432 value into the volatile object pointed to by @var{dst}. There is no
7433 guarantee that these reads and writes are atomic, especially for objects
7434 larger than @code{int}.
7436 Less obvious expressions are where something which looks like an access
7437 is used in a void context. An example would be,
7440 volatile int *src = @var{somevalue};
7444 With C, such expressions are rvalues, and as rvalues cause a read of
7445 the object, GCC interprets this as a read of the volatile being pointed
7446 to. The C++ standard specifies that such expressions do not undergo
7447 lvalue to rvalue conversion, and that the type of the dereferenced
7448 object may be incomplete. The C++ standard does not specify explicitly
7449 that it is this lvalue to rvalue conversion which is responsible for
7450 causing an access. However, there is reason to believe that it is,
7451 because otherwise certain simple expressions become undefined. However,
7452 because it would surprise most programmers, G++ treats dereferencing a
7453 pointer to volatile object of complete type in a void context as a read
7454 of the object. When the object has incomplete type, G++ issues a
7459 struct T @{int m;@};
7460 volatile S *ptr1 = @var{somevalue};
7461 volatile T *ptr2 = @var{somevalue};
7466 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
7467 causes a read of the object pointed to. If you wish to force an error on
7468 the first case, you must force a conversion to rvalue with, for instance
7469 a static cast, @code{static_cast<S>(*ptr1)}.
7471 When using a reference to volatile, G++ does not treat equivalent
7472 expressions as accesses to volatiles, but instead issues a warning that
7473 no volatile is accessed. The rationale for this is that otherwise it
7474 becomes difficult to determine where volatile access occur, and not
7475 possible to ignore the return value from functions returning volatile
7476 references. Again, if you wish to force a read, cast the reference to
7479 @node Restricted Pointers
7480 @section Restricting Pointer Aliasing
7481 @cindex restricted pointers
7482 @cindex restricted references
7483 @cindex restricted this pointer
7485 As with gcc, g++ understands the C99 feature of restricted pointers,
7486 specified with the @code{__restrict__}, or @code{__restrict} type
7487 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
7488 language flag, @code{restrict} is not a keyword in C++.
7490 In addition to allowing restricted pointers, you can specify restricted
7491 references, which indicate that the reference is not aliased in the local
7495 void fn (int *__restrict__ rptr, int &__restrict__ rref)
7502 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
7503 @var{rref} refers to a (different) unaliased integer.
7505 You may also specify whether a member function's @var{this} pointer is
7506 unaliased by using @code{__restrict__} as a member function qualifier.
7509 void T::fn () __restrict__
7516 Within the body of @code{T::fn}, @var{this} will have the effective
7517 definition @code{T *__restrict__ const this}. Notice that the
7518 interpretation of a @code{__restrict__} member function qualifier is
7519 different to that of @code{const} or @code{volatile} qualifier, in that it
7520 is applied to the pointer rather than the object. This is consistent with
7521 other compilers which implement restricted pointers.
7523 As with all outermost parameter qualifiers, @code{__restrict__} is
7524 ignored in function definition matching. This means you only need to
7525 specify @code{__restrict__} in a function definition, rather than
7526 in a function prototype as well.
7529 @section Vague Linkage
7530 @cindex vague linkage
7532 There are several constructs in C++ which require space in the object
7533 file but are not clearly tied to a single translation unit. We say that
7534 these constructs have ``vague linkage''. Typically such constructs are
7535 emitted wherever they are needed, though sometimes we can be more
7539 @item Inline Functions
7540 Inline functions are typically defined in a header file which can be
7541 included in many different compilations. Hopefully they can usually be
7542 inlined, but sometimes an out-of-line copy is necessary, if the address
7543 of the function is taken or if inlining fails. In general, we emit an
7544 out-of-line copy in all translation units where one is needed. As an
7545 exception, we only emit inline virtual functions with the vtable, since
7546 it will always require a copy.
7548 Local static variables and string constants used in an inline function
7549 are also considered to have vague linkage, since they must be shared
7550 between all inlined and out-of-line instances of the function.
7554 C++ virtual functions are implemented in most compilers using a lookup
7555 table, known as a vtable. The vtable contains pointers to the virtual
7556 functions provided by a class, and each object of the class contains a
7557 pointer to its vtable (or vtables, in some multiple-inheritance
7558 situations). If the class declares any non-inline, non-pure virtual
7559 functions, the first one is chosen as the ``key method'' for the class,
7560 and the vtable is only emitted in the translation unit where the key
7563 @emph{Note:} If the chosen key method is later defined as inline, the
7564 vtable will still be emitted in every translation unit which defines it.
7565 Make sure that any inline virtuals are declared inline in the class
7566 body, even if they are not defined there.
7568 @item type_info objects
7571 C++ requires information about types to be written out in order to
7572 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7573 For polymorphic classes (classes with virtual functions), the type_info
7574 object is written out along with the vtable so that @samp{dynamic_cast}
7575 can determine the dynamic type of a class object at runtime. For all
7576 other types, we write out the type_info object when it is used: when
7577 applying @samp{typeid} to an expression, throwing an object, or
7578 referring to a type in a catch clause or exception specification.
7580 @item Template Instantiations
7581 Most everything in this section also applies to template instantiations,
7582 but there are other options as well.
7583 @xref{Template Instantiation,,Where's the Template?}.
7587 When used with GNU ld version 2.8 or later on an ELF system such as
7588 Linux/GNU or Solaris 2, or on Microsoft Windows, duplicate copies of
7589 these constructs will be discarded at link time. This is known as
7592 On targets that don't support COMDAT, but do support weak symbols, GCC
7593 will use them. This way one copy will override all the others, but
7594 the unused copies will still take up space in the executable.
7596 For targets which do not support either COMDAT or weak symbols,
7597 most entities with vague linkage will be emitted as local symbols to
7598 avoid duplicate definition errors from the linker. This will not happen
7599 for local statics in inlines, however, as having multiple copies will
7600 almost certainly break things.
7602 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7603 another way to control placement of these constructs.
7606 @section Declarations and Definitions in One Header
7608 @cindex interface and implementation headers, C++
7609 @cindex C++ interface and implementation headers
7610 C++ object definitions can be quite complex. In principle, your source
7611 code will need two kinds of things for each object that you use across
7612 more than one source file. First, you need an @dfn{interface}
7613 specification, describing its structure with type declarations and
7614 function prototypes. Second, you need the @dfn{implementation} itself.
7615 It can be tedious to maintain a separate interface description in a
7616 header file, in parallel to the actual implementation. It is also
7617 dangerous, since separate interface and implementation definitions may
7618 not remain parallel.
7620 @cindex pragmas, interface and implementation
7621 With GNU C++, you can use a single header file for both purposes.
7624 @emph{Warning:} The mechanism to specify this is in transition. For the
7625 nonce, you must use one of two @code{#pragma} commands; in a future
7626 release of GNU C++, an alternative mechanism will make these
7627 @code{#pragma} commands unnecessary.
7630 The header file contains the full definitions, but is marked with
7631 @samp{#pragma interface} in the source code. This allows the compiler
7632 to use the header file only as an interface specification when ordinary
7633 source files incorporate it with @code{#include}. In the single source
7634 file where the full implementation belongs, you can use either a naming
7635 convention or @samp{#pragma implementation} to indicate this alternate
7636 use of the header file.
7639 @item #pragma interface
7640 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7641 @kindex #pragma interface
7642 Use this directive in @emph{header files} that define object classes, to save
7643 space in most of the object files that use those classes. Normally,
7644 local copies of certain information (backup copies of inline member
7645 functions, debugging information, and the internal tables that implement
7646 virtual functions) must be kept in each object file that includes class
7647 definitions. You can use this pragma to avoid such duplication. When a
7648 header file containing @samp{#pragma interface} is included in a
7649 compilation, this auxiliary information will not be generated (unless
7650 the main input source file itself uses @samp{#pragma implementation}).
7651 Instead, the object files will contain references to be resolved at link
7654 The second form of this directive is useful for the case where you have
7655 multiple headers with the same name in different directories. If you
7656 use this form, you must specify the same string to @samp{#pragma
7659 @item #pragma implementation
7660 @itemx #pragma implementation "@var{objects}.h"
7661 @kindex #pragma implementation
7662 Use this pragma in a @emph{main input file}, when you want full output from
7663 included header files to be generated (and made globally visible). The
7664 included header file, in turn, should use @samp{#pragma interface}.
7665 Backup copies of inline member functions, debugging information, and the
7666 internal tables used to implement virtual functions are all generated in
7667 implementation files.
7669 @cindex implied @code{#pragma implementation}
7670 @cindex @code{#pragma implementation}, implied
7671 @cindex naming convention, implementation headers
7672 If you use @samp{#pragma implementation} with no argument, it applies to
7673 an include file with the same basename@footnote{A file's @dfn{basename}
7674 was the name stripped of all leading path information and of trailing
7675 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
7676 file. For example, in @file{allclass.cc}, giving just
7677 @samp{#pragma implementation}
7678 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
7680 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
7681 an implementation file whenever you would include it from
7682 @file{allclass.cc} even if you never specified @samp{#pragma
7683 implementation}. This was deemed to be more trouble than it was worth,
7684 however, and disabled.
7686 If you use an explicit @samp{#pragma implementation}, it must appear in
7687 your source file @emph{before} you include the affected header files.
7689 Use the string argument if you want a single implementation file to
7690 include code from multiple header files. (You must also use
7691 @samp{#include} to include the header file; @samp{#pragma
7692 implementation} only specifies how to use the file---it doesn't actually
7695 There is no way to split up the contents of a single header file into
7696 multiple implementation files.
7699 @cindex inlining and C++ pragmas
7700 @cindex C++ pragmas, effect on inlining
7701 @cindex pragmas in C++, effect on inlining
7702 @samp{#pragma implementation} and @samp{#pragma interface} also have an
7703 effect on function inlining.
7705 If you define a class in a header file marked with @samp{#pragma
7706 interface}, the effect on a function defined in that class is similar to
7707 an explicit @code{extern} declaration---the compiler emits no code at
7708 all to define an independent version of the function. Its definition
7709 is used only for inlining with its callers.
7711 @opindex fno-implement-inlines
7712 Conversely, when you include the same header file in a main source file
7713 that declares it as @samp{#pragma implementation}, the compiler emits
7714 code for the function itself; this defines a version of the function
7715 that can be found via pointers (or by callers compiled without
7716 inlining). If all calls to the function can be inlined, you can avoid
7717 emitting the function by compiling with @option{-fno-implement-inlines}.
7718 If any calls were not inlined, you will get linker errors.
7720 @node Template Instantiation
7721 @section Where's the Template?
7722 @cindex template instantiation
7724 C++ templates are the first language feature to require more
7725 intelligence from the environment than one usually finds on a UNIX
7726 system. Somehow the compiler and linker have to make sure that each
7727 template instance occurs exactly once in the executable if it is needed,
7728 and not at all otherwise. There are two basic approaches to this
7729 problem, which I will refer to as the Borland model and the Cfront model.
7733 Borland C++ solved the template instantiation problem by adding the code
7734 equivalent of common blocks to their linker; the compiler emits template
7735 instances in each translation unit that uses them, and the linker
7736 collapses them together. The advantage of this model is that the linker
7737 only has to consider the object files themselves; there is no external
7738 complexity to worry about. This disadvantage is that compilation time
7739 is increased because the template code is being compiled repeatedly.
7740 Code written for this model tends to include definitions of all
7741 templates in the header file, since they must be seen to be
7745 The AT&T C++ translator, Cfront, solved the template instantiation
7746 problem by creating the notion of a template repository, an
7747 automatically maintained place where template instances are stored. A
7748 more modern version of the repository works as follows: As individual
7749 object files are built, the compiler places any template definitions and
7750 instantiations encountered in the repository. At link time, the link
7751 wrapper adds in the objects in the repository and compiles any needed
7752 instances that were not previously emitted. The advantages of this
7753 model are more optimal compilation speed and the ability to use the
7754 system linker; to implement the Borland model a compiler vendor also
7755 needs to replace the linker. The disadvantages are vastly increased
7756 complexity, and thus potential for error; for some code this can be
7757 just as transparent, but in practice it can been very difficult to build
7758 multiple programs in one directory and one program in multiple
7759 directories. Code written for this model tends to separate definitions
7760 of non-inline member templates into a separate file, which should be
7761 compiled separately.
7764 When used with GNU ld version 2.8 or later on an ELF system such as
7765 Linux/GNU or Solaris 2, or on Microsoft Windows, g++ supports the
7766 Borland model. On other systems, g++ implements neither automatic
7769 A future version of g++ will support a hybrid model whereby the compiler
7770 will emit any instantiations for which the template definition is
7771 included in the compile, and store template definitions and
7772 instantiation context information into the object file for the rest.
7773 The link wrapper will extract that information as necessary and invoke
7774 the compiler to produce the remaining instantiations. The linker will
7775 then combine duplicate instantiations.
7777 In the mean time, you have the following options for dealing with
7778 template instantiations:
7783 Compile your template-using code with @option{-frepo}. The compiler will
7784 generate files with the extension @samp{.rpo} listing all of the
7785 template instantiations used in the corresponding object files which
7786 could be instantiated there; the link wrapper, @samp{collect2}, will
7787 then update the @samp{.rpo} files to tell the compiler where to place
7788 those instantiations and rebuild any affected object files. The
7789 link-time overhead is negligible after the first pass, as the compiler
7790 will continue to place the instantiations in the same files.
7792 This is your best option for application code written for the Borland
7793 model, as it will just work. Code written for the Cfront model will
7794 need to be modified so that the template definitions are available at
7795 one or more points of instantiation; usually this is as simple as adding
7796 @code{#include <tmethods.cc>} to the end of each template header.
7798 For library code, if you want the library to provide all of the template
7799 instantiations it needs, just try to link all of its object files
7800 together; the link will fail, but cause the instantiations to be
7801 generated as a side effect. Be warned, however, that this may cause
7802 conflicts if multiple libraries try to provide the same instantiations.
7803 For greater control, use explicit instantiation as described in the next
7807 @opindex fno-implicit-templates
7808 Compile your code with @option{-fno-implicit-templates} to disable the
7809 implicit generation of template instances, and explicitly instantiate
7810 all the ones you use. This approach requires more knowledge of exactly
7811 which instances you need than do the others, but it's less
7812 mysterious and allows greater control. You can scatter the explicit
7813 instantiations throughout your program, perhaps putting them in the
7814 translation units where the instances are used or the translation units
7815 that define the templates themselves; you can put all of the explicit
7816 instantiations you need into one big file; or you can create small files
7823 template class Foo<int>;
7824 template ostream& operator <<
7825 (ostream&, const Foo<int>&);
7828 for each of the instances you need, and create a template instantiation
7831 If you are using Cfront-model code, you can probably get away with not
7832 using @option{-fno-implicit-templates} when compiling files that don't
7833 @samp{#include} the member template definitions.
7835 If you use one big file to do the instantiations, you may want to
7836 compile it without @option{-fno-implicit-templates} so you get all of the
7837 instances required by your explicit instantiations (but not by any
7838 other files) without having to specify them as well.
7840 g++ has extended the template instantiation syntax given in the ISO
7841 standard to allow forward declaration of explicit instantiations
7842 (with @code{extern}), instantiation of the compiler support data for a
7843 template class (i.e.@: the vtable) without instantiating any of its
7844 members (with @code{inline}), and instantiation of only the static data
7845 members of a template class, without the support data or member
7846 functions (with (@code{static}):
7849 extern template int max (int, int);
7850 inline template class Foo<int>;
7851 static template class Foo<int>;
7855 Do nothing. Pretend g++ does implement automatic instantiation
7856 management. Code written for the Borland model will work fine, but
7857 each translation unit will contain instances of each of the templates it
7858 uses. In a large program, this can lead to an unacceptable amount of code
7861 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7862 more discussion of these pragmas.
7865 @node Bound member functions
7866 @section Extracting the function pointer from a bound pointer to member function
7868 @cindex pointer to member function
7869 @cindex bound pointer to member function
7871 In C++, pointer to member functions (PMFs) are implemented using a wide
7872 pointer of sorts to handle all the possible call mechanisms; the PMF
7873 needs to store information about how to adjust the @samp{this} pointer,
7874 and if the function pointed to is virtual, where to find the vtable, and
7875 where in the vtable to look for the member function. If you are using
7876 PMFs in an inner loop, you should really reconsider that decision. If
7877 that is not an option, you can extract the pointer to the function that
7878 would be called for a given object/PMF pair and call it directly inside
7879 the inner loop, to save a bit of time.
7881 Note that you will still be paying the penalty for the call through a
7882 function pointer; on most modern architectures, such a call defeats the
7883 branch prediction features of the CPU@. This is also true of normal
7884 virtual function calls.
7886 The syntax for this extension is
7890 extern int (A::*fp)();
7891 typedef int (*fptr)(A *);
7893 fptr p = (fptr)(a.*fp);
7896 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
7897 no object is needed to obtain the address of the function. They can be
7898 converted to function pointers directly:
7901 fptr p1 = (fptr)(&A::foo);
7904 @opindex Wno-pmf-conversions
7905 You must specify @option{-Wno-pmf-conversions} to use this extension.
7907 @node C++ Attributes
7908 @section C++-Specific Variable, Function, and Type Attributes
7910 Some attributes only make sense for C++ programs.
7913 @item init_priority (@var{priority})
7914 @cindex init_priority attribute
7917 In Standard C++, objects defined at namespace scope are guaranteed to be
7918 initialized in an order in strict accordance with that of their definitions
7919 @emph{in a given translation unit}. No guarantee is made for initializations
7920 across translation units. However, GNU C++ allows users to control the
7921 order of initialization of objects defined at namespace scope with the
7922 @code{init_priority} attribute by specifying a relative @var{priority},
7923 a constant integral expression currently bounded between 101 and 65535
7924 inclusive. Lower numbers indicate a higher priority.
7926 In the following example, @code{A} would normally be created before
7927 @code{B}, but the @code{init_priority} attribute has reversed that order:
7930 Some_Class A __attribute__ ((init_priority (2000)));
7931 Some_Class B __attribute__ ((init_priority (543)));
7935 Note that the particular values of @var{priority} do not matter; only their
7938 @item java_interface
7939 @cindex java_interface attribute
7941 This type attribute informs C++ that the class is a Java interface. It may
7942 only be applied to classes declared within an @code{extern "Java"} block.
7943 Calls to methods declared in this interface will be dispatched using GCJ's
7944 interface table mechanism, instead of regular virtual table dispatch.
7948 @node Java Exceptions
7949 @section Java Exceptions
7951 The Java language uses a slightly different exception handling model
7952 from C++. Normally, GNU C++ will automatically detect when you are
7953 writing C++ code that uses Java exceptions, and handle them
7954 appropriately. However, if C++ code only needs to execute destructors
7955 when Java exceptions are thrown through it, GCC will guess incorrectly.
7956 Sample problematic code is:
7959 struct S @{ ~S(); @};
7960 extern void bar(); // is written in Java, and may throw exceptions
7969 The usual effect of an incorrect guess is a link failure, complaining of
7970 a missing routine called @samp{__gxx_personality_v0}.
7972 You can inform the compiler that Java exceptions are to be used in a
7973 translation unit, irrespective of what it might think, by writing
7974 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
7975 @samp{#pragma} must appear before any functions that throw or catch
7976 exceptions, or run destructors when exceptions are thrown through them.
7978 You cannot mix Java and C++ exceptions in the same translation unit. It
7979 is believed to be safe to throw a C++ exception from one file through
7980 another file compiled for the Java exception model, or vice versa, but
7981 there may be bugs in this area.
7983 @node Deprecated Features
7984 @section Deprecated Features
7986 In the past, the GNU C++ compiler was extended to experiment with new
7987 features, at a time when the C++ language was still evolving. Now that
7988 the C++ standard is complete, some of those features are superseded by
7989 superior alternatives. Using the old features might cause a warning in
7990 some cases that the feature will be dropped in the future. In other
7991 cases, the feature might be gone already.
7993 While the list below is not exhaustive, it documents some of the options
7994 that are now deprecated:
7997 @item -fexternal-templates
7998 @itemx -falt-external-templates
7999 These are two of the many ways for g++ to implement template
8000 instantiation. @xref{Template Instantiation}. The C++ standard clearly
8001 defines how template definitions have to be organized across
8002 implementation units. g++ has an implicit instantiation mechanism that
8003 should work just fine for standard-conforming code.
8005 @item -fstrict-prototype
8006 @itemx -fno-strict-prototype
8007 Previously it was possible to use an empty prototype parameter list to
8008 indicate an unspecified number of parameters (like C), rather than no
8009 parameters, as C++ demands. This feature has been removed, except where
8010 it is required for backwards compatibility @xref{Backwards Compatibility}.
8013 The named return value extension has been deprecated, and is now
8016 The use of initializer lists with new expressions has been deprecated,
8017 and is now removed from g++.
8019 Floating and complex non-type template parameters have been deprecated,
8020 and are now removed from g++.
8022 The implicit typename extension has been deprecated and is now
8025 The use of default arguments in function pointers, function typedefs and
8026 and other places where they are not permitted by the standard is
8027 deprecated and will be removed from a future version of g++.
8029 @node Backwards Compatibility
8030 @section Backwards Compatibility
8031 @cindex Backwards Compatibility
8032 @cindex ARM [Annotated C++ Reference Manual]
8034 Now that there is a definitive ISO standard C++, G++ has a specification
8035 to adhere to. The C++ language evolved over time, and features that
8036 used to be acceptable in previous drafts of the standard, such as the ARM
8037 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
8038 compilation of C++ written to such drafts, G++ contains some backwards
8039 compatibilities. @emph{All such backwards compatibility features are
8040 liable to disappear in future versions of G++.} They should be considered
8041 deprecated @xref{Deprecated Features}.
8045 If a variable is declared at for scope, it used to remain in scope until
8046 the end of the scope which contained the for statement (rather than just
8047 within the for scope). G++ retains this, but issues a warning, if such a
8048 variable is accessed outside the for scope.
8050 @item Implicit C language
8051 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
8052 scope to set the language. On such systems, all header files are
8053 implicitly scoped inside a C language scope. Also, an empty prototype
8054 @code{()} will be treated as an unspecified number of arguments, rather
8055 than no arguments, as C++ demands.