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 * Naming Types:: Giving a name to the type of some expression.
431 * Typeof:: @code{typeof}: referring to the type of an expression.
432 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
433 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
434 * Long Long:: Double-word integers---@code{long long int}.
435 * Complex:: Data types for complex numbers.
436 * Hex Floats:: Hexadecimal floating-point constants.
437 * Zero Length:: Zero-length arrays.
438 * Variable Length:: Arrays whose length is computed at run time.
439 * Variadic Macros:: Macros with a variable number of arguments.
440 * Escaped Newlines:: Slightly looser rules for escaped newlines.
441 * Multi-line Strings:: String literals with embedded newlines.
442 * Subscripting:: Any array can be subscripted, even if not an lvalue.
443 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
444 * Initializers:: Non-constant initializers.
445 * Compound Literals:: Compound literals give structures, unions
447 * Designated Inits:: Labeling elements of initializers.
448 * Cast to Union:: Casting to union type from any member of the union.
449 * Case Ranges:: `case 1 ... 9' and such.
450 * Mixed Declarations:: Mixing declarations and code.
451 * Function Attributes:: Declaring that functions have no side effects,
452 or that they can never return.
453 * Attribute Syntax:: Formal syntax for attributes.
454 * Function Prototypes:: Prototype declarations and old-style definitions.
455 * C++ Comments:: C++ comments are recognized.
456 * Dollar Signs:: Dollar sign is allowed in identifiers.
457 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
458 * Variable Attributes:: Specifying attributes of variables.
459 * Type Attributes:: Specifying attributes of types.
460 * Alignment:: Inquiring about the alignment of a type or variable.
461 * Inline:: Defining inline functions (as fast as macros).
462 * Extended Asm:: Assembler instructions with C expressions as operands.
463 (With them you can define ``built-in'' functions.)
464 * Constraints:: Constraints for asm operands
465 * Asm Labels:: Specifying the assembler name to use for a C symbol.
466 * Explicit Reg Vars:: Defining variables residing in specified registers.
467 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
468 * Incomplete Enums:: @code{enum foo;}, with details to follow.
469 * Function Names:: Printable strings which are the name of the current
471 * Return Address:: Getting the return or frame address of a function.
472 * Vector Extensions:: Using vector instructions through built-in functions.
473 * Other Builtins:: Other built-in functions.
474 * Target Builtins:: Built-in functions specific to particular targets.
475 * Pragmas:: Pragmas accepted by GCC.
476 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
477 * Thread-Local:: Per-thread variables.
480 @node Statement Exprs
481 @section Statements and Declarations in Expressions
482 @cindex statements inside expressions
483 @cindex declarations inside expressions
484 @cindex expressions containing statements
485 @cindex macros, statements in expressions
487 @c the above section title wrapped and causes an underfull hbox.. i
488 @c changed it from "within" to "in". --mew 4feb93
490 A compound statement enclosed in parentheses may appear as an expression
491 in GNU C@. This allows you to use loops, switches, and local variables
492 within an expression.
494 Recall that a compound statement is a sequence of statements surrounded
495 by braces; in this construct, parentheses go around the braces. For
499 (@{ int y = foo (); int z;
506 is a valid (though slightly more complex than necessary) expression
507 for the absolute value of @code{foo ()}.
509 The last thing in the compound statement should be an expression
510 followed by a semicolon; the value of this subexpression serves as the
511 value of the entire construct. (If you use some other kind of statement
512 last within the braces, the construct has type @code{void}, and thus
513 effectively no value.)
515 This feature is especially useful in making macro definitions ``safe'' (so
516 that they evaluate each operand exactly once). For example, the
517 ``maximum'' function is commonly defined as a macro in standard C as
521 #define max(a,b) ((a) > (b) ? (a) : (b))
525 @cindex side effects, macro argument
526 But this definition computes either @var{a} or @var{b} twice, with bad
527 results if the operand has side effects. In GNU C, if you know the
528 type of the operands (here let's assume @code{int}), you can define
529 the macro safely as follows:
532 #define maxint(a,b) \
533 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
536 Embedded statements are not allowed in constant expressions, such as
537 the value of an enumeration constant, the width of a bit-field, or
538 the initial value of a static variable.
540 If you don't know the type of the operand, you can still do this, but you
541 must use @code{typeof} (@pxref{Typeof}) or type naming (@pxref{Naming
544 Statement expressions are not supported fully in G++, and their fate
545 there is unclear. (It is possible that they will become fully supported
546 at some point, or that they will be deprecated, or that the bugs that
547 are present will continue to exist indefinitely.) Presently, statement
548 expressions do not work well as default arguments.
550 In addition, there are semantic issues with statement-expressions in
551 C++. If you try to use statement-expressions instead of inline
552 functions in C++, you may be surprised at the way object destruction is
553 handled. For example:
556 #define foo(a) (@{int b = (a); b + 3; @})
560 does not work the same way as:
563 inline int foo(int a) @{ int b = a; return b + 3; @}
567 In particular, if the expression passed into @code{foo} involves the
568 creation of temporaries, the destructors for those temporaries will be
569 run earlier in the case of the macro than in the case of the function.
571 These considerations mean that it is probably a bad idea to use
572 statement-expressions of this form in header files that are designed to
573 work with C++. (Note that some versions of the GNU C Library contained
574 header files using statement-expression that lead to precisely this
578 @section Locally Declared Labels
580 @cindex macros, local labels
582 Each statement expression is a scope in which @dfn{local labels} can be
583 declared. A local label is simply an identifier; you can jump to it
584 with an ordinary @code{goto} statement, but only from within the
585 statement expression it belongs to.
587 A local label declaration looks like this:
590 __label__ @var{label};
597 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
600 Local label declarations must come at the beginning of the statement
601 expression, right after the @samp{(@{}, before any ordinary
604 The label declaration defines the label @emph{name}, but does not define
605 the label itself. You must do this in the usual way, with
606 @code{@var{label}:}, within the statements of the statement expression.
608 The local label feature is useful because statement expressions are
609 often used in macros. If the macro contains nested loops, a @code{goto}
610 can be useful for breaking out of them. However, an ordinary label
611 whose scope is the whole function cannot be used: if the macro can be
612 expanded several times in one function, the label will be multiply
613 defined in that function. A local label avoids this problem. For
617 #define SEARCH(array, target) \
620 typeof (target) _SEARCH_target = (target); \
621 typeof (*(array)) *_SEARCH_array = (array); \
624 for (i = 0; i < max; i++) \
625 for (j = 0; j < max; j++) \
626 if (_SEARCH_array[i][j] == _SEARCH_target) \
627 @{ value = i; goto found; @} \
634 @node Labels as Values
635 @section Labels as Values
636 @cindex labels as values
637 @cindex computed gotos
638 @cindex goto with computed label
639 @cindex address of a label
641 You can get the address of a label defined in the current function
642 (or a containing function) with the unary operator @samp{&&}. The
643 value has type @code{void *}. This value is a constant and can be used
644 wherever a constant of that type is valid. For example:
652 To use these values, you need to be able to jump to one. This is done
653 with the computed goto statement@footnote{The analogous feature in
654 Fortran is called an assigned goto, but that name seems inappropriate in
655 C, where one can do more than simply store label addresses in label
656 variables.}, @code{goto *@var{exp};}. For example,
663 Any expression of type @code{void *} is allowed.
665 One way of using these constants is in initializing a static array that
666 will serve as a jump table:
669 static void *array[] = @{ &&foo, &&bar, &&hack @};
672 Then you can select a label with indexing, like this:
679 Note that this does not check whether the subscript is in bounds---array
680 indexing in C never does that.
682 Such an array of label values serves a purpose much like that of the
683 @code{switch} statement. The @code{switch} statement is cleaner, so
684 use that rather than an array unless the problem does not fit a
685 @code{switch} statement very well.
687 Another use of label values is in an interpreter for threaded code.
688 The labels within the interpreter function can be stored in the
689 threaded code for super-fast dispatching.
691 You may not use this mechanism to jump to code in a different function.
692 If you do that, totally unpredictable things will happen. The best way to
693 avoid this is to store the label address only in automatic variables and
694 never pass it as an argument.
696 An alternate way to write the above example is
699 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
701 goto *(&&foo + array[i]);
705 This is more friendly to code living in shared libraries, as it reduces
706 the number of dynamic relocations that are needed, and by consequence,
707 allows the data to be read-only.
709 @node Nested Functions
710 @section Nested Functions
711 @cindex nested functions
712 @cindex downward funargs
715 A @dfn{nested function} is a function defined inside another function.
716 (Nested functions are not supported for GNU C++.) The nested function's
717 name is local to the block where it is defined. For example, here we
718 define a nested function named @code{square}, and call it twice:
722 foo (double a, double b)
724 double square (double z) @{ return z * z; @}
726 return square (a) + square (b);
731 The nested function can access all the variables of the containing
732 function that are visible at the point of its definition. This is
733 called @dfn{lexical scoping}. For example, here we show a nested
734 function which uses an inherited variable named @code{offset}:
738 bar (int *array, int offset, int size)
740 int access (int *array, int index)
741 @{ return array[index + offset]; @}
744 for (i = 0; i < size; i++)
745 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
750 Nested function definitions are permitted within functions in the places
751 where variable definitions are allowed; that is, in any block, before
752 the first statement in the block.
754 It is possible to call the nested function from outside the scope of its
755 name by storing its address or passing the address to another function:
758 hack (int *array, int size)
760 void store (int index, int value)
761 @{ array[index] = value; @}
763 intermediate (store, size);
767 Here, the function @code{intermediate} receives the address of
768 @code{store} as an argument. If @code{intermediate} calls @code{store},
769 the arguments given to @code{store} are used to store into @code{array}.
770 But this technique works only so long as the containing function
771 (@code{hack}, in this example) does not exit.
773 If you try to call the nested function through its address after the
774 containing function has exited, all hell will break loose. If you try
775 to call it after a containing scope level has exited, and if it refers
776 to some of the variables that are no longer in scope, you may be lucky,
777 but it's not wise to take the risk. If, however, the nested function
778 does not refer to anything that has gone out of scope, you should be
781 GCC implements taking the address of a nested function using a technique
782 called @dfn{trampolines}. A paper describing them is available as
785 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
787 A nested function can jump to a label inherited from a containing
788 function, provided the label was explicitly declared in the containing
789 function (@pxref{Local Labels}). Such a jump returns instantly to the
790 containing function, exiting the nested function which did the
791 @code{goto} and any intermediate functions as well. Here is an example:
795 bar (int *array, int offset, int size)
798 int access (int *array, int index)
802 return array[index + offset];
806 for (i = 0; i < size; i++)
807 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
811 /* @r{Control comes here from @code{access}
812 if it detects an error.} */
819 A nested function always has internal linkage. Declaring one with
820 @code{extern} is erroneous. If you need to declare the nested function
821 before its definition, use @code{auto} (which is otherwise meaningless
822 for function declarations).
825 bar (int *array, int offset, int size)
828 auto int access (int *, int);
830 int access (int *array, int index)
834 return array[index + offset];
840 @node Constructing Calls
841 @section Constructing Function Calls
842 @cindex constructing calls
843 @cindex forwarding calls
845 Using the built-in functions described below, you can record
846 the arguments a function received, and call another function
847 with the same arguments, without knowing the number or types
850 You can also record the return value of that function call,
851 and later return that value, without knowing what data type
852 the function tried to return (as long as your caller expects
855 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
856 This built-in function returns a pointer to data
857 describing how to perform a call with the same arguments as were passed
858 to the current function.
860 The function saves the arg pointer register, structure value address,
861 and all registers that might be used to pass arguments to a function
862 into a block of memory allocated on the stack. Then it returns the
863 address of that block.
866 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
867 This built-in function invokes @var{function}
868 with a copy of the parameters described by @var{arguments}
871 The value of @var{arguments} should be the value returned by
872 @code{__builtin_apply_args}. The argument @var{size} specifies the size
873 of the stack argument data, in bytes.
875 This function returns a pointer to data describing
876 how to return whatever value was returned by @var{function}. The data
877 is saved in a block of memory allocated on the stack.
879 It is not always simple to compute the proper value for @var{size}. The
880 value is used by @code{__builtin_apply} to compute the amount of data
881 that should be pushed on the stack and copied from the incoming argument
885 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
886 This built-in function returns the value described by @var{result} from
887 the containing function. You should specify, for @var{result}, a value
888 returned by @code{__builtin_apply}.
892 @section Naming an Expression's Type
895 You can give a name to the type of an expression using a @code{typedef}
896 declaration with an initializer. Here is how to define @var{name} as a
897 type name for the type of @var{exp}:
900 typedef @var{name} = @var{exp};
903 This is useful in conjunction with the statements-within-expressions
904 feature. Here is how the two together can be used to define a safe
905 ``maximum'' macro that operates on any arithmetic type:
909 (@{typedef _ta = (a), _tb = (b); \
910 _ta _a = (a); _tb _b = (b); \
911 _a > _b ? _a : _b; @})
914 @cindex underscores in variables in macros
915 @cindex @samp{_} in variables in macros
916 @cindex local variables in macros
917 @cindex variables, local, in macros
918 @cindex macros, local variables in
920 The reason for using names that start with underscores for the local
921 variables is to avoid conflicts with variable names that occur within the
922 expressions that are substituted for @code{a} and @code{b}. Eventually we
923 hope to design a new form of declaration syntax that allows you to declare
924 variables whose scopes start only after their initializers; this will be a
925 more reliable way to prevent such conflicts.
928 @section Referring to a Type with @code{typeof}
931 @cindex macros, types of arguments
933 Another way to refer to the type of an expression is with @code{typeof}.
934 The syntax of using of this keyword looks like @code{sizeof}, but the
935 construct acts semantically like a type name defined with @code{typedef}.
937 There are two ways of writing the argument to @code{typeof}: with an
938 expression or with a type. Here is an example with an expression:
945 This assumes that @code{x} is an array of pointers to functions;
946 the type described is that of the values of the functions.
948 Here is an example with a typename as the argument:
955 Here the type described is that of pointers to @code{int}.
957 If you are writing a header file that must work when included in ISO C
958 programs, write @code{__typeof__} instead of @code{typeof}.
959 @xref{Alternate Keywords}.
961 A @code{typeof}-construct can be used anywhere a typedef name could be
962 used. For example, you can use it in a declaration, in a cast, or inside
963 of @code{sizeof} or @code{typeof}.
967 This declares @code{y} with the type of what @code{x} points to.
974 This declares @code{y} as an array of such values.
981 This declares @code{y} as an array of pointers to characters:
984 typeof (typeof (char *)[4]) y;
988 It is equivalent to the following traditional C declaration:
994 To see the meaning of the declaration using @code{typeof}, and why it
995 might be a useful way to write, let's rewrite it with these macros:
998 #define pointer(T) typeof(T *)
999 #define array(T, N) typeof(T [N])
1003 Now the declaration can be rewritten this way:
1006 array (pointer (char), 4) y;
1010 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1011 pointers to @code{char}.
1015 @section Generalized Lvalues
1016 @cindex compound expressions as lvalues
1017 @cindex expressions, compound, as lvalues
1018 @cindex conditional expressions as lvalues
1019 @cindex expressions, conditional, as lvalues
1020 @cindex casts as lvalues
1021 @cindex generalized lvalues
1022 @cindex lvalues, generalized
1023 @cindex extensions, @code{?:}
1024 @cindex @code{?:} extensions
1025 Compound expressions, conditional expressions and casts are allowed as
1026 lvalues provided their operands are lvalues. This means that you can take
1027 their addresses or store values into them.
1029 Standard C++ allows compound expressions and conditional expressions as
1030 lvalues, and permits casts to reference type, so use of this extension
1031 is deprecated for C++ code.
1033 For example, a compound expression can be assigned, provided the last
1034 expression in the sequence is an lvalue. These two expressions are
1042 Similarly, the address of the compound expression can be taken. These two
1043 expressions are equivalent:
1050 A conditional expression is a valid lvalue if its type is not void and the
1051 true and false branches are both valid lvalues. For example, these two
1052 expressions are equivalent:
1056 (a ? b = 5 : (c = 5))
1059 A cast is a valid lvalue if its operand is an lvalue. A simple
1060 assignment whose left-hand side is a cast works by converting the
1061 right-hand side first to the specified type, then to the type of the
1062 inner left-hand side expression. After this is stored, the value is
1063 converted back to the specified type to become the value of the
1064 assignment. Thus, if @code{a} has type @code{char *}, the following two
1065 expressions are equivalent:
1069 (int)(a = (char *)(int)5)
1072 An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
1073 performs the arithmetic using the type resulting from the cast, and then
1074 continues as in the previous case. Therefore, these two expressions are
1079 (int)(a = (char *)(int) ((int)a + 5))
1082 You cannot take the address of an lvalue cast, because the use of its
1083 address would not work out coherently. Suppose that @code{&(int)f} were
1084 permitted, where @code{f} has type @code{float}. Then the following
1085 statement would try to store an integer bit-pattern where a floating
1086 point number belongs:
1092 This is quite different from what @code{(int)f = 1} would do---that
1093 would convert 1 to floating point and store it. Rather than cause this
1094 inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
1096 If you really do want an @code{int *} pointer with the address of
1097 @code{f}, you can simply write @code{(int *)&f}.
1100 @section Conditionals with Omitted Operands
1101 @cindex conditional expressions, extensions
1102 @cindex omitted middle-operands
1103 @cindex middle-operands, omitted
1104 @cindex extensions, @code{?:}
1105 @cindex @code{?:} extensions
1107 The middle operand in a conditional expression may be omitted. Then
1108 if the first operand is nonzero, its value is the value of the conditional
1111 Therefore, the expression
1118 has the value of @code{x} if that is nonzero; otherwise, the value of
1121 This example is perfectly equivalent to
1127 @cindex side effect in ?:
1128 @cindex ?: side effect
1130 In this simple case, the ability to omit the middle operand is not
1131 especially useful. When it becomes useful is when the first operand does,
1132 or may (if it is a macro argument), contain a side effect. Then repeating
1133 the operand in the middle would perform the side effect twice. Omitting
1134 the middle operand uses the value already computed without the undesirable
1135 effects of recomputing it.
1138 @section Double-Word Integers
1139 @cindex @code{long long} data types
1140 @cindex double-word arithmetic
1141 @cindex multiprecision arithmetic
1142 @cindex @code{LL} integer suffix
1143 @cindex @code{ULL} integer suffix
1145 ISO C99 supports data types for integers that are at least 64 bits wide,
1146 and as an extension GCC supports them in C89 mode and in C++.
1147 Simply write @code{long long int} for a signed integer, or
1148 @code{unsigned long long int} for an unsigned integer. To make an
1149 integer constant of type @code{long long int}, add the suffix @samp{LL}
1150 to the integer. To make an integer constant of type @code{unsigned long
1151 long int}, add the suffix @samp{ULL} to the integer.
1153 You can use these types in arithmetic like any other integer types.
1154 Addition, subtraction, and bitwise boolean operations on these types
1155 are open-coded on all types of machines. Multiplication is open-coded
1156 if the machine supports fullword-to-doubleword a widening multiply
1157 instruction. Division and shifts are open-coded only on machines that
1158 provide special support. The operations that are not open-coded use
1159 special library routines that come with GCC@.
1161 There may be pitfalls when you use @code{long long} types for function
1162 arguments, unless you declare function prototypes. If a function
1163 expects type @code{int} for its argument, and you pass a value of type
1164 @code{long long int}, confusion will result because the caller and the
1165 subroutine will disagree about the number of bytes for the argument.
1166 Likewise, if the function expects @code{long long int} and you pass
1167 @code{int}. The best way to avoid such problems is to use prototypes.
1170 @section Complex Numbers
1171 @cindex complex numbers
1172 @cindex @code{_Complex} keyword
1173 @cindex @code{__complex__} keyword
1175 ISO C99 supports complex floating data types, and as an extension GCC
1176 supports them in C89 mode and in C++, and supports complex integer data
1177 types which are not part of ISO C99. You can declare complex types
1178 using the keyword @code{_Complex}. As an extension, the older GNU
1179 keyword @code{__complex__} is also supported.
1181 For example, @samp{_Complex double x;} declares @code{x} as a
1182 variable whose real part and imaginary part are both of type
1183 @code{double}. @samp{_Complex short int y;} declares @code{y} to
1184 have real and imaginary parts of type @code{short int}; this is not
1185 likely to be useful, but it shows that the set of complex types is
1188 To write a constant with a complex data type, use the suffix @samp{i} or
1189 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
1190 has type @code{_Complex float} and @code{3i} has type
1191 @code{_Complex int}. Such a constant always has a pure imaginary
1192 value, but you can form any complex value you like by adding one to a
1193 real constant. This is a GNU extension; if you have an ISO C99
1194 conforming C library (such as GNU libc), and want to construct complex
1195 constants of floating type, you should include @code{<complex.h>} and
1196 use the macros @code{I} or @code{_Complex_I} instead.
1198 @cindex @code{__real__} keyword
1199 @cindex @code{__imag__} keyword
1200 To extract the real part of a complex-valued expression @var{exp}, write
1201 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
1202 extract the imaginary part. This is a GNU extension; for values of
1203 floating type, you should use the ISO C99 functions @code{crealf},
1204 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1205 @code{cimagl}, declared in @code{<complex.h>} and also provided as
1206 built-in functions by GCC@.
1208 @cindex complex conjugation
1209 The operator @samp{~} performs complex conjugation when used on a value
1210 with a complex type. This is a GNU extension; for values of
1211 floating type, you should use the ISO C99 functions @code{conjf},
1212 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1213 provided as built-in functions by GCC@.
1215 GCC can allocate complex automatic variables in a noncontiguous
1216 fashion; it's even possible for the real part to be in a register while
1217 the imaginary part is on the stack (or vice-versa). None of the
1218 supported debugging info formats has a way to represent noncontiguous
1219 allocation like this, so GCC describes a noncontiguous complex
1220 variable as if it were two separate variables of noncomplex type.
1221 If the variable's actual name is @code{foo}, the two fictitious
1222 variables are named @code{foo$real} and @code{foo$imag}. You can
1223 examine and set these two fictitious variables with your debugger.
1225 A future version of GDB will know how to recognize such pairs and treat
1226 them as a single variable with a complex type.
1232 ISO C99 supports floating-point numbers written not only in the usual
1233 decimal notation, such as @code{1.55e1}, but also numbers such as
1234 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1235 supports this in C89 mode (except in some cases when strictly
1236 conforming) and in C++. In that format the
1237 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1238 mandatory. The exponent is a decimal number that indicates the power of
1239 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1246 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1247 is the same as @code{1.55e1}.
1249 Unlike for floating-point numbers in the decimal notation the exponent
1250 is always required in the hexadecimal notation. Otherwise the compiler
1251 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1252 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1253 extension for floating-point constants of type @code{float}.
1256 @section Arrays of Length Zero
1257 @cindex arrays of length zero
1258 @cindex zero-length arrays
1259 @cindex length-zero arrays
1260 @cindex flexible array members
1262 Zero-length arrays are allowed in GNU C@. They are very useful as the
1263 last element of a structure which is really a header for a variable-length
1272 struct line *thisline = (struct line *)
1273 malloc (sizeof (struct line) + this_length);
1274 thisline->length = this_length;
1277 In ISO C90, you would have to give @code{contents} a length of 1, which
1278 means either you waste space or complicate the argument to @code{malloc}.
1280 In ISO C99, you would use a @dfn{flexible array member}, which is
1281 slightly different in syntax and semantics:
1285 Flexible array members are written as @code{contents[]} without
1289 Flexible array members have incomplete type, and so the @code{sizeof}
1290 operator may not be applied. As a quirk of the original implementation
1291 of zero-length arrays, @code{sizeof} evaluates to zero.
1294 Flexible array members may only appear as the last member of a
1295 @code{struct} that is otherwise non-empty.
1298 A structure containing a flexible array member, or a union containing
1299 such a structure (possibly recursively), may not be a member of a
1300 structure or an element of an array. (However, these uses are
1301 permitted by GCC as extensions.)
1304 GCC versions before 3.0 allowed zero-length arrays to be statically
1305 initialized, as if they were flexible arrays. In addition to those
1306 cases that were useful, it also allowed initializations in situations
1307 that would corrupt later data. Non-empty initialization of zero-length
1308 arrays is now treated like any case where there are more initializer
1309 elements than the array holds, in that a suitable warning about "excess
1310 elements in array" is given, and the excess elements (all of them, in
1311 this case) are ignored.
1313 Instead GCC allows static initialization of flexible array members.
1314 This is equivalent to defining a new structure containing the original
1315 structure followed by an array of sufficient size to contain the data.
1316 I.e.@: in the following, @code{f1} is constructed as if it were declared
1322 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1325 struct f1 f1; int data[3];
1326 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1330 The convenience of this extension is that @code{f1} has the desired
1331 type, eliminating the need to consistently refer to @code{f2.f1}.
1333 This has symmetry with normal static arrays, in that an array of
1334 unknown size is also written with @code{[]}.
1336 Of course, this extension only makes sense if the extra data comes at
1337 the end of a top-level object, as otherwise we would be overwriting
1338 data at subsequent offsets. To avoid undue complication and confusion
1339 with initialization of deeply nested arrays, we simply disallow any
1340 non-empty initialization except when the structure is the top-level
1341 object. For example:
1344 struct foo @{ int x; int y[]; @};
1345 struct bar @{ struct foo z; @};
1347 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1348 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1349 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1350 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1353 @node Variable Length
1354 @section Arrays of Variable Length
1355 @cindex variable-length arrays
1356 @cindex arrays of variable length
1359 Variable-length automatic arrays are allowed in ISO C99, and as an
1360 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1361 implementation of variable-length arrays does not yet conform in detail
1362 to the ISO C99 standard.) These arrays are
1363 declared like any other automatic arrays, but with a length that is not
1364 a constant expression. The storage is allocated at the point of
1365 declaration and deallocated when the brace-level is exited. For
1370 concat_fopen (char *s1, char *s2, char *mode)
1372 char str[strlen (s1) + strlen (s2) + 1];
1375 return fopen (str, mode);
1379 @cindex scope of a variable length array
1380 @cindex variable-length array scope
1381 @cindex deallocating variable length arrays
1382 Jumping or breaking out of the scope of the array name deallocates the
1383 storage. Jumping into the scope is not allowed; you get an error
1386 @cindex @code{alloca} vs variable-length arrays
1387 You can use the function @code{alloca} to get an effect much like
1388 variable-length arrays. The function @code{alloca} is available in
1389 many other C implementations (but not in all). On the other hand,
1390 variable-length arrays are more elegant.
1392 There are other differences between these two methods. Space allocated
1393 with @code{alloca} exists until the containing @emph{function} returns.
1394 The space for a variable-length array is deallocated as soon as the array
1395 name's scope ends. (If you use both variable-length arrays and
1396 @code{alloca} in the same function, deallocation of a variable-length array
1397 will also deallocate anything more recently allocated with @code{alloca}.)
1399 You can also use variable-length arrays as arguments to functions:
1403 tester (int len, char data[len][len])
1409 The length of an array is computed once when the storage is allocated
1410 and is remembered for the scope of the array in case you access it with
1413 If you want to pass the array first and the length afterward, you can
1414 use a forward declaration in the parameter list---another GNU extension.
1418 tester (int len; char data[len][len], int len)
1424 @cindex parameter forward declaration
1425 The @samp{int len} before the semicolon is a @dfn{parameter forward
1426 declaration}, and it serves the purpose of making the name @code{len}
1427 known when the declaration of @code{data} is parsed.
1429 You can write any number of such parameter forward declarations in the
1430 parameter list. They can be separated by commas or semicolons, but the
1431 last one must end with a semicolon, which is followed by the ``real''
1432 parameter declarations. Each forward declaration must match a ``real''
1433 declaration in parameter name and data type. ISO C99 does not support
1434 parameter forward declarations.
1436 @node Variadic Macros
1437 @section Macros with a Variable Number of Arguments.
1438 @cindex variable number of arguments
1439 @cindex macro with variable arguments
1440 @cindex rest argument (in macro)
1441 @cindex variadic macros
1443 In the ISO C standard of 1999, a macro can be declared to accept a
1444 variable number of arguments much as a function can. The syntax for
1445 defining the macro is similar to that of a function. Here is an
1449 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1452 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1453 such a macro, it represents the zero or more tokens until the closing
1454 parenthesis that ends the invocation, including any commas. This set of
1455 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1456 wherever it appears. See the CPP manual for more information.
1458 GCC has long supported variadic macros, and used a different syntax that
1459 allowed you to give a name to the variable arguments just like any other
1460 argument. Here is an example:
1463 #define debug(format, args...) fprintf (stderr, format, args)
1466 This is in all ways equivalent to the ISO C example above, but arguably
1467 more readable and descriptive.
1469 GNU CPP has two further variadic macro extensions, and permits them to
1470 be used with either of the above forms of macro definition.
1472 In standard C, you are not allowed to leave the variable argument out
1473 entirely; but you are allowed to pass an empty argument. For example,
1474 this invocation is invalid in ISO C, because there is no comma after
1481 GNU CPP permits you to completely omit the variable arguments in this
1482 way. In the above examples, the compiler would complain, though since
1483 the expansion of the macro still has the extra comma after the format
1486 To help solve this problem, CPP behaves specially for variable arguments
1487 used with the token paste operator, @samp{##}. If instead you write
1490 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1493 and if the variable arguments are omitted or empty, the @samp{##}
1494 operator causes the preprocessor to remove the comma before it. If you
1495 do provide some variable arguments in your macro invocation, GNU CPP
1496 does not complain about the paste operation and instead places the
1497 variable arguments after the comma. Just like any other pasted macro
1498 argument, these arguments are not macro expanded.
1500 @node Escaped Newlines
1501 @section Slightly Looser Rules for Escaped Newlines
1502 @cindex escaped newlines
1503 @cindex newlines (escaped)
1505 Recently, the preprocessor has relaxed its treatment of escaped
1506 newlines. Previously, the newline had to immediately follow a
1507 backslash. The current implementation allows whitespace in the form of
1508 spaces, horizontal and vertical tabs, and form feeds between the
1509 backslash and the subsequent newline. The preprocessor issues a
1510 warning, but treats it as a valid escaped newline and combines the two
1511 lines to form a single logical line. This works within comments and
1512 tokens, including multi-line strings, as well as between tokens.
1513 Comments are @emph{not} treated as whitespace for the purposes of this
1514 relaxation, since they have not yet been replaced with spaces.
1516 @node Multi-line Strings
1517 @section String Literals with Embedded Newlines
1518 @cindex multi-line string literals
1520 As an extension, GNU CPP permits string literals to cross multiple lines
1521 without escaping the embedded newlines. Each embedded newline is
1522 replaced with a single @samp{\n} character in the resulting string
1523 literal, regardless of what form the newline took originally.
1525 CPP currently allows such strings in directives as well (other than the
1526 @samp{#include} family). This is deprecated and will eventually be
1530 @section Non-Lvalue Arrays May Have Subscripts
1531 @cindex subscripting
1532 @cindex arrays, non-lvalue
1534 @cindex subscripting and function values
1535 In ISO C99, arrays that are not lvalues still decay to pointers, and
1536 may be subscripted, although they may not be modified or used after
1537 the next sequence point and the unary @samp{&} operator may not be
1538 applied to them. As an extension, GCC allows such arrays to be
1539 subscripted in C89 mode, though otherwise they do not decay to
1540 pointers outside C99 mode. For example,
1541 this is valid in GNU C though not valid in C89:
1545 struct foo @{int a[4];@};
1551 return f().a[index];
1557 @section Arithmetic on @code{void}- and Function-Pointers
1558 @cindex void pointers, arithmetic
1559 @cindex void, size of pointer to
1560 @cindex function pointers, arithmetic
1561 @cindex function, size of pointer to
1563 In GNU C, addition and subtraction operations are supported on pointers to
1564 @code{void} and on pointers to functions. This is done by treating the
1565 size of a @code{void} or of a function as 1.
1567 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1568 and on function types, and returns 1.
1570 @opindex Wpointer-arith
1571 The option @option{-Wpointer-arith} requests a warning if these extensions
1575 @section Non-Constant Initializers
1576 @cindex initializers, non-constant
1577 @cindex non-constant initializers
1579 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1580 automatic variable are not required to be constant expressions in GNU C@.
1581 Here is an example of an initializer with run-time varying elements:
1584 foo (float f, float g)
1586 float beat_freqs[2] = @{ f-g, f+g @};
1591 @node Compound Literals
1592 @section Compound Literals
1593 @cindex constructor expressions
1594 @cindex initializations in expressions
1595 @cindex structures, constructor expression
1596 @cindex expressions, constructor
1597 @cindex compound literals
1598 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1600 ISO C99 supports compound literals. A compound literal looks like
1601 a cast containing an initializer. Its value is an object of the
1602 type specified in the cast, containing the elements specified in
1603 the initializer; it is an lvalue. As an extension, GCC supports
1604 compound literals in C89 mode and in C++.
1606 Usually, the specified type is a structure. Assume that
1607 @code{struct foo} and @code{structure} are declared as shown:
1610 struct foo @{int a; char b[2];@} structure;
1614 Here is an example of constructing a @code{struct foo} with a compound literal:
1617 structure = ((struct foo) @{x + y, 'a', 0@});
1621 This is equivalent to writing the following:
1625 struct foo temp = @{x + y, 'a', 0@};
1630 You can also construct an array. If all the elements of the compound literal
1631 are (made up of) simple constant expressions, suitable for use in
1632 initializers of objects of static storage duration, then the compound
1633 literal can be coerced to a pointer to its first element and used in
1634 such an initializer, as shown here:
1637 char **foo = (char *[]) @{ "x", "y", "z" @};
1640 Compound literals for scalar types and union types are is
1641 also allowed, but then the compound literal is equivalent
1644 As a GNU extension, GCC allows initialization of objects with static storage
1645 duration by compound literals (which is not possible in ISO C99, because
1646 the initializer is not a constant).
1647 It is handled as if the object was initialized only with the bracket
1648 enclosed list if compound literal's and object types match.
1649 The initializer list of the compound literal must be constant.
1650 If the object being initialized has array type of unknown size, the size is
1651 determined by compound literal size.
1654 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1655 static int y[] = (int []) @{1, 2, 3@};
1656 static int z[] = (int [3]) @{1@};
1660 The above lines are equivalent to the following:
1662 static struct foo x = @{1, 'a', 'b'@};
1663 static int y[] = @{1, 2, 3@};
1664 static int z[] = @{1, 0, 0@};
1667 @node Designated Inits
1668 @section Designated Initializers
1669 @cindex initializers with labeled elements
1670 @cindex labeled elements in initializers
1671 @cindex case labels in initializers
1672 @cindex designated initializers
1674 Standard C89 requires the elements of an initializer to appear in a fixed
1675 order, the same as the order of the elements in the array or structure
1678 In ISO C99 you can give the elements in any order, specifying the array
1679 indices or structure field names they apply to, and GNU C allows this as
1680 an extension in C89 mode as well. This extension is not
1681 implemented in GNU C++.
1683 To specify an array index, write
1684 @samp{[@var{index}] =} before the element value. For example,
1687 int a[6] = @{ [4] = 29, [2] = 15 @};
1694 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1698 The index values must be constant expressions, even if the array being
1699 initialized is automatic.
1701 An alternative syntax for this which has been obsolete since GCC 2.5 but
1702 GCC still accepts is to write @samp{[@var{index}]} before the element
1703 value, with no @samp{=}.
1705 To initialize a range of elements to the same value, write
1706 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1707 extension. For example,
1710 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1714 If the value in it has side-effects, the side-effects will happen only once,
1715 not for each initialized field by the range initializer.
1718 Note that the length of the array is the highest value specified
1721 In a structure initializer, specify the name of a field to initialize
1722 with @samp{.@var{fieldname} =} before the element value. For example,
1723 given the following structure,
1726 struct point @{ int x, y; @};
1730 the following initialization
1733 struct point p = @{ .y = yvalue, .x = xvalue @};
1740 struct point p = @{ xvalue, yvalue @};
1743 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1744 @samp{@var{fieldname}:}, as shown here:
1747 struct point p = @{ y: yvalue, x: xvalue @};
1751 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1752 @dfn{designator}. You can also use a designator (or the obsolete colon
1753 syntax) when initializing a union, to specify which element of the union
1754 should be used. For example,
1757 union foo @{ int i; double d; @};
1759 union foo f = @{ .d = 4 @};
1763 will convert 4 to a @code{double} to store it in the union using
1764 the second element. By contrast, casting 4 to type @code{union foo}
1765 would store it into the union as the integer @code{i}, since it is
1766 an integer. (@xref{Cast to Union}.)
1768 You can combine this technique of naming elements with ordinary C
1769 initialization of successive elements. Each initializer element that
1770 does not have a designator applies to the next consecutive element of the
1771 array or structure. For example,
1774 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1781 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1784 Labeling the elements of an array initializer is especially useful
1785 when the indices are characters or belong to an @code{enum} type.
1790 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1791 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1794 @cindex designator lists
1795 You can also write a series of @samp{.@var{fieldname}} and
1796 @samp{[@var{index}]} designators before an @samp{=} to specify a
1797 nested subobject to initialize; the list is taken relative to the
1798 subobject corresponding to the closest surrounding brace pair. For
1799 example, with the @samp{struct point} declaration above:
1802 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1806 If the same field is initialized multiple times, it will have value from
1807 the last initialization. If any such overridden initialization has
1808 side-effect, it is unspecified whether the side-effect happens or not.
1809 Currently, gcc will discard them and issue a warning.
1812 @section Case Ranges
1814 @cindex ranges in case statements
1816 You can specify a range of consecutive values in a single @code{case} label,
1820 case @var{low} ... @var{high}:
1824 This has the same effect as the proper number of individual @code{case}
1825 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1827 This feature is especially useful for ranges of ASCII character codes:
1833 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1834 it may be parsed wrong when you use it with integer values. For example,
1849 @section Cast to a Union Type
1850 @cindex cast to a union
1851 @cindex union, casting to a
1853 A cast to union type is similar to other casts, except that the type
1854 specified is a union type. You can specify the type either with
1855 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1856 a constructor though, not a cast, and hence does not yield an lvalue like
1857 normal casts. (@xref{Compound Literals}.)
1859 The types that may be cast to the union type are those of the members
1860 of the union. Thus, given the following union and variables:
1863 union foo @{ int i; double d; @};
1869 both @code{x} and @code{y} can be cast to type @code{union foo}.
1871 Using the cast as the right-hand side of an assignment to a variable of
1872 union type is equivalent to storing in a member of the union:
1877 u = (union foo) x @equiv{} u.i = x
1878 u = (union foo) y @equiv{} u.d = y
1881 You can also use the union cast as a function argument:
1884 void hack (union foo);
1886 hack ((union foo) x);
1889 @node Mixed Declarations
1890 @section Mixed Declarations and Code
1891 @cindex mixed declarations and code
1892 @cindex declarations, mixed with code
1893 @cindex code, mixed with declarations
1895 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1896 within compound statements. As an extension, GCC also allows this in
1897 C89 mode. For example, you could do:
1906 Each identifier is visible from where it is declared until the end of
1907 the enclosing block.
1909 @node Function Attributes
1910 @section Declaring Attributes of Functions
1911 @cindex function attributes
1912 @cindex declaring attributes of functions
1913 @cindex functions that never return
1914 @cindex functions that have no side effects
1915 @cindex functions in arbitrary sections
1916 @cindex functions that behave like malloc
1917 @cindex @code{volatile} applied to function
1918 @cindex @code{const} applied to function
1919 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1920 @cindex functions with non-null pointer arguments
1921 @cindex functions that are passed arguments in registers on the 386
1922 @cindex functions that pop the argument stack on the 386
1923 @cindex functions that do not pop the argument stack on the 386
1925 In GNU C, you declare certain things about functions called in your program
1926 which help the compiler optimize function calls and check your code more
1929 The keyword @code{__attribute__} allows you to specify special
1930 attributes when making a declaration. This keyword is followed by an
1931 attribute specification inside double parentheses. The following
1932 attributes are currently defined for functions on all targets:
1933 @code{noreturn}, @code{noinline}, @code{always_inline},
1934 @code{pure}, @code{const}, @code{nothrow},
1935 @code{format}, @code{format_arg}, @code{no_instrument_function},
1936 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1937 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1938 @code{alias}, and @code{nonnull}. Several other attributes are defined
1939 for functions on particular target systems. Other attributes, including
1940 @code{section} are supported for variables declarations
1941 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1943 You may also specify attributes with @samp{__} preceding and following
1944 each keyword. This allows you to use them in header files without
1945 being concerned about a possible macro of the same name. For example,
1946 you may use @code{__noreturn__} instead of @code{noreturn}.
1948 @xref{Attribute Syntax}, for details of the exact syntax for using
1952 @cindex @code{noreturn} function attribute
1954 A few standard library functions, such as @code{abort} and @code{exit},
1955 cannot return. GCC knows this automatically. Some programs define
1956 their own functions that never return. You can declare them
1957 @code{noreturn} to tell the compiler this fact. For example,
1961 void fatal () __attribute__ ((noreturn));
1964 fatal (/* @r{@dots{}} */)
1966 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
1972 The @code{noreturn} keyword tells the compiler to assume that
1973 @code{fatal} cannot return. It can then optimize without regard to what
1974 would happen if @code{fatal} ever did return. This makes slightly
1975 better code. More importantly, it helps avoid spurious warnings of
1976 uninitialized variables.
1978 Do not assume that registers saved by the calling function are
1979 restored before calling the @code{noreturn} function.
1981 It does not make sense for a @code{noreturn} function to have a return
1982 type other than @code{void}.
1984 The attribute @code{noreturn} is not implemented in GCC versions
1985 earlier than 2.5. An alternative way to declare that a function does
1986 not return, which works in the current version and in some older
1987 versions, is as follows:
1990 typedef void voidfn ();
1992 volatile voidfn fatal;
1995 @cindex @code{noinline} function attribute
1997 This function attribute prevents a function from being considered for
2000 @cindex @code{always_inline} function attribute
2002 Generally, functions are not inlined unless optimization is specified.
2003 For functions declared inline, this attribute inlines the function even
2004 if no optimization level was specified.
2006 @cindex @code{pure} function attribute
2008 Many functions have no effects except the return value and their
2009 return value depends only on the parameters and/or global variables.
2010 Such a function can be subject
2011 to common subexpression elimination and loop optimization just as an
2012 arithmetic operator would be. These functions should be declared
2013 with the attribute @code{pure}. For example,
2016 int square (int) __attribute__ ((pure));
2020 says that the hypothetical function @code{square} is safe to call
2021 fewer times than the program says.
2023 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2024 Interesting non-pure functions are functions with infinite loops or those
2025 depending on volatile memory or other system resource, that may change between
2026 two consecutive calls (such as @code{feof} in a multithreading environment).
2028 The attribute @code{pure} is not implemented in GCC versions earlier
2030 @cindex @code{const} function attribute
2032 Many functions do not examine any values except their arguments, and
2033 have no effects except the return value. Basically this is just slightly
2034 more strict class than the @code{pure} attribute above, since function is not
2035 allowed to read global memory.
2037 @cindex pointer arguments
2038 Note that a function that has pointer arguments and examines the data
2039 pointed to must @emph{not} be declared @code{const}. Likewise, a
2040 function that calls a non-@code{const} function usually must not be
2041 @code{const}. It does not make sense for a @code{const} function to
2044 The attribute @code{const} is not implemented in GCC versions earlier
2045 than 2.5. An alternative way to declare that a function has no side
2046 effects, which works in the current version and in some older versions,
2050 typedef int intfn ();
2052 extern const intfn square;
2055 This approach does not work in GNU C++ from 2.6.0 on, since the language
2056 specifies that the @samp{const} must be attached to the return value.
2058 @cindex @code{nothrow} function attribute
2060 The @code{nothrow} attribute is used to inform the compiler that a
2061 function cannot throw an exception. For example, most functions in
2062 the standard C library can be guaranteed not to throw an exception
2063 with the notable exceptions of @code{qsort} and @code{bsearch} that
2064 take function pointer arguments. The @code{nothrow} attribute is not
2065 implemented in GCC versions earlier than 3.2.
2067 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2068 @cindex @code{format} function attribute
2070 The @code{format} attribute specifies that a function takes @code{printf},
2071 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2072 should be type-checked against a format string. For example, the
2077 my_printf (void *my_object, const char *my_format, ...)
2078 __attribute__ ((format (printf, 2, 3)));
2082 causes the compiler to check the arguments in calls to @code{my_printf}
2083 for consistency with the @code{printf} style format string argument
2086 The parameter @var{archetype} determines how the format string is
2087 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2088 or @code{strfmon}. (You can also use @code{__printf__},
2089 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2090 parameter @var{string-index} specifies which argument is the format
2091 string argument (starting from 1), while @var{first-to-check} is the
2092 number of the first argument to check against the format string. For
2093 functions where the arguments are not available to be checked (such as
2094 @code{vprintf}), specify the third parameter as zero. In this case the
2095 compiler only checks the format string for consistency. For
2096 @code{strftime} formats, the third parameter is required to be zero.
2098 In the example above, the format string (@code{my_format}) is the second
2099 argument of the function @code{my_print}, and the arguments to check
2100 start with the third argument, so the correct parameters for the format
2101 attribute are 2 and 3.
2103 @opindex ffreestanding
2104 The @code{format} attribute allows you to identify your own functions
2105 which take format strings as arguments, so that GCC can check the
2106 calls to these functions for errors. The compiler always (unless
2107 @option{-ffreestanding} is used) checks formats
2108 for the standard library functions @code{printf}, @code{fprintf},
2109 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2110 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2111 warnings are requested (using @option{-Wformat}), so there is no need to
2112 modify the header file @file{stdio.h}. In C99 mode, the functions
2113 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2114 @code{vsscanf} are also checked. Except in strictly conforming C
2115 standard modes, the X/Open function @code{strfmon} is also checked as
2116 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2117 @xref{C Dialect Options,,Options Controlling C Dialect}.
2119 @item format_arg (@var{string-index})
2120 @cindex @code{format_arg} function attribute
2121 @opindex Wformat-nonliteral
2122 The @code{format_arg} attribute specifies that a function takes a format
2123 string for a @code{printf}, @code{scanf}, @code{strftime} or
2124 @code{strfmon} style function and modifies it (for example, to translate
2125 it into another language), so the result can be passed to a
2126 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2127 function (with the remaining arguments to the format function the same
2128 as they would have been for the unmodified string). For example, the
2133 my_dgettext (char *my_domain, const char *my_format)
2134 __attribute__ ((format_arg (2)));
2138 causes the compiler to check the arguments in calls to a @code{printf},
2139 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2140 format string argument is a call to the @code{my_dgettext} function, for
2141 consistency with the format string argument @code{my_format}. If the
2142 @code{format_arg} attribute had not been specified, all the compiler
2143 could tell in such calls to format functions would be that the format
2144 string argument is not constant; this would generate a warning when
2145 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2146 without the attribute.
2148 The parameter @var{string-index} specifies which argument is the format
2149 string argument (starting from 1).
2151 The @code{format-arg} attribute allows you to identify your own
2152 functions which modify format strings, so that GCC can check the
2153 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2154 type function whose operands are a call to one of your own function.
2155 The compiler always treats @code{gettext}, @code{dgettext}, and
2156 @code{dcgettext} in this manner except when strict ISO C support is
2157 requested by @option{-ansi} or an appropriate @option{-std} option, or
2158 @option{-ffreestanding} is used. @xref{C Dialect Options,,Options
2159 Controlling C Dialect}.
2161 @item nonnull (@var{arg-index}, @dots{})
2162 @cindex @code{nonnull} function attribute
2163 The @code{nonnull} attribute specifies that some function parameters should
2164 be non-null pointers. For instance, the declaration:
2168 my_memcpy (void *dest, const void *src, size_t len)
2169 __attribute__((nonnull (1, 2)));
2173 causes the compiler to check that, in calls to @code{my_memcpy},
2174 arguments @var{dest} and @var{src} are non-null. If the compiler
2175 determines that a null pointer is passed in an argument slot marked
2176 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2177 is issued. The compiler may also choose to make optimizations based
2178 on the knowledge that certain function arguments will not be null.
2180 If no argument index list is given to the @code{nonnull} attribute,
2181 all pointer arguments are marked as non-null. To illustrate, the
2182 following declaration is equivalent to the previous example:
2186 my_memcpy (void *dest, const void *src, size_t len)
2187 __attribute__((nonnull));
2190 @item no_instrument_function
2191 @cindex @code{no_instrument_function} function attribute
2192 @opindex finstrument-functions
2193 If @option{-finstrument-functions} is given, profiling function calls will
2194 be generated at entry and exit of most user-compiled functions.
2195 Functions with this attribute will not be so instrumented.
2197 @item section ("@var{section-name}")
2198 @cindex @code{section} function attribute
2199 Normally, the compiler places the code it generates in the @code{text} section.
2200 Sometimes, however, you need additional sections, or you need certain
2201 particular functions to appear in special sections. The @code{section}
2202 attribute specifies that a function lives in a particular section.
2203 For example, the declaration:
2206 extern void foobar (void) __attribute__ ((section ("bar")));
2210 puts the function @code{foobar} in the @code{bar} section.
2212 Some file formats do not support arbitrary sections so the @code{section}
2213 attribute is not available on all platforms.
2214 If you need to map the entire contents of a module to a particular
2215 section, consider using the facilities of the linker instead.
2219 @cindex @code{constructor} function attribute
2220 @cindex @code{destructor} function attribute
2221 The @code{constructor} attribute causes the function to be called
2222 automatically before execution enters @code{main ()}. Similarly, the
2223 @code{destructor} attribute causes the function to be called
2224 automatically after @code{main ()} has completed or @code{exit ()} has
2225 been called. Functions with these attributes are useful for
2226 initializing data that will be used implicitly during the execution of
2229 These attributes are not currently implemented for Objective-C@.
2231 @cindex @code{unused} attribute.
2233 This attribute, attached to a function, means that the function is meant
2234 to be possibly unused. GCC will not produce a warning for this
2235 function. GNU C++ does not currently support this attribute as
2236 definitions without parameters are valid in C++.
2238 @cindex @code{used} attribute.
2240 This attribute, attached to a function, means that code must be emitted
2241 for the function even if it appears that the function is not referenced.
2242 This is useful, for example, when the function is referenced only in
2245 @cindex @code{deprecated} attribute.
2247 The @code{deprecated} attribute results in a warning if the function
2248 is used anywhere in the source file. This is useful when identifying
2249 functions that are expected to be removed in a future version of a
2250 program. The warning also includes the location of the declaration
2251 of the deprecated function, to enable users to easily find further
2252 information about why the function is deprecated, or what they should
2253 do instead. Note that the warnings only occurs for uses:
2256 int old_fn () __attribute__ ((deprecated));
2258 int (*fn_ptr)() = old_fn;
2261 results in a warning on line 3 but not line 2.
2263 The @code{deprecated} attribute can also be used for variables and
2264 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2267 @cindex @code{weak} attribute
2268 The @code{weak} attribute causes the declaration to be emitted as a weak
2269 symbol rather than a global. This is primarily useful in defining
2270 library functions which can be overridden in user code, though it can
2271 also be used with non-function declarations. Weak symbols are supported
2272 for ELF targets, and also for a.out targets when using the GNU assembler
2276 @cindex @code{malloc} attribute
2277 The @code{malloc} attribute is used to tell the compiler that a function
2278 may be treated as if it were the malloc function. The compiler assumes
2279 that calls to malloc result in a pointers that cannot alias anything.
2280 This will often improve optimization.
2282 @item alias ("@var{target}")
2283 @cindex @code{alias} attribute
2284 The @code{alias} attribute causes the declaration to be emitted as an
2285 alias for another symbol, which must be specified. For instance,
2288 void __f () @{ /* @r{Do something.} */; @}
2289 void f () __attribute__ ((weak, alias ("__f")));
2292 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
2293 mangled name for the target must be used.
2295 Not all target machines support this attribute.
2297 @item visibility ("@var{visibility_type}")
2298 @cindex @code{visibility} attribute
2299 The @code{visibility} attribute on ELF targets causes the declaration
2300 to be emitted with hidden, protected or internal visibility.
2303 void __attribute__ ((visibility ("protected")))
2304 f () @{ /* @r{Do something.} */; @}
2305 int i __attribute__ ((visibility ("hidden")));
2308 See the ELF gABI for complete details, but the short story is
2312 Hidden visibility indicates that the symbol will not be placed into
2313 the dynamic symbol table, so no other @dfn{module} (executable or
2314 shared library) can reference it directly.
2317 Protected visibility indicates that the symbol will be placed in the
2318 dynamic symbol table, but that references within the defining module
2319 will bind to the local symbol. That is, the symbol cannot be overridden
2323 Internal visibility is like hidden visibility, but with additional
2324 processor specific semantics. Unless otherwise specified by the psABI,
2325 gcc defines internal visibility to mean that the function is @emph{never}
2326 called from another module. Note that hidden symbols, while then cannot
2327 be referenced directly by other modules, can be referenced indirectly via
2328 function pointers. By indicating that a symbol cannot be called from
2329 outside the module, gcc may for instance omit the load of a PIC register
2330 since it is known that the calling function loaded the correct value.
2333 Not all ELF targets support this attribute.
2335 @item regparm (@var{number})
2336 @cindex functions that are passed arguments in registers on the 386
2337 On the Intel 386, the @code{regparm} attribute causes the compiler to
2338 pass up to @var{number} integer arguments in registers EAX,
2339 EDX, and ECX instead of on the stack. Functions that take a
2340 variable number of arguments will continue to be passed all of their
2341 arguments on the stack.
2344 @cindex functions that pop the argument stack on the 386
2345 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2346 assume that the called function will pop off the stack space used to
2347 pass arguments, unless it takes a variable number of arguments.
2349 The PowerPC compiler for Windows NT currently ignores the @code{stdcall}
2353 @cindex functions that do pop the argument stack on the 386
2355 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2356 assume that the calling function will pop off the stack space used to
2357 pass arguments. This is
2358 useful to override the effects of the @option{-mrtd} switch.
2360 The PowerPC compiler for Windows NT currently ignores the @code{cdecl}
2363 @item longcall/shortcall
2364 @cindex functions called via pointer on the RS/6000 and PowerPC
2365 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2366 compiler to always call this function via a pointer, just as it would if
2367 the @option{-mlongcall} option had been specified. The @code{shortcall}
2368 attribute causes the compiler not to do this. These attributes override
2369 both the @option{-mlongcall} switch and the @code{#pragma longcall}
2372 @xref{RS/6000 and PowerPC Options}, for more information on when long
2373 calls are and are not necessary.
2375 @item long_call/short_call
2376 @cindex indirect calls on ARM
2377 This attribute allows to specify how to call a particular function on
2378 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2379 command line switch and @code{#pragma long_calls} settings. The
2380 @code{long_call} attribute causes the compiler to always call the
2381 function by first loading its address into a register and then using the
2382 contents of that register. The @code{short_call} attribute always places
2383 the offset to the function from the call site into the @samp{BL}
2384 instruction directly.
2387 @cindex functions which are imported from a dll on PowerPC Windows NT
2388 On the PowerPC running Windows NT, the @code{dllimport} attribute causes
2389 the compiler to call the function via a global pointer to the function
2390 pointer that is set up by the Windows NT dll library. The pointer name
2391 is formed by combining @code{__imp_} and the function name.
2394 @cindex functions which are exported from a dll on PowerPC Windows NT
2395 On the PowerPC running Windows NT, the @code{dllexport} attribute causes
2396 the compiler to provide a global pointer to the function pointer, so
2397 that it can be called with the @code{dllimport} attribute. The pointer
2398 name is formed by combining @code{__imp_} and the function name.
2400 @item exception (@var{except-func} [, @var{except-arg}])
2401 @cindex functions which specify exception handling on PowerPC Windows NT
2402 On the PowerPC running Windows NT, the @code{exception} attribute causes
2403 the compiler to modify the structured exception table entry it emits for
2404 the declared function. The string or identifier @var{except-func} is
2405 placed in the third entry of the structured exception table. It
2406 represents a function, which is called by the exception handling
2407 mechanism if an exception occurs. If it was specified, the string or
2408 identifier @var{except-arg} is placed in the fourth entry of the
2409 structured exception table.
2411 @item function_vector
2412 @cindex calling functions through the function vector on the H8/300 processors
2413 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2414 function should be called through the function vector. Calling a
2415 function through the function vector will reduce code size, however;
2416 the function vector has a limited size (maximum 128 entries on the H8/300
2417 and 64 entries on the H8/300H) and shares space with the interrupt vector.
2419 You must use GAS and GLD from GNU binutils version 2.7 or later for
2420 this attribute to work correctly.
2423 @cindex interrupt handler functions
2424 Use this attribute on the ARM, AVR, M32R/D and Xstormy16 ports to indicate
2425 that the specified function is an interrupt handler. The compiler will
2426 generate function entry and exit sequences suitable for use in an
2427 interrupt handler when this attribute is present.
2429 Note, interrupt handlers for the H8/300, H8/300H and SH processors can
2430 be specified via the @code{interrupt_handler} attribute.
2432 Note, on the AVR interrupts will be enabled inside the function.
2434 Note, for the ARM you can specify the kind of interrupt to be handled by
2435 adding an optional parameter to the interrupt attribute like this:
2438 void f () __attribute__ ((interrupt ("IRQ")));
2441 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2443 @item interrupt_handler
2444 @cindex interrupt handler functions on the H8/300 and SH processors
2445 Use this attribute on the H8/300, H8/300H and SH to indicate that the
2446 specified function is an interrupt handler. The compiler will generate
2447 function entry and exit sequences suitable for use in an interrupt
2448 handler when this attribute is present.
2451 Use this attribute on the SH to indicate an @code{interrupt_handler}
2452 function should switch to an alternate stack. It expects a string
2453 argument that names a global variable holding the address of the
2458 void f () __attribute__ ((interrupt_handler,
2459 sp_switch ("alt_stack")));
2463 Use this attribute on the SH for an @code{interrupt_handle} to return using
2464 @code{trapa} instead of @code{rte}. This attribute expects an integer
2465 argument specifying the trap number to be used.
2468 @cindex eight bit data on the H8/300 and H8/300H
2469 Use this attribute on the H8/300 and H8/300H to indicate that the specified
2470 variable should be placed into the eight bit data section.
2471 The compiler will generate more efficient code for certain operations
2472 on data in the eight bit data area. Note the eight bit data area is limited to
2475 You must use GAS and GLD from GNU binutils version 2.7 or later for
2476 this attribute to work correctly.
2479 @cindex tiny data section on the H8/300H
2480 Use this attribute on the H8/300H to indicate that the specified
2481 variable should be placed into the tiny data section.
2482 The compiler will generate more efficient code for loads and stores
2483 on data in the tiny data section. Note the tiny data area is limited to
2484 slightly under 32kbytes of data.
2487 @cindex signal handler functions on the AVR processors
2488 Use this attribute on the AVR to indicate that the specified
2489 function is an signal handler. The compiler will generate function
2490 entry and exit sequences suitable for use in an signal handler when this
2491 attribute is present. Interrupts will be disabled inside function.
2494 @cindex function without a prologue/epilogue code
2495 Use this attribute on the ARM, AVR and IP2K ports to indicate that the
2496 specified function do not need prologue/epilogue sequences generated by
2497 the compiler. It is up to the programmer to provide these sequences.
2499 @item model (@var{model-name})
2500 @cindex function addressability on the M32R/D
2501 Use this attribute on the M32R/D to set the addressability of an object,
2502 and the code generated for a function.
2503 The identifier @var{model-name} is one of @code{small}, @code{medium},
2504 or @code{large}, representing each of the code models.
2506 Small model objects live in the lower 16MB of memory (so that their
2507 addresses can be loaded with the @code{ld24} instruction), and are
2508 callable with the @code{bl} instruction.
2510 Medium model objects may live anywhere in the 32-bit address space (the
2511 compiler will generate @code{seth/add3} instructions to load their addresses),
2512 and are callable with the @code{bl} instruction.
2514 Large model objects may live anywhere in the 32-bit address space (the
2515 compiler will generate @code{seth/add3} instructions to load their addresses),
2516 and may not be reachable with the @code{bl} instruction (the compiler will
2517 generate the much slower @code{seth/add3/jl} instruction sequence).
2521 You can specify multiple attributes in a declaration by separating them
2522 by commas within the double parentheses or by immediately following an
2523 attribute declaration with another attribute declaration.
2525 @cindex @code{#pragma}, reason for not using
2526 @cindex pragma, reason for not using
2527 Some people object to the @code{__attribute__} feature, suggesting that
2528 ISO C's @code{#pragma} should be used instead. At the time
2529 @code{__attribute__} was designed, there were two reasons for not doing
2534 It is impossible to generate @code{#pragma} commands from a macro.
2537 There is no telling what the same @code{#pragma} might mean in another
2541 These two reasons applied to almost any application that might have been
2542 proposed for @code{#pragma}. It was basically a mistake to use
2543 @code{#pragma} for @emph{anything}.
2545 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2546 to be generated from macros. In addition, a @code{#pragma GCC}
2547 namespace is now in use for GCC-specific pragmas. However, it has been
2548 found convenient to use @code{__attribute__} to achieve a natural
2549 attachment of attributes to their corresponding declarations, whereas
2550 @code{#pragma GCC} is of use for constructs that do not naturally form
2551 part of the grammar. @xref{Other Directives,,Miscellaneous
2552 Preprocessing Directives, cpp, The C Preprocessor}.
2554 @node Attribute Syntax
2555 @section Attribute Syntax
2556 @cindex attribute syntax
2558 This section describes the syntax with which @code{__attribute__} may be
2559 used, and the constructs to which attribute specifiers bind, for the C
2560 language. Some details may vary for C++ and Objective-C@. Because of
2561 infelicities in the grammar for attributes, some forms described here
2562 may not be successfully parsed in all cases.
2564 There are some problems with the semantics of attributes in C++. For
2565 example, there are no manglings for attributes, although they may affect
2566 code generation, so problems may arise when attributed types are used in
2567 conjunction with templates or overloading. Similarly, @code{typeid}
2568 does not distinguish between types with different attributes. Support
2569 for attributes in C++ may be restricted in future to attributes on
2570 declarations only, but not on nested declarators.
2572 @xref{Function Attributes}, for details of the semantics of attributes
2573 applying to functions. @xref{Variable Attributes}, for details of the
2574 semantics of attributes applying to variables. @xref{Type Attributes},
2575 for details of the semantics of attributes applying to structure, union
2576 and enumerated types.
2578 An @dfn{attribute specifier} is of the form
2579 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2580 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2581 each attribute is one of the following:
2585 Empty. Empty attributes are ignored.
2588 A word (which may be an identifier such as @code{unused}, or a reserved
2589 word such as @code{const}).
2592 A word, followed by, in parentheses, parameters for the attribute.
2593 These parameters take one of the following forms:
2597 An identifier. For example, @code{mode} attributes use this form.
2600 An identifier followed by a comma and a non-empty comma-separated list
2601 of expressions. For example, @code{format} attributes use this form.
2604 A possibly empty comma-separated list of expressions. For example,
2605 @code{format_arg} attributes use this form with the list being a single
2606 integer constant expression, and @code{alias} attributes use this form
2607 with the list being a single string constant.
2611 An @dfn{attribute specifier list} is a sequence of one or more attribute
2612 specifiers, not separated by any other tokens.
2614 An attribute specifier list may appear after the colon following a
2615 label, other than a @code{case} or @code{default} label. The only
2616 attribute it makes sense to use after a label is @code{unused}. This
2617 feature is intended for code generated by programs which contains labels
2618 that may be unused but which is compiled with @option{-Wall}. It would
2619 not normally be appropriate to use in it human-written code, though it
2620 could be useful in cases where the code that jumps to the label is
2621 contained within an @code{#ifdef} conditional.
2623 An attribute specifier list may appear as part of a @code{struct},
2624 @code{union} or @code{enum} specifier. It may go either immediately
2625 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2626 the closing brace. It is ignored if the content of the structure, union
2627 or enumerated type is not defined in the specifier in which the
2628 attribute specifier list is used---that is, in usages such as
2629 @code{struct __attribute__((foo)) bar} with no following opening brace.
2630 Where attribute specifiers follow the closing brace, they are considered
2631 to relate to the structure, union or enumerated type defined, not to any
2632 enclosing declaration the type specifier appears in, and the type
2633 defined is not complete until after the attribute specifiers.
2634 @c Otherwise, there would be the following problems: a shift/reduce
2635 @c conflict between attributes binding the struct/union/enum and
2636 @c binding to the list of specifiers/qualifiers; and "aligned"
2637 @c attributes could use sizeof for the structure, but the size could be
2638 @c changed later by "packed" attributes.
2640 Otherwise, an attribute specifier appears as part of a declaration,
2641 counting declarations of unnamed parameters and type names, and relates
2642 to that declaration (which may be nested in another declaration, for
2643 example in the case of a parameter declaration), or to a particular declarator
2644 within a declaration. Where an
2645 attribute specifier is applied to a parameter declared as a function or
2646 an array, it should apply to the function or array rather than the
2647 pointer to which the parameter is implicitly converted, but this is not
2648 yet correctly implemented.
2650 Any list of specifiers and qualifiers at the start of a declaration may
2651 contain attribute specifiers, whether or not such a list may in that
2652 context contain storage class specifiers. (Some attributes, however,
2653 are essentially in the nature of storage class specifiers, and only make
2654 sense where storage class specifiers may be used; for example,
2655 @code{section}.) There is one necessary limitation to this syntax: the
2656 first old-style parameter declaration in a function definition cannot
2657 begin with an attribute specifier, because such an attribute applies to
2658 the function instead by syntax described below (which, however, is not
2659 yet implemented in this case). In some other cases, attribute
2660 specifiers are permitted by this grammar but not yet supported by the
2661 compiler. All attribute specifiers in this place relate to the
2662 declaration as a whole. In the obsolescent usage where a type of
2663 @code{int} is implied by the absence of type specifiers, such a list of
2664 specifiers and qualifiers may be an attribute specifier list with no
2665 other specifiers or qualifiers.
2667 An attribute specifier list may appear immediately before a declarator
2668 (other than the first) in a comma-separated list of declarators in a
2669 declaration of more than one identifier using a single list of
2670 specifiers and qualifiers. Such attribute specifiers apply
2671 only to the identifier before whose declarator they appear. For
2675 __attribute__((noreturn)) void d0 (void),
2676 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2681 the @code{noreturn} attribute applies to all the functions
2682 declared; the @code{format} attribute only applies to @code{d1}.
2684 An attribute specifier list may appear immediately before the comma,
2685 @code{=} or semicolon terminating the declaration of an identifier other
2686 than a function definition. At present, such attribute specifiers apply
2687 to the declared object or function, but in future they may attach to the
2688 outermost adjacent declarator. In simple cases there is no difference,
2689 but, for example, in
2692 void (****f)(void) __attribute__((noreturn));
2696 at present the @code{noreturn} attribute applies to @code{f}, which
2697 causes a warning since @code{f} is not a function, but in future it may
2698 apply to the function @code{****f}. The precise semantics of what
2699 attributes in such cases will apply to are not yet specified. Where an
2700 assembler name for an object or function is specified (@pxref{Asm
2701 Labels}), at present the attribute must follow the @code{asm}
2702 specification; in future, attributes before the @code{asm} specification
2703 may apply to the adjacent declarator, and those after it to the declared
2706 An attribute specifier list may, in future, be permitted to appear after
2707 the declarator in a function definition (before any old-style parameter
2708 declarations or the function body).
2710 Attribute specifiers may be mixed with type qualifiers appearing inside
2711 the @code{[]} of a parameter array declarator, in the C99 construct by
2712 which such qualifiers are applied to the pointer to which the array is
2713 implicitly converted. Such attribute specifiers apply to the pointer,
2714 not to the array, but at present this is not implemented and they are
2717 An attribute specifier list may appear at the start of a nested
2718 declarator. At present, there are some limitations in this usage: the
2719 attributes correctly apply to the declarator, but for most individual
2720 attributes the semantics this implies are not implemented.
2721 When attribute specifiers follow the @code{*} of a pointer
2722 declarator, they may be mixed with any type qualifiers present.
2723 The following describes the formal semantics of this syntax. It will make the
2724 most sense if you are familiar with the formal specification of
2725 declarators in the ISO C standard.
2727 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2728 D1}, where @code{T} contains declaration specifiers that specify a type
2729 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2730 contains an identifier @var{ident}. The type specified for @var{ident}
2731 for derived declarators whose type does not include an attribute
2732 specifier is as in the ISO C standard.
2734 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2735 and the declaration @code{T D} specifies the type
2736 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2737 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2738 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2740 If @code{D1} has the form @code{*
2741 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2742 declaration @code{T D} specifies the type
2743 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2744 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2745 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2751 void (__attribute__((noreturn)) ****f) (void);
2755 specifies the type ``pointer to pointer to pointer to pointer to
2756 non-returning function returning @code{void}''. As another example,
2759 char *__attribute__((aligned(8))) *f;
2763 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2764 Note again that this does not work with most attributes; for example,
2765 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2766 is not yet supported.
2768 For compatibility with existing code written for compiler versions that
2769 did not implement attributes on nested declarators, some laxity is
2770 allowed in the placing of attributes. If an attribute that only applies
2771 to types is applied to a declaration, it will be treated as applying to
2772 the type of that declaration. If an attribute that only applies to
2773 declarations is applied to the type of a declaration, it will be treated
2774 as applying to that declaration; and, for compatibility with code
2775 placing the attributes immediately before the identifier declared, such
2776 an attribute applied to a function return type will be treated as
2777 applying to the function type, and such an attribute applied to an array
2778 element type will be treated as applying to the array type. If an
2779 attribute that only applies to function types is applied to a
2780 pointer-to-function type, it will be treated as applying to the pointer
2781 target type; if such an attribute is applied to a function return type
2782 that is not a pointer-to-function type, it will be treated as applying
2783 to the function type.
2785 @node Function Prototypes
2786 @section Prototypes and Old-Style Function Definitions
2787 @cindex function prototype declarations
2788 @cindex old-style function definitions
2789 @cindex promotion of formal parameters
2791 GNU C extends ISO C to allow a function prototype to override a later
2792 old-style non-prototype definition. Consider the following example:
2795 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2802 /* @r{Prototype function declaration.} */
2803 int isroot P((uid_t));
2805 /* @r{Old-style function definition.} */
2807 isroot (x) /* ??? lossage here ??? */
2814 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2815 not allow this example, because subword arguments in old-style
2816 non-prototype definitions are promoted. Therefore in this example the
2817 function definition's argument is really an @code{int}, which does not
2818 match the prototype argument type of @code{short}.
2820 This restriction of ISO C makes it hard to write code that is portable
2821 to traditional C compilers, because the programmer does not know
2822 whether the @code{uid_t} type is @code{short}, @code{int}, or
2823 @code{long}. Therefore, in cases like these GNU C allows a prototype
2824 to override a later old-style definition. More precisely, in GNU C, a
2825 function prototype argument type overrides the argument type specified
2826 by a later old-style definition if the former type is the same as the
2827 latter type before promotion. Thus in GNU C the above example is
2828 equivalent to the following:
2841 GNU C++ does not support old-style function definitions, so this
2842 extension is irrelevant.
2845 @section C++ Style Comments
2847 @cindex C++ comments
2848 @cindex comments, C++ style
2850 In GNU C, you may use C++ style comments, which start with @samp{//} and
2851 continue until the end of the line. Many other C implementations allow
2852 such comments, and they are included in the 1999 C standard. However,
2853 C++ style comments are not recognized if you specify an @option{-std}
2854 option specifying a version of ISO C before C99, or @option{-ansi}
2855 (equivalent to @option{-std=c89}).
2858 @section Dollar Signs in Identifier Names
2860 @cindex dollar signs in identifier names
2861 @cindex identifier names, dollar signs in
2863 In GNU C, you may normally use dollar signs in identifier names.
2864 This is because many traditional C implementations allow such identifiers.
2865 However, dollar signs in identifiers are not supported on a few target
2866 machines, typically because the target assembler does not allow them.
2868 @node Character Escapes
2869 @section The Character @key{ESC} in Constants
2871 You can use the sequence @samp{\e} in a string or character constant to
2872 stand for the ASCII character @key{ESC}.
2875 @section Inquiring on Alignment of Types or Variables
2877 @cindex type alignment
2878 @cindex variable alignment
2880 The keyword @code{__alignof__} allows you to inquire about how an object
2881 is aligned, or the minimum alignment usually required by a type. Its
2882 syntax is just like @code{sizeof}.
2884 For example, if the target machine requires a @code{double} value to be
2885 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2886 This is true on many RISC machines. On more traditional machine
2887 designs, @code{__alignof__ (double)} is 4 or even 2.
2889 Some machines never actually require alignment; they allow reference to any
2890 data type even at an odd addresses. For these machines, @code{__alignof__}
2891 reports the @emph{recommended} alignment of a type.
2893 If the operand of @code{__alignof__} is an lvalue rather than a type,
2894 its value is the required alignment for its type, taking into account
2895 any minimum alignment specified with GCC's @code{__attribute__}
2896 extension (@pxref{Variable Attributes}). For example, after this
2900 struct foo @{ int x; char y; @} foo1;
2904 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2905 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2907 It is an error to ask for the alignment of an incomplete type.
2909 @node Variable Attributes
2910 @section Specifying Attributes of Variables
2911 @cindex attribute of variables
2912 @cindex variable attributes
2914 The keyword @code{__attribute__} allows you to specify special
2915 attributes of variables or structure fields. This keyword is followed
2916 by an attribute specification inside double parentheses. Ten
2917 attributes are currently defined for variables: @code{aligned},
2918 @code{mode}, @code{nocommon}, @code{packed}, @code{section},
2919 @code{transparent_union}, @code{unused}, @code{deprecated},
2920 @code{vector_size}, and @code{weak}. Some other attributes are defined
2921 for variables on particular target systems. Other attributes are
2922 available for functions (@pxref{Function Attributes}) and for types
2923 (@pxref{Type Attributes}). Other front ends might define more
2924 attributes (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2926 You may also specify attributes with @samp{__} preceding and following
2927 each keyword. This allows you to use them in header files without
2928 being concerned about a possible macro of the same name. For example,
2929 you may use @code{__aligned__} instead of @code{aligned}.
2931 @xref{Attribute Syntax}, for details of the exact syntax for using
2935 @cindex @code{aligned} attribute
2936 @item aligned (@var{alignment})
2937 This attribute specifies a minimum alignment for the variable or
2938 structure field, measured in bytes. For example, the declaration:
2941 int x __attribute__ ((aligned (16))) = 0;
2945 causes the compiler to allocate the global variable @code{x} on a
2946 16-byte boundary. On a 68040, this could be used in conjunction with
2947 an @code{asm} expression to access the @code{move16} instruction which
2948 requires 16-byte aligned operands.
2950 You can also specify the alignment of structure fields. For example, to
2951 create a double-word aligned @code{int} pair, you could write:
2954 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2958 This is an alternative to creating a union with a @code{double} member
2959 that forces the union to be double-word aligned.
2961 As in the preceding examples, you can explicitly specify the alignment
2962 (in bytes) that you wish the compiler to use for a given variable or
2963 structure field. Alternatively, you can leave out the alignment factor
2964 and just ask the compiler to align a variable or field to the maximum
2965 useful alignment for the target machine you are compiling for. For
2966 example, you could write:
2969 short array[3] __attribute__ ((aligned));
2972 Whenever you leave out the alignment factor in an @code{aligned} attribute
2973 specification, the compiler automatically sets the alignment for the declared
2974 variable or field to the largest alignment which is ever used for any data
2975 type on the target machine you are compiling for. Doing this can often make
2976 copy operations more efficient, because the compiler can use whatever
2977 instructions copy the biggest chunks of memory when performing copies to
2978 or from the variables or fields that you have aligned this way.
2980 The @code{aligned} attribute can only increase the alignment; but you
2981 can decrease it by specifying @code{packed} as well. See below.
2983 Note that the effectiveness of @code{aligned} attributes may be limited
2984 by inherent limitations in your linker. On many systems, the linker is
2985 only able to arrange for variables to be aligned up to a certain maximum
2986 alignment. (For some linkers, the maximum supported alignment may
2987 be very very small.) If your linker is only able to align variables
2988 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
2989 in an @code{__attribute__} will still only provide you with 8 byte
2990 alignment. See your linker documentation for further information.
2992 @item mode (@var{mode})
2993 @cindex @code{mode} attribute
2994 This attribute specifies the data type for the declaration---whichever
2995 type corresponds to the mode @var{mode}. This in effect lets you
2996 request an integer or floating point type according to its width.
2998 You may also specify a mode of @samp{byte} or @samp{__byte__} to
2999 indicate the mode corresponding to a one-byte integer, @samp{word} or
3000 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3001 or @samp{__pointer__} for the mode used to represent pointers.
3004 @cindex @code{nocommon} attribute
3006 This attribute specifies requests GCC not to place a variable
3007 ``common'' but instead to allocate space for it directly. If you
3008 specify the @option{-fno-common} flag, GCC will do this for all
3011 Specifying the @code{nocommon} attribute for a variable provides an
3012 initialization of zeros. A variable may only be initialized in one
3016 @cindex @code{packed} attribute
3017 The @code{packed} attribute specifies that a variable or structure field
3018 should have the smallest possible alignment---one byte for a variable,
3019 and one bit for a field, unless you specify a larger value with the
3020 @code{aligned} attribute.
3022 Here is a structure in which the field @code{x} is packed, so that it
3023 immediately follows @code{a}:
3029 int x[2] __attribute__ ((packed));
3033 @item section ("@var{section-name}")
3034 @cindex @code{section} variable attribute
3035 Normally, the compiler places the objects it generates in sections like
3036 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3037 or you need certain particular variables to appear in special sections,
3038 for example to map to special hardware. The @code{section}
3039 attribute specifies that a variable (or function) lives in a particular
3040 section. For example, this small program uses several specific section names:
3043 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3044 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3045 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3046 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3050 /* Initialize stack pointer */
3051 init_sp (stack + sizeof (stack));
3053 /* Initialize initialized data */
3054 memcpy (&init_data, &data, &edata - &data);
3056 /* Turn on the serial ports */
3063 Use the @code{section} attribute with an @emph{initialized} definition
3064 of a @emph{global} variable, as shown in the example. GCC issues
3065 a warning and otherwise ignores the @code{section} attribute in
3066 uninitialized variable declarations.
3068 You may only use the @code{section} attribute with a fully initialized
3069 global definition because of the way linkers work. The linker requires
3070 each object be defined once, with the exception that uninitialized
3071 variables tentatively go in the @code{common} (or @code{bss}) section
3072 and can be multiply ``defined''. You can force a variable to be
3073 initialized with the @option{-fno-common} flag or the @code{nocommon}
3076 Some file formats do not support arbitrary sections so the @code{section}
3077 attribute is not available on all platforms.
3078 If you need to map the entire contents of a module to a particular
3079 section, consider using the facilities of the linker instead.
3082 @cindex @code{shared} variable attribute
3083 On Windows NT, in addition to putting variable definitions in a named
3084 section, the section can also be shared among all running copies of an
3085 executable or DLL@. For example, this small program defines shared data
3086 by putting it in a named section @code{shared} and marking the section
3090 int foo __attribute__((section ("shared"), shared)) = 0;
3095 /* Read and write foo. All running
3096 copies see the same value. */
3102 You may only use the @code{shared} attribute along with @code{section}
3103 attribute with a fully initialized global definition because of the way
3104 linkers work. See @code{section} attribute for more information.
3106 The @code{shared} attribute is only available on Windows NT@.
3108 @item transparent_union
3109 This attribute, attached to a function parameter which is a union, means
3110 that the corresponding argument may have the type of any union member,
3111 but the argument is passed as if its type were that of the first union
3112 member. For more details see @xref{Type Attributes}. You can also use
3113 this attribute on a @code{typedef} for a union data type; then it
3114 applies to all function parameters with that type.
3117 This attribute, attached to a variable, means that the variable is meant
3118 to be possibly unused. GCC will not produce a warning for this
3122 The @code{deprecated} attribute results in a warning if the variable
3123 is used anywhere in the source file. This is useful when identifying
3124 variables that are expected to be removed in a future version of a
3125 program. The warning also includes the location of the declaration
3126 of the deprecated variable, to enable users to easily find further
3127 information about why the variable is deprecated, or what they should
3128 do instead. Note that the warnings only occurs for uses:
3131 extern int old_var __attribute__ ((deprecated));
3133 int new_fn () @{ return old_var; @}
3136 results in a warning on line 3 but not line 2.
3138 The @code{deprecated} attribute can also be used for functions and
3139 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3141 @item vector_size (@var{bytes})
3142 This attribute specifies the vector size for the variable, measured in
3143 bytes. For example, the declaration:
3146 int foo __attribute__ ((vector_size (16)));
3150 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3151 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3152 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3154 This attribute is only applicable to integral and float scalars,
3155 although arrays, pointers, and function return values are allowed in
3156 conjunction with this construct.
3158 Aggregates with this attribute are invalid, even if they are of the same
3159 size as a corresponding scalar. For example, the declaration:
3162 struct S @{ int a; @};
3163 struct S __attribute__ ((vector_size (16))) foo;
3167 is invalid even if the size of the structure is the same as the size of
3171 The @code{weak} attribute is described in @xref{Function Attributes}.
3173 @item model (@var{model-name})
3174 @cindex variable addressability on the M32R/D
3175 Use this attribute on the M32R/D to set the addressability of an object.
3176 The identifier @var{model-name} is one of @code{small}, @code{medium},
3177 or @code{large}, representing each of the code models.
3179 Small model objects live in the lower 16MB of memory (so that their
3180 addresses can be loaded with the @code{ld24} instruction).
3182 Medium and large model objects may live anywhere in the 32-bit address space
3183 (the compiler will generate @code{seth/add3} instructions to load their
3188 To specify multiple attributes, separate them by commas within the
3189 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3192 @node Type Attributes
3193 @section Specifying Attributes of Types
3194 @cindex attribute of types
3195 @cindex type attributes
3197 The keyword @code{__attribute__} allows you to specify special
3198 attributes of @code{struct} and @code{union} types when you define such
3199 types. This keyword is followed by an attribute specification inside
3200 double parentheses. Six attributes are currently defined for types:
3201 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3202 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3203 functions (@pxref{Function Attributes}) and for variables
3204 (@pxref{Variable Attributes}).
3206 You may also specify any one of these attributes with @samp{__}
3207 preceding and following its keyword. This allows you to use these
3208 attributes in header files without being concerned about a possible
3209 macro of the same name. For example, you may use @code{__aligned__}
3210 instead of @code{aligned}.
3212 You may specify the @code{aligned} and @code{transparent_union}
3213 attributes either in a @code{typedef} declaration or just past the
3214 closing curly brace of a complete enum, struct or union type
3215 @emph{definition} and the @code{packed} attribute only past the closing
3216 brace of a definition.
3218 You may also specify attributes between the enum, struct or union
3219 tag and the name of the type rather than after the closing brace.
3221 @xref{Attribute Syntax}, for details of the exact syntax for using
3225 @cindex @code{aligned} attribute
3226 @item aligned (@var{alignment})
3227 This attribute specifies a minimum alignment (in bytes) for variables
3228 of the specified type. For example, the declarations:
3231 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3232 typedef int more_aligned_int __attribute__ ((aligned (8)));
3236 force the compiler to insure (as far as it can) that each variable whose
3237 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3238 aligned @emph{at least} on a 8-byte boundary. On a Sparc, having all
3239 variables of type @code{struct S} aligned to 8-byte boundaries allows
3240 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3241 store) instructions when copying one variable of type @code{struct S} to
3242 another, thus improving run-time efficiency.
3244 Note that the alignment of any given @code{struct} or @code{union} type
3245 is required by the ISO C standard to be at least a perfect multiple of
3246 the lowest common multiple of the alignments of all of the members of
3247 the @code{struct} or @code{union} in question. This means that you @emph{can}
3248 effectively adjust the alignment of a @code{struct} or @code{union}
3249 type by attaching an @code{aligned} attribute to any one of the members
3250 of such a type, but the notation illustrated in the example above is a
3251 more obvious, intuitive, and readable way to request the compiler to
3252 adjust the alignment of an entire @code{struct} or @code{union} type.
3254 As in the preceding example, you can explicitly specify the alignment
3255 (in bytes) that you wish the compiler to use for a given @code{struct}
3256 or @code{union} type. Alternatively, you can leave out the alignment factor
3257 and just ask the compiler to align a type to the maximum
3258 useful alignment for the target machine you are compiling for. For
3259 example, you could write:
3262 struct S @{ short f[3]; @} __attribute__ ((aligned));
3265 Whenever you leave out the alignment factor in an @code{aligned}
3266 attribute specification, the compiler automatically sets the alignment
3267 for the type to the largest alignment which is ever used for any data
3268 type on the target machine you are compiling for. Doing this can often
3269 make copy operations more efficient, because the compiler can use
3270 whatever instructions copy the biggest chunks of memory when performing
3271 copies to or from the variables which have types that you have aligned
3274 In the example above, if the size of each @code{short} is 2 bytes, then
3275 the size of the entire @code{struct S} type is 6 bytes. The smallest
3276 power of two which is greater than or equal to that is 8, so the
3277 compiler sets the alignment for the entire @code{struct S} type to 8
3280 Note that although you can ask the compiler to select a time-efficient
3281 alignment for a given type and then declare only individual stand-alone
3282 objects of that type, the compiler's ability to select a time-efficient
3283 alignment is primarily useful only when you plan to create arrays of
3284 variables having the relevant (efficiently aligned) type. If you
3285 declare or use arrays of variables of an efficiently-aligned type, then
3286 it is likely that your program will also be doing pointer arithmetic (or
3287 subscripting, which amounts to the same thing) on pointers to the
3288 relevant type, and the code that the compiler generates for these
3289 pointer arithmetic operations will often be more efficient for
3290 efficiently-aligned types than for other types.
3292 The @code{aligned} attribute can only increase the alignment; but you
3293 can decrease it by specifying @code{packed} as well. See below.
3295 Note that the effectiveness of @code{aligned} attributes may be limited
3296 by inherent limitations in your linker. On many systems, the linker is
3297 only able to arrange for variables to be aligned up to a certain maximum
3298 alignment. (For some linkers, the maximum supported alignment may
3299 be very very small.) If your linker is only able to align variables
3300 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3301 in an @code{__attribute__} will still only provide you with 8 byte
3302 alignment. See your linker documentation for further information.
3305 This attribute, attached to an @code{enum}, @code{struct}, or
3306 @code{union} type definition, specified that the minimum required memory
3307 be used to represent the type.
3309 @opindex fshort-enums
3310 Specifying this attribute for @code{struct} and @code{union} types is
3311 equivalent to specifying the @code{packed} attribute on each of the
3312 structure or union members. Specifying the @option{-fshort-enums}
3313 flag on the line is equivalent to specifying the @code{packed}
3314 attribute on all @code{enum} definitions.
3316 You may only specify this attribute after a closing curly brace on an
3317 @code{enum} definition, not in a @code{typedef} declaration, unless that
3318 declaration also contains the definition of the @code{enum}.
3320 @item transparent_union
3321 This attribute, attached to a @code{union} type definition, indicates
3322 that any function parameter having that union type causes calls to that
3323 function to be treated in a special way.
3325 First, the argument corresponding to a transparent union type can be of
3326 any type in the union; no cast is required. Also, if the union contains
3327 a pointer type, the corresponding argument can be a null pointer
3328 constant or a void pointer expression; and if the union contains a void
3329 pointer type, the corresponding argument can be any pointer expression.
3330 If the union member type is a pointer, qualifiers like @code{const} on
3331 the referenced type must be respected, just as with normal pointer
3334 Second, the argument is passed to the function using the calling
3335 conventions of first member of the transparent union, not the calling
3336 conventions of the union itself. All members of the union must have the
3337 same machine representation; this is necessary for this argument passing
3340 Transparent unions are designed for library functions that have multiple
3341 interfaces for compatibility reasons. For example, suppose the
3342 @code{wait} function must accept either a value of type @code{int *} to
3343 comply with Posix, or a value of type @code{union wait *} to comply with
3344 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3345 @code{wait} would accept both kinds of arguments, but it would also
3346 accept any other pointer type and this would make argument type checking
3347 less useful. Instead, @code{<sys/wait.h>} might define the interface
3355 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3357 pid_t wait (wait_status_ptr_t);
3360 This interface allows either @code{int *} or @code{union wait *}
3361 arguments to be passed, using the @code{int *} calling convention.
3362 The program can call @code{wait} with arguments of either type:
3365 int w1 () @{ int w; return wait (&w); @}
3366 int w2 () @{ union wait w; return wait (&w); @}
3369 With this interface, @code{wait}'s implementation might look like this:
3372 pid_t wait (wait_status_ptr_t p)
3374 return waitpid (-1, p.__ip, 0);
3379 When attached to a type (including a @code{union} or a @code{struct}),
3380 this attribute means that variables of that type are meant to appear
3381 possibly unused. GCC will not produce a warning for any variables of
3382 that type, even if the variable appears to do nothing. This is often
3383 the case with lock or thread classes, which are usually defined and then
3384 not referenced, but contain constructors and destructors that have
3385 nontrivial bookkeeping functions.
3388 The @code{deprecated} attribute results in a warning if the type
3389 is used anywhere in the source file. This is useful when identifying
3390 types that are expected to be removed in a future version of a program.
3391 If possible, the warning also includes the location of the declaration
3392 of the deprecated type, to enable users to easily find further
3393 information about why the type is deprecated, or what they should do
3394 instead. Note that the warnings only occur for uses and then only
3395 if the type is being applied to an identifier that itself is not being
3396 declared as deprecated.
3399 typedef int T1 __attribute__ ((deprecated));
3403 typedef T1 T3 __attribute__ ((deprecated));
3404 T3 z __attribute__ ((deprecated));
3407 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3408 warning is issued for line 4 because T2 is not explicitly
3409 deprecated. Line 5 has no warning because T3 is explicitly
3410 deprecated. Similarly for line 6.
3412 The @code{deprecated} attribute can also be used for functions and
3413 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3416 Accesses to objects with types with this attribute are not subjected to
3417 type-based alias analysis, but are instead assumed to be able to alias
3418 any other type of objects, just like the @code{char} type. See
3419 @option{-fstrict-aliasing} for more information on aliasing issues.
3424 typedef short __attribute__((__may_alias__)) short_a;
3430 short_a *b = (short_a *) &a;
3434 if (a == 0x12345678)
3441 If you replaced @code{short_a} with @code{short} in the variable
3442 declaration, the above program would abort when compiled with
3443 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3444 above in recent GCC versions.
3447 To specify multiple attributes, separate them by commas within the
3448 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3452 @section An Inline Function is As Fast As a Macro
3453 @cindex inline functions
3454 @cindex integrating function code
3456 @cindex macros, inline alternative
3458 By declaring a function @code{inline}, you can direct GCC to
3459 integrate that function's code into the code for its callers. This
3460 makes execution faster by eliminating the function-call overhead; in
3461 addition, if any of the actual argument values are constant, their known
3462 values may permit simplifications at compile time so that not all of the
3463 inline function's code needs to be included. The effect on code size is
3464 less predictable; object code may be larger or smaller with function
3465 inlining, depending on the particular case. Inlining of functions is an
3466 optimization and it really ``works'' only in optimizing compilation. If
3467 you don't use @option{-O}, no function is really inline.
3469 Inline functions are included in the ISO C99 standard, but there are
3470 currently substantial differences between what GCC implements and what
3471 the ISO C99 standard requires.
3473 To declare a function inline, use the @code{inline} keyword in its
3474 declaration, like this:
3484 (If you are writing a header file to be included in ISO C programs, write
3485 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3486 You can also make all ``simple enough'' functions inline with the option
3487 @option{-finline-functions}.
3490 Note that certain usages in a function definition can make it unsuitable
3491 for inline substitution. Among these usages are: use of varargs, use of
3492 alloca, use of variable sized data types (@pxref{Variable Length}),
3493 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3494 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3495 will warn when a function marked @code{inline} could not be substituted,
3496 and will give the reason for the failure.
3498 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3499 does not affect the linkage of the function.
3501 @cindex automatic @code{inline} for C++ member fns
3502 @cindex @code{inline} automatic for C++ member fns
3503 @cindex member fns, automatically @code{inline}
3504 @cindex C++ member fns, automatically @code{inline}
3505 @opindex fno-default-inline
3506 GCC automatically inlines member functions defined within the class
3507 body of C++ programs even if they are not explicitly declared
3508 @code{inline}. (You can override this with @option{-fno-default-inline};
3509 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3511 @cindex inline functions, omission of
3512 @opindex fkeep-inline-functions
3513 When a function is both inline and @code{static}, if all calls to the
3514 function are integrated into the caller, and the function's address is
3515 never used, then the function's own assembler code is never referenced.
3516 In this case, GCC does not actually output assembler code for the
3517 function, unless you specify the option @option{-fkeep-inline-functions}.
3518 Some calls cannot be integrated for various reasons (in particular,
3519 calls that precede the function's definition cannot be integrated, and
3520 neither can recursive calls within the definition). If there is a
3521 nonintegrated call, then the function is compiled to assembler code as
3522 usual. The function must also be compiled as usual if the program
3523 refers to its address, because that can't be inlined.
3525 @cindex non-static inline function
3526 When an inline function is not @code{static}, then the compiler must assume
3527 that there may be calls from other source files; since a global symbol can
3528 be defined only once in any program, the function must not be defined in
3529 the other source files, so the calls therein cannot be integrated.
3530 Therefore, a non-@code{static} inline function is always compiled on its
3531 own in the usual fashion.
3533 If you specify both @code{inline} and @code{extern} in the function
3534 definition, then the definition is used only for inlining. In no case
3535 is the function compiled on its own, not even if you refer to its
3536 address explicitly. Such an address becomes an external reference, as
3537 if you had only declared the function, and had not defined it.
3539 This combination of @code{inline} and @code{extern} has almost the
3540 effect of a macro. The way to use it is to put a function definition in
3541 a header file with these keywords, and put another copy of the
3542 definition (lacking @code{inline} and @code{extern}) in a library file.
3543 The definition in the header file will cause most calls to the function
3544 to be inlined. If any uses of the function remain, they will refer to
3545 the single copy in the library.
3547 For future compatibility with when GCC implements ISO C99 semantics for
3548 inline functions, it is best to use @code{static inline} only. (The
3549 existing semantics will remain available when @option{-std=gnu89} is
3550 specified, but eventually the default will be @option{-std=gnu99} and
3551 that will implement the C99 semantics, though it does not do so yet.)
3553 GCC does not inline any functions when not optimizing unless you specify
3554 the @samp{always_inline} attribute for the function, like this:
3558 inline void foo (const char) __attribute__((always_inline));
3562 @section Assembler Instructions with C Expression Operands
3563 @cindex extended @code{asm}
3564 @cindex @code{asm} expressions
3565 @cindex assembler instructions
3568 In an assembler instruction using @code{asm}, you can specify the
3569 operands of the instruction using C expressions. This means you need not
3570 guess which registers or memory locations will contain the data you want
3573 You must specify an assembler instruction template much like what
3574 appears in a machine description, plus an operand constraint string for
3577 For example, here is how to use the 68881's @code{fsinx} instruction:
3580 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3584 Here @code{angle} is the C expression for the input operand while
3585 @code{result} is that of the output operand. Each has @samp{"f"} as its
3586 operand constraint, saying that a floating point register is required.
3587 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3588 output operands' constraints must use @samp{=}. The constraints use the
3589 same language used in the machine description (@pxref{Constraints}).
3591 Each operand is described by an operand-constraint string followed by
3592 the C expression in parentheses. A colon separates the assembler
3593 template from the first output operand and another separates the last
3594 output operand from the first input, if any. Commas separate the
3595 operands within each group. The total number of operands is currently
3596 limited to 30; this limitation may be lifted in some future version of
3599 If there are no output operands but there are input operands, you must
3600 place two consecutive colons surrounding the place where the output
3603 As of GCC version 3.1, it is also possible to specify input and output
3604 operands using symbolic names which can be referenced within the
3605 assembler code. These names are specified inside square brackets
3606 preceding the constraint string, and can be referenced inside the
3607 assembler code using @code{%[@var{name}]} instead of a percentage sign
3608 followed by the operand number. Using named operands the above example
3612 asm ("fsinx %[angle],%[output]"
3613 : [output] "=f" (result)
3614 : [angle] "f" (angle));
3618 Note that the symbolic operand names have no relation whatsoever to
3619 other C identifiers. You may use any name you like, even those of
3620 existing C symbols, but must ensure that no two operands within the same
3621 assembler construct use the same symbolic name.
3623 Output operand expressions must be lvalues; the compiler can check this.
3624 The input operands need not be lvalues. The compiler cannot check
3625 whether the operands have data types that are reasonable for the
3626 instruction being executed. It does not parse the assembler instruction
3627 template and does not know what it means or even whether it is valid
3628 assembler input. The extended @code{asm} feature is most often used for
3629 machine instructions the compiler itself does not know exist. If
3630 the output expression cannot be directly addressed (for example, it is a
3631 bit-field), your constraint must allow a register. In that case, GCC
3632 will use the register as the output of the @code{asm}, and then store
3633 that register into the output.
3635 The ordinary output operands must be write-only; GCC will assume that
3636 the values in these operands before the instruction are dead and need
3637 not be generated. Extended asm supports input-output or read-write
3638 operands. Use the constraint character @samp{+} to indicate such an
3639 operand and list it with the output operands.
3641 When the constraints for the read-write operand (or the operand in which
3642 only some of the bits are to be changed) allows a register, you may, as
3643 an alternative, logically split its function into two separate operands,
3644 one input operand and one write-only output operand. The connection
3645 between them is expressed by constraints which say they need to be in
3646 the same location when the instruction executes. You can use the same C
3647 expression for both operands, or different expressions. For example,
3648 here we write the (fictitious) @samp{combine} instruction with
3649 @code{bar} as its read-only source operand and @code{foo} as its
3650 read-write destination:
3653 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3657 The constraint @samp{"0"} for operand 1 says that it must occupy the
3658 same location as operand 0. A number in constraint is allowed only in
3659 an input operand and it must refer to an output operand.
3661 Only a number in the constraint can guarantee that one operand will be in
3662 the same place as another. The mere fact that @code{foo} is the value
3663 of both operands is not enough to guarantee that they will be in the
3664 same place in the generated assembler code. The following would not
3668 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3671 Various optimizations or reloading could cause operands 0 and 1 to be in
3672 different registers; GCC knows no reason not to do so. For example, the
3673 compiler might find a copy of the value of @code{foo} in one register and
3674 use it for operand 1, but generate the output operand 0 in a different
3675 register (copying it afterward to @code{foo}'s own address). Of course,
3676 since the register for operand 1 is not even mentioned in the assembler
3677 code, the result will not work, but GCC can't tell that.
3679 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3680 the operand number for a matching constraint. For example:
3683 asm ("cmoveq %1,%2,%[result]"
3684 : [result] "=r"(result)
3685 : "r" (test), "r"(new), "[result]"(old));
3688 Some instructions clobber specific hard registers. To describe this,
3689 write a third colon after the input operands, followed by the names of
3690 the clobbered hard registers (given as strings). Here is a realistic
3691 example for the VAX:
3694 asm volatile ("movc3 %0,%1,%2"
3696 : "g" (from), "g" (to), "g" (count)
3697 : "r0", "r1", "r2", "r3", "r4", "r5");
3700 You may not write a clobber description in a way that overlaps with an
3701 input or output operand. For example, you may not have an operand
3702 describing a register class with one member if you mention that register
3703 in the clobber list. There is no way for you to specify that an input
3704 operand is modified without also specifying it as an output
3705 operand. Note that if all the output operands you specify are for this
3706 purpose (and hence unused), you will then also need to specify
3707 @code{volatile} for the @code{asm} construct, as described below, to
3708 prevent GCC from deleting the @code{asm} statement as unused.
3710 If you refer to a particular hardware register from the assembler code,
3711 you will probably have to list the register after the third colon to
3712 tell the compiler the register's value is modified. In some assemblers,
3713 the register names begin with @samp{%}; to produce one @samp{%} in the
3714 assembler code, you must write @samp{%%} in the input.
3716 If your assembler instruction can alter the condition code register, add
3717 @samp{cc} to the list of clobbered registers. GCC on some machines
3718 represents the condition codes as a specific hardware register;
3719 @samp{cc} serves to name this register. On other machines, the
3720 condition code is handled differently, and specifying @samp{cc} has no
3721 effect. But it is valid no matter what the machine.
3723 If your assembler instruction modifies memory in an unpredictable
3724 fashion, add @samp{memory} to the list of clobbered registers. This
3725 will cause GCC to not keep memory values cached in registers across
3726 the assembler instruction. You will also want to add the
3727 @code{volatile} keyword if the memory affected is not listed in the
3728 inputs or outputs of the @code{asm}, as the @samp{memory} clobber does
3729 not count as a side-effect of the @code{asm}.
3731 You can put multiple assembler instructions together in a single
3732 @code{asm} template, separated by the characters normally used in assembly
3733 code for the system. A combination that works in most places is a newline
3734 to break the line, plus a tab character to move to the instruction field
3735 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3736 assembler allows semicolons as a line-breaking character. Note that some
3737 assembler dialects use semicolons to start a comment.
3738 The input operands are guaranteed not to use any of the clobbered
3739 registers, and neither will the output operands' addresses, so you can
3740 read and write the clobbered registers as many times as you like. Here
3741 is an example of multiple instructions in a template; it assumes the
3742 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3745 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3747 : "g" (from), "g" (to)
3751 Unless an output operand has the @samp{&} constraint modifier, GCC
3752 may allocate it in the same register as an unrelated input operand, on
3753 the assumption the inputs are consumed before the outputs are produced.
3754 This assumption may be false if the assembler code actually consists of
3755 more than one instruction. In such a case, use @samp{&} for each output
3756 operand that may not overlap an input. @xref{Modifiers}.
3758 If you want to test the condition code produced by an assembler
3759 instruction, you must include a branch and a label in the @code{asm}
3760 construct, as follows:
3763 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3769 This assumes your assembler supports local labels, as the GNU assembler
3770 and most Unix assemblers do.
3772 Speaking of labels, jumps from one @code{asm} to another are not
3773 supported. The compiler's optimizers do not know about these jumps, and
3774 therefore they cannot take account of them when deciding how to
3777 @cindex macros containing @code{asm}
3778 Usually the most convenient way to use these @code{asm} instructions is to
3779 encapsulate them in macros that look like functions. For example,
3783 (@{ double __value, __arg = (x); \
3784 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
3789 Here the variable @code{__arg} is used to make sure that the instruction
3790 operates on a proper @code{double} value, and to accept only those
3791 arguments @code{x} which can convert automatically to a @code{double}.
3793 Another way to make sure the instruction operates on the correct data
3794 type is to use a cast in the @code{asm}. This is different from using a
3795 variable @code{__arg} in that it converts more different types. For
3796 example, if the desired type were @code{int}, casting the argument to
3797 @code{int} would accept a pointer with no complaint, while assigning the
3798 argument to an @code{int} variable named @code{__arg} would warn about
3799 using a pointer unless the caller explicitly casts it.
3801 If an @code{asm} has output operands, GCC assumes for optimization
3802 purposes the instruction has no side effects except to change the output
3803 operands. This does not mean instructions with a side effect cannot be
3804 used, but you must be careful, because the compiler may eliminate them
3805 if the output operands aren't used, or move them out of loops, or
3806 replace two with one if they constitute a common subexpression. Also,
3807 if your instruction does have a side effect on a variable that otherwise
3808 appears not to change, the old value of the variable may be reused later
3809 if it happens to be found in a register.
3811 You can prevent an @code{asm} instruction from being deleted, moved
3812 significantly, or combined, by writing the keyword @code{volatile} after
3813 the @code{asm}. For example:
3816 #define get_and_set_priority(new) \
3818 asm volatile ("get_and_set_priority %0, %1" \
3819 : "=g" (__old) : "g" (new)); \
3824 If you write an @code{asm} instruction with no outputs, GCC will know
3825 the instruction has side-effects and will not delete the instruction or
3826 move it outside of loops.
3828 The @code{volatile} keyword indicates that the instruction has
3829 important side-effects. GCC will not delete a volatile @code{asm} if
3830 it is reachable. (The instruction can still be deleted if GCC can
3831 prove that control-flow will never reach the location of the
3832 instruction.) In addition, GCC will not reschedule instructions
3833 across a volatile @code{asm} instruction. For example:
3836 *(volatile int *)addr = foo;
3837 asm volatile ("eieio" : : );
3841 Assume @code{addr} contains the address of a memory mapped device
3842 register. The PowerPC @code{eieio} instruction (Enforce In-order
3843 Execution of I/O) tells the CPU to make sure that the store to that
3844 device register happens before it issues any other I/O@.
3846 Note that even a volatile @code{asm} instruction can be moved in ways
3847 that appear insignificant to the compiler, such as across jump
3848 instructions. You can't expect a sequence of volatile @code{asm}
3849 instructions to remain perfectly consecutive. If you want consecutive
3850 output, use a single @code{asm}. Also, GCC will perform some
3851 optimizations across a volatile @code{asm} instruction; GCC does not
3852 ``forget everything'' when it encounters a volatile @code{asm}
3853 instruction the way some other compilers do.
3855 An @code{asm} instruction without any operands or clobbers (an ``old
3856 style'' @code{asm}) will be treated identically to a volatile
3857 @code{asm} instruction.
3859 It is a natural idea to look for a way to give access to the condition
3860 code left by the assembler instruction. However, when we attempted to
3861 implement this, we found no way to make it work reliably. The problem
3862 is that output operands might need reloading, which would result in
3863 additional following ``store'' instructions. On most machines, these
3864 instructions would alter the condition code before there was time to
3865 test it. This problem doesn't arise for ordinary ``test'' and
3866 ``compare'' instructions because they don't have any output operands.
3868 For reasons similar to those described above, it is not possible to give
3869 an assembler instruction access to the condition code left by previous
3872 If you are writing a header file that should be includable in ISO C
3873 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
3876 @subsection i386 floating point asm operands
3878 There are several rules on the usage of stack-like regs in
3879 asm_operands insns. These rules apply only to the operands that are
3884 Given a set of input regs that die in an asm_operands, it is
3885 necessary to know which are implicitly popped by the asm, and
3886 which must be explicitly popped by gcc.
3888 An input reg that is implicitly popped by the asm must be
3889 explicitly clobbered, unless it is constrained to match an
3893 For any input reg that is implicitly popped by an asm, it is
3894 necessary to know how to adjust the stack to compensate for the pop.
3895 If any non-popped input is closer to the top of the reg-stack than
3896 the implicitly popped reg, it would not be possible to know what the
3897 stack looked like---it's not clear how the rest of the stack ``slides
3900 All implicitly popped input regs must be closer to the top of
3901 the reg-stack than any input that is not implicitly popped.
3903 It is possible that if an input dies in an insn, reload might
3904 use the input reg for an output reload. Consider this example:
3907 asm ("foo" : "=t" (a) : "f" (b));
3910 This asm says that input B is not popped by the asm, and that
3911 the asm pushes a result onto the reg-stack, i.e., the stack is one
3912 deeper after the asm than it was before. But, it is possible that
3913 reload will think that it can use the same reg for both the input and
3914 the output, if input B dies in this insn.
3916 If any input operand uses the @code{f} constraint, all output reg
3917 constraints must use the @code{&} earlyclobber.
3919 The asm above would be written as
3922 asm ("foo" : "=&t" (a) : "f" (b));
3926 Some operands need to be in particular places on the stack. All
3927 output operands fall in this category---there is no other way to
3928 know which regs the outputs appear in unless the user indicates
3929 this in the constraints.
3931 Output operands must specifically indicate which reg an output
3932 appears in after an asm. @code{=f} is not allowed: the operand
3933 constraints must select a class with a single reg.
3936 Output operands may not be ``inserted'' between existing stack regs.
3937 Since no 387 opcode uses a read/write operand, all output operands
3938 are dead before the asm_operands, and are pushed by the asm_operands.
3939 It makes no sense to push anywhere but the top of the reg-stack.
3941 Output operands must start at the top of the reg-stack: output
3942 operands may not ``skip'' a reg.
3945 Some asm statements may need extra stack space for internal
3946 calculations. This can be guaranteed by clobbering stack registers
3947 unrelated to the inputs and outputs.
3951 Here are a couple of reasonable asms to want to write. This asm
3952 takes one input, which is internally popped, and produces two outputs.
3955 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
3958 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
3959 and replaces them with one output. The user must code the @code{st(1)}
3960 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
3963 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
3969 @section Controlling Names Used in Assembler Code
3970 @cindex assembler names for identifiers
3971 @cindex names used in assembler code
3972 @cindex identifiers, names in assembler code
3974 You can specify the name to be used in the assembler code for a C
3975 function or variable by writing the @code{asm} (or @code{__asm__})
3976 keyword after the declarator as follows:
3979 int foo asm ("myfoo") = 2;
3983 This specifies that the name to be used for the variable @code{foo} in
3984 the assembler code should be @samp{myfoo} rather than the usual
3987 On systems where an underscore is normally prepended to the name of a C
3988 function or variable, this feature allows you to define names for the
3989 linker that do not start with an underscore.
3991 It does not make sense to use this feature with a non-static local
3992 variable since such variables do not have assembler names. If you are
3993 trying to put the variable in a particular register, see @ref{Explicit
3994 Reg Vars}. GCC presently accepts such code with a warning, but will
3995 probably be changed to issue an error, rather than a warning, in the
3998 You cannot use @code{asm} in this way in a function @emph{definition}; but
3999 you can get the same effect by writing a declaration for the function
4000 before its definition and putting @code{asm} there, like this:
4003 extern func () asm ("FUNC");
4010 It is up to you to make sure that the assembler names you choose do not
4011 conflict with any other assembler symbols. Also, you must not use a
4012 register name; that would produce completely invalid assembler code. GCC
4013 does not as yet have the ability to store static variables in registers.
4014 Perhaps that will be added.
4016 @node Explicit Reg Vars
4017 @section Variables in Specified Registers
4018 @cindex explicit register variables
4019 @cindex variables in specified registers
4020 @cindex specified registers
4021 @cindex registers, global allocation
4023 GNU C allows you to put a few global variables into specified hardware
4024 registers. You can also specify the register in which an ordinary
4025 register variable should be allocated.
4029 Global register variables reserve registers throughout the program.
4030 This may be useful in programs such as programming language
4031 interpreters which have a couple of global variables that are accessed
4035 Local register variables in specific registers do not reserve the
4036 registers. The compiler's data flow analysis is capable of determining
4037 where the specified registers contain live values, and where they are
4038 available for other uses. Stores into local register variables may be deleted
4039 when they appear to be dead according to dataflow analysis. References
4040 to local register variables may be deleted or moved or simplified.
4042 These local variables are sometimes convenient for use with the extended
4043 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4044 output of the assembler instruction directly into a particular register.
4045 (This will work provided the register you specify fits the constraints
4046 specified for that operand in the @code{asm}.)
4054 @node Global Reg Vars
4055 @subsection Defining Global Register Variables
4056 @cindex global register variables
4057 @cindex registers, global variables in
4059 You can define a global register variable in GNU C like this:
4062 register int *foo asm ("a5");
4066 Here @code{a5} is the name of the register which should be used. Choose a
4067 register which is normally saved and restored by function calls on your
4068 machine, so that library routines will not clobber it.
4070 Naturally the register name is cpu-dependent, so you would need to
4071 conditionalize your program according to cpu type. The register
4072 @code{a5} would be a good choice on a 68000 for a variable of pointer
4073 type. On machines with register windows, be sure to choose a ``global''
4074 register that is not affected magically by the function call mechanism.
4076 In addition, operating systems on one type of cpu may differ in how they
4077 name the registers; then you would need additional conditionals. For
4078 example, some 68000 operating systems call this register @code{%a5}.
4080 Eventually there may be a way of asking the compiler to choose a register
4081 automatically, but first we need to figure out how it should choose and
4082 how to enable you to guide the choice. No solution is evident.
4084 Defining a global register variable in a certain register reserves that
4085 register entirely for this use, at least within the current compilation.
4086 The register will not be allocated for any other purpose in the functions
4087 in the current compilation. The register will not be saved and restored by
4088 these functions. Stores into this register are never deleted even if they
4089 would appear to be dead, but references may be deleted or moved or
4092 It is not safe to access the global register variables from signal
4093 handlers, or from more than one thread of control, because the system
4094 library routines may temporarily use the register for other things (unless
4095 you recompile them specially for the task at hand).
4097 @cindex @code{qsort}, and global register variables
4098 It is not safe for one function that uses a global register variable to
4099 call another such function @code{foo} by way of a third function
4100 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4101 different source file in which the variable wasn't declared). This is
4102 because @code{lose} might save the register and put some other value there.
4103 For example, you can't expect a global register variable to be available in
4104 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4105 might have put something else in that register. (If you are prepared to
4106 recompile @code{qsort} with the same global register variable, you can
4107 solve this problem.)
4109 If you want to recompile @code{qsort} or other source files which do not
4110 actually use your global register variable, so that they will not use that
4111 register for any other purpose, then it suffices to specify the compiler
4112 option @option{-ffixed-@var{reg}}. You need not actually add a global
4113 register declaration to their source code.
4115 A function which can alter the value of a global register variable cannot
4116 safely be called from a function compiled without this variable, because it
4117 could clobber the value the caller expects to find there on return.
4118 Therefore, the function which is the entry point into the part of the
4119 program that uses the global register variable must explicitly save and
4120 restore the value which belongs to its caller.
4122 @cindex register variable after @code{longjmp}
4123 @cindex global register after @code{longjmp}
4124 @cindex value after @code{longjmp}
4127 On most machines, @code{longjmp} will restore to each global register
4128 variable the value it had at the time of the @code{setjmp}. On some
4129 machines, however, @code{longjmp} will not change the value of global
4130 register variables. To be portable, the function that called @code{setjmp}
4131 should make other arrangements to save the values of the global register
4132 variables, and to restore them in a @code{longjmp}. This way, the same
4133 thing will happen regardless of what @code{longjmp} does.
4135 All global register variable declarations must precede all function
4136 definitions. If such a declaration could appear after function
4137 definitions, the declaration would be too late to prevent the register from
4138 being used for other purposes in the preceding functions.
4140 Global register variables may not have initial values, because an
4141 executable file has no means to supply initial contents for a register.
4143 On the Sparc, there are reports that g3 @dots{} g7 are suitable
4144 registers, but certain library functions, such as @code{getwd}, as well
4145 as the subroutines for division and remainder, modify g3 and g4. g1 and
4146 g2 are local temporaries.
4148 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4149 Of course, it will not do to use more than a few of those.
4151 @node Local Reg Vars
4152 @subsection Specifying Registers for Local Variables
4153 @cindex local variables, specifying registers
4154 @cindex specifying registers for local variables
4155 @cindex registers for local variables
4157 You can define a local register variable with a specified register
4161 register int *foo asm ("a5");
4165 Here @code{a5} is the name of the register which should be used. Note
4166 that this is the same syntax used for defining global register
4167 variables, but for a local variable it would appear within a function.
4169 Naturally the register name is cpu-dependent, but this is not a
4170 problem, since specific registers are most often useful with explicit
4171 assembler instructions (@pxref{Extended Asm}). Both of these things
4172 generally require that you conditionalize your program according to
4175 In addition, operating systems on one type of cpu may differ in how they
4176 name the registers; then you would need additional conditionals. For
4177 example, some 68000 operating systems call this register @code{%a5}.
4179 Defining such a register variable does not reserve the register; it
4180 remains available for other uses in places where flow control determines
4181 the variable's value is not live. However, these registers are made
4182 unavailable for use in the reload pass; excessive use of this feature
4183 leaves the compiler too few available registers to compile certain
4186 This option does not guarantee that GCC will generate code that has
4187 this variable in the register you specify at all times. You may not
4188 code an explicit reference to this register in an @code{asm} statement
4189 and assume it will always refer to this variable.
4191 Stores into local register variables may be deleted when they appear to be dead
4192 according to dataflow analysis. References to local register variables may
4193 be deleted or moved or simplified.
4195 @node Alternate Keywords
4196 @section Alternate Keywords
4197 @cindex alternate keywords
4198 @cindex keywords, alternate
4200 @option{-ansi} and the various @option{-std} options disable certain
4201 keywords. This causes trouble when you want to use GNU C extensions, or
4202 a general-purpose header file that should be usable by all programs,
4203 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4204 @code{inline} are not available in programs compiled with
4205 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4206 program compiled with @option{-std=c99}). The ISO C99 keyword
4207 @code{restrict} is only available when @option{-std=gnu99} (which will
4208 eventually be the default) or @option{-std=c99} (or the equivalent
4209 @option{-std=iso9899:1999}) is used.
4211 The way to solve these problems is to put @samp{__} at the beginning and
4212 end of each problematical keyword. For example, use @code{__asm__}
4213 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4215 Other C compilers won't accept these alternative keywords; if you want to
4216 compile with another compiler, you can define the alternate keywords as
4217 macros to replace them with the customary keywords. It looks like this:
4225 @findex __extension__
4227 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4229 prevent such warnings within one expression by writing
4230 @code{__extension__} before the expression. @code{__extension__} has no
4231 effect aside from this.
4233 @node Incomplete Enums
4234 @section Incomplete @code{enum} Types
4236 You can define an @code{enum} tag without specifying its possible values.
4237 This results in an incomplete type, much like what you get if you write
4238 @code{struct foo} without describing the elements. A later declaration
4239 which does specify the possible values completes the type.
4241 You can't allocate variables or storage using the type while it is
4242 incomplete. However, you can work with pointers to that type.
4244 This extension may not be very useful, but it makes the handling of
4245 @code{enum} more consistent with the way @code{struct} and @code{union}
4248 This extension is not supported by GNU C++.
4250 @node Function Names
4251 @section Function Names as Strings
4252 @cindex @code{__FUNCTION__} identifier
4253 @cindex @code{__PRETTY_FUNCTION__} identifier
4254 @cindex @code{__func__} identifier
4256 GCC predefines two magic identifiers to hold the name of the current
4257 function. The identifier @code{__FUNCTION__} holds the name of the function
4258 as it appears in the source. The identifier @code{__PRETTY_FUNCTION__}
4259 holds the name of the function pretty printed in a language specific
4262 These names are always the same in a C function, but in a C++ function
4263 they may be different. For example, this program:
4267 extern int printf (char *, ...);
4274 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4275 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4293 __PRETTY_FUNCTION__ = int a::sub (int)
4296 The compiler automagically replaces the identifiers with a string
4297 literal containing the appropriate name. Thus, they are neither
4298 preprocessor macros, like @code{__FILE__} and @code{__LINE__}, nor
4299 variables. This means that they catenate with other string literals, and
4300 that they can be used to initialize char arrays. For example
4303 char here[] = "Function " __FUNCTION__ " in " __FILE__;
4306 On the other hand, @samp{#ifdef __FUNCTION__} does not have any special
4307 meaning inside a function, since the preprocessor does not do anything
4308 special with the identifier @code{__FUNCTION__}.
4310 Note that these semantics are deprecated, and that GCC 3.2 will handle
4311 @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} the same way as
4312 @code{__func__}. @code{__func__} is defined by the ISO standard C99:
4315 The identifier @code{__func__} is implicitly declared by the translator
4316 as if, immediately following the opening brace of each function
4317 definition, the declaration
4320 static const char __func__[] = "function-name";
4323 appeared, where function-name is the name of the lexically-enclosing
4324 function. This name is the unadorned name of the function.
4327 By this definition, @code{__func__} is a variable, not a string literal.
4328 In particular, @code{__func__} does not catenate with other string
4331 In @code{C++}, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} are
4332 variables, declared in the same way as @code{__func__}.
4334 @node Return Address
4335 @section Getting the Return or Frame Address of a Function
4337 These functions may be used to get information about the callers of a
4340 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4341 This function returns the return address of the current function, or of
4342 one of its callers. The @var{level} argument is number of frames to
4343 scan up the call stack. A value of @code{0} yields the return address
4344 of the current function, a value of @code{1} yields the return address
4345 of the caller of the current function, and so forth. When inlining
4346 the expected behavior is that the function will return the address of
4347 the function that will be returned to. To work around this behavior use
4348 the @code{noinline} function attribute.
4350 The @var{level} argument must be a constant integer.
4352 On some machines it may be impossible to determine the return address of
4353 any function other than the current one; in such cases, or when the top
4354 of the stack has been reached, this function will return @code{0} or a
4355 random value. In addition, @code{__builtin_frame_address} may be used
4356 to determine if the top of the stack has been reached.
4358 This function should only be used with a nonzero argument for debugging
4362 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4363 This function is similar to @code{__builtin_return_address}, but it
4364 returns the address of the function frame rather than the return address
4365 of the function. Calling @code{__builtin_frame_address} with a value of
4366 @code{0} yields the frame address of the current function, a value of
4367 @code{1} yields the frame address of the caller of the current function,
4370 The frame is the area on the stack which holds local variables and saved
4371 registers. The frame address is normally the address of the first word
4372 pushed on to the stack by the function. However, the exact definition
4373 depends upon the processor and the calling convention. If the processor
4374 has a dedicated frame pointer register, and the function has a frame,
4375 then @code{__builtin_frame_address} will return the value of the frame
4378 On some machines it may be impossible to determine the frame address of
4379 any function other than the current one; in such cases, or when the top
4380 of the stack has been reached, this function will return @code{0} if
4381 the first frame pointer is properly initialized by the startup code.
4383 This function should only be used with a nonzero argument for debugging
4387 @node Vector Extensions
4388 @section Using vector instructions through built-in functions
4390 On some targets, the instruction set contains SIMD vector instructions that
4391 operate on multiple values contained in one large register at the same time.
4392 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4395 The first step in using these extensions is to provide the necessary data
4396 types. This should be done using an appropriate @code{typedef}:
4399 typedef int v4si __attribute__ ((mode(V4SI)));
4402 The base type @code{int} is effectively ignored by the compiler, the
4403 actual properties of the new type @code{v4si} are defined by the
4404 @code{__attribute__}. It defines the machine mode to be used; for vector
4405 types these have the form @code{V@var{n}@var{B}}; @var{n} should be the
4406 number of elements in the vector, and @var{B} should be the base mode of the
4407 individual elements. The following can be used as base modes:
4411 An integer that is as wide as the smallest addressable unit, usually 8 bits.
4413 An integer, twice as wide as a QI mode integer, usually 16 bits.
4415 An integer, four times as wide as a QI mode integer, usually 32 bits.
4417 An integer, eight times as wide as a QI mode integer, usually 64 bits.
4419 A floating point value, as wide as a SI mode integer, usually 32 bits.
4421 A floating point value, as wide as a DI mode integer, usually 64 bits.
4424 There are no @code{V1xx} vector modes - they would be identical to the
4425 corresponding base mode.
4427 Specifying a combination that is not valid for the current architecture
4428 will cause gcc to synthesize the instructions using a narrower mode.
4429 For example, if you specify a variable of type @code{V4SI} and your
4430 architecture does not allow for this specific SIMD type, gcc will
4431 produce code that uses 4 @code{SIs}.
4433 The types defined in this manner can be used with a subset of normal C
4434 operations. Currently, gcc will allow using the following operators on
4435 these types: @code{+, -, *, /, unary minus}@.
4437 The operations behave like C++ @code{valarrays}. Addition is defined as
4438 the addition of the corresponding elements of the operands. For
4439 example, in the code below, each of the 4 elements in @var{a} will be
4440 added to the corresponding 4 elements in @var{b} and the resulting
4441 vector will be stored in @var{c}.
4444 typedef int v4si __attribute__ ((mode(V4SI)));
4451 Subtraction, multiplication, and division operate in a similar manner.
4452 Likewise, the result of using the unary minus operator on a vector type
4453 is a vector whose elements are the negative value of the corresponding
4454 elements in the operand.
4456 You can declare variables and use them in function calls and returns, as
4457 well as in assignments and some casts. You can specify a vector type as
4458 a return type for a function. Vector types can also be used as function
4459 arguments. It is possible to cast from one vector type to another,
4460 provided they are of the same size (in fact, you can also cast vectors
4461 to and from other datatypes of the same size).
4463 You cannot operate between vectors of different lengths or different
4464 signness without a cast.
4466 A port that supports hardware vector operations, usually provides a set
4467 of built-in functions that can be used to operate on vectors. For
4468 example, a function to add two vectors and multiply the result by a
4469 third could look like this:
4472 v4si f (v4si a, v4si b, v4si c)
4474 v4si tmp = __builtin_addv4si (a, b);
4475 return __builtin_mulv4si (tmp, c);
4480 @node Other Builtins
4481 @section Other built-in functions provided by GCC
4482 @cindex built-in functions
4483 @findex __builtin_isgreater
4484 @findex __builtin_isgreaterequal
4485 @findex __builtin_isless
4486 @findex __builtin_islessequal
4487 @findex __builtin_islessgreater
4488 @findex __builtin_isunordered
4517 @findex fprintf_unlocked
4519 @findex fputs_unlocked
4531 @findex printf_unlocked
4553 GCC provides a large number of built-in functions other than the ones
4554 mentioned above. Some of these are for internal use in the processing
4555 of exceptions or variable-length argument lists and will not be
4556 documented here because they may change from time to time; we do not
4557 recommend general use of these functions.
4559 The remaining functions are provided for optimization purposes.
4561 @opindex fno-builtin
4562 GCC includes built-in versions of many of the functions in the standard
4563 C library. The versions prefixed with @code{__builtin_} will always be
4564 treated as having the same meaning as the C library function even if you
4565 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
4566 Many of these functions are only optimized in certain cases; if they are
4567 not optimized in a particular case, a call to the library function will
4572 The functions @code{abort}, @code{exit}, @code{_Exit} and @code{_exit}
4573 are recognized and presumed not to return, but otherwise are not built
4574 in. @code{_exit} is not recognized in strict ISO C mode (@option{-ansi},
4575 @option{-std=c89} or @option{-std=c99}). @code{_Exit} is not recognized in
4576 strict C89 mode (@option{-ansi} or @option{-std=c89}). All these functions
4577 have corresponding versions prefixed with @code{__builtin_}, which may be
4578 used even in strict C89 mode.
4580 Outside strict ISO C mode, the functions @code{alloca}, @code{bcmp},
4581 @code{bzero}, @code{index}, @code{rindex}, @code{ffs}, @code{fputs_unlocked},
4582 @code{printf_unlocked} and @code{fprintf_unlocked} may be handled as
4583 built-in functions. All these functions have corresponding versions
4584 prefixed with @code{__builtin_}, which may be used even in strict C89
4587 The ISO C99 functions @code{conj}, @code{conjf}, @code{conjl},
4588 @code{creal}, @code{crealf}, @code{creall}, @code{cimag}, @code{cimagf},
4589 @code{cimagl}, @code{llabs} and @code{imaxabs} are handled as built-in
4590 functions except in strict ISO C90 mode. There are also built-in
4591 versions of the ISO C99 functions @code{cosf}, @code{cosl},
4592 @code{expf}, @code{expl}, @code{fabsf}, @code{fabsl},
4593 @code{logf}, @code{logl}, @code{sinf}, @code{sinl}, @code{sqrtf}, and
4594 @code{sqrtl}, that are recognized in any mode since ISO C90 reserves
4595 these names for the purpose to which ISO C99 puts them. All these
4596 functions have corresponding versions prefixed with @code{__builtin_}.
4598 The ISO C90 functions @code{abs}, @code{cos}, @code{exp}, @code{fabs},
4599 @code{fprintf}, @code{fputs}, @code{labs}, @code{log},
4600 @code{memcmp}, @code{memcpy},
4601 @code{memset}, @code{printf}, @code{sin}, @code{sqrt}, @code{strcat},
4602 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
4603 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
4604 @code{strpbrk}, @code{strrchr}, @code{strspn}, and @code{strstr} are all
4605 recognized as built-in functions unless @option{-fno-builtin} is
4606 specified (or @option{-fno-builtin-@var{function}} is specified for an
4607 individual function). All of these functions have corresponding
4608 versions prefixed with @code{__builtin_}.
4610 GCC provides built-in versions of the ISO C99 floating point comparison
4611 macros that avoid raising exceptions for unordered operands. They have
4612 the same names as the standard macros ( @code{isgreater},
4613 @code{isgreaterequal}, @code{isless}, @code{islessequal},
4614 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
4615 prefixed. We intend for a library implementor to be able to simply
4616 @code{#define} each standard macro to its built-in equivalent.
4618 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
4620 You can use the built-in function @code{__builtin_types_compatible_p} to
4621 determine whether two types are the same.
4623 This built-in function returns 1 if the unqualified versions of the
4624 types @var{type1} and @var{type2} (which are types, not expressions) are
4625 compatible, 0 otherwise. The result of this built-in function can be
4626 used in integer constant expressions.
4628 This built-in function ignores top level qualifiers (e.g., @code{const},
4629 @code{volatile}). For example, @code{int} is equivalent to @code{const
4632 The type @code{int[]} and @code{int[5]} are compatible. On the other
4633 hand, @code{int} and @code{char *} are not compatible, even if the size
4634 of their types, on the particular architecture are the same. Also, the
4635 amount of pointer indirection is taken into account when determining
4636 similarity. Consequently, @code{short *} is not similar to
4637 @code{short **}. Furthermore, two types that are typedefed are
4638 considered compatible if their underlying types are compatible.
4640 An @code{enum} type is considered to be compatible with another
4641 @code{enum} type. For example, @code{enum @{foo, bar@}} is similar to
4642 @code{enum @{hot, dog@}}.
4644 You would typically use this function in code whose execution varies
4645 depending on the arguments' types. For example:
4651 if (__builtin_types_compatible_p (typeof (x), long double)) \
4652 tmp = foo_long_double (tmp); \
4653 else if (__builtin_types_compatible_p (typeof (x), double)) \
4654 tmp = foo_double (tmp); \
4655 else if (__builtin_types_compatible_p (typeof (x), float)) \
4656 tmp = foo_float (tmp); \
4663 @emph{Note:} This construct is only available for C.
4667 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
4669 You can use the built-in function @code{__builtin_choose_expr} to
4670 evaluate code depending on the value of a constant expression. This
4671 built-in function returns @var{exp1} if @var{const_exp}, which is a
4672 constant expression that must be able to be determined at compile time,
4673 is nonzero. Otherwise it returns 0.
4675 This built-in function is analogous to the @samp{? :} operator in C,
4676 except that the expression returned has its type unaltered by promotion
4677 rules. Also, the built-in function does not evaluate the expression
4678 that was not chosen. For example, if @var{const_exp} evaluates to true,
4679 @var{exp2} is not evaluated even if it has side-effects.
4681 This built-in function can return an lvalue if the chosen argument is an
4684 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
4685 type. Similarly, if @var{exp2} is returned, its return type is the same
4692 __builtin_choose_expr ( \
4693 __builtin_types_compatible_p (typeof (x), double), \
4695 __builtin_choose_expr ( \
4696 __builtin_types_compatible_p (typeof (x), float), \
4698 /* @r{The void expression results in a compile-time error} \
4699 @r{when assigning the result to something.} */ \
4703 @emph{Note:} This construct is only available for C. Furthermore, the
4704 unused expression (@var{exp1} or @var{exp2} depending on the value of
4705 @var{const_exp}) may still generate syntax errors. This may change in
4710 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
4711 You can use the built-in function @code{__builtin_constant_p} to
4712 determine if a value is known to be constant at compile-time and hence
4713 that GCC can perform constant-folding on expressions involving that
4714 value. The argument of the function is the value to test. The function
4715 returns the integer 1 if the argument is known to be a compile-time
4716 constant and 0 if it is not known to be a compile-time constant. A
4717 return of 0 does not indicate that the value is @emph{not} a constant,
4718 but merely that GCC cannot prove it is a constant with the specified
4719 value of the @option{-O} option.
4721 You would typically use this function in an embedded application where
4722 memory was a critical resource. If you have some complex calculation,
4723 you may want it to be folded if it involves constants, but need to call
4724 a function if it does not. For example:
4727 #define Scale_Value(X) \
4728 (__builtin_constant_p (X) \
4729 ? ((X) * SCALE + OFFSET) : Scale (X))
4732 You may use this built-in function in either a macro or an inline
4733 function. However, if you use it in an inlined function and pass an
4734 argument of the function as the argument to the built-in, GCC will
4735 never return 1 when you call the inline function with a string constant
4736 or compound literal (@pxref{Compound Literals}) and will not return 1
4737 when you pass a constant numeric value to the inline function unless you
4738 specify the @option{-O} option.
4740 You may also use @code{__builtin_constant_p} in initializers for static
4741 data. For instance, you can write
4744 static const int table[] = @{
4745 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
4751 This is an acceptable initializer even if @var{EXPRESSION} is not a
4752 constant expression. GCC must be more conservative about evaluating the
4753 built-in in this case, because it has no opportunity to perform
4756 Previous versions of GCC did not accept this built-in in data
4757 initializers. The earliest version where it is completely safe is
4761 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
4762 @opindex fprofile-arcs
4763 You may use @code{__builtin_expect} to provide the compiler with
4764 branch prediction information. In general, you should prefer to
4765 use actual profile feedback for this (@option{-fprofile-arcs}), as
4766 programmers are notoriously bad at predicting how their programs
4767 actually perform. However, there are applications in which this
4768 data is hard to collect.
4770 The return value is the value of @var{exp}, which should be an
4771 integral expression. The value of @var{c} must be a compile-time
4772 constant. The semantics of the built-in are that it is expected
4773 that @var{exp} == @var{c}. For example:
4776 if (__builtin_expect (x, 0))
4781 would indicate that we do not expect to call @code{foo}, since
4782 we expect @code{x} to be zero. Since you are limited to integral
4783 expressions for @var{exp}, you should use constructions such as
4786 if (__builtin_expect (ptr != NULL, 1))
4791 when testing pointer or floating-point values.
4794 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
4795 This function is used to minimize cache-miss latency by moving data into
4796 a cache before it is accessed.
4797 You can insert calls to @code{__builtin_prefetch} into code for which
4798 you know addresses of data in memory that is likely to be accessed soon.
4799 If the target supports them, data prefetch instructions will be generated.
4800 If the prefetch is done early enough before the access then the data will
4801 be in the cache by the time it is accessed.
4803 The value of @var{addr} is the address of the memory to prefetch.
4804 There are two optional arguments, @var{rw} and @var{locality}.
4805 The value of @var{rw} is a compile-time constant one or zero; one
4806 means that the prefetch is preparing for a write to the memory address
4807 and zero, the default, means that the prefetch is preparing for a read.
4808 The value @var{locality} must be a compile-time constant integer between
4809 zero and three. A value of zero means that the data has no temporal
4810 locality, so it need not be left in the cache after the access. A value
4811 of three means that the data has a high degree of temporal locality and
4812 should be left in all levels of cache possible. Values of one and two
4813 mean, respectively, a low or moderate degree of temporal locality. The
4817 for (i = 0; i < n; i++)
4820 __builtin_prefetch (&a[i+j], 1, 1);
4821 __builtin_prefetch (&b[i+j], 0, 1);
4826 Data prefetch does not generate faults if @var{addr} is invalid, but
4827 the address expression itself must be valid. For example, a prefetch
4828 of @code{p->next} will not fault if @code{p->next} is not a valid
4829 address, but evaluation will fault if @code{p} is not a valid address.
4831 If the target does not support data prefetch, the address expression
4832 is evaluated if it includes side effects but no other code is generated
4833 and GCC does not issue a warning.
4836 @deftypefn {Built-in Function} double __builtin_huge_val (void)
4837 Returns a positive infinity, if supported by the floating-point format,
4838 else @code{DBL_MAX}. This function is suitable for implementing the
4839 ISO C macro @code{HUGE_VAL}.
4842 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
4843 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
4846 @deftypefn {Built-in Function} long double __builtin_huge_vall (void)
4847 Similar to @code{__builtin_huge_val}, except the return
4848 type is @code{long double}.
4851 @deftypefn {Built-in Function} double __builtin_inf (void)
4852 Similar to @code{__builtin_huge_val}, except a warning is generated
4853 if the target floating-point format does not support infinities.
4854 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
4857 @deftypefn {Built-in Function} float __builtin_inff (void)
4858 Similar to @code{__builtin_inf}, except the return type is @code{float}.
4861 @deftypefn {Built-in Function} long double __builtin_infl (void)
4862 Similar to @code{__builtin_inf}, except the return
4863 type is @code{long double}.
4866 @node Target Builtins
4867 @section Built-in Functions Specific to Particular Target Machines
4869 On some target machines, GCC supports many built-in functions specific
4870 to those machines. Generally these generate calls to specific machine
4871 instructions, but allow the compiler to schedule those calls.
4874 * Alpha Built-in Functions::
4875 * X86 Built-in Functions::
4876 * PowerPC AltiVec Built-in Functions::
4879 @node Alpha Built-in Functions
4880 @subsection Alpha Built-in Functions
4882 These built-in functions are available for the Alpha family of
4883 processors, depending on the command-line switches used.
4885 The following built-in functions are always available. They
4886 all generate the machine instruction that is part of the name.
4889 long __builtin_alpha_implver (void)
4890 long __builtin_alpha_rpcc (void)
4891 long __builtin_alpha_amask (long)
4892 long __builtin_alpha_cmpbge (long, long)
4893 long __builtin_alpha_extbl (long, long)
4894 long __builtin_alpha_extwl (long, long)
4895 long __builtin_alpha_extll (long, long)
4896 long __builtin_alpha_extql (long, long)
4897 long __builtin_alpha_extwh (long, long)
4898 long __builtin_alpha_extlh (long, long)
4899 long __builtin_alpha_extqh (long, long)
4900 long __builtin_alpha_insbl (long, long)
4901 long __builtin_alpha_inswl (long, long)
4902 long __builtin_alpha_insll (long, long)
4903 long __builtin_alpha_insql (long, long)
4904 long __builtin_alpha_inswh (long, long)
4905 long __builtin_alpha_inslh (long, long)
4906 long __builtin_alpha_insqh (long, long)
4907 long __builtin_alpha_mskbl (long, long)
4908 long __builtin_alpha_mskwl (long, long)
4909 long __builtin_alpha_mskll (long, long)
4910 long __builtin_alpha_mskql (long, long)
4911 long __builtin_alpha_mskwh (long, long)
4912 long __builtin_alpha_msklh (long, long)
4913 long __builtin_alpha_mskqh (long, long)
4914 long __builtin_alpha_umulh (long, long)
4915 long __builtin_alpha_zap (long, long)
4916 long __builtin_alpha_zapnot (long, long)
4919 The following built-in functions are always with @option{-mmax}
4920 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
4921 later. They all generate the machine instruction that is part
4925 long __builtin_alpha_pklb (long)
4926 long __builtin_alpha_pkwb (long)
4927 long __builtin_alpha_unpkbl (long)
4928 long __builtin_alpha_unpkbw (long)
4929 long __builtin_alpha_minub8 (long, long)
4930 long __builtin_alpha_minsb8 (long, long)
4931 long __builtin_alpha_minuw4 (long, long)
4932 long __builtin_alpha_minsw4 (long, long)
4933 long __builtin_alpha_maxub8 (long, long)
4934 long __builtin_alpha_maxsb8 (long, long)
4935 long __builtin_alpha_maxuw4 (long, long)
4936 long __builtin_alpha_maxsw4 (long, long)
4937 long __builtin_alpha_perr (long, long)
4940 The following built-in functions are always with @option{-mcix}
4941 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
4942 later. They all generate the machine instruction that is part
4946 long __builtin_alpha_cttz (long)
4947 long __builtin_alpha_ctlz (long)
4948 long __builtin_alpha_ctpop (long)
4951 The following builtins are available on systems that use the OSF/1
4952 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
4953 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
4954 @code{rdval} and @code{wrval}.
4957 void *__builtin_thread_pointer (void)
4958 void __builtin_set_thread_pointer (void *)
4961 @node X86 Built-in Functions
4962 @subsection X86 Built-in Functions
4964 These built-in functions are available for the i386 and x86-64 family
4965 of computers, depending on the command-line switches used.
4967 The following machine modes are available for use with MMX built-in functions
4968 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
4969 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
4970 vector of eight 8-bit integers. Some of the built-in functions operate on
4971 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
4973 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
4974 of two 32-bit floating point values.
4976 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
4977 floating point values. Some instructions use a vector of four 32-bit
4978 integers, these use @code{V4SI}. Finally, some instructions operate on an
4979 entire vector register, interpreting it as a 128-bit integer, these use mode
4982 The following built-in functions are made available by @option{-mmmx}.
4983 All of them generate the machine instruction that is part of the name.
4986 v8qi __builtin_ia32_paddb (v8qi, v8qi)
4987 v4hi __builtin_ia32_paddw (v4hi, v4hi)
4988 v2si __builtin_ia32_paddd (v2si, v2si)
4989 v8qi __builtin_ia32_psubb (v8qi, v8qi)
4990 v4hi __builtin_ia32_psubw (v4hi, v4hi)
4991 v2si __builtin_ia32_psubd (v2si, v2si)
4992 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
4993 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
4994 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
4995 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
4996 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
4997 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
4998 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
4999 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5000 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5001 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5002 di __builtin_ia32_pand (di, di)
5003 di __builtin_ia32_pandn (di,di)
5004 di __builtin_ia32_por (di, di)
5005 di __builtin_ia32_pxor (di, di)
5006 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5007 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5008 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5009 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5010 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5011 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5012 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5013 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5014 v2si __builtin_ia32_punpckhdq (v2si, v2si)
5015 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5016 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5017 v2si __builtin_ia32_punpckldq (v2si, v2si)
5018 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5019 v4hi __builtin_ia32_packssdw (v2si, v2si)
5020 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5023 The following built-in functions are made available either with
5024 @option{-msse}, or with a combination of @option{-m3dnow} and
5025 @option{-march=athlon}. All of them generate the machine
5026 instruction that is part of the name.
5029 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5030 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5031 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5032 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5033 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5034 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5035 v8qi __builtin_ia32_pminub (v8qi, v8qi)
5036 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5037 int __builtin_ia32_pextrw (v4hi, int)
5038 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5039 int __builtin_ia32_pmovmskb (v8qi)
5040 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5041 void __builtin_ia32_movntq (di *, di)
5042 void __builtin_ia32_sfence (void)
5045 The following built-in functions are available when @option{-msse} is used.
5046 All of them generate the machine instruction that is part of the name.
5049 int __builtin_ia32_comieq (v4sf, v4sf)
5050 int __builtin_ia32_comineq (v4sf, v4sf)
5051 int __builtin_ia32_comilt (v4sf, v4sf)
5052 int __builtin_ia32_comile (v4sf, v4sf)
5053 int __builtin_ia32_comigt (v4sf, v4sf)
5054 int __builtin_ia32_comige (v4sf, v4sf)
5055 int __builtin_ia32_ucomieq (v4sf, v4sf)
5056 int __builtin_ia32_ucomineq (v4sf, v4sf)
5057 int __builtin_ia32_ucomilt (v4sf, v4sf)
5058 int __builtin_ia32_ucomile (v4sf, v4sf)
5059 int __builtin_ia32_ucomigt (v4sf, v4sf)
5060 int __builtin_ia32_ucomige (v4sf, v4sf)
5061 v4sf __builtin_ia32_addps (v4sf, v4sf)
5062 v4sf __builtin_ia32_subps (v4sf, v4sf)
5063 v4sf __builtin_ia32_mulps (v4sf, v4sf)
5064 v4sf __builtin_ia32_divps (v4sf, v4sf)
5065 v4sf __builtin_ia32_addss (v4sf, v4sf)
5066 v4sf __builtin_ia32_subss (v4sf, v4sf)
5067 v4sf __builtin_ia32_mulss (v4sf, v4sf)
5068 v4sf __builtin_ia32_divss (v4sf, v4sf)
5069 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5070 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5071 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5072 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5073 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5074 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5075 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5076 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5077 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5078 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5079 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5080 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5081 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5082 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5083 v4si __builtin_ia32_cmpless (v4sf, v4sf)
5084 v4si __builtin_ia32_cmpgtss (v4sf, v4sf)
5085 v4si __builtin_ia32_cmpgess (v4sf, v4sf)
5086 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5087 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5088 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
5089 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
5090 v4si __builtin_ia32_cmpngtss (v4sf, v4sf)
5091 v4si __builtin_ia32_cmpngess (v4sf, v4sf)
5092 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
5093 v4sf __builtin_ia32_maxps (v4sf, v4sf)
5094 v4sf __builtin_ia32_maxss (v4sf, v4sf)
5095 v4sf __builtin_ia32_minps (v4sf, v4sf)
5096 v4sf __builtin_ia32_minss (v4sf, v4sf)
5097 v4sf __builtin_ia32_andps (v4sf, v4sf)
5098 v4sf __builtin_ia32_andnps (v4sf, v4sf)
5099 v4sf __builtin_ia32_orps (v4sf, v4sf)
5100 v4sf __builtin_ia32_xorps (v4sf, v4sf)
5101 v4sf __builtin_ia32_movss (v4sf, v4sf)
5102 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
5103 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
5104 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
5105 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
5106 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
5107 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
5108 v2si __builtin_ia32_cvtps2pi (v4sf)
5109 int __builtin_ia32_cvtss2si (v4sf)
5110 v2si __builtin_ia32_cvttps2pi (v4sf)
5111 int __builtin_ia32_cvttss2si (v4sf)
5112 v4sf __builtin_ia32_rcpps (v4sf)
5113 v4sf __builtin_ia32_rsqrtps (v4sf)
5114 v4sf __builtin_ia32_sqrtps (v4sf)
5115 v4sf __builtin_ia32_rcpss (v4sf)
5116 v4sf __builtin_ia32_rsqrtss (v4sf)
5117 v4sf __builtin_ia32_sqrtss (v4sf)
5118 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
5119 void __builtin_ia32_movntps (float *, v4sf)
5120 int __builtin_ia32_movmskps (v4sf)
5123 The following built-in functions are available when @option{-msse} is used.
5126 @item v4sf __builtin_ia32_loadaps (float *)
5127 Generates the @code{movaps} machine instruction as a load from memory.
5128 @item void __builtin_ia32_storeaps (float *, v4sf)
5129 Generates the @code{movaps} machine instruction as a store to memory.
5130 @item v4sf __builtin_ia32_loadups (float *)
5131 Generates the @code{movups} machine instruction as a load from memory.
5132 @item void __builtin_ia32_storeups (float *, v4sf)
5133 Generates the @code{movups} machine instruction as a store to memory.
5134 @item v4sf __builtin_ia32_loadsss (float *)
5135 Generates the @code{movss} machine instruction as a load from memory.
5136 @item void __builtin_ia32_storess (float *, v4sf)
5137 Generates the @code{movss} machine instruction as a store to memory.
5138 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
5139 Generates the @code{movhps} machine instruction as a load from memory.
5140 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
5141 Generates the @code{movlps} machine instruction as a load from memory
5142 @item void __builtin_ia32_storehps (v4sf, v2si *)
5143 Generates the @code{movhps} machine instruction as a store to memory.
5144 @item void __builtin_ia32_storelps (v4sf, v2si *)
5145 Generates the @code{movlps} machine instruction as a store to memory.
5148 The following built-in functions are available when @option{-m3dnow} is used.
5149 All of them generate the machine instruction that is part of the name.
5152 void __builtin_ia32_femms (void)
5153 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
5154 v2si __builtin_ia32_pf2id (v2sf)
5155 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
5156 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
5157 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
5158 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
5159 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
5160 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
5161 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
5162 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
5163 v2sf __builtin_ia32_pfrcp (v2sf)
5164 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
5165 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
5166 v2sf __builtin_ia32_pfrsqrt (v2sf)
5167 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
5168 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
5169 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
5170 v2sf __builtin_ia32_pi2fd (v2si)
5171 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
5174 The following built-in functions are available when both @option{-m3dnow}
5175 and @option{-march=athlon} are used. All of them generate the machine
5176 instruction that is part of the name.
5179 v2si __builtin_ia32_pf2iw (v2sf)
5180 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
5181 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
5182 v2sf __builtin_ia32_pi2fw (v2si)
5183 v2sf __builtin_ia32_pswapdsf (v2sf)
5184 v2si __builtin_ia32_pswapdsi (v2si)
5187 @node PowerPC AltiVec Built-in Functions
5188 @subsection PowerPC AltiVec Built-in Functions
5190 These built-in functions are available for the PowerPC family
5191 of computers, depending on the command-line switches used.
5193 The following machine modes are available for use with AltiVec built-in
5194 functions (@pxref{Vector Extensions}): @code{V4SI} for a vector of four
5195 32-bit integers, @code{V4SF} for a vector of four 32-bit floating point
5196 numbers, @code{V8HI} for a vector of eight 16-bit integers, and
5197 @code{V16QI} for a vector of sixteen 8-bit integers.
5199 The following functions are made available by including
5200 @code{<altivec.h>} and using @option{-maltivec} and
5201 @option{-mabi=altivec}. The functions implement the functionality
5202 described in Motorola's AltiVec Programming Interface Manual.
5204 There are a few differences from Motorola's documentation and GCC's
5205 implementation. Vector constants are done with curly braces (not
5206 parentheses). Vector initializers require no casts if the vector
5207 constant is of the same type as the variable it is initializing. The
5208 @code{vector bool} type is deprecated and will be discontinued in
5209 further revisions. Use @code{vector signed} instead. If @code{signed}
5210 or @code{unsigned} is omitted, the vector type will default to
5211 @code{signed}. Lastly, all overloaded functions are implemented with macros
5212 for the C implementation. So code the following example will not work:
5215 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
5218 Since vec_add is a macro, the vector constant in the above example will
5219 be treated as four different arguments. Wrap the entire argument in
5220 parentheses for this to work. The C++ implementation does not use
5223 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
5224 Internally, GCC uses built-in functions to achieve the functionality in
5225 the aforementioned header file, but they are not supported and are
5226 subject to change without notice.
5229 vector signed char vec_abs (vector signed char, vector signed char);
5230 vector signed short vec_abs (vector signed short, vector signed short);
5231 vector signed int vec_abs (vector signed int, vector signed int);
5232 vector signed float vec_abs (vector signed float, vector signed float);
5234 vector signed char vec_abss (vector signed char, vector signed char);
5235 vector signed short vec_abss (vector signed short, vector signed short);
5237 vector signed char vec_add (vector signed char, vector signed char);
5238 vector unsigned char vec_add (vector signed char, vector unsigned char);
5240 vector unsigned char vec_add (vector unsigned char, vector signed char);
5242 vector unsigned char vec_add (vector unsigned char,
5243 vector unsigned char);
5244 vector signed short vec_add (vector signed short, vector signed short);
5245 vector unsigned short vec_add (vector signed short,
5246 vector unsigned short);
5247 vector unsigned short vec_add (vector unsigned short,
5248 vector signed short);
5249 vector unsigned short vec_add (vector unsigned short,
5250 vector unsigned short);
5251 vector signed int vec_add (vector signed int, vector signed int);
5252 vector unsigned int vec_add (vector signed int, vector unsigned int);
5253 vector unsigned int vec_add (vector unsigned int, vector signed int);
5254 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
5255 vector float vec_add (vector float, vector float);
5257 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
5259 vector unsigned char vec_adds (vector signed char,
5260 vector unsigned char);
5261 vector unsigned char vec_adds (vector unsigned char,
5262 vector signed char);
5263 vector unsigned char vec_adds (vector unsigned char,
5264 vector unsigned char);
5265 vector signed char vec_adds (vector signed char, vector signed char);
5266 vector unsigned short vec_adds (vector signed short,
5267 vector unsigned short);
5268 vector unsigned short vec_adds (vector unsigned short,
5269 vector signed short);
5270 vector unsigned short vec_adds (vector unsigned short,
5271 vector unsigned short);
5272 vector signed short vec_adds (vector signed short, vector signed short);
5274 vector unsigned int vec_adds (vector signed int, vector unsigned int);
5275 vector unsigned int vec_adds (vector unsigned int, vector signed int);
5276 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
5278 vector signed int vec_adds (vector signed int, vector signed int);
5280 vector float vec_and (vector float, vector float);
5281 vector float vec_and (vector float, vector signed int);
5282 vector float vec_and (vector signed int, vector float);
5283 vector signed int vec_and (vector signed int, vector signed int);
5284 vector unsigned int vec_and (vector signed int, vector unsigned int);
5285 vector unsigned int vec_and (vector unsigned int, vector signed int);
5286 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
5287 vector signed short vec_and (vector signed short, vector signed short);
5288 vector unsigned short vec_and (vector signed short,
5289 vector unsigned short);
5290 vector unsigned short vec_and (vector unsigned short,
5291 vector signed short);
5292 vector unsigned short vec_and (vector unsigned short,
5293 vector unsigned short);
5294 vector signed char vec_and (vector signed char, vector signed char);
5295 vector unsigned char vec_and (vector signed char, vector unsigned char);
5297 vector unsigned char vec_and (vector unsigned char, vector signed char);
5299 vector unsigned char vec_and (vector unsigned char,
5300 vector unsigned char);
5302 vector float vec_andc (vector float, vector float);
5303 vector float vec_andc (vector float, vector signed int);
5304 vector float vec_andc (vector signed int, vector float);
5305 vector signed int vec_andc (vector signed int, vector signed int);
5306 vector unsigned int vec_andc (vector signed int, vector unsigned int);
5307 vector unsigned int vec_andc (vector unsigned int, vector signed int);
5308 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
5310 vector signed short vec_andc (vector signed short, vector signed short);
5312 vector unsigned short vec_andc (vector signed short,
5313 vector unsigned short);
5314 vector unsigned short vec_andc (vector unsigned short,
5315 vector signed short);
5316 vector unsigned short vec_andc (vector unsigned short,
5317 vector unsigned short);
5318 vector signed char vec_andc (vector signed char, vector signed char);
5319 vector unsigned char vec_andc (vector signed char,
5320 vector unsigned char);
5321 vector unsigned char vec_andc (vector unsigned char,
5322 vector signed char);
5323 vector unsigned char vec_andc (vector unsigned char,
5324 vector unsigned char);
5326 vector unsigned char vec_avg (vector unsigned char,
5327 vector unsigned char);
5328 vector signed char vec_avg (vector signed char, vector signed char);
5329 vector unsigned short vec_avg (vector unsigned short,
5330 vector unsigned short);
5331 vector signed short vec_avg (vector signed short, vector signed short);
5332 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
5333 vector signed int vec_avg (vector signed int, vector signed int);
5335 vector float vec_ceil (vector float);
5337 vector signed int vec_cmpb (vector float, vector float);
5339 vector signed char vec_cmpeq (vector signed char, vector signed char);
5340 vector signed char vec_cmpeq (vector unsigned char,
5341 vector unsigned char);
5342 vector signed short vec_cmpeq (vector signed short,
5343 vector signed short);
5344 vector signed short vec_cmpeq (vector unsigned short,
5345 vector unsigned short);
5346 vector signed int vec_cmpeq (vector signed int, vector signed int);
5347 vector signed int vec_cmpeq (vector unsigned int, vector unsigned int);
5348 vector signed int vec_cmpeq (vector float, vector float);
5350 vector signed int vec_cmpge (vector float, vector float);
5352 vector signed char vec_cmpgt (vector unsigned char,
5353 vector unsigned char);
5354 vector signed char vec_cmpgt (vector signed char, vector signed char);
5355 vector signed short vec_cmpgt (vector unsigned short,
5356 vector unsigned short);
5357 vector signed short vec_cmpgt (vector signed short,
5358 vector signed short);
5359 vector signed int vec_cmpgt (vector unsigned int, vector unsigned int);
5360 vector signed int vec_cmpgt (vector signed int, vector signed int);
5361 vector signed int vec_cmpgt (vector float, vector float);
5363 vector signed int vec_cmple (vector float, vector float);
5365 vector signed char vec_cmplt (vector unsigned char,
5366 vector unsigned char);
5367 vector signed char vec_cmplt (vector signed char, vector signed char);
5368 vector signed short vec_cmplt (vector unsigned short,
5369 vector unsigned short);
5370 vector signed short vec_cmplt (vector signed short,
5371 vector signed short);
5372 vector signed int vec_cmplt (vector unsigned int, vector unsigned int);
5373 vector signed int vec_cmplt (vector signed int, vector signed int);
5374 vector signed int vec_cmplt (vector float, vector float);
5376 vector float vec_ctf (vector unsigned int, const char);
5377 vector float vec_ctf (vector signed int, const char);
5379 vector signed int vec_cts (vector float, const char);
5381 vector unsigned int vec_ctu (vector float, const char);
5383 void vec_dss (const char);
5385 void vec_dssall (void);
5387 void vec_dst (void *, int, const char);
5389 void vec_dstst (void *, int, const char);
5391 void vec_dststt (void *, int, const char);
5393 void vec_dstt (void *, int, const char);
5395 vector float vec_expte (vector float, vector float);
5397 vector float vec_floor (vector float, vector float);
5399 vector float vec_ld (int, vector float *);
5400 vector float vec_ld (int, float *):
5401 vector signed int vec_ld (int, int *);
5402 vector signed int vec_ld (int, vector signed int *);
5403 vector unsigned int vec_ld (int, vector unsigned int *);
5404 vector unsigned int vec_ld (int, unsigned int *);
5405 vector signed short vec_ld (int, short *, vector signed short *);
5406 vector unsigned short vec_ld (int, unsigned short *,
5407 vector unsigned short *);
5408 vector signed char vec_ld (int, signed char *);
5409 vector signed char vec_ld (int, vector signed char *);
5410 vector unsigned char vec_ld (int, unsigned char *);
5411 vector unsigned char vec_ld (int, vector unsigned char *);
5413 vector signed char vec_lde (int, signed char *);
5414 vector unsigned char vec_lde (int, unsigned char *);
5415 vector signed short vec_lde (int, short *);
5416 vector unsigned short vec_lde (int, unsigned short *);
5417 vector float vec_lde (int, float *);
5418 vector signed int vec_lde (int, int *);
5419 vector unsigned int vec_lde (int, unsigned int *);
5421 void float vec_ldl (int, float *);
5422 void float vec_ldl (int, vector float *);
5423 void signed int vec_ldl (int, vector signed int *);
5424 void signed int vec_ldl (int, int *);
5425 void unsigned int vec_ldl (int, unsigned int *);
5426 void unsigned int vec_ldl (int, vector unsigned int *);
5427 void signed short vec_ldl (int, vector signed short *);
5428 void signed short vec_ldl (int, short *);
5429 void unsigned short vec_ldl (int, vector unsigned short *);
5430 void unsigned short vec_ldl (int, unsigned short *);
5431 void signed char vec_ldl (int, vector signed char *);
5432 void signed char vec_ldl (int, signed char *);
5433 void unsigned char vec_ldl (int, vector unsigned char *);
5434 void unsigned char vec_ldl (int, unsigned char *);
5436 vector float vec_loge (vector float);
5438 vector unsigned char vec_lvsl (int, void *, int *);
5440 vector unsigned char vec_lvsr (int, void *, int *);
5442 vector float vec_madd (vector float, vector float, vector float);
5444 vector signed short vec_madds (vector signed short, vector signed short,
5445 vector signed short);
5447 vector unsigned char vec_max (vector signed char, vector unsigned char);
5449 vector unsigned char vec_max (vector unsigned char, vector signed char);
5451 vector unsigned char vec_max (vector unsigned char,
5452 vector unsigned char);
5453 vector signed char vec_max (vector signed char, vector signed char);
5454 vector unsigned short vec_max (vector signed short,
5455 vector unsigned short);
5456 vector unsigned short vec_max (vector unsigned short,
5457 vector signed short);
5458 vector unsigned short vec_max (vector unsigned short,
5459 vector unsigned short);
5460 vector signed short vec_max (vector signed short, vector signed short);
5461 vector unsigned int vec_max (vector signed int, vector unsigned int);
5462 vector unsigned int vec_max (vector unsigned int, vector signed int);
5463 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
5464 vector signed int vec_max (vector signed int, vector signed int);
5465 vector float vec_max (vector float, vector float);
5467 vector signed char vec_mergeh (vector signed char, vector signed char);
5468 vector unsigned char vec_mergeh (vector unsigned char,
5469 vector unsigned char);
5470 vector signed short vec_mergeh (vector signed short,
5471 vector signed short);
5472 vector unsigned short vec_mergeh (vector unsigned short,
5473 vector unsigned short);
5474 vector float vec_mergeh (vector float, vector float);
5475 vector signed int vec_mergeh (vector signed int, vector signed int);
5476 vector unsigned int vec_mergeh (vector unsigned int,
5477 vector unsigned int);
5479 vector signed char vec_mergel (vector signed char, vector signed char);
5480 vector unsigned char vec_mergel (vector unsigned char,
5481 vector unsigned char);
5482 vector signed short vec_mergel (vector signed short,
5483 vector signed short);
5484 vector unsigned short vec_mergel (vector unsigned short,
5485 vector unsigned short);
5486 vector float vec_mergel (vector float, vector float);
5487 vector signed int vec_mergel (vector signed int, vector signed int);
5488 vector unsigned int vec_mergel (vector unsigned int,
5489 vector unsigned int);
5491 vector unsigned short vec_mfvscr (void);
5493 vector unsigned char vec_min (vector signed char, vector unsigned char);
5495 vector unsigned char vec_min (vector unsigned char, vector signed char);
5497 vector unsigned char vec_min (vector unsigned char,
5498 vector unsigned char);
5499 vector signed char vec_min (vector signed char, vector signed char);
5500 vector unsigned short vec_min (vector signed short,
5501 vector unsigned short);
5502 vector unsigned short vec_min (vector unsigned short,
5503 vector signed short);
5504 vector unsigned short vec_min (vector unsigned short,
5505 vector unsigned short);
5506 vector signed short vec_min (vector signed short, vector signed short);
5507 vector unsigned int vec_min (vector signed int, vector unsigned int);
5508 vector unsigned int vec_min (vector unsigned int, vector signed int);
5509 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
5510 vector signed int vec_min (vector signed int, vector signed int);
5511 vector float vec_min (vector float, vector float);
5513 vector signed short vec_mladd (vector signed short, vector signed short,
5514 vector signed short);
5515 vector signed short vec_mladd (vector signed short,
5516 vector unsigned short,
5517 vector unsigned short);
5518 vector signed short vec_mladd (vector unsigned short,
5519 vector signed short,
5520 vector signed short);
5521 vector unsigned short vec_mladd (vector unsigned short,
5522 vector unsigned short,
5523 vector unsigned short);
5525 vector signed short vec_mradds (vector signed short,
5526 vector signed short,
5527 vector signed short);
5529 vector unsigned int vec_msum (vector unsigned char,
5530 vector unsigned char,
5531 vector unsigned int);
5532 vector signed int vec_msum (vector signed char, vector unsigned char,
5534 vector unsigned int vec_msum (vector unsigned short,
5535 vector unsigned short,
5536 vector unsigned int);
5537 vector signed int vec_msum (vector signed short, vector signed short,
5540 vector unsigned int vec_msums (vector unsigned short,
5541 vector unsigned short,
5542 vector unsigned int);
5543 vector signed int vec_msums (vector signed short, vector signed short,
5546 void vec_mtvscr (vector signed int);
5547 void vec_mtvscr (vector unsigned int);
5548 void vec_mtvscr (vector signed short);
5549 void vec_mtvscr (vector unsigned short);
5550 void vec_mtvscr (vector signed char);
5551 void vec_mtvscr (vector unsigned char);
5553 vector unsigned short vec_mule (vector unsigned char,
5554 vector unsigned char);
5555 vector signed short vec_mule (vector signed char, vector signed char);
5556 vector unsigned int vec_mule (vector unsigned short,
5557 vector unsigned short);
5558 vector signed int vec_mule (vector signed short, vector signed short);
5560 vector unsigned short vec_mulo (vector unsigned char,
5561 vector unsigned char);
5562 vector signed short vec_mulo (vector signed char, vector signed char);
5563 vector unsigned int vec_mulo (vector unsigned short,
5564 vector unsigned short);
5565 vector signed int vec_mulo (vector signed short, vector signed short);
5567 vector float vec_nmsub (vector float, vector float, vector float);
5569 vector float vec_nor (vector float, vector float);
5570 vector signed int vec_nor (vector signed int, vector signed int);
5571 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
5572 vector signed short vec_nor (vector signed short, vector signed short);
5573 vector unsigned short vec_nor (vector unsigned short,
5574 vector unsigned short);
5575 vector signed char vec_nor (vector signed char, vector signed char);
5576 vector unsigned char vec_nor (vector unsigned char,
5577 vector unsigned char);
5579 vector float vec_or (vector float, vector float);
5580 vector float vec_or (vector float, vector signed int);
5581 vector float vec_or (vector signed int, vector float);
5582 vector signed int vec_or (vector signed int, vector signed int);
5583 vector unsigned int vec_or (vector signed int, vector unsigned int);
5584 vector unsigned int vec_or (vector unsigned int, vector signed int);
5585 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
5586 vector signed short vec_or (vector signed short, vector signed short);
5587 vector unsigned short vec_or (vector signed short,
5588 vector unsigned short);
5589 vector unsigned short vec_or (vector unsigned short,
5590 vector signed short);
5591 vector unsigned short vec_or (vector unsigned short,
5592 vector unsigned short);
5593 vector signed char vec_or (vector signed char, vector signed char);
5594 vector unsigned char vec_or (vector signed char, vector unsigned char);
5595 vector unsigned char vec_or (vector unsigned char, vector signed char);
5596 vector unsigned char vec_or (vector unsigned char,
5597 vector unsigned char);
5599 vector signed char vec_pack (vector signed short, vector signed short);
5600 vector unsigned char vec_pack (vector unsigned short,
5601 vector unsigned short);
5602 vector signed short vec_pack (vector signed int, vector signed int);
5603 vector unsigned short vec_pack (vector unsigned int,
5604 vector unsigned int);
5606 vector signed short vec_packpx (vector unsigned int,
5607 vector unsigned int);
5609 vector unsigned char vec_packs (vector unsigned short,
5610 vector unsigned short);
5611 vector signed char vec_packs (vector signed short, vector signed short);
5613 vector unsigned short vec_packs (vector unsigned int,
5614 vector unsigned int);
5615 vector signed short vec_packs (vector signed int, vector signed int);
5617 vector unsigned char vec_packsu (vector unsigned short,
5618 vector unsigned short);
5619 vector unsigned char vec_packsu (vector signed short,
5620 vector signed short);
5621 vector unsigned short vec_packsu (vector unsigned int,
5622 vector unsigned int);
5623 vector unsigned short vec_packsu (vector signed int, vector signed int);
5625 vector float vec_perm (vector float, vector float,
5626 vector unsigned char);
5627 vector signed int vec_perm (vector signed int, vector signed int,
5628 vector unsigned char);
5629 vector unsigned int vec_perm (vector unsigned int, vector unsigned int,
5630 vector unsigned char);
5631 vector signed short vec_perm (vector signed short, vector signed short,
5632 vector unsigned char);
5633 vector unsigned short vec_perm (vector unsigned short,
5634 vector unsigned short,
5635 vector unsigned char);
5636 vector signed char vec_perm (vector signed char, vector signed char,
5637 vector unsigned char);
5638 vector unsigned char vec_perm (vector unsigned char,
5639 vector unsigned char,
5640 vector unsigned char);
5642 vector float vec_re (vector float);
5644 vector signed char vec_rl (vector signed char, vector unsigned char);
5645 vector unsigned char vec_rl (vector unsigned char,
5646 vector unsigned char);
5647 vector signed short vec_rl (vector signed short, vector unsigned short);
5649 vector unsigned short vec_rl (vector unsigned short,
5650 vector unsigned short);
5651 vector signed int vec_rl (vector signed int, vector unsigned int);
5652 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
5654 vector float vec_round (vector float);
5656 vector float vec_rsqrte (vector float);
5658 vector float vec_sel (vector float, vector float, vector signed int);
5659 vector float vec_sel (vector float, vector float, vector unsigned int);
5660 vector signed int vec_sel (vector signed int, vector signed int,
5662 vector signed int vec_sel (vector signed int, vector signed int,
5663 vector unsigned int);
5664 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
5666 vector unsigned int vec_sel (vector unsigned int, vector unsigned int,
5667 vector unsigned int);
5668 vector signed short vec_sel (vector signed short, vector signed short,
5669 vector signed short);
5670 vector signed short vec_sel (vector signed short, vector signed short,
5671 vector unsigned short);
5672 vector unsigned short vec_sel (vector unsigned short,
5673 vector unsigned short,
5674 vector signed short);
5675 vector unsigned short vec_sel (vector unsigned short,
5676 vector unsigned short,
5677 vector unsigned short);
5678 vector signed char vec_sel (vector signed char, vector signed char,
5679 vector signed char);
5680 vector signed char vec_sel (vector signed char, vector signed char,
5681 vector unsigned char);
5682 vector unsigned char vec_sel (vector unsigned char,
5683 vector unsigned char,
5684 vector signed char);
5685 vector unsigned char vec_sel (vector unsigned char,
5686 vector unsigned char,
5687 vector unsigned char);
5689 vector signed char vec_sl (vector signed char, vector unsigned char);
5690 vector unsigned char vec_sl (vector unsigned char,
5691 vector unsigned char);
5692 vector signed short vec_sl (vector signed short, vector unsigned short);
5694 vector unsigned short vec_sl (vector unsigned short,
5695 vector unsigned short);
5696 vector signed int vec_sl (vector signed int, vector unsigned int);
5697 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
5699 vector float vec_sld (vector float, vector float, const char);
5700 vector signed int vec_sld (vector signed int, vector signed int,
5702 vector unsigned int vec_sld (vector unsigned int, vector unsigned int,
5704 vector signed short vec_sld (vector signed short, vector signed short,
5706 vector unsigned short vec_sld (vector unsigned short,
5707 vector unsigned short, const char);
5708 vector signed char vec_sld (vector signed char, vector signed char,
5710 vector unsigned char vec_sld (vector unsigned char,
5711 vector unsigned char,
5714 vector signed int vec_sll (vector signed int, vector unsigned int);
5715 vector signed int vec_sll (vector signed int, vector unsigned short);
5716 vector signed int vec_sll (vector signed int, vector unsigned char);
5717 vector unsigned int vec_sll (vector unsigned int, vector unsigned int);
5718 vector unsigned int vec_sll (vector unsigned int,
5719 vector unsigned short);
5720 vector unsigned int vec_sll (vector unsigned int, vector unsigned char);
5722 vector signed short vec_sll (vector signed short, vector unsigned int);
5723 vector signed short vec_sll (vector signed short,
5724 vector unsigned short);
5725 vector signed short vec_sll (vector signed short, vector unsigned char);
5727 vector unsigned short vec_sll (vector unsigned short,
5728 vector unsigned int);
5729 vector unsigned short vec_sll (vector unsigned short,
5730 vector unsigned short);
5731 vector unsigned short vec_sll (vector unsigned short,
5732 vector unsigned char);
5733 vector signed char vec_sll (vector signed char, vector unsigned int);
5734 vector signed char vec_sll (vector signed char, vector unsigned short);
5735 vector signed char vec_sll (vector signed char, vector unsigned char);
5736 vector unsigned char vec_sll (vector unsigned char,
5737 vector unsigned int);
5738 vector unsigned char vec_sll (vector unsigned char,
5739 vector unsigned short);
5740 vector unsigned char vec_sll (vector unsigned char,
5741 vector unsigned char);
5743 vector float vec_slo (vector float, vector signed char);
5744 vector float vec_slo (vector float, vector unsigned char);
5745 vector signed int vec_slo (vector signed int, vector signed char);
5746 vector signed int vec_slo (vector signed int, vector unsigned char);
5747 vector unsigned int vec_slo (vector unsigned int, vector signed char);
5748 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
5750 vector signed short vec_slo (vector signed short, vector signed char);
5751 vector signed short vec_slo (vector signed short, vector unsigned char);
5753 vector unsigned short vec_slo (vector unsigned short,
5754 vector signed char);
5755 vector unsigned short vec_slo (vector unsigned short,
5756 vector unsigned char);
5757 vector signed char vec_slo (vector signed char, vector signed char);
5758 vector signed char vec_slo (vector signed char, vector unsigned char);
5759 vector unsigned char vec_slo (vector unsigned char, vector signed char);
5761 vector unsigned char vec_slo (vector unsigned char,
5762 vector unsigned char);
5764 vector signed char vec_splat (vector signed char, const char);
5765 vector unsigned char vec_splat (vector unsigned char, const char);
5766 vector signed short vec_splat (vector signed short, const char);
5767 vector unsigned short vec_splat (vector unsigned short, const char);
5768 vector float vec_splat (vector float, const char);
5769 vector signed int vec_splat (vector signed int, const char);
5770 vector unsigned int vec_splat (vector unsigned int, const char);
5772 vector signed char vec_splat_s8 (const char);
5774 vector signed short vec_splat_s16 (const char);
5776 vector signed int vec_splat_s32 (const char);
5778 vector unsigned char vec_splat_u8 (const char);
5780 vector unsigned short vec_splat_u16 (const char);
5782 vector unsigned int vec_splat_u32 (const char);
5784 vector signed char vec_sr (vector signed char, vector unsigned char);
5785 vector unsigned char vec_sr (vector unsigned char,
5786 vector unsigned char);
5787 vector signed short vec_sr (vector signed short, vector unsigned short);
5789 vector unsigned short vec_sr (vector unsigned short,
5790 vector unsigned short);
5791 vector signed int vec_sr (vector signed int, vector unsigned int);
5792 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
5794 vector signed char vec_sra (vector signed char, vector unsigned char);
5795 vector unsigned char vec_sra (vector unsigned char,
5796 vector unsigned char);
5797 vector signed short vec_sra (vector signed short,
5798 vector unsigned short);
5799 vector unsigned short vec_sra (vector unsigned short,
5800 vector unsigned short);
5801 vector signed int vec_sra (vector signed int, vector unsigned int);
5802 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
5804 vector signed int vec_srl (vector signed int, vector unsigned int);
5805 vector signed int vec_srl (vector signed int, vector unsigned short);
5806 vector signed int vec_srl (vector signed int, vector unsigned char);
5807 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
5808 vector unsigned int vec_srl (vector unsigned int,
5809 vector unsigned short);
5810 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
5812 vector signed short vec_srl (vector signed short, vector unsigned int);
5813 vector signed short vec_srl (vector signed short,
5814 vector unsigned short);
5815 vector signed short vec_srl (vector signed short, vector unsigned char);
5817 vector unsigned short vec_srl (vector unsigned short,
5818 vector unsigned int);
5819 vector unsigned short vec_srl (vector unsigned short,
5820 vector unsigned short);
5821 vector unsigned short vec_srl (vector unsigned short,
5822 vector unsigned char);
5823 vector signed char vec_srl (vector signed char, vector unsigned int);
5824 vector signed char vec_srl (vector signed char, vector unsigned short);
5825 vector signed char vec_srl (vector signed char, vector unsigned char);
5826 vector unsigned char vec_srl (vector unsigned char,
5827 vector unsigned int);
5828 vector unsigned char vec_srl (vector unsigned char,
5829 vector unsigned short);
5830 vector unsigned char vec_srl (vector unsigned char,
5831 vector unsigned char);
5833 vector float vec_sro (vector float, vector signed char);
5834 vector float vec_sro (vector float, vector unsigned char);
5835 vector signed int vec_sro (vector signed int, vector signed char);
5836 vector signed int vec_sro (vector signed int, vector unsigned char);
5837 vector unsigned int vec_sro (vector unsigned int, vector signed char);
5838 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
5840 vector signed short vec_sro (vector signed short, vector signed char);
5841 vector signed short vec_sro (vector signed short, vector unsigned char);
5843 vector unsigned short vec_sro (vector unsigned short,
5844 vector signed char);
5845 vector unsigned short vec_sro (vector unsigned short,
5846 vector unsigned char);
5847 vector signed char vec_sro (vector signed char, vector signed char);
5848 vector signed char vec_sro (vector signed char, vector unsigned char);
5849 vector unsigned char vec_sro (vector unsigned char, vector signed char);
5851 vector unsigned char vec_sro (vector unsigned char,
5852 vector unsigned char);
5854 void vec_st (vector float, int, float *);
5855 void vec_st (vector float, int, vector float *);
5856 void vec_st (vector signed int, int, int *);
5857 void vec_st (vector signed int, int, unsigned int *);
5858 void vec_st (vector unsigned int, int, unsigned int *);
5859 void vec_st (vector unsigned int, int, vector unsigned int *);
5860 void vec_st (vector signed short, int, short *);
5861 void vec_st (vector signed short, int, vector unsigned short *);
5862 void vec_st (vector signed short, int, vector signed short *);
5863 void vec_st (vector unsigned short, int, unsigned short *);
5864 void vec_st (vector unsigned short, int, vector unsigned short *);
5865 void vec_st (vector signed char, int, signed char *);
5866 void vec_st (vector signed char, int, unsigned char *);
5867 void vec_st (vector signed char, int, vector signed char *);
5868 void vec_st (vector unsigned char, int, unsigned char *);
5869 void vec_st (vector unsigned char, int, vector unsigned char *);
5871 void vec_ste (vector signed char, int, unsigned char *);
5872 void vec_ste (vector signed char, int, signed char *);
5873 void vec_ste (vector unsigned char, int, unsigned char *);
5874 void vec_ste (vector signed short, int, short *);
5875 void vec_ste (vector signed short, int, unsigned short *);
5876 void vec_ste (vector unsigned short, int, void *);
5877 void vec_ste (vector signed int, int, unsigned int *);
5878 void vec_ste (vector signed int, int, int *);
5879 void vec_ste (vector unsigned int, int, unsigned int *);
5880 void vec_ste (vector float, int, float *);
5882 void vec_stl (vector float, int, vector float *);
5883 void vec_stl (vector float, int, float *);
5884 void vec_stl (vector signed int, int, vector signed int *);
5885 void vec_stl (vector signed int, int, int *);
5886 void vec_stl (vector signed int, int, unsigned int *);
5887 void vec_stl (vector unsigned int, int, vector unsigned int *);
5888 void vec_stl (vector unsigned int, int, unsigned int *);
5889 void vec_stl (vector signed short, int, short *);
5890 void vec_stl (vector signed short, int, unsigned short *);
5891 void vec_stl (vector signed short, int, vector signed short *);
5892 void vec_stl (vector unsigned short, int, unsigned short *);
5893 void vec_stl (vector unsigned short, int, vector signed short *);
5894 void vec_stl (vector signed char, int, signed char *);
5895 void vec_stl (vector signed char, int, unsigned char *);
5896 void vec_stl (vector signed char, int, vector signed char *);
5897 void vec_stl (vector unsigned char, int, unsigned char *);
5898 void vec_stl (vector unsigned char, int, vector unsigned char *);
5900 vector signed char vec_sub (vector signed char, vector signed char);
5901 vector unsigned char vec_sub (vector signed char, vector unsigned char);
5903 vector unsigned char vec_sub (vector unsigned char, vector signed char);
5905 vector unsigned char vec_sub (vector unsigned char,
5906 vector unsigned char);
5907 vector signed short vec_sub (vector signed short, vector signed short);
5908 vector unsigned short vec_sub (vector signed short,
5909 vector unsigned short);
5910 vector unsigned short vec_sub (vector unsigned short,
5911 vector signed short);
5912 vector unsigned short vec_sub (vector unsigned short,
5913 vector unsigned short);
5914 vector signed int vec_sub (vector signed int, vector signed int);
5915 vector unsigned int vec_sub (vector signed int, vector unsigned int);
5916 vector unsigned int vec_sub (vector unsigned int, vector signed int);
5917 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
5918 vector float vec_sub (vector float, vector float);
5920 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
5922 vector unsigned char vec_subs (vector signed char,
5923 vector unsigned char);
5924 vector unsigned char vec_subs (vector unsigned char,
5925 vector signed char);
5926 vector unsigned char vec_subs (vector unsigned char,
5927 vector unsigned char);
5928 vector signed char vec_subs (vector signed char, vector signed char);
5929 vector unsigned short vec_subs (vector signed short,
5930 vector unsigned short);
5931 vector unsigned short vec_subs (vector unsigned short,
5932 vector signed short);
5933 vector unsigned short vec_subs (vector unsigned short,
5934 vector unsigned short);
5935 vector signed short vec_subs (vector signed short, vector signed short);
5937 vector unsigned int vec_subs (vector signed int, vector unsigned int);
5938 vector unsigned int vec_subs (vector unsigned int, vector signed int);
5939 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
5941 vector signed int vec_subs (vector signed int, vector signed int);
5943 vector unsigned int vec_sum4s (vector unsigned char,
5944 vector unsigned int);
5945 vector signed int vec_sum4s (vector signed char, vector signed int);
5946 vector signed int vec_sum4s (vector signed short, vector signed int);
5948 vector signed int vec_sum2s (vector signed int, vector signed int);
5950 vector signed int vec_sums (vector signed int, vector signed int);
5952 vector float vec_trunc (vector float);
5954 vector signed short vec_unpackh (vector signed char);
5955 vector unsigned int vec_unpackh (vector signed short);
5956 vector signed int vec_unpackh (vector signed short);
5958 vector signed short vec_unpackl (vector signed char);
5959 vector unsigned int vec_unpackl (vector signed short);
5960 vector signed int vec_unpackl (vector signed short);
5962 vector float vec_xor (vector float, vector float);
5963 vector float vec_xor (vector float, vector signed int);
5964 vector float vec_xor (vector signed int, vector float);
5965 vector signed int vec_xor (vector signed int, vector signed int);
5966 vector unsigned int vec_xor (vector signed int, vector unsigned int);
5967 vector unsigned int vec_xor (vector unsigned int, vector signed int);
5968 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
5969 vector signed short vec_xor (vector signed short, vector signed short);
5970 vector unsigned short vec_xor (vector signed short,
5971 vector unsigned short);
5972 vector unsigned short vec_xor (vector unsigned short,
5973 vector signed short);
5974 vector unsigned short vec_xor (vector unsigned short,
5975 vector unsigned short);
5976 vector signed char vec_xor (vector signed char, vector signed char);
5977 vector unsigned char vec_xor (vector signed char, vector unsigned char);
5979 vector unsigned char vec_xor (vector unsigned char, vector signed char);
5981 vector unsigned char vec_xor (vector unsigned char,
5982 vector unsigned char);
5984 vector signed int vec_all_eq (vector signed char, vector unsigned char);
5986 vector signed int vec_all_eq (vector signed char, vector signed char);
5987 vector signed int vec_all_eq (vector unsigned char, vector signed char);
5989 vector signed int vec_all_eq (vector unsigned char,
5990 vector unsigned char);
5991 vector signed int vec_all_eq (vector signed short,
5992 vector unsigned short);
5993 vector signed int vec_all_eq (vector signed short, vector signed short);
5995 vector signed int vec_all_eq (vector unsigned short,
5996 vector signed short);
5997 vector signed int vec_all_eq (vector unsigned short,
5998 vector unsigned short);
5999 vector signed int vec_all_eq (vector signed int, vector unsigned int);
6000 vector signed int vec_all_eq (vector signed int, vector signed int);
6001 vector signed int vec_all_eq (vector unsigned int, vector signed int);
6002 vector signed int vec_all_eq (vector unsigned int, vector unsigned int);
6004 vector signed int vec_all_eq (vector float, vector float);
6006 vector signed int vec_all_ge (vector signed char, vector unsigned char);
6008 vector signed int vec_all_ge (vector unsigned char, vector signed char);
6010 vector signed int vec_all_ge (vector unsigned char,
6011 vector unsigned char);
6012 vector signed int vec_all_ge (vector signed char, vector signed char);
6013 vector signed int vec_all_ge (vector signed short,
6014 vector unsigned short);
6015 vector signed int vec_all_ge (vector unsigned short,
6016 vector signed short);
6017 vector signed int vec_all_ge (vector unsigned short,
6018 vector unsigned short);
6019 vector signed int vec_all_ge (vector signed short, vector signed short);
6021 vector signed int vec_all_ge (vector signed int, vector unsigned int);
6022 vector signed int vec_all_ge (vector unsigned int, vector signed int);
6023 vector signed int vec_all_ge (vector unsigned int, vector unsigned int);
6025 vector signed int vec_all_ge (vector signed int, vector signed int);
6026 vector signed int vec_all_ge (vector float, vector float);
6028 vector signed int vec_all_gt (vector signed char, vector unsigned char);
6030 vector signed int vec_all_gt (vector unsigned char, vector signed char);
6032 vector signed int vec_all_gt (vector unsigned char,
6033 vector unsigned char);
6034 vector signed int vec_all_gt (vector signed char, vector signed char);
6035 vector signed int vec_all_gt (vector signed short,
6036 vector unsigned short);
6037 vector signed int vec_all_gt (vector unsigned short,
6038 vector signed short);
6039 vector signed int vec_all_gt (vector unsigned short,
6040 vector unsigned short);
6041 vector signed int vec_all_gt (vector signed short, vector signed short);
6043 vector signed int vec_all_gt (vector signed int, vector unsigned int);
6044 vector signed int vec_all_gt (vector unsigned int, vector signed int);
6045 vector signed int vec_all_gt (vector unsigned int, vector unsigned int);
6047 vector signed int vec_all_gt (vector signed int, vector signed int);
6048 vector signed int vec_all_gt (vector float, vector float);
6050 vector signed int vec_all_in (vector float, vector float);
6052 vector signed int vec_all_le (vector signed char, vector unsigned char);
6054 vector signed int vec_all_le (vector unsigned char, vector signed char);
6056 vector signed int vec_all_le (vector unsigned char,
6057 vector unsigned char);
6058 vector signed int vec_all_le (vector signed char, vector signed char);
6059 vector signed int vec_all_le (vector signed short,
6060 vector unsigned short);
6061 vector signed int vec_all_le (vector unsigned short,
6062 vector signed short);
6063 vector signed int vec_all_le (vector unsigned short,
6064 vector unsigned short);
6065 vector signed int vec_all_le (vector signed short, vector signed short);
6067 vector signed int vec_all_le (vector signed int, vector unsigned int);
6068 vector signed int vec_all_le (vector unsigned int, vector signed int);
6069 vector signed int vec_all_le (vector unsigned int, vector unsigned int);
6071 vector signed int vec_all_le (vector signed int, vector signed int);
6072 vector signed int vec_all_le (vector float, vector float);
6074 vector signed int vec_all_lt (vector signed char, vector unsigned char);
6076 vector signed int vec_all_lt (vector unsigned char, vector signed char);
6078 vector signed int vec_all_lt (vector unsigned char,
6079 vector unsigned char);
6080 vector signed int vec_all_lt (vector signed char, vector signed char);
6081 vector signed int vec_all_lt (vector signed short,
6082 vector unsigned short);
6083 vector signed int vec_all_lt (vector unsigned short,
6084 vector signed short);
6085 vector signed int vec_all_lt (vector unsigned short,
6086 vector unsigned short);
6087 vector signed int vec_all_lt (vector signed short, vector signed short);
6089 vector signed int vec_all_lt (vector signed int, vector unsigned int);
6090 vector signed int vec_all_lt (vector unsigned int, vector signed int);
6091 vector signed int vec_all_lt (vector unsigned int, vector unsigned int);
6093 vector signed int vec_all_lt (vector signed int, vector signed int);
6094 vector signed int vec_all_lt (vector float, vector float);
6096 vector signed int vec_all_nan (vector float);
6098 vector signed int vec_all_ne (vector signed char, vector unsigned char);
6100 vector signed int vec_all_ne (vector signed char, vector signed char);
6101 vector signed int vec_all_ne (vector unsigned char, vector signed char);
6103 vector signed int vec_all_ne (vector unsigned char,
6104 vector unsigned char);
6105 vector signed int vec_all_ne (vector signed short,
6106 vector unsigned short);
6107 vector signed int vec_all_ne (vector signed short, vector signed short);
6109 vector signed int vec_all_ne (vector unsigned short,
6110 vector signed short);
6111 vector signed int vec_all_ne (vector unsigned short,
6112 vector unsigned short);
6113 vector signed int vec_all_ne (vector signed int, vector unsigned int);
6114 vector signed int vec_all_ne (vector signed int, vector signed int);
6115 vector signed int vec_all_ne (vector unsigned int, vector signed int);
6116 vector signed int vec_all_ne (vector unsigned int, vector unsigned int);
6118 vector signed int vec_all_ne (vector float, vector float);
6120 vector signed int vec_all_nge (vector float, vector float);
6122 vector signed int vec_all_ngt (vector float, vector float);
6124 vector signed int vec_all_nle (vector float, vector float);
6126 vector signed int vec_all_nlt (vector float, vector float);
6128 vector signed int vec_all_numeric (vector float);
6130 vector signed int vec_any_eq (vector signed char, vector unsigned char);
6132 vector signed int vec_any_eq (vector signed char, vector signed char);
6133 vector signed int vec_any_eq (vector unsigned char, vector signed char);
6135 vector signed int vec_any_eq (vector unsigned char,
6136 vector unsigned char);
6137 vector signed int vec_any_eq (vector signed short,
6138 vector unsigned short);
6139 vector signed int vec_any_eq (vector signed short, vector signed short);
6141 vector signed int vec_any_eq (vector unsigned short,
6142 vector signed short);
6143 vector signed int vec_any_eq (vector unsigned short,
6144 vector unsigned short);
6145 vector signed int vec_any_eq (vector signed int, vector unsigned int);
6146 vector signed int vec_any_eq (vector signed int, vector signed int);
6147 vector signed int vec_any_eq (vector unsigned int, vector signed int);
6148 vector signed int vec_any_eq (vector unsigned int, vector unsigned int);
6150 vector signed int vec_any_eq (vector float, vector float);
6152 vector signed int vec_any_ge (vector signed char, vector unsigned char);
6154 vector signed int vec_any_ge (vector unsigned char, vector signed char);
6156 vector signed int vec_any_ge (vector unsigned char,
6157 vector unsigned char);
6158 vector signed int vec_any_ge (vector signed char, vector signed char);
6159 vector signed int vec_any_ge (vector signed short,
6160 vector unsigned short);
6161 vector signed int vec_any_ge (vector unsigned short,
6162 vector signed short);
6163 vector signed int vec_any_ge (vector unsigned short,
6164 vector unsigned short);
6165 vector signed int vec_any_ge (vector signed short, vector signed short);
6167 vector signed int vec_any_ge (vector signed int, vector unsigned int);
6168 vector signed int vec_any_ge (vector unsigned int, vector signed int);
6169 vector signed int vec_any_ge (vector unsigned int, vector unsigned int);
6171 vector signed int vec_any_ge (vector signed int, vector signed int);
6172 vector signed int vec_any_ge (vector float, vector float);
6174 vector signed int vec_any_gt (vector signed char, vector unsigned char);
6176 vector signed int vec_any_gt (vector unsigned char, vector signed char);
6178 vector signed int vec_any_gt (vector unsigned char,
6179 vector unsigned char);
6180 vector signed int vec_any_gt (vector signed char, vector signed char);
6181 vector signed int vec_any_gt (vector signed short,
6182 vector unsigned short);
6183 vector signed int vec_any_gt (vector unsigned short,
6184 vector signed short);
6185 vector signed int vec_any_gt (vector unsigned short,
6186 vector unsigned short);
6187 vector signed int vec_any_gt (vector signed short, vector signed short);
6189 vector signed int vec_any_gt (vector signed int, vector unsigned int);
6190 vector signed int vec_any_gt (vector unsigned int, vector signed int);
6191 vector signed int vec_any_gt (vector unsigned int, vector unsigned int);
6193 vector signed int vec_any_gt (vector signed int, vector signed int);
6194 vector signed int vec_any_gt (vector float, vector float);
6196 vector signed int vec_any_le (vector signed char, vector unsigned char);
6198 vector signed int vec_any_le (vector unsigned char, vector signed char);
6200 vector signed int vec_any_le (vector unsigned char,
6201 vector unsigned char);
6202 vector signed int vec_any_le (vector signed char, vector signed char);
6203 vector signed int vec_any_le (vector signed short,
6204 vector unsigned short);
6205 vector signed int vec_any_le (vector unsigned short,
6206 vector signed short);
6207 vector signed int vec_any_le (vector unsigned short,
6208 vector unsigned short);
6209 vector signed int vec_any_le (vector signed short, vector signed short);
6211 vector signed int vec_any_le (vector signed int, vector unsigned int);
6212 vector signed int vec_any_le (vector unsigned int, vector signed int);
6213 vector signed int vec_any_le (vector unsigned int, vector unsigned int);
6215 vector signed int vec_any_le (vector signed int, vector signed int);
6216 vector signed int vec_any_le (vector float, vector float);
6218 vector signed int vec_any_lt (vector signed char, vector unsigned char);
6220 vector signed int vec_any_lt (vector unsigned char, vector signed char);
6222 vector signed int vec_any_lt (vector unsigned char,
6223 vector unsigned char);
6224 vector signed int vec_any_lt (vector signed char, vector signed char);
6225 vector signed int vec_any_lt (vector signed short,
6226 vector unsigned short);
6227 vector signed int vec_any_lt (vector unsigned short,
6228 vector signed short);
6229 vector signed int vec_any_lt (vector unsigned short,
6230 vector unsigned short);
6231 vector signed int vec_any_lt (vector signed short, vector signed short);
6233 vector signed int vec_any_lt (vector signed int, vector unsigned int);
6234 vector signed int vec_any_lt (vector unsigned int, vector signed int);
6235 vector signed int vec_any_lt (vector unsigned int, vector unsigned int);
6237 vector signed int vec_any_lt (vector signed int, vector signed int);
6238 vector signed int vec_any_lt (vector float, vector float);
6240 vector signed int vec_any_nan (vector float);
6242 vector signed int vec_any_ne (vector signed char, vector unsigned char);
6244 vector signed int vec_any_ne (vector signed char, vector signed char);
6245 vector signed int vec_any_ne (vector unsigned char, vector signed char);
6247 vector signed int vec_any_ne (vector unsigned char,
6248 vector unsigned char);
6249 vector signed int vec_any_ne (vector signed short,
6250 vector unsigned short);
6251 vector signed int vec_any_ne (vector signed short, vector signed short);
6253 vector signed int vec_any_ne (vector unsigned short,
6254 vector signed short);
6255 vector signed int vec_any_ne (vector unsigned short,
6256 vector unsigned short);
6257 vector signed int vec_any_ne (vector signed int, vector unsigned int);
6258 vector signed int vec_any_ne (vector signed int, vector signed int);
6259 vector signed int vec_any_ne (vector unsigned int, vector signed int);
6260 vector signed int vec_any_ne (vector unsigned int, vector unsigned int);
6262 vector signed int vec_any_ne (vector float, vector float);
6264 vector signed int vec_any_nge (vector float, vector float);
6266 vector signed int vec_any_ngt (vector float, vector float);
6268 vector signed int vec_any_nle (vector float, vector float);
6270 vector signed int vec_any_nlt (vector float, vector float);
6272 vector signed int vec_any_numeric (vector float);
6274 vector signed int vec_any_out (vector float, vector float);
6278 @section Pragmas Accepted by GCC
6282 GCC supports several types of pragmas, primarily in order to compile
6283 code originally written for other compilers. Note that in general
6284 we do not recommend the use of pragmas; @xref{Function Attributes},
6285 for further explanation.
6289 * RS/6000 and PowerPC Pragmas::
6296 @subsection ARM Pragmas
6298 The ARM target defines pragmas for controlling the default addition of
6299 @code{long_call} and @code{short_call} attributes to functions.
6300 @xref{Function Attributes}, for information about the effects of these
6305 @cindex pragma, long_calls
6306 Set all subsequent functions to have the @code{long_call} attribute.
6309 @cindex pragma, no_long_calls
6310 Set all subsequent functions to have the @code{short_call} attribute.
6312 @item long_calls_off
6313 @cindex pragma, long_calls_off
6314 Do not affect the @code{long_call} or @code{short_call} attributes of
6315 subsequent functions.
6318 @node RS/6000 and PowerPC Pragmas
6319 @subsection RS/6000 and PowerPC Pragmas
6321 The RS/6000 and PowerPC targets define one pragma for controlling
6322 whether or not the @code{longcall} attribute is added to function
6323 declarations by default. This pragma overrides the @option{-mlongcall}
6324 option, but not the @code{longcall} and @code{shortcall} attributes.
6325 @xref{RS/6000 and PowerPC Options}, for more information about when long
6326 calls are and are not necessary.
6330 @cindex pragma, longcall
6331 Apply the @code{longcall} attribute to all subsequent function
6335 Do not apply the @code{longcall} attribute to subsequent function
6339 @c Describe c4x pragmas here.
6340 @c Describe h8300 pragmas here.
6341 @c Describe i370 pragmas here.
6342 @c Describe i960 pragmas here.
6343 @c Describe sh pragmas here.
6344 @c Describe v850 pragmas here.
6346 @node Darwin Pragmas
6347 @subsection Darwin Pragmas
6349 The following pragmas are available for all architectures running the
6350 Darwin operating system. These are useful for compatibility with other
6354 @item mark @var{tokens}@dots{}
6355 @cindex pragma, mark
6356 This pragma is accepted, but has no effect.
6358 @item options align=@var{alignment}
6359 @cindex pragma, options align
6360 This pragma sets the alignment of fields in structures. The values of
6361 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
6362 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
6363 properly; to restore the previous setting, use @code{reset} for the
6366 @item segment @var{tokens}@dots{}
6367 @cindex pragma, segment
6368 This pragma is accepted, but has no effect.
6370 @item unused (@var{var} [, @var{var}]@dots{})
6371 @cindex pragma, unused
6372 This pragma declares variables to be possibly unused. GCC will not
6373 produce warnings for the listed variables. The effect is similar to
6374 that of the @code{unused} attribute, except that this pragma may appear
6375 anywhere within the variables' scopes.
6378 @node Solaris Pragmas
6379 @subsection Solaris Pragmas
6381 For compatibility with the SunPRO compiler, the following pragma
6385 @item redefine_extname @var{oldname} @var{newname}
6386 @cindex pragma, redefine_extname
6388 This pragma gives the C function @var{oldname} the assembler label
6389 @var{newname}. The pragma must appear before the function declaration.
6390 This pragma is equivalent to the asm labels extension (@pxref{Asm
6391 Labels}). The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
6392 if the pragma is available.
6396 @subsection Tru64 Pragmas
6398 For compatibility with the Compaq C compiler, the following pragma
6402 @item extern_prefix @var{string}
6403 @cindex pragma, extern_prefix
6405 This pragma renames all subsequent function and variable declarations
6406 such that @var{string} is prepended to the name. This effect may be
6407 terminated by using another @code{extern_prefix} pragma with the
6410 This pragma is similar in intent to to the asm labels extension
6411 (@pxref{Asm Labels}) in that the system programmer wants to change
6412 the assembly-level ABI without changing the source-level API. The
6413 preprocessor defines @code{__EXTERN_PREFIX} if the pragma is available.
6416 @node Unnamed Fields
6417 @section Unnamed struct/union fields within structs/unions.
6421 For compatibility with other compilers, GCC allows you to define
6422 a structure or union that contains, as fields, structures and unions
6423 without names. For example:
6436 In this example, the user would be able to access members of the unnamed
6437 union with code like @samp{foo.b}. Note that only unnamed structs and
6438 unions are allowed, you may not have, for example, an unnamed
6441 You must never create such structures that cause ambiguous field definitions.
6442 For example, this structure:
6453 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
6454 Such constructs are not supported and must be avoided. In the future,
6455 such constructs may be detected and treated as compilation errors.
6458 @section Thread-Local Storage
6459 @cindex Thread-Local Storage
6460 @cindex @acronym{TLS}
6463 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
6464 are allocated such that there is one instance of the variable per extant
6465 thread. The run-time model GCC uses to implement this originates
6466 in the IA-64 processor-specific ABI, but has since been migrated
6467 to other processors as well. It requires significant support from
6468 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
6469 system libraries (@file{libc.so} and @file{libpthread.so}), so it
6470 is not available everywhere.
6472 At the user level, the extension is visible with a new storage
6473 class keyword: @code{__thread}. For example:
6477 extern __thread struct state s;
6478 static __thread char *p;
6481 The @code{__thread} specifier may be used alone, with the @code{extern}
6482 or @code{static} specifiers, but with no other storage class specifier.
6483 When used with @code{extern} or @code{static}, @code{__thread} must appear
6484 immediately after the other storage class specifier.
6486 The @code{__thread} specifier may be applied to any global, file-scoped
6487 static, function-scoped static, or static data member of a class. It may
6488 not be applied to block-scoped automatic or non-static data member.
6490 When the address-of operator is applied to a thread-local variable, it is
6491 evaluated at run-time and returns the address of the current thread's
6492 instance of that variable. An address so obtained may be used by any
6493 thread. When a thread terminates, any pointers to thread-local variables
6494 in that thread become invalid.
6496 No static initialization may refer to the address of a thread-local variable.
6498 In C++, if an initializer is present for a thread-local variable, it must
6499 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
6502 See @uref{http://people.redhat.com/drepper/tls.pdf,
6503 ELF Handling For Thread-Local Storage} for a detailed explanation of
6504 the four thread-local storage addressing models, and how the run-time
6505 is expected to function.
6508 * C99 Thread-Local Edits::
6509 * C++98 Thread-Local Edits::
6512 @node C99 Thread-Local Edits
6513 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
6515 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
6516 that document the exact semantics of the language extension.
6520 @cite{5.1.2 Execution environments}
6522 Add new text after paragraph 1
6525 Within either execution environment, a @dfn{thread} is a flow of
6526 control within a program. It is implementation defined whether
6527 or not there may be more than one thread associated with a program.
6528 It is implementation defined how threads beyond the first are
6529 created, the name and type of the function called at thread
6530 startup, and how threads may be terminated. However, objects
6531 with thread storage duration shall be initialized before thread
6536 @cite{6.2.4 Storage durations of objects}
6538 Add new text before paragraph 3
6541 An object whose identifier is declared with the storage-class
6542 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
6543 Its lifetime is the entire execution of the thread, and its
6544 stored value is initialized only once, prior to thread startup.
6548 @cite{6.4.1 Keywords}
6550 Add @code{__thread}.
6553 @cite{6.7.1 Storage-class specifiers}
6555 Add @code{__thread} to the list of storage class specifiers in
6558 Change paragraph 2 to
6561 With the exception of @code{__thread}, at most one storage-class
6562 specifier may be given [@dots{}]. The @code{__thread} specifier may
6563 be used alone, or immediately following @code{extern} or
6567 Add new text after paragraph 6
6570 The declaration of an identifier for a variable that has
6571 block scope that specifies @code{__thread} shall also
6572 specify either @code{extern} or @code{static}.
6574 The @code{__thread} specifier shall be used only with
6579 @node C++98 Thread-Local Edits
6580 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
6582 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
6583 that document the exact semantics of the language extension.
6586 @b{[intro.execution]}
6588 New text after paragraph 4
6591 A @dfn{thread} is a flow of control within the abstract machine.
6592 It is implementation defined whether or not there may be more than
6596 New text after paragraph 7
6599 It is unspecified whether additional action must be taken to
6600 ensure when and whether side effects are visible to other threads.
6606 Add @code{__thread}.
6609 @b{[basic.start.main]}
6611 Add after paragraph 5
6614 The thread that begins execution at the @code{main} function is called
6615 the @dfn{main thread}. It is implementation defined how functions
6616 beginning threads other than the main thread are designated or typed.
6617 A function so designated, as well as the @code{main} function, is called
6618 a @dfn{thread startup function}. It is implementation defined what
6619 happens if a thread startup function returns. It is implementation
6620 defined what happens to other threads when any thread calls @code{exit}.
6624 @b{[basic.start.init]}
6626 Add after paragraph 4
6629 The storage for an object of thread storage duration shall be
6630 staticly initialized before the first statement of the thread startup
6631 function. An object of thread storage duration shall not require
6632 dynamic initialization.
6636 @b{[basic.start.term]}
6638 Add after paragraph 3
6641 The type of an object with thread storage duration shall not have a
6642 non-trivial destructor, nor shall it be an array type whose elements
6643 (directly or indirectly) have non-trivial destructors.
6649 Add ``thread storage duration'' to the list in paragraph 1.
6654 Thread, static, and automatic storage durations are associated with
6655 objects introduced by declarations [@dots{}].
6658 Add @code{__thread} to the list of specifiers in paragraph 3.
6661 @b{[basic.stc.thread]}
6663 New section before @b{[basic.stc.static]}
6666 The keyword @code{__thread} applied to an non-local object gives the
6667 object thread storage duration.
6669 A local variable or class data member declared both @code{static}
6670 and @code{__thread} gives the variable or member thread storage
6675 @b{[basic.stc.static]}
6680 All objects which have neither thread storage duration, dynamic
6681 storage duration nor are local [@dots{}].
6687 Add @code{__thread} to the list in paragraph 1.
6692 With the exception of @code{__thread}, at most one
6693 @var{storage-class-specifier} shall appear in a given
6694 @var{decl-specifier-seq}. The @code{__thread} specifier may
6695 be used alone, or immediately following the @code{extern} or
6696 @code{static} specifiers. [@dots{}]
6699 Add after paragraph 5
6702 The @code{__thread} specifier can be applied only to the names of objects
6703 and to anonymous unions.
6709 Add after paragraph 6
6712 Non-@code{static} members shall not be @code{__thread}.
6716 @node C++ Extensions
6717 @chapter Extensions to the C++ Language
6718 @cindex extensions, C++ language
6719 @cindex C++ language extensions
6721 The GNU compiler provides these extensions to the C++ language (and you
6722 can also use most of the C language extensions in your C++ programs). If you
6723 want to write code that checks whether these features are available, you can
6724 test for the GNU compiler the same way as for C programs: check for a
6725 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
6726 test specifically for GNU C++ (@pxref{Standard Predefined,,Standard
6727 Predefined Macros,cpp.info,The C Preprocessor}).
6730 * Min and Max:: C++ Minimum and maximum operators.
6731 * Volatiles:: What constitutes an access to a volatile object.
6732 * Restricted Pointers:: C99 restricted pointers and references.
6733 * Vague Linkage:: Where G++ puts inlines, vtables and such.
6734 * C++ Interface:: You can use a single C++ header file for both
6735 declarations and definitions.
6736 * Template Instantiation:: Methods for ensuring that exactly one copy of
6737 each needed template instantiation is emitted.
6738 * Bound member functions:: You can extract a function pointer to the
6739 method denoted by a @samp{->*} or @samp{.*} expression.
6740 * C++ Attributes:: Variable, function, and type attributes for C++ only.
6741 * Java Exceptions:: Tweaking exception handling to work with Java.
6742 * Deprecated Features:: Things might disappear from g++.
6743 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
6747 @section Minimum and Maximum Operators in C++
6749 It is very convenient to have operators which return the ``minimum'' or the
6750 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
6753 @item @var{a} <? @var{b}
6755 @cindex minimum operator
6756 is the @dfn{minimum}, returning the smaller of the numeric values
6757 @var{a} and @var{b};
6759 @item @var{a} >? @var{b}
6761 @cindex maximum operator
6762 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
6766 These operations are not primitive in ordinary C++, since you can
6767 use a macro to return the minimum of two things in C++, as in the
6771 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
6775 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
6776 the minimum value of variables @var{i} and @var{j}.
6778 However, side effects in @code{X} or @code{Y} may cause unintended
6779 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
6780 the smaller counter twice. A GNU C extension allows you to write safe
6781 macros that avoid this kind of problem (@pxref{Naming Types,,Naming an
6782 Expression's Type}). However, writing @code{MIN} and @code{MAX} as
6783 macros also forces you to use function-call notation for a
6784 fundamental arithmetic operation. Using GNU C++ extensions, you can
6785 write @w{@samp{int min = i <? j;}} instead.
6787 Since @code{<?} and @code{>?} are built into the compiler, they properly
6788 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
6792 @section When is a Volatile Object Accessed?
6793 @cindex accessing volatiles
6794 @cindex volatile read
6795 @cindex volatile write
6796 @cindex volatile access
6798 Both the C and C++ standard have the concept of volatile objects. These
6799 are normally accessed by pointers and used for accessing hardware. The
6800 standards encourage compilers to refrain from optimizations
6801 concerning accesses to volatile objects that it might perform on
6802 non-volatile objects. The C standard leaves it implementation defined
6803 as to what constitutes a volatile access. The C++ standard omits to
6804 specify this, except to say that C++ should behave in a similar manner
6805 to C with respect to volatiles, where possible. The minimum either
6806 standard specifies is that at a sequence point all previous accesses to
6807 volatile objects have stabilized and no subsequent accesses have
6808 occurred. Thus an implementation is free to reorder and combine
6809 volatile accesses which occur between sequence points, but cannot do so
6810 for accesses across a sequence point. The use of volatiles does not
6811 allow you to violate the restriction on updating objects multiple times
6812 within a sequence point.
6814 In most expressions, it is intuitively obvious what is a read and what is
6815 a write. For instance
6818 volatile int *dst = @var{somevalue};
6819 volatile int *src = @var{someothervalue};
6824 will cause a read of the volatile object pointed to by @var{src} and stores the
6825 value into the volatile object pointed to by @var{dst}. There is no
6826 guarantee that these reads and writes are atomic, especially for objects
6827 larger than @code{int}.
6829 Less obvious expressions are where something which looks like an access
6830 is used in a void context. An example would be,
6833 volatile int *src = @var{somevalue};
6837 With C, such expressions are rvalues, and as rvalues cause a read of
6838 the object, GCC interprets this as a read of the volatile being pointed
6839 to. The C++ standard specifies that such expressions do not undergo
6840 lvalue to rvalue conversion, and that the type of the dereferenced
6841 object may be incomplete. The C++ standard does not specify explicitly
6842 that it is this lvalue to rvalue conversion which is responsible for
6843 causing an access. However, there is reason to believe that it is,
6844 because otherwise certain simple expressions become undefined. However,
6845 because it would surprise most programmers, G++ treats dereferencing a
6846 pointer to volatile object of complete type in a void context as a read
6847 of the object. When the object has incomplete type, G++ issues a
6852 struct T @{int m;@};
6853 volatile S *ptr1 = @var{somevalue};
6854 volatile T *ptr2 = @var{somevalue};
6859 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
6860 causes a read of the object pointed to. If you wish to force an error on
6861 the first case, you must force a conversion to rvalue with, for instance
6862 a static cast, @code{static_cast<S>(*ptr1)}.
6864 When using a reference to volatile, G++ does not treat equivalent
6865 expressions as accesses to volatiles, but instead issues a warning that
6866 no volatile is accessed. The rationale for this is that otherwise it
6867 becomes difficult to determine where volatile access occur, and not
6868 possible to ignore the return value from functions returning volatile
6869 references. Again, if you wish to force a read, cast the reference to
6872 @node Restricted Pointers
6873 @section Restricting Pointer Aliasing
6874 @cindex restricted pointers
6875 @cindex restricted references
6876 @cindex restricted this pointer
6878 As with gcc, g++ understands the C99 feature of restricted pointers,
6879 specified with the @code{__restrict__}, or @code{__restrict} type
6880 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
6881 language flag, @code{restrict} is not a keyword in C++.
6883 In addition to allowing restricted pointers, you can specify restricted
6884 references, which indicate that the reference is not aliased in the local
6888 void fn (int *__restrict__ rptr, int &__restrict__ rref)
6895 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
6896 @var{rref} refers to a (different) unaliased integer.
6898 You may also specify whether a member function's @var{this} pointer is
6899 unaliased by using @code{__restrict__} as a member function qualifier.
6902 void T::fn () __restrict__
6909 Within the body of @code{T::fn}, @var{this} will have the effective
6910 definition @code{T *__restrict__ const this}. Notice that the
6911 interpretation of a @code{__restrict__} member function qualifier is
6912 different to that of @code{const} or @code{volatile} qualifier, in that it
6913 is applied to the pointer rather than the object. This is consistent with
6914 other compilers which implement restricted pointers.
6916 As with all outermost parameter qualifiers, @code{__restrict__} is
6917 ignored in function definition matching. This means you only need to
6918 specify @code{__restrict__} in a function definition, rather than
6919 in a function prototype as well.
6922 @section Vague Linkage
6923 @cindex vague linkage
6925 There are several constructs in C++ which require space in the object
6926 file but are not clearly tied to a single translation unit. We say that
6927 these constructs have ``vague linkage''. Typically such constructs are
6928 emitted wherever they are needed, though sometimes we can be more
6932 @item Inline Functions
6933 Inline functions are typically defined in a header file which can be
6934 included in many different compilations. Hopefully they can usually be
6935 inlined, but sometimes an out-of-line copy is necessary, if the address
6936 of the function is taken or if inlining fails. In general, we emit an
6937 out-of-line copy in all translation units where one is needed. As an
6938 exception, we only emit inline virtual functions with the vtable, since
6939 it will always require a copy.
6941 Local static variables and string constants used in an inline function
6942 are also considered to have vague linkage, since they must be shared
6943 between all inlined and out-of-line instances of the function.
6947 C++ virtual functions are implemented in most compilers using a lookup
6948 table, known as a vtable. The vtable contains pointers to the virtual
6949 functions provided by a class, and each object of the class contains a
6950 pointer to its vtable (or vtables, in some multiple-inheritance
6951 situations). If the class declares any non-inline, non-pure virtual
6952 functions, the first one is chosen as the ``key method'' for the class,
6953 and the vtable is only emitted in the translation unit where the key
6956 @emph{Note:} If the chosen key method is later defined as inline, the
6957 vtable will still be emitted in every translation unit which defines it.
6958 Make sure that any inline virtuals are declared inline in the class
6959 body, even if they are not defined there.
6961 @item type_info objects
6964 C++ requires information about types to be written out in order to
6965 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
6966 For polymorphic classes (classes with virtual functions), the type_info
6967 object is written out along with the vtable so that @samp{dynamic_cast}
6968 can determine the dynamic type of a class object at runtime. For all
6969 other types, we write out the type_info object when it is used: when
6970 applying @samp{typeid} to an expression, throwing an object, or
6971 referring to a type in a catch clause or exception specification.
6973 @item Template Instantiations
6974 Most everything in this section also applies to template instantiations,
6975 but there are other options as well.
6976 @xref{Template Instantiation,,Where's the Template?}.
6980 When used with GNU ld version 2.8 or later on an ELF system such as
6981 Linux/GNU or Solaris 2, or on Microsoft Windows, duplicate copies of
6982 these constructs will be discarded at link time. This is known as
6985 On targets that don't support COMDAT, but do support weak symbols, GCC
6986 will use them. This way one copy will override all the others, but
6987 the unused copies will still take up space in the executable.
6989 For targets which do not support either COMDAT or weak symbols,
6990 most entities with vague linkage will be emitted as local symbols to
6991 avoid duplicate definition errors from the linker. This will not happen
6992 for local statics in inlines, however, as having multiple copies will
6993 almost certainly break things.
6995 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
6996 another way to control placement of these constructs.
6999 @section Declarations and Definitions in One Header
7001 @cindex interface and implementation headers, C++
7002 @cindex C++ interface and implementation headers
7003 C++ object definitions can be quite complex. In principle, your source
7004 code will need two kinds of things for each object that you use across
7005 more than one source file. First, you need an @dfn{interface}
7006 specification, describing its structure with type declarations and
7007 function prototypes. Second, you need the @dfn{implementation} itself.
7008 It can be tedious to maintain a separate interface description in a
7009 header file, in parallel to the actual implementation. It is also
7010 dangerous, since separate interface and implementation definitions may
7011 not remain parallel.
7013 @cindex pragmas, interface and implementation
7014 With GNU C++, you can use a single header file for both purposes.
7017 @emph{Warning:} The mechanism to specify this is in transition. For the
7018 nonce, you must use one of two @code{#pragma} commands; in a future
7019 release of GNU C++, an alternative mechanism will make these
7020 @code{#pragma} commands unnecessary.
7023 The header file contains the full definitions, but is marked with
7024 @samp{#pragma interface} in the source code. This allows the compiler
7025 to use the header file only as an interface specification when ordinary
7026 source files incorporate it with @code{#include}. In the single source
7027 file where the full implementation belongs, you can use either a naming
7028 convention or @samp{#pragma implementation} to indicate this alternate
7029 use of the header file.
7032 @item #pragma interface
7033 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
7034 @kindex #pragma interface
7035 Use this directive in @emph{header files} that define object classes, to save
7036 space in most of the object files that use those classes. Normally,
7037 local copies of certain information (backup copies of inline member
7038 functions, debugging information, and the internal tables that implement
7039 virtual functions) must be kept in each object file that includes class
7040 definitions. You can use this pragma to avoid such duplication. When a
7041 header file containing @samp{#pragma interface} is included in a
7042 compilation, this auxiliary information will not be generated (unless
7043 the main input source file itself uses @samp{#pragma implementation}).
7044 Instead, the object files will contain references to be resolved at link
7047 The second form of this directive is useful for the case where you have
7048 multiple headers with the same name in different directories. If you
7049 use this form, you must specify the same string to @samp{#pragma
7052 @item #pragma implementation
7053 @itemx #pragma implementation "@var{objects}.h"
7054 @kindex #pragma implementation
7055 Use this pragma in a @emph{main input file}, when you want full output from
7056 included header files to be generated (and made globally visible). The
7057 included header file, in turn, should use @samp{#pragma interface}.
7058 Backup copies of inline member functions, debugging information, and the
7059 internal tables used to implement virtual functions are all generated in
7060 implementation files.
7062 @cindex implied @code{#pragma implementation}
7063 @cindex @code{#pragma implementation}, implied
7064 @cindex naming convention, implementation headers
7065 If you use @samp{#pragma implementation} with no argument, it applies to
7066 an include file with the same basename@footnote{A file's @dfn{basename}
7067 was the name stripped of all leading path information and of trailing
7068 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
7069 file. For example, in @file{allclass.cc}, giving just
7070 @samp{#pragma implementation}
7071 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
7073 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
7074 an implementation file whenever you would include it from
7075 @file{allclass.cc} even if you never specified @samp{#pragma
7076 implementation}. This was deemed to be more trouble than it was worth,
7077 however, and disabled.
7079 If you use an explicit @samp{#pragma implementation}, it must appear in
7080 your source file @emph{before} you include the affected header files.
7082 Use the string argument if you want a single implementation file to
7083 include code from multiple header files. (You must also use
7084 @samp{#include} to include the header file; @samp{#pragma
7085 implementation} only specifies how to use the file---it doesn't actually
7088 There is no way to split up the contents of a single header file into
7089 multiple implementation files.
7092 @cindex inlining and C++ pragmas
7093 @cindex C++ pragmas, effect on inlining
7094 @cindex pragmas in C++, effect on inlining
7095 @samp{#pragma implementation} and @samp{#pragma interface} also have an
7096 effect on function inlining.
7098 If you define a class in a header file marked with @samp{#pragma
7099 interface}, the effect on a function defined in that class is similar to
7100 an explicit @code{extern} declaration---the compiler emits no code at
7101 all to define an independent version of the function. Its definition
7102 is used only for inlining with its callers.
7104 @opindex fno-implement-inlines
7105 Conversely, when you include the same header file in a main source file
7106 that declares it as @samp{#pragma implementation}, the compiler emits
7107 code for the function itself; this defines a version of the function
7108 that can be found via pointers (or by callers compiled without
7109 inlining). If all calls to the function can be inlined, you can avoid
7110 emitting the function by compiling with @option{-fno-implement-inlines}.
7111 If any calls were not inlined, you will get linker errors.
7113 @node Template Instantiation
7114 @section Where's the Template?
7116 @cindex template instantiation
7118 C++ templates are the first language feature to require more
7119 intelligence from the environment than one usually finds on a UNIX
7120 system. Somehow the compiler and linker have to make sure that each
7121 template instance occurs exactly once in the executable if it is needed,
7122 and not at all otherwise. There are two basic approaches to this
7123 problem, which I will refer to as the Borland model and the Cfront model.
7127 Borland C++ solved the template instantiation problem by adding the code
7128 equivalent of common blocks to their linker; the compiler emits template
7129 instances in each translation unit that uses them, and the linker
7130 collapses them together. The advantage of this model is that the linker
7131 only has to consider the object files themselves; there is no external
7132 complexity to worry about. This disadvantage is that compilation time
7133 is increased because the template code is being compiled repeatedly.
7134 Code written for this model tends to include definitions of all
7135 templates in the header file, since they must be seen to be
7139 The AT&T C++ translator, Cfront, solved the template instantiation
7140 problem by creating the notion of a template repository, an
7141 automatically maintained place where template instances are stored. A
7142 more modern version of the repository works as follows: As individual
7143 object files are built, the compiler places any template definitions and
7144 instantiations encountered in the repository. At link time, the link
7145 wrapper adds in the objects in the repository and compiles any needed
7146 instances that were not previously emitted. The advantages of this
7147 model are more optimal compilation speed and the ability to use the
7148 system linker; to implement the Borland model a compiler vendor also
7149 needs to replace the linker. The disadvantages are vastly increased
7150 complexity, and thus potential for error; for some code this can be
7151 just as transparent, but in practice it can been very difficult to build
7152 multiple programs in one directory and one program in multiple
7153 directories. Code written for this model tends to separate definitions
7154 of non-inline member templates into a separate file, which should be
7155 compiled separately.
7158 When used with GNU ld version 2.8 or later on an ELF system such as
7159 Linux/GNU or Solaris 2, or on Microsoft Windows, g++ supports the
7160 Borland model. On other systems, g++ implements neither automatic
7163 A future version of g++ will support a hybrid model whereby the compiler
7164 will emit any instantiations for which the template definition is
7165 included in the compile, and store template definitions and
7166 instantiation context information into the object file for the rest.
7167 The link wrapper will extract that information as necessary and invoke
7168 the compiler to produce the remaining instantiations. The linker will
7169 then combine duplicate instantiations.
7171 In the mean time, you have the following options for dealing with
7172 template instantiations:
7177 Compile your template-using code with @option{-frepo}. The compiler will
7178 generate files with the extension @samp{.rpo} listing all of the
7179 template instantiations used in the corresponding object files which
7180 could be instantiated there; the link wrapper, @samp{collect2}, will
7181 then update the @samp{.rpo} files to tell the compiler where to place
7182 those instantiations and rebuild any affected object files. The
7183 link-time overhead is negligible after the first pass, as the compiler
7184 will continue to place the instantiations in the same files.
7186 This is your best option for application code written for the Borland
7187 model, as it will just work. Code written for the Cfront model will
7188 need to be modified so that the template definitions are available at
7189 one or more points of instantiation; usually this is as simple as adding
7190 @code{#include <tmethods.cc>} to the end of each template header.
7192 For library code, if you want the library to provide all of the template
7193 instantiations it needs, just try to link all of its object files
7194 together; the link will fail, but cause the instantiations to be
7195 generated as a side effect. Be warned, however, that this may cause
7196 conflicts if multiple libraries try to provide the same instantiations.
7197 For greater control, use explicit instantiation as described in the next
7201 @opindex fno-implicit-templates
7202 Compile your code with @option{-fno-implicit-templates} to disable the
7203 implicit generation of template instances, and explicitly instantiate
7204 all the ones you use. This approach requires more knowledge of exactly
7205 which instances you need than do the others, but it's less
7206 mysterious and allows greater control. You can scatter the explicit
7207 instantiations throughout your program, perhaps putting them in the
7208 translation units where the instances are used or the translation units
7209 that define the templates themselves; you can put all of the explicit
7210 instantiations you need into one big file; or you can create small files
7217 template class Foo<int>;
7218 template ostream& operator <<
7219 (ostream&, const Foo<int>&);
7222 for each of the instances you need, and create a template instantiation
7225 If you are using Cfront-model code, you can probably get away with not
7226 using @option{-fno-implicit-templates} when compiling files that don't
7227 @samp{#include} the member template definitions.
7229 If you use one big file to do the instantiations, you may want to
7230 compile it without @option{-fno-implicit-templates} so you get all of the
7231 instances required by your explicit instantiations (but not by any
7232 other files) without having to specify them as well.
7234 g++ has extended the template instantiation syntax outlined in the
7235 Working Paper to allow forward declaration of explicit instantiations
7236 (with @code{extern}), instantiation of the compiler support data for a
7237 template class (i.e.@: the vtable) without instantiating any of its
7238 members (with @code{inline}), and instantiation of only the static data
7239 members of a template class, without the support data or member
7240 functions (with (@code{static}):
7243 extern template int max (int, int);
7244 inline template class Foo<int>;
7245 static template class Foo<int>;
7249 Do nothing. Pretend g++ does implement automatic instantiation
7250 management. Code written for the Borland model will work fine, but
7251 each translation unit will contain instances of each of the templates it
7252 uses. In a large program, this can lead to an unacceptable amount of code
7255 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
7256 more discussion of these pragmas.
7259 @node Bound member functions
7260 @section Extracting the function pointer from a bound pointer to member function
7263 @cindex pointer to member function
7264 @cindex bound pointer to member function
7266 In C++, pointer to member functions (PMFs) are implemented using a wide
7267 pointer of sorts to handle all the possible call mechanisms; the PMF
7268 needs to store information about how to adjust the @samp{this} pointer,
7269 and if the function pointed to is virtual, where to find the vtable, and
7270 where in the vtable to look for the member function. If you are using
7271 PMFs in an inner loop, you should really reconsider that decision. If
7272 that is not an option, you can extract the pointer to the function that
7273 would be called for a given object/PMF pair and call it directly inside
7274 the inner loop, to save a bit of time.
7276 Note that you will still be paying the penalty for the call through a
7277 function pointer; on most modern architectures, such a call defeats the
7278 branch prediction features of the CPU@. This is also true of normal
7279 virtual function calls.
7281 The syntax for this extension is
7285 extern int (A::*fp)();
7286 typedef int (*fptr)(A *);
7288 fptr p = (fptr)(a.*fp);
7291 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
7292 no object is needed to obtain the address of the function. They can be
7293 converted to function pointers directly:
7296 fptr p1 = (fptr)(&A::foo);
7299 @opindex Wno-pmf-conversions
7300 You must specify @option{-Wno-pmf-conversions} to use this extension.
7302 @node C++ Attributes
7303 @section C++-Specific Variable, Function, and Type Attributes
7305 Some attributes only make sense for C++ programs.
7308 @item init_priority (@var{priority})
7309 @cindex init_priority attribute
7312 In Standard C++, objects defined at namespace scope are guaranteed to be
7313 initialized in an order in strict accordance with that of their definitions
7314 @emph{in a given translation unit}. No guarantee is made for initializations
7315 across translation units. However, GNU C++ allows users to control the
7316 order of initialization of objects defined at namespace scope with the
7317 @code{init_priority} attribute by specifying a relative @var{priority},
7318 a constant integral expression currently bounded between 101 and 65535
7319 inclusive. Lower numbers indicate a higher priority.
7321 In the following example, @code{A} would normally be created before
7322 @code{B}, but the @code{init_priority} attribute has reversed that order:
7325 Some_Class A __attribute__ ((init_priority (2000)));
7326 Some_Class B __attribute__ ((init_priority (543)));
7330 Note that the particular values of @var{priority} do not matter; only their
7333 @item java_interface
7334 @cindex java_interface attribute
7336 This type attribute informs C++ that the class is a Java interface. It may
7337 only be applied to classes declared within an @code{extern "Java"} block.
7338 Calls to methods declared in this interface will be dispatched using GCJ's
7339 interface table mechanism, instead of regular virtual table dispatch.
7343 @node Java Exceptions
7344 @section Java Exceptions
7346 The Java language uses a slightly different exception handling model
7347 from C++. Normally, GNU C++ will automatically detect when you are
7348 writing C++ code that uses Java exceptions, and handle them
7349 appropriately. However, if C++ code only needs to execute destructors
7350 when Java exceptions are thrown through it, GCC will guess incorrectly.
7351 Sample problematic code is:
7354 struct S @{ ~S(); @};
7355 extern void bar(); // is written in Java, and may throw exceptions
7364 The usual effect of an incorrect guess is a link failure, complaining of
7365 a missing routine called @samp{__gxx_personality_v0}.
7367 You can inform the compiler that Java exceptions are to be used in a
7368 translation unit, irrespective of what it might think, by writing
7369 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
7370 @samp{#pragma} must appear before any functions that throw or catch
7371 exceptions, or run destructors when exceptions are thrown through them.
7373 You cannot mix Java and C++ exceptions in the same translation unit. It
7374 is believed to be safe to throw a C++ exception from one file through
7375 another file compiled for the Java exception model, or vice versa, but
7376 there may be bugs in this area.
7378 @node Deprecated Features
7379 @section Deprecated Features
7381 In the past, the GNU C++ compiler was extended to experiment with new
7382 features, at a time when the C++ language was still evolving. Now that
7383 the C++ standard is complete, some of those features are superseded by
7384 superior alternatives. Using the old features might cause a warning in
7385 some cases that the feature will be dropped in the future. In other
7386 cases, the feature might be gone already.
7388 While the list below is not exhaustive, it documents some of the options
7389 that are now deprecated:
7392 @item -fexternal-templates
7393 @itemx -falt-external-templates
7394 These are two of the many ways for g++ to implement template
7395 instantiation. @xref{Template Instantiation}. The C++ standard clearly
7396 defines how template definitions have to be organized across
7397 implementation units. g++ has an implicit instantiation mechanism that
7398 should work just fine for standard-conforming code.
7400 @item -fstrict-prototype
7401 @itemx -fno-strict-prototype
7402 Previously it was possible to use an empty prototype parameter list to
7403 indicate an unspecified number of parameters (like C), rather than no
7404 parameters, as C++ demands. This feature has been removed, except where
7405 it is required for backwards compatibility @xref{Backwards Compatibility}.
7408 The named return value extension has been deprecated, and is now
7411 The use of initializer lists with new expressions has been deprecated,
7412 and is now removed from g++.
7414 Floating and complex non-type template parameters have been deprecated,
7415 and are now removed from g++.
7417 The implicit typename extension has been deprecated and will be removed
7418 from g++ at some point. In some cases g++ determines that a dependant
7419 type such as @code{TPL<T>::X} is a type without needing a
7420 @code{typename} keyword, contrary to the standard.
7422 @node Backwards Compatibility
7423 @section Backwards Compatibility
7424 @cindex Backwards Compatibility
7425 @cindex ARM [Annotated C++ Reference Manual]
7427 Now that there is a definitive ISO standard C++, G++ has a specification
7428 to adhere to. The C++ language evolved over time, and features that
7429 used to be acceptable in previous drafts of the standard, such as the ARM
7430 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
7431 compilation of C++ written to such drafts, G++ contains some backwards
7432 compatibilities. @emph{All such backwards compatibility features are
7433 liable to disappear in future versions of G++.} They should be considered
7434 deprecated @xref{Deprecated Features}.
7438 If a variable is declared at for scope, it used to remain in scope until
7439 the end of the scope which contained the for statement (rather than just
7440 within the for scope). G++ retains this, but issues a warning, if such a
7441 variable is accessed outside the for scope.
7443 @item Implicit C language
7444 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
7445 scope to set the language. On such systems, all header files are
7446 implicitly scoped inside a C language scope. Also, an empty prototype
7447 @code{()} will be treated as an unspecified number of arguments, rather
7448 than no arguments, as C++ demands.