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 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
433 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
434 * Long Long:: Double-word integers---@code{long long int}.
435 * Complex:: Data types for complex numbers.
436 * Hex Floats:: Hexadecimal floating-point constants.
437 * Zero Length:: Zero-length arrays.
438 * Variable Length:: Arrays whose length is computed at run time.
439 * Empty Structures:: Structures with no members.
440 * Variadic Macros:: Macros with a variable number of arguments.
441 * Escaped Newlines:: Slightly looser rules for escaped newlines.
442 * Subscripting:: Any array can be subscripted, even if not an lvalue.
443 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
444 * Initializers:: Non-constant initializers.
445 * Compound Literals:: Compound literals give structures, unions
447 * Designated Inits:: Labeling elements of initializers.
448 * Cast to Union:: Casting to union type from any member of the union.
449 * Case Ranges:: `case 1 ... 9' and such.
450 * Mixed Declarations:: Mixing declarations and code.
451 * Function Attributes:: Declaring that functions have no side effects,
452 or that they can never return.
453 * Attribute Syntax:: Formal syntax for attributes.
454 * Function Prototypes:: Prototype declarations and old-style definitions.
455 * C++ Comments:: C++ comments are recognized.
456 * Dollar Signs:: Dollar sign is allowed in identifiers.
457 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
458 * Variable Attributes:: Specifying attributes of variables.
459 * Type Attributes:: Specifying attributes of types.
460 * Alignment:: Inquiring about the alignment of a type or variable.
461 * Inline:: Defining inline functions (as fast as macros).
462 * Extended Asm:: Assembler instructions with C expressions as operands.
463 (With them you can define ``built-in'' functions.)
464 * Constraints:: Constraints for asm operands
465 * Asm Labels:: Specifying the assembler name to use for a C symbol.
466 * Explicit Reg Vars:: Defining variables residing in specified registers.
467 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
468 * Incomplete Enums:: @code{enum foo;}, with details to follow.
469 * Function Names:: Printable strings which are the name of the current
471 * Return Address:: Getting the return or frame address of a function.
472 * Vector Extensions:: Using vector instructions through built-in functions.
473 * Other Builtins:: Other built-in functions.
474 * Target Builtins:: Built-in functions specific to particular targets.
475 * Pragmas:: Pragmas accepted by GCC.
476 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
477 * Thread-Local:: Per-thread variables.
480 @node Statement Exprs
481 @section Statements and Declarations in Expressions
482 @cindex statements inside expressions
483 @cindex declarations inside expressions
484 @cindex expressions containing statements
485 @cindex macros, statements in expressions
487 @c the above section title wrapped and causes an underfull hbox.. i
488 @c changed it from "within" to "in". --mew 4feb93
489 A compound statement enclosed in parentheses may appear as an expression
490 in GNU C@. This allows you to use loops, switches, and local variables
491 within an expression.
493 Recall that a compound statement is a sequence of statements surrounded
494 by braces; in this construct, parentheses go around the braces. For
498 (@{ int y = foo (); int z;
505 is a valid (though slightly more complex than necessary) expression
506 for the absolute value of @code{foo ()}.
508 The last thing in the compound statement should be an expression
509 followed by a semicolon; the value of this subexpression serves as the
510 value of the entire construct. (If you use some other kind of statement
511 last within the braces, the construct has type @code{void}, and thus
512 effectively no value.)
514 This feature is especially useful in making macro definitions ``safe'' (so
515 that they evaluate each operand exactly once). For example, the
516 ``maximum'' function is commonly defined as a macro in standard C as
520 #define max(a,b) ((a) > (b) ? (a) : (b))
524 @cindex side effects, macro argument
525 But this definition computes either @var{a} or @var{b} twice, with bad
526 results if the operand has side effects. In GNU C, if you know the
527 type of the operands (here let's assume @code{int}), you can define
528 the macro safely as follows:
531 #define maxint(a,b) \
532 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
535 Embedded statements are not allowed in constant expressions, such as
536 the value of an enumeration constant, the width of a bit-field, or
537 the initial value of a static variable.
539 If you don't know the type of the operand, you can still do this, but you
540 must use @code{typeof} (@pxref{Typeof}).
542 In G++, the result value of a statement expression undergoes array and
543 function pointer decay, and is returned by value to the enclosing
544 expression. For instance, if @code{A} is a class, then
553 will construct a temporary @code{A} object to hold the result of the
554 statement expression, and that will be used to invoke @code{Foo}.
555 Therefore the @code{this} pointer observed by @code{Foo} will not be the
558 Any temporaries created within a statement within a statement expression
559 will be destroyed at the statement's end. This makes statement
560 expressions inside macros slightly different from function calls. In
561 the latter case temporaries introduced during argument evaluation will
562 be destroyed at the end of the statement that includes the function
563 call. In the statement expression case they will be destroyed during
564 the statement expression. For instance,
567 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
568 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
578 will have different places where temporaries are destroyed. For the
579 @code{macro} case, the temporary @code{X} will be destroyed just after
580 the initialization of @code{b}. In the @code{function} case that
581 temporary will be destroyed when the function returns.
583 These considerations mean that it is probably a bad idea to use
584 statement-expressions of this form in header files that are designed to
585 work with C++. (Note that some versions of the GNU C Library contained
586 header files using statement-expression that lead to precisely this
590 @section Locally Declared Labels
592 @cindex macros, local labels
594 GCC allows you to declare @dfn{local labels} in any nested block
595 scope. A local label is just like an ordinary label, but you can
596 only reference it (with a @code{goto} statement, or by taking its
597 address) within the block in which it was declared.
599 A local label declaration looks like this:
602 __label__ @var{label};
609 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
612 Local label declarations must come at the beginning of the block,
613 before any ordinary declarations or statements.
615 The label declaration defines the label @emph{name}, but does not define
616 the label itself. You must do this in the usual way, with
617 @code{@var{label}:}, within the statements of the statement expression.
619 The local label feature is useful for complex macros. If a macro
620 contains nested loops, a @code{goto} can be useful for breaking out of
621 them. However, an ordinary label whose scope is the whole function
622 cannot be used: if the macro can be expanded several times in one
623 function, the label will be multiply defined in that function. A
624 local label avoids this problem. For example:
627 #define SEARCH(value, array, target) \
630 typeof (target) _SEARCH_target = (target); \
631 typeof (*(array)) *_SEARCH_array = (array); \
634 for (i = 0; i < max; i++) \
635 for (j = 0; j < max; j++) \
636 if (_SEARCH_array[i][j] == _SEARCH_target) \
637 @{ (value) = i; goto found; @} \
643 This could also be written using a statement-expression:
646 #define SEARCH(array, target) \
649 typeof (target) _SEARCH_target = (target); \
650 typeof (*(array)) *_SEARCH_array = (array); \
653 for (i = 0; i < max; i++) \
654 for (j = 0; j < max; j++) \
655 if (_SEARCH_array[i][j] == _SEARCH_target) \
656 @{ value = i; goto found; @} \
663 Local label declarations also make the labels they declare visible to
664 nested functions, if there are any. @xref{Nested Functions}, for details.
666 @node Labels as Values
667 @section Labels as Values
668 @cindex labels as values
669 @cindex computed gotos
670 @cindex goto with computed label
671 @cindex address of a label
673 You can get the address of a label defined in the current function
674 (or a containing function) with the unary operator @samp{&&}. The
675 value has type @code{void *}. This value is a constant and can be used
676 wherever a constant of that type is valid. For example:
684 To use these values, you need to be able to jump to one. This is done
685 with the computed goto statement@footnote{The analogous feature in
686 Fortran is called an assigned goto, but that name seems inappropriate in
687 C, where one can do more than simply store label addresses in label
688 variables.}, @code{goto *@var{exp};}. For example,
695 Any expression of type @code{void *} is allowed.
697 One way of using these constants is in initializing a static array that
698 will serve as a jump table:
701 static void *array[] = @{ &&foo, &&bar, &&hack @};
704 Then you can select a label with indexing, like this:
711 Note that this does not check whether the subscript is in bounds---array
712 indexing in C never does that.
714 Such an array of label values serves a purpose much like that of the
715 @code{switch} statement. The @code{switch} statement is cleaner, so
716 use that rather than an array unless the problem does not fit a
717 @code{switch} statement very well.
719 Another use of label values is in an interpreter for threaded code.
720 The labels within the interpreter function can be stored in the
721 threaded code for super-fast dispatching.
723 You may not use this mechanism to jump to code in a different function.
724 If you do that, totally unpredictable things will happen. The best way to
725 avoid this is to store the label address only in automatic variables and
726 never pass it as an argument.
728 An alternate way to write the above example is
731 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
733 goto *(&&foo + array[i]);
737 This is more friendly to code living in shared libraries, as it reduces
738 the number of dynamic relocations that are needed, and by consequence,
739 allows the data to be read-only.
741 @node Nested Functions
742 @section Nested Functions
743 @cindex nested functions
744 @cindex downward funargs
747 A @dfn{nested function} is a function defined inside another function.
748 (Nested functions are not supported for GNU C++.) The nested function's
749 name is local to the block where it is defined. For example, here we
750 define a nested function named @code{square}, and call it twice:
754 foo (double a, double b)
756 double square (double z) @{ return z * z; @}
758 return square (a) + square (b);
763 The nested function can access all the variables of the containing
764 function that are visible at the point of its definition. This is
765 called @dfn{lexical scoping}. For example, here we show a nested
766 function which uses an inherited variable named @code{offset}:
770 bar (int *array, int offset, int size)
772 int access (int *array, int index)
773 @{ return array[index + offset]; @}
776 for (i = 0; i < size; i++)
777 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
782 Nested function definitions are permitted within functions in the places
783 where variable definitions are allowed; that is, in any block, before
784 the first statement in the block.
786 It is possible to call the nested function from outside the scope of its
787 name by storing its address or passing the address to another function:
790 hack (int *array, int size)
792 void store (int index, int value)
793 @{ array[index] = value; @}
795 intermediate (store, size);
799 Here, the function @code{intermediate} receives the address of
800 @code{store} as an argument. If @code{intermediate} calls @code{store},
801 the arguments given to @code{store} are used to store into @code{array}.
802 But this technique works only so long as the containing function
803 (@code{hack}, in this example) does not exit.
805 If you try to call the nested function through its address after the
806 containing function has exited, all hell will break loose. If you try
807 to call it after a containing scope level has exited, and if it refers
808 to some of the variables that are no longer in scope, you may be lucky,
809 but it's not wise to take the risk. If, however, the nested function
810 does not refer to anything that has gone out of scope, you should be
813 GCC implements taking the address of a nested function using a technique
814 called @dfn{trampolines}. A paper describing them is available as
817 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
819 A nested function can jump to a label inherited from a containing
820 function, provided the label was explicitly declared in the containing
821 function (@pxref{Local Labels}). Such a jump returns instantly to the
822 containing function, exiting the nested function which did the
823 @code{goto} and any intermediate functions as well. Here is an example:
827 bar (int *array, int offset, int size)
830 int access (int *array, int index)
834 return array[index + offset];
838 for (i = 0; i < size; i++)
839 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
843 /* @r{Control comes here from @code{access}
844 if it detects an error.} */
851 A nested function always has internal linkage. Declaring one with
852 @code{extern} is erroneous. If you need to declare the nested function
853 before its definition, use @code{auto} (which is otherwise meaningless
854 for function declarations).
857 bar (int *array, int offset, int size)
860 auto int access (int *, int);
862 int access (int *array, int index)
866 return array[index + offset];
872 @node Constructing Calls
873 @section Constructing Function Calls
874 @cindex constructing calls
875 @cindex forwarding calls
877 Using the built-in functions described below, you can record
878 the arguments a function received, and call another function
879 with the same arguments, without knowing the number or types
882 You can also record the return value of that function call,
883 and later return that value, without knowing what data type
884 the function tried to return (as long as your caller expects
887 However, these built-in functions may interact badly with some
888 sophisticated features or other extensions of the language. It
889 is, therefore, not recommended to use them outside very simple
890 functions acting as mere forwarders for their arguments.
892 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
893 This built-in function returns a pointer to data
894 describing how to perform a call with the same arguments as were passed
895 to the current function.
897 The function saves the arg pointer register, structure value address,
898 and all registers that might be used to pass arguments to a function
899 into a block of memory allocated on the stack. Then it returns the
900 address of that block.
903 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
904 This built-in function invokes @var{function}
905 with a copy of the parameters described by @var{arguments}
908 The value of @var{arguments} should be the value returned by
909 @code{__builtin_apply_args}. The argument @var{size} specifies the size
910 of the stack argument data, in bytes.
912 This function returns a pointer to data describing
913 how to return whatever value was returned by @var{function}. The data
914 is saved in a block of memory allocated on the stack.
916 It is not always simple to compute the proper value for @var{size}. The
917 value is used by @code{__builtin_apply} to compute the amount of data
918 that should be pushed on the stack and copied from the incoming argument
922 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
923 This built-in function returns the value described by @var{result} from
924 the containing function. You should specify, for @var{result}, a value
925 returned by @code{__builtin_apply}.
929 @section Referring to a Type with @code{typeof}
932 @cindex macros, types of arguments
934 Another way to refer to the type of an expression is with @code{typeof}.
935 The syntax of using of this keyword looks like @code{sizeof}, but the
936 construct acts semantically like a type name defined with @code{typedef}.
938 There are two ways of writing the argument to @code{typeof}: with an
939 expression or with a type. Here is an example with an expression:
946 This assumes that @code{x} is an array of pointers to functions;
947 the type described is that of the values of the functions.
949 Here is an example with a typename as the argument:
956 Here the type described is that of pointers to @code{int}.
958 If you are writing a header file that must work when included in ISO C
959 programs, write @code{__typeof__} instead of @code{typeof}.
960 @xref{Alternate Keywords}.
962 A @code{typeof}-construct can be used anywhere a typedef name could be
963 used. For example, you can use it in a declaration, in a cast, or inside
964 of @code{sizeof} or @code{typeof}.
966 @code{typeof} is often useful in conjunction with the
967 statements-within-expressions feature. Here is how the two together can
968 be used to define a safe ``maximum'' macro that operates on any
969 arithmetic type and evaluates each of its arguments exactly once:
973 (@{ typeof (a) _a = (a); \
974 typeof (b) _b = (b); \
975 _a > _b ? _a : _b; @})
978 @cindex underscores in variables in macros
979 @cindex @samp{_} in variables in macros
980 @cindex local variables in macros
981 @cindex variables, local, in macros
982 @cindex macros, local variables in
984 The reason for using names that start with underscores for the local
985 variables is to avoid conflicts with variable names that occur within the
986 expressions that are substituted for @code{a} and @code{b}. Eventually we
987 hope to design a new form of declaration syntax that allows you to declare
988 variables whose scopes start only after their initializers; this will be a
989 more reliable way to prevent such conflicts.
992 Some more examples of the use of @code{typeof}:
996 This declares @code{y} with the type of what @code{x} points to.
1003 This declares @code{y} as an array of such values.
1010 This declares @code{y} as an array of pointers to characters:
1013 typeof (typeof (char *)[4]) y;
1017 It is equivalent to the following traditional C declaration:
1023 To see the meaning of the declaration using @code{typeof}, and why it
1024 might be a useful way to write, let's rewrite it with these macros:
1027 #define pointer(T) typeof(T *)
1028 #define array(T, N) typeof(T [N])
1032 Now the declaration can be rewritten this way:
1035 array (pointer (char), 4) y;
1039 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1040 pointers to @code{char}.
1043 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1044 a more limited extension which permitted one to write
1047 typedef @var{T} = @var{expr};
1051 with the effect of declaring @var{T} to have the type of the expression
1052 @var{expr}. This extension does not work with GCC 3 (versions between
1053 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1054 relies on it should be rewritten to use @code{typeof}:
1057 typedef typeof(@var{expr}) @var{T};
1061 This will work with all versions of GCC@.
1064 @section Generalized Lvalues
1065 @cindex compound expressions as lvalues
1066 @cindex expressions, compound, as lvalues
1067 @cindex conditional expressions as lvalues
1068 @cindex expressions, conditional, as lvalues
1069 @cindex casts as lvalues
1070 @cindex generalized lvalues
1071 @cindex lvalues, generalized
1072 @cindex extensions, @code{?:}
1073 @cindex @code{?:} extensions
1075 Compound expressions, conditional expressions and casts are allowed as
1076 lvalues provided their operands are lvalues. This means that you can take
1077 their addresses or store values into them. All these extensions are
1080 Standard C++ allows compound expressions and conditional expressions
1081 as lvalues, and permits casts to reference type, so use of this
1082 extension is not supported for C++ code.
1084 For example, a compound expression can be assigned, provided the last
1085 expression in the sequence is an lvalue. These two expressions are
1093 Similarly, the address of the compound expression can be taken. These two
1094 expressions are equivalent:
1101 A conditional expression is a valid lvalue if its type is not void and the
1102 true and false branches are both valid lvalues. For example, these two
1103 expressions are equivalent:
1107 (a ? b = 5 : (c = 5))
1110 A cast is a valid lvalue if its operand is an lvalue. This extension
1111 is deprecated. A simple
1112 assignment whose left-hand side is a cast works by converting the
1113 right-hand side first to the specified type, then to the type of the
1114 inner left-hand side expression. After this is stored, the value is
1115 converted back to the specified type to become the value of the
1116 assignment. Thus, if @code{a} has type @code{char *}, the following two
1117 expressions are equivalent:
1121 (int)(a = (char *)(int)5)
1124 An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
1125 performs the arithmetic using the type resulting from the cast, and then
1126 continues as in the previous case. Therefore, these two expressions are
1131 (int)(a = (char *)(int) ((int)a + 5))
1134 You cannot take the address of an lvalue cast, because the use of its
1135 address would not work out coherently. Suppose that @code{&(int)f} were
1136 permitted, where @code{f} has type @code{float}. Then the following
1137 statement would try to store an integer bit-pattern where a floating
1138 point number belongs:
1144 This is quite different from what @code{(int)f = 1} would do---that
1145 would convert 1 to floating point and store it. Rather than cause this
1146 inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
1148 If you really do want an @code{int *} pointer with the address of
1149 @code{f}, you can simply write @code{(int *)&f}.
1152 @section Conditionals with Omitted Operands
1153 @cindex conditional expressions, extensions
1154 @cindex omitted middle-operands
1155 @cindex middle-operands, omitted
1156 @cindex extensions, @code{?:}
1157 @cindex @code{?:} extensions
1159 The middle operand in a conditional expression may be omitted. Then
1160 if the first operand is nonzero, its value is the value of the conditional
1163 Therefore, the expression
1170 has the value of @code{x} if that is nonzero; otherwise, the value of
1173 This example is perfectly equivalent to
1179 @cindex side effect in ?:
1180 @cindex ?: side effect
1182 In this simple case, the ability to omit the middle operand is not
1183 especially useful. When it becomes useful is when the first operand does,
1184 or may (if it is a macro argument), contain a side effect. Then repeating
1185 the operand in the middle would perform the side effect twice. Omitting
1186 the middle operand uses the value already computed without the undesirable
1187 effects of recomputing it.
1190 @section Double-Word Integers
1191 @cindex @code{long long} data types
1192 @cindex double-word arithmetic
1193 @cindex multiprecision arithmetic
1194 @cindex @code{LL} integer suffix
1195 @cindex @code{ULL} integer suffix
1197 ISO C99 supports data types for integers that are at least 64 bits wide,
1198 and as an extension GCC supports them in C89 mode and in C++.
1199 Simply write @code{long long int} for a signed integer, or
1200 @code{unsigned long long int} for an unsigned integer. To make an
1201 integer constant of type @code{long long int}, add the suffix @samp{LL}
1202 to the integer. To make an integer constant of type @code{unsigned long
1203 long int}, add the suffix @samp{ULL} to the integer.
1205 You can use these types in arithmetic like any other integer types.
1206 Addition, subtraction, and bitwise boolean operations on these types
1207 are open-coded on all types of machines. Multiplication is open-coded
1208 if the machine supports fullword-to-doubleword a widening multiply
1209 instruction. Division and shifts are open-coded only on machines that
1210 provide special support. The operations that are not open-coded use
1211 special library routines that come with GCC@.
1213 There may be pitfalls when you use @code{long long} types for function
1214 arguments, unless you declare function prototypes. If a function
1215 expects type @code{int} for its argument, and you pass a value of type
1216 @code{long long int}, confusion will result because the caller and the
1217 subroutine will disagree about the number of bytes for the argument.
1218 Likewise, if the function expects @code{long long int} and you pass
1219 @code{int}. The best way to avoid such problems is to use prototypes.
1222 @section Complex Numbers
1223 @cindex complex numbers
1224 @cindex @code{_Complex} keyword
1225 @cindex @code{__complex__} keyword
1227 ISO C99 supports complex floating data types, and as an extension GCC
1228 supports them in C89 mode and in C++, and supports complex integer data
1229 types which are not part of ISO C99. You can declare complex types
1230 using the keyword @code{_Complex}. As an extension, the older GNU
1231 keyword @code{__complex__} is also supported.
1233 For example, @samp{_Complex double x;} declares @code{x} as a
1234 variable whose real part and imaginary part are both of type
1235 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1236 have real and imaginary parts of type @code{short int}; this is not
1237 likely to be useful, but it shows that the set of complex types is
1240 To write a constant with a complex data type, use the suffix @samp{i} or
1241 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1242 has type @code{_Complex float} and @code{3i} has type
1243 @code{_Complex int}. Such a constant always has a pure imaginary
1244 value, but you can form any complex value you like by adding one to a
1245 real constant. This is a GNU extension; if you have an ISO C99
1246 conforming C library (such as GNU libc), and want to construct complex
1247 constants of floating type, you should include @code{<complex.h>} and
1248 use the macros @code{I} or @code{_Complex_I} instead.
1250 @cindex @code{__real__} keyword
1251 @cindex @code{__imag__} keyword
1252 To extract the real part of a complex-valued expression @var{exp}, write
1253 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1254 extract the imaginary part. This is a GNU extension; for values of
1255 floating type, you should use the ISO C99 functions @code{crealf},
1256 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1257 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1258 built-in functions by GCC@.
1260 @cindex complex conjugation
1261 The operator @samp{~} performs complex conjugation when used on a value
1262 with a complex type. This is a GNU extension; for values of
1263 floating type, you should use the ISO C99 functions @code{conjf},
1264 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1265 provided as built-in functions by GCC@.
1267 GCC can allocate complex automatic variables in a noncontiguous
1268 fashion; it's even possible for the real part to be in a register while
1269 the imaginary part is on the stack (or vice-versa). Only the DWARF2
1270 debug info format can represent this, so use of DWARF2 is recommended.
1271 If you are using the stabs debug info format, GCC describes a noncontiguous
1272 complex variable as if it were two separate variables of noncomplex type.
1273 If the variable's actual name is @code{foo}, the two fictitious
1274 variables are named @code{foo$real} and @code{foo$imag}. You can
1275 examine and set these two fictitious variables with your debugger.
1281 ISO C99 supports floating-point numbers written not only in the usual
1282 decimal notation, such as @code{1.55e1}, but also numbers such as
1283 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1284 supports this in C89 mode (except in some cases when strictly
1285 conforming) and in C++. In that format the
1286 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1287 mandatory. The exponent is a decimal number that indicates the power of
1288 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1295 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1296 is the same as @code{1.55e1}.
1298 Unlike for floating-point numbers in the decimal notation the exponent
1299 is always required in the hexadecimal notation. Otherwise the compiler
1300 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1301 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1302 extension for floating-point constants of type @code{float}.
1305 @section Arrays of Length Zero
1306 @cindex arrays of length zero
1307 @cindex zero-length arrays
1308 @cindex length-zero arrays
1309 @cindex flexible array members
1311 Zero-length arrays are allowed in GNU C@. They are very useful as the
1312 last element of a structure which is really a header for a variable-length
1321 struct line *thisline = (struct line *)
1322 malloc (sizeof (struct line) + this_length);
1323 thisline->length = this_length;
1326 In ISO C90, you would have to give @code{contents} a length of 1, which
1327 means either you waste space or complicate the argument to @code{malloc}.
1329 In ISO C99, you would use a @dfn{flexible array member}, which is
1330 slightly different in syntax and semantics:
1334 Flexible array members are written as @code{contents[]} without
1338 Flexible array members have incomplete type, and so the @code{sizeof}
1339 operator may not be applied. As a quirk of the original implementation
1340 of zero-length arrays, @code{sizeof} evaluates to zero.
1343 Flexible array members may only appear as the last member of a
1344 @code{struct} that is otherwise non-empty.
1347 A structure containing a flexible array member, or a union containing
1348 such a structure (possibly recursively), may not be a member of a
1349 structure or an element of an array. (However, these uses are
1350 permitted by GCC as extensions.)
1353 GCC versions before 3.0 allowed zero-length arrays to be statically
1354 initialized, as if they were flexible arrays. In addition to those
1355 cases that were useful, it also allowed initializations in situations
1356 that would corrupt later data. Non-empty initialization of zero-length
1357 arrays is now treated like any case where there are more initializer
1358 elements than the array holds, in that a suitable warning about "excess
1359 elements in array" is given, and the excess elements (all of them, in
1360 this case) are ignored.
1362 Instead GCC allows static initialization of flexible array members.
1363 This is equivalent to defining a new structure containing the original
1364 structure followed by an array of sufficient size to contain the data.
1365 I.e.@: in the following, @code{f1} is constructed as if it were declared
1371 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1374 struct f1 f1; int data[3];
1375 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1379 The convenience of this extension is that @code{f1} has the desired
1380 type, eliminating the need to consistently refer to @code{f2.f1}.
1382 This has symmetry with normal static arrays, in that an array of
1383 unknown size is also written with @code{[]}.
1385 Of course, this extension only makes sense if the extra data comes at
1386 the end of a top-level object, as otherwise we would be overwriting
1387 data at subsequent offsets. To avoid undue complication and confusion
1388 with initialization of deeply nested arrays, we simply disallow any
1389 non-empty initialization except when the structure is the top-level
1390 object. For example:
1393 struct foo @{ int x; int y[]; @};
1394 struct bar @{ struct foo z; @};
1396 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1397 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1398 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1399 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1402 @node Empty Structures
1403 @section Structures With No Members
1404 @cindex empty structures
1405 @cindex zero-size structures
1407 GCC permits a C structure to have no members:
1414 The structure will have size zero. In C++, empty structures are part
1415 of the language. G++ treats empty structures as if they had a single
1416 member of type @code{char}.
1418 @node Variable Length
1419 @section Arrays of Variable Length
1420 @cindex variable-length arrays
1421 @cindex arrays of variable length
1424 Variable-length automatic arrays are allowed in ISO C99, and as an
1425 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1426 implementation of variable-length arrays does not yet conform in detail
1427 to the ISO C99 standard.) These arrays are
1428 declared like any other automatic arrays, but with a length that is not
1429 a constant expression. The storage is allocated at the point of
1430 declaration and deallocated when the brace-level is exited. For
1435 concat_fopen (char *s1, char *s2, char *mode)
1437 char str[strlen (s1) + strlen (s2) + 1];
1440 return fopen (str, mode);
1444 @cindex scope of a variable length array
1445 @cindex variable-length array scope
1446 @cindex deallocating variable length arrays
1447 Jumping or breaking out of the scope of the array name deallocates the
1448 storage. Jumping into the scope is not allowed; you get an error
1451 @cindex @code{alloca} vs variable-length arrays
1452 You can use the function @code{alloca} to get an effect much like
1453 variable-length arrays. The function @code{alloca} is available in
1454 many other C implementations (but not in all). On the other hand,
1455 variable-length arrays are more elegant.
1457 There are other differences between these two methods. Space allocated
1458 with @code{alloca} exists until the containing @emph{function} returns.
1459 The space for a variable-length array is deallocated as soon as the array
1460 name's scope ends. (If you use both variable-length arrays and
1461 @code{alloca} in the same function, deallocation of a variable-length array
1462 will also deallocate anything more recently allocated with @code{alloca}.)
1464 You can also use variable-length arrays as arguments to functions:
1468 tester (int len, char data[len][len])
1474 The length of an array is computed once when the storage is allocated
1475 and is remembered for the scope of the array in case you access it with
1478 If you want to pass the array first and the length afterward, you can
1479 use a forward declaration in the parameter list---another GNU extension.
1483 tester (int len; char data[len][len], int len)
1489 @cindex parameter forward declaration
1490 The @samp{int len} before the semicolon is a @dfn{parameter forward
1491 declaration}, and it serves the purpose of making the name @code{len}
1492 known when the declaration of @code{data} is parsed.
1494 You can write any number of such parameter forward declarations in the
1495 parameter list. They can be separated by commas or semicolons, but the
1496 last one must end with a semicolon, which is followed by the ``real''
1497 parameter declarations. Each forward declaration must match a ``real''
1498 declaration in parameter name and data type. ISO C99 does not support
1499 parameter forward declarations.
1501 @node Variadic Macros
1502 @section Macros with a Variable Number of Arguments.
1503 @cindex variable number of arguments
1504 @cindex macro with variable arguments
1505 @cindex rest argument (in macro)
1506 @cindex variadic macros
1508 In the ISO C standard of 1999, a macro can be declared to accept a
1509 variable number of arguments much as a function can. The syntax for
1510 defining the macro is similar to that of a function. Here is an
1514 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1517 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1518 such a macro, it represents the zero or more tokens until the closing
1519 parenthesis that ends the invocation, including any commas. This set of
1520 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1521 wherever it appears. See the CPP manual for more information.
1523 GCC has long supported variadic macros, and used a different syntax that
1524 allowed you to give a name to the variable arguments just like any other
1525 argument. Here is an example:
1528 #define debug(format, args...) fprintf (stderr, format, args)
1531 This is in all ways equivalent to the ISO C example above, but arguably
1532 more readable and descriptive.
1534 GNU CPP has two further variadic macro extensions, and permits them to
1535 be used with either of the above forms of macro definition.
1537 In standard C, you are not allowed to leave the variable argument out
1538 entirely; but you are allowed to pass an empty argument. For example,
1539 this invocation is invalid in ISO C, because there is no comma after
1546 GNU CPP permits you to completely omit the variable arguments in this
1547 way. In the above examples, the compiler would complain, though since
1548 the expansion of the macro still has the extra comma after the format
1551 To help solve this problem, CPP behaves specially for variable arguments
1552 used with the token paste operator, @samp{##}. If instead you write
1555 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1558 and if the variable arguments are omitted or empty, the @samp{##}
1559 operator causes the preprocessor to remove the comma before it. If you
1560 do provide some variable arguments in your macro invocation, GNU CPP
1561 does not complain about the paste operation and instead places the
1562 variable arguments after the comma. Just like any other pasted macro
1563 argument, these arguments are not macro expanded.
1565 @node Escaped Newlines
1566 @section Slightly Looser Rules for Escaped Newlines
1567 @cindex escaped newlines
1568 @cindex newlines (escaped)
1570 Recently, the preprocessor has relaxed its treatment of escaped
1571 newlines. Previously, the newline had to immediately follow a
1572 backslash. The current implementation allows whitespace in the form
1573 of spaces, horizontal and vertical tabs, and form feeds between the
1574 backslash and the subsequent newline. The preprocessor issues a
1575 warning, but treats it as a valid escaped newline and combines the two
1576 lines to form a single logical line. This works within comments and
1577 tokens, as well as between tokens. Comments are @emph{not} treated as
1578 whitespace for the purposes of this relaxation, since they have not
1579 yet been replaced with spaces.
1582 @section Non-Lvalue Arrays May Have Subscripts
1583 @cindex subscripting
1584 @cindex arrays, non-lvalue
1586 @cindex subscripting and function values
1587 In ISO C99, arrays that are not lvalues still decay to pointers, and
1588 may be subscripted, although they may not be modified or used after
1589 the next sequence point and the unary @samp{&} operator may not be
1590 applied to them. As an extension, GCC allows such arrays to be
1591 subscripted in C89 mode, though otherwise they do not decay to
1592 pointers outside C99 mode. For example,
1593 this is valid in GNU C though not valid in C89:
1597 struct foo @{int a[4];@};
1603 return f().a[index];
1609 @section Arithmetic on @code{void}- and Function-Pointers
1610 @cindex void pointers, arithmetic
1611 @cindex void, size of pointer to
1612 @cindex function pointers, arithmetic
1613 @cindex function, size of pointer to
1615 In GNU C, addition and subtraction operations are supported on pointers to
1616 @code{void} and on pointers to functions. This is done by treating the
1617 size of a @code{void} or of a function as 1.
1619 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1620 and on function types, and returns 1.
1622 @opindex Wpointer-arith
1623 The option @option{-Wpointer-arith} requests a warning if these extensions
1627 @section Non-Constant Initializers
1628 @cindex initializers, non-constant
1629 @cindex non-constant initializers
1631 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1632 automatic variable are not required to be constant expressions in GNU C@.
1633 Here is an example of an initializer with run-time varying elements:
1636 foo (float f, float g)
1638 float beat_freqs[2] = @{ f-g, f+g @};
1643 @node Compound Literals
1644 @section Compound Literals
1645 @cindex constructor expressions
1646 @cindex initializations in expressions
1647 @cindex structures, constructor expression
1648 @cindex expressions, constructor
1649 @cindex compound literals
1650 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1652 ISO C99 supports compound literals. A compound literal looks like
1653 a cast containing an initializer. Its value is an object of the
1654 type specified in the cast, containing the elements specified in
1655 the initializer; it is an lvalue. As an extension, GCC supports
1656 compound literals in C89 mode and in C++.
1658 Usually, the specified type is a structure. Assume that
1659 @code{struct foo} and @code{structure} are declared as shown:
1662 struct foo @{int a; char b[2];@} structure;
1666 Here is an example of constructing a @code{struct foo} with a compound literal:
1669 structure = ((struct foo) @{x + y, 'a', 0@});
1673 This is equivalent to writing the following:
1677 struct foo temp = @{x + y, 'a', 0@};
1682 You can also construct an array. If all the elements of the compound literal
1683 are (made up of) simple constant expressions, suitable for use in
1684 initializers of objects of static storage duration, then the compound
1685 literal can be coerced to a pointer to its first element and used in
1686 such an initializer, as shown here:
1689 char **foo = (char *[]) @{ "x", "y", "z" @};
1692 Compound literals for scalar types and union types are is
1693 also allowed, but then the compound literal is equivalent
1696 As a GNU extension, GCC allows initialization of objects with static storage
1697 duration by compound literals (which is not possible in ISO C99, because
1698 the initializer is not a constant).
1699 It is handled as if the object was initialized only with the bracket
1700 enclosed list if compound literal's and object types match.
1701 The initializer list of the compound literal must be constant.
1702 If the object being initialized has array type of unknown size, the size is
1703 determined by compound literal size.
1706 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1707 static int y[] = (int []) @{1, 2, 3@};
1708 static int z[] = (int [3]) @{1@};
1712 The above lines are equivalent to the following:
1714 static struct foo x = @{1, 'a', 'b'@};
1715 static int y[] = @{1, 2, 3@};
1716 static int z[] = @{1, 0, 0@};
1719 @node Designated Inits
1720 @section Designated Initializers
1721 @cindex initializers with labeled elements
1722 @cindex labeled elements in initializers
1723 @cindex case labels in initializers
1724 @cindex designated initializers
1726 Standard C89 requires the elements of an initializer to appear in a fixed
1727 order, the same as the order of the elements in the array or structure
1730 In ISO C99 you can give the elements in any order, specifying the array
1731 indices or structure field names they apply to, and GNU C allows this as
1732 an extension in C89 mode as well. This extension is not
1733 implemented in GNU C++.
1735 To specify an array index, write
1736 @samp{[@var{index}] =} before the element value. For example,
1739 int a[6] = @{ [4] = 29, [2] = 15 @};
1746 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1750 The index values must be constant expressions, even if the array being
1751 initialized is automatic.
1753 An alternative syntax for this which has been obsolete since GCC 2.5 but
1754 GCC still accepts is to write @samp{[@var{index}]} before the element
1755 value, with no @samp{=}.
1757 To initialize a range of elements to the same value, write
1758 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1759 extension. For example,
1762 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1766 If the value in it has side-effects, the side-effects will happen only once,
1767 not for each initialized field by the range initializer.
1770 Note that the length of the array is the highest value specified
1773 In a structure initializer, specify the name of a field to initialize
1774 with @samp{.@var{fieldname} =} before the element value. For example,
1775 given the following structure,
1778 struct point @{ int x, y; @};
1782 the following initialization
1785 struct point p = @{ .y = yvalue, .x = xvalue @};
1792 struct point p = @{ xvalue, yvalue @};
1795 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1796 @samp{@var{fieldname}:}, as shown here:
1799 struct point p = @{ y: yvalue, x: xvalue @};
1803 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1804 @dfn{designator}. You can also use a designator (or the obsolete colon
1805 syntax) when initializing a union, to specify which element of the union
1806 should be used. For example,
1809 union foo @{ int i; double d; @};
1811 union foo f = @{ .d = 4 @};
1815 will convert 4 to a @code{double} to store it in the union using
1816 the second element. By contrast, casting 4 to type @code{union foo}
1817 would store it into the union as the integer @code{i}, since it is
1818 an integer. (@xref{Cast to Union}.)
1820 You can combine this technique of naming elements with ordinary C
1821 initialization of successive elements. Each initializer element that
1822 does not have a designator applies to the next consecutive element of the
1823 array or structure. For example,
1826 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1833 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1836 Labeling the elements of an array initializer is especially useful
1837 when the indices are characters or belong to an @code{enum} type.
1842 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1843 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1846 @cindex designator lists
1847 You can also write a series of @samp{.@var{fieldname}} and
1848 @samp{[@var{index}]} designators before an @samp{=} to specify a
1849 nested subobject to initialize; the list is taken relative to the
1850 subobject corresponding to the closest surrounding brace pair. For
1851 example, with the @samp{struct point} declaration above:
1854 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1858 If the same field is initialized multiple times, it will have value from
1859 the last initialization. If any such overridden initialization has
1860 side-effect, it is unspecified whether the side-effect happens or not.
1861 Currently, GCC will discard them and issue a warning.
1864 @section Case Ranges
1866 @cindex ranges in case statements
1868 You can specify a range of consecutive values in a single @code{case} label,
1872 case @var{low} ... @var{high}:
1876 This has the same effect as the proper number of individual @code{case}
1877 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1879 This feature is especially useful for ranges of ASCII character codes:
1885 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1886 it may be parsed wrong when you use it with integer values. For example,
1901 @section Cast to a Union Type
1902 @cindex cast to a union
1903 @cindex union, casting to a
1905 A cast to union type is similar to other casts, except that the type
1906 specified is a union type. You can specify the type either with
1907 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1908 a constructor though, not a cast, and hence does not yield an lvalue like
1909 normal casts. (@xref{Compound Literals}.)
1911 The types that may be cast to the union type are those of the members
1912 of the union. Thus, given the following union and variables:
1915 union foo @{ int i; double d; @};
1921 both @code{x} and @code{y} can be cast to type @code{union foo}.
1923 Using the cast as the right-hand side of an assignment to a variable of
1924 union type is equivalent to storing in a member of the union:
1929 u = (union foo) x @equiv{} u.i = x
1930 u = (union foo) y @equiv{} u.d = y
1933 You can also use the union cast as a function argument:
1936 void hack (union foo);
1938 hack ((union foo) x);
1941 @node Mixed Declarations
1942 @section Mixed Declarations and Code
1943 @cindex mixed declarations and code
1944 @cindex declarations, mixed with code
1945 @cindex code, mixed with declarations
1947 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1948 within compound statements. As an extension, GCC also allows this in
1949 C89 mode. For example, you could do:
1958 Each identifier is visible from where it is declared until the end of
1959 the enclosing block.
1961 @node Function Attributes
1962 @section Declaring Attributes of Functions
1963 @cindex function attributes
1964 @cindex declaring attributes of functions
1965 @cindex functions that never return
1966 @cindex functions that have no side effects
1967 @cindex functions in arbitrary sections
1968 @cindex functions that behave like malloc
1969 @cindex @code{volatile} applied to function
1970 @cindex @code{const} applied to function
1971 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1972 @cindex functions with non-null pointer arguments
1973 @cindex functions that are passed arguments in registers on the 386
1974 @cindex functions that pop the argument stack on the 386
1975 @cindex functions that do not pop the argument stack on the 386
1977 In GNU C, you declare certain things about functions called in your program
1978 which help the compiler optimize function calls and check your code more
1981 The keyword @code{__attribute__} allows you to specify special
1982 attributes when making a declaration. This keyword is followed by an
1983 attribute specification inside double parentheses. The following
1984 attributes are currently defined for functions on all targets:
1985 @code{noreturn}, @code{noinline}, @code{always_inline},
1986 @code{pure}, @code{const}, @code{nothrow},
1987 @code{format}, @code{format_arg}, @code{no_instrument_function},
1988 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1989 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1990 @code{alias}, @code{warn_unused_result} and @code{nonnull}. Several other
1991 attributes are defined for functions on particular target systems. Other
1992 attributes, including @code{section} are supported for variables declarations
1993 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1995 You may also specify attributes with @samp{__} preceding and following
1996 each keyword. This allows you to use them in header files without
1997 being concerned about a possible macro of the same name. For example,
1998 you may use @code{__noreturn__} instead of @code{noreturn}.
2000 @xref{Attribute Syntax}, for details of the exact syntax for using
2004 @cindex @code{noreturn} function attribute
2006 A few standard library functions, such as @code{abort} and @code{exit},
2007 cannot return. GCC knows this automatically. Some programs define
2008 their own functions that never return. You can declare them
2009 @code{noreturn} to tell the compiler this fact. For example,
2013 void fatal () __attribute__ ((noreturn));
2016 fatal (/* @r{@dots{}} */)
2018 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2024 The @code{noreturn} keyword tells the compiler to assume that
2025 @code{fatal} cannot return. It can then optimize without regard to what
2026 would happen if @code{fatal} ever did return. This makes slightly
2027 better code. More importantly, it helps avoid spurious warnings of
2028 uninitialized variables.
2030 The @code{noreturn} keyword does not affect the exceptional path when that
2031 applies: a @code{noreturn}-marked function may still return to the caller
2032 by throwing an exception.
2034 Do not assume that registers saved by the calling function are
2035 restored before calling the @code{noreturn} function.
2037 It does not make sense for a @code{noreturn} function to have a return
2038 type other than @code{void}.
2040 The attribute @code{noreturn} is not implemented in GCC versions
2041 earlier than 2.5. An alternative way to declare that a function does
2042 not return, which works in the current version and in some older
2043 versions, is as follows:
2046 typedef void voidfn ();
2048 volatile voidfn fatal;
2051 @cindex @code{noinline} function attribute
2053 This function attribute prevents a function from being considered for
2056 @cindex @code{always_inline} function attribute
2058 Generally, functions are not inlined unless optimization is specified.
2059 For functions declared inline, this attribute inlines the function even
2060 if no optimization level was specified.
2062 @cindex @code{pure} function attribute
2064 Many functions have no effects except the return value and their
2065 return value depends only on the parameters and/or global variables.
2066 Such a function can be subject
2067 to common subexpression elimination and loop optimization just as an
2068 arithmetic operator would be. These functions should be declared
2069 with the attribute @code{pure}. For example,
2072 int square (int) __attribute__ ((pure));
2076 says that the hypothetical function @code{square} is safe to call
2077 fewer times than the program says.
2079 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2080 Interesting non-pure functions are functions with infinite loops or those
2081 depending on volatile memory or other system resource, that may change between
2082 two consecutive calls (such as @code{feof} in a multithreading environment).
2084 The attribute @code{pure} is not implemented in GCC versions earlier
2086 @cindex @code{const} function attribute
2088 Many functions do not examine any values except their arguments, and
2089 have no effects except the return value. Basically this is just slightly
2090 more strict class than the @code{pure} attribute above, since function is not
2091 allowed to read global memory.
2093 @cindex pointer arguments
2094 Note that a function that has pointer arguments and examines the data
2095 pointed to must @emph{not} be declared @code{const}. Likewise, a
2096 function that calls a non-@code{const} function usually must not be
2097 @code{const}. It does not make sense for a @code{const} function to
2100 The attribute @code{const} is not implemented in GCC versions earlier
2101 than 2.5. An alternative way to declare that a function has no side
2102 effects, which works in the current version and in some older versions,
2106 typedef int intfn ();
2108 extern const intfn square;
2111 This approach does not work in GNU C++ from 2.6.0 on, since the language
2112 specifies that the @samp{const} must be attached to the return value.
2114 @cindex @code{nothrow} function attribute
2116 The @code{nothrow} attribute is used to inform the compiler that a
2117 function cannot throw an exception. For example, most functions in
2118 the standard C library can be guaranteed not to throw an exception
2119 with the notable exceptions of @code{qsort} and @code{bsearch} that
2120 take function pointer arguments. The @code{nothrow} attribute is not
2121 implemented in GCC versions earlier than 3.2.
2123 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2124 @cindex @code{format} function attribute
2126 The @code{format} attribute specifies that a function takes @code{printf},
2127 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2128 should be type-checked against a format string. For example, the
2133 my_printf (void *my_object, const char *my_format, ...)
2134 __attribute__ ((format (printf, 2, 3)));
2138 causes the compiler to check the arguments in calls to @code{my_printf}
2139 for consistency with the @code{printf} style format string argument
2142 The parameter @var{archetype} determines how the format string is
2143 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2144 or @code{strfmon}. (You can also use @code{__printf__},
2145 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2146 parameter @var{string-index} specifies which argument is the format
2147 string argument (starting from 1), while @var{first-to-check} is the
2148 number of the first argument to check against the format string. For
2149 functions where the arguments are not available to be checked (such as
2150 @code{vprintf}), specify the third parameter as zero. In this case the
2151 compiler only checks the format string for consistency. For
2152 @code{strftime} formats, the third parameter is required to be zero.
2153 Since non-static C++ methods have an implicit @code{this} argument, the
2154 arguments of such methods should be counted from two, not one, when
2155 giving values for @var{string-index} and @var{first-to-check}.
2157 In the example above, the format string (@code{my_format}) is the second
2158 argument of the function @code{my_print}, and the arguments to check
2159 start with the third argument, so the correct parameters for the format
2160 attribute are 2 and 3.
2162 @opindex ffreestanding
2163 The @code{format} attribute allows you to identify your own functions
2164 which take format strings as arguments, so that GCC can check the
2165 calls to these functions for errors. The compiler always (unless
2166 @option{-ffreestanding} is used) checks formats
2167 for the standard library functions @code{printf}, @code{fprintf},
2168 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2169 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2170 warnings are requested (using @option{-Wformat}), so there is no need to
2171 modify the header file @file{stdio.h}. In C99 mode, the functions
2172 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2173 @code{vsscanf} are also checked. Except in strictly conforming C
2174 standard modes, the X/Open function @code{strfmon} is also checked as
2175 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2176 @xref{C Dialect Options,,Options Controlling C Dialect}.
2178 @item format_arg (@var{string-index})
2179 @cindex @code{format_arg} function attribute
2180 @opindex Wformat-nonliteral
2181 The @code{format_arg} attribute specifies that a function takes a format
2182 string for a @code{printf}, @code{scanf}, @code{strftime} or
2183 @code{strfmon} style function and modifies it (for example, to translate
2184 it into another language), so the result can be passed to a
2185 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2186 function (with the remaining arguments to the format function the same
2187 as they would have been for the unmodified string). For example, the
2192 my_dgettext (char *my_domain, const char *my_format)
2193 __attribute__ ((format_arg (2)));
2197 causes the compiler to check the arguments in calls to a @code{printf},
2198 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2199 format string argument is a call to the @code{my_dgettext} function, for
2200 consistency with the format string argument @code{my_format}. If the
2201 @code{format_arg} attribute had not been specified, all the compiler
2202 could tell in such calls to format functions would be that the format
2203 string argument is not constant; this would generate a warning when
2204 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2205 without the attribute.
2207 The parameter @var{string-index} specifies which argument is the format
2208 string argument (starting from one). Since non-static C++ methods have
2209 an implicit @code{this} argument, the arguments of such methods should
2210 be counted from two.
2212 The @code{format-arg} attribute allows you to identify your own
2213 functions which modify format strings, so that GCC can check the
2214 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2215 type function whose operands are a call to one of your own function.
2216 The compiler always treats @code{gettext}, @code{dgettext}, and
2217 @code{dcgettext} in this manner except when strict ISO C support is
2218 requested by @option{-ansi} or an appropriate @option{-std} option, or
2219 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2220 Controlling C Dialect}.
2222 @item nonnull (@var{arg-index}, @dots{})
2223 @cindex @code{nonnull} function attribute
2224 The @code{nonnull} attribute specifies that some function parameters should
2225 be non-null pointers. For instance, the declaration:
2229 my_memcpy (void *dest, const void *src, size_t len)
2230 __attribute__((nonnull (1, 2)));
2234 causes the compiler to check that, in calls to @code{my_memcpy},
2235 arguments @var{dest} and @var{src} are non-null. If the compiler
2236 determines that a null pointer is passed in an argument slot marked
2237 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2238 is issued. The compiler may also choose to make optimizations based
2239 on the knowledge that certain function arguments will not be null.
2241 If no argument index list is given to the @code{nonnull} attribute,
2242 all pointer arguments are marked as non-null. To illustrate, the
2243 following declaration is equivalent to the previous example:
2247 my_memcpy (void *dest, const void *src, size_t len)
2248 __attribute__((nonnull));
2251 @item no_instrument_function
2252 @cindex @code{no_instrument_function} function attribute
2253 @opindex finstrument-functions
2254 If @option{-finstrument-functions} is given, profiling function calls will
2255 be generated at entry and exit of most user-compiled functions.
2256 Functions with this attribute will not be so instrumented.
2258 @item section ("@var{section-name}")
2259 @cindex @code{section} function attribute
2260 Normally, the compiler places the code it generates in the @code{text} section.
2261 Sometimes, however, you need additional sections, or you need certain
2262 particular functions to appear in special sections. The @code{section}
2263 attribute specifies that a function lives in a particular section.
2264 For example, the declaration:
2267 extern void foobar (void) __attribute__ ((section ("bar")));
2271 puts the function @code{foobar} in the @code{bar} section.
2273 Some file formats do not support arbitrary sections so the @code{section}
2274 attribute is not available on all platforms.
2275 If you need to map the entire contents of a module to a particular
2276 section, consider using the facilities of the linker instead.
2280 @cindex @code{constructor} function attribute
2281 @cindex @code{destructor} function attribute
2282 The @code{constructor} attribute causes the function to be called
2283 automatically before execution enters @code{main ()}. Similarly, the
2284 @code{destructor} attribute causes the function to be called
2285 automatically after @code{main ()} has completed or @code{exit ()} has
2286 been called. Functions with these attributes are useful for
2287 initializing data that will be used implicitly during the execution of
2290 These attributes are not currently implemented for Objective-C@.
2292 @cindex @code{unused} attribute.
2294 This attribute, attached to a function, means that the function is meant
2295 to be possibly unused. GCC will not produce a warning for this
2298 @cindex @code{used} attribute.
2300 This attribute, attached to a function, means that code must be emitted
2301 for the function even if it appears that the function is not referenced.
2302 This is useful, for example, when the function is referenced only in
2305 @cindex @code{deprecated} attribute.
2307 The @code{deprecated} attribute results in a warning if the function
2308 is used anywhere in the source file. This is useful when identifying
2309 functions that are expected to be removed in a future version of a
2310 program. The warning also includes the location of the declaration
2311 of the deprecated function, to enable users to easily find further
2312 information about why the function is deprecated, or what they should
2313 do instead. Note that the warnings only occurs for uses:
2316 int old_fn () __attribute__ ((deprecated));
2318 int (*fn_ptr)() = old_fn;
2321 results in a warning on line 3 but not line 2.
2323 The @code{deprecated} attribute can also be used for variables and
2324 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2326 @item warn_unused_result
2327 @cindex @code{warn_unused_result} attribute
2328 The @code{warn_unused_result} attribute causes a warning to be emitted
2329 if a caller of the function with this attribute does not use its
2330 return value. This is useful for functions where not checking
2331 the result is either a security problem or always a bug, such as
2335 int fn () __attribute__ ((warn_unused_result));
2338 if (fn () < 0) return -1;
2344 results in warning on line 5.
2347 @cindex @code{weak} attribute
2348 The @code{weak} attribute causes the declaration to be emitted as a weak
2349 symbol rather than a global. This is primarily useful in defining
2350 library functions which can be overridden in user code, though it can
2351 also be used with non-function declarations. Weak symbols are supported
2352 for ELF targets, and also for a.out targets when using the GNU assembler
2356 @cindex @code{malloc} attribute
2357 The @code{malloc} attribute is used to tell the compiler that a function
2358 may be treated as if any non-@code{NULL} pointer it returns cannot
2359 alias any other pointer valid when the function returns.
2360 This will often improve optimization.
2361 Standard functions with this property include @code{malloc} and
2362 @code{calloc}. @code{realloc}-like functions have this property as
2363 long as the old pointer is never referred to (including comparing it
2364 to the new pointer) after the function returns a non-@code{NULL}
2367 @item alias ("@var{target}")
2368 @cindex @code{alias} attribute
2369 The @code{alias} attribute causes the declaration to be emitted as an
2370 alias for another symbol, which must be specified. For instance,
2373 void __f () @{ /* @r{Do something.} */; @}
2374 void f () __attribute__ ((weak, alias ("__f")));
2377 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2378 mangled name for the target must be used.
2380 Not all target machines support this attribute.
2382 @item visibility ("@var{visibility_type}")
2383 @cindex @code{visibility} attribute
2384 The @code{visibility} attribute on ELF targets causes the declaration
2385 to be emitted with default, hidden, protected or internal visibility.
2388 void __attribute__ ((visibility ("protected")))
2389 f () @{ /* @r{Do something.} */; @}
2390 int i __attribute__ ((visibility ("hidden")));
2393 See the ELF gABI for complete details, but the short story is:
2397 Default visibility is the normal case for ELF. This value is
2398 available for the visibility attribute to override other options
2399 that may change the assumed visibility of symbols.
2402 Hidden visibility indicates that the symbol will not be placed into
2403 the dynamic symbol table, so no other @dfn{module} (executable or
2404 shared library) can reference it directly.
2407 Protected visibility indicates that the symbol will be placed in the
2408 dynamic symbol table, but that references within the defining module
2409 will bind to the local symbol. That is, the symbol cannot be overridden
2413 Internal visibility is like hidden visibility, but with additional
2414 processor specific semantics. Unless otherwise specified by the psABI,
2415 GCC defines internal visibility to mean that the function is @emph{never}
2416 called from another module. Note that hidden symbols, while they cannot
2417 be referenced directly by other modules, can be referenced indirectly via
2418 function pointers. By indicating that a symbol cannot be called from
2419 outside the module, GCC may for instance omit the load of a PIC register
2420 since it is known that the calling function loaded the correct value.
2423 Not all ELF targets support this attribute.
2425 @item regparm (@var{number})
2426 @cindex @code{regparm} attribute
2427 @cindex functions that are passed arguments in registers on the 386
2428 On the Intel 386, the @code{regparm} attribute causes the compiler to
2429 pass up to @var{number} integer arguments in registers EAX,
2430 EDX, and ECX instead of on the stack. Functions that take a
2431 variable number of arguments will continue to be passed all of their
2432 arguments on the stack.
2434 Beware that on some ELF systems this attribute is unsuitable for
2435 global functions in shared libraries with lazy binding (which is the
2436 default). Lazy binding will send the first call via resolving code in
2437 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2438 per the standard calling conventions. Solaris 8 is affected by this.
2439 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2440 safe since the loaders there save all registers. (Lazy binding can be
2441 disabled with the linker or the loader if desired, to avoid the
2445 @cindex functions that pop the argument stack on the 386
2446 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2447 assume that the called function will pop off the stack space used to
2448 pass arguments, unless it takes a variable number of arguments.
2451 @cindex functions that pop the argument stack on the 386
2452 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2453 pass the first two arguments in the registers ECX and EDX. Subsequent
2454 arguments are passed on the stack. The called function will pop the
2455 arguments off the stack. If the number of arguments is variable all
2456 arguments are pushed on the stack.
2459 @cindex functions that do pop the argument stack on the 386
2461 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2462 assume that the calling function will pop off the stack space used to
2463 pass arguments. This is
2464 useful to override the effects of the @option{-mrtd} switch.
2466 @item longcall/shortcall
2467 @cindex functions called via pointer on the RS/6000 and PowerPC
2468 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2469 compiler to always call this function via a pointer, just as it would if
2470 the @option{-mlongcall} option had been specified. The @code{shortcall}
2471 attribute causes the compiler not to do this. These attributes override
2472 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2475 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2476 calls are necessary.
2478 @item long_call/short_call
2479 @cindex indirect calls on ARM
2480 This attribute specifies how a particular function is called on
2481 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2482 command line switch and @code{#pragma long_calls} settings. The
2483 @code{long_call} attribute causes the compiler to always call the
2484 function by first loading its address into a register and then using the
2485 contents of that register. The @code{short_call} attribute always places
2486 the offset to the function from the call site into the @samp{BL}
2487 instruction directly.
2489 @item function_vector
2490 @cindex calling functions through the function vector on the H8/300 processors
2491 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2492 function should be called through the function vector. Calling a
2493 function through the function vector will reduce code size, however;
2494 the function vector has a limited size (maximum 128 entries on the H8/300
2495 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2497 You must use GAS and GLD from GNU binutils version 2.7 or later for
2498 this attribute to work correctly.
2501 @cindex interrupt handler functions
2502 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2503 that the specified function is an interrupt handler. The compiler will
2504 generate function entry and exit sequences suitable for use in an
2505 interrupt handler when this attribute is present.
2507 Note, interrupt handlers for the m68k, H8/300, H8/300H, H8S, and SH processors
2508 can be specified via the @code{interrupt_handler} attribute.
2510 Note, on the AVR, interrupts will be enabled inside the function.
2512 Note, for the ARM, you can specify the kind of interrupt to be handled by
2513 adding an optional parameter to the interrupt attribute like this:
2516 void f () __attribute__ ((interrupt ("IRQ")));
2519 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2521 @item interrupt_handler
2522 @cindex interrupt handler functions on the m68k, H8/300 and SH processors
2523 Use this attribute on the m68k, H8/300, H8/300H, H8S, and SH to indicate that
2524 the specified function is an interrupt handler. The compiler will generate
2525 function entry and exit sequences suitable for use in an interrupt
2526 handler when this attribute is present.
2529 Use this attribute on the SH to indicate an @code{interrupt_handler}
2530 function should switch to an alternate stack. It expects a string
2531 argument that names a global variable holding the address of the
2536 void f () __attribute__ ((interrupt_handler,
2537 sp_switch ("alt_stack")));
2541 Use this attribute on the SH for an @code{interrupt_handler} to return using
2542 @code{trapa} instead of @code{rte}. This attribute expects an integer
2543 argument specifying the trap number to be used.
2546 @cindex eight bit data on the H8/300, H8/300H, and H8S
2547 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2548 variable should be placed into the eight bit data section.
2549 The compiler will generate more efficient code for certain operations
2550 on data in the eight bit data area. Note the eight bit data area is limited to
2553 You must use GAS and GLD from GNU binutils version 2.7 or later for
2554 this attribute to work correctly.
2557 @cindex tiny data section on the H8/300H and H8S
2558 Use this attribute on the H8/300H and H8S to indicate that the specified
2559 variable should be placed into the tiny data section.
2560 The compiler will generate more efficient code for loads and stores
2561 on data in the tiny data section. Note the tiny data area is limited to
2562 slightly under 32kbytes of data.
2565 @cindex save all registers on the H8/300, H8/300H, and H8S
2566 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
2567 all registers except the stack pointer should be saved in the prologue
2568 regardless of whether they are used or not.
2571 @cindex signal handler functions on the AVR processors
2572 Use this attribute on the AVR to indicate that the specified
2573 function is a signal handler. The compiler will generate function
2574 entry and exit sequences suitable for use in a signal handler when this
2575 attribute is present. Interrupts will be disabled inside the function.
2578 @cindex function without a prologue/epilogue code
2579 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2580 specified function does not need prologue/epilogue sequences generated by
2581 the compiler. It is up to the programmer to provide these sequences.
2583 @item model (@var{model-name})
2584 @cindex function addressability on the M32R/D
2585 @cindex variable addressability on the IA-64
2587 On the M32R/D, use this attribute to set the addressability of an
2588 object, and of the code generated for a function. The identifier
2589 @var{model-name} is one of @code{small}, @code{medium}, or
2590 @code{large}, representing each of the code models.
2592 Small model objects live in the lower 16MB of memory (so that their
2593 addresses can be loaded with the @code{ld24} instruction), and are
2594 callable with the @code{bl} instruction.
2596 Medium model objects may live anywhere in the 32-bit address space (the
2597 compiler will generate @code{seth/add3} instructions to load their addresses),
2598 and are callable with the @code{bl} instruction.
2600 Large model objects may live anywhere in the 32-bit address space (the
2601 compiler will generate @code{seth/add3} instructions to load their addresses),
2602 and may not be reachable with the @code{bl} instruction (the compiler will
2603 generate the much slower @code{seth/add3/jl} instruction sequence).
2605 On IA-64, use this attribute to set the addressability of an object.
2606 At present, the only supported identifier for @var{model-name} is
2607 @code{small}, indicating addressability via ``small'' (22-bit)
2608 addresses (so that their addresses can be loaded with the @code{addl}
2609 instruction). Caveat: such addressing is by definition not position
2610 independent and hence this attribute must not be used for objects
2611 defined by shared libraries.
2614 @cindex functions which handle memory bank switching
2615 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2616 use a calling convention that takes care of switching memory banks when
2617 entering and leaving a function. This calling convention is also the
2618 default when using the @option{-mlong-calls} option.
2620 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2621 to call and return from a function.
2623 On 68HC11 the compiler will generate a sequence of instructions
2624 to invoke a board-specific routine to switch the memory bank and call the
2625 real function. The board-specific routine simulates a @code{call}.
2626 At the end of a function, it will jump to a board-specific routine
2627 instead of using @code{rts}. The board-specific return routine simulates
2631 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2632 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2633 use the normal calling convention based on @code{jsr} and @code{rts}.
2634 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2638 @cindex @code{__declspec(dllimport)}
2639 On Microsoft Windows targets, the @code{dllimport} attribute causes the compiler
2640 to reference a function or variable via a global pointer to a pointer
2641 that is set up by the Microsoft Windows dll library. The pointer name is formed by
2642 combining @code{_imp__} and the function or variable name. The attribute
2643 implies @code{extern} storage.
2645 Currently, the attribute is ignored for inlined functions. If the
2646 attribute is applied to a symbol @emph{definition}, an error is reported.
2647 If a symbol previously declared @code{dllimport} is later defined, the
2648 attribute is ignored in subsequent references, and a warning is emitted.
2649 The attribute is also overridden by a subsequent declaration as
2652 When applied to C++ classes, the attribute marks non-inlined
2653 member functions and static data members as imports. However, the
2654 attribute is ignored for virtual methods to allow creation of vtables
2657 On cygwin, mingw and arm-pe targets, @code{__declspec(dllimport)} is
2658 recognized as a synonym for @code{__attribute__ ((dllimport))} for
2659 compatibility with other Microsoft Windows compilers.
2661 The use of the @code{dllimport} attribute on functions is not necessary,
2662 but provides a small performance benefit by eliminating a thunk in the
2663 dll. The use of the @code{dllimport} attribute on imported variables was
2664 required on older versions of GNU ld, but can now be avoided by passing
2665 the @option{--enable-auto-import} switch to ld. As with functions, using
2666 the attribute for a variable eliminates a thunk in the dll.
2668 One drawback to using this attribute is that a pointer to a function or
2669 variable marked as dllimport cannot be used as a constant address. The
2670 attribute can be disabled for functions by setting the
2671 @option{-mnop-fun-dllimport} flag.
2674 @cindex @code{__declspec(dllexport)}
2675 On Microsoft Windows targets the @code{dllexport} attribute causes the compiler to
2676 provide a global pointer to a pointer in a dll, so that it can be
2677 referenced with the @code{dllimport} attribute. The pointer name is
2678 formed by combining @code{_imp__} and the function or variable name.
2680 Currently, the @code{dllexport}attribute is ignored for inlined
2681 functions, but export can be forced by using the
2682 @option{-fkeep-inline-functions} flag. The attribute is also ignored for
2685 When applied to C++ classes. the attribute marks defined non-inlined
2686 member functions and static data members as exports. Static consts
2687 initialized in-class are not marked unless they are also defined
2690 On cygwin, mingw and arm-pe targets, @code{__declspec(dllexport)} is
2691 recognized as a synonym for @code{__attribute__ ((dllexport))} for
2692 compatibility with other Microsoft Windows compilers.
2694 Alternative methods for including the symbol in the dll's export table
2695 are to use a .def file with an @code{EXPORTS} section or, with GNU ld,
2696 using the @option{--export-all} linker flag.
2700 You can specify multiple attributes in a declaration by separating them
2701 by commas within the double parentheses or by immediately following an
2702 attribute declaration with another attribute declaration.
2704 @cindex @code{#pragma}, reason for not using
2705 @cindex pragma, reason for not using
2706 Some people object to the @code{__attribute__} feature, suggesting that
2707 ISO C's @code{#pragma} should be used instead. At the time
2708 @code{__attribute__} was designed, there were two reasons for not doing
2713 It is impossible to generate @code{#pragma} commands from a macro.
2716 There is no telling what the same @code{#pragma} might mean in another
2720 These two reasons applied to almost any application that might have been
2721 proposed for @code{#pragma}. It was basically a mistake to use
2722 @code{#pragma} for @emph{anything}.
2724 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2725 to be generated from macros. In addition, a @code{#pragma GCC}
2726 namespace is now in use for GCC-specific pragmas. However, it has been
2727 found convenient to use @code{__attribute__} to achieve a natural
2728 attachment of attributes to their corresponding declarations, whereas
2729 @code{#pragma GCC} is of use for constructs that do not naturally form
2730 part of the grammar. @xref{Other Directives,,Miscellaneous
2731 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2733 @node Attribute Syntax
2734 @section Attribute Syntax
2735 @cindex attribute syntax
2737 This section describes the syntax with which @code{__attribute__} may be
2738 used, and the constructs to which attribute specifiers bind, for the C
2739 language. Some details may vary for C++ and Objective-C@. Because of
2740 infelicities in the grammar for attributes, some forms described here
2741 may not be successfully parsed in all cases.
2743 There are some problems with the semantics of attributes in C++. For
2744 example, there are no manglings for attributes, although they may affect
2745 code generation, so problems may arise when attributed types are used in
2746 conjunction with templates or overloading. Similarly, @code{typeid}
2747 does not distinguish between types with different attributes. Support
2748 for attributes in C++ may be restricted in future to attributes on
2749 declarations only, but not on nested declarators.
2751 @xref{Function Attributes}, for details of the semantics of attributes
2752 applying to functions. @xref{Variable Attributes}, for details of the
2753 semantics of attributes applying to variables. @xref{Type Attributes},
2754 for details of the semantics of attributes applying to structure, union
2755 and enumerated types.
2757 An @dfn{attribute specifier} is of the form
2758 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2759 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2760 each attribute is one of the following:
2764 Empty. Empty attributes are ignored.
2767 A word (which may be an identifier such as @code{unused}, or a reserved
2768 word such as @code{const}).
2771 A word, followed by, in parentheses, parameters for the attribute.
2772 These parameters take one of the following forms:
2776 An identifier. For example, @code{mode} attributes use this form.
2779 An identifier followed by a comma and a non-empty comma-separated list
2780 of expressions. For example, @code{format} attributes use this form.
2783 A possibly empty comma-separated list of expressions. For example,
2784 @code{format_arg} attributes use this form with the list being a single
2785 integer constant expression, and @code{alias} attributes use this form
2786 with the list being a single string constant.
2790 An @dfn{attribute specifier list} is a sequence of one or more attribute
2791 specifiers, not separated by any other tokens.
2793 In GNU C, an attribute specifier list may appear after the colon following a
2794 label, other than a @code{case} or @code{default} label. The only
2795 attribute it makes sense to use after a label is @code{unused}. This
2796 feature is intended for code generated by programs which contains labels
2797 that may be unused but which is compiled with @option{-Wall}. It would
2798 not normally be appropriate to use in it human-written code, though it
2799 could be useful in cases where the code that jumps to the label is
2800 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2801 such placement of attribute lists, as it is permissible for a
2802 declaration, which could begin with an attribute list, to be labelled in
2803 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2804 does not arise there.
2806 An attribute specifier list may appear as part of a @code{struct},
2807 @code{union} or @code{enum} specifier. It may go either immediately
2808 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2809 the closing brace. It is ignored if the content of the structure, union
2810 or enumerated type is not defined in the specifier in which the
2811 attribute specifier list is used---that is, in usages such as
2812 @code{struct __attribute__((foo)) bar} with no following opening brace.
2813 Where attribute specifiers follow the closing brace, they are considered
2814 to relate to the structure, union or enumerated type defined, not to any
2815 enclosing declaration the type specifier appears in, and the type
2816 defined is not complete until after the attribute specifiers.
2817 @c Otherwise, there would be the following problems: a shift/reduce
2818 @c conflict between attributes binding the struct/union/enum and
2819 @c binding to the list of specifiers/qualifiers; and "aligned"
2820 @c attributes could use sizeof for the structure, but the size could be
2821 @c changed later by "packed" attributes.
2823 Otherwise, an attribute specifier appears as part of a declaration,
2824 counting declarations of unnamed parameters and type names, and relates
2825 to that declaration (which may be nested in another declaration, for
2826 example in the case of a parameter declaration), or to a particular declarator
2827 within a declaration. Where an
2828 attribute specifier is applied to a parameter declared as a function or
2829 an array, it should apply to the function or array rather than the
2830 pointer to which the parameter is implicitly converted, but this is not
2831 yet correctly implemented.
2833 Any list of specifiers and qualifiers at the start of a declaration may
2834 contain attribute specifiers, whether or not such a list may in that
2835 context contain storage class specifiers. (Some attributes, however,
2836 are essentially in the nature of storage class specifiers, and only make
2837 sense where storage class specifiers may be used; for example,
2838 @code{section}.) There is one necessary limitation to this syntax: the
2839 first old-style parameter declaration in a function definition cannot
2840 begin with an attribute specifier, because such an attribute applies to
2841 the function instead by syntax described below (which, however, is not
2842 yet implemented in this case). In some other cases, attribute
2843 specifiers are permitted by this grammar but not yet supported by the
2844 compiler. All attribute specifiers in this place relate to the
2845 declaration as a whole. In the obsolescent usage where a type of
2846 @code{int} is implied by the absence of type specifiers, such a list of
2847 specifiers and qualifiers may be an attribute specifier list with no
2848 other specifiers or qualifiers.
2850 An attribute specifier list may appear immediately before a declarator
2851 (other than the first) in a comma-separated list of declarators in a
2852 declaration of more than one identifier using a single list of
2853 specifiers and qualifiers. Such attribute specifiers apply
2854 only to the identifier before whose declarator they appear. For
2858 __attribute__((noreturn)) void d0 (void),
2859 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2864 the @code{noreturn} attribute applies to all the functions
2865 declared; the @code{format} attribute only applies to @code{d1}.
2867 An attribute specifier list may appear immediately before the comma,
2868 @code{=} or semicolon terminating the declaration of an identifier other
2869 than a function definition. At present, such attribute specifiers apply
2870 to the declared object or function, but in future they may attach to the
2871 outermost adjacent declarator. In simple cases there is no difference,
2872 but, for example, in
2875 void (****f)(void) __attribute__((noreturn));
2879 at present the @code{noreturn} attribute applies to @code{f}, which
2880 causes a warning since @code{f} is not a function, but in future it may
2881 apply to the function @code{****f}. The precise semantics of what
2882 attributes in such cases will apply to are not yet specified. Where an
2883 assembler name for an object or function is specified (@pxref{Asm
2884 Labels}), at present the attribute must follow the @code{asm}
2885 specification; in future, attributes before the @code{asm} specification
2886 may apply to the adjacent declarator, and those after it to the declared
2889 An attribute specifier list may, in future, be permitted to appear after
2890 the declarator in a function definition (before any old-style parameter
2891 declarations or the function body).
2893 Attribute specifiers may be mixed with type qualifiers appearing inside
2894 the @code{[]} of a parameter array declarator, in the C99 construct by
2895 which such qualifiers are applied to the pointer to which the array is
2896 implicitly converted. Such attribute specifiers apply to the pointer,
2897 not to the array, but at present this is not implemented and they are
2900 An attribute specifier list may appear at the start of a nested
2901 declarator. At present, there are some limitations in this usage: the
2902 attributes correctly apply to the declarator, but for most individual
2903 attributes the semantics this implies are not implemented.
2904 When attribute specifiers follow the @code{*} of a pointer
2905 declarator, they may be mixed with any type qualifiers present.
2906 The following describes the formal semantics of this syntax. It will make the
2907 most sense if you are familiar with the formal specification of
2908 declarators in the ISO C standard.
2910 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2911 D1}, where @code{T} contains declaration specifiers that specify a type
2912 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2913 contains an identifier @var{ident}. The type specified for @var{ident}
2914 for derived declarators whose type does not include an attribute
2915 specifier is as in the ISO C standard.
2917 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2918 and the declaration @code{T D} specifies the type
2919 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2920 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2921 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2923 If @code{D1} has the form @code{*
2924 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2925 declaration @code{T D} specifies the type
2926 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2927 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2928 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2934 void (__attribute__((noreturn)) ****f) (void);
2938 specifies the type ``pointer to pointer to pointer to pointer to
2939 non-returning function returning @code{void}''. As another example,
2942 char *__attribute__((aligned(8))) *f;
2946 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2947 Note again that this does not work with most attributes; for example,
2948 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2949 is not yet supported.
2951 For compatibility with existing code written for compiler versions that
2952 did not implement attributes on nested declarators, some laxity is
2953 allowed in the placing of attributes. If an attribute that only applies
2954 to types is applied to a declaration, it will be treated as applying to
2955 the type of that declaration. If an attribute that only applies to
2956 declarations is applied to the type of a declaration, it will be treated
2957 as applying to that declaration; and, for compatibility with code
2958 placing the attributes immediately before the identifier declared, such
2959 an attribute applied to a function return type will be treated as
2960 applying to the function type, and such an attribute applied to an array
2961 element type will be treated as applying to the array type. If an
2962 attribute that only applies to function types is applied to a
2963 pointer-to-function type, it will be treated as applying to the pointer
2964 target type; if such an attribute is applied to a function return type
2965 that is not a pointer-to-function type, it will be treated as applying
2966 to the function type.
2968 @node Function Prototypes
2969 @section Prototypes and Old-Style Function Definitions
2970 @cindex function prototype declarations
2971 @cindex old-style function definitions
2972 @cindex promotion of formal parameters
2974 GNU C extends ISO C to allow a function prototype to override a later
2975 old-style non-prototype definition. Consider the following example:
2978 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2985 /* @r{Prototype function declaration.} */
2986 int isroot P((uid_t));
2988 /* @r{Old-style function definition.} */
2990 isroot (x) /* ??? lossage here ??? */
2997 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2998 not allow this example, because subword arguments in old-style
2999 non-prototype definitions are promoted. Therefore in this example the
3000 function definition's argument is really an @code{int}, which does not
3001 match the prototype argument type of @code{short}.
3003 This restriction of ISO C makes it hard to write code that is portable
3004 to traditional C compilers, because the programmer does not know
3005 whether the @code{uid_t} type is @code{short}, @code{int}, or
3006 @code{long}. Therefore, in cases like these GNU C allows a prototype
3007 to override a later old-style definition. More precisely, in GNU C, a
3008 function prototype argument type overrides the argument type specified
3009 by a later old-style definition if the former type is the same as the
3010 latter type before promotion. Thus in GNU C the above example is
3011 equivalent to the following:
3024 GNU C++ does not support old-style function definitions, so this
3025 extension is irrelevant.
3028 @section C++ Style Comments
3030 @cindex C++ comments
3031 @cindex comments, C++ style
3033 In GNU C, you may use C++ style comments, which start with @samp{//} and
3034 continue until the end of the line. Many other C implementations allow
3035 such comments, and they are included in the 1999 C standard. However,
3036 C++ style comments are not recognized if you specify an @option{-std}
3037 option specifying a version of ISO C before C99, or @option{-ansi}
3038 (equivalent to @option{-std=c89}).
3041 @section Dollar Signs in Identifier Names
3043 @cindex dollar signs in identifier names
3044 @cindex identifier names, dollar signs in
3046 In GNU C, you may normally use dollar signs in identifier names.
3047 This is because many traditional C implementations allow such identifiers.
3048 However, dollar signs in identifiers are not supported on a few target
3049 machines, typically because the target assembler does not allow them.
3051 @node Character Escapes
3052 @section The Character @key{ESC} in Constants
3054 You can use the sequence @samp{\e} in a string or character constant to
3055 stand for the ASCII character @key{ESC}.
3058 @section Inquiring on Alignment of Types or Variables
3060 @cindex type alignment
3061 @cindex variable alignment
3063 The keyword @code{__alignof__} allows you to inquire about how an object
3064 is aligned, or the minimum alignment usually required by a type. Its
3065 syntax is just like @code{sizeof}.
3067 For example, if the target machine requires a @code{double} value to be
3068 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3069 This is true on many RISC machines. On more traditional machine
3070 designs, @code{__alignof__ (double)} is 4 or even 2.
3072 Some machines never actually require alignment; they allow reference to any
3073 data type even at an odd address. For these machines, @code{__alignof__}
3074 reports the @emph{recommended} alignment of a type.
3076 If the operand of @code{__alignof__} is an lvalue rather than a type,
3077 its value is the required alignment for its type, taking into account
3078 any minimum alignment specified with GCC's @code{__attribute__}
3079 extension (@pxref{Variable Attributes}). For example, after this
3083 struct foo @{ int x; char y; @} foo1;
3087 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3088 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3090 It is an error to ask for the alignment of an incomplete type.
3092 @node Variable Attributes
3093 @section Specifying Attributes of Variables
3094 @cindex attribute of variables
3095 @cindex variable attributes
3097 The keyword @code{__attribute__} allows you to specify special
3098 attributes of variables or structure fields. This keyword is followed
3099 by an attribute specification inside double parentheses. Some
3100 attributes are currently defined generically for variables.
3101 Other attributes are defined for variables on particular target
3102 systems. Other attributes are available for functions
3103 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3104 Other front ends might define more attributes
3105 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3107 You may also specify attributes with @samp{__} preceding and following
3108 each keyword. This allows you to use them in header files without
3109 being concerned about a possible macro of the same name. For example,
3110 you may use @code{__aligned__} instead of @code{aligned}.
3112 @xref{Attribute Syntax}, for details of the exact syntax for using
3116 @cindex @code{aligned} attribute
3117 @item aligned (@var{alignment})
3118 This attribute specifies a minimum alignment for the variable or
3119 structure field, measured in bytes. For example, the declaration:
3122 int x __attribute__ ((aligned (16))) = 0;
3126 causes the compiler to allocate the global variable @code{x} on a
3127 16-byte boundary. On a 68040, this could be used in conjunction with
3128 an @code{asm} expression to access the @code{move16} instruction which
3129 requires 16-byte aligned operands.
3131 You can also specify the alignment of structure fields. For example, to
3132 create a double-word aligned @code{int} pair, you could write:
3135 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3139 This is an alternative to creating a union with a @code{double} member
3140 that forces the union to be double-word aligned.
3142 As in the preceding examples, you can explicitly specify the alignment
3143 (in bytes) that you wish the compiler to use for a given variable or
3144 structure field. Alternatively, you can leave out the alignment factor
3145 and just ask the compiler to align a variable or field to the maximum
3146 useful alignment for the target machine you are compiling for. For
3147 example, you could write:
3150 short array[3] __attribute__ ((aligned));
3153 Whenever you leave out the alignment factor in an @code{aligned} attribute
3154 specification, the compiler automatically sets the alignment for the declared
3155 variable or field to the largest alignment which is ever used for any data
3156 type on the target machine you are compiling for. Doing this can often make
3157 copy operations more efficient, because the compiler can use whatever
3158 instructions copy the biggest chunks of memory when performing copies to
3159 or from the variables or fields that you have aligned this way.
3161 The @code{aligned} attribute can only increase the alignment; but you
3162 can decrease it by specifying @code{packed} as well. See below.
3164 Note that the effectiveness of @code{aligned} attributes may be limited
3165 by inherent limitations in your linker. On many systems, the linker is
3166 only able to arrange for variables to be aligned up to a certain maximum
3167 alignment. (For some linkers, the maximum supported alignment may
3168 be very very small.) If your linker is only able to align variables
3169 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3170 in an @code{__attribute__} will still only provide you with 8 byte
3171 alignment. See your linker documentation for further information.
3173 @item cleanup (@var{cleanup_function})
3174 @cindex @code{cleanup} attribute
3175 The @code{cleanup} attribute runs a function when the variable goes
3176 out of scope. This attribute can only be applied to auto function
3177 scope variables; it may not be applied to parameters or variables
3178 with static storage duration. The function must take one parameter,
3179 a pointer to a type compatible with the variable. The return value
3180 of the function (if any) is ignored.
3182 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3183 will be run during the stack unwinding that happens during the
3184 processing of the exception. Note that the @code{cleanup} attribute
3185 does not allow the exception to be caught, only to perform an action.
3186 It is undefined what happens if @var{cleanup_function} does not
3191 @cindex @code{common} attribute
3192 @cindex @code{nocommon} attribute
3195 The @code{common} attribute requests GCC to place a variable in
3196 ``common'' storage. The @code{nocommon} attribute requests the
3197 opposite -- to allocate space for it directly.
3199 These attributes override the default chosen by the
3200 @option{-fno-common} and @option{-fcommon} flags respectively.
3203 @cindex @code{deprecated} attribute
3204 The @code{deprecated} attribute results in a warning if the variable
3205 is used anywhere in the source file. This is useful when identifying
3206 variables that are expected to be removed in a future version of a
3207 program. The warning also includes the location of the declaration
3208 of the deprecated variable, to enable users to easily find further
3209 information about why the variable is deprecated, or what they should
3210 do instead. Note that the warning only occurs for uses:
3213 extern int old_var __attribute__ ((deprecated));
3215 int new_fn () @{ return old_var; @}
3218 results in a warning on line 3 but not line 2.
3220 The @code{deprecated} attribute can also be used for functions and
3221 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3223 @item mode (@var{mode})
3224 @cindex @code{mode} attribute
3225 This attribute specifies the data type for the declaration---whichever
3226 type corresponds to the mode @var{mode}. This in effect lets you
3227 request an integer or floating point type according to its width.
3229 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3230 indicate the mode corresponding to a one-byte integer, @samp{word} or
3231 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3232 or @samp{__pointer__} for the mode used to represent pointers.
3235 @cindex @code{packed} attribute
3236 The @code{packed} attribute specifies that a variable or structure field
3237 should have the smallest possible alignment---one byte for a variable,
3238 and one bit for a field, unless you specify a larger value with the
3239 @code{aligned} attribute.
3241 Here is a structure in which the field @code{x} is packed, so that it
3242 immediately follows @code{a}:
3248 int x[2] __attribute__ ((packed));
3252 @item section ("@var{section-name}")
3253 @cindex @code{section} variable attribute
3254 Normally, the compiler places the objects it generates in sections like
3255 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3256 or you need certain particular variables to appear in special sections,
3257 for example to map to special hardware. The @code{section}
3258 attribute specifies that a variable (or function) lives in a particular
3259 section. For example, this small program uses several specific section names:
3262 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3263 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3264 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3265 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3269 /* Initialize stack pointer */
3270 init_sp (stack + sizeof (stack));
3272 /* Initialize initialized data */
3273 memcpy (&init_data, &data, &edata - &data);
3275 /* Turn on the serial ports */
3282 Use the @code{section} attribute with an @emph{initialized} definition
3283 of a @emph{global} variable, as shown in the example. GCC issues
3284 a warning and otherwise ignores the @code{section} attribute in
3285 uninitialized variable declarations.
3287 You may only use the @code{section} attribute with a fully initialized
3288 global definition because of the way linkers work. The linker requires
3289 each object be defined once, with the exception that uninitialized
3290 variables tentatively go in the @code{common} (or @code{bss}) section
3291 and can be multiply ``defined''. You can force a variable to be
3292 initialized with the @option{-fno-common} flag or the @code{nocommon}
3295 Some file formats do not support arbitrary sections so the @code{section}
3296 attribute is not available on all platforms.
3297 If you need to map the entire contents of a module to a particular
3298 section, consider using the facilities of the linker instead.
3301 @cindex @code{shared} variable attribute
3302 On Microsoft Windows, in addition to putting variable definitions in a named
3303 section, the section can also be shared among all running copies of an
3304 executable or DLL@. For example, this small program defines shared data
3305 by putting it in a named section @code{shared} and marking the section
3309 int foo __attribute__((section ("shared"), shared)) = 0;
3314 /* Read and write foo. All running
3315 copies see the same value. */
3321 You may only use the @code{shared} attribute along with @code{section}
3322 attribute with a fully initialized global definition because of the way
3323 linkers work. See @code{section} attribute for more information.
3325 The @code{shared} attribute is only available on Microsoft Windows@.
3327 @item tls_model ("@var{tls_model}")
3328 @cindex @code{tls_model} attribute
3329 The @code{tls_model} attribute sets thread-local storage model
3330 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3331 overriding @code{-ftls-model=} command line switch on a per-variable
3333 The @var{tls_model} argument should be one of @code{global-dynamic},
3334 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3336 Not all targets support this attribute.
3338 @item transparent_union
3339 This attribute, attached to a function parameter which is a union, means
3340 that the corresponding argument may have the type of any union member,
3341 but the argument is passed as if its type were that of the first union
3342 member. For more details see @xref{Type Attributes}. You can also use
3343 this attribute on a @code{typedef} for a union data type; then it
3344 applies to all function parameters with that type.
3347 This attribute, attached to a variable, means that the variable is meant
3348 to be possibly unused. GCC will not produce a warning for this
3351 @item vector_size (@var{bytes})
3352 This attribute specifies the vector size for the variable, measured in
3353 bytes. For example, the declaration:
3356 int foo __attribute__ ((vector_size (16)));
3360 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3361 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3362 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3364 This attribute is only applicable to integral and float scalars,
3365 although arrays, pointers, and function return values are allowed in
3366 conjunction with this construct.
3368 Aggregates with this attribute are invalid, even if they are of the same
3369 size as a corresponding scalar. For example, the declaration:
3372 struct S @{ int a; @};
3373 struct S __attribute__ ((vector_size (16))) foo;
3377 is invalid even if the size of the structure is the same as the size of
3381 The @code{weak} attribute is described in @xref{Function Attributes}.
3384 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3387 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3391 @subsection M32R/D Variable Attributes
3393 One attribute is currently defined for the M32R/D.
3396 @item model (@var{model-name})
3397 @cindex variable addressability on the M32R/D
3398 Use this attribute on the M32R/D to set the addressability of an object.
3399 The identifier @var{model-name} is one of @code{small}, @code{medium},
3400 or @code{large}, representing each of the code models.
3402 Small model objects live in the lower 16MB of memory (so that their
3403 addresses can be loaded with the @code{ld24} instruction).
3405 Medium and large model objects may live anywhere in the 32-bit address space
3406 (the compiler will generate @code{seth/add3} instructions to load their
3410 @subsection i386 Variable Attributes
3412 Two attributes are currently defined for i386 configurations:
3413 @code{ms_struct} and @code{gcc_struct}
3418 @cindex @code{ms_struct} attribute
3419 @cindex @code{gcc_struct} attribute
3421 If @code{packed} is used on a structure, or if bit-fields are used
3422 it may be that the Microsoft ABI packs them differently
3423 than GCC would normally pack them. Particularly when moving packed
3424 data between functions compiled with GCC and the native Microsoft compiler
3425 (either via function call or as data in a file), it may be necessary to access
3428 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3429 compilers to match the native Microsoft compiler.
3432 @node Type Attributes
3433 @section Specifying Attributes of Types
3434 @cindex attribute of types
3435 @cindex type attributes
3437 The keyword @code{__attribute__} allows you to specify special
3438 attributes of @code{struct} and @code{union} types when you define such
3439 types. This keyword is followed by an attribute specification inside
3440 double parentheses. Six attributes are currently defined for types:
3441 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3442 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3443 functions (@pxref{Function Attributes}) and for variables
3444 (@pxref{Variable Attributes}).
3446 You may also specify any one of these attributes with @samp{__}
3447 preceding and following its keyword. This allows you to use these
3448 attributes in header files without being concerned about a possible
3449 macro of the same name. For example, you may use @code{__aligned__}
3450 instead of @code{aligned}.
3452 You may specify the @code{aligned} and @code{transparent_union}
3453 attributes either in a @code{typedef} declaration or just past the
3454 closing curly brace of a complete enum, struct or union type
3455 @emph{definition} and the @code{packed} attribute only past the closing
3456 brace of a definition.
3458 You may also specify attributes between the enum, struct or union
3459 tag and the name of the type rather than after the closing brace.
3461 @xref{Attribute Syntax}, for details of the exact syntax for using
3465 @cindex @code{aligned} attribute
3466 @item aligned (@var{alignment})
3467 This attribute specifies a minimum alignment (in bytes) for variables
3468 of the specified type. For example, the declarations:
3471 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3472 typedef int more_aligned_int __attribute__ ((aligned (8)));
3476 force the compiler to insure (as far as it can) that each variable whose
3477 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3478 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3479 variables of type @code{struct S} aligned to 8-byte boundaries allows
3480 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3481 store) instructions when copying one variable of type @code{struct S} to
3482 another, thus improving run-time efficiency.
3484 Note that the alignment of any given @code{struct} or @code{union} type
3485 is required by the ISO C standard to be at least a perfect multiple of
3486 the lowest common multiple of the alignments of all of the members of
3487 the @code{struct} or @code{union} in question. This means that you @emph{can}
3488 effectively adjust the alignment of a @code{struct} or @code{union}
3489 type by attaching an @code{aligned} attribute to any one of the members
3490 of such a type, but the notation illustrated in the example above is a
3491 more obvious, intuitive, and readable way to request the compiler to
3492 adjust the alignment of an entire @code{struct} or @code{union} type.
3494 As in the preceding example, you can explicitly specify the alignment
3495 (in bytes) that you wish the compiler to use for a given @code{struct}
3496 or @code{union} type. Alternatively, you can leave out the alignment factor
3497 and just ask the compiler to align a type to the maximum
3498 useful alignment for the target machine you are compiling for. For
3499 example, you could write:
3502 struct S @{ short f[3]; @} __attribute__ ((aligned));
3505 Whenever you leave out the alignment factor in an @code{aligned}
3506 attribute specification, the compiler automatically sets the alignment
3507 for the type to the largest alignment which is ever used for any data
3508 type on the target machine you are compiling for. Doing this can often
3509 make copy operations more efficient, because the compiler can use
3510 whatever instructions copy the biggest chunks of memory when performing
3511 copies to or from the variables which have types that you have aligned
3514 In the example above, if the size of each @code{short} is 2 bytes, then
3515 the size of the entire @code{struct S} type is 6 bytes. The smallest
3516 power of two which is greater than or equal to that is 8, so the
3517 compiler sets the alignment for the entire @code{struct S} type to 8
3520 Note that although you can ask the compiler to select a time-efficient
3521 alignment for a given type and then declare only individual stand-alone
3522 objects of that type, the compiler's ability to select a time-efficient
3523 alignment is primarily useful only when you plan to create arrays of
3524 variables having the relevant (efficiently aligned) type. If you
3525 declare or use arrays of variables of an efficiently-aligned type, then
3526 it is likely that your program will also be doing pointer arithmetic (or
3527 subscripting, which amounts to the same thing) on pointers to the
3528 relevant type, and the code that the compiler generates for these
3529 pointer arithmetic operations will often be more efficient for
3530 efficiently-aligned types than for other types.
3532 The @code{aligned} attribute can only increase the alignment; but you
3533 can decrease it by specifying @code{packed} as well. See below.
3535 Note that the effectiveness of @code{aligned} attributes may be limited
3536 by inherent limitations in your linker. On many systems, the linker is
3537 only able to arrange for variables to be aligned up to a certain maximum
3538 alignment. (For some linkers, the maximum supported alignment may
3539 be very very small.) If your linker is only able to align variables
3540 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3541 in an @code{__attribute__} will still only provide you with 8 byte
3542 alignment. See your linker documentation for further information.
3545 This attribute, attached to @code{struct} or @code{union} type
3546 definition, specifies that each member of the structure or union is
3547 placed to minimize the memory required. When attached to an @code{enum}
3548 definition, it indicates that the smallest integral type should be used.
3550 @opindex fshort-enums
3551 Specifying this attribute for @code{struct} and @code{union} types is
3552 equivalent to specifying the @code{packed} attribute on each of the
3553 structure or union members. Specifying the @option{-fshort-enums}
3554 flag on the line is equivalent to specifying the @code{packed}
3555 attribute on all @code{enum} definitions.
3557 In the following example @code{struct my_packed_struct}'s members are
3558 packed closely together, but the internal layout of its @code{s} member
3559 is not packed -- to do that, @code{struct my_unpacked_struct} would need to
3563 struct my_unpacked_struct
3569 struct my_packed_struct __attribute__ ((__packed__))
3573 struct my_unpacked_struct s;
3577 You may only specify this attribute on the definition of a @code{enum},
3578 @code{struct} or @code{union}, not on a @code{typedef} which does not
3579 also define the enumerated type, structure or union.
3581 @item transparent_union
3582 This attribute, attached to a @code{union} type definition, indicates
3583 that any function parameter having that union type causes calls to that
3584 function to be treated in a special way.
3586 First, the argument corresponding to a transparent union type can be of
3587 any type in the union; no cast is required. Also, if the union contains
3588 a pointer type, the corresponding argument can be a null pointer
3589 constant or a void pointer expression; and if the union contains a void
3590 pointer type, the corresponding argument can be any pointer expression.
3591 If the union member type is a pointer, qualifiers like @code{const} on
3592 the referenced type must be respected, just as with normal pointer
3595 Second, the argument is passed to the function using the calling
3596 conventions of the first member of the transparent union, not the calling
3597 conventions of the union itself. All members of the union must have the
3598 same machine representation; this is necessary for this argument passing
3601 Transparent unions are designed for library functions that have multiple
3602 interfaces for compatibility reasons. For example, suppose the
3603 @code{wait} function must accept either a value of type @code{int *} to
3604 comply with Posix, or a value of type @code{union wait *} to comply with
3605 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3606 @code{wait} would accept both kinds of arguments, but it would also
3607 accept any other pointer type and this would make argument type checking
3608 less useful. Instead, @code{<sys/wait.h>} might define the interface
3616 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3618 pid_t wait (wait_status_ptr_t);
3621 This interface allows either @code{int *} or @code{union wait *}
3622 arguments to be passed, using the @code{int *} calling convention.
3623 The program can call @code{wait} with arguments of either type:
3626 int w1 () @{ int w; return wait (&w); @}
3627 int w2 () @{ union wait w; return wait (&w); @}
3630 With this interface, @code{wait}'s implementation might look like this:
3633 pid_t wait (wait_status_ptr_t p)
3635 return waitpid (-1, p.__ip, 0);
3640 When attached to a type (including a @code{union} or a @code{struct}),
3641 this attribute means that variables of that type are meant to appear
3642 possibly unused. GCC will not produce a warning for any variables of
3643 that type, even if the variable appears to do nothing. This is often
3644 the case with lock or thread classes, which are usually defined and then
3645 not referenced, but contain constructors and destructors that have
3646 nontrivial bookkeeping functions.
3649 The @code{deprecated} attribute results in a warning if the type
3650 is used anywhere in the source file. This is useful when identifying
3651 types that are expected to be removed in a future version of a program.
3652 If possible, the warning also includes the location of the declaration
3653 of the deprecated type, to enable users to easily find further
3654 information about why the type is deprecated, or what they should do
3655 instead. Note that the warnings only occur for uses and then only
3656 if the type is being applied to an identifier that itself is not being
3657 declared as deprecated.
3660 typedef int T1 __attribute__ ((deprecated));
3664 typedef T1 T3 __attribute__ ((deprecated));
3665 T3 z __attribute__ ((deprecated));
3668 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3669 warning is issued for line 4 because T2 is not explicitly
3670 deprecated. Line 5 has no warning because T3 is explicitly
3671 deprecated. Similarly for line 6.
3673 The @code{deprecated} attribute can also be used for functions and
3674 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3677 Accesses to objects with types with this attribute are not subjected to
3678 type-based alias analysis, but are instead assumed to be able to alias
3679 any other type of objects, just like the @code{char} type. See
3680 @option{-fstrict-aliasing} for more information on aliasing issues.
3685 typedef short __attribute__((__may_alias__)) short_a;
3691 short_a *b = (short_a *) &a;
3695 if (a == 0x12345678)
3702 If you replaced @code{short_a} with @code{short} in the variable
3703 declaration, the above program would abort when compiled with
3704 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3705 above in recent GCC versions.
3707 @subsection i386 Type Attributes
3709 Two attributes are currently defined for i386 configurations:
3710 @code{ms_struct} and @code{gcc_struct}
3714 @cindex @code{ms_struct}
3715 @cindex @code{gcc_struct}
3717 If @code{packed} is used on a structure, or if bit-fields are used
3718 it may be that the Microsoft ABI packs them differently
3719 than GCC would normally pack them. Particularly when moving packed
3720 data between functions compiled with GCC and the native Microsoft compiler
3721 (either via function call or as data in a file), it may be necessary to access
3724 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3725 compilers to match the native Microsoft compiler.
3728 To specify multiple attributes, separate them by commas within the
3729 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3733 @section An Inline Function is As Fast As a Macro
3734 @cindex inline functions
3735 @cindex integrating function code
3737 @cindex macros, inline alternative
3739 By declaring a function @code{inline}, you can direct GCC to
3740 integrate that function's code into the code for its callers. This
3741 makes execution faster by eliminating the function-call overhead; in
3742 addition, if any of the actual argument values are constant, their known
3743 values may permit simplifications at compile time so that not all of the
3744 inline function's code needs to be included. The effect on code size is
3745 less predictable; object code may be larger or smaller with function
3746 inlining, depending on the particular case. Inlining of functions is an
3747 optimization and it really ``works'' only in optimizing compilation. If
3748 you don't use @option{-O}, no function is really inline.
3750 Inline functions are included in the ISO C99 standard, but there are
3751 currently substantial differences between what GCC implements and what
3752 the ISO C99 standard requires.
3754 To declare a function inline, use the @code{inline} keyword in its
3755 declaration, like this:
3765 (If you are writing a header file to be included in ISO C programs, write
3766 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3767 You can also make all ``simple enough'' functions inline with the option
3768 @option{-finline-functions}.
3771 Note that certain usages in a function definition can make it unsuitable
3772 for inline substitution. Among these usages are: use of varargs, use of
3773 alloca, use of variable sized data types (@pxref{Variable Length}),
3774 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3775 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3776 will warn when a function marked @code{inline} could not be substituted,
3777 and will give the reason for the failure.
3779 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3780 does not affect the linkage of the function.
3782 @cindex automatic @code{inline} for C++ member fns
3783 @cindex @code{inline} automatic for C++ member fns
3784 @cindex member fns, automatically @code{inline}
3785 @cindex C++ member fns, automatically @code{inline}
3786 @opindex fno-default-inline
3787 GCC automatically inlines member functions defined within the class
3788 body of C++ programs even if they are not explicitly declared
3789 @code{inline}. (You can override this with @option{-fno-default-inline};
3790 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3792 @cindex inline functions, omission of
3793 @opindex fkeep-inline-functions
3794 When a function is both inline and @code{static}, if all calls to the
3795 function are integrated into the caller, and the function's address is
3796 never used, then the function's own assembler code is never referenced.
3797 In this case, GCC does not actually output assembler code for the
3798 function, unless you specify the option @option{-fkeep-inline-functions}.
3799 Some calls cannot be integrated for various reasons (in particular,
3800 calls that precede the function's definition cannot be integrated, and
3801 neither can recursive calls within the definition). If there is a
3802 nonintegrated call, then the function is compiled to assembler code as
3803 usual. The function must also be compiled as usual if the program
3804 refers to its address, because that can't be inlined.
3806 @cindex non-static inline function
3807 When an inline function is not @code{static}, then the compiler must assume
3808 that there may be calls from other source files; since a global symbol can
3809 be defined only once in any program, the function must not be defined in
3810 the other source files, so the calls therein cannot be integrated.
3811 Therefore, a non-@code{static} inline function is always compiled on its
3812 own in the usual fashion.
3814 If you specify both @code{inline} and @code{extern} in the function
3815 definition, then the definition is used only for inlining. In no case
3816 is the function compiled on its own, not even if you refer to its
3817 address explicitly. Such an address becomes an external reference, as
3818 if you had only declared the function, and had not defined it.
3820 This combination of @code{inline} and @code{extern} has almost the
3821 effect of a macro. The way to use it is to put a function definition in
3822 a header file with these keywords, and put another copy of the
3823 definition (lacking @code{inline} and @code{extern}) in a library file.
3824 The definition in the header file will cause most calls to the function
3825 to be inlined. If any uses of the function remain, they will refer to
3826 the single copy in the library.
3828 Since GCC eventually will implement ISO C99 semantics for
3829 inline functions, it is best to use @code{static inline} only
3830 to guarantee compatibility. (The
3831 existing semantics will remain available when @option{-std=gnu89} is
3832 specified, but eventually the default will be @option{-std=gnu99} and
3833 that will implement the C99 semantics, though it does not do so yet.)
3835 GCC does not inline any functions when not optimizing unless you specify
3836 the @samp{always_inline} attribute for the function, like this:
3840 inline void foo (const char) __attribute__((always_inline));
3844 @section Assembler Instructions with C Expression Operands
3845 @cindex extended @code{asm}
3846 @cindex @code{asm} expressions
3847 @cindex assembler instructions
3850 In an assembler instruction using @code{asm}, you can specify the
3851 operands of the instruction using C expressions. This means you need not
3852 guess which registers or memory locations will contain the data you want
3855 You must specify an assembler instruction template much like what
3856 appears in a machine description, plus an operand constraint string for
3859 For example, here is how to use the 68881's @code{fsinx} instruction:
3862 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3866 Here @code{angle} is the C expression for the input operand while
3867 @code{result} is that of the output operand. Each has @samp{"f"} as its
3868 operand constraint, saying that a floating point register is required.
3869 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3870 output operands' constraints must use @samp{=}. The constraints use the
3871 same language used in the machine description (@pxref{Constraints}).
3873 Each operand is described by an operand-constraint string followed by
3874 the C expression in parentheses. A colon separates the assembler
3875 template from the first output operand and another separates the last
3876 output operand from the first input, if any. Commas separate the
3877 operands within each group. The total number of operands is currently
3878 limited to 30; this limitation may be lifted in some future version of
3881 If there are no output operands but there are input operands, you must
3882 place two consecutive colons surrounding the place where the output
3885 As of GCC version 3.1, it is also possible to specify input and output
3886 operands using symbolic names which can be referenced within the
3887 assembler code. These names are specified inside square brackets
3888 preceding the constraint string, and can be referenced inside the
3889 assembler code using @code{%[@var{name}]} instead of a percentage sign
3890 followed by the operand number. Using named operands the above example
3894 asm ("fsinx %[angle],%[output]"
3895 : [output] "=f" (result)
3896 : [angle] "f" (angle));
3900 Note that the symbolic operand names have no relation whatsoever to
3901 other C identifiers. You may use any name you like, even those of
3902 existing C symbols, but you must ensure that no two operands within the same
3903 assembler construct use the same symbolic name.
3905 Output operand expressions must be lvalues; the compiler can check this.
3906 The input operands need not be lvalues. The compiler cannot check
3907 whether the operands have data types that are reasonable for the
3908 instruction being executed. It does not parse the assembler instruction
3909 template and does not know what it means or even whether it is valid
3910 assembler input. The extended @code{asm} feature is most often used for
3911 machine instructions the compiler itself does not know exist. If
3912 the output expression cannot be directly addressed (for example, it is a
3913 bit-field), your constraint must allow a register. In that case, GCC
3914 will use the register as the output of the @code{asm}, and then store
3915 that register into the output.
3917 The ordinary output operands must be write-only; GCC will assume that
3918 the values in these operands before the instruction are dead and need
3919 not be generated. Extended asm supports input-output or read-write
3920 operands. Use the constraint character @samp{+} to indicate such an
3921 operand and list it with the output operands. You should only use
3922 read-write operands when the constraints for the operand (or the
3923 operand in which only some of the bits are to be changed) allow a
3926 You may, as an alternative, logically split its function into two
3927 separate operands, one input operand and one write-only output
3928 operand. The connection between them is expressed by constraints
3929 which say they need to be in the same location when the instruction
3930 executes. You can use the same C expression for both operands, or
3931 different expressions. For example, here we write the (fictitious)
3932 @samp{combine} instruction with @code{bar} as its read-only source
3933 operand and @code{foo} as its read-write destination:
3936 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3940 The constraint @samp{"0"} for operand 1 says that it must occupy the
3941 same location as operand 0. A number in constraint is allowed only in
3942 an input operand and it must refer to an output operand.
3944 Only a number in the constraint can guarantee that one operand will be in
3945 the same place as another. The mere fact that @code{foo} is the value
3946 of both operands is not enough to guarantee that they will be in the
3947 same place in the generated assembler code. The following would not
3951 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3954 Various optimizations or reloading could cause operands 0 and 1 to be in
3955 different registers; GCC knows no reason not to do so. For example, the
3956 compiler might find a copy of the value of @code{foo} in one register and
3957 use it for operand 1, but generate the output operand 0 in a different
3958 register (copying it afterward to @code{foo}'s own address). Of course,
3959 since the register for operand 1 is not even mentioned in the assembler
3960 code, the result will not work, but GCC can't tell that.
3962 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3963 the operand number for a matching constraint. For example:
3966 asm ("cmoveq %1,%2,%[result]"
3967 : [result] "=r"(result)
3968 : "r" (test), "r"(new), "[result]"(old));
3971 Some instructions clobber specific hard registers. To describe this,
3972 write a third colon after the input operands, followed by the names of
3973 the clobbered hard registers (given as strings). Here is a realistic
3974 example for the VAX:
3977 asm volatile ("movc3 %0,%1,%2"
3979 : "g" (from), "g" (to), "g" (count)
3980 : "r0", "r1", "r2", "r3", "r4", "r5");
3983 You may not write a clobber description in a way that overlaps with an
3984 input or output operand. For example, you may not have an operand
3985 describing a register class with one member if you mention that register
3986 in the clobber list. Variables declared to live in specific registers
3987 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3988 have no part mentioned in the clobber description.
3989 There is no way for you to specify that an input
3990 operand is modified without also specifying it as an output
3991 operand. Note that if all the output operands you specify are for this
3992 purpose (and hence unused), you will then also need to specify
3993 @code{volatile} for the @code{asm} construct, as described below, to
3994 prevent GCC from deleting the @code{asm} statement as unused.
3996 If you refer to a particular hardware register from the assembler code,
3997 you will probably have to list the register after the third colon to
3998 tell the compiler the register's value is modified. In some assemblers,
3999 the register names begin with @samp{%}; to produce one @samp{%} in the
4000 assembler code, you must write @samp{%%} in the input.
4002 If your assembler instruction can alter the condition code register, add
4003 @samp{cc} to the list of clobbered registers. GCC on some machines
4004 represents the condition codes as a specific hardware register;
4005 @samp{cc} serves to name this register. On other machines, the
4006 condition code is handled differently, and specifying @samp{cc} has no
4007 effect. But it is valid no matter what the machine.
4009 If your assembler instructions access memory in an unpredictable
4010 fashion, add @samp{memory} to the list of clobbered registers. This
4011 will cause GCC to not keep memory values cached in registers across the
4012 assembler instruction and not optimize stores or loads to that memory.
4013 You will also want to add the @code{volatile} keyword if the memory
4014 affected is not listed in the inputs or outputs of the @code{asm}, as
4015 the @samp{memory} clobber does not count as a side-effect of the
4016 @code{asm}. If you know how large the accessed memory is, you can add
4017 it as input or output but if this is not known, you should add
4018 @samp{memory}. As an example, if you access ten bytes of a string, you
4019 can use a memory input like:
4022 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4025 Note that in the following example the memory input is necessary,
4026 otherwise GCC might optimize the store to @code{x} away:
4033 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4034 "=&d" (r) : "a" (y), "m" (*y));
4039 You can put multiple assembler instructions together in a single
4040 @code{asm} template, separated by the characters normally used in assembly
4041 code for the system. A combination that works in most places is a newline
4042 to break the line, plus a tab character to move to the instruction field
4043 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4044 assembler allows semicolons as a line-breaking character. Note that some
4045 assembler dialects use semicolons to start a comment.
4046 The input operands are guaranteed not to use any of the clobbered
4047 registers, and neither will the output operands' addresses, so you can
4048 read and write the clobbered registers as many times as you like. Here
4049 is an example of multiple instructions in a template; it assumes the
4050 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4053 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4055 : "g" (from), "g" (to)
4059 Unless an output operand has the @samp{&} constraint modifier, GCC
4060 may allocate it in the same register as an unrelated input operand, on
4061 the assumption the inputs are consumed before the outputs are produced.
4062 This assumption may be false if the assembler code actually consists of
4063 more than one instruction. In such a case, use @samp{&} for each output
4064 operand that may not overlap an input. @xref{Modifiers}.
4066 If you want to test the condition code produced by an assembler
4067 instruction, you must include a branch and a label in the @code{asm}
4068 construct, as follows:
4071 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4077 This assumes your assembler supports local labels, as the GNU assembler
4078 and most Unix assemblers do.
4080 Speaking of labels, jumps from one @code{asm} to another are not
4081 supported. The compiler's optimizers do not know about these jumps, and
4082 therefore they cannot take account of them when deciding how to
4085 @cindex macros containing @code{asm}
4086 Usually the most convenient way to use these @code{asm} instructions is to
4087 encapsulate them in macros that look like functions. For example,
4091 (@{ double __value, __arg = (x); \
4092 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4097 Here the variable @code{__arg} is used to make sure that the instruction
4098 operates on a proper @code{double} value, and to accept only those
4099 arguments @code{x} which can convert automatically to a @code{double}.
4101 Another way to make sure the instruction operates on the correct data
4102 type is to use a cast in the @code{asm}. This is different from using a
4103 variable @code{__arg} in that it converts more different types. For
4104 example, if the desired type were @code{int}, casting the argument to
4105 @code{int} would accept a pointer with no complaint, while assigning the
4106 argument to an @code{int} variable named @code{__arg} would warn about
4107 using a pointer unless the caller explicitly casts it.
4109 If an @code{asm} has output operands, GCC assumes for optimization
4110 purposes the instruction has no side effects except to change the output
4111 operands. This does not mean instructions with a side effect cannot be
4112 used, but you must be careful, because the compiler may eliminate them
4113 if the output operands aren't used, or move them out of loops, or
4114 replace two with one if they constitute a common subexpression. Also,
4115 if your instruction does have a side effect on a variable that otherwise
4116 appears not to change, the old value of the variable may be reused later
4117 if it happens to be found in a register.
4119 You can prevent an @code{asm} instruction from being deleted, moved
4120 significantly, or combined, by writing the keyword @code{volatile} after
4121 the @code{asm}. For example:
4124 #define get_and_set_priority(new) \
4126 asm volatile ("get_and_set_priority %0, %1" \
4127 : "=g" (__old) : "g" (new)); \
4132 If you write an @code{asm} instruction with no outputs, GCC will know
4133 the instruction has side-effects and will not delete the instruction or
4134 move it outside of loops.
4136 The @code{volatile} keyword indicates that the instruction has
4137 important side-effects. GCC will not delete a volatile @code{asm} if
4138 it is reachable. (The instruction can still be deleted if GCC can
4139 prove that control-flow will never reach the location of the
4140 instruction.) In addition, GCC will not reschedule instructions
4141 across a volatile @code{asm} instruction. For example:
4144 *(volatile int *)addr = foo;
4145 asm volatile ("eieio" : : );
4149 Assume @code{addr} contains the address of a memory mapped device
4150 register. The PowerPC @code{eieio} instruction (Enforce In-order
4151 Execution of I/O) tells the CPU to make sure that the store to that
4152 device register happens before it issues any other I/O@.
4154 Note that even a volatile @code{asm} instruction can be moved in ways
4155 that appear insignificant to the compiler, such as across jump
4156 instructions. You can't expect a sequence of volatile @code{asm}
4157 instructions to remain perfectly consecutive. If you want consecutive
4158 output, use a single @code{asm}. Also, GCC will perform some
4159 optimizations across a volatile @code{asm} instruction; GCC does not
4160 ``forget everything'' when it encounters a volatile @code{asm}
4161 instruction the way some other compilers do.
4163 An @code{asm} instruction without any operands or clobbers (an ``old
4164 style'' @code{asm}) will be treated identically to a volatile
4165 @code{asm} instruction.
4167 It is a natural idea to look for a way to give access to the condition
4168 code left by the assembler instruction. However, when we attempted to
4169 implement this, we found no way to make it work reliably. The problem
4170 is that output operands might need reloading, which would result in
4171 additional following ``store'' instructions. On most machines, these
4172 instructions would alter the condition code before there was time to
4173 test it. This problem doesn't arise for ordinary ``test'' and
4174 ``compare'' instructions because they don't have any output operands.
4176 For reasons similar to those described above, it is not possible to give
4177 an assembler instruction access to the condition code left by previous
4180 If you are writing a header file that should be includable in ISO C
4181 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4184 @subsection Size of an @code{asm}
4186 Some targets require that GCC track the size of each instruction used in
4187 order to generate correct code. Because the final length of an
4188 @code{asm} is only known by the assembler, GCC must make an estimate as
4189 to how big it will be. The estimate is formed by counting the number of
4190 statements in the pattern of the @code{asm} and multiplying that by the
4191 length of the longest instruction on that processor. Statements in the
4192 @code{asm} are identified by newline characters and whatever statement
4193 separator characters are supported by the assembler; on most processors
4194 this is the `@code{;}' character.
4196 Normally, GCC's estimate is perfectly adequate to ensure that correct
4197 code is generated, but it is possible to confuse the compiler if you use
4198 pseudo instructions or assembler macros that expand into multiple real
4199 instructions or if you use assembler directives that expand to more
4200 space in the object file than would be needed for a single instruction.
4201 If this happens then the assembler will produce a diagnostic saying that
4202 a label is unreachable.
4204 @subsection i386 floating point asm operands
4206 There are several rules on the usage of stack-like regs in
4207 asm_operands insns. These rules apply only to the operands that are
4212 Given a set of input regs that die in an asm_operands, it is
4213 necessary to know which are implicitly popped by the asm, and
4214 which must be explicitly popped by gcc.
4216 An input reg that is implicitly popped by the asm must be
4217 explicitly clobbered, unless it is constrained to match an
4221 For any input reg that is implicitly popped by an asm, it is
4222 necessary to know how to adjust the stack to compensate for the pop.
4223 If any non-popped input is closer to the top of the reg-stack than
4224 the implicitly popped reg, it would not be possible to know what the
4225 stack looked like---it's not clear how the rest of the stack ``slides
4228 All implicitly popped input regs must be closer to the top of
4229 the reg-stack than any input that is not implicitly popped.
4231 It is possible that if an input dies in an insn, reload might
4232 use the input reg for an output reload. Consider this example:
4235 asm ("foo" : "=t" (a) : "f" (b));
4238 This asm says that input B is not popped by the asm, and that
4239 the asm pushes a result onto the reg-stack, i.e., the stack is one
4240 deeper after the asm than it was before. But, it is possible that
4241 reload will think that it can use the same reg for both the input and
4242 the output, if input B dies in this insn.
4244 If any input operand uses the @code{f} constraint, all output reg
4245 constraints must use the @code{&} earlyclobber.
4247 The asm above would be written as
4250 asm ("foo" : "=&t" (a) : "f" (b));
4254 Some operands need to be in particular places on the stack. All
4255 output operands fall in this category---there is no other way to
4256 know which regs the outputs appear in unless the user indicates
4257 this in the constraints.
4259 Output operands must specifically indicate which reg an output
4260 appears in after an asm. @code{=f} is not allowed: the operand
4261 constraints must select a class with a single reg.
4264 Output operands may not be ``inserted'' between existing stack regs.
4265 Since no 387 opcode uses a read/write operand, all output operands
4266 are dead before the asm_operands, and are pushed by the asm_operands.
4267 It makes no sense to push anywhere but the top of the reg-stack.
4269 Output operands must start at the top of the reg-stack: output
4270 operands may not ``skip'' a reg.
4273 Some asm statements may need extra stack space for internal
4274 calculations. This can be guaranteed by clobbering stack registers
4275 unrelated to the inputs and outputs.
4279 Here are a couple of reasonable asms to want to write. This asm
4280 takes one input, which is internally popped, and produces two outputs.
4283 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4286 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4287 and replaces them with one output. The user must code the @code{st(1)}
4288 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4291 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4297 @section Controlling Names Used in Assembler Code
4298 @cindex assembler names for identifiers
4299 @cindex names used in assembler code
4300 @cindex identifiers, names in assembler code
4302 You can specify the name to be used in the assembler code for a C
4303 function or variable by writing the @code{asm} (or @code{__asm__})
4304 keyword after the declarator as follows:
4307 int foo asm ("myfoo") = 2;
4311 This specifies that the name to be used for the variable @code{foo} in
4312 the assembler code should be @samp{myfoo} rather than the usual
4315 On systems where an underscore is normally prepended to the name of a C
4316 function or variable, this feature allows you to define names for the
4317 linker that do not start with an underscore.
4319 It does not make sense to use this feature with a non-static local
4320 variable since such variables do not have assembler names. If you are
4321 trying to put the variable in a particular register, see @ref{Explicit
4322 Reg Vars}. GCC presently accepts such code with a warning, but will
4323 probably be changed to issue an error, rather than a warning, in the
4326 You cannot use @code{asm} in this way in a function @emph{definition}; but
4327 you can get the same effect by writing a declaration for the function
4328 before its definition and putting @code{asm} there, like this:
4331 extern func () asm ("FUNC");
4338 It is up to you to make sure that the assembler names you choose do not
4339 conflict with any other assembler symbols. Also, you must not use a
4340 register name; that would produce completely invalid assembler code. GCC
4341 does not as yet have the ability to store static variables in registers.
4342 Perhaps that will be added.
4344 @node Explicit Reg Vars
4345 @section Variables in Specified Registers
4346 @cindex explicit register variables
4347 @cindex variables in specified registers
4348 @cindex specified registers
4349 @cindex registers, global allocation
4351 GNU C allows you to put a few global variables into specified hardware
4352 registers. You can also specify the register in which an ordinary
4353 register variable should be allocated.
4357 Global register variables reserve registers throughout the program.
4358 This may be useful in programs such as programming language
4359 interpreters which have a couple of global variables that are accessed
4363 Local register variables in specific registers do not reserve the
4364 registers. The compiler's data flow analysis is capable of determining
4365 where the specified registers contain live values, and where they are
4366 available for other uses. Stores into local register variables may be deleted
4367 when they appear to be dead according to dataflow analysis. References
4368 to local register variables may be deleted or moved or simplified.
4370 These local variables are sometimes convenient for use with the extended
4371 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4372 output of the assembler instruction directly into a particular register.
4373 (This will work provided the register you specify fits the constraints
4374 specified for that operand in the @code{asm}.)
4382 @node Global Reg Vars
4383 @subsection Defining Global Register Variables
4384 @cindex global register variables
4385 @cindex registers, global variables in
4387 You can define a global register variable in GNU C like this:
4390 register int *foo asm ("a5");
4394 Here @code{a5} is the name of the register which should be used. Choose a
4395 register which is normally saved and restored by function calls on your
4396 machine, so that library routines will not clobber it.
4398 Naturally the register name is cpu-dependent, so you would need to
4399 conditionalize your program according to cpu type. The register
4400 @code{a5} would be a good choice on a 68000 for a variable of pointer
4401 type. On machines with register windows, be sure to choose a ``global''
4402 register that is not affected magically by the function call mechanism.
4404 In addition, operating systems on one type of cpu may differ in how they
4405 name the registers; then you would need additional conditionals. For
4406 example, some 68000 operating systems call this register @code{%a5}.
4408 Eventually there may be a way of asking the compiler to choose a register
4409 automatically, but first we need to figure out how it should choose and
4410 how to enable you to guide the choice. No solution is evident.
4412 Defining a global register variable in a certain register reserves that
4413 register entirely for this use, at least within the current compilation.
4414 The register will not be allocated for any other purpose in the functions
4415 in the current compilation. The register will not be saved and restored by
4416 these functions. Stores into this register are never deleted even if they
4417 would appear to be dead, but references may be deleted or moved or
4420 It is not safe to access the global register variables from signal
4421 handlers, or from more than one thread of control, because the system
4422 library routines may temporarily use the register for other things (unless
4423 you recompile them specially for the task at hand).
4425 @cindex @code{qsort}, and global register variables
4426 It is not safe for one function that uses a global register variable to
4427 call another such function @code{foo} by way of a third function
4428 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4429 different source file in which the variable wasn't declared). This is
4430 because @code{lose} might save the register and put some other value there.
4431 For example, you can't expect a global register variable to be available in
4432 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4433 might have put something else in that register. (If you are prepared to
4434 recompile @code{qsort} with the same global register variable, you can
4435 solve this problem.)
4437 If you want to recompile @code{qsort} or other source files which do not
4438 actually use your global register variable, so that they will not use that
4439 register for any other purpose, then it suffices to specify the compiler
4440 option @option{-ffixed-@var{reg}}. You need not actually add a global
4441 register declaration to their source code.
4443 A function which can alter the value of a global register variable cannot
4444 safely be called from a function compiled without this variable, because it
4445 could clobber the value the caller expects to find there on return.
4446 Therefore, the function which is the entry point into the part of the
4447 program that uses the global register variable must explicitly save and
4448 restore the value which belongs to its caller.
4450 @cindex register variable after @code{longjmp}
4451 @cindex global register after @code{longjmp}
4452 @cindex value after @code{longjmp}
4455 On most machines, @code{longjmp} will restore to each global register
4456 variable the value it had at the time of the @code{setjmp}. On some
4457 machines, however, @code{longjmp} will not change the value of global
4458 register variables. To be portable, the function that called @code{setjmp}
4459 should make other arrangements to save the values of the global register
4460 variables, and to restore them in a @code{longjmp}. This way, the same
4461 thing will happen regardless of what @code{longjmp} does.
4463 All global register variable declarations must precede all function
4464 definitions. If such a declaration could appear after function
4465 definitions, the declaration would be too late to prevent the register from
4466 being used for other purposes in the preceding functions.
4468 Global register variables may not have initial values, because an
4469 executable file has no means to supply initial contents for a register.
4471 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4472 registers, but certain library functions, such as @code{getwd}, as well
4473 as the subroutines for division and remainder, modify g3 and g4. g1 and
4474 g2 are local temporaries.
4476 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4477 Of course, it will not do to use more than a few of those.
4479 @node Local Reg Vars
4480 @subsection Specifying Registers for Local Variables
4481 @cindex local variables, specifying registers
4482 @cindex specifying registers for local variables
4483 @cindex registers for local variables
4485 You can define a local register variable with a specified register
4489 register int *foo asm ("a5");
4493 Here @code{a5} is the name of the register which should be used. Note
4494 that this is the same syntax used for defining global register
4495 variables, but for a local variable it would appear within a function.
4497 Naturally the register name is cpu-dependent, but this is not a
4498 problem, since specific registers are most often useful with explicit
4499 assembler instructions (@pxref{Extended Asm}). Both of these things
4500 generally require that you conditionalize your program according to
4503 In addition, operating systems on one type of cpu may differ in how they
4504 name the registers; then you would need additional conditionals. For
4505 example, some 68000 operating systems call this register @code{%a5}.
4507 Defining such a register variable does not reserve the register; it
4508 remains available for other uses in places where flow control determines
4509 the variable's value is not live. However, these registers are made
4510 unavailable for use in the reload pass; excessive use of this feature
4511 leaves the compiler too few available registers to compile certain
4514 This option does not guarantee that GCC will generate code that has
4515 this variable in the register you specify at all times. You may not
4516 code an explicit reference to this register in an @code{asm} statement
4517 and assume it will always refer to this variable.
4519 Stores into local register variables may be deleted when they appear to be dead
4520 according to dataflow analysis. References to local register variables may
4521 be deleted or moved or simplified.
4523 @node Alternate Keywords
4524 @section Alternate Keywords
4525 @cindex alternate keywords
4526 @cindex keywords, alternate
4528 @option{-ansi} and the various @option{-std} options disable certain
4529 keywords. This causes trouble when you want to use GNU C extensions, or
4530 a general-purpose header file that should be usable by all programs,
4531 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4532 @code{inline} are not available in programs compiled with
4533 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4534 program compiled with @option{-std=c99}). The ISO C99 keyword
4535 @code{restrict} is only available when @option{-std=gnu99} (which will
4536 eventually be the default) or @option{-std=c99} (or the equivalent
4537 @option{-std=iso9899:1999}) is used.
4539 The way to solve these problems is to put @samp{__} at the beginning and
4540 end of each problematical keyword. For example, use @code{__asm__}
4541 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4543 Other C compilers won't accept these alternative keywords; if you want to
4544 compile with another compiler, you can define the alternate keywords as
4545 macros to replace them with the customary keywords. It looks like this:
4553 @findex __extension__
4555 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4557 prevent such warnings within one expression by writing
4558 @code{__extension__} before the expression. @code{__extension__} has no
4559 effect aside from this.
4561 @node Incomplete Enums
4562 @section Incomplete @code{enum} Types
4564 You can define an @code{enum} tag without specifying its possible values.
4565 This results in an incomplete type, much like what you get if you write
4566 @code{struct foo} without describing the elements. A later declaration
4567 which does specify the possible values completes the type.
4569 You can't allocate variables or storage using the type while it is
4570 incomplete. However, you can work with pointers to that type.
4572 This extension may not be very useful, but it makes the handling of
4573 @code{enum} more consistent with the way @code{struct} and @code{union}
4576 This extension is not supported by GNU C++.
4578 @node Function Names
4579 @section Function Names as Strings
4580 @cindex @code{__func__} identifier
4581 @cindex @code{__FUNCTION__} identifier
4582 @cindex @code{__PRETTY_FUNCTION__} identifier
4584 GCC provides three magic variables which hold the name of the current
4585 function, as a string. The first of these is @code{__func__}, which
4586 is part of the C99 standard:
4589 The identifier @code{__func__} is implicitly declared by the translator
4590 as if, immediately following the opening brace of each function
4591 definition, the declaration
4594 static const char __func__[] = "function-name";
4597 appeared, where function-name is the name of the lexically-enclosing
4598 function. This name is the unadorned name of the function.
4601 @code{__FUNCTION__} is another name for @code{__func__}. Older
4602 versions of GCC recognize only this name. However, it is not
4603 standardized. For maximum portability, we recommend you use
4604 @code{__func__}, but provide a fallback definition with the
4608 #if __STDC_VERSION__ < 199901L
4610 # define __func__ __FUNCTION__
4612 # define __func__ "<unknown>"
4617 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4618 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4619 the type signature of the function as well as its bare name. For
4620 example, this program:
4624 extern int printf (char *, ...);
4631 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4632 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4650 __PRETTY_FUNCTION__ = void a::sub(int)
4653 These identifiers are not preprocessor macros. In GCC 3.3 and
4654 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4655 were treated as string literals; they could be used to initialize
4656 @code{char} arrays, and they could be concatenated with other string
4657 literals. GCC 3.4 and later treat them as variables, like
4658 @code{__func__}. In C++, @code{__FUNCTION__} and
4659 @code{__PRETTY_FUNCTION__} have always been variables.
4661 @node Return Address
4662 @section Getting the Return or Frame Address of a Function
4664 These functions may be used to get information about the callers of a
4667 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4668 This function returns the return address of the current function, or of
4669 one of its callers. The @var{level} argument is number of frames to
4670 scan up the call stack. A value of @code{0} yields the return address
4671 of the current function, a value of @code{1} yields the return address
4672 of the caller of the current function, and so forth. When inlining
4673 the expected behavior is that the function will return the address of
4674 the function that will be returned to. To work around this behavior use
4675 the @code{noinline} function attribute.
4677 The @var{level} argument must be a constant integer.
4679 On some machines it may be impossible to determine the return address of
4680 any function other than the current one; in such cases, or when the top
4681 of the stack has been reached, this function will return @code{0} or a
4682 random value. In addition, @code{__builtin_frame_address} may be used
4683 to determine if the top of the stack has been reached.
4685 This function should only be used with a nonzero argument for debugging
4689 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4690 This function is similar to @code{__builtin_return_address}, but it
4691 returns the address of the function frame rather than the return address
4692 of the function. Calling @code{__builtin_frame_address} with a value of
4693 @code{0} yields the frame address of the current function, a value of
4694 @code{1} yields the frame address of the caller of the current function,
4697 The frame is the area on the stack which holds local variables and saved
4698 registers. The frame address is normally the address of the first word
4699 pushed on to the stack by the function. However, the exact definition
4700 depends upon the processor and the calling convention. If the processor
4701 has a dedicated frame pointer register, and the function has a frame,
4702 then @code{__builtin_frame_address} will return the value of the frame
4705 On some machines it may be impossible to determine the frame address of
4706 any function other than the current one; in such cases, or when the top
4707 of the stack has been reached, this function will return @code{0} if
4708 the first frame pointer is properly initialized by the startup code.
4710 This function should only be used with a nonzero argument for debugging
4714 @node Vector Extensions
4715 @section Using vector instructions through built-in functions
4717 On some targets, the instruction set contains SIMD vector instructions that
4718 operate on multiple values contained in one large register at the same time.
4719 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4722 The first step in using these extensions is to provide the necessary data
4723 types. This should be done using an appropriate @code{typedef}:
4726 typedef int v4si __attribute__ ((mode(V4SI)));
4729 The base type @code{int} is effectively ignored by the compiler, the
4730 actual properties of the new type @code{v4si} are defined by the
4731 @code{__attribute__}. It defines the machine mode to be used; for vector
4732 types these have the form @code{V@var{n}@var{B}}; @var{n} should be the
4733 number of elements in the vector, and @var{B} should be the base mode of the
4734 individual elements. The following can be used as base modes:
4738 An integer that is as wide as the smallest addressable unit, usually 8 bits.
4740 An integer, twice as wide as a QI mode integer, usually 16 bits.
4742 An integer, four times as wide as a QI mode integer, usually 32 bits.
4744 An integer, eight times as wide as a QI mode integer, usually 64 bits.
4746 A floating point value, as wide as a SI mode integer, usually 32 bits.
4748 A floating point value, as wide as a DI mode integer, usually 64 bits.
4751 Specifying a combination that is not valid for the current architecture
4752 will cause GCC to synthesize the instructions using a narrower mode.
4753 For example, if you specify a variable of type @code{V4SI} and your
4754 architecture does not allow for this specific SIMD type, GCC will
4755 produce code that uses 4 @code{SIs}.
4757 The types defined in this manner can be used with a subset of normal C
4758 operations. Currently, GCC will allow using the following operators
4759 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4761 The operations behave like C++ @code{valarrays}. Addition is defined as
4762 the addition of the corresponding elements of the operands. For
4763 example, in the code below, each of the 4 elements in @var{a} will be
4764 added to the corresponding 4 elements in @var{b} and the resulting
4765 vector will be stored in @var{c}.
4768 typedef int v4si __attribute__ ((mode(V4SI)));
4775 Subtraction, multiplication, division, and the logical operations
4776 operate in a similar manner. Likewise, the result of using the unary
4777 minus or complement operators on a vector type is a vector whose
4778 elements are the negative or complemented values of the corresponding
4779 elements in the operand.
4781 You can declare variables and use them in function calls and returns, as
4782 well as in assignments and some casts. You can specify a vector type as
4783 a return type for a function. Vector types can also be used as function
4784 arguments. It is possible to cast from one vector type to another,
4785 provided they are of the same size (in fact, you can also cast vectors
4786 to and from other datatypes of the same size).
4788 You cannot operate between vectors of different lengths or different
4789 signedness without a cast.
4791 A port that supports hardware vector operations, usually provides a set
4792 of built-in functions that can be used to operate on vectors. For
4793 example, a function to add two vectors and multiply the result by a
4794 third could look like this:
4797 v4si f (v4si a, v4si b, v4si c)
4799 v4si tmp = __builtin_addv4si (a, b);
4800 return __builtin_mulv4si (tmp, c);
4805 @node Other Builtins
4806 @section Other built-in functions provided by GCC
4807 @cindex built-in functions
4808 @findex __builtin_isgreater
4809 @findex __builtin_isgreaterequal
4810 @findex __builtin_isless
4811 @findex __builtin_islessequal
4812 @findex __builtin_islessgreater
4813 @findex __builtin_isunordered
4968 @findex fprintf_unlocked
4970 @findex fputs_unlocked
5055 @findex printf_unlocked
5081 @findex significandf
5082 @findex significandl
5144 GCC provides a large number of built-in functions other than the ones
5145 mentioned above. Some of these are for internal use in the processing
5146 of exceptions or variable-length argument lists and will not be
5147 documented here because they may change from time to time; we do not
5148 recommend general use of these functions.
5150 The remaining functions are provided for optimization purposes.
5152 @opindex fno-builtin
5153 GCC includes built-in versions of many of the functions in the standard
5154 C library. The versions prefixed with @code{__builtin_} will always be
5155 treated as having the same meaning as the C library function even if you
5156 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5157 Many of these functions are only optimized in certain cases; if they are
5158 not optimized in a particular case, a call to the library function will
5163 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5164 @option{-std=c99}), the functions
5165 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5166 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5167 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5168 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5169 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5170 @code{index}, @code{j0f}, @code{j0l}, @code{j0}, @code{j1f}, @code{j1l},
5171 @code{j1}, @code{jnf}, @code{jnl}, @code{jn}, @code{mempcpy},
5172 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
5173 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
5174 @code{significandf}, @code{significandl}, @code{significand},
5175 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5176 @code{strdup}, @code{strfmon}, @code{y0f}, @code{y0l}, @code{y0},
5177 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and @code{yn}
5178 may be handled as built-in functions.
5179 All these functions have corresponding versions
5180 prefixed with @code{__builtin_}, which may be used even in strict C89
5183 The ISO C99 functions
5184 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5185 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5186 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5187 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5188 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5189 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5190 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5191 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5192 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5193 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5194 @code{cimagl}, @code{cimag},
5195 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf},
5196 @code{copysignl}, @code{copysign}, @code{cpowf}, @code{cpowl},
5197 @code{cpow}, @code{cprojf}, @code{cprojl}, @code{cproj}, @code{crealf},
5198 @code{creall}, @code{creal}, @code{csinf}, @code{csinhf}, @code{csinhl},
5199 @code{csinh}, @code{csinl}, @code{csin}, @code{csqrtf}, @code{csqrtl},
5200 @code{csqrt}, @code{ctanf}, @code{ctanhf}, @code{ctanhl}, @code{ctanh},
5201 @code{ctanl}, @code{ctan}, @code{erfcf}, @code{erfcl}, @code{erfc},
5202 @code{erff}, @code{erfl}, @code{erf}, @code{exp2f}, @code{exp2l},
5203 @code{exp2}, @code{expm1f}, @code{expm1l}, @code{expm1}, @code{fdimf},
5204 @code{fdiml}, @code{fdim}, @code{fmaf}, @code{fmal}, @code{fmaxf},
5205 @code{fmaxl}, @code{fmax}, @code{fma}, @code{fminf}, @code{fminl},
5206 @code{fmin}, @code{hypotf}, @code{hypotl}, @code{hypot}, @code{ilogbf},
5207 @code{ilogbl}, @code{ilogb}, @code{imaxabs}, @code{lgammaf},
5208 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf},
5209 @code{llrintl}, @code{llrint}, @code{llroundf}, @code{llroundl},
5210 @code{llround}, @code{log1pf}, @code{log1pl}, @code{log1p},
5211 @code{log2f}, @code{log2l}, @code{log2}, @code{logbf}, @code{logbl},
5212 @code{logb}, @code{lrintf}, @code{lrintl}, @code{lrint}, @code{lroundf},
5213 @code{lroundl}, @code{lround}, @code{nearbyintf}, @code{nearbyintl},
5214 @code{nearbyint}, @code{nextafterf}, @code{nextafterl},
5215 @code{nextafter}, @code{nexttowardf}, @code{nexttowardl},
5216 @code{nexttoward}, @code{remainderf}, @code{remainderl},
5217 @code{remainder}, @code{remquof}, @code{remquol}, @code{remquo},
5218 @code{rintf}, @code{rintl}, @code{rint}, @code{roundf}, @code{roundl},
5219 @code{round}, @code{scalblnf}, @code{scalblnl}, @code{scalbln},
5220 @code{scalbnf}, @code{scalbnl}, @code{scalbn}, @code{snprintf},
5221 @code{tgammaf}, @code{tgammal}, @code{tgamma}, @code{truncf},
5222 @code{truncl}, @code{trunc}, @code{vfscanf}, @code{vscanf},
5223 @code{vsnprintf} and @code{vsscanf}
5224 are handled as built-in functions
5225 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5227 There are also built-in versions of the ISO C99 functions
5228 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5229 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5230 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5231 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5232 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5233 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5234 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5235 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5236 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5237 that are recognized in any mode since ISO C90 reserves these names for
5238 the purpose to which ISO C99 puts them. All these functions have
5239 corresponding versions prefixed with @code{__builtin_}.
5241 The ISO C90 functions
5242 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5243 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5244 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5245 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf}, @code{labs},
5246 @code{ldexp}, @code{log10}, @code{log}, @code{malloc}, @code{memcmp},
5247 @code{memcpy}, @code{memset}, @code{modf}, @code{pow}, @code{printf},
5248 @code{putchar}, @code{puts}, @code{scanf}, @code{sinh}, @code{sin},
5249 @code{snprintf}, @code{sprintf}, @code{sqrt}, @code{sscanf},
5250 @code{strcat}, @code{strchr}, @code{strcmp}, @code{strcpy},
5251 @code{strcspn}, @code{strlen}, @code{strncat}, @code{strncmp},
5252 @code{strncpy}, @code{strpbrk}, @code{strrchr}, @code{strspn},
5253 @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf}
5255 are all recognized as built-in functions unless
5256 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5257 is specified for an individual function). All of these functions have
5258 corresponding versions prefixed with @code{__builtin_}.
5260 GCC provides built-in versions of the ISO C99 floating point comparison
5261 macros that avoid raising exceptions for unordered operands. They have
5262 the same names as the standard macros ( @code{isgreater},
5263 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5264 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5265 prefixed. We intend for a library implementor to be able to simply
5266 @code{#define} each standard macro to its built-in equivalent.
5268 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5270 You can use the built-in function @code{__builtin_types_compatible_p} to
5271 determine whether two types are the same.
5273 This built-in function returns 1 if the unqualified versions of the
5274 types @var{type1} and @var{type2} (which are types, not expressions) are
5275 compatible, 0 otherwise. The result of this built-in function can be
5276 used in integer constant expressions.
5278 This built-in function ignores top level qualifiers (e.g., @code{const},
5279 @code{volatile}). For example, @code{int} is equivalent to @code{const
5282 The type @code{int[]} and @code{int[5]} are compatible. On the other
5283 hand, @code{int} and @code{char *} are not compatible, even if the size
5284 of their types, on the particular architecture are the same. Also, the
5285 amount of pointer indirection is taken into account when determining
5286 similarity. Consequently, @code{short *} is not similar to
5287 @code{short **}. Furthermore, two types that are typedefed are
5288 considered compatible if their underlying types are compatible.
5290 An @code{enum} type is not considered to be compatible with another
5291 @code{enum} type even if both are compatible with the same integer
5292 type; this is what the C standard specifies.
5293 For example, @code{enum @{foo, bar@}} is not similar to
5294 @code{enum @{hot, dog@}}.
5296 You would typically use this function in code whose execution varies
5297 depending on the arguments' types. For example:
5303 if (__builtin_types_compatible_p (typeof (x), long double)) \
5304 tmp = foo_long_double (tmp); \
5305 else if (__builtin_types_compatible_p (typeof (x), double)) \
5306 tmp = foo_double (tmp); \
5307 else if (__builtin_types_compatible_p (typeof (x), float)) \
5308 tmp = foo_float (tmp); \
5315 @emph{Note:} This construct is only available for C.
5319 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5321 You can use the built-in function @code{__builtin_choose_expr} to
5322 evaluate code depending on the value of a constant expression. This
5323 built-in function returns @var{exp1} if @var{const_exp}, which is a
5324 constant expression that must be able to be determined at compile time,
5325 is nonzero. Otherwise it returns 0.
5327 This built-in function is analogous to the @samp{? :} operator in C,
5328 except that the expression returned has its type unaltered by promotion
5329 rules. Also, the built-in function does not evaluate the expression
5330 that was not chosen. For example, if @var{const_exp} evaluates to true,
5331 @var{exp2} is not evaluated even if it has side-effects.
5333 This built-in function can return an lvalue if the chosen argument is an
5336 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5337 type. Similarly, if @var{exp2} is returned, its return type is the same
5344 __builtin_choose_expr ( \
5345 __builtin_types_compatible_p (typeof (x), double), \
5347 __builtin_choose_expr ( \
5348 __builtin_types_compatible_p (typeof (x), float), \
5350 /* @r{The void expression results in a compile-time error} \
5351 @r{when assigning the result to something.} */ \
5355 @emph{Note:} This construct is only available for C. Furthermore, the
5356 unused expression (@var{exp1} or @var{exp2} depending on the value of
5357 @var{const_exp}) may still generate syntax errors. This may change in
5362 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5363 You can use the built-in function @code{__builtin_constant_p} to
5364 determine if a value is known to be constant at compile-time and hence
5365 that GCC can perform constant-folding on expressions involving that
5366 value. The argument of the function is the value to test. The function
5367 returns the integer 1 if the argument is known to be a compile-time
5368 constant and 0 if it is not known to be a compile-time constant. A
5369 return of 0 does not indicate that the value is @emph{not} a constant,
5370 but merely that GCC cannot prove it is a constant with the specified
5371 value of the @option{-O} option.
5373 You would typically use this function in an embedded application where
5374 memory was a critical resource. If you have some complex calculation,
5375 you may want it to be folded if it involves constants, but need to call
5376 a function if it does not. For example:
5379 #define Scale_Value(X) \
5380 (__builtin_constant_p (X) \
5381 ? ((X) * SCALE + OFFSET) : Scale (X))
5384 You may use this built-in function in either a macro or an inline
5385 function. However, if you use it in an inlined function and pass an
5386 argument of the function as the argument to the built-in, GCC will
5387 never return 1 when you call the inline function with a string constant
5388 or compound literal (@pxref{Compound Literals}) and will not return 1
5389 when you pass a constant numeric value to the inline function unless you
5390 specify the @option{-O} option.
5392 You may also use @code{__builtin_constant_p} in initializers for static
5393 data. For instance, you can write
5396 static const int table[] = @{
5397 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5403 This is an acceptable initializer even if @var{EXPRESSION} is not a
5404 constant expression. GCC must be more conservative about evaluating the
5405 built-in in this case, because it has no opportunity to perform
5408 Previous versions of GCC did not accept this built-in in data
5409 initializers. The earliest version where it is completely safe is
5413 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5414 @opindex fprofile-arcs
5415 You may use @code{__builtin_expect} to provide the compiler with
5416 branch prediction information. In general, you should prefer to
5417 use actual profile feedback for this (@option{-fprofile-arcs}), as
5418 programmers are notoriously bad at predicting how their programs
5419 actually perform. However, there are applications in which this
5420 data is hard to collect.
5422 The return value is the value of @var{exp}, which should be an
5423 integral expression. The value of @var{c} must be a compile-time
5424 constant. The semantics of the built-in are that it is expected
5425 that @var{exp} == @var{c}. For example:
5428 if (__builtin_expect (x, 0))
5433 would indicate that we do not expect to call @code{foo}, since
5434 we expect @code{x} to be zero. Since you are limited to integral
5435 expressions for @var{exp}, you should use constructions such as
5438 if (__builtin_expect (ptr != NULL, 1))
5443 when testing pointer or floating-point values.
5446 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5447 This function is used to minimize cache-miss latency by moving data into
5448 a cache before it is accessed.
5449 You can insert calls to @code{__builtin_prefetch} into code for which
5450 you know addresses of data in memory that is likely to be accessed soon.
5451 If the target supports them, data prefetch instructions will be generated.
5452 If the prefetch is done early enough before the access then the data will
5453 be in the cache by the time it is accessed.
5455 The value of @var{addr} is the address of the memory to prefetch.
5456 There are two optional arguments, @var{rw} and @var{locality}.
5457 The value of @var{rw} is a compile-time constant one or zero; one
5458 means that the prefetch is preparing for a write to the memory address
5459 and zero, the default, means that the prefetch is preparing for a read.
5460 The value @var{locality} must be a compile-time constant integer between
5461 zero and three. A value of zero means that the data has no temporal
5462 locality, so it need not be left in the cache after the access. A value
5463 of three means that the data has a high degree of temporal locality and
5464 should be left in all levels of cache possible. Values of one and two
5465 mean, respectively, a low or moderate degree of temporal locality. The
5469 for (i = 0; i < n; i++)
5472 __builtin_prefetch (&a[i+j], 1, 1);
5473 __builtin_prefetch (&b[i+j], 0, 1);
5478 Data prefetch does not generate faults if @var{addr} is invalid, but
5479 the address expression itself must be valid. For example, a prefetch
5480 of @code{p->next} will not fault if @code{p->next} is not a valid
5481 address, but evaluation will fault if @code{p} is not a valid address.
5483 If the target does not support data prefetch, the address expression
5484 is evaluated if it includes side effects but no other code is generated
5485 and GCC does not issue a warning.
5488 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5489 Returns a positive infinity, if supported by the floating-point format,
5490 else @code{DBL_MAX}. This function is suitable for implementing the
5491 ISO C macro @code{HUGE_VAL}.
5494 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5495 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5498 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5499 Similar to @code{__builtin_huge_val}, except the return
5500 type is @code{long double}.
5503 @deftypefn {Built-in Function} double __builtin_inf (void)
5504 Similar to @code{__builtin_huge_val}, except a warning is generated
5505 if the target floating-point format does not support infinities.
5506 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5509 @deftypefn {Built-in Function} float __builtin_inff (void)
5510 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5513 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5514 Similar to @code{__builtin_inf}, except the return
5515 type is @code{long double}.
5518 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5519 This is an implementation of the ISO C99 function @code{nan}.
5521 Since ISO C99 defines this function in terms of @code{strtod}, which we
5522 do not implement, a description of the parsing is in order. The string
5523 is parsed as by @code{strtol}; that is, the base is recognized by
5524 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5525 in the significand such that the least significant bit of the number
5526 is at the least significant bit of the significand. The number is
5527 truncated to fit the significand field provided. The significand is
5528 forced to be a quiet NaN.
5530 This function, if given a string literal, is evaluated early enough
5531 that it is considered a compile-time constant.
5534 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5535 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5538 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5539 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5542 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5543 Similar to @code{__builtin_nan}, except the significand is forced
5544 to be a signaling NaN. The @code{nans} function is proposed by
5545 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
5548 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5549 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5552 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5553 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5556 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5557 Returns one plus the index of the least significant 1-bit of @var{x}, or
5558 if @var{x} is zero, returns zero.
5561 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5562 Returns the number of leading 0-bits in @var{x}, starting at the most
5563 significant bit position. If @var{x} is 0, the result is undefined.
5566 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5567 Returns the number of trailing 0-bits in @var{x}, starting at the least
5568 significant bit position. If @var{x} is 0, the result is undefined.
5571 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5572 Returns the number of 1-bits in @var{x}.
5575 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5576 Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5580 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5581 Similar to @code{__builtin_ffs}, except the argument type is
5582 @code{unsigned long}.
5585 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5586 Similar to @code{__builtin_clz}, except the argument type is
5587 @code{unsigned long}.
5590 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5591 Similar to @code{__builtin_ctz}, except the argument type is
5592 @code{unsigned long}.
5595 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5596 Similar to @code{__builtin_popcount}, except the argument type is
5597 @code{unsigned long}.
5600 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5601 Similar to @code{__builtin_parity}, except the argument type is
5602 @code{unsigned long}.
5605 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5606 Similar to @code{__builtin_ffs}, except the argument type is
5607 @code{unsigned long long}.
5610 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5611 Similar to @code{__builtin_clz}, except the argument type is
5612 @code{unsigned long long}.
5615 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5616 Similar to @code{__builtin_ctz}, except the argument type is
5617 @code{unsigned long long}.
5620 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5621 Similar to @code{__builtin_popcount}, except the argument type is
5622 @code{unsigned long long}.
5625 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5626 Similar to @code{__builtin_parity}, except the argument type is
5627 @code{unsigned long long}.
5631 @node Target Builtins
5632 @section Built-in Functions Specific to Particular Target Machines
5634 On some target machines, GCC supports many built-in functions specific
5635 to those machines. Generally these generate calls to specific machine
5636 instructions, but allow the compiler to schedule those calls.
5639 * Alpha Built-in Functions::
5640 * ARM Built-in Functions::
5641 * X86 Built-in Functions::
5642 * PowerPC AltiVec Built-in Functions::
5645 @node Alpha Built-in Functions
5646 @subsection Alpha Built-in Functions
5648 These built-in functions are available for the Alpha family of
5649 processors, depending on the command-line switches used.
5651 The following built-in functions are always available. They
5652 all generate the machine instruction that is part of the name.
5655 long __builtin_alpha_implver (void)
5656 long __builtin_alpha_rpcc (void)
5657 long __builtin_alpha_amask (long)
5658 long __builtin_alpha_cmpbge (long, long)
5659 long __builtin_alpha_extbl (long, long)
5660 long __builtin_alpha_extwl (long, long)
5661 long __builtin_alpha_extll (long, long)
5662 long __builtin_alpha_extql (long, long)
5663 long __builtin_alpha_extwh (long, long)
5664 long __builtin_alpha_extlh (long, long)
5665 long __builtin_alpha_extqh (long, long)
5666 long __builtin_alpha_insbl (long, long)
5667 long __builtin_alpha_inswl (long, long)
5668 long __builtin_alpha_insll (long, long)
5669 long __builtin_alpha_insql (long, long)
5670 long __builtin_alpha_inswh (long, long)
5671 long __builtin_alpha_inslh (long, long)
5672 long __builtin_alpha_insqh (long, long)
5673 long __builtin_alpha_mskbl (long, long)
5674 long __builtin_alpha_mskwl (long, long)
5675 long __builtin_alpha_mskll (long, long)
5676 long __builtin_alpha_mskql (long, long)
5677 long __builtin_alpha_mskwh (long, long)
5678 long __builtin_alpha_msklh (long, long)
5679 long __builtin_alpha_mskqh (long, long)
5680 long __builtin_alpha_umulh (long, long)
5681 long __builtin_alpha_zap (long, long)
5682 long __builtin_alpha_zapnot (long, long)
5685 The following built-in functions are always with @option{-mmax}
5686 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5687 later. They all generate the machine instruction that is part
5691 long __builtin_alpha_pklb (long)
5692 long __builtin_alpha_pkwb (long)
5693 long __builtin_alpha_unpkbl (long)
5694 long __builtin_alpha_unpkbw (long)
5695 long __builtin_alpha_minub8 (long, long)
5696 long __builtin_alpha_minsb8 (long, long)
5697 long __builtin_alpha_minuw4 (long, long)
5698 long __builtin_alpha_minsw4 (long, long)
5699 long __builtin_alpha_maxub8 (long, long)
5700 long __builtin_alpha_maxsb8 (long, long)
5701 long __builtin_alpha_maxuw4 (long, long)
5702 long __builtin_alpha_maxsw4 (long, long)
5703 long __builtin_alpha_perr (long, long)
5706 The following built-in functions are always with @option{-mcix}
5707 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5708 later. They all generate the machine instruction that is part
5712 long __builtin_alpha_cttz (long)
5713 long __builtin_alpha_ctlz (long)
5714 long __builtin_alpha_ctpop (long)
5717 The following builtins are available on systems that use the OSF/1
5718 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5719 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5720 @code{rdval} and @code{wrval}.
5723 void *__builtin_thread_pointer (void)
5724 void __builtin_set_thread_pointer (void *)
5727 @node ARM Built-in Functions
5728 @subsection ARM Built-in Functions
5730 These built-in functions are available for the ARM family of
5731 processors, when the @option{-mcpu=iwmmxt} switch is used:
5734 typedef int v2si __attribute__ ((vector_size (8)));
5735 typedef short v4hi __attribute__ ((vector_size (8)));
5736 typedef char v8qi __attribute__ ((vector_size (8)));
5738 int __builtin_arm_getwcx (int)
5739 void __builtin_arm_setwcx (int, int)
5740 int __builtin_arm_textrmsb (v8qi, int)
5741 int __builtin_arm_textrmsh (v4hi, int)
5742 int __builtin_arm_textrmsw (v2si, int)
5743 int __builtin_arm_textrmub (v8qi, int)
5744 int __builtin_arm_textrmuh (v4hi, int)
5745 int __builtin_arm_textrmuw (v2si, int)
5746 v8qi __builtin_arm_tinsrb (v8qi, int)
5747 v4hi __builtin_arm_tinsrh (v4hi, int)
5748 v2si __builtin_arm_tinsrw (v2si, int)
5749 long long __builtin_arm_tmia (long long, int, int)
5750 long long __builtin_arm_tmiabb (long long, int, int)
5751 long long __builtin_arm_tmiabt (long long, int, int)
5752 long long __builtin_arm_tmiaph (long long, int, int)
5753 long long __builtin_arm_tmiatb (long long, int, int)
5754 long long __builtin_arm_tmiatt (long long, int, int)
5755 int __builtin_arm_tmovmskb (v8qi)
5756 int __builtin_arm_tmovmskh (v4hi)
5757 int __builtin_arm_tmovmskw (v2si)
5758 long long __builtin_arm_waccb (v8qi)
5759 long long __builtin_arm_wacch (v4hi)
5760 long long __builtin_arm_waccw (v2si)
5761 v8qi __builtin_arm_waddb (v8qi, v8qi)
5762 v8qi __builtin_arm_waddbss (v8qi, v8qi)
5763 v8qi __builtin_arm_waddbus (v8qi, v8qi)
5764 v4hi __builtin_arm_waddh (v4hi, v4hi)
5765 v4hi __builtin_arm_waddhss (v4hi, v4hi)
5766 v4hi __builtin_arm_waddhus (v4hi, v4hi)
5767 v2si __builtin_arm_waddw (v2si, v2si)
5768 v2si __builtin_arm_waddwss (v2si, v2si)
5769 v2si __builtin_arm_waddwus (v2si, v2si)
5770 v8qi __builtin_arm_walign (v8qi, v8qi, int)
5771 long long __builtin_arm_wand(long long, long long)
5772 long long __builtin_arm_wandn (long long, long long)
5773 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5774 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5775 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5776 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5777 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5778 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5779 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5780 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5781 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5782 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5783 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5784 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5785 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5786 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5787 long long __builtin_arm_wmacsz (v4hi, v4hi)
5788 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5789 long long __builtin_arm_wmacuz (v4hi, v4hi)
5790 v4hi __builtin_arm_wmadds (v4hi, v4hi)
5791 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5792 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5793 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5794 v2si __builtin_arm_wmaxsw (v2si, v2si)
5795 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5796 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5797 v2si __builtin_arm_wmaxuw (v2si, v2si)
5798 v8qi __builtin_arm_wminsb (v8qi, v8qi)
5799 v4hi __builtin_arm_wminsh (v4hi, v4hi)
5800 v2si __builtin_arm_wminsw (v2si, v2si)
5801 v8qi __builtin_arm_wminub (v8qi, v8qi)
5802 v4hi __builtin_arm_wminuh (v4hi, v4hi)
5803 v2si __builtin_arm_wminuw (v2si, v2si)
5804 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5805 v4hi __builtin_arm_wmulul (v4hi, v4hi)
5806 v4hi __builtin_arm_wmulum (v4hi, v4hi)
5807 long long __builtin_arm_wor (long long, long long)
5808 v2si __builtin_arm_wpackdss (long long, long long)
5809 v2si __builtin_arm_wpackdus (long long, long long)
5810 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5811 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5812 v4hi __builtin_arm_wpackwss (v2si, v2si)
5813 v4hi __builtin_arm_wpackwus (v2si, v2si)
5814 long long __builtin_arm_wrord (long long, long long)
5815 long long __builtin_arm_wrordi (long long, int)
5816 v4hi __builtin_arm_wrorh (v4hi, long long)
5817 v4hi __builtin_arm_wrorhi (v4hi, int)
5818 v2si __builtin_arm_wrorw (v2si, long long)
5819 v2si __builtin_arm_wrorwi (v2si, int)
5820 v2si __builtin_arm_wsadb (v8qi, v8qi)
5821 v2si __builtin_arm_wsadbz (v8qi, v8qi)
5822 v2si __builtin_arm_wsadh (v4hi, v4hi)
5823 v2si __builtin_arm_wsadhz (v4hi, v4hi)
5824 v4hi __builtin_arm_wshufh (v4hi, int)
5825 long long __builtin_arm_wslld (long long, long long)
5826 long long __builtin_arm_wslldi (long long, int)
5827 v4hi __builtin_arm_wsllh (v4hi, long long)
5828 v4hi __builtin_arm_wsllhi (v4hi, int)
5829 v2si __builtin_arm_wsllw (v2si, long long)
5830 v2si __builtin_arm_wsllwi (v2si, int)
5831 long long __builtin_arm_wsrad (long long, long long)
5832 long long __builtin_arm_wsradi (long long, int)
5833 v4hi __builtin_arm_wsrah (v4hi, long long)
5834 v4hi __builtin_arm_wsrahi (v4hi, int)
5835 v2si __builtin_arm_wsraw (v2si, long long)
5836 v2si __builtin_arm_wsrawi (v2si, int)
5837 long long __builtin_arm_wsrld (long long, long long)
5838 long long __builtin_arm_wsrldi (long long, int)
5839 v4hi __builtin_arm_wsrlh (v4hi, long long)
5840 v4hi __builtin_arm_wsrlhi (v4hi, int)
5841 v2si __builtin_arm_wsrlw (v2si, long long)
5842 v2si __builtin_arm_wsrlwi (v2si, int)
5843 v8qi __builtin_arm_wsubb (v8qi, v8qi)
5844 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5845 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5846 v4hi __builtin_arm_wsubh (v4hi, v4hi)
5847 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5848 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5849 v2si __builtin_arm_wsubw (v2si, v2si)
5850 v2si __builtin_arm_wsubwss (v2si, v2si)
5851 v2si __builtin_arm_wsubwus (v2si, v2si)
5852 v4hi __builtin_arm_wunpckehsb (v8qi)
5853 v2si __builtin_arm_wunpckehsh (v4hi)
5854 long long __builtin_arm_wunpckehsw (v2si)
5855 v4hi __builtin_arm_wunpckehub (v8qi)
5856 v2si __builtin_arm_wunpckehuh (v4hi)
5857 long long __builtin_arm_wunpckehuw (v2si)
5858 v4hi __builtin_arm_wunpckelsb (v8qi)
5859 v2si __builtin_arm_wunpckelsh (v4hi)
5860 long long __builtin_arm_wunpckelsw (v2si)
5861 v4hi __builtin_arm_wunpckelub (v8qi)
5862 v2si __builtin_arm_wunpckeluh (v4hi)
5863 long long __builtin_arm_wunpckeluw (v2si)
5864 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5865 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5866 v2si __builtin_arm_wunpckihw (v2si, v2si)
5867 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5868 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5869 v2si __builtin_arm_wunpckilw (v2si, v2si)
5870 long long __builtin_arm_wxor (long long, long long)
5871 long long __builtin_arm_wzero ()
5874 @node X86 Built-in Functions
5875 @subsection X86 Built-in Functions
5877 These built-in functions are available for the i386 and x86-64 family
5878 of computers, depending on the command-line switches used.
5880 The following machine modes are available for use with MMX built-in functions
5881 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5882 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5883 vector of eight 8-bit integers. Some of the built-in functions operate on
5884 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5886 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5887 of two 32-bit floating point values.
5889 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5890 floating point values. Some instructions use a vector of four 32-bit
5891 integers, these use @code{V4SI}. Finally, some instructions operate on an
5892 entire vector register, interpreting it as a 128-bit integer, these use mode
5895 The following built-in functions are made available by @option{-mmmx}.
5896 All of them generate the machine instruction that is part of the name.
5899 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5900 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5901 v2si __builtin_ia32_paddd (v2si, v2si)
5902 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5903 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5904 v2si __builtin_ia32_psubd (v2si, v2si)
5905 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5906 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5907 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5908 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5909 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5910 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5911 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5912 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5913 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5914 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5915 di __builtin_ia32_pand (di, di)
5916 di __builtin_ia32_pandn (di,di)
5917 di __builtin_ia32_por (di, di)
5918 di __builtin_ia32_pxor (di, di)
5919 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5920 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5921 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5922 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5923 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5924 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5925 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5926 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5927 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5928 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5929 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5930 v2si __builtin_ia32_punpckldq (v2si, v2si)
5931 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5932 v4hi __builtin_ia32_packssdw (v2si, v2si)
5933 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5936 The following built-in functions are made available either with
5937 @option{-msse}, or with a combination of @option{-m3dnow} and
5938 @option{-march=athlon}. All of them generate the machine
5939 instruction that is part of the name.
5942 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5943 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5944 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5945 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5946 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5947 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5948 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5949 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5950 int __builtin_ia32_pextrw (v4hi, int)
5951 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5952 int __builtin_ia32_pmovmskb (v8qi)
5953 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5954 void __builtin_ia32_movntq (di *, di)
5955 void __builtin_ia32_sfence (void)
5958 The following built-in functions are available when @option{-msse} is used.
5959 All of them generate the machine instruction that is part of the name.
5962 int __builtin_ia32_comieq (v4sf, v4sf)
5963 int __builtin_ia32_comineq (v4sf, v4sf)
5964 int __builtin_ia32_comilt (v4sf, v4sf)
5965 int __builtin_ia32_comile (v4sf, v4sf)
5966 int __builtin_ia32_comigt (v4sf, v4sf)
5967 int __builtin_ia32_comige (v4sf, v4sf)
5968 int __builtin_ia32_ucomieq (v4sf, v4sf)
5969 int __builtin_ia32_ucomineq (v4sf, v4sf)
5970 int __builtin_ia32_ucomilt (v4sf, v4sf)
5971 int __builtin_ia32_ucomile (v4sf, v4sf)
5972 int __builtin_ia32_ucomigt (v4sf, v4sf)
5973 int __builtin_ia32_ucomige (v4sf, v4sf)
5974 v4sf __builtin_ia32_addps (v4sf, v4sf)
5975 v4sf __builtin_ia32_subps (v4sf, v4sf)
5976 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5977 v4sf __builtin_ia32_divps (v4sf, v4sf)
5978 v4sf __builtin_ia32_addss (v4sf, v4sf)
5979 v4sf __builtin_ia32_subss (v4sf, v4sf)
5980 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5981 v4sf __builtin_ia32_divss (v4sf, v4sf)
5982 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5983 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5984 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5985 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5986 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5987 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5988 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5989 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5990 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5991 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5992 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5993 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5994 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5995 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5996 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5997 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5998 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5999 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6000 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6001 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6002 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6003 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6004 v4sf __builtin_ia32_minps (v4sf, v4sf)
6005 v4sf __builtin_ia32_minss (v4sf, v4sf)
6006 v4sf __builtin_ia32_andps (v4sf, v4sf)
6007 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6008 v4sf __builtin_ia32_orps (v4sf, v4sf)
6009 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6010 v4sf __builtin_ia32_movss (v4sf, v4sf)
6011 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6012 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6013 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6014 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6015 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6016 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6017 v2si __builtin_ia32_cvtps2pi (v4sf)
6018 int __builtin_ia32_cvtss2si (v4sf)
6019 v2si __builtin_ia32_cvttps2pi (v4sf)
6020 int __builtin_ia32_cvttss2si (v4sf)
6021 v4sf __builtin_ia32_rcpps (v4sf)
6022 v4sf __builtin_ia32_rsqrtps (v4sf)
6023 v4sf __builtin_ia32_sqrtps (v4sf)
6024 v4sf __builtin_ia32_rcpss (v4sf)
6025 v4sf __builtin_ia32_rsqrtss (v4sf)
6026 v4sf __builtin_ia32_sqrtss (v4sf)
6027 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6028 void __builtin_ia32_movntps (float *, v4sf)
6029 int __builtin_ia32_movmskps (v4sf)
6032 The following built-in functions are available when @option{-msse} is used.
6035 @item v4sf __builtin_ia32_loadaps (float *)
6036 Generates the @code{movaps} machine instruction as a load from memory.
6037 @item void __builtin_ia32_storeaps (float *, v4sf)
6038 Generates the @code{movaps} machine instruction as a store to memory.
6039 @item v4sf __builtin_ia32_loadups (float *)
6040 Generates the @code{movups} machine instruction as a load from memory.
6041 @item void __builtin_ia32_storeups (float *, v4sf)
6042 Generates the @code{movups} machine instruction as a store to memory.
6043 @item v4sf __builtin_ia32_loadsss (float *)
6044 Generates the @code{movss} machine instruction as a load from memory.
6045 @item void __builtin_ia32_storess (float *, v4sf)
6046 Generates the @code{movss} machine instruction as a store to memory.
6047 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6048 Generates the @code{movhps} machine instruction as a load from memory.
6049 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6050 Generates the @code{movlps} machine instruction as a load from memory
6051 @item void __builtin_ia32_storehps (v4sf, v2si *)
6052 Generates the @code{movhps} machine instruction as a store to memory.
6053 @item void __builtin_ia32_storelps (v4sf, v2si *)
6054 Generates the @code{movlps} machine instruction as a store to memory.
6057 The following built-in functions are available when @option{-msse3} is used.
6058 All of them generate the machine instruction that is part of the name.
6061 v2df __builtin_ia32_addsubpd (v2df, v2df)
6062 v2df __builtin_ia32_addsubps (v2df, v2df)
6063 v2df __builtin_ia32_haddpd (v2df, v2df)
6064 v2df __builtin_ia32_haddps (v2df, v2df)
6065 v2df __builtin_ia32_hsubpd (v2df, v2df)
6066 v2df __builtin_ia32_hsubps (v2df, v2df)
6067 v16qi __builtin_ia32_lddqu (char const *)
6068 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6069 v2df __builtin_ia32_movddup (v2df)
6070 v4sf __builtin_ia32_movshdup (v4sf)
6071 v4sf __builtin_ia32_movsldup (v4sf)
6072 void __builtin_ia32_mwait (unsigned int, unsigned int)
6075 The following built-in functions are available when @option{-msse3} is used.
6078 @item v2df __builtin_ia32_loadddup (double const *)
6079 Generates the @code{movddup} machine instruction as a load from memory.
6082 The following built-in functions are available when @option{-m3dnow} is used.
6083 All of them generate the machine instruction that is part of the name.
6086 void __builtin_ia32_femms (void)
6087 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6088 v2si __builtin_ia32_pf2id (v2sf)
6089 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6090 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6091 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6092 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6093 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6094 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6095 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6096 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6097 v2sf __builtin_ia32_pfrcp (v2sf)
6098 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6099 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6100 v2sf __builtin_ia32_pfrsqrt (v2sf)
6101 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6102 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6103 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6104 v2sf __builtin_ia32_pi2fd (v2si)
6105 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6108 The following built-in functions are available when both @option{-m3dnow}
6109 and @option{-march=athlon} are used. All of them generate the machine
6110 instruction that is part of the name.
6113 v2si __builtin_ia32_pf2iw (v2sf)
6114 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6115 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6116 v2sf __builtin_ia32_pi2fw (v2si)
6117 v2sf __builtin_ia32_pswapdsf (v2sf)
6118 v2si __builtin_ia32_pswapdsi (v2si)
6121 @node PowerPC AltiVec Built-in Functions
6122 @subsection PowerPC AltiVec Built-in Functions
6124 These built-in functions are available for the PowerPC family
6125 of computers, depending on the command-line switches used.
6127 The following machine modes are available for use with AltiVec built-in
6128 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
6129 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
6130 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
6131 @code{V16QI} for a vector of sixteen 8-bit integers.
6133 The following functions are made available by including
6134 @code{<altivec.h>} and using @option{-maltivec} and
6135 @option{-mabi=altivec}. The functions implement the functionality
6136 described in Motorola's AltiVec Programming Interface Manual.
6138 There are a few differences from Motorola's documentation and GCC's
6139 implementation. Vector constants are done with curly braces (not
6140 parentheses). Vector initializers require no casts if the vector
6141 constant is of the same type as the variable it is initializing. The
6142 @code{vector bool} type is deprecated and will be discontinued in
6143 further revisions. Use @code{vector signed} instead. If @code{signed}
6144 or @code{unsigned} is omitted, the vector type will default to
6145 @code{signed}. Lastly, all overloaded functions are implemented with macros
6146 for the C implementation. So code the following example will not work:
6149 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6152 Since vec_add is a macro, the vector constant in the above example will
6153 be treated as four different arguments. Wrap the entire argument in
6154 parentheses for this to work. The C++ implementation does not use
6157 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
6158 Internally, GCC uses built-in functions to achieve the functionality in
6159 the aforementioned header file, but they are not supported and are
6160 subject to change without notice.
6163 vector signed char vec_abs (vector signed char, vector signed char);
6164 vector signed short vec_abs (vector signed short, vector signed short);
6165 vector signed int vec_abs (vector signed int, vector signed int);
6166 vector signed float vec_abs (vector signed float, vector signed float);
6168 vector signed char vec_abss (vector signed char, vector signed char);
6169 vector signed short vec_abss (vector signed short, vector signed short);
6171 vector signed char vec_add (vector signed char, vector signed char);
6172 vector unsigned char vec_add (vector signed char, vector unsigned char);
6174 vector unsigned char vec_add (vector unsigned char, vector signed char);
6176 vector unsigned char vec_add (vector unsigned char,
6177 vector unsigned char);
6178 vector signed short vec_add (vector signed short, vector signed short);
6179 vector unsigned short vec_add (vector signed short,
6180 vector unsigned short);
6181 vector unsigned short vec_add (vector unsigned short,
6182 vector signed short);
6183 vector unsigned short vec_add (vector unsigned short,
6184 vector unsigned short);
6185 vector signed int vec_add (vector signed int, vector signed int);
6186 vector unsigned int vec_add (vector signed int, vector unsigned int);
6187 vector unsigned int vec_add (vector unsigned int, vector signed int);
6188 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6189 vector float vec_add (vector float, vector float);
6191 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6193 vector unsigned char vec_adds (vector signed char,
6194 vector unsigned char);
6195 vector unsigned char vec_adds (vector unsigned char,
6196 vector signed char);
6197 vector unsigned char vec_adds (vector unsigned char,
6198 vector unsigned char);
6199 vector signed char vec_adds (vector signed char, vector signed char);
6200 vector unsigned short vec_adds (vector signed short,
6201 vector unsigned short);
6202 vector unsigned short vec_adds (vector unsigned short,
6203 vector signed short);
6204 vector unsigned short vec_adds (vector unsigned short,
6205 vector unsigned short);
6206 vector signed short vec_adds (vector signed short, vector signed short);
6208 vector unsigned int vec_adds (vector signed int, vector unsigned int);
6209 vector unsigned int vec_adds (vector unsigned int, vector signed int);
6210 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6212 vector signed int vec_adds (vector signed int, vector signed int);
6214 vector float vec_and (vector float, vector float);
6215 vector float vec_and (vector float, vector signed int);
6216 vector float vec_and (vector signed int, vector float);
6217 vector signed int vec_and (vector signed int, vector signed int);
6218 vector unsigned int vec_and (vector signed int, vector unsigned int);
6219 vector unsigned int vec_and (vector unsigned int, vector signed int);
6220 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
6221 vector signed short vec_and (vector signed short, vector signed short);
6222 vector unsigned short vec_and (vector signed short,
6223 vector unsigned short);
6224 vector unsigned short vec_and (vector unsigned short,
6225 vector signed short);
6226 vector unsigned short vec_and (vector unsigned short,
6227 vector unsigned short);
6228 vector signed char vec_and (vector signed char, vector signed char);
6229 vector unsigned char vec_and (vector signed char, vector unsigned char);
6231 vector unsigned char vec_and (vector unsigned char, vector signed char);
6233 vector unsigned char vec_and (vector unsigned char,
6234 vector unsigned char);
6236 vector float vec_andc (vector float, vector float);
6237 vector float vec_andc (vector float, vector signed int);
6238 vector float vec_andc (vector signed int, vector float);
6239 vector signed int vec_andc (vector signed int, vector signed int);
6240 vector unsigned int vec_andc (vector signed int, vector unsigned int);
6241 vector unsigned int vec_andc (vector unsigned int, vector signed int);
6242 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
6244 vector signed short vec_andc (vector signed short, vector signed short);
6246 vector unsigned short vec_andc (vector signed short,
6247 vector unsigned short);
6248 vector unsigned short vec_andc (vector unsigned short,
6249 vector signed short);
6250 vector unsigned short vec_andc (vector unsigned short,
6251 vector unsigned short);
6252 vector signed char vec_andc (vector signed char, vector signed char);
6253 vector unsigned char vec_andc (vector signed char,
6254 vector unsigned char);
6255 vector unsigned char vec_andc (vector unsigned char,
6256 vector signed char);
6257 vector unsigned char vec_andc (vector unsigned char,
6258 vector unsigned char);
6260 vector unsigned char vec_avg (vector unsigned char,
6261 vector unsigned char);
6262 vector signed char vec_avg (vector signed char, vector signed char);
6263 vector unsigned short vec_avg (vector unsigned short,
6264 vector unsigned short);
6265 vector signed short vec_avg (vector signed short, vector signed short);
6266 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
6267 vector signed int vec_avg (vector signed int, vector signed int);
6269 vector float vec_ceil (vector float);
6271 vector signed int vec_cmpb (vector float, vector float);
6273 vector signed char vec_cmpeq (vector signed char, vector signed char);
6274 vector signed char vec_cmpeq (vector unsigned char,
6275 vector unsigned char);
6276 vector signed short vec_cmpeq (vector signed short,
6277 vector signed short);
6278 vector signed short vec_cmpeq (vector unsigned short,
6279 vector unsigned short);
6280 vector signed int vec_cmpeq (vector signed int, vector signed int);
6281 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
6282 vector signed int vec_cmpeq (vector float, vector float);
6284 vector signed int vec_cmpge (vector float, vector float);
6286 vector signed char vec_cmpgt (vector unsigned char,
6287 vector unsigned char);
6288 vector signed char vec_cmpgt (vector signed char, vector signed char);
6289 vector signed short vec_cmpgt (vector unsigned short,
6290 vector unsigned short);
6291 vector signed short vec_cmpgt (vector signed short,
6292 vector signed short);
6293 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
6294 vector signed int vec_cmpgt (vector signed int, vector signed int);
6295 vector signed int vec_cmpgt (vector float, vector float);
6297 vector signed int vec_cmple (vector float, vector float);
6299 vector signed char vec_cmplt (vector unsigned char,
6300 vector unsigned char);
6301 vector signed char vec_cmplt (vector signed char, vector signed char);
6302 vector signed short vec_cmplt (vector unsigned short,
6303 vector unsigned short);
6304 vector signed short vec_cmplt (vector signed short,
6305 vector signed short);
6306 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
6307 vector signed int vec_cmplt (vector signed int, vector signed int);
6308 vector signed int vec_cmplt (vector float, vector float);
6310 vector float vec_ctf (vector unsigned int, const char);
6311 vector float vec_ctf (vector signed int, const char);
6313 vector signed int vec_cts (vector float, const char);
6315 vector unsigned int vec_ctu (vector float, const char);
6317 void vec_dss (const char);
6319 void vec_dssall (void);
6321 void vec_dst (void *, int, const char);
6323 void vec_dstst (void *, int, const char);
6325 void vec_dststt (void *, int, const char);
6327 void vec_dstt (void *, int, const char);
6329 vector float vec_expte (vector float, vector float);
6331 vector float vec_floor (vector float, vector float);
6333 vector float vec_ld (int, vector float *);
6334 vector float vec_ld (int, float *):
6335 vector signed int vec_ld (int, int *);
6336 vector signed int vec_ld (int, vector signed int *);
6337 vector unsigned int vec_ld (int, vector unsigned int *);
6338 vector unsigned int vec_ld (int, unsigned int *);
6339 vector signed short vec_ld (int, short *, vector signed short *);
6340 vector unsigned short vec_ld (int, unsigned short *,
6341 vector unsigned short *);
6342 vector signed char vec_ld (int, signed char *);
6343 vector signed char vec_ld (int, vector signed char *);
6344 vector unsigned char vec_ld (int, unsigned char *);
6345 vector unsigned char vec_ld (int, vector unsigned char *);
6347 vector signed char vec_lde (int, signed char *);
6348 vector unsigned char vec_lde (int, unsigned char *);
6349 vector signed short vec_lde (int, short *);
6350 vector unsigned short vec_lde (int, unsigned short *);
6351 vector float vec_lde (int, float *);
6352 vector signed int vec_lde (int, int *);
6353 vector unsigned int vec_lde (int, unsigned int *);
6355 void float vec_ldl (int, float *);
6356 void float vec_ldl (int, vector float *);
6357 void signed int vec_ldl (int, vector signed int *);
6358 void signed int vec_ldl (int, int *);
6359 void unsigned int vec_ldl (int, unsigned int *);
6360 void unsigned int vec_ldl (int, vector unsigned int *);
6361 void signed short vec_ldl (int, vector signed short *);
6362 void signed short vec_ldl (int, short *);
6363 void unsigned short vec_ldl (int, vector unsigned short *);
6364 void unsigned short vec_ldl (int, unsigned short *);
6365 void signed char vec_ldl (int, vector signed char *);
6366 void signed char vec_ldl (int, signed char *);
6367 void unsigned char vec_ldl (int, vector unsigned char *);
6368 void unsigned char vec_ldl (int, unsigned char *);
6370 vector float vec_loge (vector float);
6372 vector unsigned char vec_lvsl (int, void *, int *);
6374 vector unsigned char vec_lvsr (int, void *, int *);
6376 vector float vec_madd (vector float, vector float, vector float);
6378 vector signed short vec_madds (vector signed short, vector signed short,
6379 vector signed short);
6381 vector unsigned char vec_max (vector signed char, vector unsigned char);
6383 vector unsigned char vec_max (vector unsigned char, vector signed char);
6385 vector unsigned char vec_max (vector unsigned char,
6386 vector unsigned char);
6387 vector signed char vec_max (vector signed char, vector signed char);
6388 vector unsigned short vec_max (vector signed short,
6389 vector unsigned short);
6390 vector unsigned short vec_max (vector unsigned short,
6391 vector signed short);
6392 vector unsigned short vec_max (vector unsigned short,
6393 vector unsigned short);
6394 vector signed short vec_max (vector signed short, vector signed short);
6395 vector unsigned int vec_max (vector signed int, vector unsigned int);
6396 vector unsigned int vec_max (vector unsigned int, vector signed int);
6397 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
6398 vector signed int vec_max (vector signed int, vector signed int);
6399 vector float vec_max (vector float, vector float);
6401 vector signed char vec_mergeh (vector signed char, vector signed char);
6402 vector unsigned char vec_mergeh (vector unsigned char,
6403 vector unsigned char);
6404 vector signed short vec_mergeh (vector signed short,
6405 vector signed short);
6406 vector unsigned short vec_mergeh (vector unsigned short,
6407 vector unsigned short);
6408 vector float vec_mergeh (vector float, vector float);
6409 vector signed int vec_mergeh (vector signed int, vector signed int);
6410 vector unsigned int vec_mergeh (vector unsigned int,
6411 vector unsigned int);
6413 vector signed char vec_mergel (vector signed char, vector signed char);
6414 vector unsigned char vec_mergel (vector unsigned char,
6415 vector unsigned char);
6416 vector signed short vec_mergel (vector signed short,
6417 vector signed short);
6418 vector unsigned short vec_mergel (vector unsigned short,
6419 vector unsigned short);
6420 vector float vec_mergel (vector float, vector float);
6421 vector signed int vec_mergel (vector signed int, vector signed int);
6422 vector unsigned int vec_mergel (vector unsigned int,
6423 vector unsigned int);
6425 vector unsigned short vec_mfvscr (void);
6427 vector unsigned char vec_min (vector signed char, vector unsigned char);
6429 vector unsigned char vec_min (vector unsigned char, vector signed char);
6431 vector unsigned char vec_min (vector unsigned char,
6432 vector unsigned char);
6433 vector signed char vec_min (vector signed char, vector signed char);
6434 vector unsigned short vec_min (vector signed short,
6435 vector unsigned short);
6436 vector unsigned short vec_min (vector unsigned short,
6437 vector signed short);
6438 vector unsigned short vec_min (vector unsigned short,
6439 vector unsigned short);
6440 vector signed short vec_min (vector signed short, vector signed short);
6441 vector unsigned int vec_min (vector signed int, vector unsigned int);
6442 vector unsigned int vec_min (vector unsigned int, vector signed int);
6443 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
6444 vector signed int vec_min (vector signed int, vector signed int);
6445 vector float vec_min (vector float, vector float);
6447 vector signed short vec_mladd (vector signed short, vector signed short,
6448 vector signed short);
6449 vector signed short vec_mladd (vector signed short,
6450 vector unsigned short,
6451 vector unsigned short);
6452 vector signed short vec_mladd (vector unsigned short,
6453 vector signed short,
6454 vector signed short);
6455 vector unsigned short vec_mladd (vector unsigned short,
6456 vector unsigned short,
6457 vector unsigned short);
6459 vector signed short vec_mradds (vector signed short,
6460 vector signed short,
6461 vector signed short);
6463 vector unsigned int vec_msum (vector unsigned char,
6464 vector unsigned char,
6465 vector unsigned int);
6466 vector signed int vec_msum (vector signed char, vector unsigned char,
6468 vector unsigned int vec_msum (vector unsigned short,
6469 vector unsigned short,
6470 vector unsigned int);
6471 vector signed int vec_msum (vector signed short, vector signed short,
6474 vector unsigned int vec_msums (vector unsigned short,
6475 vector unsigned short,
6476 vector unsigned int);
6477 vector signed int vec_msums (vector signed short, vector signed short,
6480 void vec_mtvscr (vector signed int);
6481 void vec_mtvscr (vector unsigned int);
6482 void vec_mtvscr (vector signed short);
6483 void vec_mtvscr (vector unsigned short);
6484 void vec_mtvscr (vector signed char);
6485 void vec_mtvscr (vector unsigned char);
6487 vector unsigned short vec_mule (vector unsigned char,
6488 vector unsigned char);
6489 vector signed short vec_mule (vector signed char, vector signed char);
6490 vector unsigned int vec_mule (vector unsigned short,
6491 vector unsigned short);
6492 vector signed int vec_mule (vector signed short, vector signed short);
6494 vector unsigned short vec_mulo (vector unsigned char,
6495 vector unsigned char);
6496 vector signed short vec_mulo (vector signed char, vector signed char);
6497 vector unsigned int vec_mulo (vector unsigned short,
6498 vector unsigned short);
6499 vector signed int vec_mulo (vector signed short, vector signed short);
6501 vector float vec_nmsub (vector float, vector float, vector float);
6503 vector float vec_nor (vector float, vector float);
6504 vector signed int vec_nor (vector signed int, vector signed int);
6505 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
6506 vector signed short vec_nor (vector signed short, vector signed short);
6507 vector unsigned short vec_nor (vector unsigned short,
6508 vector unsigned short);
6509 vector signed char vec_nor (vector signed char, vector signed char);
6510 vector unsigned char vec_nor (vector unsigned char,
6511 vector unsigned char);
6513 vector float vec_or (vector float, vector float);
6514 vector float vec_or (vector float, vector signed int);
6515 vector float vec_or (vector signed int, vector float);
6516 vector signed int vec_or (vector signed int, vector signed int);
6517 vector unsigned int vec_or (vector signed int, vector unsigned int);
6518 vector unsigned int vec_or (vector unsigned int, vector signed int);
6519 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
6520 vector signed short vec_or (vector signed short, vector signed short);
6521 vector unsigned short vec_or (vector signed short,
6522 vector unsigned short);
6523 vector unsigned short vec_or (vector unsigned short,
6524 vector signed short);
6525 vector unsigned short vec_or (vector unsigned short,
6526 vector unsigned short);
6527 vector signed char vec_or (vector signed char, vector signed char);
6528 vector unsigned char vec_or (vector signed char, vector unsigned char);
6529 vector unsigned char vec_or (vector unsigned char, vector signed char);
6530 vector unsigned char vec_or (vector unsigned char,
6531 vector unsigned char);
6533 vector signed char vec_pack (vector signed short, vector signed short);
6534 vector unsigned char vec_pack (vector unsigned short,
6535 vector unsigned short);
6536 vector signed short vec_pack (vector signed int, vector signed int);
6537 vector unsigned short vec_pack (vector unsigned int,
6538 vector unsigned int);
6540 vector signed short vec_packpx (vector unsigned int,
6541 vector unsigned int);
6543 vector unsigned char vec_packs (vector unsigned short,
6544 vector unsigned short);
6545 vector signed char vec_packs (vector signed short, vector signed short);
6547 vector unsigned short vec_packs (vector unsigned int,
6548 vector unsigned int);
6549 vector signed short vec_packs (vector signed int, vector signed int);
6551 vector unsigned char vec_packsu (vector unsigned short,
6552 vector unsigned short);
6553 vector unsigned char vec_packsu (vector signed short,
6554 vector signed short);
6555 vector unsigned short vec_packsu (vector unsigned int,
6556 vector unsigned int);
6557 vector unsigned short vec_packsu (vector signed int, vector signed int);
6559 vector float vec_perm (vector float, vector float,
6560 vector unsigned char);
6561 vector signed int vec_perm (vector signed int, vector signed int,
6562 vector unsigned char);
6563 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
6564 vector unsigned char);
6565 vector signed short vec_perm (vector signed short, vector signed short,
6566 vector unsigned char);
6567 vector unsigned short vec_perm (vector unsigned short,
6568 vector unsigned short,
6569 vector unsigned char);
6570 vector signed char vec_perm (vector signed char, vector signed char,
6571 vector unsigned char);
6572 vector unsigned char vec_perm (vector unsigned char,
6573 vector unsigned char,
6574 vector unsigned char);
6576 vector float vec_re (vector float);
6578 vector signed char vec_rl (vector signed char, vector unsigned char);
6579 vector unsigned char vec_rl (vector unsigned char,
6580 vector unsigned char);
6581 vector signed short vec_rl (vector signed short, vector unsigned short);
6583 vector unsigned short vec_rl (vector unsigned short,
6584 vector unsigned short);
6585 vector signed int vec_rl (vector signed int, vector unsigned int);
6586 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
6588 vector float vec_round (vector float);
6590 vector float vec_rsqrte (vector float);
6592 vector float vec_sel (vector float, vector float, vector signed int);
6593 vector float vec_sel (vector float, vector float, vector unsigned int);
6594 vector signed int vec_sel (vector signed int, vector signed int,
6596 vector signed int vec_sel (vector signed int, vector signed int,
6597 vector unsigned int);
6598 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6600 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
6601 vector unsigned int);
6602 vector signed short vec_sel (vector signed short, vector signed short,
6603 vector signed short);
6604 vector signed short vec_sel (vector signed short, vector signed short,
6605 vector unsigned short);
6606 vector unsigned short vec_sel (vector unsigned short,
6607 vector unsigned short,
6608 vector signed short);
6609 vector unsigned short vec_sel (vector unsigned short,
6610 vector unsigned short,
6611 vector unsigned short);
6612 vector signed char vec_sel (vector signed char, vector signed char,
6613 vector signed char);
6614 vector signed char vec_sel (vector signed char, vector signed char,
6615 vector unsigned char);
6616 vector unsigned char vec_sel (vector unsigned char,
6617 vector unsigned char,
6618 vector signed char);
6619 vector unsigned char vec_sel (vector unsigned char,
6620 vector unsigned char,
6621 vector unsigned char);
6623 vector signed char vec_sl (vector signed char, vector unsigned char);
6624 vector unsigned char vec_sl (vector unsigned char,
6625 vector unsigned char);
6626 vector signed short vec_sl (vector signed short, vector unsigned short);
6628 vector unsigned short vec_sl (vector unsigned short,
6629 vector unsigned short);
6630 vector signed int vec_sl (vector signed int, vector unsigned int);
6631 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
6633 vector float vec_sld (vector float, vector float, const char);
6634 vector signed int vec_sld (vector signed int, vector signed int,
6636 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
6638 vector signed short vec_sld (vector signed short, vector signed short,
6640 vector unsigned short vec_sld (vector unsigned short,
6641 vector unsigned short, const char);
6642 vector signed char vec_sld (vector signed char, vector signed char,
6644 vector unsigned char vec_sld (vector unsigned char,
6645 vector unsigned char,
6648 vector signed int vec_sll (vector signed int, vector unsigned int);
6649 vector signed int vec_sll (vector signed int, vector unsigned short);
6650 vector signed int vec_sll (vector signed int, vector unsigned char);
6651 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
6652 vector unsigned int vec_sll (vector unsigned int,
6653 vector unsigned short);
6654 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
6656 vector signed short vec_sll (vector signed short, vector unsigned int);
6657 vector signed short vec_sll (vector signed short,
6658 vector unsigned short);
6659 vector signed short vec_sll (vector signed short, vector unsigned char);
6661 vector unsigned short vec_sll (vector unsigned short,
6662 vector unsigned int);
6663 vector unsigned short vec_sll (vector unsigned short,
6664 vector unsigned short);
6665 vector unsigned short vec_sll (vector unsigned short,
6666 vector unsigned char);
6667 vector signed char vec_sll (vector signed char, vector unsigned int);
6668 vector signed char vec_sll (vector signed char, vector unsigned short);
6669 vector signed char vec_sll (vector signed char, vector unsigned char);
6670 vector unsigned char vec_sll (vector unsigned char,
6671 vector unsigned int);
6672 vector unsigned char vec_sll (vector unsigned char,
6673 vector unsigned short);
6674 vector unsigned char vec_sll (vector unsigned char,
6675 vector unsigned char);
6677 vector float vec_slo (vector float, vector signed char);
6678 vector float vec_slo (vector float, vector unsigned char);
6679 vector signed int vec_slo (vector signed int, vector signed char);
6680 vector signed int vec_slo (vector signed int, vector unsigned char);
6681 vector unsigned int vec_slo (vector unsigned int, vector signed char);
6682 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
6684 vector signed short vec_slo (vector signed short, vector signed char);
6685 vector signed short vec_slo (vector signed short, vector unsigned char);
6687 vector unsigned short vec_slo (vector unsigned short,
6688 vector signed char);
6689 vector unsigned short vec_slo (vector unsigned short,
6690 vector unsigned char);
6691 vector signed char vec_slo (vector signed char, vector signed char);
6692 vector signed char vec_slo (vector signed char, vector unsigned char);
6693 vector unsigned char vec_slo (vector unsigned char, vector signed char);
6695 vector unsigned char vec_slo (vector unsigned char,
6696 vector unsigned char);
6698 vector signed char vec_splat (vector signed char, const char);
6699 vector unsigned char vec_splat (vector unsigned char, const char);
6700 vector signed short vec_splat (vector signed short, const char);
6701 vector unsigned short vec_splat (vector unsigned short, const char);
6702 vector float vec_splat (vector float, const char);
6703 vector signed int vec_splat (vector signed int, const char);
6704 vector unsigned int vec_splat (vector unsigned int, const char);
6706 vector signed char vec_splat_s8 (const char);
6708 vector signed short vec_splat_s16 (const char);
6710 vector signed int vec_splat_s32 (const char);
6712 vector unsigned char vec_splat_u8 (const char);
6714 vector unsigned short vec_splat_u16 (const char);
6716 vector unsigned int vec_splat_u32 (const char);
6718 vector signed char vec_sr (vector signed char, vector unsigned char);
6719 vector unsigned char vec_sr (vector unsigned char,
6720 vector unsigned char);
6721 vector signed short vec_sr (vector signed short, vector unsigned short);
6723 vector unsigned short vec_sr (vector unsigned short,
6724 vector unsigned short);
6725 vector signed int vec_sr (vector signed int, vector unsigned int);
6726 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
6728 vector signed char vec_sra (vector signed char, vector unsigned char);
6729 vector unsigned char vec_sra (vector unsigned char,
6730 vector unsigned char);
6731 vector signed short vec_sra (vector signed short,
6732 vector unsigned short);
6733 vector unsigned short vec_sra (vector unsigned short,
6734 vector unsigned short);
6735 vector signed int vec_sra (vector signed int, vector unsigned int);
6736 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
6738 vector signed int vec_srl (vector signed int, vector unsigned int);
6739 vector signed int vec_srl (vector signed int, vector unsigned short);
6740 vector signed int vec_srl (vector signed int, vector unsigned char);
6741 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
6742 vector unsigned int vec_srl (vector unsigned int,
6743 vector unsigned short);
6744 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
6746 vector signed short vec_srl (vector signed short, vector unsigned int);
6747 vector signed short vec_srl (vector signed short,
6748 vector unsigned short);
6749 vector signed short vec_srl (vector signed short, vector unsigned char);
6751 vector unsigned short vec_srl (vector unsigned short,
6752 vector unsigned int);
6753 vector unsigned short vec_srl (vector unsigned short,
6754 vector unsigned short);
6755 vector unsigned short vec_srl (vector unsigned short,
6756 vector unsigned char);
6757 vector signed char vec_srl (vector signed char, vector unsigned int);
6758 vector signed char vec_srl (vector signed char, vector unsigned short);
6759 vector signed char vec_srl (vector signed char, vector unsigned char);
6760 vector unsigned char vec_srl (vector unsigned char,
6761 vector unsigned int);
6762 vector unsigned char vec_srl (vector unsigned char,
6763 vector unsigned short);
6764 vector unsigned char vec_srl (vector unsigned char,
6765 vector unsigned char);
6767 vector float vec_sro (vector float, vector signed char);
6768 vector float vec_sro (vector float, vector unsigned char);
6769 vector signed int vec_sro (vector signed int, vector signed char);
6770 vector signed int vec_sro (vector signed int, vector unsigned char);
6771 vector unsigned int vec_sro (vector unsigned int, vector signed char);
6772 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
6774 vector signed short vec_sro (vector signed short, vector signed char);
6775 vector signed short vec_sro (vector signed short, vector unsigned char);
6777 vector unsigned short vec_sro (vector unsigned short,
6778 vector signed char);
6779 vector unsigned short vec_sro (vector unsigned short,
6780 vector unsigned char);
6781 vector signed char vec_sro (vector signed char, vector signed char);
6782 vector signed char vec_sro (vector signed char, vector unsigned char);
6783 vector unsigned char vec_sro (vector unsigned char, vector signed char);
6785 vector unsigned char vec_sro (vector unsigned char,
6786 vector unsigned char);
6788 void vec_st (vector float, int, float *);
6789 void vec_st (vector float, int, vector float *);
6790 void vec_st (vector signed int, int, int *);
6791 void vec_st (vector signed int, int, unsigned int *);
6792 void vec_st (vector unsigned int, int, unsigned int *);
6793 void vec_st (vector unsigned int, int, vector unsigned int *);
6794 void vec_st (vector signed short, int, short *);
6795 void vec_st (vector signed short, int, vector unsigned short *);
6796 void vec_st (vector signed short, int, vector signed short *);
6797 void vec_st (vector unsigned short, int, unsigned short *);
6798 void vec_st (vector unsigned short, int, vector unsigned short *);
6799 void vec_st (vector signed char, int, signed char *);
6800 void vec_st (vector signed char, int, unsigned char *);
6801 void vec_st (vector signed char, int, vector signed char *);
6802 void vec_st (vector unsigned char, int, unsigned char *);
6803 void vec_st (vector unsigned char, int, vector unsigned char *);
6805 void vec_ste (vector signed char, int, unsigned char *);
6806 void vec_ste (vector signed char, int, signed char *);
6807 void vec_ste (vector unsigned char, int, unsigned char *);
6808 void vec_ste (vector signed short, int, short *);
6809 void vec_ste (vector signed short, int, unsigned short *);
6810 void vec_ste (vector unsigned short, int, void *);
6811 void vec_ste (vector signed int, int, unsigned int *);
6812 void vec_ste (vector signed int, int, int *);
6813 void vec_ste (vector unsigned int, int, unsigned int *);
6814 void vec_ste (vector float, int, float *);
6816 void vec_stl (vector float, int, vector float *);
6817 void vec_stl (vector float, int, float *);
6818 void vec_stl (vector signed int, int, vector signed int *);
6819 void vec_stl (vector signed int, int, int *);
6820 void vec_stl (vector signed int, int, unsigned int *);
6821 void vec_stl (vector unsigned int, int, vector unsigned int *);
6822 void vec_stl (vector unsigned int, int, unsigned int *);
6823 void vec_stl (vector signed short, int, short *);
6824 void vec_stl (vector signed short, int, unsigned short *);
6825 void vec_stl (vector signed short, int, vector signed short *);
6826 void vec_stl (vector unsigned short, int, unsigned short *);
6827 void vec_stl (vector unsigned short, int, vector signed short *);
6828 void vec_stl (vector signed char, int, signed char *);
6829 void vec_stl (vector signed char, int, unsigned char *);
6830 void vec_stl (vector signed char, int, vector signed char *);
6831 void vec_stl (vector unsigned char, int, unsigned char *);
6832 void vec_stl (vector unsigned char, int, vector unsigned char *);
6834 vector signed char vec_sub (vector signed char, vector signed char);
6835 vector unsigned char vec_sub (vector signed char, vector unsigned char);
6837 vector unsigned char vec_sub (vector unsigned char, vector signed char);
6839 vector unsigned char vec_sub (vector unsigned char,
6840 vector unsigned char);
6841 vector signed short vec_sub (vector signed short, vector signed short);
6842 vector unsigned short vec_sub (vector signed short,
6843 vector unsigned short);
6844 vector unsigned short vec_sub (vector unsigned short,
6845 vector signed short);
6846 vector unsigned short vec_sub (vector unsigned short,
6847 vector unsigned short);
6848 vector signed int vec_sub (vector signed int, vector signed int);
6849 vector unsigned int vec_sub (vector signed int, vector unsigned int);
6850 vector unsigned int vec_sub (vector unsigned int, vector signed int);
6851 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
6852 vector float vec_sub (vector float, vector float);
6854 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
6856 vector unsigned char vec_subs (vector signed char,
6857 vector unsigned char);
6858 vector unsigned char vec_subs (vector unsigned char,
6859 vector signed char);
6860 vector unsigned char vec_subs (vector unsigned char,
6861 vector unsigned char);
6862 vector signed char vec_subs (vector signed char, vector signed char);
6863 vector unsigned short vec_subs (vector signed short,
6864 vector unsigned short);
6865 vector unsigned short vec_subs (vector unsigned short,
6866 vector signed short);
6867 vector unsigned short vec_subs (vector unsigned short,
6868 vector unsigned short);
6869 vector signed short vec_subs (vector signed short, vector signed short);
6871 vector unsigned int vec_subs (vector signed int, vector unsigned int);
6872 vector unsigned int vec_subs (vector unsigned int, vector signed int);
6873 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
6875 vector signed int vec_subs (vector signed int, vector signed int);
6877 vector unsigned int vec_sum4s (vector unsigned char,
6878 vector unsigned int);
6879 vector signed int vec_sum4s (vector signed char, vector signed int);
6880 vector signed int vec_sum4s (vector signed short, vector signed int);
6882 vector signed int vec_sum2s (vector signed int, vector signed int);
6884 vector signed int vec_sums (vector signed int, vector signed int);
6886 vector float vec_trunc (vector float);
6888 vector signed short vec_unpackh (vector signed char);
6889 vector unsigned int vec_unpackh (vector signed short);
6890 vector signed int vec_unpackh (vector signed short);
6892 vector signed short vec_unpackl (vector signed char);
6893 vector unsigned int vec_unpackl (vector signed short);
6894 vector signed int vec_unpackl (vector signed short);
6896 vector float vec_xor (vector float, vector float);
6897 vector float vec_xor (vector float, vector signed int);
6898 vector float vec_xor (vector signed int, vector float);
6899 vector signed int vec_xor (vector signed int, vector signed int);
6900 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6901 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6902 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6903 vector signed short vec_xor (vector signed short, vector signed short);
6904 vector unsigned short vec_xor (vector signed short,
6905 vector unsigned short);
6906 vector unsigned short vec_xor (vector unsigned short,
6907 vector signed short);
6908 vector unsigned short vec_xor (vector unsigned short,
6909 vector unsigned short);
6910 vector signed char vec_xor (vector signed char, vector signed char);
6911 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6913 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6915 vector unsigned char vec_xor (vector unsigned char,
6916 vector unsigned char);
6918 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6920 vector signed int vec_all_eq (vector signed char, vector signed char);
6921 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6923 vector signed int vec_all_eq (vector unsigned char,
6924 vector unsigned char);
6925 vector signed int vec_all_eq (vector signed short,
6926 vector unsigned short);
6927 vector signed int vec_all_eq (vector signed short, vector signed short);
6929 vector signed int vec_all_eq (vector unsigned short,
6930 vector signed short);
6931 vector signed int vec_all_eq (vector unsigned short,
6932 vector unsigned short);
6933 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6934 vector signed int vec_all_eq (vector signed int, vector signed int);
6935 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6936 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6938 vector signed int vec_all_eq (vector float, vector float);
6940 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6942 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6944 vector signed int vec_all_ge (vector unsigned char,
6945 vector unsigned char);
6946 vector signed int vec_all_ge (vector signed char, vector signed char);
6947 vector signed int vec_all_ge (vector signed short,
6948 vector unsigned short);
6949 vector signed int vec_all_ge (vector unsigned short,
6950 vector signed short);
6951 vector signed int vec_all_ge (vector unsigned short,
6952 vector unsigned short);
6953 vector signed int vec_all_ge (vector signed short, vector signed short);
6955 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6956 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6957 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6959 vector signed int vec_all_ge (vector signed int, vector signed int);
6960 vector signed int vec_all_ge (vector float, vector float);
6962 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6964 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6966 vector signed int vec_all_gt (vector unsigned char,
6967 vector unsigned char);
6968 vector signed int vec_all_gt (vector signed char, vector signed char);
6969 vector signed int vec_all_gt (vector signed short,
6970 vector unsigned short);
6971 vector signed int vec_all_gt (vector unsigned short,
6972 vector signed short);
6973 vector signed int vec_all_gt (vector unsigned short,
6974 vector unsigned short);
6975 vector signed int vec_all_gt (vector signed short, vector signed short);
6977 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6978 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6979 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6981 vector signed int vec_all_gt (vector signed int, vector signed int);
6982 vector signed int vec_all_gt (vector float, vector float);
6984 vector signed int vec_all_in (vector float, vector float);
6986 vector signed int vec_all_le (vector signed char, vector unsigned char);
6988 vector signed int vec_all_le (vector unsigned char, vector signed char);
6990 vector signed int vec_all_le (vector unsigned char,
6991 vector unsigned char);
6992 vector signed int vec_all_le (vector signed char, vector signed char);
6993 vector signed int vec_all_le (vector signed short,
6994 vector unsigned short);
6995 vector signed int vec_all_le (vector unsigned short,
6996 vector signed short);
6997 vector signed int vec_all_le (vector unsigned short,
6998 vector unsigned short);
6999 vector signed int vec_all_le (vector signed short, vector signed short);
7001 vector signed int vec_all_le (vector signed int, vector unsigned int);
7002 vector signed int vec_all_le (vector unsigned int, vector signed int);
7003 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
7005 vector signed int vec_all_le (vector signed int, vector signed int);
7006 vector signed int vec_all_le (vector float, vector float);
7008 vector signed int vec_all_lt (vector signed char, vector unsigned char);
7010 vector signed int vec_all_lt (vector unsigned char, vector signed char);
7012 vector signed int vec_all_lt (vector unsigned char,
7013 vector unsigned char);
7014 vector signed int vec_all_lt (vector signed char, vector signed char);
7015 vector signed int vec_all_lt (vector signed short,
7016 vector unsigned short);
7017 vector signed int vec_all_lt (vector unsigned short,
7018 vector signed short);
7019 vector signed int vec_all_lt (vector unsigned short,
7020 vector unsigned short);
7021 vector signed int vec_all_lt (vector signed short, vector signed short);
7023 vector signed int vec_all_lt (vector signed int, vector unsigned int);
7024 vector signed int vec_all_lt (vector unsigned int, vector signed int);
7025 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
7027 vector signed int vec_all_lt (vector signed int, vector signed int);
7028 vector signed int vec_all_lt (vector float, vector float);
7030 vector signed int vec_all_nan (vector float);
7032 vector signed int vec_all_ne (vector signed char, vector unsigned char);
7034 vector signed int vec_all_ne (vector signed char, vector signed char);
7035 vector signed int vec_all_ne (vector unsigned char, vector signed char);
7037 vector signed int vec_all_ne (vector unsigned char,
7038 vector unsigned char);
7039 vector signed int vec_all_ne (vector signed short,
7040 vector unsigned short);
7041 vector signed int vec_all_ne (vector signed short, vector signed short);
7043 vector signed int vec_all_ne (vector unsigned short,
7044 vector signed short);
7045 vector signed int vec_all_ne (vector unsigned short,
7046 vector unsigned short);
7047 vector signed int vec_all_ne (vector signed int, vector unsigned int);
7048 vector signed int vec_all_ne (vector signed int, vector signed int);
7049 vector signed int vec_all_ne (vector unsigned int, vector signed int);
7050 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
7052 vector signed int vec_all_ne (vector float, vector float);
7054 vector signed int vec_all_nge (vector float, vector float);
7056 vector signed int vec_all_ngt (vector float, vector float);
7058 vector signed int vec_all_nle (vector float, vector float);
7060 vector signed int vec_all_nlt (vector float, vector float);
7062 vector signed int vec_all_numeric (vector float);
7064 vector signed int vec_any_eq (vector signed char, vector unsigned char);
7066 vector signed int vec_any_eq (vector signed char, vector signed char);
7067 vector signed int vec_any_eq (vector unsigned char, vector signed char);
7069 vector signed int vec_any_eq (vector unsigned char,
7070 vector unsigned char);
7071 vector signed int vec_any_eq (vector signed short,
7072 vector unsigned short);
7073 vector signed int vec_any_eq (vector signed short, vector signed short);
7075 vector signed int vec_any_eq (vector unsigned short,
7076 vector signed short);
7077 vector signed int vec_any_eq (vector unsigned short,
7078 vector unsigned short);
7079 vector signed int vec_any_eq (vector signed int, vector unsigned int);
7080 vector signed int vec_any_eq (vector signed int, vector signed int);
7081 vector signed int vec_any_eq (vector unsigned int, vector signed int);
7082 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
7084 vector signed int vec_any_eq (vector float, vector float);
7086 vector signed int vec_any_ge (vector signed char, vector unsigned char);
7088 vector signed int vec_any_ge (vector unsigned char, vector signed char);
7090 vector signed int vec_any_ge (vector unsigned char,
7091 vector unsigned char);
7092 vector signed int vec_any_ge (vector signed char, vector signed char);
7093 vector signed int vec_any_ge (vector signed short,
7094 vector unsigned short);
7095 vector signed int vec_any_ge (vector unsigned short,
7096 vector signed short);
7097 vector signed int vec_any_ge (vector unsigned short,
7098 vector unsigned short);
7099 vector signed int vec_any_ge (vector signed short, vector signed short);
7101 vector signed int vec_any_ge (vector signed int, vector unsigned int);
7102 vector signed int vec_any_ge (vector unsigned int, vector signed int);
7103 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
7105 vector signed int vec_any_ge (vector signed int, vector signed int);
7106 vector signed int vec_any_ge (vector float, vector float);
7108 vector signed int vec_any_gt (vector signed char, vector unsigned char);
7110 vector signed int vec_any_gt (vector unsigned char, vector signed char);
7112 vector signed int vec_any_gt (vector unsigned char,
7113 vector unsigned char);
7114 vector signed int vec_any_gt (vector signed char, vector signed char);
7115 vector signed int vec_any_gt (vector signed short,
7116 vector unsigned short);
7117 vector signed int vec_any_gt (vector unsigned short,
7118 vector signed short);
7119 vector signed int vec_any_gt (vector unsigned short,
7120 vector unsigned short);
7121 vector signed int vec_any_gt (vector signed short, vector signed short);
7123 vector signed int vec_any_gt (vector signed int, vector unsigned int);
7124 vector signed int vec_any_gt (vector unsigned int, vector signed int);
7125 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
7127 vector signed int vec_any_gt (vector signed int, vector signed int);
7128 vector signed int vec_any_gt (vector float, vector float);
7130 vector signed int vec_any_le (vector signed char, vector unsigned char);
7132 vector signed int vec_any_le (vector unsigned char, vector signed char);
7134 vector signed int vec_any_le (vector unsigned char,
7135 vector unsigned char);
7136 vector signed int vec_any_le (vector signed char, vector signed char);
7137 vector signed int vec_any_le (vector signed short,
7138 vector unsigned short);
7139 vector signed int vec_any_le (vector unsigned short,
7140 vector signed short);
7141 vector signed int vec_any_le (vector unsigned short,
7142 vector unsigned short);
7143 vector signed int vec_any_le (vector signed short, vector signed short);
7145 vector signed int vec_any_le (vector signed int, vector unsigned int);
7146 vector signed int vec_any_le (vector unsigned int, vector signed int);
7147 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
7149 vector signed int vec_any_le (vector signed int, vector signed int);
7150 vector signed int vec_any_le (vector float, vector float);
7152 vector signed int vec_any_lt (vector signed char, vector unsigned char);
7154 vector signed int vec_any_lt (vector unsigned char, vector signed char);
7156 vector signed int vec_any_lt (vector unsigned char,
7157 vector unsigned char);
7158 vector signed int vec_any_lt (vector signed char, vector signed char);
7159 vector signed int vec_any_lt (vector signed short,
7160 vector unsigned short);
7161 vector signed int vec_any_lt (vector unsigned short,
7162 vector signed short);
7163 vector signed int vec_any_lt (vector unsigned short,
7164 vector unsigned short);
7165 vector signed int vec_any_lt (vector signed short, vector signed short);
7167 vector signed int vec_any_lt (vector signed int, vector unsigned int);
7168 vector signed int vec_any_lt (vector unsigned int, vector signed int);
7169 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
7171 vector signed int vec_any_lt (vector signed int, vector signed int);
7172 vector signed int vec_any_lt (vector float, vector float);
7174 vector signed int vec_any_nan (vector float);
7176 vector signed int vec_any_ne (vector signed char, vector unsigned char);
7178 vector signed int vec_any_ne (vector signed char, vector signed char);
7179 vector signed int vec_any_ne (vector unsigned char, vector signed char);
7181 vector signed int vec_any_ne (vector unsigned char,
7182 vector unsigned char);
7183 vector signed int vec_any_ne (vector signed short,
7184 vector unsigned short);
7185 vector signed int vec_any_ne (vector signed short, vector signed short);
7187 vector signed int vec_any_ne (vector unsigned short,
7188 vector signed short);
7189 vector signed int vec_any_ne (vector unsigned short,
7190 vector unsigned short);
7191 vector signed int vec_any_ne (vector signed int, vector unsigned int);
7192 vector signed int vec_any_ne (vector signed int, vector signed int);
7193 vector signed int vec_any_ne (vector unsigned int, vector signed int);
7194 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
7196 vector signed int vec_any_ne (vector float, vector float);
7198 vector signed int vec_any_nge (vector float, vector float);
7200 vector signed int vec_any_ngt (vector float, vector float);
7202 vector signed int vec_any_nle (vector float, vector float);
7204 vector signed int vec_any_nlt (vector float, vector float);
7206 vector signed int vec_any_numeric (vector float);
7208 vector signed int vec_any_out (vector float, vector float);
7212 @section Pragmas Accepted by GCC
7216 GCC supports several types of pragmas, primarily in order to compile
7217 code originally written for other compilers. Note that in general
7218 we do not recommend the use of pragmas; @xref{Function Attributes},
7219 for further explanation.
7223 * RS/6000 and PowerPC Pragmas::
7230 @subsection ARM Pragmas
7232 The ARM target defines pragmas for controlling the default addition of
7233 @code{long_call} and @code{short_call} attributes to functions.
7234 @xref{Function Attributes}, for information about the effects of these
7239 @cindex pragma, long_calls
7240 Set all subsequent functions to have the @code{long_call} attribute.
7243 @cindex pragma, no_long_calls
7244 Set all subsequent functions to have the @code{short_call} attribute.
7246 @item long_calls_off
7247 @cindex pragma, long_calls_off
7248 Do not affect the @code{long_call} or @code{short_call} attributes of
7249 subsequent functions.
7252 @node RS/6000 and PowerPC Pragmas
7253 @subsection RS/6000 and PowerPC Pragmas
7255 The RS/6000 and PowerPC targets define one pragma for controlling
7256 whether or not the @code{longcall} attribute is added to function
7257 declarations by default. This pragma overrides the @option{-mlongcall}
7258 option, but not the @code{longcall} and @code{shortcall} attributes.
7259 @xref{RS/6000 and PowerPC Options}, for more information about when long
7260 calls are and are not necessary.
7264 @cindex pragma, longcall
7265 Apply the @code{longcall} attribute to all subsequent function
7269 Do not apply the @code{longcall} attribute to subsequent function
7273 @c Describe c4x pragmas here.
7274 @c Describe h8300 pragmas here.
7275 @c Describe i370 pragmas here.
7276 @c Describe i960 pragmas here.
7277 @c Describe sh pragmas here.
7278 @c Describe v850 pragmas here.
7280 @node Darwin Pragmas
7281 @subsection Darwin Pragmas
7283 The following pragmas are available for all architectures running the
7284 Darwin operating system. These are useful for compatibility with other
7288 @item mark @var{tokens}@dots{}
7289 @cindex pragma, mark
7290 This pragma is accepted, but has no effect.
7292 @item options align=@var{alignment}
7293 @cindex pragma, options align
7294 This pragma sets the alignment of fields in structures. The values of
7295 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
7296 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
7297 properly; to restore the previous setting, use @code{reset} for the
7300 @item segment @var{tokens}@dots{}
7301 @cindex pragma, segment
7302 This pragma is accepted, but has no effect.
7304 @item unused (@var{var} [, @var{var}]@dots{})
7305 @cindex pragma, unused
7306 This pragma declares variables to be possibly unused. GCC will not
7307 produce warnings for the listed variables. The effect is similar to
7308 that of the @code{unused} attribute, except that this pragma may appear
7309 anywhere within the variables' scopes.
7312 @node Solaris Pragmas
7313 @subsection Solaris Pragmas
7315 For compatibility with the SunPRO compiler, the following pragma
7319 @item redefine_extname @var{oldname} @var{newname}
7320 @cindex pragma, redefine_extname
7322 This pragma gives the C function @var{oldname} the assembler label
7323 @var{newname}. The pragma must appear before the function declaration.
7324 This pragma is equivalent to the asm labels extension (@pxref{Asm
7325 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
7326 if the pragma is available.
7330 @subsection Tru64 Pragmas
7332 For compatibility with the Compaq C compiler, the following pragma
7336 @item extern_prefix @var{string}
7337 @cindex pragma, extern_prefix
7339 This pragma renames all subsequent function and variable declarations
7340 such that @var{string} is prepended to the name. This effect may be
7341 terminated by using another @code{extern_prefix} pragma with the
7344 This pragma is similar in intent to to the asm labels extension
7345 (@pxref{Asm Labels}) in that the system programmer wants to change
7346 the assembly-level ABI without changing the source-level API. The
7347 preprocessor defines @code{__PRAGMA_EXTERN_PREFIX} if the pragma is
7351 @node Unnamed Fields
7352 @section Unnamed struct/union fields within structs/unions.
7356 For compatibility with other compilers, GCC allows you to define
7357 a structure or union that contains, as fields, structures and unions
7358 without names. For example:
7371 In this example, the user would be able to access members of the unnamed
7372 union with code like @samp{foo.b}. Note that only unnamed structs and
7373 unions are allowed, you may not have, for example, an unnamed
7376 You must never create such structures that cause ambiguous field definitions.
7377 For example, this structure:
7388 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
7389 Such constructs are not supported and must be avoided. In the future,
7390 such constructs may be detected and treated as compilation errors.
7393 @section Thread-Local Storage
7394 @cindex Thread-Local Storage
7395 @cindex @acronym{TLS}
7398 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
7399 are allocated such that there is one instance of the variable per extant
7400 thread. The run-time model GCC uses to implement this originates
7401 in the IA-64 processor-specific ABI, but has since been migrated
7402 to other processors as well. It requires significant support from
7403 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
7404 system libraries (@file{libc.so} and @file{libpthread.so}), so it
7405 is not available everywhere.
7407 At the user level, the extension is visible with a new storage
7408 class keyword: @code{__thread}. For example:
7412 extern __thread struct state s;
7413 static __thread char *p;
7416 The @code{__thread} specifier may be used alone, with the @code{extern}
7417 or @code{static} specifiers, but with no other storage class specifier.
7418 When used with @code{extern} or @code{static}, @code{__thread} must appear
7419 immediately after the other storage class specifier.
7421 The @code{__thread} specifier may be applied to any global, file-scoped
7422 static, function-scoped static, or static data member of a class. It may
7423 not be applied to block-scoped automatic or non-static data member.
7425 When the address-of operator is applied to a thread-local variable, it is
7426 evaluated at run-time and returns the address of the current thread's
7427 instance of that variable. An address so obtained may be used by any
7428 thread. When a thread terminates, any pointers to thread-local variables
7429 in that thread become invalid.
7431 No static initialization may refer to the address of a thread-local variable.
7433 In C++, if an initializer is present for a thread-local variable, it must
7434 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
7437 See @uref{http://people.redhat.com/drepper/tls.pdf,
7438 ELF Handling For Thread-Local Storage} for a detailed explanation of
7439 the four thread-local storage addressing models, and how the run-time
7440 is expected to function.
7443 * C99 Thread-Local Edits::
7444 * C++98 Thread-Local Edits::
7447 @node C99 Thread-Local Edits
7448 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
7450 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
7451 that document the exact semantics of the language extension.
7455 @cite{5.1.2 Execution environments}
7457 Add new text after paragraph 1
7460 Within either execution environment, a @dfn{thread} is a flow of
7461 control within a program. It is implementation defined whether
7462 or not there may be more than one thread associated with a program.
7463 It is implementation defined how threads beyond the first are
7464 created, the name and type of the function called at thread
7465 startup, and how threads may be terminated. However, objects
7466 with thread storage duration shall be initialized before thread
7471 @cite{6.2.4 Storage durations of objects}
7473 Add new text before paragraph 3
7476 An object whose identifier is declared with the storage-class
7477 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
7478 Its lifetime is the entire execution of the thread, and its
7479 stored value is initialized only once, prior to thread startup.
7483 @cite{6.4.1 Keywords}
7485 Add @code{__thread}.
7488 @cite{6.7.1 Storage-class specifiers}
7490 Add @code{__thread} to the list of storage class specifiers in
7493 Change paragraph 2 to
7496 With the exception of @code{__thread}, at most one storage-class
7497 specifier may be given [@dots{}]. The @code{__thread} specifier may
7498 be used alone, or immediately following @code{extern} or
7502 Add new text after paragraph 6
7505 The declaration of an identifier for a variable that has
7506 block scope that specifies @code{__thread} shall also
7507 specify either @code{extern} or @code{static}.
7509 The @code{__thread} specifier shall be used only with
7514 @node C++98 Thread-Local Edits
7515 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
7517 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
7518 that document the exact semantics of the language extension.
7522 @b{[intro.execution]}
7524 New text after paragraph 4
7527 A @dfn{thread} is a flow of control within the abstract machine.
7528 It is implementation defined whether or not there may be more than
7532 New text after paragraph 7
7535 It is unspecified whether additional action must be taken to
7536 ensure when and whether side effects are visible to other threads.
7542 Add @code{__thread}.
7545 @b{[basic.start.main]}
7547 Add after paragraph 5
7550 The thread that begins execution at the @code{main} function is called
7551 the @dfn{main thread}. It is implementation defined how functions
7552 beginning threads other than the main thread are designated or typed.
7553 A function so designated, as well as the @code{main} function, is called
7554 a @dfn{thread startup function}. It is implementation defined what
7555 happens if a thread startup function returns. It is implementation
7556 defined what happens to other threads when any thread calls @code{exit}.
7560 @b{[basic.start.init]}
7562 Add after paragraph 4
7565 The storage for an object of thread storage duration shall be
7566 statically initialized before the first statement of the thread startup
7567 function. An object of thread storage duration shall not require
7568 dynamic initialization.
7572 @b{[basic.start.term]}
7574 Add after paragraph 3
7577 The type of an object with thread storage duration shall not have a
7578 non-trivial destructor, nor shall it be an array type whose elements
7579 (directly or indirectly) have non-trivial destructors.
7585 Add ``thread storage duration'' to the list in paragraph 1.
7590 Thread, static, and automatic storage durations are associated with
7591 objects introduced by declarations [@dots{}].
7594 Add @code{__thread} to the list of specifiers in paragraph 3.
7597 @b{[basic.stc.thread]}
7599 New section before @b{[basic.stc.static]}
7602 The keyword @code{__thread} applied to a non-local object gives the
7603 object thread storage duration.
7605 A local variable or class data member declared both @code{static}
7606 and @code{__thread} gives the variable or member thread storage
7611 @b{[basic.stc.static]}
7616 All objects which have neither thread storage duration, dynamic
7617 storage duration nor are local [@dots{}].
7623 Add @code{__thread} to the list in paragraph 1.
7628 With the exception of @code{__thread}, at most one
7629 @var{storage-class-specifier} shall appear in a given
7630 @var{decl-specifier-seq}. The @code{__thread} specifier may
7631 be used alone, or immediately following the @code{extern} or
7632 @code{static} specifiers. [@dots{}]
7635 Add after paragraph 5
7638 The @code{__thread} specifier can be applied only to the names of objects
7639 and to anonymous unions.
7645 Add after paragraph 6
7648 Non-@code{static} members shall not be @code{__thread}.
7652 @node C++ Extensions
7653 @chapter Extensions to the C++ Language
7654 @cindex extensions, C++ language
7655 @cindex C++ language extensions
7657 The GNU compiler provides these extensions to the C++ language (and you
7658 can also use most of the C language extensions in your C++ programs). If you
7659 want to write code that checks whether these features are available, you can
7660 test for the GNU compiler the same way as for C programs: check for a
7661 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
7662 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
7663 Predefined Macros,cpp,The GNU C Preprocessor}).
7666 * Min and Max:: C++ Minimum and maximum operators.
7667 * Volatiles:: What constitutes an access to a volatile object.
7668 * Restricted Pointers:: C99 restricted pointers and references.
7669 * Vague Linkage:: Where G++ puts inlines, vtables and such.
7670 * C++ Interface:: You can use a single C++ header file for both
7671 declarations and definitions.
7672 * Template Instantiation:: Methods for ensuring that exactly one copy of
7673 each needed template instantiation is emitted.
7674 * Bound member functions:: You can extract a function pointer to the
7675 method denoted by a @samp{->*} or @samp{.*} expression.
7676 * C++ Attributes:: Variable, function, and type attributes for C++ only.
7677 * Strong Using:: Strong using-directives for namespace composition.
7678 * Offsetof:: Special syntax for implementing @code{offsetof}.
7679 * Java Exceptions:: Tweaking exception handling to work with Java.
7680 * Deprecated Features:: Things will disappear from g++.
7681 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
7685 @section Minimum and Maximum Operators in C++
7687 It is very convenient to have operators which return the ``minimum'' or the
7688 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
7691 @item @var{a} <? @var{b}
7693 @cindex minimum operator
7694 is the @dfn{minimum}, returning the smaller of the numeric values
7695 @var{a} and @var{b};
7697 @item @var{a} >? @var{b}
7699 @cindex maximum operator
7700 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
7704 These operations are not primitive in ordinary C++, since you can
7705 use a macro to return the minimum of two things in C++, as in the
7709 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
7713 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
7714 the minimum value of variables @var{i} and @var{j}.
7716 However, side effects in @code{X} or @code{Y} may cause unintended
7717 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
7718 the smaller counter twice. The GNU C @code{typeof} extension allows you
7719 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
7720 However, writing @code{MIN} and @code{MAX} as macros also forces you to
7721 use function-call notation for a fundamental arithmetic operation.
7722 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
7725 Since @code{<?} and @code{>?} are built into the compiler, they properly
7726 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
7730 @section When is a Volatile Object Accessed?
7731 @cindex accessing volatiles
7732 @cindex volatile read
7733 @cindex volatile write
7734 @cindex volatile access
7736 Both the C and C++ standard have the concept of volatile objects. These
7737 are normally accessed by pointers and used for accessing hardware. The
7738 standards encourage compilers to refrain from optimizations
7739 concerning accesses to volatile objects that it might perform on
7740 non-volatile objects. The C standard leaves it implementation defined
7741 as to what constitutes a volatile access. The C++ standard omits to
7742 specify this, except to say that C++ should behave in a similar manner
7743 to C with respect to volatiles, where possible. The minimum either
7744 standard specifies is that at a sequence point all previous accesses to
7745 volatile objects have stabilized and no subsequent accesses have
7746 occurred. Thus an implementation is free to reorder and combine
7747 volatile accesses which occur between sequence points, but cannot do so
7748 for accesses across a sequence point. The use of volatiles does not
7749 allow you to violate the restriction on updating objects multiple times
7750 within a sequence point.
7752 In most expressions, it is intuitively obvious what is a read and what is
7753 a write. For instance
7756 volatile int *dst = @var{somevalue};
7757 volatile int *src = @var{someothervalue};
7762 will cause a read of the volatile object pointed to by @var{src} and stores the
7763 value into the volatile object pointed to by @var{dst}. There is no
7764 guarantee that these reads and writes are atomic, especially for objects
7765 larger than @code{int}.
7767 Less obvious expressions are where something which looks like an access
7768 is used in a void context. An example would be,
7771 volatile int *src = @var{somevalue};
7775 With C, such expressions are rvalues, and as rvalues cause a read of
7776 the object, GCC interprets this as a read of the volatile being pointed
7777 to. The C++ standard specifies that such expressions do not undergo
7778 lvalue to rvalue conversion, and that the type of the dereferenced
7779 object may be incomplete. The C++ standard does not specify explicitly
7780 that it is this lvalue to rvalue conversion which is responsible for
7781 causing an access. However, there is reason to believe that it is,
7782 because otherwise certain simple expressions become undefined. However,
7783 because it would surprise most programmers, G++ treats dereferencing a
7784 pointer to volatile object of complete type in a void context as a read
7785 of the object. When the object has incomplete type, G++ issues a
7790 struct T @{int m;@};
7791 volatile S *ptr1 = @var{somevalue};
7792 volatile T *ptr2 = @var{somevalue};
7797 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
7798 causes a read of the object pointed to. If you wish to force an error on
7799 the first case, you must force a conversion to rvalue with, for instance
7800 a static cast, @code{static_cast<S>(*ptr1)}.
7802 When using a reference to volatile, G++ does not treat equivalent
7803 expressions as accesses to volatiles, but instead issues a warning that
7804 no volatile is accessed. The rationale for this is that otherwise it
7805 becomes difficult to determine where volatile access occur, and not
7806 possible to ignore the return value from functions returning volatile
7807 references. Again, if you wish to force a read, cast the reference to
7810 @node Restricted Pointers
7811 @section Restricting Pointer Aliasing
7812 @cindex restricted pointers
7813 @cindex restricted references
7814 @cindex restricted this pointer
7816 As with the C front end, G++ understands the C99 feature of restricted pointers,
7817 specified with the @code{__restrict__}, or @code{__restrict} type
7818 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
7819 language flag, @code{restrict} is not a keyword in C++.
7821 In addition to allowing restricted pointers, you can specify restricted
7822 references, which indicate that the reference is not aliased in the local
7826 void fn (int *__restrict__ rptr, int &__restrict__ rref)
7833 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
7834 @var{rref} refers to a (different) unaliased integer.
7836 You may also specify whether a member function's @var{this} pointer is
7837 unaliased by using @code{__restrict__} as a member function qualifier.
7840 void T::fn () __restrict__
7847 Within the body of @code{T::fn}, @var{this} will have the effective
7848 definition @code{T *__restrict__ const this}. Notice that the
7849 interpretation of a @code{__restrict__} member function qualifier is
7850 different to that of @code{const} or @code{volatile} qualifier, in that it
7851 is applied to the pointer rather than the object. This is consistent with
7852 other compilers which implement restricted pointers.
7854 As with all outermost parameter qualifiers, @code{__restrict__} is
7855 ignored in function definition matching. This means you only need to
7856 specify @code{__restrict__} in a function definition, rather than
7857 in a function prototype as well.
7860 @section Vague Linkage
7861 @cindex vague linkage
7863 There are several constructs in C++ which require space in the object
7864 file but are not clearly tied to a single translation unit. We say that
7865 these constructs have ``vague linkage''. Typically such constructs are
7866 emitted wherever they are needed, though sometimes we can be more
7870 @item Inline Functions
7871 Inline functions are typically defined in a header file which can be
7872 included in many different compilations. Hopefully they can usually be
7873 inlined, but sometimes an out-of-line copy is necessary, if the address
7874 of the function is taken or if inlining fails. In general, we emit an
7875 out-of-line copy in all translation units where one is needed. As an
7876 exception, we only emit inline virtual functions with the vtable, since
7877 it will always require a copy.
7879 Local static variables and string constants used in an inline function
7880 are also considered to have vague linkage, since they must be shared
7881 between all inlined and out-of-line instances of the function.
7885 C++ virtual functions are implemented in most compilers using a lookup
7886 table, known as a vtable. The vtable contains pointers to the virtual
7887 functions provided by a class, and each object of the class contains a
7888 pointer to its vtable (or vtables, in some multiple-inheritance
7889 situations). If the class declares any non-inline, non-pure virtual
7890 functions, the first one is chosen as the ``key method'' for the class,
7891 and the vtable is only emitted in the translation unit where the key
7894 @emph{Note:} If the chosen key method is later defined as inline, the
7895 vtable will still be emitted in every translation unit which defines it.
7896 Make sure that any inline virtuals are declared inline in the class
7897 body, even if they are not defined there.
7899 @item type_info objects
7902 C++ requires information about types to be written out in order to
7903 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7904 For polymorphic classes (classes with virtual functions), the type_info
7905 object is written out along with the vtable so that @samp{dynamic_cast}
7906 can determine the dynamic type of a class object at runtime. For all
7907 other types, we write out the type_info object when it is used: when
7908 applying @samp{typeid} to an expression, throwing an object, or
7909 referring to a type in a catch clause or exception specification.
7911 @item Template Instantiations
7912 Most everything in this section also applies to template instantiations,
7913 but there are other options as well.
7914 @xref{Template Instantiation,,Where's the Template?}.
7918 When used with GNU ld version 2.8 or later on an ELF system such as
7919 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
7920 these constructs will be discarded at link time. This is known as
7923 On targets that don't support COMDAT, but do support weak symbols, GCC
7924 will use them. This way one copy will override all the others, but
7925 the unused copies will still take up space in the executable.
7927 For targets which do not support either COMDAT or weak symbols,
7928 most entities with vague linkage will be emitted as local symbols to
7929 avoid duplicate definition errors from the linker. This will not happen
7930 for local statics in inlines, however, as having multiple copies will
7931 almost certainly break things.
7933 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7934 another way to control placement of these constructs.
7937 @section Declarations and Definitions in One Header
7939 @cindex interface and implementation headers, C++
7940 @cindex C++ interface and implementation headers
7941 C++ object definitions can be quite complex. In principle, your source
7942 code will need two kinds of things for each object that you use across
7943 more than one source file. First, you need an @dfn{interface}
7944 specification, describing its structure with type declarations and
7945 function prototypes. Second, you need the @dfn{implementation} itself.
7946 It can be tedious to maintain a separate interface description in a
7947 header file, in parallel to the actual implementation. It is also
7948 dangerous, since separate interface and implementation definitions may
7949 not remain parallel.
7951 @cindex pragmas, interface and implementation
7952 With GNU C++, you can use a single header file for both purposes.
7955 @emph{Warning:} The mechanism to specify this is in transition. For the
7956 nonce, you must use one of two @code{#pragma} commands; in a future
7957 release of GNU C++, an alternative mechanism will make these
7958 @code{#pragma} commands unnecessary.
7961 The header file contains the full definitions, but is marked with
7962 @samp{#pragma interface} in the source code. This allows the compiler
7963 to use the header file only as an interface specification when ordinary
7964 source files incorporate it with @code{#include}. In the single source
7965 file where the full implementation belongs, you can use either a naming
7966 convention or @samp{#pragma implementation} to indicate this alternate
7967 use of the header file.
7970 @item #pragma interface
7971 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7972 @kindex #pragma interface
7973 Use this directive in @emph{header files} that define object classes, to save
7974 space in most of the object files that use those classes. Normally,
7975 local copies of certain information (backup copies of inline member
7976 functions, debugging information, and the internal tables that implement
7977 virtual functions) must be kept in each object file that includes class
7978 definitions. You can use this pragma to avoid such duplication. When a
7979 header file containing @samp{#pragma interface} is included in a
7980 compilation, this auxiliary information will not be generated (unless
7981 the main input source file itself uses @samp{#pragma implementation}).
7982 Instead, the object files will contain references to be resolved at link
7985 The second form of this directive is useful for the case where you have
7986 multiple headers with the same name in different directories. If you
7987 use this form, you must specify the same string to @samp{#pragma
7990 @item #pragma implementation
7991 @itemx #pragma implementation "@var{objects}.h"
7992 @kindex #pragma implementation
7993 Use this pragma in a @emph{main input file}, when you want full output from
7994 included header files to be generated (and made globally visible). The
7995 included header file, in turn, should use @samp{#pragma interface}.
7996 Backup copies of inline member functions, debugging information, and the
7997 internal tables used to implement virtual functions are all generated in
7998 implementation files.
8000 @cindex implied @code{#pragma implementation}
8001 @cindex @code{#pragma implementation}, implied
8002 @cindex naming convention, implementation headers
8003 If you use @samp{#pragma implementation} with no argument, it applies to
8004 an include file with the same basename@footnote{A file's @dfn{basename}
8005 was the name stripped of all leading path information and of trailing
8006 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
8007 file. For example, in @file{allclass.cc}, giving just
8008 @samp{#pragma implementation}
8009 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
8011 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
8012 an implementation file whenever you would include it from
8013 @file{allclass.cc} even if you never specified @samp{#pragma
8014 implementation}. This was deemed to be more trouble than it was worth,
8015 however, and disabled.
8017 If you use an explicit @samp{#pragma implementation}, it must appear in
8018 your source file @emph{before} you include the affected header files.
8020 Use the string argument if you want a single implementation file to
8021 include code from multiple header files. (You must also use
8022 @samp{#include} to include the header file; @samp{#pragma
8023 implementation} only specifies how to use the file---it doesn't actually
8026 There is no way to split up the contents of a single header file into
8027 multiple implementation files.
8030 @cindex inlining and C++ pragmas
8031 @cindex C++ pragmas, effect on inlining
8032 @cindex pragmas in C++, effect on inlining
8033 @samp{#pragma implementation} and @samp{#pragma interface} also have an
8034 effect on function inlining.
8036 If you define a class in a header file marked with @samp{#pragma
8037 interface}, the effect on a function defined in that class is similar to
8038 an explicit @code{extern} declaration---the compiler emits no code at
8039 all to define an independent version of the function. Its definition
8040 is used only for inlining with its callers.
8042 @opindex fno-implement-inlines
8043 Conversely, when you include the same header file in a main source file
8044 that declares it as @samp{#pragma implementation}, the compiler emits
8045 code for the function itself; this defines a version of the function
8046 that can be found via pointers (or by callers compiled without
8047 inlining). If all calls to the function can be inlined, you can avoid
8048 emitting the function by compiling with @option{-fno-implement-inlines}.
8049 If any calls were not inlined, you will get linker errors.
8051 @node Template Instantiation
8052 @section Where's the Template?
8053 @cindex template instantiation
8055 C++ templates are the first language feature to require more
8056 intelligence from the environment than one usually finds on a UNIX
8057 system. Somehow the compiler and linker have to make sure that each
8058 template instance occurs exactly once in the executable if it is needed,
8059 and not at all otherwise. There are two basic approaches to this
8060 problem, which I will refer to as the Borland model and the Cfront model.
8064 Borland C++ solved the template instantiation problem by adding the code
8065 equivalent of common blocks to their linker; the compiler emits template
8066 instances in each translation unit that uses them, and the linker
8067 collapses them together. The advantage of this model is that the linker
8068 only has to consider the object files themselves; there is no external
8069 complexity to worry about. This disadvantage is that compilation time
8070 is increased because the template code is being compiled repeatedly.
8071 Code written for this model tends to include definitions of all
8072 templates in the header file, since they must be seen to be
8076 The AT&T C++ translator, Cfront, solved the template instantiation
8077 problem by creating the notion of a template repository, an
8078 automatically maintained place where template instances are stored. A
8079 more modern version of the repository works as follows: As individual
8080 object files are built, the compiler places any template definitions and
8081 instantiations encountered in the repository. At link time, the link
8082 wrapper adds in the objects in the repository and compiles any needed
8083 instances that were not previously emitted. The advantages of this
8084 model are more optimal compilation speed and the ability to use the
8085 system linker; to implement the Borland model a compiler vendor also
8086 needs to replace the linker. The disadvantages are vastly increased
8087 complexity, and thus potential for error; for some code this can be
8088 just as transparent, but in practice it can been very difficult to build
8089 multiple programs in one directory and one program in multiple
8090 directories. Code written for this model tends to separate definitions
8091 of non-inline member templates into a separate file, which should be
8092 compiled separately.
8095 When used with GNU ld version 2.8 or later on an ELF system such as
8096 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
8097 Borland model. On other systems, G++ implements neither automatic
8100 A future version of G++ will support a hybrid model whereby the compiler
8101 will emit any instantiations for which the template definition is
8102 included in the compile, and store template definitions and
8103 instantiation context information into the object file for the rest.
8104 The link wrapper will extract that information as necessary and invoke
8105 the compiler to produce the remaining instantiations. The linker will
8106 then combine duplicate instantiations.
8108 In the mean time, you have the following options for dealing with
8109 template instantiations:
8114 Compile your template-using code with @option{-frepo}. The compiler will
8115 generate files with the extension @samp{.rpo} listing all of the
8116 template instantiations used in the corresponding object files which
8117 could be instantiated there; the link wrapper, @samp{collect2}, will
8118 then update the @samp{.rpo} files to tell the compiler where to place
8119 those instantiations and rebuild any affected object files. The
8120 link-time overhead is negligible after the first pass, as the compiler
8121 will continue to place the instantiations in the same files.
8123 This is your best option for application code written for the Borland
8124 model, as it will just work. Code written for the Cfront model will
8125 need to be modified so that the template definitions are available at
8126 one or more points of instantiation; usually this is as simple as adding
8127 @code{#include <tmethods.cc>} to the end of each template header.
8129 For library code, if you want the library to provide all of the template
8130 instantiations it needs, just try to link all of its object files
8131 together; the link will fail, but cause the instantiations to be
8132 generated as a side effect. Be warned, however, that this may cause
8133 conflicts if multiple libraries try to provide the same instantiations.
8134 For greater control, use explicit instantiation as described in the next
8138 @opindex fno-implicit-templates
8139 Compile your code with @option{-fno-implicit-templates} to disable the
8140 implicit generation of template instances, and explicitly instantiate
8141 all the ones you use. This approach requires more knowledge of exactly
8142 which instances you need than do the others, but it's less
8143 mysterious and allows greater control. You can scatter the explicit
8144 instantiations throughout your program, perhaps putting them in the
8145 translation units where the instances are used or the translation units
8146 that define the templates themselves; you can put all of the explicit
8147 instantiations you need into one big file; or you can create small files
8154 template class Foo<int>;
8155 template ostream& operator <<
8156 (ostream&, const Foo<int>&);
8159 for each of the instances you need, and create a template instantiation
8162 If you are using Cfront-model code, you can probably get away with not
8163 using @option{-fno-implicit-templates} when compiling files that don't
8164 @samp{#include} the member template definitions.
8166 If you use one big file to do the instantiations, you may want to
8167 compile it without @option{-fno-implicit-templates} so you get all of the
8168 instances required by your explicit instantiations (but not by any
8169 other files) without having to specify them as well.
8171 G++ has extended the template instantiation syntax given in the ISO
8172 standard to allow forward declaration of explicit instantiations
8173 (with @code{extern}), instantiation of the compiler support data for a
8174 template class (i.e.@: the vtable) without instantiating any of its
8175 members (with @code{inline}), and instantiation of only the static data
8176 members of a template class, without the support data or member
8177 functions (with (@code{static}):
8180 extern template int max (int, int);
8181 inline template class Foo<int>;
8182 static template class Foo<int>;
8186 Do nothing. Pretend G++ does implement automatic instantiation
8187 management. Code written for the Borland model will work fine, but
8188 each translation unit will contain instances of each of the templates it
8189 uses. In a large program, this can lead to an unacceptable amount of code
8192 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
8193 more discussion of these pragmas.
8196 @node Bound member functions
8197 @section Extracting the function pointer from a bound pointer to member function
8199 @cindex pointer to member function
8200 @cindex bound pointer to member function
8202 In C++, pointer to member functions (PMFs) are implemented using a wide
8203 pointer of sorts to handle all the possible call mechanisms; the PMF
8204 needs to store information about how to adjust the @samp{this} pointer,
8205 and if the function pointed to is virtual, where to find the vtable, and
8206 where in the vtable to look for the member function. If you are using
8207 PMFs in an inner loop, you should really reconsider that decision. If
8208 that is not an option, you can extract the pointer to the function that
8209 would be called for a given object/PMF pair and call it directly inside
8210 the inner loop, to save a bit of time.
8212 Note that you will still be paying the penalty for the call through a
8213 function pointer; on most modern architectures, such a call defeats the
8214 branch prediction features of the CPU@. This is also true of normal
8215 virtual function calls.
8217 The syntax for this extension is
8221 extern int (A::*fp)();
8222 typedef int (*fptr)(A *);
8224 fptr p = (fptr)(a.*fp);
8227 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
8228 no object is needed to obtain the address of the function. They can be
8229 converted to function pointers directly:
8232 fptr p1 = (fptr)(&A::foo);
8235 @opindex Wno-pmf-conversions
8236 You must specify @option{-Wno-pmf-conversions} to use this extension.
8238 @node C++ Attributes
8239 @section C++-Specific Variable, Function, and Type Attributes
8241 Some attributes only make sense for C++ programs.
8244 @item init_priority (@var{priority})
8245 @cindex init_priority attribute
8248 In Standard C++, objects defined at namespace scope are guaranteed to be
8249 initialized in an order in strict accordance with that of their definitions
8250 @emph{in a given translation unit}. No guarantee is made for initializations
8251 across translation units. However, GNU C++ allows users to control the
8252 order of initialization of objects defined at namespace scope with the
8253 @code{init_priority} attribute by specifying a relative @var{priority},
8254 a constant integral expression currently bounded between 101 and 65535
8255 inclusive. Lower numbers indicate a higher priority.
8257 In the following example, @code{A} would normally be created before
8258 @code{B}, but the @code{init_priority} attribute has reversed that order:
8261 Some_Class A __attribute__ ((init_priority (2000)));
8262 Some_Class B __attribute__ ((init_priority (543)));
8266 Note that the particular values of @var{priority} do not matter; only their
8269 @item java_interface
8270 @cindex java_interface attribute
8272 This type attribute informs C++ that the class is a Java interface. It may
8273 only be applied to classes declared within an @code{extern "Java"} block.
8274 Calls to methods declared in this interface will be dispatched using GCJ's
8275 interface table mechanism, instead of regular virtual table dispatch.
8279 See also @xref{Strong Using}.
8282 @section Strong Using
8284 @strong{Caution:} The semantics of this extension are not fully
8285 defined. Users should refrain from using this extension as its
8286 semantics may change subtly over time. It is possible that this
8287 extension wil be removed in future versions of G++.
8289 A using-directive with @code{__attribute ((strong))} is stronger
8290 than a normal using-directive in two ways:
8294 Templates from the used namespace can be specialized as though they were members of the using namespace.
8297 The using namespace is considered an associated namespace of all
8298 templates in the used namespace for purposes of argument-dependent
8302 This is useful for composing a namespace transparently from
8303 implementation namespaces. For example:
8308 template <class T> struct A @{ @};
8310 using namespace debug __attribute ((__strong__));
8311 template <> struct A<int> @{ @}; // ok to specialize
8313 template <class T> void f (A<T>);
8318 f (std::A<float>()); // lookup finds std::f
8326 G++ uses a syntactic extension to implement the @code{offsetof} macro.
8331 __offsetof__ (expression)
8334 is equivalent to the parenthesized expression, except that the
8335 expression is considered an integral constant expression even if it
8336 contains certain operators that are not normally permitted in an
8337 integral constant expression. Users should never use
8338 @code{__offsetof__} directly; the only valid use of
8339 @code{__offsetof__} is to implement the @code{offsetof} macro in
8342 @node Java Exceptions
8343 @section Java Exceptions
8345 The Java language uses a slightly different exception handling model
8346 from C++. Normally, GNU C++ will automatically detect when you are
8347 writing C++ code that uses Java exceptions, and handle them
8348 appropriately. However, if C++ code only needs to execute destructors
8349 when Java exceptions are thrown through it, GCC will guess incorrectly.
8350 Sample problematic code is:
8353 struct S @{ ~S(); @};
8354 extern void bar(); // is written in Java, and may throw exceptions
8363 The usual effect of an incorrect guess is a link failure, complaining of
8364 a missing routine called @samp{__gxx_personality_v0}.
8366 You can inform the compiler that Java exceptions are to be used in a
8367 translation unit, irrespective of what it might think, by writing
8368 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
8369 @samp{#pragma} must appear before any functions that throw or catch
8370 exceptions, or run destructors when exceptions are thrown through them.
8372 You cannot mix Java and C++ exceptions in the same translation unit. It
8373 is believed to be safe to throw a C++ exception from one file through
8374 another file compiled for the Java exception model, or vice versa, but
8375 there may be bugs in this area.
8377 @node Deprecated Features
8378 @section Deprecated Features
8380 In the past, the GNU C++ compiler was extended to experiment with new
8381 features, at a time when the C++ language was still evolving. Now that
8382 the C++ standard is complete, some of those features are superseded by
8383 superior alternatives. Using the old features might cause a warning in
8384 some cases that the feature will be dropped in the future. In other
8385 cases, the feature might be gone already.
8387 While the list below is not exhaustive, it documents some of the options
8388 that are now deprecated:
8391 @item -fexternal-templates
8392 @itemx -falt-external-templates
8393 These are two of the many ways for G++ to implement template
8394 instantiation. @xref{Template Instantiation}. The C++ standard clearly
8395 defines how template definitions have to be organized across
8396 implementation units. G++ has an implicit instantiation mechanism that
8397 should work just fine for standard-conforming code.
8399 @item -fstrict-prototype
8400 @itemx -fno-strict-prototype
8401 Previously it was possible to use an empty prototype parameter list to
8402 indicate an unspecified number of parameters (like C), rather than no
8403 parameters, as C++ demands. This feature has been removed, except where
8404 it is required for backwards compatibility @xref{Backwards Compatibility}.
8407 The named return value extension has been deprecated, and is now
8410 The use of initializer lists with new expressions has been deprecated,
8411 and is now removed from G++.
8413 Floating and complex non-type template parameters have been deprecated,
8414 and are now removed from G++.
8416 The implicit typename extension has been deprecated and is now
8419 The use of default arguments in function pointers, function typedefs and
8420 and other places where they are not permitted by the standard is
8421 deprecated and will be removed from a future version of G++.
8423 @node Backwards Compatibility
8424 @section Backwards Compatibility
8425 @cindex Backwards Compatibility
8426 @cindex ARM [Annotated C++ Reference Manual]
8428 Now that there is a definitive ISO standard C++, G++ has a specification
8429 to adhere to. The C++ language evolved over time, and features that
8430 used to be acceptable in previous drafts of the standard, such as the ARM
8431 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
8432 compilation of C++ written to such drafts, G++ contains some backwards
8433 compatibilities. @emph{All such backwards compatibility features are
8434 liable to disappear in future versions of G++.} They should be considered
8435 deprecated @xref{Deprecated Features}.
8439 If a variable is declared at for scope, it used to remain in scope until
8440 the end of the scope which contained the for statement (rather than just
8441 within the for scope). G++ retains this, but issues a warning, if such a
8442 variable is accessed outside the for scope.
8444 @item Implicit C language
8445 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
8446 scope to set the language. On such systems, all header files are
8447 implicitly scoped inside a C language scope. Also, an empty prototype
8448 @code{()} will be treated as an unspecified number of arguments, rather
8449 than no arguments, as C++ demands.