1 @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter C Implementation-defined behavior
7 @cindex implementation-defined behavior, C language
9 A conforming implementation of ISO C is required to document its
10 choice of behavior in each of the areas that are designated
11 ``implementation defined.'' The following lists all such areas,
12 along with the section number from the ISO/IEC 9899:1999 standard.
15 * Translation implementation::
16 * Environment implementation::
17 * Identifiers implementation::
18 * Characters implementation::
19 * Integers implementation::
20 * Floating point implementation::
21 * Arrays and pointers implementation::
22 * Hints implementation::
23 * Structures unions enumerations and bit-fields implementation::
24 * Qualifiers implementation::
25 * Preprocessing directives implementation::
26 * Library functions implementation::
27 * Architecture implementation::
28 * Locale-specific behavior implementation::
31 @node Translation implementation
36 @cite{How a diagnostic is identified (3.10, 5.1.1.3).}
38 Diagnostics consist of all the output sent to stderr by GCC.
41 @cite{Whether each nonempty sequence of white-space characters other than
42 new-line is retained or replaced by one space character in translation
46 @node Environment implementation
49 The behavior of these points are dependent on the implementation
50 of the C library, and are not defined by GCC itself.
52 @node Identifiers implementation
57 @cite{Which additional multibyte characters may appear in identifiers
58 and their correspondence to universal character names (6.4.2).}
61 @cite{The number of significant initial characters in an identifier
64 For internal names, all characters are significant. For external names,
65 the number of significant characters are defined by the linker; for
66 almost all targets, all characters are significant.
70 @node Characters implementation
75 @cite{The number of bits in a byte (3.6).}
78 @cite{The values of the members of the execution character set (5.2.1).}
81 @cite{The unique value of the member of the execution character set produced
82 for each of the standard alphabetic escape sequences (5.2.2).}
85 @cite{The value of a @code{char} object into which has been stored any
86 character other than a member of the basic execution character set (6.2.5).}
89 @cite{Which of @code{signed char} or @code{unsigned char} has the same range,
90 representation, and behavior as ``plain'' @code{char} (6.2.5, 6.3.1.1).}
93 @cite{The mapping of members of the source character set (in character
94 constants and string literals) to members of the execution character
95 set (6.4.4.4, 5.1.1.2).}
98 @cite{The value of an integer character constant containing more than one
99 character or containing a character or escape sequence that does not map
100 to a single-byte execution character (6.4.4.4).}
103 @cite{The value of a wide character constant containing more than one
104 multibyte character, or containing a multibyte character or escape
105 sequence not represented in the extended execution character set (6.4.4.4).}
108 @cite{The current locale used to convert a wide character constant consisting
109 of a single multibyte character that maps to a member of the extended
110 execution character set into a corresponding wide character code (6.4.4.4).}
113 @cite{The current locale used to convert a wide string literal into
114 corresponding wide character codes (6.4.5).}
117 @cite{The value of a string literal containing a multibyte character or escape
118 sequence not represented in the execution character set (6.4.5).}
121 @node Integers implementation
126 @cite{Any extended integer types that exist in the implementation (6.2.5).}
129 @cite{Whether signed integer types are represented using sign and magnitude,
130 two's complement, or one's complement, and whether the extraordinary value
131 is a trap representation or an ordinary value (6.2.6.2).}
133 GCC supports only two's complement integer types, and all bit patterns
137 @cite{The rank of any extended integer type relative to another extended
138 integer type with the same precision (6.3.1.1).}
141 @cite{The result of, or the signal raised by, converting an integer to a
142 signed integer type when the value cannot be represented in an object of
143 that type (6.3.1.3).}
146 @cite{The results of some bitwise operations on signed integers (6.5).}
149 @node Floating point implementation
150 @section Floating point
154 @cite{The accuracy of the floating-point operations and of the library
155 functions in @code{<math.h>} and @code{<complex.h>} that return floating-point
156 results (5.2.4.2.2).}
159 @cite{The rounding behaviors characterized by non-standard values
160 of @code{FLT_ROUNDS} @gol
164 @cite{The evaluation methods characterized by non-standard negative
165 values of @code{FLT_EVAL_METHOD} (5.2.4.2.2).}
168 @cite{The direction of rounding when an integer is converted to a
169 floating-point number that cannot exactly represent the original
173 @cite{The direction of rounding when a floating-point number is
174 converted to a narrower floating-point number (6.3.1.5).}
177 @cite{How the nearest representable value or the larger or smaller
178 representable value immediately adjacent to the nearest representable
179 value is chosen for certain floating constants (6.4.4.2).}
182 @cite{Whether and how floating expressions are contracted when not
183 disallowed by the @code{FP_CONTRACT} pragma (6.5).}
186 @cite{The default state for the @code{FENV_ACCESS} pragma (7.6.1).}
189 @cite{Additional floating-point exceptions, rounding modes, environments,
190 and classifications, and their macro names (7.6, 7.12).}
193 @cite{The default state for the @code{FP_CONTRACT} pragma (7.12.2).}
196 @cite{Whether the ``inexact'' floating-point exception can be raised
197 when the rounded result actually does equal the mathematical result
198 in an IEC 60559 conformant implementation (F.9).}
201 @cite{Whether the ``underflow'' (and ``inexact'') floating-point
202 exception can be raised when a result is tiny but not inexact in an
203 IEC 60559 conformant implementation (F.9).}
207 @node Arrays and pointers implementation
208 @section Arrays and pointers
212 @cite{The result of converting a pointer to an integer or
213 vice versa (6.3.2.3).}
215 A cast from pointer to integer discards most-significant bits if the
216 pointer representation is larger than the integer type,
217 sign-extends@footnote{Future versions of GCC may zero-extend, or use
218 a target-defined @code{ptr_extend} pattern. Do not rely on sign extension.}
219 if the pointer representation is smaller than the integer type, otherwise
220 the bits are unchanged.
221 @c ??? We've always claimed that pointers were unsigned entities.
222 @c Shouldn't we therefore be doing zero-extension? If so, the bug
223 @c is in convert_to_integer, where we call type_for_size and request
224 @c a signed integral type. On the other hand, it might be most useful
225 @c for the target if we extend according to POINTERS_EXTEND_UNSIGNED.
227 A cast from integer to pointer discards most-significant bits if the
228 pointer representation is smaller than the integer type, extends according
229 to the signedness of the integer type if the pointer representation
230 is larger than the integer type, otherwise the bits are unchanged.
232 When casting from pointer to integer and back again, the resulting
233 pointer must reference the same object as the original pointer, otherwise
234 the behavior is undefined. That is, one may not use integer arithmetic to
235 avoid the undefined behavior of pointer arithmetic as proscribed in 6.5.6/8.
238 @cite{The size of the result of subtracting two pointers to elements
239 of the same array (6.5.6).}
243 @node Hints implementation
248 @cite{The extent to which suggestions made by using the @code{register}
249 storage-class specifier are effective (6.7.1).}
251 The @code{register} specifier affects code generation only in these ways:
255 When used as part of the register variable extension, see
256 @ref{Explicit Reg Vars}.
259 When @option{-O0} is in use, the compiler allocates distinct stack
260 memory for all variables that do not have the @code{register}
261 storage-class specifier; if @code{register} is specified, the variable
262 may have a shorter lifespan than the code would indicate and may never
266 On some rare x86 targets, @code{setjmp} doesn't save the registers in
267 all circumstances. In those cases, GCC doesn't allocate any variables
268 in registers unless they are marked @code{register}.
273 @cite{The extent to which suggestions made by using the inline function
274 specifier are effective (6.7.4).}
276 GCC will not inline any functions if the @option{-fno-inline} option is
277 used or if @option{-O0} is used. Otherwise, GCC may still be unable to
278 inline a function for many reasons; the @option{-Winline} option may be
279 used to determine if a function has not been inlined and why not.
283 @node Structures unions enumerations and bit-fields implementation
284 @section Structures, unions, enumerations, and bit-fields
288 @cite{Whether a ``plain'' int bit-field is treated as a @code{signed int}
289 bit-field or as an @code{unsigned int} bit-field (6.7.2, 6.7.2.1).}
292 @cite{Allowable bit-field types other than @code{_Bool}, @code{signed int},
293 and @code{unsigned int} (6.7.2.1).}
296 @cite{Whether a bit-field can straddle a storage-unit boundary (6.7.2.1).}
299 @cite{The order of allocation of bit-fields within a unit (6.7.2.1).}
302 @cite{The alignment of non-bit-field members of structures (6.7.2.1).}
305 @cite{The integer type compatible with each enumerated type (6.7.2.2).}
309 @node Qualifiers implementation
314 @cite{What constitutes an access to an object that has volatile-qualified
319 @node Preprocessing directives implementation
320 @section Preprocessing directives
324 @cite{How sequences in both forms of header names are mapped to headers
325 or external source file names (6.4.7).}
328 @cite{Whether the value of a character constant in a constant expression
329 that controls conditional inclusion matches the value of the same character
330 constant in the execution character set (6.10.1).}
333 @cite{Whether the value of a single-character character constant in a
334 constant expression that controls conditional inclusion may have a
335 negative value (6.10.1).}
338 @cite{The places that are searched for an included @samp{<>} delimited
339 header, and how the places are specified or the header is
340 identified (6.10.2).}
343 @cite{How the named source file is searched for in an included @samp{""}
344 delimited header (6.10.2).}
347 @cite{The method by which preprocessing tokens (possibly resulting from
348 macro expansion) in a @code{#include} directive are combined into a header
352 @cite{The nesting limit for @code{#include} processing (6.10.2).}
354 GCC imposes a limit of 200 nested @code{#include}s.
357 @cite{Whether the @samp{#} operator inserts a @samp{\} character before
358 the @samp{\} character that begins a universal character name in a
359 character constant or string literal (6.10.3.2).}
362 @cite{The behavior on each recognized non-@code{STDC #pragma}
366 @cite{The definitions for @code{__DATE__} and @code{__TIME__} when
367 respectively, the date and time of translation are not available (6.10.8).}
369 If the date and time are not available, @code{__DATE__} expands to
370 @code{@w{"??? ?? ????"}} and @code{__TIME__} expands to
375 @node Library functions implementation
376 @section Library functions
378 The behavior of these points are dependent on the implementation
379 of the C library, and are not defined by GCC itself.
381 @node Architecture implementation
382 @section Architecture
386 @cite{The values or expressions assigned to the macros specified in the
387 headers @code{<float.h>}, @code{<limits.h>}, and @code{<stdint.h>}
388 (5.2.4.2, 7.18.2, 7.18.3).}
391 @cite{The number, order, and encoding of bytes in any object
392 (when not explicitly specified in this International Standard) (6.2.6.1).}
395 @cite{The value of the result of the sizeof operator (6.5.3.4).}
399 @node Locale-specific behavior implementation
400 @section Locale-specific behavior
402 The behavior of these points are dependent on the implementation
403 of the C library, and are not defined by GCC itself.
406 @chapter Extensions to the C Language Family
407 @cindex extensions, C language
408 @cindex C language extensions
411 GNU C provides several language features not found in ISO standard C@.
412 (The @option{-pedantic} option directs GCC to print a warning message if
413 any of these features is used.) To test for the availability of these
414 features in conditional compilation, check for a predefined macro
415 @code{__GNUC__}, which is always defined under GCC@.
417 These extensions are available in C and Objective-C@. Most of them are
418 also available in C++. @xref{C++ Extensions,,Extensions to the
419 C++ Language}, for extensions that apply @emph{only} to C++.
421 Some features that are in ISO C99 but not C89 or C++ are also, as
422 extensions, accepted by GCC in C89 mode and in C++.
425 * Statement Exprs:: Putting statements and declarations inside expressions.
426 * Local Labels:: Labels local to a statement-expression.
427 * Labels as Values:: Getting pointers to labels, and computed gotos.
428 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
429 * Constructing Calls:: Dispatching a call to another function.
430 * Typeof:: @code{typeof}: referring to the type of an expression.
431 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
432 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
433 * Long Long:: Double-word integers---@code{long long int}.
434 * Complex:: Data types for complex numbers.
435 * Hex Floats:: Hexadecimal floating-point constants.
436 * Zero Length:: Zero-length arrays.
437 * Variable Length:: Arrays whose length is computed at run time.
438 * Variadic Macros:: Macros with a variable number of arguments.
439 * Escaped Newlines:: Slightly looser rules for escaped newlines.
440 * Multi-line Strings:: String literals with embedded newlines.
441 * Subscripting:: Any array can be subscripted, even if not an lvalue.
442 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
443 * Initializers:: Non-constant initializers.
444 * Compound Literals:: Compound literals give structures, unions
446 * Designated Inits:: Labeling elements of initializers.
447 * Cast to Union:: Casting to union type from any member of the union.
448 * Case Ranges:: `case 1 ... 9' and such.
449 * Mixed Declarations:: Mixing declarations and code.
450 * Function Attributes:: Declaring that functions have no side effects,
451 or that they can never return.
452 * Attribute Syntax:: Formal syntax for attributes.
453 * Function Prototypes:: Prototype declarations and old-style definitions.
454 * C++ Comments:: C++ comments are recognized.
455 * Dollar Signs:: Dollar sign is allowed in identifiers.
456 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
457 * Variable Attributes:: Specifying attributes of variables.
458 * Type Attributes:: Specifying attributes of types.
459 * Alignment:: Inquiring about the alignment of a type or variable.
460 * Inline:: Defining inline functions (as fast as macros).
461 * Extended Asm:: Assembler instructions with C expressions as operands.
462 (With them you can define ``built-in'' functions.)
463 * Constraints:: Constraints for asm operands
464 * Asm Labels:: Specifying the assembler name to use for a C symbol.
465 * Explicit Reg Vars:: Defining variables residing in specified registers.
466 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
467 * Incomplete Enums:: @code{enum foo;}, with details to follow.
468 * Function Names:: Printable strings which are the name of the current
470 * Return Address:: Getting the return or frame address of a function.
471 * Vector Extensions:: Using vector instructions through built-in functions.
472 * Other Builtins:: Other built-in functions.
473 * Target Builtins:: Built-in functions specific to particular targets.
474 * Pragmas:: Pragmas accepted by GCC.
475 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
476 * Thread-Local:: Per-thread variables.
479 @node Statement Exprs
480 @section Statements and Declarations in Expressions
481 @cindex statements inside expressions
482 @cindex declarations inside expressions
483 @cindex expressions containing statements
484 @cindex macros, statements in expressions
486 @c the above section title wrapped and causes an underfull hbox.. i
487 @c changed it from "within" to "in". --mew 4feb93
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 Statement expressions are not supported fully in G++, and their fate
543 there is unclear. (It is possible that they will become fully supported
544 at some point, or that they will be deprecated, or that the bugs that
545 are present will continue to exist indefinitely.) Presently, statement
546 expressions do not work well as default arguments.
548 In addition, there are semantic issues with statement-expressions in
549 C++. If you try to use statement-expressions instead of inline
550 functions in C++, you may be surprised at the way object destruction is
551 handled. For example:
554 #define foo(a) (@{int b = (a); b + 3; @})
558 does not work the same way as:
561 inline int foo(int a) @{ int b = a; return b + 3; @}
565 In particular, if the expression passed into @code{foo} involves the
566 creation of temporaries, the destructors for those temporaries will be
567 run earlier in the case of the macro than in the case of the function.
569 These considerations mean that it is probably a bad idea to use
570 statement-expressions of this form in header files that are designed to
571 work with C++. (Note that some versions of the GNU C Library contained
572 header files using statement-expression that lead to precisely this
576 @section Locally Declared Labels
578 @cindex macros, local labels
580 Each statement expression is a scope in which @dfn{local labels} can be
581 declared. A local label is simply an identifier; you can jump to it
582 with an ordinary @code{goto} statement, but only from within the
583 statement expression it belongs to.
585 A local label declaration looks like this:
588 __label__ @var{label};
595 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
598 Local label declarations must come at the beginning of the statement
599 expression, right after the @samp{(@{}, before any ordinary
602 The label declaration defines the label @emph{name}, but does not define
603 the label itself. You must do this in the usual way, with
604 @code{@var{label}:}, within the statements of the statement expression.
606 The local label feature is useful because statement expressions are
607 often used in macros. If the macro contains nested loops, a @code{goto}
608 can be useful for breaking out of them. However, an ordinary label
609 whose scope is the whole function cannot be used: if the macro can be
610 expanded several times in one function, the label will be multiply
611 defined in that function. A local label avoids this problem. For
615 #define SEARCH(array, target) \
618 typeof (target) _SEARCH_target = (target); \
619 typeof (*(array)) *_SEARCH_array = (array); \
622 for (i = 0; i < max; i++) \
623 for (j = 0; j < max; j++) \
624 if (_SEARCH_array[i][j] == _SEARCH_target) \
625 @{ value = i; goto found; @} \
632 @node Labels as Values
633 @section Labels as Values
634 @cindex labels as values
635 @cindex computed gotos
636 @cindex goto with computed label
637 @cindex address of a label
639 You can get the address of a label defined in the current function
640 (or a containing function) with the unary operator @samp{&&}. The
641 value has type @code{void *}. This value is a constant and can be used
642 wherever a constant of that type is valid. For example:
650 To use these values, you need to be able to jump to one. This is done
651 with the computed goto statement@footnote{The analogous feature in
652 Fortran is called an assigned goto, but that name seems inappropriate in
653 C, where one can do more than simply store label addresses in label
654 variables.}, @code{goto *@var{exp};}. For example,
661 Any expression of type @code{void *} is allowed.
663 One way of using these constants is in initializing a static array that
664 will serve as a jump table:
667 static void *array[] = @{ &&foo, &&bar, &&hack @};
670 Then you can select a label with indexing, like this:
677 Note that this does not check whether the subscript is in bounds---array
678 indexing in C never does that.
680 Such an array of label values serves a purpose much like that of the
681 @code{switch} statement. The @code{switch} statement is cleaner, so
682 use that rather than an array unless the problem does not fit a
683 @code{switch} statement very well.
685 Another use of label values is in an interpreter for threaded code.
686 The labels within the interpreter function can be stored in the
687 threaded code for super-fast dispatching.
689 You may not use this mechanism to jump to code in a different function.
690 If you do that, totally unpredictable things will happen. The best way to
691 avoid this is to store the label address only in automatic variables and
692 never pass it as an argument.
694 An alternate way to write the above example is
697 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
699 goto *(&&foo + array[i]);
703 This is more friendly to code living in shared libraries, as it reduces
704 the number of dynamic relocations that are needed, and by consequence,
705 allows the data to be read-only.
707 @node Nested Functions
708 @section Nested Functions
709 @cindex nested functions
710 @cindex downward funargs
713 A @dfn{nested function} is a function defined inside another function.
714 (Nested functions are not supported for GNU C++.) The nested function's
715 name is local to the block where it is defined. For example, here we
716 define a nested function named @code{square}, and call it twice:
720 foo (double a, double b)
722 double square (double z) @{ return z * z; @}
724 return square (a) + square (b);
729 The nested function can access all the variables of the containing
730 function that are visible at the point of its definition. This is
731 called @dfn{lexical scoping}. For example, here we show a nested
732 function which uses an inherited variable named @code{offset}:
736 bar (int *array, int offset, int size)
738 int access (int *array, int index)
739 @{ return array[index + offset]; @}
742 for (i = 0; i < size; i++)
743 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
748 Nested function definitions are permitted within functions in the places
749 where variable definitions are allowed; that is, in any block, before
750 the first statement in the block.
752 It is possible to call the nested function from outside the scope of its
753 name by storing its address or passing the address to another function:
756 hack (int *array, int size)
758 void store (int index, int value)
759 @{ array[index] = value; @}
761 intermediate (store, size);
765 Here, the function @code{intermediate} receives the address of
766 @code{store} as an argument. If @code{intermediate} calls @code{store},
767 the arguments given to @code{store} are used to store into @code{array}.
768 But this technique works only so long as the containing function
769 (@code{hack}, in this example) does not exit.
771 If you try to call the nested function through its address after the
772 containing function has exited, all hell will break loose. If you try
773 to call it after a containing scope level has exited, and if it refers
774 to some of the variables that are no longer in scope, you may be lucky,
775 but it's not wise to take the risk. If, however, the nested function
776 does not refer to anything that has gone out of scope, you should be
779 GCC implements taking the address of a nested function using a technique
780 called @dfn{trampolines}. A paper describing them is available as
783 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
785 A nested function can jump to a label inherited from a containing
786 function, provided the label was explicitly declared in the containing
787 function (@pxref{Local Labels}). Such a jump returns instantly to the
788 containing function, exiting the nested function which did the
789 @code{goto} and any intermediate functions as well. Here is an example:
793 bar (int *array, int offset, int size)
796 int access (int *array, int index)
800 return array[index + offset];
804 for (i = 0; i < size; i++)
805 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
809 /* @r{Control comes here from @code{access}
810 if it detects an error.} */
817 A nested function always has internal linkage. Declaring one with
818 @code{extern} is erroneous. If you need to declare the nested function
819 before its definition, use @code{auto} (which is otherwise meaningless
820 for function declarations).
823 bar (int *array, int offset, int size)
826 auto int access (int *, int);
828 int access (int *array, int index)
832 return array[index + offset];
838 @node Constructing Calls
839 @section Constructing Function Calls
840 @cindex constructing calls
841 @cindex forwarding calls
843 Using the built-in functions described below, you can record
844 the arguments a function received, and call another function
845 with the same arguments, without knowing the number or types
848 You can also record the return value of that function call,
849 and later return that value, without knowing what data type
850 the function tried to return (as long as your caller expects
853 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
854 This built-in function returns a pointer to data
855 describing how to perform a call with the same arguments as were passed
856 to the current function.
858 The function saves the arg pointer register, structure value address,
859 and all registers that might be used to pass arguments to a function
860 into a block of memory allocated on the stack. Then it returns the
861 address of that block.
864 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
865 This built-in function invokes @var{function}
866 with a copy of the parameters described by @var{arguments}
869 The value of @var{arguments} should be the value returned by
870 @code{__builtin_apply_args}. The argument @var{size} specifies the size
871 of the stack argument data, in bytes.
873 This function returns a pointer to data describing
874 how to return whatever value was returned by @var{function}. The data
875 is saved in a block of memory allocated on the stack.
877 It is not always simple to compute the proper value for @var{size}. The
878 value is used by @code{__builtin_apply} to compute the amount of data
879 that should be pushed on the stack and copied from the incoming argument
883 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
884 This built-in function returns the value described by @var{result} from
885 the containing function. You should specify, for @var{result}, a value
886 returned by @code{__builtin_apply}.
889 @cindex underscores in variables in macros
890 @cindex @samp{_} in variables in macros
891 @cindex local variables in macros
892 @cindex variables, local, in macros
893 @cindex macros, local variables in
895 The reason for using names that start with underscores for the local
896 variables is to avoid conflicts with variable names that occur within the
897 expressions that are substituted for @code{a} and @code{b}. Eventually we
898 hope to design a new form of declaration syntax that allows you to declare
899 variables whose scopes start only after their initializers; this will be a
900 more reliable way to prevent such conflicts.
903 @section Referring to a Type with @code{typeof}
906 @cindex macros, types of arguments
908 Another way to refer to the type of an expression is with @code{typeof}.
909 The syntax of using of this keyword looks like @code{sizeof}, but the
910 construct acts semantically like a type name defined with @code{typedef}.
912 There are two ways of writing the argument to @code{typeof}: with an
913 expression or with a type. Here is an example with an expression:
920 This assumes that @code{x} is an array of pointers to functions;
921 the type described is that of the values of the functions.
923 Here is an example with a typename as the argument:
930 Here the type described is that of pointers to @code{int}.
932 If you are writing a header file that must work when included in ISO C
933 programs, write @code{__typeof__} instead of @code{typeof}.
934 @xref{Alternate Keywords}.
936 A @code{typeof}-construct can be used anywhere a typedef name could be
937 used. For example, you can use it in a declaration, in a cast, or inside
938 of @code{sizeof} or @code{typeof}.
940 @code{typeof} is often useful in conjunction with the
941 statements-within-expressions feature. Here is how the two together can
942 be used to define a safe ``maximum'' macro that operates on any
943 arithmetic type and evaluates each of its arguments exactly once:
947 (@{ typeof (a) _a = (a); \
948 typeof (b) _b = (b); \
949 _a > _b ? _a : _b; @})
953 Some more examples of the use of @code{typeof}:
957 This declares @code{y} with the type of what @code{x} points to.
964 This declares @code{y} as an array of such values.
971 This declares @code{y} as an array of pointers to characters:
974 typeof (typeof (char *)[4]) y;
978 It is equivalent to the following traditional C declaration:
984 To see the meaning of the declaration using @code{typeof}, and why it
985 might be a useful way to write, let's rewrite it with these macros:
988 #define pointer(T) typeof(T *)
989 #define array(T, N) typeof(T [N])
993 Now the declaration can be rewritten this way:
996 array (pointer (char), 4) y;
1000 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1001 pointers to @code{char}.
1004 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1005 a more limited extension which permitted one to write
1008 typedef @var{T} = @var{expr};
1012 with the effect of declaring @var{T} to have the type of the expression
1013 @var{expr}. This extension does not work with GCC 3 (versions between
1014 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
1015 relies on it should be rewritten to use @code{typeof}:
1018 typedef typeof(@var{expr}) @var{T};
1022 This will work with all versions of GCC@.
1025 @section Generalized Lvalues
1026 @cindex compound expressions as lvalues
1027 @cindex expressions, compound, as lvalues
1028 @cindex conditional expressions as lvalues
1029 @cindex expressions, conditional, as lvalues
1030 @cindex casts as lvalues
1031 @cindex generalized lvalues
1032 @cindex lvalues, generalized
1033 @cindex extensions, @code{?:}
1034 @cindex @code{?:} extensions
1035 Compound expressions, conditional expressions and casts are allowed as
1036 lvalues provided their operands are lvalues. This means that you can take
1037 their addresses or store values into them.
1039 Standard C++ allows compound expressions and conditional expressions as
1040 lvalues, and permits casts to reference type, so use of this extension
1041 is deprecated for C++ code.
1043 For example, a compound expression can be assigned, provided the last
1044 expression in the sequence is an lvalue. These two expressions are
1052 Similarly, the address of the compound expression can be taken. These two
1053 expressions are equivalent:
1060 A conditional expression is a valid lvalue if its type is not void and the
1061 true and false branches are both valid lvalues. For example, these two
1062 expressions are equivalent:
1066 (a ? b = 5 : (c = 5))
1069 A cast is a valid lvalue if its operand is an lvalue. A simple
1070 assignment whose left-hand side is a cast works by converting the
1071 right-hand side first to the specified type, then to the type of the
1072 inner left-hand side expression. After this is stored, the value is
1073 converted back to the specified type to become the value of the
1074 assignment. Thus, if @code{a} has type @code{char *}, the following two
1075 expressions are equivalent:
1079 (int)(a = (char *)(int)5)
1082 An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
1083 performs the arithmetic using the type resulting from the cast, and then
1084 continues as in the previous case. Therefore, these two expressions are
1089 (int)(a = (char *)(int) ((int)a + 5))
1092 You cannot take the address of an lvalue cast, because the use of its
1093 address would not work out coherently. Suppose that @code{&(int)f} were
1094 permitted, where @code{f} has type @code{float}. Then the following
1095 statement would try to store an integer bit-pattern where a floating
1096 point number belongs:
1102 This is quite different from what @code{(int)f = 1} would do---that
1103 would convert 1 to floating point and store it. Rather than cause this
1104 inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
1106 If you really do want an @code{int *} pointer with the address of
1107 @code{f}, you can simply write @code{(int *)&f}.
1110 @section Conditionals with Omitted Operands
1111 @cindex conditional expressions, extensions
1112 @cindex omitted middle-operands
1113 @cindex middle-operands, omitted
1114 @cindex extensions, @code{?:}
1115 @cindex @code{?:} extensions
1117 The middle operand in a conditional expression may be omitted. Then
1118 if the first operand is nonzero, its value is the value of the conditional
1121 Therefore, the expression
1128 has the value of @code{x} if that is nonzero; otherwise, the value of
1131 This example is perfectly equivalent to
1137 @cindex side effect in ?:
1138 @cindex ?: side effect
1140 In this simple case, the ability to omit the middle operand is not
1141 especially useful. When it becomes useful is when the first operand does,
1142 or may (if it is a macro argument), contain a side effect. Then repeating
1143 the operand in the middle would perform the side effect twice. Omitting
1144 the middle operand uses the value already computed without the undesirable
1145 effects of recomputing it.
1148 @section Double-Word Integers
1149 @cindex @code{long long} data types
1150 @cindex double-word arithmetic
1151 @cindex multiprecision arithmetic
1152 @cindex @code{LL} integer suffix
1153 @cindex @code{ULL} integer suffix
1155 ISO C99 supports data types for integers that are at least 64 bits wide,
1156 and as an extension GCC supports them in C89 mode and in C++.
1157 Simply write @code{long long int} for a signed integer, or
1158 @code{unsigned long long int} for an unsigned integer. To make an
1159 integer constant of type @code{long long int}, add the suffix @samp{LL}
1160 to the integer. To make an integer constant of type @code{unsigned long
1161 long int}, add the suffix @samp{ULL} to the integer.
1163 You can use these types in arithmetic like any other integer types.
1164 Addition, subtraction, and bitwise boolean operations on these types
1165 are open-coded on all types of machines. Multiplication is open-coded
1166 if the machine supports fullword-to-doubleword a widening multiply
1167 instruction. Division and shifts are open-coded only on machines that
1168 provide special support. The operations that are not open-coded use
1169 special library routines that come with GCC@.
1171 There may be pitfalls when you use @code{long long} types for function
1172 arguments, unless you declare function prototypes. If a function
1173 expects type @code{int} for its argument, and you pass a value of type
1174 @code{long long int}, confusion will result because the caller and the
1175 subroutine will disagree about the number of bytes for the argument.
1176 Likewise, if the function expects @code{long long int} and you pass
1177 @code{int}. The best way to avoid such problems is to use prototypes.
1180 @section Complex Numbers
1181 @cindex complex numbers
1182 @cindex @code{_Complex} keyword
1183 @cindex @code{__complex__} keyword
1185 ISO C99 supports complex floating data types, and as an extension GCC
1186 supports them in C89 mode and in C++, and supports complex integer data
1187 types which are not part of ISO C99. You can declare complex types
1188 using the keyword @code{_Complex}. As an extension, the older GNU
1189 keyword @code{__complex__} is also supported.
1191 For example, @samp{_Complex double x;} declares @code{x} as a
1192 variable whose real part and imaginary part are both of type
1193 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1194 have real and imaginary parts of type @code{short int}; this is not
1195 likely to be useful, but it shows that the set of complex types is
1198 To write a constant with a complex data type, use the suffix @samp{i} or
1199 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1200 has type @code{_Complex float} and @code{3i} has type
1201 @code{_Complex int}. Such a constant always has a pure imaginary
1202 value, but you can form any complex value you like by adding one to a
1203 real constant. This is a GNU extension; if you have an ISO C99
1204 conforming C library (such as GNU libc), and want to construct complex
1205 constants of floating type, you should include @code{<complex.h>} and
1206 use the macros @code{I} or @code{_Complex_I} instead.
1208 @cindex @code{__real__} keyword
1209 @cindex @code{__imag__} keyword
1210 To extract the real part of a complex-valued expression @var{exp}, write
1211 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1212 extract the imaginary part. This is a GNU extension; for values of
1213 floating type, you should use the ISO C99 functions @code{crealf},
1214 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1215 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1216 built-in functions by GCC@.
1218 @cindex complex conjugation
1219 The operator @samp{~} performs complex conjugation when used on a value
1220 with a complex type. This is a GNU extension; for values of
1221 floating type, you should use the ISO C99 functions @code{conjf},
1222 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1223 provided as built-in functions by GCC@.
1225 GCC can allocate complex automatic variables in a noncontiguous
1226 fashion; it's even possible for the real part to be in a register while
1227 the imaginary part is on the stack (or vice-versa). None of the
1228 supported debugging info formats has a way to represent noncontiguous
1229 allocation like this, so GCC describes a noncontiguous complex
1230 variable as if it were two separate variables of noncomplex type.
1231 If the variable's actual name is @code{foo}, the two fictitious
1232 variables are named @code{foo$real} and @code{foo$imag}. You can
1233 examine and set these two fictitious variables with your debugger.
1235 A future version of GDB will know how to recognize such pairs and treat
1236 them as a single variable with a complex type.
1242 ISO C99 supports floating-point numbers written not only in the usual
1243 decimal notation, such as @code{1.55e1}, but also numbers such as
1244 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1245 supports this in C89 mode (except in some cases when strictly
1246 conforming) and in C++. In that format the
1247 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1248 mandatory. The exponent is a decimal number that indicates the power of
1249 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1256 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1257 is the same as @code{1.55e1}.
1259 Unlike for floating-point numbers in the decimal notation the exponent
1260 is always required in the hexadecimal notation. Otherwise the compiler
1261 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1262 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1263 extension for floating-point constants of type @code{float}.
1266 @section Arrays of Length Zero
1267 @cindex arrays of length zero
1268 @cindex zero-length arrays
1269 @cindex length-zero arrays
1270 @cindex flexible array members
1272 Zero-length arrays are allowed in GNU C@. They are very useful as the
1273 last element of a structure which is really a header for a variable-length
1282 struct line *thisline = (struct line *)
1283 malloc (sizeof (struct line) + this_length);
1284 thisline->length = this_length;
1287 In ISO C90, you would have to give @code{contents} a length of 1, which
1288 means either you waste space or complicate the argument to @code{malloc}.
1290 In ISO C99, you would use a @dfn{flexible array member}, which is
1291 slightly different in syntax and semantics:
1295 Flexible array members are written as @code{contents[]} without
1299 Flexible array members have incomplete type, and so the @code{sizeof}
1300 operator may not be applied. As a quirk of the original implementation
1301 of zero-length arrays, @code{sizeof} evaluates to zero.
1304 Flexible array members may only appear as the last member of a
1305 @code{struct} that is otherwise non-empty.
1308 A structure containing a flexible array member, or a union containing
1309 such a structure (possibly recursively), may not be a member of a
1310 structure or an element of an array. (However, these uses are
1311 permitted by GCC as extensions.)
1314 GCC versions before 3.0 allowed zero-length arrays to be statically
1315 initialized, as if they were flexible arrays. In addition to those
1316 cases that were useful, it also allowed initializations in situations
1317 that would corrupt later data. Non-empty initialization of zero-length
1318 arrays is now treated like any case where there are more initializer
1319 elements than the array holds, in that a suitable warning about "excess
1320 elements in array" is given, and the excess elements (all of them, in
1321 this case) are ignored.
1323 Instead GCC allows static initialization of flexible array members.
1324 This is equivalent to defining a new structure containing the original
1325 structure followed by an array of sufficient size to contain the data.
1326 I.e.@: in the following, @code{f1} is constructed as if it were declared
1332 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1335 struct f1 f1; int data[3];
1336 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1340 The convenience of this extension is that @code{f1} has the desired
1341 type, eliminating the need to consistently refer to @code{f2.f1}.
1343 This has symmetry with normal static arrays, in that an array of
1344 unknown size is also written with @code{[]}.
1346 Of course, this extension only makes sense if the extra data comes at
1347 the end of a top-level object, as otherwise we would be overwriting
1348 data at subsequent offsets. To avoid undue complication and confusion
1349 with initialization of deeply nested arrays, we simply disallow any
1350 non-empty initialization except when the structure is the top-level
1351 object. For example:
1354 struct foo @{ int x; int y[]; @};
1355 struct bar @{ struct foo z; @};
1357 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1358 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1359 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1360 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1363 @node Variable Length
1364 @section Arrays of Variable Length
1365 @cindex variable-length arrays
1366 @cindex arrays of variable length
1369 Variable-length automatic arrays are allowed in ISO C99, and as an
1370 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1371 implementation of variable-length arrays does not yet conform in detail
1372 to the ISO C99 standard.) These arrays are
1373 declared like any other automatic arrays, but with a length that is not
1374 a constant expression. The storage is allocated at the point of
1375 declaration and deallocated when the brace-level is exited. For
1380 concat_fopen (char *s1, char *s2, char *mode)
1382 char str[strlen (s1) + strlen (s2) + 1];
1385 return fopen (str, mode);
1389 @cindex scope of a variable length array
1390 @cindex variable-length array scope
1391 @cindex deallocating variable length arrays
1392 Jumping or breaking out of the scope of the array name deallocates the
1393 storage. Jumping into the scope is not allowed; you get an error
1396 @cindex @code{alloca} vs variable-length arrays
1397 You can use the function @code{alloca} to get an effect much like
1398 variable-length arrays. The function @code{alloca} is available in
1399 many other C implementations (but not in all). On the other hand,
1400 variable-length arrays are more elegant.
1402 There are other differences between these two methods. Space allocated
1403 with @code{alloca} exists until the containing @emph{function} returns.
1404 The space for a variable-length array is deallocated as soon as the array
1405 name's scope ends. (If you use both variable-length arrays and
1406 @code{alloca} in the same function, deallocation of a variable-length array
1407 will also deallocate anything more recently allocated with @code{alloca}.)
1409 You can also use variable-length arrays as arguments to functions:
1413 tester (int len, char data[len][len])
1419 The length of an array is computed once when the storage is allocated
1420 and is remembered for the scope of the array in case you access it with
1423 If you want to pass the array first and the length afterward, you can
1424 use a forward declaration in the parameter list---another GNU extension.
1428 tester (int len; char data[len][len], int len)
1434 @cindex parameter forward declaration
1435 The @samp{int len} before the semicolon is a @dfn{parameter forward
1436 declaration}, and it serves the purpose of making the name @code{len}
1437 known when the declaration of @code{data} is parsed.
1439 You can write any number of such parameter forward declarations in the
1440 parameter list. They can be separated by commas or semicolons, but the
1441 last one must end with a semicolon, which is followed by the ``real''
1442 parameter declarations. Each forward declaration must match a ``real''
1443 declaration in parameter name and data type. ISO C99 does not support
1444 parameter forward declarations.
1446 @node Variadic Macros
1447 @section Macros with a Variable Number of Arguments.
1448 @cindex variable number of arguments
1449 @cindex macro with variable arguments
1450 @cindex rest argument (in macro)
1451 @cindex variadic macros
1453 In the ISO C standard of 1999, a macro can be declared to accept a
1454 variable number of arguments much as a function can. The syntax for
1455 defining the macro is similar to that of a function. Here is an
1459 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1462 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1463 such a macro, it represents the zero or more tokens until the closing
1464 parenthesis that ends the invocation, including any commas. This set of
1465 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1466 wherever it appears. See the CPP manual for more information.
1468 GCC has long supported variadic macros, and used a different syntax that
1469 allowed you to give a name to the variable arguments just like any other
1470 argument. Here is an example:
1473 #define debug(format, args...) fprintf (stderr, format, args)
1476 This is in all ways equivalent to the ISO C example above, but arguably
1477 more readable and descriptive.
1479 GNU CPP has two further variadic macro extensions, and permits them to
1480 be used with either of the above forms of macro definition.
1482 In standard C, you are not allowed to leave the variable argument out
1483 entirely; but you are allowed to pass an empty argument. For example,
1484 this invocation is invalid in ISO C, because there is no comma after
1491 GNU CPP permits you to completely omit the variable arguments in this
1492 way. In the above examples, the compiler would complain, though since
1493 the expansion of the macro still has the extra comma after the format
1496 To help solve this problem, CPP behaves specially for variable arguments
1497 used with the token paste operator, @samp{##}. If instead you write
1500 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1503 and if the variable arguments are omitted or empty, the @samp{##}
1504 operator causes the preprocessor to remove the comma before it. If you
1505 do provide some variable arguments in your macro invocation, GNU CPP
1506 does not complain about the paste operation and instead places the
1507 variable arguments after the comma. Just like any other pasted macro
1508 argument, these arguments are not macro expanded.
1510 @node Escaped Newlines
1511 @section Slightly Looser Rules for Escaped Newlines
1512 @cindex escaped newlines
1513 @cindex newlines (escaped)
1515 Recently, the preprocessor has relaxed its treatment of escaped
1516 newlines. Previously, the newline had to immediately follow a
1517 backslash. The current implementation allows whitespace in the form of
1518 spaces, horizontal and vertical tabs, and form feeds between the
1519 backslash and the subsequent newline. The preprocessor issues a
1520 warning, but treats it as a valid escaped newline and combines the two
1521 lines to form a single logical line. This works within comments and
1522 tokens, including multi-line strings, as well as between tokens.
1523 Comments are @emph{not} treated as whitespace for the purposes of this
1524 relaxation, since they have not yet been replaced with spaces.
1526 @node Multi-line Strings
1527 @section String Literals with Embedded Newlines
1528 @cindex multi-line string literals
1530 As an extension, GNU CPP permits string literals to cross multiple lines
1531 without escaping the embedded newlines. Each embedded newline is
1532 replaced with a single @samp{\n} character in the resulting string
1533 literal, regardless of what form the newline took originally.
1535 CPP currently allows such strings in directives as well (other than the
1536 @samp{#include} family). This is deprecated and will eventually be
1540 @section Non-Lvalue Arrays May Have Subscripts
1541 @cindex subscripting
1542 @cindex arrays, non-lvalue
1544 @cindex subscripting and function values
1545 In ISO C99, arrays that are not lvalues still decay to pointers, and
1546 may be subscripted, although they may not be modified or used after
1547 the next sequence point and the unary @samp{&} operator may not be
1548 applied to them. As an extension, GCC allows such arrays to be
1549 subscripted in C89 mode, though otherwise they do not decay to
1550 pointers outside C99 mode. For example,
1551 this is valid in GNU C though not valid in C89:
1555 struct foo @{int a[4];@};
1561 return f().a[index];
1567 @section Arithmetic on @code{void}- and Function-Pointers
1568 @cindex void pointers, arithmetic
1569 @cindex void, size of pointer to
1570 @cindex function pointers, arithmetic
1571 @cindex function, size of pointer to
1573 In GNU C, addition and subtraction operations are supported on pointers to
1574 @code{void} and on pointers to functions. This is done by treating the
1575 size of a @code{void} or of a function as 1.
1577 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1578 and on function types, and returns 1.
1580 @opindex Wpointer-arith
1581 The option @option{-Wpointer-arith} requests a warning if these extensions
1585 @section Non-Constant Initializers
1586 @cindex initializers, non-constant
1587 @cindex non-constant initializers
1589 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1590 automatic variable are not required to be constant expressions in GNU C@.
1591 Here is an example of an initializer with run-time varying elements:
1594 foo (float f, float g)
1596 float beat_freqs[2] = @{ f-g, f+g @};
1601 @node Compound Literals
1602 @section Compound Literals
1603 @cindex constructor expressions
1604 @cindex initializations in expressions
1605 @cindex structures, constructor expression
1606 @cindex expressions, constructor
1607 @cindex compound literals
1608 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1610 ISO C99 supports compound literals. A compound literal looks like
1611 a cast containing an initializer. Its value is an object of the
1612 type specified in the cast, containing the elements specified in
1613 the initializer; it is an lvalue. As an extension, GCC supports
1614 compound literals in C89 mode and in C++.
1616 Usually, the specified type is a structure. Assume that
1617 @code{struct foo} and @code{structure} are declared as shown:
1620 struct foo @{int a; char b[2];@} structure;
1624 Here is an example of constructing a @code{struct foo} with a compound literal:
1627 structure = ((struct foo) @{x + y, 'a', 0@});
1631 This is equivalent to writing the following:
1635 struct foo temp = @{x + y, 'a', 0@};
1640 You can also construct an array. If all the elements of the compound literal
1641 are (made up of) simple constant expressions, suitable for use in
1642 initializers of objects of static storage duration, then the compound
1643 literal can be coerced to a pointer to its first element and used in
1644 such an initializer, as shown here:
1647 char **foo = (char *[]) @{ "x", "y", "z" @};
1650 Compound literals for scalar types and union types are is
1651 also allowed, but then the compound literal is equivalent
1654 As a GNU extension, GCC allows initialization of objects with static storage
1655 duration by compound literals (which is not possible in ISO C99, because
1656 the initializer is not a constant).
1657 It is handled as if the object was initialized only with the bracket
1658 enclosed list if compound literal's and object types match.
1659 The initializer list of the compound literal must be constant.
1660 If the object being initialized has array type of unknown size, the size is
1661 determined by compound literal size.
1664 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1665 static int y[] = (int []) @{1, 2, 3@};
1666 static int z[] = (int [3]) @{1@};
1670 The above lines are equivalent to the following:
1672 static struct foo x = @{1, 'a', 'b'@};
1673 static int y[] = @{1, 2, 3@};
1674 static int z[] = @{1, 0, 0@};
1677 @node Designated Inits
1678 @section Designated Initializers
1679 @cindex initializers with labeled elements
1680 @cindex labeled elements in initializers
1681 @cindex case labels in initializers
1682 @cindex designated initializers
1684 Standard C89 requires the elements of an initializer to appear in a fixed
1685 order, the same as the order of the elements in the array or structure
1688 In ISO C99 you can give the elements in any order, specifying the array
1689 indices or structure field names they apply to, and GNU C allows this as
1690 an extension in C89 mode as well. This extension is not
1691 implemented in GNU C++.
1693 To specify an array index, write
1694 @samp{[@var{index}] =} before the element value. For example,
1697 int a[6] = @{ [4] = 29, [2] = 15 @};
1704 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1708 The index values must be constant expressions, even if the array being
1709 initialized is automatic.
1711 An alternative syntax for this which has been obsolete since GCC 2.5 but
1712 GCC still accepts is to write @samp{[@var{index}]} before the element
1713 value, with no @samp{=}.
1715 To initialize a range of elements to the same value, write
1716 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1717 extension. For example,
1720 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1724 If the value in it has side-effects, the side-effects will happen only once,
1725 not for each initialized field by the range initializer.
1728 Note that the length of the array is the highest value specified
1731 In a structure initializer, specify the name of a field to initialize
1732 with @samp{.@var{fieldname} =} before the element value. For example,
1733 given the following structure,
1736 struct point @{ int x, y; @};
1740 the following initialization
1743 struct point p = @{ .y = yvalue, .x = xvalue @};
1750 struct point p = @{ xvalue, yvalue @};
1753 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1754 @samp{@var{fieldname}:}, as shown here:
1757 struct point p = @{ y: yvalue, x: xvalue @};
1761 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1762 @dfn{designator}. You can also use a designator (or the obsolete colon
1763 syntax) when initializing a union, to specify which element of the union
1764 should be used. For example,
1767 union foo @{ int i; double d; @};
1769 union foo f = @{ .d = 4 @};
1773 will convert 4 to a @code{double} to store it in the union using
1774 the second element. By contrast, casting 4 to type @code{union foo}
1775 would store it into the union as the integer @code{i}, since it is
1776 an integer. (@xref{Cast to Union}.)
1778 You can combine this technique of naming elements with ordinary C
1779 initialization of successive elements. Each initializer element that
1780 does not have a designator applies to the next consecutive element of the
1781 array or structure. For example,
1784 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1791 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1794 Labeling the elements of an array initializer is especially useful
1795 when the indices are characters or belong to an @code{enum} type.
1800 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1801 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1804 @cindex designator lists
1805 You can also write a series of @samp{.@var{fieldname}} and
1806 @samp{[@var{index}]} designators before an @samp{=} to specify a
1807 nested subobject to initialize; the list is taken relative to the
1808 subobject corresponding to the closest surrounding brace pair. For
1809 example, with the @samp{struct point} declaration above:
1812 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1816 If the same field is initialized multiple times, it will have value from
1817 the last initialization. If any such overridden initialization has
1818 side-effect, it is unspecified whether the side-effect happens or not.
1819 Currently, gcc will discard them and issue a warning.
1822 @section Case Ranges
1824 @cindex ranges in case statements
1826 You can specify a range of consecutive values in a single @code{case} label,
1830 case @var{low} ... @var{high}:
1834 This has the same effect as the proper number of individual @code{case}
1835 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1837 This feature is especially useful for ranges of ASCII character codes:
1843 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1844 it may be parsed wrong when you use it with integer values. For example,
1859 @section Cast to a Union Type
1860 @cindex cast to a union
1861 @cindex union, casting to a
1863 A cast to union type is similar to other casts, except that the type
1864 specified is a union type. You can specify the type either with
1865 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1866 a constructor though, not a cast, and hence does not yield an lvalue like
1867 normal casts. (@xref{Compound Literals}.)
1869 The types that may be cast to the union type are those of the members
1870 of the union. Thus, given the following union and variables:
1873 union foo @{ int i; double d; @};
1879 both @code{x} and @code{y} can be cast to type @code{union foo}.
1881 Using the cast as the right-hand side of an assignment to a variable of
1882 union type is equivalent to storing in a member of the union:
1887 u = (union foo) x @equiv{} u.i = x
1888 u = (union foo) y @equiv{} u.d = y
1891 You can also use the union cast as a function argument:
1894 void hack (union foo);
1896 hack ((union foo) x);
1899 @node Mixed Declarations
1900 @section Mixed Declarations and Code
1901 @cindex mixed declarations and code
1902 @cindex declarations, mixed with code
1903 @cindex code, mixed with declarations
1905 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1906 within compound statements. As an extension, GCC also allows this in
1907 C89 mode. For example, you could do:
1916 Each identifier is visible from where it is declared until the end of
1917 the enclosing block.
1919 @node Function Attributes
1920 @section Declaring Attributes of Functions
1921 @cindex function attributes
1922 @cindex declaring attributes of functions
1923 @cindex functions that never return
1924 @cindex functions that have no side effects
1925 @cindex functions in arbitrary sections
1926 @cindex functions that behave like malloc
1927 @cindex @code{volatile} applied to function
1928 @cindex @code{const} applied to function
1929 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1930 @cindex functions with non-null pointer arguments
1931 @cindex functions that are passed arguments in registers on the 386
1932 @cindex functions that pop the argument stack on the 386
1933 @cindex functions that do not pop the argument stack on the 386
1935 In GNU C, you declare certain things about functions called in your program
1936 which help the compiler optimize function calls and check your code more
1939 The keyword @code{__attribute__} allows you to specify special
1940 attributes when making a declaration. This keyword is followed by an
1941 attribute specification inside double parentheses. The following
1942 attributes are currently defined for functions on all targets:
1943 @code{noreturn}, @code{noinline}, @code{always_inline},
1944 @code{pure}, @code{const}, @code{nothrow},
1945 @code{format}, @code{format_arg}, @code{no_instrument_function},
1946 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1947 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1948 @code{alias}, and @code{nonnull}. Several other attributes are defined
1949 for functions on particular target systems. Other attributes, including
1950 @code{section} are supported for variables declarations
1951 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1953 You may also specify attributes with @samp{__} preceding and following
1954 each keyword. This allows you to use them in header files without
1955 being concerned about a possible macro of the same name. For example,
1956 you may use @code{__noreturn__} instead of @code{noreturn}.
1958 @xref{Attribute Syntax}, for details of the exact syntax for using
1962 @cindex @code{noreturn} function attribute
1964 A few standard library functions, such as @code{abort} and @code{exit},
1965 cannot return. GCC knows this automatically. Some programs define
1966 their own functions that never return. You can declare them
1967 @code{noreturn} to tell the compiler this fact. For example,
1971 void fatal () __attribute__ ((noreturn));
1974 fatal (/* @r{@dots{}} */)
1976 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
1982 The @code{noreturn} keyword tells the compiler to assume that
1983 @code{fatal} cannot return. It can then optimize without regard to what
1984 would happen if @code{fatal} ever did return. This makes slightly
1985 better code. More importantly, it helps avoid spurious warnings of
1986 uninitialized variables.
1988 Do not assume that registers saved by the calling function are
1989 restored before calling the @code{noreturn} function.
1991 It does not make sense for a @code{noreturn} function to have a return
1992 type other than @code{void}.
1994 The attribute @code{noreturn} is not implemented in GCC versions
1995 earlier than 2.5. An alternative way to declare that a function does
1996 not return, which works in the current version and in some older
1997 versions, is as follows:
2000 typedef void voidfn ();
2002 volatile voidfn fatal;
2005 @cindex @code{noinline} function attribute
2007 This function attribute prevents a function from being considered for
2010 @cindex @code{always_inline} function attribute
2012 Generally, functions are not inlined unless optimization is specified.
2013 For functions declared inline, this attribute inlines the function even
2014 if no optimization level was specified.
2016 @cindex @code{pure} function attribute
2018 Many functions have no effects except the return value and their
2019 return value depends only on the parameters and/or global variables.
2020 Such a function can be subject
2021 to common subexpression elimination and loop optimization just as an
2022 arithmetic operator would be. These functions should be declared
2023 with the attribute @code{pure}. For example,
2026 int square (int) __attribute__ ((pure));
2030 says that the hypothetical function @code{square} is safe to call
2031 fewer times than the program says.
2033 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2034 Interesting non-pure functions are functions with infinite loops or those
2035 depending on volatile memory or other system resource, that may change between
2036 two consecutive calls (such as @code{feof} in a multithreading environment).
2038 The attribute @code{pure} is not implemented in GCC versions earlier
2040 @cindex @code{const} function attribute
2042 Many functions do not examine any values except their arguments, and
2043 have no effects except the return value. Basically this is just slightly
2044 more strict class than the @code{pure} attribute above, since function is not
2045 allowed to read global memory.
2047 @cindex pointer arguments
2048 Note that a function that has pointer arguments and examines the data
2049 pointed to must @emph{not} be declared @code{const}. Likewise, a
2050 function that calls a non-@code{const} function usually must not be
2051 @code{const}. It does not make sense for a @code{const} function to
2054 The attribute @code{const} is not implemented in GCC versions earlier
2055 than 2.5. An alternative way to declare that a function has no side
2056 effects, which works in the current version and in some older versions,
2060 typedef int intfn ();
2062 extern const intfn square;
2065 This approach does not work in GNU C++ from 2.6.0 on, since the language
2066 specifies that the @samp{const} must be attached to the return value.
2068 @cindex @code{nothrow} function attribute
2070 The @code{nothrow} attribute is used to inform the compiler that a
2071 function cannot throw an exception. For example, most functions in
2072 the standard C library can be guaranteed not to throw an exception
2073 with the notable exceptions of @code{qsort} and @code{bsearch} that
2074 take function pointer arguments. The @code{nothrow} attribute is not
2075 implemented in GCC versions earlier than 3.2.
2077 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2078 @cindex @code{format} function attribute
2080 The @code{format} attribute specifies that a function takes @code{printf},
2081 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2082 should be type-checked against a format string. For example, the
2087 my_printf (void *my_object, const char *my_format, ...)
2088 __attribute__ ((format (printf, 2, 3)));
2092 causes the compiler to check the arguments in calls to @code{my_printf}
2093 for consistency with the @code{printf} style format string argument
2096 The parameter @var{archetype} determines how the format string is
2097 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2098 or @code{strfmon}. (You can also use @code{__printf__},
2099 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2100 parameter @var{string-index} specifies which argument is the format
2101 string argument (starting from 1), while @var{first-to-check} is the
2102 number of the first argument to check against the format string. For
2103 functions where the arguments are not available to be checked (such as
2104 @code{vprintf}), specify the third parameter as zero. In this case the
2105 compiler only checks the format string for consistency. For
2106 @code{strftime} formats, the third parameter is required to be zero.
2108 In the example above, the format string (@code{my_format}) is the second
2109 argument of the function @code{my_print}, and the arguments to check
2110 start with the third argument, so the correct parameters for the format
2111 attribute are 2 and 3.
2113 @opindex ffreestanding
2114 The @code{format} attribute allows you to identify your own functions
2115 which take format strings as arguments, so that GCC can check the
2116 calls to these functions for errors. The compiler always (unless
2117 @option{-ffreestanding} is used) checks formats
2118 for the standard library functions @code{printf}, @code{fprintf},
2119 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2120 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2121 warnings are requested (using @option{-Wformat}), so there is no need to
2122 modify the header file @file{stdio.h}. In C99 mode, the functions
2123 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2124 @code{vsscanf} are also checked. Except in strictly conforming C
2125 standard modes, the X/Open function @code{strfmon} is also checked as
2126 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2127 @xref{C Dialect Options,,Options Controlling C Dialect}.
2129 @item format_arg (@var{string-index})
2130 @cindex @code{format_arg} function attribute
2131 @opindex Wformat-nonliteral
2132 The @code{format_arg} attribute specifies that a function takes a format
2133 string for a @code{printf}, @code{scanf}, @code{strftime} or
2134 @code{strfmon} style function and modifies it (for example, to translate
2135 it into another language), so the result can be passed to a
2136 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2137 function (with the remaining arguments to the format function the same
2138 as they would have been for the unmodified string). For example, the
2143 my_dgettext (char *my_domain, const char *my_format)
2144 __attribute__ ((format_arg (2)));
2148 causes the compiler to check the arguments in calls to a @code{printf},
2149 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2150 format string argument is a call to the @code{my_dgettext} function, for
2151 consistency with the format string argument @code{my_format}. If the
2152 @code{format_arg} attribute had not been specified, all the compiler
2153 could tell in such calls to format functions would be that the format
2154 string argument is not constant; this would generate a warning when
2155 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2156 without the attribute.
2158 The parameter @var{string-index} specifies which argument is the format
2159 string argument (starting from 1).
2161 The @code{format-arg} attribute allows you to identify your own
2162 functions which modify format strings, so that GCC can check the
2163 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2164 type function whose operands are a call to one of your own function.
2165 The compiler always treats @code{gettext}, @code{dgettext}, and
2166 @code{dcgettext} in this manner except when strict ISO C support is
2167 requested by @option{-ansi} or an appropriate @option{-std} option, or
2168 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2169 Controlling C Dialect}.
2171 @item nonnull (@var{arg-index}, @dots{})
2172 @cindex @code{nonnull} function attribute
2173 The @code{nonnull} attribute specifies that some function parameters should
2174 be non-null pointers. For instance, the declaration:
2178 my_memcpy (void *dest, const void *src, size_t len)
2179 __attribute__((nonnull (1, 2)));
2183 causes the compiler to check that, in calls to @code{my_memcpy},
2184 arguments @var{dest} and @var{src} are non-null. If the compiler
2185 determines that a null pointer is passed in an argument slot marked
2186 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2187 is issued. The compiler may also choose to make optimizations based
2188 on the knowledge that certain function arguments will not be null.
2190 If no argument index list is given to the @code{nonnull} attribute,
2191 all pointer arguments are marked as non-null. To illustrate, the
2192 following declaration is equivalent to the previous example:
2196 my_memcpy (void *dest, const void *src, size_t len)
2197 __attribute__((nonnull));
2200 @item no_instrument_function
2201 @cindex @code{no_instrument_function} function attribute
2202 @opindex finstrument-functions
2203 If @option{-finstrument-functions} is given, profiling function calls will
2204 be generated at entry and exit of most user-compiled functions.
2205 Functions with this attribute will not be so instrumented.
2207 @item section ("@var{section-name}")
2208 @cindex @code{section} function attribute
2209 Normally, the compiler places the code it generates in the @code{text} section.
2210 Sometimes, however, you need additional sections, or you need certain
2211 particular functions to appear in special sections. The @code{section}
2212 attribute specifies that a function lives in a particular section.
2213 For example, the declaration:
2216 extern void foobar (void) __attribute__ ((section ("bar")));
2220 puts the function @code{foobar} in the @code{bar} section.
2222 Some file formats do not support arbitrary sections so the @code{section}
2223 attribute is not available on all platforms.
2224 If you need to map the entire contents of a module to a particular
2225 section, consider using the facilities of the linker instead.
2229 @cindex @code{constructor} function attribute
2230 @cindex @code{destructor} function attribute
2231 The @code{constructor} attribute causes the function to be called
2232 automatically before execution enters @code{main ()}. Similarly, the
2233 @code{destructor} attribute causes the function to be called
2234 automatically after @code{main ()} has completed or @code{exit ()} has
2235 been called. Functions with these attributes are useful for
2236 initializing data that will be used implicitly during the execution of
2239 These attributes are not currently implemented for Objective-C@.
2241 @cindex @code{unused} attribute.
2243 This attribute, attached to a function, means that the function is meant
2244 to be possibly unused. GCC will not produce a warning for this
2245 function. GNU C++ does not currently support this attribute as
2246 definitions without parameters are valid in C++.
2248 @cindex @code{used} attribute.
2250 This attribute, attached to a function, means that code must be emitted
2251 for the function even if it appears that the function is not referenced.
2252 This is useful, for example, when the function is referenced only in
2255 @cindex @code{deprecated} attribute.
2257 The @code{deprecated} attribute results in a warning if the function
2258 is used anywhere in the source file. This is useful when identifying
2259 functions that are expected to be removed in a future version of a
2260 program. The warning also includes the location of the declaration
2261 of the deprecated function, to enable users to easily find further
2262 information about why the function is deprecated, or what they should
2263 do instead. Note that the warnings only occurs for uses:
2266 int old_fn () __attribute__ ((deprecated));
2268 int (*fn_ptr)() = old_fn;
2271 results in a warning on line 3 but not line 2.
2273 The @code{deprecated} attribute can also be used for variables and
2274 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2277 @cindex @code{weak} attribute
2278 The @code{weak} attribute causes the declaration to be emitted as a weak
2279 symbol rather than a global. This is primarily useful in defining
2280 library functions which can be overridden in user code, though it can
2281 also be used with non-function declarations. Weak symbols are supported
2282 for ELF targets, and also for a.out targets when using the GNU assembler
2286 @cindex @code{malloc} attribute
2287 The @code{malloc} attribute is used to tell the compiler that a function
2288 may be treated as if it were the malloc function. The compiler assumes
2289 that calls to malloc result in a pointers that cannot alias anything.
2290 This will often improve optimization.
2292 @item alias ("@var{target}")
2293 @cindex @code{alias} attribute
2294 The @code{alias} attribute causes the declaration to be emitted as an
2295 alias for another symbol, which must be specified. For instance,
2298 void __f () @{ /* @r{Do something.} */; @}
2299 void f () __attribute__ ((weak, alias ("__f")));
2302 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2303 mangled name for the target must be used.
2305 Not all target machines support this attribute.
2307 @item visibility ("@var{visibility_type}")
2308 @cindex @code{visibility} attribute
2309 The @code{visibility} attribute on ELF targets causes the declaration
2310 to be emitted with hidden, protected or internal visibility.
2313 void __attribute__ ((visibility ("protected")))
2314 f () @{ /* @r{Do something.} */; @}
2315 int i __attribute__ ((visibility ("hidden")));
2318 See the ELF gABI for complete details, but the short story is
2322 Hidden visibility indicates that the symbol will not be placed into
2323 the dynamic symbol table, so no other @dfn{module} (executable or
2324 shared library) can reference it directly.
2327 Protected visibility indicates that the symbol will be placed in the
2328 dynamic symbol table, but that references within the defining module
2329 will bind to the local symbol. That is, the symbol cannot be overridden
2333 Internal visibility is like hidden visibility, but with additional
2334 processor specific semantics. Unless otherwise specified by the psABI,
2335 gcc defines internal visibility to mean that the function is @emph{never}
2336 called from another module. Note that hidden symbols, while then cannot
2337 be referenced directly by other modules, can be referenced indirectly via
2338 function pointers. By indicating that a symbol cannot be called from
2339 outside the module, gcc may for instance omit the load of a PIC register
2340 since it is known that the calling function loaded the correct value.
2343 Not all ELF targets support this attribute.
2345 @item tls_model ("@var{tls_model}")
2346 @cindex @code{tls_model} attribute
2347 The @code{tls_model} attribute sets thread-local storage model
2348 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
2349 overriding @code{-ftls-model=} command line switch on a per-variable
2351 The @var{tls_model} argument should be one of @code{global-dynamic},
2352 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
2354 @item regparm (@var{number})
2355 @cindex functions that are passed arguments in registers on the 386
2356 On the Intel 386, the @code{regparm} attribute causes the compiler to
2357 pass up to @var{number} integer arguments in registers EAX,
2358 EDX, and ECX instead of on the stack. Functions that take a
2359 variable number of arguments will continue to be passed all of their
2360 arguments on the stack.
2363 @cindex functions that pop the argument stack on the 386
2364 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2365 assume that the called function will pop off the stack space used to
2366 pass arguments, unless it takes a variable number of arguments.
2368 The PowerPC compiler for Windows NT currently ignores the @code{stdcall}
2372 @cindex functions that do pop the argument stack on the 386
2374 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2375 assume that the calling function will pop off the stack space used to
2376 pass arguments. This is
2377 useful to override the effects of the @option{-mrtd} switch.
2379 The PowerPC compiler for Windows NT currently ignores the @code{cdecl}
2382 @item longcall/shortcall
2383 @cindex functions called via pointer on the RS/6000 and PowerPC
2384 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2385 compiler to always call this function via a pointer, just as it would if
2386 the @option{-mlongcall} option had been specified. The @code{shortcall}
2387 attribute causes the compiler not to do this. These attributes override
2388 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2391 @xref{RS/6000 and PowerPC Options}, for more information on when long
2392 calls are and are not necessary.
2394 @item long_call/short_call
2395 @cindex indirect calls on ARM
2396 This attribute allows to specify how to call a particular function on
2397 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2398 command line switch and @code{#pragma long_calls} settings. The
2399 @code{long_call} attribute causes the compiler to always call the
2400 function by first loading its address into a register and then using the
2401 contents of that register. The @code{short_call} attribute always places
2402 the offset to the function from the call site into the @samp{BL}
2403 instruction directly.
2406 @cindex functions which are imported from a dll on PowerPC Windows NT
2407 On the PowerPC running Windows NT, the @code{dllimport} attribute causes
2408 the compiler to call the function via a global pointer to the function
2409 pointer that is set up by the Windows NT dll library. The pointer name
2410 is formed by combining @code{__imp_} and the function name.
2413 @cindex functions which are exported from a dll on PowerPC Windows NT
2414 On the PowerPC running Windows NT, the @code{dllexport} attribute causes
2415 the compiler to provide a global pointer to the function pointer, so
2416 that it can be called with the @code{dllimport} attribute. The pointer
2417 name is formed by combining @code{__imp_} and the function name.
2419 @item exception (@var{except-func} [, @var{except-arg}])
2420 @cindex functions which specify exception handling on PowerPC Windows NT
2421 On the PowerPC running Windows NT, the @code{exception} attribute causes
2422 the compiler to modify the structured exception table entry it emits for
2423 the declared function. The string or identifier @var{except-func} is
2424 placed in the third entry of the structured exception table. It
2425 represents a function, which is called by the exception handling
2426 mechanism if an exception occurs. If it was specified, the string or
2427 identifier @var{except-arg} is placed in the fourth entry of the
2428 structured exception table.
2430 @item function_vector
2431 @cindex calling functions through the function vector on the H8/300 processors
2432 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2433 function should be called through the function vector. Calling a
2434 function through the function vector will reduce code size, however;
2435 the function vector has a limited size (maximum 128 entries on the H8/300
2436 and 64 entries on the H8/300H) and shares space with the interrupt vector.
2438 You must use GAS and GLD from GNU binutils version 2.7 or later for
2439 this attribute to work correctly.
2442 @cindex interrupt handler functions
2443 Use this attribute on the ARM, AVR, M32R/D and Xstormy16 ports to indicate
2444 that the specified function is an interrupt handler. The compiler will
2445 generate function entry and exit sequences suitable for use in an
2446 interrupt handler when this attribute is present.
2448 Note, interrupt handlers for the H8/300, H8/300H and SH processors can
2449 be specified via the @code{interrupt_handler} attribute.
2451 Note, on the AVR interrupts will be enabled inside the function.
2453 Note, for the ARM you can specify the kind of interrupt to be handled by
2454 adding an optional parameter to the interrupt attribute like this:
2457 void f () __attribute__ ((interrupt ("IRQ")));
2460 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2462 @item interrupt_handler
2463 @cindex interrupt handler functions on the H8/300 and SH processors
2464 Use this attribute on the H8/300, H8/300H and SH to indicate that the
2465 specified function is an interrupt handler. The compiler will generate
2466 function entry and exit sequences suitable for use in an interrupt
2467 handler when this attribute is present.
2470 Use this attribute on the SH to indicate an @code{interrupt_handler}
2471 function should switch to an alternate stack. It expects a string
2472 argument that names a global variable holding the address of the
2477 void f () __attribute__ ((interrupt_handler,
2478 sp_switch ("alt_stack")));
2482 Use this attribute on the SH for an @code{interrupt_handle} to return using
2483 @code{trapa} instead of @code{rte}. This attribute expects an integer
2484 argument specifying the trap number to be used.
2487 @cindex eight bit data on the H8/300 and H8/300H
2488 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2489 variable should be placed into the eight bit data section.
2490 The compiler will generate more efficient code for certain operations
2491 on data in the eight bit data area. Note the eight bit data area is limited to
2494 You must use GAS and GLD from GNU binutils version 2.7 or later for
2495 this attribute to work correctly.
2498 @cindex tiny data section on the H8/300H
2499 Use this attribute on the H8/300H to indicate that the specified
2500 variable should be placed into the tiny data section.
2501 The compiler will generate more efficient code for loads and stores
2502 on data in the tiny data section. Note the tiny data area is limited to
2503 slightly under 32kbytes of data.
2506 @cindex signal handler functions on the AVR processors
2507 Use this attribute on the AVR to indicate that the specified
2508 function is an signal handler. The compiler will generate function
2509 entry and exit sequences suitable for use in an signal handler when this
2510 attribute is present. Interrupts will be disabled inside function.
2513 @cindex function without a prologue/epilogue code
2514 Use this attribute on the ARM, AVR and IP2K ports to indicate that the
2515 specified function do not need prologue/epilogue sequences generated by
2516 the compiler. It is up to the programmer to provide these sequences.
2518 @item model (@var{model-name})
2519 @cindex function addressability on the M32R/D
2520 Use this attribute on the M32R/D to set the addressability of an object,
2521 and the code generated for a function.
2522 The identifier @var{model-name} is one of @code{small}, @code{medium},
2523 or @code{large}, representing each of the code models.
2525 Small model objects live in the lower 16MB of memory (so that their
2526 addresses can be loaded with the @code{ld24} instruction), and are
2527 callable with the @code{bl} instruction.
2529 Medium model objects may live anywhere in the 32-bit address space (the
2530 compiler will generate @code{seth/add3} instructions to load their addresses),
2531 and are callable with the @code{bl} instruction.
2533 Large model objects may live anywhere in the 32-bit address space (the
2534 compiler will generate @code{seth/add3} instructions to load their addresses),
2535 and may not be reachable with the @code{bl} instruction (the compiler will
2536 generate the much slower @code{seth/add3/jl} instruction sequence).
2540 You can specify multiple attributes in a declaration by separating them
2541 by commas within the double parentheses or by immediately following an
2542 attribute declaration with another attribute declaration.
2544 @cindex @code{#pragma}, reason for not using
2545 @cindex pragma, reason for not using
2546 Some people object to the @code{__attribute__} feature, suggesting that
2547 ISO C's @code{#pragma} should be used instead. At the time
2548 @code{__attribute__} was designed, there were two reasons for not doing
2553 It is impossible to generate @code{#pragma} commands from a macro.
2556 There is no telling what the same @code{#pragma} might mean in another
2560 These two reasons applied to almost any application that might have been
2561 proposed for @code{#pragma}. It was basically a mistake to use
2562 @code{#pragma} for @emph{anything}.
2564 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2565 to be generated from macros. In addition, a @code{#pragma GCC}
2566 namespace is now in use for GCC-specific pragmas. However, it has been
2567 found convenient to use @code{__attribute__} to achieve a natural
2568 attachment of attributes to their corresponding declarations, whereas
2569 @code{#pragma GCC} is of use for constructs that do not naturally form
2570 part of the grammar. @xref{Other Directives,,Miscellaneous
2571 Preprocessing Directives, cpp, The C Preprocessor}.
2573 @node Attribute Syntax
2574 @section Attribute Syntax
2575 @cindex attribute syntax
2577 This section describes the syntax with which @code{__attribute__} may be
2578 used, and the constructs to which attribute specifiers bind, for the C
2579 language. Some details may vary for C++ and Objective-C@. Because of
2580 infelicities in the grammar for attributes, some forms described here
2581 may not be successfully parsed in all cases.
2583 There are some problems with the semantics of attributes in C++. For
2584 example, there are no manglings for attributes, although they may affect
2585 code generation, so problems may arise when attributed types are used in
2586 conjunction with templates or overloading. Similarly, @code{typeid}
2587 does not distinguish between types with different attributes. Support
2588 for attributes in C++ may be restricted in future to attributes on
2589 declarations only, but not on nested declarators.
2591 @xref{Function Attributes}, for details of the semantics of attributes
2592 applying to functions. @xref{Variable Attributes}, for details of the
2593 semantics of attributes applying to variables. @xref{Type Attributes},
2594 for details of the semantics of attributes applying to structure, union
2595 and enumerated types.
2597 An @dfn{attribute specifier} is of the form
2598 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2599 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2600 each attribute is one of the following:
2604 Empty. Empty attributes are ignored.
2607 A word (which may be an identifier such as @code{unused}, or a reserved
2608 word such as @code{const}).
2611 A word, followed by, in parentheses, parameters for the attribute.
2612 These parameters take one of the following forms:
2616 An identifier. For example, @code{mode} attributes use this form.
2619 An identifier followed by a comma and a non-empty comma-separated list
2620 of expressions. For example, @code{format} attributes use this form.
2623 A possibly empty comma-separated list of expressions. For example,
2624 @code{format_arg} attributes use this form with the list being a single
2625 integer constant expression, and @code{alias} attributes use this form
2626 with the list being a single string constant.
2630 An @dfn{attribute specifier list} is a sequence of one or more attribute
2631 specifiers, not separated by any other tokens.
2633 An attribute specifier list may appear after the colon following a
2634 label, other than a @code{case} or @code{default} label. The only
2635 attribute it makes sense to use after a label is @code{unused}. This
2636 feature is intended for code generated by programs which contains labels
2637 that may be unused but which is compiled with @option{-Wall}. It would
2638 not normally be appropriate to use in it human-written code, though it
2639 could be useful in cases where the code that jumps to the label is
2640 contained within an @code{#ifdef} conditional.
2642 An attribute specifier list may appear as part of a @code{struct},
2643 @code{union} or @code{enum} specifier. It may go either immediately
2644 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2645 the closing brace. It is ignored if the content of the structure, union
2646 or enumerated type is not defined in the specifier in which the
2647 attribute specifier list is used---that is, in usages such as
2648 @code{struct __attribute__((foo)) bar} with no following opening brace.
2649 Where attribute specifiers follow the closing brace, they are considered
2650 to relate to the structure, union or enumerated type defined, not to any
2651 enclosing declaration the type specifier appears in, and the type
2652 defined is not complete until after the attribute specifiers.
2653 @c Otherwise, there would be the following problems: a shift/reduce
2654 @c conflict between attributes binding the struct/union/enum and
2655 @c binding to the list of specifiers/qualifiers; and "aligned"
2656 @c attributes could use sizeof for the structure, but the size could be
2657 @c changed later by "packed" attributes.
2659 Otherwise, an attribute specifier appears as part of a declaration,
2660 counting declarations of unnamed parameters and type names, and relates
2661 to that declaration (which may be nested in another declaration, for
2662 example in the case of a parameter declaration), or to a particular declarator
2663 within a declaration. Where an
2664 attribute specifier is applied to a parameter declared as a function or
2665 an array, it should apply to the function or array rather than the
2666 pointer to which the parameter is implicitly converted, but this is not
2667 yet correctly implemented.
2669 Any list of specifiers and qualifiers at the start of a declaration may
2670 contain attribute specifiers, whether or not such a list may in that
2671 context contain storage class specifiers. (Some attributes, however,
2672 are essentially in the nature of storage class specifiers, and only make
2673 sense where storage class specifiers may be used; for example,
2674 @code{section}.) There is one necessary limitation to this syntax: the
2675 first old-style parameter declaration in a function definition cannot
2676 begin with an attribute specifier, because such an attribute applies to
2677 the function instead by syntax described below (which, however, is not
2678 yet implemented in this case). In some other cases, attribute
2679 specifiers are permitted by this grammar but not yet supported by the
2680 compiler. All attribute specifiers in this place relate to the
2681 declaration as a whole. In the obsolescent usage where a type of
2682 @code{int} is implied by the absence of type specifiers, such a list of
2683 specifiers and qualifiers may be an attribute specifier list with no
2684 other specifiers or qualifiers.
2686 An attribute specifier list may appear immediately before a declarator
2687 (other than the first) in a comma-separated list of declarators in a
2688 declaration of more than one identifier using a single list of
2689 specifiers and qualifiers. Such attribute specifiers apply
2690 only to the identifier before whose declarator they appear. For
2694 __attribute__((noreturn)) void d0 (void),
2695 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2700 the @code{noreturn} attribute applies to all the functions
2701 declared; the @code{format} attribute only applies to @code{d1}.
2703 An attribute specifier list may appear immediately before the comma,
2704 @code{=} or semicolon terminating the declaration of an identifier other
2705 than a function definition. At present, such attribute specifiers apply
2706 to the declared object or function, but in future they may attach to the
2707 outermost adjacent declarator. In simple cases there is no difference,
2708 but, for example, in
2711 void (****f)(void) __attribute__((noreturn));
2715 at present the @code{noreturn} attribute applies to @code{f}, which
2716 causes a warning since @code{f} is not a function, but in future it may
2717 apply to the function @code{****f}. The precise semantics of what
2718 attributes in such cases will apply to are not yet specified. Where an
2719 assembler name for an object or function is specified (@pxref{Asm
2720 Labels}), at present the attribute must follow the @code{asm}
2721 specification; in future, attributes before the @code{asm} specification
2722 may apply to the adjacent declarator, and those after it to the declared
2725 An attribute specifier list may, in future, be permitted to appear after
2726 the declarator in a function definition (before any old-style parameter
2727 declarations or the function body).
2729 Attribute specifiers may be mixed with type qualifiers appearing inside
2730 the @code{[]} of a parameter array declarator, in the C99 construct by
2731 which such qualifiers are applied to the pointer to which the array is
2732 implicitly converted. Such attribute specifiers apply to the pointer,
2733 not to the array, but at present this is not implemented and they are
2736 An attribute specifier list may appear at the start of a nested
2737 declarator. At present, there are some limitations in this usage: the
2738 attributes correctly apply to the declarator, but for most individual
2739 attributes the semantics this implies are not implemented.
2740 When attribute specifiers follow the @code{*} of a pointer
2741 declarator, they may be mixed with any type qualifiers present.
2742 The following describes the formal semantics of this syntax. It will make the
2743 most sense if you are familiar with the formal specification of
2744 declarators in the ISO C standard.
2746 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2747 D1}, where @code{T} contains declaration specifiers that specify a type
2748 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2749 contains an identifier @var{ident}. The type specified for @var{ident}
2750 for derived declarators whose type does not include an attribute
2751 specifier is as in the ISO C standard.
2753 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2754 and the declaration @code{T D} specifies the type
2755 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2756 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2757 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2759 If @code{D1} has the form @code{*
2760 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2761 declaration @code{T D} specifies the type
2762 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2763 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2764 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2770 void (__attribute__((noreturn)) ****f) (void);
2774 specifies the type ``pointer to pointer to pointer to pointer to
2775 non-returning function returning @code{void}''. As another example,
2778 char *__attribute__((aligned(8))) *f;
2782 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2783 Note again that this does not work with most attributes; for example,
2784 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2785 is not yet supported.
2787 For compatibility with existing code written for compiler versions that
2788 did not implement attributes on nested declarators, some laxity is
2789 allowed in the placing of attributes. If an attribute that only applies
2790 to types is applied to a declaration, it will be treated as applying to
2791 the type of that declaration. If an attribute that only applies to
2792 declarations is applied to the type of a declaration, it will be treated
2793 as applying to that declaration; and, for compatibility with code
2794 placing the attributes immediately before the identifier declared, such
2795 an attribute applied to a function return type will be treated as
2796 applying to the function type, and such an attribute applied to an array
2797 element type will be treated as applying to the array type. If an
2798 attribute that only applies to function types is applied to a
2799 pointer-to-function type, it will be treated as applying to the pointer
2800 target type; if such an attribute is applied to a function return type
2801 that is not a pointer-to-function type, it will be treated as applying
2802 to the function type.
2804 @node Function Prototypes
2805 @section Prototypes and Old-Style Function Definitions
2806 @cindex function prototype declarations
2807 @cindex old-style function definitions
2808 @cindex promotion of formal parameters
2810 GNU C extends ISO C to allow a function prototype to override a later
2811 old-style non-prototype definition. Consider the following example:
2814 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2821 /* @r{Prototype function declaration.} */
2822 int isroot P((uid_t));
2824 /* @r{Old-style function definition.} */
2826 isroot (x) /* ??? lossage here ??? */
2833 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2834 not allow this example, because subword arguments in old-style
2835 non-prototype definitions are promoted. Therefore in this example the
2836 function definition's argument is really an @code{int}, which does not
2837 match the prototype argument type of @code{short}.
2839 This restriction of ISO C makes it hard to write code that is portable
2840 to traditional C compilers, because the programmer does not know
2841 whether the @code{uid_t} type is @code{short}, @code{int}, or
2842 @code{long}. Therefore, in cases like these GNU C allows a prototype
2843 to override a later old-style definition. More precisely, in GNU C, a
2844 function prototype argument type overrides the argument type specified
2845 by a later old-style definition if the former type is the same as the
2846 latter type before promotion. Thus in GNU C the above example is
2847 equivalent to the following:
2860 GNU C++ does not support old-style function definitions, so this
2861 extension is irrelevant.
2864 @section C++ Style Comments
2866 @cindex C++ comments
2867 @cindex comments, C++ style
2869 In GNU C, you may use C++ style comments, which start with @samp{//} and
2870 continue until the end of the line. Many other C implementations allow
2871 such comments, and they are included in the 1999 C standard. However,
2872 C++ style comments are not recognized if you specify an @option{-std}
2873 option specifying a version of ISO C before C99, or @option{-ansi}
2874 (equivalent to @option{-std=c89}).
2877 @section Dollar Signs in Identifier Names
2879 @cindex dollar signs in identifier names
2880 @cindex identifier names, dollar signs in
2882 In GNU C, you may normally use dollar signs in identifier names.
2883 This is because many traditional C implementations allow such identifiers.
2884 However, dollar signs in identifiers are not supported on a few target
2885 machines, typically because the target assembler does not allow them.
2887 @node Character Escapes
2888 @section The Character @key{ESC} in Constants
2890 You can use the sequence @samp{\e} in a string or character constant to
2891 stand for the ASCII character @key{ESC}.
2894 @section Inquiring on Alignment of Types or Variables
2896 @cindex type alignment
2897 @cindex variable alignment
2899 The keyword @code{__alignof__} allows you to inquire about how an object
2900 is aligned, or the minimum alignment usually required by a type. Its
2901 syntax is just like @code{sizeof}.
2903 For example, if the target machine requires a @code{double} value to be
2904 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2905 This is true on many RISC machines. On more traditional machine
2906 designs, @code{__alignof__ (double)} is 4 or even 2.
2908 Some machines never actually require alignment; they allow reference to any
2909 data type even at an odd addresses. For these machines, @code{__alignof__}
2910 reports the @emph{recommended} alignment of a type.
2912 If the operand of @code{__alignof__} is an lvalue rather than a type,
2913 its value is the required alignment for its type, taking into account
2914 any minimum alignment specified with GCC's @code{__attribute__}
2915 extension (@pxref{Variable Attributes}). For example, after this
2919 struct foo @{ int x; char y; @} foo1;
2923 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2924 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2926 It is an error to ask for the alignment of an incomplete type.
2928 @node Variable Attributes
2929 @section Specifying Attributes of Variables
2930 @cindex attribute of variables
2931 @cindex variable attributes
2933 The keyword @code{__attribute__} allows you to specify special
2934 attributes of variables or structure fields. This keyword is followed
2935 by an attribute specification inside double parentheses. Ten
2936 attributes are currently defined for variables: @code{aligned},
2937 @code{mode}, @code{nocommon}, @code{packed}, @code{section},
2938 @code{transparent_union}, @code{unused}, @code{deprecated},
2939 @code{vector_size}, and @code{weak}. Some other attributes are defined
2940 for variables on particular target systems. Other attributes are
2941 available for functions (@pxref{Function Attributes}) and for types
2942 (@pxref{Type Attributes}). Other front ends might define more
2943 attributes (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2945 You may also specify attributes with @samp{__} preceding and following
2946 each keyword. This allows you to use them in header files without
2947 being concerned about a possible macro of the same name. For example,
2948 you may use @code{__aligned__} instead of @code{aligned}.
2950 @xref{Attribute Syntax}, for details of the exact syntax for using
2954 @cindex @code{aligned} attribute
2955 @item aligned (@var{alignment})
2956 This attribute specifies a minimum alignment for the variable or
2957 structure field, measured in bytes. For example, the declaration:
2960 int x __attribute__ ((aligned (16))) = 0;
2964 causes the compiler to allocate the global variable @code{x} on a
2965 16-byte boundary. On a 68040, this could be used in conjunction with
2966 an @code{asm} expression to access the @code{move16} instruction which
2967 requires 16-byte aligned operands.
2969 You can also specify the alignment of structure fields. For example, to
2970 create a double-word aligned @code{int} pair, you could write:
2973 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2977 This is an alternative to creating a union with a @code{double} member
2978 that forces the union to be double-word aligned.
2980 As in the preceding examples, you can explicitly specify the alignment
2981 (in bytes) that you wish the compiler to use for a given variable or
2982 structure field. Alternatively, you can leave out the alignment factor
2983 and just ask the compiler to align a variable or field to the maximum
2984 useful alignment for the target machine you are compiling for. For
2985 example, you could write:
2988 short array[3] __attribute__ ((aligned));
2991 Whenever you leave out the alignment factor in an @code{aligned} attribute
2992 specification, the compiler automatically sets the alignment for the declared
2993 variable or field to the largest alignment which is ever used for any data
2994 type on the target machine you are compiling for. Doing this can often make
2995 copy operations more efficient, because the compiler can use whatever
2996 instructions copy the biggest chunks of memory when performing copies to
2997 or from the variables or fields that you have aligned this way.
2999 The @code{aligned} attribute can only increase the alignment; but you
3000 can decrease it by specifying @code{packed} as well. See below.
3002 Note that the effectiveness of @code{aligned} attributes may be limited
3003 by inherent limitations in your linker. On many systems, the linker is
3004 only able to arrange for variables to be aligned up to a certain maximum
3005 alignment. (For some linkers, the maximum supported alignment may
3006 be very very small.) If your linker is only able to align variables
3007 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3008 in an @code{__attribute__} will still only provide you with 8 byte
3009 alignment. See your linker documentation for further information.
3011 @item mode (@var{mode})
3012 @cindex @code{mode} attribute
3013 This attribute specifies the data type for the declaration---whichever
3014 type corresponds to the mode @var{mode}. This in effect lets you
3015 request an integer or floating point type according to its width.
3017 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3018 indicate the mode corresponding to a one-byte integer, @samp{word} or
3019 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3020 or @samp{__pointer__} for the mode used to represent pointers.
3023 @cindex @code{nocommon} attribute
3025 This attribute specifies requests GCC not to place a variable
3026 ``common'' but instead to allocate space for it directly. If you
3027 specify the @option{-fno-common} flag, GCC will do this for all
3030 Specifying the @code{nocommon} attribute for a variable provides an
3031 initialization of zeros. A variable may only be initialized in one
3035 @cindex @code{packed} attribute
3036 The @code{packed} attribute specifies that a variable or structure field
3037 should have the smallest possible alignment---one byte for a variable,
3038 and one bit for a field, unless you specify a larger value with the
3039 @code{aligned} attribute.
3041 Here is a structure in which the field @code{x} is packed, so that it
3042 immediately follows @code{a}:
3048 int x[2] __attribute__ ((packed));
3052 @item section ("@var{section-name}")
3053 @cindex @code{section} variable attribute
3054 Normally, the compiler places the objects it generates in sections like
3055 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3056 or you need certain particular variables to appear in special sections,
3057 for example to map to special hardware. The @code{section}
3058 attribute specifies that a variable (or function) lives in a particular
3059 section. For example, this small program uses several specific section names:
3062 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3063 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3064 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3065 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3069 /* Initialize stack pointer */
3070 init_sp (stack + sizeof (stack));
3072 /* Initialize initialized data */
3073 memcpy (&init_data, &data, &edata - &data);
3075 /* Turn on the serial ports */
3082 Use the @code{section} attribute with an @emph{initialized} definition
3083 of a @emph{global} variable, as shown in the example. GCC issues
3084 a warning and otherwise ignores the @code{section} attribute in
3085 uninitialized variable declarations.
3087 You may only use the @code{section} attribute with a fully initialized
3088 global definition because of the way linkers work. The linker requires
3089 each object be defined once, with the exception that uninitialized
3090 variables tentatively go in the @code{common} (or @code{bss}) section
3091 and can be multiply ``defined''. You can force a variable to be
3092 initialized with the @option{-fno-common} flag or the @code{nocommon}
3095 Some file formats do not support arbitrary sections so the @code{section}
3096 attribute is not available on all platforms.
3097 If you need to map the entire contents of a module to a particular
3098 section, consider using the facilities of the linker instead.
3101 @cindex @code{shared} variable attribute
3102 On Windows NT, in addition to putting variable definitions in a named
3103 section, the section can also be shared among all running copies of an
3104 executable or DLL@. For example, this small program defines shared data
3105 by putting it in a named section @code{shared} and marking the section
3109 int foo __attribute__((section ("shared"), shared)) = 0;
3114 /* Read and write foo. All running
3115 copies see the same value. */
3121 You may only use the @code{shared} attribute along with @code{section}
3122 attribute with a fully initialized global definition because of the way
3123 linkers work. See @code{section} attribute for more information.
3125 The @code{shared} attribute is only available on Windows NT@.
3127 @item transparent_union
3128 This attribute, attached to a function parameter which is a union, means
3129 that the corresponding argument may have the type of any union member,
3130 but the argument is passed as if its type were that of the first union
3131 member. For more details see @xref{Type Attributes}. You can also use
3132 this attribute on a @code{typedef} for a union data type; then it
3133 applies to all function parameters with that type.
3136 This attribute, attached to a variable, means that the variable is meant
3137 to be possibly unused. GCC will not produce a warning for this
3141 The @code{deprecated} attribute results in a warning if the variable
3142 is used anywhere in the source file. This is useful when identifying
3143 variables that are expected to be removed in a future version of a
3144 program. The warning also includes the location of the declaration
3145 of the deprecated variable, to enable users to easily find further
3146 information about why the variable is deprecated, or what they should
3147 do instead. Note that the warnings only occurs for uses:
3150 extern int old_var __attribute__ ((deprecated));
3152 int new_fn () @{ return old_var; @}
3155 results in a warning on line 3 but not line 2.
3157 The @code{deprecated} attribute can also be used for functions and
3158 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3160 @item vector_size (@var{bytes})
3161 This attribute specifies the vector size for the variable, measured in
3162 bytes. For example, the declaration:
3165 int foo __attribute__ ((vector_size (16)));
3169 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3170 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3171 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3173 This attribute is only applicable to integral and float scalars,
3174 although arrays, pointers, and function return values are allowed in
3175 conjunction with this construct.
3177 Aggregates with this attribute are invalid, even if they are of the same
3178 size as a corresponding scalar. For example, the declaration:
3181 struct S @{ int a; @};
3182 struct S __attribute__ ((vector_size (16))) foo;
3186 is invalid even if the size of the structure is the same as the size of
3190 The @code{weak} attribute is described in @xref{Function Attributes}.
3192 @item model (@var{model-name})
3193 @cindex variable addressability on the M32R/D
3194 Use this attribute on the M32R/D to set the addressability of an object.
3195 The identifier @var{model-name} is one of @code{small}, @code{medium},
3196 or @code{large}, representing each of the code models.
3198 Small model objects live in the lower 16MB of memory (so that their
3199 addresses can be loaded with the @code{ld24} instruction).
3201 Medium and large model objects may live anywhere in the 32-bit address space
3202 (the compiler will generate @code{seth/add3} instructions to load their
3207 To specify multiple attributes, separate them by commas within the
3208 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3211 @node Type Attributes
3212 @section Specifying Attributes of Types
3213 @cindex attribute of types
3214 @cindex type attributes
3216 The keyword @code{__attribute__} allows you to specify special
3217 attributes of @code{struct} and @code{union} types when you define such
3218 types. This keyword is followed by an attribute specification inside
3219 double parentheses. Six attributes are currently defined for types:
3220 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3221 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3222 functions (@pxref{Function Attributes}) and for variables
3223 (@pxref{Variable Attributes}).
3225 You may also specify any one of these attributes with @samp{__}
3226 preceding and following its keyword. This allows you to use these
3227 attributes in header files without being concerned about a possible
3228 macro of the same name. For example, you may use @code{__aligned__}
3229 instead of @code{aligned}.
3231 You may specify the @code{aligned} and @code{transparent_union}
3232 attributes either in a @code{typedef} declaration or just past the
3233 closing curly brace of a complete enum, struct or union type
3234 @emph{definition} and the @code{packed} attribute only past the closing
3235 brace of a definition.
3237 You may also specify attributes between the enum, struct or union
3238 tag and the name of the type rather than after the closing brace.
3240 @xref{Attribute Syntax}, for details of the exact syntax for using
3244 @cindex @code{aligned} attribute
3245 @item aligned (@var{alignment})
3246 This attribute specifies a minimum alignment (in bytes) for variables
3247 of the specified type. For example, the declarations:
3250 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3251 typedef int more_aligned_int __attribute__ ((aligned (8)));
3255 force the compiler to insure (as far as it can) that each variable whose
3256 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3257 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3258 variables of type @code{struct S} aligned to 8-byte boundaries allows
3259 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3260 store) instructions when copying one variable of type @code{struct S} to
3261 another, thus improving run-time efficiency.
3263 Note that the alignment of any given @code{struct} or @code{union} type
3264 is required by the ISO C standard to be at least a perfect multiple of
3265 the lowest common multiple of the alignments of all of the members of
3266 the @code{struct} or @code{union} in question. This means that you @emph{can}
3267 effectively adjust the alignment of a @code{struct} or @code{union}
3268 type by attaching an @code{aligned} attribute to any one of the members
3269 of such a type, but the notation illustrated in the example above is a
3270 more obvious, intuitive, and readable way to request the compiler to
3271 adjust the alignment of an entire @code{struct} or @code{union} type.
3273 As in the preceding example, you can explicitly specify the alignment
3274 (in bytes) that you wish the compiler to use for a given @code{struct}
3275 or @code{union} type. Alternatively, you can leave out the alignment factor
3276 and just ask the compiler to align a type to the maximum
3277 useful alignment for the target machine you are compiling for. For
3278 example, you could write:
3281 struct S @{ short f[3]; @} __attribute__ ((aligned));
3284 Whenever you leave out the alignment factor in an @code{aligned}
3285 attribute specification, the compiler automatically sets the alignment
3286 for the type to the largest alignment which is ever used for any data
3287 type on the target machine you are compiling for. Doing this can often
3288 make copy operations more efficient, because the compiler can use
3289 whatever instructions copy the biggest chunks of memory when performing
3290 copies to or from the variables which have types that you have aligned
3293 In the example above, if the size of each @code{short} is 2 bytes, then
3294 the size of the entire @code{struct S} type is 6 bytes. The smallest
3295 power of two which is greater than or equal to that is 8, so the
3296 compiler sets the alignment for the entire @code{struct S} type to 8
3299 Note that although you can ask the compiler to select a time-efficient
3300 alignment for a given type and then declare only individual stand-alone
3301 objects of that type, the compiler's ability to select a time-efficient
3302 alignment is primarily useful only when you plan to create arrays of
3303 variables having the relevant (efficiently aligned) type. If you
3304 declare or use arrays of variables of an efficiently-aligned type, then
3305 it is likely that your program will also be doing pointer arithmetic (or
3306 subscripting, which amounts to the same thing) on pointers to the
3307 relevant type, and the code that the compiler generates for these
3308 pointer arithmetic operations will often be more efficient for
3309 efficiently-aligned types than for other types.
3311 The @code{aligned} attribute can only increase the alignment; but you
3312 can decrease it by specifying @code{packed} as well. See below.
3314 Note that the effectiveness of @code{aligned} attributes may be limited
3315 by inherent limitations in your linker. On many systems, the linker is
3316 only able to arrange for variables to be aligned up to a certain maximum
3317 alignment. (For some linkers, the maximum supported alignment may
3318 be very very small.) If your linker is only able to align variables
3319 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3320 in an @code{__attribute__} will still only provide you with 8 byte
3321 alignment. See your linker documentation for further information.
3324 This attribute, attached to an @code{enum}, @code{struct}, or
3325 @code{union} type definition, specified that the minimum required memory
3326 be used to represent the type.
3328 @opindex fshort-enums
3329 Specifying this attribute for @code{struct} and @code{union} types is
3330 equivalent to specifying the @code{packed} attribute on each of the
3331 structure or union members. Specifying the @option{-fshort-enums}
3332 flag on the line is equivalent to specifying the @code{packed}
3333 attribute on all @code{enum} definitions.
3335 You may only specify this attribute after a closing curly brace on an
3336 @code{enum} definition, not in a @code{typedef} declaration, unless that
3337 declaration also contains the definition of the @code{enum}.
3339 @item transparent_union
3340 This attribute, attached to a @code{union} type definition, indicates
3341 that any function parameter having that union type causes calls to that
3342 function to be treated in a special way.
3344 First, the argument corresponding to a transparent union type can be of
3345 any type in the union; no cast is required. Also, if the union contains
3346 a pointer type, the corresponding argument can be a null pointer
3347 constant or a void pointer expression; and if the union contains a void
3348 pointer type, the corresponding argument can be any pointer expression.
3349 If the union member type is a pointer, qualifiers like @code{const} on
3350 the referenced type must be respected, just as with normal pointer
3353 Second, the argument is passed to the function using the calling
3354 conventions of first member of the transparent union, not the calling
3355 conventions of the union itself. All members of the union must have the
3356 same machine representation; this is necessary for this argument passing
3359 Transparent unions are designed for library functions that have multiple
3360 interfaces for compatibility reasons. For example, suppose the
3361 @code{wait} function must accept either a value of type @code{int *} to
3362 comply with Posix, or a value of type @code{union wait *} to comply with
3363 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3364 @code{wait} would accept both kinds of arguments, but it would also
3365 accept any other pointer type and this would make argument type checking
3366 less useful. Instead, @code{<sys/wait.h>} might define the interface
3374 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3376 pid_t wait (wait_status_ptr_t);
3379 This interface allows either @code{int *} or @code{union wait *}
3380 arguments to be passed, using the @code{int *} calling convention.
3381 The program can call @code{wait} with arguments of either type:
3384 int w1 () @{ int w; return wait (&w); @}
3385 int w2 () @{ union wait w; return wait (&w); @}
3388 With this interface, @code{wait}'s implementation might look like this:
3391 pid_t wait (wait_status_ptr_t p)
3393 return waitpid (-1, p.__ip, 0);
3398 When attached to a type (including a @code{union} or a @code{struct}),
3399 this attribute means that variables of that type are meant to appear
3400 possibly unused. GCC will not produce a warning for any variables of
3401 that type, even if the variable appears to do nothing. This is often
3402 the case with lock or thread classes, which are usually defined and then
3403 not referenced, but contain constructors and destructors that have
3404 nontrivial bookkeeping functions.
3407 The @code{deprecated} attribute results in a warning if the type
3408 is used anywhere in the source file. This is useful when identifying
3409 types that are expected to be removed in a future version of a program.
3410 If possible, the warning also includes the location of the declaration
3411 of the deprecated type, to enable users to easily find further
3412 information about why the type is deprecated, or what they should do
3413 instead. Note that the warnings only occur for uses and then only
3414 if the type is being applied to an identifier that itself is not being
3415 declared as deprecated.
3418 typedef int T1 __attribute__ ((deprecated));
3422 typedef T1 T3 __attribute__ ((deprecated));
3423 T3 z __attribute__ ((deprecated));
3426 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3427 warning is issued for line 4 because T2 is not explicitly
3428 deprecated. Line 5 has no warning because T3 is explicitly
3429 deprecated. Similarly for line 6.
3431 The @code{deprecated} attribute can also be used for functions and
3432 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3435 Accesses to objects with types with this attribute are not subjected to
3436 type-based alias analysis, but are instead assumed to be able to alias
3437 any other type of objects, just like the @code{char} type. See
3438 @option{-fstrict-aliasing} for more information on aliasing issues.
3443 typedef short __attribute__((__may_alias__)) short_a;
3449 short_a *b = (short_a *) &a;
3453 if (a == 0x12345678)
3460 If you replaced @code{short_a} with @code{short} in the variable
3461 declaration, the above program would abort when compiled with
3462 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3463 above in recent GCC versions.
3466 To specify multiple attributes, separate them by commas within the
3467 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3471 @section An Inline Function is As Fast As a Macro
3472 @cindex inline functions
3473 @cindex integrating function code
3475 @cindex macros, inline alternative
3477 By declaring a function @code{inline}, you can direct GCC to
3478 integrate that function's code into the code for its callers. This
3479 makes execution faster by eliminating the function-call overhead; in
3480 addition, if any of the actual argument values are constant, their known
3481 values may permit simplifications at compile time so that not all of the
3482 inline function's code needs to be included. The effect on code size is
3483 less predictable; object code may be larger or smaller with function
3484 inlining, depending on the particular case. Inlining of functions is an
3485 optimization and it really ``works'' only in optimizing compilation. If
3486 you don't use @option{-O}, no function is really inline.
3488 Inline functions are included in the ISO C99 standard, but there are
3489 currently substantial differences between what GCC implements and what
3490 the ISO C99 standard requires.
3492 To declare a function inline, use the @code{inline} keyword in its
3493 declaration, like this:
3503 (If you are writing a header file to be included in ISO C programs, write
3504 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3505 You can also make all ``simple enough'' functions inline with the option
3506 @option{-finline-functions}.
3509 Note that certain usages in a function definition can make it unsuitable
3510 for inline substitution. Among these usages are: use of varargs, use of
3511 alloca, use of variable sized data types (@pxref{Variable Length}),
3512 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3513 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3514 will warn when a function marked @code{inline} could not be substituted,
3515 and will give the reason for the failure.
3517 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3518 does not affect the linkage of the function.
3520 @cindex automatic @code{inline} for C++ member fns
3521 @cindex @code{inline} automatic for C++ member fns
3522 @cindex member fns, automatically @code{inline}
3523 @cindex C++ member fns, automatically @code{inline}
3524 @opindex fno-default-inline
3525 GCC automatically inlines member functions defined within the class
3526 body of C++ programs even if they are not explicitly declared
3527 @code{inline}. (You can override this with @option{-fno-default-inline};
3528 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3530 @cindex inline functions, omission of
3531 @opindex fkeep-inline-functions
3532 When a function is both inline and @code{static}, if all calls to the
3533 function are integrated into the caller, and the function's address is
3534 never used, then the function's own assembler code is never referenced.
3535 In this case, GCC does not actually output assembler code for the
3536 function, unless you specify the option @option{-fkeep-inline-functions}.
3537 Some calls cannot be integrated for various reasons (in particular,
3538 calls that precede the function's definition cannot be integrated, and
3539 neither can recursive calls within the definition). If there is a
3540 nonintegrated call, then the function is compiled to assembler code as
3541 usual. The function must also be compiled as usual if the program
3542 refers to its address, because that can't be inlined.
3544 @cindex non-static inline function
3545 When an inline function is not @code{static}, then the compiler must assume
3546 that there may be calls from other source files; since a global symbol can
3547 be defined only once in any program, the function must not be defined in
3548 the other source files, so the calls therein cannot be integrated.
3549 Therefore, a non-@code{static} inline function is always compiled on its
3550 own in the usual fashion.
3552 If you specify both @code{inline} and @code{extern} in the function
3553 definition, then the definition is used only for inlining. In no case
3554 is the function compiled on its own, not even if you refer to its
3555 address explicitly. Such an address becomes an external reference, as
3556 if you had only declared the function, and had not defined it.
3558 This combination of @code{inline} and @code{extern} has almost the
3559 effect of a macro. The way to use it is to put a function definition in
3560 a header file with these keywords, and put another copy of the
3561 definition (lacking @code{inline} and @code{extern}) in a library file.
3562 The definition in the header file will cause most calls to the function
3563 to be inlined. If any uses of the function remain, they will refer to
3564 the single copy in the library.
3566 For future compatibility with when GCC implements ISO C99 semantics for
3567 inline functions, it is best to use @code{static inline} only. (The
3568 existing semantics will remain available when @option{-std=gnu89} is
3569 specified, but eventually the default will be @option{-std=gnu99} and
3570 that will implement the C99 semantics, though it does not do so yet.)
3572 GCC does not inline any functions when not optimizing unless you specify
3573 the @samp{always_inline} attribute for the function, like this:
3577 inline void foo (const char) __attribute__((always_inline));
3581 @section Assembler Instructions with C Expression Operands
3582 @cindex extended @code{asm}
3583 @cindex @code{asm} expressions
3584 @cindex assembler instructions
3587 In an assembler instruction using @code{asm}, you can specify the
3588 operands of the instruction using C expressions. This means you need not
3589 guess which registers or memory locations will contain the data you want
3592 You must specify an assembler instruction template much like what
3593 appears in a machine description, plus an operand constraint string for
3596 For example, here is how to use the 68881's @code{fsinx} instruction:
3599 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3603 Here @code{angle} is the C expression for the input operand while
3604 @code{result} is that of the output operand. Each has @samp{"f"} as its
3605 operand constraint, saying that a floating point register is required.
3606 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3607 output operands' constraints must use @samp{=}. The constraints use the
3608 same language used in the machine description (@pxref{Constraints}).
3610 Each operand is described by an operand-constraint string followed by
3611 the C expression in parentheses. A colon separates the assembler
3612 template from the first output operand and another separates the last
3613 output operand from the first input, if any. Commas separate the
3614 operands within each group. The total number of operands is currently
3615 limited to 30; this limitation may be lifted in some future version of
3618 If there are no output operands but there are input operands, you must
3619 place two consecutive colons surrounding the place where the output
3622 As of GCC version 3.1, it is also possible to specify input and output
3623 operands using symbolic names which can be referenced within the
3624 assembler code. These names are specified inside square brackets
3625 preceding the constraint string, and can be referenced inside the
3626 assembler code using @code{%[@var{name}]} instead of a percentage sign
3627 followed by the operand number. Using named operands the above example
3631 asm ("fsinx %[angle],%[output]"
3632 : [output] "=f" (result)
3633 : [angle] "f" (angle));
3637 Note that the symbolic operand names have no relation whatsoever to
3638 other C identifiers. You may use any name you like, even those of
3639 existing C symbols, but must ensure that no two operands within the same
3640 assembler construct use the same symbolic name.
3642 Output operand expressions must be lvalues; the compiler can check this.
3643 The input operands need not be lvalues. The compiler cannot check
3644 whether the operands have data types that are reasonable for the
3645 instruction being executed. It does not parse the assembler instruction
3646 template and does not know what it means or even whether it is valid
3647 assembler input. The extended @code{asm} feature is most often used for
3648 machine instructions the compiler itself does not know exist. If
3649 the output expression cannot be directly addressed (for example, it is a
3650 bit-field), your constraint must allow a register. In that case, GCC
3651 will use the register as the output of the @code{asm}, and then store
3652 that register into the output.
3654 The ordinary output operands must be write-only; GCC will assume that
3655 the values in these operands before the instruction are dead and need
3656 not be generated. Extended asm supports input-output or read-write
3657 operands. Use the constraint character @samp{+} to indicate such an
3658 operand and list it with the output operands.
3660 When the constraints for the read-write operand (or the operand in which
3661 only some of the bits are to be changed) allows a register, you may, as
3662 an alternative, logically split its function into two separate operands,
3663 one input operand and one write-only output operand. The connection
3664 between them is expressed by constraints which say they need to be in
3665 the same location when the instruction executes. You can use the same C
3666 expression for both operands, or different expressions. For example,
3667 here we write the (fictitious) @samp{combine} instruction with
3668 @code{bar} as its read-only source operand and @code{foo} as its
3669 read-write destination:
3672 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3676 The constraint @samp{"0"} for operand 1 says that it must occupy the
3677 same location as operand 0. A number in constraint is allowed only in
3678 an input operand and it must refer to an output operand.
3680 Only a number in the constraint can guarantee that one operand will be in
3681 the same place as another. The mere fact that @code{foo} is the value
3682 of both operands is not enough to guarantee that they will be in the
3683 same place in the generated assembler code. The following would not
3687 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3690 Various optimizations or reloading could cause operands 0 and 1 to be in
3691 different registers; GCC knows no reason not to do so. For example, the
3692 compiler might find a copy of the value of @code{foo} in one register and
3693 use it for operand 1, but generate the output operand 0 in a different
3694 register (copying it afterward to @code{foo}'s own address). Of course,
3695 since the register for operand 1 is not even mentioned in the assembler
3696 code, the result will not work, but GCC can't tell that.
3698 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3699 the operand number for a matching constraint. For example:
3702 asm ("cmoveq %1,%2,%[result]"
3703 : [result] "=r"(result)
3704 : "r" (test), "r"(new), "[result]"(old));
3707 Some instructions clobber specific hard registers. To describe this,
3708 write a third colon after the input operands, followed by the names of
3709 the clobbered hard registers (given as strings). Here is a realistic
3710 example for the VAX:
3713 asm volatile ("movc3 %0,%1,%2"
3715 : "g" (from), "g" (to), "g" (count)
3716 : "r0", "r1", "r2", "r3", "r4", "r5");
3719 You may not write a clobber description in a way that overlaps with an
3720 input or output operand. For example, you may not have an operand
3721 describing a register class with one member if you mention that register
3722 in the clobber list. Variables declared to live in specific registers
3723 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3724 have no part mentioned in the clobber description.
3725 There is no way for you to specify that an input
3726 operand is modified without also specifying it as an output
3727 operand. Note that if all the output operands you specify are for this
3728 purpose (and hence unused), you will then also need to specify
3729 @code{volatile} for the @code{asm} construct, as described below, to
3730 prevent GCC from deleting the @code{asm} statement as unused.
3732 If you refer to a particular hardware register from the assembler code,
3733 you will probably have to list the register after the third colon to
3734 tell the compiler the register's value is modified. In some assemblers,
3735 the register names begin with @samp{%}; to produce one @samp{%} in the
3736 assembler code, you must write @samp{%%} in the input.
3738 If your assembler instruction can alter the condition code register, add
3739 @samp{cc} to the list of clobbered registers. GCC on some machines
3740 represents the condition codes as a specific hardware register;
3741 @samp{cc} serves to name this register. On other machines, the
3742 condition code is handled differently, and specifying @samp{cc} has no
3743 effect. But it is valid no matter what the machine.
3745 If your assembler instruction modifies memory in an unpredictable
3746 fashion, add @samp{memory} to the list of clobbered registers. This
3747 will cause GCC to not keep memory values cached in registers across
3748 the assembler instruction. You will also want to add the
3749 @code{volatile} keyword if the memory affected is not listed in the
3750 inputs or outputs of the @code{asm}, as the @samp{memory} clobber does
3751 not count as a side-effect of the @code{asm}.
3753 You can put multiple assembler instructions together in a single
3754 @code{asm} template, separated by the characters normally used in assembly
3755 code for the system. A combination that works in most places is a newline
3756 to break the line, plus a tab character to move to the instruction field
3757 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3758 assembler allows semicolons as a line-breaking character. Note that some
3759 assembler dialects use semicolons to start a comment.
3760 The input operands are guaranteed not to use any of the clobbered
3761 registers, and neither will the output operands' addresses, so you can
3762 read and write the clobbered registers as many times as you like. Here
3763 is an example of multiple instructions in a template; it assumes the
3764 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3767 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3769 : "g" (from), "g" (to)
3773 Unless an output operand has the @samp{&} constraint modifier, GCC
3774 may allocate it in the same register as an unrelated input operand, on
3775 the assumption the inputs are consumed before the outputs are produced.
3776 This assumption may be false if the assembler code actually consists of
3777 more than one instruction. In such a case, use @samp{&} for each output
3778 operand that may not overlap an input. @xref{Modifiers}.
3780 If you want to test the condition code produced by an assembler
3781 instruction, you must include a branch and a label in the @code{asm}
3782 construct, as follows:
3785 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3791 This assumes your assembler supports local labels, as the GNU assembler
3792 and most Unix assemblers do.
3794 Speaking of labels, jumps from one @code{asm} to another are not
3795 supported. The compiler's optimizers do not know about these jumps, and
3796 therefore they cannot take account of them when deciding how to
3799 @cindex macros containing @code{asm}
3800 Usually the most convenient way to use these @code{asm} instructions is to
3801 encapsulate them in macros that look like functions. For example,
3805 (@{ double __value, __arg = (x); \
3806 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
3811 Here the variable @code{__arg} is used to make sure that the instruction
3812 operates on a proper @code{double} value, and to accept only those
3813 arguments @code{x} which can convert automatically to a @code{double}.
3815 Another way to make sure the instruction operates on the correct data
3816 type is to use a cast in the @code{asm}. This is different from using a
3817 variable @code{__arg} in that it converts more different types. For
3818 example, if the desired type were @code{int}, casting the argument to
3819 @code{int} would accept a pointer with no complaint, while assigning the
3820 argument to an @code{int} variable named @code{__arg} would warn about
3821 using a pointer unless the caller explicitly casts it.
3823 If an @code{asm} has output operands, GCC assumes for optimization
3824 purposes the instruction has no side effects except to change the output
3825 operands. This does not mean instructions with a side effect cannot be
3826 used, but you must be careful, because the compiler may eliminate them
3827 if the output operands aren't used, or move them out of loops, or
3828 replace two with one if they constitute a common subexpression. Also,
3829 if your instruction does have a side effect on a variable that otherwise
3830 appears not to change, the old value of the variable may be reused later
3831 if it happens to be found in a register.
3833 You can prevent an @code{asm} instruction from being deleted, moved
3834 significantly, or combined, by writing the keyword @code{volatile} after
3835 the @code{asm}. For example:
3838 #define get_and_set_priority(new) \
3840 asm volatile ("get_and_set_priority %0, %1" \
3841 : "=g" (__old) : "g" (new)); \
3846 If you write an @code{asm} instruction with no outputs, GCC will know
3847 the instruction has side-effects and will not delete the instruction or
3848 move it outside of loops.
3850 The @code{volatile} keyword indicates that the instruction has
3851 important side-effects. GCC will not delete a volatile @code{asm} if
3852 it is reachable. (The instruction can still be deleted if GCC can
3853 prove that control-flow will never reach the location of the
3854 instruction.) In addition, GCC will not reschedule instructions
3855 across a volatile @code{asm} instruction. For example:
3858 *(volatile int *)addr = foo;
3859 asm volatile ("eieio" : : );
3863 Assume @code{addr} contains the address of a memory mapped device
3864 register. The PowerPC @code{eieio} instruction (Enforce In-order
3865 Execution of I/O) tells the CPU to make sure that the store to that
3866 device register happens before it issues any other I/O@.
3868 Note that even a volatile @code{asm} instruction can be moved in ways
3869 that appear insignificant to the compiler, such as across jump
3870 instructions. You can't expect a sequence of volatile @code{asm}
3871 instructions to remain perfectly consecutive. If you want consecutive
3872 output, use a single @code{asm}. Also, GCC will perform some
3873 optimizations across a volatile @code{asm} instruction; GCC does not
3874 ``forget everything'' when it encounters a volatile @code{asm}
3875 instruction the way some other compilers do.
3877 An @code{asm} instruction without any operands or clobbers (an ``old
3878 style'' @code{asm}) will be treated identically to a volatile
3879 @code{asm} instruction.
3881 It is a natural idea to look for a way to give access to the condition
3882 code left by the assembler instruction. However, when we attempted to
3883 implement this, we found no way to make it work reliably. The problem
3884 is that output operands might need reloading, which would result in
3885 additional following ``store'' instructions. On most machines, these
3886 instructions would alter the condition code before there was time to
3887 test it. This problem doesn't arise for ordinary ``test'' and
3888 ``compare'' instructions because they don't have any output operands.
3890 For reasons similar to those described above, it is not possible to give
3891 an assembler instruction access to the condition code left by previous
3894 If you are writing a header file that should be includable in ISO C
3895 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
3898 @subsection i386 floating point asm operands
3900 There are several rules on the usage of stack-like regs in
3901 asm_operands insns. These rules apply only to the operands that are
3906 Given a set of input regs that die in an asm_operands, it is
3907 necessary to know which are implicitly popped by the asm, and
3908 which must be explicitly popped by gcc.
3910 An input reg that is implicitly popped by the asm must be
3911 explicitly clobbered, unless it is constrained to match an
3915 For any input reg that is implicitly popped by an asm, it is
3916 necessary to know how to adjust the stack to compensate for the pop.
3917 If any non-popped input is closer to the top of the reg-stack than
3918 the implicitly popped reg, it would not be possible to know what the
3919 stack looked like---it's not clear how the rest of the stack ``slides
3922 All implicitly popped input regs must be closer to the top of
3923 the reg-stack than any input that is not implicitly popped.
3925 It is possible that if an input dies in an insn, reload might
3926 use the input reg for an output reload. Consider this example:
3929 asm ("foo" : "=t" (a) : "f" (b));
3932 This asm says that input B is not popped by the asm, and that
3933 the asm pushes a result onto the reg-stack, i.e., the stack is one
3934 deeper after the asm than it was before. But, it is possible that
3935 reload will think that it can use the same reg for both the input and
3936 the output, if input B dies in this insn.
3938 If any input operand uses the @code{f} constraint, all output reg
3939 constraints must use the @code{&} earlyclobber.
3941 The asm above would be written as
3944 asm ("foo" : "=&t" (a) : "f" (b));
3948 Some operands need to be in particular places on the stack. All
3949 output operands fall in this category---there is no other way to
3950 know which regs the outputs appear in unless the user indicates
3951 this in the constraints.
3953 Output operands must specifically indicate which reg an output
3954 appears in after an asm. @code{=f} is not allowed: the operand
3955 constraints must select a class with a single reg.
3958 Output operands may not be ``inserted'' between existing stack regs.
3959 Since no 387 opcode uses a read/write operand, all output operands
3960 are dead before the asm_operands, and are pushed by the asm_operands.
3961 It makes no sense to push anywhere but the top of the reg-stack.
3963 Output operands must start at the top of the reg-stack: output
3964 operands may not ``skip'' a reg.
3967 Some asm statements may need extra stack space for internal
3968 calculations. This can be guaranteed by clobbering stack registers
3969 unrelated to the inputs and outputs.
3973 Here are a couple of reasonable asms to want to write. This asm
3974 takes one input, which is internally popped, and produces two outputs.
3977 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
3980 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
3981 and replaces them with one output. The user must code the @code{st(1)}
3982 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
3985 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
3991 @section Controlling Names Used in Assembler Code
3992 @cindex assembler names for identifiers
3993 @cindex names used in assembler code
3994 @cindex identifiers, names in assembler code
3996 You can specify the name to be used in the assembler code for a C
3997 function or variable by writing the @code{asm} (or @code{__asm__})
3998 keyword after the declarator as follows:
4001 int foo asm ("myfoo") = 2;
4005 This specifies that the name to be used for the variable @code{foo} in
4006 the assembler code should be @samp{myfoo} rather than the usual
4009 On systems where an underscore is normally prepended to the name of a C
4010 function or variable, this feature allows you to define names for the
4011 linker that do not start with an underscore.
4013 It does not make sense to use this feature with a non-static local
4014 variable since such variables do not have assembler names. If you are
4015 trying to put the variable in a particular register, see @ref{Explicit
4016 Reg Vars}. GCC presently accepts such code with a warning, but will
4017 probably be changed to issue an error, rather than a warning, in the
4020 You cannot use @code{asm} in this way in a function @emph{definition}; but
4021 you can get the same effect by writing a declaration for the function
4022 before its definition and putting @code{asm} there, like this:
4025 extern func () asm ("FUNC");
4032 It is up to you to make sure that the assembler names you choose do not
4033 conflict with any other assembler symbols. Also, you must not use a
4034 register name; that would produce completely invalid assembler code. GCC
4035 does not as yet have the ability to store static variables in registers.
4036 Perhaps that will be added.
4038 @node Explicit Reg Vars
4039 @section Variables in Specified Registers
4040 @cindex explicit register variables
4041 @cindex variables in specified registers
4042 @cindex specified registers
4043 @cindex registers, global allocation
4045 GNU C allows you to put a few global variables into specified hardware
4046 registers. You can also specify the register in which an ordinary
4047 register variable should be allocated.
4051 Global register variables reserve registers throughout the program.
4052 This may be useful in programs such as programming language
4053 interpreters which have a couple of global variables that are accessed
4057 Local register variables in specific registers do not reserve the
4058 registers. The compiler's data flow analysis is capable of determining
4059 where the specified registers contain live values, and where they are
4060 available for other uses. Stores into local register variables may be deleted
4061 when they appear to be dead according to dataflow analysis. References
4062 to local register variables may be deleted or moved or simplified.
4064 These local variables are sometimes convenient for use with the extended
4065 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4066 output of the assembler instruction directly into a particular register.
4067 (This will work provided the register you specify fits the constraints
4068 specified for that operand in the @code{asm}.)
4076 @node Global Reg Vars
4077 @subsection Defining Global Register Variables
4078 @cindex global register variables
4079 @cindex registers, global variables in
4081 You can define a global register variable in GNU C like this:
4084 register int *foo asm ("a5");
4088 Here @code{a5} is the name of the register which should be used. Choose a
4089 register which is normally saved and restored by function calls on your
4090 machine, so that library routines will not clobber it.
4092 Naturally the register name is cpu-dependent, so you would need to
4093 conditionalize your program according to cpu type. The register
4094 @code{a5} would be a good choice on a 68000 for a variable of pointer
4095 type. On machines with register windows, be sure to choose a ``global''
4096 register that is not affected magically by the function call mechanism.
4098 In addition, operating systems on one type of cpu may differ in how they
4099 name the registers; then you would need additional conditionals. For
4100 example, some 68000 operating systems call this register @code{%a5}.
4102 Eventually there may be a way of asking the compiler to choose a register
4103 automatically, but first we need to figure out how it should choose and
4104 how to enable you to guide the choice. No solution is evident.
4106 Defining a global register variable in a certain register reserves that
4107 register entirely for this use, at least within the current compilation.
4108 The register will not be allocated for any other purpose in the functions
4109 in the current compilation. The register will not be saved and restored by
4110 these functions. Stores into this register are never deleted even if they
4111 would appear to be dead, but references may be deleted or moved or
4114 It is not safe to access the global register variables from signal
4115 handlers, or from more than one thread of control, because the system
4116 library routines may temporarily use the register for other things (unless
4117 you recompile them specially for the task at hand).
4119 @cindex @code{qsort}, and global register variables
4120 It is not safe for one function that uses a global register variable to
4121 call another such function @code{foo} by way of a third function
4122 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4123 different source file in which the variable wasn't declared). This is
4124 because @code{lose} might save the register and put some other value there.
4125 For example, you can't expect a global register variable to be available in
4126 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4127 might have put something else in that register. (If you are prepared to
4128 recompile @code{qsort} with the same global register variable, you can
4129 solve this problem.)
4131 If you want to recompile @code{qsort} or other source files which do not
4132 actually use your global register variable, so that they will not use that
4133 register for any other purpose, then it suffices to specify the compiler
4134 option @option{-ffixed-@var{reg}}. You need not actually add a global
4135 register declaration to their source code.
4137 A function which can alter the value of a global register variable cannot
4138 safely be called from a function compiled without this variable, because it
4139 could clobber the value the caller expects to find there on return.
4140 Therefore, the function which is the entry point into the part of the
4141 program that uses the global register variable must explicitly save and
4142 restore the value which belongs to its caller.
4144 @cindex register variable after @code{longjmp}
4145 @cindex global register after @code{longjmp}
4146 @cindex value after @code{longjmp}
4149 On most machines, @code{longjmp} will restore to each global register
4150 variable the value it had at the time of the @code{setjmp}. On some
4151 machines, however, @code{longjmp} will not change the value of global
4152 register variables. To be portable, the function that called @code{setjmp}
4153 should make other arrangements to save the values of the global register
4154 variables, and to restore them in a @code{longjmp}. This way, the same
4155 thing will happen regardless of what @code{longjmp} does.
4157 All global register variable declarations must precede all function
4158 definitions. If such a declaration could appear after function
4159 definitions, the declaration would be too late to prevent the register from
4160 being used for other purposes in the preceding functions.
4162 Global register variables may not have initial values, because an
4163 executable file has no means to supply initial contents for a register.
4165 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4166 registers, but certain library functions, such as @code{getwd}, as well
4167 as the subroutines for division and remainder, modify g3 and g4. g1 and
4168 g2 are local temporaries.
4170 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4171 Of course, it will not do to use more than a few of those.
4173 @node Local Reg Vars
4174 @subsection Specifying Registers for Local Variables
4175 @cindex local variables, specifying registers
4176 @cindex specifying registers for local variables
4177 @cindex registers for local variables
4179 You can define a local register variable with a specified register
4183 register int *foo asm ("a5");
4187 Here @code{a5} is the name of the register which should be used. Note
4188 that this is the same syntax used for defining global register
4189 variables, but for a local variable it would appear within a function.
4191 Naturally the register name is cpu-dependent, but this is not a
4192 problem, since specific registers are most often useful with explicit
4193 assembler instructions (@pxref{Extended Asm}). Both of these things
4194 generally require that you conditionalize your program according to
4197 In addition, operating systems on one type of cpu may differ in how they
4198 name the registers; then you would need additional conditionals. For
4199 example, some 68000 operating systems call this register @code{%a5}.
4201 Defining such a register variable does not reserve the register; it
4202 remains available for other uses in places where flow control determines
4203 the variable's value is not live. However, these registers are made
4204 unavailable for use in the reload pass; excessive use of this feature
4205 leaves the compiler too few available registers to compile certain
4208 This option does not guarantee that GCC will generate code that has
4209 this variable in the register you specify at all times. You may not
4210 code an explicit reference to this register in an @code{asm} statement
4211 and assume it will always refer to this variable.
4213 Stores into local register variables may be deleted when they appear to be dead
4214 according to dataflow analysis. References to local register variables may
4215 be deleted or moved or simplified.
4217 @node Alternate Keywords
4218 @section Alternate Keywords
4219 @cindex alternate keywords
4220 @cindex keywords, alternate
4222 @option{-ansi} and the various @option{-std} options disable certain
4223 keywords. This causes trouble when you want to use GNU C extensions, or
4224 a general-purpose header file that should be usable by all programs,
4225 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4226 @code{inline} are not available in programs compiled with
4227 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4228 program compiled with @option{-std=c99}). The ISO C99 keyword
4229 @code{restrict} is only available when @option{-std=gnu99} (which will
4230 eventually be the default) or @option{-std=c99} (or the equivalent
4231 @option{-std=iso9899:1999}) is used.
4233 The way to solve these problems is to put @samp{__} at the beginning and
4234 end of each problematical keyword. For example, use @code{__asm__}
4235 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4237 Other C compilers won't accept these alternative keywords; if you want to
4238 compile with another compiler, you can define the alternate keywords as
4239 macros to replace them with the customary keywords. It looks like this:
4247 @findex __extension__
4249 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4251 prevent such warnings within one expression by writing
4252 @code{__extension__} before the expression. @code{__extension__} has no
4253 effect aside from this.
4255 @node Incomplete Enums
4256 @section Incomplete @code{enum} Types
4258 You can define an @code{enum} tag without specifying its possible values.
4259 This results in an incomplete type, much like what you get if you write
4260 @code{struct foo} without describing the elements. A later declaration
4261 which does specify the possible values completes the type.
4263 You can't allocate variables or storage using the type while it is
4264 incomplete. However, you can work with pointers to that type.
4266 This extension may not be very useful, but it makes the handling of
4267 @code{enum} more consistent with the way @code{struct} and @code{union}
4270 This extension is not supported by GNU C++.
4272 @node Function Names
4273 @section Function Names as Strings
4274 @cindex @code{__FUNCTION__} identifier
4275 @cindex @code{__PRETTY_FUNCTION__} identifier
4276 @cindex @code{__func__} identifier
4278 GCC predefines two magic identifiers to hold the name of the current
4279 function. The identifier @code{__FUNCTION__} holds the name of the function
4280 as it appears in the source. The identifier @code{__PRETTY_FUNCTION__}
4281 holds the name of the function pretty printed in a language specific
4284 These names are always the same in a C function, but in a C++ function
4285 they may be different. For example, this program:
4289 extern int printf (char *, ...);
4296 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4297 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4315 __PRETTY_FUNCTION__ = int a::sub (int)
4318 The compiler automagically replaces the identifiers with a string
4319 literal containing the appropriate name. Thus, they are neither
4320 preprocessor macros, like @code{__FILE__} and @code{__LINE__}, nor
4321 variables. This means that they catenate with other string literals, and
4322 that they can be used to initialize char arrays. For example
4325 char here[] = "Function " __FUNCTION__ " in " __FILE__;
4328 On the other hand, @samp{#ifdef __FUNCTION__} does not have any special
4329 meaning inside a function, since the preprocessor does not do anything
4330 special with the identifier @code{__FUNCTION__}.
4332 Note that these semantics are deprecated, and that GCC 3.2 will handle
4333 @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} the same way as
4334 @code{__func__}. @code{__func__} is defined by the ISO standard C99:
4337 The identifier @code{__func__} is implicitly declared by the translator
4338 as if, immediately following the opening brace of each function
4339 definition, the declaration
4342 static const char __func__[] = "function-name";
4345 appeared, where function-name is the name of the lexically-enclosing
4346 function. This name is the unadorned name of the function.
4349 By this definition, @code{__func__} is a variable, not a string literal.
4350 In particular, @code{__func__} does not catenate with other string
4353 In @code{C++}, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} are
4354 variables, declared in the same way as @code{__func__}.
4356 @node Return Address
4357 @section Getting the Return or Frame Address of a Function
4359 These functions may be used to get information about the callers of a
4362 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4363 This function returns the return address of the current function, or of
4364 one of its callers. The @var{level} argument is number of frames to
4365 scan up the call stack. A value of @code{0} yields the return address
4366 of the current function, a value of @code{1} yields the return address
4367 of the caller of the current function, and so forth. When inlining
4368 the expected behavior is that the function will return the address of
4369 the function that will be returned to. To work around this behavior use
4370 the @code{noinline} function attribute.
4372 The @var{level} argument must be a constant integer.
4374 On some machines it may be impossible to determine the return address of
4375 any function other than the current one; in such cases, or when the top
4376 of the stack has been reached, this function will return @code{0} or a
4377 random value. In addition, @code{__builtin_frame_address} may be used
4378 to determine if the top of the stack has been reached.
4380 This function should only be used with a nonzero argument for debugging
4384 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4385 This function is similar to @code{__builtin_return_address}, but it
4386 returns the address of the function frame rather than the return address
4387 of the function. Calling @code{__builtin_frame_address} with a value of
4388 @code{0} yields the frame address of the current function, a value of
4389 @code{1} yields the frame address of the caller of the current function,
4392 The frame is the area on the stack which holds local variables and saved
4393 registers. The frame address is normally the address of the first word
4394 pushed on to the stack by the function. However, the exact definition
4395 depends upon the processor and the calling convention. If the processor
4396 has a dedicated frame pointer register, and the function has a frame,
4397 then @code{__builtin_frame_address} will return the value of the frame
4400 On some machines it may be impossible to determine the frame address of
4401 any function other than the current one; in such cases, or when the top
4402 of the stack has been reached, this function will return @code{0} if
4403 the first frame pointer is properly initialized by the startup code.
4405 This function should only be used with a nonzero argument for debugging
4409 @node Vector Extensions
4410 @section Using vector instructions through built-in functions
4412 On some targets, the instruction set contains SIMD vector instructions that
4413 operate on multiple values contained in one large register at the same time.
4414 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4417 The first step in using these extensions is to provide the necessary data
4418 types. This should be done using an appropriate @code{typedef}:
4421 typedef int v4si __attribute__ ((mode(V4SI)));
4424 The base type @code{int} is effectively ignored by the compiler, the
4425 actual properties of the new type @code{v4si} are defined by the
4426 @code{__attribute__}. It defines the machine mode to be used; for vector
4427 types these have the form @code{V@var{n}@var{B}}; @var{n} should be the
4428 number of elements in the vector, and @var{B} should be the base mode of the
4429 individual elements. The following can be used as base modes:
4433 An integer that is as wide as the smallest addressable unit, usually 8 bits.
4435 An integer, twice as wide as a QI mode integer, usually 16 bits.
4437 An integer, four times as wide as a QI mode integer, usually 32 bits.
4439 An integer, eight times as wide as a QI mode integer, usually 64 bits.
4441 A floating point value, as wide as a SI mode integer, usually 32 bits.
4443 A floating point value, as wide as a DI mode integer, usually 64 bits.
4446 Specifying a combination that is not valid for the current architecture
4447 will cause gcc to synthesize the instructions using a narrower mode.
4448 For example, if you specify a variable of type @code{V4SI} and your
4449 architecture does not allow for this specific SIMD type, gcc will
4450 produce code that uses 4 @code{SIs}.
4452 The types defined in this manner can be used with a subset of normal C
4453 operations. Currently, gcc will allow using the following operators on
4454 these types: @code{+, -, *, /, unary minus}@.
4456 The operations behave like C++ @code{valarrays}. Addition is defined as
4457 the addition of the corresponding elements of the operands. For
4458 example, in the code below, each of the 4 elements in @var{a} will be
4459 added to the corresponding 4 elements in @var{b} and the resulting
4460 vector will be stored in @var{c}.
4463 typedef int v4si __attribute__ ((mode(V4SI)));
4470 Subtraction, multiplication, and division operate in a similar manner.
4471 Likewise, the result of using the unary minus operator on a vector type
4472 is a vector whose elements are the negative value of the corresponding
4473 elements in the operand.
4475 You can declare variables and use them in function calls and returns, as
4476 well as in assignments and some casts. You can specify a vector type as
4477 a return type for a function. Vector types can also be used as function
4478 arguments. It is possible to cast from one vector type to another,
4479 provided they are of the same size (in fact, you can also cast vectors
4480 to and from other datatypes of the same size).
4482 You cannot operate between vectors of different lengths or different
4483 signness without a cast.
4485 A port that supports hardware vector operations, usually provides a set
4486 of built-in functions that can be used to operate on vectors. For
4487 example, a function to add two vectors and multiply the result by a
4488 third could look like this:
4491 v4si f (v4si a, v4si b, v4si c)
4493 v4si tmp = __builtin_addv4si (a, b);
4494 return __builtin_mulv4si (tmp, c);
4499 @node Other Builtins
4500 @section Other built-in functions provided by GCC
4501 @cindex built-in functions
4502 @findex __builtin_isgreater
4503 @findex __builtin_isgreaterequal
4504 @findex __builtin_isless
4505 @findex __builtin_islessequal
4506 @findex __builtin_islessgreater
4507 @findex __builtin_isunordered
4536 @findex fprintf_unlocked
4538 @findex fputs_unlocked
4550 @findex printf_unlocked
4572 GCC provides a large number of built-in functions other than the ones
4573 mentioned above. Some of these are for internal use in the processing
4574 of exceptions or variable-length argument lists and will not be
4575 documented here because they may change from time to time; we do not
4576 recommend general use of these functions.
4578 The remaining functions are provided for optimization purposes.
4580 @opindex fno-builtin
4581 GCC includes built-in versions of many of the functions in the standard
4582 C library. The versions prefixed with @code{__builtin_} will always be
4583 treated as having the same meaning as the C library function even if you
4584 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
4585 Many of these functions are only optimized in certain cases; if they are
4586 not optimized in a particular case, a call to the library function will
4591 The functions @code{abort}, @code{exit}, @code{_Exit} and @code{_exit}
4592 are recognized and presumed not to return, but otherwise are not built
4593 in. @code{_exit} is not recognized in strict ISO C mode (@option{-ansi},
4594 @option{-std=c89} or @option{-std=c99}). @code{_Exit} is not recognized in
4595 strict C89 mode (@option{-ansi} or @option{-std=c89}). All these functions
4596 have corresponding versions prefixed with @code{__builtin_}, which may be
4597 used even in strict C89 mode.
4599 Outside strict ISO C mode, the functions @code{alloca}, @code{bcmp},
4600 @code{bzero}, @code{index}, @code{rindex}, @code{ffs}, @code{fputs_unlocked},
4601 @code{printf_unlocked} and @code{fprintf_unlocked} may be handled as
4602 built-in functions. All these functions have corresponding versions
4603 prefixed with @code{__builtin_}, which may be used even in strict C89
4606 The ISO C99 functions @code{conj}, @code{conjf}, @code{conjl},
4607 @code{creal}, @code{crealf}, @code{creall}, @code{cimag}, @code{cimagf},
4608 @code{cimagl}, @code{llabs} and @code{imaxabs} are handled as built-in
4609 functions except in strict ISO C90 mode. There are also built-in
4610 versions of the ISO C99 functions @code{cosf}, @code{cosl},
4611 @code{expf}, @code{expl}, @code{fabsf}, @code{fabsl},
4612 @code{logf}, @code{logl}, @code{sinf}, @code{sinl}, @code{sqrtf}, and
4613 @code{sqrtl}, that are recognized in any mode since ISO C90 reserves
4614 these names for the purpose to which ISO C99 puts them. All these
4615 functions have corresponding versions prefixed with @code{__builtin_}.
4617 The ISO C90 functions @code{abs}, @code{cos}, @code{exp}, @code{fabs},
4618 @code{fprintf}, @code{fputs}, @code{labs}, @code{log},
4619 @code{memcmp}, @code{memcpy},
4620 @code{memset}, @code{printf}, @code{sin}, @code{sqrt}, @code{strcat},
4621 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
4622 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
4623 @code{strpbrk}, @code{strrchr}, @code{strspn}, and @code{strstr} are all
4624 recognized as built-in functions unless @option{-fno-builtin} is
4625 specified (or @option{-fno-builtin-@var{function}} is specified for an
4626 individual function). All of these functions have corresponding
4627 versions prefixed with @code{__builtin_}.
4629 GCC provides built-in versions of the ISO C99 floating point comparison
4630 macros that avoid raising exceptions for unordered operands. They have
4631 the same names as the standard macros ( @code{isgreater},
4632 @code{isgreaterequal}, @code{isless}, @code{islessequal},
4633 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
4634 prefixed. We intend for a library implementor to be able to simply
4635 @code{#define} each standard macro to its built-in equivalent.
4637 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
4639 You can use the built-in function @code{__builtin_types_compatible_p} to
4640 determine whether two types are the same.
4642 This built-in function returns 1 if the unqualified versions of the
4643 types @var{type1} and @var{type2} (which are types, not expressions) are
4644 compatible, 0 otherwise. The result of this built-in function can be
4645 used in integer constant expressions.
4647 This built-in function ignores top level qualifiers (e.g., @code{const},
4648 @code{volatile}). For example, @code{int} is equivalent to @code{const
4651 The type @code{int[]} and @code{int[5]} are compatible. On the other
4652 hand, @code{int} and @code{char *} are not compatible, even if the size
4653 of their types, on the particular architecture are the same. Also, the
4654 amount of pointer indirection is taken into account when determining
4655 similarity. Consequently, @code{short *} is not similar to
4656 @code{short **}. Furthermore, two types that are typedefed are
4657 considered compatible if their underlying types are compatible.
4659 An @code{enum} type is considered to be compatible with another
4660 @code{enum} type. For example, @code{enum @{foo, bar@}} is similar to
4661 @code{enum @{hot, dog@}}.
4663 You would typically use this function in code whose execution varies
4664 depending on the arguments' types. For example:
4670 if (__builtin_types_compatible_p (typeof (x), long double)) \
4671 tmp = foo_long_double (tmp); \
4672 else if (__builtin_types_compatible_p (typeof (x), double)) \
4673 tmp = foo_double (tmp); \
4674 else if (__builtin_types_compatible_p (typeof (x), float)) \
4675 tmp = foo_float (tmp); \
4682 @emph{Note:} This construct is only available for C.
4686 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
4688 You can use the built-in function @code{__builtin_choose_expr} to
4689 evaluate code depending on the value of a constant expression. This
4690 built-in function returns @var{exp1} if @var{const_exp}, which is a
4691 constant expression that must be able to be determined at compile time,
4692 is nonzero. Otherwise it returns 0.
4694 This built-in function is analogous to the @samp{? :} operator in C,
4695 except that the expression returned has its type unaltered by promotion
4696 rules. Also, the built-in function does not evaluate the expression
4697 that was not chosen. For example, if @var{const_exp} evaluates to true,
4698 @var{exp2} is not evaluated even if it has side-effects.
4700 This built-in function can return an lvalue if the chosen argument is an
4703 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
4704 type. Similarly, if @var{exp2} is returned, its return type is the same
4711 __builtin_choose_expr ( \
4712 __builtin_types_compatible_p (typeof (x), double), \
4714 __builtin_choose_expr ( \
4715 __builtin_types_compatible_p (typeof (x), float), \
4717 /* @r{The void expression results in a compile-time error} \
4718 @r{when assigning the result to something.} */ \
4722 @emph{Note:} This construct is only available for C. Furthermore, the
4723 unused expression (@var{exp1} or @var{exp2} depending on the value of
4724 @var{const_exp}) may still generate syntax errors. This may change in
4729 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
4730 You can use the built-in function @code{__builtin_constant_p} to
4731 determine if a value is known to be constant at compile-time and hence
4732 that GCC can perform constant-folding on expressions involving that
4733 value. The argument of the function is the value to test. The function
4734 returns the integer 1 if the argument is known to be a compile-time
4735 constant and 0 if it is not known to be a compile-time constant. A
4736 return of 0 does not indicate that the value is @emph{not} a constant,
4737 but merely that GCC cannot prove it is a constant with the specified
4738 value of the @option{-O} option.
4740 You would typically use this function in an embedded application where
4741 memory was a critical resource. If you have some complex calculation,
4742 you may want it to be folded if it involves constants, but need to call
4743 a function if it does not. For example:
4746 #define Scale_Value(X) \
4747 (__builtin_constant_p (X) \
4748 ? ((X) * SCALE + OFFSET) : Scale (X))
4751 You may use this built-in function in either a macro or an inline
4752 function. However, if you use it in an inlined function and pass an
4753 argument of the function as the argument to the built-in, GCC will
4754 never return 1 when you call the inline function with a string constant
4755 or compound literal (@pxref{Compound Literals}) and will not return 1
4756 when you pass a constant numeric value to the inline function unless you
4757 specify the @option{-O} option.
4759 You may also use @code{__builtin_constant_p} in initializers for static
4760 data. For instance, you can write
4763 static const int table[] = @{
4764 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
4770 This is an acceptable initializer even if @var{EXPRESSION} is not a
4771 constant expression. GCC must be more conservative about evaluating the
4772 built-in in this case, because it has no opportunity to perform
4775 Previous versions of GCC did not accept this built-in in data
4776 initializers. The earliest version where it is completely safe is
4780 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
4781 @opindex fprofile-arcs
4782 You may use @code{__builtin_expect} to provide the compiler with
4783 branch prediction information. In general, you should prefer to
4784 use actual profile feedback for this (@option{-fprofile-arcs}), as
4785 programmers are notoriously bad at predicting how their programs
4786 actually perform. However, there are applications in which this
4787 data is hard to collect.
4789 The return value is the value of @var{exp}, which should be an
4790 integral expression. The value of @var{c} must be a compile-time
4791 constant. The semantics of the built-in are that it is expected
4792 that @var{exp} == @var{c}. For example:
4795 if (__builtin_expect (x, 0))
4800 would indicate that we do not expect to call @code{foo}, since
4801 we expect @code{x} to be zero. Since you are limited to integral
4802 expressions for @var{exp}, you should use constructions such as
4805 if (__builtin_expect (ptr != NULL, 1))
4810 when testing pointer or floating-point values.
4813 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
4814 This function is used to minimize cache-miss latency by moving data into
4815 a cache before it is accessed.
4816 You can insert calls to @code{__builtin_prefetch} into code for which
4817 you know addresses of data in memory that is likely to be accessed soon.
4818 If the target supports them, data prefetch instructions will be generated.
4819 If the prefetch is done early enough before the access then the data will
4820 be in the cache by the time it is accessed.
4822 The value of @var{addr} is the address of the memory to prefetch.
4823 There are two optional arguments, @var{rw} and @var{locality}.
4824 The value of @var{rw} is a compile-time constant one or zero; one
4825 means that the prefetch is preparing for a write to the memory address
4826 and zero, the default, means that the prefetch is preparing for a read.
4827 The value @var{locality} must be a compile-time constant integer between
4828 zero and three. A value of zero means that the data has no temporal
4829 locality, so it need not be left in the cache after the access. A value
4830 of three means that the data has a high degree of temporal locality and
4831 should be left in all levels of cache possible. Values of one and two
4832 mean, respectively, a low or moderate degree of temporal locality. The
4836 for (i = 0; i < n; i++)
4839 __builtin_prefetch (&a[i+j], 1, 1);
4840 __builtin_prefetch (&b[i+j], 0, 1);
4845 Data prefetch does not generate faults if @var{addr} is invalid, but
4846 the address expression itself must be valid. For example, a prefetch
4847 of @code{p->next} will not fault if @code{p->next} is not a valid
4848 address, but evaluation will fault if @code{p} is not a valid address.
4850 If the target does not support data prefetch, the address expression
4851 is evaluated if it includes side effects but no other code is generated
4852 and GCC does not issue a warning.
4855 @deftypefn {Built-in Function} double __builtin_huge_val (void)
4856 Returns a positive infinity, if supported by the floating-point format,
4857 else @code{DBL_MAX}. This function is suitable for implementing the
4858 ISO C macro @code{HUGE_VAL}.
4861 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
4862 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
4865 @deftypefn {Built-in Function} long double __builtin_huge_vall (void)
4866 Similar to @code{__builtin_huge_val}, except the return
4867 type is @code{long double}.
4870 @deftypefn {Built-in Function} double __builtin_inf (void)
4871 Similar to @code{__builtin_huge_val}, except a warning is generated
4872 if the target floating-point format does not support infinities.
4873 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
4876 @deftypefn {Built-in Function} float __builtin_inff (void)
4877 Similar to @code{__builtin_inf}, except the return type is @code{float}.
4880 @deftypefn {Built-in Function} long double __builtin_infl (void)
4881 Similar to @code{__builtin_inf}, except the return
4882 type is @code{long double}.
4885 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
4886 This is an implementation of the ISO C99 function @code{nan}.
4888 Since ISO C99 defines this function in terms of @code{strtod}, which we
4889 do not implement, a desription of the parsing is in order. The string
4890 is parsed as by @code{strtol}; that is, the base is recognized by
4891 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
4892 in the significand such that the least significant bit of the number
4893 is at the least significant bit of the significand. The number is
4894 truncated to fit the significand field provided. The significand is
4895 forced to be a quiet NaN.
4897 This function, if given a string literal, is evaluated early enough
4898 that it is considered a compile-time constant.
4901 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
4902 Similar to @code{__builtin_nan}, except the return type is @code{float}.
4905 @deftypefn {Built-in Function} long double __builtin_nanl (const char *str)
4906 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
4909 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
4910 Similar to @code{__builtin_nan}, except the significand is forced
4911 to be a signaling NaN. The @code{nans} function is proposed by
4912 @uref{http://std.dkuug.dk/JTC1/SC22/WG14/www/docs/n965.htm,,WG14 N965}.
4915 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
4916 Similar to @code{__builtin_nans}, except the return type is @code{float}.
4919 @deftypefn {Built-in Function} long double __builtin_nansl (const char *str)
4920 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
4923 @node Target Builtins
4924 @section Built-in Functions Specific to Particular Target Machines
4926 On some target machines, GCC supports many built-in functions specific
4927 to those machines. Generally these generate calls to specific machine
4928 instructions, but allow the compiler to schedule those calls.
4931 * Alpha Built-in Functions::
4932 * X86 Built-in Functions::
4933 * PowerPC AltiVec Built-in Functions::
4936 @node Alpha Built-in Functions
4937 @subsection Alpha Built-in Functions
4939 These built-in functions are available for the Alpha family of
4940 processors, depending on the command-line switches used.
4942 The following built-in functions are always available. They
4943 all generate the machine instruction that is part of the name.
4946 long __builtin_alpha_implver (void)
4947 long __builtin_alpha_rpcc (void)
4948 long __builtin_alpha_amask (long)
4949 long __builtin_alpha_cmpbge (long, long)
4950 long __builtin_alpha_extbl (long, long)
4951 long __builtin_alpha_extwl (long, long)
4952 long __builtin_alpha_extll (long, long)
4953 long __builtin_alpha_extql (long, long)
4954 long __builtin_alpha_extwh (long, long)
4955 long __builtin_alpha_extlh (long, long)
4956 long __builtin_alpha_extqh (long, long)
4957 long __builtin_alpha_insbl (long, long)
4958 long __builtin_alpha_inswl (long, long)
4959 long __builtin_alpha_insll (long, long)
4960 long __builtin_alpha_insql (long, long)
4961 long __builtin_alpha_inswh (long, long)
4962 long __builtin_alpha_inslh (long, long)
4963 long __builtin_alpha_insqh (long, long)
4964 long __builtin_alpha_mskbl (long, long)
4965 long __builtin_alpha_mskwl (long, long)
4966 long __builtin_alpha_mskll (long, long)
4967 long __builtin_alpha_mskql (long, long)
4968 long __builtin_alpha_mskwh (long, long)
4969 long __builtin_alpha_msklh (long, long)
4970 long __builtin_alpha_mskqh (long, long)
4971 long __builtin_alpha_umulh (long, long)
4972 long __builtin_alpha_zap (long, long)
4973 long __builtin_alpha_zapnot (long, long)
4976 The following built-in functions are always with @option{-mmax}
4977 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
4978 later. They all generate the machine instruction that is part
4982 long __builtin_alpha_pklb (long)
4983 long __builtin_alpha_pkwb (long)
4984 long __builtin_alpha_unpkbl (long)
4985 long __builtin_alpha_unpkbw (long)
4986 long __builtin_alpha_minub8 (long, long)
4987 long __builtin_alpha_minsb8 (long, long)
4988 long __builtin_alpha_minuw4 (long, long)
4989 long __builtin_alpha_minsw4 (long, long)
4990 long __builtin_alpha_maxub8 (long, long)
4991 long __builtin_alpha_maxsb8 (long, long)
4992 long __builtin_alpha_maxuw4 (long, long)
4993 long __builtin_alpha_maxsw4 (long, long)
4994 long __builtin_alpha_perr (long, long)
4997 The following built-in functions are always with @option{-mcix}
4998 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
4999 later. They all generate the machine instruction that is part
5003 long __builtin_alpha_cttz (long)
5004 long __builtin_alpha_ctlz (long)
5005 long __builtin_alpha_ctpop (long)
5008 The following builtins are available on systems that use the OSF/1
5009 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5010 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5011 @code{rdval} and @code{wrval}.
5014 void *__builtin_thread_pointer (void)
5015 void __builtin_set_thread_pointer (void *)
5018 @node X86 Built-in Functions
5019 @subsection X86 Built-in Functions
5021 These built-in functions are available for the i386 and x86-64 family
5022 of computers, depending on the command-line switches used.
5024 The following machine modes are available for use with MMX built-in functions
5025 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5026 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5027 vector of eight 8-bit integers. Some of the built-in functions operate on
5028 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5030 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5031 of two 32-bit floating point values.
5033 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5034 floating point values. Some instructions use a vector of four 32-bit
5035 integers, these use @code{V4SI}. Finally, some instructions operate on an
5036 entire vector register, interpreting it as a 128-bit integer, these use mode
5039 The following built-in functions are made available by @option{-mmmx}.
5040 All of them generate the machine instruction that is part of the name.
5043 v8qi __builtin_ia32_paddb (v8qi, v8qi)
5044 v4hi __builtin_ia32_paddw (v4hi, v4hi)
5045 v2si __builtin_ia32_paddd (v2si, v2si)
5046 v8qi __builtin_ia32_psubb (v8qi, v8qi)
5047 v4hi __builtin_ia32_psubw (v4hi, v4hi)
5048 v2si __builtin_ia32_psubd (v2si, v2si)
5049 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5050 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5051 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5052 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5053 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5054 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5055 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5056 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5057 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5058 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5059 di __builtin_ia32_pand (di, di)
5060 di __builtin_ia32_pandn (di,di)
5061 di __builtin_ia32_por (di, di)
5062 di __builtin_ia32_pxor (di, di)
5063 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5064 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5065 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5066 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5067 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5068 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5069 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5070 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5071 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5072 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5073 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5074 v2si __builtin_ia32_punpckldq (v2si, v2si)
5075 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5076 v4hi __builtin_ia32_packssdw (v2si, v2si)
5077 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5080 The following built-in functions are made available either with
5081 @option{-msse}, or with a combination of @option{-m3dnow} and
5082 @option{-march=athlon}. All of them generate the machine
5083 instruction that is part of the name.
5086 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5087 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5088 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5089 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5090 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5091 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5092 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5093 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5094 int __builtin_ia32_pextrw (v4hi, int)
5095 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5096 int __builtin_ia32_pmovmskb (v8qi)
5097 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5098 void __builtin_ia32_movntq (di *, di)
5099 void __builtin_ia32_sfence (void)
5102 The following built-in functions are available when @option{-msse} is used.
5103 All of them generate the machine instruction that is part of the name.
5106 int __builtin_ia32_comieq (v4sf, v4sf)
5107 int __builtin_ia32_comineq (v4sf, v4sf)
5108 int __builtin_ia32_comilt (v4sf, v4sf)
5109 int __builtin_ia32_comile (v4sf, v4sf)
5110 int __builtin_ia32_comigt (v4sf, v4sf)
5111 int __builtin_ia32_comige (v4sf, v4sf)
5112 int __builtin_ia32_ucomieq (v4sf, v4sf)
5113 int __builtin_ia32_ucomineq (v4sf, v4sf)
5114 int __builtin_ia32_ucomilt (v4sf, v4sf)
5115 int __builtin_ia32_ucomile (v4sf, v4sf)
5116 int __builtin_ia32_ucomigt (v4sf, v4sf)
5117 int __builtin_ia32_ucomige (v4sf, v4sf)
5118 v4sf __builtin_ia32_addps (v4sf, v4sf)
5119 v4sf __builtin_ia32_subps (v4sf, v4sf)
5120 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5121 v4sf __builtin_ia32_divps (v4sf, v4sf)
5122 v4sf __builtin_ia32_addss (v4sf, v4sf)
5123 v4sf __builtin_ia32_subss (v4sf, v4sf)
5124 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5125 v4sf __builtin_ia32_divss (v4sf, v4sf)
5126 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5127 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5128 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5129 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5130 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5131 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5132 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5133 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5134 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5135 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5136 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5137 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5138 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5139 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5140 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5141 v4si __builtin_ia32_cmpgtss (v4sf, v4sf)
5142 v4si __builtin_ia32_cmpgess (v4sf, v4sf)
5143 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5144 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5145 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
5146 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
5147 v4si __builtin_ia32_cmpngtss (v4sf, v4sf)
5148 v4si __builtin_ia32_cmpngess (v4sf, v4sf)
5149 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
5150 v4sf __builtin_ia32_maxps (v4sf, v4sf)
5151 v4sf __builtin_ia32_maxss (v4sf, v4sf)
5152 v4sf __builtin_ia32_minps (v4sf, v4sf)
5153 v4sf __builtin_ia32_minss (v4sf, v4sf)
5154 v4sf __builtin_ia32_andps (v4sf, v4sf)
5155 v4sf __builtin_ia32_andnps (v4sf, v4sf)
5156 v4sf __builtin_ia32_orps (v4sf, v4sf)
5157 v4sf __builtin_ia32_xorps (v4sf, v4sf)
5158 v4sf __builtin_ia32_movss (v4sf, v4sf)
5159 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
5160 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
5161 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
5162 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
5163 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
5164 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
5165 v2si __builtin_ia32_cvtps2pi (v4sf)
5166 int __builtin_ia32_cvtss2si (v4sf)
5167 v2si __builtin_ia32_cvttps2pi (v4sf)
5168 int __builtin_ia32_cvttss2si (v4sf)
5169 v4sf __builtin_ia32_rcpps (v4sf)
5170 v4sf __builtin_ia32_rsqrtps (v4sf)
5171 v4sf __builtin_ia32_sqrtps (v4sf)
5172 v4sf __builtin_ia32_rcpss (v4sf)
5173 v4sf __builtin_ia32_rsqrtss (v4sf)
5174 v4sf __builtin_ia32_sqrtss (v4sf)
5175 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
5176 void __builtin_ia32_movntps (float *, v4sf)
5177 int __builtin_ia32_movmskps (v4sf)
5180 The following built-in functions are available when @option{-msse} is used.
5183 @item v4sf __builtin_ia32_loadaps (float *)
5184 Generates the @code{movaps} machine instruction as a load from memory.
5185 @item void __builtin_ia32_storeaps (float *, v4sf)
5186 Generates the @code{movaps} machine instruction as a store to memory.
5187 @item v4sf __builtin_ia32_loadups (float *)
5188 Generates the @code{movups} machine instruction as a load from memory.
5189 @item void __builtin_ia32_storeups (float *, v4sf)
5190 Generates the @code{movups} machine instruction as a store to memory.
5191 @item v4sf __builtin_ia32_loadsss (float *)
5192 Generates the @code{movss} machine instruction as a load from memory.
5193 @item void __builtin_ia32_storess (float *, v4sf)
5194 Generates the @code{movss} machine instruction as a store to memory.
5195 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
5196 Generates the @code{movhps} machine instruction as a load from memory.
5197 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
5198 Generates the @code{movlps} machine instruction as a load from memory
5199 @item void __builtin_ia32_storehps (v4sf, v2si *)
5200 Generates the @code{movhps} machine instruction as a store to memory.
5201 @item void __builtin_ia32_storelps (v4sf, v2si *)
5202 Generates the @code{movlps} machine instruction as a store to memory.
5205 The following built-in functions are available when @option{-m3dnow} is used.
5206 All of them generate the machine instruction that is part of the name.
5209 void __builtin_ia32_femms (void)
5210 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
5211 v2si __builtin_ia32_pf2id (v2sf)
5212 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
5213 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
5214 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
5215 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
5216 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
5217 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
5218 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
5219 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
5220 v2sf __builtin_ia32_pfrcp (v2sf)
5221 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
5222 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
5223 v2sf __builtin_ia32_pfrsqrt (v2sf)
5224 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
5225 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
5226 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
5227 v2sf __builtin_ia32_pi2fd (v2si)
5228 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
5231 The following built-in functions are available when both @option{-m3dnow}
5232 and @option{-march=athlon} are used. All of them generate the machine
5233 instruction that is part of the name.
5236 v2si __builtin_ia32_pf2iw (v2sf)
5237 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
5238 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
5239 v2sf __builtin_ia32_pi2fw (v2si)
5240 v2sf __builtin_ia32_pswapdsf (v2sf)
5241 v2si __builtin_ia32_pswapdsi (v2si)
5244 @node PowerPC AltiVec Built-in Functions
5245 @subsection PowerPC AltiVec Built-in Functions
5247 These built-in functions are available for the PowerPC family
5248 of computers, depending on the command-line switches used.
5250 The following machine modes are available for use with AltiVec built-in
5251 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
5252 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
5253 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
5254 @code{V16QI} for a vector of sixteen 8-bit integers.
5256 The following functions are made available by including
5257 @code{<altivec.h>} and using @option{-maltivec} and
5258 @option{-mabi=altivec}. The functions implement the functionality
5259 described in Motorola's AltiVec Programming Interface Manual.
5261 There are a few differences from Motorola's documentation and GCC's
5262 implementation. Vector constants are done with curly braces (not
5263 parentheses). Vector initializers require no casts if the vector
5264 constant is of the same type as the variable it is initializing. The
5265 @code{vector bool} type is deprecated and will be discontinued in
5266 further revisions. Use @code{vector signed} instead. If @code{signed}
5267 or @code{unsigned} is omitted, the vector type will default to
5268 @code{signed}. Lastly, all overloaded functions are implemented with macros
5269 for the C implementation. So code the following example will not work:
5272 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
5275 Since vec_add is a macro, the vector constant in the above example will
5276 be treated as four different arguments. Wrap the entire argument in
5277 parentheses for this to work. The C++ implementation does not use
5280 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
5281 Internally, GCC uses built-in functions to achieve the functionality in
5282 the aforementioned header file, but they are not supported and are
5283 subject to change without notice.
5286 vector signed char vec_abs (vector signed char, vector signed char);
5287 vector signed short vec_abs (vector signed short, vector signed short);
5288 vector signed int vec_abs (vector signed int, vector signed int);
5289 vector signed float vec_abs (vector signed float, vector signed float);
5291 vector signed char vec_abss (vector signed char, vector signed char);
5292 vector signed short vec_abss (vector signed short, vector signed short);
5294 vector signed char vec_add (vector signed char, vector signed char);
5295 vector unsigned char vec_add (vector signed char, vector unsigned char);
5297 vector unsigned char vec_add (vector unsigned char, vector signed char);
5299 vector unsigned char vec_add (vector unsigned char,
5300 vector unsigned char);
5301 vector signed short vec_add (vector signed short, vector signed short);
5302 vector unsigned short vec_add (vector signed short,
5303 vector unsigned short);
5304 vector unsigned short vec_add (vector unsigned short,
5305 vector signed short);
5306 vector unsigned short vec_add (vector unsigned short,
5307 vector unsigned short);
5308 vector signed int vec_add (vector signed int, vector signed int);
5309 vector unsigned int vec_add (vector signed int, vector unsigned int);
5310 vector unsigned int vec_add (vector unsigned int, vector signed int);
5311 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
5312 vector float vec_add (vector float, vector float);
5314 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
5316 vector unsigned char vec_adds (vector signed char,
5317 vector unsigned char);
5318 vector unsigned char vec_adds (vector unsigned char,
5319 vector signed char);
5320 vector unsigned char vec_adds (vector unsigned char,
5321 vector unsigned char);
5322 vector signed char vec_adds (vector signed char, vector signed char);
5323 vector unsigned short vec_adds (vector signed short,
5324 vector unsigned short);
5325 vector unsigned short vec_adds (vector unsigned short,
5326 vector signed short);
5327 vector unsigned short vec_adds (vector unsigned short,
5328 vector unsigned short);
5329 vector signed short vec_adds (vector signed short, vector signed short);
5331 vector unsigned int vec_adds (vector signed int, vector unsigned int);
5332 vector unsigned int vec_adds (vector unsigned int, vector signed int);
5333 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
5335 vector signed int vec_adds (vector signed int, vector signed int);
5337 vector float vec_and (vector float, vector float);
5338 vector float vec_and (vector float, vector signed int);
5339 vector float vec_and (vector signed int, vector float);
5340 vector signed int vec_and (vector signed int, vector signed int);
5341 vector unsigned int vec_and (vector signed int, vector unsigned int);
5342 vector unsigned int vec_and (vector unsigned int, vector signed int);
5343 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
5344 vector signed short vec_and (vector signed short, vector signed short);
5345 vector unsigned short vec_and (vector signed short,
5346 vector unsigned short);
5347 vector unsigned short vec_and (vector unsigned short,
5348 vector signed short);
5349 vector unsigned short vec_and (vector unsigned short,
5350 vector unsigned short);
5351 vector signed char vec_and (vector signed char, vector signed char);
5352 vector unsigned char vec_and (vector signed char, vector unsigned char);
5354 vector unsigned char vec_and (vector unsigned char, vector signed char);
5356 vector unsigned char vec_and (vector unsigned char,
5357 vector unsigned char);
5359 vector float vec_andc (vector float, vector float);
5360 vector float vec_andc (vector float, vector signed int);
5361 vector float vec_andc (vector signed int, vector float);
5362 vector signed int vec_andc (vector signed int, vector signed int);
5363 vector unsigned int vec_andc (vector signed int, vector unsigned int);
5364 vector unsigned int vec_andc (vector unsigned int, vector signed int);
5365 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
5367 vector signed short vec_andc (vector signed short, vector signed short);
5369 vector unsigned short vec_andc (vector signed short,
5370 vector unsigned short);
5371 vector unsigned short vec_andc (vector unsigned short,
5372 vector signed short);
5373 vector unsigned short vec_andc (vector unsigned short,
5374 vector unsigned short);
5375 vector signed char vec_andc (vector signed char, vector signed char);
5376 vector unsigned char vec_andc (vector signed char,
5377 vector unsigned char);
5378 vector unsigned char vec_andc (vector unsigned char,
5379 vector signed char);
5380 vector unsigned char vec_andc (vector unsigned char,
5381 vector unsigned char);
5383 vector unsigned char vec_avg (vector unsigned char,
5384 vector unsigned char);
5385 vector signed char vec_avg (vector signed char, vector signed char);
5386 vector unsigned short vec_avg (vector unsigned short,
5387 vector unsigned short);
5388 vector signed short vec_avg (vector signed short, vector signed short);
5389 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
5390 vector signed int vec_avg (vector signed int, vector signed int);
5392 vector float vec_ceil (vector float);
5394 vector signed int vec_cmpb (vector float, vector float);
5396 vector signed char vec_cmpeq (vector signed char, vector signed char);
5397 vector signed char vec_cmpeq (vector unsigned char,
5398 vector unsigned char);
5399 vector signed short vec_cmpeq (vector signed short,
5400 vector signed short);
5401 vector signed short vec_cmpeq (vector unsigned short,
5402 vector unsigned short);
5403 vector signed int vec_cmpeq (vector signed int, vector signed int);
5404 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
5405 vector signed int vec_cmpeq (vector float, vector float);
5407 vector signed int vec_cmpge (vector float, vector float);
5409 vector signed char vec_cmpgt (vector unsigned char,
5410 vector unsigned char);
5411 vector signed char vec_cmpgt (vector signed char, vector signed char);
5412 vector signed short vec_cmpgt (vector unsigned short,
5413 vector unsigned short);
5414 vector signed short vec_cmpgt (vector signed short,
5415 vector signed short);
5416 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
5417 vector signed int vec_cmpgt (vector signed int, vector signed int);
5418 vector signed int vec_cmpgt (vector float, vector float);
5420 vector signed int vec_cmple (vector float, vector float);
5422 vector signed char vec_cmplt (vector unsigned char,
5423 vector unsigned char);
5424 vector signed char vec_cmplt (vector signed char, vector signed char);
5425 vector signed short vec_cmplt (vector unsigned short,
5426 vector unsigned short);
5427 vector signed short vec_cmplt (vector signed short,
5428 vector signed short);
5429 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
5430 vector signed int vec_cmplt (vector signed int, vector signed int);
5431 vector signed int vec_cmplt (vector float, vector float);
5433 vector float vec_ctf (vector unsigned int, const char);
5434 vector float vec_ctf (vector signed int, const char);
5436 vector signed int vec_cts (vector float, const char);
5438 vector unsigned int vec_ctu (vector float, const char);
5440 void vec_dss (const char);
5442 void vec_dssall (void);
5444 void vec_dst (void *, int, const char);
5446 void vec_dstst (void *, int, const char);
5448 void vec_dststt (void *, int, const char);
5450 void vec_dstt (void *, int, const char);
5452 vector float vec_expte (vector float, vector float);
5454 vector float vec_floor (vector float, vector float);
5456 vector float vec_ld (int, vector float *);
5457 vector float vec_ld (int, float *):
5458 vector signed int vec_ld (int, int *);
5459 vector signed int vec_ld (int, vector signed int *);
5460 vector unsigned int vec_ld (int, vector unsigned int *);
5461 vector unsigned int vec_ld (int, unsigned int *);
5462 vector signed short vec_ld (int, short *, vector signed short *);
5463 vector unsigned short vec_ld (int, unsigned short *,
5464 vector unsigned short *);
5465 vector signed char vec_ld (int, signed char *);
5466 vector signed char vec_ld (int, vector signed char *);
5467 vector unsigned char vec_ld (int, unsigned char *);
5468 vector unsigned char vec_ld (int, vector unsigned char *);
5470 vector signed char vec_lde (int, signed char *);
5471 vector unsigned char vec_lde (int, unsigned char *);
5472 vector signed short vec_lde (int, short *);
5473 vector unsigned short vec_lde (int, unsigned short *);
5474 vector float vec_lde (int, float *);
5475 vector signed int vec_lde (int, int *);
5476 vector unsigned int vec_lde (int, unsigned int *);
5478 void float vec_ldl (int, float *);
5479 void float vec_ldl (int, vector float *);
5480 void signed int vec_ldl (int, vector signed int *);
5481 void signed int vec_ldl (int, int *);
5482 void unsigned int vec_ldl (int, unsigned int *);
5483 void unsigned int vec_ldl (int, vector unsigned int *);
5484 void signed short vec_ldl (int, vector signed short *);
5485 void signed short vec_ldl (int, short *);
5486 void unsigned short vec_ldl (int, vector unsigned short *);
5487 void unsigned short vec_ldl (int, unsigned short *);
5488 void signed char vec_ldl (int, vector signed char *);
5489 void signed char vec_ldl (int, signed char *);
5490 void unsigned char vec_ldl (int, vector unsigned char *);
5491 void unsigned char vec_ldl (int, unsigned char *);
5493 vector float vec_loge (vector float);
5495 vector unsigned char vec_lvsl (int, void *, int *);
5497 vector unsigned char vec_lvsr (int, void *, int *);
5499 vector float vec_madd (vector float, vector float, vector float);
5501 vector signed short vec_madds (vector signed short, vector signed short,
5502 vector signed short);
5504 vector unsigned char vec_max (vector signed char, vector unsigned char);
5506 vector unsigned char vec_max (vector unsigned char, vector signed char);
5508 vector unsigned char vec_max (vector unsigned char,
5509 vector unsigned char);
5510 vector signed char vec_max (vector signed char, vector signed char);
5511 vector unsigned short vec_max (vector signed short,
5512 vector unsigned short);
5513 vector unsigned short vec_max (vector unsigned short,
5514 vector signed short);
5515 vector unsigned short vec_max (vector unsigned short,
5516 vector unsigned short);
5517 vector signed short vec_max (vector signed short, vector signed short);
5518 vector unsigned int vec_max (vector signed int, vector unsigned int);
5519 vector unsigned int vec_max (vector unsigned int, vector signed int);
5520 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
5521 vector signed int vec_max (vector signed int, vector signed int);
5522 vector float vec_max (vector float, vector float);
5524 vector signed char vec_mergeh (vector signed char, vector signed char);
5525 vector unsigned char vec_mergeh (vector unsigned char,
5526 vector unsigned char);
5527 vector signed short vec_mergeh (vector signed short,
5528 vector signed short);
5529 vector unsigned short vec_mergeh (vector unsigned short,
5530 vector unsigned short);
5531 vector float vec_mergeh (vector float, vector float);
5532 vector signed int vec_mergeh (vector signed int, vector signed int);
5533 vector unsigned int vec_mergeh (vector unsigned int,
5534 vector unsigned int);
5536 vector signed char vec_mergel (vector signed char, vector signed char);
5537 vector unsigned char vec_mergel (vector unsigned char,
5538 vector unsigned char);
5539 vector signed short vec_mergel (vector signed short,
5540 vector signed short);
5541 vector unsigned short vec_mergel (vector unsigned short,
5542 vector unsigned short);
5543 vector float vec_mergel (vector float, vector float);
5544 vector signed int vec_mergel (vector signed int, vector signed int);
5545 vector unsigned int vec_mergel (vector unsigned int,
5546 vector unsigned int);
5548 vector unsigned short vec_mfvscr (void);
5550 vector unsigned char vec_min (vector signed char, vector unsigned char);
5552 vector unsigned char vec_min (vector unsigned char, vector signed char);
5554 vector unsigned char vec_min (vector unsigned char,
5555 vector unsigned char);
5556 vector signed char vec_min (vector signed char, vector signed char);
5557 vector unsigned short vec_min (vector signed short,
5558 vector unsigned short);
5559 vector unsigned short vec_min (vector unsigned short,
5560 vector signed short);
5561 vector unsigned short vec_min (vector unsigned short,
5562 vector unsigned short);
5563 vector signed short vec_min (vector signed short, vector signed short);
5564 vector unsigned int vec_min (vector signed int, vector unsigned int);
5565 vector unsigned int vec_min (vector unsigned int, vector signed int);
5566 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
5567 vector signed int vec_min (vector signed int, vector signed int);
5568 vector float vec_min (vector float, vector float);
5570 vector signed short vec_mladd (vector signed short, vector signed short,
5571 vector signed short);
5572 vector signed short vec_mladd (vector signed short,
5573 vector unsigned short,
5574 vector unsigned short);
5575 vector signed short vec_mladd (vector unsigned short,
5576 vector signed short,
5577 vector signed short);
5578 vector unsigned short vec_mladd (vector unsigned short,
5579 vector unsigned short,
5580 vector unsigned short);
5582 vector signed short vec_mradds (vector signed short,
5583 vector signed short,
5584 vector signed short);
5586 vector unsigned int vec_msum (vector unsigned char,
5587 vector unsigned char,
5588 vector unsigned int);
5589 vector signed int vec_msum (vector signed char, vector unsigned char,
5591 vector unsigned int vec_msum (vector unsigned short,
5592 vector unsigned short,
5593 vector unsigned int);
5594 vector signed int vec_msum (vector signed short, vector signed short,
5597 vector unsigned int vec_msums (vector unsigned short,
5598 vector unsigned short,
5599 vector unsigned int);
5600 vector signed int vec_msums (vector signed short, vector signed short,
5603 void vec_mtvscr (vector signed int);
5604 void vec_mtvscr (vector unsigned int);
5605 void vec_mtvscr (vector signed short);
5606 void vec_mtvscr (vector unsigned short);
5607 void vec_mtvscr (vector signed char);
5608 void vec_mtvscr (vector unsigned char);
5610 vector unsigned short vec_mule (vector unsigned char,
5611 vector unsigned char);
5612 vector signed short vec_mule (vector signed char, vector signed char);
5613 vector unsigned int vec_mule (vector unsigned short,
5614 vector unsigned short);
5615 vector signed int vec_mule (vector signed short, vector signed short);
5617 vector unsigned short vec_mulo (vector unsigned char,
5618 vector unsigned char);
5619 vector signed short vec_mulo (vector signed char, vector signed char);
5620 vector unsigned int vec_mulo (vector unsigned short,
5621 vector unsigned short);
5622 vector signed int vec_mulo (vector signed short, vector signed short);
5624 vector float vec_nmsub (vector float, vector float, vector float);
5626 vector float vec_nor (vector float, vector float);
5627 vector signed int vec_nor (vector signed int, vector signed int);
5628 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
5629 vector signed short vec_nor (vector signed short, vector signed short);
5630 vector unsigned short vec_nor (vector unsigned short,
5631 vector unsigned short);
5632 vector signed char vec_nor (vector signed char, vector signed char);
5633 vector unsigned char vec_nor (vector unsigned char,
5634 vector unsigned char);
5636 vector float vec_or (vector float, vector float);
5637 vector float vec_or (vector float, vector signed int);
5638 vector float vec_or (vector signed int, vector float);
5639 vector signed int vec_or (vector signed int, vector signed int);
5640 vector unsigned int vec_or (vector signed int, vector unsigned int);
5641 vector unsigned int vec_or (vector unsigned int, vector signed int);
5642 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
5643 vector signed short vec_or (vector signed short, vector signed short);
5644 vector unsigned short vec_or (vector signed short,
5645 vector unsigned short);
5646 vector unsigned short vec_or (vector unsigned short,
5647 vector signed short);
5648 vector unsigned short vec_or (vector unsigned short,
5649 vector unsigned short);
5650 vector signed char vec_or (vector signed char, vector signed char);
5651 vector unsigned char vec_or (vector signed char, vector unsigned char);
5652 vector unsigned char vec_or (vector unsigned char, vector signed char);
5653 vector unsigned char vec_or (vector unsigned char,
5654 vector unsigned char);
5656 vector signed char vec_pack (vector signed short, vector signed short);
5657 vector unsigned char vec_pack (vector unsigned short,
5658 vector unsigned short);
5659 vector signed short vec_pack (vector signed int, vector signed int);
5660 vector unsigned short vec_pack (vector unsigned int,
5661 vector unsigned int);
5663 vector signed short vec_packpx (vector unsigned int,
5664 vector unsigned int);
5666 vector unsigned char vec_packs (vector unsigned short,
5667 vector unsigned short);
5668 vector signed char vec_packs (vector signed short, vector signed short);
5670 vector unsigned short vec_packs (vector unsigned int,
5671 vector unsigned int);
5672 vector signed short vec_packs (vector signed int, vector signed int);
5674 vector unsigned char vec_packsu (vector unsigned short,
5675 vector unsigned short);
5676 vector unsigned char vec_packsu (vector signed short,
5677 vector signed short);
5678 vector unsigned short vec_packsu (vector unsigned int,
5679 vector unsigned int);
5680 vector unsigned short vec_packsu (vector signed int, vector signed int);
5682 vector float vec_perm (vector float, vector float,
5683 vector unsigned char);
5684 vector signed int vec_perm (vector signed int, vector signed int,
5685 vector unsigned char);
5686 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
5687 vector unsigned char);
5688 vector signed short vec_perm (vector signed short, vector signed short,
5689 vector unsigned char);
5690 vector unsigned short vec_perm (vector unsigned short,
5691 vector unsigned short,
5692 vector unsigned char);
5693 vector signed char vec_perm (vector signed char, vector signed char,
5694 vector unsigned char);
5695 vector unsigned char vec_perm (vector unsigned char,
5696 vector unsigned char,
5697 vector unsigned char);
5699 vector float vec_re (vector float);
5701 vector signed char vec_rl (vector signed char, vector unsigned char);
5702 vector unsigned char vec_rl (vector unsigned char,
5703 vector unsigned char);
5704 vector signed short vec_rl (vector signed short, vector unsigned short);
5706 vector unsigned short vec_rl (vector unsigned short,
5707 vector unsigned short);
5708 vector signed int vec_rl (vector signed int, vector unsigned int);
5709 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
5711 vector float vec_round (vector float);
5713 vector float vec_rsqrte (vector float);
5715 vector float vec_sel (vector float, vector float, vector signed int);
5716 vector float vec_sel (vector float, vector float, vector unsigned int);
5717 vector signed int vec_sel (vector signed int, vector signed int,
5719 vector signed int vec_sel (vector signed int, vector signed int,
5720 vector unsigned int);
5721 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
5723 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
5724 vector unsigned int);
5725 vector signed short vec_sel (vector signed short, vector signed short,
5726 vector signed short);
5727 vector signed short vec_sel (vector signed short, vector signed short,
5728 vector unsigned short);
5729 vector unsigned short vec_sel (vector unsigned short,
5730 vector unsigned short,
5731 vector signed short);
5732 vector unsigned short vec_sel (vector unsigned short,
5733 vector unsigned short,
5734 vector unsigned short);
5735 vector signed char vec_sel (vector signed char, vector signed char,
5736 vector signed char);
5737 vector signed char vec_sel (vector signed char, vector signed char,
5738 vector unsigned char);
5739 vector unsigned char vec_sel (vector unsigned char,
5740 vector unsigned char,
5741 vector signed char);
5742 vector unsigned char vec_sel (vector unsigned char,
5743 vector unsigned char,
5744 vector unsigned char);
5746 vector signed char vec_sl (vector signed char, vector unsigned char);
5747 vector unsigned char vec_sl (vector unsigned char,
5748 vector unsigned char);
5749 vector signed short vec_sl (vector signed short, vector unsigned short);
5751 vector unsigned short vec_sl (vector unsigned short,
5752 vector unsigned short);
5753 vector signed int vec_sl (vector signed int, vector unsigned int);
5754 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
5756 vector float vec_sld (vector float, vector float, const char);
5757 vector signed int vec_sld (vector signed int, vector signed int,
5759 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
5761 vector signed short vec_sld (vector signed short, vector signed short,
5763 vector unsigned short vec_sld (vector unsigned short,
5764 vector unsigned short, const char);
5765 vector signed char vec_sld (vector signed char, vector signed char,
5767 vector unsigned char vec_sld (vector unsigned char,
5768 vector unsigned char,
5771 vector signed int vec_sll (vector signed int, vector unsigned int);
5772 vector signed int vec_sll (vector signed int, vector unsigned short);
5773 vector signed int vec_sll (vector signed int, vector unsigned char);
5774 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
5775 vector unsigned int vec_sll (vector unsigned int,
5776 vector unsigned short);
5777 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
5779 vector signed short vec_sll (vector signed short, vector unsigned int);
5780 vector signed short vec_sll (vector signed short,
5781 vector unsigned short);
5782 vector signed short vec_sll (vector signed short, vector unsigned char);
5784 vector unsigned short vec_sll (vector unsigned short,
5785 vector unsigned int);
5786 vector unsigned short vec_sll (vector unsigned short,
5787 vector unsigned short);
5788 vector unsigned short vec_sll (vector unsigned short,
5789 vector unsigned char);
5790 vector signed char vec_sll (vector signed char, vector unsigned int);
5791 vector signed char vec_sll (vector signed char, vector unsigned short);
5792 vector signed char vec_sll (vector signed char, vector unsigned char);
5793 vector unsigned char vec_sll (vector unsigned char,
5794 vector unsigned int);
5795 vector unsigned char vec_sll (vector unsigned char,
5796 vector unsigned short);
5797 vector unsigned char vec_sll (vector unsigned char,
5798 vector unsigned char);
5800 vector float vec_slo (vector float, vector signed char);
5801 vector float vec_slo (vector float, vector unsigned char);
5802 vector signed int vec_slo (vector signed int, vector signed char);
5803 vector signed int vec_slo (vector signed int, vector unsigned char);
5804 vector unsigned int vec_slo (vector unsigned int, vector signed char);
5805 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
5807 vector signed short vec_slo (vector signed short, vector signed char);
5808 vector signed short vec_slo (vector signed short, vector unsigned char);
5810 vector unsigned short vec_slo (vector unsigned short,
5811 vector signed char);
5812 vector unsigned short vec_slo (vector unsigned short,
5813 vector unsigned char);
5814 vector signed char vec_slo (vector signed char, vector signed char);
5815 vector signed char vec_slo (vector signed char, vector unsigned char);
5816 vector unsigned char vec_slo (vector unsigned char, vector signed char);
5818 vector unsigned char vec_slo (vector unsigned char,
5819 vector unsigned char);
5821 vector signed char vec_splat (vector signed char, const char);
5822 vector unsigned char vec_splat (vector unsigned char, const char);
5823 vector signed short vec_splat (vector signed short, const char);
5824 vector unsigned short vec_splat (vector unsigned short, const char);
5825 vector float vec_splat (vector float, const char);
5826 vector signed int vec_splat (vector signed int, const char);
5827 vector unsigned int vec_splat (vector unsigned int, const char);
5829 vector signed char vec_splat_s8 (const char);
5831 vector signed short vec_splat_s16 (const char);
5833 vector signed int vec_splat_s32 (const char);
5835 vector unsigned char vec_splat_u8 (const char);
5837 vector unsigned short vec_splat_u16 (const char);
5839 vector unsigned int vec_splat_u32 (const char);
5841 vector signed char vec_sr (vector signed char, vector unsigned char);
5842 vector unsigned char vec_sr (vector unsigned char,
5843 vector unsigned char);
5844 vector signed short vec_sr (vector signed short, vector unsigned short);
5846 vector unsigned short vec_sr (vector unsigned short,
5847 vector unsigned short);
5848 vector signed int vec_sr (vector signed int, vector unsigned int);
5849 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
5851 vector signed char vec_sra (vector signed char, vector unsigned char);
5852 vector unsigned char vec_sra (vector unsigned char,
5853 vector unsigned char);
5854 vector signed short vec_sra (vector signed short,
5855 vector unsigned short);
5856 vector unsigned short vec_sra (vector unsigned short,
5857 vector unsigned short);
5858 vector signed int vec_sra (vector signed int, vector unsigned int);
5859 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
5861 vector signed int vec_srl (vector signed int, vector unsigned int);
5862 vector signed int vec_srl (vector signed int, vector unsigned short);
5863 vector signed int vec_srl (vector signed int, vector unsigned char);
5864 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
5865 vector unsigned int vec_srl (vector unsigned int,
5866 vector unsigned short);
5867 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
5869 vector signed short vec_srl (vector signed short, vector unsigned int);
5870 vector signed short vec_srl (vector signed short,
5871 vector unsigned short);
5872 vector signed short vec_srl (vector signed short, vector unsigned char);
5874 vector unsigned short vec_srl (vector unsigned short,
5875 vector unsigned int);
5876 vector unsigned short vec_srl (vector unsigned short,
5877 vector unsigned short);
5878 vector unsigned short vec_srl (vector unsigned short,
5879 vector unsigned char);
5880 vector signed char vec_srl (vector signed char, vector unsigned int);
5881 vector signed char vec_srl (vector signed char, vector unsigned short);
5882 vector signed char vec_srl (vector signed char, vector unsigned char);
5883 vector unsigned char vec_srl (vector unsigned char,
5884 vector unsigned int);
5885 vector unsigned char vec_srl (vector unsigned char,
5886 vector unsigned short);
5887 vector unsigned char vec_srl (vector unsigned char,
5888 vector unsigned char);
5890 vector float vec_sro (vector float, vector signed char);
5891 vector float vec_sro (vector float, vector unsigned char);
5892 vector signed int vec_sro (vector signed int, vector signed char);
5893 vector signed int vec_sro (vector signed int, vector unsigned char);
5894 vector unsigned int vec_sro (vector unsigned int, vector signed char);
5895 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
5897 vector signed short vec_sro (vector signed short, vector signed char);
5898 vector signed short vec_sro (vector signed short, vector unsigned char);
5900 vector unsigned short vec_sro (vector unsigned short,
5901 vector signed char);
5902 vector unsigned short vec_sro (vector unsigned short,
5903 vector unsigned char);
5904 vector signed char vec_sro (vector signed char, vector signed char);
5905 vector signed char vec_sro (vector signed char, vector unsigned char);
5906 vector unsigned char vec_sro (vector unsigned char, vector signed char);
5908 vector unsigned char vec_sro (vector unsigned char,
5909 vector unsigned char);
5911 void vec_st (vector float, int, float *);
5912 void vec_st (vector float, int, vector float *);
5913 void vec_st (vector signed int, int, int *);
5914 void vec_st (vector signed int, int, unsigned int *);
5915 void vec_st (vector unsigned int, int, unsigned int *);
5916 void vec_st (vector unsigned int, int, vector unsigned int *);
5917 void vec_st (vector signed short, int, short *);
5918 void vec_st (vector signed short, int, vector unsigned short *);
5919 void vec_st (vector signed short, int, vector signed short *);
5920 void vec_st (vector unsigned short, int, unsigned short *);
5921 void vec_st (vector unsigned short, int, vector unsigned short *);
5922 void vec_st (vector signed char, int, signed char *);
5923 void vec_st (vector signed char, int, unsigned char *);
5924 void vec_st (vector signed char, int, vector signed char *);
5925 void vec_st (vector unsigned char, int, unsigned char *);
5926 void vec_st (vector unsigned char, int, vector unsigned char *);
5928 void vec_ste (vector signed char, int, unsigned char *);
5929 void vec_ste (vector signed char, int, signed char *);
5930 void vec_ste (vector unsigned char, int, unsigned char *);
5931 void vec_ste (vector signed short, int, short *);
5932 void vec_ste (vector signed short, int, unsigned short *);
5933 void vec_ste (vector unsigned short, int, void *);
5934 void vec_ste (vector signed int, int, unsigned int *);
5935 void vec_ste (vector signed int, int, int *);
5936 void vec_ste (vector unsigned int, int, unsigned int *);
5937 void vec_ste (vector float, int, float *);
5939 void vec_stl (vector float, int, vector float *);
5940 void vec_stl (vector float, int, float *);
5941 void vec_stl (vector signed int, int, vector signed int *);
5942 void vec_stl (vector signed int, int, int *);
5943 void vec_stl (vector signed int, int, unsigned int *);
5944 void vec_stl (vector unsigned int, int, vector unsigned int *);
5945 void vec_stl (vector unsigned int, int, unsigned int *);
5946 void vec_stl (vector signed short, int, short *);
5947 void vec_stl (vector signed short, int, unsigned short *);
5948 void vec_stl (vector signed short, int, vector signed short *);
5949 void vec_stl (vector unsigned short, int, unsigned short *);
5950 void vec_stl (vector unsigned short, int, vector signed short *);
5951 void vec_stl (vector signed char, int, signed char *);
5952 void vec_stl (vector signed char, int, unsigned char *);
5953 void vec_stl (vector signed char, int, vector signed char *);
5954 void vec_stl (vector unsigned char, int, unsigned char *);
5955 void vec_stl (vector unsigned char, int, vector unsigned char *);
5957 vector signed char vec_sub (vector signed char, vector signed char);
5958 vector unsigned char vec_sub (vector signed char, vector unsigned char);
5960 vector unsigned char vec_sub (vector unsigned char, vector signed char);
5962 vector unsigned char vec_sub (vector unsigned char,
5963 vector unsigned char);
5964 vector signed short vec_sub (vector signed short, vector signed short);
5965 vector unsigned short vec_sub (vector signed short,
5966 vector unsigned short);
5967 vector unsigned short vec_sub (vector unsigned short,
5968 vector signed short);
5969 vector unsigned short vec_sub (vector unsigned short,
5970 vector unsigned short);
5971 vector signed int vec_sub (vector signed int, vector signed int);
5972 vector unsigned int vec_sub (vector signed int, vector unsigned int);
5973 vector unsigned int vec_sub (vector unsigned int, vector signed int);
5974 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
5975 vector float vec_sub (vector float, vector float);
5977 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
5979 vector unsigned char vec_subs (vector signed char,
5980 vector unsigned char);
5981 vector unsigned char vec_subs (vector unsigned char,
5982 vector signed char);
5983 vector unsigned char vec_subs (vector unsigned char,
5984 vector unsigned char);
5985 vector signed char vec_subs (vector signed char, vector signed char);
5986 vector unsigned short vec_subs (vector signed short,
5987 vector unsigned short);
5988 vector unsigned short vec_subs (vector unsigned short,
5989 vector signed short);
5990 vector unsigned short vec_subs (vector unsigned short,
5991 vector unsigned short);
5992 vector signed short vec_subs (vector signed short, vector signed short);
5994 vector unsigned int vec_subs (vector signed int, vector unsigned int);
5995 vector unsigned int vec_subs (vector unsigned int, vector signed int);
5996 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
5998 vector signed int vec_subs (vector signed int, vector signed int);
6000 vector unsigned int vec_sum4s (vector unsigned char,
6001 vector unsigned int);
6002 vector signed int vec_sum4s (vector signed char, vector signed int);
6003 vector signed int vec_sum4s (vector signed short, vector signed int);
6005 vector signed int vec_sum2s (vector signed int, vector signed int);
6007 vector signed int vec_sums (vector signed int, vector signed int);
6009 vector float vec_trunc (vector float);
6011 vector signed short vec_unpackh (vector signed char);
6012 vector unsigned int vec_unpackh (vector signed short);
6013 vector signed int vec_unpackh (vector signed short);
6015 vector signed short vec_unpackl (vector signed char);
6016 vector unsigned int vec_unpackl (vector signed short);
6017 vector signed int vec_unpackl (vector signed short);
6019 vector float vec_xor (vector float, vector float);
6020 vector float vec_xor (vector float, vector signed int);
6021 vector float vec_xor (vector signed int, vector float);
6022 vector signed int vec_xor (vector signed int, vector signed int);
6023 vector unsigned int vec_xor (vector signed int, vector unsigned int);
6024 vector unsigned int vec_xor (vector unsigned int, vector signed int);
6025 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
6026 vector signed short vec_xor (vector signed short, vector signed short);
6027 vector unsigned short vec_xor (vector signed short,
6028 vector unsigned short);
6029 vector unsigned short vec_xor (vector unsigned short,
6030 vector signed short);
6031 vector unsigned short vec_xor (vector unsigned short,
6032 vector unsigned short);
6033 vector signed char vec_xor (vector signed char, vector signed char);
6034 vector unsigned char vec_xor (vector signed char, vector unsigned char);
6036 vector unsigned char vec_xor (vector unsigned char, vector signed char);
6038 vector unsigned char vec_xor (vector unsigned char,
6039 vector unsigned char);
6041 vector signed int vec_all_eq (vector signed char, vector unsigned char);
6043 vector signed int vec_all_eq (vector signed char, vector signed char);
6044 vector signed int vec_all_eq (vector unsigned char, vector signed char);
6046 vector signed int vec_all_eq (vector unsigned char,
6047 vector unsigned char);
6048 vector signed int vec_all_eq (vector signed short,
6049 vector unsigned short);
6050 vector signed int vec_all_eq (vector signed short, vector signed short);
6052 vector signed int vec_all_eq (vector unsigned short,
6053 vector signed short);
6054 vector signed int vec_all_eq (vector unsigned short,
6055 vector unsigned short);
6056 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6057 vector signed int vec_all_eq (vector signed int, vector signed int);
6058 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6059 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6061 vector signed int vec_all_eq (vector float, vector float);
6063 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6065 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6067 vector signed int vec_all_ge (vector unsigned char,
6068 vector unsigned char);
6069 vector signed int vec_all_ge (vector signed char, vector signed char);
6070 vector signed int vec_all_ge (vector signed short,
6071 vector unsigned short);
6072 vector signed int vec_all_ge (vector unsigned short,
6073 vector signed short);
6074 vector signed int vec_all_ge (vector unsigned short,
6075 vector unsigned short);
6076 vector signed int vec_all_ge (vector signed short, vector signed short);
6078 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6079 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6080 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6082 vector signed int vec_all_ge (vector signed int, vector signed int);
6083 vector signed int vec_all_ge (vector float, vector float);
6085 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6087 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6089 vector signed int vec_all_gt (vector unsigned char,
6090 vector unsigned char);
6091 vector signed int vec_all_gt (vector signed char, vector signed char);
6092 vector signed int vec_all_gt (vector signed short,
6093 vector unsigned short);
6094 vector signed int vec_all_gt (vector unsigned short,
6095 vector signed short);
6096 vector signed int vec_all_gt (vector unsigned short,
6097 vector unsigned short);
6098 vector signed int vec_all_gt (vector signed short, vector signed short);
6100 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6101 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6102 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6104 vector signed int vec_all_gt (vector signed int, vector signed int);
6105 vector signed int vec_all_gt (vector float, vector float);
6107 vector signed int vec_all_in (vector float, vector float);
6109 vector signed int vec_all_le (vector signed char, vector unsigned char);
6111 vector signed int vec_all_le (vector unsigned char, vector signed char);
6113 vector signed int vec_all_le (vector unsigned char,
6114 vector unsigned char);
6115 vector signed int vec_all_le (vector signed char, vector signed char);
6116 vector signed int vec_all_le (vector signed short,
6117 vector unsigned short);
6118 vector signed int vec_all_le (vector unsigned short,
6119 vector signed short);
6120 vector signed int vec_all_le (vector unsigned short,
6121 vector unsigned short);
6122 vector signed int vec_all_le (vector signed short, vector signed short);
6124 vector signed int vec_all_le (vector signed int, vector unsigned int);
6125 vector signed int vec_all_le (vector unsigned int, vector signed int);
6126 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
6128 vector signed int vec_all_le (vector signed int, vector signed int);
6129 vector signed int vec_all_le (vector float, vector float);
6131 vector signed int vec_all_lt (vector signed char, vector unsigned char);
6133 vector signed int vec_all_lt (vector unsigned char, vector signed char);
6135 vector signed int vec_all_lt (vector unsigned char,
6136 vector unsigned char);
6137 vector signed int vec_all_lt (vector signed char, vector signed char);
6138 vector signed int vec_all_lt (vector signed short,
6139 vector unsigned short);
6140 vector signed int vec_all_lt (vector unsigned short,
6141 vector signed short);
6142 vector signed int vec_all_lt (vector unsigned short,
6143 vector unsigned short);
6144 vector signed int vec_all_lt (vector signed short, vector signed short);
6146 vector signed int vec_all_lt (vector signed int, vector unsigned int);
6147 vector signed int vec_all_lt (vector unsigned int, vector signed int);
6148 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
6150 vector signed int vec_all_lt (vector signed int, vector signed int);
6151 vector signed int vec_all_lt (vector float, vector float);
6153 vector signed int vec_all_nan (vector float);
6155 vector signed int vec_all_ne (vector signed char, vector unsigned char);
6157 vector signed int vec_all_ne (vector signed char, vector signed char);
6158 vector signed int vec_all_ne (vector unsigned char, vector signed char);
6160 vector signed int vec_all_ne (vector unsigned char,
6161 vector unsigned char);
6162 vector signed int vec_all_ne (vector signed short,
6163 vector unsigned short);
6164 vector signed int vec_all_ne (vector signed short, vector signed short);
6166 vector signed int vec_all_ne (vector unsigned short,
6167 vector signed short);
6168 vector signed int vec_all_ne (vector unsigned short,
6169 vector unsigned short);
6170 vector signed int vec_all_ne (vector signed int, vector unsigned int);
6171 vector signed int vec_all_ne (vector signed int, vector signed int);
6172 vector signed int vec_all_ne (vector unsigned int, vector signed int);
6173 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
6175 vector signed int vec_all_ne (vector float, vector float);
6177 vector signed int vec_all_nge (vector float, vector float);
6179 vector signed int vec_all_ngt (vector float, vector float);
6181 vector signed int vec_all_nle (vector float, vector float);
6183 vector signed int vec_all_nlt (vector float, vector float);
6185 vector signed int vec_all_numeric (vector float);
6187 vector signed int vec_any_eq (vector signed char, vector unsigned char);
6189 vector signed int vec_any_eq (vector signed char, vector signed char);
6190 vector signed int vec_any_eq (vector unsigned char, vector signed char);
6192 vector signed int vec_any_eq (vector unsigned char,
6193 vector unsigned char);
6194 vector signed int vec_any_eq (vector signed short,
6195 vector unsigned short);
6196 vector signed int vec_any_eq (vector signed short, vector signed short);
6198 vector signed int vec_any_eq (vector unsigned short,
6199 vector signed short);
6200 vector signed int vec_any_eq (vector unsigned short,
6201 vector unsigned short);
6202 vector signed int vec_any_eq (vector signed int, vector unsigned int);
6203 vector signed int vec_any_eq (vector signed int, vector signed int);
6204 vector signed int vec_any_eq (vector unsigned int, vector signed int);
6205 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
6207 vector signed int vec_any_eq (vector float, vector float);
6209 vector signed int vec_any_ge (vector signed char, vector unsigned char);
6211 vector signed int vec_any_ge (vector unsigned char, vector signed char);
6213 vector signed int vec_any_ge (vector unsigned char,
6214 vector unsigned char);
6215 vector signed int vec_any_ge (vector signed char, vector signed char);
6216 vector signed int vec_any_ge (vector signed short,
6217 vector unsigned short);
6218 vector signed int vec_any_ge (vector unsigned short,
6219 vector signed short);
6220 vector signed int vec_any_ge (vector unsigned short,
6221 vector unsigned short);
6222 vector signed int vec_any_ge (vector signed short, vector signed short);
6224 vector signed int vec_any_ge (vector signed int, vector unsigned int);
6225 vector signed int vec_any_ge (vector unsigned int, vector signed int);
6226 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
6228 vector signed int vec_any_ge (vector signed int, vector signed int);
6229 vector signed int vec_any_ge (vector float, vector float);
6231 vector signed int vec_any_gt (vector signed char, vector unsigned char);
6233 vector signed int vec_any_gt (vector unsigned char, vector signed char);
6235 vector signed int vec_any_gt (vector unsigned char,
6236 vector unsigned char);
6237 vector signed int vec_any_gt (vector signed char, vector signed char);
6238 vector signed int vec_any_gt (vector signed short,
6239 vector unsigned short);
6240 vector signed int vec_any_gt (vector unsigned short,
6241 vector signed short);
6242 vector signed int vec_any_gt (vector unsigned short,
6243 vector unsigned short);
6244 vector signed int vec_any_gt (vector signed short, vector signed short);
6246 vector signed int vec_any_gt (vector signed int, vector unsigned int);
6247 vector signed int vec_any_gt (vector unsigned int, vector signed int);
6248 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
6250 vector signed int vec_any_gt (vector signed int, vector signed int);
6251 vector signed int vec_any_gt (vector float, vector float);
6253 vector signed int vec_any_le (vector signed char, vector unsigned char);
6255 vector signed int vec_any_le (vector unsigned char, vector signed char);
6257 vector signed int vec_any_le (vector unsigned char,
6258 vector unsigned char);
6259 vector signed int vec_any_le (vector signed char, vector signed char);
6260 vector signed int vec_any_le (vector signed short,
6261 vector unsigned short);
6262 vector signed int vec_any_le (vector unsigned short,
6263 vector signed short);
6264 vector signed int vec_any_le (vector unsigned short,
6265 vector unsigned short);
6266 vector signed int vec_any_le (vector signed short, vector signed short);
6268 vector signed int vec_any_le (vector signed int, vector unsigned int);
6269 vector signed int vec_any_le (vector unsigned int, vector signed int);
6270 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
6272 vector signed int vec_any_le (vector signed int, vector signed int);
6273 vector signed int vec_any_le (vector float, vector float);
6275 vector signed int vec_any_lt (vector signed char, vector unsigned char);
6277 vector signed int vec_any_lt (vector unsigned char, vector signed char);
6279 vector signed int vec_any_lt (vector unsigned char,
6280 vector unsigned char);
6281 vector signed int vec_any_lt (vector signed char, vector signed char);
6282 vector signed int vec_any_lt (vector signed short,
6283 vector unsigned short);
6284 vector signed int vec_any_lt (vector unsigned short,
6285 vector signed short);
6286 vector signed int vec_any_lt (vector unsigned short,
6287 vector unsigned short);
6288 vector signed int vec_any_lt (vector signed short, vector signed short);
6290 vector signed int vec_any_lt (vector signed int, vector unsigned int);
6291 vector signed int vec_any_lt (vector unsigned int, vector signed int);
6292 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
6294 vector signed int vec_any_lt (vector signed int, vector signed int);
6295 vector signed int vec_any_lt (vector float, vector float);
6297 vector signed int vec_any_nan (vector float);
6299 vector signed int vec_any_ne (vector signed char, vector unsigned char);
6301 vector signed int vec_any_ne (vector signed char, vector signed char);
6302 vector signed int vec_any_ne (vector unsigned char, vector signed char);
6304 vector signed int vec_any_ne (vector unsigned char,
6305 vector unsigned char);
6306 vector signed int vec_any_ne (vector signed short,
6307 vector unsigned short);
6308 vector signed int vec_any_ne (vector signed short, vector signed short);
6310 vector signed int vec_any_ne (vector unsigned short,
6311 vector signed short);
6312 vector signed int vec_any_ne (vector unsigned short,
6313 vector unsigned short);
6314 vector signed int vec_any_ne (vector signed int, vector unsigned int);
6315 vector signed int vec_any_ne (vector signed int, vector signed int);
6316 vector signed int vec_any_ne (vector unsigned int, vector signed int);
6317 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
6319 vector signed int vec_any_ne (vector float, vector float);
6321 vector signed int vec_any_nge (vector float, vector float);
6323 vector signed int vec_any_ngt (vector float, vector float);
6325 vector signed int vec_any_nle (vector float, vector float);
6327 vector signed int vec_any_nlt (vector float, vector float);
6329 vector signed int vec_any_numeric (vector float);
6331 vector signed int vec_any_out (vector float, vector float);
6335 @section Pragmas Accepted by GCC
6339 GCC supports several types of pragmas, primarily in order to compile
6340 code originally written for other compilers. Note that in general
6341 we do not recommend the use of pragmas; @xref{Function Attributes},
6342 for further explanation.
6346 * RS/6000 and PowerPC Pragmas::
6353 @subsection ARM Pragmas
6355 The ARM target defines pragmas for controlling the default addition of
6356 @code{long_call} and @code{short_call} attributes to functions.
6357 @xref{Function Attributes}, for information about the effects of these
6362 @cindex pragma, long_calls
6363 Set all subsequent functions to have the @code{long_call} attribute.
6366 @cindex pragma, no_long_calls
6367 Set all subsequent functions to have the @code{short_call} attribute.
6369 @item long_calls_off
6370 @cindex pragma, long_calls_off
6371 Do not affect the @code{long_call} or @code{short_call} attributes of
6372 subsequent functions.
6375 @node RS/6000 and PowerPC Pragmas
6376 @subsection RS/6000 and PowerPC Pragmas
6378 The RS/6000 and PowerPC targets define one pragma for controlling
6379 whether or not the @code{longcall} attribute is added to function
6380 declarations by default. This pragma overrides the @option{-mlongcall}
6381 option, but not the @code{longcall} and @code{shortcall} attributes.
6382 @xref{RS/6000 and PowerPC Options}, for more information about when long
6383 calls are and are not necessary.
6387 @cindex pragma, longcall
6388 Apply the @code{longcall} attribute to all subsequent function
6392 Do not apply the @code{longcall} attribute to subsequent function
6396 @c Describe c4x pragmas here.
6397 @c Describe h8300 pragmas here.
6398 @c Describe i370 pragmas here.
6399 @c Describe i960 pragmas here.
6400 @c Describe sh pragmas here.
6401 @c Describe v850 pragmas here.
6403 @node Darwin Pragmas
6404 @subsection Darwin Pragmas
6406 The following pragmas are available for all architectures running the
6407 Darwin operating system. These are useful for compatibility with other
6411 @item mark @var{tokens}@dots{}
6412 @cindex pragma, mark
6413 This pragma is accepted, but has no effect.
6415 @item options align=@var{alignment}
6416 @cindex pragma, options align
6417 This pragma sets the alignment of fields in structures. The values of
6418 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
6419 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
6420 properly; to restore the previous setting, use @code{reset} for the
6423 @item segment @var{tokens}@dots{}
6424 @cindex pragma, segment
6425 This pragma is accepted, but has no effect.
6427 @item unused (@var{var} [, @var{var}]@dots{})
6428 @cindex pragma, unused
6429 This pragma declares variables to be possibly unused. GCC will not
6430 produce warnings for the listed variables. The effect is similar to
6431 that of the @code{unused} attribute, except that this pragma may appear
6432 anywhere within the variables' scopes.
6435 @node Solaris Pragmas
6436 @subsection Solaris Pragmas
6438 For compatibility with the SunPRO compiler, the following pragma
6442 @item redefine_extname @var{oldname} @var{newname}
6443 @cindex pragma, redefine_extname
6445 This pragma gives the C function @var{oldname} the assembler label
6446 @var{newname}. The pragma must appear before the function declaration.
6447 This pragma is equivalent to the asm labels extension (@pxref{Asm
6448 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
6449 if the pragma is available.
6453 @subsection Tru64 Pragmas
6455 For compatibility with the Compaq C compiler, the following pragma
6459 @item extern_prefix @var{string}
6460 @cindex pragma, extern_prefix
6462 This pragma renames all subsequent function and variable declarations
6463 such that @var{string} is prepended to the name. This effect may be
6464 terminated by using another @code{extern_prefix} pragma with the
6467 This pragma is similar in intent to to the asm labels extension
6468 (@pxref{Asm Labels}) in that the system programmer wants to change
6469 the assembly-level ABI without changing the source-level API. The
6470 preprocessor defines @code{__EXTERN_PREFIX} if the pragma is available.
6473 @node Unnamed Fields
6474 @section Unnamed struct/union fields within structs/unions.
6478 For compatibility with other compilers, GCC allows you to define
6479 a structure or union that contains, as fields, structures and unions
6480 without names. For example:
6493 In this example, the user would be able to access members of the unnamed
6494 union with code like @samp{foo.b}. Note that only unnamed structs and
6495 unions are allowed, you may not have, for example, an unnamed
6498 You must never create such structures that cause ambiguous field definitions.
6499 For example, this structure:
6510 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
6511 Such constructs are not supported and must be avoided. In the future,
6512 such constructs may be detected and treated as compilation errors.
6515 @section Thread-Local Storage
6516 @cindex Thread-Local Storage
6517 @cindex @acronym{TLS}
6520 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
6521 are allocated such that there is one instance of the variable per extant
6522 thread. The run-time model GCC uses to implement this originates
6523 in the IA-64 processor-specific ABI, but has since been migrated
6524 to other processors as well. It requires significant support from
6525 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
6526 system libraries (@file{libc.so} and @file{libpthread.so}), so it
6527 is not available everywhere.
6529 At the user level, the extension is visible with a new storage
6530 class keyword: @code{__thread}. For example:
6534 extern __thread struct state s;
6535 static __thread char *p;
6538 The @code{__thread} specifier may be used alone, with the @code{extern}
6539 or @code{static} specifiers, but with no other storage class specifier.
6540 When used with @code{extern} or @code{static}, @code{__thread} must appear
6541 immediately after the other storage class specifier.
6543 The @code{__thread} specifier may be applied to any global, file-scoped
6544 static, function-scoped static, or static data member of a class. It may
6545 not be applied to block-scoped automatic or non-static data member.
6547 When the address-of operator is applied to a thread-local variable, it is
6548 evaluated at run-time and returns the address of the current thread's
6549 instance of that variable. An address so obtained may be used by any
6550 thread. When a thread terminates, any pointers to thread-local variables
6551 in that thread become invalid.
6553 No static initialization may refer to the address of a thread-local variable.
6555 In C++, if an initializer is present for a thread-local variable, it must
6556 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
6559 See @uref{http://people.redhat.com/drepper/tls.pdf,
6560 ELF Handling For Thread-Local Storage} for a detailed explanation of
6561 the four thread-local storage addressing models, and how the run-time
6562 is expected to function.
6565 * C99 Thread-Local Edits::
6566 * C++98 Thread-Local Edits::
6569 @node C99 Thread-Local Edits
6570 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
6572 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
6573 that document the exact semantics of the language extension.
6577 @cite{5.1.2 Execution environments}
6579 Add new text after paragraph 1
6582 Within either execution environment, a @dfn{thread} is a flow of
6583 control within a program. It is implementation defined whether
6584 or not there may be more than one thread associated with a program.
6585 It is implementation defined how threads beyond the first are
6586 created, the name and type of the function called at thread
6587 startup, and how threads may be terminated. However, objects
6588 with thread storage duration shall be initialized before thread
6593 @cite{6.2.4 Storage durations of objects}
6595 Add new text before paragraph 3
6598 An object whose identifier is declared with the storage-class
6599 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
6600 Its lifetime is the entire execution of the thread, and its
6601 stored value is initialized only once, prior to thread startup.
6605 @cite{6.4.1 Keywords}
6607 Add @code{__thread}.
6610 @cite{6.7.1 Storage-class specifiers}
6612 Add @code{__thread} to the list of storage class specifiers in
6615 Change paragraph 2 to
6618 With the exception of @code{__thread}, at most one storage-class
6619 specifier may be given [@dots{}]. The @code{__thread} specifier may
6620 be used alone, or immediately following @code{extern} or
6624 Add new text after paragraph 6
6627 The declaration of an identifier for a variable that has
6628 block scope that specifies @code{__thread} shall also
6629 specify either @code{extern} or @code{static}.
6631 The @code{__thread} specifier shall be used only with
6636 @node C++98 Thread-Local Edits
6637 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
6639 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
6640 that document the exact semantics of the language extension.
6643 @b{[intro.execution]}
6645 New text after paragraph 4
6648 A @dfn{thread} is a flow of control within the abstract machine.
6649 It is implementation defined whether or not there may be more than
6653 New text after paragraph 7
6656 It is unspecified whether additional action must be taken to
6657 ensure when and whether side effects are visible to other threads.
6663 Add @code{__thread}.
6666 @b{[basic.start.main]}
6668 Add after paragraph 5
6671 The thread that begins execution at the @code{main} function is called
6672 the @dfn{main thread}. It is implementation defined how functions
6673 beginning threads other than the main thread are designated or typed.
6674 A function so designated, as well as the @code{main} function, is called
6675 a @dfn{thread startup function}. It is implementation defined what
6676 happens if a thread startup function returns. It is implementation
6677 defined what happens to other threads when any thread calls @code{exit}.
6681 @b{[basic.start.init]}
6683 Add after paragraph 4
6686 The storage for an object of thread storage duration shall be
6687 staticly initialized before the first statement of the thread startup
6688 function. An object of thread storage duration shall not require
6689 dynamic initialization.
6693 @b{[basic.start.term]}
6695 Add after paragraph 3
6698 The type of an object with thread storage duration shall not have a
6699 non-trivial destructor, nor shall it be an array type whose elements
6700 (directly or indirectly) have non-trivial destructors.
6706 Add ``thread storage duration'' to the list in paragraph 1.
6711 Thread, static, and automatic storage durations are associated with
6712 objects introduced by declarations [@dots{}].
6715 Add @code{__thread} to the list of specifiers in paragraph 3.
6718 @b{[basic.stc.thread]}
6720 New section before @b{[basic.stc.static]}
6723 The keyword @code{__thread} applied to an non-local object gives the
6724 object thread storage duration.
6726 A local variable or class data member declared both @code{static}
6727 and @code{__thread} gives the variable or member thread storage
6732 @b{[basic.stc.static]}
6737 All objects which have neither thread storage duration, dynamic
6738 storage duration nor are local [@dots{}].
6744 Add @code{__thread} to the list in paragraph 1.
6749 With the exception of @code{__thread}, at most one
6750 @var{storage-class-specifier} shall appear in a given
6751 @var{decl-specifier-seq}. The @code{__thread} specifier may
6752 be used alone, or immediately following the @code{extern} or
6753 @code{static} specifiers. [@dots{}]
6756 Add after paragraph 5
6759 The @code{__thread} specifier can be applied only to the names of objects
6760 and to anonymous unions.
6766 Add after paragraph 6
6769 Non-@code{static} members shall not be @code{__thread}.
6773 @node C++ Extensions
6774 @chapter Extensions to the C++ Language
6775 @cindex extensions, C++ language
6776 @cindex C++ language extensions
6778 The GNU compiler provides these extensions to the C++ language (and you
6779 can also use most of the C language extensions in your C++ programs). If you
6780 want to write code that checks whether these features are available, you can
6781 test for the GNU compiler the same way as for C programs: check for a
6782 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
6783 test specifically for GNU C++ (@pxref{Standard Predefined,,Standard
6784 Predefined Macros,cpp.info,The C Preprocessor}).
6787 * Min and Max:: C++ Minimum and maximum operators.
6788 * Volatiles:: What constitutes an access to a volatile object.
6789 * Restricted Pointers:: C99 restricted pointers and references.
6790 * Vague Linkage:: Where G++ puts inlines, vtables and such.
6791 * C++ Interface:: You can use a single C++ header file for both
6792 declarations and definitions.
6793 * Template Instantiation:: Methods for ensuring that exactly one copy of
6794 each needed template instantiation is emitted.
6795 * Bound member functions:: You can extract a function pointer to the
6796 method denoted by a @samp{->*} or @samp{.*} expression.
6797 * C++ Attributes:: Variable, function, and type attributes for C++ only.
6798 * Java Exceptions:: Tweaking exception handling to work with Java.
6799 * Deprecated Features:: Things might disappear from g++.
6800 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
6804 @section Minimum and Maximum Operators in C++
6806 It is very convenient to have operators which return the ``minimum'' or the
6807 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
6810 @item @var{a} <? @var{b}
6812 @cindex minimum operator
6813 is the @dfn{minimum}, returning the smaller of the numeric values
6814 @var{a} and @var{b};
6816 @item @var{a} >? @var{b}
6818 @cindex maximum operator
6819 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
6823 These operations are not primitive in ordinary C++, since you can
6824 use a macro to return the minimum of two things in C++, as in the
6828 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
6832 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
6833 the minimum value of variables @var{i} and @var{j}.
6835 However, side effects in @code{X} or @code{Y} may cause unintended
6836 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
6837 the smaller counter twice. The GNU C @code{typeof} extension allows you
6838 to write safe macros that avoid this kind of problem (@pxref{Typeof}).
6839 However, writing @code{MIN} and @code{MAX} as macros also forces you to
6840 use function-call notation for a fundamental arithmetic operation.
6841 Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
6844 Since @code{<?} and @code{>?} are built into the compiler, they properly
6845 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
6849 @section When is a Volatile Object Accessed?
6850 @cindex accessing volatiles
6851 @cindex volatile read
6852 @cindex volatile write
6853 @cindex volatile access
6855 Both the C and C++ standard have the concept of volatile objects. These
6856 are normally accessed by pointers and used for accessing hardware. The
6857 standards encourage compilers to refrain from optimizations
6858 concerning accesses to volatile objects that it might perform on
6859 non-volatile objects. The C standard leaves it implementation defined
6860 as to what constitutes a volatile access. The C++ standard omits to
6861 specify this, except to say that C++ should behave in a similar manner
6862 to C with respect to volatiles, where possible. The minimum either
6863 standard specifies is that at a sequence point all previous accesses to
6864 volatile objects have stabilized and no subsequent accesses have
6865 occurred. Thus an implementation is free to reorder and combine
6866 volatile accesses which occur between sequence points, but cannot do so
6867 for accesses across a sequence point. The use of volatiles does not
6868 allow you to violate the restriction on updating objects multiple times
6869 within a sequence point.
6871 In most expressions, it is intuitively obvious what is a read and what is
6872 a write. For instance
6875 volatile int *dst = @var{somevalue};
6876 volatile int *src = @var{someothervalue};
6881 will cause a read of the volatile object pointed to by @var{src} and stores the
6882 value into the volatile object pointed to by @var{dst}. There is no
6883 guarantee that these reads and writes are atomic, especially for objects
6884 larger than @code{int}.
6886 Less obvious expressions are where something which looks like an access
6887 is used in a void context. An example would be,
6890 volatile int *src = @var{somevalue};
6894 With C, such expressions are rvalues, and as rvalues cause a read of
6895 the object, GCC interprets this as a read of the volatile being pointed
6896 to. The C++ standard specifies that such expressions do not undergo
6897 lvalue to rvalue conversion, and that the type of the dereferenced
6898 object may be incomplete. The C++ standard does not specify explicitly
6899 that it is this lvalue to rvalue conversion which is responsible for
6900 causing an access. However, there is reason to believe that it is,
6901 because otherwise certain simple expressions become undefined. However,
6902 because it would surprise most programmers, G++ treats dereferencing a
6903 pointer to volatile object of complete type in a void context as a read
6904 of the object. When the object has incomplete type, G++ issues a
6909 struct T @{int m;@};
6910 volatile S *ptr1 = @var{somevalue};
6911 volatile T *ptr2 = @var{somevalue};
6916 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
6917 causes a read of the object pointed to. If you wish to force an error on
6918 the first case, you must force a conversion to rvalue with, for instance
6919 a static cast, @code{static_cast<S>(*ptr1)}.
6921 When using a reference to volatile, G++ does not treat equivalent
6922 expressions as accesses to volatiles, but instead issues a warning that
6923 no volatile is accessed. The rationale for this is that otherwise it
6924 becomes difficult to determine where volatile access occur, and not
6925 possible to ignore the return value from functions returning volatile
6926 references. Again, if you wish to force a read, cast the reference to
6929 @node Restricted Pointers
6930 @section Restricting Pointer Aliasing
6931 @cindex restricted pointers
6932 @cindex restricted references
6933 @cindex restricted this pointer
6935 As with gcc, g++ understands the C99 feature of restricted pointers,
6936 specified with the @code{__restrict__}, or @code{__restrict} type
6937 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
6938 language flag, @code{restrict} is not a keyword in C++.
6940 In addition to allowing restricted pointers, you can specify restricted
6941 references, which indicate that the reference is not aliased in the local
6945 void fn (int *__restrict__ rptr, int &__restrict__ rref)
6952 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
6953 @var{rref} refers to a (different) unaliased integer.
6955 You may also specify whether a member function's @var{this} pointer is
6956 unaliased by using @code{__restrict__} as a member function qualifier.
6959 void T::fn () __restrict__
6966 Within the body of @code{T::fn}, @var{this} will have the effective
6967 definition @code{T *__restrict__ const this}. Notice that the
6968 interpretation of a @code{__restrict__} member function qualifier is
6969 different to that of @code{const} or @code{volatile} qualifier, in that it
6970 is applied to the pointer rather than the object. This is consistent with
6971 other compilers which implement restricted pointers.
6973 As with all outermost parameter qualifiers, @code{__restrict__} is
6974 ignored in function definition matching. This means you only need to
6975 specify @code{__restrict__} in a function definition, rather than
6976 in a function prototype as well.
6979 @section Vague Linkage
6980 @cindex vague linkage
6982 There are several constructs in C++ which require space in the object
6983 file but are not clearly tied to a single translation unit. We say that
6984 these constructs have ``vague linkage''. Typically such constructs are
6985 emitted wherever they are needed, though sometimes we can be more
6989 @item Inline Functions
6990 Inline functions are typically defined in a header file which can be
6991 included in many different compilations. Hopefully they can usually be
6992 inlined, but sometimes an out-of-line copy is necessary, if the address
6993 of the function is taken or if inlining fails. In general, we emit an
6994 out-of-line copy in all translation units where one is needed. As an
6995 exception, we only emit inline virtual functions with the vtable, since
6996 it will always require a copy.
6998 Local static variables and string constants used in an inline function
6999 are also considered to have vague linkage, since they must be shared
7000 between all inlined and out-of-line instances of the function.
7004 C++ virtual functions are implemented in most compilers using a lookup
7005 table, known as a vtable. The vtable contains pointers to the virtual
7006 functions provided by a class, and each object of the class contains a
7007 pointer to its vtable (or vtables, in some multiple-inheritance
7008 situations). If the class declares any non-inline, non-pure virtual
7009 functions, the first one is chosen as the ``key method'' for the class,
7010 and the vtable is only emitted in the translation unit where the key
7013 @emph{Note:} If the chosen key method is later defined as inline, the
7014 vtable will still be emitted in every translation unit which defines it.
7015 Make sure that any inline virtuals are declared inline in the class
7016 body, even if they are not defined there.
7018 @item type_info objects
7021 C++ requires information about types to be written out in order to
7022 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
7023 For polymorphic classes (classes with virtual functions), the type_info
7024 object is written out along with the vtable so that @samp{dynamic_cast}
7025 can determine the dynamic type of a class object at runtime. For all
7026 other types, we write out the type_info object when it is used: when
7027 applying @samp{typeid} to an expression, throwing an object, or
7028 referring to a type in a catch clause or exception specification.
7030 @item Template Instantiations
7031 Most everything in this section also applies to template instantiations,
7032 but there are other options as well.
7033 @xref{Template Instantiation,,Where's the Template?}.
7037 When used with GNU ld version 2.8 or later on an ELF system such as
7038 Linux/GNU or Solaris 2, or on Microsoft Windows, duplicate copies of
7039 these constructs will be discarded at link time. This is known as
7042 On targets that don't support COMDAT, but do support weak symbols, GCC
7043 will use them. This way one copy will override all the others, but
7044 the unused copies will still take up space in the executable.
7046 For targets which do not support either COMDAT or weak symbols,
7047 most entities with vague linkage will be emitted as local symbols to
7048 avoid duplicate definition errors from the linker. This will not happen
7049 for local statics in inlines, however, as having multiple copies will
7050 almost certainly break things.
7052 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7053 another way to control placement of these constructs.
7056 @section Declarations and Definitions in One Header
7058 @cindex interface and implementation headers, C++
7059 @cindex C++ interface and implementation headers
7060 C++ object definitions can be quite complex. In principle, your source
7061 code will need two kinds of things for each object that you use across
7062 more than one source file. First, you need an @dfn{interface}
7063 specification, describing its structure with type declarations and
7064 function prototypes. Second, you need the @dfn{implementation} itself.
7065 It can be tedious to maintain a separate interface description in a
7066 header file, in parallel to the actual implementation. It is also
7067 dangerous, since separate interface and implementation definitions may
7068 not remain parallel.
7070 @cindex pragmas, interface and implementation
7071 With GNU C++, you can use a single header file for both purposes.
7074 @emph{Warning:} The mechanism to specify this is in transition. For the
7075 nonce, you must use one of two @code{#pragma} commands; in a future
7076 release of GNU C++, an alternative mechanism will make these
7077 @code{#pragma} commands unnecessary.
7080 The header file contains the full definitions, but is marked with
7081 @samp{#pragma interface} in the source code. This allows the compiler
7082 to use the header file only as an interface specification when ordinary
7083 source files incorporate it with @code{#include}. In the single source
7084 file where the full implementation belongs, you can use either a naming
7085 convention or @samp{#pragma implementation} to indicate this alternate
7086 use of the header file.
7089 @item #pragma interface
7090 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7091 @kindex #pragma interface
7092 Use this directive in @emph{header files} that define object classes, to save
7093 space in most of the object files that use those classes. Normally,
7094 local copies of certain information (backup copies of inline member
7095 functions, debugging information, and the internal tables that implement
7096 virtual functions) must be kept in each object file that includes class
7097 definitions. You can use this pragma to avoid such duplication. When a
7098 header file containing @samp{#pragma interface} is included in a
7099 compilation, this auxiliary information will not be generated (unless
7100 the main input source file itself uses @samp{#pragma implementation}).
7101 Instead, the object files will contain references to be resolved at link
7104 The second form of this directive is useful for the case where you have
7105 multiple headers with the same name in different directories. If you
7106 use this form, you must specify the same string to @samp{#pragma
7109 @item #pragma implementation
7110 @itemx #pragma implementation "@var{objects}.h"
7111 @kindex #pragma implementation
7112 Use this pragma in a @emph{main input file}, when you want full output from
7113 included header files to be generated (and made globally visible). The
7114 included header file, in turn, should use @samp{#pragma interface}.
7115 Backup copies of inline member functions, debugging information, and the
7116 internal tables used to implement virtual functions are all generated in
7117 implementation files.
7119 @cindex implied @code{#pragma implementation}
7120 @cindex @code{#pragma implementation}, implied
7121 @cindex naming convention, implementation headers
7122 If you use @samp{#pragma implementation} with no argument, it applies to
7123 an include file with the same basename@footnote{A file's @dfn{basename}
7124 was the name stripped of all leading path information and of trailing
7125 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
7126 file. For example, in @file{allclass.cc}, giving just
7127 @samp{#pragma implementation}
7128 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
7130 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
7131 an implementation file whenever you would include it from
7132 @file{allclass.cc} even if you never specified @samp{#pragma
7133 implementation}. This was deemed to be more trouble than it was worth,
7134 however, and disabled.
7136 If you use an explicit @samp{#pragma implementation}, it must appear in
7137 your source file @emph{before} you include the affected header files.
7139 Use the string argument if you want a single implementation file to
7140 include code from multiple header files. (You must also use
7141 @samp{#include} to include the header file; @samp{#pragma
7142 implementation} only specifies how to use the file---it doesn't actually
7145 There is no way to split up the contents of a single header file into
7146 multiple implementation files.
7149 @cindex inlining and C++ pragmas
7150 @cindex C++ pragmas, effect on inlining
7151 @cindex pragmas in C++, effect on inlining
7152 @samp{#pragma implementation} and @samp{#pragma interface} also have an
7153 effect on function inlining.
7155 If you define a class in a header file marked with @samp{#pragma
7156 interface}, the effect on a function defined in that class is similar to
7157 an explicit @code{extern} declaration---the compiler emits no code at
7158 all to define an independent version of the function. Its definition
7159 is used only for inlining with its callers.
7161 @opindex fno-implement-inlines
7162 Conversely, when you include the same header file in a main source file
7163 that declares it as @samp{#pragma implementation}, the compiler emits
7164 code for the function itself; this defines a version of the function
7165 that can be found via pointers (or by callers compiled without
7166 inlining). If all calls to the function can be inlined, you can avoid
7167 emitting the function by compiling with @option{-fno-implement-inlines}.
7168 If any calls were not inlined, you will get linker errors.
7170 @node Template Instantiation
7171 @section Where's the Template?
7173 @cindex template instantiation
7175 C++ templates are the first language feature to require more
7176 intelligence from the environment than one usually finds on a UNIX
7177 system. Somehow the compiler and linker have to make sure that each
7178 template instance occurs exactly once in the executable if it is needed,
7179 and not at all otherwise. There are two basic approaches to this
7180 problem, which I will refer to as the Borland model and the Cfront model.
7184 Borland C++ solved the template instantiation problem by adding the code
7185 equivalent of common blocks to their linker; the compiler emits template
7186 instances in each translation unit that uses them, and the linker
7187 collapses them together. The advantage of this model is that the linker
7188 only has to consider the object files themselves; there is no external
7189 complexity to worry about. This disadvantage is that compilation time
7190 is increased because the template code is being compiled repeatedly.
7191 Code written for this model tends to include definitions of all
7192 templates in the header file, since they must be seen to be
7196 The AT&T C++ translator, Cfront, solved the template instantiation
7197 problem by creating the notion of a template repository, an
7198 automatically maintained place where template instances are stored. A
7199 more modern version of the repository works as follows: As individual
7200 object files are built, the compiler places any template definitions and
7201 instantiations encountered in the repository. At link time, the link
7202 wrapper adds in the objects in the repository and compiles any needed
7203 instances that were not previously emitted. The advantages of this
7204 model are more optimal compilation speed and the ability to use the
7205 system linker; to implement the Borland model a compiler vendor also
7206 needs to replace the linker. The disadvantages are vastly increased
7207 complexity, and thus potential for error; for some code this can be
7208 just as transparent, but in practice it can been very difficult to build
7209 multiple programs in one directory and one program in multiple
7210 directories. Code written for this model tends to separate definitions
7211 of non-inline member templates into a separate file, which should be
7212 compiled separately.
7215 When used with GNU ld version 2.8 or later on an ELF system such as
7216 Linux/GNU or Solaris 2, or on Microsoft Windows, g++ supports the
7217 Borland model. On other systems, g++ implements neither automatic
7220 A future version of g++ will support a hybrid model whereby the compiler
7221 will emit any instantiations for which the template definition is
7222 included in the compile, and store template definitions and
7223 instantiation context information into the object file for the rest.
7224 The link wrapper will extract that information as necessary and invoke
7225 the compiler to produce the remaining instantiations. The linker will
7226 then combine duplicate instantiations.
7228 In the mean time, you have the following options for dealing with
7229 template instantiations:
7234 Compile your template-using code with @option{-frepo}. The compiler will
7235 generate files with the extension @samp{.rpo} listing all of the
7236 template instantiations used in the corresponding object files which
7237 could be instantiated there; the link wrapper, @samp{collect2}, will
7238 then update the @samp{.rpo} files to tell the compiler where to place
7239 those instantiations and rebuild any affected object files. The
7240 link-time overhead is negligible after the first pass, as the compiler
7241 will continue to place the instantiations in the same files.
7243 This is your best option for application code written for the Borland
7244 model, as it will just work. Code written for the Cfront model will
7245 need to be modified so that the template definitions are available at
7246 one or more points of instantiation; usually this is as simple as adding
7247 @code{#include <tmethods.cc>} to the end of each template header.
7249 For library code, if you want the library to provide all of the template
7250 instantiations it needs, just try to link all of its object files
7251 together; the link will fail, but cause the instantiations to be
7252 generated as a side effect. Be warned, however, that this may cause
7253 conflicts if multiple libraries try to provide the same instantiations.
7254 For greater control, use explicit instantiation as described in the next
7258 @opindex fno-implicit-templates
7259 Compile your code with @option{-fno-implicit-templates} to disable the
7260 implicit generation of template instances, and explicitly instantiate
7261 all the ones you use. This approach requires more knowledge of exactly
7262 which instances you need than do the others, but it's less
7263 mysterious and allows greater control. You can scatter the explicit
7264 instantiations throughout your program, perhaps putting them in the
7265 translation units where the instances are used or the translation units
7266 that define the templates themselves; you can put all of the explicit
7267 instantiations you need into one big file; or you can create small files
7274 template class Foo<int>;
7275 template ostream& operator <<
7276 (ostream&, const Foo<int>&);
7279 for each of the instances you need, and create a template instantiation
7282 If you are using Cfront-model code, you can probably get away with not
7283 using @option{-fno-implicit-templates} when compiling files that don't
7284 @samp{#include} the member template definitions.
7286 If you use one big file to do the instantiations, you may want to
7287 compile it without @option{-fno-implicit-templates} so you get all of the
7288 instances required by your explicit instantiations (but not by any
7289 other files) without having to specify them as well.
7291 g++ has extended the template instantiation syntax outlined in the
7292 Working Paper to allow forward declaration of explicit instantiations
7293 (with @code{extern}), instantiation of the compiler support data for a
7294 template class (i.e.@: the vtable) without instantiating any of its
7295 members (with @code{inline}), and instantiation of only the static data
7296 members of a template class, without the support data or member
7297 functions (with (@code{static}):
7300 extern template int max (int, int);
7301 inline template class Foo<int>;
7302 static template class Foo<int>;
7306 Do nothing. Pretend g++ does implement automatic instantiation
7307 management. Code written for the Borland model will work fine, but
7308 each translation unit will contain instances of each of the templates it
7309 uses. In a large program, this can lead to an unacceptable amount of code
7312 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7313 more discussion of these pragmas.
7316 @node Bound member functions
7317 @section Extracting the function pointer from a bound pointer to member function
7320 @cindex pointer to member function
7321 @cindex bound pointer to member function
7323 In C++, pointer to member functions (PMFs) are implemented using a wide
7324 pointer of sorts to handle all the possible call mechanisms; the PMF
7325 needs to store information about how to adjust the @samp{this} pointer,
7326 and if the function pointed to is virtual, where to find the vtable, and
7327 where in the vtable to look for the member function. If you are using
7328 PMFs in an inner loop, you should really reconsider that decision. If
7329 that is not an option, you can extract the pointer to the function that
7330 would be called for a given object/PMF pair and call it directly inside
7331 the inner loop, to save a bit of time.
7333 Note that you will still be paying the penalty for the call through a
7334 function pointer; on most modern architectures, such a call defeats the
7335 branch prediction features of the CPU@. This is also true of normal
7336 virtual function calls.
7338 The syntax for this extension is
7342 extern int (A::*fp)();
7343 typedef int (*fptr)(A *);
7345 fptr p = (fptr)(a.*fp);
7348 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
7349 no object is needed to obtain the address of the function. They can be
7350 converted to function pointers directly:
7353 fptr p1 = (fptr)(&A::foo);
7356 @opindex Wno-pmf-conversions
7357 You must specify @option{-Wno-pmf-conversions} to use this extension.
7359 @node C++ Attributes
7360 @section C++-Specific Variable, Function, and Type Attributes
7362 Some attributes only make sense for C++ programs.
7365 @item init_priority (@var{priority})
7366 @cindex init_priority attribute
7369 In Standard C++, objects defined at namespace scope are guaranteed to be
7370 initialized in an order in strict accordance with that of their definitions
7371 @emph{in a given translation unit}. No guarantee is made for initializations
7372 across translation units. However, GNU C++ allows users to control the
7373 order of initialization of objects defined at namespace scope with the
7374 @code{init_priority} attribute by specifying a relative @var{priority},
7375 a constant integral expression currently bounded between 101 and 65535
7376 inclusive. Lower numbers indicate a higher priority.
7378 In the following example, @code{A} would normally be created before
7379 @code{B}, but the @code{init_priority} attribute has reversed that order:
7382 Some_Class A __attribute__ ((init_priority (2000)));
7383 Some_Class B __attribute__ ((init_priority (543)));
7387 Note that the particular values of @var{priority} do not matter; only their
7390 @item java_interface
7391 @cindex java_interface attribute
7393 This type attribute informs C++ that the class is a Java interface. It may
7394 only be applied to classes declared within an @code{extern "Java"} block.
7395 Calls to methods declared in this interface will be dispatched using GCJ's
7396 interface table mechanism, instead of regular virtual table dispatch.
7400 @node Java Exceptions
7401 @section Java Exceptions
7403 The Java language uses a slightly different exception handling model
7404 from C++. Normally, GNU C++ will automatically detect when you are
7405 writing C++ code that uses Java exceptions, and handle them
7406 appropriately. However, if C++ code only needs to execute destructors
7407 when Java exceptions are thrown through it, GCC will guess incorrectly.
7408 Sample problematic code is:
7411 struct S @{ ~S(); @};
7412 extern void bar(); // is written in Java, and may throw exceptions
7421 The usual effect of an incorrect guess is a link failure, complaining of
7422 a missing routine called @samp{__gxx_personality_v0}.
7424 You can inform the compiler that Java exceptions are to be used in a
7425 translation unit, irrespective of what it might think, by writing
7426 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
7427 @samp{#pragma} must appear before any functions that throw or catch
7428 exceptions, or run destructors when exceptions are thrown through them.
7430 You cannot mix Java and C++ exceptions in the same translation unit. It
7431 is believed to be safe to throw a C++ exception from one file through
7432 another file compiled for the Java exception model, or vice versa, but
7433 there may be bugs in this area.
7435 @node Deprecated Features
7436 @section Deprecated Features
7438 In the past, the GNU C++ compiler was extended to experiment with new
7439 features, at a time when the C++ language was still evolving. Now that
7440 the C++ standard is complete, some of those features are superseded by
7441 superior alternatives. Using the old features might cause a warning in
7442 some cases that the feature will be dropped in the future. In other
7443 cases, the feature might be gone already.
7445 While the list below is not exhaustive, it documents some of the options
7446 that are now deprecated:
7449 @item -fexternal-templates
7450 @itemx -falt-external-templates
7451 These are two of the many ways for g++ to implement template
7452 instantiation. @xref{Template Instantiation}. The C++ standard clearly
7453 defines how template definitions have to be organized across
7454 implementation units. g++ has an implicit instantiation mechanism that
7455 should work just fine for standard-conforming code.
7457 @item -fstrict-prototype
7458 @itemx -fno-strict-prototype
7459 Previously it was possible to use an empty prototype parameter list to
7460 indicate an unspecified number of parameters (like C), rather than no
7461 parameters, as C++ demands. This feature has been removed, except where
7462 it is required for backwards compatibility @xref{Backwards Compatibility}.
7465 The named return value extension has been deprecated, and is now
7468 The use of initializer lists with new expressions has been deprecated,
7469 and is now removed from g++.
7471 Floating and complex non-type template parameters have been deprecated,
7472 and are now removed from g++.
7474 The implicit typename extension has been deprecated and will be removed
7475 from g++ at some point. In some cases g++ determines that a dependent
7476 type such as @code{TPL<T>::X} is a type without needing a
7477 @code{typename} keyword, contrary to the standard.
7479 @node Backwards Compatibility
7480 @section Backwards Compatibility
7481 @cindex Backwards Compatibility
7482 @cindex ARM [Annotated C++ Reference Manual]
7484 Now that there is a definitive ISO standard C++, G++ has a specification
7485 to adhere to. The C++ language evolved over time, and features that
7486 used to be acceptable in previous drafts of the standard, such as the ARM
7487 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
7488 compilation of C++ written to such drafts, G++ contains some backwards
7489 compatibilities. @emph{All such backwards compatibility features are
7490 liable to disappear in future versions of G++.} They should be considered
7491 deprecated @xref{Deprecated Features}.
7495 If a variable is declared at for scope, it used to remain in scope until
7496 the end of the scope which contained the for statement (rather than just
7497 within the for scope). G++ retains this, but issues a warning, if such a
7498 variable is accessed outside the for scope.
7500 @item Implicit C language
7501 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
7502 scope to set the language. On such systems, all header files are
7503 implicitly scoped inside a C language scope. Also, an empty prototype
7504 @code{()} will be treated as an unspecified number of arguments, rather
7505 than no arguments, as C++ demands.