1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
5 @c This is part of the GCC manual.
6 @c For copying conditions, see the file gcc.texi.
9 @chapter Extensions to the C Language Family
10 @cindex extensions, C language
11 @cindex C language extensions
14 GNU C provides several language features not found in ISO standard C@.
15 (The @option{-pedantic} option directs GCC to print a warning message if
16 any of these features is used.) To test for the availability of these
17 features in conditional compilation, check for a predefined macro
18 @code{__GNUC__}, which is always defined under GCC@.
20 These extensions are available in C and Objective-C@. Most of them are
21 also available in C++. @xref{C++ Extensions,,Extensions to the
22 C++ Language}, for extensions that apply @emph{only} to C++.
24 Some features that are in ISO C99 but not C90 or C++ are also, as
25 extensions, accepted by GCC in C90 mode and in C++.
28 * Statement Exprs:: Putting statements and declarations inside expressions.
29 * Local Labels:: Labels local to a block.
30 * Labels as Values:: Getting pointers to labels, and computed gotos.
31 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
32 * Constructing Calls:: Dispatching a call to another function.
33 * Typeof:: @code{typeof}: referring to the type of an expression.
34 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
35 * Long Long:: Double-word integers---@code{long long int}.
36 * __int128:: 128-bit integers---@code{__int128}.
37 * Complex:: Data types for complex numbers.
38 * Floating Types:: Additional Floating Types.
39 * Half-Precision:: Half-Precision Floating Point.
40 * Decimal Float:: Decimal Floating Types.
41 * Hex Floats:: Hexadecimal floating-point constants.
42 * Fixed-Point:: Fixed-Point Types.
43 * Named Address Spaces::Named address spaces.
44 * Zero Length:: Zero-length arrays.
45 * Variable Length:: Arrays whose length is computed at run time.
46 * Empty Structures:: Structures with no members.
47 * Variadic Macros:: Macros with a variable number of arguments.
48 * Escaped Newlines:: Slightly looser rules for escaped newlines.
49 * Subscripting:: Any array can be subscripted, even if not an lvalue.
50 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
51 * Initializers:: Non-constant initializers.
52 * Compound Literals:: Compound literals give structures, unions
54 * Designated Inits:: Labeling elements of initializers.
55 * Cast to Union:: Casting to union type from any member of the union.
56 * Case Ranges:: `case 1 ... 9' and such.
57 * Mixed Declarations:: Mixing declarations and code.
58 * Function Attributes:: Declaring that functions have no side effects,
59 or that they can never return.
60 * Attribute Syntax:: Formal syntax for attributes.
61 * Function Prototypes:: Prototype declarations and old-style definitions.
62 * C++ Comments:: C++ comments are recognized.
63 * Dollar Signs:: Dollar sign is allowed in identifiers.
64 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
65 * Variable Attributes:: Specifying attributes of variables.
66 * Type Attributes:: Specifying attributes of types.
67 * Alignment:: Inquiring about the alignment of a type or variable.
68 * Inline:: Defining inline functions (as fast as macros).
69 * Volatiles:: What constitutes an access to a volatile object.
70 * Extended Asm:: Assembler instructions with C expressions as operands.
71 (With them you can define ``built-in'' functions.)
72 * Constraints:: Constraints for asm operands
73 * Asm Labels:: Specifying the assembler name to use for a C symbol.
74 * Explicit Reg Vars:: Defining variables residing in specified registers.
75 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
76 * Incomplete Enums:: @code{enum foo;}, with details to follow.
77 * Function Names:: Printable strings which are the name of the current
79 * Return Address:: Getting the return or frame address of a function.
80 * Vector Extensions:: Using vector instructions through built-in functions.
81 * Offsetof:: Special syntax for implementing @code{offsetof}.
82 * Atomic Builtins:: Built-in functions for atomic memory access.
83 * Object Size Checking:: Built-in functions for limited buffer overflow
85 * Other Builtins:: Other built-in functions.
86 * Target Builtins:: Built-in functions specific to particular targets.
87 * Target Format Checks:: Format checks specific to particular targets.
88 * Pragmas:: Pragmas accepted by GCC.
89 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
90 * Thread-Local:: Per-thread variables.
91 * Binary constants:: Binary constants using the @samp{0b} prefix.
95 @section Statements and Declarations in Expressions
96 @cindex statements inside expressions
97 @cindex declarations inside expressions
98 @cindex expressions containing statements
99 @cindex macros, statements in expressions
101 @c the above section title wrapped and causes an underfull hbox.. i
102 @c changed it from "within" to "in". --mew 4feb93
103 A compound statement enclosed in parentheses may appear as an expression
104 in GNU C@. This allows you to use loops, switches, and local variables
105 within an expression.
107 Recall that a compound statement is a sequence of statements surrounded
108 by braces; in this construct, parentheses go around the braces. For
112 (@{ int y = foo (); int z;
119 is a valid (though slightly more complex than necessary) expression
120 for the absolute value of @code{foo ()}.
122 The last thing in the compound statement should be an expression
123 followed by a semicolon; the value of this subexpression serves as the
124 value of the entire construct. (If you use some other kind of statement
125 last within the braces, the construct has type @code{void}, and thus
126 effectively no value.)
128 This feature is especially useful in making macro definitions ``safe'' (so
129 that they evaluate each operand exactly once). For example, the
130 ``maximum'' function is commonly defined as a macro in standard C as
134 #define max(a,b) ((a) > (b) ? (a) : (b))
138 @cindex side effects, macro argument
139 But this definition computes either @var{a} or @var{b} twice, with bad
140 results if the operand has side effects. In GNU C, if you know the
141 type of the operands (here taken as @code{int}), you can define
142 the macro safely as follows:
145 #define maxint(a,b) \
146 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
149 Embedded statements are not allowed in constant expressions, such as
150 the value of an enumeration constant, the width of a bit-field, or
151 the initial value of a static variable.
153 If you don't know the type of the operand, you can still do this, but you
154 must use @code{typeof} (@pxref{Typeof}).
156 In G++, the result value of a statement expression undergoes array and
157 function pointer decay, and is returned by value to the enclosing
158 expression. For instance, if @code{A} is a class, then
167 will construct a temporary @code{A} object to hold the result of the
168 statement expression, and that will be used to invoke @code{Foo}.
169 Therefore the @code{this} pointer observed by @code{Foo} will not be the
172 Any temporaries created within a statement within a statement expression
173 will be destroyed at the statement's end. This makes statement
174 expressions inside macros slightly different from function calls. In
175 the latter case temporaries introduced during argument evaluation will
176 be destroyed at the end of the statement that includes the function
177 call. In the statement expression case they will be destroyed during
178 the statement expression. For instance,
181 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
182 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
192 will have different places where temporaries are destroyed. For the
193 @code{macro} case, the temporary @code{X} will be destroyed just after
194 the initialization of @code{b}. In the @code{function} case that
195 temporary will be destroyed when the function returns.
197 These considerations mean that it is probably a bad idea to use
198 statement-expressions of this form in header files that are designed to
199 work with C++. (Note that some versions of the GNU C Library contained
200 header files using statement-expression that lead to precisely this
203 Jumping into a statement expression with @code{goto} or using a
204 @code{switch} statement outside the statement expression with a
205 @code{case} or @code{default} label inside the statement expression is
206 not permitted. Jumping into a statement expression with a computed
207 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
208 Jumping out of a statement expression is permitted, but if the
209 statement expression is part of a larger expression then it is
210 unspecified which other subexpressions of that expression have been
211 evaluated except where the language definition requires certain
212 subexpressions to be evaluated before or after the statement
213 expression. In any case, as with a function call the evaluation of a
214 statement expression is not interleaved with the evaluation of other
215 parts of the containing expression. For example,
218 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
222 will call @code{foo} and @code{bar1} and will not call @code{baz} but
223 may or may not call @code{bar2}. If @code{bar2} is called, it will be
224 called after @code{foo} and before @code{bar1}
227 @section Locally Declared Labels
229 @cindex macros, local labels
231 GCC allows you to declare @dfn{local labels} in any nested block
232 scope. A local label is just like an ordinary label, but you can
233 only reference it (with a @code{goto} statement, or by taking its
234 address) within the block in which it was declared.
236 A local label declaration looks like this:
239 __label__ @var{label};
246 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
249 Local label declarations must come at the beginning of the block,
250 before any ordinary declarations or statements.
252 The label declaration defines the label @emph{name}, but does not define
253 the label itself. You must do this in the usual way, with
254 @code{@var{label}:}, within the statements of the statement expression.
256 The local label feature is useful for complex macros. If a macro
257 contains nested loops, a @code{goto} can be useful for breaking out of
258 them. However, an ordinary label whose scope is the whole function
259 cannot be used: if the macro can be expanded several times in one
260 function, the label will be multiply defined in that function. A
261 local label avoids this problem. For example:
264 #define SEARCH(value, array, target) \
267 typeof (target) _SEARCH_target = (target); \
268 typeof (*(array)) *_SEARCH_array = (array); \
271 for (i = 0; i < max; i++) \
272 for (j = 0; j < max; j++) \
273 if (_SEARCH_array[i][j] == _SEARCH_target) \
274 @{ (value) = i; goto found; @} \
280 This could also be written using a statement-expression:
283 #define SEARCH(array, target) \
286 typeof (target) _SEARCH_target = (target); \
287 typeof (*(array)) *_SEARCH_array = (array); \
290 for (i = 0; i < max; i++) \
291 for (j = 0; j < max; j++) \
292 if (_SEARCH_array[i][j] == _SEARCH_target) \
293 @{ value = i; goto found; @} \
300 Local label declarations also make the labels they declare visible to
301 nested functions, if there are any. @xref{Nested Functions}, for details.
303 @node Labels as Values
304 @section Labels as Values
305 @cindex labels as values
306 @cindex computed gotos
307 @cindex goto with computed label
308 @cindex address of a label
310 You can get the address of a label defined in the current function
311 (or a containing function) with the unary operator @samp{&&}. The
312 value has type @code{void *}. This value is a constant and can be used
313 wherever a constant of that type is valid. For example:
321 To use these values, you need to be able to jump to one. This is done
322 with the computed goto statement@footnote{The analogous feature in
323 Fortran is called an assigned goto, but that name seems inappropriate in
324 C, where one can do more than simply store label addresses in label
325 variables.}, @code{goto *@var{exp};}. For example,
332 Any expression of type @code{void *} is allowed.
334 One way of using these constants is in initializing a static array that
335 will serve as a jump table:
338 static void *array[] = @{ &&foo, &&bar, &&hack @};
341 Then you can select a label with indexing, like this:
348 Note that this does not check whether the subscript is in bounds---array
349 indexing in C never does that.
351 Such an array of label values serves a purpose much like that of the
352 @code{switch} statement. The @code{switch} statement is cleaner, so
353 use that rather than an array unless the problem does not fit a
354 @code{switch} statement very well.
356 Another use of label values is in an interpreter for threaded code.
357 The labels within the interpreter function can be stored in the
358 threaded code for super-fast dispatching.
360 You may not use this mechanism to jump to code in a different function.
361 If you do that, totally unpredictable things will happen. The best way to
362 avoid this is to store the label address only in automatic variables and
363 never pass it as an argument.
365 An alternate way to write the above example is
368 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
370 goto *(&&foo + array[i]);
374 This is more friendly to code living in shared libraries, as it reduces
375 the number of dynamic relocations that are needed, and by consequence,
376 allows the data to be read-only.
378 The @code{&&foo} expressions for the same label might have different
379 values if the containing function is inlined or cloned. If a program
380 relies on them being always the same,
381 @code{__attribute__((__noinline__,__noclone__))} should be used to
382 prevent inlining and cloning. If @code{&&foo} is used in a static
383 variable initializer, inlining and cloning is forbidden.
385 @node Nested Functions
386 @section Nested Functions
387 @cindex nested functions
388 @cindex downward funargs
391 A @dfn{nested function} is a function defined inside another function.
392 (Nested functions are not supported for GNU C++.) The nested function's
393 name is local to the block where it is defined. For example, here we
394 define a nested function named @code{square}, and call it twice:
398 foo (double a, double b)
400 double square (double z) @{ return z * z; @}
402 return square (a) + square (b);
407 The nested function can access all the variables of the containing
408 function that are visible at the point of its definition. This is
409 called @dfn{lexical scoping}. For example, here we show a nested
410 function which uses an inherited variable named @code{offset}:
414 bar (int *array, int offset, int size)
416 int access (int *array, int index)
417 @{ return array[index + offset]; @}
420 for (i = 0; i < size; i++)
421 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
426 Nested function definitions are permitted within functions in the places
427 where variable definitions are allowed; that is, in any block, mixed
428 with the other declarations and statements in the block.
430 It is possible to call the nested function from outside the scope of its
431 name by storing its address or passing the address to another function:
434 hack (int *array, int size)
436 void store (int index, int value)
437 @{ array[index] = value; @}
439 intermediate (store, size);
443 Here, the function @code{intermediate} receives the address of
444 @code{store} as an argument. If @code{intermediate} calls @code{store},
445 the arguments given to @code{store} are used to store into @code{array}.
446 But this technique works only so long as the containing function
447 (@code{hack}, in this example) does not exit.
449 If you try to call the nested function through its address after the
450 containing function has exited, all hell will break loose. If you try
451 to call it after a containing scope level has exited, and if it refers
452 to some of the variables that are no longer in scope, you may be lucky,
453 but it's not wise to take the risk. If, however, the nested function
454 does not refer to anything that has gone out of scope, you should be
457 GCC implements taking the address of a nested function using a technique
458 called @dfn{trampolines}. This technique was described in
459 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
460 C++ Conference Proceedings, October 17-21, 1988).
462 A nested function can jump to a label inherited from a containing
463 function, provided the label was explicitly declared in the containing
464 function (@pxref{Local Labels}). Such a jump returns instantly to the
465 containing function, exiting the nested function which did the
466 @code{goto} and any intermediate functions as well. Here is an example:
470 bar (int *array, int offset, int size)
473 int access (int *array, int index)
477 return array[index + offset];
481 for (i = 0; i < size; i++)
482 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
486 /* @r{Control comes here from @code{access}
487 if it detects an error.} */
494 A nested function always has no linkage. Declaring one with
495 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
496 before its definition, use @code{auto} (which is otherwise meaningless
497 for function declarations).
500 bar (int *array, int offset, int size)
503 auto int access (int *, int);
505 int access (int *array, int index)
509 return array[index + offset];
515 @node Constructing Calls
516 @section Constructing Function Calls
517 @cindex constructing calls
518 @cindex forwarding calls
520 Using the built-in functions described below, you can record
521 the arguments a function received, and call another function
522 with the same arguments, without knowing the number or types
525 You can also record the return value of that function call,
526 and later return that value, without knowing what data type
527 the function tried to return (as long as your caller expects
530 However, these built-in functions may interact badly with some
531 sophisticated features or other extensions of the language. It
532 is, therefore, not recommended to use them outside very simple
533 functions acting as mere forwarders for their arguments.
535 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
536 This built-in function returns a pointer to data
537 describing how to perform a call with the same arguments as were passed
538 to the current function.
540 The function saves the arg pointer register, structure value address,
541 and all registers that might be used to pass arguments to a function
542 into a block of memory allocated on the stack. Then it returns the
543 address of that block.
546 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
547 This built-in function invokes @var{function}
548 with a copy of the parameters described by @var{arguments}
551 The value of @var{arguments} should be the value returned by
552 @code{__builtin_apply_args}. The argument @var{size} specifies the size
553 of the stack argument data, in bytes.
555 This function returns a pointer to data describing
556 how to return whatever value was returned by @var{function}. The data
557 is saved in a block of memory allocated on the stack.
559 It is not always simple to compute the proper value for @var{size}. The
560 value is used by @code{__builtin_apply} to compute the amount of data
561 that should be pushed on the stack and copied from the incoming argument
565 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
566 This built-in function returns the value described by @var{result} from
567 the containing function. You should specify, for @var{result}, a value
568 returned by @code{__builtin_apply}.
571 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
572 This built-in function represents all anonymous arguments of an inline
573 function. It can be used only in inline functions which will be always
574 inlined, never compiled as a separate function, such as those using
575 @code{__attribute__ ((__always_inline__))} or
576 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
577 It must be only passed as last argument to some other function
578 with variable arguments. This is useful for writing small wrapper
579 inlines for variable argument functions, when using preprocessor
580 macros is undesirable. For example:
582 extern int myprintf (FILE *f, const char *format, ...);
583 extern inline __attribute__ ((__gnu_inline__)) int
584 myprintf (FILE *f, const char *format, ...)
586 int r = fprintf (f, "myprintf: ");
589 int s = fprintf (f, format, __builtin_va_arg_pack ());
597 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
598 This built-in function returns the number of anonymous arguments of
599 an inline function. It can be used only in inline functions which
600 will be always inlined, never compiled as a separate function, such
601 as those using @code{__attribute__ ((__always_inline__))} or
602 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
603 For example following will do link or runtime checking of open
604 arguments for optimized code:
607 extern inline __attribute__((__gnu_inline__)) int
608 myopen (const char *path, int oflag, ...)
610 if (__builtin_va_arg_pack_len () > 1)
611 warn_open_too_many_arguments ();
613 if (__builtin_constant_p (oflag))
615 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
617 warn_open_missing_mode ();
618 return __open_2 (path, oflag);
620 return open (path, oflag, __builtin_va_arg_pack ());
623 if (__builtin_va_arg_pack_len () < 1)
624 return __open_2 (path, oflag);
626 return open (path, oflag, __builtin_va_arg_pack ());
633 @section Referring to a Type with @code{typeof}
636 @cindex macros, types of arguments
638 Another way to refer to the type of an expression is with @code{typeof}.
639 The syntax of using of this keyword looks like @code{sizeof}, but the
640 construct acts semantically like a type name defined with @code{typedef}.
642 There are two ways of writing the argument to @code{typeof}: with an
643 expression or with a type. Here is an example with an expression:
650 This assumes that @code{x} is an array of pointers to functions;
651 the type described is that of the values of the functions.
653 Here is an example with a typename as the argument:
660 Here the type described is that of pointers to @code{int}.
662 If you are writing a header file that must work when included in ISO C
663 programs, write @code{__typeof__} instead of @code{typeof}.
664 @xref{Alternate Keywords}.
666 A @code{typeof}-construct can be used anywhere a typedef name could be
667 used. For example, you can use it in a declaration, in a cast, or inside
668 of @code{sizeof} or @code{typeof}.
670 The operand of @code{typeof} is evaluated for its side effects if and
671 only if it is an expression of variably modified type or the name of
674 @code{typeof} is often useful in conjunction with the
675 statements-within-expressions feature. Here is how the two together can
676 be used to define a safe ``maximum'' macro that operates on any
677 arithmetic type and evaluates each of its arguments exactly once:
681 (@{ typeof (a) _a = (a); \
682 typeof (b) _b = (b); \
683 _a > _b ? _a : _b; @})
686 @cindex underscores in variables in macros
687 @cindex @samp{_} in variables in macros
688 @cindex local variables in macros
689 @cindex variables, local, in macros
690 @cindex macros, local variables in
692 The reason for using names that start with underscores for the local
693 variables is to avoid conflicts with variable names that occur within the
694 expressions that are substituted for @code{a} and @code{b}. Eventually we
695 hope to design a new form of declaration syntax that allows you to declare
696 variables whose scopes start only after their initializers; this will be a
697 more reliable way to prevent such conflicts.
700 Some more examples of the use of @code{typeof}:
704 This declares @code{y} with the type of what @code{x} points to.
711 This declares @code{y} as an array of such values.
718 This declares @code{y} as an array of pointers to characters:
721 typeof (typeof (char *)[4]) y;
725 It is equivalent to the following traditional C declaration:
731 To see the meaning of the declaration using @code{typeof}, and why it
732 might be a useful way to write, rewrite it with these macros:
735 #define pointer(T) typeof(T *)
736 #define array(T, N) typeof(T [N])
740 Now the declaration can be rewritten this way:
743 array (pointer (char), 4) y;
747 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
748 pointers to @code{char}.
751 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
752 a more limited extension which permitted one to write
755 typedef @var{T} = @var{expr};
759 with the effect of declaring @var{T} to have the type of the expression
760 @var{expr}. This extension does not work with GCC 3 (versions between
761 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
762 relies on it should be rewritten to use @code{typeof}:
765 typedef typeof(@var{expr}) @var{T};
769 This will work with all versions of GCC@.
772 @section Conditionals with Omitted Operands
773 @cindex conditional expressions, extensions
774 @cindex omitted middle-operands
775 @cindex middle-operands, omitted
776 @cindex extensions, @code{?:}
777 @cindex @code{?:} extensions
779 The middle operand in a conditional expression may be omitted. Then
780 if the first operand is nonzero, its value is the value of the conditional
783 Therefore, the expression
790 has the value of @code{x} if that is nonzero; otherwise, the value of
793 This example is perfectly equivalent to
799 @cindex side effect in @code{?:}
800 @cindex @code{?:} side effect
802 In this simple case, the ability to omit the middle operand is not
803 especially useful. When it becomes useful is when the first operand does,
804 or may (if it is a macro argument), contain a side effect. Then repeating
805 the operand in the middle would perform the side effect twice. Omitting
806 the middle operand uses the value already computed without the undesirable
807 effects of recomputing it.
810 @section 128-bits integers
811 @cindex @code{__int128} data types
813 As an extension the integer scalar type @code{__int128} is supported for
814 targets having an integer mode wide enough to hold 128-bit.
815 Simply write @code{__int128} for a signed 128-bit integer, or
816 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
817 support in GCC to express an integer constant of type @code{__int128}
818 for targets having @code{long long} integer with less then 128 bit width.
821 @section Double-Word Integers
822 @cindex @code{long long} data types
823 @cindex double-word arithmetic
824 @cindex multiprecision arithmetic
825 @cindex @code{LL} integer suffix
826 @cindex @code{ULL} integer suffix
828 ISO C99 supports data types for integers that are at least 64 bits wide,
829 and as an extension GCC supports them in C90 mode and in C++.
830 Simply write @code{long long int} for a signed integer, or
831 @code{unsigned long long int} for an unsigned integer. To make an
832 integer constant of type @code{long long int}, add the suffix @samp{LL}
833 to the integer. To make an integer constant of type @code{unsigned long
834 long int}, add the suffix @samp{ULL} to the integer.
836 You can use these types in arithmetic like any other integer types.
837 Addition, subtraction, and bitwise boolean operations on these types
838 are open-coded on all types of machines. Multiplication is open-coded
839 if the machine supports fullword-to-doubleword a widening multiply
840 instruction. Division and shifts are open-coded only on machines that
841 provide special support. The operations that are not open-coded use
842 special library routines that come with GCC@.
844 There may be pitfalls when you use @code{long long} types for function
845 arguments, unless you declare function prototypes. If a function
846 expects type @code{int} for its argument, and you pass a value of type
847 @code{long long int}, confusion will result because the caller and the
848 subroutine will disagree about the number of bytes for the argument.
849 Likewise, if the function expects @code{long long int} and you pass
850 @code{int}. The best way to avoid such problems is to use prototypes.
853 @section Complex Numbers
854 @cindex complex numbers
855 @cindex @code{_Complex} keyword
856 @cindex @code{__complex__} keyword
858 ISO C99 supports complex floating data types, and as an extension GCC
859 supports them in C90 mode and in C++, and supports complex integer data
860 types which are not part of ISO C99. You can declare complex types
861 using the keyword @code{_Complex}. As an extension, the older GNU
862 keyword @code{__complex__} is also supported.
864 For example, @samp{_Complex double x;} declares @code{x} as a
865 variable whose real part and imaginary part are both of type
866 @code{double}. @samp{_Complex short int y;} declares @code{y} to
867 have real and imaginary parts of type @code{short int}; this is not
868 likely to be useful, but it shows that the set of complex types is
871 To write a constant with a complex data type, use the suffix @samp{i} or
872 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
873 has type @code{_Complex float} and @code{3i} has type
874 @code{_Complex int}. Such a constant always has a pure imaginary
875 value, but you can form any complex value you like by adding one to a
876 real constant. This is a GNU extension; if you have an ISO C99
877 conforming C library (such as GNU libc), and want to construct complex
878 constants of floating type, you should include @code{<complex.h>} and
879 use the macros @code{I} or @code{_Complex_I} instead.
881 @cindex @code{__real__} keyword
882 @cindex @code{__imag__} keyword
883 To extract the real part of a complex-valued expression @var{exp}, write
884 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
885 extract the imaginary part. This is a GNU extension; for values of
886 floating type, you should use the ISO C99 functions @code{crealf},
887 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
888 @code{cimagl}, declared in @code{<complex.h>} and also provided as
889 built-in functions by GCC@.
891 @cindex complex conjugation
892 The operator @samp{~} performs complex conjugation when used on a value
893 with a complex type. This is a GNU extension; for values of
894 floating type, you should use the ISO C99 functions @code{conjf},
895 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
896 provided as built-in functions by GCC@.
898 GCC can allocate complex automatic variables in a noncontiguous
899 fashion; it's even possible for the real part to be in a register while
900 the imaginary part is on the stack (or vice-versa). Only the DWARF2
901 debug info format can represent this, so use of DWARF2 is recommended.
902 If you are using the stabs debug info format, GCC describes a noncontiguous
903 complex variable as if it were two separate variables of noncomplex type.
904 If the variable's actual name is @code{foo}, the two fictitious
905 variables are named @code{foo$real} and @code{foo$imag}. You can
906 examine and set these two fictitious variables with your debugger.
909 @section Additional Floating Types
910 @cindex additional floating types
911 @cindex @code{__float80} data type
912 @cindex @code{__float128} data type
913 @cindex @code{w} floating point suffix
914 @cindex @code{q} floating point suffix
915 @cindex @code{W} floating point suffix
916 @cindex @code{Q} floating point suffix
918 As an extension, the GNU C compiler supports additional floating
919 types, @code{__float80} and @code{__float128} to support 80bit
920 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
921 Support for additional types includes the arithmetic operators:
922 add, subtract, multiply, divide; unary arithmetic operators;
923 relational operators; equality operators; and conversions to and from
924 integer and other floating types. Use a suffix @samp{w} or @samp{W}
925 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
926 for @code{_float128}. You can declare complex types using the
927 corresponding internal complex type, @code{XCmode} for @code{__float80}
928 type and @code{TCmode} for @code{__float128} type:
931 typedef _Complex float __attribute__((mode(TC))) _Complex128;
932 typedef _Complex float __attribute__((mode(XC))) _Complex80;
935 Not all targets support additional floating point types. @code{__float80}
936 and @code{__float128} types are supported on i386, x86_64 and ia64 targets.
937 The @code{__float128} type is supported on hppa HP-UX targets.
940 @section Half-Precision Floating Point
941 @cindex half-precision floating point
942 @cindex @code{__fp16} data type
944 On ARM targets, GCC supports half-precision (16-bit) floating point via
945 the @code{__fp16} type. You must enable this type explicitly
946 with the @option{-mfp16-format} command-line option in order to use it.
948 ARM supports two incompatible representations for half-precision
949 floating-point values. You must choose one of the representations and
950 use it consistently in your program.
952 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
953 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
954 There are 11 bits of significand precision, approximately 3
957 Specifying @option{-mfp16-format=alternative} selects the ARM
958 alternative format. This representation is similar to the IEEE
959 format, but does not support infinities or NaNs. Instead, the range
960 of exponents is extended, so that this format can represent normalized
961 values in the range of @math{2^{-14}} to 131008.
963 The @code{__fp16} type is a storage format only. For purposes
964 of arithmetic and other operations, @code{__fp16} values in C or C++
965 expressions are automatically promoted to @code{float}. In addition,
966 you cannot declare a function with a return value or parameters
967 of type @code{__fp16}.
969 Note that conversions from @code{double} to @code{__fp16}
970 involve an intermediate conversion to @code{float}. Because
971 of rounding, this can sometimes produce a different result than a
974 ARM provides hardware support for conversions between
975 @code{__fp16} and @code{float} values
976 as an extension to VFP and NEON (Advanced SIMD). GCC generates
977 code using these hardware instructions if you compile with
978 options to select an FPU that provides them;
979 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
980 in addition to the @option{-mfp16-format} option to select
981 a half-precision format.
983 Language-level support for the @code{__fp16} data type is
984 independent of whether GCC generates code using hardware floating-point
985 instructions. In cases where hardware support is not specified, GCC
986 implements conversions between @code{__fp16} and @code{float} values
990 @section Decimal Floating Types
991 @cindex decimal floating types
992 @cindex @code{_Decimal32} data type
993 @cindex @code{_Decimal64} data type
994 @cindex @code{_Decimal128} data type
995 @cindex @code{df} integer suffix
996 @cindex @code{dd} integer suffix
997 @cindex @code{dl} integer suffix
998 @cindex @code{DF} integer suffix
999 @cindex @code{DD} integer suffix
1000 @cindex @code{DL} integer suffix
1002 As an extension, the GNU C compiler supports decimal floating types as
1003 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1004 floating types in GCC will evolve as the draft technical report changes.
1005 Calling conventions for any target might also change. Not all targets
1006 support decimal floating types.
1008 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1009 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1010 @code{float}, @code{double}, and @code{long double} whose radix is not
1011 specified by the C standard but is usually two.
1013 Support for decimal floating types includes the arithmetic operators
1014 add, subtract, multiply, divide; unary arithmetic operators;
1015 relational operators; equality operators; and conversions to and from
1016 integer and other floating types. Use a suffix @samp{df} or
1017 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1018 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1021 GCC support of decimal float as specified by the draft technical report
1026 When the value of a decimal floating type cannot be represented in the
1027 integer type to which it is being converted, the result is undefined
1028 rather than the result value specified by the draft technical report.
1031 GCC does not provide the C library functionality associated with
1032 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1033 @file{wchar.h}, which must come from a separate C library implementation.
1034 Because of this the GNU C compiler does not define macro
1035 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1036 the technical report.
1039 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1040 are supported by the DWARF2 debug information format.
1046 ISO C99 supports floating-point numbers written not only in the usual
1047 decimal notation, such as @code{1.55e1}, but also numbers such as
1048 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1049 supports this in C90 mode (except in some cases when strictly
1050 conforming) and in C++. In that format the
1051 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1052 mandatory. The exponent is a decimal number that indicates the power of
1053 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1060 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1061 is the same as @code{1.55e1}.
1063 Unlike for floating-point numbers in the decimal notation the exponent
1064 is always required in the hexadecimal notation. Otherwise the compiler
1065 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1066 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1067 extension for floating-point constants of type @code{float}.
1070 @section Fixed-Point Types
1071 @cindex fixed-point types
1072 @cindex @code{_Fract} data type
1073 @cindex @code{_Accum} data type
1074 @cindex @code{_Sat} data type
1075 @cindex @code{hr} fixed-suffix
1076 @cindex @code{r} fixed-suffix
1077 @cindex @code{lr} fixed-suffix
1078 @cindex @code{llr} fixed-suffix
1079 @cindex @code{uhr} fixed-suffix
1080 @cindex @code{ur} fixed-suffix
1081 @cindex @code{ulr} fixed-suffix
1082 @cindex @code{ullr} fixed-suffix
1083 @cindex @code{hk} fixed-suffix
1084 @cindex @code{k} fixed-suffix
1085 @cindex @code{lk} fixed-suffix
1086 @cindex @code{llk} fixed-suffix
1087 @cindex @code{uhk} fixed-suffix
1088 @cindex @code{uk} fixed-suffix
1089 @cindex @code{ulk} fixed-suffix
1090 @cindex @code{ullk} fixed-suffix
1091 @cindex @code{HR} fixed-suffix
1092 @cindex @code{R} fixed-suffix
1093 @cindex @code{LR} fixed-suffix
1094 @cindex @code{LLR} fixed-suffix
1095 @cindex @code{UHR} fixed-suffix
1096 @cindex @code{UR} fixed-suffix
1097 @cindex @code{ULR} fixed-suffix
1098 @cindex @code{ULLR} fixed-suffix
1099 @cindex @code{HK} fixed-suffix
1100 @cindex @code{K} fixed-suffix
1101 @cindex @code{LK} fixed-suffix
1102 @cindex @code{LLK} fixed-suffix
1103 @cindex @code{UHK} fixed-suffix
1104 @cindex @code{UK} fixed-suffix
1105 @cindex @code{ULK} fixed-suffix
1106 @cindex @code{ULLK} fixed-suffix
1108 As an extension, the GNU C compiler supports fixed-point types as
1109 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1110 types in GCC will evolve as the draft technical report changes.
1111 Calling conventions for any target might also change. Not all targets
1112 support fixed-point types.
1114 The fixed-point types are
1115 @code{short _Fract},
1118 @code{long long _Fract},
1119 @code{unsigned short _Fract},
1120 @code{unsigned _Fract},
1121 @code{unsigned long _Fract},
1122 @code{unsigned long long _Fract},
1123 @code{_Sat short _Fract},
1125 @code{_Sat long _Fract},
1126 @code{_Sat long long _Fract},
1127 @code{_Sat unsigned short _Fract},
1128 @code{_Sat unsigned _Fract},
1129 @code{_Sat unsigned long _Fract},
1130 @code{_Sat unsigned long long _Fract},
1131 @code{short _Accum},
1134 @code{long long _Accum},
1135 @code{unsigned short _Accum},
1136 @code{unsigned _Accum},
1137 @code{unsigned long _Accum},
1138 @code{unsigned long long _Accum},
1139 @code{_Sat short _Accum},
1141 @code{_Sat long _Accum},
1142 @code{_Sat long long _Accum},
1143 @code{_Sat unsigned short _Accum},
1144 @code{_Sat unsigned _Accum},
1145 @code{_Sat unsigned long _Accum},
1146 @code{_Sat unsigned long long _Accum}.
1148 Fixed-point data values contain fractional and optional integral parts.
1149 The format of fixed-point data varies and depends on the target machine.
1151 Support for fixed-point types includes:
1154 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1156 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1158 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1160 binary shift operators (@code{<<}, @code{>>})
1162 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1164 equality operators (@code{==}, @code{!=})
1166 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1167 @code{<<=}, @code{>>=})
1169 conversions to and from integer, floating-point, or fixed-point types
1172 Use a suffix in a fixed-point literal constant:
1174 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1175 @code{_Sat short _Fract}
1176 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1177 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1178 @code{_Sat long _Fract}
1179 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1180 @code{_Sat long long _Fract}
1181 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1182 @code{_Sat unsigned short _Fract}
1183 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1184 @code{_Sat unsigned _Fract}
1185 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1186 @code{_Sat unsigned long _Fract}
1187 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1188 and @code{_Sat unsigned long long _Fract}
1189 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1190 @code{_Sat short _Accum}
1191 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1192 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1193 @code{_Sat long _Accum}
1194 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1195 @code{_Sat long long _Accum}
1196 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1197 @code{_Sat unsigned short _Accum}
1198 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1199 @code{_Sat unsigned _Accum}
1200 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1201 @code{_Sat unsigned long _Accum}
1202 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1203 and @code{_Sat unsigned long long _Accum}
1206 GCC support of fixed-point types as specified by the draft technical report
1211 Pragmas to control overflow and rounding behaviors are not implemented.
1214 Fixed-point types are supported by the DWARF2 debug information format.
1216 @node Named Address Spaces
1217 @section Named address spaces
1218 @cindex named address spaces
1220 As an extension, the GNU C compiler supports named address spaces as
1221 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1222 address spaces in GCC will evolve as the draft technical report changes.
1223 Calling conventions for any target might also change. At present, only
1224 the SPU and M32C targets support other address spaces. On the SPU target, for
1225 example, variables may be declared as belonging to another address space
1226 by qualifying the type with the @code{__ea} address space identifier:
1232 When the variable @code{i} is accessed, the compiler will generate
1233 special code to access this variable. It may use runtime library
1234 support, or generate special machine instructions to access that address
1237 The @code{__ea} identifier may be used exactly like any other C type
1238 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1239 document for more details.
1241 On the M32C target, with the R8C and M16C cpu variants, variables
1242 qualified with @code{__far} are accessed using 32-bit addresses in
1243 order to access memory beyond the first 64k bytes. If @code{__far} is
1244 used with the M32CM or M32C cpu variants, it has no effect.
1247 @section Arrays of Length Zero
1248 @cindex arrays of length zero
1249 @cindex zero-length arrays
1250 @cindex length-zero arrays
1251 @cindex flexible array members
1253 Zero-length arrays are allowed in GNU C@. They are very useful as the
1254 last element of a structure which is really a header for a variable-length
1263 struct line *thisline = (struct line *)
1264 malloc (sizeof (struct line) + this_length);
1265 thisline->length = this_length;
1268 In ISO C90, you would have to give @code{contents} a length of 1, which
1269 means either you waste space or complicate the argument to @code{malloc}.
1271 In ISO C99, you would use a @dfn{flexible array member}, which is
1272 slightly different in syntax and semantics:
1276 Flexible array members are written as @code{contents[]} without
1280 Flexible array members have incomplete type, and so the @code{sizeof}
1281 operator may not be applied. As a quirk of the original implementation
1282 of zero-length arrays, @code{sizeof} evaluates to zero.
1285 Flexible array members may only appear as the last member of a
1286 @code{struct} that is otherwise non-empty.
1289 A structure containing a flexible array member, or a union containing
1290 such a structure (possibly recursively), may not be a member of a
1291 structure or an element of an array. (However, these uses are
1292 permitted by GCC as extensions.)
1295 GCC versions before 3.0 allowed zero-length arrays to be statically
1296 initialized, as if they were flexible arrays. In addition to those
1297 cases that were useful, it also allowed initializations in situations
1298 that would corrupt later data. Non-empty initialization of zero-length
1299 arrays is now treated like any case where there are more initializer
1300 elements than the array holds, in that a suitable warning about "excess
1301 elements in array" is given, and the excess elements (all of them, in
1302 this case) are ignored.
1304 Instead GCC allows static initialization of flexible array members.
1305 This is equivalent to defining a new structure containing the original
1306 structure followed by an array of sufficient size to contain the data.
1307 I.e.@: in the following, @code{f1} is constructed as if it were declared
1313 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1316 struct f1 f1; int data[3];
1317 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1321 The convenience of this extension is that @code{f1} has the desired
1322 type, eliminating the need to consistently refer to @code{f2.f1}.
1324 This has symmetry with normal static arrays, in that an array of
1325 unknown size is also written with @code{[]}.
1327 Of course, this extension only makes sense if the extra data comes at
1328 the end of a top-level object, as otherwise we would be overwriting
1329 data at subsequent offsets. To avoid undue complication and confusion
1330 with initialization of deeply nested arrays, we simply disallow any
1331 non-empty initialization except when the structure is the top-level
1332 object. For example:
1335 struct foo @{ int x; int y[]; @};
1336 struct bar @{ struct foo z; @};
1338 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1339 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1340 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1341 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1344 @node Empty Structures
1345 @section Structures With No Members
1346 @cindex empty structures
1347 @cindex zero-size structures
1349 GCC permits a C structure to have no members:
1356 The structure will have size zero. In C++, empty structures are part
1357 of the language. G++ treats empty structures as if they had a single
1358 member of type @code{char}.
1360 @node Variable Length
1361 @section Arrays of Variable Length
1362 @cindex variable-length arrays
1363 @cindex arrays of variable length
1366 Variable-length automatic arrays are allowed in ISO C99, and as an
1367 extension GCC accepts them in C90 mode and in C++. These arrays are
1368 declared like any other automatic arrays, but with a length that is not
1369 a constant expression. The storage is allocated at the point of
1370 declaration and deallocated when the brace-level is exited. For
1375 concat_fopen (char *s1, char *s2, char *mode)
1377 char str[strlen (s1) + strlen (s2) + 1];
1380 return fopen (str, mode);
1384 @cindex scope of a variable length array
1385 @cindex variable-length array scope
1386 @cindex deallocating variable length arrays
1387 Jumping or breaking out of the scope of the array name deallocates the
1388 storage. Jumping into the scope is not allowed; you get an error
1391 @cindex @code{alloca} vs variable-length arrays
1392 You can use the function @code{alloca} to get an effect much like
1393 variable-length arrays. The function @code{alloca} is available in
1394 many other C implementations (but not in all). On the other hand,
1395 variable-length arrays are more elegant.
1397 There are other differences between these two methods. Space allocated
1398 with @code{alloca} exists until the containing @emph{function} returns.
1399 The space for a variable-length array is deallocated as soon as the array
1400 name's scope ends. (If you use both variable-length arrays and
1401 @code{alloca} in the same function, deallocation of a variable-length array
1402 will also deallocate anything more recently allocated with @code{alloca}.)
1404 You can also use variable-length arrays as arguments to functions:
1408 tester (int len, char data[len][len])
1414 The length of an array is computed once when the storage is allocated
1415 and is remembered for the scope of the array in case you access it with
1418 If you want to pass the array first and the length afterward, you can
1419 use a forward declaration in the parameter list---another GNU extension.
1423 tester (int len; char data[len][len], int len)
1429 @cindex parameter forward declaration
1430 The @samp{int len} before the semicolon is a @dfn{parameter forward
1431 declaration}, and it serves the purpose of making the name @code{len}
1432 known when the declaration of @code{data} is parsed.
1434 You can write any number of such parameter forward declarations in the
1435 parameter list. They can be separated by commas or semicolons, but the
1436 last one must end with a semicolon, which is followed by the ``real''
1437 parameter declarations. Each forward declaration must match a ``real''
1438 declaration in parameter name and data type. ISO C99 does not support
1439 parameter forward declarations.
1441 @node Variadic Macros
1442 @section Macros with a Variable Number of Arguments.
1443 @cindex variable number of arguments
1444 @cindex macro with variable arguments
1445 @cindex rest argument (in macro)
1446 @cindex variadic macros
1448 In the ISO C standard of 1999, a macro can be declared to accept a
1449 variable number of arguments much as a function can. The syntax for
1450 defining the macro is similar to that of a function. Here is an
1454 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1457 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1458 such a macro, it represents the zero or more tokens until the closing
1459 parenthesis that ends the invocation, including any commas. This set of
1460 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1461 wherever it appears. See the CPP manual for more information.
1463 GCC has long supported variadic macros, and used a different syntax that
1464 allowed you to give a name to the variable arguments just like any other
1465 argument. Here is an example:
1468 #define debug(format, args...) fprintf (stderr, format, args)
1471 This is in all ways equivalent to the ISO C example above, but arguably
1472 more readable and descriptive.
1474 GNU CPP has two further variadic macro extensions, and permits them to
1475 be used with either of the above forms of macro definition.
1477 In standard C, you are not allowed to leave the variable argument out
1478 entirely; but you are allowed to pass an empty argument. For example,
1479 this invocation is invalid in ISO C, because there is no comma after
1486 GNU CPP permits you to completely omit the variable arguments in this
1487 way. In the above examples, the compiler would complain, though since
1488 the expansion of the macro still has the extra comma after the format
1491 To help solve this problem, CPP behaves specially for variable arguments
1492 used with the token paste operator, @samp{##}. If instead you write
1495 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1498 and if the variable arguments are omitted or empty, the @samp{##}
1499 operator causes the preprocessor to remove the comma before it. If you
1500 do provide some variable arguments in your macro invocation, GNU CPP
1501 does not complain about the paste operation and instead places the
1502 variable arguments after the comma. Just like any other pasted macro
1503 argument, these arguments are not macro expanded.
1505 @node Escaped Newlines
1506 @section Slightly Looser Rules for Escaped Newlines
1507 @cindex escaped newlines
1508 @cindex newlines (escaped)
1510 Recently, the preprocessor has relaxed its treatment of escaped
1511 newlines. Previously, the newline had to immediately follow a
1512 backslash. The current implementation allows whitespace in the form
1513 of spaces, horizontal and vertical tabs, and form feeds between the
1514 backslash and the subsequent newline. The preprocessor issues a
1515 warning, but treats it as a valid escaped newline and combines the two
1516 lines to form a single logical line. This works within comments and
1517 tokens, as well as between tokens. Comments are @emph{not} treated as
1518 whitespace for the purposes of this relaxation, since they have not
1519 yet been replaced with spaces.
1522 @section Non-Lvalue Arrays May Have Subscripts
1523 @cindex subscripting
1524 @cindex arrays, non-lvalue
1526 @cindex subscripting and function values
1527 In ISO C99, arrays that are not lvalues still decay to pointers, and
1528 may be subscripted, although they may not be modified or used after
1529 the next sequence point and the unary @samp{&} operator may not be
1530 applied to them. As an extension, GCC allows such arrays to be
1531 subscripted in C90 mode, though otherwise they do not decay to
1532 pointers outside C99 mode. For example,
1533 this is valid in GNU C though not valid in C90:
1537 struct foo @{int a[4];@};
1543 return f().a[index];
1549 @section Arithmetic on @code{void}- and Function-Pointers
1550 @cindex void pointers, arithmetic
1551 @cindex void, size of pointer to
1552 @cindex function pointers, arithmetic
1553 @cindex function, size of pointer to
1555 In GNU C, addition and subtraction operations are supported on pointers to
1556 @code{void} and on pointers to functions. This is done by treating the
1557 size of a @code{void} or of a function as 1.
1559 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1560 and on function types, and returns 1.
1562 @opindex Wpointer-arith
1563 The option @option{-Wpointer-arith} requests a warning if these extensions
1567 @section Non-Constant Initializers
1568 @cindex initializers, non-constant
1569 @cindex non-constant initializers
1571 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1572 automatic variable are not required to be constant expressions in GNU C@.
1573 Here is an example of an initializer with run-time varying elements:
1576 foo (float f, float g)
1578 float beat_freqs[2] = @{ f-g, f+g @};
1583 @node Compound Literals
1584 @section Compound Literals
1585 @cindex constructor expressions
1586 @cindex initializations in expressions
1587 @cindex structures, constructor expression
1588 @cindex expressions, constructor
1589 @cindex compound literals
1590 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1592 ISO C99 supports compound literals. A compound literal looks like
1593 a cast containing an initializer. Its value is an object of the
1594 type specified in the cast, containing the elements specified in
1595 the initializer; it is an lvalue. As an extension, GCC supports
1596 compound literals in C90 mode and in C++.
1598 Usually, the specified type is a structure. Assume that
1599 @code{struct foo} and @code{structure} are declared as shown:
1602 struct foo @{int a; char b[2];@} structure;
1606 Here is an example of constructing a @code{struct foo} with a compound literal:
1609 structure = ((struct foo) @{x + y, 'a', 0@});
1613 This is equivalent to writing the following:
1617 struct foo temp = @{x + y, 'a', 0@};
1622 You can also construct an array. If all the elements of the compound literal
1623 are (made up of) simple constant expressions, suitable for use in
1624 initializers of objects of static storage duration, then the compound
1625 literal can be coerced to a pointer to its first element and used in
1626 such an initializer, as shown here:
1629 char **foo = (char *[]) @{ "x", "y", "z" @};
1632 Compound literals for scalar types and union types are is
1633 also allowed, but then the compound literal is equivalent
1636 As a GNU extension, GCC allows initialization of objects with static storage
1637 duration by compound literals (which is not possible in ISO C99, because
1638 the initializer is not a constant).
1639 It is handled as if the object was initialized only with the bracket
1640 enclosed list if the types of the compound literal and the object match.
1641 The initializer list of the compound literal must be constant.
1642 If the object being initialized has array type of unknown size, the size is
1643 determined by compound literal size.
1646 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1647 static int y[] = (int []) @{1, 2, 3@};
1648 static int z[] = (int [3]) @{1@};
1652 The above lines are equivalent to the following:
1654 static struct foo x = @{1, 'a', 'b'@};
1655 static int y[] = @{1, 2, 3@};
1656 static int z[] = @{1, 0, 0@};
1659 @node Designated Inits
1660 @section Designated Initializers
1661 @cindex initializers with labeled elements
1662 @cindex labeled elements in initializers
1663 @cindex case labels in initializers
1664 @cindex designated initializers
1666 Standard C90 requires the elements of an initializer to appear in a fixed
1667 order, the same as the order of the elements in the array or structure
1670 In ISO C99 you can give the elements in any order, specifying the array
1671 indices or structure field names they apply to, and GNU C allows this as
1672 an extension in C90 mode as well. This extension is not
1673 implemented in GNU C++.
1675 To specify an array index, write
1676 @samp{[@var{index}] =} before the element value. For example,
1679 int a[6] = @{ [4] = 29, [2] = 15 @};
1686 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1690 The index values must be constant expressions, even if the array being
1691 initialized is automatic.
1693 An alternative syntax for this which has been obsolete since GCC 2.5 but
1694 GCC still accepts is to write @samp{[@var{index}]} before the element
1695 value, with no @samp{=}.
1697 To initialize a range of elements to the same value, write
1698 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1699 extension. For example,
1702 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1706 If the value in it has side-effects, the side-effects will happen only once,
1707 not for each initialized field by the range initializer.
1710 Note that the length of the array is the highest value specified
1713 In a structure initializer, specify the name of a field to initialize
1714 with @samp{.@var{fieldname} =} before the element value. For example,
1715 given the following structure,
1718 struct point @{ int x, y; @};
1722 the following initialization
1725 struct point p = @{ .y = yvalue, .x = xvalue @};
1732 struct point p = @{ xvalue, yvalue @};
1735 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1736 @samp{@var{fieldname}:}, as shown here:
1739 struct point p = @{ y: yvalue, x: xvalue @};
1743 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1744 @dfn{designator}. You can also use a designator (or the obsolete colon
1745 syntax) when initializing a union, to specify which element of the union
1746 should be used. For example,
1749 union foo @{ int i; double d; @};
1751 union foo f = @{ .d = 4 @};
1755 will convert 4 to a @code{double} to store it in the union using
1756 the second element. By contrast, casting 4 to type @code{union foo}
1757 would store it into the union as the integer @code{i}, since it is
1758 an integer. (@xref{Cast to Union}.)
1760 You can combine this technique of naming elements with ordinary C
1761 initialization of successive elements. Each initializer element that
1762 does not have a designator applies to the next consecutive element of the
1763 array or structure. For example,
1766 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1773 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1776 Labeling the elements of an array initializer is especially useful
1777 when the indices are characters or belong to an @code{enum} type.
1782 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1783 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1786 @cindex designator lists
1787 You can also write a series of @samp{.@var{fieldname}} and
1788 @samp{[@var{index}]} designators before an @samp{=} to specify a
1789 nested subobject to initialize; the list is taken relative to the
1790 subobject corresponding to the closest surrounding brace pair. For
1791 example, with the @samp{struct point} declaration above:
1794 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1798 If the same field is initialized multiple times, it will have value from
1799 the last initialization. If any such overridden initialization has
1800 side-effect, it is unspecified whether the side-effect happens or not.
1801 Currently, GCC will discard them and issue a warning.
1804 @section Case Ranges
1806 @cindex ranges in case statements
1808 You can specify a range of consecutive values in a single @code{case} label,
1812 case @var{low} ... @var{high}:
1816 This has the same effect as the proper number of individual @code{case}
1817 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1819 This feature is especially useful for ranges of ASCII character codes:
1825 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1826 it may be parsed wrong when you use it with integer values. For example,
1841 @section Cast to a Union Type
1842 @cindex cast to a union
1843 @cindex union, casting to a
1845 A cast to union type is similar to other casts, except that the type
1846 specified is a union type. You can specify the type either with
1847 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1848 a constructor though, not a cast, and hence does not yield an lvalue like
1849 normal casts. (@xref{Compound Literals}.)
1851 The types that may be cast to the union type are those of the members
1852 of the union. Thus, given the following union and variables:
1855 union foo @{ int i; double d; @};
1861 both @code{x} and @code{y} can be cast to type @code{union foo}.
1863 Using the cast as the right-hand side of an assignment to a variable of
1864 union type is equivalent to storing in a member of the union:
1869 u = (union foo) x @equiv{} u.i = x
1870 u = (union foo) y @equiv{} u.d = y
1873 You can also use the union cast as a function argument:
1876 void hack (union foo);
1878 hack ((union foo) x);
1881 @node Mixed Declarations
1882 @section Mixed Declarations and Code
1883 @cindex mixed declarations and code
1884 @cindex declarations, mixed with code
1885 @cindex code, mixed with declarations
1887 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1888 within compound statements. As an extension, GCC also allows this in
1889 C90 mode. For example, you could do:
1898 Each identifier is visible from where it is declared until the end of
1899 the enclosing block.
1901 @node Function Attributes
1902 @section Declaring Attributes of Functions
1903 @cindex function attributes
1904 @cindex declaring attributes of functions
1905 @cindex functions that never return
1906 @cindex functions that return more than once
1907 @cindex functions that have no side effects
1908 @cindex functions in arbitrary sections
1909 @cindex functions that behave like malloc
1910 @cindex @code{volatile} applied to function
1911 @cindex @code{const} applied to function
1912 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1913 @cindex functions with non-null pointer arguments
1914 @cindex functions that are passed arguments in registers on the 386
1915 @cindex functions that pop the argument stack on the 386
1916 @cindex functions that do not pop the argument stack on the 386
1917 @cindex functions that have different compilation options on the 386
1918 @cindex functions that have different optimization options
1919 @cindex functions that are dynamically resolved
1921 In GNU C, you declare certain things about functions called in your program
1922 which help the compiler optimize function calls and check your code more
1925 The keyword @code{__attribute__} allows you to specify special
1926 attributes when making a declaration. This keyword is followed by an
1927 attribute specification inside double parentheses. The following
1928 attributes are currently defined for functions on all targets:
1929 @code{aligned}, @code{alloc_size}, @code{noreturn},
1930 @code{returns_twice}, @code{noinline}, @code{noclone},
1931 @code{always_inline}, @code{flatten}, @code{pure}, @code{const},
1932 @code{nothrow}, @code{sentinel}, @code{format}, @code{format_arg},
1933 @code{no_instrument_function}, @code{no_split_stack},
1934 @code{section}, @code{constructor},
1935 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1936 @code{weak}, @code{malloc}, @code{alias}, @code{ifunc},
1937 @code{warn_unused_result}, @code{nonnull}, @code{gnu_inline},
1938 @code{externally_visible}, @code{hot}, @code{cold}, @code{artificial},
1939 @code{error} and @code{warning}. Several other attributes are defined
1940 for functions on particular target systems. Other attributes,
1941 including @code{section} are supported for variables declarations
1942 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1944 GCC plugins may provide their own attributes.
1946 You may also specify attributes with @samp{__} preceding and following
1947 each keyword. This allows you to use them in header files without
1948 being concerned about a possible macro of the same name. For example,
1949 you may use @code{__noreturn__} instead of @code{noreturn}.
1951 @xref{Attribute Syntax}, for details of the exact syntax for using
1955 @c Keep this table alphabetized by attribute name. Treat _ as space.
1957 @item alias ("@var{target}")
1958 @cindex @code{alias} attribute
1959 The @code{alias} attribute causes the declaration to be emitted as an
1960 alias for another symbol, which must be specified. For instance,
1963 void __f () @{ /* @r{Do something.} */; @}
1964 void f () __attribute__ ((weak, alias ("__f")));
1967 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1968 mangled name for the target must be used. It is an error if @samp{__f}
1969 is not defined in the same translation unit.
1971 Not all target machines support this attribute.
1973 @item aligned (@var{alignment})
1974 @cindex @code{aligned} attribute
1975 This attribute specifies a minimum alignment for the function,
1978 You cannot use this attribute to decrease the alignment of a function,
1979 only to increase it. However, when you explicitly specify a function
1980 alignment this will override the effect of the
1981 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1984 Note that the effectiveness of @code{aligned} attributes may be
1985 limited by inherent limitations in your linker. On many systems, the
1986 linker is only able to arrange for functions to be aligned up to a
1987 certain maximum alignment. (For some linkers, the maximum supported
1988 alignment may be very very small.) See your linker documentation for
1989 further information.
1991 The @code{aligned} attribute can also be used for variables and fields
1992 (@pxref{Variable Attributes}.)
1995 @cindex @code{alloc_size} attribute
1996 The @code{alloc_size} attribute is used to tell the compiler that the
1997 function return value points to memory, where the size is given by
1998 one or two of the functions parameters. GCC uses this
1999 information to improve the correctness of @code{__builtin_object_size}.
2001 The function parameter(s) denoting the allocated size are specified by
2002 one or two integer arguments supplied to the attribute. The allocated size
2003 is either the value of the single function argument specified or the product
2004 of the two function arguments specified. Argument numbering starts at
2010 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2011 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2014 declares that my_calloc will return memory of the size given by
2015 the product of parameter 1 and 2 and that my_realloc will return memory
2016 of the size given by parameter 2.
2019 @cindex @code{always_inline} function attribute
2020 Generally, functions are not inlined unless optimization is specified.
2021 For functions declared inline, this attribute inlines the function even
2022 if no optimization level was specified.
2025 @cindex @code{gnu_inline} function attribute
2026 This attribute should be used with a function which is also declared
2027 with the @code{inline} keyword. It directs GCC to treat the function
2028 as if it were defined in gnu90 mode even when compiling in C99 or
2031 If the function is declared @code{extern}, then this definition of the
2032 function is used only for inlining. In no case is the function
2033 compiled as a standalone function, not even if you take its address
2034 explicitly. Such an address becomes an external reference, as if you
2035 had only declared the function, and had not defined it. This has
2036 almost the effect of a macro. The way to use this is to put a
2037 function definition in a header file with this attribute, and put
2038 another copy of the function, without @code{extern}, in a library
2039 file. The definition in the header file will cause most calls to the
2040 function to be inlined. If any uses of the function remain, they will
2041 refer to the single copy in the library. Note that the two
2042 definitions of the functions need not be precisely the same, although
2043 if they do not have the same effect your program may behave oddly.
2045 In C, if the function is neither @code{extern} nor @code{static}, then
2046 the function is compiled as a standalone function, as well as being
2047 inlined where possible.
2049 This is how GCC traditionally handled functions declared
2050 @code{inline}. Since ISO C99 specifies a different semantics for
2051 @code{inline}, this function attribute is provided as a transition
2052 measure and as a useful feature in its own right. This attribute is
2053 available in GCC 4.1.3 and later. It is available if either of the
2054 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2055 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2056 Function is As Fast As a Macro}.
2058 In C++, this attribute does not depend on @code{extern} in any way,
2059 but it still requires the @code{inline} keyword to enable its special
2063 @cindex @code{artificial} function attribute
2064 This attribute is useful for small inline wrappers which if possible
2065 should appear during debugging as a unit, depending on the debug
2066 info format it will either mean marking the function as artificial
2067 or using the caller location for all instructions within the inlined
2071 @cindex interrupt handler functions
2072 When added to an interrupt handler with the M32C port, causes the
2073 prologue and epilogue to use bank switching to preserve the registers
2074 rather than saving them on the stack.
2077 @cindex @code{flatten} function attribute
2078 Generally, inlining into a function is limited. For a function marked with
2079 this attribute, every call inside this function will be inlined, if possible.
2080 Whether the function itself is considered for inlining depends on its size and
2081 the current inlining parameters.
2083 @item error ("@var{message}")
2084 @cindex @code{error} function attribute
2085 If this attribute is used on a function declaration and a call to such a function
2086 is not eliminated through dead code elimination or other optimizations, an error
2087 which will include @var{message} will be diagnosed. This is useful
2088 for compile time checking, especially together with @code{__builtin_constant_p}
2089 and inline functions where checking the inline function arguments is not
2090 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2091 While it is possible to leave the function undefined and thus invoke
2092 a link failure, when using this attribute the problem will be diagnosed
2093 earlier and with exact location of the call even in presence of inline
2094 functions or when not emitting debugging information.
2096 @item warning ("@var{message}")
2097 @cindex @code{warning} function attribute
2098 If this attribute is used on a function declaration and a call to such a function
2099 is not eliminated through dead code elimination or other optimizations, a warning
2100 which will include @var{message} will be diagnosed. This is useful
2101 for compile time checking, especially together with @code{__builtin_constant_p}
2102 and inline functions. While it is possible to define the function with
2103 a message in @code{.gnu.warning*} section, when using this attribute the problem
2104 will be diagnosed earlier and with exact location of the call even in presence
2105 of inline functions or when not emitting debugging information.
2108 @cindex functions that do pop the argument stack on the 386
2110 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2111 assume that the calling function will pop off the stack space used to
2112 pass arguments. This is
2113 useful to override the effects of the @option{-mrtd} switch.
2116 @cindex @code{const} function attribute
2117 Many functions do not examine any values except their arguments, and
2118 have no effects except the return value. Basically this is just slightly
2119 more strict class than the @code{pure} attribute below, since function is not
2120 allowed to read global memory.
2122 @cindex pointer arguments
2123 Note that a function that has pointer arguments and examines the data
2124 pointed to must @emph{not} be declared @code{const}. Likewise, a
2125 function that calls a non-@code{const} function usually must not be
2126 @code{const}. It does not make sense for a @code{const} function to
2129 The attribute @code{const} is not implemented in GCC versions earlier
2130 than 2.5. An alternative way to declare that a function has no side
2131 effects, which works in the current version and in some older versions,
2135 typedef int intfn ();
2137 extern const intfn square;
2140 This approach does not work in GNU C++ from 2.6.0 on, since the language
2141 specifies that the @samp{const} must be attached to the return value.
2145 @itemx constructor (@var{priority})
2146 @itemx destructor (@var{priority})
2147 @cindex @code{constructor} function attribute
2148 @cindex @code{destructor} function attribute
2149 The @code{constructor} attribute causes the function to be called
2150 automatically before execution enters @code{main ()}. Similarly, the
2151 @code{destructor} attribute causes the function to be called
2152 automatically after @code{main ()} has completed or @code{exit ()} has
2153 been called. Functions with these attributes are useful for
2154 initializing data that will be used implicitly during the execution of
2157 You may provide an optional integer priority to control the order in
2158 which constructor and destructor functions are run. A constructor
2159 with a smaller priority number runs before a constructor with a larger
2160 priority number; the opposite relationship holds for destructors. So,
2161 if you have a constructor that allocates a resource and a destructor
2162 that deallocates the same resource, both functions typically have the
2163 same priority. The priorities for constructor and destructor
2164 functions are the same as those specified for namespace-scope C++
2165 objects (@pxref{C++ Attributes}).
2167 These attributes are not currently implemented for Objective-C@.
2170 @itemx deprecated (@var{msg})
2171 @cindex @code{deprecated} attribute.
2172 The @code{deprecated} attribute results in a warning if the function
2173 is used anywhere in the source file. This is useful when identifying
2174 functions that are expected to be removed in a future version of a
2175 program. The warning also includes the location of the declaration
2176 of the deprecated function, to enable users to easily find further
2177 information about why the function is deprecated, or what they should
2178 do instead. Note that the warnings only occurs for uses:
2181 int old_fn () __attribute__ ((deprecated));
2183 int (*fn_ptr)() = old_fn;
2186 results in a warning on line 3 but not line 2. The optional msg
2187 argument, which must be a string, will be printed in the warning if
2190 The @code{deprecated} attribute can also be used for variables and
2191 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2194 @cindex @code{disinterrupt} attribute
2195 On MeP targets, this attribute causes the compiler to emit
2196 instructions to disable interrupts for the duration of the given
2200 @cindex @code{__declspec(dllexport)}
2201 On Microsoft Windows targets and Symbian OS targets the
2202 @code{dllexport} attribute causes the compiler to provide a global
2203 pointer to a pointer in a DLL, so that it can be referenced with the
2204 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2205 name is formed by combining @code{_imp__} and the function or variable
2208 You can use @code{__declspec(dllexport)} as a synonym for
2209 @code{__attribute__ ((dllexport))} for compatibility with other
2212 On systems that support the @code{visibility} attribute, this
2213 attribute also implies ``default'' visibility. It is an error to
2214 explicitly specify any other visibility.
2216 In previous versions of GCC, the @code{dllexport} attribute was ignored
2217 for inlined functions, unless the @option{-fkeep-inline-functions} flag
2218 had been used. The default behaviour now is to emit all dllexported
2219 inline functions; however, this can cause object file-size bloat, in
2220 which case the old behaviour can be restored by using
2221 @option{-fno-keep-inline-dllexport}.
2223 The attribute is also ignored for undefined symbols.
2225 When applied to C++ classes, the attribute marks defined non-inlined
2226 member functions and static data members as exports. Static consts
2227 initialized in-class are not marked unless they are also defined
2230 For Microsoft Windows targets there are alternative methods for
2231 including the symbol in the DLL's export table such as using a
2232 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2233 the @option{--export-all} linker flag.
2236 @cindex @code{__declspec(dllimport)}
2237 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2238 attribute causes the compiler to reference a function or variable via
2239 a global pointer to a pointer that is set up by the DLL exporting the
2240 symbol. The attribute implies @code{extern}. On Microsoft Windows
2241 targets, the pointer name is formed by combining @code{_imp__} and the
2242 function or variable name.
2244 You can use @code{__declspec(dllimport)} as a synonym for
2245 @code{__attribute__ ((dllimport))} for compatibility with other
2248 On systems that support the @code{visibility} attribute, this
2249 attribute also implies ``default'' visibility. It is an error to
2250 explicitly specify any other visibility.
2252 Currently, the attribute is ignored for inlined functions. If the
2253 attribute is applied to a symbol @emph{definition}, an error is reported.
2254 If a symbol previously declared @code{dllimport} is later defined, the
2255 attribute is ignored in subsequent references, and a warning is emitted.
2256 The attribute is also overridden by a subsequent declaration as
2259 When applied to C++ classes, the attribute marks non-inlined
2260 member functions and static data members as imports. However, the
2261 attribute is ignored for virtual methods to allow creation of vtables
2264 On the SH Symbian OS target the @code{dllimport} attribute also has
2265 another affect---it can cause the vtable and run-time type information
2266 for a class to be exported. This happens when the class has a
2267 dllimport'ed constructor or a non-inline, non-pure virtual function
2268 and, for either of those two conditions, the class also has an inline
2269 constructor or destructor and has a key function that is defined in
2270 the current translation unit.
2272 For Microsoft Windows based targets the use of the @code{dllimport}
2273 attribute on functions is not necessary, but provides a small
2274 performance benefit by eliminating a thunk in the DLL@. The use of the
2275 @code{dllimport} attribute on imported variables was required on older
2276 versions of the GNU linker, but can now be avoided by passing the
2277 @option{--enable-auto-import} switch to the GNU linker. As with
2278 functions, using the attribute for a variable eliminates a thunk in
2281 One drawback to using this attribute is that a pointer to a
2282 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2283 address. However, a pointer to a @emph{function} with the
2284 @code{dllimport} attribute can be used as a constant initializer; in
2285 this case, the address of a stub function in the import lib is
2286 referenced. On Microsoft Windows targets, the attribute can be disabled
2287 for functions by setting the @option{-mnop-fun-dllimport} flag.
2290 @cindex eight bit data on the H8/300, H8/300H, and H8S
2291 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2292 variable should be placed into the eight bit data section.
2293 The compiler will generate more efficient code for certain operations
2294 on data in the eight bit data area. Note the eight bit data area is limited to
2297 You must use GAS and GLD from GNU binutils version 2.7 or later for
2298 this attribute to work correctly.
2300 @item exception_handler
2301 @cindex exception handler functions on the Blackfin processor
2302 Use this attribute on the Blackfin to indicate that the specified function
2303 is an exception handler. The compiler will generate function entry and
2304 exit sequences suitable for use in an exception handler when this
2305 attribute is present.
2307 @item externally_visible
2308 @cindex @code{externally_visible} attribute.
2309 This attribute, attached to a global variable or function, nullifies
2310 the effect of the @option{-fwhole-program} command-line option, so the
2311 object remains visible outside the current compilation unit. If @option{-fwhole-program} is used together with @option{-flto} and @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
2314 @cindex functions which handle memory bank switching
2315 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2316 use a calling convention that takes care of switching memory banks when
2317 entering and leaving a function. This calling convention is also the
2318 default when using the @option{-mlong-calls} option.
2320 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2321 to call and return from a function.
2323 On 68HC11 the compiler will generate a sequence of instructions
2324 to invoke a board-specific routine to switch the memory bank and call the
2325 real function. The board-specific routine simulates a @code{call}.
2326 At the end of a function, it will jump to a board-specific routine
2327 instead of using @code{rts}. The board-specific return routine simulates
2330 On MeP targets this causes the compiler to use a calling convention
2331 which assumes the called function is too far away for the built-in
2334 @item fast_interrupt
2335 @cindex interrupt handler functions
2336 Use this attribute on the M32C and RX ports to indicate that the specified
2337 function is a fast interrupt handler. This is just like the
2338 @code{interrupt} attribute, except that @code{freit} is used to return
2339 instead of @code{reit}.
2342 @cindex functions that pop the argument stack on the 386
2343 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2344 pass the first argument (if of integral type) in the register ECX and
2345 the second argument (if of integral type) in the register EDX@. Subsequent
2346 and other typed arguments are passed on the stack. The called function will
2347 pop the arguments off the stack. If the number of arguments is variable all
2348 arguments are pushed on the stack.
2351 @cindex functions that pop the argument stack on the 386
2352 On the Intel 386, the @code{thiscall} attribute causes the compiler to
2353 pass the first argument (if of integral type) in the register ECX.
2354 Subsequent and other typed arguments are passed on the stack. The called
2355 function will pop the arguments off the stack.
2356 If the number of arguments is variable all arguments are pushed on the
2358 The @code{thiscall} attribute is intended for C++ non-static member functions.
2359 As gcc extension this calling convention can be used for C-functions
2360 and for static member methods.
2362 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2363 @cindex @code{format} function attribute
2365 The @code{format} attribute specifies that a function takes @code{printf},
2366 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2367 should be type-checked against a format string. For example, the
2372 my_printf (void *my_object, const char *my_format, ...)
2373 __attribute__ ((format (printf, 2, 3)));
2377 causes the compiler to check the arguments in calls to @code{my_printf}
2378 for consistency with the @code{printf} style format string argument
2381 The parameter @var{archetype} determines how the format string is
2382 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2383 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2384 @code{strfmon}. (You can also use @code{__printf__},
2385 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2386 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2387 @code{ms_strftime} are also present.
2388 @var{archtype} values such as @code{printf} refer to the formats accepted
2389 by the system's C run-time library, while @code{gnu_} values always refer
2390 to the formats accepted by the GNU C Library. On Microsoft Windows
2391 targets, @code{ms_} values refer to the formats accepted by the
2392 @file{msvcrt.dll} library.
2393 The parameter @var{string-index}
2394 specifies which argument is the format string argument (starting
2395 from 1), while @var{first-to-check} is the number of the first
2396 argument to check against the format string. For functions
2397 where the arguments are not available to be checked (such as
2398 @code{vprintf}), specify the third parameter as zero. In this case the
2399 compiler only checks the format string for consistency. For
2400 @code{strftime} formats, the third parameter is required to be zero.
2401 Since non-static C++ methods have an implicit @code{this} argument, the
2402 arguments of such methods should be counted from two, not one, when
2403 giving values for @var{string-index} and @var{first-to-check}.
2405 In the example above, the format string (@code{my_format}) is the second
2406 argument of the function @code{my_print}, and the arguments to check
2407 start with the third argument, so the correct parameters for the format
2408 attribute are 2 and 3.
2410 @opindex ffreestanding
2411 @opindex fno-builtin
2412 The @code{format} attribute allows you to identify your own functions
2413 which take format strings as arguments, so that GCC can check the
2414 calls to these functions for errors. The compiler always (unless
2415 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2416 for the standard library functions @code{printf}, @code{fprintf},
2417 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2418 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2419 warnings are requested (using @option{-Wformat}), so there is no need to
2420 modify the header file @file{stdio.h}. In C99 mode, the functions
2421 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2422 @code{vsscanf} are also checked. Except in strictly conforming C
2423 standard modes, the X/Open function @code{strfmon} is also checked as
2424 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2425 @xref{C Dialect Options,,Options Controlling C Dialect}.
2427 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2428 recognized in the same context. Declarations including these format attributes
2429 will be parsed for correct syntax, however the result of checking of such format
2430 strings is not yet defined, and will not be carried out by this version of the
2433 The target may also provide additional types of format checks.
2434 @xref{Target Format Checks,,Format Checks Specific to Particular
2437 @item format_arg (@var{string-index})
2438 @cindex @code{format_arg} function attribute
2439 @opindex Wformat-nonliteral
2440 The @code{format_arg} attribute specifies that a function takes a format
2441 string for a @code{printf}, @code{scanf}, @code{strftime} or
2442 @code{strfmon} style function and modifies it (for example, to translate
2443 it into another language), so the result can be passed to a
2444 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2445 function (with the remaining arguments to the format function the same
2446 as they would have been for the unmodified string). For example, the
2451 my_dgettext (char *my_domain, const char *my_format)
2452 __attribute__ ((format_arg (2)));
2456 causes the compiler to check the arguments in calls to a @code{printf},
2457 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2458 format string argument is a call to the @code{my_dgettext} function, for
2459 consistency with the format string argument @code{my_format}. If the
2460 @code{format_arg} attribute had not been specified, all the compiler
2461 could tell in such calls to format functions would be that the format
2462 string argument is not constant; this would generate a warning when
2463 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2464 without the attribute.
2466 The parameter @var{string-index} specifies which argument is the format
2467 string argument (starting from one). Since non-static C++ methods have
2468 an implicit @code{this} argument, the arguments of such methods should
2469 be counted from two.
2471 The @code{format-arg} attribute allows you to identify your own
2472 functions which modify format strings, so that GCC can check the
2473 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2474 type function whose operands are a call to one of your own function.
2475 The compiler always treats @code{gettext}, @code{dgettext}, and
2476 @code{dcgettext} in this manner except when strict ISO C support is
2477 requested by @option{-ansi} or an appropriate @option{-std} option, or
2478 @option{-ffreestanding} or @option{-fno-builtin}
2479 is used. @xref{C Dialect Options,,Options
2480 Controlling C Dialect}.
2482 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2483 @code{NSString} reference for compatibility with the @code{format} attribute
2486 The target may also allow additional types in @code{format-arg} attributes.
2487 @xref{Target Format Checks,,Format Checks Specific to Particular
2490 @item function_vector
2491 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2492 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2493 function should be called through the function vector. Calling a
2494 function through the function vector will reduce code size, however;
2495 the function vector has a limited size (maximum 128 entries on the H8/300
2496 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2498 In SH2A target, this attribute declares a function to be called using the
2499 TBR relative addressing mode. The argument to this attribute is the entry
2500 number of the same function in a vector table containing all the TBR
2501 relative addressable functions. For the successful jump, register TBR
2502 should contain the start address of this TBR relative vector table.
2503 In the startup routine of the user application, user needs to care of this
2504 TBR register initialization. The TBR relative vector table can have at
2505 max 256 function entries. The jumps to these functions will be generated
2506 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2507 You must use GAS and GLD from GNU binutils version 2.7 or later for
2508 this attribute to work correctly.
2510 Please refer the example of M16C target, to see the use of this
2511 attribute while declaring a function,
2513 In an application, for a function being called once, this attribute will
2514 save at least 8 bytes of code; and if other successive calls are being
2515 made to the same function, it will save 2 bytes of code per each of these
2518 On M16C/M32C targets, the @code{function_vector} attribute declares a
2519 special page subroutine call function. Use of this attribute reduces
2520 the code size by 2 bytes for each call generated to the
2521 subroutine. The argument to the attribute is the vector number entry
2522 from the special page vector table which contains the 16 low-order
2523 bits of the subroutine's entry address. Each vector table has special
2524 page number (18 to 255) which are used in @code{jsrs} instruction.
2525 Jump addresses of the routines are generated by adding 0x0F0000 (in
2526 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2527 byte addresses set in the vector table. Therefore you need to ensure
2528 that all the special page vector routines should get mapped within the
2529 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2532 In the following example 2 bytes will be saved for each call to
2533 function @code{foo}.
2536 void foo (void) __attribute__((function_vector(0x18)));
2547 If functions are defined in one file and are called in another file,
2548 then be sure to write this declaration in both files.
2550 This attribute is ignored for R8C target.
2553 @cindex interrupt handler functions
2554 Use this attribute on the ARM, AVR, M32C, M32R/D, m68k, MeP, MIPS,
2555 RX and Xstormy16 ports to indicate that the specified function is an
2556 interrupt handler. The compiler will generate function entry and exit
2557 sequences suitable for use in an interrupt handler when this attribute
2560 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, MicroBlaze,
2561 and SH processors can be specified via the @code{interrupt_handler} attribute.
2563 Note, on the AVR, interrupts will be enabled inside the function.
2565 Note, for the ARM, you can specify the kind of interrupt to be handled by
2566 adding an optional parameter to the interrupt attribute like this:
2569 void f () __attribute__ ((interrupt ("IRQ")));
2572 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2574 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2575 may be called with a word aligned stack pointer.
2577 On MIPS targets, you can use the following attributes to modify the behavior
2578 of an interrupt handler:
2580 @item use_shadow_register_set
2581 @cindex @code{use_shadow_register_set} attribute
2582 Assume that the handler uses a shadow register set, instead of
2583 the main general-purpose registers.
2585 @item keep_interrupts_masked
2586 @cindex @code{keep_interrupts_masked} attribute
2587 Keep interrupts masked for the whole function. Without this attribute,
2588 GCC tries to reenable interrupts for as much of the function as it can.
2590 @item use_debug_exception_return
2591 @cindex @code{use_debug_exception_return} attribute
2592 Return using the @code{deret} instruction. Interrupt handlers that don't
2593 have this attribute return using @code{eret} instead.
2596 You can use any combination of these attributes, as shown below:
2598 void __attribute__ ((interrupt)) v0 ();
2599 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
2600 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
2601 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
2602 void __attribute__ ((interrupt, use_shadow_register_set,
2603 keep_interrupts_masked)) v4 ();
2604 void __attribute__ ((interrupt, use_shadow_register_set,
2605 use_debug_exception_return)) v5 ();
2606 void __attribute__ ((interrupt, keep_interrupts_masked,
2607 use_debug_exception_return)) v6 ();
2608 void __attribute__ ((interrupt, use_shadow_register_set,
2609 keep_interrupts_masked,
2610 use_debug_exception_return)) v7 ();
2613 @item ifunc ("@var{resolver}")
2614 @cindex @code{ifunc} attribute
2615 The @code{ifunc} attribute is used to mark a function as an indirect
2616 function using the STT_GNU_IFUNC symbol type extension to the ELF
2617 standard. This allows the resolution of the symbol value to be
2618 determined dynamically at load time, and an optimized version of the
2619 routine can be selected for the particular processor or other system
2620 characteristics determined then. To use this attribute, first define
2621 the implementation functions available, and a resolver function that
2622 returns a pointer to the selected implementation function. The
2623 implementation functions' declarations must match the API of the
2624 function being implemented, the resolver's declaration is be a
2625 function returning pointer to void function returning void:
2628 void *my_memcpy (void *dst, const void *src, size_t len)
2633 static void (*resolve_memcpy (void)) (void)
2635 return my_memcpy; // we'll just always select this routine
2639 The exported header file declaring the function the user calls would
2643 extern void *memcpy (void *, const void *, size_t);
2646 allowing the user to call this as a regular function, unaware of the
2647 implementation. Finally, the indirect function needs to be defined in
2648 the same translation unit as the resolver function:
2651 void *memcpy (void *, const void *, size_t)
2652 __attribute__ ((ifunc ("resolve_memcpy")));
2655 Indirect functions cannot be weak, and require a recent binutils (at
2656 least version 2.20.1), and GNU C library (at least version 2.11.1).
2658 @item interrupt_handler
2659 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2660 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2661 indicate that the specified function is an interrupt handler. The compiler
2662 will generate function entry and exit sequences suitable for use in an
2663 interrupt handler when this attribute is present.
2665 @item interrupt_thread
2666 @cindex interrupt thread functions on fido
2667 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2668 that the specified function is an interrupt handler that is designed
2669 to run as a thread. The compiler omits generate prologue/epilogue
2670 sequences and replaces the return instruction with a @code{sleep}
2671 instruction. This attribute is available only on fido.
2674 @cindex interrupt service routines on ARM
2675 Use this attribute on ARM to write Interrupt Service Routines. This is an
2676 alias to the @code{interrupt} attribute above.
2679 @cindex User stack pointer in interrupts on the Blackfin
2680 When used together with @code{interrupt_handler}, @code{exception_handler}
2681 or @code{nmi_handler}, code will be generated to load the stack pointer
2682 from the USP register in the function prologue.
2685 @cindex @code{l1_text} function attribute
2686 This attribute specifies a function to be placed into L1 Instruction
2687 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2688 With @option{-mfdpic}, function calls with a such function as the callee
2689 or caller will use inlined PLT.
2692 @cindex @code{l2} function attribute
2693 On the Blackfin, this attribute specifies a function to be placed into L2
2694 SRAM. The function will be put into a specific section named
2695 @code{.l1.text}. With @option{-mfdpic}, callers of such functions will use
2699 @cindex @code{leaf} function attribute
2700 Calls to external functions with this attribute must return to the current
2701 compilation unit only by return or by exception handling. In particular, leaf
2702 functions are not allowed to call callback function passed to it from the current
2703 compilation unit or directly call functions exported by the unit or longjmp
2704 into the unit. Leaf function might still call functions from other compilation
2705 units and thus they are not necessarily leaf in the sense that they contain no
2706 function calls at all.
2708 The attribute is intended for library functions to improve dataflow analysis.
2709 The compiler takes the hint that any data not escaping the current compilation unit can
2710 not be used or modified by the leaf function. For example, the @code{sin} function
2711 is a leaf function, but @code{qsort} is not.
2713 Note that leaf functions might invoke signals and signal handlers might be
2714 defined in the current compilation unit and use static variables. The only
2715 compliant way to write such a signal handler is to declare such variables
2718 The attribute has no effect on functions defined within the current compilation
2719 unit. This is to allow easy merging of multiple compilation units into one,
2720 for example, by using the link time optimization. For this reason the
2721 attribute is not allowed on types to annotate indirect calls.
2723 @item long_call/short_call
2724 @cindex indirect calls on ARM
2725 This attribute specifies how a particular function is called on
2726 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2727 command-line switch and @code{#pragma long_calls} settings. The
2728 @code{long_call} attribute indicates that the function might be far
2729 away from the call site and require a different (more expensive)
2730 calling sequence. The @code{short_call} attribute always places
2731 the offset to the function from the call site into the @samp{BL}
2732 instruction directly.
2734 @item longcall/shortcall
2735 @cindex functions called via pointer on the RS/6000 and PowerPC
2736 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2737 indicates that the function might be far away from the call site and
2738 require a different (more expensive) calling sequence. The
2739 @code{shortcall} attribute indicates that the function is always close
2740 enough for the shorter calling sequence to be used. These attributes
2741 override both the @option{-mlongcall} switch and, on the RS/6000 and
2742 PowerPC, the @code{#pragma longcall} setting.
2744 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2745 calls are necessary.
2747 @item long_call/near/far
2748 @cindex indirect calls on MIPS
2749 These attributes specify how a particular function is called on MIPS@.
2750 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2751 command-line switch. The @code{long_call} and @code{far} attributes are
2752 synonyms, and cause the compiler to always call
2753 the function by first loading its address into a register, and then using
2754 the contents of that register. The @code{near} attribute has the opposite
2755 effect; it specifies that non-PIC calls should be made using the more
2756 efficient @code{jal} instruction.
2759 @cindex @code{malloc} attribute
2760 The @code{malloc} attribute is used to tell the compiler that a function
2761 may be treated as if any non-@code{NULL} pointer it returns cannot
2762 alias any other pointer valid when the function returns.
2763 This will often improve optimization.
2764 Standard functions with this property include @code{malloc} and
2765 @code{calloc}. @code{realloc}-like functions have this property as
2766 long as the old pointer is never referred to (including comparing it
2767 to the new pointer) after the function returns a non-@code{NULL}
2770 @item mips16/nomips16
2771 @cindex @code{mips16} attribute
2772 @cindex @code{nomips16} attribute
2774 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2775 function attributes to locally select or turn off MIPS16 code generation.
2776 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2777 while MIPS16 code generation is disabled for functions with the
2778 @code{nomips16} attribute. These attributes override the
2779 @option{-mips16} and @option{-mno-mips16} options on the command line
2780 (@pxref{MIPS Options}).
2782 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2783 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2784 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2785 may interact badly with some GCC extensions such as @code{__builtin_apply}
2786 (@pxref{Constructing Calls}).
2788 @item model (@var{model-name})
2789 @cindex function addressability on the M32R/D
2790 @cindex variable addressability on the IA-64
2792 On the M32R/D, use this attribute to set the addressability of an
2793 object, and of the code generated for a function. The identifier
2794 @var{model-name} is one of @code{small}, @code{medium}, or
2795 @code{large}, representing each of the code models.
2797 Small model objects live in the lower 16MB of memory (so that their
2798 addresses can be loaded with the @code{ld24} instruction), and are
2799 callable with the @code{bl} instruction.
2801 Medium model objects may live anywhere in the 32-bit address space (the
2802 compiler will generate @code{seth/add3} instructions to load their addresses),
2803 and are callable with the @code{bl} instruction.
2805 Large model objects may live anywhere in the 32-bit address space (the
2806 compiler will generate @code{seth/add3} instructions to load their addresses),
2807 and may not be reachable with the @code{bl} instruction (the compiler will
2808 generate the much slower @code{seth/add3/jl} instruction sequence).
2810 On IA-64, use this attribute to set the addressability of an object.
2811 At present, the only supported identifier for @var{model-name} is
2812 @code{small}, indicating addressability via ``small'' (22-bit)
2813 addresses (so that their addresses can be loaded with the @code{addl}
2814 instruction). Caveat: such addressing is by definition not position
2815 independent and hence this attribute must not be used for objects
2816 defined by shared libraries.
2818 @item ms_abi/sysv_abi
2819 @cindex @code{ms_abi} attribute
2820 @cindex @code{sysv_abi} attribute
2822 On 64-bit x86_64-*-* targets, you can use an ABI attribute to indicate
2823 which calling convention should be used for a function. The @code{ms_abi}
2824 attribute tells the compiler to use the Microsoft ABI, while the
2825 @code{sysv_abi} attribute tells the compiler to use the ABI used on
2826 GNU/Linux and other systems. The default is to use the Microsoft ABI
2827 when targeting Windows. On all other systems, the default is the AMD ABI.
2829 Note, the @code{ms_abi} attribute for Windows targets currently requires
2830 the @option{-maccumulate-outgoing-args} option.
2832 @item callee_pop_aggregate_return (@var{number})
2833 @cindex @code{callee_pop_aggregate_return} attribute
2835 On 32-bit i?86-*-* targets, you can control by those attribute for
2836 aggregate return in memory, if the caller is responsible to pop the hidden
2837 pointer together with the rest of the arguments - @var{number} equal to
2838 zero -, or if the callee is responsible to pop hidden pointer - @var{number}
2841 For i?86-netware, the caller pops the stack for the hidden arguments pointing
2842 to aggregate return value. This differs from the default i386 ABI which assumes
2843 that the callee pops the stack for hidden pointer.
2845 @item ms_hook_prologue
2846 @cindex @code{ms_hook_prologue} attribute
2848 On 32 bit i[34567]86-*-* targets and 64 bit x86_64-*-* targets, you can use
2849 this function attribute to make gcc generate the "hot-patching" function
2850 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
2854 @cindex function without a prologue/epilogue code
2855 Use this attribute on the ARM, AVR, MCORE, RX and SPU ports to indicate that
2856 the specified function does not need prologue/epilogue sequences generated by
2857 the compiler. It is up to the programmer to provide these sequences. The
2858 only statements that can be safely included in naked functions are
2859 @code{asm} statements that do not have operands. All other statements,
2860 including declarations of local variables, @code{if} statements, and so
2861 forth, should be avoided. Naked functions should be used to implement the
2862 body of an assembly function, while allowing the compiler to construct
2863 the requisite function declaration for the assembler.
2866 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2867 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2868 use the normal calling convention based on @code{jsr} and @code{rts}.
2869 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2872 On MeP targets this attribute causes the compiler to assume the called
2873 function is close enough to use the normal calling convention,
2874 overriding the @code{-mtf} command line option.
2877 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2878 Use this attribute together with @code{interrupt_handler},
2879 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2880 entry code should enable nested interrupts or exceptions.
2883 @cindex NMI handler functions on the Blackfin processor
2884 Use this attribute on the Blackfin to indicate that the specified function
2885 is an NMI handler. The compiler will generate function entry and
2886 exit sequences suitable for use in an NMI handler when this
2887 attribute is present.
2889 @item no_instrument_function
2890 @cindex @code{no_instrument_function} function attribute
2891 @opindex finstrument-functions
2892 If @option{-finstrument-functions} is given, profiling function calls will
2893 be generated at entry and exit of most user-compiled functions.
2894 Functions with this attribute will not be so instrumented.
2896 @item no_split_stack
2897 @cindex @code{no_split_stack} function attribute
2898 @opindex fsplit-stack
2899 If @option{-fsplit-stack} is given, functions will have a small
2900 prologue which decides whether to split the stack. Functions with the
2901 @code{no_split_stack} attribute will not have that prologue, and thus
2902 may run with only a small amount of stack space available.
2905 @cindex @code{noinline} function attribute
2906 This function attribute prevents a function from being considered for
2908 @c Don't enumerate the optimizations by name here; we try to be
2909 @c future-compatible with this mechanism.
2910 If the function does not have side-effects, there are optimizations
2911 other than inlining that causes function calls to be optimized away,
2912 although the function call is live. To keep such calls from being
2917 (@pxref{Extended Asm}) in the called function, to serve as a special
2921 @cindex @code{noclone} function attribute
2922 This function attribute prevents a function from being considered for
2923 cloning - a mechanism which produces specialized copies of functions
2924 and which is (currently) performed by interprocedural constant
2927 @item nonnull (@var{arg-index}, @dots{})
2928 @cindex @code{nonnull} function attribute
2929 The @code{nonnull} attribute specifies that some function parameters should
2930 be non-null pointers. For instance, the declaration:
2934 my_memcpy (void *dest, const void *src, size_t len)
2935 __attribute__((nonnull (1, 2)));
2939 causes the compiler to check that, in calls to @code{my_memcpy},
2940 arguments @var{dest} and @var{src} are non-null. If the compiler
2941 determines that a null pointer is passed in an argument slot marked
2942 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2943 is issued. The compiler may also choose to make optimizations based
2944 on the knowledge that certain function arguments will not be null.
2946 If no argument index list is given to the @code{nonnull} attribute,
2947 all pointer arguments are marked as non-null. To illustrate, the
2948 following declaration is equivalent to the previous example:
2952 my_memcpy (void *dest, const void *src, size_t len)
2953 __attribute__((nonnull));
2957 @cindex @code{noreturn} function attribute
2958 A few standard library functions, such as @code{abort} and @code{exit},
2959 cannot return. GCC knows this automatically. Some programs define
2960 their own functions that never return. You can declare them
2961 @code{noreturn} to tell the compiler this fact. For example,
2965 void fatal () __attribute__ ((noreturn));
2968 fatal (/* @r{@dots{}} */)
2970 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2976 The @code{noreturn} keyword tells the compiler to assume that
2977 @code{fatal} cannot return. It can then optimize without regard to what
2978 would happen if @code{fatal} ever did return. This makes slightly
2979 better code. More importantly, it helps avoid spurious warnings of
2980 uninitialized variables.
2982 The @code{noreturn} keyword does not affect the exceptional path when that
2983 applies: a @code{noreturn}-marked function may still return to the caller
2984 by throwing an exception or calling @code{longjmp}.
2986 Do not assume that registers saved by the calling function are
2987 restored before calling the @code{noreturn} function.
2989 It does not make sense for a @code{noreturn} function to have a return
2990 type other than @code{void}.
2992 The attribute @code{noreturn} is not implemented in GCC versions
2993 earlier than 2.5. An alternative way to declare that a function does
2994 not return, which works in the current version and in some older
2995 versions, is as follows:
2998 typedef void voidfn ();
3000 volatile voidfn fatal;
3003 This approach does not work in GNU C++.
3006 @cindex @code{nothrow} function attribute
3007 The @code{nothrow} attribute is used to inform the compiler that a
3008 function cannot throw an exception. For example, most functions in
3009 the standard C library can be guaranteed not to throw an exception
3010 with the notable exceptions of @code{qsort} and @code{bsearch} that
3011 take function pointer arguments. The @code{nothrow} attribute is not
3012 implemented in GCC versions earlier than 3.3.
3015 @cindex @code{optimize} function attribute
3016 The @code{optimize} attribute is used to specify that a function is to
3017 be compiled with different optimization options than specified on the
3018 command line. Arguments can either be numbers or strings. Numbers
3019 are assumed to be an optimization level. Strings that begin with
3020 @code{O} are assumed to be an optimization option, while other options
3021 are assumed to be used with a @code{-f} prefix. You can also use the
3022 @samp{#pragma GCC optimize} pragma to set the optimization options
3023 that affect more than one function.
3024 @xref{Function Specific Option Pragmas}, for details about the
3025 @samp{#pragma GCC optimize} pragma.
3027 This can be used for instance to have frequently executed functions
3028 compiled with more aggressive optimization options that produce faster
3029 and larger code, while other functions can be called with less
3033 @cindex @code{pcs} function attribute
3035 The @code{pcs} attribute can be used to control the calling convention
3036 used for a function on ARM. The attribute takes an argument that specifies
3037 the calling convention to use.
3039 When compiling using the AAPCS ABI (or a variant of that) then valid
3040 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3041 order to use a variant other than @code{"aapcs"} then the compiler must
3042 be permitted to use the appropriate co-processor registers (i.e., the
3043 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3047 /* Argument passed in r0, and result returned in r0+r1. */
3048 double f2d (float) __attribute__((pcs("aapcs")));
3051 Variadic functions always use the @code{"aapcs"} calling convention and
3052 the compiler will reject attempts to specify an alternative.
3055 @cindex @code{pure} function attribute
3056 Many functions have no effects except the return value and their
3057 return value depends only on the parameters and/or global variables.
3058 Such a function can be subject
3059 to common subexpression elimination and loop optimization just as an
3060 arithmetic operator would be. These functions should be declared
3061 with the attribute @code{pure}. For example,
3064 int square (int) __attribute__ ((pure));
3068 says that the hypothetical function @code{square} is safe to call
3069 fewer times than the program says.
3071 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3072 Interesting non-pure functions are functions with infinite loops or those
3073 depending on volatile memory or other system resource, that may change between
3074 two consecutive calls (such as @code{feof} in a multithreading environment).
3076 The attribute @code{pure} is not implemented in GCC versions earlier
3080 @cindex @code{hot} function attribute
3081 The @code{hot} attribute is used to inform the compiler that a function is a
3082 hot spot of the compiled program. The function is optimized more aggressively
3083 and on many target it is placed into special subsection of the text section so
3084 all hot functions appears close together improving locality.
3086 When profile feedback is available, via @option{-fprofile-use}, hot functions
3087 are automatically detected and this attribute is ignored.
3089 The @code{hot} attribute is not implemented in GCC versions earlier
3093 @cindex @code{cold} function attribute
3094 The @code{cold} attribute is used to inform the compiler that a function is
3095 unlikely executed. The function is optimized for size rather than speed and on
3096 many targets it is placed into special subsection of the text section so all
3097 cold functions appears close together improving code locality of non-cold parts
3098 of program. The paths leading to call of cold functions within code are marked
3099 as unlikely by the branch prediction mechanism. It is thus useful to mark
3100 functions used to handle unlikely conditions, such as @code{perror}, as cold to
3101 improve optimization of hot functions that do call marked functions in rare
3104 When profile feedback is available, via @option{-fprofile-use}, hot functions
3105 are automatically detected and this attribute is ignored.
3107 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
3109 @item regparm (@var{number})
3110 @cindex @code{regparm} attribute
3111 @cindex functions that are passed arguments in registers on the 386
3112 On the Intel 386, the @code{regparm} attribute causes the compiler to
3113 pass arguments number one to @var{number} if they are of integral type
3114 in registers EAX, EDX, and ECX instead of on the stack. Functions that
3115 take a variable number of arguments will continue to be passed all of their
3116 arguments on the stack.
3118 Beware that on some ELF systems this attribute is unsuitable for
3119 global functions in shared libraries with lazy binding (which is the
3120 default). Lazy binding will send the first call via resolving code in
3121 the loader, which might assume EAX, EDX and ECX can be clobbered, as
3122 per the standard calling conventions. Solaris 8 is affected by this.
3123 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
3124 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
3125 disabled with the linker or the loader if desired, to avoid the
3129 @cindex @code{sseregparm} attribute
3130 On the Intel 386 with SSE support, the @code{sseregparm} attribute
3131 causes the compiler to pass up to 3 floating point arguments in
3132 SSE registers instead of on the stack. Functions that take a
3133 variable number of arguments will continue to pass all of their
3134 floating point arguments on the stack.
3136 @item force_align_arg_pointer
3137 @cindex @code{force_align_arg_pointer} attribute
3138 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
3139 applied to individual function definitions, generating an alternate
3140 prologue and epilogue that realigns the runtime stack if necessary.
3141 This supports mixing legacy codes that run with a 4-byte aligned stack
3142 with modern codes that keep a 16-byte stack for SSE compatibility.
3145 @cindex @code{resbank} attribute
3146 On the SH2A target, this attribute enables the high-speed register
3147 saving and restoration using a register bank for @code{interrupt_handler}
3148 routines. Saving to the bank is performed automatically after the CPU
3149 accepts an interrupt that uses a register bank.
3151 The nineteen 32-bit registers comprising general register R0 to R14,
3152 control register GBR, and system registers MACH, MACL, and PR and the
3153 vector table address offset are saved into a register bank. Register
3154 banks are stacked in first-in last-out (FILO) sequence. Restoration
3155 from the bank is executed by issuing a RESBANK instruction.
3158 @cindex @code{returns_twice} attribute
3159 The @code{returns_twice} attribute tells the compiler that a function may
3160 return more than one time. The compiler will ensure that all registers
3161 are dead before calling such a function and will emit a warning about
3162 the variables that may be clobbered after the second return from the
3163 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3164 The @code{longjmp}-like counterpart of such function, if any, might need
3165 to be marked with the @code{noreturn} attribute.
3168 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
3169 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
3170 all registers except the stack pointer should be saved in the prologue
3171 regardless of whether they are used or not.
3173 @item save_volatiles
3174 @cindex save volatile registers on the MicroBlaze
3175 Use this attribute on the MicroBlaze to indicate that the function is
3176 an interrupt handler. All volatile registers (in addition to non-volatile
3177 registers) will be saved in the function prologue. If the function is a leaf
3178 function, only volatiles used by the function are saved. A normal function
3179 return is generated instead of a return from interrupt.
3181 @item section ("@var{section-name}")
3182 @cindex @code{section} function attribute
3183 Normally, the compiler places the code it generates in the @code{text} section.
3184 Sometimes, however, you need additional sections, or you need certain
3185 particular functions to appear in special sections. The @code{section}
3186 attribute specifies that a function lives in a particular section.
3187 For example, the declaration:
3190 extern void foobar (void) __attribute__ ((section ("bar")));
3194 puts the function @code{foobar} in the @code{bar} section.
3196 Some file formats do not support arbitrary sections so the @code{section}
3197 attribute is not available on all platforms.
3198 If you need to map the entire contents of a module to a particular
3199 section, consider using the facilities of the linker instead.
3202 @cindex @code{sentinel} function attribute
3203 This function attribute ensures that a parameter in a function call is
3204 an explicit @code{NULL}. The attribute is only valid on variadic
3205 functions. By default, the sentinel is located at position zero, the
3206 last parameter of the function call. If an optional integer position
3207 argument P is supplied to the attribute, the sentinel must be located at
3208 position P counting backwards from the end of the argument list.
3211 __attribute__ ((sentinel))
3213 __attribute__ ((sentinel(0)))
3216 The attribute is automatically set with a position of 0 for the built-in
3217 functions @code{execl} and @code{execlp}. The built-in function
3218 @code{execle} has the attribute set with a position of 1.
3220 A valid @code{NULL} in this context is defined as zero with any pointer
3221 type. If your system defines the @code{NULL} macro with an integer type
3222 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3223 with a copy that redefines NULL appropriately.
3225 The warnings for missing or incorrect sentinels are enabled with
3229 See long_call/short_call.
3232 See longcall/shortcall.
3235 @cindex signal handler functions on the AVR processors
3236 Use this attribute on the AVR to indicate that the specified
3237 function is a signal handler. The compiler will generate function
3238 entry and exit sequences suitable for use in a signal handler when this
3239 attribute is present. Interrupts will be disabled inside the function.
3242 Use this attribute on the SH to indicate an @code{interrupt_handler}
3243 function should switch to an alternate stack. It expects a string
3244 argument that names a global variable holding the address of the
3249 void f () __attribute__ ((interrupt_handler,
3250 sp_switch ("alt_stack")));
3254 @cindex functions that pop the argument stack on the 386
3255 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3256 assume that the called function will pop off the stack space used to
3257 pass arguments, unless it takes a variable number of arguments.
3259 @item syscall_linkage
3260 @cindex @code{syscall_linkage} attribute
3261 This attribute is used to modify the IA64 calling convention by marking
3262 all input registers as live at all function exits. This makes it possible
3263 to restart a system call after an interrupt without having to save/restore
3264 the input registers. This also prevents kernel data from leaking into
3268 @cindex @code{target} function attribute
3269 The @code{target} attribute is used to specify that a function is to
3270 be compiled with different target options than specified on the
3271 command line. This can be used for instance to have functions
3272 compiled with a different ISA (instruction set architecture) than the
3273 default. You can also use the @samp{#pragma GCC target} pragma to set
3274 more than one function to be compiled with specific target options.
3275 @xref{Function Specific Option Pragmas}, for details about the
3276 @samp{#pragma GCC target} pragma.
3278 For instance on a 386, you could compile one function with
3279 @code{target("sse4.1,arch=core2")} and another with
3280 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
3281 compiling the first function with @option{-msse4.1} and
3282 @option{-march=core2} options, and the second function with
3283 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
3284 user to make sure that a function is only invoked on a machine that
3285 supports the particular ISA it was compiled for (for example by using
3286 @code{cpuid} on 386 to determine what feature bits and architecture
3290 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3291 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3294 On the 386, the following options are allowed:
3299 @cindex @code{target("abm")} attribute
3300 Enable/disable the generation of the advanced bit instructions.
3304 @cindex @code{target("aes")} attribute
3305 Enable/disable the generation of the AES instructions.
3309 @cindex @code{target("mmx")} attribute
3310 Enable/disable the generation of the MMX instructions.
3314 @cindex @code{target("pclmul")} attribute
3315 Enable/disable the generation of the PCLMUL instructions.
3319 @cindex @code{target("popcnt")} attribute
3320 Enable/disable the generation of the POPCNT instruction.
3324 @cindex @code{target("sse")} attribute
3325 Enable/disable the generation of the SSE instructions.
3329 @cindex @code{target("sse2")} attribute
3330 Enable/disable the generation of the SSE2 instructions.
3334 @cindex @code{target("sse3")} attribute
3335 Enable/disable the generation of the SSE3 instructions.
3339 @cindex @code{target("sse4")} attribute
3340 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3345 @cindex @code{target("sse4.1")} attribute
3346 Enable/disable the generation of the sse4.1 instructions.
3350 @cindex @code{target("sse4.2")} attribute
3351 Enable/disable the generation of the sse4.2 instructions.
3355 @cindex @code{target("sse4a")} attribute
3356 Enable/disable the generation of the SSE4A instructions.
3360 @cindex @code{target("fma4")} attribute
3361 Enable/disable the generation of the FMA4 instructions.
3365 @cindex @code{target("xop")} attribute
3366 Enable/disable the generation of the XOP instructions.
3370 @cindex @code{target("lwp")} attribute
3371 Enable/disable the generation of the LWP instructions.
3375 @cindex @code{target("ssse3")} attribute
3376 Enable/disable the generation of the SSSE3 instructions.
3380 @cindex @code{target("cld")} attribute
3381 Enable/disable the generation of the CLD before string moves.
3383 @item fancy-math-387
3384 @itemx no-fancy-math-387
3385 @cindex @code{target("fancy-math-387")} attribute
3386 Enable/disable the generation of the @code{sin}, @code{cos}, and
3387 @code{sqrt} instructions on the 387 floating point unit.
3390 @itemx no-fused-madd
3391 @cindex @code{target("fused-madd")} attribute
3392 Enable/disable the generation of the fused multiply/add instructions.
3396 @cindex @code{target("ieee-fp")} attribute
3397 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3399 @item inline-all-stringops
3400 @itemx no-inline-all-stringops
3401 @cindex @code{target("inline-all-stringops")} attribute
3402 Enable/disable inlining of string operations.
3404 @item inline-stringops-dynamically
3405 @itemx no-inline-stringops-dynamically
3406 @cindex @code{target("inline-stringops-dynamically")} attribute
3407 Enable/disable the generation of the inline code to do small string
3408 operations and calling the library routines for large operations.
3410 @item align-stringops
3411 @itemx no-align-stringops
3412 @cindex @code{target("align-stringops")} attribute
3413 Do/do not align destination of inlined string operations.
3417 @cindex @code{target("recip")} attribute
3418 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3419 instructions followed an additional Newton-Raphson step instead of
3420 doing a floating point division.
3422 @item arch=@var{ARCH}
3423 @cindex @code{target("arch=@var{ARCH}")} attribute
3424 Specify the architecture to generate code for in compiling the function.
3426 @item tune=@var{TUNE}
3427 @cindex @code{target("tune=@var{TUNE}")} attribute
3428 Specify the architecture to tune for in compiling the function.
3430 @item fpmath=@var{FPMATH}
3431 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3432 Specify which floating point unit to use. The
3433 @code{target("fpmath=sse,387")} option must be specified as
3434 @code{target("fpmath=sse+387")} because the comma would separate
3438 On the PowerPC, the following options are allowed:
3443 @cindex @code{target("altivec")} attribute
3444 Generate code that uses (does not use) AltiVec instructions. In
3445 32-bit code, you cannot enable Altivec instructions unless
3446 @option{-mabi=altivec} was used on the command line.
3450 @cindex @code{target("cmpb")} attribute
3451 Generate code that uses (does not use) the compare bytes instruction
3452 implemented on the POWER6 processor and other processors that support
3453 the PowerPC V2.05 architecture.
3457 @cindex @code{target("dlmzb")} attribute
3458 Generate code that uses (does not use) the string-search @samp{dlmzb}
3459 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
3460 generated by default when targetting those processors.
3464 @cindex @code{target("fprnd")} attribute
3465 Generate code that uses (does not use) the FP round to integer
3466 instructions implemented on the POWER5+ processor and other processors
3467 that support the PowerPC V2.03 architecture.
3471 @cindex @code{target("hard-dfp")} attribute
3472 Generate code that uses (does not use) the decimal floating point
3473 instructions implemented on some POWER processors.
3477 @cindex @code{target("isel")} attribute
3478 Generate code that uses (does not use) ISEL instruction.
3482 @cindex @code{target("mfcrf")} attribute
3483 Generate code that uses (does not use) the move from condition
3484 register field instruction implemented on the POWER4 processor and
3485 other processors that support the PowerPC V2.01 architecture.
3489 @cindex @code{target("mfpgpr")} attribute
3490 Generate code that uses (does not use) the FP move to/from general
3491 purpose register instructions implemented on the POWER6X processor and
3492 other processors that support the extended PowerPC V2.05 architecture.
3496 @cindex @code{target("mulhw")} attribute
3497 Generate code that uses (does not use) the half-word multiply and
3498 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
3499 These instructions are generated by default when targetting those
3504 @cindex @code{target("multiple")} attribute
3505 Generate code that uses (does not use) the load multiple word
3506 instructions and the store multiple word instructions.
3510 @cindex @code{target("update")} attribute
3511 Generate code that uses (does not use) the load or store instructions
3512 that update the base register to the address of the calculated memory
3517 @cindex @code{target("popcntb")} attribute
3518 Generate code that uses (does not use) the popcount and double
3519 precision FP reciprocal estimate instruction implemented on the POWER5
3520 processor and other processors that support the PowerPC V2.02
3525 @cindex @code{target("popcntd")} attribute
3526 Generate code that uses (does not use) the popcount instruction
3527 implemented on the POWER7 processor and other processors that support
3528 the PowerPC V2.06 architecture.
3530 @item powerpc-gfxopt
3531 @itemx no-powerpc-gfxopt
3532 @cindex @code{target("powerpc-gfxopt")} attribute
3533 Generate code that uses (does not use) the optional PowerPC
3534 architecture instructions in the Graphics group, including
3535 floating-point select.
3538 @itemx no-powerpc-gpopt
3539 @cindex @code{target("powerpc-gpopt")} attribute
3540 Generate code that uses (does not use) the optional PowerPC
3541 architecture instructions in the General Purpose group, including
3542 floating-point square root.
3544 @item recip-precision
3545 @itemx no-recip-precision
3546 @cindex @code{target("recip-precision")} attribute
3547 Assume (do not assume) that the reciprocal estimate instructions
3548 provide higher precision estimates than is mandated by the powerpc
3553 @cindex @code{target("string")} attribute
3554 Generate code that uses (does not use) the load string instructions
3555 and the store string word instructions to save multiple registers and
3556 do small block moves.
3560 @cindex @code{target("vsx")} attribute
3561 Generate code that uses (does not use) vector/scalar (VSX)
3562 instructions, and also enable the use of built-in functions that allow
3563 more direct access to the VSX instruction set. In 32-bit code, you
3564 cannot enable VSX or Altivec instructions unless
3565 @option{-mabi=altivec} was used on the command line.
3569 @cindex @code{target("friz")} attribute
3570 Generate (do not generate) the @code{friz} instruction when the
3571 @option{-funsafe-math-optimizations} option is used to optimize
3572 rounding a floating point value to 64-bit integer and back to floating
3573 point. The @code{friz} instruction does not return the same value if
3574 the floating point number is too large to fit in an integer.
3576 @item avoid-indexed-addresses
3577 @itemx no-avoid-indexed-addresses
3578 @cindex @code{target("avoid-indexed-addresses")} attribute
3579 Generate code that tries to avoid (not avoid) the use of indexed load
3580 or store instructions.
3584 @cindex @code{target("paired")} attribute
3585 Generate code that uses (does not use) the generation of PAIRED simd
3590 @cindex @code{target("longcall")} attribute
3591 Generate code that assumes (does not assume) that all calls are far
3592 away so that a longer more expensive calling sequence is required.
3595 @cindex @code{target("cpu=@var{CPU}")} attribute
3596 Specify the architecture to generate code for when compiling the
3597 function. If you select the @code{"target("cpu=power7)"} attribute when
3598 generating 32-bit code, VSX and Altivec instructions are not generated
3599 unless you use the @option{-mabi=altivec} option on the command line.
3601 @item tune=@var{TUNE}
3602 @cindex @code{target("tune=@var{TUNE}")} attribute
3603 Specify the architecture to tune for when compiling the function. If
3604 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
3605 you do specify the @code{target("cpu=@var{CPU}")} attribute,
3606 compilation will tune for the @var{CPU} architecture, and not the
3607 default tuning specified on the command line.
3610 On the 386/x86_64 and PowerPC backends, you can use either multiple
3611 strings to specify multiple options, or you can separate the option
3612 with a comma (@code{,}).
3614 On the 386/x86_64 and PowerPC backends, the inliner will not inline a
3615 function that has different target options than the caller, unless the
3616 callee has a subset of the target options of the caller. For example
3617 a function declared with @code{target("sse3")} can inline a function
3618 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
3620 The @code{target} attribute is not implemented in GCC versions earlier
3621 than 4.4 for the i386/x86_64 and 4.6 for the PowerPC backends. It is
3622 not currently implemented for other backends.
3625 @cindex tiny data section on the H8/300H and H8S
3626 Use this attribute on the H8/300H and H8S to indicate that the specified
3627 variable should be placed into the tiny data section.
3628 The compiler will generate more efficient code for loads and stores
3629 on data in the tiny data section. Note the tiny data area is limited to
3630 slightly under 32kbytes of data.
3633 Use this attribute on the SH for an @code{interrupt_handler} to return using
3634 @code{trapa} instead of @code{rte}. This attribute expects an integer
3635 argument specifying the trap number to be used.
3638 @cindex @code{unused} attribute.
3639 This attribute, attached to a function, means that the function is meant
3640 to be possibly unused. GCC will not produce a warning for this
3644 @cindex @code{used} attribute.
3645 This attribute, attached to a function, means that code must be emitted
3646 for the function even if it appears that the function is not referenced.
3647 This is useful, for example, when the function is referenced only in
3651 @cindex @code{version_id} attribute
3652 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3653 symbol to contain a version string, thus allowing for function level
3654 versioning. HP-UX system header files may use version level functioning
3655 for some system calls.
3658 extern int foo () __attribute__((version_id ("20040821")));
3661 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3663 @item visibility ("@var{visibility_type}")
3664 @cindex @code{visibility} attribute
3665 This attribute affects the linkage of the declaration to which it is attached.
3666 There are four supported @var{visibility_type} values: default,
3667 hidden, protected or internal visibility.
3670 void __attribute__ ((visibility ("protected")))
3671 f () @{ /* @r{Do something.} */; @}
3672 int i __attribute__ ((visibility ("hidden")));
3675 The possible values of @var{visibility_type} correspond to the
3676 visibility settings in the ELF gABI.
3679 @c keep this list of visibilities in alphabetical order.
3682 Default visibility is the normal case for the object file format.
3683 This value is available for the visibility attribute to override other
3684 options that may change the assumed visibility of entities.
3686 On ELF, default visibility means that the declaration is visible to other
3687 modules and, in shared libraries, means that the declared entity may be
3690 On Darwin, default visibility means that the declaration is visible to
3693 Default visibility corresponds to ``external linkage'' in the language.
3696 Hidden visibility indicates that the entity declared will have a new
3697 form of linkage, which we'll call ``hidden linkage''. Two
3698 declarations of an object with hidden linkage refer to the same object
3699 if they are in the same shared object.
3702 Internal visibility is like hidden visibility, but with additional
3703 processor specific semantics. Unless otherwise specified by the
3704 psABI, GCC defines internal visibility to mean that a function is
3705 @emph{never} called from another module. Compare this with hidden
3706 functions which, while they cannot be referenced directly by other
3707 modules, can be referenced indirectly via function pointers. By
3708 indicating that a function cannot be called from outside the module,
3709 GCC may for instance omit the load of a PIC register since it is known
3710 that the calling function loaded the correct value.
3713 Protected visibility is like default visibility except that it
3714 indicates that references within the defining module will bind to the
3715 definition in that module. That is, the declared entity cannot be
3716 overridden by another module.
3720 All visibilities are supported on many, but not all, ELF targets
3721 (supported when the assembler supports the @samp{.visibility}
3722 pseudo-op). Default visibility is supported everywhere. Hidden
3723 visibility is supported on Darwin targets.
3725 The visibility attribute should be applied only to declarations which
3726 would otherwise have external linkage. The attribute should be applied
3727 consistently, so that the same entity should not be declared with
3728 different settings of the attribute.
3730 In C++, the visibility attribute applies to types as well as functions
3731 and objects, because in C++ types have linkage. A class must not have
3732 greater visibility than its non-static data member types and bases,
3733 and class members default to the visibility of their class. Also, a
3734 declaration without explicit visibility is limited to the visibility
3737 In C++, you can mark member functions and static member variables of a
3738 class with the visibility attribute. This is useful if you know a
3739 particular method or static member variable should only be used from
3740 one shared object; then you can mark it hidden while the rest of the
3741 class has default visibility. Care must be taken to avoid breaking
3742 the One Definition Rule; for example, it is usually not useful to mark
3743 an inline method as hidden without marking the whole class as hidden.
3745 A C++ namespace declaration can also have the visibility attribute.
3746 This attribute applies only to the particular namespace body, not to
3747 other definitions of the same namespace; it is equivalent to using
3748 @samp{#pragma GCC visibility} before and after the namespace
3749 definition (@pxref{Visibility Pragmas}).
3751 In C++, if a template argument has limited visibility, this
3752 restriction is implicitly propagated to the template instantiation.
3753 Otherwise, template instantiations and specializations default to the
3754 visibility of their template.
3756 If both the template and enclosing class have explicit visibility, the
3757 visibility from the template is used.
3760 @cindex @code{vliw} attribute
3761 On MeP, the @code{vliw} attribute tells the compiler to emit
3762 instructions in VLIW mode instead of core mode. Note that this
3763 attribute is not allowed unless a VLIW coprocessor has been configured
3764 and enabled through command line options.
3766 @item warn_unused_result
3767 @cindex @code{warn_unused_result} attribute
3768 The @code{warn_unused_result} attribute causes a warning to be emitted
3769 if a caller of the function with this attribute does not use its
3770 return value. This is useful for functions where not checking
3771 the result is either a security problem or always a bug, such as
3775 int fn () __attribute__ ((warn_unused_result));
3778 if (fn () < 0) return -1;
3784 results in warning on line 5.
3787 @cindex @code{weak} attribute
3788 The @code{weak} attribute causes the declaration to be emitted as a weak
3789 symbol rather than a global. This is primarily useful in defining
3790 library functions which can be overridden in user code, though it can
3791 also be used with non-function declarations. Weak symbols are supported
3792 for ELF targets, and also for a.out targets when using the GNU assembler
3796 @itemx weakref ("@var{target}")
3797 @cindex @code{weakref} attribute
3798 The @code{weakref} attribute marks a declaration as a weak reference.
3799 Without arguments, it should be accompanied by an @code{alias} attribute
3800 naming the target symbol. Optionally, the @var{target} may be given as
3801 an argument to @code{weakref} itself. In either case, @code{weakref}
3802 implicitly marks the declaration as @code{weak}. Without a
3803 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3804 @code{weakref} is equivalent to @code{weak}.
3807 static int x() __attribute__ ((weakref ("y")));
3808 /* is equivalent to... */
3809 static int x() __attribute__ ((weak, weakref, alias ("y")));
3811 static int x() __attribute__ ((weakref));
3812 static int x() __attribute__ ((alias ("y")));
3815 A weak reference is an alias that does not by itself require a
3816 definition to be given for the target symbol. If the target symbol is
3817 only referenced through weak references, then it becomes a @code{weak}
3818 undefined symbol. If it is directly referenced, however, then such
3819 strong references prevail, and a definition will be required for the
3820 symbol, not necessarily in the same translation unit.
3822 The effect is equivalent to moving all references to the alias to a
3823 separate translation unit, renaming the alias to the aliased symbol,
3824 declaring it as weak, compiling the two separate translation units and
3825 performing a reloadable link on them.
3827 At present, a declaration to which @code{weakref} is attached can
3828 only be @code{static}.
3832 You can specify multiple attributes in a declaration by separating them
3833 by commas within the double parentheses or by immediately following an
3834 attribute declaration with another attribute declaration.
3836 @cindex @code{#pragma}, reason for not using
3837 @cindex pragma, reason for not using
3838 Some people object to the @code{__attribute__} feature, suggesting that
3839 ISO C's @code{#pragma} should be used instead. At the time
3840 @code{__attribute__} was designed, there were two reasons for not doing
3845 It is impossible to generate @code{#pragma} commands from a macro.
3848 There is no telling what the same @code{#pragma} might mean in another
3852 These two reasons applied to almost any application that might have been
3853 proposed for @code{#pragma}. It was basically a mistake to use
3854 @code{#pragma} for @emph{anything}.
3856 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
3857 to be generated from macros. In addition, a @code{#pragma GCC}
3858 namespace is now in use for GCC-specific pragmas. However, it has been
3859 found convenient to use @code{__attribute__} to achieve a natural
3860 attachment of attributes to their corresponding declarations, whereas
3861 @code{#pragma GCC} is of use for constructs that do not naturally form
3862 part of the grammar. @xref{Other Directives,,Miscellaneous
3863 Preprocessing Directives, cpp, The GNU C Preprocessor}.
3865 @node Attribute Syntax
3866 @section Attribute Syntax
3867 @cindex attribute syntax
3869 This section describes the syntax with which @code{__attribute__} may be
3870 used, and the constructs to which attribute specifiers bind, for the C
3871 language. Some details may vary for C++ and Objective-C@. Because of
3872 infelicities in the grammar for attributes, some forms described here
3873 may not be successfully parsed in all cases.
3875 There are some problems with the semantics of attributes in C++. For
3876 example, there are no manglings for attributes, although they may affect
3877 code generation, so problems may arise when attributed types are used in
3878 conjunction with templates or overloading. Similarly, @code{typeid}
3879 does not distinguish between types with different attributes. Support
3880 for attributes in C++ may be restricted in future to attributes on
3881 declarations only, but not on nested declarators.
3883 @xref{Function Attributes}, for details of the semantics of attributes
3884 applying to functions. @xref{Variable Attributes}, for details of the
3885 semantics of attributes applying to variables. @xref{Type Attributes},
3886 for details of the semantics of attributes applying to structure, union
3887 and enumerated types.
3889 An @dfn{attribute specifier} is of the form
3890 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
3891 is a possibly empty comma-separated sequence of @dfn{attributes}, where
3892 each attribute is one of the following:
3896 Empty. Empty attributes are ignored.
3899 A word (which may be an identifier such as @code{unused}, or a reserved
3900 word such as @code{const}).
3903 A word, followed by, in parentheses, parameters for the attribute.
3904 These parameters take one of the following forms:
3908 An identifier. For example, @code{mode} attributes use this form.
3911 An identifier followed by a comma and a non-empty comma-separated list
3912 of expressions. For example, @code{format} attributes use this form.
3915 A possibly empty comma-separated list of expressions. For example,
3916 @code{format_arg} attributes use this form with the list being a single
3917 integer constant expression, and @code{alias} attributes use this form
3918 with the list being a single string constant.
3922 An @dfn{attribute specifier list} is a sequence of one or more attribute
3923 specifiers, not separated by any other tokens.
3925 In GNU C, an attribute specifier list may appear after the colon following a
3926 label, other than a @code{case} or @code{default} label. The only
3927 attribute it makes sense to use after a label is @code{unused}. This
3928 feature is intended for code generated by programs which contains labels
3929 that may be unused but which is compiled with @option{-Wall}. It would
3930 not normally be appropriate to use in it human-written code, though it
3931 could be useful in cases where the code that jumps to the label is
3932 contained within an @code{#ifdef} conditional. GNU C++ only permits
3933 attributes on labels if the attribute specifier is immediately
3934 followed by a semicolon (i.e., the label applies to an empty
3935 statement). If the semicolon is missing, C++ label attributes are
3936 ambiguous, as it is permissible for a declaration, which could begin
3937 with an attribute list, to be labelled in C++. Declarations cannot be
3938 labelled in C90 or C99, so the ambiguity does not arise there.
3940 An attribute specifier list may appear as part of a @code{struct},
3941 @code{union} or @code{enum} specifier. It may go either immediately
3942 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3943 the closing brace. The former syntax is preferred.
3944 Where attribute specifiers follow the closing brace, they are considered
3945 to relate to the structure, union or enumerated type defined, not to any
3946 enclosing declaration the type specifier appears in, and the type
3947 defined is not complete until after the attribute specifiers.
3948 @c Otherwise, there would be the following problems: a shift/reduce
3949 @c conflict between attributes binding the struct/union/enum and
3950 @c binding to the list of specifiers/qualifiers; and "aligned"
3951 @c attributes could use sizeof for the structure, but the size could be
3952 @c changed later by "packed" attributes.
3954 Otherwise, an attribute specifier appears as part of a declaration,
3955 counting declarations of unnamed parameters and type names, and relates
3956 to that declaration (which may be nested in another declaration, for
3957 example in the case of a parameter declaration), or to a particular declarator
3958 within a declaration. Where an
3959 attribute specifier is applied to a parameter declared as a function or
3960 an array, it should apply to the function or array rather than the
3961 pointer to which the parameter is implicitly converted, but this is not
3962 yet correctly implemented.
3964 Any list of specifiers and qualifiers at the start of a declaration may
3965 contain attribute specifiers, whether or not such a list may in that
3966 context contain storage class specifiers. (Some attributes, however,
3967 are essentially in the nature of storage class specifiers, and only make
3968 sense where storage class specifiers may be used; for example,
3969 @code{section}.) There is one necessary limitation to this syntax: the
3970 first old-style parameter declaration in a function definition cannot
3971 begin with an attribute specifier, because such an attribute applies to
3972 the function instead by syntax described below (which, however, is not
3973 yet implemented in this case). In some other cases, attribute
3974 specifiers are permitted by this grammar but not yet supported by the
3975 compiler. All attribute specifiers in this place relate to the
3976 declaration as a whole. In the obsolescent usage where a type of
3977 @code{int} is implied by the absence of type specifiers, such a list of
3978 specifiers and qualifiers may be an attribute specifier list with no
3979 other specifiers or qualifiers.
3981 At present, the first parameter in a function prototype must have some
3982 type specifier which is not an attribute specifier; this resolves an
3983 ambiguity in the interpretation of @code{void f(int
3984 (__attribute__((foo)) x))}, but is subject to change. At present, if
3985 the parentheses of a function declarator contain only attributes then
3986 those attributes are ignored, rather than yielding an error or warning
3987 or implying a single parameter of type int, but this is subject to
3990 An attribute specifier list may appear immediately before a declarator
3991 (other than the first) in a comma-separated list of declarators in a
3992 declaration of more than one identifier using a single list of
3993 specifiers and qualifiers. Such attribute specifiers apply
3994 only to the identifier before whose declarator they appear. For
3998 __attribute__((noreturn)) void d0 (void),
3999 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
4004 the @code{noreturn} attribute applies to all the functions
4005 declared; the @code{format} attribute only applies to @code{d1}.
4007 An attribute specifier list may appear immediately before the comma,
4008 @code{=} or semicolon terminating the declaration of an identifier other
4009 than a function definition. Such attribute specifiers apply
4010 to the declared object or function. Where an
4011 assembler name for an object or function is specified (@pxref{Asm
4012 Labels}), the attribute must follow the @code{asm}
4015 An attribute specifier list may, in future, be permitted to appear after
4016 the declarator in a function definition (before any old-style parameter
4017 declarations or the function body).
4019 Attribute specifiers may be mixed with type qualifiers appearing inside
4020 the @code{[]} of a parameter array declarator, in the C99 construct by
4021 which such qualifiers are applied to the pointer to which the array is
4022 implicitly converted. Such attribute specifiers apply to the pointer,
4023 not to the array, but at present this is not implemented and they are
4026 An attribute specifier list may appear at the start of a nested
4027 declarator. At present, there are some limitations in this usage: the
4028 attributes correctly apply to the declarator, but for most individual
4029 attributes the semantics this implies are not implemented.
4030 When attribute specifiers follow the @code{*} of a pointer
4031 declarator, they may be mixed with any type qualifiers present.
4032 The following describes the formal semantics of this syntax. It will make the
4033 most sense if you are familiar with the formal specification of
4034 declarators in the ISO C standard.
4036 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
4037 D1}, where @code{T} contains declaration specifiers that specify a type
4038 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
4039 contains an identifier @var{ident}. The type specified for @var{ident}
4040 for derived declarators whose type does not include an attribute
4041 specifier is as in the ISO C standard.
4043 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
4044 and the declaration @code{T D} specifies the type
4045 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4046 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4047 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
4049 If @code{D1} has the form @code{*
4050 @var{type-qualifier-and-attribute-specifier-list} D}, and the
4051 declaration @code{T D} specifies the type
4052 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4053 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4054 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
4060 void (__attribute__((noreturn)) ****f) (void);
4064 specifies the type ``pointer to pointer to pointer to pointer to
4065 non-returning function returning @code{void}''. As another example,
4068 char *__attribute__((aligned(8))) *f;
4072 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
4073 Note again that this does not work with most attributes; for example,
4074 the usage of @samp{aligned} and @samp{noreturn} attributes given above
4075 is not yet supported.
4077 For compatibility with existing code written for compiler versions that
4078 did not implement attributes on nested declarators, some laxity is
4079 allowed in the placing of attributes. If an attribute that only applies
4080 to types is applied to a declaration, it will be treated as applying to
4081 the type of that declaration. If an attribute that only applies to
4082 declarations is applied to the type of a declaration, it will be treated
4083 as applying to that declaration; and, for compatibility with code
4084 placing the attributes immediately before the identifier declared, such
4085 an attribute applied to a function return type will be treated as
4086 applying to the function type, and such an attribute applied to an array
4087 element type will be treated as applying to the array type. If an
4088 attribute that only applies to function types is applied to a
4089 pointer-to-function type, it will be treated as applying to the pointer
4090 target type; if such an attribute is applied to a function return type
4091 that is not a pointer-to-function type, it will be treated as applying
4092 to the function type.
4094 @node Function Prototypes
4095 @section Prototypes and Old-Style Function Definitions
4096 @cindex function prototype declarations
4097 @cindex old-style function definitions
4098 @cindex promotion of formal parameters
4100 GNU C extends ISO C to allow a function prototype to override a later
4101 old-style non-prototype definition. Consider the following example:
4104 /* @r{Use prototypes unless the compiler is old-fashioned.} */
4111 /* @r{Prototype function declaration.} */
4112 int isroot P((uid_t));
4114 /* @r{Old-style function definition.} */
4116 isroot (x) /* @r{??? lossage here ???} */
4123 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
4124 not allow this example, because subword arguments in old-style
4125 non-prototype definitions are promoted. Therefore in this example the
4126 function definition's argument is really an @code{int}, which does not
4127 match the prototype argument type of @code{short}.
4129 This restriction of ISO C makes it hard to write code that is portable
4130 to traditional C compilers, because the programmer does not know
4131 whether the @code{uid_t} type is @code{short}, @code{int}, or
4132 @code{long}. Therefore, in cases like these GNU C allows a prototype
4133 to override a later old-style definition. More precisely, in GNU C, a
4134 function prototype argument type overrides the argument type specified
4135 by a later old-style definition if the former type is the same as the
4136 latter type before promotion. Thus in GNU C the above example is
4137 equivalent to the following:
4150 GNU C++ does not support old-style function definitions, so this
4151 extension is irrelevant.
4154 @section C++ Style Comments
4156 @cindex C++ comments
4157 @cindex comments, C++ style
4159 In GNU C, you may use C++ style comments, which start with @samp{//} and
4160 continue until the end of the line. Many other C implementations allow
4161 such comments, and they are included in the 1999 C standard. However,
4162 C++ style comments are not recognized if you specify an @option{-std}
4163 option specifying a version of ISO C before C99, or @option{-ansi}
4164 (equivalent to @option{-std=c90}).
4167 @section Dollar Signs in Identifier Names
4169 @cindex dollar signs in identifier names
4170 @cindex identifier names, dollar signs in
4172 In GNU C, you may normally use dollar signs in identifier names.
4173 This is because many traditional C implementations allow such identifiers.
4174 However, dollar signs in identifiers are not supported on a few target
4175 machines, typically because the target assembler does not allow them.
4177 @node Character Escapes
4178 @section The Character @key{ESC} in Constants
4180 You can use the sequence @samp{\e} in a string or character constant to
4181 stand for the ASCII character @key{ESC}.
4183 @node Variable Attributes
4184 @section Specifying Attributes of Variables
4185 @cindex attribute of variables
4186 @cindex variable attributes
4188 The keyword @code{__attribute__} allows you to specify special
4189 attributes of variables or structure fields. This keyword is followed
4190 by an attribute specification inside double parentheses. Some
4191 attributes are currently defined generically for variables.
4192 Other attributes are defined for variables on particular target
4193 systems. Other attributes are available for functions
4194 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
4195 Other front ends might define more attributes
4196 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
4198 You may also specify attributes with @samp{__} preceding and following
4199 each keyword. This allows you to use them in header files without
4200 being concerned about a possible macro of the same name. For example,
4201 you may use @code{__aligned__} instead of @code{aligned}.
4203 @xref{Attribute Syntax}, for details of the exact syntax for using
4207 @cindex @code{aligned} attribute
4208 @item aligned (@var{alignment})
4209 This attribute specifies a minimum alignment for the variable or
4210 structure field, measured in bytes. For example, the declaration:
4213 int x __attribute__ ((aligned (16))) = 0;
4217 causes the compiler to allocate the global variable @code{x} on a
4218 16-byte boundary. On a 68040, this could be used in conjunction with
4219 an @code{asm} expression to access the @code{move16} instruction which
4220 requires 16-byte aligned operands.
4222 You can also specify the alignment of structure fields. For example, to
4223 create a double-word aligned @code{int} pair, you could write:
4226 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
4230 This is an alternative to creating a union with a @code{double} member
4231 that forces the union to be double-word aligned.
4233 As in the preceding examples, you can explicitly specify the alignment
4234 (in bytes) that you wish the compiler to use for a given variable or
4235 structure field. Alternatively, you can leave out the alignment factor
4236 and just ask the compiler to align a variable or field to the
4237 default alignment for the target architecture you are compiling for.
4238 The default alignment is sufficient for all scalar types, but may not be
4239 enough for all vector types on a target which supports vector operations.
4240 The default alignment is fixed for a particular target ABI.
4242 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
4243 which is the largest alignment ever used for any data type on the
4244 target machine you are compiling for. For example, you could write:
4247 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
4250 The compiler automatically sets the alignment for the declared
4251 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
4252 often make copy operations more efficient, because the compiler can
4253 use whatever instructions copy the biggest chunks of memory when
4254 performing copies to or from the variables or fields that you have
4255 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
4256 may change depending on command line options.
4258 When used on a struct, or struct member, the @code{aligned} attribute can
4259 only increase the alignment; in order to decrease it, the @code{packed}
4260 attribute must be specified as well. When used as part of a typedef, the
4261 @code{aligned} attribute can both increase and decrease alignment, and
4262 specifying the @code{packed} attribute will generate a warning.
4264 Note that the effectiveness of @code{aligned} attributes may be limited
4265 by inherent limitations in your linker. On many systems, the linker is
4266 only able to arrange for variables to be aligned up to a certain maximum
4267 alignment. (For some linkers, the maximum supported alignment may
4268 be very very small.) If your linker is only able to align variables
4269 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4270 in an @code{__attribute__} will still only provide you with 8 byte
4271 alignment. See your linker documentation for further information.
4273 The @code{aligned} attribute can also be used for functions
4274 (@pxref{Function Attributes}.)
4276 @item cleanup (@var{cleanup_function})
4277 @cindex @code{cleanup} attribute
4278 The @code{cleanup} attribute runs a function when the variable goes
4279 out of scope. This attribute can only be applied to auto function
4280 scope variables; it may not be applied to parameters or variables
4281 with static storage duration. The function must take one parameter,
4282 a pointer to a type compatible with the variable. The return value
4283 of the function (if any) is ignored.
4285 If @option{-fexceptions} is enabled, then @var{cleanup_function}
4286 will be run during the stack unwinding that happens during the
4287 processing of the exception. Note that the @code{cleanup} attribute
4288 does not allow the exception to be caught, only to perform an action.
4289 It is undefined what happens if @var{cleanup_function} does not
4294 @cindex @code{common} attribute
4295 @cindex @code{nocommon} attribute
4298 The @code{common} attribute requests GCC to place a variable in
4299 ``common'' storage. The @code{nocommon} attribute requests the
4300 opposite---to allocate space for it directly.
4302 These attributes override the default chosen by the
4303 @option{-fno-common} and @option{-fcommon} flags respectively.
4306 @itemx deprecated (@var{msg})
4307 @cindex @code{deprecated} attribute
4308 The @code{deprecated} attribute results in a warning if the variable
4309 is used anywhere in the source file. This is useful when identifying
4310 variables that are expected to be removed in a future version of a
4311 program. The warning also includes the location of the declaration
4312 of the deprecated variable, to enable users to easily find further
4313 information about why the variable is deprecated, or what they should
4314 do instead. Note that the warning only occurs for uses:
4317 extern int old_var __attribute__ ((deprecated));
4319 int new_fn () @{ return old_var; @}
4322 results in a warning on line 3 but not line 2. The optional msg
4323 argument, which must be a string, will be printed in the warning if
4326 The @code{deprecated} attribute can also be used for functions and
4327 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
4329 @item mode (@var{mode})
4330 @cindex @code{mode} attribute
4331 This attribute specifies the data type for the declaration---whichever
4332 type corresponds to the mode @var{mode}. This in effect lets you
4333 request an integer or floating point type according to its width.
4335 You may also specify a mode of @samp{byte} or @samp{__byte__} to
4336 indicate the mode corresponding to a one-byte integer, @samp{word} or
4337 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
4338 or @samp{__pointer__} for the mode used to represent pointers.
4341 @cindex @code{packed} attribute
4342 The @code{packed} attribute specifies that a variable or structure field
4343 should have the smallest possible alignment---one byte for a variable,
4344 and one bit for a field, unless you specify a larger value with the
4345 @code{aligned} attribute.
4347 Here is a structure in which the field @code{x} is packed, so that it
4348 immediately follows @code{a}:
4354 int x[2] __attribute__ ((packed));
4358 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
4359 @code{packed} attribute on bit-fields of type @code{char}. This has
4360 been fixed in GCC 4.4 but the change can lead to differences in the
4361 structure layout. See the documentation of
4362 @option{-Wpacked-bitfield-compat} for more information.
4364 @item section ("@var{section-name}")
4365 @cindex @code{section} variable attribute
4366 Normally, the compiler places the objects it generates in sections like
4367 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
4368 or you need certain particular variables to appear in special sections,
4369 for example to map to special hardware. The @code{section}
4370 attribute specifies that a variable (or function) lives in a particular
4371 section. For example, this small program uses several specific section names:
4374 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
4375 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
4376 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
4377 int init_data __attribute__ ((section ("INITDATA")));
4381 /* @r{Initialize stack pointer} */
4382 init_sp (stack + sizeof (stack));
4384 /* @r{Initialize initialized data} */
4385 memcpy (&init_data, &data, &edata - &data);
4387 /* @r{Turn on the serial ports} */
4394 Use the @code{section} attribute with
4395 @emph{global} variables and not @emph{local} variables,
4396 as shown in the example.
4398 You may use the @code{section} attribute with initialized or
4399 uninitialized global variables but the linker requires
4400 each object be defined once, with the exception that uninitialized
4401 variables tentatively go in the @code{common} (or @code{bss}) section
4402 and can be multiply ``defined''. Using the @code{section} attribute
4403 will change what section the variable goes into and may cause the
4404 linker to issue an error if an uninitialized variable has multiple
4405 definitions. You can force a variable to be initialized with the
4406 @option{-fno-common} flag or the @code{nocommon} attribute.
4408 Some file formats do not support arbitrary sections so the @code{section}
4409 attribute is not available on all platforms.
4410 If you need to map the entire contents of a module to a particular
4411 section, consider using the facilities of the linker instead.
4414 @cindex @code{shared} variable attribute
4415 On Microsoft Windows, in addition to putting variable definitions in a named
4416 section, the section can also be shared among all running copies of an
4417 executable or DLL@. For example, this small program defines shared data
4418 by putting it in a named section @code{shared} and marking the section
4422 int foo __attribute__((section ("shared"), shared)) = 0;
4427 /* @r{Read and write foo. All running
4428 copies see the same value.} */
4434 You may only use the @code{shared} attribute along with @code{section}
4435 attribute with a fully initialized global definition because of the way
4436 linkers work. See @code{section} attribute for more information.
4438 The @code{shared} attribute is only available on Microsoft Windows@.
4440 @item tls_model ("@var{tls_model}")
4441 @cindex @code{tls_model} attribute
4442 The @code{tls_model} attribute sets thread-local storage model
4443 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
4444 overriding @option{-ftls-model=} command-line switch on a per-variable
4446 The @var{tls_model} argument should be one of @code{global-dynamic},
4447 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
4449 Not all targets support this attribute.
4452 This attribute, attached to a variable, means that the variable is meant
4453 to be possibly unused. GCC will not produce a warning for this
4457 This attribute, attached to a variable, means that the variable must be
4458 emitted even if it appears that the variable is not referenced.
4460 @item vector_size (@var{bytes})
4461 This attribute specifies the vector size for the variable, measured in
4462 bytes. For example, the declaration:
4465 int foo __attribute__ ((vector_size (16)));
4469 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
4470 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
4471 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
4473 This attribute is only applicable to integral and float scalars,
4474 although arrays, pointers, and function return values are allowed in
4475 conjunction with this construct.
4477 Aggregates with this attribute are invalid, even if they are of the same
4478 size as a corresponding scalar. For example, the declaration:
4481 struct S @{ int a; @};
4482 struct S __attribute__ ((vector_size (16))) foo;
4486 is invalid even if the size of the structure is the same as the size of
4490 The @code{selectany} attribute causes an initialized global variable to
4491 have link-once semantics. When multiple definitions of the variable are
4492 encountered by the linker, the first is selected and the remainder are
4493 discarded. Following usage by the Microsoft compiler, the linker is told
4494 @emph{not} to warn about size or content differences of the multiple
4497 Although the primary usage of this attribute is for POD types, the
4498 attribute can also be applied to global C++ objects that are initialized
4499 by a constructor. In this case, the static initialization and destruction
4500 code for the object is emitted in each translation defining the object,
4501 but the calls to the constructor and destructor are protected by a
4502 link-once guard variable.
4504 The @code{selectany} attribute is only available on Microsoft Windows
4505 targets. You can use @code{__declspec (selectany)} as a synonym for
4506 @code{__attribute__ ((selectany))} for compatibility with other
4510 The @code{weak} attribute is described in @ref{Function Attributes}.
4513 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4516 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4520 @subsection Blackfin Variable Attributes
4522 Three attributes are currently defined for the Blackfin.
4528 @cindex @code{l1_data} variable attribute
4529 @cindex @code{l1_data_A} variable attribute
4530 @cindex @code{l1_data_B} variable attribute
4531 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4532 Variables with @code{l1_data} attribute will be put into the specific section
4533 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4534 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4535 attribute will be put into the specific section named @code{.l1.data.B}.
4538 @cindex @code{l2} variable attribute
4539 Use this attribute on the Blackfin to place the variable into L2 SRAM.
4540 Variables with @code{l2} attribute will be put into the specific section
4541 named @code{.l2.data}.
4544 @subsection M32R/D Variable Attributes
4546 One attribute is currently defined for the M32R/D@.
4549 @item model (@var{model-name})
4550 @cindex variable addressability on the M32R/D
4551 Use this attribute on the M32R/D to set the addressability of an object.
4552 The identifier @var{model-name} is one of @code{small}, @code{medium},
4553 or @code{large}, representing each of the code models.
4555 Small model objects live in the lower 16MB of memory (so that their
4556 addresses can be loaded with the @code{ld24} instruction).
4558 Medium and large model objects may live anywhere in the 32-bit address space
4559 (the compiler will generate @code{seth/add3} instructions to load their
4563 @anchor{MeP Variable Attributes}
4564 @subsection MeP Variable Attributes
4566 The MeP target has a number of addressing modes and busses. The
4567 @code{near} space spans the standard memory space's first 16 megabytes
4568 (24 bits). The @code{far} space spans the entire 32-bit memory space.
4569 The @code{based} space is a 128 byte region in the memory space which
4570 is addressed relative to the @code{$tp} register. The @code{tiny}
4571 space is a 65536 byte region relative to the @code{$gp} register. In
4572 addition to these memory regions, the MeP target has a separate 16-bit
4573 control bus which is specified with @code{cb} attributes.
4578 Any variable with the @code{based} attribute will be assigned to the
4579 @code{.based} section, and will be accessed with relative to the
4580 @code{$tp} register.
4583 Likewise, the @code{tiny} attribute assigned variables to the
4584 @code{.tiny} section, relative to the @code{$gp} register.
4587 Variables with the @code{near} attribute are assumed to have addresses
4588 that fit in a 24-bit addressing mode. This is the default for large
4589 variables (@code{-mtiny=4} is the default) but this attribute can
4590 override @code{-mtiny=} for small variables, or override @code{-ml}.
4593 Variables with the @code{far} attribute are addressed using a full
4594 32-bit address. Since this covers the entire memory space, this
4595 allows modules to make no assumptions about where variables might be
4599 @itemx io (@var{addr})
4600 Variables with the @code{io} attribute are used to address
4601 memory-mapped peripherals. If an address is specified, the variable
4602 is assigned that address, else it is not assigned an address (it is
4603 assumed some other module will assign an address). Example:
4606 int timer_count __attribute__((io(0x123)));
4610 @itemx cb (@var{addr})
4611 Variables with the @code{cb} attribute are used to access the control
4612 bus, using special instructions. @code{addr} indicates the control bus
4616 int cpu_clock __attribute__((cb(0x123)));
4621 @anchor{i386 Variable Attributes}
4622 @subsection i386 Variable Attributes
4624 Two attributes are currently defined for i386 configurations:
4625 @code{ms_struct} and @code{gcc_struct}
4630 @cindex @code{ms_struct} attribute
4631 @cindex @code{gcc_struct} attribute
4633 If @code{packed} is used on a structure, or if bit-fields are used
4634 it may be that the Microsoft ABI packs them differently
4635 than GCC would normally pack them. Particularly when moving packed
4636 data between functions compiled with GCC and the native Microsoft compiler
4637 (either via function call or as data in a file), it may be necessary to access
4640 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4641 compilers to match the native Microsoft compiler.
4643 The Microsoft structure layout algorithm is fairly simple with the exception
4644 of the bitfield packing:
4646 The padding and alignment of members of structures and whether a bit field
4647 can straddle a storage-unit boundary
4650 @item Structure members are stored sequentially in the order in which they are
4651 declared: the first member has the lowest memory address and the last member
4654 @item Every data object has an alignment-requirement. The alignment-requirement
4655 for all data except structures, unions, and arrays is either the size of the
4656 object or the current packing size (specified with either the aligned attribute
4657 or the pack pragma), whichever is less. For structures, unions, and arrays,
4658 the alignment-requirement is the largest alignment-requirement of its members.
4659 Every object is allocated an offset so that:
4661 offset % alignment-requirement == 0
4663 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4664 unit if the integral types are the same size and if the next bit field fits
4665 into the current allocation unit without crossing the boundary imposed by the
4666 common alignment requirements of the bit fields.
4669 Handling of zero-length bitfields:
4671 MSVC interprets zero-length bitfields in the following ways:
4674 @item If a zero-length bitfield is inserted between two bitfields that would
4675 normally be coalesced, the bitfields will not be coalesced.
4682 unsigned long bf_1 : 12;
4684 unsigned long bf_2 : 12;
4688 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4689 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4691 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4692 alignment of the zero-length bitfield is greater than the member that follows it,
4693 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4713 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4714 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4715 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4718 Taking this into account, it is important to note the following:
4721 @item If a zero-length bitfield follows a normal bitfield, the type of the
4722 zero-length bitfield may affect the alignment of the structure as whole. For
4723 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4724 normal bitfield, and is of type short.
4726 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4727 still affect the alignment of the structure:
4737 Here, @code{t4} will take up 4 bytes.
4740 @item Zero-length bitfields following non-bitfield members are ignored:
4751 Here, @code{t5} will take up 2 bytes.
4755 @subsection PowerPC Variable Attributes
4757 Three attributes currently are defined for PowerPC configurations:
4758 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4760 For full documentation of the struct attributes please see the
4761 documentation in @ref{i386 Variable Attributes}.
4763 For documentation of @code{altivec} attribute please see the
4764 documentation in @ref{PowerPC Type Attributes}.
4766 @subsection SPU Variable Attributes
4768 The SPU supports the @code{spu_vector} attribute for variables. For
4769 documentation of this attribute please see the documentation in
4770 @ref{SPU Type Attributes}.
4772 @subsection Xstormy16 Variable Attributes
4774 One attribute is currently defined for xstormy16 configurations:
4779 @cindex @code{below100} attribute
4781 If a variable has the @code{below100} attribute (@code{BELOW100} is
4782 allowed also), GCC will place the variable in the first 0x100 bytes of
4783 memory and use special opcodes to access it. Such variables will be
4784 placed in either the @code{.bss_below100} section or the
4785 @code{.data_below100} section.
4789 @subsection AVR Variable Attributes
4793 @cindex @code{progmem} variable attribute
4794 The @code{progmem} attribute is used on the AVR to place data in the Program
4795 Memory address space. The AVR is a Harvard Architecture processor and data
4796 normally resides in the Data Memory address space.
4799 @node Type Attributes
4800 @section Specifying Attributes of Types
4801 @cindex attribute of types
4802 @cindex type attributes
4804 The keyword @code{__attribute__} allows you to specify special
4805 attributes of @code{struct} and @code{union} types when you define
4806 such types. This keyword is followed by an attribute specification
4807 inside double parentheses. Seven attributes are currently defined for
4808 types: @code{aligned}, @code{packed}, @code{transparent_union},
4809 @code{unused}, @code{deprecated}, @code{visibility}, and
4810 @code{may_alias}. Other attributes are defined for functions
4811 (@pxref{Function Attributes}) and for variables (@pxref{Variable
4814 You may also specify any one of these attributes with @samp{__}
4815 preceding and following its keyword. This allows you to use these
4816 attributes in header files without being concerned about a possible
4817 macro of the same name. For example, you may use @code{__aligned__}
4818 instead of @code{aligned}.
4820 You may specify type attributes in an enum, struct or union type
4821 declaration or definition, or for other types in a @code{typedef}
4824 For an enum, struct or union type, you may specify attributes either
4825 between the enum, struct or union tag and the name of the type, or
4826 just past the closing curly brace of the @emph{definition}. The
4827 former syntax is preferred.
4829 @xref{Attribute Syntax}, for details of the exact syntax for using
4833 @cindex @code{aligned} attribute
4834 @item aligned (@var{alignment})
4835 This attribute specifies a minimum alignment (in bytes) for variables
4836 of the specified type. For example, the declarations:
4839 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
4840 typedef int more_aligned_int __attribute__ ((aligned (8)));
4844 force the compiler to insure (as far as it can) that each variable whose
4845 type is @code{struct S} or @code{more_aligned_int} will be allocated and
4846 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
4847 variables of type @code{struct S} aligned to 8-byte boundaries allows
4848 the compiler to use the @code{ldd} and @code{std} (doubleword load and
4849 store) instructions when copying one variable of type @code{struct S} to
4850 another, thus improving run-time efficiency.
4852 Note that the alignment of any given @code{struct} or @code{union} type
4853 is required by the ISO C standard to be at least a perfect multiple of
4854 the lowest common multiple of the alignments of all of the members of
4855 the @code{struct} or @code{union} in question. This means that you @emph{can}
4856 effectively adjust the alignment of a @code{struct} or @code{union}
4857 type by attaching an @code{aligned} attribute to any one of the members
4858 of such a type, but the notation illustrated in the example above is a
4859 more obvious, intuitive, and readable way to request the compiler to
4860 adjust the alignment of an entire @code{struct} or @code{union} type.
4862 As in the preceding example, you can explicitly specify the alignment
4863 (in bytes) that you wish the compiler to use for a given @code{struct}
4864 or @code{union} type. Alternatively, you can leave out the alignment factor
4865 and just ask the compiler to align a type to the maximum
4866 useful alignment for the target machine you are compiling for. For
4867 example, you could write:
4870 struct S @{ short f[3]; @} __attribute__ ((aligned));
4873 Whenever you leave out the alignment factor in an @code{aligned}
4874 attribute specification, the compiler automatically sets the alignment
4875 for the type to the largest alignment which is ever used for any data
4876 type on the target machine you are compiling for. Doing this can often
4877 make copy operations more efficient, because the compiler can use
4878 whatever instructions copy the biggest chunks of memory when performing
4879 copies to or from the variables which have types that you have aligned
4882 In the example above, if the size of each @code{short} is 2 bytes, then
4883 the size of the entire @code{struct S} type is 6 bytes. The smallest
4884 power of two which is greater than or equal to that is 8, so the
4885 compiler sets the alignment for the entire @code{struct S} type to 8
4888 Note that although you can ask the compiler to select a time-efficient
4889 alignment for a given type and then declare only individual stand-alone
4890 objects of that type, the compiler's ability to select a time-efficient
4891 alignment is primarily useful only when you plan to create arrays of
4892 variables having the relevant (efficiently aligned) type. If you
4893 declare or use arrays of variables of an efficiently-aligned type, then
4894 it is likely that your program will also be doing pointer arithmetic (or
4895 subscripting, which amounts to the same thing) on pointers to the
4896 relevant type, and the code that the compiler generates for these
4897 pointer arithmetic operations will often be more efficient for
4898 efficiently-aligned types than for other types.
4900 The @code{aligned} attribute can only increase the alignment; but you
4901 can decrease it by specifying @code{packed} as well. See below.
4903 Note that the effectiveness of @code{aligned} attributes may be limited
4904 by inherent limitations in your linker. On many systems, the linker is
4905 only able to arrange for variables to be aligned up to a certain maximum
4906 alignment. (For some linkers, the maximum supported alignment may
4907 be very very small.) If your linker is only able to align variables
4908 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4909 in an @code{__attribute__} will still only provide you with 8 byte
4910 alignment. See your linker documentation for further information.
4913 This attribute, attached to @code{struct} or @code{union} type
4914 definition, specifies that each member (other than zero-width bitfields)
4915 of the structure or union is placed to minimize the memory required. When
4916 attached to an @code{enum} definition, it indicates that the smallest
4917 integral type should be used.
4919 @opindex fshort-enums
4920 Specifying this attribute for @code{struct} and @code{union} types is
4921 equivalent to specifying the @code{packed} attribute on each of the
4922 structure or union members. Specifying the @option{-fshort-enums}
4923 flag on the line is equivalent to specifying the @code{packed}
4924 attribute on all @code{enum} definitions.
4926 In the following example @code{struct my_packed_struct}'s members are
4927 packed closely together, but the internal layout of its @code{s} member
4928 is not packed---to do that, @code{struct my_unpacked_struct} would need to
4932 struct my_unpacked_struct
4938 struct __attribute__ ((__packed__)) my_packed_struct
4942 struct my_unpacked_struct s;
4946 You may only specify this attribute on the definition of an @code{enum},
4947 @code{struct} or @code{union}, not on a @code{typedef} which does not
4948 also define the enumerated type, structure or union.
4950 @item transparent_union
4951 This attribute, attached to a @code{union} type definition, indicates
4952 that any function parameter having that union type causes calls to that
4953 function to be treated in a special way.
4955 First, the argument corresponding to a transparent union type can be of
4956 any type in the union; no cast is required. Also, if the union contains
4957 a pointer type, the corresponding argument can be a null pointer
4958 constant or a void pointer expression; and if the union contains a void
4959 pointer type, the corresponding argument can be any pointer expression.
4960 If the union member type is a pointer, qualifiers like @code{const} on
4961 the referenced type must be respected, just as with normal pointer
4964 Second, the argument is passed to the function using the calling
4965 conventions of the first member of the transparent union, not the calling
4966 conventions of the union itself. All members of the union must have the
4967 same machine representation; this is necessary for this argument passing
4970 Transparent unions are designed for library functions that have multiple
4971 interfaces for compatibility reasons. For example, suppose the
4972 @code{wait} function must accept either a value of type @code{int *} to
4973 comply with Posix, or a value of type @code{union wait *} to comply with
4974 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4975 @code{wait} would accept both kinds of arguments, but it would also
4976 accept any other pointer type and this would make argument type checking
4977 less useful. Instead, @code{<sys/wait.h>} might define the interface
4981 typedef union __attribute__ ((__transparent_union__))
4985 @} wait_status_ptr_t;
4987 pid_t wait (wait_status_ptr_t);
4990 This interface allows either @code{int *} or @code{union wait *}
4991 arguments to be passed, using the @code{int *} calling convention.
4992 The program can call @code{wait} with arguments of either type:
4995 int w1 () @{ int w; return wait (&w); @}
4996 int w2 () @{ union wait w; return wait (&w); @}
4999 With this interface, @code{wait}'s implementation might look like this:
5002 pid_t wait (wait_status_ptr_t p)
5004 return waitpid (-1, p.__ip, 0);
5009 When attached to a type (including a @code{union} or a @code{struct}),
5010 this attribute means that variables of that type are meant to appear
5011 possibly unused. GCC will not produce a warning for any variables of
5012 that type, even if the variable appears to do nothing. This is often
5013 the case with lock or thread classes, which are usually defined and then
5014 not referenced, but contain constructors and destructors that have
5015 nontrivial bookkeeping functions.
5018 @itemx deprecated (@var{msg})
5019 The @code{deprecated} attribute results in a warning if the type
5020 is used anywhere in the source file. This is useful when identifying
5021 types that are expected to be removed in a future version of a program.
5022 If possible, the warning also includes the location of the declaration
5023 of the deprecated type, to enable users to easily find further
5024 information about why the type is deprecated, or what they should do
5025 instead. Note that the warnings only occur for uses and then only
5026 if the type is being applied to an identifier that itself is not being
5027 declared as deprecated.
5030 typedef int T1 __attribute__ ((deprecated));
5034 typedef T1 T3 __attribute__ ((deprecated));
5035 T3 z __attribute__ ((deprecated));
5038 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
5039 warning is issued for line 4 because T2 is not explicitly
5040 deprecated. Line 5 has no warning because T3 is explicitly
5041 deprecated. Similarly for line 6. The optional msg
5042 argument, which must be a string, will be printed in the warning if
5045 The @code{deprecated} attribute can also be used for functions and
5046 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
5049 Accesses through pointers to types with this attribute are not subject
5050 to type-based alias analysis, but are instead assumed to be able to alias
5051 any other type of objects. In the context of 6.5/7 an lvalue expression
5052 dereferencing such a pointer is treated like having a character type.
5053 See @option{-fstrict-aliasing} for more information on aliasing issues.
5054 This extension exists to support some vector APIs, in which pointers to
5055 one vector type are permitted to alias pointers to a different vector type.
5057 Note that an object of a type with this attribute does not have any
5063 typedef short __attribute__((__may_alias__)) short_a;
5069 short_a *b = (short_a *) &a;
5073 if (a == 0x12345678)
5080 If you replaced @code{short_a} with @code{short} in the variable
5081 declaration, the above program would abort when compiled with
5082 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
5083 above in recent GCC versions.
5086 In C++, attribute visibility (@pxref{Function Attributes}) can also be
5087 applied to class, struct, union and enum types. Unlike other type
5088 attributes, the attribute must appear between the initial keyword and
5089 the name of the type; it cannot appear after the body of the type.
5091 Note that the type visibility is applied to vague linkage entities
5092 associated with the class (vtable, typeinfo node, etc.). In
5093 particular, if a class is thrown as an exception in one shared object
5094 and caught in another, the class must have default visibility.
5095 Otherwise the two shared objects will be unable to use the same
5096 typeinfo node and exception handling will break.
5100 @subsection ARM Type Attributes
5102 On those ARM targets that support @code{dllimport} (such as Symbian
5103 OS), you can use the @code{notshared} attribute to indicate that the
5104 virtual table and other similar data for a class should not be
5105 exported from a DLL@. For example:
5108 class __declspec(notshared) C @{
5110 __declspec(dllimport) C();
5114 __declspec(dllexport)
5118 In this code, @code{C::C} is exported from the current DLL, but the
5119 virtual table for @code{C} is not exported. (You can use
5120 @code{__attribute__} instead of @code{__declspec} if you prefer, but
5121 most Symbian OS code uses @code{__declspec}.)
5123 @anchor{MeP Type Attributes}
5124 @subsection MeP Type Attributes
5126 Many of the MeP variable attributes may be applied to types as well.
5127 Specifically, the @code{based}, @code{tiny}, @code{near}, and
5128 @code{far} attributes may be applied to either. The @code{io} and
5129 @code{cb} attributes may not be applied to types.
5131 @anchor{i386 Type Attributes}
5132 @subsection i386 Type Attributes
5134 Two attributes are currently defined for i386 configurations:
5135 @code{ms_struct} and @code{gcc_struct}.
5141 @cindex @code{ms_struct}
5142 @cindex @code{gcc_struct}
5144 If @code{packed} is used on a structure, or if bit-fields are used
5145 it may be that the Microsoft ABI packs them differently
5146 than GCC would normally pack them. Particularly when moving packed
5147 data between functions compiled with GCC and the native Microsoft compiler
5148 (either via function call or as data in a file), it may be necessary to access
5151 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
5152 compilers to match the native Microsoft compiler.
5155 To specify multiple attributes, separate them by commas within the
5156 double parentheses: for example, @samp{__attribute__ ((aligned (16),
5159 @anchor{PowerPC Type Attributes}
5160 @subsection PowerPC Type Attributes
5162 Three attributes currently are defined for PowerPC configurations:
5163 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5165 For full documentation of the @code{ms_struct} and @code{gcc_struct}
5166 attributes please see the documentation in @ref{i386 Type Attributes}.
5168 The @code{altivec} attribute allows one to declare AltiVec vector data
5169 types supported by the AltiVec Programming Interface Manual. The
5170 attribute requires an argument to specify one of three vector types:
5171 @code{vector__}, @code{pixel__} (always followed by unsigned short),
5172 and @code{bool__} (always followed by unsigned).
5175 __attribute__((altivec(vector__)))
5176 __attribute__((altivec(pixel__))) unsigned short
5177 __attribute__((altivec(bool__))) unsigned
5180 These attributes mainly are intended to support the @code{__vector},
5181 @code{__pixel}, and @code{__bool} AltiVec keywords.
5183 @anchor{SPU Type Attributes}
5184 @subsection SPU Type Attributes
5186 The SPU supports the @code{spu_vector} attribute for types. This attribute
5187 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
5188 Language Extensions Specification. It is intended to support the
5189 @code{__vector} keyword.
5192 @section Inquiring on Alignment of Types or Variables
5194 @cindex type alignment
5195 @cindex variable alignment
5197 The keyword @code{__alignof__} allows you to inquire about how an object
5198 is aligned, or the minimum alignment usually required by a type. Its
5199 syntax is just like @code{sizeof}.
5201 For example, if the target machine requires a @code{double} value to be
5202 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
5203 This is true on many RISC machines. On more traditional machine
5204 designs, @code{__alignof__ (double)} is 4 or even 2.
5206 Some machines never actually require alignment; they allow reference to any
5207 data type even at an odd address. For these machines, @code{__alignof__}
5208 reports the smallest alignment that GCC will give the data type, usually as
5209 mandated by the target ABI.
5211 If the operand of @code{__alignof__} is an lvalue rather than a type,
5212 its value is the required alignment for its type, taking into account
5213 any minimum alignment specified with GCC's @code{__attribute__}
5214 extension (@pxref{Variable Attributes}). For example, after this
5218 struct foo @{ int x; char y; @} foo1;
5222 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
5223 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
5225 It is an error to ask for the alignment of an incomplete type.
5229 @section An Inline Function is As Fast As a Macro
5230 @cindex inline functions
5231 @cindex integrating function code
5233 @cindex macros, inline alternative
5235 By declaring a function inline, you can direct GCC to make
5236 calls to that function faster. One way GCC can achieve this is to
5237 integrate that function's code into the code for its callers. This
5238 makes execution faster by eliminating the function-call overhead; in
5239 addition, if any of the actual argument values are constant, their
5240 known values may permit simplifications at compile time so that not
5241 all of the inline function's code needs to be included. The effect on
5242 code size is less predictable; object code may be larger or smaller
5243 with function inlining, depending on the particular case. You can
5244 also direct GCC to try to integrate all ``simple enough'' functions
5245 into their callers with the option @option{-finline-functions}.
5247 GCC implements three different semantics of declaring a function
5248 inline. One is available with @option{-std=gnu89} or
5249 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
5250 on all inline declarations, another when
5251 @option{-std=c99}, @option{-std=c1x},
5252 @option{-std=gnu99} or @option{-std=gnu1x}
5253 (without @option{-fgnu89-inline}), and the third
5254 is used when compiling C++.
5256 To declare a function inline, use the @code{inline} keyword in its
5257 declaration, like this:
5267 If you are writing a header file to be included in ISO C90 programs, write
5268 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
5270 The three types of inlining behave similarly in two important cases:
5271 when the @code{inline} keyword is used on a @code{static} function,
5272 like the example above, and when a function is first declared without
5273 using the @code{inline} keyword and then is defined with
5274 @code{inline}, like this:
5277 extern int inc (int *a);
5285 In both of these common cases, the program behaves the same as if you
5286 had not used the @code{inline} keyword, except for its speed.
5288 @cindex inline functions, omission of
5289 @opindex fkeep-inline-functions
5290 When a function is both inline and @code{static}, if all calls to the
5291 function are integrated into the caller, and the function's address is
5292 never used, then the function's own assembler code is never referenced.
5293 In this case, GCC does not actually output assembler code for the
5294 function, unless you specify the option @option{-fkeep-inline-functions}.
5295 Some calls cannot be integrated for various reasons (in particular,
5296 calls that precede the function's definition cannot be integrated, and
5297 neither can recursive calls within the definition). If there is a
5298 nonintegrated call, then the function is compiled to assembler code as
5299 usual. The function must also be compiled as usual if the program
5300 refers to its address, because that can't be inlined.
5303 Note that certain usages in a function definition can make it unsuitable
5304 for inline substitution. Among these usages are: use of varargs, use of
5305 alloca, use of variable sized data types (@pxref{Variable Length}),
5306 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
5307 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
5308 will warn when a function marked @code{inline} could not be substituted,
5309 and will give the reason for the failure.
5311 @cindex automatic @code{inline} for C++ member fns
5312 @cindex @code{inline} automatic for C++ member fns
5313 @cindex member fns, automatically @code{inline}
5314 @cindex C++ member fns, automatically @code{inline}
5315 @opindex fno-default-inline
5316 As required by ISO C++, GCC considers member functions defined within
5317 the body of a class to be marked inline even if they are
5318 not explicitly declared with the @code{inline} keyword. You can
5319 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
5320 Options,,Options Controlling C++ Dialect}.
5322 GCC does not inline any functions when not optimizing unless you specify
5323 the @samp{always_inline} attribute for the function, like this:
5326 /* @r{Prototype.} */
5327 inline void foo (const char) __attribute__((always_inline));
5330 The remainder of this section is specific to GNU C90 inlining.
5332 @cindex non-static inline function
5333 When an inline function is not @code{static}, then the compiler must assume
5334 that there may be calls from other source files; since a global symbol can
5335 be defined only once in any program, the function must not be defined in
5336 the other source files, so the calls therein cannot be integrated.
5337 Therefore, a non-@code{static} inline function is always compiled on its
5338 own in the usual fashion.
5340 If you specify both @code{inline} and @code{extern} in the function
5341 definition, then the definition is used only for inlining. In no case
5342 is the function compiled on its own, not even if you refer to its
5343 address explicitly. Such an address becomes an external reference, as
5344 if you had only declared the function, and had not defined it.
5346 This combination of @code{inline} and @code{extern} has almost the
5347 effect of a macro. The way to use it is to put a function definition in
5348 a header file with these keywords, and put another copy of the
5349 definition (lacking @code{inline} and @code{extern}) in a library file.
5350 The definition in the header file will cause most calls to the function
5351 to be inlined. If any uses of the function remain, they will refer to
5352 the single copy in the library.
5355 @section When is a Volatile Object Accessed?
5356 @cindex accessing volatiles
5357 @cindex volatile read
5358 @cindex volatile write
5359 @cindex volatile access
5361 C has the concept of volatile objects. These are normally accessed by
5362 pointers and used for accessing hardware or inter-thread
5363 communication. The standard encourages compilers to refrain from
5364 optimizations concerning accesses to volatile objects, but leaves it
5365 implementation defined as to what constitutes a volatile access. The
5366 minimum requirement is that at a sequence point all previous accesses
5367 to volatile objects have stabilized and no subsequent accesses have
5368 occurred. Thus an implementation is free to reorder and combine
5369 volatile accesses which occur between sequence points, but cannot do
5370 so for accesses across a sequence point. The use of volatile does
5371 not allow you to violate the restriction on updating objects multiple
5372 times between two sequence points.
5374 Accesses to non-volatile objects are not ordered with respect to
5375 volatile accesses. You cannot use a volatile object as a memory
5376 barrier to order a sequence of writes to non-volatile memory. For
5380 int *ptr = @var{something};
5382 *ptr = @var{something};
5386 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
5387 that the write to @var{*ptr} will have occurred by the time the update
5388 of @var{vobj} has happened. If you need this guarantee, you must use
5389 a stronger memory barrier such as:
5392 int *ptr = @var{something};
5394 *ptr = @var{something};
5395 asm volatile ("" : : : "memory");
5399 A scalar volatile object is read when it is accessed in a void context:
5402 volatile int *src = @var{somevalue};
5406 Such expressions are rvalues, and GCC implements this as a
5407 read of the volatile object being pointed to.
5409 Assignments are also expressions and have an rvalue. However when
5410 assigning to a scalar volatile, the volatile object is not reread,
5411 regardless of whether the assignment expression's rvalue is used or
5412 not. If the assignment's rvalue is used, the value is that assigned
5413 to the volatile object. For instance, there is no read of @var{vobj}
5414 in all the following cases:
5419 vobj = @var{something};
5420 obj = vobj = @var{something};
5421 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
5422 obj = (@var{something}, vobj = @var{anotherthing});
5425 If you need to read the volatile object after an assignment has
5426 occurred, you must use a separate expression with an intervening
5429 As bitfields are not individually addressable, volatile bitfields may
5430 be implicitly read when written to, or when adjacent bitfields are
5431 accessed. Bitfield operations may be optimized such that adjacent
5432 bitfields are only partially accessed, if they straddle a storage unit
5433 boundary. For these reasons it is unwise to use volatile bitfields to
5437 @section Assembler Instructions with C Expression Operands
5438 @cindex extended @code{asm}
5439 @cindex @code{asm} expressions
5440 @cindex assembler instructions
5443 In an assembler instruction using @code{asm}, you can specify the
5444 operands of the instruction using C expressions. This means you need not
5445 guess which registers or memory locations will contain the data you want
5448 You must specify an assembler instruction template much like what
5449 appears in a machine description, plus an operand constraint string for
5452 For example, here is how to use the 68881's @code{fsinx} instruction:
5455 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
5459 Here @code{angle} is the C expression for the input operand while
5460 @code{result} is that of the output operand. Each has @samp{"f"} as its
5461 operand constraint, saying that a floating point register is required.
5462 The @samp{=} in @samp{=f} indicates that the operand is an output; all
5463 output operands' constraints must use @samp{=}. The constraints use the
5464 same language used in the machine description (@pxref{Constraints}).
5466 Each operand is described by an operand-constraint string followed by
5467 the C expression in parentheses. A colon separates the assembler
5468 template from the first output operand and another separates the last
5469 output operand from the first input, if any. Commas separate the
5470 operands within each group. The total number of operands is currently
5471 limited to 30; this limitation may be lifted in some future version of
5474 If there are no output operands but there are input operands, you must
5475 place two consecutive colons surrounding the place where the output
5478 As of GCC version 3.1, it is also possible to specify input and output
5479 operands using symbolic names which can be referenced within the
5480 assembler code. These names are specified inside square brackets
5481 preceding the constraint string, and can be referenced inside the
5482 assembler code using @code{%[@var{name}]} instead of a percentage sign
5483 followed by the operand number. Using named operands the above example
5487 asm ("fsinx %[angle],%[output]"
5488 : [output] "=f" (result)
5489 : [angle] "f" (angle));
5493 Note that the symbolic operand names have no relation whatsoever to
5494 other C identifiers. You may use any name you like, even those of
5495 existing C symbols, but you must ensure that no two operands within the same
5496 assembler construct use the same symbolic name.
5498 Output operand expressions must be lvalues; the compiler can check this.
5499 The input operands need not be lvalues. The compiler cannot check
5500 whether the operands have data types that are reasonable for the
5501 instruction being executed. It does not parse the assembler instruction
5502 template and does not know what it means or even whether it is valid
5503 assembler input. The extended @code{asm} feature is most often used for
5504 machine instructions the compiler itself does not know exist. If
5505 the output expression cannot be directly addressed (for example, it is a
5506 bit-field), your constraint must allow a register. In that case, GCC
5507 will use the register as the output of the @code{asm}, and then store
5508 that register into the output.
5510 The ordinary output operands must be write-only; GCC will assume that
5511 the values in these operands before the instruction are dead and need
5512 not be generated. Extended asm supports input-output or read-write
5513 operands. Use the constraint character @samp{+} to indicate such an
5514 operand and list it with the output operands. You should only use
5515 read-write operands when the constraints for the operand (or the
5516 operand in which only some of the bits are to be changed) allow a
5519 You may, as an alternative, logically split its function into two
5520 separate operands, one input operand and one write-only output
5521 operand. The connection between them is expressed by constraints
5522 which say they need to be in the same location when the instruction
5523 executes. You can use the same C expression for both operands, or
5524 different expressions. For example, here we write the (fictitious)
5525 @samp{combine} instruction with @code{bar} as its read-only source
5526 operand and @code{foo} as its read-write destination:
5529 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
5533 The constraint @samp{"0"} for operand 1 says that it must occupy the
5534 same location as operand 0. A number in constraint is allowed only in
5535 an input operand and it must refer to an output operand.
5537 Only a number in the constraint can guarantee that one operand will be in
5538 the same place as another. The mere fact that @code{foo} is the value
5539 of both operands is not enough to guarantee that they will be in the
5540 same place in the generated assembler code. The following would not
5544 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
5547 Various optimizations or reloading could cause operands 0 and 1 to be in
5548 different registers; GCC knows no reason not to do so. For example, the
5549 compiler might find a copy of the value of @code{foo} in one register and
5550 use it for operand 1, but generate the output operand 0 in a different
5551 register (copying it afterward to @code{foo}'s own address). Of course,
5552 since the register for operand 1 is not even mentioned in the assembler
5553 code, the result will not work, but GCC can't tell that.
5555 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
5556 the operand number for a matching constraint. For example:
5559 asm ("cmoveq %1,%2,%[result]"
5560 : [result] "=r"(result)
5561 : "r" (test), "r"(new), "[result]"(old));
5564 Sometimes you need to make an @code{asm} operand be a specific register,
5565 but there's no matching constraint letter for that register @emph{by
5566 itself}. To force the operand into that register, use a local variable
5567 for the operand and specify the register in the variable declaration.
5568 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
5569 register constraint letter that matches the register:
5572 register int *p1 asm ("r0") = @dots{};
5573 register int *p2 asm ("r1") = @dots{};
5574 register int *result asm ("r0");
5575 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5578 @anchor{Example of asm with clobbered asm reg}
5579 In the above example, beware that a register that is call-clobbered by
5580 the target ABI will be overwritten by any function call in the
5581 assignment, including library calls for arithmetic operators.
5582 Also a register may be clobbered when generating some operations,
5583 like variable shift, memory copy or memory move on x86.
5584 Assuming it is a call-clobbered register, this may happen to @code{r0}
5585 above by the assignment to @code{p2}. If you have to use such a
5586 register, use temporary variables for expressions between the register
5591 register int *p1 asm ("r0") = @dots{};
5592 register int *p2 asm ("r1") = t1;
5593 register int *result asm ("r0");
5594 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5597 Some instructions clobber specific hard registers. To describe this,
5598 write a third colon after the input operands, followed by the names of
5599 the clobbered hard registers (given as strings). Here is a realistic
5600 example for the VAX:
5603 asm volatile ("movc3 %0,%1,%2"
5604 : /* @r{no outputs} */
5605 : "g" (from), "g" (to), "g" (count)
5606 : "r0", "r1", "r2", "r3", "r4", "r5");
5609 You may not write a clobber description in a way that overlaps with an
5610 input or output operand. For example, you may not have an operand
5611 describing a register class with one member if you mention that register
5612 in the clobber list. Variables declared to live in specific registers
5613 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
5614 have no part mentioned in the clobber description.
5615 There is no way for you to specify that an input
5616 operand is modified without also specifying it as an output
5617 operand. Note that if all the output operands you specify are for this
5618 purpose (and hence unused), you will then also need to specify
5619 @code{volatile} for the @code{asm} construct, as described below, to
5620 prevent GCC from deleting the @code{asm} statement as unused.
5622 If you refer to a particular hardware register from the assembler code,
5623 you will probably have to list the register after the third colon to
5624 tell the compiler the register's value is modified. In some assemblers,
5625 the register names begin with @samp{%}; to produce one @samp{%} in the
5626 assembler code, you must write @samp{%%} in the input.
5628 If your assembler instruction can alter the condition code register, add
5629 @samp{cc} to the list of clobbered registers. GCC on some machines
5630 represents the condition codes as a specific hardware register;
5631 @samp{cc} serves to name this register. On other machines, the
5632 condition code is handled differently, and specifying @samp{cc} has no
5633 effect. But it is valid no matter what the machine.
5635 If your assembler instructions access memory in an unpredictable
5636 fashion, add @samp{memory} to the list of clobbered registers. This
5637 will cause GCC to not keep memory values cached in registers across the
5638 assembler instruction and not optimize stores or loads to that memory.
5639 You will also want to add the @code{volatile} keyword if the memory
5640 affected is not listed in the inputs or outputs of the @code{asm}, as
5641 the @samp{memory} clobber does not count as a side-effect of the
5642 @code{asm}. If you know how large the accessed memory is, you can add
5643 it as input or output but if this is not known, you should add
5644 @samp{memory}. As an example, if you access ten bytes of a string, you
5645 can use a memory input like:
5648 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
5651 Note that in the following example the memory input is necessary,
5652 otherwise GCC might optimize the store to @code{x} away:
5659 asm ("magic stuff accessing an 'int' pointed to by '%1'"
5660 "=&d" (r) : "a" (y), "m" (*y));
5665 You can put multiple assembler instructions together in a single
5666 @code{asm} template, separated by the characters normally used in assembly
5667 code for the system. A combination that works in most places is a newline
5668 to break the line, plus a tab character to move to the instruction field
5669 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
5670 assembler allows semicolons as a line-breaking character. Note that some
5671 assembler dialects use semicolons to start a comment.
5672 The input operands are guaranteed not to use any of the clobbered
5673 registers, and neither will the output operands' addresses, so you can
5674 read and write the clobbered registers as many times as you like. Here
5675 is an example of multiple instructions in a template; it assumes the
5676 subroutine @code{_foo} accepts arguments in registers 9 and 10:
5679 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
5681 : "g" (from), "g" (to)
5685 Unless an output operand has the @samp{&} constraint modifier, GCC
5686 may allocate it in the same register as an unrelated input operand, on
5687 the assumption the inputs are consumed before the outputs are produced.
5688 This assumption may be false if the assembler code actually consists of
5689 more than one instruction. In such a case, use @samp{&} for each output
5690 operand that may not overlap an input. @xref{Modifiers}.
5692 If you want to test the condition code produced by an assembler
5693 instruction, you must include a branch and a label in the @code{asm}
5694 construct, as follows:
5697 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
5703 This assumes your assembler supports local labels, as the GNU assembler
5704 and most Unix assemblers do.
5706 Speaking of labels, jumps from one @code{asm} to another are not
5707 supported. The compiler's optimizers do not know about these jumps, and
5708 therefore they cannot take account of them when deciding how to
5709 optimize. @xref{Extended asm with goto}.
5711 @cindex macros containing @code{asm}
5712 Usually the most convenient way to use these @code{asm} instructions is to
5713 encapsulate them in macros that look like functions. For example,
5717 (@{ double __value, __arg = (x); \
5718 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
5723 Here the variable @code{__arg} is used to make sure that the instruction
5724 operates on a proper @code{double} value, and to accept only those
5725 arguments @code{x} which can convert automatically to a @code{double}.
5727 Another way to make sure the instruction operates on the correct data
5728 type is to use a cast in the @code{asm}. This is different from using a
5729 variable @code{__arg} in that it converts more different types. For
5730 example, if the desired type were @code{int}, casting the argument to
5731 @code{int} would accept a pointer with no complaint, while assigning the
5732 argument to an @code{int} variable named @code{__arg} would warn about
5733 using a pointer unless the caller explicitly casts it.
5735 If an @code{asm} has output operands, GCC assumes for optimization
5736 purposes the instruction has no side effects except to change the output
5737 operands. This does not mean instructions with a side effect cannot be
5738 used, but you must be careful, because the compiler may eliminate them
5739 if the output operands aren't used, or move them out of loops, or
5740 replace two with one if they constitute a common subexpression. Also,
5741 if your instruction does have a side effect on a variable that otherwise
5742 appears not to change, the old value of the variable may be reused later
5743 if it happens to be found in a register.
5745 You can prevent an @code{asm} instruction from being deleted
5746 by writing the keyword @code{volatile} after
5747 the @code{asm}. For example:
5750 #define get_and_set_priority(new) \
5752 asm volatile ("get_and_set_priority %0, %1" \
5753 : "=g" (__old) : "g" (new)); \
5758 The @code{volatile} keyword indicates that the instruction has
5759 important side-effects. GCC will not delete a volatile @code{asm} if
5760 it is reachable. (The instruction can still be deleted if GCC can
5761 prove that control-flow will never reach the location of the
5762 instruction.) Note that even a volatile @code{asm} instruction
5763 can be moved relative to other code, including across jump
5764 instructions. For example, on many targets there is a system
5765 register which can be set to control the rounding mode of
5766 floating point operations. You might try
5767 setting it with a volatile @code{asm}, like this PowerPC example:
5770 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
5775 This will not work reliably, as the compiler may move the addition back
5776 before the volatile @code{asm}. To make it work you need to add an
5777 artificial dependency to the @code{asm} referencing a variable in the code
5778 you don't want moved, for example:
5781 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
5785 Similarly, you can't expect a
5786 sequence of volatile @code{asm} instructions to remain perfectly
5787 consecutive. If you want consecutive output, use a single @code{asm}.
5788 Also, GCC will perform some optimizations across a volatile @code{asm}
5789 instruction; GCC does not ``forget everything'' when it encounters
5790 a volatile @code{asm} instruction the way some other compilers do.
5792 An @code{asm} instruction without any output operands will be treated
5793 identically to a volatile @code{asm} instruction.
5795 It is a natural idea to look for a way to give access to the condition
5796 code left by the assembler instruction. However, when we attempted to
5797 implement this, we found no way to make it work reliably. The problem
5798 is that output operands might need reloading, which would result in
5799 additional following ``store'' instructions. On most machines, these
5800 instructions would alter the condition code before there was time to
5801 test it. This problem doesn't arise for ordinary ``test'' and
5802 ``compare'' instructions because they don't have any output operands.
5804 For reasons similar to those described above, it is not possible to give
5805 an assembler instruction access to the condition code left by previous
5808 @anchor{Extended asm with goto}
5809 As of GCC version 4.5, @code{asm goto} may be used to have the assembly
5810 jump to one or more C labels. In this form, a fifth section after the
5811 clobber list contains a list of all C labels to which the assembly may jump.
5812 Each label operand is implicitly self-named. The @code{asm} is also assumed
5813 to fall through to the next statement.
5815 This form of @code{asm} is restricted to not have outputs. This is due
5816 to a internal restriction in the compiler that control transfer instructions
5817 cannot have outputs. This restriction on @code{asm goto} may be lifted
5818 in some future version of the compiler. In the mean time, @code{asm goto}
5819 may include a memory clobber, and so leave outputs in memory.
5825 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
5826 : : "r"(x), "r"(&y) : "r5", "memory" : error);
5833 In this (inefficient) example, the @code{frob} instruction sets the
5834 carry bit to indicate an error. The @code{jc} instruction detects
5835 this and branches to the @code{error} label. Finally, the output
5836 of the @code{frob} instruction (@code{%r5}) is stored into the memory
5837 for variable @code{y}, which is later read by the @code{return} statement.
5843 asm goto ("mfsr %%r1, 123; jmp %%r1;"
5844 ".pushsection doit_table;"
5845 ".long %l0, %l1, %l2, %l3;"
5847 : : : "r1" : label1, label2, label3, label4);
5848 __builtin_unreachable ();
5863 In this (also inefficient) example, the @code{mfsr} instruction reads
5864 an address from some out-of-band machine register, and the following
5865 @code{jmp} instruction branches to that address. The address read by
5866 the @code{mfsr} instruction is assumed to have been previously set via
5867 some application-specific mechanism to be one of the four values stored
5868 in the @code{doit_table} section. Finally, the @code{asm} is followed
5869 by a call to @code{__builtin_unreachable} to indicate that the @code{asm}
5870 does not in fact fall through.
5873 #define TRACE1(NUM) \
5875 asm goto ("0: nop;" \
5876 ".pushsection trace_table;" \
5879 : : : : trace#NUM); \
5880 if (0) @{ trace#NUM: trace(); @} \
5882 #define TRACE TRACE1(__COUNTER__)
5885 In this example (which in fact inspired the @code{asm goto} feature)
5886 we want on rare occasions to call the @code{trace} function; on other
5887 occasions we'd like to keep the overhead to the absolute minimum.
5888 The normal code path consists of a single @code{nop} instruction.
5889 However, we record the address of this @code{nop} together with the
5890 address of a label that calls the @code{trace} function. This allows
5891 the @code{nop} instruction to be patched at runtime to be an
5892 unconditional branch to the stored label. It is assumed that an
5893 optimizing compiler will move the labeled block out of line, to
5894 optimize the fall through path from the @code{asm}.
5896 If you are writing a header file that should be includable in ISO C
5897 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
5900 @subsection Size of an @code{asm}
5902 Some targets require that GCC track the size of each instruction used in
5903 order to generate correct code. Because the final length of an
5904 @code{asm} is only known by the assembler, GCC must make an estimate as
5905 to how big it will be. The estimate is formed by counting the number of
5906 statements in the pattern of the @code{asm} and multiplying that by the
5907 length of the longest instruction on that processor. Statements in the
5908 @code{asm} are identified by newline characters and whatever statement
5909 separator characters are supported by the assembler; on most processors
5910 this is the `@code{;}' character.
5912 Normally, GCC's estimate is perfectly adequate to ensure that correct
5913 code is generated, but it is possible to confuse the compiler if you use
5914 pseudo instructions or assembler macros that expand into multiple real
5915 instructions or if you use assembler directives that expand to more
5916 space in the object file than would be needed for a single instruction.
5917 If this happens then the assembler will produce a diagnostic saying that
5918 a label is unreachable.
5920 @subsection i386 floating point asm operands
5922 There are several rules on the usage of stack-like regs in
5923 asm_operands insns. These rules apply only to the operands that are
5928 Given a set of input regs that die in an asm_operands, it is
5929 necessary to know which are implicitly popped by the asm, and
5930 which must be explicitly popped by gcc.
5932 An input reg that is implicitly popped by the asm must be
5933 explicitly clobbered, unless it is constrained to match an
5937 For any input reg that is implicitly popped by an asm, it is
5938 necessary to know how to adjust the stack to compensate for the pop.
5939 If any non-popped input is closer to the top of the reg-stack than
5940 the implicitly popped reg, it would not be possible to know what the
5941 stack looked like---it's not clear how the rest of the stack ``slides
5944 All implicitly popped input regs must be closer to the top of
5945 the reg-stack than any input that is not implicitly popped.
5947 It is possible that if an input dies in an insn, reload might
5948 use the input reg for an output reload. Consider this example:
5951 asm ("foo" : "=t" (a) : "f" (b));
5954 This asm says that input B is not popped by the asm, and that
5955 the asm pushes a result onto the reg-stack, i.e., the stack is one
5956 deeper after the asm than it was before. But, it is possible that
5957 reload will think that it can use the same reg for both the input and
5958 the output, if input B dies in this insn.
5960 If any input operand uses the @code{f} constraint, all output reg
5961 constraints must use the @code{&} earlyclobber.
5963 The asm above would be written as
5966 asm ("foo" : "=&t" (a) : "f" (b));
5970 Some operands need to be in particular places on the stack. All
5971 output operands fall in this category---there is no other way to
5972 know which regs the outputs appear in unless the user indicates
5973 this in the constraints.
5975 Output operands must specifically indicate which reg an output
5976 appears in after an asm. @code{=f} is not allowed: the operand
5977 constraints must select a class with a single reg.
5980 Output operands may not be ``inserted'' between existing stack regs.
5981 Since no 387 opcode uses a read/write operand, all output operands
5982 are dead before the asm_operands, and are pushed by the asm_operands.
5983 It makes no sense to push anywhere but the top of the reg-stack.
5985 Output operands must start at the top of the reg-stack: output
5986 operands may not ``skip'' a reg.
5989 Some asm statements may need extra stack space for internal
5990 calculations. This can be guaranteed by clobbering stack registers
5991 unrelated to the inputs and outputs.
5995 Here are a couple of reasonable asms to want to write. This asm
5996 takes one input, which is internally popped, and produces two outputs.
5999 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
6002 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
6003 and replaces them with one output. The user must code the @code{st(1)}
6004 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
6007 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
6013 @section Controlling Names Used in Assembler Code
6014 @cindex assembler names for identifiers
6015 @cindex names used in assembler code
6016 @cindex identifiers, names in assembler code
6018 You can specify the name to be used in the assembler code for a C
6019 function or variable by writing the @code{asm} (or @code{__asm__})
6020 keyword after the declarator as follows:
6023 int foo asm ("myfoo") = 2;
6027 This specifies that the name to be used for the variable @code{foo} in
6028 the assembler code should be @samp{myfoo} rather than the usual
6031 On systems where an underscore is normally prepended to the name of a C
6032 function or variable, this feature allows you to define names for the
6033 linker that do not start with an underscore.
6035 It does not make sense to use this feature with a non-static local
6036 variable since such variables do not have assembler names. If you are
6037 trying to put the variable in a particular register, see @ref{Explicit
6038 Reg Vars}. GCC presently accepts such code with a warning, but will
6039 probably be changed to issue an error, rather than a warning, in the
6042 You cannot use @code{asm} in this way in a function @emph{definition}; but
6043 you can get the same effect by writing a declaration for the function
6044 before its definition and putting @code{asm} there, like this:
6047 extern func () asm ("FUNC");
6054 It is up to you to make sure that the assembler names you choose do not
6055 conflict with any other assembler symbols. Also, you must not use a
6056 register name; that would produce completely invalid assembler code. GCC
6057 does not as yet have the ability to store static variables in registers.
6058 Perhaps that will be added.
6060 @node Explicit Reg Vars
6061 @section Variables in Specified Registers
6062 @cindex explicit register variables
6063 @cindex variables in specified registers
6064 @cindex specified registers
6065 @cindex registers, global allocation
6067 GNU C allows you to put a few global variables into specified hardware
6068 registers. You can also specify the register in which an ordinary
6069 register variable should be allocated.
6073 Global register variables reserve registers throughout the program.
6074 This may be useful in programs such as programming language
6075 interpreters which have a couple of global variables that are accessed
6079 Local register variables in specific registers do not reserve the
6080 registers, except at the point where they are used as input or output
6081 operands in an @code{asm} statement and the @code{asm} statement itself is
6082 not deleted. The compiler's data flow analysis is capable of determining
6083 where the specified registers contain live values, and where they are
6084 available for other uses. Stores into local register variables may be deleted
6085 when they appear to be dead according to dataflow analysis. References
6086 to local register variables may be deleted or moved or simplified.
6088 These local variables are sometimes convenient for use with the extended
6089 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
6090 output of the assembler instruction directly into a particular register.
6091 (This will work provided the register you specify fits the constraints
6092 specified for that operand in the @code{asm}.)
6100 @node Global Reg Vars
6101 @subsection Defining Global Register Variables
6102 @cindex global register variables
6103 @cindex registers, global variables in
6105 You can define a global register variable in GNU C like this:
6108 register int *foo asm ("a5");
6112 Here @code{a5} is the name of the register which should be used. Choose a
6113 register which is normally saved and restored by function calls on your
6114 machine, so that library routines will not clobber it.
6116 Naturally the register name is cpu-dependent, so you would need to
6117 conditionalize your program according to cpu type. The register
6118 @code{a5} would be a good choice on a 68000 for a variable of pointer
6119 type. On machines with register windows, be sure to choose a ``global''
6120 register that is not affected magically by the function call mechanism.
6122 In addition, operating systems on one type of cpu may differ in how they
6123 name the registers; then you would need additional conditionals. For
6124 example, some 68000 operating systems call this register @code{%a5}.
6126 Eventually there may be a way of asking the compiler to choose a register
6127 automatically, but first we need to figure out how it should choose and
6128 how to enable you to guide the choice. No solution is evident.
6130 Defining a global register variable in a certain register reserves that
6131 register entirely for this use, at least within the current compilation.
6132 The register will not be allocated for any other purpose in the functions
6133 in the current compilation. The register will not be saved and restored by
6134 these functions. Stores into this register are never deleted even if they
6135 would appear to be dead, but references may be deleted or moved or
6138 It is not safe to access the global register variables from signal
6139 handlers, or from more than one thread of control, because the system
6140 library routines may temporarily use the register for other things (unless
6141 you recompile them specially for the task at hand).
6143 @cindex @code{qsort}, and global register variables
6144 It is not safe for one function that uses a global register variable to
6145 call another such function @code{foo} by way of a third function
6146 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
6147 different source file in which the variable wasn't declared). This is
6148 because @code{lose} might save the register and put some other value there.
6149 For example, you can't expect a global register variable to be available in
6150 the comparison-function that you pass to @code{qsort}, since @code{qsort}
6151 might have put something else in that register. (If you are prepared to
6152 recompile @code{qsort} with the same global register variable, you can
6153 solve this problem.)
6155 If you want to recompile @code{qsort} or other source files which do not
6156 actually use your global register variable, so that they will not use that
6157 register for any other purpose, then it suffices to specify the compiler
6158 option @option{-ffixed-@var{reg}}. You need not actually add a global
6159 register declaration to their source code.
6161 A function which can alter the value of a global register variable cannot
6162 safely be called from a function compiled without this variable, because it
6163 could clobber the value the caller expects to find there on return.
6164 Therefore, the function which is the entry point into the part of the
6165 program that uses the global register variable must explicitly save and
6166 restore the value which belongs to its caller.
6168 @cindex register variable after @code{longjmp}
6169 @cindex global register after @code{longjmp}
6170 @cindex value after @code{longjmp}
6173 On most machines, @code{longjmp} will restore to each global register
6174 variable the value it had at the time of the @code{setjmp}. On some
6175 machines, however, @code{longjmp} will not change the value of global
6176 register variables. To be portable, the function that called @code{setjmp}
6177 should make other arrangements to save the values of the global register
6178 variables, and to restore them in a @code{longjmp}. This way, the same
6179 thing will happen regardless of what @code{longjmp} does.
6181 All global register variable declarations must precede all function
6182 definitions. If such a declaration could appear after function
6183 definitions, the declaration would be too late to prevent the register from
6184 being used for other purposes in the preceding functions.
6186 Global register variables may not have initial values, because an
6187 executable file has no means to supply initial contents for a register.
6189 On the SPARC, there are reports that g3 @dots{} g7 are suitable
6190 registers, but certain library functions, such as @code{getwd}, as well
6191 as the subroutines for division and remainder, modify g3 and g4. g1 and
6192 g2 are local temporaries.
6194 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
6195 Of course, it will not do to use more than a few of those.
6197 @node Local Reg Vars
6198 @subsection Specifying Registers for Local Variables
6199 @cindex local variables, specifying registers
6200 @cindex specifying registers for local variables
6201 @cindex registers for local variables
6203 You can define a local register variable with a specified register
6207 register int *foo asm ("a5");
6211 Here @code{a5} is the name of the register which should be used. Note
6212 that this is the same syntax used for defining global register
6213 variables, but for a local variable it would appear within a function.
6215 Naturally the register name is cpu-dependent, but this is not a
6216 problem, since specific registers are most often useful with explicit
6217 assembler instructions (@pxref{Extended Asm}). Both of these things
6218 generally require that you conditionalize your program according to
6221 In addition, operating systems on one type of cpu may differ in how they
6222 name the registers; then you would need additional conditionals. For
6223 example, some 68000 operating systems call this register @code{%a5}.
6225 Defining such a register variable does not reserve the register; it
6226 remains available for other uses in places where flow control determines
6227 the variable's value is not live.
6229 This option does not guarantee that GCC will generate code that has
6230 this variable in the register you specify at all times. You may not
6231 code an explicit reference to this register in the @emph{assembler
6232 instruction template} part of an @code{asm} statement and assume it will
6233 always refer to this variable. However, using the variable as an
6234 @code{asm} @emph{operand} guarantees that the specified register is used
6237 Stores into local register variables may be deleted when they appear to be dead
6238 according to dataflow analysis. References to local register variables may
6239 be deleted or moved or simplified.
6241 As for global register variables, it's recommended that you choose a
6242 register which is normally saved and restored by function calls on
6243 your machine, so that library routines will not clobber it. A common
6244 pitfall is to initialize multiple call-clobbered registers with
6245 arbitrary expressions, where a function call or library call for an
6246 arithmetic operator will overwrite a register value from a previous
6247 assignment, for example @code{r0} below:
6249 register int *p1 asm ("r0") = @dots{};
6250 register int *p2 asm ("r1") = @dots{};
6252 In those cases, a solution is to use a temporary variable for
6253 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
6255 @node Alternate Keywords
6256 @section Alternate Keywords
6257 @cindex alternate keywords
6258 @cindex keywords, alternate
6260 @option{-ansi} and the various @option{-std} options disable certain
6261 keywords. This causes trouble when you want to use GNU C extensions, or
6262 a general-purpose header file that should be usable by all programs,
6263 including ISO C programs. The keywords @code{asm}, @code{typeof} and
6264 @code{inline} are not available in programs compiled with
6265 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
6266 program compiled with @option{-std=c99} or @option{-std=c1x}). The
6268 @code{restrict} is only available when @option{-std=gnu99} (which will
6269 eventually be the default) or @option{-std=c99} (or the equivalent
6270 @option{-std=iso9899:1999}), or an option for a later standard
6273 The way to solve these problems is to put @samp{__} at the beginning and
6274 end of each problematical keyword. For example, use @code{__asm__}
6275 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
6277 Other C compilers won't accept these alternative keywords; if you want to
6278 compile with another compiler, you can define the alternate keywords as
6279 macros to replace them with the customary keywords. It looks like this:
6287 @findex __extension__
6289 @option{-pedantic} and other options cause warnings for many GNU C extensions.
6291 prevent such warnings within one expression by writing
6292 @code{__extension__} before the expression. @code{__extension__} has no
6293 effect aside from this.
6295 @node Incomplete Enums
6296 @section Incomplete @code{enum} Types
6298 You can define an @code{enum} tag without specifying its possible values.
6299 This results in an incomplete type, much like what you get if you write
6300 @code{struct foo} without describing the elements. A later declaration
6301 which does specify the possible values completes the type.
6303 You can't allocate variables or storage using the type while it is
6304 incomplete. However, you can work with pointers to that type.
6306 This extension may not be very useful, but it makes the handling of
6307 @code{enum} more consistent with the way @code{struct} and @code{union}
6310 This extension is not supported by GNU C++.
6312 @node Function Names
6313 @section Function Names as Strings
6314 @cindex @code{__func__} identifier
6315 @cindex @code{__FUNCTION__} identifier
6316 @cindex @code{__PRETTY_FUNCTION__} identifier
6318 GCC provides three magic variables which hold the name of the current
6319 function, as a string. The first of these is @code{__func__}, which
6320 is part of the C99 standard:
6322 The identifier @code{__func__} is implicitly declared by the translator
6323 as if, immediately following the opening brace of each function
6324 definition, the declaration
6327 static const char __func__[] = "function-name";
6331 appeared, where function-name is the name of the lexically-enclosing
6332 function. This name is the unadorned name of the function.
6334 @code{__FUNCTION__} is another name for @code{__func__}. Older
6335 versions of GCC recognize only this name. However, it is not
6336 standardized. For maximum portability, we recommend you use
6337 @code{__func__}, but provide a fallback definition with the
6341 #if __STDC_VERSION__ < 199901L
6343 # define __func__ __FUNCTION__
6345 # define __func__ "<unknown>"
6350 In C, @code{__PRETTY_FUNCTION__} is yet another name for
6351 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
6352 the type signature of the function as well as its bare name. For
6353 example, this program:
6357 extern int printf (char *, ...);
6364 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
6365 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
6383 __PRETTY_FUNCTION__ = void a::sub(int)
6386 These identifiers are not preprocessor macros. In GCC 3.3 and
6387 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
6388 were treated as string literals; they could be used to initialize
6389 @code{char} arrays, and they could be concatenated with other string
6390 literals. GCC 3.4 and later treat them as variables, like
6391 @code{__func__}. In C++, @code{__FUNCTION__} and
6392 @code{__PRETTY_FUNCTION__} have always been variables.
6394 @node Return Address
6395 @section Getting the Return or Frame Address of a Function
6397 These functions may be used to get information about the callers of a
6400 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
6401 This function returns the return address of the current function, or of
6402 one of its callers. The @var{level} argument is number of frames to
6403 scan up the call stack. A value of @code{0} yields the return address
6404 of the current function, a value of @code{1} yields the return address
6405 of the caller of the current function, and so forth. When inlining
6406 the expected behavior is that the function will return the address of
6407 the function that will be returned to. To work around this behavior use
6408 the @code{noinline} function attribute.
6410 The @var{level} argument must be a constant integer.
6412 On some machines it may be impossible to determine the return address of
6413 any function other than the current one; in such cases, or when the top
6414 of the stack has been reached, this function will return @code{0} or a
6415 random value. In addition, @code{__builtin_frame_address} may be used
6416 to determine if the top of the stack has been reached.
6418 Additional post-processing of the returned value may be needed, see
6419 @code{__builtin_extract_return_address}.
6421 This function should only be used with a nonzero argument for debugging
6425 @deftypefn {Built-in Function} {void *} __builtin_extract_return_address (void *@var{addr})
6426 The address as returned by @code{__builtin_return_address} may have to be fed
6427 through this function to get the actual encoded address. For example, on the
6428 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
6429 platforms an offset has to be added for the true next instruction to be
6432 If no fixup is needed, this function simply passes through @var{addr}.
6435 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
6436 This function does the reverse of @code{__builtin_extract_return_address}.
6439 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
6440 This function is similar to @code{__builtin_return_address}, but it
6441 returns the address of the function frame rather than the return address
6442 of the function. Calling @code{__builtin_frame_address} with a value of
6443 @code{0} yields the frame address of the current function, a value of
6444 @code{1} yields the frame address of the caller of the current function,
6447 The frame is the area on the stack which holds local variables and saved
6448 registers. The frame address is normally the address of the first word
6449 pushed on to the stack by the function. However, the exact definition
6450 depends upon the processor and the calling convention. If the processor
6451 has a dedicated frame pointer register, and the function has a frame,
6452 then @code{__builtin_frame_address} will return the value of the frame
6455 On some machines it may be impossible to determine the frame address of
6456 any function other than the current one; in such cases, or when the top
6457 of the stack has been reached, this function will return @code{0} if
6458 the first frame pointer is properly initialized by the startup code.
6460 This function should only be used with a nonzero argument for debugging
6464 @node Vector Extensions
6465 @section Using vector instructions through built-in functions
6467 On some targets, the instruction set contains SIMD vector instructions that
6468 operate on multiple values contained in one large register at the same time.
6469 For example, on the i386 the MMX, 3DNow!@: and SSE extensions can be used
6472 The first step in using these extensions is to provide the necessary data
6473 types. This should be done using an appropriate @code{typedef}:
6476 typedef int v4si __attribute__ ((vector_size (16)));
6479 The @code{int} type specifies the base type, while the attribute specifies
6480 the vector size for the variable, measured in bytes. For example, the
6481 declaration above causes the compiler to set the mode for the @code{v4si}
6482 type to be 16 bytes wide and divided into @code{int} sized units. For
6483 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
6484 corresponding mode of @code{foo} will be @acronym{V4SI}.
6486 The @code{vector_size} attribute is only applicable to integral and
6487 float scalars, although arrays, pointers, and function return values
6488 are allowed in conjunction with this construct.
6490 All the basic integer types can be used as base types, both as signed
6491 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
6492 @code{long long}. In addition, @code{float} and @code{double} can be
6493 used to build floating-point vector types.
6495 Specifying a combination that is not valid for the current architecture
6496 will cause GCC to synthesize the instructions using a narrower mode.
6497 For example, if you specify a variable of type @code{V4SI} and your
6498 architecture does not allow for this specific SIMD type, GCC will
6499 produce code that uses 4 @code{SIs}.
6501 The types defined in this manner can be used with a subset of normal C
6502 operations. Currently, GCC will allow using the following operators
6503 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
6505 The operations behave like C++ @code{valarrays}. Addition is defined as
6506 the addition of the corresponding elements of the operands. For
6507 example, in the code below, each of the 4 elements in @var{a} will be
6508 added to the corresponding 4 elements in @var{b} and the resulting
6509 vector will be stored in @var{c}.
6512 typedef int v4si __attribute__ ((vector_size (16)));
6519 Subtraction, multiplication, division, and the logical operations
6520 operate in a similar manner. Likewise, the result of using the unary
6521 minus or complement operators on a vector type is a vector whose
6522 elements are the negative or complemented values of the corresponding
6523 elements in the operand.
6525 In C it is possible to use shifting operators @code{<<}, @code{>>} on
6526 integer-type vectors. The operation is defined as following: @code{@{a0,
6527 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
6528 @dots{}, an >> bn@}}@. Vector operands must have the same number of
6529 elements. Additionally second operands can be a scalar integer in which
6530 case the scalar is converted to the type used by the vector operand (with
6531 possible truncation) and each element of this new vector is the scalar's
6533 Consider the following code.
6536 typedef int v4si __attribute__ ((vector_size (16)));
6540 b = a >> 1; /* b = a >> @{1,1,1,1@}; */
6543 In C vectors can be subscripted as if the vector were an array with
6544 the same number of elements and base type. Out of bound accesses
6545 invoke undefined behavior at runtime. Warnings for out of bound
6546 accesses for vector subscription can be enabled with
6547 @option{-Warray-bounds}.
6549 You can declare variables and use them in function calls and returns, as
6550 well as in assignments and some casts. You can specify a vector type as
6551 a return type for a function. Vector types can also be used as function
6552 arguments. It is possible to cast from one vector type to another,
6553 provided they are of the same size (in fact, you can also cast vectors
6554 to and from other datatypes of the same size).
6556 You cannot operate between vectors of different lengths or different
6557 signedness without a cast.
6559 A port that supports hardware vector operations, usually provides a set
6560 of built-in functions that can be used to operate on vectors. For
6561 example, a function to add two vectors and multiply the result by a
6562 third could look like this:
6565 v4si f (v4si a, v4si b, v4si c)
6567 v4si tmp = __builtin_addv4si (a, b);
6568 return __builtin_mulv4si (tmp, c);
6575 @findex __builtin_offsetof
6577 GCC implements for both C and C++ a syntactic extension to implement
6578 the @code{offsetof} macro.
6582 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
6584 offsetof_member_designator:
6586 | offsetof_member_designator "." @code{identifier}
6587 | offsetof_member_designator "[" @code{expr} "]"
6590 This extension is sufficient such that
6593 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
6596 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
6597 may be dependent. In either case, @var{member} may consist of a single
6598 identifier, or a sequence of member accesses and array references.
6600 @node Atomic Builtins
6601 @section Built-in functions for atomic memory access
6603 The following builtins are intended to be compatible with those described
6604 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
6605 section 7.4. As such, they depart from the normal GCC practice of using
6606 the ``__builtin_'' prefix, and further that they are overloaded such that
6607 they work on multiple types.
6609 The definition given in the Intel documentation allows only for the use of
6610 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
6611 counterparts. GCC will allow any integral scalar or pointer type that is
6612 1, 2, 4 or 8 bytes in length.
6614 Not all operations are supported by all target processors. If a particular
6615 operation cannot be implemented on the target processor, a warning will be
6616 generated and a call an external function will be generated. The external
6617 function will carry the same name as the builtin, with an additional suffix
6618 @samp{_@var{n}} where @var{n} is the size of the data type.
6620 @c ??? Should we have a mechanism to suppress this warning? This is almost
6621 @c useful for implementing the operation under the control of an external
6624 In most cases, these builtins are considered a @dfn{full barrier}. That is,
6625 no memory operand will be moved across the operation, either forward or
6626 backward. Further, instructions will be issued as necessary to prevent the
6627 processor from speculating loads across the operation and from queuing stores
6628 after the operation.
6630 All of the routines are described in the Intel documentation to take
6631 ``an optional list of variables protected by the memory barrier''. It's
6632 not clear what is meant by that; it could mean that @emph{only} the
6633 following variables are protected, or it could mean that these variables
6634 should in addition be protected. At present GCC ignores this list and
6635 protects all variables which are globally accessible. If in the future
6636 we make some use of this list, an empty list will continue to mean all
6637 globally accessible variables.
6640 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
6641 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
6642 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
6643 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
6644 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
6645 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
6646 @findex __sync_fetch_and_add
6647 @findex __sync_fetch_and_sub
6648 @findex __sync_fetch_and_or
6649 @findex __sync_fetch_and_and
6650 @findex __sync_fetch_and_xor
6651 @findex __sync_fetch_and_nand
6652 These builtins perform the operation suggested by the name, and
6653 returns the value that had previously been in memory. That is,
6656 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
6657 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
6660 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
6661 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
6663 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
6664 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
6665 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
6666 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
6667 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
6668 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
6669 @findex __sync_add_and_fetch
6670 @findex __sync_sub_and_fetch
6671 @findex __sync_or_and_fetch
6672 @findex __sync_and_and_fetch
6673 @findex __sync_xor_and_fetch
6674 @findex __sync_nand_and_fetch
6675 These builtins perform the operation suggested by the name, and
6676 return the new value. That is,
6679 @{ *ptr @var{op}= value; return *ptr; @}
6680 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
6683 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
6684 builtin as @code{*ptr = ~(*ptr & value)} instead of
6685 @code{*ptr = ~*ptr & value}.
6687 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
6688 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
6689 @findex __sync_bool_compare_and_swap
6690 @findex __sync_val_compare_and_swap
6691 These builtins perform an atomic compare and swap. That is, if the current
6692 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
6695 The ``bool'' version returns true if the comparison is successful and
6696 @var{newval} was written. The ``val'' version returns the contents
6697 of @code{*@var{ptr}} before the operation.
6699 @item __sync_synchronize (...)
6700 @findex __sync_synchronize
6701 This builtin issues a full memory barrier.
6703 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
6704 @findex __sync_lock_test_and_set
6705 This builtin, as described by Intel, is not a traditional test-and-set
6706 operation, but rather an atomic exchange operation. It writes @var{value}
6707 into @code{*@var{ptr}}, and returns the previous contents of
6710 Many targets have only minimal support for such locks, and do not support
6711 a full exchange operation. In this case, a target may support reduced
6712 functionality here by which the @emph{only} valid value to store is the
6713 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
6714 is implementation defined.
6716 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
6717 This means that references after the builtin cannot move to (or be
6718 speculated to) before the builtin, but previous memory stores may not
6719 be globally visible yet, and previous memory loads may not yet be
6722 @item void __sync_lock_release (@var{type} *ptr, ...)
6723 @findex __sync_lock_release
6724 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
6725 Normally this means writing the constant 0 to @code{*@var{ptr}}.
6727 This builtin is not a full barrier, but rather a @dfn{release barrier}.
6728 This means that all previous memory stores are globally visible, and all
6729 previous memory loads have been satisfied, but following memory reads
6730 are not prevented from being speculated to before the barrier.
6733 @node Object Size Checking
6734 @section Object Size Checking Builtins
6735 @findex __builtin_object_size
6736 @findex __builtin___memcpy_chk
6737 @findex __builtin___mempcpy_chk
6738 @findex __builtin___memmove_chk
6739 @findex __builtin___memset_chk
6740 @findex __builtin___strcpy_chk
6741 @findex __builtin___stpcpy_chk
6742 @findex __builtin___strncpy_chk
6743 @findex __builtin___strcat_chk
6744 @findex __builtin___strncat_chk
6745 @findex __builtin___sprintf_chk
6746 @findex __builtin___snprintf_chk
6747 @findex __builtin___vsprintf_chk
6748 @findex __builtin___vsnprintf_chk
6749 @findex __builtin___printf_chk
6750 @findex __builtin___vprintf_chk
6751 @findex __builtin___fprintf_chk
6752 @findex __builtin___vfprintf_chk
6754 GCC implements a limited buffer overflow protection mechanism
6755 that can prevent some buffer overflow attacks.
6757 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
6758 is a built-in construct that returns a constant number of bytes from
6759 @var{ptr} to the end of the object @var{ptr} pointer points to
6760 (if known at compile time). @code{__builtin_object_size} never evaluates
6761 its arguments for side-effects. If there are any side-effects in them, it
6762 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
6763 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
6764 point to and all of them are known at compile time, the returned number
6765 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
6766 0 and minimum if nonzero. If it is not possible to determine which objects
6767 @var{ptr} points to at compile time, @code{__builtin_object_size} should
6768 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
6769 for @var{type} 2 or 3.
6771 @var{type} is an integer constant from 0 to 3. If the least significant
6772 bit is clear, objects are whole variables, if it is set, a closest
6773 surrounding subobject is considered the object a pointer points to.
6774 The second bit determines if maximum or minimum of remaining bytes
6778 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
6779 char *p = &var.buf1[1], *q = &var.b;
6781 /* Here the object p points to is var. */
6782 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
6783 /* The subobject p points to is var.buf1. */
6784 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
6785 /* The object q points to is var. */
6786 assert (__builtin_object_size (q, 0)
6787 == (char *) (&var + 1) - (char *) &var.b);
6788 /* The subobject q points to is var.b. */
6789 assert (__builtin_object_size (q, 1) == sizeof (var.b));
6793 There are built-in functions added for many common string operation
6794 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
6795 built-in is provided. This built-in has an additional last argument,
6796 which is the number of bytes remaining in object the @var{dest}
6797 argument points to or @code{(size_t) -1} if the size is not known.
6799 The built-in functions are optimized into the normal string functions
6800 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
6801 it is known at compile time that the destination object will not
6802 be overflown. If the compiler can determine at compile time the
6803 object will be always overflown, it issues a warning.
6805 The intended use can be e.g.
6809 #define bos0(dest) __builtin_object_size (dest, 0)
6810 #define memcpy(dest, src, n) \
6811 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
6815 /* It is unknown what object p points to, so this is optimized
6816 into plain memcpy - no checking is possible. */
6817 memcpy (p, "abcde", n);
6818 /* Destination is known and length too. It is known at compile
6819 time there will be no overflow. */
6820 memcpy (&buf[5], "abcde", 5);
6821 /* Destination is known, but the length is not known at compile time.
6822 This will result in __memcpy_chk call that can check for overflow
6824 memcpy (&buf[5], "abcde", n);
6825 /* Destination is known and it is known at compile time there will
6826 be overflow. There will be a warning and __memcpy_chk call that
6827 will abort the program at runtime. */
6828 memcpy (&buf[6], "abcde", 5);
6831 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
6832 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
6833 @code{strcat} and @code{strncat}.
6835 There are also checking built-in functions for formatted output functions.
6837 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
6838 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6839 const char *fmt, ...);
6840 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
6842 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6843 const char *fmt, va_list ap);
6846 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
6847 etc.@: functions and can contain implementation specific flags on what
6848 additional security measures the checking function might take, such as
6849 handling @code{%n} differently.
6851 The @var{os} argument is the object size @var{s} points to, like in the
6852 other built-in functions. There is a small difference in the behavior
6853 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
6854 optimized into the non-checking functions only if @var{flag} is 0, otherwise
6855 the checking function is called with @var{os} argument set to
6858 In addition to this, there are checking built-in functions
6859 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
6860 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
6861 These have just one additional argument, @var{flag}, right before
6862 format string @var{fmt}. If the compiler is able to optimize them to
6863 @code{fputc} etc.@: functions, it will, otherwise the checking function
6864 should be called and the @var{flag} argument passed to it.
6866 @node Other Builtins
6867 @section Other built-in functions provided by GCC
6868 @cindex built-in functions
6869 @findex __builtin_fpclassify
6870 @findex __builtin_isfinite
6871 @findex __builtin_isnormal
6872 @findex __builtin_isgreater
6873 @findex __builtin_isgreaterequal
6874 @findex __builtin_isinf_sign
6875 @findex __builtin_isless
6876 @findex __builtin_islessequal
6877 @findex __builtin_islessgreater
6878 @findex __builtin_isunordered
6879 @findex __builtin_powi
6880 @findex __builtin_powif
6881 @findex __builtin_powil
7039 @findex fprintf_unlocked
7041 @findex fputs_unlocked
7158 @findex printf_unlocked
7190 @findex significandf
7191 @findex significandl
7262 GCC provides a large number of built-in functions other than the ones
7263 mentioned above. Some of these are for internal use in the processing
7264 of exceptions or variable-length argument lists and will not be
7265 documented here because they may change from time to time; we do not
7266 recommend general use of these functions.
7268 The remaining functions are provided for optimization purposes.
7270 @opindex fno-builtin
7271 GCC includes built-in versions of many of the functions in the standard
7272 C library. The versions prefixed with @code{__builtin_} will always be
7273 treated as having the same meaning as the C library function even if you
7274 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
7275 Many of these functions are only optimized in certain cases; if they are
7276 not optimized in a particular case, a call to the library function will
7281 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
7282 @option{-std=c99} or @option{-std=c1x}), the functions
7283 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
7284 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
7285 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
7286 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
7287 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
7288 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
7289 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
7290 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
7291 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
7292 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
7293 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
7294 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
7295 @code{signbitd64}, @code{signbitd128}, @code{significandf},
7296 @code{significandl}, @code{significand}, @code{sincosf},
7297 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
7298 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
7299 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
7300 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
7302 may be handled as built-in functions.
7303 All these functions have corresponding versions
7304 prefixed with @code{__builtin_}, which may be used even in strict C90
7307 The ISO C99 functions
7308 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
7309 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
7310 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
7311 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
7312 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
7313 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
7314 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
7315 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
7316 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
7317 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
7318 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
7319 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
7320 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
7321 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
7322 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
7323 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
7324 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
7325 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
7326 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
7327 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
7328 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
7329 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
7330 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
7331 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
7332 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
7333 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
7334 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
7335 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
7336 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
7337 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
7338 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
7339 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
7340 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
7341 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
7342 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
7343 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
7344 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
7345 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
7346 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
7347 are handled as built-in functions
7348 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7350 There are also built-in versions of the ISO C99 functions
7351 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
7352 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
7353 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
7354 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
7355 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
7356 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
7357 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
7358 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
7359 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
7360 that are recognized in any mode since ISO C90 reserves these names for
7361 the purpose to which ISO C99 puts them. All these functions have
7362 corresponding versions prefixed with @code{__builtin_}.
7364 The ISO C94 functions
7365 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
7366 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
7367 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
7369 are handled as built-in functions
7370 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7372 The ISO C90 functions
7373 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
7374 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
7375 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
7376 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
7377 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
7378 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
7379 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
7380 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
7381 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
7382 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
7383 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
7384 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
7385 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
7386 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
7387 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
7388 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
7389 are all recognized as built-in functions unless
7390 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
7391 is specified for an individual function). All of these functions have
7392 corresponding versions prefixed with @code{__builtin_}.
7394 GCC provides built-in versions of the ISO C99 floating point comparison
7395 macros that avoid raising exceptions for unordered operands. They have
7396 the same names as the standard macros ( @code{isgreater},
7397 @code{isgreaterequal}, @code{isless}, @code{islessequal},
7398 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
7399 prefixed. We intend for a library implementor to be able to simply
7400 @code{#define} each standard macro to its built-in equivalent.
7401 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
7402 @code{isinf_sign} and @code{isnormal} built-ins used with
7403 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
7404 builtins appear both with and without the @code{__builtin_} prefix.
7406 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
7408 You can use the built-in function @code{__builtin_types_compatible_p} to
7409 determine whether two types are the same.
7411 This built-in function returns 1 if the unqualified versions of the
7412 types @var{type1} and @var{type2} (which are types, not expressions) are
7413 compatible, 0 otherwise. The result of this built-in function can be
7414 used in integer constant expressions.
7416 This built-in function ignores top level qualifiers (e.g., @code{const},
7417 @code{volatile}). For example, @code{int} is equivalent to @code{const
7420 The type @code{int[]} and @code{int[5]} are compatible. On the other
7421 hand, @code{int} and @code{char *} are not compatible, even if the size
7422 of their types, on the particular architecture are the same. Also, the
7423 amount of pointer indirection is taken into account when determining
7424 similarity. Consequently, @code{short *} is not similar to
7425 @code{short **}. Furthermore, two types that are typedefed are
7426 considered compatible if their underlying types are compatible.
7428 An @code{enum} type is not considered to be compatible with another
7429 @code{enum} type even if both are compatible with the same integer
7430 type; this is what the C standard specifies.
7431 For example, @code{enum @{foo, bar@}} is not similar to
7432 @code{enum @{hot, dog@}}.
7434 You would typically use this function in code whose execution varies
7435 depending on the arguments' types. For example:
7440 typeof (x) tmp = (x); \
7441 if (__builtin_types_compatible_p (typeof (x), long double)) \
7442 tmp = foo_long_double (tmp); \
7443 else if (__builtin_types_compatible_p (typeof (x), double)) \
7444 tmp = foo_double (tmp); \
7445 else if (__builtin_types_compatible_p (typeof (x), float)) \
7446 tmp = foo_float (tmp); \
7453 @emph{Note:} This construct is only available for C@.
7457 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
7459 You can use the built-in function @code{__builtin_choose_expr} to
7460 evaluate code depending on the value of a constant expression. This
7461 built-in function returns @var{exp1} if @var{const_exp}, which is an
7462 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
7464 This built-in function is analogous to the @samp{? :} operator in C,
7465 except that the expression returned has its type unaltered by promotion
7466 rules. Also, the built-in function does not evaluate the expression
7467 that was not chosen. For example, if @var{const_exp} evaluates to true,
7468 @var{exp2} is not evaluated even if it has side-effects.
7470 This built-in function can return an lvalue if the chosen argument is an
7473 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
7474 type. Similarly, if @var{exp2} is returned, its return type is the same
7481 __builtin_choose_expr ( \
7482 __builtin_types_compatible_p (typeof (x), double), \
7484 __builtin_choose_expr ( \
7485 __builtin_types_compatible_p (typeof (x), float), \
7487 /* @r{The void expression results in a compile-time error} \
7488 @r{when assigning the result to something.} */ \
7492 @emph{Note:} This construct is only available for C@. Furthermore, the
7493 unused expression (@var{exp1} or @var{exp2} depending on the value of
7494 @var{const_exp}) may still generate syntax errors. This may change in
7499 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
7500 You can use the built-in function @code{__builtin_constant_p} to
7501 determine if a value is known to be constant at compile-time and hence
7502 that GCC can perform constant-folding on expressions involving that
7503 value. The argument of the function is the value to test. The function
7504 returns the integer 1 if the argument is known to be a compile-time
7505 constant and 0 if it is not known to be a compile-time constant. A
7506 return of 0 does not indicate that the value is @emph{not} a constant,
7507 but merely that GCC cannot prove it is a constant with the specified
7508 value of the @option{-O} option.
7510 You would typically use this function in an embedded application where
7511 memory was a critical resource. If you have some complex calculation,
7512 you may want it to be folded if it involves constants, but need to call
7513 a function if it does not. For example:
7516 #define Scale_Value(X) \
7517 (__builtin_constant_p (X) \
7518 ? ((X) * SCALE + OFFSET) : Scale (X))
7521 You may use this built-in function in either a macro or an inline
7522 function. However, if you use it in an inlined function and pass an
7523 argument of the function as the argument to the built-in, GCC will
7524 never return 1 when you call the inline function with a string constant
7525 or compound literal (@pxref{Compound Literals}) and will not return 1
7526 when you pass a constant numeric value to the inline function unless you
7527 specify the @option{-O} option.
7529 You may also use @code{__builtin_constant_p} in initializers for static
7530 data. For instance, you can write
7533 static const int table[] = @{
7534 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
7540 This is an acceptable initializer even if @var{EXPRESSION} is not a
7541 constant expression, including the case where
7542 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
7543 folded to a constant but @var{EXPRESSION} contains operands that would
7544 not otherwise be permitted in a static initializer (for example,
7545 @code{0 && foo ()}). GCC must be more conservative about evaluating the
7546 built-in in this case, because it has no opportunity to perform
7549 Previous versions of GCC did not accept this built-in in data
7550 initializers. The earliest version where it is completely safe is
7554 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
7555 @opindex fprofile-arcs
7556 You may use @code{__builtin_expect} to provide the compiler with
7557 branch prediction information. In general, you should prefer to
7558 use actual profile feedback for this (@option{-fprofile-arcs}), as
7559 programmers are notoriously bad at predicting how their programs
7560 actually perform. However, there are applications in which this
7561 data is hard to collect.
7563 The return value is the value of @var{exp}, which should be an integral
7564 expression. The semantics of the built-in are that it is expected that
7565 @var{exp} == @var{c}. For example:
7568 if (__builtin_expect (x, 0))
7573 would indicate that we do not expect to call @code{foo}, since
7574 we expect @code{x} to be zero. Since you are limited to integral
7575 expressions for @var{exp}, you should use constructions such as
7578 if (__builtin_expect (ptr != NULL, 1))
7583 when testing pointer or floating-point values.
7586 @deftypefn {Built-in Function} void __builtin_trap (void)
7587 This function causes the program to exit abnormally. GCC implements
7588 this function by using a target-dependent mechanism (such as
7589 intentionally executing an illegal instruction) or by calling
7590 @code{abort}. The mechanism used may vary from release to release so
7591 you should not rely on any particular implementation.
7594 @deftypefn {Built-in Function} void __builtin_unreachable (void)
7595 If control flow reaches the point of the @code{__builtin_unreachable},
7596 the program is undefined. It is useful in situations where the
7597 compiler cannot deduce the unreachability of the code.
7599 One such case is immediately following an @code{asm} statement that
7600 will either never terminate, or one that transfers control elsewhere
7601 and never returns. In this example, without the
7602 @code{__builtin_unreachable}, GCC would issue a warning that control
7603 reaches the end of a non-void function. It would also generate code
7604 to return after the @code{asm}.
7607 int f (int c, int v)
7615 asm("jmp error_handler");
7616 __builtin_unreachable ();
7621 Because the @code{asm} statement unconditionally transfers control out
7622 of the function, control will never reach the end of the function
7623 body. The @code{__builtin_unreachable} is in fact unreachable and
7624 communicates this fact to the compiler.
7626 Another use for @code{__builtin_unreachable} is following a call a
7627 function that never returns but that is not declared
7628 @code{__attribute__((noreturn))}, as in this example:
7631 void function_that_never_returns (void);
7641 function_that_never_returns ();
7642 __builtin_unreachable ();
7649 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
7650 This function is used to flush the processor's instruction cache for
7651 the region of memory between @var{begin} inclusive and @var{end}
7652 exclusive. Some targets require that the instruction cache be
7653 flushed, after modifying memory containing code, in order to obtain
7654 deterministic behavior.
7656 If the target does not require instruction cache flushes,
7657 @code{__builtin___clear_cache} has no effect. Otherwise either
7658 instructions are emitted in-line to clear the instruction cache or a
7659 call to the @code{__clear_cache} function in libgcc is made.
7662 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
7663 This function is used to minimize cache-miss latency by moving data into
7664 a cache before it is accessed.
7665 You can insert calls to @code{__builtin_prefetch} into code for which
7666 you know addresses of data in memory that is likely to be accessed soon.
7667 If the target supports them, data prefetch instructions will be generated.
7668 If the prefetch is done early enough before the access then the data will
7669 be in the cache by the time it is accessed.
7671 The value of @var{addr} is the address of the memory to prefetch.
7672 There are two optional arguments, @var{rw} and @var{locality}.
7673 The value of @var{rw} is a compile-time constant one or zero; one
7674 means that the prefetch is preparing for a write to the memory address
7675 and zero, the default, means that the prefetch is preparing for a read.
7676 The value @var{locality} must be a compile-time constant integer between
7677 zero and three. A value of zero means that the data has no temporal
7678 locality, so it need not be left in the cache after the access. A value
7679 of three means that the data has a high degree of temporal locality and
7680 should be left in all levels of cache possible. Values of one and two
7681 mean, respectively, a low or moderate degree of temporal locality. The
7685 for (i = 0; i < n; i++)
7688 __builtin_prefetch (&a[i+j], 1, 1);
7689 __builtin_prefetch (&b[i+j], 0, 1);
7694 Data prefetch does not generate faults if @var{addr} is invalid, but
7695 the address expression itself must be valid. For example, a prefetch
7696 of @code{p->next} will not fault if @code{p->next} is not a valid
7697 address, but evaluation will fault if @code{p} is not a valid address.
7699 If the target does not support data prefetch, the address expression
7700 is evaluated if it includes side effects but no other code is generated
7701 and GCC does not issue a warning.
7704 @deftypefn {Built-in Function} double __builtin_huge_val (void)
7705 Returns a positive infinity, if supported by the floating-point format,
7706 else @code{DBL_MAX}. This function is suitable for implementing the
7707 ISO C macro @code{HUGE_VAL}.
7710 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
7711 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
7714 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
7715 Similar to @code{__builtin_huge_val}, except the return
7716 type is @code{long double}.
7719 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
7720 This built-in implements the C99 fpclassify functionality. The first
7721 five int arguments should be the target library's notion of the
7722 possible FP classes and are used for return values. They must be
7723 constant values and they must appear in this order: @code{FP_NAN},
7724 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
7725 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
7726 to classify. GCC treats the last argument as type-generic, which
7727 means it does not do default promotion from float to double.
7730 @deftypefn {Built-in Function} double __builtin_inf (void)
7731 Similar to @code{__builtin_huge_val}, except a warning is generated
7732 if the target floating-point format does not support infinities.
7735 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
7736 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
7739 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
7740 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
7743 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
7744 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
7747 @deftypefn {Built-in Function} float __builtin_inff (void)
7748 Similar to @code{__builtin_inf}, except the return type is @code{float}.
7749 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
7752 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
7753 Similar to @code{__builtin_inf}, except the return
7754 type is @code{long double}.
7757 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
7758 Similar to @code{isinf}, except the return value will be negative for
7759 an argument of @code{-Inf}. Note while the parameter list is an
7760 ellipsis, this function only accepts exactly one floating point
7761 argument. GCC treats this parameter as type-generic, which means it
7762 does not do default promotion from float to double.
7765 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
7766 This is an implementation of the ISO C99 function @code{nan}.
7768 Since ISO C99 defines this function in terms of @code{strtod}, which we
7769 do not implement, a description of the parsing is in order. The string
7770 is parsed as by @code{strtol}; that is, the base is recognized by
7771 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
7772 in the significand such that the least significant bit of the number
7773 is at the least significant bit of the significand. The number is
7774 truncated to fit the significand field provided. The significand is
7775 forced to be a quiet NaN@.
7777 This function, if given a string literal all of which would have been
7778 consumed by strtol, is evaluated early enough that it is considered a
7779 compile-time constant.
7782 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
7783 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
7786 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
7787 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
7790 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
7791 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
7794 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
7795 Similar to @code{__builtin_nan}, except the return type is @code{float}.
7798 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
7799 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
7802 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
7803 Similar to @code{__builtin_nan}, except the significand is forced
7804 to be a signaling NaN@. The @code{nans} function is proposed by
7805 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
7808 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
7809 Similar to @code{__builtin_nans}, except the return type is @code{float}.
7812 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
7813 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
7816 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
7817 Returns one plus the index of the least significant 1-bit of @var{x}, or
7818 if @var{x} is zero, returns zero.
7821 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
7822 Returns the number of leading 0-bits in @var{x}, starting at the most
7823 significant bit position. If @var{x} is 0, the result is undefined.
7826 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
7827 Returns the number of trailing 0-bits in @var{x}, starting at the least
7828 significant bit position. If @var{x} is 0, the result is undefined.
7831 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
7832 Returns the number of 1-bits in @var{x}.
7835 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
7836 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
7840 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
7841 Similar to @code{__builtin_ffs}, except the argument type is
7842 @code{unsigned long}.
7845 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
7846 Similar to @code{__builtin_clz}, except the argument type is
7847 @code{unsigned long}.
7850 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
7851 Similar to @code{__builtin_ctz}, except the argument type is
7852 @code{unsigned long}.
7855 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
7856 Similar to @code{__builtin_popcount}, except the argument type is
7857 @code{unsigned long}.
7860 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
7861 Similar to @code{__builtin_parity}, except the argument type is
7862 @code{unsigned long}.
7865 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
7866 Similar to @code{__builtin_ffs}, except the argument type is
7867 @code{unsigned long long}.
7870 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
7871 Similar to @code{__builtin_clz}, except the argument type is
7872 @code{unsigned long long}.
7875 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
7876 Similar to @code{__builtin_ctz}, except the argument type is
7877 @code{unsigned long long}.
7880 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
7881 Similar to @code{__builtin_popcount}, except the argument type is
7882 @code{unsigned long long}.
7885 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
7886 Similar to @code{__builtin_parity}, except the argument type is
7887 @code{unsigned long long}.
7890 @deftypefn {Built-in Function} double __builtin_powi (double, int)
7891 Returns the first argument raised to the power of the second. Unlike the
7892 @code{pow} function no guarantees about precision and rounding are made.
7895 @deftypefn {Built-in Function} float __builtin_powif (float, int)
7896 Similar to @code{__builtin_powi}, except the argument and return types
7900 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
7901 Similar to @code{__builtin_powi}, except the argument and return types
7902 are @code{long double}.
7905 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
7906 Returns @var{x} with the order of the bytes reversed; for example,
7907 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
7911 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
7912 Similar to @code{__builtin_bswap32}, except the argument and return types
7916 @node Target Builtins
7917 @section Built-in Functions Specific to Particular Target Machines
7919 On some target machines, GCC supports many built-in functions specific
7920 to those machines. Generally these generate calls to specific machine
7921 instructions, but allow the compiler to schedule those calls.
7924 * Alpha Built-in Functions::
7925 * ARM iWMMXt Built-in Functions::
7926 * ARM NEON Intrinsics::
7927 * AVR Built-in Functions::
7928 * Blackfin Built-in Functions::
7929 * FR-V Built-in Functions::
7930 * X86 Built-in Functions::
7931 * MIPS DSP Built-in Functions::
7932 * MIPS Paired-Single Support::
7933 * MIPS Loongson Built-in Functions::
7934 * Other MIPS Built-in Functions::
7935 * picoChip Built-in Functions::
7936 * PowerPC AltiVec/VSX Built-in Functions::
7937 * RX Built-in Functions::
7938 * SPARC VIS Built-in Functions::
7939 * SPU Built-in Functions::
7942 @node Alpha Built-in Functions
7943 @subsection Alpha Built-in Functions
7945 These built-in functions are available for the Alpha family of
7946 processors, depending on the command-line switches used.
7948 The following built-in functions are always available. They
7949 all generate the machine instruction that is part of the name.
7952 long __builtin_alpha_implver (void)
7953 long __builtin_alpha_rpcc (void)
7954 long __builtin_alpha_amask (long)
7955 long __builtin_alpha_cmpbge (long, long)
7956 long __builtin_alpha_extbl (long, long)
7957 long __builtin_alpha_extwl (long, long)
7958 long __builtin_alpha_extll (long, long)
7959 long __builtin_alpha_extql (long, long)
7960 long __builtin_alpha_extwh (long, long)
7961 long __builtin_alpha_extlh (long, long)
7962 long __builtin_alpha_extqh (long, long)
7963 long __builtin_alpha_insbl (long, long)
7964 long __builtin_alpha_inswl (long, long)
7965 long __builtin_alpha_insll (long, long)
7966 long __builtin_alpha_insql (long, long)
7967 long __builtin_alpha_inswh (long, long)
7968 long __builtin_alpha_inslh (long, long)
7969 long __builtin_alpha_insqh (long, long)
7970 long __builtin_alpha_mskbl (long, long)
7971 long __builtin_alpha_mskwl (long, long)
7972 long __builtin_alpha_mskll (long, long)
7973 long __builtin_alpha_mskql (long, long)
7974 long __builtin_alpha_mskwh (long, long)
7975 long __builtin_alpha_msklh (long, long)
7976 long __builtin_alpha_mskqh (long, long)
7977 long __builtin_alpha_umulh (long, long)
7978 long __builtin_alpha_zap (long, long)
7979 long __builtin_alpha_zapnot (long, long)
7982 The following built-in functions are always with @option{-mmax}
7983 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
7984 later. They all generate the machine instruction that is part
7988 long __builtin_alpha_pklb (long)
7989 long __builtin_alpha_pkwb (long)
7990 long __builtin_alpha_unpkbl (long)
7991 long __builtin_alpha_unpkbw (long)
7992 long __builtin_alpha_minub8 (long, long)
7993 long __builtin_alpha_minsb8 (long, long)
7994 long __builtin_alpha_minuw4 (long, long)
7995 long __builtin_alpha_minsw4 (long, long)
7996 long __builtin_alpha_maxub8 (long, long)
7997 long __builtin_alpha_maxsb8 (long, long)
7998 long __builtin_alpha_maxuw4 (long, long)
7999 long __builtin_alpha_maxsw4 (long, long)
8000 long __builtin_alpha_perr (long, long)
8003 The following built-in functions are always with @option{-mcix}
8004 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
8005 later. They all generate the machine instruction that is part
8009 long __builtin_alpha_cttz (long)
8010 long __builtin_alpha_ctlz (long)
8011 long __builtin_alpha_ctpop (long)
8014 The following builtins are available on systems that use the OSF/1
8015 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
8016 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
8017 @code{rdval} and @code{wrval}.
8020 void *__builtin_thread_pointer (void)
8021 void __builtin_set_thread_pointer (void *)
8024 @node ARM iWMMXt Built-in Functions
8025 @subsection ARM iWMMXt Built-in Functions
8027 These built-in functions are available for the ARM family of
8028 processors when the @option{-mcpu=iwmmxt} switch is used:
8031 typedef int v2si __attribute__ ((vector_size (8)));
8032 typedef short v4hi __attribute__ ((vector_size (8)));
8033 typedef char v8qi __attribute__ ((vector_size (8)));
8035 int __builtin_arm_getwcx (int)
8036 void __builtin_arm_setwcx (int, int)
8037 int __builtin_arm_textrmsb (v8qi, int)
8038 int __builtin_arm_textrmsh (v4hi, int)
8039 int __builtin_arm_textrmsw (v2si, int)
8040 int __builtin_arm_textrmub (v8qi, int)
8041 int __builtin_arm_textrmuh (v4hi, int)
8042 int __builtin_arm_textrmuw (v2si, int)
8043 v8qi __builtin_arm_tinsrb (v8qi, int)
8044 v4hi __builtin_arm_tinsrh (v4hi, int)
8045 v2si __builtin_arm_tinsrw (v2si, int)
8046 long long __builtin_arm_tmia (long long, int, int)
8047 long long __builtin_arm_tmiabb (long long, int, int)
8048 long long __builtin_arm_tmiabt (long long, int, int)
8049 long long __builtin_arm_tmiaph (long long, int, int)
8050 long long __builtin_arm_tmiatb (long long, int, int)
8051 long long __builtin_arm_tmiatt (long long, int, int)
8052 int __builtin_arm_tmovmskb (v8qi)
8053 int __builtin_arm_tmovmskh (v4hi)
8054 int __builtin_arm_tmovmskw (v2si)
8055 long long __builtin_arm_waccb (v8qi)
8056 long long __builtin_arm_wacch (v4hi)
8057 long long __builtin_arm_waccw (v2si)
8058 v8qi __builtin_arm_waddb (v8qi, v8qi)
8059 v8qi __builtin_arm_waddbss (v8qi, v8qi)
8060 v8qi __builtin_arm_waddbus (v8qi, v8qi)
8061 v4hi __builtin_arm_waddh (v4hi, v4hi)
8062 v4hi __builtin_arm_waddhss (v4hi, v4hi)
8063 v4hi __builtin_arm_waddhus (v4hi, v4hi)
8064 v2si __builtin_arm_waddw (v2si, v2si)
8065 v2si __builtin_arm_waddwss (v2si, v2si)
8066 v2si __builtin_arm_waddwus (v2si, v2si)
8067 v8qi __builtin_arm_walign (v8qi, v8qi, int)
8068 long long __builtin_arm_wand(long long, long long)
8069 long long __builtin_arm_wandn (long long, long long)
8070 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
8071 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
8072 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
8073 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
8074 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
8075 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
8076 v2si __builtin_arm_wcmpeqw (v2si, v2si)
8077 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
8078 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
8079 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
8080 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
8081 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
8082 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
8083 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
8084 long long __builtin_arm_wmacsz (v4hi, v4hi)
8085 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
8086 long long __builtin_arm_wmacuz (v4hi, v4hi)
8087 v4hi __builtin_arm_wmadds (v4hi, v4hi)
8088 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
8089 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
8090 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
8091 v2si __builtin_arm_wmaxsw (v2si, v2si)
8092 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
8093 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
8094 v2si __builtin_arm_wmaxuw (v2si, v2si)
8095 v8qi __builtin_arm_wminsb (v8qi, v8qi)
8096 v4hi __builtin_arm_wminsh (v4hi, v4hi)
8097 v2si __builtin_arm_wminsw (v2si, v2si)
8098 v8qi __builtin_arm_wminub (v8qi, v8qi)
8099 v4hi __builtin_arm_wminuh (v4hi, v4hi)
8100 v2si __builtin_arm_wminuw (v2si, v2si)
8101 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
8102 v4hi __builtin_arm_wmulul (v4hi, v4hi)
8103 v4hi __builtin_arm_wmulum (v4hi, v4hi)
8104 long long __builtin_arm_wor (long long, long long)
8105 v2si __builtin_arm_wpackdss (long long, long long)
8106 v2si __builtin_arm_wpackdus (long long, long long)
8107 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
8108 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
8109 v4hi __builtin_arm_wpackwss (v2si, v2si)
8110 v4hi __builtin_arm_wpackwus (v2si, v2si)
8111 long long __builtin_arm_wrord (long long, long long)
8112 long long __builtin_arm_wrordi (long long, int)
8113 v4hi __builtin_arm_wrorh (v4hi, long long)
8114 v4hi __builtin_arm_wrorhi (v4hi, int)
8115 v2si __builtin_arm_wrorw (v2si, long long)
8116 v2si __builtin_arm_wrorwi (v2si, int)
8117 v2si __builtin_arm_wsadb (v8qi, v8qi)
8118 v2si __builtin_arm_wsadbz (v8qi, v8qi)
8119 v2si __builtin_arm_wsadh (v4hi, v4hi)
8120 v2si __builtin_arm_wsadhz (v4hi, v4hi)
8121 v4hi __builtin_arm_wshufh (v4hi, int)
8122 long long __builtin_arm_wslld (long long, long long)
8123 long long __builtin_arm_wslldi (long long, int)
8124 v4hi __builtin_arm_wsllh (v4hi, long long)
8125 v4hi __builtin_arm_wsllhi (v4hi, int)
8126 v2si __builtin_arm_wsllw (v2si, long long)
8127 v2si __builtin_arm_wsllwi (v2si, int)
8128 long long __builtin_arm_wsrad (long long, long long)
8129 long long __builtin_arm_wsradi (long long, int)
8130 v4hi __builtin_arm_wsrah (v4hi, long long)
8131 v4hi __builtin_arm_wsrahi (v4hi, int)
8132 v2si __builtin_arm_wsraw (v2si, long long)
8133 v2si __builtin_arm_wsrawi (v2si, int)
8134 long long __builtin_arm_wsrld (long long, long long)
8135 long long __builtin_arm_wsrldi (long long, int)
8136 v4hi __builtin_arm_wsrlh (v4hi, long long)
8137 v4hi __builtin_arm_wsrlhi (v4hi, int)
8138 v2si __builtin_arm_wsrlw (v2si, long long)
8139 v2si __builtin_arm_wsrlwi (v2si, int)
8140 v8qi __builtin_arm_wsubb (v8qi, v8qi)
8141 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
8142 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
8143 v4hi __builtin_arm_wsubh (v4hi, v4hi)
8144 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
8145 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
8146 v2si __builtin_arm_wsubw (v2si, v2si)
8147 v2si __builtin_arm_wsubwss (v2si, v2si)
8148 v2si __builtin_arm_wsubwus (v2si, v2si)
8149 v4hi __builtin_arm_wunpckehsb (v8qi)
8150 v2si __builtin_arm_wunpckehsh (v4hi)
8151 long long __builtin_arm_wunpckehsw (v2si)
8152 v4hi __builtin_arm_wunpckehub (v8qi)
8153 v2si __builtin_arm_wunpckehuh (v4hi)
8154 long long __builtin_arm_wunpckehuw (v2si)
8155 v4hi __builtin_arm_wunpckelsb (v8qi)
8156 v2si __builtin_arm_wunpckelsh (v4hi)
8157 long long __builtin_arm_wunpckelsw (v2si)
8158 v4hi __builtin_arm_wunpckelub (v8qi)
8159 v2si __builtin_arm_wunpckeluh (v4hi)
8160 long long __builtin_arm_wunpckeluw (v2si)
8161 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
8162 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
8163 v2si __builtin_arm_wunpckihw (v2si, v2si)
8164 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
8165 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
8166 v2si __builtin_arm_wunpckilw (v2si, v2si)
8167 long long __builtin_arm_wxor (long long, long long)
8168 long long __builtin_arm_wzero ()
8171 @node ARM NEON Intrinsics
8172 @subsection ARM NEON Intrinsics
8174 These built-in intrinsics for the ARM Advanced SIMD extension are available
8175 when the @option{-mfpu=neon} switch is used:
8177 @include arm-neon-intrinsics.texi
8179 @node AVR Built-in Functions
8180 @subsection AVR Built-in Functions
8182 For each built-in function for AVR, there is an equally named,
8183 uppercase built-in macro defined. That way users can easily query if
8184 or if not a specific built-in is implemented or not. For example, if
8185 @code{__builtin_avr_nop} is available the macro
8186 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
8188 The following built-in functions map to the respective machine
8189 instruction, i.e. @code{nop}, @code{sei}, @code{cli}, @code{sleep},
8190 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
8191 resp. @code{fmulsu}. The latter three are only available if the AVR
8192 device actually supports multiplication.
8195 void __builtin_avr_nop (void)
8196 void __builtin_avr_sei (void)
8197 void __builtin_avr_cli (void)
8198 void __builtin_avr_sleep (void)
8199 void __builtin_avr_wdr (void)
8200 unsigned char __builtin_avr_swap (unsigned char)
8201 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
8202 int __builtin_avr_fmuls (char, char)
8203 int __builtin_avr_fmulsu (char, unsigned char)
8206 In order to delay execution for a specific number of cycles, GCC
8209 void __builtin_avr_delay_cycles (unsigned long ticks)
8212 @code{ticks} is the number of ticks to delay execution. Note that this
8213 built-in does not take into account the effect of interrupts which
8214 might increase delay time. @code{ticks} must be a compile time
8215 integer constant; delays with a variable number of cycles are not supported.
8217 @node Blackfin Built-in Functions
8218 @subsection Blackfin Built-in Functions
8220 Currently, there are two Blackfin-specific built-in functions. These are
8221 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
8222 using inline assembly; by using these built-in functions the compiler can
8223 automatically add workarounds for hardware errata involving these
8224 instructions. These functions are named as follows:
8227 void __builtin_bfin_csync (void)
8228 void __builtin_bfin_ssync (void)
8231 @node FR-V Built-in Functions
8232 @subsection FR-V Built-in Functions
8234 GCC provides many FR-V-specific built-in functions. In general,
8235 these functions are intended to be compatible with those described
8236 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
8237 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
8238 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
8239 pointer rather than by value.
8241 Most of the functions are named after specific FR-V instructions.
8242 Such functions are said to be ``directly mapped'' and are summarized
8243 here in tabular form.
8247 * Directly-mapped Integer Functions::
8248 * Directly-mapped Media Functions::
8249 * Raw read/write Functions::
8250 * Other Built-in Functions::
8253 @node Argument Types
8254 @subsubsection Argument Types
8256 The arguments to the built-in functions can be divided into three groups:
8257 register numbers, compile-time constants and run-time values. In order
8258 to make this classification clear at a glance, the arguments and return
8259 values are given the following pseudo types:
8261 @multitable @columnfractions .20 .30 .15 .35
8262 @item Pseudo type @tab Real C type @tab Constant? @tab Description
8263 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
8264 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
8265 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
8266 @item @code{uw2} @tab @code{unsigned long long} @tab No
8267 @tab an unsigned doubleword
8268 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
8269 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
8270 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
8271 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
8274 These pseudo types are not defined by GCC, they are simply a notational
8275 convenience used in this manual.
8277 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
8278 and @code{sw2} are evaluated at run time. They correspond to
8279 register operands in the underlying FR-V instructions.
8281 @code{const} arguments represent immediate operands in the underlying
8282 FR-V instructions. They must be compile-time constants.
8284 @code{acc} arguments are evaluated at compile time and specify the number
8285 of an accumulator register. For example, an @code{acc} argument of 2
8286 will select the ACC2 register.
8288 @code{iacc} arguments are similar to @code{acc} arguments but specify the
8289 number of an IACC register. See @pxref{Other Built-in Functions}
8292 @node Directly-mapped Integer Functions
8293 @subsubsection Directly-mapped Integer Functions
8295 The functions listed below map directly to FR-V I-type instructions.
8297 @multitable @columnfractions .45 .32 .23
8298 @item Function prototype @tab Example usage @tab Assembly output
8299 @item @code{sw1 __ADDSS (sw1, sw1)}
8300 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
8301 @tab @code{ADDSS @var{a},@var{b},@var{c}}
8302 @item @code{sw1 __SCAN (sw1, sw1)}
8303 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
8304 @tab @code{SCAN @var{a},@var{b},@var{c}}
8305 @item @code{sw1 __SCUTSS (sw1)}
8306 @tab @code{@var{b} = __SCUTSS (@var{a})}
8307 @tab @code{SCUTSS @var{a},@var{b}}
8308 @item @code{sw1 __SLASS (sw1, sw1)}
8309 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
8310 @tab @code{SLASS @var{a},@var{b},@var{c}}
8311 @item @code{void __SMASS (sw1, sw1)}
8312 @tab @code{__SMASS (@var{a}, @var{b})}
8313 @tab @code{SMASS @var{a},@var{b}}
8314 @item @code{void __SMSSS (sw1, sw1)}
8315 @tab @code{__SMSSS (@var{a}, @var{b})}
8316 @tab @code{SMSSS @var{a},@var{b}}
8317 @item @code{void __SMU (sw1, sw1)}
8318 @tab @code{__SMU (@var{a}, @var{b})}
8319 @tab @code{SMU @var{a},@var{b}}
8320 @item @code{sw2 __SMUL (sw1, sw1)}
8321 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
8322 @tab @code{SMUL @var{a},@var{b},@var{c}}
8323 @item @code{sw1 __SUBSS (sw1, sw1)}
8324 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
8325 @tab @code{SUBSS @var{a},@var{b},@var{c}}
8326 @item @code{uw2 __UMUL (uw1, uw1)}
8327 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
8328 @tab @code{UMUL @var{a},@var{b},@var{c}}
8331 @node Directly-mapped Media Functions
8332 @subsubsection Directly-mapped Media Functions
8334 The functions listed below map directly to FR-V M-type instructions.
8336 @multitable @columnfractions .45 .32 .23
8337 @item Function prototype @tab Example usage @tab Assembly output
8338 @item @code{uw1 __MABSHS (sw1)}
8339 @tab @code{@var{b} = __MABSHS (@var{a})}
8340 @tab @code{MABSHS @var{a},@var{b}}
8341 @item @code{void __MADDACCS (acc, acc)}
8342 @tab @code{__MADDACCS (@var{b}, @var{a})}
8343 @tab @code{MADDACCS @var{a},@var{b}}
8344 @item @code{sw1 __MADDHSS (sw1, sw1)}
8345 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
8346 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
8347 @item @code{uw1 __MADDHUS (uw1, uw1)}
8348 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
8349 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
8350 @item @code{uw1 __MAND (uw1, uw1)}
8351 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
8352 @tab @code{MAND @var{a},@var{b},@var{c}}
8353 @item @code{void __MASACCS (acc, acc)}
8354 @tab @code{__MASACCS (@var{b}, @var{a})}
8355 @tab @code{MASACCS @var{a},@var{b}}
8356 @item @code{uw1 __MAVEH (uw1, uw1)}
8357 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
8358 @tab @code{MAVEH @var{a},@var{b},@var{c}}
8359 @item @code{uw2 __MBTOH (uw1)}
8360 @tab @code{@var{b} = __MBTOH (@var{a})}
8361 @tab @code{MBTOH @var{a},@var{b}}
8362 @item @code{void __MBTOHE (uw1 *, uw1)}
8363 @tab @code{__MBTOHE (&@var{b}, @var{a})}
8364 @tab @code{MBTOHE @var{a},@var{b}}
8365 @item @code{void __MCLRACC (acc)}
8366 @tab @code{__MCLRACC (@var{a})}
8367 @tab @code{MCLRACC @var{a}}
8368 @item @code{void __MCLRACCA (void)}
8369 @tab @code{__MCLRACCA ()}
8370 @tab @code{MCLRACCA}
8371 @item @code{uw1 __Mcop1 (uw1, uw1)}
8372 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
8373 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
8374 @item @code{uw1 __Mcop2 (uw1, uw1)}
8375 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
8376 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
8377 @item @code{uw1 __MCPLHI (uw2, const)}
8378 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
8379 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
8380 @item @code{uw1 __MCPLI (uw2, const)}
8381 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
8382 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
8383 @item @code{void __MCPXIS (acc, sw1, sw1)}
8384 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
8385 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
8386 @item @code{void __MCPXIU (acc, uw1, uw1)}
8387 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
8388 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
8389 @item @code{void __MCPXRS (acc, sw1, sw1)}
8390 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
8391 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
8392 @item @code{void __MCPXRU (acc, uw1, uw1)}
8393 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
8394 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
8395 @item @code{uw1 __MCUT (acc, uw1)}
8396 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
8397 @tab @code{MCUT @var{a},@var{b},@var{c}}
8398 @item @code{uw1 __MCUTSS (acc, sw1)}
8399 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
8400 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
8401 @item @code{void __MDADDACCS (acc, acc)}
8402 @tab @code{__MDADDACCS (@var{b}, @var{a})}
8403 @tab @code{MDADDACCS @var{a},@var{b}}
8404 @item @code{void __MDASACCS (acc, acc)}
8405 @tab @code{__MDASACCS (@var{b}, @var{a})}
8406 @tab @code{MDASACCS @var{a},@var{b}}
8407 @item @code{uw2 __MDCUTSSI (acc, const)}
8408 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
8409 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
8410 @item @code{uw2 __MDPACKH (uw2, uw2)}
8411 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
8412 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
8413 @item @code{uw2 __MDROTLI (uw2, const)}
8414 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
8415 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
8416 @item @code{void __MDSUBACCS (acc, acc)}
8417 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
8418 @tab @code{MDSUBACCS @var{a},@var{b}}
8419 @item @code{void __MDUNPACKH (uw1 *, uw2)}
8420 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
8421 @tab @code{MDUNPACKH @var{a},@var{b}}
8422 @item @code{uw2 __MEXPDHD (uw1, const)}
8423 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
8424 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
8425 @item @code{uw1 __MEXPDHW (uw1, const)}
8426 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
8427 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
8428 @item @code{uw1 __MHDSETH (uw1, const)}
8429 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
8430 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
8431 @item @code{sw1 __MHDSETS (const)}
8432 @tab @code{@var{b} = __MHDSETS (@var{a})}
8433 @tab @code{MHDSETS #@var{a},@var{b}}
8434 @item @code{uw1 __MHSETHIH (uw1, const)}
8435 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
8436 @tab @code{MHSETHIH #@var{a},@var{b}}
8437 @item @code{sw1 __MHSETHIS (sw1, const)}
8438 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
8439 @tab @code{MHSETHIS #@var{a},@var{b}}
8440 @item @code{uw1 __MHSETLOH (uw1, const)}
8441 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
8442 @tab @code{MHSETLOH #@var{a},@var{b}}
8443 @item @code{sw1 __MHSETLOS (sw1, const)}
8444 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
8445 @tab @code{MHSETLOS #@var{a},@var{b}}
8446 @item @code{uw1 __MHTOB (uw2)}
8447 @tab @code{@var{b} = __MHTOB (@var{a})}
8448 @tab @code{MHTOB @var{a},@var{b}}
8449 @item @code{void __MMACHS (acc, sw1, sw1)}
8450 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
8451 @tab @code{MMACHS @var{a},@var{b},@var{c}}
8452 @item @code{void __MMACHU (acc, uw1, uw1)}
8453 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
8454 @tab @code{MMACHU @var{a},@var{b},@var{c}}
8455 @item @code{void __MMRDHS (acc, sw1, sw1)}
8456 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
8457 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
8458 @item @code{void __MMRDHU (acc, uw1, uw1)}
8459 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
8460 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
8461 @item @code{void __MMULHS (acc, sw1, sw1)}
8462 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
8463 @tab @code{MMULHS @var{a},@var{b},@var{c}}
8464 @item @code{void __MMULHU (acc, uw1, uw1)}
8465 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
8466 @tab @code{MMULHU @var{a},@var{b},@var{c}}
8467 @item @code{void __MMULXHS (acc, sw1, sw1)}
8468 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
8469 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
8470 @item @code{void __MMULXHU (acc, uw1, uw1)}
8471 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
8472 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
8473 @item @code{uw1 __MNOT (uw1)}
8474 @tab @code{@var{b} = __MNOT (@var{a})}
8475 @tab @code{MNOT @var{a},@var{b}}
8476 @item @code{uw1 __MOR (uw1, uw1)}
8477 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
8478 @tab @code{MOR @var{a},@var{b},@var{c}}
8479 @item @code{uw1 __MPACKH (uh, uh)}
8480 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
8481 @tab @code{MPACKH @var{a},@var{b},@var{c}}
8482 @item @code{sw2 __MQADDHSS (sw2, sw2)}
8483 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
8484 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
8485 @item @code{uw2 __MQADDHUS (uw2, uw2)}
8486 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
8487 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
8488 @item @code{void __MQCPXIS (acc, sw2, sw2)}
8489 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
8490 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
8491 @item @code{void __MQCPXIU (acc, uw2, uw2)}
8492 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
8493 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
8494 @item @code{void __MQCPXRS (acc, sw2, sw2)}
8495 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
8496 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
8497 @item @code{void __MQCPXRU (acc, uw2, uw2)}
8498 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
8499 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
8500 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
8501 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
8502 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
8503 @item @code{sw2 __MQLMTHS (sw2, sw2)}
8504 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
8505 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
8506 @item @code{void __MQMACHS (acc, sw2, sw2)}
8507 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
8508 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
8509 @item @code{void __MQMACHU (acc, uw2, uw2)}
8510 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
8511 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
8512 @item @code{void __MQMACXHS (acc, sw2, sw2)}
8513 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
8514 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
8515 @item @code{void __MQMULHS (acc, sw2, sw2)}
8516 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
8517 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
8518 @item @code{void __MQMULHU (acc, uw2, uw2)}
8519 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
8520 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
8521 @item @code{void __MQMULXHS (acc, sw2, sw2)}
8522 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
8523 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
8524 @item @code{void __MQMULXHU (acc, uw2, uw2)}
8525 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
8526 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
8527 @item @code{sw2 __MQSATHS (sw2, sw2)}
8528 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
8529 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
8530 @item @code{uw2 __MQSLLHI (uw2, int)}
8531 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
8532 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
8533 @item @code{sw2 __MQSRAHI (sw2, int)}
8534 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
8535 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
8536 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
8537 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
8538 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
8539 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
8540 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
8541 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
8542 @item @code{void __MQXMACHS (acc, sw2, sw2)}
8543 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
8544 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
8545 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
8546 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
8547 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
8548 @item @code{uw1 __MRDACC (acc)}
8549 @tab @code{@var{b} = __MRDACC (@var{a})}
8550 @tab @code{MRDACC @var{a},@var{b}}
8551 @item @code{uw1 __MRDACCG (acc)}
8552 @tab @code{@var{b} = __MRDACCG (@var{a})}
8553 @tab @code{MRDACCG @var{a},@var{b}}
8554 @item @code{uw1 __MROTLI (uw1, const)}
8555 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
8556 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
8557 @item @code{uw1 __MROTRI (uw1, const)}
8558 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
8559 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
8560 @item @code{sw1 __MSATHS (sw1, sw1)}
8561 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
8562 @tab @code{MSATHS @var{a},@var{b},@var{c}}
8563 @item @code{uw1 __MSATHU (uw1, uw1)}
8564 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
8565 @tab @code{MSATHU @var{a},@var{b},@var{c}}
8566 @item @code{uw1 __MSLLHI (uw1, const)}
8567 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
8568 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
8569 @item @code{sw1 __MSRAHI (sw1, const)}
8570 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
8571 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
8572 @item @code{uw1 __MSRLHI (uw1, const)}
8573 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
8574 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
8575 @item @code{void __MSUBACCS (acc, acc)}
8576 @tab @code{__MSUBACCS (@var{b}, @var{a})}
8577 @tab @code{MSUBACCS @var{a},@var{b}}
8578 @item @code{sw1 __MSUBHSS (sw1, sw1)}
8579 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
8580 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
8581 @item @code{uw1 __MSUBHUS (uw1, uw1)}
8582 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
8583 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
8584 @item @code{void __MTRAP (void)}
8585 @tab @code{__MTRAP ()}
8587 @item @code{uw2 __MUNPACKH (uw1)}
8588 @tab @code{@var{b} = __MUNPACKH (@var{a})}
8589 @tab @code{MUNPACKH @var{a},@var{b}}
8590 @item @code{uw1 __MWCUT (uw2, uw1)}
8591 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
8592 @tab @code{MWCUT @var{a},@var{b},@var{c}}
8593 @item @code{void __MWTACC (acc, uw1)}
8594 @tab @code{__MWTACC (@var{b}, @var{a})}
8595 @tab @code{MWTACC @var{a},@var{b}}
8596 @item @code{void __MWTACCG (acc, uw1)}
8597 @tab @code{__MWTACCG (@var{b}, @var{a})}
8598 @tab @code{MWTACCG @var{a},@var{b}}
8599 @item @code{uw1 __MXOR (uw1, uw1)}
8600 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
8601 @tab @code{MXOR @var{a},@var{b},@var{c}}
8604 @node Raw read/write Functions
8605 @subsubsection Raw read/write Functions
8607 This sections describes built-in functions related to read and write
8608 instructions to access memory. These functions generate
8609 @code{membar} instructions to flush the I/O load and stores where
8610 appropriate, as described in Fujitsu's manual described above.
8614 @item unsigned char __builtin_read8 (void *@var{data})
8615 @item unsigned short __builtin_read16 (void *@var{data})
8616 @item unsigned long __builtin_read32 (void *@var{data})
8617 @item unsigned long long __builtin_read64 (void *@var{data})
8619 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
8620 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
8621 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
8622 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
8625 @node Other Built-in Functions
8626 @subsubsection Other Built-in Functions
8628 This section describes built-in functions that are not named after
8629 a specific FR-V instruction.
8632 @item sw2 __IACCreadll (iacc @var{reg})
8633 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
8634 for future expansion and must be 0.
8636 @item sw1 __IACCreadl (iacc @var{reg})
8637 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
8638 Other values of @var{reg} are rejected as invalid.
8640 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
8641 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
8642 is reserved for future expansion and must be 0.
8644 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
8645 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
8646 is 1. Other values of @var{reg} are rejected as invalid.
8648 @item void __data_prefetch0 (const void *@var{x})
8649 Use the @code{dcpl} instruction to load the contents of address @var{x}
8650 into the data cache.
8652 @item void __data_prefetch (const void *@var{x})
8653 Use the @code{nldub} instruction to load the contents of address @var{x}
8654 into the data cache. The instruction will be issued in slot I1@.
8657 @node X86 Built-in Functions
8658 @subsection X86 Built-in Functions
8660 These built-in functions are available for the i386 and x86-64 family
8661 of computers, depending on the command-line switches used.
8663 Note that, if you specify command-line switches such as @option{-msse},
8664 the compiler could use the extended instruction sets even if the built-ins
8665 are not used explicitly in the program. For this reason, applications
8666 which perform runtime CPU detection must compile separate files for each
8667 supported architecture, using the appropriate flags. In particular,
8668 the file containing the CPU detection code should be compiled without
8671 The following machine modes are available for use with MMX built-in functions
8672 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
8673 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
8674 vector of eight 8-bit integers. Some of the built-in functions operate on
8675 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
8677 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
8678 of two 32-bit floating point values.
8680 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
8681 floating point values. Some instructions use a vector of four 32-bit
8682 integers, these use @code{V4SI}. Finally, some instructions operate on an
8683 entire vector register, interpreting it as a 128-bit integer, these use mode
8686 In 64-bit mode, the x86-64 family of processors uses additional built-in
8687 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
8688 floating point and @code{TC} 128-bit complex floating point values.
8690 The following floating point built-in functions are available in 64-bit
8691 mode. All of them implement the function that is part of the name.
8694 __float128 __builtin_fabsq (__float128)
8695 __float128 __builtin_copysignq (__float128, __float128)
8698 The following built-in function is always available.
8701 @item void __builtin_ia32_pause (void)
8702 Generates the @code{pause} machine instruction with a compiler memory
8706 The following floating point built-in functions are made available in the
8710 @item __float128 __builtin_infq (void)
8711 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
8712 @findex __builtin_infq
8714 @item __float128 __builtin_huge_valq (void)
8715 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
8716 @findex __builtin_huge_valq
8719 The following built-in functions are made available by @option{-mmmx}.
8720 All of them generate the machine instruction that is part of the name.
8723 v8qi __builtin_ia32_paddb (v8qi, v8qi)
8724 v4hi __builtin_ia32_paddw (v4hi, v4hi)
8725 v2si __builtin_ia32_paddd (v2si, v2si)
8726 v8qi __builtin_ia32_psubb (v8qi, v8qi)
8727 v4hi __builtin_ia32_psubw (v4hi, v4hi)
8728 v2si __builtin_ia32_psubd (v2si, v2si)
8729 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
8730 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
8731 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
8732 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
8733 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
8734 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
8735 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
8736 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
8737 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
8738 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
8739 di __builtin_ia32_pand (di, di)
8740 di __builtin_ia32_pandn (di,di)
8741 di __builtin_ia32_por (di, di)
8742 di __builtin_ia32_pxor (di, di)
8743 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
8744 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
8745 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
8746 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
8747 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
8748 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
8749 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
8750 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
8751 v2si __builtin_ia32_punpckhdq (v2si, v2si)
8752 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
8753 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
8754 v2si __builtin_ia32_punpckldq (v2si, v2si)
8755 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
8756 v4hi __builtin_ia32_packssdw (v2si, v2si)
8757 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
8759 v4hi __builtin_ia32_psllw (v4hi, v4hi)
8760 v2si __builtin_ia32_pslld (v2si, v2si)
8761 v1di __builtin_ia32_psllq (v1di, v1di)
8762 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
8763 v2si __builtin_ia32_psrld (v2si, v2si)
8764 v1di __builtin_ia32_psrlq (v1di, v1di)
8765 v4hi __builtin_ia32_psraw (v4hi, v4hi)
8766 v2si __builtin_ia32_psrad (v2si, v2si)
8767 v4hi __builtin_ia32_psllwi (v4hi, int)
8768 v2si __builtin_ia32_pslldi (v2si, int)
8769 v1di __builtin_ia32_psllqi (v1di, int)
8770 v4hi __builtin_ia32_psrlwi (v4hi, int)
8771 v2si __builtin_ia32_psrldi (v2si, int)
8772 v1di __builtin_ia32_psrlqi (v1di, int)
8773 v4hi __builtin_ia32_psrawi (v4hi, int)
8774 v2si __builtin_ia32_psradi (v2si, int)
8778 The following built-in functions are made available either with
8779 @option{-msse}, or with a combination of @option{-m3dnow} and
8780 @option{-march=athlon}. All of them generate the machine
8781 instruction that is part of the name.
8784 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
8785 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
8786 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
8787 v1di __builtin_ia32_psadbw (v8qi, v8qi)
8788 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
8789 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
8790 v8qi __builtin_ia32_pminub (v8qi, v8qi)
8791 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
8792 int __builtin_ia32_pextrw (v4hi, int)
8793 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
8794 int __builtin_ia32_pmovmskb (v8qi)
8795 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
8796 void __builtin_ia32_movntq (di *, di)
8797 void __builtin_ia32_sfence (void)
8800 The following built-in functions are available when @option{-msse} is used.
8801 All of them generate the machine instruction that is part of the name.
8804 int __builtin_ia32_comieq (v4sf, v4sf)
8805 int __builtin_ia32_comineq (v4sf, v4sf)
8806 int __builtin_ia32_comilt (v4sf, v4sf)
8807 int __builtin_ia32_comile (v4sf, v4sf)
8808 int __builtin_ia32_comigt (v4sf, v4sf)
8809 int __builtin_ia32_comige (v4sf, v4sf)
8810 int __builtin_ia32_ucomieq (v4sf, v4sf)
8811 int __builtin_ia32_ucomineq (v4sf, v4sf)
8812 int __builtin_ia32_ucomilt (v4sf, v4sf)
8813 int __builtin_ia32_ucomile (v4sf, v4sf)
8814 int __builtin_ia32_ucomigt (v4sf, v4sf)
8815 int __builtin_ia32_ucomige (v4sf, v4sf)
8816 v4sf __builtin_ia32_addps (v4sf, v4sf)
8817 v4sf __builtin_ia32_subps (v4sf, v4sf)
8818 v4sf __builtin_ia32_mulps (v4sf, v4sf)
8819 v4sf __builtin_ia32_divps (v4sf, v4sf)
8820 v4sf __builtin_ia32_addss (v4sf, v4sf)
8821 v4sf __builtin_ia32_subss (v4sf, v4sf)
8822 v4sf __builtin_ia32_mulss (v4sf, v4sf)
8823 v4sf __builtin_ia32_divss (v4sf, v4sf)
8824 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
8825 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
8826 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
8827 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
8828 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
8829 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
8830 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
8831 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
8832 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
8833 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
8834 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
8835 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
8836 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
8837 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
8838 v4si __builtin_ia32_cmpless (v4sf, v4sf)
8839 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
8840 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
8841 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
8842 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
8843 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
8844 v4sf __builtin_ia32_maxps (v4sf, v4sf)
8845 v4sf __builtin_ia32_maxss (v4sf, v4sf)
8846 v4sf __builtin_ia32_minps (v4sf, v4sf)
8847 v4sf __builtin_ia32_minss (v4sf, v4sf)
8848 v4sf __builtin_ia32_andps (v4sf, v4sf)
8849 v4sf __builtin_ia32_andnps (v4sf, v4sf)
8850 v4sf __builtin_ia32_orps (v4sf, v4sf)
8851 v4sf __builtin_ia32_xorps (v4sf, v4sf)
8852 v4sf __builtin_ia32_movss (v4sf, v4sf)
8853 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
8854 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
8855 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
8856 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
8857 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
8858 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
8859 v2si __builtin_ia32_cvtps2pi (v4sf)
8860 int __builtin_ia32_cvtss2si (v4sf)
8861 v2si __builtin_ia32_cvttps2pi (v4sf)
8862 int __builtin_ia32_cvttss2si (v4sf)
8863 v4sf __builtin_ia32_rcpps (v4sf)
8864 v4sf __builtin_ia32_rsqrtps (v4sf)
8865 v4sf __builtin_ia32_sqrtps (v4sf)
8866 v4sf __builtin_ia32_rcpss (v4sf)
8867 v4sf __builtin_ia32_rsqrtss (v4sf)
8868 v4sf __builtin_ia32_sqrtss (v4sf)
8869 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
8870 void __builtin_ia32_movntps (float *, v4sf)
8871 int __builtin_ia32_movmskps (v4sf)
8874 The following built-in functions are available when @option{-msse} is used.
8877 @item v4sf __builtin_ia32_loadaps (float *)
8878 Generates the @code{movaps} machine instruction as a load from memory.
8879 @item void __builtin_ia32_storeaps (float *, v4sf)
8880 Generates the @code{movaps} machine instruction as a store to memory.
8881 @item v4sf __builtin_ia32_loadups (float *)
8882 Generates the @code{movups} machine instruction as a load from memory.
8883 @item void __builtin_ia32_storeups (float *, v4sf)
8884 Generates the @code{movups} machine instruction as a store to memory.
8885 @item v4sf __builtin_ia32_loadsss (float *)
8886 Generates the @code{movss} machine instruction as a load from memory.
8887 @item void __builtin_ia32_storess (float *, v4sf)
8888 Generates the @code{movss} machine instruction as a store to memory.
8889 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
8890 Generates the @code{movhps} machine instruction as a load from memory.
8891 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
8892 Generates the @code{movlps} machine instruction as a load from memory
8893 @item void __builtin_ia32_storehps (v2sf *, v4sf)
8894 Generates the @code{movhps} machine instruction as a store to memory.
8895 @item void __builtin_ia32_storelps (v2sf *, v4sf)
8896 Generates the @code{movlps} machine instruction as a store to memory.
8899 The following built-in functions are available when @option{-msse2} is used.
8900 All of them generate the machine instruction that is part of the name.
8903 int __builtin_ia32_comisdeq (v2df, v2df)
8904 int __builtin_ia32_comisdlt (v2df, v2df)
8905 int __builtin_ia32_comisdle (v2df, v2df)
8906 int __builtin_ia32_comisdgt (v2df, v2df)
8907 int __builtin_ia32_comisdge (v2df, v2df)
8908 int __builtin_ia32_comisdneq (v2df, v2df)
8909 int __builtin_ia32_ucomisdeq (v2df, v2df)
8910 int __builtin_ia32_ucomisdlt (v2df, v2df)
8911 int __builtin_ia32_ucomisdle (v2df, v2df)
8912 int __builtin_ia32_ucomisdgt (v2df, v2df)
8913 int __builtin_ia32_ucomisdge (v2df, v2df)
8914 int __builtin_ia32_ucomisdneq (v2df, v2df)
8915 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
8916 v2df __builtin_ia32_cmpltpd (v2df, v2df)
8917 v2df __builtin_ia32_cmplepd (v2df, v2df)
8918 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
8919 v2df __builtin_ia32_cmpgepd (v2df, v2df)
8920 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
8921 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
8922 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
8923 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
8924 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
8925 v2df __builtin_ia32_cmpngepd (v2df, v2df)
8926 v2df __builtin_ia32_cmpordpd (v2df, v2df)
8927 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
8928 v2df __builtin_ia32_cmpltsd (v2df, v2df)
8929 v2df __builtin_ia32_cmplesd (v2df, v2df)
8930 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
8931 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
8932 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
8933 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
8934 v2df __builtin_ia32_cmpordsd (v2df, v2df)
8935 v2di __builtin_ia32_paddq (v2di, v2di)
8936 v2di __builtin_ia32_psubq (v2di, v2di)
8937 v2df __builtin_ia32_addpd (v2df, v2df)
8938 v2df __builtin_ia32_subpd (v2df, v2df)
8939 v2df __builtin_ia32_mulpd (v2df, v2df)
8940 v2df __builtin_ia32_divpd (v2df, v2df)
8941 v2df __builtin_ia32_addsd (v2df, v2df)
8942 v2df __builtin_ia32_subsd (v2df, v2df)
8943 v2df __builtin_ia32_mulsd (v2df, v2df)
8944 v2df __builtin_ia32_divsd (v2df, v2df)
8945 v2df __builtin_ia32_minpd (v2df, v2df)
8946 v2df __builtin_ia32_maxpd (v2df, v2df)
8947 v2df __builtin_ia32_minsd (v2df, v2df)
8948 v2df __builtin_ia32_maxsd (v2df, v2df)
8949 v2df __builtin_ia32_andpd (v2df, v2df)
8950 v2df __builtin_ia32_andnpd (v2df, v2df)
8951 v2df __builtin_ia32_orpd (v2df, v2df)
8952 v2df __builtin_ia32_xorpd (v2df, v2df)
8953 v2df __builtin_ia32_movsd (v2df, v2df)
8954 v2df __builtin_ia32_unpckhpd (v2df, v2df)
8955 v2df __builtin_ia32_unpcklpd (v2df, v2df)
8956 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
8957 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
8958 v4si __builtin_ia32_paddd128 (v4si, v4si)
8959 v2di __builtin_ia32_paddq128 (v2di, v2di)
8960 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
8961 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
8962 v4si __builtin_ia32_psubd128 (v4si, v4si)
8963 v2di __builtin_ia32_psubq128 (v2di, v2di)
8964 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
8965 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
8966 v2di __builtin_ia32_pand128 (v2di, v2di)
8967 v2di __builtin_ia32_pandn128 (v2di, v2di)
8968 v2di __builtin_ia32_por128 (v2di, v2di)
8969 v2di __builtin_ia32_pxor128 (v2di, v2di)
8970 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
8971 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
8972 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
8973 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
8974 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
8975 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
8976 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
8977 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
8978 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
8979 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
8980 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
8981 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
8982 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
8983 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
8984 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
8985 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
8986 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
8987 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
8988 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
8989 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
8990 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
8991 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
8992 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
8993 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
8994 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
8995 v2df __builtin_ia32_loadupd (double *)
8996 void __builtin_ia32_storeupd (double *, v2df)
8997 v2df __builtin_ia32_loadhpd (v2df, double const *)
8998 v2df __builtin_ia32_loadlpd (v2df, double const *)
8999 int __builtin_ia32_movmskpd (v2df)
9000 int __builtin_ia32_pmovmskb128 (v16qi)
9001 void __builtin_ia32_movnti (int *, int)
9002 void __builtin_ia32_movntpd (double *, v2df)
9003 void __builtin_ia32_movntdq (v2df *, v2df)
9004 v4si __builtin_ia32_pshufd (v4si, int)
9005 v8hi __builtin_ia32_pshuflw (v8hi, int)
9006 v8hi __builtin_ia32_pshufhw (v8hi, int)
9007 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
9008 v2df __builtin_ia32_sqrtpd (v2df)
9009 v2df __builtin_ia32_sqrtsd (v2df)
9010 v2df __builtin_ia32_shufpd (v2df, v2df, int)
9011 v2df __builtin_ia32_cvtdq2pd (v4si)
9012 v4sf __builtin_ia32_cvtdq2ps (v4si)
9013 v4si __builtin_ia32_cvtpd2dq (v2df)
9014 v2si __builtin_ia32_cvtpd2pi (v2df)
9015 v4sf __builtin_ia32_cvtpd2ps (v2df)
9016 v4si __builtin_ia32_cvttpd2dq (v2df)
9017 v2si __builtin_ia32_cvttpd2pi (v2df)
9018 v2df __builtin_ia32_cvtpi2pd (v2si)
9019 int __builtin_ia32_cvtsd2si (v2df)
9020 int __builtin_ia32_cvttsd2si (v2df)
9021 long long __builtin_ia32_cvtsd2si64 (v2df)
9022 long long __builtin_ia32_cvttsd2si64 (v2df)
9023 v4si __builtin_ia32_cvtps2dq (v4sf)
9024 v2df __builtin_ia32_cvtps2pd (v4sf)
9025 v4si __builtin_ia32_cvttps2dq (v4sf)
9026 v2df __builtin_ia32_cvtsi2sd (v2df, int)
9027 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
9028 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
9029 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
9030 void __builtin_ia32_clflush (const void *)
9031 void __builtin_ia32_lfence (void)
9032 void __builtin_ia32_mfence (void)
9033 v16qi __builtin_ia32_loaddqu (const char *)
9034 void __builtin_ia32_storedqu (char *, v16qi)
9035 v1di __builtin_ia32_pmuludq (v2si, v2si)
9036 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
9037 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
9038 v4si __builtin_ia32_pslld128 (v4si, v4si)
9039 v2di __builtin_ia32_psllq128 (v2di, v2di)
9040 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
9041 v4si __builtin_ia32_psrld128 (v4si, v4si)
9042 v2di __builtin_ia32_psrlq128 (v2di, v2di)
9043 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
9044 v4si __builtin_ia32_psrad128 (v4si, v4si)
9045 v2di __builtin_ia32_pslldqi128 (v2di, int)
9046 v8hi __builtin_ia32_psllwi128 (v8hi, int)
9047 v4si __builtin_ia32_pslldi128 (v4si, int)
9048 v2di __builtin_ia32_psllqi128 (v2di, int)
9049 v2di __builtin_ia32_psrldqi128 (v2di, int)
9050 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
9051 v4si __builtin_ia32_psrldi128 (v4si, int)
9052 v2di __builtin_ia32_psrlqi128 (v2di, int)
9053 v8hi __builtin_ia32_psrawi128 (v8hi, int)
9054 v4si __builtin_ia32_psradi128 (v4si, int)
9055 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
9056 v2di __builtin_ia32_movq128 (v2di)
9059 The following built-in functions are available when @option{-msse3} is used.
9060 All of them generate the machine instruction that is part of the name.
9063 v2df __builtin_ia32_addsubpd (v2df, v2df)
9064 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
9065 v2df __builtin_ia32_haddpd (v2df, v2df)
9066 v4sf __builtin_ia32_haddps (v4sf, v4sf)
9067 v2df __builtin_ia32_hsubpd (v2df, v2df)
9068 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
9069 v16qi __builtin_ia32_lddqu (char const *)
9070 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
9071 v2df __builtin_ia32_movddup (v2df)
9072 v4sf __builtin_ia32_movshdup (v4sf)
9073 v4sf __builtin_ia32_movsldup (v4sf)
9074 void __builtin_ia32_mwait (unsigned int, unsigned int)
9077 The following built-in functions are available when @option{-msse3} is used.
9080 @item v2df __builtin_ia32_loadddup (double const *)
9081 Generates the @code{movddup} machine instruction as a load from memory.
9084 The following built-in functions are available when @option{-mssse3} is used.
9085 All of them generate the machine instruction that is part of the name
9089 v2si __builtin_ia32_phaddd (v2si, v2si)
9090 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
9091 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
9092 v2si __builtin_ia32_phsubd (v2si, v2si)
9093 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
9094 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
9095 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
9096 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
9097 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
9098 v8qi __builtin_ia32_psignb (v8qi, v8qi)
9099 v2si __builtin_ia32_psignd (v2si, v2si)
9100 v4hi __builtin_ia32_psignw (v4hi, v4hi)
9101 v1di __builtin_ia32_palignr (v1di, v1di, int)
9102 v8qi __builtin_ia32_pabsb (v8qi)
9103 v2si __builtin_ia32_pabsd (v2si)
9104 v4hi __builtin_ia32_pabsw (v4hi)
9107 The following built-in functions are available when @option{-mssse3} is used.
9108 All of them generate the machine instruction that is part of the name
9112 v4si __builtin_ia32_phaddd128 (v4si, v4si)
9113 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
9114 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
9115 v4si __builtin_ia32_phsubd128 (v4si, v4si)
9116 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
9117 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
9118 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
9119 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
9120 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
9121 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
9122 v4si __builtin_ia32_psignd128 (v4si, v4si)
9123 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
9124 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
9125 v16qi __builtin_ia32_pabsb128 (v16qi)
9126 v4si __builtin_ia32_pabsd128 (v4si)
9127 v8hi __builtin_ia32_pabsw128 (v8hi)
9130 The following built-in functions are available when @option{-msse4.1} is
9131 used. All of them generate the machine instruction that is part of the
9135 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
9136 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
9137 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
9138 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
9139 v2df __builtin_ia32_dppd (v2df, v2df, const int)
9140 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
9141 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
9142 v2di __builtin_ia32_movntdqa (v2di *);
9143 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
9144 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
9145 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
9146 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
9147 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
9148 v8hi __builtin_ia32_phminposuw128 (v8hi)
9149 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
9150 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
9151 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
9152 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
9153 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
9154 v4si __builtin_ia32_pminsd128 (v4si, v4si)
9155 v4si __builtin_ia32_pminud128 (v4si, v4si)
9156 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
9157 v4si __builtin_ia32_pmovsxbd128 (v16qi)
9158 v2di __builtin_ia32_pmovsxbq128 (v16qi)
9159 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
9160 v2di __builtin_ia32_pmovsxdq128 (v4si)
9161 v4si __builtin_ia32_pmovsxwd128 (v8hi)
9162 v2di __builtin_ia32_pmovsxwq128 (v8hi)
9163 v4si __builtin_ia32_pmovzxbd128 (v16qi)
9164 v2di __builtin_ia32_pmovzxbq128 (v16qi)
9165 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
9166 v2di __builtin_ia32_pmovzxdq128 (v4si)
9167 v4si __builtin_ia32_pmovzxwd128 (v8hi)
9168 v2di __builtin_ia32_pmovzxwq128 (v8hi)
9169 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
9170 v4si __builtin_ia32_pmulld128 (v4si, v4si)
9171 int __builtin_ia32_ptestc128 (v2di, v2di)
9172 int __builtin_ia32_ptestnzc128 (v2di, v2di)
9173 int __builtin_ia32_ptestz128 (v2di, v2di)
9174 v2df __builtin_ia32_roundpd (v2df, const int)
9175 v4sf __builtin_ia32_roundps (v4sf, const int)
9176 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
9177 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
9180 The following built-in functions are available when @option{-msse4.1} is
9184 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
9185 Generates the @code{insertps} machine instruction.
9186 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
9187 Generates the @code{pextrb} machine instruction.
9188 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
9189 Generates the @code{pinsrb} machine instruction.
9190 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
9191 Generates the @code{pinsrd} machine instruction.
9192 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
9193 Generates the @code{pinsrq} machine instruction in 64bit mode.
9196 The following built-in functions are changed to generate new SSE4.1
9197 instructions when @option{-msse4.1} is used.
9200 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
9201 Generates the @code{extractps} machine instruction.
9202 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
9203 Generates the @code{pextrd} machine instruction.
9204 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
9205 Generates the @code{pextrq} machine instruction in 64bit mode.
9208 The following built-in functions are available when @option{-msse4.2} is
9209 used. All of them generate the machine instruction that is part of the
9213 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
9214 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
9215 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
9216 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
9217 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
9218 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
9219 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
9220 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
9221 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
9222 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
9223 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
9224 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
9225 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
9226 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
9227 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
9230 The following built-in functions are available when @option{-msse4.2} is
9234 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
9235 Generates the @code{crc32b} machine instruction.
9236 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
9237 Generates the @code{crc32w} machine instruction.
9238 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
9239 Generates the @code{crc32l} machine instruction.
9240 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
9241 Generates the @code{crc32q} machine instruction.
9244 The following built-in functions are changed to generate new SSE4.2
9245 instructions when @option{-msse4.2} is used.
9248 @item int __builtin_popcount (unsigned int)
9249 Generates the @code{popcntl} machine instruction.
9250 @item int __builtin_popcountl (unsigned long)
9251 Generates the @code{popcntl} or @code{popcntq} machine instruction,
9252 depending on the size of @code{unsigned long}.
9253 @item int __builtin_popcountll (unsigned long long)
9254 Generates the @code{popcntq} machine instruction.
9257 The following built-in functions are available when @option{-mavx} is
9258 used. All of them generate the machine instruction that is part of the
9262 v4df __builtin_ia32_addpd256 (v4df,v4df)
9263 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
9264 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
9265 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
9266 v4df __builtin_ia32_andnpd256 (v4df,v4df)
9267 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
9268 v4df __builtin_ia32_andpd256 (v4df,v4df)
9269 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
9270 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
9271 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
9272 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
9273 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
9274 v2df __builtin_ia32_cmppd (v2df,v2df,int)
9275 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
9276 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
9277 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
9278 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
9279 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
9280 v4df __builtin_ia32_cvtdq2pd256 (v4si)
9281 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
9282 v4si __builtin_ia32_cvtpd2dq256 (v4df)
9283 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
9284 v8si __builtin_ia32_cvtps2dq256 (v8sf)
9285 v4df __builtin_ia32_cvtps2pd256 (v4sf)
9286 v4si __builtin_ia32_cvttpd2dq256 (v4df)
9287 v8si __builtin_ia32_cvttps2dq256 (v8sf)
9288 v4df __builtin_ia32_divpd256 (v4df,v4df)
9289 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
9290 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
9291 v4df __builtin_ia32_haddpd256 (v4df,v4df)
9292 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
9293 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
9294 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
9295 v32qi __builtin_ia32_lddqu256 (pcchar)
9296 v32qi __builtin_ia32_loaddqu256 (pcchar)
9297 v4df __builtin_ia32_loadupd256 (pcdouble)
9298 v8sf __builtin_ia32_loadups256 (pcfloat)
9299 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
9300 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
9301 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
9302 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
9303 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
9304 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
9305 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
9306 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
9307 v4df __builtin_ia32_maxpd256 (v4df,v4df)
9308 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
9309 v4df __builtin_ia32_minpd256 (v4df,v4df)
9310 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
9311 v4df __builtin_ia32_movddup256 (v4df)
9312 int __builtin_ia32_movmskpd256 (v4df)
9313 int __builtin_ia32_movmskps256 (v8sf)
9314 v8sf __builtin_ia32_movshdup256 (v8sf)
9315 v8sf __builtin_ia32_movsldup256 (v8sf)
9316 v4df __builtin_ia32_mulpd256 (v4df,v4df)
9317 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
9318 v4df __builtin_ia32_orpd256 (v4df,v4df)
9319 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
9320 v2df __builtin_ia32_pd_pd256 (v4df)
9321 v4df __builtin_ia32_pd256_pd (v2df)
9322 v4sf __builtin_ia32_ps_ps256 (v8sf)
9323 v8sf __builtin_ia32_ps256_ps (v4sf)
9324 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
9325 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
9326 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
9327 v8sf __builtin_ia32_rcpps256 (v8sf)
9328 v4df __builtin_ia32_roundpd256 (v4df,int)
9329 v8sf __builtin_ia32_roundps256 (v8sf,int)
9330 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
9331 v8sf __builtin_ia32_rsqrtps256 (v8sf)
9332 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
9333 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
9334 v4si __builtin_ia32_si_si256 (v8si)
9335 v8si __builtin_ia32_si256_si (v4si)
9336 v4df __builtin_ia32_sqrtpd256 (v4df)
9337 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
9338 v8sf __builtin_ia32_sqrtps256 (v8sf)
9339 void __builtin_ia32_storedqu256 (pchar,v32qi)
9340 void __builtin_ia32_storeupd256 (pdouble,v4df)
9341 void __builtin_ia32_storeups256 (pfloat,v8sf)
9342 v4df __builtin_ia32_subpd256 (v4df,v4df)
9343 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
9344 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
9345 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
9346 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
9347 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
9348 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
9349 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
9350 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
9351 v4sf __builtin_ia32_vbroadcastss (pcfloat)
9352 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
9353 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
9354 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
9355 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
9356 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
9357 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
9358 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
9359 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
9360 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
9361 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
9362 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
9363 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
9364 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
9365 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
9366 v2df __builtin_ia32_vpermilpd (v2df,int)
9367 v4df __builtin_ia32_vpermilpd256 (v4df,int)
9368 v4sf __builtin_ia32_vpermilps (v4sf,int)
9369 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
9370 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
9371 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
9372 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
9373 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
9374 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
9375 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
9376 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
9377 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
9378 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
9379 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
9380 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
9381 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
9382 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
9383 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
9384 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
9385 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
9386 void __builtin_ia32_vzeroall (void)
9387 void __builtin_ia32_vzeroupper (void)
9388 v4df __builtin_ia32_xorpd256 (v4df,v4df)
9389 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
9392 The following built-in functions are available when @option{-maes} is
9393 used. All of them generate the machine instruction that is part of the
9397 v2di __builtin_ia32_aesenc128 (v2di, v2di)
9398 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
9399 v2di __builtin_ia32_aesdec128 (v2di, v2di)
9400 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
9401 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
9402 v2di __builtin_ia32_aesimc128 (v2di)
9405 The following built-in function is available when @option{-mpclmul} is
9409 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
9410 Generates the @code{pclmulqdq} machine instruction.
9413 The following built-in function is available when @option{-mfsgsbase} is
9414 used. All of them generate the machine instruction that is part of the
9418 unsigned int __builtin_ia32_rdfsbase32 (void)
9419 unsigned long long __builtin_ia32_rdfsbase64 (void)
9420 unsigned int __builtin_ia32_rdgsbase32 (void)
9421 unsigned long long __builtin_ia32_rdgsbase64 (void)
9422 void _writefsbase_u32 (unsigned int)
9423 void _writefsbase_u64 (unsigned long long)
9424 void _writegsbase_u32 (unsigned int)
9425 void _writegsbase_u64 (unsigned long long)
9428 The following built-in function is available when @option{-mrdrnd} is
9429 used. All of them generate the machine instruction that is part of the
9433 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
9434 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
9435 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
9438 The following built-in functions are available when @option{-msse4a} is used.
9439 All of them generate the machine instruction that is part of the name.
9442 void __builtin_ia32_movntsd (double *, v2df)
9443 void __builtin_ia32_movntss (float *, v4sf)
9444 v2di __builtin_ia32_extrq (v2di, v16qi)
9445 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
9446 v2di __builtin_ia32_insertq (v2di, v2di)
9447 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
9450 The following built-in functions are available when @option{-mxop} is used.
9452 v2df __builtin_ia32_vfrczpd (v2df)
9453 v4sf __builtin_ia32_vfrczps (v4sf)
9454 v2df __builtin_ia32_vfrczsd (v2df, v2df)
9455 v4sf __builtin_ia32_vfrczss (v4sf, v4sf)
9456 v4df __builtin_ia32_vfrczpd256 (v4df)
9457 v8sf __builtin_ia32_vfrczps256 (v8sf)
9458 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
9459 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
9460 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
9461 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
9462 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
9463 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
9464 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
9465 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
9466 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
9467 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
9468 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
9469 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
9470 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
9471 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
9472 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
9473 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
9474 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
9475 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
9476 v4si __builtin_ia32_vpcomequd (v4si, v4si)
9477 v2di __builtin_ia32_vpcomequq (v2di, v2di)
9478 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
9479 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
9480 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
9481 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
9482 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
9483 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
9484 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
9485 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
9486 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
9487 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
9488 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
9489 v4si __builtin_ia32_vpcomged (v4si, v4si)
9490 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
9491 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
9492 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
9493 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
9494 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
9495 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
9496 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
9497 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
9498 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
9499 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
9500 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
9501 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
9502 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
9503 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
9504 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
9505 v4si __builtin_ia32_vpcomled (v4si, v4si)
9506 v2di __builtin_ia32_vpcomleq (v2di, v2di)
9507 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
9508 v4si __builtin_ia32_vpcomleud (v4si, v4si)
9509 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
9510 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
9511 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
9512 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
9513 v4si __builtin_ia32_vpcomltd (v4si, v4si)
9514 v2di __builtin_ia32_vpcomltq (v2di, v2di)
9515 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
9516 v4si __builtin_ia32_vpcomltud (v4si, v4si)
9517 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
9518 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
9519 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
9520 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
9521 v4si __builtin_ia32_vpcomned (v4si, v4si)
9522 v2di __builtin_ia32_vpcomneq (v2di, v2di)
9523 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
9524 v4si __builtin_ia32_vpcomneud (v4si, v4si)
9525 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
9526 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
9527 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
9528 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
9529 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
9530 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
9531 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
9532 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
9533 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
9534 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
9535 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
9536 v4si __builtin_ia32_vphaddbd (v16qi)
9537 v2di __builtin_ia32_vphaddbq (v16qi)
9538 v8hi __builtin_ia32_vphaddbw (v16qi)
9539 v2di __builtin_ia32_vphadddq (v4si)
9540 v4si __builtin_ia32_vphaddubd (v16qi)
9541 v2di __builtin_ia32_vphaddubq (v16qi)
9542 v8hi __builtin_ia32_vphaddubw (v16qi)
9543 v2di __builtin_ia32_vphaddudq (v4si)
9544 v4si __builtin_ia32_vphadduwd (v8hi)
9545 v2di __builtin_ia32_vphadduwq (v8hi)
9546 v4si __builtin_ia32_vphaddwd (v8hi)
9547 v2di __builtin_ia32_vphaddwq (v8hi)
9548 v8hi __builtin_ia32_vphsubbw (v16qi)
9549 v2di __builtin_ia32_vphsubdq (v4si)
9550 v4si __builtin_ia32_vphsubwd (v8hi)
9551 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
9552 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
9553 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
9554 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
9555 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
9556 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
9557 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
9558 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
9559 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
9560 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
9561 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
9562 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
9563 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
9564 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
9565 v4si __builtin_ia32_vprotd (v4si, v4si)
9566 v2di __builtin_ia32_vprotq (v2di, v2di)
9567 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
9568 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
9569 v4si __builtin_ia32_vpshad (v4si, v4si)
9570 v2di __builtin_ia32_vpshaq (v2di, v2di)
9571 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
9572 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
9573 v4si __builtin_ia32_vpshld (v4si, v4si)
9574 v2di __builtin_ia32_vpshlq (v2di, v2di)
9575 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
9578 The following built-in functions are available when @option{-mfma4} is used.
9579 All of them generate the machine instruction that is part of the name
9583 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
9584 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
9585 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
9586 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
9587 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
9588 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
9589 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
9590 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
9591 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
9592 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
9593 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
9594 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
9595 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
9596 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
9597 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
9598 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
9599 v2df __builtin_ia32_fmaddsubpd (v2df, v2df, v2df)
9600 v4sf __builtin_ia32_fmaddsubps (v4sf, v4sf, v4sf)
9601 v2df __builtin_ia32_fmsubaddpd (v2df, v2df, v2df)
9602 v4sf __builtin_ia32_fmsubaddps (v4sf, v4sf, v4sf)
9603 v4df __builtin_ia32_fmaddpd256 (v4df, v4df, v4df)
9604 v8sf __builtin_ia32_fmaddps256 (v8sf, v8sf, v8sf)
9605 v4df __builtin_ia32_fmsubpd256 (v4df, v4df, v4df)
9606 v8sf __builtin_ia32_fmsubps256 (v8sf, v8sf, v8sf)
9607 v4df __builtin_ia32_fnmaddpd256 (v4df, v4df, v4df)
9608 v8sf __builtin_ia32_fnmaddps256 (v8sf, v8sf, v8sf)
9609 v4df __builtin_ia32_fnmsubpd256 (v4df, v4df, v4df)
9610 v8sf __builtin_ia32_fnmsubps256 (v8sf, v8sf, v8sf)
9611 v4df __builtin_ia32_fmaddsubpd256 (v4df, v4df, v4df)
9612 v8sf __builtin_ia32_fmaddsubps256 (v8sf, v8sf, v8sf)
9613 v4df __builtin_ia32_fmsubaddpd256 (v4df, v4df, v4df)
9614 v8sf __builtin_ia32_fmsubaddps256 (v8sf, v8sf, v8sf)
9618 The following built-in functions are available when @option{-mlwp} is used.
9621 void __builtin_ia32_llwpcb16 (void *);
9622 void __builtin_ia32_llwpcb32 (void *);
9623 void __builtin_ia32_llwpcb64 (void *);
9624 void * __builtin_ia32_llwpcb16 (void);
9625 void * __builtin_ia32_llwpcb32 (void);
9626 void * __builtin_ia32_llwpcb64 (void);
9627 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
9628 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
9629 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
9630 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
9631 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
9632 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
9635 The following built-in functions are available when @option{-mbmi} is used.
9636 All of them generate the machine instruction that is part of the name.
9638 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
9639 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
9640 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
9641 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
9642 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
9645 The following built-in functions are available when @option{-mtbm} is used.
9646 Both of them generate the immediate form of the bextr machine instruction.
9648 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
9649 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
9653 The following built-in functions are available when @option{-m3dnow} is used.
9654 All of them generate the machine instruction that is part of the name.
9657 void __builtin_ia32_femms (void)
9658 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
9659 v2si __builtin_ia32_pf2id (v2sf)
9660 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
9661 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
9662 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
9663 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
9664 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
9665 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
9666 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
9667 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
9668 v2sf __builtin_ia32_pfrcp (v2sf)
9669 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
9670 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
9671 v2sf __builtin_ia32_pfrsqrt (v2sf)
9672 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
9673 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
9674 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
9675 v2sf __builtin_ia32_pi2fd (v2si)
9676 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
9679 The following built-in functions are available when both @option{-m3dnow}
9680 and @option{-march=athlon} are used. All of them generate the machine
9681 instruction that is part of the name.
9684 v2si __builtin_ia32_pf2iw (v2sf)
9685 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
9686 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
9687 v2sf __builtin_ia32_pi2fw (v2si)
9688 v2sf __builtin_ia32_pswapdsf (v2sf)
9689 v2si __builtin_ia32_pswapdsi (v2si)
9692 @node MIPS DSP Built-in Functions
9693 @subsection MIPS DSP Built-in Functions
9695 The MIPS DSP Application-Specific Extension (ASE) includes new
9696 instructions that are designed to improve the performance of DSP and
9697 media applications. It provides instructions that operate on packed
9698 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
9700 GCC supports MIPS DSP operations using both the generic
9701 vector extensions (@pxref{Vector Extensions}) and a collection of
9702 MIPS-specific built-in functions. Both kinds of support are
9703 enabled by the @option{-mdsp} command-line option.
9705 Revision 2 of the ASE was introduced in the second half of 2006.
9706 This revision adds extra instructions to the original ASE, but is
9707 otherwise backwards-compatible with it. You can select revision 2
9708 using the command-line option @option{-mdspr2}; this option implies
9711 The SCOUNT and POS bits of the DSP control register are global. The
9712 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
9713 POS bits. During optimization, the compiler will not delete these
9714 instructions and it will not delete calls to functions containing
9717 At present, GCC only provides support for operations on 32-bit
9718 vectors. The vector type associated with 8-bit integer data is
9719 usually called @code{v4i8}, the vector type associated with Q7
9720 is usually called @code{v4q7}, the vector type associated with 16-bit
9721 integer data is usually called @code{v2i16}, and the vector type
9722 associated with Q15 is usually called @code{v2q15}. They can be
9723 defined in C as follows:
9726 typedef signed char v4i8 __attribute__ ((vector_size(4)));
9727 typedef signed char v4q7 __attribute__ ((vector_size(4)));
9728 typedef short v2i16 __attribute__ ((vector_size(4)));
9729 typedef short v2q15 __attribute__ ((vector_size(4)));
9732 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
9733 initialized in the same way as aggregates. For example:
9736 v4i8 a = @{1, 2, 3, 4@};
9738 b = (v4i8) @{5, 6, 7, 8@};
9740 v2q15 c = @{0x0fcb, 0x3a75@};
9742 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
9745 @emph{Note:} The CPU's endianness determines the order in which values
9746 are packed. On little-endian targets, the first value is the least
9747 significant and the last value is the most significant. The opposite
9748 order applies to big-endian targets. For example, the code above will
9749 set the lowest byte of @code{a} to @code{1} on little-endian targets
9750 and @code{4} on big-endian targets.
9752 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
9753 representation. As shown in this example, the integer representation
9754 of a Q7 value can be obtained by multiplying the fractional value by
9755 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
9756 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
9759 The table below lists the @code{v4i8} and @code{v2q15} operations for which
9760 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
9761 and @code{c} and @code{d} are @code{v2q15} values.
9763 @multitable @columnfractions .50 .50
9764 @item C code @tab MIPS instruction
9765 @item @code{a + b} @tab @code{addu.qb}
9766 @item @code{c + d} @tab @code{addq.ph}
9767 @item @code{a - b} @tab @code{subu.qb}
9768 @item @code{c - d} @tab @code{subq.ph}
9771 The table below lists the @code{v2i16} operation for which
9772 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
9773 @code{v2i16} values.
9775 @multitable @columnfractions .50 .50
9776 @item C code @tab MIPS instruction
9777 @item @code{e * f} @tab @code{mul.ph}
9780 It is easier to describe the DSP built-in functions if we first define
9781 the following types:
9786 typedef unsigned int ui32;
9787 typedef long long a64;
9790 @code{q31} and @code{i32} are actually the same as @code{int}, but we
9791 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
9792 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
9793 @code{long long}, but we use @code{a64} to indicate values that will
9794 be placed in one of the four DSP accumulators (@code{$ac0},
9795 @code{$ac1}, @code{$ac2} or @code{$ac3}).
9797 Also, some built-in functions prefer or require immediate numbers as
9798 parameters, because the corresponding DSP instructions accept both immediate
9799 numbers and register operands, or accept immediate numbers only. The
9800 immediate parameters are listed as follows.
9809 imm_n32_31: -32 to 31.
9810 imm_n512_511: -512 to 511.
9813 The following built-in functions map directly to a particular MIPS DSP
9814 instruction. Please refer to the architecture specification
9815 for details on what each instruction does.
9818 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
9819 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
9820 q31 __builtin_mips_addq_s_w (q31, q31)
9821 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
9822 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
9823 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
9824 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
9825 q31 __builtin_mips_subq_s_w (q31, q31)
9826 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
9827 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
9828 i32 __builtin_mips_addsc (i32, i32)
9829 i32 __builtin_mips_addwc (i32, i32)
9830 i32 __builtin_mips_modsub (i32, i32)
9831 i32 __builtin_mips_raddu_w_qb (v4i8)
9832 v2q15 __builtin_mips_absq_s_ph (v2q15)
9833 q31 __builtin_mips_absq_s_w (q31)
9834 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
9835 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
9836 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
9837 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
9838 q31 __builtin_mips_preceq_w_phl (v2q15)
9839 q31 __builtin_mips_preceq_w_phr (v2q15)
9840 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
9841 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
9842 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
9843 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
9844 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
9845 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
9846 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
9847 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
9848 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
9849 v4i8 __builtin_mips_shll_qb (v4i8, i32)
9850 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
9851 v2q15 __builtin_mips_shll_ph (v2q15, i32)
9852 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
9853 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
9854 q31 __builtin_mips_shll_s_w (q31, imm0_31)
9855 q31 __builtin_mips_shll_s_w (q31, i32)
9856 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
9857 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
9858 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
9859 v2q15 __builtin_mips_shra_ph (v2q15, i32)
9860 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
9861 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
9862 q31 __builtin_mips_shra_r_w (q31, imm0_31)
9863 q31 __builtin_mips_shra_r_w (q31, i32)
9864 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
9865 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
9866 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
9867 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
9868 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
9869 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
9870 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
9871 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
9872 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
9873 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
9874 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
9875 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
9876 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
9877 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
9878 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
9879 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
9880 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
9881 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
9882 i32 __builtin_mips_bitrev (i32)
9883 i32 __builtin_mips_insv (i32, i32)
9884 v4i8 __builtin_mips_repl_qb (imm0_255)
9885 v4i8 __builtin_mips_repl_qb (i32)
9886 v2q15 __builtin_mips_repl_ph (imm_n512_511)
9887 v2q15 __builtin_mips_repl_ph (i32)
9888 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
9889 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
9890 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
9891 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
9892 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
9893 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
9894 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
9895 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
9896 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
9897 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
9898 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
9899 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
9900 i32 __builtin_mips_extr_w (a64, imm0_31)
9901 i32 __builtin_mips_extr_w (a64, i32)
9902 i32 __builtin_mips_extr_r_w (a64, imm0_31)
9903 i32 __builtin_mips_extr_s_h (a64, i32)
9904 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
9905 i32 __builtin_mips_extr_rs_w (a64, i32)
9906 i32 __builtin_mips_extr_s_h (a64, imm0_31)
9907 i32 __builtin_mips_extr_r_w (a64, i32)
9908 i32 __builtin_mips_extp (a64, imm0_31)
9909 i32 __builtin_mips_extp (a64, i32)
9910 i32 __builtin_mips_extpdp (a64, imm0_31)
9911 i32 __builtin_mips_extpdp (a64, i32)
9912 a64 __builtin_mips_shilo (a64, imm_n32_31)
9913 a64 __builtin_mips_shilo (a64, i32)
9914 a64 __builtin_mips_mthlip (a64, i32)
9915 void __builtin_mips_wrdsp (i32, imm0_63)
9916 i32 __builtin_mips_rddsp (imm0_63)
9917 i32 __builtin_mips_lbux (void *, i32)
9918 i32 __builtin_mips_lhx (void *, i32)
9919 i32 __builtin_mips_lwx (void *, i32)
9920 i32 __builtin_mips_bposge32 (void)
9921 a64 __builtin_mips_madd (a64, i32, i32);
9922 a64 __builtin_mips_maddu (a64, ui32, ui32);
9923 a64 __builtin_mips_msub (a64, i32, i32);
9924 a64 __builtin_mips_msubu (a64, ui32, ui32);
9925 a64 __builtin_mips_mult (i32, i32);
9926 a64 __builtin_mips_multu (ui32, ui32);
9929 The following built-in functions map directly to a particular MIPS DSP REV 2
9930 instruction. Please refer to the architecture specification
9931 for details on what each instruction does.
9934 v4q7 __builtin_mips_absq_s_qb (v4q7);
9935 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
9936 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
9937 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
9938 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
9939 i32 __builtin_mips_append (i32, i32, imm0_31);
9940 i32 __builtin_mips_balign (i32, i32, imm0_3);
9941 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
9942 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
9943 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
9944 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
9945 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
9946 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
9947 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
9948 q31 __builtin_mips_mulq_rs_w (q31, q31);
9949 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
9950 q31 __builtin_mips_mulq_s_w (q31, q31);
9951 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
9952 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
9953 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
9954 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
9955 i32 __builtin_mips_prepend (i32, i32, imm0_31);
9956 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
9957 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
9958 v4i8 __builtin_mips_shra_qb (v4i8, i32);
9959 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
9960 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
9961 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
9962 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
9963 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
9964 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
9965 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
9966 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
9967 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
9968 q31 __builtin_mips_addqh_w (q31, q31);
9969 q31 __builtin_mips_addqh_r_w (q31, q31);
9970 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
9971 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
9972 q31 __builtin_mips_subqh_w (q31, q31);
9973 q31 __builtin_mips_subqh_r_w (q31, q31);
9974 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
9975 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
9976 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
9977 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
9978 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
9979 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
9983 @node MIPS Paired-Single Support
9984 @subsection MIPS Paired-Single Support
9986 The MIPS64 architecture includes a number of instructions that
9987 operate on pairs of single-precision floating-point values.
9988 Each pair is packed into a 64-bit floating-point register,
9989 with one element being designated the ``upper half'' and
9990 the other being designated the ``lower half''.
9992 GCC supports paired-single operations using both the generic
9993 vector extensions (@pxref{Vector Extensions}) and a collection of
9994 MIPS-specific built-in functions. Both kinds of support are
9995 enabled by the @option{-mpaired-single} command-line option.
9997 The vector type associated with paired-single values is usually
9998 called @code{v2sf}. It can be defined in C as follows:
10001 typedef float v2sf __attribute__ ((vector_size (8)));
10004 @code{v2sf} values are initialized in the same way as aggregates.
10008 v2sf a = @{1.5, 9.1@};
10011 b = (v2sf) @{e, f@};
10014 @emph{Note:} The CPU's endianness determines which value is stored in
10015 the upper half of a register and which value is stored in the lower half.
10016 On little-endian targets, the first value is the lower one and the second
10017 value is the upper one. The opposite order applies to big-endian targets.
10018 For example, the code above will set the lower half of @code{a} to
10019 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
10021 @node MIPS Loongson Built-in Functions
10022 @subsection MIPS Loongson Built-in Functions
10024 GCC provides intrinsics to access the SIMD instructions provided by the
10025 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
10026 available after inclusion of the @code{loongson.h} header file,
10027 operate on the following 64-bit vector types:
10030 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
10031 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
10032 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
10033 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
10034 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
10035 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
10038 The intrinsics provided are listed below; each is named after the
10039 machine instruction to which it corresponds, with suffixes added as
10040 appropriate to distinguish intrinsics that expand to the same machine
10041 instruction yet have different argument types. Refer to the architecture
10042 documentation for a description of the functionality of each
10046 int16x4_t packsswh (int32x2_t s, int32x2_t t);
10047 int8x8_t packsshb (int16x4_t s, int16x4_t t);
10048 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
10049 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
10050 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
10051 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
10052 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
10053 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
10054 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
10055 uint64_t paddd_u (uint64_t s, uint64_t t);
10056 int64_t paddd_s (int64_t s, int64_t t);
10057 int16x4_t paddsh (int16x4_t s, int16x4_t t);
10058 int8x8_t paddsb (int8x8_t s, int8x8_t t);
10059 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
10060 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
10061 uint64_t pandn_ud (uint64_t s, uint64_t t);
10062 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
10063 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
10064 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
10065 int64_t pandn_sd (int64_t s, int64_t t);
10066 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
10067 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
10068 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
10069 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
10070 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
10071 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
10072 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
10073 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
10074 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
10075 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
10076 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
10077 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
10078 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
10079 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
10080 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
10081 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
10082 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
10083 uint16x4_t pextrh_u (uint16x4_t s, int field);
10084 int16x4_t pextrh_s (int16x4_t s, int field);
10085 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
10086 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
10087 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
10088 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
10089 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
10090 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
10091 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
10092 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
10093 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
10094 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
10095 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
10096 int16x4_t pminsh (int16x4_t s, int16x4_t t);
10097 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
10098 uint8x8_t pmovmskb_u (uint8x8_t s);
10099 int8x8_t pmovmskb_s (int8x8_t s);
10100 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
10101 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
10102 int16x4_t pmullh (int16x4_t s, int16x4_t t);
10103 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
10104 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
10105 uint16x4_t biadd (uint8x8_t s);
10106 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
10107 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
10108 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
10109 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
10110 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
10111 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
10112 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
10113 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
10114 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
10115 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
10116 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
10117 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
10118 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
10119 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
10120 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
10121 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
10122 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
10123 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
10124 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
10125 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
10126 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
10127 uint64_t psubd_u (uint64_t s, uint64_t t);
10128 int64_t psubd_s (int64_t s, int64_t t);
10129 int16x4_t psubsh (int16x4_t s, int16x4_t t);
10130 int8x8_t psubsb (int8x8_t s, int8x8_t t);
10131 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
10132 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
10133 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
10134 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
10135 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
10136 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
10137 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
10138 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
10139 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
10140 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
10141 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
10142 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
10143 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
10144 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
10148 * Paired-Single Arithmetic::
10149 * Paired-Single Built-in Functions::
10150 * MIPS-3D Built-in Functions::
10153 @node Paired-Single Arithmetic
10154 @subsubsection Paired-Single Arithmetic
10156 The table below lists the @code{v2sf} operations for which hardware
10157 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
10158 values and @code{x} is an integral value.
10160 @multitable @columnfractions .50 .50
10161 @item C code @tab MIPS instruction
10162 @item @code{a + b} @tab @code{add.ps}
10163 @item @code{a - b} @tab @code{sub.ps}
10164 @item @code{-a} @tab @code{neg.ps}
10165 @item @code{a * b} @tab @code{mul.ps}
10166 @item @code{a * b + c} @tab @code{madd.ps}
10167 @item @code{a * b - c} @tab @code{msub.ps}
10168 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
10169 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
10170 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
10173 Note that the multiply-accumulate instructions can be disabled
10174 using the command-line option @code{-mno-fused-madd}.
10176 @node Paired-Single Built-in Functions
10177 @subsubsection Paired-Single Built-in Functions
10179 The following paired-single functions map directly to a particular
10180 MIPS instruction. Please refer to the architecture specification
10181 for details on what each instruction does.
10184 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
10185 Pair lower lower (@code{pll.ps}).
10187 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
10188 Pair upper lower (@code{pul.ps}).
10190 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
10191 Pair lower upper (@code{plu.ps}).
10193 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
10194 Pair upper upper (@code{puu.ps}).
10196 @item v2sf __builtin_mips_cvt_ps_s (float, float)
10197 Convert pair to paired single (@code{cvt.ps.s}).
10199 @item float __builtin_mips_cvt_s_pl (v2sf)
10200 Convert pair lower to single (@code{cvt.s.pl}).
10202 @item float __builtin_mips_cvt_s_pu (v2sf)
10203 Convert pair upper to single (@code{cvt.s.pu}).
10205 @item v2sf __builtin_mips_abs_ps (v2sf)
10206 Absolute value (@code{abs.ps}).
10208 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
10209 Align variable (@code{alnv.ps}).
10211 @emph{Note:} The value of the third parameter must be 0 or 4
10212 modulo 8, otherwise the result will be unpredictable. Please read the
10213 instruction description for details.
10216 The following multi-instruction functions are also available.
10217 In each case, @var{cond} can be any of the 16 floating-point conditions:
10218 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
10219 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
10220 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
10223 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10224 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10225 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
10226 @code{movt.ps}/@code{movf.ps}).
10228 The @code{movt} functions return the value @var{x} computed by:
10231 c.@var{cond}.ps @var{cc},@var{a},@var{b}
10232 mov.ps @var{x},@var{c}
10233 movt.ps @var{x},@var{d},@var{cc}
10236 The @code{movf} functions are similar but use @code{movf.ps} instead
10239 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10240 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10241 Comparison of two paired-single values (@code{c.@var{cond}.ps},
10242 @code{bc1t}/@code{bc1f}).
10244 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
10245 and return either the upper or lower half of the result. For example:
10249 if (__builtin_mips_upper_c_eq_ps (a, b))
10250 upper_halves_are_equal ();
10252 upper_halves_are_unequal ();
10254 if (__builtin_mips_lower_c_eq_ps (a, b))
10255 lower_halves_are_equal ();
10257 lower_halves_are_unequal ();
10261 @node MIPS-3D Built-in Functions
10262 @subsubsection MIPS-3D Built-in Functions
10264 The MIPS-3D Application-Specific Extension (ASE) includes additional
10265 paired-single instructions that are designed to improve the performance
10266 of 3D graphics operations. Support for these instructions is controlled
10267 by the @option{-mips3d} command-line option.
10269 The functions listed below map directly to a particular MIPS-3D
10270 instruction. Please refer to the architecture specification for
10271 more details on what each instruction does.
10274 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
10275 Reduction add (@code{addr.ps}).
10277 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
10278 Reduction multiply (@code{mulr.ps}).
10280 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
10281 Convert paired single to paired word (@code{cvt.pw.ps}).
10283 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
10284 Convert paired word to paired single (@code{cvt.ps.pw}).
10286 @item float __builtin_mips_recip1_s (float)
10287 @itemx double __builtin_mips_recip1_d (double)
10288 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
10289 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
10291 @item float __builtin_mips_recip2_s (float, float)
10292 @itemx double __builtin_mips_recip2_d (double, double)
10293 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
10294 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
10296 @item float __builtin_mips_rsqrt1_s (float)
10297 @itemx double __builtin_mips_rsqrt1_d (double)
10298 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
10299 Reduced precision reciprocal square root (sequence step 1)
10300 (@code{rsqrt1.@var{fmt}}).
10302 @item float __builtin_mips_rsqrt2_s (float, float)
10303 @itemx double __builtin_mips_rsqrt2_d (double, double)
10304 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
10305 Reduced precision reciprocal square root (sequence step 2)
10306 (@code{rsqrt2.@var{fmt}}).
10309 The following multi-instruction functions are also available.
10310 In each case, @var{cond} can be any of the 16 floating-point conditions:
10311 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
10312 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
10313 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
10316 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
10317 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
10318 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
10319 @code{bc1t}/@code{bc1f}).
10321 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
10322 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
10327 if (__builtin_mips_cabs_eq_s (a, b))
10333 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10334 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10335 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
10336 @code{bc1t}/@code{bc1f}).
10338 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
10339 and return either the upper or lower half of the result. For example:
10343 if (__builtin_mips_upper_cabs_eq_ps (a, b))
10344 upper_halves_are_equal ();
10346 upper_halves_are_unequal ();
10348 if (__builtin_mips_lower_cabs_eq_ps (a, b))
10349 lower_halves_are_equal ();
10351 lower_halves_are_unequal ();
10354 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10355 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10356 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
10357 @code{movt.ps}/@code{movf.ps}).
10359 The @code{movt} functions return the value @var{x} computed by:
10362 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
10363 mov.ps @var{x},@var{c}
10364 movt.ps @var{x},@var{d},@var{cc}
10367 The @code{movf} functions are similar but use @code{movf.ps} instead
10370 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10371 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10372 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10373 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10374 Comparison of two paired-single values
10375 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
10376 @code{bc1any2t}/@code{bc1any2f}).
10378 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
10379 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
10380 result is true and the @code{all} forms return true if both results are true.
10385 if (__builtin_mips_any_c_eq_ps (a, b))
10390 if (__builtin_mips_all_c_eq_ps (a, b))
10396 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10397 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10398 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10399 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10400 Comparison of four paired-single values
10401 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
10402 @code{bc1any4t}/@code{bc1any4f}).
10404 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
10405 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
10406 The @code{any} forms return true if any of the four results are true
10407 and the @code{all} forms return true if all four results are true.
10412 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
10417 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
10424 @node picoChip Built-in Functions
10425 @subsection picoChip Built-in Functions
10427 GCC provides an interface to selected machine instructions from the
10428 picoChip instruction set.
10431 @item int __builtin_sbc (int @var{value})
10432 Sign bit count. Return the number of consecutive bits in @var{value}
10433 which have the same value as the sign-bit. The result is the number of
10434 leading sign bits minus one, giving the number of redundant sign bits in
10437 @item int __builtin_byteswap (int @var{value})
10438 Byte swap. Return the result of swapping the upper and lower bytes of
10441 @item int __builtin_brev (int @var{value})
10442 Bit reversal. Return the result of reversing the bits in
10443 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
10446 @item int __builtin_adds (int @var{x}, int @var{y})
10447 Saturating addition. Return the result of adding @var{x} and @var{y},
10448 storing the value 32767 if the result overflows.
10450 @item int __builtin_subs (int @var{x}, int @var{y})
10451 Saturating subtraction. Return the result of subtracting @var{y} from
10452 @var{x}, storing the value @minus{}32768 if the result overflows.
10454 @item void __builtin_halt (void)
10455 Halt. The processor will stop execution. This built-in is useful for
10456 implementing assertions.
10460 @node Other MIPS Built-in Functions
10461 @subsection Other MIPS Built-in Functions
10463 GCC provides other MIPS-specific built-in functions:
10466 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
10467 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
10468 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
10469 when this function is available.
10472 @node PowerPC AltiVec/VSX Built-in Functions
10473 @subsection PowerPC AltiVec Built-in Functions
10475 GCC provides an interface for the PowerPC family of processors to access
10476 the AltiVec operations described in Motorola's AltiVec Programming
10477 Interface Manual. The interface is made available by including
10478 @code{<altivec.h>} and using @option{-maltivec} and
10479 @option{-mabi=altivec}. The interface supports the following vector
10483 vector unsigned char
10487 vector unsigned short
10488 vector signed short
10492 vector unsigned int
10498 If @option{-mvsx} is used the following additional vector types are
10502 vector unsigned long
10507 The long types are only implemented for 64-bit code generation, and
10508 the long type is only used in the floating point/integer conversion
10511 GCC's implementation of the high-level language interface available from
10512 C and C++ code differs from Motorola's documentation in several ways.
10517 A vector constant is a list of constant expressions within curly braces.
10520 A vector initializer requires no cast if the vector constant is of the
10521 same type as the variable it is initializing.
10524 If @code{signed} or @code{unsigned} is omitted, the signedness of the
10525 vector type is the default signedness of the base type. The default
10526 varies depending on the operating system, so a portable program should
10527 always specify the signedness.
10530 Compiling with @option{-maltivec} adds keywords @code{__vector},
10531 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
10532 @code{bool}. When compiling ISO C, the context-sensitive substitution
10533 of the keywords @code{vector}, @code{pixel} and @code{bool} is
10534 disabled. To use them, you must include @code{<altivec.h>} instead.
10537 GCC allows using a @code{typedef} name as the type specifier for a
10541 For C, overloaded functions are implemented with macros so the following
10545 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
10548 Since @code{vec_add} is a macro, the vector constant in the example
10549 is treated as four separate arguments. Wrap the entire argument in
10550 parentheses for this to work.
10553 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
10554 Internally, GCC uses built-in functions to achieve the functionality in
10555 the aforementioned header file, but they are not supported and are
10556 subject to change without notice.
10558 The following interfaces are supported for the generic and specific
10559 AltiVec operations and the AltiVec predicates. In cases where there
10560 is a direct mapping between generic and specific operations, only the
10561 generic names are shown here, although the specific operations can also
10564 Arguments that are documented as @code{const int} require literal
10565 integral values within the range required for that operation.
10568 vector signed char vec_abs (vector signed char);
10569 vector signed short vec_abs (vector signed short);
10570 vector signed int vec_abs (vector signed int);
10571 vector float vec_abs (vector float);
10573 vector signed char vec_abss (vector signed char);
10574 vector signed short vec_abss (vector signed short);
10575 vector signed int vec_abss (vector signed int);
10577 vector signed char vec_add (vector bool char, vector signed char);
10578 vector signed char vec_add (vector signed char, vector bool char);
10579 vector signed char vec_add (vector signed char, vector signed char);
10580 vector unsigned char vec_add (vector bool char, vector unsigned char);
10581 vector unsigned char vec_add (vector unsigned char, vector bool char);
10582 vector unsigned char vec_add (vector unsigned char,
10583 vector unsigned char);
10584 vector signed short vec_add (vector bool short, vector signed short);
10585 vector signed short vec_add (vector signed short, vector bool short);
10586 vector signed short vec_add (vector signed short, vector signed short);
10587 vector unsigned short vec_add (vector bool short,
10588 vector unsigned short);
10589 vector unsigned short vec_add (vector unsigned short,
10590 vector bool short);
10591 vector unsigned short vec_add (vector unsigned short,
10592 vector unsigned short);
10593 vector signed int vec_add (vector bool int, vector signed int);
10594 vector signed int vec_add (vector signed int, vector bool int);
10595 vector signed int vec_add (vector signed int, vector signed int);
10596 vector unsigned int vec_add (vector bool int, vector unsigned int);
10597 vector unsigned int vec_add (vector unsigned int, vector bool int);
10598 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
10599 vector float vec_add (vector float, vector float);
10601 vector float vec_vaddfp (vector float, vector float);
10603 vector signed int vec_vadduwm (vector bool int, vector signed int);
10604 vector signed int vec_vadduwm (vector signed int, vector bool int);
10605 vector signed int vec_vadduwm (vector signed int, vector signed int);
10606 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
10607 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
10608 vector unsigned int vec_vadduwm (vector unsigned int,
10609 vector unsigned int);
10611 vector signed short vec_vadduhm (vector bool short,
10612 vector signed short);
10613 vector signed short vec_vadduhm (vector signed short,
10614 vector bool short);
10615 vector signed short vec_vadduhm (vector signed short,
10616 vector signed short);
10617 vector unsigned short vec_vadduhm (vector bool short,
10618 vector unsigned short);
10619 vector unsigned short vec_vadduhm (vector unsigned short,
10620 vector bool short);
10621 vector unsigned short vec_vadduhm (vector unsigned short,
10622 vector unsigned short);
10624 vector signed char vec_vaddubm (vector bool char, vector signed char);
10625 vector signed char vec_vaddubm (vector signed char, vector bool char);
10626 vector signed char vec_vaddubm (vector signed char, vector signed char);
10627 vector unsigned char vec_vaddubm (vector bool char,
10628 vector unsigned char);
10629 vector unsigned char vec_vaddubm (vector unsigned char,
10631 vector unsigned char vec_vaddubm (vector unsigned char,
10632 vector unsigned char);
10634 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
10636 vector unsigned char vec_adds (vector bool char, vector unsigned char);
10637 vector unsigned char vec_adds (vector unsigned char, vector bool char);
10638 vector unsigned char vec_adds (vector unsigned char,
10639 vector unsigned char);
10640 vector signed char vec_adds (vector bool char, vector signed char);
10641 vector signed char vec_adds (vector signed char, vector bool char);
10642 vector signed char vec_adds (vector signed char, vector signed char);
10643 vector unsigned short vec_adds (vector bool short,
10644 vector unsigned short);
10645 vector unsigned short vec_adds (vector unsigned short,
10646 vector bool short);
10647 vector unsigned short vec_adds (vector unsigned short,
10648 vector unsigned short);
10649 vector signed short vec_adds (vector bool short, vector signed short);
10650 vector signed short vec_adds (vector signed short, vector bool short);
10651 vector signed short vec_adds (vector signed short, vector signed short);
10652 vector unsigned int vec_adds (vector bool int, vector unsigned int);
10653 vector unsigned int vec_adds (vector unsigned int, vector bool int);
10654 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
10655 vector signed int vec_adds (vector bool int, vector signed int);
10656 vector signed int vec_adds (vector signed int, vector bool int);
10657 vector signed int vec_adds (vector signed int, vector signed int);
10659 vector signed int vec_vaddsws (vector bool int, vector signed int);
10660 vector signed int vec_vaddsws (vector signed int, vector bool int);
10661 vector signed int vec_vaddsws (vector signed int, vector signed int);
10663 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
10664 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
10665 vector unsigned int vec_vadduws (vector unsigned int,
10666 vector unsigned int);
10668 vector signed short vec_vaddshs (vector bool short,
10669 vector signed short);
10670 vector signed short vec_vaddshs (vector signed short,
10671 vector bool short);
10672 vector signed short vec_vaddshs (vector signed short,
10673 vector signed short);
10675 vector unsigned short vec_vadduhs (vector bool short,
10676 vector unsigned short);
10677 vector unsigned short vec_vadduhs (vector unsigned short,
10678 vector bool short);
10679 vector unsigned short vec_vadduhs (vector unsigned short,
10680 vector unsigned short);
10682 vector signed char vec_vaddsbs (vector bool char, vector signed char);
10683 vector signed char vec_vaddsbs (vector signed char, vector bool char);
10684 vector signed char vec_vaddsbs (vector signed char, vector signed char);
10686 vector unsigned char vec_vaddubs (vector bool char,
10687 vector unsigned char);
10688 vector unsigned char vec_vaddubs (vector unsigned char,
10690 vector unsigned char vec_vaddubs (vector unsigned char,
10691 vector unsigned char);
10693 vector float vec_and (vector float, vector float);
10694 vector float vec_and (vector float, vector bool int);
10695 vector float vec_and (vector bool int, vector float);
10696 vector bool int vec_and (vector bool int, vector bool int);
10697 vector signed int vec_and (vector bool int, vector signed int);
10698 vector signed int vec_and (vector signed int, vector bool int);
10699 vector signed int vec_and (vector signed int, vector signed int);
10700 vector unsigned int vec_and (vector bool int, vector unsigned int);
10701 vector unsigned int vec_and (vector unsigned int, vector bool int);
10702 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
10703 vector bool short vec_and (vector bool short, vector bool short);
10704 vector signed short vec_and (vector bool short, vector signed short);
10705 vector signed short vec_and (vector signed short, vector bool short);
10706 vector signed short vec_and (vector signed short, vector signed short);
10707 vector unsigned short vec_and (vector bool short,
10708 vector unsigned short);
10709 vector unsigned short vec_and (vector unsigned short,
10710 vector bool short);
10711 vector unsigned short vec_and (vector unsigned short,
10712 vector unsigned short);
10713 vector signed char vec_and (vector bool char, vector signed char);
10714 vector bool char vec_and (vector bool char, vector bool char);
10715 vector signed char vec_and (vector signed char, vector bool char);
10716 vector signed char vec_and (vector signed char, vector signed char);
10717 vector unsigned char vec_and (vector bool char, vector unsigned char);
10718 vector unsigned char vec_and (vector unsigned char, vector bool char);
10719 vector unsigned char vec_and (vector unsigned char,
10720 vector unsigned char);
10722 vector float vec_andc (vector float, vector float);
10723 vector float vec_andc (vector float, vector bool int);
10724 vector float vec_andc (vector bool int, vector float);
10725 vector bool int vec_andc (vector bool int, vector bool int);
10726 vector signed int vec_andc (vector bool int, vector signed int);
10727 vector signed int vec_andc (vector signed int, vector bool int);
10728 vector signed int vec_andc (vector signed int, vector signed int);
10729 vector unsigned int vec_andc (vector bool int, vector unsigned int);
10730 vector unsigned int vec_andc (vector unsigned int, vector bool int);
10731 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
10732 vector bool short vec_andc (vector bool short, vector bool short);
10733 vector signed short vec_andc (vector bool short, vector signed short);
10734 vector signed short vec_andc (vector signed short, vector bool short);
10735 vector signed short vec_andc (vector signed short, vector signed short);
10736 vector unsigned short vec_andc (vector bool short,
10737 vector unsigned short);
10738 vector unsigned short vec_andc (vector unsigned short,
10739 vector bool short);
10740 vector unsigned short vec_andc (vector unsigned short,
10741 vector unsigned short);
10742 vector signed char vec_andc (vector bool char, vector signed char);
10743 vector bool char vec_andc (vector bool char, vector bool char);
10744 vector signed char vec_andc (vector signed char, vector bool char);
10745 vector signed char vec_andc (vector signed char, vector signed char);
10746 vector unsigned char vec_andc (vector bool char, vector unsigned char);
10747 vector unsigned char vec_andc (vector unsigned char, vector bool char);
10748 vector unsigned char vec_andc (vector unsigned char,
10749 vector unsigned char);
10751 vector unsigned char vec_avg (vector unsigned char,
10752 vector unsigned char);
10753 vector signed char vec_avg (vector signed char, vector signed char);
10754 vector unsigned short vec_avg (vector unsigned short,
10755 vector unsigned short);
10756 vector signed short vec_avg (vector signed short, vector signed short);
10757 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
10758 vector signed int vec_avg (vector signed int, vector signed int);
10760 vector signed int vec_vavgsw (vector signed int, vector signed int);
10762 vector unsigned int vec_vavguw (vector unsigned int,
10763 vector unsigned int);
10765 vector signed short vec_vavgsh (vector signed short,
10766 vector signed short);
10768 vector unsigned short vec_vavguh (vector unsigned short,
10769 vector unsigned short);
10771 vector signed char vec_vavgsb (vector signed char, vector signed char);
10773 vector unsigned char vec_vavgub (vector unsigned char,
10774 vector unsigned char);
10776 vector float vec_copysign (vector float);
10778 vector float vec_ceil (vector float);
10780 vector signed int vec_cmpb (vector float, vector float);
10782 vector bool char vec_cmpeq (vector signed char, vector signed char);
10783 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
10784 vector bool short vec_cmpeq (vector signed short, vector signed short);
10785 vector bool short vec_cmpeq (vector unsigned short,
10786 vector unsigned short);
10787 vector bool int vec_cmpeq (vector signed int, vector signed int);
10788 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
10789 vector bool int vec_cmpeq (vector float, vector float);
10791 vector bool int vec_vcmpeqfp (vector float, vector float);
10793 vector bool int vec_vcmpequw (vector signed int, vector signed int);
10794 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
10796 vector bool short vec_vcmpequh (vector signed short,
10797 vector signed short);
10798 vector bool short vec_vcmpequh (vector unsigned short,
10799 vector unsigned short);
10801 vector bool char vec_vcmpequb (vector signed char, vector signed char);
10802 vector bool char vec_vcmpequb (vector unsigned char,
10803 vector unsigned char);
10805 vector bool int vec_cmpge (vector float, vector float);
10807 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
10808 vector bool char vec_cmpgt (vector signed char, vector signed char);
10809 vector bool short vec_cmpgt (vector unsigned short,
10810 vector unsigned short);
10811 vector bool short vec_cmpgt (vector signed short, vector signed short);
10812 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
10813 vector bool int vec_cmpgt (vector signed int, vector signed int);
10814 vector bool int vec_cmpgt (vector float, vector float);
10816 vector bool int vec_vcmpgtfp (vector float, vector float);
10818 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
10820 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
10822 vector bool short vec_vcmpgtsh (vector signed short,
10823 vector signed short);
10825 vector bool short vec_vcmpgtuh (vector unsigned short,
10826 vector unsigned short);
10828 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
10830 vector bool char vec_vcmpgtub (vector unsigned char,
10831 vector unsigned char);
10833 vector bool int vec_cmple (vector float, vector float);
10835 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
10836 vector bool char vec_cmplt (vector signed char, vector signed char);
10837 vector bool short vec_cmplt (vector unsigned short,
10838 vector unsigned short);
10839 vector bool short vec_cmplt (vector signed short, vector signed short);
10840 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
10841 vector bool int vec_cmplt (vector signed int, vector signed int);
10842 vector bool int vec_cmplt (vector float, vector float);
10844 vector float vec_ctf (vector unsigned int, const int);
10845 vector float vec_ctf (vector signed int, const int);
10847 vector float vec_vcfsx (vector signed int, const int);
10849 vector float vec_vcfux (vector unsigned int, const int);
10851 vector signed int vec_cts (vector float, const int);
10853 vector unsigned int vec_ctu (vector float, const int);
10855 void vec_dss (const int);
10857 void vec_dssall (void);
10859 void vec_dst (const vector unsigned char *, int, const int);
10860 void vec_dst (const vector signed char *, int, const int);
10861 void vec_dst (const vector bool char *, int, const int);
10862 void vec_dst (const vector unsigned short *, int, const int);
10863 void vec_dst (const vector signed short *, int, const int);
10864 void vec_dst (const vector bool short *, int, const int);
10865 void vec_dst (const vector pixel *, int, const int);
10866 void vec_dst (const vector unsigned int *, int, const int);
10867 void vec_dst (const vector signed int *, int, const int);
10868 void vec_dst (const vector bool int *, int, const int);
10869 void vec_dst (const vector float *, int, const int);
10870 void vec_dst (const unsigned char *, int, const int);
10871 void vec_dst (const signed char *, int, const int);
10872 void vec_dst (const unsigned short *, int, const int);
10873 void vec_dst (const short *, int, const int);
10874 void vec_dst (const unsigned int *, int, const int);
10875 void vec_dst (const int *, int, const int);
10876 void vec_dst (const unsigned long *, int, const int);
10877 void vec_dst (const long *, int, const int);
10878 void vec_dst (const float *, int, const int);
10880 void vec_dstst (const vector unsigned char *, int, const int);
10881 void vec_dstst (const vector signed char *, int, const int);
10882 void vec_dstst (const vector bool char *, int, const int);
10883 void vec_dstst (const vector unsigned short *, int, const int);
10884 void vec_dstst (const vector signed short *, int, const int);
10885 void vec_dstst (const vector bool short *, int, const int);
10886 void vec_dstst (const vector pixel *, int, const int);
10887 void vec_dstst (const vector unsigned int *, int, const int);
10888 void vec_dstst (const vector signed int *, int, const int);
10889 void vec_dstst (const vector bool int *, int, const int);
10890 void vec_dstst (const vector float *, int, const int);
10891 void vec_dstst (const unsigned char *, int, const int);
10892 void vec_dstst (const signed char *, int, const int);
10893 void vec_dstst (const unsigned short *, int, const int);
10894 void vec_dstst (const short *, int, const int);
10895 void vec_dstst (const unsigned int *, int, const int);
10896 void vec_dstst (const int *, int, const int);
10897 void vec_dstst (const unsigned long *, int, const int);
10898 void vec_dstst (const long *, int, const int);
10899 void vec_dstst (const float *, int, const int);
10901 void vec_dststt (const vector unsigned char *, int, const int);
10902 void vec_dststt (const vector signed char *, int, const int);
10903 void vec_dststt (const vector bool char *, int, const int);
10904 void vec_dststt (const vector unsigned short *, int, const int);
10905 void vec_dststt (const vector signed short *, int, const int);
10906 void vec_dststt (const vector bool short *, int, const int);
10907 void vec_dststt (const vector pixel *, int, const int);
10908 void vec_dststt (const vector unsigned int *, int, const int);
10909 void vec_dststt (const vector signed int *, int, const int);
10910 void vec_dststt (const vector bool int *, int, const int);
10911 void vec_dststt (const vector float *, int, const int);
10912 void vec_dststt (const unsigned char *, int, const int);
10913 void vec_dststt (const signed char *, int, const int);
10914 void vec_dststt (const unsigned short *, int, const int);
10915 void vec_dststt (const short *, int, const int);
10916 void vec_dststt (const unsigned int *, int, const int);
10917 void vec_dststt (const int *, int, const int);
10918 void vec_dststt (const unsigned long *, int, const int);
10919 void vec_dststt (const long *, int, const int);
10920 void vec_dststt (const float *, int, const int);
10922 void vec_dstt (const vector unsigned char *, int, const int);
10923 void vec_dstt (const vector signed char *, int, const int);
10924 void vec_dstt (const vector bool char *, int, const int);
10925 void vec_dstt (const vector unsigned short *, int, const int);
10926 void vec_dstt (const vector signed short *, int, const int);
10927 void vec_dstt (const vector bool short *, int, const int);
10928 void vec_dstt (const vector pixel *, int, const int);
10929 void vec_dstt (const vector unsigned int *, int, const int);
10930 void vec_dstt (const vector signed int *, int, const int);
10931 void vec_dstt (const vector bool int *, int, const int);
10932 void vec_dstt (const vector float *, int, const int);
10933 void vec_dstt (const unsigned char *, int, const int);
10934 void vec_dstt (const signed char *, int, const int);
10935 void vec_dstt (const unsigned short *, int, const int);
10936 void vec_dstt (const short *, int, const int);
10937 void vec_dstt (const unsigned int *, int, const int);
10938 void vec_dstt (const int *, int, const int);
10939 void vec_dstt (const unsigned long *, int, const int);
10940 void vec_dstt (const long *, int, const int);
10941 void vec_dstt (const float *, int, const int);
10943 vector float vec_expte (vector float);
10945 vector float vec_floor (vector float);
10947 vector float vec_ld (int, const vector float *);
10948 vector float vec_ld (int, const float *);
10949 vector bool int vec_ld (int, const vector bool int *);
10950 vector signed int vec_ld (int, const vector signed int *);
10951 vector signed int vec_ld (int, const int *);
10952 vector signed int vec_ld (int, const long *);
10953 vector unsigned int vec_ld (int, const vector unsigned int *);
10954 vector unsigned int vec_ld (int, const unsigned int *);
10955 vector unsigned int vec_ld (int, const unsigned long *);
10956 vector bool short vec_ld (int, const vector bool short *);
10957 vector pixel vec_ld (int, const vector pixel *);
10958 vector signed short vec_ld (int, const vector signed short *);
10959 vector signed short vec_ld (int, const short *);
10960 vector unsigned short vec_ld (int, const vector unsigned short *);
10961 vector unsigned short vec_ld (int, const unsigned short *);
10962 vector bool char vec_ld (int, const vector bool char *);
10963 vector signed char vec_ld (int, const vector signed char *);
10964 vector signed char vec_ld (int, const signed char *);
10965 vector unsigned char vec_ld (int, const vector unsigned char *);
10966 vector unsigned char vec_ld (int, const unsigned char *);
10968 vector signed char vec_lde (int, const signed char *);
10969 vector unsigned char vec_lde (int, const unsigned char *);
10970 vector signed short vec_lde (int, const short *);
10971 vector unsigned short vec_lde (int, const unsigned short *);
10972 vector float vec_lde (int, const float *);
10973 vector signed int vec_lde (int, const int *);
10974 vector unsigned int vec_lde (int, const unsigned int *);
10975 vector signed int vec_lde (int, const long *);
10976 vector unsigned int vec_lde (int, const unsigned long *);
10978 vector float vec_lvewx (int, float *);
10979 vector signed int vec_lvewx (int, int *);
10980 vector unsigned int vec_lvewx (int, unsigned int *);
10981 vector signed int vec_lvewx (int, long *);
10982 vector unsigned int vec_lvewx (int, unsigned long *);
10984 vector signed short vec_lvehx (int, short *);
10985 vector unsigned short vec_lvehx (int, unsigned short *);
10987 vector signed char vec_lvebx (int, char *);
10988 vector unsigned char vec_lvebx (int, unsigned char *);
10990 vector float vec_ldl (int, const vector float *);
10991 vector float vec_ldl (int, const float *);
10992 vector bool int vec_ldl (int, const vector bool int *);
10993 vector signed int vec_ldl (int, const vector signed int *);
10994 vector signed int vec_ldl (int, const int *);
10995 vector signed int vec_ldl (int, const long *);
10996 vector unsigned int vec_ldl (int, const vector unsigned int *);
10997 vector unsigned int vec_ldl (int, const unsigned int *);
10998 vector unsigned int vec_ldl (int, const unsigned long *);
10999 vector bool short vec_ldl (int, const vector bool short *);
11000 vector pixel vec_ldl (int, const vector pixel *);
11001 vector signed short vec_ldl (int, const vector signed short *);
11002 vector signed short vec_ldl (int, const short *);
11003 vector unsigned short vec_ldl (int, const vector unsigned short *);
11004 vector unsigned short vec_ldl (int, const unsigned short *);
11005 vector bool char vec_ldl (int, const vector bool char *);
11006 vector signed char vec_ldl (int, const vector signed char *);
11007 vector signed char vec_ldl (int, const signed char *);
11008 vector unsigned char vec_ldl (int, const vector unsigned char *);
11009 vector unsigned char vec_ldl (int, const unsigned char *);
11011 vector float vec_loge (vector float);
11013 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
11014 vector unsigned char vec_lvsl (int, const volatile signed char *);
11015 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
11016 vector unsigned char vec_lvsl (int, const volatile short *);
11017 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
11018 vector unsigned char vec_lvsl (int, const volatile int *);
11019 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
11020 vector unsigned char vec_lvsl (int, const volatile long *);
11021 vector unsigned char vec_lvsl (int, const volatile float *);
11023 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
11024 vector unsigned char vec_lvsr (int, const volatile signed char *);
11025 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
11026 vector unsigned char vec_lvsr (int, const volatile short *);
11027 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
11028 vector unsigned char vec_lvsr (int, const volatile int *);
11029 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
11030 vector unsigned char vec_lvsr (int, const volatile long *);
11031 vector unsigned char vec_lvsr (int, const volatile float *);
11033 vector float vec_madd (vector float, vector float, vector float);
11035 vector signed short vec_madds (vector signed short,
11036 vector signed short,
11037 vector signed short);
11039 vector unsigned char vec_max (vector bool char, vector unsigned char);
11040 vector unsigned char vec_max (vector unsigned char, vector bool char);
11041 vector unsigned char vec_max (vector unsigned char,
11042 vector unsigned char);
11043 vector signed char vec_max (vector bool char, vector signed char);
11044 vector signed char vec_max (vector signed char, vector bool char);
11045 vector signed char vec_max (vector signed char, vector signed char);
11046 vector unsigned short vec_max (vector bool short,
11047 vector unsigned short);
11048 vector unsigned short vec_max (vector unsigned short,
11049 vector bool short);
11050 vector unsigned short vec_max (vector unsigned short,
11051 vector unsigned short);
11052 vector signed short vec_max (vector bool short, vector signed short);
11053 vector signed short vec_max (vector signed short, vector bool short);
11054 vector signed short vec_max (vector signed short, vector signed short);
11055 vector unsigned int vec_max (vector bool int, vector unsigned int);
11056 vector unsigned int vec_max (vector unsigned int, vector bool int);
11057 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
11058 vector signed int vec_max (vector bool int, vector signed int);
11059 vector signed int vec_max (vector signed int, vector bool int);
11060 vector signed int vec_max (vector signed int, vector signed int);
11061 vector float vec_max (vector float, vector float);
11063 vector float vec_vmaxfp (vector float, vector float);
11065 vector signed int vec_vmaxsw (vector bool int, vector signed int);
11066 vector signed int vec_vmaxsw (vector signed int, vector bool int);
11067 vector signed int vec_vmaxsw (vector signed int, vector signed int);
11069 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
11070 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
11071 vector unsigned int vec_vmaxuw (vector unsigned int,
11072 vector unsigned int);
11074 vector signed short vec_vmaxsh (vector bool short, vector signed short);
11075 vector signed short vec_vmaxsh (vector signed short, vector bool short);
11076 vector signed short vec_vmaxsh (vector signed short,
11077 vector signed short);
11079 vector unsigned short vec_vmaxuh (vector bool short,
11080 vector unsigned short);
11081 vector unsigned short vec_vmaxuh (vector unsigned short,
11082 vector bool short);
11083 vector unsigned short vec_vmaxuh (vector unsigned short,
11084 vector unsigned short);
11086 vector signed char vec_vmaxsb (vector bool char, vector signed char);
11087 vector signed char vec_vmaxsb (vector signed char, vector bool char);
11088 vector signed char vec_vmaxsb (vector signed char, vector signed char);
11090 vector unsigned char vec_vmaxub (vector bool char,
11091 vector unsigned char);
11092 vector unsigned char vec_vmaxub (vector unsigned char,
11094 vector unsigned char vec_vmaxub (vector unsigned char,
11095 vector unsigned char);
11097 vector bool char vec_mergeh (vector bool char, vector bool char);
11098 vector signed char vec_mergeh (vector signed char, vector signed char);
11099 vector unsigned char vec_mergeh (vector unsigned char,
11100 vector unsigned char);
11101 vector bool short vec_mergeh (vector bool short, vector bool short);
11102 vector pixel vec_mergeh (vector pixel, vector pixel);
11103 vector signed short vec_mergeh (vector signed short,
11104 vector signed short);
11105 vector unsigned short vec_mergeh (vector unsigned short,
11106 vector unsigned short);
11107 vector float vec_mergeh (vector float, vector float);
11108 vector bool int vec_mergeh (vector bool int, vector bool int);
11109 vector signed int vec_mergeh (vector signed int, vector signed int);
11110 vector unsigned int vec_mergeh (vector unsigned int,
11111 vector unsigned int);
11113 vector float vec_vmrghw (vector float, vector float);
11114 vector bool int vec_vmrghw (vector bool int, vector bool int);
11115 vector signed int vec_vmrghw (vector signed int, vector signed int);
11116 vector unsigned int vec_vmrghw (vector unsigned int,
11117 vector unsigned int);
11119 vector bool short vec_vmrghh (vector bool short, vector bool short);
11120 vector signed short vec_vmrghh (vector signed short,
11121 vector signed short);
11122 vector unsigned short vec_vmrghh (vector unsigned short,
11123 vector unsigned short);
11124 vector pixel vec_vmrghh (vector pixel, vector pixel);
11126 vector bool char vec_vmrghb (vector bool char, vector bool char);
11127 vector signed char vec_vmrghb (vector signed char, vector signed char);
11128 vector unsigned char vec_vmrghb (vector unsigned char,
11129 vector unsigned char);
11131 vector bool char vec_mergel (vector bool char, vector bool char);
11132 vector signed char vec_mergel (vector signed char, vector signed char);
11133 vector unsigned char vec_mergel (vector unsigned char,
11134 vector unsigned char);
11135 vector bool short vec_mergel (vector bool short, vector bool short);
11136 vector pixel vec_mergel (vector pixel, vector pixel);
11137 vector signed short vec_mergel (vector signed short,
11138 vector signed short);
11139 vector unsigned short vec_mergel (vector unsigned short,
11140 vector unsigned short);
11141 vector float vec_mergel (vector float, vector float);
11142 vector bool int vec_mergel (vector bool int, vector bool int);
11143 vector signed int vec_mergel (vector signed int, vector signed int);
11144 vector unsigned int vec_mergel (vector unsigned int,
11145 vector unsigned int);
11147 vector float vec_vmrglw (vector float, vector float);
11148 vector signed int vec_vmrglw (vector signed int, vector signed int);
11149 vector unsigned int vec_vmrglw (vector unsigned int,
11150 vector unsigned int);
11151 vector bool int vec_vmrglw (vector bool int, vector bool int);
11153 vector bool short vec_vmrglh (vector bool short, vector bool short);
11154 vector signed short vec_vmrglh (vector signed short,
11155 vector signed short);
11156 vector unsigned short vec_vmrglh (vector unsigned short,
11157 vector unsigned short);
11158 vector pixel vec_vmrglh (vector pixel, vector pixel);
11160 vector bool char vec_vmrglb (vector bool char, vector bool char);
11161 vector signed char vec_vmrglb (vector signed char, vector signed char);
11162 vector unsigned char vec_vmrglb (vector unsigned char,
11163 vector unsigned char);
11165 vector unsigned short vec_mfvscr (void);
11167 vector unsigned char vec_min (vector bool char, vector unsigned char);
11168 vector unsigned char vec_min (vector unsigned char, vector bool char);
11169 vector unsigned char vec_min (vector unsigned char,
11170 vector unsigned char);
11171 vector signed char vec_min (vector bool char, vector signed char);
11172 vector signed char vec_min (vector signed char, vector bool char);
11173 vector signed char vec_min (vector signed char, vector signed char);
11174 vector unsigned short vec_min (vector bool short,
11175 vector unsigned short);
11176 vector unsigned short vec_min (vector unsigned short,
11177 vector bool short);
11178 vector unsigned short vec_min (vector unsigned short,
11179 vector unsigned short);
11180 vector signed short vec_min (vector bool short, vector signed short);
11181 vector signed short vec_min (vector signed short, vector bool short);
11182 vector signed short vec_min (vector signed short, vector signed short);
11183 vector unsigned int vec_min (vector bool int, vector unsigned int);
11184 vector unsigned int vec_min (vector unsigned int, vector bool int);
11185 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
11186 vector signed int vec_min (vector bool int, vector signed int);
11187 vector signed int vec_min (vector signed int, vector bool int);
11188 vector signed int vec_min (vector signed int, vector signed int);
11189 vector float vec_min (vector float, vector float);
11191 vector float vec_vminfp (vector float, vector float);
11193 vector signed int vec_vminsw (vector bool int, vector signed int);
11194 vector signed int vec_vminsw (vector signed int, vector bool int);
11195 vector signed int vec_vminsw (vector signed int, vector signed int);
11197 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
11198 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
11199 vector unsigned int vec_vminuw (vector unsigned int,
11200 vector unsigned int);
11202 vector signed short vec_vminsh (vector bool short, vector signed short);
11203 vector signed short vec_vminsh (vector signed short, vector bool short);
11204 vector signed short vec_vminsh (vector signed short,
11205 vector signed short);
11207 vector unsigned short vec_vminuh (vector bool short,
11208 vector unsigned short);
11209 vector unsigned short vec_vminuh (vector unsigned short,
11210 vector bool short);
11211 vector unsigned short vec_vminuh (vector unsigned short,
11212 vector unsigned short);
11214 vector signed char vec_vminsb (vector bool char, vector signed char);
11215 vector signed char vec_vminsb (vector signed char, vector bool char);
11216 vector signed char vec_vminsb (vector signed char, vector signed char);
11218 vector unsigned char vec_vminub (vector bool char,
11219 vector unsigned char);
11220 vector unsigned char vec_vminub (vector unsigned char,
11222 vector unsigned char vec_vminub (vector unsigned char,
11223 vector unsigned char);
11225 vector signed short vec_mladd (vector signed short,
11226 vector signed short,
11227 vector signed short);
11228 vector signed short vec_mladd (vector signed short,
11229 vector unsigned short,
11230 vector unsigned short);
11231 vector signed short vec_mladd (vector unsigned short,
11232 vector signed short,
11233 vector signed short);
11234 vector unsigned short vec_mladd (vector unsigned short,
11235 vector unsigned short,
11236 vector unsigned short);
11238 vector signed short vec_mradds (vector signed short,
11239 vector signed short,
11240 vector signed short);
11242 vector unsigned int vec_msum (vector unsigned char,
11243 vector unsigned char,
11244 vector unsigned int);
11245 vector signed int vec_msum (vector signed char,
11246 vector unsigned char,
11247 vector signed int);
11248 vector unsigned int vec_msum (vector unsigned short,
11249 vector unsigned short,
11250 vector unsigned int);
11251 vector signed int vec_msum (vector signed short,
11252 vector signed short,
11253 vector signed int);
11255 vector signed int vec_vmsumshm (vector signed short,
11256 vector signed short,
11257 vector signed int);
11259 vector unsigned int vec_vmsumuhm (vector unsigned short,
11260 vector unsigned short,
11261 vector unsigned int);
11263 vector signed int vec_vmsummbm (vector signed char,
11264 vector unsigned char,
11265 vector signed int);
11267 vector unsigned int vec_vmsumubm (vector unsigned char,
11268 vector unsigned char,
11269 vector unsigned int);
11271 vector unsigned int vec_msums (vector unsigned short,
11272 vector unsigned short,
11273 vector unsigned int);
11274 vector signed int vec_msums (vector signed short,
11275 vector signed short,
11276 vector signed int);
11278 vector signed int vec_vmsumshs (vector signed short,
11279 vector signed short,
11280 vector signed int);
11282 vector unsigned int vec_vmsumuhs (vector unsigned short,
11283 vector unsigned short,
11284 vector unsigned int);
11286 void vec_mtvscr (vector signed int);
11287 void vec_mtvscr (vector unsigned int);
11288 void vec_mtvscr (vector bool int);
11289 void vec_mtvscr (vector signed short);
11290 void vec_mtvscr (vector unsigned short);
11291 void vec_mtvscr (vector bool short);
11292 void vec_mtvscr (vector pixel);
11293 void vec_mtvscr (vector signed char);
11294 void vec_mtvscr (vector unsigned char);
11295 void vec_mtvscr (vector bool char);
11297 vector unsigned short vec_mule (vector unsigned char,
11298 vector unsigned char);
11299 vector signed short vec_mule (vector signed char,
11300 vector signed char);
11301 vector unsigned int vec_mule (vector unsigned short,
11302 vector unsigned short);
11303 vector signed int vec_mule (vector signed short, vector signed short);
11305 vector signed int vec_vmulesh (vector signed short,
11306 vector signed short);
11308 vector unsigned int vec_vmuleuh (vector unsigned short,
11309 vector unsigned short);
11311 vector signed short vec_vmulesb (vector signed char,
11312 vector signed char);
11314 vector unsigned short vec_vmuleub (vector unsigned char,
11315 vector unsigned char);
11317 vector unsigned short vec_mulo (vector unsigned char,
11318 vector unsigned char);
11319 vector signed short vec_mulo (vector signed char, vector signed char);
11320 vector unsigned int vec_mulo (vector unsigned short,
11321 vector unsigned short);
11322 vector signed int vec_mulo (vector signed short, vector signed short);
11324 vector signed int vec_vmulosh (vector signed short,
11325 vector signed short);
11327 vector unsigned int vec_vmulouh (vector unsigned short,
11328 vector unsigned short);
11330 vector signed short vec_vmulosb (vector signed char,
11331 vector signed char);
11333 vector unsigned short vec_vmuloub (vector unsigned char,
11334 vector unsigned char);
11336 vector float vec_nmsub (vector float, vector float, vector float);
11338 vector float vec_nor (vector float, vector float);
11339 vector signed int vec_nor (vector signed int, vector signed int);
11340 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
11341 vector bool int vec_nor (vector bool int, vector bool int);
11342 vector signed short vec_nor (vector signed short, vector signed short);
11343 vector unsigned short vec_nor (vector unsigned short,
11344 vector unsigned short);
11345 vector bool short vec_nor (vector bool short, vector bool short);
11346 vector signed char vec_nor (vector signed char, vector signed char);
11347 vector unsigned char vec_nor (vector unsigned char,
11348 vector unsigned char);
11349 vector bool char vec_nor (vector bool char, vector bool char);
11351 vector float vec_or (vector float, vector float);
11352 vector float vec_or (vector float, vector bool int);
11353 vector float vec_or (vector bool int, vector float);
11354 vector bool int vec_or (vector bool int, vector bool int);
11355 vector signed int vec_or (vector bool int, vector signed int);
11356 vector signed int vec_or (vector signed int, vector bool int);
11357 vector signed int vec_or (vector signed int, vector signed int);
11358 vector unsigned int vec_or (vector bool int, vector unsigned int);
11359 vector unsigned int vec_or (vector unsigned int, vector bool int);
11360 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
11361 vector bool short vec_or (vector bool short, vector bool short);
11362 vector signed short vec_or (vector bool short, vector signed short);
11363 vector signed short vec_or (vector signed short, vector bool short);
11364 vector signed short vec_or (vector signed short, vector signed short);
11365 vector unsigned short vec_or (vector bool short, vector unsigned short);
11366 vector unsigned short vec_or (vector unsigned short, vector bool short);
11367 vector unsigned short vec_or (vector unsigned short,
11368 vector unsigned short);
11369 vector signed char vec_or (vector bool char, vector signed char);
11370 vector bool char vec_or (vector bool char, vector bool char);
11371 vector signed char vec_or (vector signed char, vector bool char);
11372 vector signed char vec_or (vector signed char, vector signed char);
11373 vector unsigned char vec_or (vector bool char, vector unsigned char);
11374 vector unsigned char vec_or (vector unsigned char, vector bool char);
11375 vector unsigned char vec_or (vector unsigned char,
11376 vector unsigned char);
11378 vector signed char vec_pack (vector signed short, vector signed short);
11379 vector unsigned char vec_pack (vector unsigned short,
11380 vector unsigned short);
11381 vector bool char vec_pack (vector bool short, vector bool short);
11382 vector signed short vec_pack (vector signed int, vector signed int);
11383 vector unsigned short vec_pack (vector unsigned int,
11384 vector unsigned int);
11385 vector bool short vec_pack (vector bool int, vector bool int);
11387 vector bool short vec_vpkuwum (vector bool int, vector bool int);
11388 vector signed short vec_vpkuwum (vector signed int, vector signed int);
11389 vector unsigned short vec_vpkuwum (vector unsigned int,
11390 vector unsigned int);
11392 vector bool char vec_vpkuhum (vector bool short, vector bool short);
11393 vector signed char vec_vpkuhum (vector signed short,
11394 vector signed short);
11395 vector unsigned char vec_vpkuhum (vector unsigned short,
11396 vector unsigned short);
11398 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
11400 vector unsigned char vec_packs (vector unsigned short,
11401 vector unsigned short);
11402 vector signed char vec_packs (vector signed short, vector signed short);
11403 vector unsigned short vec_packs (vector unsigned int,
11404 vector unsigned int);
11405 vector signed short vec_packs (vector signed int, vector signed int);
11407 vector signed short vec_vpkswss (vector signed int, vector signed int);
11409 vector unsigned short vec_vpkuwus (vector unsigned int,
11410 vector unsigned int);
11412 vector signed char vec_vpkshss (vector signed short,
11413 vector signed short);
11415 vector unsigned char vec_vpkuhus (vector unsigned short,
11416 vector unsigned short);
11418 vector unsigned char vec_packsu (vector unsigned short,
11419 vector unsigned short);
11420 vector unsigned char vec_packsu (vector signed short,
11421 vector signed short);
11422 vector unsigned short vec_packsu (vector unsigned int,
11423 vector unsigned int);
11424 vector unsigned short vec_packsu (vector signed int, vector signed int);
11426 vector unsigned short vec_vpkswus (vector signed int,
11427 vector signed int);
11429 vector unsigned char vec_vpkshus (vector signed short,
11430 vector signed short);
11432 vector float vec_perm (vector float,
11434 vector unsigned char);
11435 vector signed int vec_perm (vector signed int,
11437 vector unsigned char);
11438 vector unsigned int vec_perm (vector unsigned int,
11439 vector unsigned int,
11440 vector unsigned char);
11441 vector bool int vec_perm (vector bool int,
11443 vector unsigned char);
11444 vector signed short vec_perm (vector signed short,
11445 vector signed short,
11446 vector unsigned char);
11447 vector unsigned short vec_perm (vector unsigned short,
11448 vector unsigned short,
11449 vector unsigned char);
11450 vector bool short vec_perm (vector bool short,
11452 vector unsigned char);
11453 vector pixel vec_perm (vector pixel,
11455 vector unsigned char);
11456 vector signed char vec_perm (vector signed char,
11457 vector signed char,
11458 vector unsigned char);
11459 vector unsigned char vec_perm (vector unsigned char,
11460 vector unsigned char,
11461 vector unsigned char);
11462 vector bool char vec_perm (vector bool char,
11464 vector unsigned char);
11466 vector float vec_re (vector float);
11468 vector signed char vec_rl (vector signed char,
11469 vector unsigned char);
11470 vector unsigned char vec_rl (vector unsigned char,
11471 vector unsigned char);
11472 vector signed short vec_rl (vector signed short, vector unsigned short);
11473 vector unsigned short vec_rl (vector unsigned short,
11474 vector unsigned short);
11475 vector signed int vec_rl (vector signed int, vector unsigned int);
11476 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
11478 vector signed int vec_vrlw (vector signed int, vector unsigned int);
11479 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
11481 vector signed short vec_vrlh (vector signed short,
11482 vector unsigned short);
11483 vector unsigned short vec_vrlh (vector unsigned short,
11484 vector unsigned short);
11486 vector signed char vec_vrlb (vector signed char, vector unsigned char);
11487 vector unsigned char vec_vrlb (vector unsigned char,
11488 vector unsigned char);
11490 vector float vec_round (vector float);
11492 vector float vec_recip (vector float, vector float);
11494 vector float vec_rsqrt (vector float);
11496 vector float vec_rsqrte (vector float);
11498 vector float vec_sel (vector float, vector float, vector bool int);
11499 vector float vec_sel (vector float, vector float, vector unsigned int);
11500 vector signed int vec_sel (vector signed int,
11503 vector signed int vec_sel (vector signed int,
11505 vector unsigned int);
11506 vector unsigned int vec_sel (vector unsigned int,
11507 vector unsigned int,
11509 vector unsigned int vec_sel (vector unsigned int,
11510 vector unsigned int,
11511 vector unsigned int);
11512 vector bool int vec_sel (vector bool int,
11515 vector bool int vec_sel (vector bool int,
11517 vector unsigned int);
11518 vector signed short vec_sel (vector signed short,
11519 vector signed short,
11520 vector bool short);
11521 vector signed short vec_sel (vector signed short,
11522 vector signed short,
11523 vector unsigned short);
11524 vector unsigned short vec_sel (vector unsigned short,
11525 vector unsigned short,
11526 vector bool short);
11527 vector unsigned short vec_sel (vector unsigned short,
11528 vector unsigned short,
11529 vector unsigned short);
11530 vector bool short vec_sel (vector bool short,
11532 vector bool short);
11533 vector bool short vec_sel (vector bool short,
11535 vector unsigned short);
11536 vector signed char vec_sel (vector signed char,
11537 vector signed char,
11539 vector signed char vec_sel (vector signed char,
11540 vector signed char,
11541 vector unsigned char);
11542 vector unsigned char vec_sel (vector unsigned char,
11543 vector unsigned char,
11545 vector unsigned char vec_sel (vector unsigned char,
11546 vector unsigned char,
11547 vector unsigned char);
11548 vector bool char vec_sel (vector bool char,
11551 vector bool char vec_sel (vector bool char,
11553 vector unsigned char);
11555 vector signed char vec_sl (vector signed char,
11556 vector unsigned char);
11557 vector unsigned char vec_sl (vector unsigned char,
11558 vector unsigned char);
11559 vector signed short vec_sl (vector signed short, vector unsigned short);
11560 vector unsigned short vec_sl (vector unsigned short,
11561 vector unsigned short);
11562 vector signed int vec_sl (vector signed int, vector unsigned int);
11563 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
11565 vector signed int vec_vslw (vector signed int, vector unsigned int);
11566 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
11568 vector signed short vec_vslh (vector signed short,
11569 vector unsigned short);
11570 vector unsigned short vec_vslh (vector unsigned short,
11571 vector unsigned short);
11573 vector signed char vec_vslb (vector signed char, vector unsigned char);
11574 vector unsigned char vec_vslb (vector unsigned char,
11575 vector unsigned char);
11577 vector float vec_sld (vector float, vector float, const int);
11578 vector signed int vec_sld (vector signed int,
11581 vector unsigned int vec_sld (vector unsigned int,
11582 vector unsigned int,
11584 vector bool int vec_sld (vector bool int,
11587 vector signed short vec_sld (vector signed short,
11588 vector signed short,
11590 vector unsigned short vec_sld (vector unsigned short,
11591 vector unsigned short,
11593 vector bool short vec_sld (vector bool short,
11596 vector pixel vec_sld (vector pixel,
11599 vector signed char vec_sld (vector signed char,
11600 vector signed char,
11602 vector unsigned char vec_sld (vector unsigned char,
11603 vector unsigned char,
11605 vector bool char vec_sld (vector bool char,
11609 vector signed int vec_sll (vector signed int,
11610 vector unsigned int);
11611 vector signed int vec_sll (vector signed int,
11612 vector unsigned short);
11613 vector signed int vec_sll (vector signed int,
11614 vector unsigned char);
11615 vector unsigned int vec_sll (vector unsigned int,
11616 vector unsigned int);
11617 vector unsigned int vec_sll (vector unsigned int,
11618 vector unsigned short);
11619 vector unsigned int vec_sll (vector unsigned int,
11620 vector unsigned char);
11621 vector bool int vec_sll (vector bool int,
11622 vector unsigned int);
11623 vector bool int vec_sll (vector bool int,
11624 vector unsigned short);
11625 vector bool int vec_sll (vector bool int,
11626 vector unsigned char);
11627 vector signed short vec_sll (vector signed short,
11628 vector unsigned int);
11629 vector signed short vec_sll (vector signed short,
11630 vector unsigned short);
11631 vector signed short vec_sll (vector signed short,
11632 vector unsigned char);
11633 vector unsigned short vec_sll (vector unsigned short,
11634 vector unsigned int);
11635 vector unsigned short vec_sll (vector unsigned short,
11636 vector unsigned short);
11637 vector unsigned short vec_sll (vector unsigned short,
11638 vector unsigned char);
11639 vector bool short vec_sll (vector bool short, vector unsigned int);
11640 vector bool short vec_sll (vector bool short, vector unsigned short);
11641 vector bool short vec_sll (vector bool short, vector unsigned char);
11642 vector pixel vec_sll (vector pixel, vector unsigned int);
11643 vector pixel vec_sll (vector pixel, vector unsigned short);
11644 vector pixel vec_sll (vector pixel, vector unsigned char);
11645 vector signed char vec_sll (vector signed char, vector unsigned int);
11646 vector signed char vec_sll (vector signed char, vector unsigned short);
11647 vector signed char vec_sll (vector signed char, vector unsigned char);
11648 vector unsigned char vec_sll (vector unsigned char,
11649 vector unsigned int);
11650 vector unsigned char vec_sll (vector unsigned char,
11651 vector unsigned short);
11652 vector unsigned char vec_sll (vector unsigned char,
11653 vector unsigned char);
11654 vector bool char vec_sll (vector bool char, vector unsigned int);
11655 vector bool char vec_sll (vector bool char, vector unsigned short);
11656 vector bool char vec_sll (vector bool char, vector unsigned char);
11658 vector float vec_slo (vector float, vector signed char);
11659 vector float vec_slo (vector float, vector unsigned char);
11660 vector signed int vec_slo (vector signed int, vector signed char);
11661 vector signed int vec_slo (vector signed int, vector unsigned char);
11662 vector unsigned int vec_slo (vector unsigned int, vector signed char);
11663 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
11664 vector signed short vec_slo (vector signed short, vector signed char);
11665 vector signed short vec_slo (vector signed short, vector unsigned char);
11666 vector unsigned short vec_slo (vector unsigned short,
11667 vector signed char);
11668 vector unsigned short vec_slo (vector unsigned short,
11669 vector unsigned char);
11670 vector pixel vec_slo (vector pixel, vector signed char);
11671 vector pixel vec_slo (vector pixel, vector unsigned char);
11672 vector signed char vec_slo (vector signed char, vector signed char);
11673 vector signed char vec_slo (vector signed char, vector unsigned char);
11674 vector unsigned char vec_slo (vector unsigned char, vector signed char);
11675 vector unsigned char vec_slo (vector unsigned char,
11676 vector unsigned char);
11678 vector signed char vec_splat (vector signed char, const int);
11679 vector unsigned char vec_splat (vector unsigned char, const int);
11680 vector bool char vec_splat (vector bool char, const int);
11681 vector signed short vec_splat (vector signed short, const int);
11682 vector unsigned short vec_splat (vector unsigned short, const int);
11683 vector bool short vec_splat (vector bool short, const int);
11684 vector pixel vec_splat (vector pixel, const int);
11685 vector float vec_splat (vector float, const int);
11686 vector signed int vec_splat (vector signed int, const int);
11687 vector unsigned int vec_splat (vector unsigned int, const int);
11688 vector bool int vec_splat (vector bool int, const int);
11690 vector float vec_vspltw (vector float, const int);
11691 vector signed int vec_vspltw (vector signed int, const int);
11692 vector unsigned int vec_vspltw (vector unsigned int, const int);
11693 vector bool int vec_vspltw (vector bool int, const int);
11695 vector bool short vec_vsplth (vector bool short, const int);
11696 vector signed short vec_vsplth (vector signed short, const int);
11697 vector unsigned short vec_vsplth (vector unsigned short, const int);
11698 vector pixel vec_vsplth (vector pixel, const int);
11700 vector signed char vec_vspltb (vector signed char, const int);
11701 vector unsigned char vec_vspltb (vector unsigned char, const int);
11702 vector bool char vec_vspltb (vector bool char, const int);
11704 vector signed char vec_splat_s8 (const int);
11706 vector signed short vec_splat_s16 (const int);
11708 vector signed int vec_splat_s32 (const int);
11710 vector unsigned char vec_splat_u8 (const int);
11712 vector unsigned short vec_splat_u16 (const int);
11714 vector unsigned int vec_splat_u32 (const int);
11716 vector signed char vec_sr (vector signed char, vector unsigned char);
11717 vector unsigned char vec_sr (vector unsigned char,
11718 vector unsigned char);
11719 vector signed short vec_sr (vector signed short,
11720 vector unsigned short);
11721 vector unsigned short vec_sr (vector unsigned short,
11722 vector unsigned short);
11723 vector signed int vec_sr (vector signed int, vector unsigned int);
11724 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
11726 vector signed int vec_vsrw (vector signed int, vector unsigned int);
11727 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
11729 vector signed short vec_vsrh (vector signed short,
11730 vector unsigned short);
11731 vector unsigned short vec_vsrh (vector unsigned short,
11732 vector unsigned short);
11734 vector signed char vec_vsrb (vector signed char, vector unsigned char);
11735 vector unsigned char vec_vsrb (vector unsigned char,
11736 vector unsigned char);
11738 vector signed char vec_sra (vector signed char, vector unsigned char);
11739 vector unsigned char vec_sra (vector unsigned char,
11740 vector unsigned char);
11741 vector signed short vec_sra (vector signed short,
11742 vector unsigned short);
11743 vector unsigned short vec_sra (vector unsigned short,
11744 vector unsigned short);
11745 vector signed int vec_sra (vector signed int, vector unsigned int);
11746 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
11748 vector signed int vec_vsraw (vector signed int, vector unsigned int);
11749 vector unsigned int vec_vsraw (vector unsigned int,
11750 vector unsigned int);
11752 vector signed short vec_vsrah (vector signed short,
11753 vector unsigned short);
11754 vector unsigned short vec_vsrah (vector unsigned short,
11755 vector unsigned short);
11757 vector signed char vec_vsrab (vector signed char, vector unsigned char);
11758 vector unsigned char vec_vsrab (vector unsigned char,
11759 vector unsigned char);
11761 vector signed int vec_srl (vector signed int, vector unsigned int);
11762 vector signed int vec_srl (vector signed int, vector unsigned short);
11763 vector signed int vec_srl (vector signed int, vector unsigned char);
11764 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
11765 vector unsigned int vec_srl (vector unsigned int,
11766 vector unsigned short);
11767 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
11768 vector bool int vec_srl (vector bool int, vector unsigned int);
11769 vector bool int vec_srl (vector bool int, vector unsigned short);
11770 vector bool int vec_srl (vector bool int, vector unsigned char);
11771 vector signed short vec_srl (vector signed short, vector unsigned int);
11772 vector signed short vec_srl (vector signed short,
11773 vector unsigned short);
11774 vector signed short vec_srl (vector signed short, vector unsigned char);
11775 vector unsigned short vec_srl (vector unsigned short,
11776 vector unsigned int);
11777 vector unsigned short vec_srl (vector unsigned short,
11778 vector unsigned short);
11779 vector unsigned short vec_srl (vector unsigned short,
11780 vector unsigned char);
11781 vector bool short vec_srl (vector bool short, vector unsigned int);
11782 vector bool short vec_srl (vector bool short, vector unsigned short);
11783 vector bool short vec_srl (vector bool short, vector unsigned char);
11784 vector pixel vec_srl (vector pixel, vector unsigned int);
11785 vector pixel vec_srl (vector pixel, vector unsigned short);
11786 vector pixel vec_srl (vector pixel, vector unsigned char);
11787 vector signed char vec_srl (vector signed char, vector unsigned int);
11788 vector signed char vec_srl (vector signed char, vector unsigned short);
11789 vector signed char vec_srl (vector signed char, vector unsigned char);
11790 vector unsigned char vec_srl (vector unsigned char,
11791 vector unsigned int);
11792 vector unsigned char vec_srl (vector unsigned char,
11793 vector unsigned short);
11794 vector unsigned char vec_srl (vector unsigned char,
11795 vector unsigned char);
11796 vector bool char vec_srl (vector bool char, vector unsigned int);
11797 vector bool char vec_srl (vector bool char, vector unsigned short);
11798 vector bool char vec_srl (vector bool char, vector unsigned char);
11800 vector float vec_sro (vector float, vector signed char);
11801 vector float vec_sro (vector float, vector unsigned char);
11802 vector signed int vec_sro (vector signed int, vector signed char);
11803 vector signed int vec_sro (vector signed int, vector unsigned char);
11804 vector unsigned int vec_sro (vector unsigned int, vector signed char);
11805 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
11806 vector signed short vec_sro (vector signed short, vector signed char);
11807 vector signed short vec_sro (vector signed short, vector unsigned char);
11808 vector unsigned short vec_sro (vector unsigned short,
11809 vector signed char);
11810 vector unsigned short vec_sro (vector unsigned short,
11811 vector unsigned char);
11812 vector pixel vec_sro (vector pixel, vector signed char);
11813 vector pixel vec_sro (vector pixel, vector unsigned char);
11814 vector signed char vec_sro (vector signed char, vector signed char);
11815 vector signed char vec_sro (vector signed char, vector unsigned char);
11816 vector unsigned char vec_sro (vector unsigned char, vector signed char);
11817 vector unsigned char vec_sro (vector unsigned char,
11818 vector unsigned char);
11820 void vec_st (vector float, int, vector float *);
11821 void vec_st (vector float, int, float *);
11822 void vec_st (vector signed int, int, vector signed int *);
11823 void vec_st (vector signed int, int, int *);
11824 void vec_st (vector unsigned int, int, vector unsigned int *);
11825 void vec_st (vector unsigned int, int, unsigned int *);
11826 void vec_st (vector bool int, int, vector bool int *);
11827 void vec_st (vector bool int, int, unsigned int *);
11828 void vec_st (vector bool int, int, int *);
11829 void vec_st (vector signed short, int, vector signed short *);
11830 void vec_st (vector signed short, int, short *);
11831 void vec_st (vector unsigned short, int, vector unsigned short *);
11832 void vec_st (vector unsigned short, int, unsigned short *);
11833 void vec_st (vector bool short, int, vector bool short *);
11834 void vec_st (vector bool short, int, unsigned short *);
11835 void vec_st (vector pixel, int, vector pixel *);
11836 void vec_st (vector pixel, int, unsigned short *);
11837 void vec_st (vector pixel, int, short *);
11838 void vec_st (vector bool short, int, short *);
11839 void vec_st (vector signed char, int, vector signed char *);
11840 void vec_st (vector signed char, int, signed char *);
11841 void vec_st (vector unsigned char, int, vector unsigned char *);
11842 void vec_st (vector unsigned char, int, unsigned char *);
11843 void vec_st (vector bool char, int, vector bool char *);
11844 void vec_st (vector bool char, int, unsigned char *);
11845 void vec_st (vector bool char, int, signed char *);
11847 void vec_ste (vector signed char, int, signed char *);
11848 void vec_ste (vector unsigned char, int, unsigned char *);
11849 void vec_ste (vector bool char, int, signed char *);
11850 void vec_ste (vector bool char, int, unsigned char *);
11851 void vec_ste (vector signed short, int, short *);
11852 void vec_ste (vector unsigned short, int, unsigned short *);
11853 void vec_ste (vector bool short, int, short *);
11854 void vec_ste (vector bool short, int, unsigned short *);
11855 void vec_ste (vector pixel, int, short *);
11856 void vec_ste (vector pixel, int, unsigned short *);
11857 void vec_ste (vector float, int, float *);
11858 void vec_ste (vector signed int, int, int *);
11859 void vec_ste (vector unsigned int, int, unsigned int *);
11860 void vec_ste (vector bool int, int, int *);
11861 void vec_ste (vector bool int, int, unsigned int *);
11863 void vec_stvewx (vector float, int, float *);
11864 void vec_stvewx (vector signed int, int, int *);
11865 void vec_stvewx (vector unsigned int, int, unsigned int *);
11866 void vec_stvewx (vector bool int, int, int *);
11867 void vec_stvewx (vector bool int, int, unsigned int *);
11869 void vec_stvehx (vector signed short, int, short *);
11870 void vec_stvehx (vector unsigned short, int, unsigned short *);
11871 void vec_stvehx (vector bool short, int, short *);
11872 void vec_stvehx (vector bool short, int, unsigned short *);
11873 void vec_stvehx (vector pixel, int, short *);
11874 void vec_stvehx (vector pixel, int, unsigned short *);
11876 void vec_stvebx (vector signed char, int, signed char *);
11877 void vec_stvebx (vector unsigned char, int, unsigned char *);
11878 void vec_stvebx (vector bool char, int, signed char *);
11879 void vec_stvebx (vector bool char, int, unsigned char *);
11881 void vec_stl (vector float, int, vector float *);
11882 void vec_stl (vector float, int, float *);
11883 void vec_stl (vector signed int, int, vector signed int *);
11884 void vec_stl (vector signed int, int, int *);
11885 void vec_stl (vector unsigned int, int, vector unsigned int *);
11886 void vec_stl (vector unsigned int, int, unsigned int *);
11887 void vec_stl (vector bool int, int, vector bool int *);
11888 void vec_stl (vector bool int, int, unsigned int *);
11889 void vec_stl (vector bool int, int, int *);
11890 void vec_stl (vector signed short, int, vector signed short *);
11891 void vec_stl (vector signed short, int, short *);
11892 void vec_stl (vector unsigned short, int, vector unsigned short *);
11893 void vec_stl (vector unsigned short, int, unsigned short *);
11894 void vec_stl (vector bool short, int, vector bool short *);
11895 void vec_stl (vector bool short, int, unsigned short *);
11896 void vec_stl (vector bool short, int, short *);
11897 void vec_stl (vector pixel, int, vector pixel *);
11898 void vec_stl (vector pixel, int, unsigned short *);
11899 void vec_stl (vector pixel, int, short *);
11900 void vec_stl (vector signed char, int, vector signed char *);
11901 void vec_stl (vector signed char, int, signed char *);
11902 void vec_stl (vector unsigned char, int, vector unsigned char *);
11903 void vec_stl (vector unsigned char, int, unsigned char *);
11904 void vec_stl (vector bool char, int, vector bool char *);
11905 void vec_stl (vector bool char, int, unsigned char *);
11906 void vec_stl (vector bool char, int, signed char *);
11908 vector signed char vec_sub (vector bool char, vector signed char);
11909 vector signed char vec_sub (vector signed char, vector bool char);
11910 vector signed char vec_sub (vector signed char, vector signed char);
11911 vector unsigned char vec_sub (vector bool char, vector unsigned char);
11912 vector unsigned char vec_sub (vector unsigned char, vector bool char);
11913 vector unsigned char vec_sub (vector unsigned char,
11914 vector unsigned char);
11915 vector signed short vec_sub (vector bool short, vector signed short);
11916 vector signed short vec_sub (vector signed short, vector bool short);
11917 vector signed short vec_sub (vector signed short, vector signed short);
11918 vector unsigned short vec_sub (vector bool short,
11919 vector unsigned short);
11920 vector unsigned short vec_sub (vector unsigned short,
11921 vector bool short);
11922 vector unsigned short vec_sub (vector unsigned short,
11923 vector unsigned short);
11924 vector signed int vec_sub (vector bool int, vector signed int);
11925 vector signed int vec_sub (vector signed int, vector bool int);
11926 vector signed int vec_sub (vector signed int, vector signed int);
11927 vector unsigned int vec_sub (vector bool int, vector unsigned int);
11928 vector unsigned int vec_sub (vector unsigned int, vector bool int);
11929 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
11930 vector float vec_sub (vector float, vector float);
11932 vector float vec_vsubfp (vector float, vector float);
11934 vector signed int vec_vsubuwm (vector bool int, vector signed int);
11935 vector signed int vec_vsubuwm (vector signed int, vector bool int);
11936 vector signed int vec_vsubuwm (vector signed int, vector signed int);
11937 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
11938 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
11939 vector unsigned int vec_vsubuwm (vector unsigned int,
11940 vector unsigned int);
11942 vector signed short vec_vsubuhm (vector bool short,
11943 vector signed short);
11944 vector signed short vec_vsubuhm (vector signed short,
11945 vector bool short);
11946 vector signed short vec_vsubuhm (vector signed short,
11947 vector signed short);
11948 vector unsigned short vec_vsubuhm (vector bool short,
11949 vector unsigned short);
11950 vector unsigned short vec_vsubuhm (vector unsigned short,
11951 vector bool short);
11952 vector unsigned short vec_vsubuhm (vector unsigned short,
11953 vector unsigned short);
11955 vector signed char vec_vsububm (vector bool char, vector signed char);
11956 vector signed char vec_vsububm (vector signed char, vector bool char);
11957 vector signed char vec_vsububm (vector signed char, vector signed char);
11958 vector unsigned char vec_vsububm (vector bool char,
11959 vector unsigned char);
11960 vector unsigned char vec_vsububm (vector unsigned char,
11962 vector unsigned char vec_vsububm (vector unsigned char,
11963 vector unsigned char);
11965 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
11967 vector unsigned char vec_subs (vector bool char, vector unsigned char);
11968 vector unsigned char vec_subs (vector unsigned char, vector bool char);
11969 vector unsigned char vec_subs (vector unsigned char,
11970 vector unsigned char);
11971 vector signed char vec_subs (vector bool char, vector signed char);
11972 vector signed char vec_subs (vector signed char, vector bool char);
11973 vector signed char vec_subs (vector signed char, vector signed char);
11974 vector unsigned short vec_subs (vector bool short,
11975 vector unsigned short);
11976 vector unsigned short vec_subs (vector unsigned short,
11977 vector bool short);
11978 vector unsigned short vec_subs (vector unsigned short,
11979 vector unsigned short);
11980 vector signed short vec_subs (vector bool short, vector signed short);
11981 vector signed short vec_subs (vector signed short, vector bool short);
11982 vector signed short vec_subs (vector signed short, vector signed short);
11983 vector unsigned int vec_subs (vector bool int, vector unsigned int);
11984 vector unsigned int vec_subs (vector unsigned int, vector bool int);
11985 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
11986 vector signed int vec_subs (vector bool int, vector signed int);
11987 vector signed int vec_subs (vector signed int, vector bool int);
11988 vector signed int vec_subs (vector signed int, vector signed int);
11990 vector signed int vec_vsubsws (vector bool int, vector signed int);
11991 vector signed int vec_vsubsws (vector signed int, vector bool int);
11992 vector signed int vec_vsubsws (vector signed int, vector signed int);
11994 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
11995 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
11996 vector unsigned int vec_vsubuws (vector unsigned int,
11997 vector unsigned int);
11999 vector signed short vec_vsubshs (vector bool short,
12000 vector signed short);
12001 vector signed short vec_vsubshs (vector signed short,
12002 vector bool short);
12003 vector signed short vec_vsubshs (vector signed short,
12004 vector signed short);
12006 vector unsigned short vec_vsubuhs (vector bool short,
12007 vector unsigned short);
12008 vector unsigned short vec_vsubuhs (vector unsigned short,
12009 vector bool short);
12010 vector unsigned short vec_vsubuhs (vector unsigned short,
12011 vector unsigned short);
12013 vector signed char vec_vsubsbs (vector bool char, vector signed char);
12014 vector signed char vec_vsubsbs (vector signed char, vector bool char);
12015 vector signed char vec_vsubsbs (vector signed char, vector signed char);
12017 vector unsigned char vec_vsububs (vector bool char,
12018 vector unsigned char);
12019 vector unsigned char vec_vsububs (vector unsigned char,
12021 vector unsigned char vec_vsububs (vector unsigned char,
12022 vector unsigned char);
12024 vector unsigned int vec_sum4s (vector unsigned char,
12025 vector unsigned int);
12026 vector signed int vec_sum4s (vector signed char, vector signed int);
12027 vector signed int vec_sum4s (vector signed short, vector signed int);
12029 vector signed int vec_vsum4shs (vector signed short, vector signed int);
12031 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
12033 vector unsigned int vec_vsum4ubs (vector unsigned char,
12034 vector unsigned int);
12036 vector signed int vec_sum2s (vector signed int, vector signed int);
12038 vector signed int vec_sums (vector signed int, vector signed int);
12040 vector float vec_trunc (vector float);
12042 vector signed short vec_unpackh (vector signed char);
12043 vector bool short vec_unpackh (vector bool char);
12044 vector signed int vec_unpackh (vector signed short);
12045 vector bool int vec_unpackh (vector bool short);
12046 vector unsigned int vec_unpackh (vector pixel);
12048 vector bool int vec_vupkhsh (vector bool short);
12049 vector signed int vec_vupkhsh (vector signed short);
12051 vector unsigned int vec_vupkhpx (vector pixel);
12053 vector bool short vec_vupkhsb (vector bool char);
12054 vector signed short vec_vupkhsb (vector signed char);
12056 vector signed short vec_unpackl (vector signed char);
12057 vector bool short vec_unpackl (vector bool char);
12058 vector unsigned int vec_unpackl (vector pixel);
12059 vector signed int vec_unpackl (vector signed short);
12060 vector bool int vec_unpackl (vector bool short);
12062 vector unsigned int vec_vupklpx (vector pixel);
12064 vector bool int vec_vupklsh (vector bool short);
12065 vector signed int vec_vupklsh (vector signed short);
12067 vector bool short vec_vupklsb (vector bool char);
12068 vector signed short vec_vupklsb (vector signed char);
12070 vector float vec_xor (vector float, vector float);
12071 vector float vec_xor (vector float, vector bool int);
12072 vector float vec_xor (vector bool int, vector float);
12073 vector bool int vec_xor (vector bool int, vector bool int);
12074 vector signed int vec_xor (vector bool int, vector signed int);
12075 vector signed int vec_xor (vector signed int, vector bool int);
12076 vector signed int vec_xor (vector signed int, vector signed int);
12077 vector unsigned int vec_xor (vector bool int, vector unsigned int);
12078 vector unsigned int vec_xor (vector unsigned int, vector bool int);
12079 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
12080 vector bool short vec_xor (vector bool short, vector bool short);
12081 vector signed short vec_xor (vector bool short, vector signed short);
12082 vector signed short vec_xor (vector signed short, vector bool short);
12083 vector signed short vec_xor (vector signed short, vector signed short);
12084 vector unsigned short vec_xor (vector bool short,
12085 vector unsigned short);
12086 vector unsigned short vec_xor (vector unsigned short,
12087 vector bool short);
12088 vector unsigned short vec_xor (vector unsigned short,
12089 vector unsigned short);
12090 vector signed char vec_xor (vector bool char, vector signed char);
12091 vector bool char vec_xor (vector bool char, vector bool char);
12092 vector signed char vec_xor (vector signed char, vector bool char);
12093 vector signed char vec_xor (vector signed char, vector signed char);
12094 vector unsigned char vec_xor (vector bool char, vector unsigned char);
12095 vector unsigned char vec_xor (vector unsigned char, vector bool char);
12096 vector unsigned char vec_xor (vector unsigned char,
12097 vector unsigned char);
12099 int vec_all_eq (vector signed char, vector bool char);
12100 int vec_all_eq (vector signed char, vector signed char);
12101 int vec_all_eq (vector unsigned char, vector bool char);
12102 int vec_all_eq (vector unsigned char, vector unsigned char);
12103 int vec_all_eq (vector bool char, vector bool char);
12104 int vec_all_eq (vector bool char, vector unsigned char);
12105 int vec_all_eq (vector bool char, vector signed char);
12106 int vec_all_eq (vector signed short, vector bool short);
12107 int vec_all_eq (vector signed short, vector signed short);
12108 int vec_all_eq (vector unsigned short, vector bool short);
12109 int vec_all_eq (vector unsigned short, vector unsigned short);
12110 int vec_all_eq (vector bool short, vector bool short);
12111 int vec_all_eq (vector bool short, vector unsigned short);
12112 int vec_all_eq (vector bool short, vector signed short);
12113 int vec_all_eq (vector pixel, vector pixel);
12114 int vec_all_eq (vector signed int, vector bool int);
12115 int vec_all_eq (vector signed int, vector signed int);
12116 int vec_all_eq (vector unsigned int, vector bool int);
12117 int vec_all_eq (vector unsigned int, vector unsigned int);
12118 int vec_all_eq (vector bool int, vector bool int);
12119 int vec_all_eq (vector bool int, vector unsigned int);
12120 int vec_all_eq (vector bool int, vector signed int);
12121 int vec_all_eq (vector float, vector float);
12123 int vec_all_ge (vector bool char, vector unsigned char);
12124 int vec_all_ge (vector unsigned char, vector bool char);
12125 int vec_all_ge (vector unsigned char, vector unsigned char);
12126 int vec_all_ge (vector bool char, vector signed char);
12127 int vec_all_ge (vector signed char, vector bool char);
12128 int vec_all_ge (vector signed char, vector signed char);
12129 int vec_all_ge (vector bool short, vector unsigned short);
12130 int vec_all_ge (vector unsigned short, vector bool short);
12131 int vec_all_ge (vector unsigned short, vector unsigned short);
12132 int vec_all_ge (vector signed short, vector signed short);
12133 int vec_all_ge (vector bool short, vector signed short);
12134 int vec_all_ge (vector signed short, vector bool short);
12135 int vec_all_ge (vector bool int, vector unsigned int);
12136 int vec_all_ge (vector unsigned int, vector bool int);
12137 int vec_all_ge (vector unsigned int, vector unsigned int);
12138 int vec_all_ge (vector bool int, vector signed int);
12139 int vec_all_ge (vector signed int, vector bool int);
12140 int vec_all_ge (vector signed int, vector signed int);
12141 int vec_all_ge (vector float, vector float);
12143 int vec_all_gt (vector bool char, vector unsigned char);
12144 int vec_all_gt (vector unsigned char, vector bool char);
12145 int vec_all_gt (vector unsigned char, vector unsigned char);
12146 int vec_all_gt (vector bool char, vector signed char);
12147 int vec_all_gt (vector signed char, vector bool char);
12148 int vec_all_gt (vector signed char, vector signed char);
12149 int vec_all_gt (vector bool short, vector unsigned short);
12150 int vec_all_gt (vector unsigned short, vector bool short);
12151 int vec_all_gt (vector unsigned short, vector unsigned short);
12152 int vec_all_gt (vector bool short, vector signed short);
12153 int vec_all_gt (vector signed short, vector bool short);
12154 int vec_all_gt (vector signed short, vector signed short);
12155 int vec_all_gt (vector bool int, vector unsigned int);
12156 int vec_all_gt (vector unsigned int, vector bool int);
12157 int vec_all_gt (vector unsigned int, vector unsigned int);
12158 int vec_all_gt (vector bool int, vector signed int);
12159 int vec_all_gt (vector signed int, vector bool int);
12160 int vec_all_gt (vector signed int, vector signed int);
12161 int vec_all_gt (vector float, vector float);
12163 int vec_all_in (vector float, vector float);
12165 int vec_all_le (vector bool char, vector unsigned char);
12166 int vec_all_le (vector unsigned char, vector bool char);
12167 int vec_all_le (vector unsigned char, vector unsigned char);
12168 int vec_all_le (vector bool char, vector signed char);
12169 int vec_all_le (vector signed char, vector bool char);
12170 int vec_all_le (vector signed char, vector signed char);
12171 int vec_all_le (vector bool short, vector unsigned short);
12172 int vec_all_le (vector unsigned short, vector bool short);
12173 int vec_all_le (vector unsigned short, vector unsigned short);
12174 int vec_all_le (vector bool short, vector signed short);
12175 int vec_all_le (vector signed short, vector bool short);
12176 int vec_all_le (vector signed short, vector signed short);
12177 int vec_all_le (vector bool int, vector unsigned int);
12178 int vec_all_le (vector unsigned int, vector bool int);
12179 int vec_all_le (vector unsigned int, vector unsigned int);
12180 int vec_all_le (vector bool int, vector signed int);
12181 int vec_all_le (vector signed int, vector bool int);
12182 int vec_all_le (vector signed int, vector signed int);
12183 int vec_all_le (vector float, vector float);
12185 int vec_all_lt (vector bool char, vector unsigned char);
12186 int vec_all_lt (vector unsigned char, vector bool char);
12187 int vec_all_lt (vector unsigned char, vector unsigned char);
12188 int vec_all_lt (vector bool char, vector signed char);
12189 int vec_all_lt (vector signed char, vector bool char);
12190 int vec_all_lt (vector signed char, vector signed char);
12191 int vec_all_lt (vector bool short, vector unsigned short);
12192 int vec_all_lt (vector unsigned short, vector bool short);
12193 int vec_all_lt (vector unsigned short, vector unsigned short);
12194 int vec_all_lt (vector bool short, vector signed short);
12195 int vec_all_lt (vector signed short, vector bool short);
12196 int vec_all_lt (vector signed short, vector signed short);
12197 int vec_all_lt (vector bool int, vector unsigned int);
12198 int vec_all_lt (vector unsigned int, vector bool int);
12199 int vec_all_lt (vector unsigned int, vector unsigned int);
12200 int vec_all_lt (vector bool int, vector signed int);
12201 int vec_all_lt (vector signed int, vector bool int);
12202 int vec_all_lt (vector signed int, vector signed int);
12203 int vec_all_lt (vector float, vector float);
12205 int vec_all_nan (vector float);
12207 int vec_all_ne (vector signed char, vector bool char);
12208 int vec_all_ne (vector signed char, vector signed char);
12209 int vec_all_ne (vector unsigned char, vector bool char);
12210 int vec_all_ne (vector unsigned char, vector unsigned char);
12211 int vec_all_ne (vector bool char, vector bool char);
12212 int vec_all_ne (vector bool char, vector unsigned char);
12213 int vec_all_ne (vector bool char, vector signed char);
12214 int vec_all_ne (vector signed short, vector bool short);
12215 int vec_all_ne (vector signed short, vector signed short);
12216 int vec_all_ne (vector unsigned short, vector bool short);
12217 int vec_all_ne (vector unsigned short, vector unsigned short);
12218 int vec_all_ne (vector bool short, vector bool short);
12219 int vec_all_ne (vector bool short, vector unsigned short);
12220 int vec_all_ne (vector bool short, vector signed short);
12221 int vec_all_ne (vector pixel, vector pixel);
12222 int vec_all_ne (vector signed int, vector bool int);
12223 int vec_all_ne (vector signed int, vector signed int);
12224 int vec_all_ne (vector unsigned int, vector bool int);
12225 int vec_all_ne (vector unsigned int, vector unsigned int);
12226 int vec_all_ne (vector bool int, vector bool int);
12227 int vec_all_ne (vector bool int, vector unsigned int);
12228 int vec_all_ne (vector bool int, vector signed int);
12229 int vec_all_ne (vector float, vector float);
12231 int vec_all_nge (vector float, vector float);
12233 int vec_all_ngt (vector float, vector float);
12235 int vec_all_nle (vector float, vector float);
12237 int vec_all_nlt (vector float, vector float);
12239 int vec_all_numeric (vector float);
12241 int vec_any_eq (vector signed char, vector bool char);
12242 int vec_any_eq (vector signed char, vector signed char);
12243 int vec_any_eq (vector unsigned char, vector bool char);
12244 int vec_any_eq (vector unsigned char, vector unsigned char);
12245 int vec_any_eq (vector bool char, vector bool char);
12246 int vec_any_eq (vector bool char, vector unsigned char);
12247 int vec_any_eq (vector bool char, vector signed char);
12248 int vec_any_eq (vector signed short, vector bool short);
12249 int vec_any_eq (vector signed short, vector signed short);
12250 int vec_any_eq (vector unsigned short, vector bool short);
12251 int vec_any_eq (vector unsigned short, vector unsigned short);
12252 int vec_any_eq (vector bool short, vector bool short);
12253 int vec_any_eq (vector bool short, vector unsigned short);
12254 int vec_any_eq (vector bool short, vector signed short);
12255 int vec_any_eq (vector pixel, vector pixel);
12256 int vec_any_eq (vector signed int, vector bool int);
12257 int vec_any_eq (vector signed int, vector signed int);
12258 int vec_any_eq (vector unsigned int, vector bool int);
12259 int vec_any_eq (vector unsigned int, vector unsigned int);
12260 int vec_any_eq (vector bool int, vector bool int);
12261 int vec_any_eq (vector bool int, vector unsigned int);
12262 int vec_any_eq (vector bool int, vector signed int);
12263 int vec_any_eq (vector float, vector float);
12265 int vec_any_ge (vector signed char, vector bool char);
12266 int vec_any_ge (vector unsigned char, vector bool char);
12267 int vec_any_ge (vector unsigned char, vector unsigned char);
12268 int vec_any_ge (vector signed char, vector signed char);
12269 int vec_any_ge (vector bool char, vector unsigned char);
12270 int vec_any_ge (vector bool char, vector signed char);
12271 int vec_any_ge (vector unsigned short, vector bool short);
12272 int vec_any_ge (vector unsigned short, vector unsigned short);
12273 int vec_any_ge (vector signed short, vector signed short);
12274 int vec_any_ge (vector signed short, vector bool short);
12275 int vec_any_ge (vector bool short, vector unsigned short);
12276 int vec_any_ge (vector bool short, vector signed short);
12277 int vec_any_ge (vector signed int, vector bool int);
12278 int vec_any_ge (vector unsigned int, vector bool int);
12279 int vec_any_ge (vector unsigned int, vector unsigned int);
12280 int vec_any_ge (vector signed int, vector signed int);
12281 int vec_any_ge (vector bool int, vector unsigned int);
12282 int vec_any_ge (vector bool int, vector signed int);
12283 int vec_any_ge (vector float, vector float);
12285 int vec_any_gt (vector bool char, vector unsigned char);
12286 int vec_any_gt (vector unsigned char, vector bool char);
12287 int vec_any_gt (vector unsigned char, vector unsigned char);
12288 int vec_any_gt (vector bool char, vector signed char);
12289 int vec_any_gt (vector signed char, vector bool char);
12290 int vec_any_gt (vector signed char, vector signed char);
12291 int vec_any_gt (vector bool short, vector unsigned short);
12292 int vec_any_gt (vector unsigned short, vector bool short);
12293 int vec_any_gt (vector unsigned short, vector unsigned short);
12294 int vec_any_gt (vector bool short, vector signed short);
12295 int vec_any_gt (vector signed short, vector bool short);
12296 int vec_any_gt (vector signed short, vector signed short);
12297 int vec_any_gt (vector bool int, vector unsigned int);
12298 int vec_any_gt (vector unsigned int, vector bool int);
12299 int vec_any_gt (vector unsigned int, vector unsigned int);
12300 int vec_any_gt (vector bool int, vector signed int);
12301 int vec_any_gt (vector signed int, vector bool int);
12302 int vec_any_gt (vector signed int, vector signed int);
12303 int vec_any_gt (vector float, vector float);
12305 int vec_any_le (vector bool char, vector unsigned char);
12306 int vec_any_le (vector unsigned char, vector bool char);
12307 int vec_any_le (vector unsigned char, vector unsigned char);
12308 int vec_any_le (vector bool char, vector signed char);
12309 int vec_any_le (vector signed char, vector bool char);
12310 int vec_any_le (vector signed char, vector signed char);
12311 int vec_any_le (vector bool short, vector unsigned short);
12312 int vec_any_le (vector unsigned short, vector bool short);
12313 int vec_any_le (vector unsigned short, vector unsigned short);
12314 int vec_any_le (vector bool short, vector signed short);
12315 int vec_any_le (vector signed short, vector bool short);
12316 int vec_any_le (vector signed short, vector signed short);
12317 int vec_any_le (vector bool int, vector unsigned int);
12318 int vec_any_le (vector unsigned int, vector bool int);
12319 int vec_any_le (vector unsigned int, vector unsigned int);
12320 int vec_any_le (vector bool int, vector signed int);
12321 int vec_any_le (vector signed int, vector bool int);
12322 int vec_any_le (vector signed int, vector signed int);
12323 int vec_any_le (vector float, vector float);
12325 int vec_any_lt (vector bool char, vector unsigned char);
12326 int vec_any_lt (vector unsigned char, vector bool char);
12327 int vec_any_lt (vector unsigned char, vector unsigned char);
12328 int vec_any_lt (vector bool char, vector signed char);
12329 int vec_any_lt (vector signed char, vector bool char);
12330 int vec_any_lt (vector signed char, vector signed char);
12331 int vec_any_lt (vector bool short, vector unsigned short);
12332 int vec_any_lt (vector unsigned short, vector bool short);
12333 int vec_any_lt (vector unsigned short, vector unsigned short);
12334 int vec_any_lt (vector bool short, vector signed short);
12335 int vec_any_lt (vector signed short, vector bool short);
12336 int vec_any_lt (vector signed short, vector signed short);
12337 int vec_any_lt (vector bool int, vector unsigned int);
12338 int vec_any_lt (vector unsigned int, vector bool int);
12339 int vec_any_lt (vector unsigned int, vector unsigned int);
12340 int vec_any_lt (vector bool int, vector signed int);
12341 int vec_any_lt (vector signed int, vector bool int);
12342 int vec_any_lt (vector signed int, vector signed int);
12343 int vec_any_lt (vector float, vector float);
12345 int vec_any_nan (vector float);
12347 int vec_any_ne (vector signed char, vector bool char);
12348 int vec_any_ne (vector signed char, vector signed char);
12349 int vec_any_ne (vector unsigned char, vector bool char);
12350 int vec_any_ne (vector unsigned char, vector unsigned char);
12351 int vec_any_ne (vector bool char, vector bool char);
12352 int vec_any_ne (vector bool char, vector unsigned char);
12353 int vec_any_ne (vector bool char, vector signed char);
12354 int vec_any_ne (vector signed short, vector bool short);
12355 int vec_any_ne (vector signed short, vector signed short);
12356 int vec_any_ne (vector unsigned short, vector bool short);
12357 int vec_any_ne (vector unsigned short, vector unsigned short);
12358 int vec_any_ne (vector bool short, vector bool short);
12359 int vec_any_ne (vector bool short, vector unsigned short);
12360 int vec_any_ne (vector bool short, vector signed short);
12361 int vec_any_ne (vector pixel, vector pixel);
12362 int vec_any_ne (vector signed int, vector bool int);
12363 int vec_any_ne (vector signed int, vector signed int);
12364 int vec_any_ne (vector unsigned int, vector bool int);
12365 int vec_any_ne (vector unsigned int, vector unsigned int);
12366 int vec_any_ne (vector bool int, vector bool int);
12367 int vec_any_ne (vector bool int, vector unsigned int);
12368 int vec_any_ne (vector bool int, vector signed int);
12369 int vec_any_ne (vector float, vector float);
12371 int vec_any_nge (vector float, vector float);
12373 int vec_any_ngt (vector float, vector float);
12375 int vec_any_nle (vector float, vector float);
12377 int vec_any_nlt (vector float, vector float);
12379 int vec_any_numeric (vector float);
12381 int vec_any_out (vector float, vector float);
12384 If the vector/scalar (VSX) instruction set is available, the following
12385 additional functions are available:
12388 vector double vec_abs (vector double);
12389 vector double vec_add (vector double, vector double);
12390 vector double vec_and (vector double, vector double);
12391 vector double vec_and (vector double, vector bool long);
12392 vector double vec_and (vector bool long, vector double);
12393 vector double vec_andc (vector double, vector double);
12394 vector double vec_andc (vector double, vector bool long);
12395 vector double vec_andc (vector bool long, vector double);
12396 vector double vec_ceil (vector double);
12397 vector bool long vec_cmpeq (vector double, vector double);
12398 vector bool long vec_cmpge (vector double, vector double);
12399 vector bool long vec_cmpgt (vector double, vector double);
12400 vector bool long vec_cmple (vector double, vector double);
12401 vector bool long vec_cmplt (vector double, vector double);
12402 vector float vec_div (vector float, vector float);
12403 vector double vec_div (vector double, vector double);
12404 vector double vec_floor (vector double);
12405 vector double vec_ld (int, const vector double *);
12406 vector double vec_ld (int, const double *);
12407 vector double vec_ldl (int, const vector double *);
12408 vector double vec_ldl (int, const double *);
12409 vector unsigned char vec_lvsl (int, const volatile double *);
12410 vector unsigned char vec_lvsr (int, const volatile double *);
12411 vector double vec_madd (vector double, vector double, vector double);
12412 vector double vec_max (vector double, vector double);
12413 vector double vec_min (vector double, vector double);
12414 vector float vec_msub (vector float, vector float, vector float);
12415 vector double vec_msub (vector double, vector double, vector double);
12416 vector float vec_mul (vector float, vector float);
12417 vector double vec_mul (vector double, vector double);
12418 vector float vec_nearbyint (vector float);
12419 vector double vec_nearbyint (vector double);
12420 vector float vec_nmadd (vector float, vector float, vector float);
12421 vector double vec_nmadd (vector double, vector double, vector double);
12422 vector double vec_nmsub (vector double, vector double, vector double);
12423 vector double vec_nor (vector double, vector double);
12424 vector double vec_or (vector double, vector double);
12425 vector double vec_or (vector double, vector bool long);
12426 vector double vec_or (vector bool long, vector double);
12427 vector double vec_perm (vector double,
12429 vector unsigned char);
12430 vector double vec_rint (vector double);
12431 vector double vec_recip (vector double, vector double);
12432 vector double vec_rsqrt (vector double);
12433 vector double vec_rsqrte (vector double);
12434 vector double vec_sel (vector double, vector double, vector bool long);
12435 vector double vec_sel (vector double, vector double, vector unsigned long);
12436 vector double vec_sub (vector double, vector double);
12437 vector float vec_sqrt (vector float);
12438 vector double vec_sqrt (vector double);
12439 void vec_st (vector double, int, vector double *);
12440 void vec_st (vector double, int, double *);
12441 vector double vec_trunc (vector double);
12442 vector double vec_xor (vector double, vector double);
12443 vector double vec_xor (vector double, vector bool long);
12444 vector double vec_xor (vector bool long, vector double);
12445 int vec_all_eq (vector double, vector double);
12446 int vec_all_ge (vector double, vector double);
12447 int vec_all_gt (vector double, vector double);
12448 int vec_all_le (vector double, vector double);
12449 int vec_all_lt (vector double, vector double);
12450 int vec_all_nan (vector double);
12451 int vec_all_ne (vector double, vector double);
12452 int vec_all_nge (vector double, vector double);
12453 int vec_all_ngt (vector double, vector double);
12454 int vec_all_nle (vector double, vector double);
12455 int vec_all_nlt (vector double, vector double);
12456 int vec_all_numeric (vector double);
12457 int vec_any_eq (vector double, vector double);
12458 int vec_any_ge (vector double, vector double);
12459 int vec_any_gt (vector double, vector double);
12460 int vec_any_le (vector double, vector double);
12461 int vec_any_lt (vector double, vector double);
12462 int vec_any_nan (vector double);
12463 int vec_any_ne (vector double, vector double);
12464 int vec_any_nge (vector double, vector double);
12465 int vec_any_ngt (vector double, vector double);
12466 int vec_any_nle (vector double, vector double);
12467 int vec_any_nlt (vector double, vector double);
12468 int vec_any_numeric (vector double);
12470 vector double vec_vsx_ld (int, const vector double *);
12471 vector double vec_vsx_ld (int, const double *);
12472 vector float vec_vsx_ld (int, const vector float *);
12473 vector float vec_vsx_ld (int, const float *);
12474 vector bool int vec_vsx_ld (int, const vector bool int *);
12475 vector signed int vec_vsx_ld (int, const vector signed int *);
12476 vector signed int vec_vsx_ld (int, const int *);
12477 vector signed int vec_vsx_ld (int, const long *);
12478 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
12479 vector unsigned int vec_vsx_ld (int, const unsigned int *);
12480 vector unsigned int vec_vsx_ld (int, const unsigned long *);
12481 vector bool short vec_vsx_ld (int, const vector bool short *);
12482 vector pixel vec_vsx_ld (int, const vector pixel *);
12483 vector signed short vec_vsx_ld (int, const vector signed short *);
12484 vector signed short vec_vsx_ld (int, const short *);
12485 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
12486 vector unsigned short vec_vsx_ld (int, const unsigned short *);
12487 vector bool char vec_vsx_ld (int, const vector bool char *);
12488 vector signed char vec_vsx_ld (int, const vector signed char *);
12489 vector signed char vec_vsx_ld (int, const signed char *);
12490 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
12491 vector unsigned char vec_vsx_ld (int, const unsigned char *);
12493 void vec_vsx_st (vector double, int, vector double *);
12494 void vec_vsx_st (vector double, int, double *);
12495 void vec_vsx_st (vector float, int, vector float *);
12496 void vec_vsx_st (vector float, int, float *);
12497 void vec_vsx_st (vector signed int, int, vector signed int *);
12498 void vec_vsx_st (vector signed int, int, int *);
12499 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
12500 void vec_vsx_st (vector unsigned int, int, unsigned int *);
12501 void vec_vsx_st (vector bool int, int, vector bool int *);
12502 void vec_vsx_st (vector bool int, int, unsigned int *);
12503 void vec_vsx_st (vector bool int, int, int *);
12504 void vec_vsx_st (vector signed short, int, vector signed short *);
12505 void vec_vsx_st (vector signed short, int, short *);
12506 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
12507 void vec_vsx_st (vector unsigned short, int, unsigned short *);
12508 void vec_vsx_st (vector bool short, int, vector bool short *);
12509 void vec_vsx_st (vector bool short, int, unsigned short *);
12510 void vec_vsx_st (vector pixel, int, vector pixel *);
12511 void vec_vsx_st (vector pixel, int, unsigned short *);
12512 void vec_vsx_st (vector pixel, int, short *);
12513 void vec_vsx_st (vector bool short, int, short *);
12514 void vec_vsx_st (vector signed char, int, vector signed char *);
12515 void vec_vsx_st (vector signed char, int, signed char *);
12516 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
12517 void vec_vsx_st (vector unsigned char, int, unsigned char *);
12518 void vec_vsx_st (vector bool char, int, vector bool char *);
12519 void vec_vsx_st (vector bool char, int, unsigned char *);
12520 void vec_vsx_st (vector bool char, int, signed char *);
12523 Note that the @samp{vec_ld} and @samp{vec_st} builtins will always
12524 generate the Altivec @samp{LVX} and @samp{STVX} instructions even
12525 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
12526 @samp{vec_vsx_st} builtins will always generate the VSX @samp{LXVD2X},
12527 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
12529 GCC provides a few other builtins on Powerpc to access certain instructions:
12531 float __builtin_recipdivf (float, float);
12532 float __builtin_rsqrtf (float);
12533 double __builtin_recipdiv (double, double);
12534 double __builtin_rsqrt (double);
12535 long __builtin_bpermd (long, long);
12536 int __builtin_bswap16 (int);
12539 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
12540 @code{__builtin_rsqrtf} functions generate multiple instructions to
12541 implement the reciprocal sqrt functionality using reciprocal sqrt
12542 estimate instructions.
12544 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
12545 functions generate multiple instructions to implement division using
12546 the reciprocal estimate instructions.
12548 @node RX Built-in Functions
12549 @subsection RX Built-in Functions
12550 GCC supports some of the RX instructions which cannot be expressed in
12551 the C programming language via the use of built-in functions. The
12552 following functions are supported:
12554 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
12555 Generates the @code{brk} machine instruction.
12558 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
12559 Generates the @code{clrpsw} machine instruction to clear the specified
12560 bit in the processor status word.
12563 @deftypefn {Built-in Function} void __builtin_rx_int (int)
12564 Generates the @code{int} machine instruction to generate an interrupt
12565 with the specified value.
12568 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
12569 Generates the @code{machi} machine instruction to add the result of
12570 multiplying the top 16-bits of the two arguments into the
12574 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
12575 Generates the @code{maclo} machine instruction to add the result of
12576 multiplying the bottom 16-bits of the two arguments into the
12580 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
12581 Generates the @code{mulhi} machine instruction to place the result of
12582 multiplying the top 16-bits of the two arguments into the
12586 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
12587 Generates the @code{mullo} machine instruction to place the result of
12588 multiplying the bottom 16-bits of the two arguments into the
12592 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
12593 Generates the @code{mvfachi} machine instruction to read the top
12594 32-bits of the accumulator.
12597 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
12598 Generates the @code{mvfacmi} machine instruction to read the middle
12599 32-bits of the accumulator.
12602 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
12603 Generates the @code{mvfc} machine instruction which reads the control
12604 register specified in its argument and returns its value.
12607 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
12608 Generates the @code{mvtachi} machine instruction to set the top
12609 32-bits of the accumulator.
12612 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
12613 Generates the @code{mvtaclo} machine instruction to set the bottom
12614 32-bits of the accumulator.
12617 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
12618 Generates the @code{mvtc} machine instruction which sets control
12619 register number @code{reg} to @code{val}.
12622 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
12623 Generates the @code{mvtipl} machine instruction set the interrupt
12627 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
12628 Generates the @code{racw} machine instruction to round the accumulator
12629 according to the specified mode.
12632 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
12633 Generates the @code{revw} machine instruction which swaps the bytes in
12634 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
12635 and also bits 16--23 occupy bits 24--31 and vice versa.
12638 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
12639 Generates the @code{rmpa} machine instruction which initiates a
12640 repeated multiply and accumulate sequence.
12643 @deftypefn {Built-in Function} void __builtin_rx_round (float)
12644 Generates the @code{round} machine instruction which returns the
12645 floating point argument rounded according to the current rounding mode
12646 set in the floating point status word register.
12649 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
12650 Generates the @code{sat} machine instruction which returns the
12651 saturated value of the argument.
12654 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
12655 Generates the @code{setpsw} machine instruction to set the specified
12656 bit in the processor status word.
12659 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
12660 Generates the @code{wait} machine instruction.
12663 @node SPARC VIS Built-in Functions
12664 @subsection SPARC VIS Built-in Functions
12666 GCC supports SIMD operations on the SPARC using both the generic vector
12667 extensions (@pxref{Vector Extensions}) as well as built-in functions for
12668 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
12669 switch, the VIS extension is exposed as the following built-in functions:
12672 typedef int v2si __attribute__ ((vector_size (8)));
12673 typedef short v4hi __attribute__ ((vector_size (8)));
12674 typedef short v2hi __attribute__ ((vector_size (4)));
12675 typedef char v8qi __attribute__ ((vector_size (8)));
12676 typedef char v4qi __attribute__ ((vector_size (4)));
12678 void * __builtin_vis_alignaddr (void *, long);
12679 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
12680 v2si __builtin_vis_faligndatav2si (v2si, v2si);
12681 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
12682 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
12684 v4hi __builtin_vis_fexpand (v4qi);
12686 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
12687 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
12688 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
12689 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
12690 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
12691 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
12692 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
12694 v4qi __builtin_vis_fpack16 (v4hi);
12695 v8qi __builtin_vis_fpack32 (v2si, v2si);
12696 v2hi __builtin_vis_fpackfix (v2si);
12697 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
12699 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
12702 @node SPU Built-in Functions
12703 @subsection SPU Built-in Functions
12705 GCC provides extensions for the SPU processor as described in the
12706 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
12707 found at @uref{http://cell.scei.co.jp/} or
12708 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
12709 implementation differs in several ways.
12714 The optional extension of specifying vector constants in parentheses is
12718 A vector initializer requires no cast if the vector constant is of the
12719 same type as the variable it is initializing.
12722 If @code{signed} or @code{unsigned} is omitted, the signedness of the
12723 vector type is the default signedness of the base type. The default
12724 varies depending on the operating system, so a portable program should
12725 always specify the signedness.
12728 By default, the keyword @code{__vector} is added. The macro
12729 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
12733 GCC allows using a @code{typedef} name as the type specifier for a
12737 For C, overloaded functions are implemented with macros so the following
12741 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
12744 Since @code{spu_add} is a macro, the vector constant in the example
12745 is treated as four separate arguments. Wrap the entire argument in
12746 parentheses for this to work.
12749 The extended version of @code{__builtin_expect} is not supported.
12753 @emph{Note:} Only the interface described in the aforementioned
12754 specification is supported. Internally, GCC uses built-in functions to
12755 implement the required functionality, but these are not supported and
12756 are subject to change without notice.
12758 @node Target Format Checks
12759 @section Format Checks Specific to Particular Target Machines
12761 For some target machines, GCC supports additional options to the
12763 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
12766 * Solaris Format Checks::
12767 * Darwin Format Checks::
12770 @node Solaris Format Checks
12771 @subsection Solaris Format Checks
12773 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
12774 check. @code{cmn_err} accepts a subset of the standard @code{printf}
12775 conversions, and the two-argument @code{%b} conversion for displaying
12776 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
12778 @node Darwin Format Checks
12779 @subsection Darwin Format Checks
12781 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
12782 attribute context. Declarations made with such attribution will be parsed for correct syntax
12783 and format argument types. However, parsing of the format string itself is currently undefined
12784 and will not be carried out by this version of the compiler.
12786 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
12787 also be used as format arguments. Note that the relevant headers are only likely to be
12788 available on Darwin (OSX) installations. On such installations, the XCode and system
12789 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
12790 associated functions.
12793 @section Pragmas Accepted by GCC
12795 @cindex @code{#pragma}
12797 GCC supports several types of pragmas, primarily in order to compile
12798 code originally written for other compilers. Note that in general
12799 we do not recommend the use of pragmas; @xref{Function Attributes},
12800 for further explanation.
12806 * RS/6000 and PowerPC Pragmas::
12808 * Solaris Pragmas::
12809 * Symbol-Renaming Pragmas::
12810 * Structure-Packing Pragmas::
12812 * Diagnostic Pragmas::
12813 * Visibility Pragmas::
12814 * Push/Pop Macro Pragmas::
12815 * Function Specific Option Pragmas::
12819 @subsection ARM Pragmas
12821 The ARM target defines pragmas for controlling the default addition of
12822 @code{long_call} and @code{short_call} attributes to functions.
12823 @xref{Function Attributes}, for information about the effects of these
12828 @cindex pragma, long_calls
12829 Set all subsequent functions to have the @code{long_call} attribute.
12831 @item no_long_calls
12832 @cindex pragma, no_long_calls
12833 Set all subsequent functions to have the @code{short_call} attribute.
12835 @item long_calls_off
12836 @cindex pragma, long_calls_off
12837 Do not affect the @code{long_call} or @code{short_call} attributes of
12838 subsequent functions.
12842 @subsection M32C Pragmas
12845 @item GCC memregs @var{number}
12846 @cindex pragma, memregs
12847 Overrides the command-line option @code{-memregs=} for the current
12848 file. Use with care! This pragma must be before any function in the
12849 file, and mixing different memregs values in different objects may
12850 make them incompatible. This pragma is useful when a
12851 performance-critical function uses a memreg for temporary values,
12852 as it may allow you to reduce the number of memregs used.
12854 @item ADDRESS @var{name} @var{address}
12855 @cindex pragma, address
12856 For any declared symbols matching @var{name}, this does three things
12857 to that symbol: it forces the symbol to be located at the given
12858 address (a number), it forces the symbol to be volatile, and it
12859 changes the symbol's scope to be static. This pragma exists for
12860 compatibility with other compilers, but note that the common
12861 @code{1234H} numeric syntax is not supported (use @code{0x1234}
12865 #pragma ADDRESS port3 0x103
12872 @subsection MeP Pragmas
12876 @item custom io_volatile (on|off)
12877 @cindex pragma, custom io_volatile
12878 Overrides the command line option @code{-mio-volatile} for the current
12879 file. Note that for compatibility with future GCC releases, this
12880 option should only be used once before any @code{io} variables in each
12883 @item GCC coprocessor available @var{registers}
12884 @cindex pragma, coprocessor available
12885 Specifies which coprocessor registers are available to the register
12886 allocator. @var{registers} may be a single register, register range
12887 separated by ellipses, or comma-separated list of those. Example:
12890 #pragma GCC coprocessor available $c0...$c10, $c28
12893 @item GCC coprocessor call_saved @var{registers}
12894 @cindex pragma, coprocessor call_saved
12895 Specifies which coprocessor registers are to be saved and restored by
12896 any function using them. @var{registers} may be a single register,
12897 register range separated by ellipses, or comma-separated list of
12901 #pragma GCC coprocessor call_saved $c4...$c6, $c31
12904 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
12905 @cindex pragma, coprocessor subclass
12906 Creates and defines a register class. These register classes can be
12907 used by inline @code{asm} constructs. @var{registers} may be a single
12908 register, register range separated by ellipses, or comma-separated
12909 list of those. Example:
12912 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
12914 asm ("cpfoo %0" : "=B" (x));
12917 @item GCC disinterrupt @var{name} , @var{name} @dots{}
12918 @cindex pragma, disinterrupt
12919 For the named functions, the compiler adds code to disable interrupts
12920 for the duration of those functions. Any functions so named, which
12921 are not encountered in the source, cause a warning that the pragma was
12922 not used. Examples:
12925 #pragma disinterrupt foo
12926 #pragma disinterrupt bar, grill
12927 int foo () @{ @dots{} @}
12930 @item GCC call @var{name} , @var{name} @dots{}
12931 @cindex pragma, call
12932 For the named functions, the compiler always uses a register-indirect
12933 call model when calling the named functions. Examples:
12942 @node RS/6000 and PowerPC Pragmas
12943 @subsection RS/6000 and PowerPC Pragmas
12945 The RS/6000 and PowerPC targets define one pragma for controlling
12946 whether or not the @code{longcall} attribute is added to function
12947 declarations by default. This pragma overrides the @option{-mlongcall}
12948 option, but not the @code{longcall} and @code{shortcall} attributes.
12949 @xref{RS/6000 and PowerPC Options}, for more information about when long
12950 calls are and are not necessary.
12954 @cindex pragma, longcall
12955 Apply the @code{longcall} attribute to all subsequent function
12959 Do not apply the @code{longcall} attribute to subsequent function
12963 @c Describe h8300 pragmas here.
12964 @c Describe sh pragmas here.
12965 @c Describe v850 pragmas here.
12967 @node Darwin Pragmas
12968 @subsection Darwin Pragmas
12970 The following pragmas are available for all architectures running the
12971 Darwin operating system. These are useful for compatibility with other
12975 @item mark @var{tokens}@dots{}
12976 @cindex pragma, mark
12977 This pragma is accepted, but has no effect.
12979 @item options align=@var{alignment}
12980 @cindex pragma, options align
12981 This pragma sets the alignment of fields in structures. The values of
12982 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
12983 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
12984 properly; to restore the previous setting, use @code{reset} for the
12987 @item segment @var{tokens}@dots{}
12988 @cindex pragma, segment
12989 This pragma is accepted, but has no effect.
12991 @item unused (@var{var} [, @var{var}]@dots{})
12992 @cindex pragma, unused
12993 This pragma declares variables to be possibly unused. GCC will not
12994 produce warnings for the listed variables. The effect is similar to
12995 that of the @code{unused} attribute, except that this pragma may appear
12996 anywhere within the variables' scopes.
12999 @node Solaris Pragmas
13000 @subsection Solaris Pragmas
13002 The Solaris target supports @code{#pragma redefine_extname}
13003 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
13004 @code{#pragma} directives for compatibility with the system compiler.
13007 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
13008 @cindex pragma, align
13010 Increase the minimum alignment of each @var{variable} to @var{alignment}.
13011 This is the same as GCC's @code{aligned} attribute @pxref{Variable
13012 Attributes}). Macro expansion occurs on the arguments to this pragma
13013 when compiling C and Objective-C@. It does not currently occur when
13014 compiling C++, but this is a bug which may be fixed in a future
13017 @item fini (@var{function} [, @var{function}]...)
13018 @cindex pragma, fini
13020 This pragma causes each listed @var{function} to be called after
13021 main, or during shared module unloading, by adding a call to the
13022 @code{.fini} section.
13024 @item init (@var{function} [, @var{function}]...)
13025 @cindex pragma, init
13027 This pragma causes each listed @var{function} to be called during
13028 initialization (before @code{main}) or during shared module loading, by
13029 adding a call to the @code{.init} section.
13033 @node Symbol-Renaming Pragmas
13034 @subsection Symbol-Renaming Pragmas
13036 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
13037 supports two @code{#pragma} directives which change the name used in
13038 assembly for a given declaration. @code{#pragma extern_prefix} is only
13039 available on platforms whose system headers need it. To get this effect
13040 on all platforms supported by GCC, use the asm labels extension (@pxref{Asm
13044 @item redefine_extname @var{oldname} @var{newname}
13045 @cindex pragma, redefine_extname
13047 This pragma gives the C function @var{oldname} the assembly symbol
13048 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
13049 will be defined if this pragma is available (currently on all platforms).
13051 @item extern_prefix @var{string}
13052 @cindex pragma, extern_prefix
13054 This pragma causes all subsequent external function and variable
13055 declarations to have @var{string} prepended to their assembly symbols.
13056 This effect may be terminated with another @code{extern_prefix} pragma
13057 whose argument is an empty string. The preprocessor macro
13058 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
13059 available (currently only on Tru64 UNIX)@.
13062 These pragmas and the asm labels extension interact in a complicated
13063 manner. Here are some corner cases you may want to be aware of.
13066 @item Both pragmas silently apply only to declarations with external
13067 linkage. Asm labels do not have this restriction.
13069 @item In C++, both pragmas silently apply only to declarations with
13070 ``C'' linkage. Again, asm labels do not have this restriction.
13072 @item If any of the three ways of changing the assembly name of a
13073 declaration is applied to a declaration whose assembly name has
13074 already been determined (either by a previous use of one of these
13075 features, or because the compiler needed the assembly name in order to
13076 generate code), and the new name is different, a warning issues and
13077 the name does not change.
13079 @item The @var{oldname} used by @code{#pragma redefine_extname} is
13080 always the C-language name.
13082 @item If @code{#pragma extern_prefix} is in effect, and a declaration
13083 occurs with an asm label attached, the prefix is silently ignored for
13086 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
13087 apply to the same declaration, whichever triggered first wins, and a
13088 warning issues if they contradict each other. (We would like to have
13089 @code{#pragma redefine_extname} always win, for consistency with asm
13090 labels, but if @code{#pragma extern_prefix} triggers first we have no
13091 way of knowing that that happened.)
13094 @node Structure-Packing Pragmas
13095 @subsection Structure-Packing Pragmas
13097 For compatibility with Microsoft Windows compilers, GCC supports a
13098 set of @code{#pragma} directives which change the maximum alignment of
13099 members of structures (other than zero-width bitfields), unions, and
13100 classes subsequently defined. The @var{n} value below always is required
13101 to be a small power of two and specifies the new alignment in bytes.
13104 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
13105 @item @code{#pragma pack()} sets the alignment to the one that was in
13106 effect when compilation started (see also command-line option
13107 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
13108 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
13109 setting on an internal stack and then optionally sets the new alignment.
13110 @item @code{#pragma pack(pop)} restores the alignment setting to the one
13111 saved at the top of the internal stack (and removes that stack entry).
13112 Note that @code{#pragma pack([@var{n}])} does not influence this internal
13113 stack; thus it is possible to have @code{#pragma pack(push)} followed by
13114 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
13115 @code{#pragma pack(pop)}.
13118 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
13119 @code{#pragma} which lays out a structure as the documented
13120 @code{__attribute__ ((ms_struct))}.
13122 @item @code{#pragma ms_struct on} turns on the layout for structures
13124 @item @code{#pragma ms_struct off} turns off the layout for structures
13126 @item @code{#pragma ms_struct reset} goes back to the default layout.
13130 @subsection Weak Pragmas
13132 For compatibility with SVR4, GCC supports a set of @code{#pragma}
13133 directives for declaring symbols to be weak, and defining weak
13137 @item #pragma weak @var{symbol}
13138 @cindex pragma, weak
13139 This pragma declares @var{symbol} to be weak, as if the declaration
13140 had the attribute of the same name. The pragma may appear before
13141 or after the declaration of @var{symbol}. It is not an error for
13142 @var{symbol} to never be defined at all.
13144 @item #pragma weak @var{symbol1} = @var{symbol2}
13145 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
13146 It is an error if @var{symbol2} is not defined in the current
13150 @node Diagnostic Pragmas
13151 @subsection Diagnostic Pragmas
13153 GCC allows the user to selectively enable or disable certain types of
13154 diagnostics, and change the kind of the diagnostic. For example, a
13155 project's policy might require that all sources compile with
13156 @option{-Werror} but certain files might have exceptions allowing
13157 specific types of warnings. Or, a project might selectively enable
13158 diagnostics and treat them as errors depending on which preprocessor
13159 macros are defined.
13162 @item #pragma GCC diagnostic @var{kind} @var{option}
13163 @cindex pragma, diagnostic
13165 Modifies the disposition of a diagnostic. Note that not all
13166 diagnostics are modifiable; at the moment only warnings (normally
13167 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
13168 Use @option{-fdiagnostics-show-option} to determine which diagnostics
13169 are controllable and which option controls them.
13171 @var{kind} is @samp{error} to treat this diagnostic as an error,
13172 @samp{warning} to treat it like a warning (even if @option{-Werror} is
13173 in effect), or @samp{ignored} if the diagnostic is to be ignored.
13174 @var{option} is a double quoted string which matches the command-line
13178 #pragma GCC diagnostic warning "-Wformat"
13179 #pragma GCC diagnostic error "-Wformat"
13180 #pragma GCC diagnostic ignored "-Wformat"
13183 Note that these pragmas override any command-line options. GCC keeps
13184 track of the location of each pragma, and issues diagnostics according
13185 to the state as of that point in the source file. Thus, pragmas occurring
13186 after a line do not affect diagnostics caused by that line.
13188 @item #pragma GCC diagnostic push
13189 @itemx #pragma GCC diagnostic pop
13191 Causes GCC to remember the state of the diagnostics as of each
13192 @code{push}, and restore to that point at each @code{pop}. If a
13193 @code{pop} has no matching @code{push}, the command line options are
13197 #pragma GCC diagnostic error "-Wuninitialized"
13198 foo(a); /* error is given for this one */
13199 #pragma GCC diagnostic push
13200 #pragma GCC diagnostic ignored "-Wuninitialized"
13201 foo(b); /* no diagnostic for this one */
13202 #pragma GCC diagnostic pop
13203 foo(c); /* error is given for this one */
13204 #pragma GCC diagnostic pop
13205 foo(d); /* depends on command line options */
13210 GCC also offers a simple mechanism for printing messages during
13214 @item #pragma message @var{string}
13215 @cindex pragma, diagnostic
13217 Prints @var{string} as a compiler message on compilation. The message
13218 is informational only, and is neither a compilation warning nor an error.
13221 #pragma message "Compiling " __FILE__ "..."
13224 @var{string} may be parenthesized, and is printed with location
13225 information. For example,
13228 #define DO_PRAGMA(x) _Pragma (#x)
13229 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
13231 TODO(Remember to fix this)
13234 prints @samp{/tmp/file.c:4: note: #pragma message:
13235 TODO - Remember to fix this}.
13239 @node Visibility Pragmas
13240 @subsection Visibility Pragmas
13243 @item #pragma GCC visibility push(@var{visibility})
13244 @itemx #pragma GCC visibility pop
13245 @cindex pragma, visibility
13247 This pragma allows the user to set the visibility for multiple
13248 declarations without having to give each a visibility attribute
13249 @xref{Function Attributes}, for more information about visibility and
13250 the attribute syntax.
13252 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
13253 declarations. Class members and template specializations are not
13254 affected; if you want to override the visibility for a particular
13255 member or instantiation, you must use an attribute.
13260 @node Push/Pop Macro Pragmas
13261 @subsection Push/Pop Macro Pragmas
13263 For compatibility with Microsoft Windows compilers, GCC supports
13264 @samp{#pragma push_macro(@var{"macro_name"})}
13265 and @samp{#pragma pop_macro(@var{"macro_name"})}.
13268 @item #pragma push_macro(@var{"macro_name"})
13269 @cindex pragma, push_macro
13270 This pragma saves the value of the macro named as @var{macro_name} to
13271 the top of the stack for this macro.
13273 @item #pragma pop_macro(@var{"macro_name"})
13274 @cindex pragma, pop_macro
13275 This pragma sets the value of the macro named as @var{macro_name} to
13276 the value on top of the stack for this macro. If the stack for
13277 @var{macro_name} is empty, the value of the macro remains unchanged.
13284 #pragma push_macro("X")
13287 #pragma pop_macro("X")
13291 In this example, the definition of X as 1 is saved by @code{#pragma
13292 push_macro} and restored by @code{#pragma pop_macro}.
13294 @node Function Specific Option Pragmas
13295 @subsection Function Specific Option Pragmas
13298 @item #pragma GCC target (@var{"string"}...)
13299 @cindex pragma GCC target
13301 This pragma allows you to set target specific options for functions
13302 defined later in the source file. One or more strings can be
13303 specified. Each function that is defined after this point will be as
13304 if @code{attribute((target("STRING")))} was specified for that
13305 function. The parenthesis around the options is optional.
13306 @xref{Function Attributes}, for more information about the
13307 @code{target} attribute and the attribute syntax.
13309 The @code{#pragma GCC target} attribute is not implemented in GCC versions earlier
13310 than 4.4 for the i386/x86_64 and 4.6 for the PowerPC backends. At
13311 present, it is not implemented for other backends.
13315 @item #pragma GCC optimize (@var{"string"}...)
13316 @cindex pragma GCC optimize
13318 This pragma allows you to set global optimization options for functions
13319 defined later in the source file. One or more strings can be
13320 specified. Each function that is defined after this point will be as
13321 if @code{attribute((optimize("STRING")))} was specified for that
13322 function. The parenthesis around the options is optional.
13323 @xref{Function Attributes}, for more information about the
13324 @code{optimize} attribute and the attribute syntax.
13326 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
13327 versions earlier than 4.4.
13331 @item #pragma GCC push_options
13332 @itemx #pragma GCC pop_options
13333 @cindex pragma GCC push_options
13334 @cindex pragma GCC pop_options
13336 These pragmas maintain a stack of the current target and optimization
13337 options. It is intended for include files where you temporarily want
13338 to switch to using a different @samp{#pragma GCC target} or
13339 @samp{#pragma GCC optimize} and then to pop back to the previous
13342 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
13343 pragmas are not implemented in GCC versions earlier than 4.4.
13347 @item #pragma GCC reset_options
13348 @cindex pragma GCC reset_options
13350 This pragma clears the current @code{#pragma GCC target} and
13351 @code{#pragma GCC optimize} to use the default switches as specified
13352 on the command line.
13354 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
13355 versions earlier than 4.4.
13358 @node Unnamed Fields
13359 @section Unnamed struct/union fields within structs/unions
13360 @cindex @code{struct}
13361 @cindex @code{union}
13363 As permitted by ISO C1X and for compatibility with other compilers,
13364 GCC allows you to define
13365 a structure or union that contains, as fields, structures and unions
13366 without names. For example:
13379 In this example, the user would be able to access members of the unnamed
13380 union with code like @samp{foo.b}. Note that only unnamed structs and
13381 unions are allowed, you may not have, for example, an unnamed
13384 You must never create such structures that cause ambiguous field definitions.
13385 For example, this structure:
13396 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
13397 The compiler gives errors for such constructs.
13399 @opindex fms-extensions
13400 Unless @option{-fms-extensions} is used, the unnamed field must be a
13401 structure or union definition without a tag (for example, @samp{struct
13402 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
13403 also be a definition with a tag such as @samp{struct foo @{ int a;
13404 @};}, a reference to a previously defined structure or union such as
13405 @samp{struct foo;}, or a reference to a @code{typedef} name for a
13406 previously defined structure or union type.
13408 @opindex fplan9-extensions
13409 The option @option{-fplan9-extensions} enables
13410 @option{-fms-extensions} as well as two other extensions. First, a
13411 pointer to a structure is automatically converted to a pointer to an
13412 anonymous field for assignments and function calls. For example:
13415 struct s1 @{ int a; @};
13416 struct s2 @{ struct s1; @};
13417 extern void f1 (struct s1 *);
13418 void f2 (struct s2 *p) @{ f1 (p); @}
13421 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
13422 converted into a pointer to the anonymous field.
13424 Second, when the type of an anonymous field is a @code{typedef} for a
13425 @code{struct} or @code{union}, code may refer to the field using the
13426 name of the @code{typedef}.
13429 typedef struct @{ int a; @} s1;
13430 struct s2 @{ s1; @};
13431 s1 f1 (struct s2 *p) @{ return p->s1; @}
13434 These usages are only permitted when they are not ambiguous.
13437 @section Thread-Local Storage
13438 @cindex Thread-Local Storage
13439 @cindex @acronym{TLS}
13440 @cindex @code{__thread}
13442 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
13443 are allocated such that there is one instance of the variable per extant
13444 thread. The run-time model GCC uses to implement this originates
13445 in the IA-64 processor-specific ABI, but has since been migrated
13446 to other processors as well. It requires significant support from
13447 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
13448 system libraries (@file{libc.so} and @file{libpthread.so}), so it
13449 is not available everywhere.
13451 At the user level, the extension is visible with a new storage
13452 class keyword: @code{__thread}. For example:
13456 extern __thread struct state s;
13457 static __thread char *p;
13460 The @code{__thread} specifier may be used alone, with the @code{extern}
13461 or @code{static} specifiers, but with no other storage class specifier.
13462 When used with @code{extern} or @code{static}, @code{__thread} must appear
13463 immediately after the other storage class specifier.
13465 The @code{__thread} specifier may be applied to any global, file-scoped
13466 static, function-scoped static, or static data member of a class. It may
13467 not be applied to block-scoped automatic or non-static data member.
13469 When the address-of operator is applied to a thread-local variable, it is
13470 evaluated at run-time and returns the address of the current thread's
13471 instance of that variable. An address so obtained may be used by any
13472 thread. When a thread terminates, any pointers to thread-local variables
13473 in that thread become invalid.
13475 No static initialization may refer to the address of a thread-local variable.
13477 In C++, if an initializer is present for a thread-local variable, it must
13478 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
13481 See @uref{http://www.akkadia.org/drepper/tls.pdf,
13482 ELF Handling For Thread-Local Storage} for a detailed explanation of
13483 the four thread-local storage addressing models, and how the run-time
13484 is expected to function.
13487 * C99 Thread-Local Edits::
13488 * C++98 Thread-Local Edits::
13491 @node C99 Thread-Local Edits
13492 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
13494 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
13495 that document the exact semantics of the language extension.
13499 @cite{5.1.2 Execution environments}
13501 Add new text after paragraph 1
13504 Within either execution environment, a @dfn{thread} is a flow of
13505 control within a program. It is implementation defined whether
13506 or not there may be more than one thread associated with a program.
13507 It is implementation defined how threads beyond the first are
13508 created, the name and type of the function called at thread
13509 startup, and how threads may be terminated. However, objects
13510 with thread storage duration shall be initialized before thread
13515 @cite{6.2.4 Storage durations of objects}
13517 Add new text before paragraph 3
13520 An object whose identifier is declared with the storage-class
13521 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
13522 Its lifetime is the entire execution of the thread, and its
13523 stored value is initialized only once, prior to thread startup.
13527 @cite{6.4.1 Keywords}
13529 Add @code{__thread}.
13532 @cite{6.7.1 Storage-class specifiers}
13534 Add @code{__thread} to the list of storage class specifiers in
13537 Change paragraph 2 to
13540 With the exception of @code{__thread}, at most one storage-class
13541 specifier may be given [@dots{}]. The @code{__thread} specifier may
13542 be used alone, or immediately following @code{extern} or
13546 Add new text after paragraph 6
13549 The declaration of an identifier for a variable that has
13550 block scope that specifies @code{__thread} shall also
13551 specify either @code{extern} or @code{static}.
13553 The @code{__thread} specifier shall be used only with
13558 @node C++98 Thread-Local Edits
13559 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
13561 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
13562 that document the exact semantics of the language extension.
13566 @b{[intro.execution]}
13568 New text after paragraph 4
13571 A @dfn{thread} is a flow of control within the abstract machine.
13572 It is implementation defined whether or not there may be more than
13576 New text after paragraph 7
13579 It is unspecified whether additional action must be taken to
13580 ensure when and whether side effects are visible to other threads.
13586 Add @code{__thread}.
13589 @b{[basic.start.main]}
13591 Add after paragraph 5
13594 The thread that begins execution at the @code{main} function is called
13595 the @dfn{main thread}. It is implementation defined how functions
13596 beginning threads other than the main thread are designated or typed.
13597 A function so designated, as well as the @code{main} function, is called
13598 a @dfn{thread startup function}. It is implementation defined what
13599 happens if a thread startup function returns. It is implementation
13600 defined what happens to other threads when any thread calls @code{exit}.
13604 @b{[basic.start.init]}
13606 Add after paragraph 4
13609 The storage for an object of thread storage duration shall be
13610 statically initialized before the first statement of the thread startup
13611 function. An object of thread storage duration shall not require
13612 dynamic initialization.
13616 @b{[basic.start.term]}
13618 Add after paragraph 3
13621 The type of an object with thread storage duration shall not have a
13622 non-trivial destructor, nor shall it be an array type whose elements
13623 (directly or indirectly) have non-trivial destructors.
13629 Add ``thread storage duration'' to the list in paragraph 1.
13634 Thread, static, and automatic storage durations are associated with
13635 objects introduced by declarations [@dots{}].
13638 Add @code{__thread} to the list of specifiers in paragraph 3.
13641 @b{[basic.stc.thread]}
13643 New section before @b{[basic.stc.static]}
13646 The keyword @code{__thread} applied to a non-local object gives the
13647 object thread storage duration.
13649 A local variable or class data member declared both @code{static}
13650 and @code{__thread} gives the variable or member thread storage
13655 @b{[basic.stc.static]}
13660 All objects which have neither thread storage duration, dynamic
13661 storage duration nor are local [@dots{}].
13667 Add @code{__thread} to the list in paragraph 1.
13672 With the exception of @code{__thread}, at most one
13673 @var{storage-class-specifier} shall appear in a given
13674 @var{decl-specifier-seq}. The @code{__thread} specifier may
13675 be used alone, or immediately following the @code{extern} or
13676 @code{static} specifiers. [@dots{}]
13679 Add after paragraph 5
13682 The @code{__thread} specifier can be applied only to the names of objects
13683 and to anonymous unions.
13689 Add after paragraph 6
13692 Non-@code{static} members shall not be @code{__thread}.
13696 @node Binary constants
13697 @section Binary constants using the @samp{0b} prefix
13698 @cindex Binary constants using the @samp{0b} prefix
13700 Integer constants can be written as binary constants, consisting of a
13701 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
13702 @samp{0B}. This is particularly useful in environments that operate a
13703 lot on the bit-level (like microcontrollers).
13705 The following statements are identical:
13714 The type of these constants follows the same rules as for octal or
13715 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
13718 @node C++ Extensions
13719 @chapter Extensions to the C++ Language
13720 @cindex extensions, C++ language
13721 @cindex C++ language extensions
13723 The GNU compiler provides these extensions to the C++ language (and you
13724 can also use most of the C language extensions in your C++ programs). If you
13725 want to write code that checks whether these features are available, you can
13726 test for the GNU compiler the same way as for C programs: check for a
13727 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
13728 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
13729 Predefined Macros,cpp,The GNU C Preprocessor}).
13732 * C++ Volatiles:: What constitutes an access to a volatile object.
13733 * Restricted Pointers:: C99 restricted pointers and references.
13734 * Vague Linkage:: Where G++ puts inlines, vtables and such.
13735 * C++ Interface:: You can use a single C++ header file for both
13736 declarations and definitions.
13737 * Template Instantiation:: Methods for ensuring that exactly one copy of
13738 each needed template instantiation is emitted.
13739 * Bound member functions:: You can extract a function pointer to the
13740 method denoted by a @samp{->*} or @samp{.*} expression.
13741 * C++ Attributes:: Variable, function, and type attributes for C++ only.
13742 * Namespace Association:: Strong using-directives for namespace association.
13743 * Type Traits:: Compiler support for type traits
13744 * Java Exceptions:: Tweaking exception handling to work with Java.
13745 * Deprecated Features:: Things will disappear from g++.
13746 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
13749 @node C++ Volatiles
13750 @section When is a Volatile C++ Object Accessed?
13751 @cindex accessing volatiles
13752 @cindex volatile read
13753 @cindex volatile write
13754 @cindex volatile access
13756 The C++ standard differs from the C standard in its treatment of
13757 volatile objects. It fails to specify what constitutes a volatile
13758 access, except to say that C++ should behave in a similar manner to C
13759 with respect to volatiles, where possible. However, the different
13760 lvalueness of expressions between C and C++ complicate the behavior.
13761 G++ behaves the same as GCC for volatile access, @xref{C
13762 Extensions,,Volatiles}, for a description of GCC's behavior.
13764 The C and C++ language specifications differ when an object is
13765 accessed in a void context:
13768 volatile int *src = @var{somevalue};
13772 The C++ standard specifies that such expressions do not undergo lvalue
13773 to rvalue conversion, and that the type of the dereferenced object may
13774 be incomplete. The C++ standard does not specify explicitly that it
13775 is lvalue to rvalue conversion which is responsible for causing an
13776 access. There is reason to believe that it is, because otherwise
13777 certain simple expressions become undefined. However, because it
13778 would surprise most programmers, G++ treats dereferencing a pointer to
13779 volatile object of complete type as GCC would do for an equivalent
13780 type in C@. When the object has incomplete type, G++ issues a
13781 warning; if you wish to force an error, you must force a conversion to
13782 rvalue with, for instance, a static cast.
13784 When using a reference to volatile, G++ does not treat equivalent
13785 expressions as accesses to volatiles, but instead issues a warning that
13786 no volatile is accessed. The rationale for this is that otherwise it
13787 becomes difficult to determine where volatile access occur, and not
13788 possible to ignore the return value from functions returning volatile
13789 references. Again, if you wish to force a read, cast the reference to
13792 G++ implements the same behavior as GCC does when assigning to a
13793 volatile object -- there is no reread of the assigned-to object, the
13794 assigned rvalue is reused. Note that in C++ assignment expressions
13795 are lvalues, and if used as an lvalue, the volatile object will be
13796 referred to. For instance, @var{vref} will refer to @var{vobj}, as
13797 expected, in the following example:
13801 volatile int &vref = vobj = @var{something};
13804 @node Restricted Pointers
13805 @section Restricting Pointer Aliasing
13806 @cindex restricted pointers
13807 @cindex restricted references
13808 @cindex restricted this pointer
13810 As with the C front end, G++ understands the C99 feature of restricted pointers,
13811 specified with the @code{__restrict__}, or @code{__restrict} type
13812 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
13813 language flag, @code{restrict} is not a keyword in C++.
13815 In addition to allowing restricted pointers, you can specify restricted
13816 references, which indicate that the reference is not aliased in the local
13820 void fn (int *__restrict__ rptr, int &__restrict__ rref)
13827 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
13828 @var{rref} refers to a (different) unaliased integer.
13830 You may also specify whether a member function's @var{this} pointer is
13831 unaliased by using @code{__restrict__} as a member function qualifier.
13834 void T::fn () __restrict__
13841 Within the body of @code{T::fn}, @var{this} will have the effective
13842 definition @code{T *__restrict__ const this}. Notice that the
13843 interpretation of a @code{__restrict__} member function qualifier is
13844 different to that of @code{const} or @code{volatile} qualifier, in that it
13845 is applied to the pointer rather than the object. This is consistent with
13846 other compilers which implement restricted pointers.
13848 As with all outermost parameter qualifiers, @code{__restrict__} is
13849 ignored in function definition matching. This means you only need to
13850 specify @code{__restrict__} in a function definition, rather than
13851 in a function prototype as well.
13853 @node Vague Linkage
13854 @section Vague Linkage
13855 @cindex vague linkage
13857 There are several constructs in C++ which require space in the object
13858 file but are not clearly tied to a single translation unit. We say that
13859 these constructs have ``vague linkage''. Typically such constructs are
13860 emitted wherever they are needed, though sometimes we can be more
13864 @item Inline Functions
13865 Inline functions are typically defined in a header file which can be
13866 included in many different compilations. Hopefully they can usually be
13867 inlined, but sometimes an out-of-line copy is necessary, if the address
13868 of the function is taken or if inlining fails. In general, we emit an
13869 out-of-line copy in all translation units where one is needed. As an
13870 exception, we only emit inline virtual functions with the vtable, since
13871 it will always require a copy.
13873 Local static variables and string constants used in an inline function
13874 are also considered to have vague linkage, since they must be shared
13875 between all inlined and out-of-line instances of the function.
13879 C++ virtual functions are implemented in most compilers using a lookup
13880 table, known as a vtable. The vtable contains pointers to the virtual
13881 functions provided by a class, and each object of the class contains a
13882 pointer to its vtable (or vtables, in some multiple-inheritance
13883 situations). If the class declares any non-inline, non-pure virtual
13884 functions, the first one is chosen as the ``key method'' for the class,
13885 and the vtable is only emitted in the translation unit where the key
13888 @emph{Note:} If the chosen key method is later defined as inline, the
13889 vtable will still be emitted in every translation unit which defines it.
13890 Make sure that any inline virtuals are declared inline in the class
13891 body, even if they are not defined there.
13893 @item @code{type_info} objects
13894 @cindex @code{type_info}
13896 C++ requires information about types to be written out in order to
13897 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
13898 For polymorphic classes (classes with virtual functions), the @samp{type_info}
13899 object is written out along with the vtable so that @samp{dynamic_cast}
13900 can determine the dynamic type of a class object at runtime. For all
13901 other types, we write out the @samp{type_info} object when it is used: when
13902 applying @samp{typeid} to an expression, throwing an object, or
13903 referring to a type in a catch clause or exception specification.
13905 @item Template Instantiations
13906 Most everything in this section also applies to template instantiations,
13907 but there are other options as well.
13908 @xref{Template Instantiation,,Where's the Template?}.
13912 When used with GNU ld version 2.8 or later on an ELF system such as
13913 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
13914 these constructs will be discarded at link time. This is known as
13917 On targets that don't support COMDAT, but do support weak symbols, GCC
13918 will use them. This way one copy will override all the others, but
13919 the unused copies will still take up space in the executable.
13921 For targets which do not support either COMDAT or weak symbols,
13922 most entities with vague linkage will be emitted as local symbols to
13923 avoid duplicate definition errors from the linker. This will not happen
13924 for local statics in inlines, however, as having multiple copies will
13925 almost certainly break things.
13927 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
13928 another way to control placement of these constructs.
13930 @node C++ Interface
13931 @section #pragma interface and implementation
13933 @cindex interface and implementation headers, C++
13934 @cindex C++ interface and implementation headers
13935 @cindex pragmas, interface and implementation
13937 @code{#pragma interface} and @code{#pragma implementation} provide the
13938 user with a way of explicitly directing the compiler to emit entities
13939 with vague linkage (and debugging information) in a particular
13942 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
13943 most cases, because of COMDAT support and the ``key method'' heuristic
13944 mentioned in @ref{Vague Linkage}. Using them can actually cause your
13945 program to grow due to unnecessary out-of-line copies of inline
13946 functions. Currently (3.4) the only benefit of these
13947 @code{#pragma}s is reduced duplication of debugging information, and
13948 that should be addressed soon on DWARF 2 targets with the use of
13952 @item #pragma interface
13953 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
13954 @kindex #pragma interface
13955 Use this directive in @emph{header files} that define object classes, to save
13956 space in most of the object files that use those classes. Normally,
13957 local copies of certain information (backup copies of inline member
13958 functions, debugging information, and the internal tables that implement
13959 virtual functions) must be kept in each object file that includes class
13960 definitions. You can use this pragma to avoid such duplication. When a
13961 header file containing @samp{#pragma interface} is included in a
13962 compilation, this auxiliary information will not be generated (unless
13963 the main input source file itself uses @samp{#pragma implementation}).
13964 Instead, the object files will contain references to be resolved at link
13967 The second form of this directive is useful for the case where you have
13968 multiple headers with the same name in different directories. If you
13969 use this form, you must specify the same string to @samp{#pragma
13972 @item #pragma implementation
13973 @itemx #pragma implementation "@var{objects}.h"
13974 @kindex #pragma implementation
13975 Use this pragma in a @emph{main input file}, when you want full output from
13976 included header files to be generated (and made globally visible). The
13977 included header file, in turn, should use @samp{#pragma interface}.
13978 Backup copies of inline member functions, debugging information, and the
13979 internal tables used to implement virtual functions are all generated in
13980 implementation files.
13982 @cindex implied @code{#pragma implementation}
13983 @cindex @code{#pragma implementation}, implied
13984 @cindex naming convention, implementation headers
13985 If you use @samp{#pragma implementation} with no argument, it applies to
13986 an include file with the same basename@footnote{A file's @dfn{basename}
13987 was the name stripped of all leading path information and of trailing
13988 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
13989 file. For example, in @file{allclass.cc}, giving just
13990 @samp{#pragma implementation}
13991 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
13993 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
13994 an implementation file whenever you would include it from
13995 @file{allclass.cc} even if you never specified @samp{#pragma
13996 implementation}. This was deemed to be more trouble than it was worth,
13997 however, and disabled.
13999 Use the string argument if you want a single implementation file to
14000 include code from multiple header files. (You must also use
14001 @samp{#include} to include the header file; @samp{#pragma
14002 implementation} only specifies how to use the file---it doesn't actually
14005 There is no way to split up the contents of a single header file into
14006 multiple implementation files.
14009 @cindex inlining and C++ pragmas
14010 @cindex C++ pragmas, effect on inlining
14011 @cindex pragmas in C++, effect on inlining
14012 @samp{#pragma implementation} and @samp{#pragma interface} also have an
14013 effect on function inlining.
14015 If you define a class in a header file marked with @samp{#pragma
14016 interface}, the effect on an inline function defined in that class is
14017 similar to an explicit @code{extern} declaration---the compiler emits
14018 no code at all to define an independent version of the function. Its
14019 definition is used only for inlining with its callers.
14021 @opindex fno-implement-inlines
14022 Conversely, when you include the same header file in a main source file
14023 that declares it as @samp{#pragma implementation}, the compiler emits
14024 code for the function itself; this defines a version of the function
14025 that can be found via pointers (or by callers compiled without
14026 inlining). If all calls to the function can be inlined, you can avoid
14027 emitting the function by compiling with @option{-fno-implement-inlines}.
14028 If any calls were not inlined, you will get linker errors.
14030 @node Template Instantiation
14031 @section Where's the Template?
14032 @cindex template instantiation
14034 C++ templates are the first language feature to require more
14035 intelligence from the environment than one usually finds on a UNIX
14036 system. Somehow the compiler and linker have to make sure that each
14037 template instance occurs exactly once in the executable if it is needed,
14038 and not at all otherwise. There are two basic approaches to this
14039 problem, which are referred to as the Borland model and the Cfront model.
14042 @item Borland model
14043 Borland C++ solved the template instantiation problem by adding the code
14044 equivalent of common blocks to their linker; the compiler emits template
14045 instances in each translation unit that uses them, and the linker
14046 collapses them together. The advantage of this model is that the linker
14047 only has to consider the object files themselves; there is no external
14048 complexity to worry about. This disadvantage is that compilation time
14049 is increased because the template code is being compiled repeatedly.
14050 Code written for this model tends to include definitions of all
14051 templates in the header file, since they must be seen to be
14055 The AT&T C++ translator, Cfront, solved the template instantiation
14056 problem by creating the notion of a template repository, an
14057 automatically maintained place where template instances are stored. A
14058 more modern version of the repository works as follows: As individual
14059 object files are built, the compiler places any template definitions and
14060 instantiations encountered in the repository. At link time, the link
14061 wrapper adds in the objects in the repository and compiles any needed
14062 instances that were not previously emitted. The advantages of this
14063 model are more optimal compilation speed and the ability to use the
14064 system linker; to implement the Borland model a compiler vendor also
14065 needs to replace the linker. The disadvantages are vastly increased
14066 complexity, and thus potential for error; for some code this can be
14067 just as transparent, but in practice it can been very difficult to build
14068 multiple programs in one directory and one program in multiple
14069 directories. Code written for this model tends to separate definitions
14070 of non-inline member templates into a separate file, which should be
14071 compiled separately.
14074 When used with GNU ld version 2.8 or later on an ELF system such as
14075 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
14076 Borland model. On other systems, G++ implements neither automatic
14079 A future version of G++ will support a hybrid model whereby the compiler
14080 will emit any instantiations for which the template definition is
14081 included in the compile, and store template definitions and
14082 instantiation context information into the object file for the rest.
14083 The link wrapper will extract that information as necessary and invoke
14084 the compiler to produce the remaining instantiations. The linker will
14085 then combine duplicate instantiations.
14087 In the mean time, you have the following options for dealing with
14088 template instantiations:
14093 Compile your template-using code with @option{-frepo}. The compiler will
14094 generate files with the extension @samp{.rpo} listing all of the
14095 template instantiations used in the corresponding object files which
14096 could be instantiated there; the link wrapper, @samp{collect2}, will
14097 then update the @samp{.rpo} files to tell the compiler where to place
14098 those instantiations and rebuild any affected object files. The
14099 link-time overhead is negligible after the first pass, as the compiler
14100 will continue to place the instantiations in the same files.
14102 This is your best option for application code written for the Borland
14103 model, as it will just work. Code written for the Cfront model will
14104 need to be modified so that the template definitions are available at
14105 one or more points of instantiation; usually this is as simple as adding
14106 @code{#include <tmethods.cc>} to the end of each template header.
14108 For library code, if you want the library to provide all of the template
14109 instantiations it needs, just try to link all of its object files
14110 together; the link will fail, but cause the instantiations to be
14111 generated as a side effect. Be warned, however, that this may cause
14112 conflicts if multiple libraries try to provide the same instantiations.
14113 For greater control, use explicit instantiation as described in the next
14117 @opindex fno-implicit-templates
14118 Compile your code with @option{-fno-implicit-templates} to disable the
14119 implicit generation of template instances, and explicitly instantiate
14120 all the ones you use. This approach requires more knowledge of exactly
14121 which instances you need than do the others, but it's less
14122 mysterious and allows greater control. You can scatter the explicit
14123 instantiations throughout your program, perhaps putting them in the
14124 translation units where the instances are used or the translation units
14125 that define the templates themselves; you can put all of the explicit
14126 instantiations you need into one big file; or you can create small files
14133 template class Foo<int>;
14134 template ostream& operator <<
14135 (ostream&, const Foo<int>&);
14138 for each of the instances you need, and create a template instantiation
14139 library from those.
14141 If you are using Cfront-model code, you can probably get away with not
14142 using @option{-fno-implicit-templates} when compiling files that don't
14143 @samp{#include} the member template definitions.
14145 If you use one big file to do the instantiations, you may want to
14146 compile it without @option{-fno-implicit-templates} so you get all of the
14147 instances required by your explicit instantiations (but not by any
14148 other files) without having to specify them as well.
14150 G++ has extended the template instantiation syntax given in the ISO
14151 standard to allow forward declaration of explicit instantiations
14152 (with @code{extern}), instantiation of the compiler support data for a
14153 template class (i.e.@: the vtable) without instantiating any of its
14154 members (with @code{inline}), and instantiation of only the static data
14155 members of a template class, without the support data or member
14156 functions (with (@code{static}):
14159 extern template int max (int, int);
14160 inline template class Foo<int>;
14161 static template class Foo<int>;
14165 Do nothing. Pretend G++ does implement automatic instantiation
14166 management. Code written for the Borland model will work fine, but
14167 each translation unit will contain instances of each of the templates it
14168 uses. In a large program, this can lead to an unacceptable amount of code
14172 @node Bound member functions
14173 @section Extracting the function pointer from a bound pointer to member function
14175 @cindex pointer to member function
14176 @cindex bound pointer to member function
14178 In C++, pointer to member functions (PMFs) are implemented using a wide
14179 pointer of sorts to handle all the possible call mechanisms; the PMF
14180 needs to store information about how to adjust the @samp{this} pointer,
14181 and if the function pointed to is virtual, where to find the vtable, and
14182 where in the vtable to look for the member function. If you are using
14183 PMFs in an inner loop, you should really reconsider that decision. If
14184 that is not an option, you can extract the pointer to the function that
14185 would be called for a given object/PMF pair and call it directly inside
14186 the inner loop, to save a bit of time.
14188 Note that you will still be paying the penalty for the call through a
14189 function pointer; on most modern architectures, such a call defeats the
14190 branch prediction features of the CPU@. This is also true of normal
14191 virtual function calls.
14193 The syntax for this extension is
14197 extern int (A::*fp)();
14198 typedef int (*fptr)(A *);
14200 fptr p = (fptr)(a.*fp);
14203 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
14204 no object is needed to obtain the address of the function. They can be
14205 converted to function pointers directly:
14208 fptr p1 = (fptr)(&A::foo);
14211 @opindex Wno-pmf-conversions
14212 You must specify @option{-Wno-pmf-conversions} to use this extension.
14214 @node C++ Attributes
14215 @section C++-Specific Variable, Function, and Type Attributes
14217 Some attributes only make sense for C++ programs.
14220 @item init_priority (@var{priority})
14221 @cindex @code{init_priority} attribute
14224 In Standard C++, objects defined at namespace scope are guaranteed to be
14225 initialized in an order in strict accordance with that of their definitions
14226 @emph{in a given translation unit}. No guarantee is made for initializations
14227 across translation units. However, GNU C++ allows users to control the
14228 order of initialization of objects defined at namespace scope with the
14229 @code{init_priority} attribute by specifying a relative @var{priority},
14230 a constant integral expression currently bounded between 101 and 65535
14231 inclusive. Lower numbers indicate a higher priority.
14233 In the following example, @code{A} would normally be created before
14234 @code{B}, but the @code{init_priority} attribute has reversed that order:
14237 Some_Class A __attribute__ ((init_priority (2000)));
14238 Some_Class B __attribute__ ((init_priority (543)));
14242 Note that the particular values of @var{priority} do not matter; only their
14245 @item java_interface
14246 @cindex @code{java_interface} attribute
14248 This type attribute informs C++ that the class is a Java interface. It may
14249 only be applied to classes declared within an @code{extern "Java"} block.
14250 Calls to methods declared in this interface will be dispatched using GCJ's
14251 interface table mechanism, instead of regular virtual table dispatch.
14255 See also @ref{Namespace Association}.
14257 @node Namespace Association
14258 @section Namespace Association
14260 @strong{Caution:} The semantics of this extension are not fully
14261 defined. Users should refrain from using this extension as its
14262 semantics may change subtly over time. It is possible that this
14263 extension will be removed in future versions of G++.
14265 A using-directive with @code{__attribute ((strong))} is stronger
14266 than a normal using-directive in two ways:
14270 Templates from the used namespace can be specialized and explicitly
14271 instantiated as though they were members of the using namespace.
14274 The using namespace is considered an associated namespace of all
14275 templates in the used namespace for purposes of argument-dependent
14279 The used namespace must be nested within the using namespace so that
14280 normal unqualified lookup works properly.
14282 This is useful for composing a namespace transparently from
14283 implementation namespaces. For example:
14288 template <class T> struct A @{ @};
14290 using namespace debug __attribute ((__strong__));
14291 template <> struct A<int> @{ @}; // @r{ok to specialize}
14293 template <class T> void f (A<T>);
14298 f (std::A<float>()); // @r{lookup finds} std::f
14304 @section Type Traits
14306 The C++ front-end implements syntactic extensions that allow to
14307 determine at compile time various characteristics of a type (or of a
14311 @item __has_nothrow_assign (type)
14312 If @code{type} is const qualified or is a reference type then the trait is
14313 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
14314 is true, else if @code{type} is a cv class or union type with copy assignment
14315 operators that are known not to throw an exception then the trait is true,
14316 else it is false. Requires: @code{type} shall be a complete type,
14317 (possibly cv-qualified) @code{void}, or an array of unknown bound.
14319 @item __has_nothrow_copy (type)
14320 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
14321 @code{type} is a cv class or union type with copy constructors that
14322 are known not to throw an exception then the trait is true, else it is false.
14323 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
14324 @code{void}, or an array of unknown bound.
14326 @item __has_nothrow_constructor (type)
14327 If @code{__has_trivial_constructor (type)} is true then the trait is
14328 true, else if @code{type} is a cv class or union type (or array
14329 thereof) with a default constructor that is known not to throw an
14330 exception then the trait is true, else it is false. Requires:
14331 @code{type} shall be a complete type, (possibly cv-qualified)
14332 @code{void}, or an array of unknown bound.
14334 @item __has_trivial_assign (type)
14335 If @code{type} is const qualified or is a reference type then the trait is
14336 false. Otherwise if @code{__is_pod (type)} is true then the trait is
14337 true, else if @code{type} is a cv class or union type with a trivial
14338 copy assignment ([class.copy]) then the trait is true, else it is
14339 false. Requires: @code{type} shall be a complete type, (possibly
14340 cv-qualified) @code{void}, or an array of unknown bound.
14342 @item __has_trivial_copy (type)
14343 If @code{__is_pod (type)} is true or @code{type} is a reference type
14344 then the trait is true, else if @code{type} is a cv class or union type
14345 with a trivial copy constructor ([class.copy]) then the trait
14346 is true, else it is false. Requires: @code{type} shall be a complete
14347 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14349 @item __has_trivial_constructor (type)
14350 If @code{__is_pod (type)} is true then the trait is true, else if
14351 @code{type} is a cv class or union type (or array thereof) with a
14352 trivial default constructor ([class.ctor]) then the trait is true,
14353 else it is false. Requires: @code{type} shall be a complete
14354 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14356 @item __has_trivial_destructor (type)
14357 If @code{__is_pod (type)} is true or @code{type} is a reference type then
14358 the trait is true, else if @code{type} is a cv class or union type (or
14359 array thereof) with a trivial destructor ([class.dtor]) then the trait
14360 is true, else it is false. Requires: @code{type} shall be a complete
14361 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14363 @item __has_virtual_destructor (type)
14364 If @code{type} is a class type with a virtual destructor
14365 ([class.dtor]) then the trait is true, else it is false. Requires:
14366 @code{type} shall be a complete type, (possibly cv-qualified)
14367 @code{void}, or an array of unknown bound.
14369 @item __is_abstract (type)
14370 If @code{type} is an abstract class ([class.abstract]) then the trait
14371 is true, else it is false. Requires: @code{type} shall be a complete
14372 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14374 @item __is_base_of (base_type, derived_type)
14375 If @code{base_type} is a base class of @code{derived_type}
14376 ([class.derived]) then the trait is true, otherwise it is false.
14377 Top-level cv qualifications of @code{base_type} and
14378 @code{derived_type} are ignored. For the purposes of this trait, a
14379 class type is considered is own base. Requires: if @code{__is_class
14380 (base_type)} and @code{__is_class (derived_type)} are true and
14381 @code{base_type} and @code{derived_type} are not the same type
14382 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
14383 type. Diagnostic is produced if this requirement is not met.
14385 @item __is_class (type)
14386 If @code{type} is a cv class type, and not a union type
14387 ([basic.compound]) the trait is true, else it is false.
14389 @item __is_empty (type)
14390 If @code{__is_class (type)} is false then the trait is false.
14391 Otherwise @code{type} is considered empty if and only if: @code{type}
14392 has no non-static data members, or all non-static data members, if
14393 any, are bit-fields of length 0, and @code{type} has no virtual
14394 members, and @code{type} has no virtual base classes, and @code{type}
14395 has no base classes @code{base_type} for which
14396 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
14397 be a complete type, (possibly cv-qualified) @code{void}, or an array
14400 @item __is_enum (type)
14401 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
14402 true, else it is false.
14404 @item __is_literal_type (type)
14405 If @code{type} is a literal type ([basic.types]) the trait is
14406 true, else it is false. Requires: @code{type} shall be a complete type,
14407 (possibly cv-qualified) @code{void}, or an array of unknown bound.
14409 @item __is_pod (type)
14410 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
14411 else it is false. Requires: @code{type} shall be a complete type,
14412 (possibly cv-qualified) @code{void}, or an array of unknown bound.
14414 @item __is_polymorphic (type)
14415 If @code{type} is a polymorphic class ([class.virtual]) then the trait
14416 is true, else it is false. Requires: @code{type} shall be a complete
14417 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14419 @item __is_standard_layout (type)
14420 If @code{type} is a standard-layout type ([basic.types]) the trait is
14421 true, else it is false. Requires: @code{type} shall be a complete
14422 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14424 @item __is_trivial (type)
14425 If @code{type} is a trivial type ([basic.types]) the trait is
14426 true, else it is false. Requires: @code{type} shall be a complete
14427 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
14429 @item __is_union (type)
14430 If @code{type} is a cv union type ([basic.compound]) the trait is
14431 true, else it is false.
14433 @item __underlying_type (type)
14434 The underlying type of @code{type}. Requires: @code{type} shall be
14435 an enumeration type ([dcl.enum]).
14439 @node Java Exceptions
14440 @section Java Exceptions
14442 The Java language uses a slightly different exception handling model
14443 from C++. Normally, GNU C++ will automatically detect when you are
14444 writing C++ code that uses Java exceptions, and handle them
14445 appropriately. However, if C++ code only needs to execute destructors
14446 when Java exceptions are thrown through it, GCC will guess incorrectly.
14447 Sample problematic code is:
14450 struct S @{ ~S(); @};
14451 extern void bar(); // @r{is written in Java, and may throw exceptions}
14460 The usual effect of an incorrect guess is a link failure, complaining of
14461 a missing routine called @samp{__gxx_personality_v0}.
14463 You can inform the compiler that Java exceptions are to be used in a
14464 translation unit, irrespective of what it might think, by writing
14465 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
14466 @samp{#pragma} must appear before any functions that throw or catch
14467 exceptions, or run destructors when exceptions are thrown through them.
14469 You cannot mix Java and C++ exceptions in the same translation unit. It
14470 is believed to be safe to throw a C++ exception from one file through
14471 another file compiled for the Java exception model, or vice versa, but
14472 there may be bugs in this area.
14474 @node Deprecated Features
14475 @section Deprecated Features
14477 In the past, the GNU C++ compiler was extended to experiment with new
14478 features, at a time when the C++ language was still evolving. Now that
14479 the C++ standard is complete, some of those features are superseded by
14480 superior alternatives. Using the old features might cause a warning in
14481 some cases that the feature will be dropped in the future. In other
14482 cases, the feature might be gone already.
14484 While the list below is not exhaustive, it documents some of the options
14485 that are now deprecated:
14488 @item -fexternal-templates
14489 @itemx -falt-external-templates
14490 These are two of the many ways for G++ to implement template
14491 instantiation. @xref{Template Instantiation}. The C++ standard clearly
14492 defines how template definitions have to be organized across
14493 implementation units. G++ has an implicit instantiation mechanism that
14494 should work just fine for standard-conforming code.
14496 @item -fstrict-prototype
14497 @itemx -fno-strict-prototype
14498 Previously it was possible to use an empty prototype parameter list to
14499 indicate an unspecified number of parameters (like C), rather than no
14500 parameters, as C++ demands. This feature has been removed, except where
14501 it is required for backwards compatibility. @xref{Backwards Compatibility}.
14504 G++ allows a virtual function returning @samp{void *} to be overridden
14505 by one returning a different pointer type. This extension to the
14506 covariant return type rules is now deprecated and will be removed from a
14509 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
14510 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
14511 and are now removed from G++. Code using these operators should be
14512 modified to use @code{std::min} and @code{std::max} instead.
14514 The named return value extension has been deprecated, and is now
14517 The use of initializer lists with new expressions has been deprecated,
14518 and is now removed from G++.
14520 Floating and complex non-type template parameters have been deprecated,
14521 and are now removed from G++.
14523 The implicit typename extension has been deprecated and is now
14526 The use of default arguments in function pointers, function typedefs
14527 and other places where they are not permitted by the standard is
14528 deprecated and will be removed from a future version of G++.
14530 G++ allows floating-point literals to appear in integral constant expressions,
14531 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
14532 This extension is deprecated and will be removed from a future version.
14534 G++ allows static data members of const floating-point type to be declared
14535 with an initializer in a class definition. The standard only allows
14536 initializers for static members of const integral types and const
14537 enumeration types so this extension has been deprecated and will be removed
14538 from a future version.
14540 @node Backwards Compatibility
14541 @section Backwards Compatibility
14542 @cindex Backwards Compatibility
14543 @cindex ARM [Annotated C++ Reference Manual]
14545 Now that there is a definitive ISO standard C++, G++ has a specification
14546 to adhere to. The C++ language evolved over time, and features that
14547 used to be acceptable in previous drafts of the standard, such as the ARM
14548 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
14549 compilation of C++ written to such drafts, G++ contains some backwards
14550 compatibilities. @emph{All such backwards compatibility features are
14551 liable to disappear in future versions of G++.} They should be considered
14552 deprecated. @xref{Deprecated Features}.
14556 If a variable is declared at for scope, it used to remain in scope until
14557 the end of the scope which contained the for statement (rather than just
14558 within the for scope). G++ retains this, but issues a warning, if such a
14559 variable is accessed outside the for scope.
14561 @item Implicit C language
14562 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
14563 scope to set the language. On such systems, all header files are
14564 implicitly scoped inside a C language scope. Also, an empty prototype
14565 @code{()} will be treated as an unspecified number of arguments, rather
14566 than no arguments, as C++ demands.