1 @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002,2003,2004
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter C Implementation-defined behavior
8 @cindex implementation-defined behavior, C language
10 A conforming implementation of ISO C is required to document its
11 choice of behavior in each of the areas that are designated
12 ``implementation defined.'' The following lists all such areas,
13 along with the section number from the ISO/IEC 9899:1999 standard.
16 * Translation implementation::
17 * Environment implementation::
18 * Identifiers implementation::
19 * Characters implementation::
20 * Integers implementation::
21 * Floating point implementation::
22 * Arrays and pointers implementation::
23 * Hints implementation::
24 * Structures unions enumerations and bit-fields implementation::
25 * Qualifiers implementation::
26 * Preprocessing directives implementation::
27 * Library functions implementation::
28 * Architecture implementation::
29 * Locale-specific behavior implementation::
32 @node Translation implementation
37 @cite{How a diagnostic is identified (3.10, 5.1.1.3).}
39 Diagnostics consist of all the output sent to stderr by GCC.
42 @cite{Whether each nonempty sequence of white-space characters other than
43 new-line is retained or replaced by one space character in translation
47 @node Environment implementation
50 The behavior of these points are dependent on the implementation
51 of the C library, and are not defined by GCC itself.
53 @node Identifiers implementation
58 @cite{Which additional multibyte characters may appear in identifiers
59 and their correspondence to universal character names (6.4.2).}
62 @cite{The number of significant initial characters in an identifier
65 For internal names, all characters are significant. For external names,
66 the number of significant characters are defined by the linker; for
67 almost all targets, all characters are significant.
71 @node Characters implementation
76 @cite{The number of bits in a byte (3.6).}
79 @cite{The values of the members of the execution character set (5.2.1).}
82 @cite{The unique value of the member of the execution character set produced
83 for each of the standard alphabetic escape sequences (5.2.2).}
86 @cite{The value of a @code{char} object into which has been stored any
87 character other than a member of the basic execution character set (6.2.5).}
90 @cite{Which of @code{signed char} or @code{unsigned char} has the same range,
91 representation, and behavior as ``plain'' @code{char} (6.2.5, 6.3.1.1).}
94 @cite{The mapping of members of the source character set (in character
95 constants and string literals) to members of the execution character
96 set (6.4.4.4, 5.1.1.2).}
99 @cite{The value of an integer character constant containing more than one
100 character or containing a character or escape sequence that does not map
101 to a single-byte execution character (6.4.4.4).}
104 @cite{The value of a wide character constant containing more than one
105 multibyte character, or containing a multibyte character or escape
106 sequence not represented in the extended execution character set (6.4.4.4).}
109 @cite{The current locale used to convert a wide character constant consisting
110 of a single multibyte character that maps to a member of the extended
111 execution character set into a corresponding wide character code (6.4.4.4).}
114 @cite{The current locale used to convert a wide string literal into
115 corresponding wide character codes (6.4.5).}
118 @cite{The value of a string literal containing a multibyte character or escape
119 sequence not represented in the execution character set (6.4.5).}
122 @node Integers implementation
127 @cite{Any extended integer types that exist in the implementation (6.2.5).}
130 @cite{Whether signed integer types are represented using sign and magnitude,
131 two's complement, or one's complement, and whether the extraordinary value
132 is a trap representation or an ordinary value (6.2.6.2).}
134 GCC supports only two's complement integer types, and all bit patterns
138 @cite{The rank of any extended integer type relative to another extended
139 integer type with the same precision (6.3.1.1).}
142 @cite{The result of, or the signal raised by, converting an integer to a
143 signed integer type when the value cannot be represented in an object of
144 that type (6.3.1.3).}
147 @cite{The results of some bitwise operations on signed integers (6.5).}
150 @node Floating point implementation
151 @section Floating point
155 @cite{The accuracy of the floating-point operations and of the library
156 functions in @code{<math.h>} and @code{<complex.h>} that return floating-point
157 results (5.2.4.2.2).}
160 @cite{The rounding behaviors characterized by non-standard values
161 of @code{FLT_ROUNDS} @gol
165 @cite{The evaluation methods characterized by non-standard negative
166 values of @code{FLT_EVAL_METHOD} (5.2.4.2.2).}
169 @cite{The direction of rounding when an integer is converted to a
170 floating-point number that cannot exactly represent the original
174 @cite{The direction of rounding when a floating-point number is
175 converted to a narrower floating-point number (6.3.1.5).}
178 @cite{How the nearest representable value or the larger or smaller
179 representable value immediately adjacent to the nearest representable
180 value is chosen for certain floating constants (6.4.4.2).}
183 @cite{Whether and how floating expressions are contracted when not
184 disallowed by the @code{FP_CONTRACT} pragma (6.5).}
187 @cite{The default state for the @code{FENV_ACCESS} pragma (7.6.1).}
190 @cite{Additional floating-point exceptions, rounding modes, environments,
191 and classifications, and their macro names (7.6, 7.12).}
194 @cite{The default state for the @code{FP_CONTRACT} pragma (7.12.2).}
197 @cite{Whether the ``inexact'' floating-point exception can be raised
198 when the rounded result actually does equal the mathematical result
199 in an IEC 60559 conformant implementation (F.9).}
202 @cite{Whether the ``underflow'' (and ``inexact'') floating-point
203 exception can be raised when a result is tiny but not inexact in an
204 IEC 60559 conformant implementation (F.9).}
208 @node Arrays and pointers implementation
209 @section Arrays and pointers
213 @cite{The result of converting a pointer to an integer or
214 vice versa (6.3.2.3).}
216 A cast from pointer to integer discards most-significant bits if the
217 pointer representation is larger than the integer type,
218 sign-extends@footnote{Future versions of GCC may zero-extend, or use
219 a target-defined @code{ptr_extend} pattern. Do not rely on sign extension.}
220 if the pointer representation is smaller than the integer type, otherwise
221 the bits are unchanged.
222 @c ??? We've always claimed that pointers were unsigned entities.
223 @c Shouldn't we therefore be doing zero-extension? If so, the bug
224 @c is in convert_to_integer, where we call type_for_size and request
225 @c a signed integral type. On the other hand, it might be most useful
226 @c for the target if we extend according to POINTERS_EXTEND_UNSIGNED.
228 A cast from integer to pointer discards most-significant bits if the
229 pointer representation is smaller than the integer type, extends according
230 to the signedness of the integer type if the pointer representation
231 is larger than the integer type, otherwise the bits are unchanged.
233 When casting from pointer to integer and back again, the resulting
234 pointer must reference the same object as the original pointer, otherwise
235 the behavior is undefined. That is, one may not use integer arithmetic to
236 avoid the undefined behavior of pointer arithmetic as proscribed in 6.5.6/8.
239 @cite{The size of the result of subtracting two pointers to elements
240 of the same array (6.5.6).}
244 @node Hints implementation
249 @cite{The extent to which suggestions made by using the @code{register}
250 storage-class specifier are effective (6.7.1).}
252 The @code{register} specifier affects code generation only in these ways:
256 When used as part of the register variable extension, see
257 @ref{Explicit Reg Vars}.
260 When @option{-O0} is in use, the compiler allocates distinct stack
261 memory for all variables that do not have the @code{register}
262 storage-class specifier; if @code{register} is specified, the variable
263 may have a shorter lifespan than the code would indicate and may never
267 On some rare x86 targets, @code{setjmp} doesn't save the registers in
268 all circumstances. In those cases, GCC doesn't allocate any variables
269 in registers unless they are marked @code{register}.
274 @cite{The extent to which suggestions made by using the inline function
275 specifier are effective (6.7.4).}
277 GCC will not inline any functions if the @option{-fno-inline} option is
278 used or if @option{-O0} is used. Otherwise, GCC may still be unable to
279 inline a function for many reasons; the @option{-Winline} option may be
280 used to determine if a function has not been inlined and why not.
284 @node Structures unions enumerations and bit-fields implementation
285 @section Structures, unions, enumerations, and bit-fields
289 @cite{Whether a ``plain'' int bit-field is treated as a @code{signed int}
290 bit-field or as an @code{unsigned int} bit-field (6.7.2, 6.7.2.1).}
293 @cite{Allowable bit-field types other than @code{_Bool}, @code{signed int},
294 and @code{unsigned int} (6.7.2.1).}
297 @cite{Whether a bit-field can straddle a storage-unit boundary (6.7.2.1).}
300 @cite{The order of allocation of bit-fields within a unit (6.7.2.1).}
303 @cite{The alignment of non-bit-field members of structures (6.7.2.1).}
306 @cite{The integer type compatible with each enumerated type (6.7.2.2).}
310 @node Qualifiers implementation
315 @cite{What constitutes an access to an object that has volatile-qualified
320 @node Preprocessing directives implementation
321 @section Preprocessing directives
325 @cite{How sequences in both forms of header names are mapped to headers
326 or external source file names (6.4.7).}
329 @cite{Whether the value of a character constant in a constant expression
330 that controls conditional inclusion matches the value of the same character
331 constant in the execution character set (6.10.1).}
334 @cite{Whether the value of a single-character character constant in a
335 constant expression that controls conditional inclusion may have a
336 negative value (6.10.1).}
339 @cite{The places that are searched for an included @samp{<>} delimited
340 header, and how the places are specified or the header is
341 identified (6.10.2).}
344 @cite{How the named source file is searched for in an included @samp{""}
345 delimited header (6.10.2).}
348 @cite{The method by which preprocessing tokens (possibly resulting from
349 macro expansion) in a @code{#include} directive are combined into a header
353 @cite{The nesting limit for @code{#include} processing (6.10.2).}
355 GCC imposes a limit of 200 nested @code{#include}s.
358 @cite{Whether the @samp{#} operator inserts a @samp{\} character before
359 the @samp{\} character that begins a universal character name in a
360 character constant or string literal (6.10.3.2).}
363 @cite{The behavior on each recognized non-@code{STDC #pragma}
367 @cite{The definitions for @code{__DATE__} and @code{__TIME__} when
368 respectively, the date and time of translation are not available (6.10.8).}
370 If the date and time are not available, @code{__DATE__} expands to
371 @code{@w{"??? ?? ????"}} and @code{__TIME__} expands to
376 @node Library functions implementation
377 @section Library functions
379 The behavior of these points are dependent on the implementation
380 of the C library, and are not defined by GCC itself.
382 @node Architecture implementation
383 @section Architecture
387 @cite{The values or expressions assigned to the macros specified in the
388 headers @code{<float.h>}, @code{<limits.h>}, and @code{<stdint.h>}
389 (5.2.4.2, 7.18.2, 7.18.3).}
392 @cite{The number, order, and encoding of bytes in any object
393 (when not explicitly specified in this International Standard) (6.2.6.1).}
396 @cite{The value of the result of the sizeof operator (6.5.3.4).}
400 @node Locale-specific behavior implementation
401 @section Locale-specific behavior
403 The behavior of these points are dependent on the implementation
404 of the C library, and are not defined by GCC itself.
407 @chapter Extensions to the C Language Family
408 @cindex extensions, C language
409 @cindex C language extensions
412 GNU C provides several language features not found in ISO standard C@.
413 (The @option{-pedantic} option directs GCC to print a warning message if
414 any of these features is used.) To test for the availability of these
415 features in conditional compilation, check for a predefined macro
416 @code{__GNUC__}, which is always defined under GCC@.
418 These extensions are available in C and Objective-C@. Most of them are
419 also available in C++. @xref{C++ Extensions,,Extensions to the
420 C++ Language}, for extensions that apply @emph{only} to C++.
422 Some features that are in ISO C99 but not C89 or C++ are also, as
423 extensions, accepted by GCC in C89 mode and in C++.
426 * Statement Exprs:: Putting statements and declarations inside expressions.
427 * Local Labels:: Labels local to a block.
428 * Labels as Values:: Getting pointers to labels, and computed gotos.
429 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
430 * Constructing Calls:: Dispatching a call to another function.
431 * Typeof:: @code{typeof}: referring to the type of an expression.
432 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
433 * Long Long:: Double-word integers---@code{long long int}.
434 * Complex:: Data types for complex numbers.
435 * Hex Floats:: Hexadecimal floating-point constants.
436 * Zero Length:: Zero-length arrays.
437 * Variable Length:: Arrays whose length is computed at run time.
438 * Empty Structures:: Structures with no members.
439 * Variadic Macros:: Macros with a variable number of arguments.
440 * Escaped Newlines:: Slightly looser rules for escaped newlines.
441 * Subscripting:: Any array can be subscripted, even if not an lvalue.
442 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
443 * Initializers:: Non-constant initializers.
444 * Compound Literals:: Compound literals give structures, unions
446 * Designated Inits:: Labeling elements of initializers.
447 * Cast to Union:: Casting to union type from any member of the union.
448 * Case Ranges:: `case 1 ... 9' and such.
449 * Mixed Declarations:: Mixing declarations and code.
450 * Function Attributes:: Declaring that functions have no side effects,
451 or that they can never return.
452 * Attribute Syntax:: Formal syntax for attributes.
453 * Function Prototypes:: Prototype declarations and old-style definitions.
454 * C++ Comments:: C++ comments are recognized.
455 * Dollar Signs:: Dollar sign is allowed in identifiers.
456 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
457 * Variable Attributes:: Specifying attributes of variables.
458 * Type Attributes:: Specifying attributes of types.
459 * Alignment:: Inquiring about the alignment of a type or variable.
460 * Inline:: Defining inline functions (as fast as macros).
461 * Extended Asm:: Assembler instructions with C expressions as operands.
462 (With them you can define ``built-in'' functions.)
463 * Constraints:: Constraints for asm operands
464 * Asm Labels:: Specifying the assembler name to use for a C symbol.
465 * Explicit Reg Vars:: Defining variables residing in specified registers.
466 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
467 * Incomplete Enums:: @code{enum foo;}, with details to follow.
468 * Function Names:: Printable strings which are the name of the current
470 * Return Address:: Getting the return or frame address of a function.
471 * Vector Extensions:: Using vector instructions through built-in functions.
472 * Offsetof:: Special syntax for implementing @code{offsetof}.
473 * Other Builtins:: Other built-in functions.
474 * Target Builtins:: Built-in functions specific to particular targets.
475 * Target Format Checks:: Format checks specific to particular targets.
476 * Pragmas:: Pragmas accepted by GCC.
477 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
478 * Thread-Local:: Per-thread variables.
481 @node Statement Exprs
482 @section Statements and Declarations in Expressions
483 @cindex statements inside expressions
484 @cindex declarations inside expressions
485 @cindex expressions containing statements
486 @cindex macros, statements in expressions
488 @c the above section title wrapped and causes an underfull hbox.. i
489 @c changed it from "within" to "in". --mew 4feb93
490 A compound statement enclosed in parentheses may appear as an expression
491 in GNU C@. This allows you to use loops, switches, and local variables
492 within an expression.
494 Recall that a compound statement is a sequence of statements surrounded
495 by braces; in this construct, parentheses go around the braces. For
499 (@{ int y = foo (); int z;
506 is a valid (though slightly more complex than necessary) expression
507 for the absolute value of @code{foo ()}.
509 The last thing in the compound statement should be an expression
510 followed by a semicolon; the value of this subexpression serves as the
511 value of the entire construct. (If you use some other kind of statement
512 last within the braces, the construct has type @code{void}, and thus
513 effectively no value.)
515 This feature is especially useful in making macro definitions ``safe'' (so
516 that they evaluate each operand exactly once). For example, the
517 ``maximum'' function is commonly defined as a macro in standard C as
521 #define max(a,b) ((a) > (b) ? (a) : (b))
525 @cindex side effects, macro argument
526 But this definition computes either @var{a} or @var{b} twice, with bad
527 results if the operand has side effects. In GNU C, if you know the
528 type of the operands (here taken as @code{int}), you can define
529 the macro safely as follows:
532 #define maxint(a,b) \
533 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
536 Embedded statements are not allowed in constant expressions, such as
537 the value of an enumeration constant, the width of a bit-field, or
538 the initial value of a static variable.
540 If you don't know the type of the operand, you can still do this, but you
541 must use @code{typeof} (@pxref{Typeof}).
543 In G++, the result value of a statement expression undergoes array and
544 function pointer decay, and is returned by value to the enclosing
545 expression. For instance, if @code{A} is a class, then
554 will construct a temporary @code{A} object to hold the result of the
555 statement expression, and that will be used to invoke @code{Foo}.
556 Therefore the @code{this} pointer observed by @code{Foo} will not be the
559 Any temporaries created within a statement within a statement expression
560 will be destroyed at the statement's end. This makes statement
561 expressions inside macros slightly different from function calls. In
562 the latter case temporaries introduced during argument evaluation will
563 be destroyed at the end of the statement that includes the function
564 call. In the statement expression case they will be destroyed during
565 the statement expression. For instance,
568 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
569 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
579 will have different places where temporaries are destroyed. For the
580 @code{macro} case, the temporary @code{X} will be destroyed just after
581 the initialization of @code{b}. In the @code{function} case that
582 temporary will be destroyed when the function returns.
584 These considerations mean that it is probably a bad idea to use
585 statement-expressions of this form in header files that are designed to
586 work with C++. (Note that some versions of the GNU C Library contained
587 header files using statement-expression that lead to precisely this
591 @section Locally Declared Labels
593 @cindex macros, local labels
595 GCC allows you to declare @dfn{local labels} in any nested block
596 scope. A local label is just like an ordinary label, but you can
597 only reference it (with a @code{goto} statement, or by taking its
598 address) within the block in which it was declared.
600 A local label declaration looks like this:
603 __label__ @var{label};
610 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
613 Local label declarations must come at the beginning of the block,
614 before any ordinary declarations or statements.
616 The label declaration defines the label @emph{name}, but does not define
617 the label itself. You must do this in the usual way, with
618 @code{@var{label}:}, within the statements of the statement expression.
620 The local label feature is useful for complex macros. If a macro
621 contains nested loops, a @code{goto} can be useful for breaking out of
622 them. However, an ordinary label whose scope is the whole function
623 cannot be used: if the macro can be expanded several times in one
624 function, the label will be multiply defined in that function. A
625 local label avoids this problem. For example:
628 #define SEARCH(value, array, target) \
631 typeof (target) _SEARCH_target = (target); \
632 typeof (*(array)) *_SEARCH_array = (array); \
635 for (i = 0; i < max; i++) \
636 for (j = 0; j < max; j++) \
637 if (_SEARCH_array[i][j] == _SEARCH_target) \
638 @{ (value) = i; goto found; @} \
644 This could also be written using a statement-expression:
647 #define SEARCH(array, target) \
650 typeof (target) _SEARCH_target = (target); \
651 typeof (*(array)) *_SEARCH_array = (array); \
654 for (i = 0; i < max; i++) \
655 for (j = 0; j < max; j++) \
656 if (_SEARCH_array[i][j] == _SEARCH_target) \
657 @{ value = i; goto found; @} \
664 Local label declarations also make the labels they declare visible to
665 nested functions, if there are any. @xref{Nested Functions}, for details.
667 @node Labels as Values
668 @section Labels as Values
669 @cindex labels as values
670 @cindex computed gotos
671 @cindex goto with computed label
672 @cindex address of a label
674 You can get the address of a label defined in the current function
675 (or a containing function) with the unary operator @samp{&&}. The
676 value has type @code{void *}. This value is a constant and can be used
677 wherever a constant of that type is valid. For example:
685 To use these values, you need to be able to jump to one. This is done
686 with the computed goto statement@footnote{The analogous feature in
687 Fortran is called an assigned goto, but that name seems inappropriate in
688 C, where one can do more than simply store label addresses in label
689 variables.}, @code{goto *@var{exp};}. For example,
696 Any expression of type @code{void *} is allowed.
698 One way of using these constants is in initializing a static array that
699 will serve as a jump table:
702 static void *array[] = @{ &&foo, &&bar, &&hack @};
705 Then you can select a label with indexing, like this:
712 Note that this does not check whether the subscript is in bounds---array
713 indexing in C never does that.
715 Such an array of label values serves a purpose much like that of the
716 @code{switch} statement. The @code{switch} statement is cleaner, so
717 use that rather than an array unless the problem does not fit a
718 @code{switch} statement very well.
720 Another use of label values is in an interpreter for threaded code.
721 The labels within the interpreter function can be stored in the
722 threaded code for super-fast dispatching.
724 You may not use this mechanism to jump to code in a different function.
725 If you do that, totally unpredictable things will happen. The best way to
726 avoid this is to store the label address only in automatic variables and
727 never pass it as an argument.
729 An alternate way to write the above example is
732 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
734 goto *(&&foo + array[i]);
738 This is more friendly to code living in shared libraries, as it reduces
739 the number of dynamic relocations that are needed, and by consequence,
740 allows the data to be read-only.
742 @node Nested Functions
743 @section Nested Functions
744 @cindex nested functions
745 @cindex downward funargs
748 A @dfn{nested function} is a function defined inside another function.
749 (Nested functions are not supported for GNU C++.) The nested function's
750 name is local to the block where it is defined. For example, here we
751 define a nested function named @code{square}, and call it twice:
755 foo (double a, double b)
757 double square (double z) @{ return z * z; @}
759 return square (a) + square (b);
764 The nested function can access all the variables of the containing
765 function that are visible at the point of its definition. This is
766 called @dfn{lexical scoping}. For example, here we show a nested
767 function which uses an inherited variable named @code{offset}:
771 bar (int *array, int offset, int size)
773 int access (int *array, int index)
774 @{ return array[index + offset]; @}
777 for (i = 0; i < size; i++)
778 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
783 Nested function definitions are permitted within functions in the places
784 where variable definitions are allowed; that is, in any block, before
785 the first statement in the block.
787 It is possible to call the nested function from outside the scope of its
788 name by storing its address or passing the address to another function:
791 hack (int *array, int size)
793 void store (int index, int value)
794 @{ array[index] = value; @}
796 intermediate (store, size);
800 Here, the function @code{intermediate} receives the address of
801 @code{store} as an argument. If @code{intermediate} calls @code{store},
802 the arguments given to @code{store} are used to store into @code{array}.
803 But this technique works only so long as the containing function
804 (@code{hack}, in this example) does not exit.
806 If you try to call the nested function through its address after the
807 containing function has exited, all hell will break loose. If you try
808 to call it after a containing scope level has exited, and if it refers
809 to some of the variables that are no longer in scope, you may be lucky,
810 but it's not wise to take the risk. If, however, the nested function
811 does not refer to anything that has gone out of scope, you should be
814 GCC implements taking the address of a nested function using a technique
815 called @dfn{trampolines}. A paper describing them is available as
818 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
820 A nested function can jump to a label inherited from a containing
821 function, provided the label was explicitly declared in the containing
822 function (@pxref{Local Labels}). Such a jump returns instantly to the
823 containing function, exiting the nested function which did the
824 @code{goto} and any intermediate functions as well. Here is an example:
828 bar (int *array, int offset, int size)
831 int access (int *array, int index)
835 return array[index + offset];
839 for (i = 0; i < size; i++)
840 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
844 /* @r{Control comes here from @code{access}
845 if it detects an error.} */
852 A nested function always has internal linkage. Declaring one with
853 @code{extern} is erroneous. If you need to declare the nested function
854 before its definition, use @code{auto} (which is otherwise meaningless
855 for function declarations).
858 bar (int *array, int offset, int size)
861 auto int access (int *, int);
863 int access (int *array, int index)
867 return array[index + offset];
873 @node Constructing Calls
874 @section Constructing Function Calls
875 @cindex constructing calls
876 @cindex forwarding calls
878 Using the built-in functions described below, you can record
879 the arguments a function received, and call another function
880 with the same arguments, without knowing the number or types
883 You can also record the return value of that function call,
884 and later return that value, without knowing what data type
885 the function tried to return (as long as your caller expects
888 However, these built-in functions may interact badly with some
889 sophisticated features or other extensions of the language. It
890 is, therefore, not recommended to use them outside very simple
891 functions acting as mere forwarders for their arguments.
893 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
894 This built-in function returns a pointer to data
895 describing how to perform a call with the same arguments as were passed
896 to the current function.
898 The function saves the arg pointer register, structure value address,
899 and all registers that might be used to pass arguments to a function
900 into a block of memory allocated on the stack. Then it returns the
901 address of that block.
904 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
905 This built-in function invokes @var{function}
906 with a copy of the parameters described by @var{arguments}
909 The value of @var{arguments} should be the value returned by
910 @code{__builtin_apply_args}. The argument @var{size} specifies the size
911 of the stack argument data, in bytes.
913 This function returns a pointer to data describing
914 how to return whatever value was returned by @var{function}. The data
915 is saved in a block of memory allocated on the stack.
917 It is not always simple to compute the proper value for @var{size}. The
918 value is used by @code{__builtin_apply} to compute the amount of data
919 that should be pushed on the stack and copied from the incoming argument
923 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
924 This built-in function returns the value described by @var{result} from
925 the containing function. You should specify, for @var{result}, a value
926 returned by @code{__builtin_apply}.
930 @section Referring to a Type with @code{typeof}
933 @cindex macros, types of arguments
935 Another way to refer to the type of an expression is with @code{typeof}.
936 The syntax of using of this keyword looks like @code{sizeof}, but the
937 construct acts semantically like a type name defined with @code{typedef}.
939 There are two ways of writing the argument to @code{typeof}: with an
940 expression or with a type. Here is an example with an expression:
947 This assumes that @code{x} is an array of pointers to functions;
948 the type described is that of the values of the functions.
950 Here is an example with a typename as the argument:
957 Here the type described is that of pointers to @code{int}.
959 If you are writing a header file that must work when included in ISO C
960 programs, write @code{__typeof__} instead of @code{typeof}.
961 @xref{Alternate Keywords}.
963 A @code{typeof}-construct can be used anywhere a typedef name could be
964 used. For example, you can use it in a declaration, in a cast, or inside
965 of @code{sizeof} or @code{typeof}.
967 @code{typeof} is often useful in conjunction with the
968 statements-within-expressions feature. Here is how the two together can
969 be used to define a safe ``maximum'' macro that operates on any
970 arithmetic type and evaluates each of its arguments exactly once:
974 (@{ typeof (a) _a = (a); \
975 typeof (b) _b = (b); \
976 _a > _b ? _a : _b; @})
979 @cindex underscores in variables in macros
980 @cindex @samp{_} in variables in macros
981 @cindex local variables in macros
982 @cindex variables, local, in macros
983 @cindex macros, local variables in
985 The reason for using names that start with underscores for the local
986 variables is to avoid conflicts with variable names that occur within the
987 expressions that are substituted for @code{a} and @code{b}. Eventually we
988 hope to design a new form of declaration syntax that allows you to declare
989 variables whose scopes start only after their initializers; this will be a
990 more reliable way to prevent such conflicts.
993 Some more examples of the use of @code{typeof}:
997 This declares @code{y} with the type of what @code{x} points to.
1004 This declares @code{y} as an array of such values.
1011 This declares @code{y} as an array of pointers to characters:
1014 typeof (typeof (char *)[4]) y;
1018 It is equivalent to the following traditional C declaration:
1024 To see the meaning of the declaration using @code{typeof}, and why it
1025 might be a useful way to write, rewrite it with these macros:
1028 #define pointer(T) typeof(T *)
1029 #define array(T, N) typeof(T [N])
1033 Now the declaration can be rewritten this way:
1036 array (pointer (char), 4) y;
1040 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1041 pointers to @code{char}.
1044 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1045 a more limited extension which permitted one to write
1048 typedef @var{T} = @var{expr};
1052 with the effect of declaring @var{T} to have the type of the expression
1053 @var{expr}. This extension does not work with GCC 3 (versions between
1054 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1055 relies on it should be rewritten to use @code{typeof}:
1058 typedef typeof(@var{expr}) @var{T};
1062 This will work with all versions of GCC@.
1065 @section Conditionals with Omitted Operands
1066 @cindex conditional expressions, extensions
1067 @cindex omitted middle-operands
1068 @cindex middle-operands, omitted
1069 @cindex extensions, @code{?:}
1070 @cindex @code{?:} extensions
1072 The middle operand in a conditional expression may be omitted. Then
1073 if the first operand is nonzero, its value is the value of the conditional
1076 Therefore, the expression
1083 has the value of @code{x} if that is nonzero; otherwise, the value of
1086 This example is perfectly equivalent to
1092 @cindex side effect in ?:
1093 @cindex ?: side effect
1095 In this simple case, the ability to omit the middle operand is not
1096 especially useful. When it becomes useful is when the first operand does,
1097 or may (if it is a macro argument), contain a side effect. Then repeating
1098 the operand in the middle would perform the side effect twice. Omitting
1099 the middle operand uses the value already computed without the undesirable
1100 effects of recomputing it.
1103 @section Double-Word Integers
1104 @cindex @code{long long} data types
1105 @cindex double-word arithmetic
1106 @cindex multiprecision arithmetic
1107 @cindex @code{LL} integer suffix
1108 @cindex @code{ULL} integer suffix
1110 ISO C99 supports data types for integers that are at least 64 bits wide,
1111 and as an extension GCC supports them in C89 mode and in C++.
1112 Simply write @code{long long int} for a signed integer, or
1113 @code{unsigned long long int} for an unsigned integer. To make an
1114 integer constant of type @code{long long int}, add the suffix @samp{LL}
1115 to the integer. To make an integer constant of type @code{unsigned long
1116 long int}, add the suffix @samp{ULL} to the integer.
1118 You can use these types in arithmetic like any other integer types.
1119 Addition, subtraction, and bitwise boolean operations on these types
1120 are open-coded on all types of machines. Multiplication is open-coded
1121 if the machine supports fullword-to-doubleword a widening multiply
1122 instruction. Division and shifts are open-coded only on machines that
1123 provide special support. The operations that are not open-coded use
1124 special library routines that come with GCC@.
1126 There may be pitfalls when you use @code{long long} types for function
1127 arguments, unless you declare function prototypes. If a function
1128 expects type @code{int} for its argument, and you pass a value of type
1129 @code{long long int}, confusion will result because the caller and the
1130 subroutine will disagree about the number of bytes for the argument.
1131 Likewise, if the function expects @code{long long int} and you pass
1132 @code{int}. The best way to avoid such problems is to use prototypes.
1135 @section Complex Numbers
1136 @cindex complex numbers
1137 @cindex @code{_Complex} keyword
1138 @cindex @code{__complex__} keyword
1140 ISO C99 supports complex floating data types, and as an extension GCC
1141 supports them in C89 mode and in C++, and supports complex integer data
1142 types which are not part of ISO C99. You can declare complex types
1143 using the keyword @code{_Complex}. As an extension, the older GNU
1144 keyword @code{__complex__} is also supported.
1146 For example, @samp{_Complex double x;} declares @code{x} as a
1147 variable whose real part and imaginary part are both of type
1148 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1149 have real and imaginary parts of type @code{short int}; this is not
1150 likely to be useful, but it shows that the set of complex types is
1153 To write a constant with a complex data type, use the suffix @samp{i} or
1154 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1155 has type @code{_Complex float} and @code{3i} has type
1156 @code{_Complex int}. Such a constant always has a pure imaginary
1157 value, but you can form any complex value you like by adding one to a
1158 real constant. This is a GNU extension; if you have an ISO C99
1159 conforming C library (such as GNU libc), and want to construct complex
1160 constants of floating type, you should include @code{<complex.h>} and
1161 use the macros @code{I} or @code{_Complex_I} instead.
1163 @cindex @code{__real__} keyword
1164 @cindex @code{__imag__} keyword
1165 To extract the real part of a complex-valued expression @var{exp}, write
1166 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1167 extract the imaginary part. This is a GNU extension; for values of
1168 floating type, you should use the ISO C99 functions @code{crealf},
1169 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1170 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1171 built-in functions by GCC@.
1173 @cindex complex conjugation
1174 The operator @samp{~} performs complex conjugation when used on a value
1175 with a complex type. This is a GNU extension; for values of
1176 floating type, you should use the ISO C99 functions @code{conjf},
1177 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1178 provided as built-in functions by GCC@.
1180 GCC can allocate complex automatic variables in a noncontiguous
1181 fashion; it's even possible for the real part to be in a register while
1182 the imaginary part is on the stack (or vice-versa). Only the DWARF2
1183 debug info format can represent this, so use of DWARF2 is recommended.
1184 If you are using the stabs debug info format, GCC describes a noncontiguous
1185 complex variable as if it were two separate variables of noncomplex type.
1186 If the variable's actual name is @code{foo}, the two fictitious
1187 variables are named @code{foo$real} and @code{foo$imag}. You can
1188 examine and set these two fictitious variables with your debugger.
1194 ISO C99 supports floating-point numbers written not only in the usual
1195 decimal notation, such as @code{1.55e1}, but also numbers such as
1196 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1197 supports this in C89 mode (except in some cases when strictly
1198 conforming) and in C++. In that format the
1199 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1200 mandatory. The exponent is a decimal number that indicates the power of
1201 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1208 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1209 is the same as @code{1.55e1}.
1211 Unlike for floating-point numbers in the decimal notation the exponent
1212 is always required in the hexadecimal notation. Otherwise the compiler
1213 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1214 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1215 extension for floating-point constants of type @code{float}.
1218 @section Arrays of Length Zero
1219 @cindex arrays of length zero
1220 @cindex zero-length arrays
1221 @cindex length-zero arrays
1222 @cindex flexible array members
1224 Zero-length arrays are allowed in GNU C@. They are very useful as the
1225 last element of a structure which is really a header for a variable-length
1234 struct line *thisline = (struct line *)
1235 malloc (sizeof (struct line) + this_length);
1236 thisline->length = this_length;
1239 In ISO C90, you would have to give @code{contents} a length of 1, which
1240 means either you waste space or complicate the argument to @code{malloc}.
1242 In ISO C99, you would use a @dfn{flexible array member}, which is
1243 slightly different in syntax and semantics:
1247 Flexible array members are written as @code{contents[]} without
1251 Flexible array members have incomplete type, and so the @code{sizeof}
1252 operator may not be applied. As a quirk of the original implementation
1253 of zero-length arrays, @code{sizeof} evaluates to zero.
1256 Flexible array members may only appear as the last member of a
1257 @code{struct} that is otherwise non-empty.
1260 A structure containing a flexible array member, or a union containing
1261 such a structure (possibly recursively), may not be a member of a
1262 structure or an element of an array. (However, these uses are
1263 permitted by GCC as extensions.)
1266 GCC versions before 3.0 allowed zero-length arrays to be statically
1267 initialized, as if they were flexible arrays. In addition to those
1268 cases that were useful, it also allowed initializations in situations
1269 that would corrupt later data. Non-empty initialization of zero-length
1270 arrays is now treated like any case where there are more initializer
1271 elements than the array holds, in that a suitable warning about "excess
1272 elements in array" is given, and the excess elements (all of them, in
1273 this case) are ignored.
1275 Instead GCC allows static initialization of flexible array members.
1276 This is equivalent to defining a new structure containing the original
1277 structure followed by an array of sufficient size to contain the data.
1278 I.e.@: in the following, @code{f1} is constructed as if it were declared
1284 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1287 struct f1 f1; int data[3];
1288 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1292 The convenience of this extension is that @code{f1} has the desired
1293 type, eliminating the need to consistently refer to @code{f2.f1}.
1295 This has symmetry with normal static arrays, in that an array of
1296 unknown size is also written with @code{[]}.
1298 Of course, this extension only makes sense if the extra data comes at
1299 the end of a top-level object, as otherwise we would be overwriting
1300 data at subsequent offsets. To avoid undue complication and confusion
1301 with initialization of deeply nested arrays, we simply disallow any
1302 non-empty initialization except when the structure is the top-level
1303 object. For example:
1306 struct foo @{ int x; int y[]; @};
1307 struct bar @{ struct foo z; @};
1309 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1310 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1311 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1312 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1315 @node Empty Structures
1316 @section Structures With No Members
1317 @cindex empty structures
1318 @cindex zero-size structures
1320 GCC permits a C structure to have no members:
1327 The structure will have size zero. In C++, empty structures are part
1328 of the language. G++ treats empty structures as if they had a single
1329 member of type @code{char}.
1331 @node Variable Length
1332 @section Arrays of Variable Length
1333 @cindex variable-length arrays
1334 @cindex arrays of variable length
1337 Variable-length automatic arrays are allowed in ISO C99, and as an
1338 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1339 implementation of variable-length arrays does not yet conform in detail
1340 to the ISO C99 standard.) These arrays are
1341 declared like any other automatic arrays, but with a length that is not
1342 a constant expression. The storage is allocated at the point of
1343 declaration and deallocated when the brace-level is exited. For
1348 concat_fopen (char *s1, char *s2, char *mode)
1350 char str[strlen (s1) + strlen (s2) + 1];
1353 return fopen (str, mode);
1357 @cindex scope of a variable length array
1358 @cindex variable-length array scope
1359 @cindex deallocating variable length arrays
1360 Jumping or breaking out of the scope of the array name deallocates the
1361 storage. Jumping into the scope is not allowed; you get an error
1364 @cindex @code{alloca} vs variable-length arrays
1365 You can use the function @code{alloca} to get an effect much like
1366 variable-length arrays. The function @code{alloca} is available in
1367 many other C implementations (but not in all). On the other hand,
1368 variable-length arrays are more elegant.
1370 There are other differences between these two methods. Space allocated
1371 with @code{alloca} exists until the containing @emph{function} returns.
1372 The space for a variable-length array is deallocated as soon as the array
1373 name's scope ends. (If you use both variable-length arrays and
1374 @code{alloca} in the same function, deallocation of a variable-length array
1375 will also deallocate anything more recently allocated with @code{alloca}.)
1377 You can also use variable-length arrays as arguments to functions:
1381 tester (int len, char data[len][len])
1387 The length of an array is computed once when the storage is allocated
1388 and is remembered for the scope of the array in case you access it with
1391 If you want to pass the array first and the length afterward, you can
1392 use a forward declaration in the parameter list---another GNU extension.
1396 tester (int len; char data[len][len], int len)
1402 @cindex parameter forward declaration
1403 The @samp{int len} before the semicolon is a @dfn{parameter forward
1404 declaration}, and it serves the purpose of making the name @code{len}
1405 known when the declaration of @code{data} is parsed.
1407 You can write any number of such parameter forward declarations in the
1408 parameter list. They can be separated by commas or semicolons, but the
1409 last one must end with a semicolon, which is followed by the ``real''
1410 parameter declarations. Each forward declaration must match a ``real''
1411 declaration in parameter name and data type. ISO C99 does not support
1412 parameter forward declarations.
1414 @node Variadic Macros
1415 @section Macros with a Variable Number of Arguments.
1416 @cindex variable number of arguments
1417 @cindex macro with variable arguments
1418 @cindex rest argument (in macro)
1419 @cindex variadic macros
1421 In the ISO C standard of 1999, a macro can be declared to accept a
1422 variable number of arguments much as a function can. The syntax for
1423 defining the macro is similar to that of a function. Here is an
1427 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1430 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1431 such a macro, it represents the zero or more tokens until the closing
1432 parenthesis that ends the invocation, including any commas. This set of
1433 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1434 wherever it appears. See the CPP manual for more information.
1436 GCC has long supported variadic macros, and used a different syntax that
1437 allowed you to give a name to the variable arguments just like any other
1438 argument. Here is an example:
1441 #define debug(format, args...) fprintf (stderr, format, args)
1444 This is in all ways equivalent to the ISO C example above, but arguably
1445 more readable and descriptive.
1447 GNU CPP has two further variadic macro extensions, and permits them to
1448 be used with either of the above forms of macro definition.
1450 In standard C, you are not allowed to leave the variable argument out
1451 entirely; but you are allowed to pass an empty argument. For example,
1452 this invocation is invalid in ISO C, because there is no comma after
1459 GNU CPP permits you to completely omit the variable arguments in this
1460 way. In the above examples, the compiler would complain, though since
1461 the expansion of the macro still has the extra comma after the format
1464 To help solve this problem, CPP behaves specially for variable arguments
1465 used with the token paste operator, @samp{##}. If instead you write
1468 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1471 and if the variable arguments are omitted or empty, the @samp{##}
1472 operator causes the preprocessor to remove the comma before it. If you
1473 do provide some variable arguments in your macro invocation, GNU CPP
1474 does not complain about the paste operation and instead places the
1475 variable arguments after the comma. Just like any other pasted macro
1476 argument, these arguments are not macro expanded.
1478 @node Escaped Newlines
1479 @section Slightly Looser Rules for Escaped Newlines
1480 @cindex escaped newlines
1481 @cindex newlines (escaped)
1483 Recently, the preprocessor has relaxed its treatment of escaped
1484 newlines. Previously, the newline had to immediately follow a
1485 backslash. The current implementation allows whitespace in the form
1486 of spaces, horizontal and vertical tabs, and form feeds between the
1487 backslash and the subsequent newline. The preprocessor issues a
1488 warning, but treats it as a valid escaped newline and combines the two
1489 lines to form a single logical line. This works within comments and
1490 tokens, as well as between tokens. Comments are @emph{not} treated as
1491 whitespace for the purposes of this relaxation, since they have not
1492 yet been replaced with spaces.
1495 @section Non-Lvalue Arrays May Have Subscripts
1496 @cindex subscripting
1497 @cindex arrays, non-lvalue
1499 @cindex subscripting and function values
1500 In ISO C99, arrays that are not lvalues still decay to pointers, and
1501 may be subscripted, although they may not be modified or used after
1502 the next sequence point and the unary @samp{&} operator may not be
1503 applied to them. As an extension, GCC allows such arrays to be
1504 subscripted in C89 mode, though otherwise they do not decay to
1505 pointers outside C99 mode. For example,
1506 this is valid in GNU C though not valid in C89:
1510 struct foo @{int a[4];@};
1516 return f().a[index];
1522 @section Arithmetic on @code{void}- and Function-Pointers
1523 @cindex void pointers, arithmetic
1524 @cindex void, size of pointer to
1525 @cindex function pointers, arithmetic
1526 @cindex function, size of pointer to
1528 In GNU C, addition and subtraction operations are supported on pointers to
1529 @code{void} and on pointers to functions. This is done by treating the
1530 size of a @code{void} or of a function as 1.
1532 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1533 and on function types, and returns 1.
1535 @opindex Wpointer-arith
1536 The option @option{-Wpointer-arith} requests a warning if these extensions
1540 @section Non-Constant Initializers
1541 @cindex initializers, non-constant
1542 @cindex non-constant initializers
1544 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1545 automatic variable are not required to be constant expressions in GNU C@.
1546 Here is an example of an initializer with run-time varying elements:
1549 foo (float f, float g)
1551 float beat_freqs[2] = @{ f-g, f+g @};
1556 @node Compound Literals
1557 @section Compound Literals
1558 @cindex constructor expressions
1559 @cindex initializations in expressions
1560 @cindex structures, constructor expression
1561 @cindex expressions, constructor
1562 @cindex compound literals
1563 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1565 ISO C99 supports compound literals. A compound literal looks like
1566 a cast containing an initializer. Its value is an object of the
1567 type specified in the cast, containing the elements specified in
1568 the initializer; it is an lvalue. As an extension, GCC supports
1569 compound literals in C89 mode and in C++.
1571 Usually, the specified type is a structure. Assume that
1572 @code{struct foo} and @code{structure} are declared as shown:
1575 struct foo @{int a; char b[2];@} structure;
1579 Here is an example of constructing a @code{struct foo} with a compound literal:
1582 structure = ((struct foo) @{x + y, 'a', 0@});
1586 This is equivalent to writing the following:
1590 struct foo temp = @{x + y, 'a', 0@};
1595 You can also construct an array. If all the elements of the compound literal
1596 are (made up of) simple constant expressions, suitable for use in
1597 initializers of objects of static storage duration, then the compound
1598 literal can be coerced to a pointer to its first element and used in
1599 such an initializer, as shown here:
1602 char **foo = (char *[]) @{ "x", "y", "z" @};
1605 Compound literals for scalar types and union types are is
1606 also allowed, but then the compound literal is equivalent
1609 As a GNU extension, GCC allows initialization of objects with static storage
1610 duration by compound literals (which is not possible in ISO C99, because
1611 the initializer is not a constant).
1612 It is handled as if the object was initialized only with the bracket
1613 enclosed list if compound literal's and object types match.
1614 The initializer list of the compound literal must be constant.
1615 If the object being initialized has array type of unknown size, the size is
1616 determined by compound literal size.
1619 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1620 static int y[] = (int []) @{1, 2, 3@};
1621 static int z[] = (int [3]) @{1@};
1625 The above lines are equivalent to the following:
1627 static struct foo x = @{1, 'a', 'b'@};
1628 static int y[] = @{1, 2, 3@};
1629 static int z[] = @{1, 0, 0@};
1632 @node Designated Inits
1633 @section Designated Initializers
1634 @cindex initializers with labeled elements
1635 @cindex labeled elements in initializers
1636 @cindex case labels in initializers
1637 @cindex designated initializers
1639 Standard C89 requires the elements of an initializer to appear in a fixed
1640 order, the same as the order of the elements in the array or structure
1643 In ISO C99 you can give the elements in any order, specifying the array
1644 indices or structure field names they apply to, and GNU C allows this as
1645 an extension in C89 mode as well. This extension is not
1646 implemented in GNU C++.
1648 To specify an array index, write
1649 @samp{[@var{index}] =} before the element value. For example,
1652 int a[6] = @{ [4] = 29, [2] = 15 @};
1659 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1663 The index values must be constant expressions, even if the array being
1664 initialized is automatic.
1666 An alternative syntax for this which has been obsolete since GCC 2.5 but
1667 GCC still accepts is to write @samp{[@var{index}]} before the element
1668 value, with no @samp{=}.
1670 To initialize a range of elements to the same value, write
1671 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1672 extension. For example,
1675 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1679 If the value in it has side-effects, the side-effects will happen only once,
1680 not for each initialized field by the range initializer.
1683 Note that the length of the array is the highest value specified
1686 In a structure initializer, specify the name of a field to initialize
1687 with @samp{.@var{fieldname} =} before the element value. For example,
1688 given the following structure,
1691 struct point @{ int x, y; @};
1695 the following initialization
1698 struct point p = @{ .y = yvalue, .x = xvalue @};
1705 struct point p = @{ xvalue, yvalue @};
1708 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1709 @samp{@var{fieldname}:}, as shown here:
1712 struct point p = @{ y: yvalue, x: xvalue @};
1716 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1717 @dfn{designator}. You can also use a designator (or the obsolete colon
1718 syntax) when initializing a union, to specify which element of the union
1719 should be used. For example,
1722 union foo @{ int i; double d; @};
1724 union foo f = @{ .d = 4 @};
1728 will convert 4 to a @code{double} to store it in the union using
1729 the second element. By contrast, casting 4 to type @code{union foo}
1730 would store it into the union as the integer @code{i}, since it is
1731 an integer. (@xref{Cast to Union}.)
1733 You can combine this technique of naming elements with ordinary C
1734 initialization of successive elements. Each initializer element that
1735 does not have a designator applies to the next consecutive element of the
1736 array or structure. For example,
1739 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1746 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1749 Labeling the elements of an array initializer is especially useful
1750 when the indices are characters or belong to an @code{enum} type.
1755 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1756 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1759 @cindex designator lists
1760 You can also write a series of @samp{.@var{fieldname}} and
1761 @samp{[@var{index}]} designators before an @samp{=} to specify a
1762 nested subobject to initialize; the list is taken relative to the
1763 subobject corresponding to the closest surrounding brace pair. For
1764 example, with the @samp{struct point} declaration above:
1767 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1771 If the same field is initialized multiple times, it will have value from
1772 the last initialization. If any such overridden initialization has
1773 side-effect, it is unspecified whether the side-effect happens or not.
1774 Currently, GCC will discard them and issue a warning.
1777 @section Case Ranges
1779 @cindex ranges in case statements
1781 You can specify a range of consecutive values in a single @code{case} label,
1785 case @var{low} ... @var{high}:
1789 This has the same effect as the proper number of individual @code{case}
1790 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1792 This feature is especially useful for ranges of ASCII character codes:
1798 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1799 it may be parsed wrong when you use it with integer values. For example,
1814 @section Cast to a Union Type
1815 @cindex cast to a union
1816 @cindex union, casting to a
1818 A cast to union type is similar to other casts, except that the type
1819 specified is a union type. You can specify the type either with
1820 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1821 a constructor though, not a cast, and hence does not yield an lvalue like
1822 normal casts. (@xref{Compound Literals}.)
1824 The types that may be cast to the union type are those of the members
1825 of the union. Thus, given the following union and variables:
1828 union foo @{ int i; double d; @};
1834 both @code{x} and @code{y} can be cast to type @code{union foo}.
1836 Using the cast as the right-hand side of an assignment to a variable of
1837 union type is equivalent to storing in a member of the union:
1842 u = (union foo) x @equiv{} u.i = x
1843 u = (union foo) y @equiv{} u.d = y
1846 You can also use the union cast as a function argument:
1849 void hack (union foo);
1851 hack ((union foo) x);
1854 @node Mixed Declarations
1855 @section Mixed Declarations and Code
1856 @cindex mixed declarations and code
1857 @cindex declarations, mixed with code
1858 @cindex code, mixed with declarations
1860 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1861 within compound statements. As an extension, GCC also allows this in
1862 C89 mode. For example, you could do:
1871 Each identifier is visible from where it is declared until the end of
1872 the enclosing block.
1874 @node Function Attributes
1875 @section Declaring Attributes of Functions
1876 @cindex function attributes
1877 @cindex declaring attributes of functions
1878 @cindex functions that never return
1879 @cindex functions that have no side effects
1880 @cindex functions in arbitrary sections
1881 @cindex functions that behave like malloc
1882 @cindex @code{volatile} applied to function
1883 @cindex @code{const} applied to function
1884 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1885 @cindex functions with non-null pointer arguments
1886 @cindex functions that are passed arguments in registers on the 386
1887 @cindex functions that pop the argument stack on the 386
1888 @cindex functions that do not pop the argument stack on the 386
1890 In GNU C, you declare certain things about functions called in your program
1891 which help the compiler optimize function calls and check your code more
1894 The keyword @code{__attribute__} allows you to specify special
1895 attributes when making a declaration. This keyword is followed by an
1896 attribute specification inside double parentheses. The following
1897 attributes are currently defined for functions on all targets:
1898 @code{noreturn}, @code{noinline}, @code{always_inline},
1899 @code{pure}, @code{const}, @code{nothrow},
1900 @code{format}, @code{format_arg}, @code{no_instrument_function},
1901 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1902 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1903 @code{alias}, @code{warn_unused_result} and @code{nonnull}. Several other
1904 attributes are defined for functions on particular target systems. Other
1905 attributes, including @code{section} are supported for variables declarations
1906 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1908 You may also specify attributes with @samp{__} preceding and following
1909 each keyword. This allows you to use them in header files without
1910 being concerned about a possible macro of the same name. For example,
1911 you may use @code{__noreturn__} instead of @code{noreturn}.
1913 @xref{Attribute Syntax}, for details of the exact syntax for using
1917 @c Keep this table alphabetized by attribute name. Treat _ as space.
1919 @item alias ("@var{target}")
1920 @cindex @code{alias} attribute
1921 The @code{alias} attribute causes the declaration to be emitted as an
1922 alias for another symbol, which must be specified. For instance,
1925 void __f () @{ /* @r{Do something.} */; @}
1926 void f () __attribute__ ((weak, alias ("__f")));
1929 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
1930 mangled name for the target must be used.
1932 Not all target machines support this attribute.
1935 @cindex @code{always_inline} function attribute
1936 Generally, functions are not inlined unless optimization is specified.
1937 For functions declared inline, this attribute inlines the function even
1938 if no optimization level was specified.
1941 @cindex functions that do pop the argument stack on the 386
1943 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1944 assume that the calling function will pop off the stack space used to
1945 pass arguments. This is
1946 useful to override the effects of the @option{-mrtd} switch.
1949 @cindex @code{const} function attribute
1950 Many functions do not examine any values except their arguments, and
1951 have no effects except the return value. Basically this is just slightly
1952 more strict class than the @code{pure} attribute above, since function is not
1953 allowed to read global memory.
1955 @cindex pointer arguments
1956 Note that a function that has pointer arguments and examines the data
1957 pointed to must @emph{not} be declared @code{const}. Likewise, a
1958 function that calls a non-@code{const} function usually must not be
1959 @code{const}. It does not make sense for a @code{const} function to
1962 The attribute @code{const} is not implemented in GCC versions earlier
1963 than 2.5. An alternative way to declare that a function has no side
1964 effects, which works in the current version and in some older versions,
1968 typedef int intfn ();
1970 extern const intfn square;
1973 This approach does not work in GNU C++ from 2.6.0 on, since the language
1974 specifies that the @samp{const} must be attached to the return value.
1978 @cindex @code{constructor} function attribute
1979 @cindex @code{destructor} function attribute
1980 The @code{constructor} attribute causes the function to be called
1981 automatically before execution enters @code{main ()}. Similarly, the
1982 @code{destructor} attribute causes the function to be called
1983 automatically after @code{main ()} has completed or @code{exit ()} has
1984 been called. Functions with these attributes are useful for
1985 initializing data that will be used implicitly during the execution of
1988 These attributes are not currently implemented for Objective-C@.
1991 @cindex @code{deprecated} attribute.
1992 The @code{deprecated} attribute results in a warning if the function
1993 is used anywhere in the source file. This is useful when identifying
1994 functions that are expected to be removed in a future version of a
1995 program. The warning also includes the location of the declaration
1996 of the deprecated function, to enable users to easily find further
1997 information about why the function is deprecated, or what they should
1998 do instead. Note that the warnings only occurs for uses:
2001 int old_fn () __attribute__ ((deprecated));
2003 int (*fn_ptr)() = old_fn;
2006 results in a warning on line 3 but not line 2.
2008 The @code{deprecated} attribute can also be used for variables and
2009 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2012 @cindex @code{__declspec(dllexport)}
2013 On Microsoft Windows targets and Symbian targets the @code{dllexport}
2014 attribute causes the compiler to provide a global pointer to a pointer
2015 in a dll, so that it can be referenced with the @code{dllimport}
2016 attribute. The pointer name is formed by combining @code{_imp__} and
2017 the function or variable name.
2019 Currently, the @code{dllexport}attribute is ignored for inlined
2020 functions, but export can be forced by using the
2021 @option{-fkeep-inline-functions} flag. The attribute is also ignored for
2024 When applied to C++ classes. the attribute marks defined non-inlined
2025 member functions and static data members as exports. Static consts
2026 initialized in-class are not marked unless they are also defined
2029 On cygwin, mingw, arm-pe and sh-symbianelf targets,
2030 @code{__declspec(dllexport)} is recognized as a synonym for
2031 @code{__attribute__ ((dllexport))} for compatibility with other
2032 Microsoft Windows and Symbian compilers.
2034 For Microsoft Windows targets there are alternative methods for
2035 including the symbol in the dll's export table such as using a
2036 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2037 the @option{--export-all} linker flag.
2040 @cindex @code{__declspec(dllimport)}
2041 On Microsoft Windows and Symbian targets, the @code{dllimport}
2042 attribute causes the compiler to reference a function or variable via
2043 a global pointer to a pointer that is set up by the Microsoft Windows
2044 dll library. The pointer name is formed by combining @code{_imp__} and
2045 the function or variable name. The attribute implies @code{extern}
2048 Currently, the attribute is ignored for inlined functions. If the
2049 attribute is applied to a symbol @emph{definition}, an error is reported.
2050 If a symbol previously declared @code{dllimport} is later defined, the
2051 attribute is ignored in subsequent references, and a warning is emitted.
2052 The attribute is also overridden by a subsequent declaration as
2055 When applied to C++ classes, the attribute marks non-inlined
2056 member functions and static data members as imports. However, the
2057 attribute is ignored for virtual methods to allow creation of vtables
2060 For Symbian targets the @code{dllimport} attribute also has another
2061 affect - it can cause the vtable and run-time type information for a
2062 class to be exported. This happens when the class has a dllimport'ed
2063 constructor or a non-inline, non-pure virtual function and, for either
2064 of those two conditions, the class also has a inline constructor or
2065 destructor and has a key function that is defined in the current
2068 On cygwin, mingw, arm-pe sh-symbianelf targets,
2069 @code{__declspec(dllimport)} is recognized as a synonym for
2070 @code{__attribute__ ((dllimport))} for compatibility with other
2071 Microsoft Windows and Symbian compilers.
2073 For Microsoft Windows based targets the use of the @code{dllimport}
2074 attribute on functions is not necessary, but provides a small
2075 performance benefit by eliminating a thunk in the dll. The use of the
2076 @code{dllimport} attribute on imported variables was required on older
2077 versions of GNU ld, but can now be avoided by passing the
2078 @option{--enable-auto-import} switch to ld. As with functions, using
2079 the attribute for a variable eliminates a thunk in the dll.
2081 One drawback to using this attribute is that a pointer to a function or
2082 variable marked as dllimport cannot be used as a constant address. The
2083 attribute can be disabled for functions by setting the
2084 @option{-mnop-fun-dllimport} flag.
2087 @cindex eight bit data on the H8/300, H8/300H, and H8S
2088 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2089 variable should be placed into the eight bit data section.
2090 The compiler will generate more efficient code for certain operations
2091 on data in the eight bit data area. Note the eight bit data area is limited to
2094 You must use GAS and GLD from GNU binutils version 2.7 or later for
2095 this attribute to work correctly.
2098 @cindex functions which handle memory bank switching
2099 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2100 use a calling convention that takes care of switching memory banks when
2101 entering and leaving a function. This calling convention is also the
2102 default when using the @option{-mlong-calls} option.
2104 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2105 to call and return from a function.
2107 On 68HC11 the compiler will generate a sequence of instructions
2108 to invoke a board-specific routine to switch the memory bank and call the
2109 real function. The board-specific routine simulates a @code{call}.
2110 At the end of a function, it will jump to a board-specific routine
2111 instead of using @code{rts}. The board-specific return routine simulates
2115 @cindex functions that pop the argument stack on the 386
2116 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2117 pass the first two arguments in the registers ECX and EDX. Subsequent
2118 arguments are passed on the stack. The called function will pop the
2119 arguments off the stack. If the number of arguments is variable all
2120 arguments are pushed on the stack.
2122 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2123 @cindex @code{format} function attribute
2125 The @code{format} attribute specifies that a function takes @code{printf},
2126 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2127 should be type-checked against a format string. For example, the
2132 my_printf (void *my_object, const char *my_format, ...)
2133 __attribute__ ((format (printf, 2, 3)));
2137 causes the compiler to check the arguments in calls to @code{my_printf}
2138 for consistency with the @code{printf} style format string argument
2141 The parameter @var{archetype} determines how the format string is
2142 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2143 or @code{strfmon}. (You can also use @code{__printf__},
2144 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2145 parameter @var{string-index} specifies which argument is the format
2146 string argument (starting from 1), while @var{first-to-check} is the
2147 number of the first argument to check against the format string. For
2148 functions where the arguments are not available to be checked (such as
2149 @code{vprintf}), specify the third parameter as zero. In this case the
2150 compiler only checks the format string for consistency. For
2151 @code{strftime} formats, the third parameter is required to be zero.
2152 Since non-static C++ methods have an implicit @code{this} argument, the
2153 arguments of such methods should be counted from two, not one, when
2154 giving values for @var{string-index} and @var{first-to-check}.
2156 In the example above, the format string (@code{my_format}) is the second
2157 argument of the function @code{my_print}, and the arguments to check
2158 start with the third argument, so the correct parameters for the format
2159 attribute are 2 and 3.
2161 @opindex ffreestanding
2162 The @code{format} attribute allows you to identify your own functions
2163 which take format strings as arguments, so that GCC can check the
2164 calls to these functions for errors. The compiler always (unless
2165 @option{-ffreestanding} is used) checks formats
2166 for the standard library functions @code{printf}, @code{fprintf},
2167 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2168 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2169 warnings are requested (using @option{-Wformat}), so there is no need to
2170 modify the header file @file{stdio.h}. In C99 mode, the functions
2171 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2172 @code{vsscanf} are also checked. Except in strictly conforming C
2173 standard modes, the X/Open function @code{strfmon} is also checked as
2174 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2175 @xref{C Dialect Options,,Options Controlling C Dialect}.
2177 The target may provide additional types of format checks.
2178 @xref{Target Format Checks,,Format Checks Specific to Particular
2181 @item format_arg (@var{string-index})
2182 @cindex @code{format_arg} function attribute
2183 @opindex Wformat-nonliteral
2184 The @code{format_arg} attribute specifies that a function takes a format
2185 string for a @code{printf}, @code{scanf}, @code{strftime} or
2186 @code{strfmon} style function and modifies it (for example, to translate
2187 it into another language), so the result can be passed to a
2188 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2189 function (with the remaining arguments to the format function the same
2190 as they would have been for the unmodified string). For example, the
2195 my_dgettext (char *my_domain, const char *my_format)
2196 __attribute__ ((format_arg (2)));
2200 causes the compiler to check the arguments in calls to a @code{printf},
2201 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2202 format string argument is a call to the @code{my_dgettext} function, for
2203 consistency with the format string argument @code{my_format}. If the
2204 @code{format_arg} attribute had not been specified, all the compiler
2205 could tell in such calls to format functions would be that the format
2206 string argument is not constant; this would generate a warning when
2207 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2208 without the attribute.
2210 The parameter @var{string-index} specifies which argument is the format
2211 string argument (starting from one). Since non-static C++ methods have
2212 an implicit @code{this} argument, the arguments of such methods should
2213 be counted from two.
2215 The @code{format-arg} attribute allows you to identify your own
2216 functions which modify format strings, so that GCC can check the
2217 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2218 type function whose operands are a call to one of your own function.
2219 The compiler always treats @code{gettext}, @code{dgettext}, and
2220 @code{dcgettext} in this manner except when strict ISO C support is
2221 requested by @option{-ansi} or an appropriate @option{-std} option, or
2222 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2223 Controlling C Dialect}.
2225 @item function_vector
2226 @cindex calling functions through the function vector on the H8/300 processors
2227 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2228 function should be called through the function vector. Calling a
2229 function through the function vector will reduce code size, however;
2230 the function vector has a limited size (maximum 128 entries on the H8/300
2231 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2233 You must use GAS and GLD from GNU binutils version 2.7 or later for
2234 this attribute to work correctly.
2237 @cindex interrupt handler functions
2238 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2239 that the specified function is an interrupt handler. The compiler will
2240 generate function entry and exit sequences suitable for use in an
2241 interrupt handler when this attribute is present.
2243 Note, interrupt handlers for the m68k, H8/300, H8/300H, H8S, and SH processors
2244 can be specified via the @code{interrupt_handler} attribute.
2246 Note, on the AVR, interrupts will be enabled inside the function.
2248 Note, for the ARM, you can specify the kind of interrupt to be handled by
2249 adding an optional parameter to the interrupt attribute like this:
2252 void f () __attribute__ ((interrupt ("IRQ")));
2255 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2257 @item interrupt_handler
2258 @cindex interrupt handler functions on the m68k, H8/300 and SH processors
2259 Use this attribute on the m68k, H8/300, H8/300H, H8S, and SH to indicate that
2260 the specified function is an interrupt handler. The compiler will generate
2261 function entry and exit sequences suitable for use in an interrupt
2262 handler when this attribute is present.
2264 @item long_call/short_call
2265 @cindex indirect calls on ARM
2266 This attribute specifies how a particular function is called on
2267 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2268 command line switch and @code{#pragma long_calls} settings. The
2269 @code{long_call} attribute causes the compiler to always call the
2270 function by first loading its address into a register and then using the
2271 contents of that register. The @code{short_call} attribute always places
2272 the offset to the function from the call site into the @samp{BL}
2273 instruction directly.
2275 @item longcall/shortcall
2276 @cindex functions called via pointer on the RS/6000 and PowerPC
2277 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2278 compiler to always call this function via a pointer, just as it would if
2279 the @option{-mlongcall} option had been specified. The @code{shortcall}
2280 attribute causes the compiler not to do this. These attributes override
2281 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2284 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2285 calls are necessary.
2288 @cindex @code{malloc} attribute
2289 The @code{malloc} attribute is used to tell the compiler that a function
2290 may be treated as if any non-@code{NULL} pointer it returns cannot
2291 alias any other pointer valid when the function returns.
2292 This will often improve optimization.
2293 Standard functions with this property include @code{malloc} and
2294 @code{calloc}. @code{realloc}-like functions have this property as
2295 long as the old pointer is never referred to (including comparing it
2296 to the new pointer) after the function returns a non-@code{NULL}
2299 @item model (@var{model-name})
2300 @cindex function addressability on the M32R/D
2301 @cindex variable addressability on the IA-64
2303 On the M32R/D, use this attribute to set the addressability of an
2304 object, and of the code generated for a function. The identifier
2305 @var{model-name} is one of @code{small}, @code{medium}, or
2306 @code{large}, representing each of the code models.
2308 Small model objects live in the lower 16MB of memory (so that their
2309 addresses can be loaded with the @code{ld24} instruction), and are
2310 callable with the @code{bl} instruction.
2312 Medium model objects may live anywhere in the 32-bit address space (the
2313 compiler will generate @code{seth/add3} instructions to load their addresses),
2314 and are callable with the @code{bl} instruction.
2316 Large model objects may live anywhere in the 32-bit address space (the
2317 compiler will generate @code{seth/add3} instructions to load their addresses),
2318 and may not be reachable with the @code{bl} instruction (the compiler will
2319 generate the much slower @code{seth/add3/jl} instruction sequence).
2321 On IA-64, use this attribute to set the addressability of an object.
2322 At present, the only supported identifier for @var{model-name} is
2323 @code{small}, indicating addressability via ``small'' (22-bit)
2324 addresses (so that their addresses can be loaded with the @code{addl}
2325 instruction). Caveat: such addressing is by definition not position
2326 independent and hence this attribute must not be used for objects
2327 defined by shared libraries.
2330 @cindex function without a prologue/epilogue code
2331 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2332 specified function does not need prologue/epilogue sequences generated by
2333 the compiler. It is up to the programmer to provide these sequences.
2336 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2337 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2338 use the normal calling convention based on @code{jsr} and @code{rts}.
2339 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2342 @item no_instrument_function
2343 @cindex @code{no_instrument_function} function attribute
2344 @opindex finstrument-functions
2345 If @option{-finstrument-functions} is given, profiling function calls will
2346 be generated at entry and exit of most user-compiled functions.
2347 Functions with this attribute will not be so instrumented.
2350 @cindex @code{noinline} function attribute
2351 This function attribute prevents a function from being considered for
2354 @item nonnull (@var{arg-index}, @dots{})
2355 @cindex @code{nonnull} function attribute
2356 The @code{nonnull} attribute specifies that some function parameters should
2357 be non-null pointers. For instance, the declaration:
2361 my_memcpy (void *dest, const void *src, size_t len)
2362 __attribute__((nonnull (1, 2)));
2366 causes the compiler to check that, in calls to @code{my_memcpy},
2367 arguments @var{dest} and @var{src} are non-null. If the compiler
2368 determines that a null pointer is passed in an argument slot marked
2369 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2370 is issued. The compiler may also choose to make optimizations based
2371 on the knowledge that certain function arguments will not be null.
2373 If no argument index list is given to the @code{nonnull} attribute,
2374 all pointer arguments are marked as non-null. To illustrate, the
2375 following declaration is equivalent to the previous example:
2379 my_memcpy (void *dest, const void *src, size_t len)
2380 __attribute__((nonnull));
2384 @cindex @code{noreturn} function attribute
2385 A few standard library functions, such as @code{abort} and @code{exit},
2386 cannot return. GCC knows this automatically. Some programs define
2387 their own functions that never return. You can declare them
2388 @code{noreturn} to tell the compiler this fact. For example,
2392 void fatal () __attribute__ ((noreturn));
2395 fatal (/* @r{@dots{}} */)
2397 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2403 The @code{noreturn} keyword tells the compiler to assume that
2404 @code{fatal} cannot return. It can then optimize without regard to what
2405 would happen if @code{fatal} ever did return. This makes slightly
2406 better code. More importantly, it helps avoid spurious warnings of
2407 uninitialized variables.
2409 The @code{noreturn} keyword does not affect the exceptional path when that
2410 applies: a @code{noreturn}-marked function may still return to the caller
2411 by throwing an exception.
2413 Do not assume that registers saved by the calling function are
2414 restored before calling the @code{noreturn} function.
2416 It does not make sense for a @code{noreturn} function to have a return
2417 type other than @code{void}.
2419 The attribute @code{noreturn} is not implemented in GCC versions
2420 earlier than 2.5. An alternative way to declare that a function does
2421 not return, which works in the current version and in some older
2422 versions, is as follows:
2425 typedef void voidfn ();
2427 volatile voidfn fatal;
2431 @cindex @code{nothrow} function attribute
2432 The @code{nothrow} attribute is used to inform the compiler that a
2433 function cannot throw an exception. For example, most functions in
2434 the standard C library can be guaranteed not to throw an exception
2435 with the notable exceptions of @code{qsort} and @code{bsearch} that
2436 take function pointer arguments. The @code{nothrow} attribute is not
2437 implemented in GCC versions earlier than 3.2.
2440 @cindex @code{pure} function attribute
2441 Many functions have no effects except the return value and their
2442 return value depends only on the parameters and/or global variables.
2443 Such a function can be subject
2444 to common subexpression elimination and loop optimization just as an
2445 arithmetic operator would be. These functions should be declared
2446 with the attribute @code{pure}. For example,
2449 int square (int) __attribute__ ((pure));
2453 says that the hypothetical function @code{square} is safe to call
2454 fewer times than the program says.
2456 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2457 Interesting non-pure functions are functions with infinite loops or those
2458 depending on volatile memory or other system resource, that may change between
2459 two consecutive calls (such as @code{feof} in a multithreading environment).
2461 The attribute @code{pure} is not implemented in GCC versions earlier
2464 @item regparm (@var{number})
2465 @cindex @code{regparm} attribute
2466 @cindex functions that are passed arguments in registers on the 386
2467 On the Intel 386, the @code{regparm} attribute causes the compiler to
2468 pass up to @var{number} integer arguments in registers EAX,
2469 EDX, and ECX instead of on the stack. Functions that take a
2470 variable number of arguments will continue to be passed all of their
2471 arguments on the stack.
2473 Beware that on some ELF systems this attribute is unsuitable for
2474 global functions in shared libraries with lazy binding (which is the
2475 default). Lazy binding will send the first call via resolving code in
2476 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2477 per the standard calling conventions. Solaris 8 is affected by this.
2478 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2479 safe since the loaders there save all registers. (Lazy binding can be
2480 disabled with the linker or the loader if desired, to avoid the
2484 @cindex save all registers on the H8/300, H8/300H, and H8S
2485 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
2486 all registers except the stack pointer should be saved in the prologue
2487 regardless of whether they are used or not.
2489 @item section ("@var{section-name}")
2490 @cindex @code{section} function attribute
2491 Normally, the compiler places the code it generates in the @code{text} section.
2492 Sometimes, however, you need additional sections, or you need certain
2493 particular functions to appear in special sections. The @code{section}
2494 attribute specifies that a function lives in a particular section.
2495 For example, the declaration:
2498 extern void foobar (void) __attribute__ ((section ("bar")));
2502 puts the function @code{foobar} in the @code{bar} section.
2504 Some file formats do not support arbitrary sections so the @code{section}
2505 attribute is not available on all platforms.
2506 If you need to map the entire contents of a module to a particular
2507 section, consider using the facilities of the linker instead.
2510 See long_call/short_call.
2513 See longcall/shortcall.
2516 @cindex signal handler functions on the AVR processors
2517 Use this attribute on the AVR to indicate that the specified
2518 function is a signal handler. The compiler will generate function
2519 entry and exit sequences suitable for use in a signal handler when this
2520 attribute is present. Interrupts will be disabled inside the function.
2523 Use this attribute on the SH to indicate an @code{interrupt_handler}
2524 function should switch to an alternate stack. It expects a string
2525 argument that names a global variable holding the address of the
2530 void f () __attribute__ ((interrupt_handler,
2531 sp_switch ("alt_stack")));
2535 @cindex functions that pop the argument stack on the 386
2536 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2537 assume that the called function will pop off the stack space used to
2538 pass arguments, unless it takes a variable number of arguments.
2541 @cindex tiny data section on the H8/300H and H8S
2542 Use this attribute on the H8/300H and H8S to indicate that the specified
2543 variable should be placed into the tiny data section.
2544 The compiler will generate more efficient code for loads and stores
2545 on data in the tiny data section. Note the tiny data area is limited to
2546 slightly under 32kbytes of data.
2549 Use this attribute on the SH for an @code{interrupt_handler} to return using
2550 @code{trapa} instead of @code{rte}. This attribute expects an integer
2551 argument specifying the trap number to be used.
2554 @cindex @code{unused} attribute.
2555 This attribute, attached to a function, means that the function is meant
2556 to be possibly unused. GCC will not produce a warning for this
2560 @cindex @code{used} attribute.
2561 This attribute, attached to a function, means that code must be emitted
2562 for the function even if it appears that the function is not referenced.
2563 This is useful, for example, when the function is referenced only in
2566 @item visibility ("@var{visibility_type}")
2567 @cindex @code{visibility} attribute
2568 The @code{visibility} attribute on ELF targets causes the declaration
2569 to be emitted with default, hidden, protected or internal visibility.
2572 void __attribute__ ((visibility ("protected")))
2573 f () @{ /* @r{Do something.} */; @}
2574 int i __attribute__ ((visibility ("hidden")));
2577 See the ELF gABI for complete details, but the short story is:
2580 @c keep this list of visibilies in alphabetical order.
2583 Default visibility is the normal case for ELF. This value is
2584 available for the visibility attribute to override other options
2585 that may change the assumed visibility of symbols.
2588 Hidden visibility indicates that the symbol will not be placed into
2589 the dynamic symbol table, so no other @dfn{module} (executable or
2590 shared library) can reference it directly.
2593 Internal visibility is like hidden visibility, but with additional
2594 processor specific semantics. Unless otherwise specified by the psABI,
2595 GCC defines internal visibility to mean that the function is @emph{never}
2596 called from another module. Note that hidden symbols, while they cannot
2597 be referenced directly by other modules, can be referenced indirectly via
2598 function pointers. By indicating that a symbol cannot be called from
2599 outside the module, GCC may for instance omit the load of a PIC register
2600 since it is known that the calling function loaded the correct value.
2603 Protected visibility indicates that the symbol will be placed in the
2604 dynamic symbol table, but that references within the defining module
2605 will bind to the local symbol. That is, the symbol cannot be overridden
2610 Not all ELF targets support this attribute.
2612 @item warn_unused_result
2613 @cindex @code{warn_unused_result} attribute
2614 The @code{warn_unused_result} attribute causes a warning to be emitted
2615 if a caller of the function with this attribute does not use its
2616 return value. This is useful for functions where not checking
2617 the result is either a security problem or always a bug, such as
2621 int fn () __attribute__ ((warn_unused_result));
2624 if (fn () < 0) return -1;
2630 results in warning on line 5.
2633 @cindex @code{weak} attribute
2634 The @code{weak} attribute causes the declaration to be emitted as a weak
2635 symbol rather than a global. This is primarily useful in defining
2636 library functions which can be overridden in user code, though it can
2637 also be used with non-function declarations. Weak symbols are supported
2638 for ELF targets, and also for a.out targets when using the GNU assembler
2643 You can specify multiple attributes in a declaration by separating them
2644 by commas within the double parentheses or by immediately following an
2645 attribute declaration with another attribute declaration.
2647 @cindex @code{#pragma}, reason for not using
2648 @cindex pragma, reason for not using
2649 Some people object to the @code{__attribute__} feature, suggesting that
2650 ISO C's @code{#pragma} should be used instead. At the time
2651 @code{__attribute__} was designed, there were two reasons for not doing
2656 It is impossible to generate @code{#pragma} commands from a macro.
2659 There is no telling what the same @code{#pragma} might mean in another
2663 These two reasons applied to almost any application that might have been
2664 proposed for @code{#pragma}. It was basically a mistake to use
2665 @code{#pragma} for @emph{anything}.
2667 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2668 to be generated from macros. In addition, a @code{#pragma GCC}
2669 namespace is now in use for GCC-specific pragmas. However, it has been
2670 found convenient to use @code{__attribute__} to achieve a natural
2671 attachment of attributes to their corresponding declarations, whereas
2672 @code{#pragma GCC} is of use for constructs that do not naturally form
2673 part of the grammar. @xref{Other Directives,,Miscellaneous
2674 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2676 @node Attribute Syntax
2677 @section Attribute Syntax
2678 @cindex attribute syntax
2680 This section describes the syntax with which @code{__attribute__} may be
2681 used, and the constructs to which attribute specifiers bind, for the C
2682 language. Some details may vary for C++ and Objective-C@. Because of
2683 infelicities in the grammar for attributes, some forms described here
2684 may not be successfully parsed in all cases.
2686 There are some problems with the semantics of attributes in C++. For
2687 example, there are no manglings for attributes, although they may affect
2688 code generation, so problems may arise when attributed types are used in
2689 conjunction with templates or overloading. Similarly, @code{typeid}
2690 does not distinguish between types with different attributes. Support
2691 for attributes in C++ may be restricted in future to attributes on
2692 declarations only, but not on nested declarators.
2694 @xref{Function Attributes}, for details of the semantics of attributes
2695 applying to functions. @xref{Variable Attributes}, for details of the
2696 semantics of attributes applying to variables. @xref{Type Attributes},
2697 for details of the semantics of attributes applying to structure, union
2698 and enumerated types.
2700 An @dfn{attribute specifier} is of the form
2701 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2702 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2703 each attribute is one of the following:
2707 Empty. Empty attributes are ignored.
2710 A word (which may be an identifier such as @code{unused}, or a reserved
2711 word such as @code{const}).
2714 A word, followed by, in parentheses, parameters for the attribute.
2715 These parameters take one of the following forms:
2719 An identifier. For example, @code{mode} attributes use this form.
2722 An identifier followed by a comma and a non-empty comma-separated list
2723 of expressions. For example, @code{format} attributes use this form.
2726 A possibly empty comma-separated list of expressions. For example,
2727 @code{format_arg} attributes use this form with the list being a single
2728 integer constant expression, and @code{alias} attributes use this form
2729 with the list being a single string constant.
2733 An @dfn{attribute specifier list} is a sequence of one or more attribute
2734 specifiers, not separated by any other tokens.
2736 In GNU C, an attribute specifier list may appear after the colon following a
2737 label, other than a @code{case} or @code{default} label. The only
2738 attribute it makes sense to use after a label is @code{unused}. This
2739 feature is intended for code generated by programs which contains labels
2740 that may be unused but which is compiled with @option{-Wall}. It would
2741 not normally be appropriate to use in it human-written code, though it
2742 could be useful in cases where the code that jumps to the label is
2743 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2744 such placement of attribute lists, as it is permissible for a
2745 declaration, which could begin with an attribute list, to be labelled in
2746 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2747 does not arise there.
2749 An attribute specifier list may appear as part of a @code{struct},
2750 @code{union} or @code{enum} specifier. It may go either immediately
2751 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2752 the closing brace. It is ignored if the content of the structure, union
2753 or enumerated type is not defined in the specifier in which the
2754 attribute specifier list is used---that is, in usages such as
2755 @code{struct __attribute__((foo)) bar} with no following opening brace.
2756 Where attribute specifiers follow the closing brace, they are considered
2757 to relate to the structure, union or enumerated type defined, not to any
2758 enclosing declaration the type specifier appears in, and the type
2759 defined is not complete until after the attribute specifiers.
2760 @c Otherwise, there would be the following problems: a shift/reduce
2761 @c conflict between attributes binding the struct/union/enum and
2762 @c binding to the list of specifiers/qualifiers; and "aligned"
2763 @c attributes could use sizeof for the structure, but the size could be
2764 @c changed later by "packed" attributes.
2766 Otherwise, an attribute specifier appears as part of a declaration,
2767 counting declarations of unnamed parameters and type names, and relates
2768 to that declaration (which may be nested in another declaration, for
2769 example in the case of a parameter declaration), or to a particular declarator
2770 within a declaration. Where an
2771 attribute specifier is applied to a parameter declared as a function or
2772 an array, it should apply to the function or array rather than the
2773 pointer to which the parameter is implicitly converted, but this is not
2774 yet correctly implemented.
2776 Any list of specifiers and qualifiers at the start of a declaration may
2777 contain attribute specifiers, whether or not such a list may in that
2778 context contain storage class specifiers. (Some attributes, however,
2779 are essentially in the nature of storage class specifiers, and only make
2780 sense where storage class specifiers may be used; for example,
2781 @code{section}.) There is one necessary limitation to this syntax: the
2782 first old-style parameter declaration in a function definition cannot
2783 begin with an attribute specifier, because such an attribute applies to
2784 the function instead by syntax described below (which, however, is not
2785 yet implemented in this case). In some other cases, attribute
2786 specifiers are permitted by this grammar but not yet supported by the
2787 compiler. All attribute specifiers in this place relate to the
2788 declaration as a whole. In the obsolescent usage where a type of
2789 @code{int} is implied by the absence of type specifiers, such a list of
2790 specifiers and qualifiers may be an attribute specifier list with no
2791 other specifiers or qualifiers.
2793 An attribute specifier list may appear immediately before a declarator
2794 (other than the first) in a comma-separated list of declarators in a
2795 declaration of more than one identifier using a single list of
2796 specifiers and qualifiers. Such attribute specifiers apply
2797 only to the identifier before whose declarator they appear. For
2801 __attribute__((noreturn)) void d0 (void),
2802 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2807 the @code{noreturn} attribute applies to all the functions
2808 declared; the @code{format} attribute only applies to @code{d1}.
2810 An attribute specifier list may appear immediately before the comma,
2811 @code{=} or semicolon terminating the declaration of an identifier other
2812 than a function definition. At present, such attribute specifiers apply
2813 to the declared object or function, but in future they may attach to the
2814 outermost adjacent declarator. In simple cases there is no difference,
2815 but, for example, in
2818 void (****f)(void) __attribute__((noreturn));
2822 at present the @code{noreturn} attribute applies to @code{f}, which
2823 causes a warning since @code{f} is not a function, but in future it may
2824 apply to the function @code{****f}. The precise semantics of what
2825 attributes in such cases will apply to are not yet specified. Where an
2826 assembler name for an object or function is specified (@pxref{Asm
2827 Labels}), at present the attribute must follow the @code{asm}
2828 specification; in future, attributes before the @code{asm} specification
2829 may apply to the adjacent declarator, and those after it to the declared
2832 An attribute specifier list may, in future, be permitted to appear after
2833 the declarator in a function definition (before any old-style parameter
2834 declarations or the function body).
2836 Attribute specifiers may be mixed with type qualifiers appearing inside
2837 the @code{[]} of a parameter array declarator, in the C99 construct by
2838 which such qualifiers are applied to the pointer to which the array is
2839 implicitly converted. Such attribute specifiers apply to the pointer,
2840 not to the array, but at present this is not implemented and they are
2843 An attribute specifier list may appear at the start of a nested
2844 declarator. At present, there are some limitations in this usage: the
2845 attributes correctly apply to the declarator, but for most individual
2846 attributes the semantics this implies are not implemented.
2847 When attribute specifiers follow the @code{*} of a pointer
2848 declarator, they may be mixed with any type qualifiers present.
2849 The following describes the formal semantics of this syntax. It will make the
2850 most sense if you are familiar with the formal specification of
2851 declarators in the ISO C standard.
2853 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2854 D1}, where @code{T} contains declaration specifiers that specify a type
2855 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2856 contains an identifier @var{ident}. The type specified for @var{ident}
2857 for derived declarators whose type does not include an attribute
2858 specifier is as in the ISO C standard.
2860 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2861 and the declaration @code{T D} specifies the type
2862 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2863 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2864 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2866 If @code{D1} has the form @code{*
2867 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2868 declaration @code{T D} specifies the type
2869 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2870 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2871 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2877 void (__attribute__((noreturn)) ****f) (void);
2881 specifies the type ``pointer to pointer to pointer to pointer to
2882 non-returning function returning @code{void}''. As another example,
2885 char *__attribute__((aligned(8))) *f;
2889 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2890 Note again that this does not work with most attributes; for example,
2891 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2892 is not yet supported.
2894 For compatibility with existing code written for compiler versions that
2895 did not implement attributes on nested declarators, some laxity is
2896 allowed in the placing of attributes. If an attribute that only applies
2897 to types is applied to a declaration, it will be treated as applying to
2898 the type of that declaration. If an attribute that only applies to
2899 declarations is applied to the type of a declaration, it will be treated
2900 as applying to that declaration; and, for compatibility with code
2901 placing the attributes immediately before the identifier declared, such
2902 an attribute applied to a function return type will be treated as
2903 applying to the function type, and such an attribute applied to an array
2904 element type will be treated as applying to the array type. If an
2905 attribute that only applies to function types is applied to a
2906 pointer-to-function type, it will be treated as applying to the pointer
2907 target type; if such an attribute is applied to a function return type
2908 that is not a pointer-to-function type, it will be treated as applying
2909 to the function type.
2911 @node Function Prototypes
2912 @section Prototypes and Old-Style Function Definitions
2913 @cindex function prototype declarations
2914 @cindex old-style function definitions
2915 @cindex promotion of formal parameters
2917 GNU C extends ISO C to allow a function prototype to override a later
2918 old-style non-prototype definition. Consider the following example:
2921 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2928 /* @r{Prototype function declaration.} */
2929 int isroot P((uid_t));
2931 /* @r{Old-style function definition.} */
2933 isroot (x) /* ??? lossage here ??? */
2940 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2941 not allow this example, because subword arguments in old-style
2942 non-prototype definitions are promoted. Therefore in this example the
2943 function definition's argument is really an @code{int}, which does not
2944 match the prototype argument type of @code{short}.
2946 This restriction of ISO C makes it hard to write code that is portable
2947 to traditional C compilers, because the programmer does not know
2948 whether the @code{uid_t} type is @code{short}, @code{int}, or
2949 @code{long}. Therefore, in cases like these GNU C allows a prototype
2950 to override a later old-style definition. More precisely, in GNU C, a
2951 function prototype argument type overrides the argument type specified
2952 by a later old-style definition if the former type is the same as the
2953 latter type before promotion. Thus in GNU C the above example is
2954 equivalent to the following:
2967 GNU C++ does not support old-style function definitions, so this
2968 extension is irrelevant.
2971 @section C++ Style Comments
2973 @cindex C++ comments
2974 @cindex comments, C++ style
2976 In GNU C, you may use C++ style comments, which start with @samp{//} and
2977 continue until the end of the line. Many other C implementations allow
2978 such comments, and they are included in the 1999 C standard. However,
2979 C++ style comments are not recognized if you specify an @option{-std}
2980 option specifying a version of ISO C before C99, or @option{-ansi}
2981 (equivalent to @option{-std=c89}).
2984 @section Dollar Signs in Identifier Names
2986 @cindex dollar signs in identifier names
2987 @cindex identifier names, dollar signs in
2989 In GNU C, you may normally use dollar signs in identifier names.
2990 This is because many traditional C implementations allow such identifiers.
2991 However, dollar signs in identifiers are not supported on a few target
2992 machines, typically because the target assembler does not allow them.
2994 @node Character Escapes
2995 @section The Character @key{ESC} in Constants
2997 You can use the sequence @samp{\e} in a string or character constant to
2998 stand for the ASCII character @key{ESC}.
3001 @section Inquiring on Alignment of Types or Variables
3003 @cindex type alignment
3004 @cindex variable alignment
3006 The keyword @code{__alignof__} allows you to inquire about how an object
3007 is aligned, or the minimum alignment usually required by a type. Its
3008 syntax is just like @code{sizeof}.
3010 For example, if the target machine requires a @code{double} value to be
3011 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3012 This is true on many RISC machines. On more traditional machine
3013 designs, @code{__alignof__ (double)} is 4 or even 2.
3015 Some machines never actually require alignment; they allow reference to any
3016 data type even at an odd address. For these machines, @code{__alignof__}
3017 reports the @emph{recommended} alignment of a type.
3019 If the operand of @code{__alignof__} is an lvalue rather than a type,
3020 its value is the required alignment for its type, taking into account
3021 any minimum alignment specified with GCC's @code{__attribute__}
3022 extension (@pxref{Variable Attributes}). For example, after this
3026 struct foo @{ int x; char y; @} foo1;
3030 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3031 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3033 It is an error to ask for the alignment of an incomplete type.
3035 @node Variable Attributes
3036 @section Specifying Attributes of Variables
3037 @cindex attribute of variables
3038 @cindex variable attributes
3040 The keyword @code{__attribute__} allows you to specify special
3041 attributes of variables or structure fields. This keyword is followed
3042 by an attribute specification inside double parentheses. Some
3043 attributes are currently defined generically for variables.
3044 Other attributes are defined for variables on particular target
3045 systems. Other attributes are available for functions
3046 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3047 Other front ends might define more attributes
3048 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3050 You may also specify attributes with @samp{__} preceding and following
3051 each keyword. This allows you to use them in header files without
3052 being concerned about a possible macro of the same name. For example,
3053 you may use @code{__aligned__} instead of @code{aligned}.
3055 @xref{Attribute Syntax}, for details of the exact syntax for using
3059 @cindex @code{aligned} attribute
3060 @item aligned (@var{alignment})
3061 This attribute specifies a minimum alignment for the variable or
3062 structure field, measured in bytes. For example, the declaration:
3065 int x __attribute__ ((aligned (16))) = 0;
3069 causes the compiler to allocate the global variable @code{x} on a
3070 16-byte boundary. On a 68040, this could be used in conjunction with
3071 an @code{asm} expression to access the @code{move16} instruction which
3072 requires 16-byte aligned operands.
3074 You can also specify the alignment of structure fields. For example, to
3075 create a double-word aligned @code{int} pair, you could write:
3078 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3082 This is an alternative to creating a union with a @code{double} member
3083 that forces the union to be double-word aligned.
3085 As in the preceding examples, you can explicitly specify the alignment
3086 (in bytes) that you wish the compiler to use for a given variable or
3087 structure field. Alternatively, you can leave out the alignment factor
3088 and just ask the compiler to align a variable or field to the maximum
3089 useful alignment for the target machine you are compiling for. For
3090 example, you could write:
3093 short array[3] __attribute__ ((aligned));
3096 Whenever you leave out the alignment factor in an @code{aligned} attribute
3097 specification, the compiler automatically sets the alignment for the declared
3098 variable or field to the largest alignment which is ever used for any data
3099 type on the target machine you are compiling for. Doing this can often make
3100 copy operations more efficient, because the compiler can use whatever
3101 instructions copy the biggest chunks of memory when performing copies to
3102 or from the variables or fields that you have aligned this way.
3104 The @code{aligned} attribute can only increase the alignment; but you
3105 can decrease it by specifying @code{packed} as well. See below.
3107 Note that the effectiveness of @code{aligned} attributes may be limited
3108 by inherent limitations in your linker. On many systems, the linker is
3109 only able to arrange for variables to be aligned up to a certain maximum
3110 alignment. (For some linkers, the maximum supported alignment may
3111 be very very small.) If your linker is only able to align variables
3112 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3113 in an @code{__attribute__} will still only provide you with 8 byte
3114 alignment. See your linker documentation for further information.
3116 @item cleanup (@var{cleanup_function})
3117 @cindex @code{cleanup} attribute
3118 The @code{cleanup} attribute runs a function when the variable goes
3119 out of scope. This attribute can only be applied to auto function
3120 scope variables; it may not be applied to parameters or variables
3121 with static storage duration. The function must take one parameter,
3122 a pointer to a type compatible with the variable. The return value
3123 of the function (if any) is ignored.
3125 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3126 will be run during the stack unwinding that happens during the
3127 processing of the exception. Note that the @code{cleanup} attribute
3128 does not allow the exception to be caught, only to perform an action.
3129 It is undefined what happens if @var{cleanup_function} does not
3134 @cindex @code{common} attribute
3135 @cindex @code{nocommon} attribute
3138 The @code{common} attribute requests GCC to place a variable in
3139 ``common'' storage. The @code{nocommon} attribute requests the
3140 opposite -- to allocate space for it directly.
3142 These attributes override the default chosen by the
3143 @option{-fno-common} and @option{-fcommon} flags respectively.
3146 @cindex @code{deprecated} attribute
3147 The @code{deprecated} attribute results in a warning if the variable
3148 is used anywhere in the source file. This is useful when identifying
3149 variables that are expected to be removed in a future version of a
3150 program. The warning also includes the location of the declaration
3151 of the deprecated variable, to enable users to easily find further
3152 information about why the variable is deprecated, or what they should
3153 do instead. Note that the warning only occurs for uses:
3156 extern int old_var __attribute__ ((deprecated));
3158 int new_fn () @{ return old_var; @}
3161 results in a warning on line 3 but not line 2.
3163 The @code{deprecated} attribute can also be used for functions and
3164 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3166 @item mode (@var{mode})
3167 @cindex @code{mode} attribute
3168 This attribute specifies the data type for the declaration---whichever
3169 type corresponds to the mode @var{mode}. This in effect lets you
3170 request an integer or floating point type according to its width.
3172 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3173 indicate the mode corresponding to a one-byte integer, @samp{word} or
3174 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3175 or @samp{__pointer__} for the mode used to represent pointers.
3178 @cindex @code{packed} attribute
3179 The @code{packed} attribute specifies that a variable or structure field
3180 should have the smallest possible alignment---one byte for a variable,
3181 and one bit for a field, unless you specify a larger value with the
3182 @code{aligned} attribute.
3184 Here is a structure in which the field @code{x} is packed, so that it
3185 immediately follows @code{a}:
3191 int x[2] __attribute__ ((packed));
3195 @item section ("@var{section-name}")
3196 @cindex @code{section} variable attribute
3197 Normally, the compiler places the objects it generates in sections like
3198 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3199 or you need certain particular variables to appear in special sections,
3200 for example to map to special hardware. The @code{section}
3201 attribute specifies that a variable (or function) lives in a particular
3202 section. For example, this small program uses several specific section names:
3205 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3206 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3207 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3208 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3212 /* Initialize stack pointer */
3213 init_sp (stack + sizeof (stack));
3215 /* Initialize initialized data */
3216 memcpy (&init_data, &data, &edata - &data);
3218 /* Turn on the serial ports */
3225 Use the @code{section} attribute with an @emph{initialized} definition
3226 of a @emph{global} variable, as shown in the example. GCC issues
3227 a warning and otherwise ignores the @code{section} attribute in
3228 uninitialized variable declarations.
3230 You may only use the @code{section} attribute with a fully initialized
3231 global definition because of the way linkers work. The linker requires
3232 each object be defined once, with the exception that uninitialized
3233 variables tentatively go in the @code{common} (or @code{bss}) section
3234 and can be multiply ``defined''. You can force a variable to be
3235 initialized with the @option{-fno-common} flag or the @code{nocommon}
3238 Some file formats do not support arbitrary sections so the @code{section}
3239 attribute is not available on all platforms.
3240 If you need to map the entire contents of a module to a particular
3241 section, consider using the facilities of the linker instead.
3244 @cindex @code{shared} variable attribute
3245 On Microsoft Windows, in addition to putting variable definitions in a named
3246 section, the section can also be shared among all running copies of an
3247 executable or DLL@. For example, this small program defines shared data
3248 by putting it in a named section @code{shared} and marking the section
3252 int foo __attribute__((section ("shared"), shared)) = 0;
3257 /* Read and write foo. All running
3258 copies see the same value. */
3264 You may only use the @code{shared} attribute along with @code{section}
3265 attribute with a fully initialized global definition because of the way
3266 linkers work. See @code{section} attribute for more information.
3268 The @code{shared} attribute is only available on Microsoft Windows@.
3270 @item tls_model ("@var{tls_model}")
3271 @cindex @code{tls_model} attribute
3272 The @code{tls_model} attribute sets thread-local storage model
3273 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3274 overriding @code{-ftls-model=} command line switch on a per-variable
3276 The @var{tls_model} argument should be one of @code{global-dynamic},
3277 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3279 Not all targets support this attribute.
3281 @item transparent_union
3282 This attribute, attached to a function parameter which is a union, means
3283 that the corresponding argument may have the type of any union member,
3284 but the argument is passed as if its type were that of the first union
3285 member. For more details see @xref{Type Attributes}. You can also use
3286 this attribute on a @code{typedef} for a union data type; then it
3287 applies to all function parameters with that type.
3290 This attribute, attached to a variable, means that the variable is meant
3291 to be possibly unused. GCC will not produce a warning for this
3294 @item vector_size (@var{bytes})
3295 This attribute specifies the vector size for the variable, measured in
3296 bytes. For example, the declaration:
3299 int foo __attribute__ ((vector_size (16)));
3303 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3304 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3305 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3307 This attribute is only applicable to integral and float scalars,
3308 although arrays, pointers, and function return values are allowed in
3309 conjunction with this construct.
3311 Aggregates with this attribute are invalid, even if they are of the same
3312 size as a corresponding scalar. For example, the declaration:
3315 struct S @{ int a; @};
3316 struct S __attribute__ ((vector_size (16))) foo;
3320 is invalid even if the size of the structure is the same as the size of
3324 The @code{weak} attribute is described in @xref{Function Attributes}.
3327 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3330 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3334 @subsection M32R/D Variable Attributes
3336 One attribute is currently defined for the M32R/D.
3339 @item model (@var{model-name})
3340 @cindex variable addressability on the M32R/D
3341 Use this attribute on the M32R/D to set the addressability of an object.
3342 The identifier @var{model-name} is one of @code{small}, @code{medium},
3343 or @code{large}, representing each of the code models.
3345 Small model objects live in the lower 16MB of memory (so that their
3346 addresses can be loaded with the @code{ld24} instruction).
3348 Medium and large model objects may live anywhere in the 32-bit address space
3349 (the compiler will generate @code{seth/add3} instructions to load their
3353 @subsection i386 Variable Attributes
3355 Two attributes are currently defined for i386 configurations:
3356 @code{ms_struct} and @code{gcc_struct}
3361 @cindex @code{ms_struct} attribute
3362 @cindex @code{gcc_struct} attribute
3364 If @code{packed} is used on a structure, or if bit-fields are used
3365 it may be that the Microsoft ABI packs them differently
3366 than GCC would normally pack them. Particularly when moving packed
3367 data between functions compiled with GCC and the native Microsoft compiler
3368 (either via function call or as data in a file), it may be necessary to access
3371 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3372 compilers to match the native Microsoft compiler.
3375 @node Type Attributes
3376 @section Specifying Attributes of Types
3377 @cindex attribute of types
3378 @cindex type attributes
3380 The keyword @code{__attribute__} allows you to specify special
3381 attributes of @code{struct} and @code{union} types when you define such
3382 types. This keyword is followed by an attribute specification inside
3383 double parentheses. Six attributes are currently defined for types:
3384 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3385 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3386 functions (@pxref{Function Attributes}) and for variables
3387 (@pxref{Variable Attributes}).
3389 You may also specify any one of these attributes with @samp{__}
3390 preceding and following its keyword. This allows you to use these
3391 attributes in header files without being concerned about a possible
3392 macro of the same name. For example, you may use @code{__aligned__}
3393 instead of @code{aligned}.
3395 You may specify the @code{aligned} and @code{transparent_union}
3396 attributes either in a @code{typedef} declaration or just past the
3397 closing curly brace of a complete enum, struct or union type
3398 @emph{definition} and the @code{packed} attribute only past the closing
3399 brace of a definition.
3401 You may also specify attributes between the enum, struct or union
3402 tag and the name of the type rather than after the closing brace.
3404 @xref{Attribute Syntax}, for details of the exact syntax for using
3408 @cindex @code{aligned} attribute
3409 @item aligned (@var{alignment})
3410 This attribute specifies a minimum alignment (in bytes) for variables
3411 of the specified type. For example, the declarations:
3414 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3415 typedef int more_aligned_int __attribute__ ((aligned (8)));
3419 force the compiler to insure (as far as it can) that each variable whose
3420 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3421 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3422 variables of type @code{struct S} aligned to 8-byte boundaries allows
3423 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3424 store) instructions when copying one variable of type @code{struct S} to
3425 another, thus improving run-time efficiency.
3427 Note that the alignment of any given @code{struct} or @code{union} type
3428 is required by the ISO C standard to be at least a perfect multiple of
3429 the lowest common multiple of the alignments of all of the members of
3430 the @code{struct} or @code{union} in question. This means that you @emph{can}
3431 effectively adjust the alignment of a @code{struct} or @code{union}
3432 type by attaching an @code{aligned} attribute to any one of the members
3433 of such a type, but the notation illustrated in the example above is a
3434 more obvious, intuitive, and readable way to request the compiler to
3435 adjust the alignment of an entire @code{struct} or @code{union} type.
3437 As in the preceding example, you can explicitly specify the alignment
3438 (in bytes) that you wish the compiler to use for a given @code{struct}
3439 or @code{union} type. Alternatively, you can leave out the alignment factor
3440 and just ask the compiler to align a type to the maximum
3441 useful alignment for the target machine you are compiling for. For
3442 example, you could write:
3445 struct S @{ short f[3]; @} __attribute__ ((aligned));
3448 Whenever you leave out the alignment factor in an @code{aligned}
3449 attribute specification, the compiler automatically sets the alignment
3450 for the type to the largest alignment which is ever used for any data
3451 type on the target machine you are compiling for. Doing this can often
3452 make copy operations more efficient, because the compiler can use
3453 whatever instructions copy the biggest chunks of memory when performing
3454 copies to or from the variables which have types that you have aligned
3457 In the example above, if the size of each @code{short} is 2 bytes, then
3458 the size of the entire @code{struct S} type is 6 bytes. The smallest
3459 power of two which is greater than or equal to that is 8, so the
3460 compiler sets the alignment for the entire @code{struct S} type to 8
3463 Note that although you can ask the compiler to select a time-efficient
3464 alignment for a given type and then declare only individual stand-alone
3465 objects of that type, the compiler's ability to select a time-efficient
3466 alignment is primarily useful only when you plan to create arrays of
3467 variables having the relevant (efficiently aligned) type. If you
3468 declare or use arrays of variables of an efficiently-aligned type, then
3469 it is likely that your program will also be doing pointer arithmetic (or
3470 subscripting, which amounts to the same thing) on pointers to the
3471 relevant type, and the code that the compiler generates for these
3472 pointer arithmetic operations will often be more efficient for
3473 efficiently-aligned types than for other types.
3475 The @code{aligned} attribute can only increase the alignment; but you
3476 can decrease it by specifying @code{packed} as well. See below.
3478 Note that the effectiveness of @code{aligned} attributes may be limited
3479 by inherent limitations in your linker. On many systems, the linker is
3480 only able to arrange for variables to be aligned up to a certain maximum
3481 alignment. (For some linkers, the maximum supported alignment may
3482 be very very small.) If your linker is only able to align variables
3483 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3484 in an @code{__attribute__} will still only provide you with 8 byte
3485 alignment. See your linker documentation for further information.
3488 This attribute, attached to @code{struct} or @code{union} type
3489 definition, specifies that each member of the structure or union is
3490 placed to minimize the memory required. When attached to an @code{enum}
3491 definition, it indicates that the smallest integral type should be used.
3493 @opindex fshort-enums
3494 Specifying this attribute for @code{struct} and @code{union} types is
3495 equivalent to specifying the @code{packed} attribute on each of the
3496 structure or union members. Specifying the @option{-fshort-enums}
3497 flag on the line is equivalent to specifying the @code{packed}
3498 attribute on all @code{enum} definitions.
3500 In the following example @code{struct my_packed_struct}'s members are
3501 packed closely together, but the internal layout of its @code{s} member
3502 is not packed -- to do that, @code{struct my_unpacked_struct} would need to
3506 struct my_unpacked_struct
3512 struct my_packed_struct __attribute__ ((__packed__))
3516 struct my_unpacked_struct s;
3520 You may only specify this attribute on the definition of a @code{enum},
3521 @code{struct} or @code{union}, not on a @code{typedef} which does not
3522 also define the enumerated type, structure or union.
3524 @item transparent_union
3525 This attribute, attached to a @code{union} type definition, indicates
3526 that any function parameter having that union type causes calls to that
3527 function to be treated in a special way.
3529 First, the argument corresponding to a transparent union type can be of
3530 any type in the union; no cast is required. Also, if the union contains
3531 a pointer type, the corresponding argument can be a null pointer
3532 constant or a void pointer expression; and if the union contains a void
3533 pointer type, the corresponding argument can be any pointer expression.
3534 If the union member type is a pointer, qualifiers like @code{const} on
3535 the referenced type must be respected, just as with normal pointer
3538 Second, the argument is passed to the function using the calling
3539 conventions of the first member of the transparent union, not the calling
3540 conventions of the union itself. All members of the union must have the
3541 same machine representation; this is necessary for this argument passing
3544 Transparent unions are designed for library functions that have multiple
3545 interfaces for compatibility reasons. For example, suppose the
3546 @code{wait} function must accept either a value of type @code{int *} to
3547 comply with Posix, or a value of type @code{union wait *} to comply with
3548 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3549 @code{wait} would accept both kinds of arguments, but it would also
3550 accept any other pointer type and this would make argument type checking
3551 less useful. Instead, @code{<sys/wait.h>} might define the interface
3559 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3561 pid_t wait (wait_status_ptr_t);
3564 This interface allows either @code{int *} or @code{union wait *}
3565 arguments to be passed, using the @code{int *} calling convention.
3566 The program can call @code{wait} with arguments of either type:
3569 int w1 () @{ int w; return wait (&w); @}
3570 int w2 () @{ union wait w; return wait (&w); @}
3573 With this interface, @code{wait}'s implementation might look like this:
3576 pid_t wait (wait_status_ptr_t p)
3578 return waitpid (-1, p.__ip, 0);
3583 When attached to a type (including a @code{union} or a @code{struct}),
3584 this attribute means that variables of that type are meant to appear
3585 possibly unused. GCC will not produce a warning for any variables of
3586 that type, even if the variable appears to do nothing. This is often
3587 the case with lock or thread classes, which are usually defined and then
3588 not referenced, but contain constructors and destructors that have
3589 nontrivial bookkeeping functions.
3592 The @code{deprecated} attribute results in a warning if the type
3593 is used anywhere in the source file. This is useful when identifying
3594 types that are expected to be removed in a future version of a program.
3595 If possible, the warning also includes the location of the declaration
3596 of the deprecated type, to enable users to easily find further
3597 information about why the type is deprecated, or what they should do
3598 instead. Note that the warnings only occur for uses and then only
3599 if the type is being applied to an identifier that itself is not being
3600 declared as deprecated.
3603 typedef int T1 __attribute__ ((deprecated));
3607 typedef T1 T3 __attribute__ ((deprecated));
3608 T3 z __attribute__ ((deprecated));
3611 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3612 warning is issued for line 4 because T2 is not explicitly
3613 deprecated. Line 5 has no warning because T3 is explicitly
3614 deprecated. Similarly for line 6.
3616 The @code{deprecated} attribute can also be used for functions and
3617 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3620 Accesses to objects with types with this attribute are not subjected to
3621 type-based alias analysis, but are instead assumed to be able to alias
3622 any other type of objects, just like the @code{char} type. See
3623 @option{-fstrict-aliasing} for more information on aliasing issues.
3628 typedef short __attribute__((__may_alias__)) short_a;
3634 short_a *b = (short_a *) &a;
3638 if (a == 0x12345678)
3645 If you replaced @code{short_a} with @code{short} in the variable
3646 declaration, the above program would abort when compiled with
3647 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3648 above in recent GCC versions.
3650 @subsection i386 Type Attributes
3652 Two attributes are currently defined for i386 configurations:
3653 @code{ms_struct} and @code{gcc_struct}
3657 @cindex @code{ms_struct}
3658 @cindex @code{gcc_struct}
3660 If @code{packed} is used on a structure, or if bit-fields are used
3661 it may be that the Microsoft ABI packs them differently
3662 than GCC would normally pack them. Particularly when moving packed
3663 data between functions compiled with GCC and the native Microsoft compiler
3664 (either via function call or as data in a file), it may be necessary to access
3667 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3668 compilers to match the native Microsoft compiler.
3671 To specify multiple attributes, separate them by commas within the
3672 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3676 @section An Inline Function is As Fast As a Macro
3677 @cindex inline functions
3678 @cindex integrating function code
3680 @cindex macros, inline alternative
3682 By declaring a function @code{inline}, you can direct GCC to
3683 integrate that function's code into the code for its callers. This
3684 makes execution faster by eliminating the function-call overhead; in
3685 addition, if any of the actual argument values are constant, their known
3686 values may permit simplifications at compile time so that not all of the
3687 inline function's code needs to be included. The effect on code size is
3688 less predictable; object code may be larger or smaller with function
3689 inlining, depending on the particular case. Inlining of functions is an
3690 optimization and it really ``works'' only in optimizing compilation. If
3691 you don't use @option{-O}, no function is really inline.
3693 Inline functions are included in the ISO C99 standard, but there are
3694 currently substantial differences between what GCC implements and what
3695 the ISO C99 standard requires.
3697 To declare a function inline, use the @code{inline} keyword in its
3698 declaration, like this:
3708 (If you are writing a header file to be included in ISO C programs, write
3709 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3710 You can also make all ``simple enough'' functions inline with the option
3711 @option{-finline-functions}.
3714 Note that certain usages in a function definition can make it unsuitable
3715 for inline substitution. Among these usages are: use of varargs, use of
3716 alloca, use of variable sized data types (@pxref{Variable Length}),
3717 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3718 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3719 will warn when a function marked @code{inline} could not be substituted,
3720 and will give the reason for the failure.
3722 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3723 does not affect the linkage of the function.
3725 @cindex automatic @code{inline} for C++ member fns
3726 @cindex @code{inline} automatic for C++ member fns
3727 @cindex member fns, automatically @code{inline}
3728 @cindex C++ member fns, automatically @code{inline}
3729 @opindex fno-default-inline
3730 GCC automatically inlines member functions defined within the class
3731 body of C++ programs even if they are not explicitly declared
3732 @code{inline}. (You can override this with @option{-fno-default-inline};
3733 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3735 @cindex inline functions, omission of
3736 @opindex fkeep-inline-functions
3737 When a function is both inline and @code{static}, if all calls to the
3738 function are integrated into the caller, and the function's address is
3739 never used, then the function's own assembler code is never referenced.
3740 In this case, GCC does not actually output assembler code for the
3741 function, unless you specify the option @option{-fkeep-inline-functions}.
3742 Some calls cannot be integrated for various reasons (in particular,
3743 calls that precede the function's definition cannot be integrated, and
3744 neither can recursive calls within the definition). If there is a
3745 nonintegrated call, then the function is compiled to assembler code as
3746 usual. The function must also be compiled as usual if the program
3747 refers to its address, because that can't be inlined.
3749 @cindex non-static inline function
3750 When an inline function is not @code{static}, then the compiler must assume
3751 that there may be calls from other source files; since a global symbol can
3752 be defined only once in any program, the function must not be defined in
3753 the other source files, so the calls therein cannot be integrated.
3754 Therefore, a non-@code{static} inline function is always compiled on its
3755 own in the usual fashion.
3757 If you specify both @code{inline} and @code{extern} in the function
3758 definition, then the definition is used only for inlining. In no case
3759 is the function compiled on its own, not even if you refer to its
3760 address explicitly. Such an address becomes an external reference, as
3761 if you had only declared the function, and had not defined it.
3763 This combination of @code{inline} and @code{extern} has almost the
3764 effect of a macro. The way to use it is to put a function definition in
3765 a header file with these keywords, and put another copy of the
3766 definition (lacking @code{inline} and @code{extern}) in a library file.
3767 The definition in the header file will cause most calls to the function
3768 to be inlined. If any uses of the function remain, they will refer to
3769 the single copy in the library.
3771 Since GCC eventually will implement ISO C99 semantics for
3772 inline functions, it is best to use @code{static inline} only
3773 to guarantee compatibility. (The
3774 existing semantics will remain available when @option{-std=gnu89} is
3775 specified, but eventually the default will be @option{-std=gnu99} and
3776 that will implement the C99 semantics, though it does not do so yet.)
3778 GCC does not inline any functions when not optimizing unless you specify
3779 the @samp{always_inline} attribute for the function, like this:
3783 inline void foo (const char) __attribute__((always_inline));
3787 @section Assembler Instructions with C Expression Operands
3788 @cindex extended @code{asm}
3789 @cindex @code{asm} expressions
3790 @cindex assembler instructions
3793 In an assembler instruction using @code{asm}, you can specify the
3794 operands of the instruction using C expressions. This means you need not
3795 guess which registers or memory locations will contain the data you want
3798 You must specify an assembler instruction template much like what
3799 appears in a machine description, plus an operand constraint string for
3802 For example, here is how to use the 68881's @code{fsinx} instruction:
3805 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3809 Here @code{angle} is the C expression for the input operand while
3810 @code{result} is that of the output operand. Each has @samp{"f"} as its
3811 operand constraint, saying that a floating point register is required.
3812 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3813 output operands' constraints must use @samp{=}. The constraints use the
3814 same language used in the machine description (@pxref{Constraints}).
3816 Each operand is described by an operand-constraint string followed by
3817 the C expression in parentheses. A colon separates the assembler
3818 template from the first output operand and another separates the last
3819 output operand from the first input, if any. Commas separate the
3820 operands within each group. The total number of operands is currently
3821 limited to 30; this limitation may be lifted in some future version of
3824 If there are no output operands but there are input operands, you must
3825 place two consecutive colons surrounding the place where the output
3828 As of GCC version 3.1, it is also possible to specify input and output
3829 operands using symbolic names which can be referenced within the
3830 assembler code. These names are specified inside square brackets
3831 preceding the constraint string, and can be referenced inside the
3832 assembler code using @code{%[@var{name}]} instead of a percentage sign
3833 followed by the operand number. Using named operands the above example
3837 asm ("fsinx %[angle],%[output]"
3838 : [output] "=f" (result)
3839 : [angle] "f" (angle));
3843 Note that the symbolic operand names have no relation whatsoever to
3844 other C identifiers. You may use any name you like, even those of
3845 existing C symbols, but you must ensure that no two operands within the same
3846 assembler construct use the same symbolic name.
3848 Output operand expressions must be lvalues; the compiler can check this.
3849 The input operands need not be lvalues. The compiler cannot check
3850 whether the operands have data types that are reasonable for the
3851 instruction being executed. It does not parse the assembler instruction
3852 template and does not know what it means or even whether it is valid
3853 assembler input. The extended @code{asm} feature is most often used for
3854 machine instructions the compiler itself does not know exist. If
3855 the output expression cannot be directly addressed (for example, it is a
3856 bit-field), your constraint must allow a register. In that case, GCC
3857 will use the register as the output of the @code{asm}, and then store
3858 that register into the output.
3860 The ordinary output operands must be write-only; GCC will assume that
3861 the values in these operands before the instruction are dead and need
3862 not be generated. Extended asm supports input-output or read-write
3863 operands. Use the constraint character @samp{+} to indicate such an
3864 operand and list it with the output operands. You should only use
3865 read-write operands when the constraints for the operand (or the
3866 operand in which only some of the bits are to be changed) allow a
3869 You may, as an alternative, logically split its function into two
3870 separate operands, one input operand and one write-only output
3871 operand. The connection between them is expressed by constraints
3872 which say they need to be in the same location when the instruction
3873 executes. You can use the same C expression for both operands, or
3874 different expressions. For example, here we write the (fictitious)
3875 @samp{combine} instruction with @code{bar} as its read-only source
3876 operand and @code{foo} as its read-write destination:
3879 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3883 The constraint @samp{"0"} for operand 1 says that it must occupy the
3884 same location as operand 0. A number in constraint is allowed only in
3885 an input operand and it must refer to an output operand.
3887 Only a number in the constraint can guarantee that one operand will be in
3888 the same place as another. The mere fact that @code{foo} is the value
3889 of both operands is not enough to guarantee that they will be in the
3890 same place in the generated assembler code. The following would not
3894 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3897 Various optimizations or reloading could cause operands 0 and 1 to be in
3898 different registers; GCC knows no reason not to do so. For example, the
3899 compiler might find a copy of the value of @code{foo} in one register and
3900 use it for operand 1, but generate the output operand 0 in a different
3901 register (copying it afterward to @code{foo}'s own address). Of course,
3902 since the register for operand 1 is not even mentioned in the assembler
3903 code, the result will not work, but GCC can't tell that.
3905 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3906 the operand number for a matching constraint. For example:
3909 asm ("cmoveq %1,%2,%[result]"
3910 : [result] "=r"(result)
3911 : "r" (test), "r"(new), "[result]"(old));
3914 Some instructions clobber specific hard registers. To describe this,
3915 write a third colon after the input operands, followed by the names of
3916 the clobbered hard registers (given as strings). Here is a realistic
3917 example for the VAX:
3920 asm volatile ("movc3 %0,%1,%2"
3922 : "g" (from), "g" (to), "g" (count)
3923 : "r0", "r1", "r2", "r3", "r4", "r5");
3926 You may not write a clobber description in a way that overlaps with an
3927 input or output operand. For example, you may not have an operand
3928 describing a register class with one member if you mention that register
3929 in the clobber list. Variables declared to live in specific registers
3930 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3931 have no part mentioned in the clobber description.
3932 There is no way for you to specify that an input
3933 operand is modified without also specifying it as an output
3934 operand. Note that if all the output operands you specify are for this
3935 purpose (and hence unused), you will then also need to specify
3936 @code{volatile} for the @code{asm} construct, as described below, to
3937 prevent GCC from deleting the @code{asm} statement as unused.
3939 If you refer to a particular hardware register from the assembler code,
3940 you will probably have to list the register after the third colon to
3941 tell the compiler the register's value is modified. In some assemblers,
3942 the register names begin with @samp{%}; to produce one @samp{%} in the
3943 assembler code, you must write @samp{%%} in the input.
3945 If your assembler instruction can alter the condition code register, add
3946 @samp{cc} to the list of clobbered registers. GCC on some machines
3947 represents the condition codes as a specific hardware register;
3948 @samp{cc} serves to name this register. On other machines, the
3949 condition code is handled differently, and specifying @samp{cc} has no
3950 effect. But it is valid no matter what the machine.
3952 If your assembler instructions access memory in an unpredictable
3953 fashion, add @samp{memory} to the list of clobbered registers. This
3954 will cause GCC to not keep memory values cached in registers across the
3955 assembler instruction and not optimize stores or loads to that memory.
3956 You will also want to add the @code{volatile} keyword if the memory
3957 affected is not listed in the inputs or outputs of the @code{asm}, as
3958 the @samp{memory} clobber does not count as a side-effect of the
3959 @code{asm}. If you know how large the accessed memory is, you can add
3960 it as input or output but if this is not known, you should add
3961 @samp{memory}. As an example, if you access ten bytes of a string, you
3962 can use a memory input like:
3965 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
3968 Note that in the following example the memory input is necessary,
3969 otherwise GCC might optimize the store to @code{x} away:
3976 asm ("magic stuff accessing an 'int' pointed to by '%1'"
3977 "=&d" (r) : "a" (y), "m" (*y));
3982 You can put multiple assembler instructions together in a single
3983 @code{asm} template, separated by the characters normally used in assembly
3984 code for the system. A combination that works in most places is a newline
3985 to break the line, plus a tab character to move to the instruction field
3986 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3987 assembler allows semicolons as a line-breaking character. Note that some
3988 assembler dialects use semicolons to start a comment.
3989 The input operands are guaranteed not to use any of the clobbered
3990 registers, and neither will the output operands' addresses, so you can
3991 read and write the clobbered registers as many times as you like. Here
3992 is an example of multiple instructions in a template; it assumes the
3993 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3996 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3998 : "g" (from), "g" (to)
4002 Unless an output operand has the @samp{&} constraint modifier, GCC
4003 may allocate it in the same register as an unrelated input operand, on
4004 the assumption the inputs are consumed before the outputs are produced.
4005 This assumption may be false if the assembler code actually consists of
4006 more than one instruction. In such a case, use @samp{&} for each output
4007 operand that may not overlap an input. @xref{Modifiers}.
4009 If you want to test the condition code produced by an assembler
4010 instruction, you must include a branch and a label in the @code{asm}
4011 construct, as follows:
4014 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4020 This assumes your assembler supports local labels, as the GNU assembler
4021 and most Unix assemblers do.
4023 Speaking of labels, jumps from one @code{asm} to another are not
4024 supported. The compiler's optimizers do not know about these jumps, and
4025 therefore they cannot take account of them when deciding how to
4028 @cindex macros containing @code{asm}
4029 Usually the most convenient way to use these @code{asm} instructions is to
4030 encapsulate them in macros that look like functions. For example,
4034 (@{ double __value, __arg = (x); \
4035 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4040 Here the variable @code{__arg} is used to make sure that the instruction
4041 operates on a proper @code{double} value, and to accept only those
4042 arguments @code{x} which can convert automatically to a @code{double}.
4044 Another way to make sure the instruction operates on the correct data
4045 type is to use a cast in the @code{asm}. This is different from using a
4046 variable @code{__arg} in that it converts more different types. For
4047 example, if the desired type were @code{int}, casting the argument to
4048 @code{int} would accept a pointer with no complaint, while assigning the
4049 argument to an @code{int} variable named @code{__arg} would warn about
4050 using a pointer unless the caller explicitly casts it.
4052 If an @code{asm} has output operands, GCC assumes for optimization
4053 purposes the instruction has no side effects except to change the output
4054 operands. This does not mean instructions with a side effect cannot be
4055 used, but you must be careful, because the compiler may eliminate them
4056 if the output operands aren't used, or move them out of loops, or
4057 replace two with one if they constitute a common subexpression. Also,
4058 if your instruction does have a side effect on a variable that otherwise
4059 appears not to change, the old value of the variable may be reused later
4060 if it happens to be found in a register.
4062 You can prevent an @code{asm} instruction from being deleted, moved
4063 significantly, or combined, by writing the keyword @code{volatile} after
4064 the @code{asm}. For example:
4067 #define get_and_set_priority(new) \
4069 asm volatile ("get_and_set_priority %0, %1" \
4070 : "=g" (__old) : "g" (new)); \
4075 If you write an @code{asm} instruction with no outputs, GCC will know
4076 the instruction has side-effects and will not delete the instruction or
4077 move it outside of loops.
4079 The @code{volatile} keyword indicates that the instruction has
4080 important side-effects. GCC will not delete a volatile @code{asm} if
4081 it is reachable. (The instruction can still be deleted if GCC can
4082 prove that control-flow will never reach the location of the
4083 instruction.) In addition, GCC will not reschedule instructions
4084 across a volatile @code{asm} instruction. For example:
4087 *(volatile int *)addr = foo;
4088 asm volatile ("eieio" : : );
4092 Assume @code{addr} contains the address of a memory mapped device
4093 register. The PowerPC @code{eieio} instruction (Enforce In-order
4094 Execution of I/O) tells the CPU to make sure that the store to that
4095 device register happens before it issues any other I/O@.
4097 Note that even a volatile @code{asm} instruction can be moved in ways
4098 that appear insignificant to the compiler, such as across jump
4099 instructions. You can't expect a sequence of volatile @code{asm}
4100 instructions to remain perfectly consecutive. If you want consecutive
4101 output, use a single @code{asm}. Also, GCC will perform some
4102 optimizations across a volatile @code{asm} instruction; GCC does not
4103 ``forget everything'' when it encounters a volatile @code{asm}
4104 instruction the way some other compilers do.
4106 An @code{asm} instruction without any operands or clobbers (an ``old
4107 style'' @code{asm}) will be treated identically to a volatile
4108 @code{asm} instruction.
4110 It is a natural idea to look for a way to give access to the condition
4111 code left by the assembler instruction. However, when we attempted to
4112 implement this, we found no way to make it work reliably. The problem
4113 is that output operands might need reloading, which would result in
4114 additional following ``store'' instructions. On most machines, these
4115 instructions would alter the condition code before there was time to
4116 test it. This problem doesn't arise for ordinary ``test'' and
4117 ``compare'' instructions because they don't have any output operands.
4119 For reasons similar to those described above, it is not possible to give
4120 an assembler instruction access to the condition code left by previous
4123 If you are writing a header file that should be includable in ISO C
4124 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4127 @subsection Size of an @code{asm}
4129 Some targets require that GCC track the size of each instruction used in
4130 order to generate correct code. Because the final length of an
4131 @code{asm} is only known by the assembler, GCC must make an estimate as
4132 to how big it will be. The estimate is formed by counting the number of
4133 statements in the pattern of the @code{asm} and multiplying that by the
4134 length of the longest instruction on that processor. Statements in the
4135 @code{asm} are identified by newline characters and whatever statement
4136 separator characters are supported by the assembler; on most processors
4137 this is the `@code{;}' character.
4139 Normally, GCC's estimate is perfectly adequate to ensure that correct
4140 code is generated, but it is possible to confuse the compiler if you use
4141 pseudo instructions or assembler macros that expand into multiple real
4142 instructions or if you use assembler directives that expand to more
4143 space in the object file than would be needed for a single instruction.
4144 If this happens then the assembler will produce a diagnostic saying that
4145 a label is unreachable.
4147 @subsection i386 floating point asm operands
4149 There are several rules on the usage of stack-like regs in
4150 asm_operands insns. These rules apply only to the operands that are
4155 Given a set of input regs that die in an asm_operands, it is
4156 necessary to know which are implicitly popped by the asm, and
4157 which must be explicitly popped by gcc.
4159 An input reg that is implicitly popped by the asm must be
4160 explicitly clobbered, unless it is constrained to match an
4164 For any input reg that is implicitly popped by an asm, it is
4165 necessary to know how to adjust the stack to compensate for the pop.
4166 If any non-popped input is closer to the top of the reg-stack than
4167 the implicitly popped reg, it would not be possible to know what the
4168 stack looked like---it's not clear how the rest of the stack ``slides
4171 All implicitly popped input regs must be closer to the top of
4172 the reg-stack than any input that is not implicitly popped.
4174 It is possible that if an input dies in an insn, reload might
4175 use the input reg for an output reload. Consider this example:
4178 asm ("foo" : "=t" (a) : "f" (b));
4181 This asm says that input B is not popped by the asm, and that
4182 the asm pushes a result onto the reg-stack, i.e., the stack is one
4183 deeper after the asm than it was before. But, it is possible that
4184 reload will think that it can use the same reg for both the input and
4185 the output, if input B dies in this insn.
4187 If any input operand uses the @code{f} constraint, all output reg
4188 constraints must use the @code{&} earlyclobber.
4190 The asm above would be written as
4193 asm ("foo" : "=&t" (a) : "f" (b));
4197 Some operands need to be in particular places on the stack. All
4198 output operands fall in this category---there is no other way to
4199 know which regs the outputs appear in unless the user indicates
4200 this in the constraints.
4202 Output operands must specifically indicate which reg an output
4203 appears in after an asm. @code{=f} is not allowed: the operand
4204 constraints must select a class with a single reg.
4207 Output operands may not be ``inserted'' between existing stack regs.
4208 Since no 387 opcode uses a read/write operand, all output operands
4209 are dead before the asm_operands, and are pushed by the asm_operands.
4210 It makes no sense to push anywhere but the top of the reg-stack.
4212 Output operands must start at the top of the reg-stack: output
4213 operands may not ``skip'' a reg.
4216 Some asm statements may need extra stack space for internal
4217 calculations. This can be guaranteed by clobbering stack registers
4218 unrelated to the inputs and outputs.
4222 Here are a couple of reasonable asms to want to write. This asm
4223 takes one input, which is internally popped, and produces two outputs.
4226 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4229 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4230 and replaces them with one output. The user must code the @code{st(1)}
4231 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4234 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4240 @section Controlling Names Used in Assembler Code
4241 @cindex assembler names for identifiers
4242 @cindex names used in assembler code
4243 @cindex identifiers, names in assembler code
4245 You can specify the name to be used in the assembler code for a C
4246 function or variable by writing the @code{asm} (or @code{__asm__})
4247 keyword after the declarator as follows:
4250 int foo asm ("myfoo") = 2;
4254 This specifies that the name to be used for the variable @code{foo} in
4255 the assembler code should be @samp{myfoo} rather than the usual
4258 On systems where an underscore is normally prepended to the name of a C
4259 function or variable, this feature allows you to define names for the
4260 linker that do not start with an underscore.
4262 It does not make sense to use this feature with a non-static local
4263 variable since such variables do not have assembler names. If you are
4264 trying to put the variable in a particular register, see @ref{Explicit
4265 Reg Vars}. GCC presently accepts such code with a warning, but will
4266 probably be changed to issue an error, rather than a warning, in the
4269 You cannot use @code{asm} in this way in a function @emph{definition}; but
4270 you can get the same effect by writing a declaration for the function
4271 before its definition and putting @code{asm} there, like this:
4274 extern func () asm ("FUNC");
4281 It is up to you to make sure that the assembler names you choose do not
4282 conflict with any other assembler symbols. Also, you must not use a
4283 register name; that would produce completely invalid assembler code. GCC
4284 does not as yet have the ability to store static variables in registers.
4285 Perhaps that will be added.
4287 @node Explicit Reg Vars
4288 @section Variables in Specified Registers
4289 @cindex explicit register variables
4290 @cindex variables in specified registers
4291 @cindex specified registers
4292 @cindex registers, global allocation
4294 GNU C allows you to put a few global variables into specified hardware
4295 registers. You can also specify the register in which an ordinary
4296 register variable should be allocated.
4300 Global register variables reserve registers throughout the program.
4301 This may be useful in programs such as programming language
4302 interpreters which have a couple of global variables that are accessed
4306 Local register variables in specific registers do not reserve the
4307 registers. The compiler's data flow analysis is capable of determining
4308 where the specified registers contain live values, and where they are
4309 available for other uses. Stores into local register variables may be deleted
4310 when they appear to be dead according to dataflow analysis. References
4311 to local register variables may be deleted or moved or simplified.
4313 These local variables are sometimes convenient for use with the extended
4314 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4315 output of the assembler instruction directly into a particular register.
4316 (This will work provided the register you specify fits the constraints
4317 specified for that operand in the @code{asm}.)
4325 @node Global Reg Vars
4326 @subsection Defining Global Register Variables
4327 @cindex global register variables
4328 @cindex registers, global variables in
4330 You can define a global register variable in GNU C like this:
4333 register int *foo asm ("a5");
4337 Here @code{a5} is the name of the register which should be used. Choose a
4338 register which is normally saved and restored by function calls on your
4339 machine, so that library routines will not clobber it.
4341 Naturally the register name is cpu-dependent, so you would need to
4342 conditionalize your program according to cpu type. The register
4343 @code{a5} would be a good choice on a 68000 for a variable of pointer
4344 type. On machines with register windows, be sure to choose a ``global''
4345 register that is not affected magically by the function call mechanism.
4347 In addition, operating systems on one type of cpu may differ in how they
4348 name the registers; then you would need additional conditionals. For
4349 example, some 68000 operating systems call this register @code{%a5}.
4351 Eventually there may be a way of asking the compiler to choose a register
4352 automatically, but first we need to figure out how it should choose and
4353 how to enable you to guide the choice. No solution is evident.
4355 Defining a global register variable in a certain register reserves that
4356 register entirely for this use, at least within the current compilation.
4357 The register will not be allocated for any other purpose in the functions
4358 in the current compilation. The register will not be saved and restored by
4359 these functions. Stores into this register are never deleted even if they
4360 would appear to be dead, but references may be deleted or moved or
4363 It is not safe to access the global register variables from signal
4364 handlers, or from more than one thread of control, because the system
4365 library routines may temporarily use the register for other things (unless
4366 you recompile them specially for the task at hand).
4368 @cindex @code{qsort}, and global register variables
4369 It is not safe for one function that uses a global register variable to
4370 call another such function @code{foo} by way of a third function
4371 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4372 different source file in which the variable wasn't declared). This is
4373 because @code{lose} might save the register and put some other value there.
4374 For example, you can't expect a global register variable to be available in
4375 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4376 might have put something else in that register. (If you are prepared to
4377 recompile @code{qsort} with the same global register variable, you can
4378 solve this problem.)
4380 If you want to recompile @code{qsort} or other source files which do not
4381 actually use your global register variable, so that they will not use that
4382 register for any other purpose, then it suffices to specify the compiler
4383 option @option{-ffixed-@var{reg}}. You need not actually add a global
4384 register declaration to their source code.
4386 A function which can alter the value of a global register variable cannot
4387 safely be called from a function compiled without this variable, because it
4388 could clobber the value the caller expects to find there on return.
4389 Therefore, the function which is the entry point into the part of the
4390 program that uses the global register variable must explicitly save and
4391 restore the value which belongs to its caller.
4393 @cindex register variable after @code{longjmp}
4394 @cindex global register after @code{longjmp}
4395 @cindex value after @code{longjmp}
4398 On most machines, @code{longjmp} will restore to each global register
4399 variable the value it had at the time of the @code{setjmp}. On some
4400 machines, however, @code{longjmp} will not change the value of global
4401 register variables. To be portable, the function that called @code{setjmp}
4402 should make other arrangements to save the values of the global register
4403 variables, and to restore them in a @code{longjmp}. This way, the same
4404 thing will happen regardless of what @code{longjmp} does.
4406 All global register variable declarations must precede all function
4407 definitions. If such a declaration could appear after function
4408 definitions, the declaration would be too late to prevent the register from
4409 being used for other purposes in the preceding functions.
4411 Global register variables may not have initial values, because an
4412 executable file has no means to supply initial contents for a register.
4414 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4415 registers, but certain library functions, such as @code{getwd}, as well
4416 as the subroutines for division and remainder, modify g3 and g4. g1 and
4417 g2 are local temporaries.
4419 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4420 Of course, it will not do to use more than a few of those.
4422 @node Local Reg Vars
4423 @subsection Specifying Registers for Local Variables
4424 @cindex local variables, specifying registers
4425 @cindex specifying registers for local variables
4426 @cindex registers for local variables
4428 You can define a local register variable with a specified register
4432 register int *foo asm ("a5");
4436 Here @code{a5} is the name of the register which should be used. Note
4437 that this is the same syntax used for defining global register
4438 variables, but for a local variable it would appear within a function.
4440 Naturally the register name is cpu-dependent, but this is not a
4441 problem, since specific registers are most often useful with explicit
4442 assembler instructions (@pxref{Extended Asm}). Both of these things
4443 generally require that you conditionalize your program according to
4446 In addition, operating systems on one type of cpu may differ in how they
4447 name the registers; then you would need additional conditionals. For
4448 example, some 68000 operating systems call this register @code{%a5}.
4450 Defining such a register variable does not reserve the register; it
4451 remains available for other uses in places where flow control determines
4452 the variable's value is not live.
4454 This option does not guarantee that GCC will generate code that has
4455 this variable in the register you specify at all times. You may not
4456 code an explicit reference to this register in an @code{asm} statement
4457 and assume it will always refer to this variable.
4459 Stores into local register variables may be deleted when they appear to be dead
4460 according to dataflow analysis. References to local register variables may
4461 be deleted or moved or simplified.
4463 @node Alternate Keywords
4464 @section Alternate Keywords
4465 @cindex alternate keywords
4466 @cindex keywords, alternate
4468 @option{-ansi} and the various @option{-std} options disable certain
4469 keywords. This causes trouble when you want to use GNU C extensions, or
4470 a general-purpose header file that should be usable by all programs,
4471 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4472 @code{inline} are not available in programs compiled with
4473 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4474 program compiled with @option{-std=c99}). The ISO C99 keyword
4475 @code{restrict} is only available when @option{-std=gnu99} (which will
4476 eventually be the default) or @option{-std=c99} (or the equivalent
4477 @option{-std=iso9899:1999}) is used.
4479 The way to solve these problems is to put @samp{__} at the beginning and
4480 end of each problematical keyword. For example, use @code{__asm__}
4481 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4483 Other C compilers won't accept these alternative keywords; if you want to
4484 compile with another compiler, you can define the alternate keywords as
4485 macros to replace them with the customary keywords. It looks like this:
4493 @findex __extension__
4495 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4497 prevent such warnings within one expression by writing
4498 @code{__extension__} before the expression. @code{__extension__} has no
4499 effect aside from this.
4501 @node Incomplete Enums
4502 @section Incomplete @code{enum} Types
4504 You can define an @code{enum} tag without specifying its possible values.
4505 This results in an incomplete type, much like what you get if you write
4506 @code{struct foo} without describing the elements. A later declaration
4507 which does specify the possible values completes the type.
4509 You can't allocate variables or storage using the type while it is
4510 incomplete. However, you can work with pointers to that type.
4512 This extension may not be very useful, but it makes the handling of
4513 @code{enum} more consistent with the way @code{struct} and @code{union}
4516 This extension is not supported by GNU C++.
4518 @node Function Names
4519 @section Function Names as Strings
4520 @cindex @code{__func__} identifier
4521 @cindex @code{__FUNCTION__} identifier
4522 @cindex @code{__PRETTY_FUNCTION__} identifier
4524 GCC provides three magic variables which hold the name of the current
4525 function, as a string. The first of these is @code{__func__}, which
4526 is part of the C99 standard:
4529 The identifier @code{__func__} is implicitly declared by the translator
4530 as if, immediately following the opening brace of each function
4531 definition, the declaration
4534 static const char __func__[] = "function-name";
4537 appeared, where function-name is the name of the lexically-enclosing
4538 function. This name is the unadorned name of the function.
4541 @code{__FUNCTION__} is another name for @code{__func__}. Older
4542 versions of GCC recognize only this name. However, it is not
4543 standardized. For maximum portability, we recommend you use
4544 @code{__func__}, but provide a fallback definition with the
4548 #if __STDC_VERSION__ < 199901L
4550 # define __func__ __FUNCTION__
4552 # define __func__ "<unknown>"
4557 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4558 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4559 the type signature of the function as well as its bare name. For
4560 example, this program:
4564 extern int printf (char *, ...);
4571 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4572 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4590 __PRETTY_FUNCTION__ = void a::sub(int)
4593 These identifiers are not preprocessor macros. In GCC 3.3 and
4594 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4595 were treated as string literals; they could be used to initialize
4596 @code{char} arrays, and they could be concatenated with other string
4597 literals. GCC 3.4 and later treat them as variables, like
4598 @code{__func__}. In C++, @code{__FUNCTION__} and
4599 @code{__PRETTY_FUNCTION__} have always been variables.
4601 @node Return Address
4602 @section Getting the Return or Frame Address of a Function
4604 These functions may be used to get information about the callers of a
4607 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4608 This function returns the return address of the current function, or of
4609 one of its callers. The @var{level} argument is number of frames to
4610 scan up the call stack. A value of @code{0} yields the return address
4611 of the current function, a value of @code{1} yields the return address
4612 of the caller of the current function, and so forth. When inlining
4613 the expected behavior is that the function will return the address of
4614 the function that will be returned to. To work around this behavior use
4615 the @code{noinline} function attribute.
4617 The @var{level} argument must be a constant integer.
4619 On some machines it may be impossible to determine the return address of
4620 any function other than the current one; in such cases, or when the top
4621 of the stack has been reached, this function will return @code{0} or a
4622 random value. In addition, @code{__builtin_frame_address} may be used
4623 to determine if the top of the stack has been reached.
4625 This function should only be used with a nonzero argument for debugging
4629 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4630 This function is similar to @code{__builtin_return_address}, but it
4631 returns the address of the function frame rather than the return address
4632 of the function. Calling @code{__builtin_frame_address} with a value of
4633 @code{0} yields the frame address of the current function, a value of
4634 @code{1} yields the frame address of the caller of the current function,
4637 The frame is the area on the stack which holds local variables and saved
4638 registers. The frame address is normally the address of the first word
4639 pushed on to the stack by the function. However, the exact definition
4640 depends upon the processor and the calling convention. If the processor
4641 has a dedicated frame pointer register, and the function has a frame,
4642 then @code{__builtin_frame_address} will return the value of the frame
4645 On some machines it may be impossible to determine the frame address of
4646 any function other than the current one; in such cases, or when the top
4647 of the stack has been reached, this function will return @code{0} if
4648 the first frame pointer is properly initialized by the startup code.
4650 This function should only be used with a nonzero argument for debugging
4654 @node Vector Extensions
4655 @section Using vector instructions through built-in functions
4657 On some targets, the instruction set contains SIMD vector instructions that
4658 operate on multiple values contained in one large register at the same time.
4659 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4662 The first step in using these extensions is to provide the necessary data
4663 types. This should be done using an appropriate @code{typedef}:
4666 typedef int v4si __attribute__ ((vector_size (16)));
4669 The @code{int} type specifies the base type, while the attribute specifies
4670 the vector size for the variable, measured in bytes. For example, the
4671 declaration above causes the compiler to set the mode for the @code{v4si}
4672 type to be 16 bytes wide and divided into @code{int} sized units. For
4673 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4674 corresponding mode of @code{foo} will be @acronym{V4SI}.
4676 The @code{vector_size} attribute is only applicable to integral and
4677 float scalars, although arrays, pointers, and function return values
4678 are allowed in conjunction with this construct.
4680 All the basic integer types can be used as base types, both as signed
4681 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4682 @code{long long}. In addition, @code{float} and @code{double} can be
4683 used to build floating-point vector types.
4685 Specifying a combination that is not valid for the current architecture
4686 will cause GCC to synthesize the instructions using a narrower mode.
4687 For example, if you specify a variable of type @code{V4SI} and your
4688 architecture does not allow for this specific SIMD type, GCC will
4689 produce code that uses 4 @code{SIs}.
4691 The types defined in this manner can be used with a subset of normal C
4692 operations. Currently, GCC will allow using the following operators
4693 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4695 The operations behave like C++ @code{valarrays}. Addition is defined as
4696 the addition of the corresponding elements of the operands. For
4697 example, in the code below, each of the 4 elements in @var{a} will be
4698 added to the corresponding 4 elements in @var{b} and the resulting
4699 vector will be stored in @var{c}.
4702 typedef int v4si __attribute__ ((vector_size (16)));
4709 Subtraction, multiplication, division, and the logical operations
4710 operate in a similar manner. Likewise, the result of using the unary
4711 minus or complement operators on a vector type is a vector whose
4712 elements are the negative or complemented values of the corresponding
4713 elements in the operand.
4715 You can declare variables and use them in function calls and returns, as
4716 well as in assignments and some casts. You can specify a vector type as
4717 a return type for a function. Vector types can also be used as function
4718 arguments. It is possible to cast from one vector type to another,
4719 provided they are of the same size (in fact, you can also cast vectors
4720 to and from other datatypes of the same size).
4722 You cannot operate between vectors of different lengths or different
4723 signedness without a cast.
4725 A port that supports hardware vector operations, usually provides a set
4726 of built-in functions that can be used to operate on vectors. For
4727 example, a function to add two vectors and multiply the result by a
4728 third could look like this:
4731 v4si f (v4si a, v4si b, v4si c)
4733 v4si tmp = __builtin_addv4si (a, b);
4734 return __builtin_mulv4si (tmp, c);
4741 @findex __builtin_offsetof
4743 GCC implements for both C and C++ a syntactic extension to implement
4744 the @code{offsetof} macro.
4748 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4750 offsetof_member_designator:
4752 | offsetof_member_designator "." @code{identifier}
4753 | offsetof_member_designator "[" @code{expr} "]"
4756 This extension is sufficient such that
4759 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4762 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4763 may be dependent. In either case, @var{member} may consist of a single
4764 identifier, or a sequence of member accesses and array references.
4766 @node Other Builtins
4767 @section Other built-in functions provided by GCC
4768 @cindex built-in functions
4769 @findex __builtin_isgreater
4770 @findex __builtin_isgreaterequal
4771 @findex __builtin_isless
4772 @findex __builtin_islessequal
4773 @findex __builtin_islessgreater
4774 @findex __builtin_isunordered
4929 @findex fprintf_unlocked
4931 @findex fputs_unlocked
5041 @findex printf_unlocked
5070 @findex significandf
5071 @findex significandl
5138 GCC provides a large number of built-in functions other than the ones
5139 mentioned above. Some of these are for internal use in the processing
5140 of exceptions or variable-length argument lists and will not be
5141 documented here because they may change from time to time; we do not
5142 recommend general use of these functions.
5144 The remaining functions are provided for optimization purposes.
5146 @opindex fno-builtin
5147 GCC includes built-in versions of many of the functions in the standard
5148 C library. The versions prefixed with @code{__builtin_} will always be
5149 treated as having the same meaning as the C library function even if you
5150 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5151 Many of these functions are only optimized in certain cases; if they are
5152 not optimized in a particular case, a call to the library function will
5157 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5158 @option{-std=c99}), the functions
5159 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5160 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5161 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5162 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5163 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5164 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5165 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5166 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5167 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5168 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5169 @code{significandf}, @code{significandl}, @code{significand},
5170 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5171 @code{strdup}, @code{strfmon}, @code{toascii}, @code{y0f}, @code{y0l},
5172 @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
5174 may be handled as built-in functions.
5175 All these functions have corresponding versions
5176 prefixed with @code{__builtin_}, which may be used even in strict C89
5179 The ISO C99 functions
5180 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5181 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5182 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5183 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5184 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5185 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5186 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5187 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5188 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5189 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5190 @code{cimagl}, @code{cimag}, @code{conjf}, @code{conjl}, @code{conj},
5191 @code{copysignf}, @code{copysignl}, @code{copysign}, @code{cpowf},
5192 @code{cpowl}, @code{cpow}, @code{cprojf}, @code{cprojl}, @code{cproj},
5193 @code{crealf}, @code{creall}, @code{creal}, @code{csinf}, @code{csinhf},
5194 @code{csinhl}, @code{csinh}, @code{csinl}, @code{csin}, @code{csqrtf},
5195 @code{csqrtl}, @code{csqrt}, @code{ctanf}, @code{ctanhf}, @code{ctanhl},
5196 @code{ctanh}, @code{ctanl}, @code{ctan}, @code{erfcf}, @code{erfcl},
5197 @code{erfc}, @code{erff}, @code{erfl}, @code{erf}, @code{exp2f},
5198 @code{exp2l}, @code{exp2}, @code{expm1f}, @code{expm1l}, @code{expm1},
5199 @code{fdimf}, @code{fdiml}, @code{fdim}, @code{fmaf}, @code{fmal},
5200 @code{fmaxf}, @code{fmaxl}, @code{fmax}, @code{fma}, @code{fminf},
5201 @code{fminl}, @code{fmin}, @code{hypotf}, @code{hypotl}, @code{hypot},
5202 @code{ilogbf}, @code{ilogbl}, @code{ilogb}, @code{imaxabs},
5203 @code{isblank}, @code{iswblank}, @code{lgammaf}, @code{lgammal},
5204 @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5205 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5206 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5207 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5208 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5209 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5210 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5211 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5212 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5213 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5214 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5215 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5216 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5217 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5218 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5219 are handled as built-in functions
5220 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5222 There are also built-in versions of the ISO C99 functions
5223 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5224 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5225 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5226 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5227 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5228 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5229 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5230 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5231 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5232 that are recognized in any mode since ISO C90 reserves these names for
5233 the purpose to which ISO C99 puts them. All these functions have
5234 corresponding versions prefixed with @code{__builtin_}.
5236 The ISO C94 functions
5237 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5238 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5239 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5241 are handled as built-in functions
5242 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5244 The ISO C90 functions
5245 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5246 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5247 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5248 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5249 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5250 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5251 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5252 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5253 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5254 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5255 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5256 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5257 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5258 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5259 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5260 @code{vprintf} and @code{vsprintf}
5261 are all recognized as built-in functions unless
5262 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5263 is specified for an individual function). All of these functions have
5264 corresponding versions prefixed with @code{__builtin_}.
5266 GCC provides built-in versions of the ISO C99 floating point comparison
5267 macros that avoid raising exceptions for unordered operands. They have
5268 the same names as the standard macros ( @code{isgreater},
5269 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5270 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5271 prefixed. We intend for a library implementor to be able to simply
5272 @code{#define} each standard macro to its built-in equivalent.
5274 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5276 You can use the built-in function @code{__builtin_types_compatible_p} to
5277 determine whether two types are the same.
5279 This built-in function returns 1 if the unqualified versions of the
5280 types @var{type1} and @var{type2} (which are types, not expressions) are
5281 compatible, 0 otherwise. The result of this built-in function can be
5282 used in integer constant expressions.
5284 This built-in function ignores top level qualifiers (e.g., @code{const},
5285 @code{volatile}). For example, @code{int} is equivalent to @code{const
5288 The type @code{int[]} and @code{int[5]} are compatible. On the other
5289 hand, @code{int} and @code{char *} are not compatible, even if the size
5290 of their types, on the particular architecture are the same. Also, the
5291 amount of pointer indirection is taken into account when determining
5292 similarity. Consequently, @code{short *} is not similar to
5293 @code{short **}. Furthermore, two types that are typedefed are
5294 considered compatible if their underlying types are compatible.
5296 An @code{enum} type is not considered to be compatible with another
5297 @code{enum} type even if both are compatible with the same integer
5298 type; this is what the C standard specifies.
5299 For example, @code{enum @{foo, bar@}} is not similar to
5300 @code{enum @{hot, dog@}}.
5302 You would typically use this function in code whose execution varies
5303 depending on the arguments' types. For example:
5309 if (__builtin_types_compatible_p (typeof (x), long double)) \
5310 tmp = foo_long_double (tmp); \
5311 else if (__builtin_types_compatible_p (typeof (x), double)) \
5312 tmp = foo_double (tmp); \
5313 else if (__builtin_types_compatible_p (typeof (x), float)) \
5314 tmp = foo_float (tmp); \
5321 @emph{Note:} This construct is only available for C.
5325 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5327 You can use the built-in function @code{__builtin_choose_expr} to
5328 evaluate code depending on the value of a constant expression. This
5329 built-in function returns @var{exp1} if @var{const_exp}, which is a
5330 constant expression that must be able to be determined at compile time,
5331 is nonzero. Otherwise it returns 0.
5333 This built-in function is analogous to the @samp{? :} operator in C,
5334 except that the expression returned has its type unaltered by promotion
5335 rules. Also, the built-in function does not evaluate the expression
5336 that was not chosen. For example, if @var{const_exp} evaluates to true,
5337 @var{exp2} is not evaluated even if it has side-effects.
5339 This built-in function can return an lvalue if the chosen argument is an
5342 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5343 type. Similarly, if @var{exp2} is returned, its return type is the same
5350 __builtin_choose_expr ( \
5351 __builtin_types_compatible_p (typeof (x), double), \
5353 __builtin_choose_expr ( \
5354 __builtin_types_compatible_p (typeof (x), float), \
5356 /* @r{The void expression results in a compile-time error} \
5357 @r{when assigning the result to something.} */ \
5361 @emph{Note:} This construct is only available for C. Furthermore, the
5362 unused expression (@var{exp1} or @var{exp2} depending on the value of
5363 @var{const_exp}) may still generate syntax errors. This may change in
5368 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5369 You can use the built-in function @code{__builtin_constant_p} to
5370 determine if a value is known to be constant at compile-time and hence
5371 that GCC can perform constant-folding on expressions involving that
5372 value. The argument of the function is the value to test. The function
5373 returns the integer 1 if the argument is known to be a compile-time
5374 constant and 0 if it is not known to be a compile-time constant. A
5375 return of 0 does not indicate that the value is @emph{not} a constant,
5376 but merely that GCC cannot prove it is a constant with the specified
5377 value of the @option{-O} option.
5379 You would typically use this function in an embedded application where
5380 memory was a critical resource. If you have some complex calculation,
5381 you may want it to be folded if it involves constants, but need to call
5382 a function if it does not. For example:
5385 #define Scale_Value(X) \
5386 (__builtin_constant_p (X) \
5387 ? ((X) * SCALE + OFFSET) : Scale (X))
5390 You may use this built-in function in either a macro or an inline
5391 function. However, if you use it in an inlined function and pass an
5392 argument of the function as the argument to the built-in, GCC will
5393 never return 1 when you call the inline function with a string constant
5394 or compound literal (@pxref{Compound Literals}) and will not return 1
5395 when you pass a constant numeric value to the inline function unless you
5396 specify the @option{-O} option.
5398 You may also use @code{__builtin_constant_p} in initializers for static
5399 data. For instance, you can write
5402 static const int table[] = @{
5403 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5409 This is an acceptable initializer even if @var{EXPRESSION} is not a
5410 constant expression. GCC must be more conservative about evaluating the
5411 built-in in this case, because it has no opportunity to perform
5414 Previous versions of GCC did not accept this built-in in data
5415 initializers. The earliest version where it is completely safe is
5419 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5420 @opindex fprofile-arcs
5421 You may use @code{__builtin_expect} to provide the compiler with
5422 branch prediction information. In general, you should prefer to
5423 use actual profile feedback for this (@option{-fprofile-arcs}), as
5424 programmers are notoriously bad at predicting how their programs
5425 actually perform. However, there are applications in which this
5426 data is hard to collect.
5428 The return value is the value of @var{exp}, which should be an
5429 integral expression. The value of @var{c} must be a compile-time
5430 constant. The semantics of the built-in are that it is expected
5431 that @var{exp} == @var{c}. For example:
5434 if (__builtin_expect (x, 0))
5439 would indicate that we do not expect to call @code{foo}, since
5440 we expect @code{x} to be zero. Since you are limited to integral
5441 expressions for @var{exp}, you should use constructions such as
5444 if (__builtin_expect (ptr != NULL, 1))
5449 when testing pointer or floating-point values.
5452 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5453 This function is used to minimize cache-miss latency by moving data into
5454 a cache before it is accessed.
5455 You can insert calls to @code{__builtin_prefetch} into code for which
5456 you know addresses of data in memory that is likely to be accessed soon.
5457 If the target supports them, data prefetch instructions will be generated.
5458 If the prefetch is done early enough before the access then the data will
5459 be in the cache by the time it is accessed.
5461 The value of @var{addr} is the address of the memory to prefetch.
5462 There are two optional arguments, @var{rw} and @var{locality}.
5463 The value of @var{rw} is a compile-time constant one or zero; one
5464 means that the prefetch is preparing for a write to the memory address
5465 and zero, the default, means that the prefetch is preparing for a read.
5466 The value @var{locality} must be a compile-time constant integer between
5467 zero and three. A value of zero means that the data has no temporal
5468 locality, so it need not be left in the cache after the access. A value
5469 of three means that the data has a high degree of temporal locality and
5470 should be left in all levels of cache possible. Values of one and two
5471 mean, respectively, a low or moderate degree of temporal locality. The
5475 for (i = 0; i < n; i++)
5478 __builtin_prefetch (&a[i+j], 1, 1);
5479 __builtin_prefetch (&b[i+j], 0, 1);
5484 Data prefetch does not generate faults if @var{addr} is invalid, but
5485 the address expression itself must be valid. For example, a prefetch
5486 of @code{p->next} will not fault if @code{p->next} is not a valid
5487 address, but evaluation will fault if @code{p} is not a valid address.
5489 If the target does not support data prefetch, the address expression
5490 is evaluated if it includes side effects but no other code is generated
5491 and GCC does not issue a warning.
5494 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5495 Returns a positive infinity, if supported by the floating-point format,
5496 else @code{DBL_MAX}. This function is suitable for implementing the
5497 ISO C macro @code{HUGE_VAL}.
5500 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5501 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5504 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5505 Similar to @code{__builtin_huge_val}, except the return
5506 type is @code{long double}.
5509 @deftypefn {Built-in Function} double __builtin_inf (void)
5510 Similar to @code{__builtin_huge_val}, except a warning is generated
5511 if the target floating-point format does not support infinities.
5512 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5515 @deftypefn {Built-in Function} float __builtin_inff (void)
5516 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5519 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5520 Similar to @code{__builtin_inf}, except the return
5521 type is @code{long double}.
5524 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5525 This is an implementation of the ISO C99 function @code{nan}.
5527 Since ISO C99 defines this function in terms of @code{strtod}, which we
5528 do not implement, a description of the parsing is in order. The string
5529 is parsed as by @code{strtol}; that is, the base is recognized by
5530 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5531 in the significand such that the least significant bit of the number
5532 is at the least significant bit of the significand. The number is
5533 truncated to fit the significand field provided. The significand is
5534 forced to be a quiet NaN.
5536 This function, if given a string literal, is evaluated early enough
5537 that it is considered a compile-time constant.
5540 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5541 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5544 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5545 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5548 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5549 Similar to @code{__builtin_nan}, except the significand is forced
5550 to be a signaling NaN. The @code{nans} function is proposed by
5551 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
5554 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5555 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5558 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5559 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5562 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5563 Returns one plus the index of the least significant 1-bit of @var{x}, or
5564 if @var{x} is zero, returns zero.
5567 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5568 Returns the number of leading 0-bits in @var{x}, starting at the most
5569 significant bit position. If @var{x} is 0, the result is undefined.
5572 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5573 Returns the number of trailing 0-bits in @var{x}, starting at the least
5574 significant bit position. If @var{x} is 0, the result is undefined.
5577 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5578 Returns the number of 1-bits in @var{x}.
5581 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5582 Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5586 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5587 Similar to @code{__builtin_ffs}, except the argument type is
5588 @code{unsigned long}.
5591 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5592 Similar to @code{__builtin_clz}, except the argument type is
5593 @code{unsigned long}.
5596 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5597 Similar to @code{__builtin_ctz}, except the argument type is
5598 @code{unsigned long}.
5601 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5602 Similar to @code{__builtin_popcount}, except the argument type is
5603 @code{unsigned long}.
5606 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5607 Similar to @code{__builtin_parity}, except the argument type is
5608 @code{unsigned long}.
5611 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5612 Similar to @code{__builtin_ffs}, except the argument type is
5613 @code{unsigned long long}.
5616 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5617 Similar to @code{__builtin_clz}, except the argument type is
5618 @code{unsigned long long}.
5621 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5622 Similar to @code{__builtin_ctz}, except the argument type is
5623 @code{unsigned long long}.
5626 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5627 Similar to @code{__builtin_popcount}, except the argument type is
5628 @code{unsigned long long}.
5631 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5632 Similar to @code{__builtin_parity}, except the argument type is
5633 @code{unsigned long long}.
5637 @node Target Builtins
5638 @section Built-in Functions Specific to Particular Target Machines
5640 On some target machines, GCC supports many built-in functions specific
5641 to those machines. Generally these generate calls to specific machine
5642 instructions, but allow the compiler to schedule those calls.
5645 * Alpha Built-in Functions::
5646 * ARM Built-in Functions::
5647 * X86 Built-in Functions::
5648 * PowerPC AltiVec Built-in Functions::
5651 @node Alpha Built-in Functions
5652 @subsection Alpha Built-in Functions
5654 These built-in functions are available for the Alpha family of
5655 processors, depending on the command-line switches used.
5657 The following built-in functions are always available. They
5658 all generate the machine instruction that is part of the name.
5661 long __builtin_alpha_implver (void)
5662 long __builtin_alpha_rpcc (void)
5663 long __builtin_alpha_amask (long)
5664 long __builtin_alpha_cmpbge (long, long)
5665 long __builtin_alpha_extbl (long, long)
5666 long __builtin_alpha_extwl (long, long)
5667 long __builtin_alpha_extll (long, long)
5668 long __builtin_alpha_extql (long, long)
5669 long __builtin_alpha_extwh (long, long)
5670 long __builtin_alpha_extlh (long, long)
5671 long __builtin_alpha_extqh (long, long)
5672 long __builtin_alpha_insbl (long, long)
5673 long __builtin_alpha_inswl (long, long)
5674 long __builtin_alpha_insll (long, long)
5675 long __builtin_alpha_insql (long, long)
5676 long __builtin_alpha_inswh (long, long)
5677 long __builtin_alpha_inslh (long, long)
5678 long __builtin_alpha_insqh (long, long)
5679 long __builtin_alpha_mskbl (long, long)
5680 long __builtin_alpha_mskwl (long, long)
5681 long __builtin_alpha_mskll (long, long)
5682 long __builtin_alpha_mskql (long, long)
5683 long __builtin_alpha_mskwh (long, long)
5684 long __builtin_alpha_msklh (long, long)
5685 long __builtin_alpha_mskqh (long, long)
5686 long __builtin_alpha_umulh (long, long)
5687 long __builtin_alpha_zap (long, long)
5688 long __builtin_alpha_zapnot (long, long)
5691 The following built-in functions are always with @option{-mmax}
5692 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5693 later. They all generate the machine instruction that is part
5697 long __builtin_alpha_pklb (long)
5698 long __builtin_alpha_pkwb (long)
5699 long __builtin_alpha_unpkbl (long)
5700 long __builtin_alpha_unpkbw (long)
5701 long __builtin_alpha_minub8 (long, long)
5702 long __builtin_alpha_minsb8 (long, long)
5703 long __builtin_alpha_minuw4 (long, long)
5704 long __builtin_alpha_minsw4 (long, long)
5705 long __builtin_alpha_maxub8 (long, long)
5706 long __builtin_alpha_maxsb8 (long, long)
5707 long __builtin_alpha_maxuw4 (long, long)
5708 long __builtin_alpha_maxsw4 (long, long)
5709 long __builtin_alpha_perr (long, long)
5712 The following built-in functions are always with @option{-mcix}
5713 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5714 later. They all generate the machine instruction that is part
5718 long __builtin_alpha_cttz (long)
5719 long __builtin_alpha_ctlz (long)
5720 long __builtin_alpha_ctpop (long)
5723 The following builtins are available on systems that use the OSF/1
5724 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5725 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5726 @code{rdval} and @code{wrval}.
5729 void *__builtin_thread_pointer (void)
5730 void __builtin_set_thread_pointer (void *)
5733 @node ARM Built-in Functions
5734 @subsection ARM Built-in Functions
5736 These built-in functions are available for the ARM family of
5737 processors, when the @option{-mcpu=iwmmxt} switch is used:
5740 typedef int v2si __attribute__ ((vector_size (8)));
5741 typedef short v4hi __attribute__ ((vector_size (8)));
5742 typedef char v8qi __attribute__ ((vector_size (8)));
5744 int __builtin_arm_getwcx (int)
5745 void __builtin_arm_setwcx (int, int)
5746 int __builtin_arm_textrmsb (v8qi, int)
5747 int __builtin_arm_textrmsh (v4hi, int)
5748 int __builtin_arm_textrmsw (v2si, int)
5749 int __builtin_arm_textrmub (v8qi, int)
5750 int __builtin_arm_textrmuh (v4hi, int)
5751 int __builtin_arm_textrmuw (v2si, int)
5752 v8qi __builtin_arm_tinsrb (v8qi, int)
5753 v4hi __builtin_arm_tinsrh (v4hi, int)
5754 v2si __builtin_arm_tinsrw (v2si, int)
5755 long long __builtin_arm_tmia (long long, int, int)
5756 long long __builtin_arm_tmiabb (long long, int, int)
5757 long long __builtin_arm_tmiabt (long long, int, int)
5758 long long __builtin_arm_tmiaph (long long, int, int)
5759 long long __builtin_arm_tmiatb (long long, int, int)
5760 long long __builtin_arm_tmiatt (long long, int, int)
5761 int __builtin_arm_tmovmskb (v8qi)
5762 int __builtin_arm_tmovmskh (v4hi)
5763 int __builtin_arm_tmovmskw (v2si)
5764 long long __builtin_arm_waccb (v8qi)
5765 long long __builtin_arm_wacch (v4hi)
5766 long long __builtin_arm_waccw (v2si)
5767 v8qi __builtin_arm_waddb (v8qi, v8qi)
5768 v8qi __builtin_arm_waddbss (v8qi, v8qi)
5769 v8qi __builtin_arm_waddbus (v8qi, v8qi)
5770 v4hi __builtin_arm_waddh (v4hi, v4hi)
5771 v4hi __builtin_arm_waddhss (v4hi, v4hi)
5772 v4hi __builtin_arm_waddhus (v4hi, v4hi)
5773 v2si __builtin_arm_waddw (v2si, v2si)
5774 v2si __builtin_arm_waddwss (v2si, v2si)
5775 v2si __builtin_arm_waddwus (v2si, v2si)
5776 v8qi __builtin_arm_walign (v8qi, v8qi, int)
5777 long long __builtin_arm_wand(long long, long long)
5778 long long __builtin_arm_wandn (long long, long long)
5779 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5780 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5781 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5782 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5783 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5784 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5785 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5786 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5787 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5788 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5789 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5790 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5791 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5792 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5793 long long __builtin_arm_wmacsz (v4hi, v4hi)
5794 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5795 long long __builtin_arm_wmacuz (v4hi, v4hi)
5796 v4hi __builtin_arm_wmadds (v4hi, v4hi)
5797 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5798 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5799 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5800 v2si __builtin_arm_wmaxsw (v2si, v2si)
5801 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5802 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5803 v2si __builtin_arm_wmaxuw (v2si, v2si)
5804 v8qi __builtin_arm_wminsb (v8qi, v8qi)
5805 v4hi __builtin_arm_wminsh (v4hi, v4hi)
5806 v2si __builtin_arm_wminsw (v2si, v2si)
5807 v8qi __builtin_arm_wminub (v8qi, v8qi)
5808 v4hi __builtin_arm_wminuh (v4hi, v4hi)
5809 v2si __builtin_arm_wminuw (v2si, v2si)
5810 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5811 v4hi __builtin_arm_wmulul (v4hi, v4hi)
5812 v4hi __builtin_arm_wmulum (v4hi, v4hi)
5813 long long __builtin_arm_wor (long long, long long)
5814 v2si __builtin_arm_wpackdss (long long, long long)
5815 v2si __builtin_arm_wpackdus (long long, long long)
5816 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5817 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5818 v4hi __builtin_arm_wpackwss (v2si, v2si)
5819 v4hi __builtin_arm_wpackwus (v2si, v2si)
5820 long long __builtin_arm_wrord (long long, long long)
5821 long long __builtin_arm_wrordi (long long, int)
5822 v4hi __builtin_arm_wrorh (v4hi, long long)
5823 v4hi __builtin_arm_wrorhi (v4hi, int)
5824 v2si __builtin_arm_wrorw (v2si, long long)
5825 v2si __builtin_arm_wrorwi (v2si, int)
5826 v2si __builtin_arm_wsadb (v8qi, v8qi)
5827 v2si __builtin_arm_wsadbz (v8qi, v8qi)
5828 v2si __builtin_arm_wsadh (v4hi, v4hi)
5829 v2si __builtin_arm_wsadhz (v4hi, v4hi)
5830 v4hi __builtin_arm_wshufh (v4hi, int)
5831 long long __builtin_arm_wslld (long long, long long)
5832 long long __builtin_arm_wslldi (long long, int)
5833 v4hi __builtin_arm_wsllh (v4hi, long long)
5834 v4hi __builtin_arm_wsllhi (v4hi, int)
5835 v2si __builtin_arm_wsllw (v2si, long long)
5836 v2si __builtin_arm_wsllwi (v2si, int)
5837 long long __builtin_arm_wsrad (long long, long long)
5838 long long __builtin_arm_wsradi (long long, int)
5839 v4hi __builtin_arm_wsrah (v4hi, long long)
5840 v4hi __builtin_arm_wsrahi (v4hi, int)
5841 v2si __builtin_arm_wsraw (v2si, long long)
5842 v2si __builtin_arm_wsrawi (v2si, int)
5843 long long __builtin_arm_wsrld (long long, long long)
5844 long long __builtin_arm_wsrldi (long long, int)
5845 v4hi __builtin_arm_wsrlh (v4hi, long long)
5846 v4hi __builtin_arm_wsrlhi (v4hi, int)
5847 v2si __builtin_arm_wsrlw (v2si, long long)
5848 v2si __builtin_arm_wsrlwi (v2si, int)
5849 v8qi __builtin_arm_wsubb (v8qi, v8qi)
5850 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5851 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5852 v4hi __builtin_arm_wsubh (v4hi, v4hi)
5853 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5854 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5855 v2si __builtin_arm_wsubw (v2si, v2si)
5856 v2si __builtin_arm_wsubwss (v2si, v2si)
5857 v2si __builtin_arm_wsubwus (v2si, v2si)
5858 v4hi __builtin_arm_wunpckehsb (v8qi)
5859 v2si __builtin_arm_wunpckehsh (v4hi)
5860 long long __builtin_arm_wunpckehsw (v2si)
5861 v4hi __builtin_arm_wunpckehub (v8qi)
5862 v2si __builtin_arm_wunpckehuh (v4hi)
5863 long long __builtin_arm_wunpckehuw (v2si)
5864 v4hi __builtin_arm_wunpckelsb (v8qi)
5865 v2si __builtin_arm_wunpckelsh (v4hi)
5866 long long __builtin_arm_wunpckelsw (v2si)
5867 v4hi __builtin_arm_wunpckelub (v8qi)
5868 v2si __builtin_arm_wunpckeluh (v4hi)
5869 long long __builtin_arm_wunpckeluw (v2si)
5870 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5871 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5872 v2si __builtin_arm_wunpckihw (v2si, v2si)
5873 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5874 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5875 v2si __builtin_arm_wunpckilw (v2si, v2si)
5876 long long __builtin_arm_wxor (long long, long long)
5877 long long __builtin_arm_wzero ()
5880 @node X86 Built-in Functions
5881 @subsection X86 Built-in Functions
5883 These built-in functions are available for the i386 and x86-64 family
5884 of computers, depending on the command-line switches used.
5886 The following machine modes are available for use with MMX built-in functions
5887 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5888 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5889 vector of eight 8-bit integers. Some of the built-in functions operate on
5890 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5892 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5893 of two 32-bit floating point values.
5895 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5896 floating point values. Some instructions use a vector of four 32-bit
5897 integers, these use @code{V4SI}. Finally, some instructions operate on an
5898 entire vector register, interpreting it as a 128-bit integer, these use mode
5901 The following built-in functions are made available by @option{-mmmx}.
5902 All of them generate the machine instruction that is part of the name.
5905 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5906 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5907 v2si __builtin_ia32_paddd (v2si, v2si)
5908 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5909 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5910 v2si __builtin_ia32_psubd (v2si, v2si)
5911 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5912 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5913 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5914 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5915 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5916 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5917 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5918 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5919 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5920 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5921 di __builtin_ia32_pand (di, di)
5922 di __builtin_ia32_pandn (di,di)
5923 di __builtin_ia32_por (di, di)
5924 di __builtin_ia32_pxor (di, di)
5925 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5926 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5927 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5928 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5929 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5930 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5931 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5932 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5933 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5934 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5935 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5936 v2si __builtin_ia32_punpckldq (v2si, v2si)
5937 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5938 v4hi __builtin_ia32_packssdw (v2si, v2si)
5939 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5942 The following built-in functions are made available either with
5943 @option{-msse}, or with a combination of @option{-m3dnow} and
5944 @option{-march=athlon}. All of them generate the machine
5945 instruction that is part of the name.
5948 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5949 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5950 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5951 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5952 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5953 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5954 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5955 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5956 int __builtin_ia32_pextrw (v4hi, int)
5957 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5958 int __builtin_ia32_pmovmskb (v8qi)
5959 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5960 void __builtin_ia32_movntq (di *, di)
5961 void __builtin_ia32_sfence (void)
5964 The following built-in functions are available when @option{-msse} is used.
5965 All of them generate the machine instruction that is part of the name.
5968 int __builtin_ia32_comieq (v4sf, v4sf)
5969 int __builtin_ia32_comineq (v4sf, v4sf)
5970 int __builtin_ia32_comilt (v4sf, v4sf)
5971 int __builtin_ia32_comile (v4sf, v4sf)
5972 int __builtin_ia32_comigt (v4sf, v4sf)
5973 int __builtin_ia32_comige (v4sf, v4sf)
5974 int __builtin_ia32_ucomieq (v4sf, v4sf)
5975 int __builtin_ia32_ucomineq (v4sf, v4sf)
5976 int __builtin_ia32_ucomilt (v4sf, v4sf)
5977 int __builtin_ia32_ucomile (v4sf, v4sf)
5978 int __builtin_ia32_ucomigt (v4sf, v4sf)
5979 int __builtin_ia32_ucomige (v4sf, v4sf)
5980 v4sf __builtin_ia32_addps (v4sf, v4sf)
5981 v4sf __builtin_ia32_subps (v4sf, v4sf)
5982 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5983 v4sf __builtin_ia32_divps (v4sf, v4sf)
5984 v4sf __builtin_ia32_addss (v4sf, v4sf)
5985 v4sf __builtin_ia32_subss (v4sf, v4sf)
5986 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5987 v4sf __builtin_ia32_divss (v4sf, v4sf)
5988 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5989 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5990 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5991 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5992 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5993 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5994 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5995 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5996 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5997 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5998 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5999 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6000 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6001 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6002 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6003 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6004 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6005 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6006 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6007 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6008 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6009 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6010 v4sf __builtin_ia32_minps (v4sf, v4sf)
6011 v4sf __builtin_ia32_minss (v4sf, v4sf)
6012 v4sf __builtin_ia32_andps (v4sf, v4sf)
6013 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6014 v4sf __builtin_ia32_orps (v4sf, v4sf)
6015 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6016 v4sf __builtin_ia32_movss (v4sf, v4sf)
6017 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6018 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6019 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6020 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6021 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6022 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6023 v2si __builtin_ia32_cvtps2pi (v4sf)
6024 int __builtin_ia32_cvtss2si (v4sf)
6025 v2si __builtin_ia32_cvttps2pi (v4sf)
6026 int __builtin_ia32_cvttss2si (v4sf)
6027 v4sf __builtin_ia32_rcpps (v4sf)
6028 v4sf __builtin_ia32_rsqrtps (v4sf)
6029 v4sf __builtin_ia32_sqrtps (v4sf)
6030 v4sf __builtin_ia32_rcpss (v4sf)
6031 v4sf __builtin_ia32_rsqrtss (v4sf)
6032 v4sf __builtin_ia32_sqrtss (v4sf)
6033 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6034 void __builtin_ia32_movntps (float *, v4sf)
6035 int __builtin_ia32_movmskps (v4sf)
6038 The following built-in functions are available when @option{-msse} is used.
6041 @item v4sf __builtin_ia32_loadaps (float *)
6042 Generates the @code{movaps} machine instruction as a load from memory.
6043 @item void __builtin_ia32_storeaps (float *, v4sf)
6044 Generates the @code{movaps} machine instruction as a store to memory.
6045 @item v4sf __builtin_ia32_loadups (float *)
6046 Generates the @code{movups} machine instruction as a load from memory.
6047 @item void __builtin_ia32_storeups (float *, v4sf)
6048 Generates the @code{movups} machine instruction as a store to memory.
6049 @item v4sf __builtin_ia32_loadsss (float *)
6050 Generates the @code{movss} machine instruction as a load from memory.
6051 @item void __builtin_ia32_storess (float *, v4sf)
6052 Generates the @code{movss} machine instruction as a store to memory.
6053 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6054 Generates the @code{movhps} machine instruction as a load from memory.
6055 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6056 Generates the @code{movlps} machine instruction as a load from memory
6057 @item void __builtin_ia32_storehps (v4sf, v2si *)
6058 Generates the @code{movhps} machine instruction as a store to memory.
6059 @item void __builtin_ia32_storelps (v4sf, v2si *)
6060 Generates the @code{movlps} machine instruction as a store to memory.
6063 The following built-in functions are available when @option{-msse3} is used.
6064 All of them generate the machine instruction that is part of the name.
6067 v2df __builtin_ia32_addsubpd (v2df, v2df)
6068 v2df __builtin_ia32_addsubps (v2df, v2df)
6069 v2df __builtin_ia32_haddpd (v2df, v2df)
6070 v2df __builtin_ia32_haddps (v2df, v2df)
6071 v2df __builtin_ia32_hsubpd (v2df, v2df)
6072 v2df __builtin_ia32_hsubps (v2df, v2df)
6073 v16qi __builtin_ia32_lddqu (char const *)
6074 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6075 v2df __builtin_ia32_movddup (v2df)
6076 v4sf __builtin_ia32_movshdup (v4sf)
6077 v4sf __builtin_ia32_movsldup (v4sf)
6078 void __builtin_ia32_mwait (unsigned int, unsigned int)
6081 The following built-in functions are available when @option{-msse3} is used.
6084 @item v2df __builtin_ia32_loadddup (double const *)
6085 Generates the @code{movddup} machine instruction as a load from memory.
6088 The following built-in functions are available when @option{-m3dnow} is used.
6089 All of them generate the machine instruction that is part of the name.
6092 void __builtin_ia32_femms (void)
6093 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6094 v2si __builtin_ia32_pf2id (v2sf)
6095 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6096 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6097 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6098 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6099 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6100 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6101 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6102 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6103 v2sf __builtin_ia32_pfrcp (v2sf)
6104 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6105 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6106 v2sf __builtin_ia32_pfrsqrt (v2sf)
6107 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6108 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6109 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6110 v2sf __builtin_ia32_pi2fd (v2si)
6111 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6114 The following built-in functions are available when both @option{-m3dnow}
6115 and @option{-march=athlon} are used. All of them generate the machine
6116 instruction that is part of the name.
6119 v2si __builtin_ia32_pf2iw (v2sf)
6120 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6121 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6122 v2sf __builtin_ia32_pi2fw (v2si)
6123 v2sf __builtin_ia32_pswapdsf (v2sf)
6124 v2si __builtin_ia32_pswapdsi (v2si)
6127 @node PowerPC AltiVec Built-in Functions
6128 @subsection PowerPC AltiVec Built-in Functions
6130 These built-in functions are available for the PowerPC family
6131 of computers, depending on the command-line switches used.
6133 The following machine modes are available for use with AltiVec built-in
6134 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
6135 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
6136 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
6137 @code{V16QI} for a vector of sixteen 8-bit integers.
6139 The following functions are made available by including
6140 @code{<altivec.h>} and using @option{-maltivec} and
6141 @option{-mabi=altivec}. The functions implement the functionality
6142 described in Motorola's AltiVec Programming Interface Manual.
6144 There are a few differences from Motorola's documentation and GCC's
6145 implementation. Vector constants are done with curly braces (not
6146 parentheses). Vector initializers require no casts if the vector
6147 constant is of the same type as the variable it is initializing. The
6148 @code{vector bool} type is deprecated and will be discontinued in
6149 further revisions. Use @code{vector signed} instead. If @code{signed}
6150 or @code{unsigned} is omitted, the vector type will default to
6151 @code{signed}. Lastly, all overloaded functions are implemented with macros
6152 for the C implementation. So code the following example will not work:
6155 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6158 Since vec_add is a macro, the vector constant in the above example will
6159 be treated as four different arguments. Wrap the entire argument in
6160 parentheses for this to work. The C++ implementation does not use
6163 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
6164 Internally, GCC uses built-in functions to achieve the functionality in
6165 the aforementioned header file, but they are not supported and are
6166 subject to change without notice.
6169 vector signed char vec_abs (vector signed char, vector signed char);
6170 vector signed short vec_abs (vector signed short, vector signed short);
6171 vector signed int vec_abs (vector signed int, vector signed int);
6172 vector signed float vec_abs (vector signed float, vector signed float);
6174 vector signed char vec_abss (vector signed char, vector signed char);
6175 vector signed short vec_abss (vector signed short, vector signed short);
6177 vector signed char vec_add (vector signed char, vector signed char);
6178 vector unsigned char vec_add (vector signed char, vector unsigned char);
6180 vector unsigned char vec_add (vector unsigned char, vector signed char);
6182 vector unsigned char vec_add (vector unsigned char,
6183 vector unsigned char);
6184 vector signed short vec_add (vector signed short, vector signed short);
6185 vector unsigned short vec_add (vector signed short,
6186 vector unsigned short);
6187 vector unsigned short vec_add (vector unsigned short,
6188 vector signed short);
6189 vector unsigned short vec_add (vector unsigned short,
6190 vector unsigned short);
6191 vector signed int vec_add (vector signed int, vector signed int);
6192 vector unsigned int vec_add (vector signed int, vector unsigned int);
6193 vector unsigned int vec_add (vector unsigned int, vector signed int);
6194 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6195 vector float vec_add (vector float, vector float);
6197 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6199 vector unsigned char vec_adds (vector signed char,
6200 vector unsigned char);
6201 vector unsigned char vec_adds (vector unsigned char,
6202 vector signed char);
6203 vector unsigned char vec_adds (vector unsigned char,
6204 vector unsigned char);
6205 vector signed char vec_adds (vector signed char, vector signed char);
6206 vector unsigned short vec_adds (vector signed short,
6207 vector unsigned short);
6208 vector unsigned short vec_adds (vector unsigned short,
6209 vector signed short);
6210 vector unsigned short vec_adds (vector unsigned short,
6211 vector unsigned short);
6212 vector signed short vec_adds (vector signed short, vector signed short);
6214 vector unsigned int vec_adds (vector signed int, vector unsigned int);
6215 vector unsigned int vec_adds (vector unsigned int, vector signed int);
6216 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6218 vector signed int vec_adds (vector signed int, vector signed int);
6220 vector float vec_and (vector float, vector float);
6221 vector float vec_and (vector float, vector signed int);
6222 vector float vec_and (vector signed int, vector float);
6223 vector signed int vec_and (vector signed int, vector signed int);
6224 vector unsigned int vec_and (vector signed int, vector unsigned int);
6225 vector unsigned int vec_and (vector unsigned int, vector signed int);
6226 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
6227 vector signed short vec_and (vector signed short, vector signed short);
6228 vector unsigned short vec_and (vector signed short,
6229 vector unsigned short);
6230 vector unsigned short vec_and (vector unsigned short,
6231 vector signed short);
6232 vector unsigned short vec_and (vector unsigned short,
6233 vector unsigned short);
6234 vector signed char vec_and (vector signed char, vector signed char);
6235 vector unsigned char vec_and (vector signed char, vector unsigned char);
6237 vector unsigned char vec_and (vector unsigned char, vector signed char);
6239 vector unsigned char vec_and (vector unsigned char,
6240 vector unsigned char);
6242 vector float vec_andc (vector float, vector float);
6243 vector float vec_andc (vector float, vector signed int);
6244 vector float vec_andc (vector signed int, vector float);
6245 vector signed int vec_andc (vector signed int, vector signed int);
6246 vector unsigned int vec_andc (vector signed int, vector unsigned int);
6247 vector unsigned int vec_andc (vector unsigned int, vector signed int);
6248 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
6250 vector signed short vec_andc (vector signed short, vector signed short);
6252 vector unsigned short vec_andc (vector signed short,
6253 vector unsigned short);
6254 vector unsigned short vec_andc (vector unsigned short,
6255 vector signed short);
6256 vector unsigned short vec_andc (vector unsigned short,
6257 vector unsigned short);
6258 vector signed char vec_andc (vector signed char, vector signed char);
6259 vector unsigned char vec_andc (vector signed char,
6260 vector unsigned char);
6261 vector unsigned char vec_andc (vector unsigned char,
6262 vector signed char);
6263 vector unsigned char vec_andc (vector unsigned char,
6264 vector unsigned char);
6266 vector unsigned char vec_avg (vector unsigned char,
6267 vector unsigned char);
6268 vector signed char vec_avg (vector signed char, vector signed char);
6269 vector unsigned short vec_avg (vector unsigned short,
6270 vector unsigned short);
6271 vector signed short vec_avg (vector signed short, vector signed short);
6272 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
6273 vector signed int vec_avg (vector signed int, vector signed int);
6275 vector float vec_ceil (vector float);
6277 vector signed int vec_cmpb (vector float, vector float);
6279 vector signed char vec_cmpeq (vector signed char, vector signed char);
6280 vector signed char vec_cmpeq (vector unsigned char,
6281 vector unsigned char);
6282 vector signed short vec_cmpeq (vector signed short,
6283 vector signed short);
6284 vector signed short vec_cmpeq (vector unsigned short,
6285 vector unsigned short);
6286 vector signed int vec_cmpeq (vector signed int, vector signed int);
6287 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
6288 vector signed int vec_cmpeq (vector float, vector float);
6290 vector signed int vec_cmpge (vector float, vector float);
6292 vector signed char vec_cmpgt (vector unsigned char,
6293 vector unsigned char);
6294 vector signed char vec_cmpgt (vector signed char, vector signed char);
6295 vector signed short vec_cmpgt (vector unsigned short,
6296 vector unsigned short);
6297 vector signed short vec_cmpgt (vector signed short,
6298 vector signed short);
6299 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
6300 vector signed int vec_cmpgt (vector signed int, vector signed int);
6301 vector signed int vec_cmpgt (vector float, vector float);
6303 vector signed int vec_cmple (vector float, vector float);
6305 vector signed char vec_cmplt (vector unsigned char,
6306 vector unsigned char);
6307 vector signed char vec_cmplt (vector signed char, vector signed char);
6308 vector signed short vec_cmplt (vector unsigned short,
6309 vector unsigned short);
6310 vector signed short vec_cmplt (vector signed short,
6311 vector signed short);
6312 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
6313 vector signed int vec_cmplt (vector signed int, vector signed int);
6314 vector signed int vec_cmplt (vector float, vector float);
6316 vector float vec_ctf (vector unsigned int, const char);
6317 vector float vec_ctf (vector signed int, const char);
6319 vector signed int vec_cts (vector float, const char);
6321 vector unsigned int vec_ctu (vector float, const char);
6323 void vec_dss (const char);
6325 void vec_dssall (void);
6327 void vec_dst (void *, int, const char);
6329 void vec_dstst (void *, int, const char);
6331 void vec_dststt (void *, int, const char);
6333 void vec_dstt (void *, int, const char);
6335 vector float vec_expte (vector float, vector float);
6337 vector float vec_floor (vector float, vector float);
6339 vector float vec_ld (int, vector float *);
6340 vector float vec_ld (int, float *):
6341 vector signed int vec_ld (int, int *);
6342 vector signed int vec_ld (int, vector signed int *);
6343 vector unsigned int vec_ld (int, vector unsigned int *);
6344 vector unsigned int vec_ld (int, unsigned int *);
6345 vector signed short vec_ld (int, short *, vector signed short *);
6346 vector unsigned short vec_ld (int, unsigned short *,
6347 vector unsigned short *);
6348 vector signed char vec_ld (int, signed char *);
6349 vector signed char vec_ld (int, vector signed char *);
6350 vector unsigned char vec_ld (int, unsigned char *);
6351 vector unsigned char vec_ld (int, vector unsigned char *);
6353 vector signed char vec_lde (int, signed char *);
6354 vector unsigned char vec_lde (int, unsigned char *);
6355 vector signed short vec_lde (int, short *);
6356 vector unsigned short vec_lde (int, unsigned short *);
6357 vector float vec_lde (int, float *);
6358 vector signed int vec_lde (int, int *);
6359 vector unsigned int vec_lde (int, unsigned int *);
6361 void float vec_ldl (int, float *);
6362 void float vec_ldl (int, vector float *);
6363 void signed int vec_ldl (int, vector signed int *);
6364 void signed int vec_ldl (int, int *);
6365 void unsigned int vec_ldl (int, unsigned int *);
6366 void unsigned int vec_ldl (int, vector unsigned int *);
6367 void signed short vec_ldl (int, vector signed short *);
6368 void signed short vec_ldl (int, short *);
6369 void unsigned short vec_ldl (int, vector unsigned short *);
6370 void unsigned short vec_ldl (int, unsigned short *);
6371 void signed char vec_ldl (int, vector signed char *);
6372 void signed char vec_ldl (int, signed char *);
6373 void unsigned char vec_ldl (int, vector unsigned char *);
6374 void unsigned char vec_ldl (int, unsigned char *);
6376 vector float vec_loge (vector float);
6378 vector unsigned char vec_lvsl (int, void *, int *);
6380 vector unsigned char vec_lvsr (int, void *, int *);
6382 vector float vec_madd (vector float, vector float, vector float);
6384 vector signed short vec_madds (vector signed short, vector signed short,
6385 vector signed short);
6387 vector unsigned char vec_max (vector signed char, vector unsigned char);
6389 vector unsigned char vec_max (vector unsigned char, vector signed char);
6391 vector unsigned char vec_max (vector unsigned char,
6392 vector unsigned char);
6393 vector signed char vec_max (vector signed char, vector signed char);
6394 vector unsigned short vec_max (vector signed short,
6395 vector unsigned short);
6396 vector unsigned short vec_max (vector unsigned short,
6397 vector signed short);
6398 vector unsigned short vec_max (vector unsigned short,
6399 vector unsigned short);
6400 vector signed short vec_max (vector signed short, vector signed short);
6401 vector unsigned int vec_max (vector signed int, vector unsigned int);
6402 vector unsigned int vec_max (vector unsigned int, vector signed int);
6403 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
6404 vector signed int vec_max (vector signed int, vector signed int);
6405 vector float vec_max (vector float, vector float);
6407 vector signed char vec_mergeh (vector signed char, vector signed char);
6408 vector unsigned char vec_mergeh (vector unsigned char,
6409 vector unsigned char);
6410 vector signed short vec_mergeh (vector signed short,
6411 vector signed short);
6412 vector unsigned short vec_mergeh (vector unsigned short,
6413 vector unsigned short);
6414 vector float vec_mergeh (vector float, vector float);
6415 vector signed int vec_mergeh (vector signed int, vector signed int);
6416 vector unsigned int vec_mergeh (vector unsigned int,
6417 vector unsigned int);
6419 vector signed char vec_mergel (vector signed char, vector signed char);
6420 vector unsigned char vec_mergel (vector unsigned char,
6421 vector unsigned char);
6422 vector signed short vec_mergel (vector signed short,
6423 vector signed short);
6424 vector unsigned short vec_mergel (vector unsigned short,
6425 vector unsigned short);
6426 vector float vec_mergel (vector float, vector float);
6427 vector signed int vec_mergel (vector signed int, vector signed int);
6428 vector unsigned int vec_mergel (vector unsigned int,
6429 vector unsigned int);
6431 vector unsigned short vec_mfvscr (void);
6433 vector unsigned char vec_min (vector signed char, vector unsigned char);
6435 vector unsigned char vec_min (vector unsigned char, vector signed char);
6437 vector unsigned char vec_min (vector unsigned char,
6438 vector unsigned char);
6439 vector signed char vec_min (vector signed char, vector signed char);
6440 vector unsigned short vec_min (vector signed short,
6441 vector unsigned short);
6442 vector unsigned short vec_min (vector unsigned short,
6443 vector signed short);
6444 vector unsigned short vec_min (vector unsigned short,
6445 vector unsigned short);
6446 vector signed short vec_min (vector signed short, vector signed short);
6447 vector unsigned int vec_min (vector signed int, vector unsigned int);
6448 vector unsigned int vec_min (vector unsigned int, vector signed int);
6449 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
6450 vector signed int vec_min (vector signed int, vector signed int);
6451 vector float vec_min (vector float, vector float);
6453 vector signed short vec_mladd (vector signed short, vector signed short,
6454 vector signed short);
6455 vector signed short vec_mladd (vector signed short,
6456 vector unsigned short,
6457 vector unsigned short);
6458 vector signed short vec_mladd (vector unsigned short,
6459 vector signed short,
6460 vector signed short);
6461 vector unsigned short vec_mladd (vector unsigned short,
6462 vector unsigned short,
6463 vector unsigned short);
6465 vector signed short vec_mradds (vector signed short,
6466 vector signed short,
6467 vector signed short);
6469 vector unsigned int vec_msum (vector unsigned char,
6470 vector unsigned char,
6471 vector unsigned int);
6472 vector signed int vec_msum (vector signed char, vector unsigned char,
6474 vector unsigned int vec_msum (vector unsigned short,
6475 vector unsigned short,
6476 vector unsigned int);
6477 vector signed int vec_msum (vector signed short, vector signed short,
6480 vector unsigned int vec_msums (vector unsigned short,
6481 vector unsigned short,
6482 vector unsigned int);
6483 vector signed int vec_msums (vector signed short, vector signed short,
6486 void vec_mtvscr (vector signed int);
6487 void vec_mtvscr (vector unsigned int);
6488 void vec_mtvscr (vector signed short);
6489 void vec_mtvscr (vector unsigned short);
6490 void vec_mtvscr (vector signed char);
6491 void vec_mtvscr (vector unsigned char);
6493 vector unsigned short vec_mule (vector unsigned char,
6494 vector unsigned char);
6495 vector signed short vec_mule (vector signed char, vector signed char);
6496 vector unsigned int vec_mule (vector unsigned short,
6497 vector unsigned short);
6498 vector signed int vec_mule (vector signed short, vector signed short);
6500 vector unsigned short vec_mulo (vector unsigned char,
6501 vector unsigned char);
6502 vector signed short vec_mulo (vector signed char, vector signed char);
6503 vector unsigned int vec_mulo (vector unsigned short,
6504 vector unsigned short);
6505 vector signed int vec_mulo (vector signed short, vector signed short);
6507 vector float vec_nmsub (vector float, vector float, vector float);
6509 vector float vec_nor (vector float, vector float);
6510 vector signed int vec_nor (vector signed int, vector signed int);
6511 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
6512 vector signed short vec_nor (vector signed short, vector signed short);
6513 vector unsigned short vec_nor (vector unsigned short,
6514 vector unsigned short);
6515 vector signed char vec_nor (vector signed char, vector signed char);
6516 vector unsigned char vec_nor (vector unsigned char,
6517 vector unsigned char);
6519 vector float vec_or (vector float, vector float);
6520 vector float vec_or (vector float, vector signed int);
6521 vector float vec_or (vector signed int, vector float);
6522 vector signed int vec_or (vector signed int, vector signed int);
6523 vector unsigned int vec_or (vector signed int, vector unsigned int);
6524 vector unsigned int vec_or (vector unsigned int, vector signed int);
6525 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
6526 vector signed short vec_or (vector signed short, vector signed short);
6527 vector unsigned short vec_or (vector signed short,
6528 vector unsigned short);
6529 vector unsigned short vec_or (vector unsigned short,
6530 vector signed short);
6531 vector unsigned short vec_or (vector unsigned short,
6532 vector unsigned short);
6533 vector signed char vec_or (vector signed char, vector signed char);
6534 vector unsigned char vec_or (vector signed char, vector unsigned char);
6535 vector unsigned char vec_or (vector unsigned char, vector signed char);
6536 vector unsigned char vec_or (vector unsigned char,
6537 vector unsigned char);
6539 vector signed char vec_pack (vector signed short, vector signed short);
6540 vector unsigned char vec_pack (vector unsigned short,
6541 vector unsigned short);
6542 vector signed short vec_pack (vector signed int, vector signed int);
6543 vector unsigned short vec_pack (vector unsigned int,
6544 vector unsigned int);
6546 vector signed short vec_packpx (vector unsigned int,
6547 vector unsigned int);
6549 vector unsigned char vec_packs (vector unsigned short,
6550 vector unsigned short);
6551 vector signed char vec_packs (vector signed short, vector signed short);
6553 vector unsigned short vec_packs (vector unsigned int,
6554 vector unsigned int);
6555 vector signed short vec_packs (vector signed int, vector signed int);
6557 vector unsigned char vec_packsu (vector unsigned short,
6558 vector unsigned short);
6559 vector unsigned char vec_packsu (vector signed short,
6560 vector signed short);
6561 vector unsigned short vec_packsu (vector unsigned int,
6562 vector unsigned int);
6563 vector unsigned short vec_packsu (vector signed int, vector signed int);
6565 vector float vec_perm (vector float, vector float,
6566 vector unsigned char);
6567 vector signed int vec_perm (vector signed int, vector signed int,
6568 vector unsigned char);
6569 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
6570 vector unsigned char);
6571 vector signed short vec_perm (vector signed short, vector signed short,
6572 vector unsigned char);
6573 vector unsigned short vec_perm (vector unsigned short,
6574 vector unsigned short,
6575 vector unsigned char);
6576 vector signed char vec_perm (vector signed char, vector signed char,
6577 vector unsigned char);
6578 vector unsigned char vec_perm (vector unsigned char,
6579 vector unsigned char,
6580 vector unsigned char);
6582 vector float vec_re (vector float);
6584 vector signed char vec_rl (vector signed char, vector unsigned char);
6585 vector unsigned char vec_rl (vector unsigned char,
6586 vector unsigned char);
6587 vector signed short vec_rl (vector signed short, vector unsigned short);
6589 vector unsigned short vec_rl (vector unsigned short,
6590 vector unsigned short);
6591 vector signed int vec_rl (vector signed int, vector unsigned int);
6592 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
6594 vector float vec_round (vector float);
6596 vector float vec_rsqrte (vector float);
6598 vector float vec_sel (vector float, vector float, vector signed int);
6599 vector float vec_sel (vector float, vector float, vector unsigned int);
6600 vector signed int vec_sel (vector signed int, vector signed int,
6602 vector signed int vec_sel (vector signed int, vector signed int,
6603 vector unsigned int);
6604 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6606 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6607 vector unsigned int);
6608 vector signed short vec_sel (vector signed short, vector signed short,
6609 vector signed short);
6610 vector signed short vec_sel (vector signed short, vector signed short,
6611 vector unsigned short);
6612 vector unsigned short vec_sel (vector unsigned short,
6613 vector unsigned short,
6614 vector signed short);
6615 vector unsigned short vec_sel (vector unsigned short,
6616 vector unsigned short,
6617 vector unsigned short);
6618 vector signed char vec_sel (vector signed char, vector signed char,
6619 vector signed char);
6620 vector signed char vec_sel (vector signed char, vector signed char,
6621 vector unsigned char);
6622 vector unsigned char vec_sel (vector unsigned char,
6623 vector unsigned char,
6624 vector signed char);
6625 vector unsigned char vec_sel (vector unsigned char,
6626 vector unsigned char,
6627 vector unsigned char);
6629 vector signed char vec_sl (vector signed char, vector unsigned char);
6630 vector unsigned char vec_sl (vector unsigned char,
6631 vector unsigned char);
6632 vector signed short vec_sl (vector signed short, vector unsigned short);
6634 vector unsigned short vec_sl (vector unsigned short,
6635 vector unsigned short);
6636 vector signed int vec_sl (vector signed int, vector unsigned int);
6637 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
6639 vector float vec_sld (vector float, vector float, const char);
6640 vector signed int vec_sld (vector signed int, vector signed int,
6642 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
6644 vector signed short vec_sld (vector signed short, vector signed short,
6646 vector unsigned short vec_sld (vector unsigned short,
6647 vector unsigned short, const char);
6648 vector signed char vec_sld (vector signed char, vector signed char,
6650 vector unsigned char vec_sld (vector unsigned char,
6651 vector unsigned char,
6654 vector signed int vec_sll (vector signed int, vector unsigned int);
6655 vector signed int vec_sll (vector signed int, vector unsigned short);
6656 vector signed int vec_sll (vector signed int, vector unsigned char);
6657 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
6658 vector unsigned int vec_sll (vector unsigned int,
6659 vector unsigned short);
6660 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
6662 vector signed short vec_sll (vector signed short, vector unsigned int);
6663 vector signed short vec_sll (vector signed short,
6664 vector unsigned short);
6665 vector signed short vec_sll (vector signed short, vector unsigned char);
6667 vector unsigned short vec_sll (vector unsigned short,
6668 vector unsigned int);
6669 vector unsigned short vec_sll (vector unsigned short,
6670 vector unsigned short);
6671 vector unsigned short vec_sll (vector unsigned short,
6672 vector unsigned char);
6673 vector signed char vec_sll (vector signed char, vector unsigned int);
6674 vector signed char vec_sll (vector signed char, vector unsigned short);
6675 vector signed char vec_sll (vector signed char, vector unsigned char);
6676 vector unsigned char vec_sll (vector unsigned char,
6677 vector unsigned int);
6678 vector unsigned char vec_sll (vector unsigned char,
6679 vector unsigned short);
6680 vector unsigned char vec_sll (vector unsigned char,
6681 vector unsigned char);
6683 vector float vec_slo (vector float, vector signed char);
6684 vector float vec_slo (vector float, vector unsigned char);
6685 vector signed int vec_slo (vector signed int, vector signed char);
6686 vector signed int vec_slo (vector signed int, vector unsigned char);
6687 vector unsigned int vec_slo (vector unsigned int, vector signed char);
6688 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
6690 vector signed short vec_slo (vector signed short, vector signed char);
6691 vector signed short vec_slo (vector signed short, vector unsigned char);
6693 vector unsigned short vec_slo (vector unsigned short,
6694 vector signed char);
6695 vector unsigned short vec_slo (vector unsigned short,
6696 vector unsigned char);
6697 vector signed char vec_slo (vector signed char, vector signed char);
6698 vector signed char vec_slo (vector signed char, vector unsigned char);
6699 vector unsigned char vec_slo (vector unsigned char, vector signed char);
6701 vector unsigned char vec_slo (vector unsigned char,
6702 vector unsigned char);
6704 vector signed char vec_splat (vector signed char, const char);
6705 vector unsigned char vec_splat (vector unsigned char, const char);
6706 vector signed short vec_splat (vector signed short, const char);
6707 vector unsigned short vec_splat (vector unsigned short, const char);
6708 vector float vec_splat (vector float, const char);
6709 vector signed int vec_splat (vector signed int, const char);
6710 vector unsigned int vec_splat (vector unsigned int, const char);
6712 vector signed char vec_splat_s8 (const char);
6714 vector signed short vec_splat_s16 (const char);
6716 vector signed int vec_splat_s32 (const char);
6718 vector unsigned char vec_splat_u8 (const char);
6720 vector unsigned short vec_splat_u16 (const char);
6722 vector unsigned int vec_splat_u32 (const char);
6724 vector signed char vec_sr (vector signed char, vector unsigned char);
6725 vector unsigned char vec_sr (vector unsigned char,
6726 vector unsigned char);
6727 vector signed short vec_sr (vector signed short, vector unsigned short);
6729 vector unsigned short vec_sr (vector unsigned short,
6730 vector unsigned short);
6731 vector signed int vec_sr (vector signed int, vector unsigned int);
6732 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
6734 vector signed char vec_sra (vector signed char, vector unsigned char);
6735 vector unsigned char vec_sra (vector unsigned char,
6736 vector unsigned char);
6737 vector signed short vec_sra (vector signed short,
6738 vector unsigned short);
6739 vector unsigned short vec_sra (vector unsigned short,
6740 vector unsigned short);
6741 vector signed int vec_sra (vector signed int, vector unsigned int);
6742 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
6744 vector signed int vec_srl (vector signed int, vector unsigned int);
6745 vector signed int vec_srl (vector signed int, vector unsigned short);
6746 vector signed int vec_srl (vector signed int, vector unsigned char);
6747 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
6748 vector unsigned int vec_srl (vector unsigned int,
6749 vector unsigned short);
6750 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
6752 vector signed short vec_srl (vector signed short, vector unsigned int);
6753 vector signed short vec_srl (vector signed short,
6754 vector unsigned short);
6755 vector signed short vec_srl (vector signed short, vector unsigned char);
6757 vector unsigned short vec_srl (vector unsigned short,
6758 vector unsigned int);
6759 vector unsigned short vec_srl (vector unsigned short,
6760 vector unsigned short);
6761 vector unsigned short vec_srl (vector unsigned short,
6762 vector unsigned char);
6763 vector signed char vec_srl (vector signed char, vector unsigned int);
6764 vector signed char vec_srl (vector signed char, vector unsigned short);
6765 vector signed char vec_srl (vector signed char, vector unsigned char);
6766 vector unsigned char vec_srl (vector unsigned char,
6767 vector unsigned int);
6768 vector unsigned char vec_srl (vector unsigned char,
6769 vector unsigned short);
6770 vector unsigned char vec_srl (vector unsigned char,
6771 vector unsigned char);
6773 vector float vec_sro (vector float, vector signed char);
6774 vector float vec_sro (vector float, vector unsigned char);
6775 vector signed int vec_sro (vector signed int, vector signed char);
6776 vector signed int vec_sro (vector signed int, vector unsigned char);
6777 vector unsigned int vec_sro (vector unsigned int, vector signed char);
6778 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
6780 vector signed short vec_sro (vector signed short, vector signed char);
6781 vector signed short vec_sro (vector signed short, vector unsigned char);
6783 vector unsigned short vec_sro (vector unsigned short,
6784 vector signed char);
6785 vector unsigned short vec_sro (vector unsigned short,
6786 vector unsigned char);
6787 vector signed char vec_sro (vector signed char, vector signed char);
6788 vector signed char vec_sro (vector signed char, vector unsigned char);
6789 vector unsigned char vec_sro (vector unsigned char, vector signed char);
6791 vector unsigned char vec_sro (vector unsigned char,
6792 vector unsigned char);
6794 void vec_st (vector float, int, float *);
6795 void vec_st (vector float, int, vector float *);
6796 void vec_st (vector signed int, int, int *);
6797 void vec_st (vector signed int, int, unsigned int *);
6798 void vec_st (vector unsigned int, int, unsigned int *);
6799 void vec_st (vector unsigned int, int, vector unsigned int *);
6800 void vec_st (vector signed short, int, short *);
6801 void vec_st (vector signed short, int, vector unsigned short *);
6802 void vec_st (vector signed short, int, vector signed short *);
6803 void vec_st (vector unsigned short, int, unsigned short *);
6804 void vec_st (vector unsigned short, int, vector unsigned short *);
6805 void vec_st (vector signed char, int, signed char *);
6806 void vec_st (vector signed char, int, unsigned char *);
6807 void vec_st (vector signed char, int, vector signed char *);
6808 void vec_st (vector unsigned char, int, unsigned char *);
6809 void vec_st (vector unsigned char, int, vector unsigned char *);
6811 void vec_ste (vector signed char, int, unsigned char *);
6812 void vec_ste (vector signed char, int, signed char *);
6813 void vec_ste (vector unsigned char, int, unsigned char *);
6814 void vec_ste (vector signed short, int, short *);
6815 void vec_ste (vector signed short, int, unsigned short *);
6816 void vec_ste (vector unsigned short, int, void *);
6817 void vec_ste (vector signed int, int, unsigned int *);
6818 void vec_ste (vector signed int, int, int *);
6819 void vec_ste (vector unsigned int, int, unsigned int *);
6820 void vec_ste (vector float, int, float *);
6822 void vec_stl (vector float, int, vector float *);
6823 void vec_stl (vector float, int, float *);
6824 void vec_stl (vector signed int, int, vector signed int *);
6825 void vec_stl (vector signed int, int, int *);
6826 void vec_stl (vector signed int, int, unsigned int *);
6827 void vec_stl (vector unsigned int, int, vector unsigned int *);
6828 void vec_stl (vector unsigned int, int, unsigned int *);
6829 void vec_stl (vector signed short, int, short *);
6830 void vec_stl (vector signed short, int, unsigned short *);
6831 void vec_stl (vector signed short, int, vector signed short *);
6832 void vec_stl (vector unsigned short, int, unsigned short *);
6833 void vec_stl (vector unsigned short, int, vector signed short *);
6834 void vec_stl (vector signed char, int, signed char *);
6835 void vec_stl (vector signed char, int, unsigned char *);
6836 void vec_stl (vector signed char, int, vector signed char *);
6837 void vec_stl (vector unsigned char, int, unsigned char *);
6838 void vec_stl (vector unsigned char, int, vector unsigned char *);
6840 vector signed char vec_sub (vector signed char, vector signed char);
6841 vector unsigned char vec_sub (vector signed char, vector unsigned char);
6843 vector unsigned char vec_sub (vector unsigned char, vector signed char);
6845 vector unsigned char vec_sub (vector unsigned char,
6846 vector unsigned char);
6847 vector signed short vec_sub (vector signed short, vector signed short);
6848 vector unsigned short vec_sub (vector signed short,
6849 vector unsigned short);
6850 vector unsigned short vec_sub (vector unsigned short,
6851 vector signed short);
6852 vector unsigned short vec_sub (vector unsigned short,
6853 vector unsigned short);
6854 vector signed int vec_sub (vector signed int, vector signed int);
6855 vector unsigned int vec_sub (vector signed int, vector unsigned int);
6856 vector unsigned int vec_sub (vector unsigned int, vector signed int);
6857 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
6858 vector float vec_sub (vector float, vector float);
6860 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
6862 vector unsigned char vec_subs (vector signed char,
6863 vector unsigned char);
6864 vector unsigned char vec_subs (vector unsigned char,
6865 vector signed char);
6866 vector unsigned char vec_subs (vector unsigned char,
6867 vector unsigned char);
6868 vector signed char vec_subs (vector signed char, vector signed char);
6869 vector unsigned short vec_subs (vector signed short,
6870 vector unsigned short);
6871 vector unsigned short vec_subs (vector unsigned short,
6872 vector signed short);
6873 vector unsigned short vec_subs (vector unsigned short,
6874 vector unsigned short);
6875 vector signed short vec_subs (vector signed short, vector signed short);
6877 vector unsigned int vec_subs (vector signed int, vector unsigned int);
6878 vector unsigned int vec_subs (vector unsigned int, vector signed int);
6879 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
6881 vector signed int vec_subs (vector signed int, vector signed int);
6883 vector unsigned int vec_sum4s (vector unsigned char,
6884 vector unsigned int);
6885 vector signed int vec_sum4s (vector signed char, vector signed int);
6886 vector signed int vec_sum4s (vector signed short, vector signed int);
6888 vector signed int vec_sum2s (vector signed int, vector signed int);
6890 vector signed int vec_sums (vector signed int, vector signed int);
6892 vector float vec_trunc (vector float);
6894 vector signed short vec_unpackh (vector signed char);
6895 vector unsigned int vec_unpackh (vector signed short);
6896 vector signed int vec_unpackh (vector signed short);
6898 vector signed short vec_unpackl (vector signed char);
6899 vector unsigned int vec_unpackl (vector signed short);
6900 vector signed int vec_unpackl (vector signed short);
6902 vector float vec_xor (vector float, vector float);
6903 vector float vec_xor (vector float, vector signed int);
6904 vector float vec_xor (vector signed int, vector float);
6905 vector signed int vec_xor (vector signed int, vector signed int);
6906 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6907 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6908 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6909 vector signed short vec_xor (vector signed short, vector signed short);
6910 vector unsigned short vec_xor (vector signed short,
6911 vector unsigned short);
6912 vector unsigned short vec_xor (vector unsigned short,
6913 vector signed short);
6914 vector unsigned short vec_xor (vector unsigned short,
6915 vector unsigned short);
6916 vector signed char vec_xor (vector signed char, vector signed char);
6917 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6919 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6921 vector unsigned char vec_xor (vector unsigned char,
6922 vector unsigned char);
6924 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6926 vector signed int vec_all_eq (vector signed char, vector signed char);
6927 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6929 vector signed int vec_all_eq (vector unsigned char,
6930 vector unsigned char);
6931 vector signed int vec_all_eq (vector signed short,
6932 vector unsigned short);
6933 vector signed int vec_all_eq (vector signed short, vector signed short);
6935 vector signed int vec_all_eq (vector unsigned short,
6936 vector signed short);
6937 vector signed int vec_all_eq (vector unsigned short,
6938 vector unsigned short);
6939 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6940 vector signed int vec_all_eq (vector signed int, vector signed int);
6941 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6942 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6944 vector signed int vec_all_eq (vector float, vector float);
6946 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6948 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6950 vector signed int vec_all_ge (vector unsigned char,
6951 vector unsigned char);
6952 vector signed int vec_all_ge (vector signed char, vector signed char);
6953 vector signed int vec_all_ge (vector signed short,
6954 vector unsigned short);
6955 vector signed int vec_all_ge (vector unsigned short,
6956 vector signed short);
6957 vector signed int vec_all_ge (vector unsigned short,
6958 vector unsigned short);
6959 vector signed int vec_all_ge (vector signed short, vector signed short);
6961 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6962 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6963 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6965 vector signed int vec_all_ge (vector signed int, vector signed int);
6966 vector signed int vec_all_ge (vector float, vector float);
6968 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6970 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6972 vector signed int vec_all_gt (vector unsigned char,
6973 vector unsigned char);
6974 vector signed int vec_all_gt (vector signed char, vector signed char);
6975 vector signed int vec_all_gt (vector signed short,
6976 vector unsigned short);
6977 vector signed int vec_all_gt (vector unsigned short,
6978 vector signed short);
6979 vector signed int vec_all_gt (vector unsigned short,
6980 vector unsigned short);
6981 vector signed int vec_all_gt (vector signed short, vector signed short);
6983 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6984 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6985 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6987 vector signed int vec_all_gt (vector signed int, vector signed int);
6988 vector signed int vec_all_gt (vector float, vector float);
6990 vector signed int vec_all_in (vector float, vector float);
6992 vector signed int vec_all_le (vector signed char, vector unsigned char);
6994 vector signed int vec_all_le (vector unsigned char, vector signed char);
6996 vector signed int vec_all_le (vector unsigned char,
6997 vector unsigned char);
6998 vector signed int vec_all_le (vector signed char, vector signed char);
6999 vector signed int vec_all_le (vector signed short,
7000 vector unsigned short);
7001 vector signed int vec_all_le (vector unsigned short,
7002 vector signed short);
7003 vector signed int vec_all_le (vector unsigned short,
7004 vector unsigned short);
7005 vector signed int vec_all_le (vector signed short, vector signed short);
7007 vector signed int vec_all_le (vector signed int, vector unsigned int);
7008 vector signed int vec_all_le (vector unsigned int, vector signed int);
7009 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
7011 vector signed int vec_all_le (vector signed int, vector signed int);
7012 vector signed int vec_all_le (vector float, vector float);
7014 vector signed int vec_all_lt (vector signed char, vector unsigned char);
7016 vector signed int vec_all_lt (vector unsigned char, vector signed char);
7018 vector signed int vec_all_lt (vector unsigned char,
7019 vector unsigned char);
7020 vector signed int vec_all_lt (vector signed char, vector signed char);
7021 vector signed int vec_all_lt (vector signed short,
7022 vector unsigned short);
7023 vector signed int vec_all_lt (vector unsigned short,
7024 vector signed short);
7025 vector signed int vec_all_lt (vector unsigned short,
7026 vector unsigned short);
7027 vector signed int vec_all_lt (vector signed short, vector signed short);
7029 vector signed int vec_all_lt (vector signed int, vector unsigned int);
7030 vector signed int vec_all_lt (vector unsigned int, vector signed int);
7031 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
7033 vector signed int vec_all_lt (vector signed int, vector signed int);
7034 vector signed int vec_all_lt (vector float, vector float);
7036 vector signed int vec_all_nan (vector float);
7038 vector signed int vec_all_ne (vector signed char, vector unsigned char);
7040 vector signed int vec_all_ne (vector signed char, vector signed char);
7041 vector signed int vec_all_ne (vector unsigned char, vector signed char);
7043 vector signed int vec_all_ne (vector unsigned char,
7044 vector unsigned char);
7045 vector signed int vec_all_ne (vector signed short,
7046 vector unsigned short);
7047 vector signed int vec_all_ne (vector signed short, vector signed short);
7049 vector signed int vec_all_ne (vector unsigned short,
7050 vector signed short);
7051 vector signed int vec_all_ne (vector unsigned short,
7052 vector unsigned short);
7053 vector signed int vec_all_ne (vector signed int, vector unsigned int);
7054 vector signed int vec_all_ne (vector signed int, vector signed int);
7055 vector signed int vec_all_ne (vector unsigned int, vector signed int);
7056 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
7058 vector signed int vec_all_ne (vector float, vector float);
7060 vector signed int vec_all_nge (vector float, vector float);
7062 vector signed int vec_all_ngt (vector float, vector float);
7064 vector signed int vec_all_nle (vector float, vector float);
7066 vector signed int vec_all_nlt (vector float, vector float);
7068 vector signed int vec_all_numeric (vector float);
7070 vector signed int vec_any_eq (vector signed char, vector unsigned char);
7072 vector signed int vec_any_eq (vector signed char, vector signed char);
7073 vector signed int vec_any_eq (vector unsigned char, vector signed char);
7075 vector signed int vec_any_eq (vector unsigned char,
7076 vector unsigned char);
7077 vector signed int vec_any_eq (vector signed short,
7078 vector unsigned short);
7079 vector signed int vec_any_eq (vector signed short, vector signed short);
7081 vector signed int vec_any_eq (vector unsigned short,
7082 vector signed short);
7083 vector signed int vec_any_eq (vector unsigned short,
7084 vector unsigned short);
7085 vector signed int vec_any_eq (vector signed int, vector unsigned int);
7086 vector signed int vec_any_eq (vector signed int, vector signed int);
7087 vector signed int vec_any_eq (vector unsigned int, vector signed int);
7088 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
7090 vector signed int vec_any_eq (vector float, vector float);
7092 vector signed int vec_any_ge (vector signed char, vector unsigned char);
7094 vector signed int vec_any_ge (vector unsigned char, vector signed char);
7096 vector signed int vec_any_ge (vector unsigned char,
7097 vector unsigned char);
7098 vector signed int vec_any_ge (vector signed char, vector signed char);
7099 vector signed int vec_any_ge (vector signed short,
7100 vector unsigned short);
7101 vector signed int vec_any_ge (vector unsigned short,
7102 vector signed short);
7103 vector signed int vec_any_ge (vector unsigned short,
7104 vector unsigned short);
7105 vector signed int vec_any_ge (vector signed short, vector signed short);
7107 vector signed int vec_any_ge (vector signed int, vector unsigned int);
7108 vector signed int vec_any_ge (vector unsigned int, vector signed int);
7109 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
7111 vector signed int vec_any_ge (vector signed int, vector signed int);
7112 vector signed int vec_any_ge (vector float, vector float);
7114 vector signed int vec_any_gt (vector signed char, vector unsigned char);
7116 vector signed int vec_any_gt (vector unsigned char, vector signed char);
7118 vector signed int vec_any_gt (vector unsigned char,
7119 vector unsigned char);
7120 vector signed int vec_any_gt (vector signed char, vector signed char);
7121 vector signed int vec_any_gt (vector signed short,
7122 vector unsigned short);
7123 vector signed int vec_any_gt (vector unsigned short,
7124 vector signed short);
7125 vector signed int vec_any_gt (vector unsigned short,
7126 vector unsigned short);
7127 vector signed int vec_any_gt (vector signed short, vector signed short);
7129 vector signed int vec_any_gt (vector signed int, vector unsigned int);
7130 vector signed int vec_any_gt (vector unsigned int, vector signed int);
7131 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
7133 vector signed int vec_any_gt (vector signed int, vector signed int);
7134 vector signed int vec_any_gt (vector float, vector float);
7136 vector signed int vec_any_le (vector signed char, vector unsigned char);
7138 vector signed int vec_any_le (vector unsigned char, vector signed char);
7140 vector signed int vec_any_le (vector unsigned char,
7141 vector unsigned char);
7142 vector signed int vec_any_le (vector signed char, vector signed char);
7143 vector signed int vec_any_le (vector signed short,
7144 vector unsigned short);
7145 vector signed int vec_any_le (vector unsigned short,
7146 vector signed short);
7147 vector signed int vec_any_le (vector unsigned short,
7148 vector unsigned short);
7149 vector signed int vec_any_le (vector signed short, vector signed short);
7151 vector signed int vec_any_le (vector signed int, vector unsigned int);
7152 vector signed int vec_any_le (vector unsigned int, vector signed int);
7153 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
7155 vector signed int vec_any_le (vector signed int, vector signed int);
7156 vector signed int vec_any_le (vector float, vector float);
7158 vector signed int vec_any_lt (vector signed char, vector unsigned char);
7160 vector signed int vec_any_lt (vector unsigned char, vector signed char);
7162 vector signed int vec_any_lt (vector unsigned char,
7163 vector unsigned char);
7164 vector signed int vec_any_lt (vector signed char, vector signed char);
7165 vector signed int vec_any_lt (vector signed short,
7166 vector unsigned short);
7167 vector signed int vec_any_lt (vector unsigned short,
7168 vector signed short);
7169 vector signed int vec_any_lt (vector unsigned short,
7170 vector unsigned short);
7171 vector signed int vec_any_lt (vector signed short, vector signed short);
7173 vector signed int vec_any_lt (vector signed int, vector unsigned int);
7174 vector signed int vec_any_lt (vector unsigned int, vector signed int);
7175 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
7177 vector signed int vec_any_lt (vector signed int, vector signed int);
7178 vector signed int vec_any_lt (vector float, vector float);
7180 vector signed int vec_any_nan (vector float);
7182 vector signed int vec_any_ne (vector signed char, vector unsigned char);
7184 vector signed int vec_any_ne (vector signed char, vector signed char);
7185 vector signed int vec_any_ne (vector unsigned char, vector signed char);
7187 vector signed int vec_any_ne (vector unsigned char,
7188 vector unsigned char);
7189 vector signed int vec_any_ne (vector signed short,
7190 vector unsigned short);
7191 vector signed int vec_any_ne (vector signed short, vector signed short);
7193 vector signed int vec_any_ne (vector unsigned short,
7194 vector signed short);
7195 vector signed int vec_any_ne (vector unsigned short,
7196 vector unsigned short);
7197 vector signed int vec_any_ne (vector signed int, vector unsigned int);
7198 vector signed int vec_any_ne (vector signed int, vector signed int);
7199 vector signed int vec_any_ne (vector unsigned int, vector signed int);
7200 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
7202 vector signed int vec_any_ne (vector float, vector float);
7204 vector signed int vec_any_nge (vector float, vector float);
7206 vector signed int vec_any_ngt (vector float, vector float);
7208 vector signed int vec_any_nle (vector float, vector float);
7210 vector signed int vec_any_nlt (vector float, vector float);
7212 vector signed int vec_any_numeric (vector float);
7214 vector signed int vec_any_out (vector float, vector float);
7217 @node Target Format Checks
7218 @section Format Checks Specific to Particular Target Machines
7220 For some target machines, GCC supports additional options to the
7222 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
7225 * Solaris Format Checks::
7228 @node Solaris Format Checks
7229 @subsection Solaris Format Checks
7231 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
7232 check. @code{cmn_err} accepts a subset of the standard @code{printf}
7233 conversions, and the two-argument @code{%b} conversion for displaying
7234 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
7237 @section Pragmas Accepted by GCC
7241 GCC supports several types of pragmas, primarily in order to compile
7242 code originally written for other compilers. Note that in general
7243 we do not recommend the use of pragmas; @xref{Function Attributes},
7244 for further explanation.
7248 * RS/6000 and PowerPC Pragmas::
7250 * Symbol-Renaming Pragmas::
7254 @subsection ARM Pragmas
7256 The ARM target defines pragmas for controlling the default addition of
7257 @code{long_call} and @code{short_call} attributes to functions.
7258 @xref{Function Attributes}, for information about the effects of these
7263 @cindex pragma, long_calls
7264 Set all subsequent functions to have the @code{long_call} attribute.
7267 @cindex pragma, no_long_calls
7268 Set all subsequent functions to have the @code{short_call} attribute.
7270 @item long_calls_off
7271 @cindex pragma, long_calls_off
7272 Do not affect the @code{long_call} or @code{short_call} attributes of
7273 subsequent functions.
7276 @node RS/6000 and PowerPC Pragmas
7277 @subsection RS/6000 and PowerPC Pragmas
7279 The RS/6000 and PowerPC targets define one pragma for controlling
7280 whether or not the @code{longcall} attribute is added to function
7281 declarations by default. This pragma overrides the @option{-mlongcall}
7282 option, but not the @code{longcall} and @code{shortcall} attributes.
7283 @xref{RS/6000 and PowerPC Options}, for more information about when long
7284 calls are and are not necessary.
7288 @cindex pragma, longcall
7289 Apply the @code{longcall} attribute to all subsequent function
7293 Do not apply the @code{longcall} attribute to subsequent function
7297 @c Describe c4x pragmas here.
7298 @c Describe h8300 pragmas here.
7299 @c Describe sh pragmas here.
7300 @c Describe v850 pragmas here.
7302 @node Darwin Pragmas
7303 @subsection Darwin Pragmas
7305 The following pragmas are available for all architectures running the
7306 Darwin operating system. These are useful for compatibility with other
7310 @item mark @var{tokens}@dots{}
7311 @cindex pragma, mark
7312 This pragma is accepted, but has no effect.
7314 @item options align=@var{alignment}
7315 @cindex pragma, options align
7316 This pragma sets the alignment of fields in structures. The values of
7317 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
7318 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
7319 properly; to restore the previous setting, use @code{reset} for the
7322 @item segment @var{tokens}@dots{}
7323 @cindex pragma, segment
7324 This pragma is accepted, but has no effect.
7326 @item unused (@var{var} [, @var{var}]@dots{})
7327 @cindex pragma, unused
7328 This pragma declares variables to be possibly unused. GCC will not
7329 produce warnings for the listed variables. The effect is similar to
7330 that of the @code{unused} attribute, except that this pragma may appear
7331 anywhere within the variables' scopes.
7334 @node Symbol-Renaming Pragmas
7335 @subsection Symbol-Renaming Pragmas
7337 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
7338 supports two @code{#pragma} directives which change the name used in
7339 assembly for a given declaration. These pragmas are only available on
7340 platforms whose system headers need them. To get this effect on all
7341 platforms supported by GCC, use the asm labels extension (@pxref{Asm
7345 @item redefine_extname @var{oldname} @var{newname}
7346 @cindex pragma, redefine_extname
7348 This pragma gives the C function @var{oldname} the assembly symbol
7349 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
7350 will be defined if this pragma is available (currently only on
7353 @item extern_prefix @var{string}
7354 @cindex pragma, extern_prefix
7356 This pragma causes all subsequent external function and variable
7357 declarations to have @var{string} prepended to their assembly symbols.
7358 This effect may be terminated with another @code{extern_prefix} pragma
7359 whose argument is an empty string. The preprocessor macro
7360 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
7361 available (currently only on Tru64 UNIX).
7364 These pragmas and the asm labels extension interact in a complicated
7365 manner. Here are some corner cases you may want to be aware of.
7368 @item Both pragmas silently apply only to declarations with external
7369 linkage. Asm labels do not have this restriction.
7371 @item In C++, both pragmas silently apply only to declarations with
7372 ``C'' linkage. Again, asm labels do not have this restriction.
7374 @item If any of the three ways of changing the assembly name of a
7375 declaration is applied to a declaration whose assembly name has
7376 already been determined (either by a previous use of one of these
7377 features, or because the compiler needed the assembly name in order to
7378 generate code), and the new name is different, a warning issues and
7379 the name does not change.
7381 @item The @var{oldname} used by @code{#pragma redefine_extname} is
7382 always the C-language name.
7384 @item If @code{#pragma extern_prefix} is in effect, and a declaration
7385 occurs with an asm label attached, the prefix is silently ignored for
7388 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
7389 apply to the same declaration, whichever triggered first wins, and a
7390 warning issues if they contradict each other. (We would like to have
7391 @code{#pragma redefine_extname} always win, for consistency with asm
7392 labels, but if @code{#pragma extern_prefix} triggers first we have no
7393 way of knowing that that happened.)
7396 @node Unnamed Fields
7397 @section Unnamed struct/union fields within structs/unions.
7401 For compatibility with other compilers, GCC allows you to define
7402 a structure or union that contains, as fields, structures and unions
7403 without names. For example:
7416 In this example, the user would be able to access members of the unnamed
7417 union with code like @samp{foo.b}. Note that only unnamed structs and
7418 unions are allowed, you may not have, for example, an unnamed
7421 You must never create such structures that cause ambiguous field definitions.
7422 For example, this structure:
7433 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
7434 Such constructs are not supported and must be avoided. In the future,
7435 such constructs may be detected and treated as compilation errors.
7438 @section Thread-Local Storage
7439 @cindex Thread-Local Storage
7440 @cindex @acronym{TLS}
7443 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
7444 are allocated such that there is one instance of the variable per extant
7445 thread. The run-time model GCC uses to implement this originates
7446 in the IA-64 processor-specific ABI, but has since been migrated
7447 to other processors as well. It requires significant support from
7448 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
7449 system libraries (@file{libc.so} and @file{libpthread.so}), so it
7450 is not available everywhere.
7452 At the user level, the extension is visible with a new storage
7453 class keyword: @code{__thread}. For example:
7457 extern __thread struct state s;
7458 static __thread char *p;
7461 The @code{__thread} specifier may be used alone, with the @code{extern}
7462 or @code{static} specifiers, but with no other storage class specifier.
7463 When used with @code{extern} or @code{static}, @code{__thread} must appear
7464 immediately after the other storage class specifier.
7466 The @code{__thread} specifier may be applied to any global, file-scoped
7467 static, function-scoped static, or static data member of a class. It may
7468 not be applied to block-scoped automatic or non-static data member.
7470 When the address-of operator is applied to a thread-local variable, it is
7471 evaluated at run-time and returns the address of the current thread's
7472 instance of that variable. An address so obtained may be used by any
7473 thread. When a thread terminates, any pointers to thread-local variables
7474 in that thread become invalid.
7476 No static initialization may refer to the address of a thread-local variable.
7478 In C++, if an initializer is present for a thread-local variable, it must
7479 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
7482 See @uref{http://people.redhat.com/drepper/tls.pdf,
7483 ELF Handling For Thread-Local Storage} for a detailed explanation of
7484 the four thread-local storage addressing models, and how the run-time
7485 is expected to function.
7488 * C99 Thread-Local Edits::
7489 * C++98 Thread-Local Edits::
7492 @node C99 Thread-Local Edits
7493 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
7495 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
7496 that document the exact semantics of the language extension.
7500 @cite{5.1.2 Execution environments}
7502 Add new text after paragraph 1
7505 Within either execution environment, a @dfn{thread} is a flow of
7506 control within a program. It is implementation defined whether
7507 or not there may be more than one thread associated with a program.
7508 It is implementation defined how threads beyond the first are
7509 created, the name and type of the function called at thread
7510 startup, and how threads may be terminated. However, objects
7511 with thread storage duration shall be initialized before thread
7516 @cite{6.2.4 Storage durations of objects}
7518 Add new text before paragraph 3
7521 An object whose identifier is declared with the storage-class
7522 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
7523 Its lifetime is the entire execution of the thread, and its
7524 stored value is initialized only once, prior to thread startup.
7528 @cite{6.4.1 Keywords}
7530 Add @code{__thread}.
7533 @cite{6.7.1 Storage-class specifiers}
7535 Add @code{__thread} to the list of storage class specifiers in
7538 Change paragraph 2 to
7541 With the exception of @code{__thread}, at most one storage-class
7542 specifier may be given [@dots{}]. The @code{__thread} specifier may
7543 be used alone, or immediately following @code{extern} or
7547 Add new text after paragraph 6
7550 The declaration of an identifier for a variable that has
7551 block scope that specifies @code{__thread} shall also
7552 specify either @code{extern} or @code{static}.
7554 The @code{__thread} specifier shall be used only with
7559 @node C++98 Thread-Local Edits
7560 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
7562 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
7563 that document the exact semantics of the language extension.
7567 @b{[intro.execution]}
7569 New text after paragraph 4
7572 A @dfn{thread} is a flow of control within the abstract machine.
7573 It is implementation defined whether or not there may be more than
7577 New text after paragraph 7
7580 It is unspecified whether additional action must be taken to
7581 ensure when and whether side effects are visible to other threads.
7587 Add @code{__thread}.
7590 @b{[basic.start.main]}
7592 Add after paragraph 5
7595 The thread that begins execution at the @code{main} function is called
7596 the @dfn{main thread}. It is implementation defined how functions
7597 beginning threads other than the main thread are designated or typed.
7598 A function so designated, as well as the @code{main} function, is called
7599 a @dfn{thread startup function}. It is implementation defined what
7600 happens if a thread startup function returns. It is implementation
7601 defined what happens to other threads when any thread calls @code{exit}.
7605 @b{[basic.start.init]}
7607 Add after paragraph 4
7610 The storage for an object of thread storage duration shall be
7611 statically initialized before the first statement of the thread startup
7612 function. An object of thread storage duration shall not require
7613 dynamic initialization.
7617 @b{[basic.start.term]}
7619 Add after paragraph 3
7622 The type of an object with thread storage duration shall not have a
7623 non-trivial destructor, nor shall it be an array type whose elements
7624 (directly or indirectly) have non-trivial destructors.
7630 Add ``thread storage duration'' to the list in paragraph 1.
7635 Thread, static, and automatic storage durations are associated with
7636 objects introduced by declarations [@dots{}].
7639 Add @code{__thread} to the list of specifiers in paragraph 3.
7642 @b{[basic.stc.thread]}
7644 New section before @b{[basic.stc.static]}
7647 The keyword @code{__thread} applied to a non-local object gives the
7648 object thread storage duration.
7650 A local variable or class data member declared both @code{static}
7651 and @code{__thread} gives the variable or member thread storage
7656 @b{[basic.stc.static]}
7661 All objects which have neither thread storage duration, dynamic
7662 storage duration nor are local [@dots{}].
7668 Add @code{__thread} to the list in paragraph 1.
7673 With the exception of @code{__thread}, at most one
7674 @var{storage-class-specifier} shall appear in a given
7675 @var{decl-specifier-seq}. The @code{__thread} specifier may
7676 be used alone, or immediately following the @code{extern} or
7677 @code{static} specifiers. [@dots{}]
7680 Add after paragraph 5
7683 The @code{__thread} specifier can be applied only to the names of objects
7684 and to anonymous unions.
7690 Add after paragraph 6
7693 Non-@code{static} members shall not be @code{__thread}.
7697 @node C++ Extensions
7698 @chapter Extensions to the C++ Language
7699 @cindex extensions, C++ language
7700 @cindex C++ language extensions
7702 The GNU compiler provides these extensions to the C++ language (and you
7703 can also use most of the C language extensions in your C++ programs). If you
7704 want to write code that checks whether these features are available, you can
7705 test for the GNU compiler the same way as for C programs: check for a
7706 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
7707 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
7708 Predefined Macros,cpp,The GNU C Preprocessor}).
7711 * Min and Max:: C++ Minimum and maximum operators.
7712 * Volatiles:: What constitutes an access to a volatile object.
7713 * Restricted Pointers:: C99 restricted pointers and references.
7714 * Vague Linkage:: Where G++ puts inlines, vtables and such.
7715 * C++ Interface:: You can use a single C++ header file for both
7716 declarations and definitions.
7717 * Template Instantiation:: Methods for ensuring that exactly one copy of
7718 each needed template instantiation is emitted.
7719 * Bound member functions:: You can extract a function pointer to the
7720 method denoted by a @samp{->*} or @samp{.*} expression.
7721 * C++ Attributes:: Variable, function, and type attributes for C++ only.
7722 * Strong Using:: Strong using-directives for namespace composition.
7723 * Java Exceptions:: Tweaking exception handling to work with Java.
7724 * Deprecated Features:: Things will disappear from g++.
7725 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
7729 @section Minimum and Maximum Operators in C++
7731 It is very convenient to have operators which return the ``minimum'' or the
7732 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
7735 @item @var{a} <? @var{b}
7737 @cindex minimum operator
7738 is the @dfn{minimum}, returning the smaller of the numeric values
7739 @var{a} and @var{b};
7741 @item @var{a} >? @var{b}
7743 @cindex maximum operator
7744 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
7748 These operations are not primitive in ordinary C++, since you can
7749 use a macro to return the minimum of two things in C++, as in the
7753 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
7757 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
7758 the minimum value of variables @var{i} and @var{j}.
7760 However, side effects in @code{X} or @code{Y} may cause unintended
7761 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
7762 the smaller counter twice. The GNU C @code{typeof} extension allows you
7763 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
7764 However, writing @code{MIN} and @code{MAX} as macros also forces you to
7765 use function-call notation for a fundamental arithmetic operation.
7766 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
7769 Since @code{<?} and @code{>?} are built into the compiler, they properly
7770 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
7774 @section When is a Volatile Object Accessed?
7775 @cindex accessing volatiles
7776 @cindex volatile read
7777 @cindex volatile write
7778 @cindex volatile access
7780 Both the C and C++ standard have the concept of volatile objects. These
7781 are normally accessed by pointers and used for accessing hardware. The
7782 standards encourage compilers to refrain from optimizations
7783 concerning accesses to volatile objects that it might perform on
7784 non-volatile objects. The C standard leaves it implementation defined
7785 as to what constitutes a volatile access. The C++ standard omits to
7786 specify this, except to say that C++ should behave in a similar manner
7787 to C with respect to volatiles, where possible. The minimum either
7788 standard specifies is that at a sequence point all previous accesses to
7789 volatile objects have stabilized and no subsequent accesses have
7790 occurred. Thus an implementation is free to reorder and combine
7791 volatile accesses which occur between sequence points, but cannot do so
7792 for accesses across a sequence point. The use of volatiles does not
7793 allow you to violate the restriction on updating objects multiple times
7794 within a sequence point.
7796 In most expressions, it is intuitively obvious what is a read and what is
7797 a write. For instance
7800 volatile int *dst = @var{somevalue};
7801 volatile int *src = @var{someothervalue};
7806 will cause a read of the volatile object pointed to by @var{src} and stores the
7807 value into the volatile object pointed to by @var{dst}. There is no
7808 guarantee that these reads and writes are atomic, especially for objects
7809 larger than @code{int}.
7811 Less obvious expressions are where something which looks like an access
7812 is used in a void context. An example would be,
7815 volatile int *src = @var{somevalue};
7819 With C, such expressions are rvalues, and as rvalues cause a read of
7820 the object, GCC interprets this as a read of the volatile being pointed
7821 to. The C++ standard specifies that such expressions do not undergo
7822 lvalue to rvalue conversion, and that the type of the dereferenced
7823 object may be incomplete. The C++ standard does not specify explicitly
7824 that it is this lvalue to rvalue conversion which is responsible for
7825 causing an access. However, there is reason to believe that it is,
7826 because otherwise certain simple expressions become undefined. However,
7827 because it would surprise most programmers, G++ treats dereferencing a
7828 pointer to volatile object of complete type in a void context as a read
7829 of the object. When the object has incomplete type, G++ issues a
7834 struct T @{int m;@};
7835 volatile S *ptr1 = @var{somevalue};
7836 volatile T *ptr2 = @var{somevalue};
7841 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
7842 causes a read of the object pointed to. If you wish to force an error on
7843 the first case, you must force a conversion to rvalue with, for instance
7844 a static cast, @code{static_cast<S>(*ptr1)}.
7846 When using a reference to volatile, G++ does not treat equivalent
7847 expressions as accesses to volatiles, but instead issues a warning that
7848 no volatile is accessed. The rationale for this is that otherwise it
7849 becomes difficult to determine where volatile access occur, and not
7850 possible to ignore the return value from functions returning volatile
7851 references. Again, if you wish to force a read, cast the reference to
7854 @node Restricted Pointers
7855 @section Restricting Pointer Aliasing
7856 @cindex restricted pointers
7857 @cindex restricted references
7858 @cindex restricted this pointer
7860 As with the C front end, G++ understands the C99 feature of restricted pointers,
7861 specified with the @code{__restrict__}, or @code{__restrict} type
7862 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
7863 language flag, @code{restrict} is not a keyword in C++.
7865 In addition to allowing restricted pointers, you can specify restricted
7866 references, which indicate that the reference is not aliased in the local
7870 void fn (int *__restrict__ rptr, int &__restrict__ rref)
7877 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
7878 @var{rref} refers to a (different) unaliased integer.
7880 You may also specify whether a member function's @var{this} pointer is
7881 unaliased by using @code{__restrict__} as a member function qualifier.
7884 void T::fn () __restrict__
7891 Within the body of @code{T::fn}, @var{this} will have the effective
7892 definition @code{T *__restrict__ const this}. Notice that the
7893 interpretation of a @code{__restrict__} member function qualifier is
7894 different to that of @code{const} or @code{volatile} qualifier, in that it
7895 is applied to the pointer rather than the object. This is consistent with
7896 other compilers which implement restricted pointers.
7898 As with all outermost parameter qualifiers, @code{__restrict__} is
7899 ignored in function definition matching. This means you only need to
7900 specify @code{__restrict__} in a function definition, rather than
7901 in a function prototype as well.
7904 @section Vague Linkage
7905 @cindex vague linkage
7907 There are several constructs in C++ which require space in the object
7908 file but are not clearly tied to a single translation unit. We say that
7909 these constructs have ``vague linkage''. Typically such constructs are
7910 emitted wherever they are needed, though sometimes we can be more
7914 @item Inline Functions
7915 Inline functions are typically defined in a header file which can be
7916 included in many different compilations. Hopefully they can usually be
7917 inlined, but sometimes an out-of-line copy is necessary, if the address
7918 of the function is taken or if inlining fails. In general, we emit an
7919 out-of-line copy in all translation units where one is needed. As an
7920 exception, we only emit inline virtual functions with the vtable, since
7921 it will always require a copy.
7923 Local static variables and string constants used in an inline function
7924 are also considered to have vague linkage, since they must be shared
7925 between all inlined and out-of-line instances of the function.
7929 C++ virtual functions are implemented in most compilers using a lookup
7930 table, known as a vtable. The vtable contains pointers to the virtual
7931 functions provided by a class, and each object of the class contains a
7932 pointer to its vtable (or vtables, in some multiple-inheritance
7933 situations). If the class declares any non-inline, non-pure virtual
7934 functions, the first one is chosen as the ``key method'' for the class,
7935 and the vtable is only emitted in the translation unit where the key
7938 @emph{Note:} If the chosen key method is later defined as inline, the
7939 vtable will still be emitted in every translation unit which defines it.
7940 Make sure that any inline virtuals are declared inline in the class
7941 body, even if they are not defined there.
7943 @item type_info objects
7946 C++ requires information about types to be written out in order to
7947 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7948 For polymorphic classes (classes with virtual functions), the type_info
7949 object is written out along with the vtable so that @samp{dynamic_cast}
7950 can determine the dynamic type of a class object at runtime. For all
7951 other types, we write out the type_info object when it is used: when
7952 applying @samp{typeid} to an expression, throwing an object, or
7953 referring to a type in a catch clause or exception specification.
7955 @item Template Instantiations
7956 Most everything in this section also applies to template instantiations,
7957 but there are other options as well.
7958 @xref{Template Instantiation,,Where's the Template?}.
7962 When used with GNU ld version 2.8 or later on an ELF system such as
7963 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
7964 these constructs will be discarded at link time. This is known as
7967 On targets that don't support COMDAT, but do support weak symbols, GCC
7968 will use them. This way one copy will override all the others, but
7969 the unused copies will still take up space in the executable.
7971 For targets which do not support either COMDAT or weak symbols,
7972 most entities with vague linkage will be emitted as local symbols to
7973 avoid duplicate definition errors from the linker. This will not happen
7974 for local statics in inlines, however, as having multiple copies will
7975 almost certainly break things.
7977 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7978 another way to control placement of these constructs.
7981 @section #pragma interface and implementation
7983 @cindex interface and implementation headers, C++
7984 @cindex C++ interface and implementation headers
7985 @cindex pragmas, interface and implementation
7987 @code{#pragma interface} and @code{#pragma implementation} provide the
7988 user with a way of explicitly directing the compiler to emit entities
7989 with vague linkage (and debugging information) in a particular
7992 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
7993 most cases, because of COMDAT support and the ``key method'' heuristic
7994 mentioned in @ref{Vague Linkage}. Using them can actually cause your
7995 program to grow due to unnecesary out-of-line copies of inline
7996 functions. Currently (3.4) the only benefit of these
7997 @code{#pragma}s is reduced duplication of debugging information, and
7998 that should be addressed soon on DWARF 2 targets with the use of
8002 @item #pragma interface
8003 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
8004 @kindex #pragma interface
8005 Use this directive in @emph{header files} that define object classes, to save
8006 space in most of the object files that use those classes. Normally,
8007 local copies of certain information (backup copies of inline member
8008 functions, debugging information, and the internal tables that implement
8009 virtual functions) must be kept in each object file that includes class
8010 definitions. You can use this pragma to avoid such duplication. When a
8011 header file containing @samp{#pragma interface} is included in a
8012 compilation, this auxiliary information will not be generated (unless
8013 the main input source file itself uses @samp{#pragma implementation}).
8014 Instead, the object files will contain references to be resolved at link
8017 The second form of this directive is useful for the case where you have
8018 multiple headers with the same name in different directories. If you
8019 use this form, you must specify the same string to @samp{#pragma
8022 @item #pragma implementation
8023 @itemx #pragma implementation "@var{objects}.h"
8024 @kindex #pragma implementation
8025 Use this pragma in a @emph{main input file}, when you want full output from
8026 included header files to be generated (and made globally visible). The
8027 included header file, in turn, should use @samp{#pragma interface}.
8028 Backup copies of inline member functions, debugging information, and the
8029 internal tables used to implement virtual functions are all generated in
8030 implementation files.
8032 @cindex implied @code{#pragma implementation}
8033 @cindex @code{#pragma implementation}, implied
8034 @cindex naming convention, implementation headers
8035 If you use @samp{#pragma implementation} with no argument, it applies to
8036 an include file with the same basename@footnote{A file's @dfn{basename}
8037 was the name stripped of all leading path information and of trailing
8038 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
8039 file. For example, in @file{allclass.cc}, giving just
8040 @samp{#pragma implementation}
8041 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
8043 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
8044 an implementation file whenever you would include it from
8045 @file{allclass.cc} even if you never specified @samp{#pragma
8046 implementation}. This was deemed to be more trouble than it was worth,
8047 however, and disabled.
8049 Use the string argument if you want a single implementation file to
8050 include code from multiple header files. (You must also use
8051 @samp{#include} to include the header file; @samp{#pragma
8052 implementation} only specifies how to use the file---it doesn't actually
8055 There is no way to split up the contents of a single header file into
8056 multiple implementation files.
8059 @cindex inlining and C++ pragmas
8060 @cindex C++ pragmas, effect on inlining
8061 @cindex pragmas in C++, effect on inlining
8062 @samp{#pragma implementation} and @samp{#pragma interface} also have an
8063 effect on function inlining.
8065 If you define a class in a header file marked with @samp{#pragma
8066 interface}, the effect on an inline function defined in that class is
8067 similar to an explicit @code{extern} declaration---the compiler emits
8068 no code at all to define an independent version of the function. Its
8069 definition is used only for inlining with its callers.
8071 @opindex fno-implement-inlines
8072 Conversely, when you include the same header file in a main source file
8073 that declares it as @samp{#pragma implementation}, the compiler emits
8074 code for the function itself; this defines a version of the function
8075 that can be found via pointers (or by callers compiled without
8076 inlining). If all calls to the function can be inlined, you can avoid
8077 emitting the function by compiling with @option{-fno-implement-inlines}.
8078 If any calls were not inlined, you will get linker errors.
8080 @node Template Instantiation
8081 @section Where's the Template?
8082 @cindex template instantiation
8084 C++ templates are the first language feature to require more
8085 intelligence from the environment than one usually finds on a UNIX
8086 system. Somehow the compiler and linker have to make sure that each
8087 template instance occurs exactly once in the executable if it is needed,
8088 and not at all otherwise. There are two basic approaches to this
8089 problem, which are referred to as the Borland model and the Cfront model.
8093 Borland C++ solved the template instantiation problem by adding the code
8094 equivalent of common blocks to their linker; the compiler emits template
8095 instances in each translation unit that uses them, and the linker
8096 collapses them together. The advantage of this model is that the linker
8097 only has to consider the object files themselves; there is no external
8098 complexity to worry about. This disadvantage is that compilation time
8099 is increased because the template code is being compiled repeatedly.
8100 Code written for this model tends to include definitions of all
8101 templates in the header file, since they must be seen to be
8105 The AT&T C++ translator, Cfront, solved the template instantiation
8106 problem by creating the notion of a template repository, an
8107 automatically maintained place where template instances are stored. A
8108 more modern version of the repository works as follows: As individual
8109 object files are built, the compiler places any template definitions and
8110 instantiations encountered in the repository. At link time, the link
8111 wrapper adds in the objects in the repository and compiles any needed
8112 instances that were not previously emitted. The advantages of this
8113 model are more optimal compilation speed and the ability to use the
8114 system linker; to implement the Borland model a compiler vendor also
8115 needs to replace the linker. The disadvantages are vastly increased
8116 complexity, and thus potential for error; for some code this can be
8117 just as transparent, but in practice it can been very difficult to build
8118 multiple programs in one directory and one program in multiple
8119 directories. Code written for this model tends to separate definitions
8120 of non-inline member templates into a separate file, which should be
8121 compiled separately.
8124 When used with GNU ld version 2.8 or later on an ELF system such as
8125 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
8126 Borland model. On other systems, G++ implements neither automatic
8129 A future version of G++ will support a hybrid model whereby the compiler
8130 will emit any instantiations for which the template definition is
8131 included in the compile, and store template definitions and
8132 instantiation context information into the object file for the rest.
8133 The link wrapper will extract that information as necessary and invoke
8134 the compiler to produce the remaining instantiations. The linker will
8135 then combine duplicate instantiations.
8137 In the mean time, you have the following options for dealing with
8138 template instantiations:
8143 Compile your template-using code with @option{-frepo}. The compiler will
8144 generate files with the extension @samp{.rpo} listing all of the
8145 template instantiations used in the corresponding object files which
8146 could be instantiated there; the link wrapper, @samp{collect2}, will
8147 then update the @samp{.rpo} files to tell the compiler where to place
8148 those instantiations and rebuild any affected object files. The
8149 link-time overhead is negligible after the first pass, as the compiler
8150 will continue to place the instantiations in the same files.
8152 This is your best option for application code written for the Borland
8153 model, as it will just work. Code written for the Cfront model will
8154 need to be modified so that the template definitions are available at
8155 one or more points of instantiation; usually this is as simple as adding
8156 @code{#include <tmethods.cc>} to the end of each template header.
8158 For library code, if you want the library to provide all of the template
8159 instantiations it needs, just try to link all of its object files
8160 together; the link will fail, but cause the instantiations to be
8161 generated as a side effect. Be warned, however, that this may cause
8162 conflicts if multiple libraries try to provide the same instantiations.
8163 For greater control, use explicit instantiation as described in the next
8167 @opindex fno-implicit-templates
8168 Compile your code with @option{-fno-implicit-templates} to disable the
8169 implicit generation of template instances, and explicitly instantiate
8170 all the ones you use. This approach requires more knowledge of exactly
8171 which instances you need than do the others, but it's less
8172 mysterious and allows greater control. You can scatter the explicit
8173 instantiations throughout your program, perhaps putting them in the
8174 translation units where the instances are used or the translation units
8175 that define the templates themselves; you can put all of the explicit
8176 instantiations you need into one big file; or you can create small files
8183 template class Foo<int>;
8184 template ostream& operator <<
8185 (ostream&, const Foo<int>&);
8188 for each of the instances you need, and create a template instantiation
8191 If you are using Cfront-model code, you can probably get away with not
8192 using @option{-fno-implicit-templates} when compiling files that don't
8193 @samp{#include} the member template definitions.
8195 If you use one big file to do the instantiations, you may want to
8196 compile it without @option{-fno-implicit-templates} so you get all of the
8197 instances required by your explicit instantiations (but not by any
8198 other files) without having to specify them as well.
8200 G++ has extended the template instantiation syntax given in the ISO
8201 standard to allow forward declaration of explicit instantiations
8202 (with @code{extern}), instantiation of the compiler support data for a
8203 template class (i.e.@: the vtable) without instantiating any of its
8204 members (with @code{inline}), and instantiation of only the static data
8205 members of a template class, without the support data or member
8206 functions (with (@code{static}):
8209 extern template int max (int, int);
8210 inline template class Foo<int>;
8211 static template class Foo<int>;
8215 Do nothing. Pretend G++ does implement automatic instantiation
8216 management. Code written for the Borland model will work fine, but
8217 each translation unit will contain instances of each of the templates it
8218 uses. In a large program, this can lead to an unacceptable amount of code
8222 @node Bound member functions
8223 @section Extracting the function pointer from a bound pointer to member function
8225 @cindex pointer to member function
8226 @cindex bound pointer to member function
8228 In C++, pointer to member functions (PMFs) are implemented using a wide
8229 pointer of sorts to handle all the possible call mechanisms; the PMF
8230 needs to store information about how to adjust the @samp{this} pointer,
8231 and if the function pointed to is virtual, where to find the vtable, and
8232 where in the vtable to look for the member function. If you are using
8233 PMFs in an inner loop, you should really reconsider that decision. If
8234 that is not an option, you can extract the pointer to the function that
8235 would be called for a given object/PMF pair and call it directly inside
8236 the inner loop, to save a bit of time.
8238 Note that you will still be paying the penalty for the call through a
8239 function pointer; on most modern architectures, such a call defeats the
8240 branch prediction features of the CPU@. This is also true of normal
8241 virtual function calls.
8243 The syntax for this extension is
8247 extern int (A::*fp)();
8248 typedef int (*fptr)(A *);
8250 fptr p = (fptr)(a.*fp);
8253 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
8254 no object is needed to obtain the address of the function. They can be
8255 converted to function pointers directly:
8258 fptr p1 = (fptr)(&A::foo);
8261 @opindex Wno-pmf-conversions
8262 You must specify @option{-Wno-pmf-conversions} to use this extension.
8264 @node C++ Attributes
8265 @section C++-Specific Variable, Function, and Type Attributes
8267 Some attributes only make sense for C++ programs.
8270 @item init_priority (@var{priority})
8271 @cindex init_priority attribute
8274 In Standard C++, objects defined at namespace scope are guaranteed to be
8275 initialized in an order in strict accordance with that of their definitions
8276 @emph{in a given translation unit}. No guarantee is made for initializations
8277 across translation units. However, GNU C++ allows users to control the
8278 order of initialization of objects defined at namespace scope with the
8279 @code{init_priority} attribute by specifying a relative @var{priority},
8280 a constant integral expression currently bounded between 101 and 65535
8281 inclusive. Lower numbers indicate a higher priority.
8283 In the following example, @code{A} would normally be created before
8284 @code{B}, but the @code{init_priority} attribute has reversed that order:
8287 Some_Class A __attribute__ ((init_priority (2000)));
8288 Some_Class B __attribute__ ((init_priority (543)));
8292 Note that the particular values of @var{priority} do not matter; only their
8295 @item java_interface
8296 @cindex java_interface attribute
8298 This type attribute informs C++ that the class is a Java interface. It may
8299 only be applied to classes declared within an @code{extern "Java"} block.
8300 Calls to methods declared in this interface will be dispatched using GCJ's
8301 interface table mechanism, instead of regular virtual table dispatch.
8305 See also @xref{Strong Using}.
8308 @section Strong Using
8310 @strong{Caution:} The semantics of this extension are not fully
8311 defined. Users should refrain from using this extension as its
8312 semantics may change subtly over time. It is possible that this
8313 extension wil be removed in future versions of G++.
8315 A using-directive with @code{__attribute ((strong))} is stronger
8316 than a normal using-directive in two ways:
8320 Templates from the used namespace can be specialized as though they were members of the using namespace.
8323 The using namespace is considered an associated namespace of all
8324 templates in the used namespace for purposes of argument-dependent
8328 This is useful for composing a namespace transparently from
8329 implementation namespaces. For example:
8334 template <class T> struct A @{ @};
8336 using namespace debug __attribute ((__strong__));
8337 template <> struct A<int> @{ @}; // ok to specialize
8339 template <class T> void f (A<T>);
8344 f (std::A<float>()); // lookup finds std::f
8349 @node Java Exceptions
8350 @section Java Exceptions
8352 The Java language uses a slightly different exception handling model
8353 from C++. Normally, GNU C++ will automatically detect when you are
8354 writing C++ code that uses Java exceptions, and handle them
8355 appropriately. However, if C++ code only needs to execute destructors
8356 when Java exceptions are thrown through it, GCC will guess incorrectly.
8357 Sample problematic code is:
8360 struct S @{ ~S(); @};
8361 extern void bar(); // is written in Java, and may throw exceptions
8370 The usual effect of an incorrect guess is a link failure, complaining of
8371 a missing routine called @samp{__gxx_personality_v0}.
8373 You can inform the compiler that Java exceptions are to be used in a
8374 translation unit, irrespective of what it might think, by writing
8375 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
8376 @samp{#pragma} must appear before any functions that throw or catch
8377 exceptions, or run destructors when exceptions are thrown through them.
8379 You cannot mix Java and C++ exceptions in the same translation unit. It
8380 is believed to be safe to throw a C++ exception from one file through
8381 another file compiled for the Java exception model, or vice versa, but
8382 there may be bugs in this area.
8384 @node Deprecated Features
8385 @section Deprecated Features
8387 In the past, the GNU C++ compiler was extended to experiment with new
8388 features, at a time when the C++ language was still evolving. Now that
8389 the C++ standard is complete, some of those features are superseded by
8390 superior alternatives. Using the old features might cause a warning in
8391 some cases that the feature will be dropped in the future. In other
8392 cases, the feature might be gone already.
8394 While the list below is not exhaustive, it documents some of the options
8395 that are now deprecated:
8398 @item -fexternal-templates
8399 @itemx -falt-external-templates
8400 These are two of the many ways for G++ to implement template
8401 instantiation. @xref{Template Instantiation}. The C++ standard clearly
8402 defines how template definitions have to be organized across
8403 implementation units. G++ has an implicit instantiation mechanism that
8404 should work just fine for standard-conforming code.
8406 @item -fstrict-prototype
8407 @itemx -fno-strict-prototype
8408 Previously it was possible to use an empty prototype parameter list to
8409 indicate an unspecified number of parameters (like C), rather than no
8410 parameters, as C++ demands. This feature has been removed, except where
8411 it is required for backwards compatibility @xref{Backwards Compatibility}.
8414 The named return value extension has been deprecated, and is now
8417 The use of initializer lists with new expressions has been deprecated,
8418 and is now removed from G++.
8420 Floating and complex non-type template parameters have been deprecated,
8421 and are now removed from G++.
8423 The implicit typename extension has been deprecated and is now
8426 The use of default arguments in function pointers, function typedefs and
8427 and other places where they are not permitted by the standard is
8428 deprecated and will be removed from a future version of G++.
8430 @node Backwards Compatibility
8431 @section Backwards Compatibility
8432 @cindex Backwards Compatibility
8433 @cindex ARM [Annotated C++ Reference Manual]
8435 Now that there is a definitive ISO standard C++, G++ has a specification
8436 to adhere to. The C++ language evolved over time, and features that
8437 used to be acceptable in previous drafts of the standard, such as the ARM
8438 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
8439 compilation of C++ written to such drafts, G++ contains some backwards
8440 compatibilities. @emph{All such backwards compatibility features are
8441 liable to disappear in future versions of G++.} They should be considered
8442 deprecated @xref{Deprecated Features}.
8446 If a variable is declared at for scope, it used to remain in scope until
8447 the end of the scope which contained the for statement (rather than just
8448 within the for scope). G++ retains this, but issues a warning, if such a
8449 variable is accessed outside the for scope.
8451 @item Implicit C language
8452 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
8453 scope to set the language. On such systems, all header files are
8454 implicitly scoped inside a C language scope. Also, an empty prototype
8455 @code{()} will be treated as an unspecified number of arguments, rather
8456 than no arguments, as C++ demands.