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
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.
939 @section Half-Precision Floating Point
940 @cindex half-precision floating point
941 @cindex @code{__fp16} data type
943 On ARM targets, GCC supports half-precision (16-bit) floating point via
944 the @code{__fp16} type. You must enable this type explicitly
945 with the @option{-mfp16-format} command-line option in order to use it.
947 ARM supports two incompatible representations for half-precision
948 floating-point values. You must choose one of the representations and
949 use it consistently in your program.
951 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
952 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
953 There are 11 bits of significand precision, approximately 3
956 Specifying @option{-mfp16-format=alternative} selects the ARM
957 alternative format. This representation is similar to the IEEE
958 format, but does not support infinities or NaNs. Instead, the range
959 of exponents is extended, so that this format can represent normalized
960 values in the range of @math{2^{-14}} to 131008.
962 The @code{__fp16} type is a storage format only. For purposes
963 of arithmetic and other operations, @code{__fp16} values in C or C++
964 expressions are automatically promoted to @code{float}. In addition,
965 you cannot declare a function with a return value or parameters
966 of type @code{__fp16}.
968 Note that conversions from @code{double} to @code{__fp16}
969 involve an intermediate conversion to @code{float}. Because
970 of rounding, this can sometimes produce a different result than a
973 ARM provides hardware support for conversions between
974 @code{__fp16} and @code{float} values
975 as an extension to VFP and NEON (Advanced SIMD). GCC generates
976 code using these hardware instructions if you compile with
977 options to select an FPU that provides them;
978 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
979 in addition to the @option{-mfp16-format} option to select
980 a half-precision format.
982 Language-level support for the @code{__fp16} data type is
983 independent of whether GCC generates code using hardware floating-point
984 instructions. In cases where hardware support is not specified, GCC
985 implements conversions between @code{__fp16} and @code{float} values
989 @section Decimal Floating Types
990 @cindex decimal floating types
991 @cindex @code{_Decimal32} data type
992 @cindex @code{_Decimal64} data type
993 @cindex @code{_Decimal128} data type
994 @cindex @code{df} integer suffix
995 @cindex @code{dd} integer suffix
996 @cindex @code{dl} integer suffix
997 @cindex @code{DF} integer suffix
998 @cindex @code{DD} integer suffix
999 @cindex @code{DL} integer suffix
1001 As an extension, the GNU C compiler supports decimal floating types as
1002 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1003 floating types in GCC will evolve as the draft technical report changes.
1004 Calling conventions for any target might also change. Not all targets
1005 support decimal floating types.
1007 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1008 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1009 @code{float}, @code{double}, and @code{long double} whose radix is not
1010 specified by the C standard but is usually two.
1012 Support for decimal floating types includes the arithmetic operators
1013 add, subtract, multiply, divide; unary arithmetic operators;
1014 relational operators; equality operators; and conversions to and from
1015 integer and other floating types. Use a suffix @samp{df} or
1016 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1017 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1020 GCC support of decimal float as specified by the draft technical report
1025 When the value of a decimal floating type cannot be represented in the
1026 integer type to which it is being converted, the result is undefined
1027 rather than the result value specified by the draft technical report.
1030 GCC does not provide the C library functionality associated with
1031 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1032 @file{wchar.h}, which must come from a separate C library implementation.
1033 Because of this the GNU C compiler does not define macro
1034 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1035 the technical report.
1038 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1039 are supported by the DWARF2 debug information format.
1045 ISO C99 supports floating-point numbers written not only in the usual
1046 decimal notation, such as @code{1.55e1}, but also numbers such as
1047 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1048 supports this in C90 mode (except in some cases when strictly
1049 conforming) and in C++. In that format the
1050 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1051 mandatory. The exponent is a decimal number that indicates the power of
1052 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1059 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1060 is the same as @code{1.55e1}.
1062 Unlike for floating-point numbers in the decimal notation the exponent
1063 is always required in the hexadecimal notation. Otherwise the compiler
1064 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1065 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1066 extension for floating-point constants of type @code{float}.
1069 @section Fixed-Point Types
1070 @cindex fixed-point types
1071 @cindex @code{_Fract} data type
1072 @cindex @code{_Accum} data type
1073 @cindex @code{_Sat} data type
1074 @cindex @code{hr} fixed-suffix
1075 @cindex @code{r} fixed-suffix
1076 @cindex @code{lr} fixed-suffix
1077 @cindex @code{llr} fixed-suffix
1078 @cindex @code{uhr} fixed-suffix
1079 @cindex @code{ur} fixed-suffix
1080 @cindex @code{ulr} fixed-suffix
1081 @cindex @code{ullr} fixed-suffix
1082 @cindex @code{hk} fixed-suffix
1083 @cindex @code{k} fixed-suffix
1084 @cindex @code{lk} fixed-suffix
1085 @cindex @code{llk} fixed-suffix
1086 @cindex @code{uhk} fixed-suffix
1087 @cindex @code{uk} fixed-suffix
1088 @cindex @code{ulk} fixed-suffix
1089 @cindex @code{ullk} fixed-suffix
1090 @cindex @code{HR} fixed-suffix
1091 @cindex @code{R} fixed-suffix
1092 @cindex @code{LR} fixed-suffix
1093 @cindex @code{LLR} fixed-suffix
1094 @cindex @code{UHR} fixed-suffix
1095 @cindex @code{UR} fixed-suffix
1096 @cindex @code{ULR} fixed-suffix
1097 @cindex @code{ULLR} fixed-suffix
1098 @cindex @code{HK} fixed-suffix
1099 @cindex @code{K} fixed-suffix
1100 @cindex @code{LK} fixed-suffix
1101 @cindex @code{LLK} fixed-suffix
1102 @cindex @code{UHK} fixed-suffix
1103 @cindex @code{UK} fixed-suffix
1104 @cindex @code{ULK} fixed-suffix
1105 @cindex @code{ULLK} fixed-suffix
1107 As an extension, the GNU C compiler supports fixed-point types as
1108 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1109 types in GCC will evolve as the draft technical report changes.
1110 Calling conventions for any target might also change. Not all targets
1111 support fixed-point types.
1113 The fixed-point types are
1114 @code{short _Fract},
1117 @code{long long _Fract},
1118 @code{unsigned short _Fract},
1119 @code{unsigned _Fract},
1120 @code{unsigned long _Fract},
1121 @code{unsigned long long _Fract},
1122 @code{_Sat short _Fract},
1124 @code{_Sat long _Fract},
1125 @code{_Sat long long _Fract},
1126 @code{_Sat unsigned short _Fract},
1127 @code{_Sat unsigned _Fract},
1128 @code{_Sat unsigned long _Fract},
1129 @code{_Sat unsigned long long _Fract},
1130 @code{short _Accum},
1133 @code{long long _Accum},
1134 @code{unsigned short _Accum},
1135 @code{unsigned _Accum},
1136 @code{unsigned long _Accum},
1137 @code{unsigned long long _Accum},
1138 @code{_Sat short _Accum},
1140 @code{_Sat long _Accum},
1141 @code{_Sat long long _Accum},
1142 @code{_Sat unsigned short _Accum},
1143 @code{_Sat unsigned _Accum},
1144 @code{_Sat unsigned long _Accum},
1145 @code{_Sat unsigned long long _Accum}.
1147 Fixed-point data values contain fractional and optional integral parts.
1148 The format of fixed-point data varies and depends on the target machine.
1150 Support for fixed-point types includes:
1153 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1155 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1157 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1159 binary shift operators (@code{<<}, @code{>>})
1161 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1163 equality operators (@code{==}, @code{!=})
1165 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1166 @code{<<=}, @code{>>=})
1168 conversions to and from integer, floating-point, or fixed-point types
1171 Use a suffix in a fixed-point literal constant:
1173 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1174 @code{_Sat short _Fract}
1175 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1176 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1177 @code{_Sat long _Fract}
1178 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1179 @code{_Sat long long _Fract}
1180 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1181 @code{_Sat unsigned short _Fract}
1182 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1183 @code{_Sat unsigned _Fract}
1184 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1185 @code{_Sat unsigned long _Fract}
1186 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1187 and @code{_Sat unsigned long long _Fract}
1188 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1189 @code{_Sat short _Accum}
1190 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1191 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1192 @code{_Sat long _Accum}
1193 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1194 @code{_Sat long long _Accum}
1195 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1196 @code{_Sat unsigned short _Accum}
1197 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1198 @code{_Sat unsigned _Accum}
1199 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1200 @code{_Sat unsigned long _Accum}
1201 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1202 and @code{_Sat unsigned long long _Accum}
1205 GCC support of fixed-point types as specified by the draft technical report
1210 Pragmas to control overflow and rounding behaviors are not implemented.
1213 Fixed-point types are supported by the DWARF2 debug information format.
1215 @node Named Address Spaces
1216 @section Named address spaces
1217 @cindex named address spaces
1219 As an extension, the GNU C compiler supports named address spaces as
1220 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1221 address spaces in GCC will evolve as the draft technical report changes.
1222 Calling conventions for any target might also change. At present, only
1223 the SPU target supports other address spaces. On the SPU target, for
1224 example, variables may be declared as belonging to another address space
1225 by qualifying the type with the @code{__ea} address space identifier:
1231 When the variable @code{i} is accessed, the compiler will generate
1232 special code to access this variable. It may use runtime library
1233 support, or generate special machine instructions to access that address
1236 The @code{__ea} identifier may be used exactly like any other C type
1237 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1238 document for more details.
1241 @section Arrays of Length Zero
1242 @cindex arrays of length zero
1243 @cindex zero-length arrays
1244 @cindex length-zero arrays
1245 @cindex flexible array members
1247 Zero-length arrays are allowed in GNU C@. They are very useful as the
1248 last element of a structure which is really a header for a variable-length
1257 struct line *thisline = (struct line *)
1258 malloc (sizeof (struct line) + this_length);
1259 thisline->length = this_length;
1262 In ISO C90, you would have to give @code{contents} a length of 1, which
1263 means either you waste space or complicate the argument to @code{malloc}.
1265 In ISO C99, you would use a @dfn{flexible array member}, which is
1266 slightly different in syntax and semantics:
1270 Flexible array members are written as @code{contents[]} without
1274 Flexible array members have incomplete type, and so the @code{sizeof}
1275 operator may not be applied. As a quirk of the original implementation
1276 of zero-length arrays, @code{sizeof} evaluates to zero.
1279 Flexible array members may only appear as the last member of a
1280 @code{struct} that is otherwise non-empty.
1283 A structure containing a flexible array member, or a union containing
1284 such a structure (possibly recursively), may not be a member of a
1285 structure or an element of an array. (However, these uses are
1286 permitted by GCC as extensions.)
1289 GCC versions before 3.0 allowed zero-length arrays to be statically
1290 initialized, as if they were flexible arrays. In addition to those
1291 cases that were useful, it also allowed initializations in situations
1292 that would corrupt later data. Non-empty initialization of zero-length
1293 arrays is now treated like any case where there are more initializer
1294 elements than the array holds, in that a suitable warning about "excess
1295 elements in array" is given, and the excess elements (all of them, in
1296 this case) are ignored.
1298 Instead GCC allows static initialization of flexible array members.
1299 This is equivalent to defining a new structure containing the original
1300 structure followed by an array of sufficient size to contain the data.
1301 I.e.@: in the following, @code{f1} is constructed as if it were declared
1307 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1310 struct f1 f1; int data[3];
1311 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1315 The convenience of this extension is that @code{f1} has the desired
1316 type, eliminating the need to consistently refer to @code{f2.f1}.
1318 This has symmetry with normal static arrays, in that an array of
1319 unknown size is also written with @code{[]}.
1321 Of course, this extension only makes sense if the extra data comes at
1322 the end of a top-level object, as otherwise we would be overwriting
1323 data at subsequent offsets. To avoid undue complication and confusion
1324 with initialization of deeply nested arrays, we simply disallow any
1325 non-empty initialization except when the structure is the top-level
1326 object. For example:
1329 struct foo @{ int x; int y[]; @};
1330 struct bar @{ struct foo z; @};
1332 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1333 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1334 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1335 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1338 @node Empty Structures
1339 @section Structures With No Members
1340 @cindex empty structures
1341 @cindex zero-size structures
1343 GCC permits a C structure to have no members:
1350 The structure will have size zero. In C++, empty structures are part
1351 of the language. G++ treats empty structures as if they had a single
1352 member of type @code{char}.
1354 @node Variable Length
1355 @section Arrays of Variable Length
1356 @cindex variable-length arrays
1357 @cindex arrays of variable length
1360 Variable-length automatic arrays are allowed in ISO C99, and as an
1361 extension GCC accepts them in C90 mode and in C++. (However, GCC's
1362 implementation of variable-length arrays does not yet conform in detail
1363 to the ISO C99 standard.) These arrays are
1364 declared like any other automatic arrays, but with a length that is not
1365 a constant expression. The storage is allocated at the point of
1366 declaration and deallocated when the brace-level is exited. For
1371 concat_fopen (char *s1, char *s2, char *mode)
1373 char str[strlen (s1) + strlen (s2) + 1];
1376 return fopen (str, mode);
1380 @cindex scope of a variable length array
1381 @cindex variable-length array scope
1382 @cindex deallocating variable length arrays
1383 Jumping or breaking out of the scope of the array name deallocates the
1384 storage. Jumping into the scope is not allowed; you get an error
1387 @cindex @code{alloca} vs variable-length arrays
1388 You can use the function @code{alloca} to get an effect much like
1389 variable-length arrays. The function @code{alloca} is available in
1390 many other C implementations (but not in all). On the other hand,
1391 variable-length arrays are more elegant.
1393 There are other differences between these two methods. Space allocated
1394 with @code{alloca} exists until the containing @emph{function} returns.
1395 The space for a variable-length array is deallocated as soon as the array
1396 name's scope ends. (If you use both variable-length arrays and
1397 @code{alloca} in the same function, deallocation of a variable-length array
1398 will also deallocate anything more recently allocated with @code{alloca}.)
1400 You can also use variable-length arrays as arguments to functions:
1404 tester (int len, char data[len][len])
1410 The length of an array is computed once when the storage is allocated
1411 and is remembered for the scope of the array in case you access it with
1414 If you want to pass the array first and the length afterward, you can
1415 use a forward declaration in the parameter list---another GNU extension.
1419 tester (int len; char data[len][len], int len)
1425 @cindex parameter forward declaration
1426 The @samp{int len} before the semicolon is a @dfn{parameter forward
1427 declaration}, and it serves the purpose of making the name @code{len}
1428 known when the declaration of @code{data} is parsed.
1430 You can write any number of such parameter forward declarations in the
1431 parameter list. They can be separated by commas or semicolons, but the
1432 last one must end with a semicolon, which is followed by the ``real''
1433 parameter declarations. Each forward declaration must match a ``real''
1434 declaration in parameter name and data type. ISO C99 does not support
1435 parameter forward declarations.
1437 @node Variadic Macros
1438 @section Macros with a Variable Number of Arguments.
1439 @cindex variable number of arguments
1440 @cindex macro with variable arguments
1441 @cindex rest argument (in macro)
1442 @cindex variadic macros
1444 In the ISO C standard of 1999, a macro can be declared to accept a
1445 variable number of arguments much as a function can. The syntax for
1446 defining the macro is similar to that of a function. Here is an
1450 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1453 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1454 such a macro, it represents the zero or more tokens until the closing
1455 parenthesis that ends the invocation, including any commas. This set of
1456 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1457 wherever it appears. See the CPP manual for more information.
1459 GCC has long supported variadic macros, and used a different syntax that
1460 allowed you to give a name to the variable arguments just like any other
1461 argument. Here is an example:
1464 #define debug(format, args...) fprintf (stderr, format, args)
1467 This is in all ways equivalent to the ISO C example above, but arguably
1468 more readable and descriptive.
1470 GNU CPP has two further variadic macro extensions, and permits them to
1471 be used with either of the above forms of macro definition.
1473 In standard C, you are not allowed to leave the variable argument out
1474 entirely; but you are allowed to pass an empty argument. For example,
1475 this invocation is invalid in ISO C, because there is no comma after
1482 GNU CPP permits you to completely omit the variable arguments in this
1483 way. In the above examples, the compiler would complain, though since
1484 the expansion of the macro still has the extra comma after the format
1487 To help solve this problem, CPP behaves specially for variable arguments
1488 used with the token paste operator, @samp{##}. If instead you write
1491 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1494 and if the variable arguments are omitted or empty, the @samp{##}
1495 operator causes the preprocessor to remove the comma before it. If you
1496 do provide some variable arguments in your macro invocation, GNU CPP
1497 does not complain about the paste operation and instead places the
1498 variable arguments after the comma. Just like any other pasted macro
1499 argument, these arguments are not macro expanded.
1501 @node Escaped Newlines
1502 @section Slightly Looser Rules for Escaped Newlines
1503 @cindex escaped newlines
1504 @cindex newlines (escaped)
1506 Recently, the preprocessor has relaxed its treatment of escaped
1507 newlines. Previously, the newline had to immediately follow a
1508 backslash. The current implementation allows whitespace in the form
1509 of spaces, horizontal and vertical tabs, and form feeds between the
1510 backslash and the subsequent newline. The preprocessor issues a
1511 warning, but treats it as a valid escaped newline and combines the two
1512 lines to form a single logical line. This works within comments and
1513 tokens, as well as between tokens. Comments are @emph{not} treated as
1514 whitespace for the purposes of this relaxation, since they have not
1515 yet been replaced with spaces.
1518 @section Non-Lvalue Arrays May Have Subscripts
1519 @cindex subscripting
1520 @cindex arrays, non-lvalue
1522 @cindex subscripting and function values
1523 In ISO C99, arrays that are not lvalues still decay to pointers, and
1524 may be subscripted, although they may not be modified or used after
1525 the next sequence point and the unary @samp{&} operator may not be
1526 applied to them. As an extension, GCC allows such arrays to be
1527 subscripted in C90 mode, though otherwise they do not decay to
1528 pointers outside C99 mode. For example,
1529 this is valid in GNU C though not valid in C90:
1533 struct foo @{int a[4];@};
1539 return f().a[index];
1545 @section Arithmetic on @code{void}- and Function-Pointers
1546 @cindex void pointers, arithmetic
1547 @cindex void, size of pointer to
1548 @cindex function pointers, arithmetic
1549 @cindex function, size of pointer to
1551 In GNU C, addition and subtraction operations are supported on pointers to
1552 @code{void} and on pointers to functions. This is done by treating the
1553 size of a @code{void} or of a function as 1.
1555 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1556 and on function types, and returns 1.
1558 @opindex Wpointer-arith
1559 The option @option{-Wpointer-arith} requests a warning if these extensions
1563 @section Non-Constant Initializers
1564 @cindex initializers, non-constant
1565 @cindex non-constant initializers
1567 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1568 automatic variable are not required to be constant expressions in GNU C@.
1569 Here is an example of an initializer with run-time varying elements:
1572 foo (float f, float g)
1574 float beat_freqs[2] = @{ f-g, f+g @};
1579 @node Compound Literals
1580 @section Compound Literals
1581 @cindex constructor expressions
1582 @cindex initializations in expressions
1583 @cindex structures, constructor expression
1584 @cindex expressions, constructor
1585 @cindex compound literals
1586 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1588 ISO C99 supports compound literals. A compound literal looks like
1589 a cast containing an initializer. Its value is an object of the
1590 type specified in the cast, containing the elements specified in
1591 the initializer; it is an lvalue. As an extension, GCC supports
1592 compound literals in C90 mode and in C++.
1594 Usually, the specified type is a structure. Assume that
1595 @code{struct foo} and @code{structure} are declared as shown:
1598 struct foo @{int a; char b[2];@} structure;
1602 Here is an example of constructing a @code{struct foo} with a compound literal:
1605 structure = ((struct foo) @{x + y, 'a', 0@});
1609 This is equivalent to writing the following:
1613 struct foo temp = @{x + y, 'a', 0@};
1618 You can also construct an array. If all the elements of the compound literal
1619 are (made up of) simple constant expressions, suitable for use in
1620 initializers of objects of static storage duration, then the compound
1621 literal can be coerced to a pointer to its first element and used in
1622 such an initializer, as shown here:
1625 char **foo = (char *[]) @{ "x", "y", "z" @};
1628 Compound literals for scalar types and union types are is
1629 also allowed, but then the compound literal is equivalent
1632 As a GNU extension, GCC allows initialization of objects with static storage
1633 duration by compound literals (which is not possible in ISO C99, because
1634 the initializer is not a constant).
1635 It is handled as if the object was initialized only with the bracket
1636 enclosed list if the types of the compound literal and the object match.
1637 The initializer list of the compound literal must be constant.
1638 If the object being initialized has array type of unknown size, the size is
1639 determined by compound literal size.
1642 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1643 static int y[] = (int []) @{1, 2, 3@};
1644 static int z[] = (int [3]) @{1@};
1648 The above lines are equivalent to the following:
1650 static struct foo x = @{1, 'a', 'b'@};
1651 static int y[] = @{1, 2, 3@};
1652 static int z[] = @{1, 0, 0@};
1655 @node Designated Inits
1656 @section Designated Initializers
1657 @cindex initializers with labeled elements
1658 @cindex labeled elements in initializers
1659 @cindex case labels in initializers
1660 @cindex designated initializers
1662 Standard C90 requires the elements of an initializer to appear in a fixed
1663 order, the same as the order of the elements in the array or structure
1666 In ISO C99 you can give the elements in any order, specifying the array
1667 indices or structure field names they apply to, and GNU C allows this as
1668 an extension in C90 mode as well. This extension is not
1669 implemented in GNU C++.
1671 To specify an array index, write
1672 @samp{[@var{index}] =} before the element value. For example,
1675 int a[6] = @{ [4] = 29, [2] = 15 @};
1682 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1686 The index values must be constant expressions, even if the array being
1687 initialized is automatic.
1689 An alternative syntax for this which has been obsolete since GCC 2.5 but
1690 GCC still accepts is to write @samp{[@var{index}]} before the element
1691 value, with no @samp{=}.
1693 To initialize a range of elements to the same value, write
1694 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1695 extension. For example,
1698 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1702 If the value in it has side-effects, the side-effects will happen only once,
1703 not for each initialized field by the range initializer.
1706 Note that the length of the array is the highest value specified
1709 In a structure initializer, specify the name of a field to initialize
1710 with @samp{.@var{fieldname} =} before the element value. For example,
1711 given the following structure,
1714 struct point @{ int x, y; @};
1718 the following initialization
1721 struct point p = @{ .y = yvalue, .x = xvalue @};
1728 struct point p = @{ xvalue, yvalue @};
1731 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1732 @samp{@var{fieldname}:}, as shown here:
1735 struct point p = @{ y: yvalue, x: xvalue @};
1739 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1740 @dfn{designator}. You can also use a designator (or the obsolete colon
1741 syntax) when initializing a union, to specify which element of the union
1742 should be used. For example,
1745 union foo @{ int i; double d; @};
1747 union foo f = @{ .d = 4 @};
1751 will convert 4 to a @code{double} to store it in the union using
1752 the second element. By contrast, casting 4 to type @code{union foo}
1753 would store it into the union as the integer @code{i}, since it is
1754 an integer. (@xref{Cast to Union}.)
1756 You can combine this technique of naming elements with ordinary C
1757 initialization of successive elements. Each initializer element that
1758 does not have a designator applies to the next consecutive element of the
1759 array or structure. For example,
1762 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1769 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1772 Labeling the elements of an array initializer is especially useful
1773 when the indices are characters or belong to an @code{enum} type.
1778 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1779 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1782 @cindex designator lists
1783 You can also write a series of @samp{.@var{fieldname}} and
1784 @samp{[@var{index}]} designators before an @samp{=} to specify a
1785 nested subobject to initialize; the list is taken relative to the
1786 subobject corresponding to the closest surrounding brace pair. For
1787 example, with the @samp{struct point} declaration above:
1790 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1794 If the same field is initialized multiple times, it will have value from
1795 the last initialization. If any such overridden initialization has
1796 side-effect, it is unspecified whether the side-effect happens or not.
1797 Currently, GCC will discard them and issue a warning.
1800 @section Case Ranges
1802 @cindex ranges in case statements
1804 You can specify a range of consecutive values in a single @code{case} label,
1808 case @var{low} ... @var{high}:
1812 This has the same effect as the proper number of individual @code{case}
1813 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1815 This feature is especially useful for ranges of ASCII character codes:
1821 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1822 it may be parsed wrong when you use it with integer values. For example,
1837 @section Cast to a Union Type
1838 @cindex cast to a union
1839 @cindex union, casting to a
1841 A cast to union type is similar to other casts, except that the type
1842 specified is a union type. You can specify the type either with
1843 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1844 a constructor though, not a cast, and hence does not yield an lvalue like
1845 normal casts. (@xref{Compound Literals}.)
1847 The types that may be cast to the union type are those of the members
1848 of the union. Thus, given the following union and variables:
1851 union foo @{ int i; double d; @};
1857 both @code{x} and @code{y} can be cast to type @code{union foo}.
1859 Using the cast as the right-hand side of an assignment to a variable of
1860 union type is equivalent to storing in a member of the union:
1865 u = (union foo) x @equiv{} u.i = x
1866 u = (union foo) y @equiv{} u.d = y
1869 You can also use the union cast as a function argument:
1872 void hack (union foo);
1874 hack ((union foo) x);
1877 @node Mixed Declarations
1878 @section Mixed Declarations and Code
1879 @cindex mixed declarations and code
1880 @cindex declarations, mixed with code
1881 @cindex code, mixed with declarations
1883 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1884 within compound statements. As an extension, GCC also allows this in
1885 C90 mode. For example, you could do:
1894 Each identifier is visible from where it is declared until the end of
1895 the enclosing block.
1897 @node Function Attributes
1898 @section Declaring Attributes of Functions
1899 @cindex function attributes
1900 @cindex declaring attributes of functions
1901 @cindex functions that never return
1902 @cindex functions that return more than once
1903 @cindex functions that have no side effects
1904 @cindex functions in arbitrary sections
1905 @cindex functions that behave like malloc
1906 @cindex @code{volatile} applied to function
1907 @cindex @code{const} applied to function
1908 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1909 @cindex functions with non-null pointer arguments
1910 @cindex functions that are passed arguments in registers on the 386
1911 @cindex functions that pop the argument stack on the 386
1912 @cindex functions that do not pop the argument stack on the 386
1913 @cindex functions that have different compilation options on the 386
1914 @cindex functions that have different optimization options
1915 @cindex functions that are dynamically resolved
1917 In GNU C, you declare certain things about functions called in your program
1918 which help the compiler optimize function calls and check your code more
1921 The keyword @code{__attribute__} allows you to specify special
1922 attributes when making a declaration. This keyword is followed by an
1923 attribute specification inside double parentheses. The following
1924 attributes are currently defined for functions on all targets:
1925 @code{aligned}, @code{alloc_size}, @code{noreturn},
1926 @code{returns_twice}, @code{noinline}, @code{noclone},
1927 @code{always_inline}, @code{flatten}, @code{pure}, @code{const},
1928 @code{nothrow}, @code{sentinel}, @code{format}, @code{format_arg},
1929 @code{no_instrument_function}, @code{no_split_stack},
1930 @code{section}, @code{constructor},
1931 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1932 @code{weak}, @code{malloc}, @code{alias}, @code{ifunc},
1933 @code{warn_unused_result}, @code{nonnull}, @code{gnu_inline},
1934 @code{externally_visible}, @code{hot}, @code{cold}, @code{artificial},
1935 @code{error} and @code{warning}. Several other attributes are defined
1936 for functions on particular target systems. Other attributes,
1937 including @code{section} are supported for variables declarations
1938 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1940 GCC plugins may provide their own attributes.
1942 You may also specify attributes with @samp{__} preceding and following
1943 each keyword. This allows you to use them in header files without
1944 being concerned about a possible macro of the same name. For example,
1945 you may use @code{__noreturn__} instead of @code{noreturn}.
1947 @xref{Attribute Syntax}, for details of the exact syntax for using
1951 @c Keep this table alphabetized by attribute name. Treat _ as space.
1953 @item alias ("@var{target}")
1954 @cindex @code{alias} attribute
1955 The @code{alias} attribute causes the declaration to be emitted as an
1956 alias for another symbol, which must be specified. For instance,
1959 void __f () @{ /* @r{Do something.} */; @}
1960 void f () __attribute__ ((weak, alias ("__f")));
1963 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1964 mangled name for the target must be used. It is an error if @samp{__f}
1965 is not defined in the same translation unit.
1967 Not all target machines support this attribute.
1969 @item aligned (@var{alignment})
1970 @cindex @code{aligned} attribute
1971 This attribute specifies a minimum alignment for the function,
1974 You cannot use this attribute to decrease the alignment of a function,
1975 only to increase it. However, when you explicitly specify a function
1976 alignment this will override the effect of the
1977 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1980 Note that the effectiveness of @code{aligned} attributes may be
1981 limited by inherent limitations in your linker. On many systems, the
1982 linker is only able to arrange for functions to be aligned up to a
1983 certain maximum alignment. (For some linkers, the maximum supported
1984 alignment may be very very small.) See your linker documentation for
1985 further information.
1987 The @code{aligned} attribute can also be used for variables and fields
1988 (@pxref{Variable Attributes}.)
1991 @cindex @code{alloc_size} attribute
1992 The @code{alloc_size} attribute is used to tell the compiler that the
1993 function return value points to memory, where the size is given by
1994 one or two of the functions parameters. GCC uses this
1995 information to improve the correctness of @code{__builtin_object_size}.
1997 The function parameter(s) denoting the allocated size are specified by
1998 one or two integer arguments supplied to the attribute. The allocated size
1999 is either the value of the single function argument specified or the product
2000 of the two function arguments specified. Argument numbering starts at
2006 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2007 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2010 declares that my_calloc will return memory of the size given by
2011 the product of parameter 1 and 2 and that my_realloc will return memory
2012 of the size given by parameter 2.
2015 @cindex @code{always_inline} function attribute
2016 Generally, functions are not inlined unless optimization is specified.
2017 For functions declared inline, this attribute inlines the function even
2018 if no optimization level was specified.
2021 @cindex @code{gnu_inline} function attribute
2022 This attribute should be used with a function which is also declared
2023 with the @code{inline} keyword. It directs GCC to treat the function
2024 as if it were defined in gnu90 mode even when compiling in C99 or
2027 If the function is declared @code{extern}, then this definition of the
2028 function is used only for inlining. In no case is the function
2029 compiled as a standalone function, not even if you take its address
2030 explicitly. Such an address becomes an external reference, as if you
2031 had only declared the function, and had not defined it. This has
2032 almost the effect of a macro. The way to use this is to put a
2033 function definition in a header file with this attribute, and put
2034 another copy of the function, without @code{extern}, in a library
2035 file. The definition in the header file will cause most calls to the
2036 function to be inlined. If any uses of the function remain, they will
2037 refer to the single copy in the library. Note that the two
2038 definitions of the functions need not be precisely the same, although
2039 if they do not have the same effect your program may behave oddly.
2041 In C, if the function is neither @code{extern} nor @code{static}, then
2042 the function is compiled as a standalone function, as well as being
2043 inlined where possible.
2045 This is how GCC traditionally handled functions declared
2046 @code{inline}. Since ISO C99 specifies a different semantics for
2047 @code{inline}, this function attribute is provided as a transition
2048 measure and as a useful feature in its own right. This attribute is
2049 available in GCC 4.1.3 and later. It is available if either of the
2050 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2051 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2052 Function is As Fast As a Macro}.
2054 In C++, this attribute does not depend on @code{extern} in any way,
2055 but it still requires the @code{inline} keyword to enable its special
2059 @cindex @code{artificial} function attribute
2060 This attribute is useful for small inline wrappers which if possible
2061 should appear during debugging as a unit, depending on the debug
2062 info format it will either mean marking the function as artificial
2063 or using the caller location for all instructions within the inlined
2067 @cindex interrupt handler functions
2068 When added to an interrupt handler with the M32C port, causes the
2069 prologue and epilogue to use bank switching to preserve the registers
2070 rather than saving them on the stack.
2073 @cindex @code{flatten} function attribute
2074 Generally, inlining into a function is limited. For a function marked with
2075 this attribute, every call inside this function will be inlined, if possible.
2076 Whether the function itself is considered for inlining depends on its size and
2077 the current inlining parameters.
2079 @item error ("@var{message}")
2080 @cindex @code{error} function attribute
2081 If this attribute is used on a function declaration and a call to such a function
2082 is not eliminated through dead code elimination or other optimizations, an error
2083 which will include @var{message} will be diagnosed. This is useful
2084 for compile time checking, especially together with @code{__builtin_constant_p}
2085 and inline functions where checking the inline function arguments is not
2086 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2087 While it is possible to leave the function undefined and thus invoke
2088 a link failure, when using this attribute the problem will be diagnosed
2089 earlier and with exact location of the call even in presence of inline
2090 functions or when not emitting debugging information.
2092 @item warning ("@var{message}")
2093 @cindex @code{warning} function attribute
2094 If this attribute is used on a function declaration and a call to such a function
2095 is not eliminated through dead code elimination or other optimizations, a warning
2096 which will include @var{message} will be diagnosed. This is useful
2097 for compile time checking, especially together with @code{__builtin_constant_p}
2098 and inline functions. While it is possible to define the function with
2099 a message in @code{.gnu.warning*} section, when using this attribute the problem
2100 will be diagnosed earlier and with exact location of the call even in presence
2101 of inline functions or when not emitting debugging information.
2104 @cindex functions that do pop the argument stack on the 386
2106 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2107 assume that the calling function will pop off the stack space used to
2108 pass arguments. This is
2109 useful to override the effects of the @option{-mrtd} switch.
2112 @cindex @code{const} function attribute
2113 Many functions do not examine any values except their arguments, and
2114 have no effects except the return value. Basically this is just slightly
2115 more strict class than the @code{pure} attribute below, since function is not
2116 allowed to read global memory.
2118 @cindex pointer arguments
2119 Note that a function that has pointer arguments and examines the data
2120 pointed to must @emph{not} be declared @code{const}. Likewise, a
2121 function that calls a non-@code{const} function usually must not be
2122 @code{const}. It does not make sense for a @code{const} function to
2125 The attribute @code{const} is not implemented in GCC versions earlier
2126 than 2.5. An alternative way to declare that a function has no side
2127 effects, which works in the current version and in some older versions,
2131 typedef int intfn ();
2133 extern const intfn square;
2136 This approach does not work in GNU C++ from 2.6.0 on, since the language
2137 specifies that the @samp{const} must be attached to the return value.
2141 @itemx constructor (@var{priority})
2142 @itemx destructor (@var{priority})
2143 @cindex @code{constructor} function attribute
2144 @cindex @code{destructor} function attribute
2145 The @code{constructor} attribute causes the function to be called
2146 automatically before execution enters @code{main ()}. Similarly, the
2147 @code{destructor} attribute causes the function to be called
2148 automatically after @code{main ()} has completed or @code{exit ()} has
2149 been called. Functions with these attributes are useful for
2150 initializing data that will be used implicitly during the execution of
2153 You may provide an optional integer priority to control the order in
2154 which constructor and destructor functions are run. A constructor
2155 with a smaller priority number runs before a constructor with a larger
2156 priority number; the opposite relationship holds for destructors. So,
2157 if you have a constructor that allocates a resource and a destructor
2158 that deallocates the same resource, both functions typically have the
2159 same priority. The priorities for constructor and destructor
2160 functions are the same as those specified for namespace-scope C++
2161 objects (@pxref{C++ Attributes}).
2163 These attributes are not currently implemented for Objective-C@.
2166 @itemx deprecated (@var{msg})
2167 @cindex @code{deprecated} attribute.
2168 The @code{deprecated} attribute results in a warning if the function
2169 is used anywhere in the source file. This is useful when identifying
2170 functions that are expected to be removed in a future version of a
2171 program. The warning also includes the location of the declaration
2172 of the deprecated function, to enable users to easily find further
2173 information about why the function is deprecated, or what they should
2174 do instead. Note that the warnings only occurs for uses:
2177 int old_fn () __attribute__ ((deprecated));
2179 int (*fn_ptr)() = old_fn;
2182 results in a warning on line 3 but not line 2. The optional msg
2183 argument, which must be a string, will be printed in the warning if
2186 The @code{deprecated} attribute can also be used for variables and
2187 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2190 @cindex @code{disinterrupt} attribute
2191 On MeP targets, this attribute causes the compiler to emit
2192 instructions to disable interrupts for the duration of the given
2196 @cindex @code{__declspec(dllexport)}
2197 On Microsoft Windows targets and Symbian OS targets the
2198 @code{dllexport} attribute causes the compiler to provide a global
2199 pointer to a pointer in a DLL, so that it can be referenced with the
2200 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2201 name is formed by combining @code{_imp__} and the function or variable
2204 You can use @code{__declspec(dllexport)} as a synonym for
2205 @code{__attribute__ ((dllexport))} for compatibility with other
2208 On systems that support the @code{visibility} attribute, this
2209 attribute also implies ``default'' visibility. It is an error to
2210 explicitly specify any other visibility.
2212 Currently, the @code{dllexport} attribute is ignored for inlined
2213 functions, unless the @option{-fkeep-inline-functions} flag has been
2214 used. The attribute is also ignored for undefined symbols.
2216 When applied to C++ classes, the attribute marks defined non-inlined
2217 member functions and static data members as exports. Static consts
2218 initialized in-class are not marked unless they are also defined
2221 For Microsoft Windows targets there are alternative methods for
2222 including the symbol in the DLL's export table such as using a
2223 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2224 the @option{--export-all} linker flag.
2227 @cindex @code{__declspec(dllimport)}
2228 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2229 attribute causes the compiler to reference a function or variable via
2230 a global pointer to a pointer that is set up by the DLL exporting the
2231 symbol. The attribute implies @code{extern}. On Microsoft Windows
2232 targets, the pointer name is formed by combining @code{_imp__} and the
2233 function or variable name.
2235 You can use @code{__declspec(dllimport)} as a synonym for
2236 @code{__attribute__ ((dllimport))} for compatibility with other
2239 On systems that support the @code{visibility} attribute, this
2240 attribute also implies ``default'' visibility. It is an error to
2241 explicitly specify any other visibility.
2243 Currently, the attribute is ignored for inlined functions. If the
2244 attribute is applied to a symbol @emph{definition}, an error is reported.
2245 If a symbol previously declared @code{dllimport} is later defined, the
2246 attribute is ignored in subsequent references, and a warning is emitted.
2247 The attribute is also overridden by a subsequent declaration as
2250 When applied to C++ classes, the attribute marks non-inlined
2251 member functions and static data members as imports. However, the
2252 attribute is ignored for virtual methods to allow creation of vtables
2255 On the SH Symbian OS target the @code{dllimport} attribute also has
2256 another affect---it can cause the vtable and run-time type information
2257 for a class to be exported. This happens when the class has a
2258 dllimport'ed constructor or a non-inline, non-pure virtual function
2259 and, for either of those two conditions, the class also has an inline
2260 constructor or destructor and has a key function that is defined in
2261 the current translation unit.
2263 For Microsoft Windows based targets the use of the @code{dllimport}
2264 attribute on functions is not necessary, but provides a small
2265 performance benefit by eliminating a thunk in the DLL@. The use of the
2266 @code{dllimport} attribute on imported variables was required on older
2267 versions of the GNU linker, but can now be avoided by passing the
2268 @option{--enable-auto-import} switch to the GNU linker. As with
2269 functions, using the attribute for a variable eliminates a thunk in
2272 One drawback to using this attribute is that a pointer to a
2273 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2274 address. However, a pointer to a @emph{function} with the
2275 @code{dllimport} attribute can be used as a constant initializer; in
2276 this case, the address of a stub function in the import lib is
2277 referenced. On Microsoft Windows targets, the attribute can be disabled
2278 for functions by setting the @option{-mnop-fun-dllimport} flag.
2281 @cindex eight bit data on the H8/300, H8/300H, and H8S
2282 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2283 variable should be placed into the eight bit data section.
2284 The compiler will generate more efficient code for certain operations
2285 on data in the eight bit data area. Note the eight bit data area is limited to
2288 You must use GAS and GLD from GNU binutils version 2.7 or later for
2289 this attribute to work correctly.
2291 @item exception_handler
2292 @cindex exception handler functions on the Blackfin processor
2293 Use this attribute on the Blackfin to indicate that the specified function
2294 is an exception handler. The compiler will generate function entry and
2295 exit sequences suitable for use in an exception handler when this
2296 attribute is present.
2298 @item externally_visible
2299 @cindex @code{externally_visible} attribute.
2300 This attribute, attached to a global variable or function, nullifies
2301 the effect of the @option{-fwhole-program} command-line option, so the
2302 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.
2305 @cindex functions which handle memory bank switching
2306 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2307 use a calling convention that takes care of switching memory banks when
2308 entering and leaving a function. This calling convention is also the
2309 default when using the @option{-mlong-calls} option.
2311 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2312 to call and return from a function.
2314 On 68HC11 the compiler will generate a sequence of instructions
2315 to invoke a board-specific routine to switch the memory bank and call the
2316 real function. The board-specific routine simulates a @code{call}.
2317 At the end of a function, it will jump to a board-specific routine
2318 instead of using @code{rts}. The board-specific return routine simulates
2321 On MeP targets this causes the compiler to use a calling convention
2322 which assumes the called function is too far away for the built-in
2325 @item fast_interrupt
2326 @cindex interrupt handler functions
2327 Use this attribute on the M32C and RX ports to indicate that the specified
2328 function is a fast interrupt handler. This is just like the
2329 @code{interrupt} attribute, except that @code{freit} is used to return
2330 instead of @code{reit}.
2333 @cindex functions that pop the argument stack on the 386
2334 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2335 pass the first argument (if of integral type) in the register ECX and
2336 the second argument (if of integral type) in the register EDX@. Subsequent
2337 and other typed arguments are passed on the stack. The called function will
2338 pop the arguments off the stack. If the number of arguments is variable all
2339 arguments are pushed on the stack.
2342 @cindex functions that pop the argument stack on the 386
2343 On the Intel 386, the @code{thiscall} attribute causes the compiler to
2344 pass the first argument (if of integral type) in the register ECX.
2345 Subsequent and other typed arguments are passed on the stack. The called
2346 function will pop the arguments off the stack.
2347 If the number of arguments is variable all arguments are pushed on the
2349 The @code{thiscall} attribute is intended for C++ non-static member functions.
2350 As gcc extension this calling convention can be used for C-functions
2351 and for static member methods.
2353 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2354 @cindex @code{format} function attribute
2356 The @code{format} attribute specifies that a function takes @code{printf},
2357 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2358 should be type-checked against a format string. For example, the
2363 my_printf (void *my_object, const char *my_format, ...)
2364 __attribute__ ((format (printf, 2, 3)));
2368 causes the compiler to check the arguments in calls to @code{my_printf}
2369 for consistency with the @code{printf} style format string argument
2372 The parameter @var{archetype} determines how the format string is
2373 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2374 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2375 @code{strfmon}. (You can also use @code{__printf__},
2376 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2377 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2378 @code{ms_strftime} are also present.
2379 @var{archtype} values such as @code{printf} refer to the formats accepted
2380 by the system's C run-time library, while @code{gnu_} values always refer
2381 to the formats accepted by the GNU C Library. On Microsoft Windows
2382 targets, @code{ms_} values refer to the formats accepted by the
2383 @file{msvcrt.dll} library.
2384 The parameter @var{string-index}
2385 specifies which argument is the format string argument (starting
2386 from 1), while @var{first-to-check} is the number of the first
2387 argument to check against the format string. For functions
2388 where the arguments are not available to be checked (such as
2389 @code{vprintf}), specify the third parameter as zero. In this case the
2390 compiler only checks the format string for consistency. For
2391 @code{strftime} formats, the third parameter is required to be zero.
2392 Since non-static C++ methods have an implicit @code{this} argument, the
2393 arguments of such methods should be counted from two, not one, when
2394 giving values for @var{string-index} and @var{first-to-check}.
2396 In the example above, the format string (@code{my_format}) is the second
2397 argument of the function @code{my_print}, and the arguments to check
2398 start with the third argument, so the correct parameters for the format
2399 attribute are 2 and 3.
2401 @opindex ffreestanding
2402 @opindex fno-builtin
2403 The @code{format} attribute allows you to identify your own functions
2404 which take format strings as arguments, so that GCC can check the
2405 calls to these functions for errors. The compiler always (unless
2406 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2407 for the standard library functions @code{printf}, @code{fprintf},
2408 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2409 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2410 warnings are requested (using @option{-Wformat}), so there is no need to
2411 modify the header file @file{stdio.h}. In C99 mode, the functions
2412 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2413 @code{vsscanf} are also checked. Except in strictly conforming C
2414 standard modes, the X/Open function @code{strfmon} is also checked as
2415 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2416 @xref{C Dialect Options,,Options Controlling C Dialect}.
2418 The target may provide additional types of format checks.
2419 @xref{Target Format Checks,,Format Checks Specific to Particular
2422 @item format_arg (@var{string-index})
2423 @cindex @code{format_arg} function attribute
2424 @opindex Wformat-nonliteral
2425 The @code{format_arg} attribute specifies that a function takes a format
2426 string for a @code{printf}, @code{scanf}, @code{strftime} or
2427 @code{strfmon} style function and modifies it (for example, to translate
2428 it into another language), so the result can be passed to a
2429 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2430 function (with the remaining arguments to the format function the same
2431 as they would have been for the unmodified string). For example, the
2436 my_dgettext (char *my_domain, const char *my_format)
2437 __attribute__ ((format_arg (2)));
2441 causes the compiler to check the arguments in calls to a @code{printf},
2442 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2443 format string argument is a call to the @code{my_dgettext} function, for
2444 consistency with the format string argument @code{my_format}. If the
2445 @code{format_arg} attribute had not been specified, all the compiler
2446 could tell in such calls to format functions would be that the format
2447 string argument is not constant; this would generate a warning when
2448 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2449 without the attribute.
2451 The parameter @var{string-index} specifies which argument is the format
2452 string argument (starting from one). Since non-static C++ methods have
2453 an implicit @code{this} argument, the arguments of such methods should
2454 be counted from two.
2456 The @code{format-arg} attribute allows you to identify your own
2457 functions which modify format strings, so that GCC can check the
2458 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2459 type function whose operands are a call to one of your own function.
2460 The compiler always treats @code{gettext}, @code{dgettext}, and
2461 @code{dcgettext} in this manner except when strict ISO C support is
2462 requested by @option{-ansi} or an appropriate @option{-std} option, or
2463 @option{-ffreestanding} or @option{-fno-builtin}
2464 is used. @xref{C Dialect Options,,Options
2465 Controlling C Dialect}.
2467 @item function_vector
2468 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2469 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2470 function should be called through the function vector. Calling a
2471 function through the function vector will reduce code size, however;
2472 the function vector has a limited size (maximum 128 entries on the H8/300
2473 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2475 In SH2A target, this attribute declares a function to be called using the
2476 TBR relative addressing mode. The argument to this attribute is the entry
2477 number of the same function in a vector table containing all the TBR
2478 relative addressable functions. For the successful jump, register TBR
2479 should contain the start address of this TBR relative vector table.
2480 In the startup routine of the user application, user needs to care of this
2481 TBR register initialization. The TBR relative vector table can have at
2482 max 256 function entries. The jumps to these functions will be generated
2483 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2484 You must use GAS and GLD from GNU binutils version 2.7 or later for
2485 this attribute to work correctly.
2487 Please refer the example of M16C target, to see the use of this
2488 attribute while declaring a function,
2490 In an application, for a function being called once, this attribute will
2491 save at least 8 bytes of code; and if other successive calls are being
2492 made to the same function, it will save 2 bytes of code per each of these
2495 On M16C/M32C targets, the @code{function_vector} attribute declares a
2496 special page subroutine call function. Use of this attribute reduces
2497 the code size by 2 bytes for each call generated to the
2498 subroutine. The argument to the attribute is the vector number entry
2499 from the special page vector table which contains the 16 low-order
2500 bits of the subroutine's entry address. Each vector table has special
2501 page number (18 to 255) which are used in @code{jsrs} instruction.
2502 Jump addresses of the routines are generated by adding 0x0F0000 (in
2503 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2504 byte addresses set in the vector table. Therefore you need to ensure
2505 that all the special page vector routines should get mapped within the
2506 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2509 In the following example 2 bytes will be saved for each call to
2510 function @code{foo}.
2513 void foo (void) __attribute__((function_vector(0x18)));
2524 If functions are defined in one file and are called in another file,
2525 then be sure to write this declaration in both files.
2527 This attribute is ignored for R8C target.
2530 @cindex interrupt handler functions
2531 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k, MeP, MIPS,
2532 RX and Xstormy16 ports to indicate that the specified function is an
2533 interrupt handler. The compiler will generate function entry and exit
2534 sequences suitable for use in an interrupt handler when this attribute
2537 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2538 SH processors can be specified via the @code{interrupt_handler} attribute.
2540 Note, on the AVR, interrupts will be enabled inside the function.
2542 Note, for the ARM, you can specify the kind of interrupt to be handled by
2543 adding an optional parameter to the interrupt attribute like this:
2546 void f () __attribute__ ((interrupt ("IRQ")));
2549 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2551 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2552 may be called with a word aligned stack pointer.
2554 On MIPS targets, you can use the following attributes to modify the behavior
2555 of an interrupt handler:
2557 @item use_shadow_register_set
2558 @cindex @code{use_shadow_register_set} attribute
2559 Assume that the handler uses a shadow register set, instead of
2560 the main general-purpose registers.
2562 @item keep_interrupts_masked
2563 @cindex @code{keep_interrupts_masked} attribute
2564 Keep interrupts masked for the whole function. Without this attribute,
2565 GCC tries to reenable interrupts for as much of the function as it can.
2567 @item use_debug_exception_return
2568 @cindex @code{use_debug_exception_return} attribute
2569 Return using the @code{deret} instruction. Interrupt handlers that don't
2570 have this attribute return using @code{eret} instead.
2573 You can use any combination of these attributes, as shown below:
2575 void __attribute__ ((interrupt)) v0 ();
2576 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
2577 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
2578 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
2579 void __attribute__ ((interrupt, use_shadow_register_set,
2580 keep_interrupts_masked)) v4 ();
2581 void __attribute__ ((interrupt, use_shadow_register_set,
2582 use_debug_exception_return)) v5 ();
2583 void __attribute__ ((interrupt, keep_interrupts_masked,
2584 use_debug_exception_return)) v6 ();
2585 void __attribute__ ((interrupt, use_shadow_register_set,
2586 keep_interrupts_masked,
2587 use_debug_exception_return)) v7 ();
2590 @item ifunc ("@var{resolver}")
2591 @cindex @code{ifunc} attribute
2592 The @code{ifunc} attribute is used to mark a function as an indirect
2593 function using the STT_GNU_IFUNC symbol type extension to the ELF
2594 standard. This allows the resolution of the symbol value to be
2595 determined dynamically at load time, and an optimized version of the
2596 routine can be selected for the particular processor or other system
2597 characteristics determined then. To use this attribute, first define
2598 the implementation functions available, and a resolver function that
2599 returns a pointer to the selected implementation function. The
2600 implementation functions' declarations must match the API of the
2601 function being implemented, the resolver's declaration is be a
2602 function returning pointer to void function returning void:
2605 void *my_memcpy (void *dst, const void *src, size_t len)
2610 static void (*resolve_memcpy (void)) (void)
2612 return my_memcpy; // we'll just always select this routine
2616 The exported header file declaring the function the user calls would
2620 extern void *memcpy (void *, const void *, size_t);
2623 allowing the user to call this as a regular function, unaware of the
2624 implementation. Finally, the indirect function needs to be defined in
2625 the same translation unit as the resolver function:
2628 void *memcpy (void *, const void *, size_t)
2629 __attribute__ ((ifunc ("resolve_memcpy")));
2632 Indirect functions cannot be weak, and require a recent binutils (at
2633 least version 2.20.1), and GNU C library (at least version 2.11.1).
2635 @item interrupt_handler
2636 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2637 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2638 indicate that the specified function is an interrupt handler. The compiler
2639 will generate function entry and exit sequences suitable for use in an
2640 interrupt handler when this attribute is present.
2642 @item interrupt_thread
2643 @cindex interrupt thread functions on fido
2644 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2645 that the specified function is an interrupt handler that is designed
2646 to run as a thread. The compiler omits generate prologue/epilogue
2647 sequences and replaces the return instruction with a @code{sleep}
2648 instruction. This attribute is available only on fido.
2651 @cindex interrupt service routines on ARM
2652 Use this attribute on ARM to write Interrupt Service Routines. This is an
2653 alias to the @code{interrupt} attribute above.
2656 @cindex User stack pointer in interrupts on the Blackfin
2657 When used together with @code{interrupt_handler}, @code{exception_handler}
2658 or @code{nmi_handler}, code will be generated to load the stack pointer
2659 from the USP register in the function prologue.
2662 @cindex @code{l1_text} function attribute
2663 This attribute specifies a function to be placed into L1 Instruction
2664 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2665 With @option{-mfdpic}, function calls with a such function as the callee
2666 or caller will use inlined PLT.
2669 @cindex @code{l2} function attribute
2670 On the Blackfin, this attribute specifies a function to be placed into L2
2671 SRAM. The function will be put into a specific section named
2672 @code{.l1.text}. With @option{-mfdpic}, callers of such functions will use
2676 @cindex @code{leaf} function attribute
2677 Calls to external functions with this attribute must return to the current
2678 compilation unit only by return or by exception handling. In particular, leaf
2679 functions are not allowed to call callback function passed to it from current
2680 compilation unit or directly call functions exported by the unit or longjmp
2681 into the unit. Still leaf function might call functions from other complation
2682 units and thus they are not neccesarily leaf in the sense that they contains no
2683 function calls at all.
2685 The attribute is intended for library functions to improve dataflow analysis.
2686 Compiler takes the hint that any data not escaping current compilation unit can
2687 not be used or modified by the leaf function. For example, function @code{sin}
2688 is leaf, function @code{qsort} is not.
2690 Note that the leaf functions might invoke signals and signal handlers might be
2691 defined in the current compilation unit and use static variables. Only
2692 compliant way to write such a signal handler is to declare such variables
2695 The attribute has no effect on functions defined within current compilation
2696 unit. This is to allow easy merging of multiple compilation units into one,
2697 for example, by using the link time optimization. For this reason the
2698 attribute is not allowed on types to annotate indirect calls.
2700 @item long_call/short_call
2701 @cindex indirect calls on ARM
2702 This attribute specifies how a particular function is called on
2703 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2704 command-line switch and @code{#pragma long_calls} settings. The
2705 @code{long_call} attribute indicates that the function might be far
2706 away from the call site and require a different (more expensive)
2707 calling sequence. The @code{short_call} attribute always places
2708 the offset to the function from the call site into the @samp{BL}
2709 instruction directly.
2711 @item longcall/shortcall
2712 @cindex functions called via pointer on the RS/6000 and PowerPC
2713 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2714 indicates that the function might be far away from the call site and
2715 require a different (more expensive) calling sequence. The
2716 @code{shortcall} attribute indicates that the function is always close
2717 enough for the shorter calling sequence to be used. These attributes
2718 override both the @option{-mlongcall} switch and, on the RS/6000 and
2719 PowerPC, the @code{#pragma longcall} setting.
2721 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2722 calls are necessary.
2724 @item long_call/near/far
2725 @cindex indirect calls on MIPS
2726 These attributes specify how a particular function is called on MIPS@.
2727 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2728 command-line switch. The @code{long_call} and @code{far} attributes are
2729 synonyms, and cause the compiler to always call
2730 the function by first loading its address into a register, and then using
2731 the contents of that register. The @code{near} attribute has the opposite
2732 effect; it specifies that non-PIC calls should be made using the more
2733 efficient @code{jal} instruction.
2736 @cindex @code{malloc} attribute
2737 The @code{malloc} attribute is used to tell the compiler that a function
2738 may be treated as if any non-@code{NULL} pointer it returns cannot
2739 alias any other pointer valid when the function returns.
2740 This will often improve optimization.
2741 Standard functions with this property include @code{malloc} and
2742 @code{calloc}. @code{realloc}-like functions have this property as
2743 long as the old pointer is never referred to (including comparing it
2744 to the new pointer) after the function returns a non-@code{NULL}
2747 @item mips16/nomips16
2748 @cindex @code{mips16} attribute
2749 @cindex @code{nomips16} attribute
2751 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2752 function attributes to locally select or turn off MIPS16 code generation.
2753 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2754 while MIPS16 code generation is disabled for functions with the
2755 @code{nomips16} attribute. These attributes override the
2756 @option{-mips16} and @option{-mno-mips16} options on the command line
2757 (@pxref{MIPS Options}).
2759 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2760 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2761 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2762 may interact badly with some GCC extensions such as @code{__builtin_apply}
2763 (@pxref{Constructing Calls}).
2765 @item model (@var{model-name})
2766 @cindex function addressability on the M32R/D
2767 @cindex variable addressability on the IA-64
2769 On the M32R/D, use this attribute to set the addressability of an
2770 object, and of the code generated for a function. The identifier
2771 @var{model-name} is one of @code{small}, @code{medium}, or
2772 @code{large}, representing each of the code models.
2774 Small model objects live in the lower 16MB of memory (so that their
2775 addresses can be loaded with the @code{ld24} instruction), and are
2776 callable with the @code{bl} instruction.
2778 Medium model objects may live anywhere in the 32-bit address space (the
2779 compiler will generate @code{seth/add3} instructions to load their addresses),
2780 and are callable with the @code{bl} instruction.
2782 Large model objects may live anywhere in the 32-bit address space (the
2783 compiler will generate @code{seth/add3} instructions to load their addresses),
2784 and may not be reachable with the @code{bl} instruction (the compiler will
2785 generate the much slower @code{seth/add3/jl} instruction sequence).
2787 On IA-64, use this attribute to set the addressability of an object.
2788 At present, the only supported identifier for @var{model-name} is
2789 @code{small}, indicating addressability via ``small'' (22-bit)
2790 addresses (so that their addresses can be loaded with the @code{addl}
2791 instruction). Caveat: such addressing is by definition not position
2792 independent and hence this attribute must not be used for objects
2793 defined by shared libraries.
2795 @item ms_abi/sysv_abi
2796 @cindex @code{ms_abi} attribute
2797 @cindex @code{sysv_abi} attribute
2799 On 64-bit x86_64-*-* targets, you can use an ABI attribute to indicate
2800 which calling convention should be used for a function. The @code{ms_abi}
2801 attribute tells the compiler to use the Microsoft ABI, while the
2802 @code{sysv_abi} attribute tells the compiler to use the ABI used on
2803 GNU/Linux and other systems. The default is to use the Microsoft ABI
2804 when targeting Windows. On all other systems, the default is the AMD ABI.
2806 Note, the @code{ms_abi} attribute for Windows targets currently requires
2807 the @option{-maccumulate-outgoing-args} option.
2809 @item ms_hook_prologue
2810 @cindex @code{ms_hook_prologue} attribute
2812 On 32 bit i[34567]86-*-* targets and 64 bit x86_64-*-* targets, you can use
2813 this function attribute to make gcc generate the "hot-patching" function
2814 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
2818 @cindex function without a prologue/epilogue code
2819 Use this attribute on the ARM, AVR, MCORE, RX and SPU ports to indicate that
2820 the specified function does not need prologue/epilogue sequences generated by
2821 the compiler. It is up to the programmer to provide these sequences. The
2822 only statements that can be safely included in naked functions are
2823 @code{asm} statements that do not have operands. All other statements,
2824 including declarations of local variables, @code{if} statements, and so
2825 forth, should be avoided. Naked functions should be used to implement the
2826 body of an assembly function, while allowing the compiler to construct
2827 the requisite function declaration for the assembler.
2830 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2831 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2832 use the normal calling convention based on @code{jsr} and @code{rts}.
2833 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2836 On MeP targets this attribute causes the compiler to assume the called
2837 function is close enough to use the normal calling convention,
2838 overriding the @code{-mtf} command line option.
2841 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2842 Use this attribute together with @code{interrupt_handler},
2843 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2844 entry code should enable nested interrupts or exceptions.
2847 @cindex NMI handler functions on the Blackfin processor
2848 Use this attribute on the Blackfin to indicate that the specified function
2849 is an NMI handler. The compiler will generate function entry and
2850 exit sequences suitable for use in an NMI handler when this
2851 attribute is present.
2853 @item no_instrument_function
2854 @cindex @code{no_instrument_function} function attribute
2855 @opindex finstrument-functions
2856 If @option{-finstrument-functions} is given, profiling function calls will
2857 be generated at entry and exit of most user-compiled functions.
2858 Functions with this attribute will not be so instrumented.
2860 @item no_split_stack
2861 @cindex @code{no_split_stack} function attribute
2862 @opindex fsplit-stack
2863 If @option{-fsplit-stack} is given, functions will have a small
2864 prologue which decides whether to split the stack. Functions with the
2865 @code{no_split_stack} attribute will not have that prologue, and thus
2866 may run with only a small amount of stack space available.
2869 @cindex @code{noinline} function attribute
2870 This function attribute prevents a function from being considered for
2872 @c Don't enumerate the optimizations by name here; we try to be
2873 @c future-compatible with this mechanism.
2874 If the function does not have side-effects, there are optimizations
2875 other than inlining that causes function calls to be optimized away,
2876 although the function call is live. To keep such calls from being
2881 (@pxref{Extended Asm}) in the called function, to serve as a special
2885 @cindex @code{noclone} function attribute
2886 This function attribute prevents a function from being considered for
2887 cloning - a mechanism which produces specialized copies of functions
2888 and which is (currently) performed by interprocedural constant
2891 @item nonnull (@var{arg-index}, @dots{})
2892 @cindex @code{nonnull} function attribute
2893 The @code{nonnull} attribute specifies that some function parameters should
2894 be non-null pointers. For instance, the declaration:
2898 my_memcpy (void *dest, const void *src, size_t len)
2899 __attribute__((nonnull (1, 2)));
2903 causes the compiler to check that, in calls to @code{my_memcpy},
2904 arguments @var{dest} and @var{src} are non-null. If the compiler
2905 determines that a null pointer is passed in an argument slot marked
2906 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2907 is issued. The compiler may also choose to make optimizations based
2908 on the knowledge that certain function arguments will not be null.
2910 If no argument index list is given to the @code{nonnull} attribute,
2911 all pointer arguments are marked as non-null. To illustrate, the
2912 following declaration is equivalent to the previous example:
2916 my_memcpy (void *dest, const void *src, size_t len)
2917 __attribute__((nonnull));
2921 @cindex @code{noreturn} function attribute
2922 A few standard library functions, such as @code{abort} and @code{exit},
2923 cannot return. GCC knows this automatically. Some programs define
2924 their own functions that never return. You can declare them
2925 @code{noreturn} to tell the compiler this fact. For example,
2929 void fatal () __attribute__ ((noreturn));
2932 fatal (/* @r{@dots{}} */)
2934 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2940 The @code{noreturn} keyword tells the compiler to assume that
2941 @code{fatal} cannot return. It can then optimize without regard to what
2942 would happen if @code{fatal} ever did return. This makes slightly
2943 better code. More importantly, it helps avoid spurious warnings of
2944 uninitialized variables.
2946 The @code{noreturn} keyword does not affect the exceptional path when that
2947 applies: a @code{noreturn}-marked function may still return to the caller
2948 by throwing an exception or calling @code{longjmp}.
2950 Do not assume that registers saved by the calling function are
2951 restored before calling the @code{noreturn} function.
2953 It does not make sense for a @code{noreturn} function to have a return
2954 type other than @code{void}.
2956 The attribute @code{noreturn} is not implemented in GCC versions
2957 earlier than 2.5. An alternative way to declare that a function does
2958 not return, which works in the current version and in some older
2959 versions, is as follows:
2962 typedef void voidfn ();
2964 volatile voidfn fatal;
2967 This approach does not work in GNU C++.
2970 @cindex @code{nothrow} function attribute
2971 The @code{nothrow} attribute is used to inform the compiler that a
2972 function cannot throw an exception. For example, most functions in
2973 the standard C library can be guaranteed not to throw an exception
2974 with the notable exceptions of @code{qsort} and @code{bsearch} that
2975 take function pointer arguments. The @code{nothrow} attribute is not
2976 implemented in GCC versions earlier than 3.3.
2979 @cindex @code{optimize} function attribute
2980 The @code{optimize} attribute is used to specify that a function is to
2981 be compiled with different optimization options than specified on the
2982 command line. Arguments can either be numbers or strings. Numbers
2983 are assumed to be an optimization level. Strings that begin with
2984 @code{O} are assumed to be an optimization option, while other options
2985 are assumed to be used with a @code{-f} prefix. You can also use the
2986 @samp{#pragma GCC optimize} pragma to set the optimization options
2987 that affect more than one function.
2988 @xref{Function Specific Option Pragmas}, for details about the
2989 @samp{#pragma GCC optimize} pragma.
2991 This can be used for instance to have frequently executed functions
2992 compiled with more aggressive optimization options that produce faster
2993 and larger code, while other functions can be called with less
2997 @cindex @code{pcs} function attribute
2999 The @code{pcs} attribute can be used to control the calling convention
3000 used for a function on ARM. The attribute takes an argument that specifies
3001 the calling convention to use.
3003 When compiling using the AAPCS ABI (or a variant of that) then valid
3004 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3005 order to use a variant other than @code{"aapcs"} then the compiler must
3006 be permitted to use the appropriate co-processor registers (i.e., the
3007 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3011 /* Argument passed in r0, and result returned in r0+r1. */
3012 double f2d (float) __attribute__((pcs("aapcs")));
3015 Variadic functions always use the @code{"aapcs"} calling convention and
3016 the compiler will reject attempts to specify an alternative.
3019 @cindex @code{pure} function attribute
3020 Many functions have no effects except the return value and their
3021 return value depends only on the parameters and/or global variables.
3022 Such a function can be subject
3023 to common subexpression elimination and loop optimization just as an
3024 arithmetic operator would be. These functions should be declared
3025 with the attribute @code{pure}. For example,
3028 int square (int) __attribute__ ((pure));
3032 says that the hypothetical function @code{square} is safe to call
3033 fewer times than the program says.
3035 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3036 Interesting non-pure functions are functions with infinite loops or those
3037 depending on volatile memory or other system resource, that may change between
3038 two consecutive calls (such as @code{feof} in a multithreading environment).
3040 The attribute @code{pure} is not implemented in GCC versions earlier
3044 @cindex @code{hot} function attribute
3045 The @code{hot} attribute is used to inform the compiler that a function is a
3046 hot spot of the compiled program. The function is optimized more aggressively
3047 and on many target it is placed into special subsection of the text section so
3048 all hot functions appears close together improving locality.
3050 When profile feedback is available, via @option{-fprofile-use}, hot functions
3051 are automatically detected and this attribute is ignored.
3053 The @code{hot} attribute is not implemented in GCC versions earlier
3057 @cindex @code{cold} function attribute
3058 The @code{cold} attribute is used to inform the compiler that a function is
3059 unlikely executed. The function is optimized for size rather than speed and on
3060 many targets it is placed into special subsection of the text section so all
3061 cold functions appears close together improving code locality of non-cold parts
3062 of program. The paths leading to call of cold functions within code are marked
3063 as unlikely by the branch prediction mechanism. It is thus useful to mark
3064 functions used to handle unlikely conditions, such as @code{perror}, as cold to
3065 improve optimization of hot functions that do call marked functions in rare
3068 When profile feedback is available, via @option{-fprofile-use}, hot functions
3069 are automatically detected and this attribute is ignored.
3071 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
3073 @item regparm (@var{number})
3074 @cindex @code{regparm} attribute
3075 @cindex functions that are passed arguments in registers on the 386
3076 On the Intel 386, the @code{regparm} attribute causes the compiler to
3077 pass arguments number one to @var{number} if they are of integral type
3078 in registers EAX, EDX, and ECX instead of on the stack. Functions that
3079 take a variable number of arguments will continue to be passed all of their
3080 arguments on the stack.
3082 Beware that on some ELF systems this attribute is unsuitable for
3083 global functions in shared libraries with lazy binding (which is the
3084 default). Lazy binding will send the first call via resolving code in
3085 the loader, which might assume EAX, EDX and ECX can be clobbered, as
3086 per the standard calling conventions. Solaris 8 is affected by this.
3087 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
3088 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
3089 disabled with the linker or the loader if desired, to avoid the
3093 @cindex @code{sseregparm} attribute
3094 On the Intel 386 with SSE support, the @code{sseregparm} attribute
3095 causes the compiler to pass up to 3 floating point arguments in
3096 SSE registers instead of on the stack. Functions that take a
3097 variable number of arguments will continue to pass all of their
3098 floating point arguments on the stack.
3100 @item force_align_arg_pointer
3101 @cindex @code{force_align_arg_pointer} attribute
3102 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
3103 applied to individual function definitions, generating an alternate
3104 prologue and epilogue that realigns the runtime stack if necessary.
3105 This supports mixing legacy codes that run with a 4-byte aligned stack
3106 with modern codes that keep a 16-byte stack for SSE compatibility.
3109 @cindex @code{resbank} attribute
3110 On the SH2A target, this attribute enables the high-speed register
3111 saving and restoration using a register bank for @code{interrupt_handler}
3112 routines. Saving to the bank is performed automatically after the CPU
3113 accepts an interrupt that uses a register bank.
3115 The nineteen 32-bit registers comprising general register R0 to R14,
3116 control register GBR, and system registers MACH, MACL, and PR and the
3117 vector table address offset are saved into a register bank. Register
3118 banks are stacked in first-in last-out (FILO) sequence. Restoration
3119 from the bank is executed by issuing a RESBANK instruction.
3122 @cindex @code{returns_twice} attribute
3123 The @code{returns_twice} attribute tells the compiler that a function may
3124 return more than one time. The compiler will ensure that all registers
3125 are dead before calling such a function and will emit a warning about
3126 the variables that may be clobbered after the second return from the
3127 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3128 The @code{longjmp}-like counterpart of such function, if any, might need
3129 to be marked with the @code{noreturn} attribute.
3132 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
3133 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
3134 all registers except the stack pointer should be saved in the prologue
3135 regardless of whether they are used or not.
3137 @item section ("@var{section-name}")
3138 @cindex @code{section} function attribute
3139 Normally, the compiler places the code it generates in the @code{text} section.
3140 Sometimes, however, you need additional sections, or you need certain
3141 particular functions to appear in special sections. The @code{section}
3142 attribute specifies that a function lives in a particular section.
3143 For example, the declaration:
3146 extern void foobar (void) __attribute__ ((section ("bar")));
3150 puts the function @code{foobar} in the @code{bar} section.
3152 Some file formats do not support arbitrary sections so the @code{section}
3153 attribute is not available on all platforms.
3154 If you need to map the entire contents of a module to a particular
3155 section, consider using the facilities of the linker instead.
3158 @cindex @code{sentinel} function attribute
3159 This function attribute ensures that a parameter in a function call is
3160 an explicit @code{NULL}. The attribute is only valid on variadic
3161 functions. By default, the sentinel is located at position zero, the
3162 last parameter of the function call. If an optional integer position
3163 argument P is supplied to the attribute, the sentinel must be located at
3164 position P counting backwards from the end of the argument list.
3167 __attribute__ ((sentinel))
3169 __attribute__ ((sentinel(0)))
3172 The attribute is automatically set with a position of 0 for the built-in
3173 functions @code{execl} and @code{execlp}. The built-in function
3174 @code{execle} has the attribute set with a position of 1.
3176 A valid @code{NULL} in this context is defined as zero with any pointer
3177 type. If your system defines the @code{NULL} macro with an integer type
3178 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3179 with a copy that redefines NULL appropriately.
3181 The warnings for missing or incorrect sentinels are enabled with
3185 See long_call/short_call.
3188 See longcall/shortcall.
3191 @cindex signal handler functions on the AVR processors
3192 Use this attribute on the AVR to indicate that the specified
3193 function is a signal handler. The compiler will generate function
3194 entry and exit sequences suitable for use in a signal handler when this
3195 attribute is present. Interrupts will be disabled inside the function.
3198 Use this attribute on the SH to indicate an @code{interrupt_handler}
3199 function should switch to an alternate stack. It expects a string
3200 argument that names a global variable holding the address of the
3205 void f () __attribute__ ((interrupt_handler,
3206 sp_switch ("alt_stack")));
3210 @cindex functions that pop the argument stack on the 386
3211 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3212 assume that the called function will pop off the stack space used to
3213 pass arguments, unless it takes a variable number of arguments.
3215 @item syscall_linkage
3216 @cindex @code{syscall_linkage} attribute
3217 This attribute is used to modify the IA64 calling convention by marking
3218 all input registers as live at all function exits. This makes it possible
3219 to restart a system call after an interrupt without having to save/restore
3220 the input registers. This also prevents kernel data from leaking into
3224 @cindex @code{target} function attribute
3225 The @code{target} attribute is used to specify that a function is to
3226 be compiled with different target options than specified on the
3227 command line. This can be used for instance to have functions
3228 compiled with a different ISA (instruction set architecture) than the
3229 default. You can also use the @samp{#pragma GCC target} pragma to set
3230 more than one function to be compiled with specific target options.
3231 @xref{Function Specific Option Pragmas}, for details about the
3232 @samp{#pragma GCC target} pragma.
3234 For instance on a 386, you could compile one function with
3235 @code{target("sse4.1,arch=core2")} and another with
3236 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
3237 compiling the first function with @option{-msse4.1} and
3238 @option{-march=core2} options, and the second function with
3239 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
3240 user to make sure that a function is only invoked on a machine that
3241 supports the particular ISA it was compiled for (for example by using
3242 @code{cpuid} on 386 to determine what feature bits and architecture
3246 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3247 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3250 On the 386, the following options are allowed:
3255 @cindex @code{target("abm")} attribute
3256 Enable/disable the generation of the advanced bit instructions.
3260 @cindex @code{target("aes")} attribute
3261 Enable/disable the generation of the AES instructions.
3265 @cindex @code{target("mmx")} attribute
3266 Enable/disable the generation of the MMX instructions.
3270 @cindex @code{target("pclmul")} attribute
3271 Enable/disable the generation of the PCLMUL instructions.
3275 @cindex @code{target("popcnt")} attribute
3276 Enable/disable the generation of the POPCNT instruction.
3280 @cindex @code{target("sse")} attribute
3281 Enable/disable the generation of the SSE instructions.
3285 @cindex @code{target("sse2")} attribute
3286 Enable/disable the generation of the SSE2 instructions.
3290 @cindex @code{target("sse3")} attribute
3291 Enable/disable the generation of the SSE3 instructions.
3295 @cindex @code{target("sse4")} attribute
3296 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3301 @cindex @code{target("sse4.1")} attribute
3302 Enable/disable the generation of the sse4.1 instructions.
3306 @cindex @code{target("sse4.2")} attribute
3307 Enable/disable the generation of the sse4.2 instructions.
3311 @cindex @code{target("sse4a")} attribute
3312 Enable/disable the generation of the SSE4A instructions.
3316 @cindex @code{target("fma4")} attribute
3317 Enable/disable the generation of the FMA4 instructions.
3321 @cindex @code{target("xop")} attribute
3322 Enable/disable the generation of the XOP instructions.
3326 @cindex @code{target("lwp")} attribute
3327 Enable/disable the generation of the LWP instructions.
3331 @cindex @code{target("ssse3")} attribute
3332 Enable/disable the generation of the SSSE3 instructions.
3336 @cindex @code{target("cld")} attribute
3337 Enable/disable the generation of the CLD before string moves.
3339 @item fancy-math-387
3340 @itemx no-fancy-math-387
3341 @cindex @code{target("fancy-math-387")} attribute
3342 Enable/disable the generation of the @code{sin}, @code{cos}, and
3343 @code{sqrt} instructions on the 387 floating point unit.
3346 @itemx no-fused-madd
3347 @cindex @code{target("fused-madd")} attribute
3348 Enable/disable the generation of the fused multiply/add instructions.
3352 @cindex @code{target("ieee-fp")} attribute
3353 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3355 @item inline-all-stringops
3356 @itemx no-inline-all-stringops
3357 @cindex @code{target("inline-all-stringops")} attribute
3358 Enable/disable inlining of string operations.
3360 @item inline-stringops-dynamically
3361 @itemx no-inline-stringops-dynamically
3362 @cindex @code{target("inline-stringops-dynamically")} attribute
3363 Enable/disable the generation of the inline code to do small string
3364 operations and calling the library routines for large operations.
3366 @item align-stringops
3367 @itemx no-align-stringops
3368 @cindex @code{target("align-stringops")} attribute
3369 Do/do not align destination of inlined string operations.
3373 @cindex @code{target("recip")} attribute
3374 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3375 instructions followed an additional Newton-Raphson step instead of
3376 doing a floating point division.
3378 @item arch=@var{ARCH}
3379 @cindex @code{target("arch=@var{ARCH}")} attribute
3380 Specify the architecture to generate code for in compiling the function.
3382 @item tune=@var{TUNE}
3383 @cindex @code{target("tune=@var{TUNE}")} attribute
3384 Specify the architecture to tune for in compiling the function.
3386 @item fpmath=@var{FPMATH}
3387 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3388 Specify which floating point unit to use. The
3389 @code{target("fpmath=sse,387")} option must be specified as
3390 @code{target("fpmath=sse+387")} because the comma would separate
3394 On the 386, you can use either multiple strings to specify multiple
3395 options, or you can separate the option with a comma (@code{,}).
3397 On the 386, the inliner will not inline a function that has different
3398 target options than the caller, unless the callee has a subset of the
3399 target options of the caller. For example a function declared with
3400 @code{target("sse3")} can inline a function with
3401 @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
3403 The @code{target} attribute is not implemented in GCC versions earlier
3404 than 4.4, and at present only the 386 uses it.
3407 @cindex tiny data section on the H8/300H and H8S
3408 Use this attribute on the H8/300H and H8S to indicate that the specified
3409 variable should be placed into the tiny data section.
3410 The compiler will generate more efficient code for loads and stores
3411 on data in the tiny data section. Note the tiny data area is limited to
3412 slightly under 32kbytes of data.
3415 Use this attribute on the SH for an @code{interrupt_handler} to return using
3416 @code{trapa} instead of @code{rte}. This attribute expects an integer
3417 argument specifying the trap number to be used.
3420 @cindex @code{unused} attribute.
3421 This attribute, attached to a function, means that the function is meant
3422 to be possibly unused. GCC will not produce a warning for this
3426 @cindex @code{used} attribute.
3427 This attribute, attached to a function, means that code must be emitted
3428 for the function even if it appears that the function is not referenced.
3429 This is useful, for example, when the function is referenced only in
3433 @cindex @code{version_id} attribute
3434 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3435 symbol to contain a version string, thus allowing for function level
3436 versioning. HP-UX system header files may use version level functioning
3437 for some system calls.
3440 extern int foo () __attribute__((version_id ("20040821")));
3443 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3445 @item visibility ("@var{visibility_type}")
3446 @cindex @code{visibility} attribute
3447 This attribute affects the linkage of the declaration to which it is attached.
3448 There are four supported @var{visibility_type} values: default,
3449 hidden, protected or internal visibility.
3452 void __attribute__ ((visibility ("protected")))
3453 f () @{ /* @r{Do something.} */; @}
3454 int i __attribute__ ((visibility ("hidden")));
3457 The possible values of @var{visibility_type} correspond to the
3458 visibility settings in the ELF gABI.
3461 @c keep this list of visibilities in alphabetical order.
3464 Default visibility is the normal case for the object file format.
3465 This value is available for the visibility attribute to override other
3466 options that may change the assumed visibility of entities.
3468 On ELF, default visibility means that the declaration is visible to other
3469 modules and, in shared libraries, means that the declared entity may be
3472 On Darwin, default visibility means that the declaration is visible to
3475 Default visibility corresponds to ``external linkage'' in the language.
3478 Hidden visibility indicates that the entity declared will have a new
3479 form of linkage, which we'll call ``hidden linkage''. Two
3480 declarations of an object with hidden linkage refer to the same object
3481 if they are in the same shared object.
3484 Internal visibility is like hidden visibility, but with additional
3485 processor specific semantics. Unless otherwise specified by the
3486 psABI, GCC defines internal visibility to mean that a function is
3487 @emph{never} called from another module. Compare this with hidden
3488 functions which, while they cannot be referenced directly by other
3489 modules, can be referenced indirectly via function pointers. By
3490 indicating that a function cannot be called from outside the module,
3491 GCC may for instance omit the load of a PIC register since it is known
3492 that the calling function loaded the correct value.
3495 Protected visibility is like default visibility except that it
3496 indicates that references within the defining module will bind to the
3497 definition in that module. That is, the declared entity cannot be
3498 overridden by another module.
3502 All visibilities are supported on many, but not all, ELF targets
3503 (supported when the assembler supports the @samp{.visibility}
3504 pseudo-op). Default visibility is supported everywhere. Hidden
3505 visibility is supported on Darwin targets.
3507 The visibility attribute should be applied only to declarations which
3508 would otherwise have external linkage. The attribute should be applied
3509 consistently, so that the same entity should not be declared with
3510 different settings of the attribute.
3512 In C++, the visibility attribute applies to types as well as functions
3513 and objects, because in C++ types have linkage. A class must not have
3514 greater visibility than its non-static data member types and bases,
3515 and class members default to the visibility of their class. Also, a
3516 declaration without explicit visibility is limited to the visibility
3519 In C++, you can mark member functions and static member variables of a
3520 class with the visibility attribute. This is useful if you know a
3521 particular method or static member variable should only be used from
3522 one shared object; then you can mark it hidden while the rest of the
3523 class has default visibility. Care must be taken to avoid breaking
3524 the One Definition Rule; for example, it is usually not useful to mark
3525 an inline method as hidden without marking the whole class as hidden.
3527 A C++ namespace declaration can also have the visibility attribute.
3528 This attribute applies only to the particular namespace body, not to
3529 other definitions of the same namespace; it is equivalent to using
3530 @samp{#pragma GCC visibility} before and after the namespace
3531 definition (@pxref{Visibility Pragmas}).
3533 In C++, if a template argument has limited visibility, this
3534 restriction is implicitly propagated to the template instantiation.
3535 Otherwise, template instantiations and specializations default to the
3536 visibility of their template.
3538 If both the template and enclosing class have explicit visibility, the
3539 visibility from the template is used.
3542 @cindex @code{vliw} attribute
3543 On MeP, the @code{vliw} attribute tells the compiler to emit
3544 instructions in VLIW mode instead of core mode. Note that this
3545 attribute is not allowed unless a VLIW coprocessor has been configured
3546 and enabled through command line options.
3548 @item warn_unused_result
3549 @cindex @code{warn_unused_result} attribute
3550 The @code{warn_unused_result} attribute causes a warning to be emitted
3551 if a caller of the function with this attribute does not use its
3552 return value. This is useful for functions where not checking
3553 the result is either a security problem or always a bug, such as
3557 int fn () __attribute__ ((warn_unused_result));
3560 if (fn () < 0) return -1;
3566 results in warning on line 5.
3569 @cindex @code{weak} attribute
3570 The @code{weak} attribute causes the declaration to be emitted as a weak
3571 symbol rather than a global. This is primarily useful in defining
3572 library functions which can be overridden in user code, though it can
3573 also be used with non-function declarations. Weak symbols are supported
3574 for ELF targets, and also for a.out targets when using the GNU assembler
3578 @itemx weakref ("@var{target}")
3579 @cindex @code{weakref} attribute
3580 The @code{weakref} attribute marks a declaration as a weak reference.
3581 Without arguments, it should be accompanied by an @code{alias} attribute
3582 naming the target symbol. Optionally, the @var{target} may be given as
3583 an argument to @code{weakref} itself. In either case, @code{weakref}
3584 implicitly marks the declaration as @code{weak}. Without a
3585 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3586 @code{weakref} is equivalent to @code{weak}.
3589 static int x() __attribute__ ((weakref ("y")));
3590 /* is equivalent to... */
3591 static int x() __attribute__ ((weak, weakref, alias ("y")));
3593 static int x() __attribute__ ((weakref));
3594 static int x() __attribute__ ((alias ("y")));
3597 A weak reference is an alias that does not by itself require a
3598 definition to be given for the target symbol. If the target symbol is
3599 only referenced through weak references, then it becomes a @code{weak}
3600 undefined symbol. If it is directly referenced, however, then such
3601 strong references prevail, and a definition will be required for the
3602 symbol, not necessarily in the same translation unit.
3604 The effect is equivalent to moving all references to the alias to a
3605 separate translation unit, renaming the alias to the aliased symbol,
3606 declaring it as weak, compiling the two separate translation units and
3607 performing a reloadable link on them.
3609 At present, a declaration to which @code{weakref} is attached can
3610 only be @code{static}.
3614 You can specify multiple attributes in a declaration by separating them
3615 by commas within the double parentheses or by immediately following an
3616 attribute declaration with another attribute declaration.
3618 @cindex @code{#pragma}, reason for not using
3619 @cindex pragma, reason for not using
3620 Some people object to the @code{__attribute__} feature, suggesting that
3621 ISO C's @code{#pragma} should be used instead. At the time
3622 @code{__attribute__} was designed, there were two reasons for not doing
3627 It is impossible to generate @code{#pragma} commands from a macro.
3630 There is no telling what the same @code{#pragma} might mean in another
3634 These two reasons applied to almost any application that might have been
3635 proposed for @code{#pragma}. It was basically a mistake to use
3636 @code{#pragma} for @emph{anything}.
3638 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
3639 to be generated from macros. In addition, a @code{#pragma GCC}
3640 namespace is now in use for GCC-specific pragmas. However, it has been
3641 found convenient to use @code{__attribute__} to achieve a natural
3642 attachment of attributes to their corresponding declarations, whereas
3643 @code{#pragma GCC} is of use for constructs that do not naturally form
3644 part of the grammar. @xref{Other Directives,,Miscellaneous
3645 Preprocessing Directives, cpp, The GNU C Preprocessor}.
3647 @node Attribute Syntax
3648 @section Attribute Syntax
3649 @cindex attribute syntax
3651 This section describes the syntax with which @code{__attribute__} may be
3652 used, and the constructs to which attribute specifiers bind, for the C
3653 language. Some details may vary for C++ and Objective-C@. Because of
3654 infelicities in the grammar for attributes, some forms described here
3655 may not be successfully parsed in all cases.
3657 There are some problems with the semantics of attributes in C++. For
3658 example, there are no manglings for attributes, although they may affect
3659 code generation, so problems may arise when attributed types are used in
3660 conjunction with templates or overloading. Similarly, @code{typeid}
3661 does not distinguish between types with different attributes. Support
3662 for attributes in C++ may be restricted in future to attributes on
3663 declarations only, but not on nested declarators.
3665 @xref{Function Attributes}, for details of the semantics of attributes
3666 applying to functions. @xref{Variable Attributes}, for details of the
3667 semantics of attributes applying to variables. @xref{Type Attributes},
3668 for details of the semantics of attributes applying to structure, union
3669 and enumerated types.
3671 An @dfn{attribute specifier} is of the form
3672 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
3673 is a possibly empty comma-separated sequence of @dfn{attributes}, where
3674 each attribute is one of the following:
3678 Empty. Empty attributes are ignored.
3681 A word (which may be an identifier such as @code{unused}, or a reserved
3682 word such as @code{const}).
3685 A word, followed by, in parentheses, parameters for the attribute.
3686 These parameters take one of the following forms:
3690 An identifier. For example, @code{mode} attributes use this form.
3693 An identifier followed by a comma and a non-empty comma-separated list
3694 of expressions. For example, @code{format} attributes use this form.
3697 A possibly empty comma-separated list of expressions. For example,
3698 @code{format_arg} attributes use this form with the list being a single
3699 integer constant expression, and @code{alias} attributes use this form
3700 with the list being a single string constant.
3704 An @dfn{attribute specifier list} is a sequence of one or more attribute
3705 specifiers, not separated by any other tokens.
3707 In GNU C, an attribute specifier list may appear after the colon following a
3708 label, other than a @code{case} or @code{default} label. The only
3709 attribute it makes sense to use after a label is @code{unused}. This
3710 feature is intended for code generated by programs which contains labels
3711 that may be unused but which is compiled with @option{-Wall}. It would
3712 not normally be appropriate to use in it human-written code, though it
3713 could be useful in cases where the code that jumps to the label is
3714 contained within an @code{#ifdef} conditional. GNU C++ only permits
3715 attributes on labels if the attribute specifier is immediately
3716 followed by a semicolon (i.e., the label applies to an empty
3717 statement). If the semicolon is missing, C++ label attributes are
3718 ambiguous, as it is permissible for a declaration, which could begin
3719 with an attribute list, to be labelled in C++. Declarations cannot be
3720 labelled in C90 or C99, so the ambiguity does not arise there.
3722 An attribute specifier list may appear as part of a @code{struct},
3723 @code{union} or @code{enum} specifier. It may go either immediately
3724 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3725 the closing brace. The former syntax is preferred.
3726 Where attribute specifiers follow the closing brace, they are considered
3727 to relate to the structure, union or enumerated type defined, not to any
3728 enclosing declaration the type specifier appears in, and the type
3729 defined is not complete until after the attribute specifiers.
3730 @c Otherwise, there would be the following problems: a shift/reduce
3731 @c conflict between attributes binding the struct/union/enum and
3732 @c binding to the list of specifiers/qualifiers; and "aligned"
3733 @c attributes could use sizeof for the structure, but the size could be
3734 @c changed later by "packed" attributes.
3736 Otherwise, an attribute specifier appears as part of a declaration,
3737 counting declarations of unnamed parameters and type names, and relates
3738 to that declaration (which may be nested in another declaration, for
3739 example in the case of a parameter declaration), or to a particular declarator
3740 within a declaration. Where an
3741 attribute specifier is applied to a parameter declared as a function or
3742 an array, it should apply to the function or array rather than the
3743 pointer to which the parameter is implicitly converted, but this is not
3744 yet correctly implemented.
3746 Any list of specifiers and qualifiers at the start of a declaration may
3747 contain attribute specifiers, whether or not such a list may in that
3748 context contain storage class specifiers. (Some attributes, however,
3749 are essentially in the nature of storage class specifiers, and only make
3750 sense where storage class specifiers may be used; for example,
3751 @code{section}.) There is one necessary limitation to this syntax: the
3752 first old-style parameter declaration in a function definition cannot
3753 begin with an attribute specifier, because such an attribute applies to
3754 the function instead by syntax described below (which, however, is not
3755 yet implemented in this case). In some other cases, attribute
3756 specifiers are permitted by this grammar but not yet supported by the
3757 compiler. All attribute specifiers in this place relate to the
3758 declaration as a whole. In the obsolescent usage where a type of
3759 @code{int} is implied by the absence of type specifiers, such a list of
3760 specifiers and qualifiers may be an attribute specifier list with no
3761 other specifiers or qualifiers.
3763 At present, the first parameter in a function prototype must have some
3764 type specifier which is not an attribute specifier; this resolves an
3765 ambiguity in the interpretation of @code{void f(int
3766 (__attribute__((foo)) x))}, but is subject to change. At present, if
3767 the parentheses of a function declarator contain only attributes then
3768 those attributes are ignored, rather than yielding an error or warning
3769 or implying a single parameter of type int, but this is subject to
3772 An attribute specifier list may appear immediately before a declarator
3773 (other than the first) in a comma-separated list of declarators in a
3774 declaration of more than one identifier using a single list of
3775 specifiers and qualifiers. Such attribute specifiers apply
3776 only to the identifier before whose declarator they appear. For
3780 __attribute__((noreturn)) void d0 (void),
3781 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
3786 the @code{noreturn} attribute applies to all the functions
3787 declared; the @code{format} attribute only applies to @code{d1}.
3789 An attribute specifier list may appear immediately before the comma,
3790 @code{=} or semicolon terminating the declaration of an identifier other
3791 than a function definition. Such attribute specifiers apply
3792 to the declared object or function. Where an
3793 assembler name for an object or function is specified (@pxref{Asm
3794 Labels}), the attribute must follow the @code{asm}
3797 An attribute specifier list may, in future, be permitted to appear after
3798 the declarator in a function definition (before any old-style parameter
3799 declarations or the function body).
3801 Attribute specifiers may be mixed with type qualifiers appearing inside
3802 the @code{[]} of a parameter array declarator, in the C99 construct by
3803 which such qualifiers are applied to the pointer to which the array is
3804 implicitly converted. Such attribute specifiers apply to the pointer,
3805 not to the array, but at present this is not implemented and they are
3808 An attribute specifier list may appear at the start of a nested
3809 declarator. At present, there are some limitations in this usage: the
3810 attributes correctly apply to the declarator, but for most individual
3811 attributes the semantics this implies are not implemented.
3812 When attribute specifiers follow the @code{*} of a pointer
3813 declarator, they may be mixed with any type qualifiers present.
3814 The following describes the formal semantics of this syntax. It will make the
3815 most sense if you are familiar with the formal specification of
3816 declarators in the ISO C standard.
3818 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
3819 D1}, where @code{T} contains declaration specifiers that specify a type
3820 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
3821 contains an identifier @var{ident}. The type specified for @var{ident}
3822 for derived declarators whose type does not include an attribute
3823 specifier is as in the ISO C standard.
3825 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
3826 and the declaration @code{T D} specifies the type
3827 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3828 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3829 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
3831 If @code{D1} has the form @code{*
3832 @var{type-qualifier-and-attribute-specifier-list} D}, and the
3833 declaration @code{T D} specifies the type
3834 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3835 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3836 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
3842 void (__attribute__((noreturn)) ****f) (void);
3846 specifies the type ``pointer to pointer to pointer to pointer to
3847 non-returning function returning @code{void}''. As another example,
3850 char *__attribute__((aligned(8))) *f;
3854 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
3855 Note again that this does not work with most attributes; for example,
3856 the usage of @samp{aligned} and @samp{noreturn} attributes given above
3857 is not yet supported.
3859 For compatibility with existing code written for compiler versions that
3860 did not implement attributes on nested declarators, some laxity is
3861 allowed in the placing of attributes. If an attribute that only applies
3862 to types is applied to a declaration, it will be treated as applying to
3863 the type of that declaration. If an attribute that only applies to
3864 declarations is applied to the type of a declaration, it will be treated
3865 as applying to that declaration; and, for compatibility with code
3866 placing the attributes immediately before the identifier declared, such
3867 an attribute applied to a function return type will be treated as
3868 applying to the function type, and such an attribute applied to an array
3869 element type will be treated as applying to the array type. If an
3870 attribute that only applies to function types is applied to a
3871 pointer-to-function type, it will be treated as applying to the pointer
3872 target type; if such an attribute is applied to a function return type
3873 that is not a pointer-to-function type, it will be treated as applying
3874 to the function type.
3876 @node Function Prototypes
3877 @section Prototypes and Old-Style Function Definitions
3878 @cindex function prototype declarations
3879 @cindex old-style function definitions
3880 @cindex promotion of formal parameters
3882 GNU C extends ISO C to allow a function prototype to override a later
3883 old-style non-prototype definition. Consider the following example:
3886 /* @r{Use prototypes unless the compiler is old-fashioned.} */
3893 /* @r{Prototype function declaration.} */
3894 int isroot P((uid_t));
3896 /* @r{Old-style function definition.} */
3898 isroot (x) /* @r{??? lossage here ???} */
3905 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
3906 not allow this example, because subword arguments in old-style
3907 non-prototype definitions are promoted. Therefore in this example the
3908 function definition's argument is really an @code{int}, which does not
3909 match the prototype argument type of @code{short}.
3911 This restriction of ISO C makes it hard to write code that is portable
3912 to traditional C compilers, because the programmer does not know
3913 whether the @code{uid_t} type is @code{short}, @code{int}, or
3914 @code{long}. Therefore, in cases like these GNU C allows a prototype
3915 to override a later old-style definition. More precisely, in GNU C, a
3916 function prototype argument type overrides the argument type specified
3917 by a later old-style definition if the former type is the same as the
3918 latter type before promotion. Thus in GNU C the above example is
3919 equivalent to the following:
3932 GNU C++ does not support old-style function definitions, so this
3933 extension is irrelevant.
3936 @section C++ Style Comments
3938 @cindex C++ comments
3939 @cindex comments, C++ style
3941 In GNU C, you may use C++ style comments, which start with @samp{//} and
3942 continue until the end of the line. Many other C implementations allow
3943 such comments, and they are included in the 1999 C standard. However,
3944 C++ style comments are not recognized if you specify an @option{-std}
3945 option specifying a version of ISO C before C99, or @option{-ansi}
3946 (equivalent to @option{-std=c90}).
3949 @section Dollar Signs in Identifier Names
3951 @cindex dollar signs in identifier names
3952 @cindex identifier names, dollar signs in
3954 In GNU C, you may normally use dollar signs in identifier names.
3955 This is because many traditional C implementations allow such identifiers.
3956 However, dollar signs in identifiers are not supported on a few target
3957 machines, typically because the target assembler does not allow them.
3959 @node Character Escapes
3960 @section The Character @key{ESC} in Constants
3962 You can use the sequence @samp{\e} in a string or character constant to
3963 stand for the ASCII character @key{ESC}.
3966 @section Inquiring on Alignment of Types or Variables
3968 @cindex type alignment
3969 @cindex variable alignment
3971 The keyword @code{__alignof__} allows you to inquire about how an object
3972 is aligned, or the minimum alignment usually required by a type. Its
3973 syntax is just like @code{sizeof}.
3975 For example, if the target machine requires a @code{double} value to be
3976 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3977 This is true on many RISC machines. On more traditional machine
3978 designs, @code{__alignof__ (double)} is 4 or even 2.
3980 Some machines never actually require alignment; they allow reference to any
3981 data type even at an odd address. For these machines, @code{__alignof__}
3982 reports the smallest alignment that GCC will give the data type, usually as
3983 mandated by the target ABI.
3985 If the operand of @code{__alignof__} is an lvalue rather than a type,
3986 its value is the required alignment for its type, taking into account
3987 any minimum alignment specified with GCC's @code{__attribute__}
3988 extension (@pxref{Variable Attributes}). For example, after this
3992 struct foo @{ int x; char y; @} foo1;
3996 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3997 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3999 It is an error to ask for the alignment of an incomplete type.
4001 @node Variable Attributes
4002 @section Specifying Attributes of Variables
4003 @cindex attribute of variables
4004 @cindex variable attributes
4006 The keyword @code{__attribute__} allows you to specify special
4007 attributes of variables or structure fields. This keyword is followed
4008 by an attribute specification inside double parentheses. Some
4009 attributes are currently defined generically for variables.
4010 Other attributes are defined for variables on particular target
4011 systems. Other attributes are available for functions
4012 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
4013 Other front ends might define more attributes
4014 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
4016 You may also specify attributes with @samp{__} preceding and following
4017 each keyword. This allows you to use them in header files without
4018 being concerned about a possible macro of the same name. For example,
4019 you may use @code{__aligned__} instead of @code{aligned}.
4021 @xref{Attribute Syntax}, for details of the exact syntax for using
4025 @cindex @code{aligned} attribute
4026 @item aligned (@var{alignment})
4027 This attribute specifies a minimum alignment for the variable or
4028 structure field, measured in bytes. For example, the declaration:
4031 int x __attribute__ ((aligned (16))) = 0;
4035 causes the compiler to allocate the global variable @code{x} on a
4036 16-byte boundary. On a 68040, this could be used in conjunction with
4037 an @code{asm} expression to access the @code{move16} instruction which
4038 requires 16-byte aligned operands.
4040 You can also specify the alignment of structure fields. For example, to
4041 create a double-word aligned @code{int} pair, you could write:
4044 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
4048 This is an alternative to creating a union with a @code{double} member
4049 that forces the union to be double-word aligned.
4051 As in the preceding examples, you can explicitly specify the alignment
4052 (in bytes) that you wish the compiler to use for a given variable or
4053 structure field. Alternatively, you can leave out the alignment factor
4054 and just ask the compiler to align a variable or field to the
4055 default alignment for the target architecture you are compiling for.
4056 The default alignment is sufficient for all scalar types, but may not be
4057 enough for all vector types on a target which supports vector operations.
4058 The default alignment is fixed for a particular target ABI.
4060 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
4061 which is the largest alignment ever used for any data type on the
4062 target machine you are compiling for. For example, you could write:
4065 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
4068 The compiler automatically sets the alignment for the declared
4069 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
4070 often make copy operations more efficient, because the compiler can
4071 use whatever instructions copy the biggest chunks of memory when
4072 performing copies to or from the variables or fields that you have
4073 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
4074 may change depending on command line options.
4076 When used on a struct, or struct member, the @code{aligned} attribute can
4077 only increase the alignment; in order to decrease it, the @code{packed}
4078 attribute must be specified as well. When used as part of a typedef, the
4079 @code{aligned} attribute can both increase and decrease alignment, and
4080 specifying the @code{packed} attribute will generate a warning.
4082 Note that the effectiveness of @code{aligned} attributes may be limited
4083 by inherent limitations in your linker. On many systems, the linker is
4084 only able to arrange for variables to be aligned up to a certain maximum
4085 alignment. (For some linkers, the maximum supported alignment may
4086 be very very small.) If your linker is only able to align variables
4087 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4088 in an @code{__attribute__} will still only provide you with 8 byte
4089 alignment. See your linker documentation for further information.
4091 The @code{aligned} attribute can also be used for functions
4092 (@pxref{Function Attributes}.)
4094 @item cleanup (@var{cleanup_function})
4095 @cindex @code{cleanup} attribute
4096 The @code{cleanup} attribute runs a function when the variable goes
4097 out of scope. This attribute can only be applied to auto function
4098 scope variables; it may not be applied to parameters or variables
4099 with static storage duration. The function must take one parameter,
4100 a pointer to a type compatible with the variable. The return value
4101 of the function (if any) is ignored.
4103 If @option{-fexceptions} is enabled, then @var{cleanup_function}
4104 will be run during the stack unwinding that happens during the
4105 processing of the exception. Note that the @code{cleanup} attribute
4106 does not allow the exception to be caught, only to perform an action.
4107 It is undefined what happens if @var{cleanup_function} does not
4112 @cindex @code{common} attribute
4113 @cindex @code{nocommon} attribute
4116 The @code{common} attribute requests GCC to place a variable in
4117 ``common'' storage. The @code{nocommon} attribute requests the
4118 opposite---to allocate space for it directly.
4120 These attributes override the default chosen by the
4121 @option{-fno-common} and @option{-fcommon} flags respectively.
4124 @itemx deprecated (@var{msg})
4125 @cindex @code{deprecated} attribute
4126 The @code{deprecated} attribute results in a warning if the variable
4127 is used anywhere in the source file. This is useful when identifying
4128 variables that are expected to be removed in a future version of a
4129 program. The warning also includes the location of the declaration
4130 of the deprecated variable, to enable users to easily find further
4131 information about why the variable is deprecated, or what they should
4132 do instead. Note that the warning only occurs for uses:
4135 extern int old_var __attribute__ ((deprecated));
4137 int new_fn () @{ return old_var; @}
4140 results in a warning on line 3 but not line 2. The optional msg
4141 argument, which must be a string, will be printed in the warning if
4144 The @code{deprecated} attribute can also be used for functions and
4145 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
4147 @item mode (@var{mode})
4148 @cindex @code{mode} attribute
4149 This attribute specifies the data type for the declaration---whichever
4150 type corresponds to the mode @var{mode}. This in effect lets you
4151 request an integer or floating point type according to its width.
4153 You may also specify a mode of @samp{byte} or @samp{__byte__} to
4154 indicate the mode corresponding to a one-byte integer, @samp{word} or
4155 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
4156 or @samp{__pointer__} for the mode used to represent pointers.
4159 @cindex @code{packed} attribute
4160 The @code{packed} attribute specifies that a variable or structure field
4161 should have the smallest possible alignment---one byte for a variable,
4162 and one bit for a field, unless you specify a larger value with the
4163 @code{aligned} attribute.
4165 Here is a structure in which the field @code{x} is packed, so that it
4166 immediately follows @code{a}:
4172 int x[2] __attribute__ ((packed));
4176 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
4177 @code{packed} attribute on bit-fields of type @code{char}. This has
4178 been fixed in GCC 4.4 but the change can lead to differences in the
4179 structure layout. See the documentation of
4180 @option{-Wpacked-bitfield-compat} for more information.
4182 @item section ("@var{section-name}")
4183 @cindex @code{section} variable attribute
4184 Normally, the compiler places the objects it generates in sections like
4185 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
4186 or you need certain particular variables to appear in special sections,
4187 for example to map to special hardware. The @code{section}
4188 attribute specifies that a variable (or function) lives in a particular
4189 section. For example, this small program uses several specific section names:
4192 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
4193 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
4194 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
4195 int init_data __attribute__ ((section ("INITDATA")));
4199 /* @r{Initialize stack pointer} */
4200 init_sp (stack + sizeof (stack));
4202 /* @r{Initialize initialized data} */
4203 memcpy (&init_data, &data, &edata - &data);
4205 /* @r{Turn on the serial ports} */
4212 Use the @code{section} attribute with
4213 @emph{global} variables and not @emph{local} variables,
4214 as shown in the example.
4216 You may use the @code{section} attribute with initialized or
4217 uninitialized global variables but the linker requires
4218 each object be defined once, with the exception that uninitialized
4219 variables tentatively go in the @code{common} (or @code{bss}) section
4220 and can be multiply ``defined''. Using the @code{section} attribute
4221 will change what section the variable goes into and may cause the
4222 linker to issue an error if an uninitialized variable has multiple
4223 definitions. You can force a variable to be initialized with the
4224 @option{-fno-common} flag or the @code{nocommon} attribute.
4226 Some file formats do not support arbitrary sections so the @code{section}
4227 attribute is not available on all platforms.
4228 If you need to map the entire contents of a module to a particular
4229 section, consider using the facilities of the linker instead.
4232 @cindex @code{shared} variable attribute
4233 On Microsoft Windows, in addition to putting variable definitions in a named
4234 section, the section can also be shared among all running copies of an
4235 executable or DLL@. For example, this small program defines shared data
4236 by putting it in a named section @code{shared} and marking the section
4240 int foo __attribute__((section ("shared"), shared)) = 0;
4245 /* @r{Read and write foo. All running
4246 copies see the same value.} */
4252 You may only use the @code{shared} attribute along with @code{section}
4253 attribute with a fully initialized global definition because of the way
4254 linkers work. See @code{section} attribute for more information.
4256 The @code{shared} attribute is only available on Microsoft Windows@.
4258 @item tls_model ("@var{tls_model}")
4259 @cindex @code{tls_model} attribute
4260 The @code{tls_model} attribute sets thread-local storage model
4261 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
4262 overriding @option{-ftls-model=} command-line switch on a per-variable
4264 The @var{tls_model} argument should be one of @code{global-dynamic},
4265 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
4267 Not all targets support this attribute.
4270 This attribute, attached to a variable, means that the variable is meant
4271 to be possibly unused. GCC will not produce a warning for this
4275 This attribute, attached to a variable, means that the variable must be
4276 emitted even if it appears that the variable is not referenced.
4278 @item vector_size (@var{bytes})
4279 This attribute specifies the vector size for the variable, measured in
4280 bytes. For example, the declaration:
4283 int foo __attribute__ ((vector_size (16)));
4287 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
4288 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
4289 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
4291 This attribute is only applicable to integral and float scalars,
4292 although arrays, pointers, and function return values are allowed in
4293 conjunction with this construct.
4295 Aggregates with this attribute are invalid, even if they are of the same
4296 size as a corresponding scalar. For example, the declaration:
4299 struct S @{ int a; @};
4300 struct S __attribute__ ((vector_size (16))) foo;
4304 is invalid even if the size of the structure is the same as the size of
4308 The @code{selectany} attribute causes an initialized global variable to
4309 have link-once semantics. When multiple definitions of the variable are
4310 encountered by the linker, the first is selected and the remainder are
4311 discarded. Following usage by the Microsoft compiler, the linker is told
4312 @emph{not} to warn about size or content differences of the multiple
4315 Although the primary usage of this attribute is for POD types, the
4316 attribute can also be applied to global C++ objects that are initialized
4317 by a constructor. In this case, the static initialization and destruction
4318 code for the object is emitted in each translation defining the object,
4319 but the calls to the constructor and destructor are protected by a
4320 link-once guard variable.
4322 The @code{selectany} attribute is only available on Microsoft Windows
4323 targets. You can use @code{__declspec (selectany)} as a synonym for
4324 @code{__attribute__ ((selectany))} for compatibility with other
4328 The @code{weak} attribute is described in @ref{Function Attributes}.
4331 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4334 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4338 @subsection Blackfin Variable Attributes
4340 Three attributes are currently defined for the Blackfin.
4346 @cindex @code{l1_data} variable attribute
4347 @cindex @code{l1_data_A} variable attribute
4348 @cindex @code{l1_data_B} variable attribute
4349 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4350 Variables with @code{l1_data} attribute will be put into the specific section
4351 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4352 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4353 attribute will be put into the specific section named @code{.l1.data.B}.
4356 @cindex @code{l2} variable attribute
4357 Use this attribute on the Blackfin to place the variable into L2 SRAM.
4358 Variables with @code{l2} attribute will be put into the specific section
4359 named @code{.l2.data}.
4362 @subsection M32R/D Variable Attributes
4364 One attribute is currently defined for the M32R/D@.
4367 @item model (@var{model-name})
4368 @cindex variable addressability on the M32R/D
4369 Use this attribute on the M32R/D to set the addressability of an object.
4370 The identifier @var{model-name} is one of @code{small}, @code{medium},
4371 or @code{large}, representing each of the code models.
4373 Small model objects live in the lower 16MB of memory (so that their
4374 addresses can be loaded with the @code{ld24} instruction).
4376 Medium and large model objects may live anywhere in the 32-bit address space
4377 (the compiler will generate @code{seth/add3} instructions to load their
4381 @anchor{MeP Variable Attributes}
4382 @subsection MeP Variable Attributes
4384 The MeP target has a number of addressing modes and busses. The
4385 @code{near} space spans the standard memory space's first 16 megabytes
4386 (24 bits). The @code{far} space spans the entire 32-bit memory space.
4387 The @code{based} space is a 128 byte region in the memory space which
4388 is addressed relative to the @code{$tp} register. The @code{tiny}
4389 space is a 65536 byte region relative to the @code{$gp} register. In
4390 addition to these memory regions, the MeP target has a separate 16-bit
4391 control bus which is specified with @code{cb} attributes.
4396 Any variable with the @code{based} attribute will be assigned to the
4397 @code{.based} section, and will be accessed with relative to the
4398 @code{$tp} register.
4401 Likewise, the @code{tiny} attribute assigned variables to the
4402 @code{.tiny} section, relative to the @code{$gp} register.
4405 Variables with the @code{near} attribute are assumed to have addresses
4406 that fit in a 24-bit addressing mode. This is the default for large
4407 variables (@code{-mtiny=4} is the default) but this attribute can
4408 override @code{-mtiny=} for small variables, or override @code{-ml}.
4411 Variables with the @code{far} attribute are addressed using a full
4412 32-bit address. Since this covers the entire memory space, this
4413 allows modules to make no assumptions about where variables might be
4417 @itemx io (@var{addr})
4418 Variables with the @code{io} attribute are used to address
4419 memory-mapped peripherals. If an address is specified, the variable
4420 is assigned that address, else it is not assigned an address (it is
4421 assumed some other module will assign an address). Example:
4424 int timer_count __attribute__((io(0x123)));
4428 @itemx cb (@var{addr})
4429 Variables with the @code{cb} attribute are used to access the control
4430 bus, using special instructions. @code{addr} indicates the control bus
4434 int cpu_clock __attribute__((cb(0x123)));
4439 @anchor{i386 Variable Attributes}
4440 @subsection i386 Variable Attributes
4442 Two attributes are currently defined for i386 configurations:
4443 @code{ms_struct} and @code{gcc_struct}
4448 @cindex @code{ms_struct} attribute
4449 @cindex @code{gcc_struct} attribute
4451 If @code{packed} is used on a structure, or if bit-fields are used
4452 it may be that the Microsoft ABI packs them differently
4453 than GCC would normally pack them. Particularly when moving packed
4454 data between functions compiled with GCC and the native Microsoft compiler
4455 (either via function call or as data in a file), it may be necessary to access
4458 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4459 compilers to match the native Microsoft compiler.
4461 The Microsoft structure layout algorithm is fairly simple with the exception
4462 of the bitfield packing:
4464 The padding and alignment of members of structures and whether a bit field
4465 can straddle a storage-unit boundary
4468 @item Structure members are stored sequentially in the order in which they are
4469 declared: the first member has the lowest memory address and the last member
4472 @item Every data object has an alignment-requirement. The alignment-requirement
4473 for all data except structures, unions, and arrays is either the size of the
4474 object or the current packing size (specified with either the aligned attribute
4475 or the pack pragma), whichever is less. For structures, unions, and arrays,
4476 the alignment-requirement is the largest alignment-requirement of its members.
4477 Every object is allocated an offset so that:
4479 offset % alignment-requirement == 0
4481 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4482 unit if the integral types are the same size and if the next bit field fits
4483 into the current allocation unit without crossing the boundary imposed by the
4484 common alignment requirements of the bit fields.
4487 Handling of zero-length bitfields:
4489 MSVC interprets zero-length bitfields in the following ways:
4492 @item If a zero-length bitfield is inserted between two bitfields that would
4493 normally be coalesced, the bitfields will not be coalesced.
4500 unsigned long bf_1 : 12;
4502 unsigned long bf_2 : 12;
4506 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4507 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4509 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4510 alignment of the zero-length bitfield is greater than the member that follows it,
4511 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4531 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4532 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4533 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4536 Taking this into account, it is important to note the following:
4539 @item If a zero-length bitfield follows a normal bitfield, the type of the
4540 zero-length bitfield may affect the alignment of the structure as whole. For
4541 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4542 normal bitfield, and is of type short.
4544 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4545 still affect the alignment of the structure:
4555 Here, @code{t4} will take up 4 bytes.
4558 @item Zero-length bitfields following non-bitfield members are ignored:
4569 Here, @code{t5} will take up 2 bytes.
4573 @subsection PowerPC Variable Attributes
4575 Three attributes currently are defined for PowerPC configurations:
4576 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4578 For full documentation of the struct attributes please see the
4579 documentation in @ref{i386 Variable Attributes}.
4581 For documentation of @code{altivec} attribute please see the
4582 documentation in @ref{PowerPC Type Attributes}.
4584 @subsection SPU Variable Attributes
4586 The SPU supports the @code{spu_vector} attribute for variables. For
4587 documentation of this attribute please see the documentation in
4588 @ref{SPU Type Attributes}.
4590 @subsection Xstormy16 Variable Attributes
4592 One attribute is currently defined for xstormy16 configurations:
4597 @cindex @code{below100} attribute
4599 If a variable has the @code{below100} attribute (@code{BELOW100} is
4600 allowed also), GCC will place the variable in the first 0x100 bytes of
4601 memory and use special opcodes to access it. Such variables will be
4602 placed in either the @code{.bss_below100} section or the
4603 @code{.data_below100} section.
4607 @subsection AVR Variable Attributes
4611 @cindex @code{progmem} variable attribute
4612 The @code{progmem} attribute is used on the AVR to place data in the Program
4613 Memory address space. The AVR is a Harvard Architecture processor and data
4614 normally resides in the Data Memory address space.
4617 @node Type Attributes
4618 @section Specifying Attributes of Types
4619 @cindex attribute of types
4620 @cindex type attributes
4622 The keyword @code{__attribute__} allows you to specify special
4623 attributes of @code{struct} and @code{union} types when you define
4624 such types. This keyword is followed by an attribute specification
4625 inside double parentheses. Seven attributes are currently defined for
4626 types: @code{aligned}, @code{packed}, @code{transparent_union},
4627 @code{unused}, @code{deprecated}, @code{visibility}, and
4628 @code{may_alias}. Other attributes are defined for functions
4629 (@pxref{Function Attributes}) and for variables (@pxref{Variable
4632 You may also specify any one of these attributes with @samp{__}
4633 preceding and following its keyword. This allows you to use these
4634 attributes in header files without being concerned about a possible
4635 macro of the same name. For example, you may use @code{__aligned__}
4636 instead of @code{aligned}.
4638 You may specify type attributes in an enum, struct or union type
4639 declaration or definition, or for other types in a @code{typedef}
4642 For an enum, struct or union type, you may specify attributes either
4643 between the enum, struct or union tag and the name of the type, or
4644 just past the closing curly brace of the @emph{definition}. The
4645 former syntax is preferred.
4647 @xref{Attribute Syntax}, for details of the exact syntax for using
4651 @cindex @code{aligned} attribute
4652 @item aligned (@var{alignment})
4653 This attribute specifies a minimum alignment (in bytes) for variables
4654 of the specified type. For example, the declarations:
4657 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
4658 typedef int more_aligned_int __attribute__ ((aligned (8)));
4662 force the compiler to insure (as far as it can) that each variable whose
4663 type is @code{struct S} or @code{more_aligned_int} will be allocated and
4664 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
4665 variables of type @code{struct S} aligned to 8-byte boundaries allows
4666 the compiler to use the @code{ldd} and @code{std} (doubleword load and
4667 store) instructions when copying one variable of type @code{struct S} to
4668 another, thus improving run-time efficiency.
4670 Note that the alignment of any given @code{struct} or @code{union} type
4671 is required by the ISO C standard to be at least a perfect multiple of
4672 the lowest common multiple of the alignments of all of the members of
4673 the @code{struct} or @code{union} in question. This means that you @emph{can}
4674 effectively adjust the alignment of a @code{struct} or @code{union}
4675 type by attaching an @code{aligned} attribute to any one of the members
4676 of such a type, but the notation illustrated in the example above is a
4677 more obvious, intuitive, and readable way to request the compiler to
4678 adjust the alignment of an entire @code{struct} or @code{union} type.
4680 As in the preceding example, you can explicitly specify the alignment
4681 (in bytes) that you wish the compiler to use for a given @code{struct}
4682 or @code{union} type. Alternatively, you can leave out the alignment factor
4683 and just ask the compiler to align a type to the maximum
4684 useful alignment for the target machine you are compiling for. For
4685 example, you could write:
4688 struct S @{ short f[3]; @} __attribute__ ((aligned));
4691 Whenever you leave out the alignment factor in an @code{aligned}
4692 attribute specification, the compiler automatically sets the alignment
4693 for the type to the largest alignment which is ever used for any data
4694 type on the target machine you are compiling for. Doing this can often
4695 make copy operations more efficient, because the compiler can use
4696 whatever instructions copy the biggest chunks of memory when performing
4697 copies to or from the variables which have types that you have aligned
4700 In the example above, if the size of each @code{short} is 2 bytes, then
4701 the size of the entire @code{struct S} type is 6 bytes. The smallest
4702 power of two which is greater than or equal to that is 8, so the
4703 compiler sets the alignment for the entire @code{struct S} type to 8
4706 Note that although you can ask the compiler to select a time-efficient
4707 alignment for a given type and then declare only individual stand-alone
4708 objects of that type, the compiler's ability to select a time-efficient
4709 alignment is primarily useful only when you plan to create arrays of
4710 variables having the relevant (efficiently aligned) type. If you
4711 declare or use arrays of variables of an efficiently-aligned type, then
4712 it is likely that your program will also be doing pointer arithmetic (or
4713 subscripting, which amounts to the same thing) on pointers to the
4714 relevant type, and the code that the compiler generates for these
4715 pointer arithmetic operations will often be more efficient for
4716 efficiently-aligned types than for other types.
4718 The @code{aligned} attribute can only increase the alignment; but you
4719 can decrease it by specifying @code{packed} as well. See below.
4721 Note that the effectiveness of @code{aligned} attributes may be limited
4722 by inherent limitations in your linker. On many systems, the linker is
4723 only able to arrange for variables to be aligned up to a certain maximum
4724 alignment. (For some linkers, the maximum supported alignment may
4725 be very very small.) If your linker is only able to align variables
4726 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4727 in an @code{__attribute__} will still only provide you with 8 byte
4728 alignment. See your linker documentation for further information.
4731 This attribute, attached to @code{struct} or @code{union} type
4732 definition, specifies that each member (other than zero-width bitfields)
4733 of the structure or union is placed to minimize the memory required. When
4734 attached to an @code{enum} definition, it indicates that the smallest
4735 integral type should be used.
4737 @opindex fshort-enums
4738 Specifying this attribute for @code{struct} and @code{union} types is
4739 equivalent to specifying the @code{packed} attribute on each of the
4740 structure or union members. Specifying the @option{-fshort-enums}
4741 flag on the line is equivalent to specifying the @code{packed}
4742 attribute on all @code{enum} definitions.
4744 In the following example @code{struct my_packed_struct}'s members are
4745 packed closely together, but the internal layout of its @code{s} member
4746 is not packed---to do that, @code{struct my_unpacked_struct} would need to
4750 struct my_unpacked_struct
4756 struct __attribute__ ((__packed__)) my_packed_struct
4760 struct my_unpacked_struct s;
4764 You may only specify this attribute on the definition of an @code{enum},
4765 @code{struct} or @code{union}, not on a @code{typedef} which does not
4766 also define the enumerated type, structure or union.
4768 @item transparent_union
4769 This attribute, attached to a @code{union} type definition, indicates
4770 that any function parameter having that union type causes calls to that
4771 function to be treated in a special way.
4773 First, the argument corresponding to a transparent union type can be of
4774 any type in the union; no cast is required. Also, if the union contains
4775 a pointer type, the corresponding argument can be a null pointer
4776 constant or a void pointer expression; and if the union contains a void
4777 pointer type, the corresponding argument can be any pointer expression.
4778 If the union member type is a pointer, qualifiers like @code{const} on
4779 the referenced type must be respected, just as with normal pointer
4782 Second, the argument is passed to the function using the calling
4783 conventions of the first member of the transparent union, not the calling
4784 conventions of the union itself. All members of the union must have the
4785 same machine representation; this is necessary for this argument passing
4788 Transparent unions are designed for library functions that have multiple
4789 interfaces for compatibility reasons. For example, suppose the
4790 @code{wait} function must accept either a value of type @code{int *} to
4791 comply with Posix, or a value of type @code{union wait *} to comply with
4792 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4793 @code{wait} would accept both kinds of arguments, but it would also
4794 accept any other pointer type and this would make argument type checking
4795 less useful. Instead, @code{<sys/wait.h>} might define the interface
4799 typedef union __attribute__ ((__transparent_union__))
4803 @} wait_status_ptr_t;
4805 pid_t wait (wait_status_ptr_t);
4808 This interface allows either @code{int *} or @code{union wait *}
4809 arguments to be passed, using the @code{int *} calling convention.
4810 The program can call @code{wait} with arguments of either type:
4813 int w1 () @{ int w; return wait (&w); @}
4814 int w2 () @{ union wait w; return wait (&w); @}
4817 With this interface, @code{wait}'s implementation might look like this:
4820 pid_t wait (wait_status_ptr_t p)
4822 return waitpid (-1, p.__ip, 0);
4827 When attached to a type (including a @code{union} or a @code{struct}),
4828 this attribute means that variables of that type are meant to appear
4829 possibly unused. GCC will not produce a warning for any variables of
4830 that type, even if the variable appears to do nothing. This is often
4831 the case with lock or thread classes, which are usually defined and then
4832 not referenced, but contain constructors and destructors that have
4833 nontrivial bookkeeping functions.
4836 @itemx deprecated (@var{msg})
4837 The @code{deprecated} attribute results in a warning if the type
4838 is used anywhere in the source file. This is useful when identifying
4839 types that are expected to be removed in a future version of a program.
4840 If possible, the warning also includes the location of the declaration
4841 of the deprecated type, to enable users to easily find further
4842 information about why the type is deprecated, or what they should do
4843 instead. Note that the warnings only occur for uses and then only
4844 if the type is being applied to an identifier that itself is not being
4845 declared as deprecated.
4848 typedef int T1 __attribute__ ((deprecated));
4852 typedef T1 T3 __attribute__ ((deprecated));
4853 T3 z __attribute__ ((deprecated));
4856 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
4857 warning is issued for line 4 because T2 is not explicitly
4858 deprecated. Line 5 has no warning because T3 is explicitly
4859 deprecated. Similarly for line 6. The optional msg
4860 argument, which must be a string, will be printed in the warning if
4863 The @code{deprecated} attribute can also be used for functions and
4864 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
4867 Accesses through pointers to types with this attribute are not subject
4868 to type-based alias analysis, but are instead assumed to be able to alias
4869 any other type of objects. In the context of 6.5/7 an lvalue expression
4870 dereferencing such a pointer is treated like having a character type.
4871 See @option{-fstrict-aliasing} for more information on aliasing issues.
4872 This extension exists to support some vector APIs, in which pointers to
4873 one vector type are permitted to alias pointers to a different vector type.
4875 Note that an object of a type with this attribute does not have any
4881 typedef short __attribute__((__may_alias__)) short_a;
4887 short_a *b = (short_a *) &a;
4891 if (a == 0x12345678)
4898 If you replaced @code{short_a} with @code{short} in the variable
4899 declaration, the above program would abort when compiled with
4900 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
4901 above in recent GCC versions.
4904 In C++, attribute visibility (@pxref{Function Attributes}) can also be
4905 applied to class, struct, union and enum types. Unlike other type
4906 attributes, the attribute must appear between the initial keyword and
4907 the name of the type; it cannot appear after the body of the type.
4909 Note that the type visibility is applied to vague linkage entities
4910 associated with the class (vtable, typeinfo node, etc.). In
4911 particular, if a class is thrown as an exception in one shared object
4912 and caught in another, the class must have default visibility.
4913 Otherwise the two shared objects will be unable to use the same
4914 typeinfo node and exception handling will break.
4918 @subsection ARM Type Attributes
4920 On those ARM targets that support @code{dllimport} (such as Symbian
4921 OS), you can use the @code{notshared} attribute to indicate that the
4922 virtual table and other similar data for a class should not be
4923 exported from a DLL@. For example:
4926 class __declspec(notshared) C @{
4928 __declspec(dllimport) C();
4932 __declspec(dllexport)
4936 In this code, @code{C::C} is exported from the current DLL, but the
4937 virtual table for @code{C} is not exported. (You can use
4938 @code{__attribute__} instead of @code{__declspec} if you prefer, but
4939 most Symbian OS code uses @code{__declspec}.)
4941 @anchor{MeP Type Attributes}
4942 @subsection MeP Type Attributes
4944 Many of the MeP variable attributes may be applied to types as well.
4945 Specifically, the @code{based}, @code{tiny}, @code{near}, and
4946 @code{far} attributes may be applied to either. The @code{io} and
4947 @code{cb} attributes may not be applied to types.
4949 @anchor{i386 Type Attributes}
4950 @subsection i386 Type Attributes
4952 Two attributes are currently defined for i386 configurations:
4953 @code{ms_struct} and @code{gcc_struct}.
4959 @cindex @code{ms_struct}
4960 @cindex @code{gcc_struct}
4962 If @code{packed} is used on a structure, or if bit-fields are used
4963 it may be that the Microsoft ABI packs them differently
4964 than GCC would normally pack them. Particularly when moving packed
4965 data between functions compiled with GCC and the native Microsoft compiler
4966 (either via function call or as data in a file), it may be necessary to access
4969 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4970 compilers to match the native Microsoft compiler.
4973 To specify multiple attributes, separate them by commas within the
4974 double parentheses: for example, @samp{__attribute__ ((aligned (16),
4977 @anchor{PowerPC Type Attributes}
4978 @subsection PowerPC Type Attributes
4980 Three attributes currently are defined for PowerPC configurations:
4981 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4983 For full documentation of the @code{ms_struct} and @code{gcc_struct}
4984 attributes please see the documentation in @ref{i386 Type Attributes}.
4986 The @code{altivec} attribute allows one to declare AltiVec vector data
4987 types supported by the AltiVec Programming Interface Manual. The
4988 attribute requires an argument to specify one of three vector types:
4989 @code{vector__}, @code{pixel__} (always followed by unsigned short),
4990 and @code{bool__} (always followed by unsigned).
4993 __attribute__((altivec(vector__)))
4994 __attribute__((altivec(pixel__))) unsigned short
4995 __attribute__((altivec(bool__))) unsigned
4998 These attributes mainly are intended to support the @code{__vector},
4999 @code{__pixel}, and @code{__bool} AltiVec keywords.
5001 @anchor{SPU Type Attributes}
5002 @subsection SPU Type Attributes
5004 The SPU supports the @code{spu_vector} attribute for types. This attribute
5005 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
5006 Language Extensions Specification. It is intended to support the
5007 @code{__vector} keyword.
5011 @section An Inline Function is As Fast As a Macro
5012 @cindex inline functions
5013 @cindex integrating function code
5015 @cindex macros, inline alternative
5017 By declaring a function inline, you can direct GCC to make
5018 calls to that function faster. One way GCC can achieve this is to
5019 integrate that function's code into the code for its callers. This
5020 makes execution faster by eliminating the function-call overhead; in
5021 addition, if any of the actual argument values are constant, their
5022 known values may permit simplifications at compile time so that not
5023 all of the inline function's code needs to be included. The effect on
5024 code size is less predictable; object code may be larger or smaller
5025 with function inlining, depending on the particular case. You can
5026 also direct GCC to try to integrate all ``simple enough'' functions
5027 into their callers with the option @option{-finline-functions}.
5029 GCC implements three different semantics of declaring a function
5030 inline. One is available with @option{-std=gnu89} or
5031 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
5032 on all inline declarations, another when
5033 @option{-std=c99}, @option{-std=c1x},
5034 @option{-std=gnu99} or @option{-std=gnu1x}
5035 (without @option{-fgnu89-inline}), and the third
5036 is used when compiling C++.
5038 To declare a function inline, use the @code{inline} keyword in its
5039 declaration, like this:
5049 If you are writing a header file to be included in ISO C90 programs, write
5050 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
5052 The three types of inlining behave similarly in two important cases:
5053 when the @code{inline} keyword is used on a @code{static} function,
5054 like the example above, and when a function is first declared without
5055 using the @code{inline} keyword and then is defined with
5056 @code{inline}, like this:
5059 extern int inc (int *a);
5067 In both of these common cases, the program behaves the same as if you
5068 had not used the @code{inline} keyword, except for its speed.
5070 @cindex inline functions, omission of
5071 @opindex fkeep-inline-functions
5072 When a function is both inline and @code{static}, if all calls to the
5073 function are integrated into the caller, and the function's address is
5074 never used, then the function's own assembler code is never referenced.
5075 In this case, GCC does not actually output assembler code for the
5076 function, unless you specify the option @option{-fkeep-inline-functions}.
5077 Some calls cannot be integrated for various reasons (in particular,
5078 calls that precede the function's definition cannot be integrated, and
5079 neither can recursive calls within the definition). If there is a
5080 nonintegrated call, then the function is compiled to assembler code as
5081 usual. The function must also be compiled as usual if the program
5082 refers to its address, because that can't be inlined.
5085 Note that certain usages in a function definition can make it unsuitable
5086 for inline substitution. Among these usages are: use of varargs, use of
5087 alloca, use of variable sized data types (@pxref{Variable Length}),
5088 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
5089 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
5090 will warn when a function marked @code{inline} could not be substituted,
5091 and will give the reason for the failure.
5093 @cindex automatic @code{inline} for C++ member fns
5094 @cindex @code{inline} automatic for C++ member fns
5095 @cindex member fns, automatically @code{inline}
5096 @cindex C++ member fns, automatically @code{inline}
5097 @opindex fno-default-inline
5098 As required by ISO C++, GCC considers member functions defined within
5099 the body of a class to be marked inline even if they are
5100 not explicitly declared with the @code{inline} keyword. You can
5101 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
5102 Options,,Options Controlling C++ Dialect}.
5104 GCC does not inline any functions when not optimizing unless you specify
5105 the @samp{always_inline} attribute for the function, like this:
5108 /* @r{Prototype.} */
5109 inline void foo (const char) __attribute__((always_inline));
5112 The remainder of this section is specific to GNU C90 inlining.
5114 @cindex non-static inline function
5115 When an inline function is not @code{static}, then the compiler must assume
5116 that there may be calls from other source files; since a global symbol can
5117 be defined only once in any program, the function must not be defined in
5118 the other source files, so the calls therein cannot be integrated.
5119 Therefore, a non-@code{static} inline function is always compiled on its
5120 own in the usual fashion.
5122 If you specify both @code{inline} and @code{extern} in the function
5123 definition, then the definition is used only for inlining. In no case
5124 is the function compiled on its own, not even if you refer to its
5125 address explicitly. Such an address becomes an external reference, as
5126 if you had only declared the function, and had not defined it.
5128 This combination of @code{inline} and @code{extern} has almost the
5129 effect of a macro. The way to use it is to put a function definition in
5130 a header file with these keywords, and put another copy of the
5131 definition (lacking @code{inline} and @code{extern}) in a library file.
5132 The definition in the header file will cause most calls to the function
5133 to be inlined. If any uses of the function remain, they will refer to
5134 the single copy in the library.
5137 @section When is a Volatile Object Accessed?
5138 @cindex accessing volatiles
5139 @cindex volatile read
5140 @cindex volatile write
5141 @cindex volatile access
5143 C has the concept of volatile objects. These are normally accessed by
5144 pointers and used for accessing hardware or inter-thread
5145 communication. The standard encourage compilers to refrain from
5146 optimizations concerning accesses to volatile objects, but leaves it
5147 implementation defined as to what constitutes a volatile access. The
5148 minimum requirement is that at a sequence point all previous accesses
5149 to volatile objects have stabilized and no subsequent accesses have
5150 occurred. Thus an implementation is free to reorder and combine
5151 volatile accesses which occur between sequence points, but cannot do
5152 so for accesses across a sequence point. The use of volatiles does
5153 not allow you to violate the restriction on updating objects multiple
5154 times between two sequence points.
5156 Accesses to non-volatile objects are not ordered with respect to
5157 volatile accesses. You cannot use a volatile object as a memory
5158 barrier to order a sequence of writes to non-volatile memory. For
5162 int *ptr = @var{something};
5164 *ptr = @var{something};
5168 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
5169 that the write to @var{*ptr} will have occurred by the time the update
5170 of @var{vobj} has happened. If you need this guarantee, you must use
5171 a stronger memory barrier such as:
5174 int *ptr = @var{something};
5176 *ptr = @var{something};
5177 asm volatile ("" : : : "memory");
5181 A scalar volatile object is read, when it is accessed in a void context:
5184 volatile int *src = @var{somevalue};
5188 Such expressions are rvalues, and GCC implements this as a
5189 read of the volatile object being pointed to.
5191 Assignments are also expressions and have an rvalue. However when
5192 assigning to a scalar volatile, the volatile object is not reread,
5193 regardless of whether the assignment expression's rvalue is used or
5194 not. If the assignment's rvalue is used, the value is that assigned
5195 to the volatile object. For instance, there is no read of @var{vobj}
5196 in all the following cases:
5201 vobj = @var{something};
5202 obj = vobj = @var{something};
5203 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
5204 obj = (@var{something}, vobj = @var{anotherthing});
5207 If you need to read the volatile object after an assignment has
5208 occurred, you must use a separate expression with an intervening
5211 As bitfields are not individually addressable, volatile bitfields may
5212 be implicitly read when written to, or when adjacent bitfields are
5213 accessed. Bitfield operations may be optimized such that adjacent
5214 bitfields are only partially accessed, if they straddle a storage unit
5215 boundary. For these reasons it is unwise to use volatile bitfields to
5219 @section Assembler Instructions with C Expression Operands
5220 @cindex extended @code{asm}
5221 @cindex @code{asm} expressions
5222 @cindex assembler instructions
5225 In an assembler instruction using @code{asm}, you can specify the
5226 operands of the instruction using C expressions. This means you need not
5227 guess which registers or memory locations will contain the data you want
5230 You must specify an assembler instruction template much like what
5231 appears in a machine description, plus an operand constraint string for
5234 For example, here is how to use the 68881's @code{fsinx} instruction:
5237 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
5241 Here @code{angle} is the C expression for the input operand while
5242 @code{result} is that of the output operand. Each has @samp{"f"} as its
5243 operand constraint, saying that a floating point register is required.
5244 The @samp{=} in @samp{=f} indicates that the operand is an output; all
5245 output operands' constraints must use @samp{=}. The constraints use the
5246 same language used in the machine description (@pxref{Constraints}).
5248 Each operand is described by an operand-constraint string followed by
5249 the C expression in parentheses. A colon separates the assembler
5250 template from the first output operand and another separates the last
5251 output operand from the first input, if any. Commas separate the
5252 operands within each group. The total number of operands is currently
5253 limited to 30; this limitation may be lifted in some future version of
5256 If there are no output operands but there are input operands, you must
5257 place two consecutive colons surrounding the place where the output
5260 As of GCC version 3.1, it is also possible to specify input and output
5261 operands using symbolic names which can be referenced within the
5262 assembler code. These names are specified inside square brackets
5263 preceding the constraint string, and can be referenced inside the
5264 assembler code using @code{%[@var{name}]} instead of a percentage sign
5265 followed by the operand number. Using named operands the above example
5269 asm ("fsinx %[angle],%[output]"
5270 : [output] "=f" (result)
5271 : [angle] "f" (angle));
5275 Note that the symbolic operand names have no relation whatsoever to
5276 other C identifiers. You may use any name you like, even those of
5277 existing C symbols, but you must ensure that no two operands within the same
5278 assembler construct use the same symbolic name.
5280 Output operand expressions must be lvalues; the compiler can check this.
5281 The input operands need not be lvalues. The compiler cannot check
5282 whether the operands have data types that are reasonable for the
5283 instruction being executed. It does not parse the assembler instruction
5284 template and does not know what it means or even whether it is valid
5285 assembler input. The extended @code{asm} feature is most often used for
5286 machine instructions the compiler itself does not know exist. If
5287 the output expression cannot be directly addressed (for example, it is a
5288 bit-field), your constraint must allow a register. In that case, GCC
5289 will use the register as the output of the @code{asm}, and then store
5290 that register into the output.
5292 The ordinary output operands must be write-only; GCC will assume that
5293 the values in these operands before the instruction are dead and need
5294 not be generated. Extended asm supports input-output or read-write
5295 operands. Use the constraint character @samp{+} to indicate such an
5296 operand and list it with the output operands. You should only use
5297 read-write operands when the constraints for the operand (or the
5298 operand in which only some of the bits are to be changed) allow a
5301 You may, as an alternative, logically split its function into two
5302 separate operands, one input operand and one write-only output
5303 operand. The connection between them is expressed by constraints
5304 which say they need to be in the same location when the instruction
5305 executes. You can use the same C expression for both operands, or
5306 different expressions. For example, here we write the (fictitious)
5307 @samp{combine} instruction with @code{bar} as its read-only source
5308 operand and @code{foo} as its read-write destination:
5311 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
5315 The constraint @samp{"0"} for operand 1 says that it must occupy the
5316 same location as operand 0. A number in constraint is allowed only in
5317 an input operand and it must refer to an output operand.
5319 Only a number in the constraint can guarantee that one operand will be in
5320 the same place as another. The mere fact that @code{foo} is the value
5321 of both operands is not enough to guarantee that they will be in the
5322 same place in the generated assembler code. The following would not
5326 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
5329 Various optimizations or reloading could cause operands 0 and 1 to be in
5330 different registers; GCC knows no reason not to do so. For example, the
5331 compiler might find a copy of the value of @code{foo} in one register and
5332 use it for operand 1, but generate the output operand 0 in a different
5333 register (copying it afterward to @code{foo}'s own address). Of course,
5334 since the register for operand 1 is not even mentioned in the assembler
5335 code, the result will not work, but GCC can't tell that.
5337 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
5338 the operand number for a matching constraint. For example:
5341 asm ("cmoveq %1,%2,%[result]"
5342 : [result] "=r"(result)
5343 : "r" (test), "r"(new), "[result]"(old));
5346 Sometimes you need to make an @code{asm} operand be a specific register,
5347 but there's no matching constraint letter for that register @emph{by
5348 itself}. To force the operand into that register, use a local variable
5349 for the operand and specify the register in the variable declaration.
5350 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
5351 register constraint letter that matches the register:
5354 register int *p1 asm ("r0") = @dots{};
5355 register int *p2 asm ("r1") = @dots{};
5356 register int *result asm ("r0");
5357 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5360 @anchor{Example of asm with clobbered asm reg}
5361 In the above example, beware that a register that is call-clobbered by
5362 the target ABI will be overwritten by any function call in the
5363 assignment, including library calls for arithmetic operators.
5364 Also a register may be clobbered when generating some operations,
5365 like variable shift, memory copy or memory move on x86.
5366 Assuming it is a call-clobbered register, this may happen to @code{r0}
5367 above by the assignment to @code{p2}. If you have to use such a
5368 register, use temporary variables for expressions between the register
5373 register int *p1 asm ("r0") = @dots{};
5374 register int *p2 asm ("r1") = t1;
5375 register int *result asm ("r0");
5376 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5379 Some instructions clobber specific hard registers. To describe this,
5380 write a third colon after the input operands, followed by the names of
5381 the clobbered hard registers (given as strings). Here is a realistic
5382 example for the VAX:
5385 asm volatile ("movc3 %0,%1,%2"
5386 : /* @r{no outputs} */
5387 : "g" (from), "g" (to), "g" (count)
5388 : "r0", "r1", "r2", "r3", "r4", "r5");
5391 You may not write a clobber description in a way that overlaps with an
5392 input or output operand. For example, you may not have an operand
5393 describing a register class with one member if you mention that register
5394 in the clobber list. Variables declared to live in specific registers
5395 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
5396 have no part mentioned in the clobber description.
5397 There is no way for you to specify that an input
5398 operand is modified without also specifying it as an output
5399 operand. Note that if all the output operands you specify are for this
5400 purpose (and hence unused), you will then also need to specify
5401 @code{volatile} for the @code{asm} construct, as described below, to
5402 prevent GCC from deleting the @code{asm} statement as unused.
5404 If you refer to a particular hardware register from the assembler code,
5405 you will probably have to list the register after the third colon to
5406 tell the compiler the register's value is modified. In some assemblers,
5407 the register names begin with @samp{%}; to produce one @samp{%} in the
5408 assembler code, you must write @samp{%%} in the input.
5410 If your assembler instruction can alter the condition code register, add
5411 @samp{cc} to the list of clobbered registers. GCC on some machines
5412 represents the condition codes as a specific hardware register;
5413 @samp{cc} serves to name this register. On other machines, the
5414 condition code is handled differently, and specifying @samp{cc} has no
5415 effect. But it is valid no matter what the machine.
5417 If your assembler instructions access memory in an unpredictable
5418 fashion, add @samp{memory} to the list of clobbered registers. This
5419 will cause GCC to not keep memory values cached in registers across the
5420 assembler instruction and not optimize stores or loads to that memory.
5421 You will also want to add the @code{volatile} keyword if the memory
5422 affected is not listed in the inputs or outputs of the @code{asm}, as
5423 the @samp{memory} clobber does not count as a side-effect of the
5424 @code{asm}. If you know how large the accessed memory is, you can add
5425 it as input or output but if this is not known, you should add
5426 @samp{memory}. As an example, if you access ten bytes of a string, you
5427 can use a memory input like:
5430 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
5433 Note that in the following example the memory input is necessary,
5434 otherwise GCC might optimize the store to @code{x} away:
5441 asm ("magic stuff accessing an 'int' pointed to by '%1'"
5442 "=&d" (r) : "a" (y), "m" (*y));
5447 You can put multiple assembler instructions together in a single
5448 @code{asm} template, separated by the characters normally used in assembly
5449 code for the system. A combination that works in most places is a newline
5450 to break the line, plus a tab character to move to the instruction field
5451 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
5452 assembler allows semicolons as a line-breaking character. Note that some
5453 assembler dialects use semicolons to start a comment.
5454 The input operands are guaranteed not to use any of the clobbered
5455 registers, and neither will the output operands' addresses, so you can
5456 read and write the clobbered registers as many times as you like. Here
5457 is an example of multiple instructions in a template; it assumes the
5458 subroutine @code{_foo} accepts arguments in registers 9 and 10:
5461 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
5463 : "g" (from), "g" (to)
5467 Unless an output operand has the @samp{&} constraint modifier, GCC
5468 may allocate it in the same register as an unrelated input operand, on
5469 the assumption the inputs are consumed before the outputs are produced.
5470 This assumption may be false if the assembler code actually consists of
5471 more than one instruction. In such a case, use @samp{&} for each output
5472 operand that may not overlap an input. @xref{Modifiers}.
5474 If you want to test the condition code produced by an assembler
5475 instruction, you must include a branch and a label in the @code{asm}
5476 construct, as follows:
5479 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
5485 This assumes your assembler supports local labels, as the GNU assembler
5486 and most Unix assemblers do.
5488 Speaking of labels, jumps from one @code{asm} to another are not
5489 supported. The compiler's optimizers do not know about these jumps, and
5490 therefore they cannot take account of them when deciding how to
5491 optimize. @xref{Extended asm with goto}.
5493 @cindex macros containing @code{asm}
5494 Usually the most convenient way to use these @code{asm} instructions is to
5495 encapsulate them in macros that look like functions. For example,
5499 (@{ double __value, __arg = (x); \
5500 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
5505 Here the variable @code{__arg} is used to make sure that the instruction
5506 operates on a proper @code{double} value, and to accept only those
5507 arguments @code{x} which can convert automatically to a @code{double}.
5509 Another way to make sure the instruction operates on the correct data
5510 type is to use a cast in the @code{asm}. This is different from using a
5511 variable @code{__arg} in that it converts more different types. For
5512 example, if the desired type were @code{int}, casting the argument to
5513 @code{int} would accept a pointer with no complaint, while assigning the
5514 argument to an @code{int} variable named @code{__arg} would warn about
5515 using a pointer unless the caller explicitly casts it.
5517 If an @code{asm} has output operands, GCC assumes for optimization
5518 purposes the instruction has no side effects except to change the output
5519 operands. This does not mean instructions with a side effect cannot be
5520 used, but you must be careful, because the compiler may eliminate them
5521 if the output operands aren't used, or move them out of loops, or
5522 replace two with one if they constitute a common subexpression. Also,
5523 if your instruction does have a side effect on a variable that otherwise
5524 appears not to change, the old value of the variable may be reused later
5525 if it happens to be found in a register.
5527 You can prevent an @code{asm} instruction from being deleted
5528 by writing the keyword @code{volatile} after
5529 the @code{asm}. For example:
5532 #define get_and_set_priority(new) \
5534 asm volatile ("get_and_set_priority %0, %1" \
5535 : "=g" (__old) : "g" (new)); \
5540 The @code{volatile} keyword indicates that the instruction has
5541 important side-effects. GCC will not delete a volatile @code{asm} if
5542 it is reachable. (The instruction can still be deleted if GCC can
5543 prove that control-flow will never reach the location of the
5544 instruction.) Note that even a volatile @code{asm} instruction
5545 can be moved relative to other code, including across jump
5546 instructions. For example, on many targets there is a system
5547 register which can be set to control the rounding mode of
5548 floating point operations. You might try
5549 setting it with a volatile @code{asm}, like this PowerPC example:
5552 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
5557 This will not work reliably, as the compiler may move the addition back
5558 before the volatile @code{asm}. To make it work you need to add an
5559 artificial dependency to the @code{asm} referencing a variable in the code
5560 you don't want moved, for example:
5563 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
5567 Similarly, you can't expect a
5568 sequence of volatile @code{asm} instructions to remain perfectly
5569 consecutive. If you want consecutive output, use a single @code{asm}.
5570 Also, GCC will perform some optimizations across a volatile @code{asm}
5571 instruction; GCC does not ``forget everything'' when it encounters
5572 a volatile @code{asm} instruction the way some other compilers do.
5574 An @code{asm} instruction without any output operands will be treated
5575 identically to a volatile @code{asm} instruction.
5577 It is a natural idea to look for a way to give access to the condition
5578 code left by the assembler instruction. However, when we attempted to
5579 implement this, we found no way to make it work reliably. The problem
5580 is that output operands might need reloading, which would result in
5581 additional following ``store'' instructions. On most machines, these
5582 instructions would alter the condition code before there was time to
5583 test it. This problem doesn't arise for ordinary ``test'' and
5584 ``compare'' instructions because they don't have any output operands.
5586 For reasons similar to those described above, it is not possible to give
5587 an assembler instruction access to the condition code left by previous
5590 @anchor{Extended asm with goto}
5591 As of GCC version 4.5, @code{asm goto} may be used to have the assembly
5592 jump to one or more C labels. In this form, a fifth section after the
5593 clobber list contains a list of all C labels to which the assembly may jump.
5594 Each label operand is implicitly self-named. The @code{asm} is also assumed
5595 to fall through to the next statement.
5597 This form of @code{asm} is restricted to not have outputs. This is due
5598 to a internal restriction in the compiler that control transfer instructions
5599 cannot have outputs. This restriction on @code{asm goto} may be lifted
5600 in some future version of the compiler. In the mean time, @code{asm goto}
5601 may include a memory clobber, and so leave outputs in memory.
5607 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
5608 : : "r"(x), "r"(&y) : "r5", "memory" : error);
5615 In this (inefficient) example, the @code{frob} instruction sets the
5616 carry bit to indicate an error. The @code{jc} instruction detects
5617 this and branches to the @code{error} label. Finally, the output
5618 of the @code{frob} instruction (@code{%r5}) is stored into the memory
5619 for variable @code{y}, which is later read by the @code{return} statement.
5625 asm goto ("mfsr %%r1, 123; jmp %%r1;"
5626 ".pushsection doit_table;"
5627 ".long %l0, %l1, %l2, %l3;"
5629 : : : "r1" : label1, label2, label3, label4);
5630 __builtin_unreachable ();
5645 In this (also inefficient) example, the @code{mfsr} instruction reads
5646 an address from some out-of-band machine register, and the following
5647 @code{jmp} instruction branches to that address. The address read by
5648 the @code{mfsr} instruction is assumed to have been previously set via
5649 some application-specific mechanism to be one of the four values stored
5650 in the @code{doit_table} section. Finally, the @code{asm} is followed
5651 by a call to @code{__builtin_unreachable} to indicate that the @code{asm}
5652 does not in fact fall through.
5655 #define TRACE1(NUM) \
5657 asm goto ("0: nop;" \
5658 ".pushsection trace_table;" \
5661 : : : : trace#NUM); \
5662 if (0) @{ trace#NUM: trace(); @} \
5664 #define TRACE TRACE1(__COUNTER__)
5667 In this example (which in fact inspired the @code{asm goto} feature)
5668 we want on rare occasions to call the @code{trace} function; on other
5669 occasions we'd like to keep the overhead to the absolute minimum.
5670 The normal code path consists of a single @code{nop} instruction.
5671 However, we record the address of this @code{nop} together with the
5672 address of a label that calls the @code{trace} function. This allows
5673 the @code{nop} instruction to be patched at runtime to be an
5674 unconditional branch to the stored label. It is assumed that an
5675 optimizing compiler will move the labeled block out of line, to
5676 optimize the fall through path from the @code{asm}.
5678 If you are writing a header file that should be includable in ISO C
5679 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
5682 @subsection Size of an @code{asm}
5684 Some targets require that GCC track the size of each instruction used in
5685 order to generate correct code. Because the final length of an
5686 @code{asm} is only known by the assembler, GCC must make an estimate as
5687 to how big it will be. The estimate is formed by counting the number of
5688 statements in the pattern of the @code{asm} and multiplying that by the
5689 length of the longest instruction on that processor. Statements in the
5690 @code{asm} are identified by newline characters and whatever statement
5691 separator characters are supported by the assembler; on most processors
5692 this is the `@code{;}' character.
5694 Normally, GCC's estimate is perfectly adequate to ensure that correct
5695 code is generated, but it is possible to confuse the compiler if you use
5696 pseudo instructions or assembler macros that expand into multiple real
5697 instructions or if you use assembler directives that expand to more
5698 space in the object file than would be needed for a single instruction.
5699 If this happens then the assembler will produce a diagnostic saying that
5700 a label is unreachable.
5702 @subsection i386 floating point asm operands
5704 There are several rules on the usage of stack-like regs in
5705 asm_operands insns. These rules apply only to the operands that are
5710 Given a set of input regs that die in an asm_operands, it is
5711 necessary to know which are implicitly popped by the asm, and
5712 which must be explicitly popped by gcc.
5714 An input reg that is implicitly popped by the asm must be
5715 explicitly clobbered, unless it is constrained to match an
5719 For any input reg that is implicitly popped by an asm, it is
5720 necessary to know how to adjust the stack to compensate for the pop.
5721 If any non-popped input is closer to the top of the reg-stack than
5722 the implicitly popped reg, it would not be possible to know what the
5723 stack looked like---it's not clear how the rest of the stack ``slides
5726 All implicitly popped input regs must be closer to the top of
5727 the reg-stack than any input that is not implicitly popped.
5729 It is possible that if an input dies in an insn, reload might
5730 use the input reg for an output reload. Consider this example:
5733 asm ("foo" : "=t" (a) : "f" (b));
5736 This asm says that input B is not popped by the asm, and that
5737 the asm pushes a result onto the reg-stack, i.e., the stack is one
5738 deeper after the asm than it was before. But, it is possible that
5739 reload will think that it can use the same reg for both the input and
5740 the output, if input B dies in this insn.
5742 If any input operand uses the @code{f} constraint, all output reg
5743 constraints must use the @code{&} earlyclobber.
5745 The asm above would be written as
5748 asm ("foo" : "=&t" (a) : "f" (b));
5752 Some operands need to be in particular places on the stack. All
5753 output operands fall in this category---there is no other way to
5754 know which regs the outputs appear in unless the user indicates
5755 this in the constraints.
5757 Output operands must specifically indicate which reg an output
5758 appears in after an asm. @code{=f} is not allowed: the operand
5759 constraints must select a class with a single reg.
5762 Output operands may not be ``inserted'' between existing stack regs.
5763 Since no 387 opcode uses a read/write operand, all output operands
5764 are dead before the asm_operands, and are pushed by the asm_operands.
5765 It makes no sense to push anywhere but the top of the reg-stack.
5767 Output operands must start at the top of the reg-stack: output
5768 operands may not ``skip'' a reg.
5771 Some asm statements may need extra stack space for internal
5772 calculations. This can be guaranteed by clobbering stack registers
5773 unrelated to the inputs and outputs.
5777 Here are a couple of reasonable asms to want to write. This asm
5778 takes one input, which is internally popped, and produces two outputs.
5781 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
5784 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
5785 and replaces them with one output. The user must code the @code{st(1)}
5786 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
5789 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
5795 @section Controlling Names Used in Assembler Code
5796 @cindex assembler names for identifiers
5797 @cindex names used in assembler code
5798 @cindex identifiers, names in assembler code
5800 You can specify the name to be used in the assembler code for a C
5801 function or variable by writing the @code{asm} (or @code{__asm__})
5802 keyword after the declarator as follows:
5805 int foo asm ("myfoo") = 2;
5809 This specifies that the name to be used for the variable @code{foo} in
5810 the assembler code should be @samp{myfoo} rather than the usual
5813 On systems where an underscore is normally prepended to the name of a C
5814 function or variable, this feature allows you to define names for the
5815 linker that do not start with an underscore.
5817 It does not make sense to use this feature with a non-static local
5818 variable since such variables do not have assembler names. If you are
5819 trying to put the variable in a particular register, see @ref{Explicit
5820 Reg Vars}. GCC presently accepts such code with a warning, but will
5821 probably be changed to issue an error, rather than a warning, in the
5824 You cannot use @code{asm} in this way in a function @emph{definition}; but
5825 you can get the same effect by writing a declaration for the function
5826 before its definition and putting @code{asm} there, like this:
5829 extern func () asm ("FUNC");
5836 It is up to you to make sure that the assembler names you choose do not
5837 conflict with any other assembler symbols. Also, you must not use a
5838 register name; that would produce completely invalid assembler code. GCC
5839 does not as yet have the ability to store static variables in registers.
5840 Perhaps that will be added.
5842 @node Explicit Reg Vars
5843 @section Variables in Specified Registers
5844 @cindex explicit register variables
5845 @cindex variables in specified registers
5846 @cindex specified registers
5847 @cindex registers, global allocation
5849 GNU C allows you to put a few global variables into specified hardware
5850 registers. You can also specify the register in which an ordinary
5851 register variable should be allocated.
5855 Global register variables reserve registers throughout the program.
5856 This may be useful in programs such as programming language
5857 interpreters which have a couple of global variables that are accessed
5861 Local register variables in specific registers do not reserve the
5862 registers, except at the point where they are used as input or output
5863 operands in an @code{asm} statement and the @code{asm} statement itself is
5864 not deleted. The compiler's data flow analysis is capable of determining
5865 where the specified registers contain live values, and where they are
5866 available for other uses. Stores into local register variables may be deleted
5867 when they appear to be dead according to dataflow analysis. References
5868 to local register variables may be deleted or moved or simplified.
5870 These local variables are sometimes convenient for use with the extended
5871 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
5872 output of the assembler instruction directly into a particular register.
5873 (This will work provided the register you specify fits the constraints
5874 specified for that operand in the @code{asm}.)
5882 @node Global Reg Vars
5883 @subsection Defining Global Register Variables
5884 @cindex global register variables
5885 @cindex registers, global variables in
5887 You can define a global register variable in GNU C like this:
5890 register int *foo asm ("a5");
5894 Here @code{a5} is the name of the register which should be used. Choose a
5895 register which is normally saved and restored by function calls on your
5896 machine, so that library routines will not clobber it.
5898 Naturally the register name is cpu-dependent, so you would need to
5899 conditionalize your program according to cpu type. The register
5900 @code{a5} would be a good choice on a 68000 for a variable of pointer
5901 type. On machines with register windows, be sure to choose a ``global''
5902 register that is not affected magically by the function call mechanism.
5904 In addition, operating systems on one type of cpu may differ in how they
5905 name the registers; then you would need additional conditionals. For
5906 example, some 68000 operating systems call this register @code{%a5}.
5908 Eventually there may be a way of asking the compiler to choose a register
5909 automatically, but first we need to figure out how it should choose and
5910 how to enable you to guide the choice. No solution is evident.
5912 Defining a global register variable in a certain register reserves that
5913 register entirely for this use, at least within the current compilation.
5914 The register will not be allocated for any other purpose in the functions
5915 in the current compilation. The register will not be saved and restored by
5916 these functions. Stores into this register are never deleted even if they
5917 would appear to be dead, but references may be deleted or moved or
5920 It is not safe to access the global register variables from signal
5921 handlers, or from more than one thread of control, because the system
5922 library routines may temporarily use the register for other things (unless
5923 you recompile them specially for the task at hand).
5925 @cindex @code{qsort}, and global register variables
5926 It is not safe for one function that uses a global register variable to
5927 call another such function @code{foo} by way of a third function
5928 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
5929 different source file in which the variable wasn't declared). This is
5930 because @code{lose} might save the register and put some other value there.
5931 For example, you can't expect a global register variable to be available in
5932 the comparison-function that you pass to @code{qsort}, since @code{qsort}
5933 might have put something else in that register. (If you are prepared to
5934 recompile @code{qsort} with the same global register variable, you can
5935 solve this problem.)
5937 If you want to recompile @code{qsort} or other source files which do not
5938 actually use your global register variable, so that they will not use that
5939 register for any other purpose, then it suffices to specify the compiler
5940 option @option{-ffixed-@var{reg}}. You need not actually add a global
5941 register declaration to their source code.
5943 A function which can alter the value of a global register variable cannot
5944 safely be called from a function compiled without this variable, because it
5945 could clobber the value the caller expects to find there on return.
5946 Therefore, the function which is the entry point into the part of the
5947 program that uses the global register variable must explicitly save and
5948 restore the value which belongs to its caller.
5950 @cindex register variable after @code{longjmp}
5951 @cindex global register after @code{longjmp}
5952 @cindex value after @code{longjmp}
5955 On most machines, @code{longjmp} will restore to each global register
5956 variable the value it had at the time of the @code{setjmp}. On some
5957 machines, however, @code{longjmp} will not change the value of global
5958 register variables. To be portable, the function that called @code{setjmp}
5959 should make other arrangements to save the values of the global register
5960 variables, and to restore them in a @code{longjmp}. This way, the same
5961 thing will happen regardless of what @code{longjmp} does.
5963 All global register variable declarations must precede all function
5964 definitions. If such a declaration could appear after function
5965 definitions, the declaration would be too late to prevent the register from
5966 being used for other purposes in the preceding functions.
5968 Global register variables may not have initial values, because an
5969 executable file has no means to supply initial contents for a register.
5971 On the SPARC, there are reports that g3 @dots{} g7 are suitable
5972 registers, but certain library functions, such as @code{getwd}, as well
5973 as the subroutines for division and remainder, modify g3 and g4. g1 and
5974 g2 are local temporaries.
5976 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
5977 Of course, it will not do to use more than a few of those.
5979 @node Local Reg Vars
5980 @subsection Specifying Registers for Local Variables
5981 @cindex local variables, specifying registers
5982 @cindex specifying registers for local variables
5983 @cindex registers for local variables
5985 You can define a local register variable with a specified register
5989 register int *foo asm ("a5");
5993 Here @code{a5} is the name of the register which should be used. Note
5994 that this is the same syntax used for defining global register
5995 variables, but for a local variable it would appear within a function.
5997 Naturally the register name is cpu-dependent, but this is not a
5998 problem, since specific registers are most often useful with explicit
5999 assembler instructions (@pxref{Extended Asm}). Both of these things
6000 generally require that you conditionalize your program according to
6003 In addition, operating systems on one type of cpu may differ in how they
6004 name the registers; then you would need additional conditionals. For
6005 example, some 68000 operating systems call this register @code{%a5}.
6007 Defining such a register variable does not reserve the register; it
6008 remains available for other uses in places where flow control determines
6009 the variable's value is not live.
6011 This option does not guarantee that GCC will generate code that has
6012 this variable in the register you specify at all times. You may not
6013 code an explicit reference to this register in the @emph{assembler
6014 instruction template} part of an @code{asm} statement and assume it will
6015 always refer to this variable. However, using the variable as an
6016 @code{asm} @emph{operand} guarantees that the specified register is used
6019 Stores into local register variables may be deleted when they appear to be dead
6020 according to dataflow analysis. References to local register variables may
6021 be deleted or moved or simplified.
6023 As for global register variables, it's recommended that you choose a
6024 register which is normally saved and restored by function calls on
6025 your machine, so that library routines will not clobber it. A common
6026 pitfall is to initialize multiple call-clobbered registers with
6027 arbitrary expressions, where a function call or library call for an
6028 arithmetic operator will overwrite a register value from a previous
6029 assignment, for example @code{r0} below:
6031 register int *p1 asm ("r0") = @dots{};
6032 register int *p2 asm ("r1") = @dots{};
6034 In those cases, a solution is to use a temporary variable for
6035 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
6037 @node Alternate Keywords
6038 @section Alternate Keywords
6039 @cindex alternate keywords
6040 @cindex keywords, alternate
6042 @option{-ansi} and the various @option{-std} options disable certain
6043 keywords. This causes trouble when you want to use GNU C extensions, or
6044 a general-purpose header file that should be usable by all programs,
6045 including ISO C programs. The keywords @code{asm}, @code{typeof} and
6046 @code{inline} are not available in programs compiled with
6047 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
6048 program compiled with @option{-std=c99} or @option{-std=c1x}). The
6050 @code{restrict} is only available when @option{-std=gnu99} (which will
6051 eventually be the default) or @option{-std=c99} (or the equivalent
6052 @option{-std=iso9899:1999}), or an option for a later standard
6055 The way to solve these problems is to put @samp{__} at the beginning and
6056 end of each problematical keyword. For example, use @code{__asm__}
6057 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
6059 Other C compilers won't accept these alternative keywords; if you want to
6060 compile with another compiler, you can define the alternate keywords as
6061 macros to replace them with the customary keywords. It looks like this:
6069 @findex __extension__
6071 @option{-pedantic} and other options cause warnings for many GNU C extensions.
6073 prevent such warnings within one expression by writing
6074 @code{__extension__} before the expression. @code{__extension__} has no
6075 effect aside from this.
6077 @node Incomplete Enums
6078 @section Incomplete @code{enum} Types
6080 You can define an @code{enum} tag without specifying its possible values.
6081 This results in an incomplete type, much like what you get if you write
6082 @code{struct foo} without describing the elements. A later declaration
6083 which does specify the possible values completes the type.
6085 You can't allocate variables or storage using the type while it is
6086 incomplete. However, you can work with pointers to that type.
6088 This extension may not be very useful, but it makes the handling of
6089 @code{enum} more consistent with the way @code{struct} and @code{union}
6092 This extension is not supported by GNU C++.
6094 @node Function Names
6095 @section Function Names as Strings
6096 @cindex @code{__func__} identifier
6097 @cindex @code{__FUNCTION__} identifier
6098 @cindex @code{__PRETTY_FUNCTION__} identifier
6100 GCC provides three magic variables which hold the name of the current
6101 function, as a string. The first of these is @code{__func__}, which
6102 is part of the C99 standard:
6104 The identifier @code{__func__} is implicitly declared by the translator
6105 as if, immediately following the opening brace of each function
6106 definition, the declaration
6109 static const char __func__[] = "function-name";
6113 appeared, where function-name is the name of the lexically-enclosing
6114 function. This name is the unadorned name of the function.
6116 @code{__FUNCTION__} is another name for @code{__func__}. Older
6117 versions of GCC recognize only this name. However, it is not
6118 standardized. For maximum portability, we recommend you use
6119 @code{__func__}, but provide a fallback definition with the
6123 #if __STDC_VERSION__ < 199901L
6125 # define __func__ __FUNCTION__
6127 # define __func__ "<unknown>"
6132 In C, @code{__PRETTY_FUNCTION__} is yet another name for
6133 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
6134 the type signature of the function as well as its bare name. For
6135 example, this program:
6139 extern int printf (char *, ...);
6146 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
6147 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
6165 __PRETTY_FUNCTION__ = void a::sub(int)
6168 These identifiers are not preprocessor macros. In GCC 3.3 and
6169 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
6170 were treated as string literals; they could be used to initialize
6171 @code{char} arrays, and they could be concatenated with other string
6172 literals. GCC 3.4 and later treat them as variables, like
6173 @code{__func__}. In C++, @code{__FUNCTION__} and
6174 @code{__PRETTY_FUNCTION__} have always been variables.
6176 @node Return Address
6177 @section Getting the Return or Frame Address of a Function
6179 These functions may be used to get information about the callers of a
6182 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
6183 This function returns the return address of the current function, or of
6184 one of its callers. The @var{level} argument is number of frames to
6185 scan up the call stack. A value of @code{0} yields the return address
6186 of the current function, a value of @code{1} yields the return address
6187 of the caller of the current function, and so forth. When inlining
6188 the expected behavior is that the function will return the address of
6189 the function that will be returned to. To work around this behavior use
6190 the @code{noinline} function attribute.
6192 The @var{level} argument must be a constant integer.
6194 On some machines it may be impossible to determine the return address of
6195 any function other than the current one; in such cases, or when the top
6196 of the stack has been reached, this function will return @code{0} or a
6197 random value. In addition, @code{__builtin_frame_address} may be used
6198 to determine if the top of the stack has been reached.
6200 Additional post-processing of the returned value may be needed, see
6201 @code{__builtin_extract_return_address}.
6203 This function should only be used with a nonzero argument for debugging
6207 @deftypefn {Built-in Function} {void *} __builtin_extract_return_address (void *@var{addr})
6208 The address as returned by @code{__builtin_return_address} may have to be fed
6209 through this function to get the actual encoded address. For example, on the
6210 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
6211 platforms an offset has to be added for the true next instruction to be
6214 If no fixup is needed, this function simply passes through @var{addr}.
6217 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
6218 This function does the reverse of @code{__builtin_extract_return_address}.
6221 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
6222 This function is similar to @code{__builtin_return_address}, but it
6223 returns the address of the function frame rather than the return address
6224 of the function. Calling @code{__builtin_frame_address} with a value of
6225 @code{0} yields the frame address of the current function, a value of
6226 @code{1} yields the frame address of the caller of the current function,
6229 The frame is the area on the stack which holds local variables and saved
6230 registers. The frame address is normally the address of the first word
6231 pushed on to the stack by the function. However, the exact definition
6232 depends upon the processor and the calling convention. If the processor
6233 has a dedicated frame pointer register, and the function has a frame,
6234 then @code{__builtin_frame_address} will return the value of the frame
6237 On some machines it may be impossible to determine the frame address of
6238 any function other than the current one; in such cases, or when the top
6239 of the stack has been reached, this function will return @code{0} if
6240 the first frame pointer is properly initialized by the startup code.
6242 This function should only be used with a nonzero argument for debugging
6246 @node Vector Extensions
6247 @section Using vector instructions through built-in functions
6249 On some targets, the instruction set contains SIMD vector instructions that
6250 operate on multiple values contained in one large register at the same time.
6251 For example, on the i386 the MMX, 3DNow!@: and SSE extensions can be used
6254 The first step in using these extensions is to provide the necessary data
6255 types. This should be done using an appropriate @code{typedef}:
6258 typedef int v4si __attribute__ ((vector_size (16)));
6261 The @code{int} type specifies the base type, while the attribute specifies
6262 the vector size for the variable, measured in bytes. For example, the
6263 declaration above causes the compiler to set the mode for the @code{v4si}
6264 type to be 16 bytes wide and divided into @code{int} sized units. For
6265 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
6266 corresponding mode of @code{foo} will be @acronym{V4SI}.
6268 The @code{vector_size} attribute is only applicable to integral and
6269 float scalars, although arrays, pointers, and function return values
6270 are allowed in conjunction with this construct.
6272 All the basic integer types can be used as base types, both as signed
6273 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
6274 @code{long long}. In addition, @code{float} and @code{double} can be
6275 used to build floating-point vector types.
6277 Specifying a combination that is not valid for the current architecture
6278 will cause GCC to synthesize the instructions using a narrower mode.
6279 For example, if you specify a variable of type @code{V4SI} and your
6280 architecture does not allow for this specific SIMD type, GCC will
6281 produce code that uses 4 @code{SIs}.
6283 The types defined in this manner can be used with a subset of normal C
6284 operations. Currently, GCC will allow using the following operators
6285 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
6287 The operations behave like C++ @code{valarrays}. Addition is defined as
6288 the addition of the corresponding elements of the operands. For
6289 example, in the code below, each of the 4 elements in @var{a} will be
6290 added to the corresponding 4 elements in @var{b} and the resulting
6291 vector will be stored in @var{c}.
6294 typedef int v4si __attribute__ ((vector_size (16)));
6301 Subtraction, multiplication, division, and the logical operations
6302 operate in a similar manner. Likewise, the result of using the unary
6303 minus or complement operators on a vector type is a vector whose
6304 elements are the negative or complemented values of the corresponding
6305 elements in the operand.
6307 You can declare variables and use them in function calls and returns, as
6308 well as in assignments and some casts. You can specify a vector type as
6309 a return type for a function. Vector types can also be used as function
6310 arguments. It is possible to cast from one vector type to another,
6311 provided they are of the same size (in fact, you can also cast vectors
6312 to and from other datatypes of the same size).
6314 You cannot operate between vectors of different lengths or different
6315 signedness without a cast.
6317 A port that supports hardware vector operations, usually provides a set
6318 of built-in functions that can be used to operate on vectors. For
6319 example, a function to add two vectors and multiply the result by a
6320 third could look like this:
6323 v4si f (v4si a, v4si b, v4si c)
6325 v4si tmp = __builtin_addv4si (a, b);
6326 return __builtin_mulv4si (tmp, c);
6333 @findex __builtin_offsetof
6335 GCC implements for both C and C++ a syntactic extension to implement
6336 the @code{offsetof} macro.
6340 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
6342 offsetof_member_designator:
6344 | offsetof_member_designator "." @code{identifier}
6345 | offsetof_member_designator "[" @code{expr} "]"
6348 This extension is sufficient such that
6351 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
6354 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
6355 may be dependent. In either case, @var{member} may consist of a single
6356 identifier, or a sequence of member accesses and array references.
6358 @node Atomic Builtins
6359 @section Built-in functions for atomic memory access
6361 The following builtins are intended to be compatible with those described
6362 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
6363 section 7.4. As such, they depart from the normal GCC practice of using
6364 the ``__builtin_'' prefix, and further that they are overloaded such that
6365 they work on multiple types.
6367 The definition given in the Intel documentation allows only for the use of
6368 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
6369 counterparts. GCC will allow any integral scalar or pointer type that is
6370 1, 2, 4 or 8 bytes in length.
6372 Not all operations are supported by all target processors. If a particular
6373 operation cannot be implemented on the target processor, a warning will be
6374 generated and a call an external function will be generated. The external
6375 function will carry the same name as the builtin, with an additional suffix
6376 @samp{_@var{n}} where @var{n} is the size of the data type.
6378 @c ??? Should we have a mechanism to suppress this warning? This is almost
6379 @c useful for implementing the operation under the control of an external
6382 In most cases, these builtins are considered a @dfn{full barrier}. That is,
6383 no memory operand will be moved across the operation, either forward or
6384 backward. Further, instructions will be issued as necessary to prevent the
6385 processor from speculating loads across the operation and from queuing stores
6386 after the operation.
6388 All of the routines are described in the Intel documentation to take
6389 ``an optional list of variables protected by the memory barrier''. It's
6390 not clear what is meant by that; it could mean that @emph{only} the
6391 following variables are protected, or it could mean that these variables
6392 should in addition be protected. At present GCC ignores this list and
6393 protects all variables which are globally accessible. If in the future
6394 we make some use of this list, an empty list will continue to mean all
6395 globally accessible variables.
6398 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
6399 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
6400 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
6401 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
6402 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
6403 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
6404 @findex __sync_fetch_and_add
6405 @findex __sync_fetch_and_sub
6406 @findex __sync_fetch_and_or
6407 @findex __sync_fetch_and_and
6408 @findex __sync_fetch_and_xor
6409 @findex __sync_fetch_and_nand
6410 These builtins perform the operation suggested by the name, and
6411 returns the value that had previously been in memory. That is,
6414 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
6415 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
6418 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
6419 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
6421 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
6422 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
6423 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
6424 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
6425 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
6426 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
6427 @findex __sync_add_and_fetch
6428 @findex __sync_sub_and_fetch
6429 @findex __sync_or_and_fetch
6430 @findex __sync_and_and_fetch
6431 @findex __sync_xor_and_fetch
6432 @findex __sync_nand_and_fetch
6433 These builtins perform the operation suggested by the name, and
6434 return the new value. That is,
6437 @{ *ptr @var{op}= value; return *ptr; @}
6438 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
6441 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
6442 builtin as @code{*ptr = ~(*ptr & value)} instead of
6443 @code{*ptr = ~*ptr & value}.
6445 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
6446 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
6447 @findex __sync_bool_compare_and_swap
6448 @findex __sync_val_compare_and_swap
6449 These builtins perform an atomic compare and swap. That is, if the current
6450 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
6453 The ``bool'' version returns true if the comparison is successful and
6454 @var{newval} was written. The ``val'' version returns the contents
6455 of @code{*@var{ptr}} before the operation.
6457 @item __sync_synchronize (...)
6458 @findex __sync_synchronize
6459 This builtin issues a full memory barrier.
6461 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
6462 @findex __sync_lock_test_and_set
6463 This builtin, as described by Intel, is not a traditional test-and-set
6464 operation, but rather an atomic exchange operation. It writes @var{value}
6465 into @code{*@var{ptr}}, and returns the previous contents of
6468 Many targets have only minimal support for such locks, and do not support
6469 a full exchange operation. In this case, a target may support reduced
6470 functionality here by which the @emph{only} valid value to store is the
6471 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
6472 is implementation defined.
6474 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
6475 This means that references after the builtin cannot move to (or be
6476 speculated to) before the builtin, but previous memory stores may not
6477 be globally visible yet, and previous memory loads may not yet be
6480 @item void __sync_lock_release (@var{type} *ptr, ...)
6481 @findex __sync_lock_release
6482 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
6483 Normally this means writing the constant 0 to @code{*@var{ptr}}.
6485 This builtin is not a full barrier, but rather a @dfn{release barrier}.
6486 This means that all previous memory stores are globally visible, and all
6487 previous memory loads have been satisfied, but following memory reads
6488 are not prevented from being speculated to before the barrier.
6491 @node Object Size Checking
6492 @section Object Size Checking Builtins
6493 @findex __builtin_object_size
6494 @findex __builtin___memcpy_chk
6495 @findex __builtin___mempcpy_chk
6496 @findex __builtin___memmove_chk
6497 @findex __builtin___memset_chk
6498 @findex __builtin___strcpy_chk
6499 @findex __builtin___stpcpy_chk
6500 @findex __builtin___strncpy_chk
6501 @findex __builtin___strcat_chk
6502 @findex __builtin___strncat_chk
6503 @findex __builtin___sprintf_chk
6504 @findex __builtin___snprintf_chk
6505 @findex __builtin___vsprintf_chk
6506 @findex __builtin___vsnprintf_chk
6507 @findex __builtin___printf_chk
6508 @findex __builtin___vprintf_chk
6509 @findex __builtin___fprintf_chk
6510 @findex __builtin___vfprintf_chk
6512 GCC implements a limited buffer overflow protection mechanism
6513 that can prevent some buffer overflow attacks.
6515 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
6516 is a built-in construct that returns a constant number of bytes from
6517 @var{ptr} to the end of the object @var{ptr} pointer points to
6518 (if known at compile time). @code{__builtin_object_size} never evaluates
6519 its arguments for side-effects. If there are any side-effects in them, it
6520 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
6521 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
6522 point to and all of them are known at compile time, the returned number
6523 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
6524 0 and minimum if nonzero. If it is not possible to determine which objects
6525 @var{ptr} points to at compile time, @code{__builtin_object_size} should
6526 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
6527 for @var{type} 2 or 3.
6529 @var{type} is an integer constant from 0 to 3. If the least significant
6530 bit is clear, objects are whole variables, if it is set, a closest
6531 surrounding subobject is considered the object a pointer points to.
6532 The second bit determines if maximum or minimum of remaining bytes
6536 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
6537 char *p = &var.buf1[1], *q = &var.b;
6539 /* Here the object p points to is var. */
6540 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
6541 /* The subobject p points to is var.buf1. */
6542 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
6543 /* The object q points to is var. */
6544 assert (__builtin_object_size (q, 0)
6545 == (char *) (&var + 1) - (char *) &var.b);
6546 /* The subobject q points to is var.b. */
6547 assert (__builtin_object_size (q, 1) == sizeof (var.b));
6551 There are built-in functions added for many common string operation
6552 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
6553 built-in is provided. This built-in has an additional last argument,
6554 which is the number of bytes remaining in object the @var{dest}
6555 argument points to or @code{(size_t) -1} if the size is not known.
6557 The built-in functions are optimized into the normal string functions
6558 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
6559 it is known at compile time that the destination object will not
6560 be overflown. If the compiler can determine at compile time the
6561 object will be always overflown, it issues a warning.
6563 The intended use can be e.g.
6567 #define bos0(dest) __builtin_object_size (dest, 0)
6568 #define memcpy(dest, src, n) \
6569 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
6573 /* It is unknown what object p points to, so this is optimized
6574 into plain memcpy - no checking is possible. */
6575 memcpy (p, "abcde", n);
6576 /* Destination is known and length too. It is known at compile
6577 time there will be no overflow. */
6578 memcpy (&buf[5], "abcde", 5);
6579 /* Destination is known, but the length is not known at compile time.
6580 This will result in __memcpy_chk call that can check for overflow
6582 memcpy (&buf[5], "abcde", n);
6583 /* Destination is known and it is known at compile time there will
6584 be overflow. There will be a warning and __memcpy_chk call that
6585 will abort the program at runtime. */
6586 memcpy (&buf[6], "abcde", 5);
6589 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
6590 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
6591 @code{strcat} and @code{strncat}.
6593 There are also checking built-in functions for formatted output functions.
6595 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
6596 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6597 const char *fmt, ...);
6598 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
6600 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6601 const char *fmt, va_list ap);
6604 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
6605 etc.@: functions and can contain implementation specific flags on what
6606 additional security measures the checking function might take, such as
6607 handling @code{%n} differently.
6609 The @var{os} argument is the object size @var{s} points to, like in the
6610 other built-in functions. There is a small difference in the behavior
6611 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
6612 optimized into the non-checking functions only if @var{flag} is 0, otherwise
6613 the checking function is called with @var{os} argument set to
6616 In addition to this, there are checking built-in functions
6617 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
6618 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
6619 These have just one additional argument, @var{flag}, right before
6620 format string @var{fmt}. If the compiler is able to optimize them to
6621 @code{fputc} etc.@: functions, it will, otherwise the checking function
6622 should be called and the @var{flag} argument passed to it.
6624 @node Other Builtins
6625 @section Other built-in functions provided by GCC
6626 @cindex built-in functions
6627 @findex __builtin_fpclassify
6628 @findex __builtin_isfinite
6629 @findex __builtin_isnormal
6630 @findex __builtin_isgreater
6631 @findex __builtin_isgreaterequal
6632 @findex __builtin_isinf_sign
6633 @findex __builtin_isless
6634 @findex __builtin_islessequal
6635 @findex __builtin_islessgreater
6636 @findex __builtin_isunordered
6637 @findex __builtin_powi
6638 @findex __builtin_powif
6639 @findex __builtin_powil
6797 @findex fprintf_unlocked
6799 @findex fputs_unlocked
6916 @findex printf_unlocked
6948 @findex significandf
6949 @findex significandl
7020 GCC provides a large number of built-in functions other than the ones
7021 mentioned above. Some of these are for internal use in the processing
7022 of exceptions or variable-length argument lists and will not be
7023 documented here because they may change from time to time; we do not
7024 recommend general use of these functions.
7026 The remaining functions are provided for optimization purposes.
7028 @opindex fno-builtin
7029 GCC includes built-in versions of many of the functions in the standard
7030 C library. The versions prefixed with @code{__builtin_} will always be
7031 treated as having the same meaning as the C library function even if you
7032 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
7033 Many of these functions are only optimized in certain cases; if they are
7034 not optimized in a particular case, a call to the library function will
7039 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
7040 @option{-std=c99} or @option{-std=c1x}), the functions
7041 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
7042 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
7043 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
7044 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
7045 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
7046 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
7047 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
7048 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
7049 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
7050 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
7051 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
7052 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
7053 @code{signbitd64}, @code{signbitd128}, @code{significandf},
7054 @code{significandl}, @code{significand}, @code{sincosf},
7055 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
7056 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
7057 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
7058 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
7060 may be handled as built-in functions.
7061 All these functions have corresponding versions
7062 prefixed with @code{__builtin_}, which may be used even in strict C90
7065 The ISO C99 functions
7066 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
7067 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
7068 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
7069 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
7070 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
7071 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
7072 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
7073 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
7074 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
7075 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
7076 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
7077 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
7078 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
7079 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
7080 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
7081 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
7082 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
7083 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
7084 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
7085 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
7086 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
7087 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
7088 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
7089 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
7090 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
7091 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
7092 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
7093 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
7094 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
7095 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
7096 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
7097 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
7098 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
7099 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
7100 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
7101 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
7102 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
7103 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
7104 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
7105 are handled as built-in functions
7106 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7108 There are also built-in versions of the ISO C99 functions
7109 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
7110 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
7111 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
7112 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
7113 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
7114 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
7115 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
7116 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
7117 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
7118 that are recognized in any mode since ISO C90 reserves these names for
7119 the purpose to which ISO C99 puts them. All these functions have
7120 corresponding versions prefixed with @code{__builtin_}.
7122 The ISO C94 functions
7123 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
7124 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
7125 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
7127 are handled as built-in functions
7128 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7130 The ISO C90 functions
7131 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
7132 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
7133 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
7134 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
7135 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
7136 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
7137 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
7138 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
7139 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
7140 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
7141 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
7142 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
7143 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
7144 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
7145 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
7146 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
7147 are all recognized as built-in functions unless
7148 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
7149 is specified for an individual function). All of these functions have
7150 corresponding versions prefixed with @code{__builtin_}.
7152 GCC provides built-in versions of the ISO C99 floating point comparison
7153 macros that avoid raising exceptions for unordered operands. They have
7154 the same names as the standard macros ( @code{isgreater},
7155 @code{isgreaterequal}, @code{isless}, @code{islessequal},
7156 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
7157 prefixed. We intend for a library implementor to be able to simply
7158 @code{#define} each standard macro to its built-in equivalent.
7159 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
7160 @code{isinf_sign} and @code{isnormal} built-ins used with
7161 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
7162 builtins appear both with and without the @code{__builtin_} prefix.
7164 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
7166 You can use the built-in function @code{__builtin_types_compatible_p} to
7167 determine whether two types are the same.
7169 This built-in function returns 1 if the unqualified versions of the
7170 types @var{type1} and @var{type2} (which are types, not expressions) are
7171 compatible, 0 otherwise. The result of this built-in function can be
7172 used in integer constant expressions.
7174 This built-in function ignores top level qualifiers (e.g., @code{const},
7175 @code{volatile}). For example, @code{int} is equivalent to @code{const
7178 The type @code{int[]} and @code{int[5]} are compatible. On the other
7179 hand, @code{int} and @code{char *} are not compatible, even if the size
7180 of their types, on the particular architecture are the same. Also, the
7181 amount of pointer indirection is taken into account when determining
7182 similarity. Consequently, @code{short *} is not similar to
7183 @code{short **}. Furthermore, two types that are typedefed are
7184 considered compatible if their underlying types are compatible.
7186 An @code{enum} type is not considered to be compatible with another
7187 @code{enum} type even if both are compatible with the same integer
7188 type; this is what the C standard specifies.
7189 For example, @code{enum @{foo, bar@}} is not similar to
7190 @code{enum @{hot, dog@}}.
7192 You would typically use this function in code whose execution varies
7193 depending on the arguments' types. For example:
7198 typeof (x) tmp = (x); \
7199 if (__builtin_types_compatible_p (typeof (x), long double)) \
7200 tmp = foo_long_double (tmp); \
7201 else if (__builtin_types_compatible_p (typeof (x), double)) \
7202 tmp = foo_double (tmp); \
7203 else if (__builtin_types_compatible_p (typeof (x), float)) \
7204 tmp = foo_float (tmp); \
7211 @emph{Note:} This construct is only available for C@.
7215 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
7217 You can use the built-in function @code{__builtin_choose_expr} to
7218 evaluate code depending on the value of a constant expression. This
7219 built-in function returns @var{exp1} if @var{const_exp}, which is an
7220 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
7222 This built-in function is analogous to the @samp{? :} operator in C,
7223 except that the expression returned has its type unaltered by promotion
7224 rules. Also, the built-in function does not evaluate the expression
7225 that was not chosen. For example, if @var{const_exp} evaluates to true,
7226 @var{exp2} is not evaluated even if it has side-effects.
7228 This built-in function can return an lvalue if the chosen argument is an
7231 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
7232 type. Similarly, if @var{exp2} is returned, its return type is the same
7239 __builtin_choose_expr ( \
7240 __builtin_types_compatible_p (typeof (x), double), \
7242 __builtin_choose_expr ( \
7243 __builtin_types_compatible_p (typeof (x), float), \
7245 /* @r{The void expression results in a compile-time error} \
7246 @r{when assigning the result to something.} */ \
7250 @emph{Note:} This construct is only available for C@. Furthermore, the
7251 unused expression (@var{exp1} or @var{exp2} depending on the value of
7252 @var{const_exp}) may still generate syntax errors. This may change in
7257 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
7258 You can use the built-in function @code{__builtin_constant_p} to
7259 determine if a value is known to be constant at compile-time and hence
7260 that GCC can perform constant-folding on expressions involving that
7261 value. The argument of the function is the value to test. The function
7262 returns the integer 1 if the argument is known to be a compile-time
7263 constant and 0 if it is not known to be a compile-time constant. A
7264 return of 0 does not indicate that the value is @emph{not} a constant,
7265 but merely that GCC cannot prove it is a constant with the specified
7266 value of the @option{-O} option.
7268 You would typically use this function in an embedded application where
7269 memory was a critical resource. If you have some complex calculation,
7270 you may want it to be folded if it involves constants, but need to call
7271 a function if it does not. For example:
7274 #define Scale_Value(X) \
7275 (__builtin_constant_p (X) \
7276 ? ((X) * SCALE + OFFSET) : Scale (X))
7279 You may use this built-in function in either a macro or an inline
7280 function. However, if you use it in an inlined function and pass an
7281 argument of the function as the argument to the built-in, GCC will
7282 never return 1 when you call the inline function with a string constant
7283 or compound literal (@pxref{Compound Literals}) and will not return 1
7284 when you pass a constant numeric value to the inline function unless you
7285 specify the @option{-O} option.
7287 You may also use @code{__builtin_constant_p} in initializers for static
7288 data. For instance, you can write
7291 static const int table[] = @{
7292 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
7298 This is an acceptable initializer even if @var{EXPRESSION} is not a
7299 constant expression, including the case where
7300 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
7301 folded to a constant but @var{EXPRESSION} contains operands that would
7302 not otherwise be permitted in a static initializer (for example,
7303 @code{0 && foo ()}). GCC must be more conservative about evaluating the
7304 built-in in this case, because it has no opportunity to perform
7307 Previous versions of GCC did not accept this built-in in data
7308 initializers. The earliest version where it is completely safe is
7312 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
7313 @opindex fprofile-arcs
7314 You may use @code{__builtin_expect} to provide the compiler with
7315 branch prediction information. In general, you should prefer to
7316 use actual profile feedback for this (@option{-fprofile-arcs}), as
7317 programmers are notoriously bad at predicting how their programs
7318 actually perform. However, there are applications in which this
7319 data is hard to collect.
7321 The return value is the value of @var{exp}, which should be an integral
7322 expression. The semantics of the built-in are that it is expected that
7323 @var{exp} == @var{c}. For example:
7326 if (__builtin_expect (x, 0))
7331 would indicate that we do not expect to call @code{foo}, since
7332 we expect @code{x} to be zero. Since you are limited to integral
7333 expressions for @var{exp}, you should use constructions such as
7336 if (__builtin_expect (ptr != NULL, 1))
7341 when testing pointer or floating-point values.
7344 @deftypefn {Built-in Function} void __builtin_trap (void)
7345 This function causes the program to exit abnormally. GCC implements
7346 this function by using a target-dependent mechanism (such as
7347 intentionally executing an illegal instruction) or by calling
7348 @code{abort}. The mechanism used may vary from release to release so
7349 you should not rely on any particular implementation.
7352 @deftypefn {Built-in Function} void __builtin_unreachable (void)
7353 If control flow reaches the point of the @code{__builtin_unreachable},
7354 the program is undefined. It is useful in situations where the
7355 compiler cannot deduce the unreachability of the code.
7357 One such case is immediately following an @code{asm} statement that
7358 will either never terminate, or one that transfers control elsewhere
7359 and never returns. In this example, without the
7360 @code{__builtin_unreachable}, GCC would issue a warning that control
7361 reaches the end of a non-void function. It would also generate code
7362 to return after the @code{asm}.
7365 int f (int c, int v)
7373 asm("jmp error_handler");
7374 __builtin_unreachable ();
7379 Because the @code{asm} statement unconditionally transfers control out
7380 of the function, control will never reach the end of the function
7381 body. The @code{__builtin_unreachable} is in fact unreachable and
7382 communicates this fact to the compiler.
7384 Another use for @code{__builtin_unreachable} is following a call a
7385 function that never returns but that is not declared
7386 @code{__attribute__((noreturn))}, as in this example:
7389 void function_that_never_returns (void);
7399 function_that_never_returns ();
7400 __builtin_unreachable ();
7407 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
7408 This function is used to flush the processor's instruction cache for
7409 the region of memory between @var{begin} inclusive and @var{end}
7410 exclusive. Some targets require that the instruction cache be
7411 flushed, after modifying memory containing code, in order to obtain
7412 deterministic behavior.
7414 If the target does not require instruction cache flushes,
7415 @code{__builtin___clear_cache} has no effect. Otherwise either
7416 instructions are emitted in-line to clear the instruction cache or a
7417 call to the @code{__clear_cache} function in libgcc is made.
7420 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
7421 This function is used to minimize cache-miss latency by moving data into
7422 a cache before it is accessed.
7423 You can insert calls to @code{__builtin_prefetch} into code for which
7424 you know addresses of data in memory that is likely to be accessed soon.
7425 If the target supports them, data prefetch instructions will be generated.
7426 If the prefetch is done early enough before the access then the data will
7427 be in the cache by the time it is accessed.
7429 The value of @var{addr} is the address of the memory to prefetch.
7430 There are two optional arguments, @var{rw} and @var{locality}.
7431 The value of @var{rw} is a compile-time constant one or zero; one
7432 means that the prefetch is preparing for a write to the memory address
7433 and zero, the default, means that the prefetch is preparing for a read.
7434 The value @var{locality} must be a compile-time constant integer between
7435 zero and three. A value of zero means that the data has no temporal
7436 locality, so it need not be left in the cache after the access. A value
7437 of three means that the data has a high degree of temporal locality and
7438 should be left in all levels of cache possible. Values of one and two
7439 mean, respectively, a low or moderate degree of temporal locality. The
7443 for (i = 0; i < n; i++)
7446 __builtin_prefetch (&a[i+j], 1, 1);
7447 __builtin_prefetch (&b[i+j], 0, 1);
7452 Data prefetch does not generate faults if @var{addr} is invalid, but
7453 the address expression itself must be valid. For example, a prefetch
7454 of @code{p->next} will not fault if @code{p->next} is not a valid
7455 address, but evaluation will fault if @code{p} is not a valid address.
7457 If the target does not support data prefetch, the address expression
7458 is evaluated if it includes side effects but no other code is generated
7459 and GCC does not issue a warning.
7462 @deftypefn {Built-in Function} double __builtin_huge_val (void)
7463 Returns a positive infinity, if supported by the floating-point format,
7464 else @code{DBL_MAX}. This function is suitable for implementing the
7465 ISO C macro @code{HUGE_VAL}.
7468 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
7469 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
7472 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
7473 Similar to @code{__builtin_huge_val}, except the return
7474 type is @code{long double}.
7477 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
7478 This built-in implements the C99 fpclassify functionality. The first
7479 five int arguments should be the target library's notion of the
7480 possible FP classes and are used for return values. They must be
7481 constant values and they must appear in this order: @code{FP_NAN},
7482 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
7483 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
7484 to classify. GCC treats the last argument as type-generic, which
7485 means it does not do default promotion from float to double.
7488 @deftypefn {Built-in Function} double __builtin_inf (void)
7489 Similar to @code{__builtin_huge_val}, except a warning is generated
7490 if the target floating-point format does not support infinities.
7493 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
7494 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
7497 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
7498 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
7501 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
7502 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
7505 @deftypefn {Built-in Function} float __builtin_inff (void)
7506 Similar to @code{__builtin_inf}, except the return type is @code{float}.
7507 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
7510 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
7511 Similar to @code{__builtin_inf}, except the return
7512 type is @code{long double}.
7515 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
7516 Similar to @code{isinf}, except the return value will be negative for
7517 an argument of @code{-Inf}. Note while the parameter list is an
7518 ellipsis, this function only accepts exactly one floating point
7519 argument. GCC treats this parameter as type-generic, which means it
7520 does not do default promotion from float to double.
7523 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
7524 This is an implementation of the ISO C99 function @code{nan}.
7526 Since ISO C99 defines this function in terms of @code{strtod}, which we
7527 do not implement, a description of the parsing is in order. The string
7528 is parsed as by @code{strtol}; that is, the base is recognized by
7529 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
7530 in the significand such that the least significant bit of the number
7531 is at the least significant bit of the significand. The number is
7532 truncated to fit the significand field provided. The significand is
7533 forced to be a quiet NaN@.
7535 This function, if given a string literal all of which would have been
7536 consumed by strtol, is evaluated early enough that it is considered a
7537 compile-time constant.
7540 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
7541 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
7544 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
7545 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
7548 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
7549 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
7552 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
7553 Similar to @code{__builtin_nan}, except the return type is @code{float}.
7556 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
7557 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
7560 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
7561 Similar to @code{__builtin_nan}, except the significand is forced
7562 to be a signaling NaN@. The @code{nans} function is proposed by
7563 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
7566 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
7567 Similar to @code{__builtin_nans}, except the return type is @code{float}.
7570 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
7571 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
7574 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
7575 Returns one plus the index of the least significant 1-bit of @var{x}, or
7576 if @var{x} is zero, returns zero.
7579 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
7580 Returns the number of leading 0-bits in @var{x}, starting at the most
7581 significant bit position. If @var{x} is 0, the result is undefined.
7584 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
7585 Returns the number of trailing 0-bits in @var{x}, starting at the least
7586 significant bit position. If @var{x} is 0, the result is undefined.
7589 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
7590 Returns the number of 1-bits in @var{x}.
7593 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
7594 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
7598 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
7599 Similar to @code{__builtin_ffs}, except the argument type is
7600 @code{unsigned long}.
7603 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
7604 Similar to @code{__builtin_clz}, except the argument type is
7605 @code{unsigned long}.
7608 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
7609 Similar to @code{__builtin_ctz}, except the argument type is
7610 @code{unsigned long}.
7613 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
7614 Similar to @code{__builtin_popcount}, except the argument type is
7615 @code{unsigned long}.
7618 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
7619 Similar to @code{__builtin_parity}, except the argument type is
7620 @code{unsigned long}.
7623 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
7624 Similar to @code{__builtin_ffs}, except the argument type is
7625 @code{unsigned long long}.
7628 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
7629 Similar to @code{__builtin_clz}, except the argument type is
7630 @code{unsigned long long}.
7633 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
7634 Similar to @code{__builtin_ctz}, except the argument type is
7635 @code{unsigned long long}.
7638 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
7639 Similar to @code{__builtin_popcount}, except the argument type is
7640 @code{unsigned long long}.
7643 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
7644 Similar to @code{__builtin_parity}, except the argument type is
7645 @code{unsigned long long}.
7648 @deftypefn {Built-in Function} double __builtin_powi (double, int)
7649 Returns the first argument raised to the power of the second. Unlike the
7650 @code{pow} function no guarantees about precision and rounding are made.
7653 @deftypefn {Built-in Function} float __builtin_powif (float, int)
7654 Similar to @code{__builtin_powi}, except the argument and return types
7658 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
7659 Similar to @code{__builtin_powi}, except the argument and return types
7660 are @code{long double}.
7663 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
7664 Returns @var{x} with the order of the bytes reversed; for example,
7665 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
7669 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
7670 Similar to @code{__builtin_bswap32}, except the argument and return types
7674 @node Target Builtins
7675 @section Built-in Functions Specific to Particular Target Machines
7677 On some target machines, GCC supports many built-in functions specific
7678 to those machines. Generally these generate calls to specific machine
7679 instructions, but allow the compiler to schedule those calls.
7682 * Alpha Built-in Functions::
7683 * ARM iWMMXt Built-in Functions::
7684 * ARM NEON Intrinsics::
7685 * Blackfin Built-in Functions::
7686 * FR-V Built-in Functions::
7687 * X86 Built-in Functions::
7688 * MIPS DSP Built-in Functions::
7689 * MIPS Paired-Single Support::
7690 * MIPS Loongson Built-in Functions::
7691 * Other MIPS Built-in Functions::
7692 * picoChip Built-in Functions::
7693 * PowerPC AltiVec/VSX Built-in Functions::
7694 * RX Built-in Functions::
7695 * SPARC VIS Built-in Functions::
7696 * SPU Built-in Functions::
7699 @node Alpha Built-in Functions
7700 @subsection Alpha Built-in Functions
7702 These built-in functions are available for the Alpha family of
7703 processors, depending on the command-line switches used.
7705 The following built-in functions are always available. They
7706 all generate the machine instruction that is part of the name.
7709 long __builtin_alpha_implver (void)
7710 long __builtin_alpha_rpcc (void)
7711 long __builtin_alpha_amask (long)
7712 long __builtin_alpha_cmpbge (long, long)
7713 long __builtin_alpha_extbl (long, long)
7714 long __builtin_alpha_extwl (long, long)
7715 long __builtin_alpha_extll (long, long)
7716 long __builtin_alpha_extql (long, long)
7717 long __builtin_alpha_extwh (long, long)
7718 long __builtin_alpha_extlh (long, long)
7719 long __builtin_alpha_extqh (long, long)
7720 long __builtin_alpha_insbl (long, long)
7721 long __builtin_alpha_inswl (long, long)
7722 long __builtin_alpha_insll (long, long)
7723 long __builtin_alpha_insql (long, long)
7724 long __builtin_alpha_inswh (long, long)
7725 long __builtin_alpha_inslh (long, long)
7726 long __builtin_alpha_insqh (long, long)
7727 long __builtin_alpha_mskbl (long, long)
7728 long __builtin_alpha_mskwl (long, long)
7729 long __builtin_alpha_mskll (long, long)
7730 long __builtin_alpha_mskql (long, long)
7731 long __builtin_alpha_mskwh (long, long)
7732 long __builtin_alpha_msklh (long, long)
7733 long __builtin_alpha_mskqh (long, long)
7734 long __builtin_alpha_umulh (long, long)
7735 long __builtin_alpha_zap (long, long)
7736 long __builtin_alpha_zapnot (long, long)
7739 The following built-in functions are always with @option{-mmax}
7740 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
7741 later. They all generate the machine instruction that is part
7745 long __builtin_alpha_pklb (long)
7746 long __builtin_alpha_pkwb (long)
7747 long __builtin_alpha_unpkbl (long)
7748 long __builtin_alpha_unpkbw (long)
7749 long __builtin_alpha_minub8 (long, long)
7750 long __builtin_alpha_minsb8 (long, long)
7751 long __builtin_alpha_minuw4 (long, long)
7752 long __builtin_alpha_minsw4 (long, long)
7753 long __builtin_alpha_maxub8 (long, long)
7754 long __builtin_alpha_maxsb8 (long, long)
7755 long __builtin_alpha_maxuw4 (long, long)
7756 long __builtin_alpha_maxsw4 (long, long)
7757 long __builtin_alpha_perr (long, long)
7760 The following built-in functions are always with @option{-mcix}
7761 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
7762 later. They all generate the machine instruction that is part
7766 long __builtin_alpha_cttz (long)
7767 long __builtin_alpha_ctlz (long)
7768 long __builtin_alpha_ctpop (long)
7771 The following builtins are available on systems that use the OSF/1
7772 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
7773 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
7774 @code{rdval} and @code{wrval}.
7777 void *__builtin_thread_pointer (void)
7778 void __builtin_set_thread_pointer (void *)
7781 @node ARM iWMMXt Built-in Functions
7782 @subsection ARM iWMMXt Built-in Functions
7784 These built-in functions are available for the ARM family of
7785 processors when the @option{-mcpu=iwmmxt} switch is used:
7788 typedef int v2si __attribute__ ((vector_size (8)));
7789 typedef short v4hi __attribute__ ((vector_size (8)));
7790 typedef char v8qi __attribute__ ((vector_size (8)));
7792 int __builtin_arm_getwcx (int)
7793 void __builtin_arm_setwcx (int, int)
7794 int __builtin_arm_textrmsb (v8qi, int)
7795 int __builtin_arm_textrmsh (v4hi, int)
7796 int __builtin_arm_textrmsw (v2si, int)
7797 int __builtin_arm_textrmub (v8qi, int)
7798 int __builtin_arm_textrmuh (v4hi, int)
7799 int __builtin_arm_textrmuw (v2si, int)
7800 v8qi __builtin_arm_tinsrb (v8qi, int)
7801 v4hi __builtin_arm_tinsrh (v4hi, int)
7802 v2si __builtin_arm_tinsrw (v2si, int)
7803 long long __builtin_arm_tmia (long long, int, int)
7804 long long __builtin_arm_tmiabb (long long, int, int)
7805 long long __builtin_arm_tmiabt (long long, int, int)
7806 long long __builtin_arm_tmiaph (long long, int, int)
7807 long long __builtin_arm_tmiatb (long long, int, int)
7808 long long __builtin_arm_tmiatt (long long, int, int)
7809 int __builtin_arm_tmovmskb (v8qi)
7810 int __builtin_arm_tmovmskh (v4hi)
7811 int __builtin_arm_tmovmskw (v2si)
7812 long long __builtin_arm_waccb (v8qi)
7813 long long __builtin_arm_wacch (v4hi)
7814 long long __builtin_arm_waccw (v2si)
7815 v8qi __builtin_arm_waddb (v8qi, v8qi)
7816 v8qi __builtin_arm_waddbss (v8qi, v8qi)
7817 v8qi __builtin_arm_waddbus (v8qi, v8qi)
7818 v4hi __builtin_arm_waddh (v4hi, v4hi)
7819 v4hi __builtin_arm_waddhss (v4hi, v4hi)
7820 v4hi __builtin_arm_waddhus (v4hi, v4hi)
7821 v2si __builtin_arm_waddw (v2si, v2si)
7822 v2si __builtin_arm_waddwss (v2si, v2si)
7823 v2si __builtin_arm_waddwus (v2si, v2si)
7824 v8qi __builtin_arm_walign (v8qi, v8qi, int)
7825 long long __builtin_arm_wand(long long, long long)
7826 long long __builtin_arm_wandn (long long, long long)
7827 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
7828 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
7829 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
7830 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
7831 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
7832 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
7833 v2si __builtin_arm_wcmpeqw (v2si, v2si)
7834 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
7835 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
7836 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
7837 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
7838 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
7839 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
7840 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
7841 long long __builtin_arm_wmacsz (v4hi, v4hi)
7842 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
7843 long long __builtin_arm_wmacuz (v4hi, v4hi)
7844 v4hi __builtin_arm_wmadds (v4hi, v4hi)
7845 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
7846 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
7847 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
7848 v2si __builtin_arm_wmaxsw (v2si, v2si)
7849 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
7850 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
7851 v2si __builtin_arm_wmaxuw (v2si, v2si)
7852 v8qi __builtin_arm_wminsb (v8qi, v8qi)
7853 v4hi __builtin_arm_wminsh (v4hi, v4hi)
7854 v2si __builtin_arm_wminsw (v2si, v2si)
7855 v8qi __builtin_arm_wminub (v8qi, v8qi)
7856 v4hi __builtin_arm_wminuh (v4hi, v4hi)
7857 v2si __builtin_arm_wminuw (v2si, v2si)
7858 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
7859 v4hi __builtin_arm_wmulul (v4hi, v4hi)
7860 v4hi __builtin_arm_wmulum (v4hi, v4hi)
7861 long long __builtin_arm_wor (long long, long long)
7862 v2si __builtin_arm_wpackdss (long long, long long)
7863 v2si __builtin_arm_wpackdus (long long, long long)
7864 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
7865 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
7866 v4hi __builtin_arm_wpackwss (v2si, v2si)
7867 v4hi __builtin_arm_wpackwus (v2si, v2si)
7868 long long __builtin_arm_wrord (long long, long long)
7869 long long __builtin_arm_wrordi (long long, int)
7870 v4hi __builtin_arm_wrorh (v4hi, long long)
7871 v4hi __builtin_arm_wrorhi (v4hi, int)
7872 v2si __builtin_arm_wrorw (v2si, long long)
7873 v2si __builtin_arm_wrorwi (v2si, int)
7874 v2si __builtin_arm_wsadb (v8qi, v8qi)
7875 v2si __builtin_arm_wsadbz (v8qi, v8qi)
7876 v2si __builtin_arm_wsadh (v4hi, v4hi)
7877 v2si __builtin_arm_wsadhz (v4hi, v4hi)
7878 v4hi __builtin_arm_wshufh (v4hi, int)
7879 long long __builtin_arm_wslld (long long, long long)
7880 long long __builtin_arm_wslldi (long long, int)
7881 v4hi __builtin_arm_wsllh (v4hi, long long)
7882 v4hi __builtin_arm_wsllhi (v4hi, int)
7883 v2si __builtin_arm_wsllw (v2si, long long)
7884 v2si __builtin_arm_wsllwi (v2si, int)
7885 long long __builtin_arm_wsrad (long long, long long)
7886 long long __builtin_arm_wsradi (long long, int)
7887 v4hi __builtin_arm_wsrah (v4hi, long long)
7888 v4hi __builtin_arm_wsrahi (v4hi, int)
7889 v2si __builtin_arm_wsraw (v2si, long long)
7890 v2si __builtin_arm_wsrawi (v2si, int)
7891 long long __builtin_arm_wsrld (long long, long long)
7892 long long __builtin_arm_wsrldi (long long, int)
7893 v4hi __builtin_arm_wsrlh (v4hi, long long)
7894 v4hi __builtin_arm_wsrlhi (v4hi, int)
7895 v2si __builtin_arm_wsrlw (v2si, long long)
7896 v2si __builtin_arm_wsrlwi (v2si, int)
7897 v8qi __builtin_arm_wsubb (v8qi, v8qi)
7898 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
7899 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
7900 v4hi __builtin_arm_wsubh (v4hi, v4hi)
7901 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
7902 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
7903 v2si __builtin_arm_wsubw (v2si, v2si)
7904 v2si __builtin_arm_wsubwss (v2si, v2si)
7905 v2si __builtin_arm_wsubwus (v2si, v2si)
7906 v4hi __builtin_arm_wunpckehsb (v8qi)
7907 v2si __builtin_arm_wunpckehsh (v4hi)
7908 long long __builtin_arm_wunpckehsw (v2si)
7909 v4hi __builtin_arm_wunpckehub (v8qi)
7910 v2si __builtin_arm_wunpckehuh (v4hi)
7911 long long __builtin_arm_wunpckehuw (v2si)
7912 v4hi __builtin_arm_wunpckelsb (v8qi)
7913 v2si __builtin_arm_wunpckelsh (v4hi)
7914 long long __builtin_arm_wunpckelsw (v2si)
7915 v4hi __builtin_arm_wunpckelub (v8qi)
7916 v2si __builtin_arm_wunpckeluh (v4hi)
7917 long long __builtin_arm_wunpckeluw (v2si)
7918 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
7919 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
7920 v2si __builtin_arm_wunpckihw (v2si, v2si)
7921 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
7922 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
7923 v2si __builtin_arm_wunpckilw (v2si, v2si)
7924 long long __builtin_arm_wxor (long long, long long)
7925 long long __builtin_arm_wzero ()
7928 @node ARM NEON Intrinsics
7929 @subsection ARM NEON Intrinsics
7931 These built-in intrinsics for the ARM Advanced SIMD extension are available
7932 when the @option{-mfpu=neon} switch is used:
7934 @include arm-neon-intrinsics.texi
7936 @node Blackfin Built-in Functions
7937 @subsection Blackfin Built-in Functions
7939 Currently, there are two Blackfin-specific built-in functions. These are
7940 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
7941 using inline assembly; by using these built-in functions the compiler can
7942 automatically add workarounds for hardware errata involving these
7943 instructions. These functions are named as follows:
7946 void __builtin_bfin_csync (void)
7947 void __builtin_bfin_ssync (void)
7950 @node FR-V Built-in Functions
7951 @subsection FR-V Built-in Functions
7953 GCC provides many FR-V-specific built-in functions. In general,
7954 these functions are intended to be compatible with those described
7955 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
7956 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
7957 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
7958 pointer rather than by value.
7960 Most of the functions are named after specific FR-V instructions.
7961 Such functions are said to be ``directly mapped'' and are summarized
7962 here in tabular form.
7966 * Directly-mapped Integer Functions::
7967 * Directly-mapped Media Functions::
7968 * Raw read/write Functions::
7969 * Other Built-in Functions::
7972 @node Argument Types
7973 @subsubsection Argument Types
7975 The arguments to the built-in functions can be divided into three groups:
7976 register numbers, compile-time constants and run-time values. In order
7977 to make this classification clear at a glance, the arguments and return
7978 values are given the following pseudo types:
7980 @multitable @columnfractions .20 .30 .15 .35
7981 @item Pseudo type @tab Real C type @tab Constant? @tab Description
7982 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
7983 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
7984 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
7985 @item @code{uw2} @tab @code{unsigned long long} @tab No
7986 @tab an unsigned doubleword
7987 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
7988 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
7989 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
7990 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
7993 These pseudo types are not defined by GCC, they are simply a notational
7994 convenience used in this manual.
7996 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
7997 and @code{sw2} are evaluated at run time. They correspond to
7998 register operands in the underlying FR-V instructions.
8000 @code{const} arguments represent immediate operands in the underlying
8001 FR-V instructions. They must be compile-time constants.
8003 @code{acc} arguments are evaluated at compile time and specify the number
8004 of an accumulator register. For example, an @code{acc} argument of 2
8005 will select the ACC2 register.
8007 @code{iacc} arguments are similar to @code{acc} arguments but specify the
8008 number of an IACC register. See @pxref{Other Built-in Functions}
8011 @node Directly-mapped Integer Functions
8012 @subsubsection Directly-mapped Integer Functions
8014 The functions listed below map directly to FR-V I-type instructions.
8016 @multitable @columnfractions .45 .32 .23
8017 @item Function prototype @tab Example usage @tab Assembly output
8018 @item @code{sw1 __ADDSS (sw1, sw1)}
8019 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
8020 @tab @code{ADDSS @var{a},@var{b},@var{c}}
8021 @item @code{sw1 __SCAN (sw1, sw1)}
8022 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
8023 @tab @code{SCAN @var{a},@var{b},@var{c}}
8024 @item @code{sw1 __SCUTSS (sw1)}
8025 @tab @code{@var{b} = __SCUTSS (@var{a})}
8026 @tab @code{SCUTSS @var{a},@var{b}}
8027 @item @code{sw1 __SLASS (sw1, sw1)}
8028 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
8029 @tab @code{SLASS @var{a},@var{b},@var{c}}
8030 @item @code{void __SMASS (sw1, sw1)}
8031 @tab @code{__SMASS (@var{a}, @var{b})}
8032 @tab @code{SMASS @var{a},@var{b}}
8033 @item @code{void __SMSSS (sw1, sw1)}
8034 @tab @code{__SMSSS (@var{a}, @var{b})}
8035 @tab @code{SMSSS @var{a},@var{b}}
8036 @item @code{void __SMU (sw1, sw1)}
8037 @tab @code{__SMU (@var{a}, @var{b})}
8038 @tab @code{SMU @var{a},@var{b}}
8039 @item @code{sw2 __SMUL (sw1, sw1)}
8040 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
8041 @tab @code{SMUL @var{a},@var{b},@var{c}}
8042 @item @code{sw1 __SUBSS (sw1, sw1)}
8043 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
8044 @tab @code{SUBSS @var{a},@var{b},@var{c}}
8045 @item @code{uw2 __UMUL (uw1, uw1)}
8046 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
8047 @tab @code{UMUL @var{a},@var{b},@var{c}}
8050 @node Directly-mapped Media Functions
8051 @subsubsection Directly-mapped Media Functions
8053 The functions listed below map directly to FR-V M-type instructions.
8055 @multitable @columnfractions .45 .32 .23
8056 @item Function prototype @tab Example usage @tab Assembly output
8057 @item @code{uw1 __MABSHS (sw1)}
8058 @tab @code{@var{b} = __MABSHS (@var{a})}
8059 @tab @code{MABSHS @var{a},@var{b}}
8060 @item @code{void __MADDACCS (acc, acc)}
8061 @tab @code{__MADDACCS (@var{b}, @var{a})}
8062 @tab @code{MADDACCS @var{a},@var{b}}
8063 @item @code{sw1 __MADDHSS (sw1, sw1)}
8064 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
8065 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
8066 @item @code{uw1 __MADDHUS (uw1, uw1)}
8067 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
8068 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
8069 @item @code{uw1 __MAND (uw1, uw1)}
8070 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
8071 @tab @code{MAND @var{a},@var{b},@var{c}}
8072 @item @code{void __MASACCS (acc, acc)}
8073 @tab @code{__MASACCS (@var{b}, @var{a})}
8074 @tab @code{MASACCS @var{a},@var{b}}
8075 @item @code{uw1 __MAVEH (uw1, uw1)}
8076 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
8077 @tab @code{MAVEH @var{a},@var{b},@var{c}}
8078 @item @code{uw2 __MBTOH (uw1)}
8079 @tab @code{@var{b} = __MBTOH (@var{a})}
8080 @tab @code{MBTOH @var{a},@var{b}}
8081 @item @code{void __MBTOHE (uw1 *, uw1)}
8082 @tab @code{__MBTOHE (&@var{b}, @var{a})}
8083 @tab @code{MBTOHE @var{a},@var{b}}
8084 @item @code{void __MCLRACC (acc)}
8085 @tab @code{__MCLRACC (@var{a})}
8086 @tab @code{MCLRACC @var{a}}
8087 @item @code{void __MCLRACCA (void)}
8088 @tab @code{__MCLRACCA ()}
8089 @tab @code{MCLRACCA}
8090 @item @code{uw1 __Mcop1 (uw1, uw1)}
8091 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
8092 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
8093 @item @code{uw1 __Mcop2 (uw1, uw1)}
8094 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
8095 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
8096 @item @code{uw1 __MCPLHI (uw2, const)}
8097 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
8098 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
8099 @item @code{uw1 __MCPLI (uw2, const)}
8100 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
8101 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
8102 @item @code{void __MCPXIS (acc, sw1, sw1)}
8103 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
8104 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
8105 @item @code{void __MCPXIU (acc, uw1, uw1)}
8106 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
8107 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
8108 @item @code{void __MCPXRS (acc, sw1, sw1)}
8109 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
8110 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
8111 @item @code{void __MCPXRU (acc, uw1, uw1)}
8112 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
8113 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
8114 @item @code{uw1 __MCUT (acc, uw1)}
8115 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
8116 @tab @code{MCUT @var{a},@var{b},@var{c}}
8117 @item @code{uw1 __MCUTSS (acc, sw1)}
8118 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
8119 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
8120 @item @code{void __MDADDACCS (acc, acc)}
8121 @tab @code{__MDADDACCS (@var{b}, @var{a})}
8122 @tab @code{MDADDACCS @var{a},@var{b}}
8123 @item @code{void __MDASACCS (acc, acc)}
8124 @tab @code{__MDASACCS (@var{b}, @var{a})}
8125 @tab @code{MDASACCS @var{a},@var{b}}
8126 @item @code{uw2 __MDCUTSSI (acc, const)}
8127 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
8128 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
8129 @item @code{uw2 __MDPACKH (uw2, uw2)}
8130 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
8131 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
8132 @item @code{uw2 __MDROTLI (uw2, const)}
8133 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
8134 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
8135 @item @code{void __MDSUBACCS (acc, acc)}
8136 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
8137 @tab @code{MDSUBACCS @var{a},@var{b}}
8138 @item @code{void __MDUNPACKH (uw1 *, uw2)}
8139 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
8140 @tab @code{MDUNPACKH @var{a},@var{b}}
8141 @item @code{uw2 __MEXPDHD (uw1, const)}
8142 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
8143 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
8144 @item @code{uw1 __MEXPDHW (uw1, const)}
8145 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
8146 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
8147 @item @code{uw1 __MHDSETH (uw1, const)}
8148 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
8149 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
8150 @item @code{sw1 __MHDSETS (const)}
8151 @tab @code{@var{b} = __MHDSETS (@var{a})}
8152 @tab @code{MHDSETS #@var{a},@var{b}}
8153 @item @code{uw1 __MHSETHIH (uw1, const)}
8154 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
8155 @tab @code{MHSETHIH #@var{a},@var{b}}
8156 @item @code{sw1 __MHSETHIS (sw1, const)}
8157 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
8158 @tab @code{MHSETHIS #@var{a},@var{b}}
8159 @item @code{uw1 __MHSETLOH (uw1, const)}
8160 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
8161 @tab @code{MHSETLOH #@var{a},@var{b}}
8162 @item @code{sw1 __MHSETLOS (sw1, const)}
8163 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
8164 @tab @code{MHSETLOS #@var{a},@var{b}}
8165 @item @code{uw1 __MHTOB (uw2)}
8166 @tab @code{@var{b} = __MHTOB (@var{a})}
8167 @tab @code{MHTOB @var{a},@var{b}}
8168 @item @code{void __MMACHS (acc, sw1, sw1)}
8169 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
8170 @tab @code{MMACHS @var{a},@var{b},@var{c}}
8171 @item @code{void __MMACHU (acc, uw1, uw1)}
8172 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
8173 @tab @code{MMACHU @var{a},@var{b},@var{c}}
8174 @item @code{void __MMRDHS (acc, sw1, sw1)}
8175 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
8176 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
8177 @item @code{void __MMRDHU (acc, uw1, uw1)}
8178 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
8179 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
8180 @item @code{void __MMULHS (acc, sw1, sw1)}
8181 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
8182 @tab @code{MMULHS @var{a},@var{b},@var{c}}
8183 @item @code{void __MMULHU (acc, uw1, uw1)}
8184 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
8185 @tab @code{MMULHU @var{a},@var{b},@var{c}}
8186 @item @code{void __MMULXHS (acc, sw1, sw1)}
8187 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
8188 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
8189 @item @code{void __MMULXHU (acc, uw1, uw1)}
8190 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
8191 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
8192 @item @code{uw1 __MNOT (uw1)}
8193 @tab @code{@var{b} = __MNOT (@var{a})}
8194 @tab @code{MNOT @var{a},@var{b}}
8195 @item @code{uw1 __MOR (uw1, uw1)}
8196 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
8197 @tab @code{MOR @var{a},@var{b},@var{c}}
8198 @item @code{uw1 __MPACKH (uh, uh)}
8199 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
8200 @tab @code{MPACKH @var{a},@var{b},@var{c}}
8201 @item @code{sw2 __MQADDHSS (sw2, sw2)}
8202 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
8203 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
8204 @item @code{uw2 __MQADDHUS (uw2, uw2)}
8205 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
8206 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
8207 @item @code{void __MQCPXIS (acc, sw2, sw2)}
8208 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
8209 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
8210 @item @code{void __MQCPXIU (acc, uw2, uw2)}
8211 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
8212 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
8213 @item @code{void __MQCPXRS (acc, sw2, sw2)}
8214 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
8215 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
8216 @item @code{void __MQCPXRU (acc, uw2, uw2)}
8217 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
8218 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
8219 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
8220 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
8221 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
8222 @item @code{sw2 __MQLMTHS (sw2, sw2)}
8223 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
8224 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
8225 @item @code{void __MQMACHS (acc, sw2, sw2)}
8226 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
8227 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
8228 @item @code{void __MQMACHU (acc, uw2, uw2)}
8229 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
8230 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
8231 @item @code{void __MQMACXHS (acc, sw2, sw2)}
8232 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
8233 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
8234 @item @code{void __MQMULHS (acc, sw2, sw2)}
8235 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
8236 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
8237 @item @code{void __MQMULHU (acc, uw2, uw2)}
8238 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
8239 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
8240 @item @code{void __MQMULXHS (acc, sw2, sw2)}
8241 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
8242 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
8243 @item @code{void __MQMULXHU (acc, uw2, uw2)}
8244 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
8245 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
8246 @item @code{sw2 __MQSATHS (sw2, sw2)}
8247 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
8248 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
8249 @item @code{uw2 __MQSLLHI (uw2, int)}
8250 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
8251 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
8252 @item @code{sw2 __MQSRAHI (sw2, int)}
8253 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
8254 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
8255 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
8256 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
8257 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
8258 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
8259 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
8260 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
8261 @item @code{void __MQXMACHS (acc, sw2, sw2)}
8262 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
8263 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
8264 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
8265 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
8266 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
8267 @item @code{uw1 __MRDACC (acc)}
8268 @tab @code{@var{b} = __MRDACC (@var{a})}
8269 @tab @code{MRDACC @var{a},@var{b}}
8270 @item @code{uw1 __MRDACCG (acc)}
8271 @tab @code{@var{b} = __MRDACCG (@var{a})}
8272 @tab @code{MRDACCG @var{a},@var{b}}
8273 @item @code{uw1 __MROTLI (uw1, const)}
8274 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
8275 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
8276 @item @code{uw1 __MROTRI (uw1, const)}
8277 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
8278 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
8279 @item @code{sw1 __MSATHS (sw1, sw1)}
8280 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
8281 @tab @code{MSATHS @var{a},@var{b},@var{c}}
8282 @item @code{uw1 __MSATHU (uw1, uw1)}
8283 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
8284 @tab @code{MSATHU @var{a},@var{b},@var{c}}
8285 @item @code{uw1 __MSLLHI (uw1, const)}
8286 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
8287 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
8288 @item @code{sw1 __MSRAHI (sw1, const)}
8289 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
8290 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
8291 @item @code{uw1 __MSRLHI (uw1, const)}
8292 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
8293 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
8294 @item @code{void __MSUBACCS (acc, acc)}
8295 @tab @code{__MSUBACCS (@var{b}, @var{a})}
8296 @tab @code{MSUBACCS @var{a},@var{b}}
8297 @item @code{sw1 __MSUBHSS (sw1, sw1)}
8298 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
8299 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
8300 @item @code{uw1 __MSUBHUS (uw1, uw1)}
8301 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
8302 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
8303 @item @code{void __MTRAP (void)}
8304 @tab @code{__MTRAP ()}
8306 @item @code{uw2 __MUNPACKH (uw1)}
8307 @tab @code{@var{b} = __MUNPACKH (@var{a})}
8308 @tab @code{MUNPACKH @var{a},@var{b}}
8309 @item @code{uw1 __MWCUT (uw2, uw1)}
8310 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
8311 @tab @code{MWCUT @var{a},@var{b},@var{c}}
8312 @item @code{void __MWTACC (acc, uw1)}
8313 @tab @code{__MWTACC (@var{b}, @var{a})}
8314 @tab @code{MWTACC @var{a},@var{b}}
8315 @item @code{void __MWTACCG (acc, uw1)}
8316 @tab @code{__MWTACCG (@var{b}, @var{a})}
8317 @tab @code{MWTACCG @var{a},@var{b}}
8318 @item @code{uw1 __MXOR (uw1, uw1)}
8319 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
8320 @tab @code{MXOR @var{a},@var{b},@var{c}}
8323 @node Raw read/write Functions
8324 @subsubsection Raw read/write Functions
8326 This sections describes built-in functions related to read and write
8327 instructions to access memory. These functions generate
8328 @code{membar} instructions to flush the I/O load and stores where
8329 appropriate, as described in Fujitsu's manual described above.
8333 @item unsigned char __builtin_read8 (void *@var{data})
8334 @item unsigned short __builtin_read16 (void *@var{data})
8335 @item unsigned long __builtin_read32 (void *@var{data})
8336 @item unsigned long long __builtin_read64 (void *@var{data})
8338 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
8339 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
8340 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
8341 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
8344 @node Other Built-in Functions
8345 @subsubsection Other Built-in Functions
8347 This section describes built-in functions that are not named after
8348 a specific FR-V instruction.
8351 @item sw2 __IACCreadll (iacc @var{reg})
8352 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
8353 for future expansion and must be 0.
8355 @item sw1 __IACCreadl (iacc @var{reg})
8356 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
8357 Other values of @var{reg} are rejected as invalid.
8359 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
8360 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
8361 is reserved for future expansion and must be 0.
8363 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
8364 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
8365 is 1. Other values of @var{reg} are rejected as invalid.
8367 @item void __data_prefetch0 (const void *@var{x})
8368 Use the @code{dcpl} instruction to load the contents of address @var{x}
8369 into the data cache.
8371 @item void __data_prefetch (const void *@var{x})
8372 Use the @code{nldub} instruction to load the contents of address @var{x}
8373 into the data cache. The instruction will be issued in slot I1@.
8376 @node X86 Built-in Functions
8377 @subsection X86 Built-in Functions
8379 These built-in functions are available for the i386 and x86-64 family
8380 of computers, depending on the command-line switches used.
8382 Note that, if you specify command-line switches such as @option{-msse},
8383 the compiler could use the extended instruction sets even if the built-ins
8384 are not used explicitly in the program. For this reason, applications
8385 which perform runtime CPU detection must compile separate files for each
8386 supported architecture, using the appropriate flags. In particular,
8387 the file containing the CPU detection code should be compiled without
8390 The following machine modes are available for use with MMX built-in functions
8391 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
8392 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
8393 vector of eight 8-bit integers. Some of the built-in functions operate on
8394 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
8396 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
8397 of two 32-bit floating point values.
8399 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
8400 floating point values. Some instructions use a vector of four 32-bit
8401 integers, these use @code{V4SI}. Finally, some instructions operate on an
8402 entire vector register, interpreting it as a 128-bit integer, these use mode
8405 In 64-bit mode, the x86-64 family of processors uses additional built-in
8406 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
8407 floating point and @code{TC} 128-bit complex floating point values.
8409 The following floating point built-in functions are available in 64-bit
8410 mode. All of them implement the function that is part of the name.
8413 __float128 __builtin_fabsq (__float128)
8414 __float128 __builtin_copysignq (__float128, __float128)
8417 The following floating point built-in functions are made available in the
8421 @item __float128 __builtin_infq (void)
8422 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
8423 @findex __builtin_infq
8425 @item __float128 __builtin_huge_valq (void)
8426 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
8427 @findex __builtin_huge_valq
8430 The following built-in functions are made available by @option{-mmmx}.
8431 All of them generate the machine instruction that is part of the name.
8434 v8qi __builtin_ia32_paddb (v8qi, v8qi)
8435 v4hi __builtin_ia32_paddw (v4hi, v4hi)
8436 v2si __builtin_ia32_paddd (v2si, v2si)
8437 v8qi __builtin_ia32_psubb (v8qi, v8qi)
8438 v4hi __builtin_ia32_psubw (v4hi, v4hi)
8439 v2si __builtin_ia32_psubd (v2si, v2si)
8440 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
8441 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
8442 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
8443 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
8444 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
8445 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
8446 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
8447 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
8448 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
8449 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
8450 di __builtin_ia32_pand (di, di)
8451 di __builtin_ia32_pandn (di,di)
8452 di __builtin_ia32_por (di, di)
8453 di __builtin_ia32_pxor (di, di)
8454 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
8455 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
8456 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
8457 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
8458 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
8459 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
8460 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
8461 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
8462 v2si __builtin_ia32_punpckhdq (v2si, v2si)
8463 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
8464 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
8465 v2si __builtin_ia32_punpckldq (v2si, v2si)
8466 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
8467 v4hi __builtin_ia32_packssdw (v2si, v2si)
8468 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
8470 v4hi __builtin_ia32_psllw (v4hi, v4hi)
8471 v2si __builtin_ia32_pslld (v2si, v2si)
8472 v1di __builtin_ia32_psllq (v1di, v1di)
8473 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
8474 v2si __builtin_ia32_psrld (v2si, v2si)
8475 v1di __builtin_ia32_psrlq (v1di, v1di)
8476 v4hi __builtin_ia32_psraw (v4hi, v4hi)
8477 v2si __builtin_ia32_psrad (v2si, v2si)
8478 v4hi __builtin_ia32_psllwi (v4hi, int)
8479 v2si __builtin_ia32_pslldi (v2si, int)
8480 v1di __builtin_ia32_psllqi (v1di, int)
8481 v4hi __builtin_ia32_psrlwi (v4hi, int)
8482 v2si __builtin_ia32_psrldi (v2si, int)
8483 v1di __builtin_ia32_psrlqi (v1di, int)
8484 v4hi __builtin_ia32_psrawi (v4hi, int)
8485 v2si __builtin_ia32_psradi (v2si, int)
8489 The following built-in functions are made available either with
8490 @option{-msse}, or with a combination of @option{-m3dnow} and
8491 @option{-march=athlon}. All of them generate the machine
8492 instruction that is part of the name.
8495 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
8496 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
8497 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
8498 v1di __builtin_ia32_psadbw (v8qi, v8qi)
8499 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
8500 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
8501 v8qi __builtin_ia32_pminub (v8qi, v8qi)
8502 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
8503 int __builtin_ia32_pextrw (v4hi, int)
8504 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
8505 int __builtin_ia32_pmovmskb (v8qi)
8506 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
8507 void __builtin_ia32_movntq (di *, di)
8508 void __builtin_ia32_sfence (void)
8511 The following built-in functions are available when @option{-msse} is used.
8512 All of them generate the machine instruction that is part of the name.
8515 int __builtin_ia32_comieq (v4sf, v4sf)
8516 int __builtin_ia32_comineq (v4sf, v4sf)
8517 int __builtin_ia32_comilt (v4sf, v4sf)
8518 int __builtin_ia32_comile (v4sf, v4sf)
8519 int __builtin_ia32_comigt (v4sf, v4sf)
8520 int __builtin_ia32_comige (v4sf, v4sf)
8521 int __builtin_ia32_ucomieq (v4sf, v4sf)
8522 int __builtin_ia32_ucomineq (v4sf, v4sf)
8523 int __builtin_ia32_ucomilt (v4sf, v4sf)
8524 int __builtin_ia32_ucomile (v4sf, v4sf)
8525 int __builtin_ia32_ucomigt (v4sf, v4sf)
8526 int __builtin_ia32_ucomige (v4sf, v4sf)
8527 v4sf __builtin_ia32_addps (v4sf, v4sf)
8528 v4sf __builtin_ia32_subps (v4sf, v4sf)
8529 v4sf __builtin_ia32_mulps (v4sf, v4sf)
8530 v4sf __builtin_ia32_divps (v4sf, v4sf)
8531 v4sf __builtin_ia32_addss (v4sf, v4sf)
8532 v4sf __builtin_ia32_subss (v4sf, v4sf)
8533 v4sf __builtin_ia32_mulss (v4sf, v4sf)
8534 v4sf __builtin_ia32_divss (v4sf, v4sf)
8535 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
8536 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
8537 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
8538 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
8539 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
8540 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
8541 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
8542 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
8543 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
8544 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
8545 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
8546 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
8547 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
8548 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
8549 v4si __builtin_ia32_cmpless (v4sf, v4sf)
8550 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
8551 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
8552 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
8553 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
8554 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
8555 v4sf __builtin_ia32_maxps (v4sf, v4sf)
8556 v4sf __builtin_ia32_maxss (v4sf, v4sf)
8557 v4sf __builtin_ia32_minps (v4sf, v4sf)
8558 v4sf __builtin_ia32_minss (v4sf, v4sf)
8559 v4sf __builtin_ia32_andps (v4sf, v4sf)
8560 v4sf __builtin_ia32_andnps (v4sf, v4sf)
8561 v4sf __builtin_ia32_orps (v4sf, v4sf)
8562 v4sf __builtin_ia32_xorps (v4sf, v4sf)
8563 v4sf __builtin_ia32_movss (v4sf, v4sf)
8564 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
8565 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
8566 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
8567 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
8568 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
8569 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
8570 v2si __builtin_ia32_cvtps2pi (v4sf)
8571 int __builtin_ia32_cvtss2si (v4sf)
8572 v2si __builtin_ia32_cvttps2pi (v4sf)
8573 int __builtin_ia32_cvttss2si (v4sf)
8574 v4sf __builtin_ia32_rcpps (v4sf)
8575 v4sf __builtin_ia32_rsqrtps (v4sf)
8576 v4sf __builtin_ia32_sqrtps (v4sf)
8577 v4sf __builtin_ia32_rcpss (v4sf)
8578 v4sf __builtin_ia32_rsqrtss (v4sf)
8579 v4sf __builtin_ia32_sqrtss (v4sf)
8580 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
8581 void __builtin_ia32_movntps (float *, v4sf)
8582 int __builtin_ia32_movmskps (v4sf)
8585 The following built-in functions are available when @option{-msse} is used.
8588 @item v4sf __builtin_ia32_loadaps (float *)
8589 Generates the @code{movaps} machine instruction as a load from memory.
8590 @item void __builtin_ia32_storeaps (float *, v4sf)
8591 Generates the @code{movaps} machine instruction as a store to memory.
8592 @item v4sf __builtin_ia32_loadups (float *)
8593 Generates the @code{movups} machine instruction as a load from memory.
8594 @item void __builtin_ia32_storeups (float *, v4sf)
8595 Generates the @code{movups} machine instruction as a store to memory.
8596 @item v4sf __builtin_ia32_loadsss (float *)
8597 Generates the @code{movss} machine instruction as a load from memory.
8598 @item void __builtin_ia32_storess (float *, v4sf)
8599 Generates the @code{movss} machine instruction as a store to memory.
8600 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
8601 Generates the @code{movhps} machine instruction as a load from memory.
8602 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
8603 Generates the @code{movlps} machine instruction as a load from memory
8604 @item void __builtin_ia32_storehps (v2sf *, v4sf)
8605 Generates the @code{movhps} machine instruction as a store to memory.
8606 @item void __builtin_ia32_storelps (v2sf *, v4sf)
8607 Generates the @code{movlps} machine instruction as a store to memory.
8610 The following built-in functions are available when @option{-msse2} is used.
8611 All of them generate the machine instruction that is part of the name.
8614 int __builtin_ia32_comisdeq (v2df, v2df)
8615 int __builtin_ia32_comisdlt (v2df, v2df)
8616 int __builtin_ia32_comisdle (v2df, v2df)
8617 int __builtin_ia32_comisdgt (v2df, v2df)
8618 int __builtin_ia32_comisdge (v2df, v2df)
8619 int __builtin_ia32_comisdneq (v2df, v2df)
8620 int __builtin_ia32_ucomisdeq (v2df, v2df)
8621 int __builtin_ia32_ucomisdlt (v2df, v2df)
8622 int __builtin_ia32_ucomisdle (v2df, v2df)
8623 int __builtin_ia32_ucomisdgt (v2df, v2df)
8624 int __builtin_ia32_ucomisdge (v2df, v2df)
8625 int __builtin_ia32_ucomisdneq (v2df, v2df)
8626 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
8627 v2df __builtin_ia32_cmpltpd (v2df, v2df)
8628 v2df __builtin_ia32_cmplepd (v2df, v2df)
8629 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
8630 v2df __builtin_ia32_cmpgepd (v2df, v2df)
8631 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
8632 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
8633 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
8634 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
8635 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
8636 v2df __builtin_ia32_cmpngepd (v2df, v2df)
8637 v2df __builtin_ia32_cmpordpd (v2df, v2df)
8638 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
8639 v2df __builtin_ia32_cmpltsd (v2df, v2df)
8640 v2df __builtin_ia32_cmplesd (v2df, v2df)
8641 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
8642 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
8643 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
8644 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
8645 v2df __builtin_ia32_cmpordsd (v2df, v2df)
8646 v2di __builtin_ia32_paddq (v2di, v2di)
8647 v2di __builtin_ia32_psubq (v2di, v2di)
8648 v2df __builtin_ia32_addpd (v2df, v2df)
8649 v2df __builtin_ia32_subpd (v2df, v2df)
8650 v2df __builtin_ia32_mulpd (v2df, v2df)
8651 v2df __builtin_ia32_divpd (v2df, v2df)
8652 v2df __builtin_ia32_addsd (v2df, v2df)
8653 v2df __builtin_ia32_subsd (v2df, v2df)
8654 v2df __builtin_ia32_mulsd (v2df, v2df)
8655 v2df __builtin_ia32_divsd (v2df, v2df)
8656 v2df __builtin_ia32_minpd (v2df, v2df)
8657 v2df __builtin_ia32_maxpd (v2df, v2df)
8658 v2df __builtin_ia32_minsd (v2df, v2df)
8659 v2df __builtin_ia32_maxsd (v2df, v2df)
8660 v2df __builtin_ia32_andpd (v2df, v2df)
8661 v2df __builtin_ia32_andnpd (v2df, v2df)
8662 v2df __builtin_ia32_orpd (v2df, v2df)
8663 v2df __builtin_ia32_xorpd (v2df, v2df)
8664 v2df __builtin_ia32_movsd (v2df, v2df)
8665 v2df __builtin_ia32_unpckhpd (v2df, v2df)
8666 v2df __builtin_ia32_unpcklpd (v2df, v2df)
8667 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
8668 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
8669 v4si __builtin_ia32_paddd128 (v4si, v4si)
8670 v2di __builtin_ia32_paddq128 (v2di, v2di)
8671 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
8672 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
8673 v4si __builtin_ia32_psubd128 (v4si, v4si)
8674 v2di __builtin_ia32_psubq128 (v2di, v2di)
8675 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
8676 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
8677 v2di __builtin_ia32_pand128 (v2di, v2di)
8678 v2di __builtin_ia32_pandn128 (v2di, v2di)
8679 v2di __builtin_ia32_por128 (v2di, v2di)
8680 v2di __builtin_ia32_pxor128 (v2di, v2di)
8681 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
8682 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
8683 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
8684 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
8685 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
8686 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
8687 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
8688 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
8689 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
8690 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
8691 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
8692 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
8693 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
8694 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
8695 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
8696 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
8697 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
8698 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
8699 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
8700 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
8701 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
8702 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
8703 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
8704 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
8705 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
8706 v2df __builtin_ia32_loadupd (double *)
8707 void __builtin_ia32_storeupd (double *, v2df)
8708 v2df __builtin_ia32_loadhpd (v2df, double const *)
8709 v2df __builtin_ia32_loadlpd (v2df, double const *)
8710 int __builtin_ia32_movmskpd (v2df)
8711 int __builtin_ia32_pmovmskb128 (v16qi)
8712 void __builtin_ia32_movnti (int *, int)
8713 void __builtin_ia32_movntpd (double *, v2df)
8714 void __builtin_ia32_movntdq (v2df *, v2df)
8715 v4si __builtin_ia32_pshufd (v4si, int)
8716 v8hi __builtin_ia32_pshuflw (v8hi, int)
8717 v8hi __builtin_ia32_pshufhw (v8hi, int)
8718 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
8719 v2df __builtin_ia32_sqrtpd (v2df)
8720 v2df __builtin_ia32_sqrtsd (v2df)
8721 v2df __builtin_ia32_shufpd (v2df, v2df, int)
8722 v2df __builtin_ia32_cvtdq2pd (v4si)
8723 v4sf __builtin_ia32_cvtdq2ps (v4si)
8724 v4si __builtin_ia32_cvtpd2dq (v2df)
8725 v2si __builtin_ia32_cvtpd2pi (v2df)
8726 v4sf __builtin_ia32_cvtpd2ps (v2df)
8727 v4si __builtin_ia32_cvttpd2dq (v2df)
8728 v2si __builtin_ia32_cvttpd2pi (v2df)
8729 v2df __builtin_ia32_cvtpi2pd (v2si)
8730 int __builtin_ia32_cvtsd2si (v2df)
8731 int __builtin_ia32_cvttsd2si (v2df)
8732 long long __builtin_ia32_cvtsd2si64 (v2df)
8733 long long __builtin_ia32_cvttsd2si64 (v2df)
8734 v4si __builtin_ia32_cvtps2dq (v4sf)
8735 v2df __builtin_ia32_cvtps2pd (v4sf)
8736 v4si __builtin_ia32_cvttps2dq (v4sf)
8737 v2df __builtin_ia32_cvtsi2sd (v2df, int)
8738 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
8739 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
8740 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
8741 void __builtin_ia32_clflush (const void *)
8742 void __builtin_ia32_lfence (void)
8743 void __builtin_ia32_mfence (void)
8744 v16qi __builtin_ia32_loaddqu (const char *)
8745 void __builtin_ia32_storedqu (char *, v16qi)
8746 v1di __builtin_ia32_pmuludq (v2si, v2si)
8747 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
8748 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
8749 v4si __builtin_ia32_pslld128 (v4si, v4si)
8750 v2di __builtin_ia32_psllq128 (v2di, v2di)
8751 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
8752 v4si __builtin_ia32_psrld128 (v4si, v4si)
8753 v2di __builtin_ia32_psrlq128 (v2di, v2di)
8754 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
8755 v4si __builtin_ia32_psrad128 (v4si, v4si)
8756 v2di __builtin_ia32_pslldqi128 (v2di, int)
8757 v8hi __builtin_ia32_psllwi128 (v8hi, int)
8758 v4si __builtin_ia32_pslldi128 (v4si, int)
8759 v2di __builtin_ia32_psllqi128 (v2di, int)
8760 v2di __builtin_ia32_psrldqi128 (v2di, int)
8761 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
8762 v4si __builtin_ia32_psrldi128 (v4si, int)
8763 v2di __builtin_ia32_psrlqi128 (v2di, int)
8764 v8hi __builtin_ia32_psrawi128 (v8hi, int)
8765 v4si __builtin_ia32_psradi128 (v4si, int)
8766 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
8767 v2di __builtin_ia32_movq128 (v2di)
8770 The following built-in functions are available when @option{-msse3} is used.
8771 All of them generate the machine instruction that is part of the name.
8774 v2df __builtin_ia32_addsubpd (v2df, v2df)
8775 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
8776 v2df __builtin_ia32_haddpd (v2df, v2df)
8777 v4sf __builtin_ia32_haddps (v4sf, v4sf)
8778 v2df __builtin_ia32_hsubpd (v2df, v2df)
8779 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
8780 v16qi __builtin_ia32_lddqu (char const *)
8781 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
8782 v2df __builtin_ia32_movddup (v2df)
8783 v4sf __builtin_ia32_movshdup (v4sf)
8784 v4sf __builtin_ia32_movsldup (v4sf)
8785 void __builtin_ia32_mwait (unsigned int, unsigned int)
8788 The following built-in functions are available when @option{-msse3} is used.
8791 @item v2df __builtin_ia32_loadddup (double const *)
8792 Generates the @code{movddup} machine instruction as a load from memory.
8795 The following built-in functions are available when @option{-mssse3} is used.
8796 All of them generate the machine instruction that is part of the name
8800 v2si __builtin_ia32_phaddd (v2si, v2si)
8801 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
8802 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
8803 v2si __builtin_ia32_phsubd (v2si, v2si)
8804 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
8805 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
8806 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
8807 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
8808 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
8809 v8qi __builtin_ia32_psignb (v8qi, v8qi)
8810 v2si __builtin_ia32_psignd (v2si, v2si)
8811 v4hi __builtin_ia32_psignw (v4hi, v4hi)
8812 v1di __builtin_ia32_palignr (v1di, v1di, int)
8813 v8qi __builtin_ia32_pabsb (v8qi)
8814 v2si __builtin_ia32_pabsd (v2si)
8815 v4hi __builtin_ia32_pabsw (v4hi)
8818 The following built-in functions are available when @option{-mssse3} is used.
8819 All of them generate the machine instruction that is part of the name
8823 v4si __builtin_ia32_phaddd128 (v4si, v4si)
8824 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
8825 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
8826 v4si __builtin_ia32_phsubd128 (v4si, v4si)
8827 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
8828 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
8829 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
8830 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
8831 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
8832 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
8833 v4si __builtin_ia32_psignd128 (v4si, v4si)
8834 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
8835 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
8836 v16qi __builtin_ia32_pabsb128 (v16qi)
8837 v4si __builtin_ia32_pabsd128 (v4si)
8838 v8hi __builtin_ia32_pabsw128 (v8hi)
8841 The following built-in functions are available when @option{-msse4.1} is
8842 used. All of them generate the machine instruction that is part of the
8846 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
8847 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
8848 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
8849 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
8850 v2df __builtin_ia32_dppd (v2df, v2df, const int)
8851 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
8852 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
8853 v2di __builtin_ia32_movntdqa (v2di *);
8854 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
8855 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
8856 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
8857 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
8858 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
8859 v8hi __builtin_ia32_phminposuw128 (v8hi)
8860 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
8861 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
8862 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
8863 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
8864 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
8865 v4si __builtin_ia32_pminsd128 (v4si, v4si)
8866 v4si __builtin_ia32_pminud128 (v4si, v4si)
8867 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
8868 v4si __builtin_ia32_pmovsxbd128 (v16qi)
8869 v2di __builtin_ia32_pmovsxbq128 (v16qi)
8870 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
8871 v2di __builtin_ia32_pmovsxdq128 (v4si)
8872 v4si __builtin_ia32_pmovsxwd128 (v8hi)
8873 v2di __builtin_ia32_pmovsxwq128 (v8hi)
8874 v4si __builtin_ia32_pmovzxbd128 (v16qi)
8875 v2di __builtin_ia32_pmovzxbq128 (v16qi)
8876 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
8877 v2di __builtin_ia32_pmovzxdq128 (v4si)
8878 v4si __builtin_ia32_pmovzxwd128 (v8hi)
8879 v2di __builtin_ia32_pmovzxwq128 (v8hi)
8880 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
8881 v4si __builtin_ia32_pmulld128 (v4si, v4si)
8882 int __builtin_ia32_ptestc128 (v2di, v2di)
8883 int __builtin_ia32_ptestnzc128 (v2di, v2di)
8884 int __builtin_ia32_ptestz128 (v2di, v2di)
8885 v2df __builtin_ia32_roundpd (v2df, const int)
8886 v4sf __builtin_ia32_roundps (v4sf, const int)
8887 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
8888 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
8891 The following built-in functions are available when @option{-msse4.1} is
8895 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
8896 Generates the @code{insertps} machine instruction.
8897 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
8898 Generates the @code{pextrb} machine instruction.
8899 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
8900 Generates the @code{pinsrb} machine instruction.
8901 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
8902 Generates the @code{pinsrd} machine instruction.
8903 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
8904 Generates the @code{pinsrq} machine instruction in 64bit mode.
8907 The following built-in functions are changed to generate new SSE4.1
8908 instructions when @option{-msse4.1} is used.
8911 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
8912 Generates the @code{extractps} machine instruction.
8913 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
8914 Generates the @code{pextrd} machine instruction.
8915 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
8916 Generates the @code{pextrq} machine instruction in 64bit mode.
8919 The following built-in functions are available when @option{-msse4.2} is
8920 used. All of them generate the machine instruction that is part of the
8924 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
8925 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
8926 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
8927 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
8928 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
8929 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
8930 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
8931 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
8932 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
8933 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
8934 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
8935 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
8936 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
8937 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
8938 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
8941 The following built-in functions are available when @option{-msse4.2} is
8945 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
8946 Generates the @code{crc32b} machine instruction.
8947 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
8948 Generates the @code{crc32w} machine instruction.
8949 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
8950 Generates the @code{crc32l} machine instruction.
8951 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
8952 Generates the @code{crc32q} machine instruction.
8955 The following built-in functions are changed to generate new SSE4.2
8956 instructions when @option{-msse4.2} is used.
8959 @item int __builtin_popcount (unsigned int)
8960 Generates the @code{popcntl} machine instruction.
8961 @item int __builtin_popcountl (unsigned long)
8962 Generates the @code{popcntl} or @code{popcntq} machine instruction,
8963 depending on the size of @code{unsigned long}.
8964 @item int __builtin_popcountll (unsigned long long)
8965 Generates the @code{popcntq} machine instruction.
8968 The following built-in functions are available when @option{-mavx} is
8969 used. All of them generate the machine instruction that is part of the
8973 v4df __builtin_ia32_addpd256 (v4df,v4df)
8974 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
8975 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
8976 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
8977 v4df __builtin_ia32_andnpd256 (v4df,v4df)
8978 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
8979 v4df __builtin_ia32_andpd256 (v4df,v4df)
8980 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
8981 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
8982 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
8983 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
8984 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
8985 v2df __builtin_ia32_cmppd (v2df,v2df,int)
8986 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
8987 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
8988 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
8989 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
8990 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
8991 v4df __builtin_ia32_cvtdq2pd256 (v4si)
8992 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
8993 v4si __builtin_ia32_cvtpd2dq256 (v4df)
8994 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
8995 v8si __builtin_ia32_cvtps2dq256 (v8sf)
8996 v4df __builtin_ia32_cvtps2pd256 (v4sf)
8997 v4si __builtin_ia32_cvttpd2dq256 (v4df)
8998 v8si __builtin_ia32_cvttps2dq256 (v8sf)
8999 v4df __builtin_ia32_divpd256 (v4df,v4df)
9000 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
9001 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
9002 v4df __builtin_ia32_haddpd256 (v4df,v4df)
9003 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
9004 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
9005 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
9006 v32qi __builtin_ia32_lddqu256 (pcchar)
9007 v32qi __builtin_ia32_loaddqu256 (pcchar)
9008 v4df __builtin_ia32_loadupd256 (pcdouble)
9009 v8sf __builtin_ia32_loadups256 (pcfloat)
9010 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
9011 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
9012 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
9013 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
9014 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
9015 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
9016 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
9017 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
9018 v4df __builtin_ia32_maxpd256 (v4df,v4df)
9019 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
9020 v4df __builtin_ia32_minpd256 (v4df,v4df)
9021 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
9022 v4df __builtin_ia32_movddup256 (v4df)
9023 int __builtin_ia32_movmskpd256 (v4df)
9024 int __builtin_ia32_movmskps256 (v8sf)
9025 v8sf __builtin_ia32_movshdup256 (v8sf)
9026 v8sf __builtin_ia32_movsldup256 (v8sf)
9027 v4df __builtin_ia32_mulpd256 (v4df,v4df)
9028 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
9029 v4df __builtin_ia32_orpd256 (v4df,v4df)
9030 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
9031 v2df __builtin_ia32_pd_pd256 (v4df)
9032 v4df __builtin_ia32_pd256_pd (v2df)
9033 v4sf __builtin_ia32_ps_ps256 (v8sf)
9034 v8sf __builtin_ia32_ps256_ps (v4sf)
9035 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
9036 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
9037 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
9038 v8sf __builtin_ia32_rcpps256 (v8sf)
9039 v4df __builtin_ia32_roundpd256 (v4df,int)
9040 v8sf __builtin_ia32_roundps256 (v8sf,int)
9041 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
9042 v8sf __builtin_ia32_rsqrtps256 (v8sf)
9043 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
9044 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
9045 v4si __builtin_ia32_si_si256 (v8si)
9046 v8si __builtin_ia32_si256_si (v4si)
9047 v4df __builtin_ia32_sqrtpd256 (v4df)
9048 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
9049 v8sf __builtin_ia32_sqrtps256 (v8sf)
9050 void __builtin_ia32_storedqu256 (pchar,v32qi)
9051 void __builtin_ia32_storeupd256 (pdouble,v4df)
9052 void __builtin_ia32_storeups256 (pfloat,v8sf)
9053 v4df __builtin_ia32_subpd256 (v4df,v4df)
9054 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
9055 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
9056 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
9057 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
9058 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
9059 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
9060 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
9061 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
9062 v4sf __builtin_ia32_vbroadcastss (pcfloat)
9063 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
9064 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
9065 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
9066 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
9067 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
9068 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
9069 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
9070 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
9071 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
9072 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
9073 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
9074 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
9075 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
9076 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
9077 v2df __builtin_ia32_vpermilpd (v2df,int)
9078 v4df __builtin_ia32_vpermilpd256 (v4df,int)
9079 v4sf __builtin_ia32_vpermilps (v4sf,int)
9080 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
9081 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
9082 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
9083 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
9084 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
9085 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
9086 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
9087 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
9088 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
9089 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
9090 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
9091 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
9092 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
9093 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
9094 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
9095 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
9096 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
9097 void __builtin_ia32_vzeroall (void)
9098 void __builtin_ia32_vzeroupper (void)
9099 v4df __builtin_ia32_xorpd256 (v4df,v4df)
9100 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
9103 The following built-in functions are available when @option{-maes} is
9104 used. All of them generate the machine instruction that is part of the
9108 v2di __builtin_ia32_aesenc128 (v2di, v2di)
9109 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
9110 v2di __builtin_ia32_aesdec128 (v2di, v2di)
9111 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
9112 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
9113 v2di __builtin_ia32_aesimc128 (v2di)
9116 The following built-in function is available when @option{-mpclmul} is
9120 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
9121 Generates the @code{pclmulqdq} machine instruction.
9124 The following built-in function is available when @option{-mfsgsbase} is
9125 used. All of them generate the machine instruction that is part of the
9129 unsigned int __builtin_ia32_rdfsbase32 (void)
9130 unsigned long long __builtin_ia32_rdfsbase64 (void)
9131 unsigned int __builtin_ia32_rdgsbase32 (void)
9132 unsigned long long __builtin_ia32_rdgsbase64 (void)
9133 void _writefsbase_u32 (unsigned int)
9134 void _writefsbase_u64 (unsigned long long)
9135 void _writegsbase_u32 (unsigned int)
9136 void _writegsbase_u64 (unsigned long long)
9139 The following built-in function is available when @option{-mrdrnd} is
9140 used. All of them generate the machine instruction that is part of the
9144 unsigned short __builtin_ia32_rdrand16 (void)
9145 unsigned int __builtin_ia32_rdrand32 (void)
9146 unsigned long long __builtin_ia32_rdrand64 (void)
9149 The following built-in functions are available when @option{-msse4a} is used.
9150 All of them generate the machine instruction that is part of the name.
9153 void __builtin_ia32_movntsd (double *, v2df)
9154 void __builtin_ia32_movntss (float *, v4sf)
9155 v2di __builtin_ia32_extrq (v2di, v16qi)
9156 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
9157 v2di __builtin_ia32_insertq (v2di, v2di)
9158 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
9161 The following built-in functions are available when @option{-mxop} is used.
9163 v2df __builtin_ia32_vfrczpd (v2df)
9164 v4sf __builtin_ia32_vfrczps (v4sf)
9165 v2df __builtin_ia32_vfrczsd (v2df, v2df)
9166 v4sf __builtin_ia32_vfrczss (v4sf, v4sf)
9167 v4df __builtin_ia32_vfrczpd256 (v4df)
9168 v8sf __builtin_ia32_vfrczps256 (v8sf)
9169 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
9170 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
9171 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
9172 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
9173 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
9174 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
9175 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
9176 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
9177 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
9178 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
9179 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
9180 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
9181 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
9182 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
9183 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
9184 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
9185 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
9186 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
9187 v4si __builtin_ia32_vpcomequd (v4si, v4si)
9188 v2di __builtin_ia32_vpcomequq (v2di, v2di)
9189 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
9190 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
9191 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
9192 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
9193 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
9194 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
9195 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
9196 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
9197 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
9198 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
9199 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
9200 v4si __builtin_ia32_vpcomged (v4si, v4si)
9201 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
9202 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
9203 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
9204 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
9205 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
9206 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
9207 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
9208 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
9209 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
9210 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
9211 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
9212 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
9213 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
9214 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
9215 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
9216 v4si __builtin_ia32_vpcomled (v4si, v4si)
9217 v2di __builtin_ia32_vpcomleq (v2di, v2di)
9218 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
9219 v4si __builtin_ia32_vpcomleud (v4si, v4si)
9220 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
9221 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
9222 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
9223 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
9224 v4si __builtin_ia32_vpcomltd (v4si, v4si)
9225 v2di __builtin_ia32_vpcomltq (v2di, v2di)
9226 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
9227 v4si __builtin_ia32_vpcomltud (v4si, v4si)
9228 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
9229 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
9230 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
9231 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
9232 v4si __builtin_ia32_vpcomned (v4si, v4si)
9233 v2di __builtin_ia32_vpcomneq (v2di, v2di)
9234 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
9235 v4si __builtin_ia32_vpcomneud (v4si, v4si)
9236 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
9237 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
9238 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
9239 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
9240 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
9241 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
9242 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
9243 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
9244 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
9245 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
9246 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
9247 v4si __builtin_ia32_vphaddbd (v16qi)
9248 v2di __builtin_ia32_vphaddbq (v16qi)
9249 v8hi __builtin_ia32_vphaddbw (v16qi)
9250 v2di __builtin_ia32_vphadddq (v4si)
9251 v4si __builtin_ia32_vphaddubd (v16qi)
9252 v2di __builtin_ia32_vphaddubq (v16qi)
9253 v8hi __builtin_ia32_vphaddubw (v16qi)
9254 v2di __builtin_ia32_vphaddudq (v4si)
9255 v4si __builtin_ia32_vphadduwd (v8hi)
9256 v2di __builtin_ia32_vphadduwq (v8hi)
9257 v4si __builtin_ia32_vphaddwd (v8hi)
9258 v2di __builtin_ia32_vphaddwq (v8hi)
9259 v8hi __builtin_ia32_vphsubbw (v16qi)
9260 v2di __builtin_ia32_vphsubdq (v4si)
9261 v4si __builtin_ia32_vphsubwd (v8hi)
9262 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
9263 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
9264 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
9265 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
9266 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
9267 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
9268 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
9269 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
9270 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
9271 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
9272 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
9273 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
9274 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
9275 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
9276 v4si __builtin_ia32_vprotd (v4si, v4si)
9277 v2di __builtin_ia32_vprotq (v2di, v2di)
9278 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
9279 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
9280 v4si __builtin_ia32_vpshad (v4si, v4si)
9281 v2di __builtin_ia32_vpshaq (v2di, v2di)
9282 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
9283 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
9284 v4si __builtin_ia32_vpshld (v4si, v4si)
9285 v2di __builtin_ia32_vpshlq (v2di, v2di)
9286 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
9289 The following built-in functions are available when @option{-mfma4} is used.
9290 All of them generate the machine instruction that is part of the name
9294 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
9295 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
9296 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
9297 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
9298 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
9299 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
9300 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
9301 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
9302 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
9303 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
9304 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
9305 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
9306 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
9307 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
9308 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
9309 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
9310 v2df __builtin_ia32_fmaddsubpd (v2df, v2df, v2df)
9311 v4sf __builtin_ia32_fmaddsubps (v4sf, v4sf, v4sf)
9312 v2df __builtin_ia32_fmsubaddpd (v2df, v2df, v2df)
9313 v4sf __builtin_ia32_fmsubaddps (v4sf, v4sf, v4sf)
9314 v4df __builtin_ia32_fmaddpd256 (v4df, v4df, v4df)
9315 v8sf __builtin_ia32_fmaddps256 (v8sf, v8sf, v8sf)
9316 v4df __builtin_ia32_fmsubpd256 (v4df, v4df, v4df)
9317 v8sf __builtin_ia32_fmsubps256 (v8sf, v8sf, v8sf)
9318 v4df __builtin_ia32_fnmaddpd256 (v4df, v4df, v4df)
9319 v8sf __builtin_ia32_fnmaddps256 (v8sf, v8sf, v8sf)
9320 v4df __builtin_ia32_fnmsubpd256 (v4df, v4df, v4df)
9321 v8sf __builtin_ia32_fnmsubps256 (v8sf, v8sf, v8sf)
9322 v4df __builtin_ia32_fmaddsubpd256 (v4df, v4df, v4df)
9323 v8sf __builtin_ia32_fmaddsubps256 (v8sf, v8sf, v8sf)
9324 v4df __builtin_ia32_fmsubaddpd256 (v4df, v4df, v4df)
9325 v8sf __builtin_ia32_fmsubaddps256 (v8sf, v8sf, v8sf)
9329 The following built-in functions are available when @option{-mlwp} is used.
9332 void __builtin_ia32_llwpcb16 (void *);
9333 void __builtin_ia32_llwpcb32 (void *);
9334 void __builtin_ia32_llwpcb64 (void *);
9335 void * __builtin_ia32_llwpcb16 (void);
9336 void * __builtin_ia32_llwpcb32 (void);
9337 void * __builtin_ia32_llwpcb64 (void);
9338 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
9339 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
9340 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
9341 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
9342 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
9343 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
9346 The following built-in functions are available when @option{-m3dnow} is used.
9347 All of them generate the machine instruction that is part of the name.
9350 void __builtin_ia32_femms (void)
9351 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
9352 v2si __builtin_ia32_pf2id (v2sf)
9353 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
9354 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
9355 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
9356 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
9357 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
9358 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
9359 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
9360 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
9361 v2sf __builtin_ia32_pfrcp (v2sf)
9362 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
9363 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
9364 v2sf __builtin_ia32_pfrsqrt (v2sf)
9365 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
9366 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
9367 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
9368 v2sf __builtin_ia32_pi2fd (v2si)
9369 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
9372 The following built-in functions are available when both @option{-m3dnow}
9373 and @option{-march=athlon} are used. All of them generate the machine
9374 instruction that is part of the name.
9377 v2si __builtin_ia32_pf2iw (v2sf)
9378 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
9379 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
9380 v2sf __builtin_ia32_pi2fw (v2si)
9381 v2sf __builtin_ia32_pswapdsf (v2sf)
9382 v2si __builtin_ia32_pswapdsi (v2si)
9385 @node MIPS DSP Built-in Functions
9386 @subsection MIPS DSP Built-in Functions
9388 The MIPS DSP Application-Specific Extension (ASE) includes new
9389 instructions that are designed to improve the performance of DSP and
9390 media applications. It provides instructions that operate on packed
9391 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
9393 GCC supports MIPS DSP operations using both the generic
9394 vector extensions (@pxref{Vector Extensions}) and a collection of
9395 MIPS-specific built-in functions. Both kinds of support are
9396 enabled by the @option{-mdsp} command-line option.
9398 Revision 2 of the ASE was introduced in the second half of 2006.
9399 This revision adds extra instructions to the original ASE, but is
9400 otherwise backwards-compatible with it. You can select revision 2
9401 using the command-line option @option{-mdspr2}; this option implies
9404 The SCOUNT and POS bits of the DSP control register are global. The
9405 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
9406 POS bits. During optimization, the compiler will not delete these
9407 instructions and it will not delete calls to functions containing
9410 At present, GCC only provides support for operations on 32-bit
9411 vectors. The vector type associated with 8-bit integer data is
9412 usually called @code{v4i8}, the vector type associated with Q7
9413 is usually called @code{v4q7}, the vector type associated with 16-bit
9414 integer data is usually called @code{v2i16}, and the vector type
9415 associated with Q15 is usually called @code{v2q15}. They can be
9416 defined in C as follows:
9419 typedef signed char v4i8 __attribute__ ((vector_size(4)));
9420 typedef signed char v4q7 __attribute__ ((vector_size(4)));
9421 typedef short v2i16 __attribute__ ((vector_size(4)));
9422 typedef short v2q15 __attribute__ ((vector_size(4)));
9425 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
9426 initialized in the same way as aggregates. For example:
9429 v4i8 a = @{1, 2, 3, 4@};
9431 b = (v4i8) @{5, 6, 7, 8@};
9433 v2q15 c = @{0x0fcb, 0x3a75@};
9435 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
9438 @emph{Note:} The CPU's endianness determines the order in which values
9439 are packed. On little-endian targets, the first value is the least
9440 significant and the last value is the most significant. The opposite
9441 order applies to big-endian targets. For example, the code above will
9442 set the lowest byte of @code{a} to @code{1} on little-endian targets
9443 and @code{4} on big-endian targets.
9445 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
9446 representation. As shown in this example, the integer representation
9447 of a Q7 value can be obtained by multiplying the fractional value by
9448 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
9449 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
9452 The table below lists the @code{v4i8} and @code{v2q15} operations for which
9453 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
9454 and @code{c} and @code{d} are @code{v2q15} values.
9456 @multitable @columnfractions .50 .50
9457 @item C code @tab MIPS instruction
9458 @item @code{a + b} @tab @code{addu.qb}
9459 @item @code{c + d} @tab @code{addq.ph}
9460 @item @code{a - b} @tab @code{subu.qb}
9461 @item @code{c - d} @tab @code{subq.ph}
9464 The table below lists the @code{v2i16} operation for which
9465 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
9466 @code{v2i16} values.
9468 @multitable @columnfractions .50 .50
9469 @item C code @tab MIPS instruction
9470 @item @code{e * f} @tab @code{mul.ph}
9473 It is easier to describe the DSP built-in functions if we first define
9474 the following types:
9479 typedef unsigned int ui32;
9480 typedef long long a64;
9483 @code{q31} and @code{i32} are actually the same as @code{int}, but we
9484 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
9485 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
9486 @code{long long}, but we use @code{a64} to indicate values that will
9487 be placed in one of the four DSP accumulators (@code{$ac0},
9488 @code{$ac1}, @code{$ac2} or @code{$ac3}).
9490 Also, some built-in functions prefer or require immediate numbers as
9491 parameters, because the corresponding DSP instructions accept both immediate
9492 numbers and register operands, or accept immediate numbers only. The
9493 immediate parameters are listed as follows.
9502 imm_n32_31: -32 to 31.
9503 imm_n512_511: -512 to 511.
9506 The following built-in functions map directly to a particular MIPS DSP
9507 instruction. Please refer to the architecture specification
9508 for details on what each instruction does.
9511 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
9512 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
9513 q31 __builtin_mips_addq_s_w (q31, q31)
9514 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
9515 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
9516 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
9517 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
9518 q31 __builtin_mips_subq_s_w (q31, q31)
9519 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
9520 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
9521 i32 __builtin_mips_addsc (i32, i32)
9522 i32 __builtin_mips_addwc (i32, i32)
9523 i32 __builtin_mips_modsub (i32, i32)
9524 i32 __builtin_mips_raddu_w_qb (v4i8)
9525 v2q15 __builtin_mips_absq_s_ph (v2q15)
9526 q31 __builtin_mips_absq_s_w (q31)
9527 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
9528 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
9529 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
9530 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
9531 q31 __builtin_mips_preceq_w_phl (v2q15)
9532 q31 __builtin_mips_preceq_w_phr (v2q15)
9533 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
9534 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
9535 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
9536 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
9537 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
9538 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
9539 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
9540 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
9541 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
9542 v4i8 __builtin_mips_shll_qb (v4i8, i32)
9543 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
9544 v2q15 __builtin_mips_shll_ph (v2q15, i32)
9545 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
9546 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
9547 q31 __builtin_mips_shll_s_w (q31, imm0_31)
9548 q31 __builtin_mips_shll_s_w (q31, i32)
9549 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
9550 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
9551 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
9552 v2q15 __builtin_mips_shra_ph (v2q15, i32)
9553 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
9554 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
9555 q31 __builtin_mips_shra_r_w (q31, imm0_31)
9556 q31 __builtin_mips_shra_r_w (q31, i32)
9557 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
9558 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
9559 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
9560 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
9561 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
9562 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
9563 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
9564 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
9565 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
9566 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
9567 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
9568 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
9569 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
9570 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
9571 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
9572 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
9573 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
9574 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
9575 i32 __builtin_mips_bitrev (i32)
9576 i32 __builtin_mips_insv (i32, i32)
9577 v4i8 __builtin_mips_repl_qb (imm0_255)
9578 v4i8 __builtin_mips_repl_qb (i32)
9579 v2q15 __builtin_mips_repl_ph (imm_n512_511)
9580 v2q15 __builtin_mips_repl_ph (i32)
9581 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
9582 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
9583 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
9584 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
9585 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
9586 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
9587 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
9588 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
9589 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
9590 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
9591 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
9592 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
9593 i32 __builtin_mips_extr_w (a64, imm0_31)
9594 i32 __builtin_mips_extr_w (a64, i32)
9595 i32 __builtin_mips_extr_r_w (a64, imm0_31)
9596 i32 __builtin_mips_extr_s_h (a64, i32)
9597 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
9598 i32 __builtin_mips_extr_rs_w (a64, i32)
9599 i32 __builtin_mips_extr_s_h (a64, imm0_31)
9600 i32 __builtin_mips_extr_r_w (a64, i32)
9601 i32 __builtin_mips_extp (a64, imm0_31)
9602 i32 __builtin_mips_extp (a64, i32)
9603 i32 __builtin_mips_extpdp (a64, imm0_31)
9604 i32 __builtin_mips_extpdp (a64, i32)
9605 a64 __builtin_mips_shilo (a64, imm_n32_31)
9606 a64 __builtin_mips_shilo (a64, i32)
9607 a64 __builtin_mips_mthlip (a64, i32)
9608 void __builtin_mips_wrdsp (i32, imm0_63)
9609 i32 __builtin_mips_rddsp (imm0_63)
9610 i32 __builtin_mips_lbux (void *, i32)
9611 i32 __builtin_mips_lhx (void *, i32)
9612 i32 __builtin_mips_lwx (void *, i32)
9613 i32 __builtin_mips_bposge32 (void)
9616 The following built-in functions map directly to a particular MIPS DSP REV 2
9617 instruction. Please refer to the architecture specification
9618 for details on what each instruction does.
9621 v4q7 __builtin_mips_absq_s_qb (v4q7);
9622 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
9623 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
9624 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
9625 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
9626 i32 __builtin_mips_append (i32, i32, imm0_31);
9627 i32 __builtin_mips_balign (i32, i32, imm0_3);
9628 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
9629 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
9630 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
9631 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
9632 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
9633 a64 __builtin_mips_madd (a64, i32, i32);
9634 a64 __builtin_mips_maddu (a64, ui32, ui32);
9635 a64 __builtin_mips_msub (a64, i32, i32);
9636 a64 __builtin_mips_msubu (a64, ui32, ui32);
9637 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
9638 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
9639 q31 __builtin_mips_mulq_rs_w (q31, q31);
9640 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
9641 q31 __builtin_mips_mulq_s_w (q31, q31);
9642 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
9643 a64 __builtin_mips_mult (i32, i32);
9644 a64 __builtin_mips_multu (ui32, ui32);
9645 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
9646 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
9647 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
9648 i32 __builtin_mips_prepend (i32, i32, imm0_31);
9649 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
9650 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
9651 v4i8 __builtin_mips_shra_qb (v4i8, i32);
9652 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
9653 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
9654 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
9655 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
9656 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
9657 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
9658 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
9659 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
9660 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
9661 q31 __builtin_mips_addqh_w (q31, q31);
9662 q31 __builtin_mips_addqh_r_w (q31, q31);
9663 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
9664 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
9665 q31 __builtin_mips_subqh_w (q31, q31);
9666 q31 __builtin_mips_subqh_r_w (q31, q31);
9667 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
9668 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
9669 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
9670 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
9671 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
9672 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
9676 @node MIPS Paired-Single Support
9677 @subsection MIPS Paired-Single Support
9679 The MIPS64 architecture includes a number of instructions that
9680 operate on pairs of single-precision floating-point values.
9681 Each pair is packed into a 64-bit floating-point register,
9682 with one element being designated the ``upper half'' and
9683 the other being designated the ``lower half''.
9685 GCC supports paired-single operations using both the generic
9686 vector extensions (@pxref{Vector Extensions}) and a collection of
9687 MIPS-specific built-in functions. Both kinds of support are
9688 enabled by the @option{-mpaired-single} command-line option.
9690 The vector type associated with paired-single values is usually
9691 called @code{v2sf}. It can be defined in C as follows:
9694 typedef float v2sf __attribute__ ((vector_size (8)));
9697 @code{v2sf} values are initialized in the same way as aggregates.
9701 v2sf a = @{1.5, 9.1@};
9704 b = (v2sf) @{e, f@};
9707 @emph{Note:} The CPU's endianness determines which value is stored in
9708 the upper half of a register and which value is stored in the lower half.
9709 On little-endian targets, the first value is the lower one and the second
9710 value is the upper one. The opposite order applies to big-endian targets.
9711 For example, the code above will set the lower half of @code{a} to
9712 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
9714 @node MIPS Loongson Built-in Functions
9715 @subsection MIPS Loongson Built-in Functions
9717 GCC provides intrinsics to access the SIMD instructions provided by the
9718 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
9719 available after inclusion of the @code{loongson.h} header file,
9720 operate on the following 64-bit vector types:
9723 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
9724 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
9725 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
9726 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
9727 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
9728 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
9731 The intrinsics provided are listed below; each is named after the
9732 machine instruction to which it corresponds, with suffixes added as
9733 appropriate to distinguish intrinsics that expand to the same machine
9734 instruction yet have different argument types. Refer to the architecture
9735 documentation for a description of the functionality of each
9739 int16x4_t packsswh (int32x2_t s, int32x2_t t);
9740 int8x8_t packsshb (int16x4_t s, int16x4_t t);
9741 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
9742 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
9743 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
9744 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
9745 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
9746 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
9747 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
9748 uint64_t paddd_u (uint64_t s, uint64_t t);
9749 int64_t paddd_s (int64_t s, int64_t t);
9750 int16x4_t paddsh (int16x4_t s, int16x4_t t);
9751 int8x8_t paddsb (int8x8_t s, int8x8_t t);
9752 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
9753 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
9754 uint64_t pandn_ud (uint64_t s, uint64_t t);
9755 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
9756 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
9757 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
9758 int64_t pandn_sd (int64_t s, int64_t t);
9759 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
9760 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
9761 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
9762 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
9763 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
9764 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
9765 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
9766 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
9767 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
9768 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
9769 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
9770 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
9771 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
9772 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
9773 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
9774 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
9775 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
9776 uint16x4_t pextrh_u (uint16x4_t s, int field);
9777 int16x4_t pextrh_s (int16x4_t s, int field);
9778 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
9779 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
9780 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
9781 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
9782 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
9783 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
9784 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
9785 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
9786 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
9787 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
9788 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
9789 int16x4_t pminsh (int16x4_t s, int16x4_t t);
9790 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
9791 uint8x8_t pmovmskb_u (uint8x8_t s);
9792 int8x8_t pmovmskb_s (int8x8_t s);
9793 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
9794 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
9795 int16x4_t pmullh (int16x4_t s, int16x4_t t);
9796 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
9797 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
9798 uint16x4_t biadd (uint8x8_t s);
9799 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
9800 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
9801 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
9802 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
9803 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
9804 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
9805 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
9806 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
9807 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
9808 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
9809 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
9810 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
9811 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
9812 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
9813 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
9814 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
9815 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
9816 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
9817 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
9818 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
9819 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
9820 uint64_t psubd_u (uint64_t s, uint64_t t);
9821 int64_t psubd_s (int64_t s, int64_t t);
9822 int16x4_t psubsh (int16x4_t s, int16x4_t t);
9823 int8x8_t psubsb (int8x8_t s, int8x8_t t);
9824 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
9825 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
9826 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
9827 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
9828 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
9829 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
9830 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
9831 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
9832 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
9833 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
9834 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
9835 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
9836 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
9837 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
9841 * Paired-Single Arithmetic::
9842 * Paired-Single Built-in Functions::
9843 * MIPS-3D Built-in Functions::
9846 @node Paired-Single Arithmetic
9847 @subsubsection Paired-Single Arithmetic
9849 The table below lists the @code{v2sf} operations for which hardware
9850 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
9851 values and @code{x} is an integral value.
9853 @multitable @columnfractions .50 .50
9854 @item C code @tab MIPS instruction
9855 @item @code{a + b} @tab @code{add.ps}
9856 @item @code{a - b} @tab @code{sub.ps}
9857 @item @code{-a} @tab @code{neg.ps}
9858 @item @code{a * b} @tab @code{mul.ps}
9859 @item @code{a * b + c} @tab @code{madd.ps}
9860 @item @code{a * b - c} @tab @code{msub.ps}
9861 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
9862 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
9863 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
9866 Note that the multiply-accumulate instructions can be disabled
9867 using the command-line option @code{-mno-fused-madd}.
9869 @node Paired-Single Built-in Functions
9870 @subsubsection Paired-Single Built-in Functions
9872 The following paired-single functions map directly to a particular
9873 MIPS instruction. Please refer to the architecture specification
9874 for details on what each instruction does.
9877 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
9878 Pair lower lower (@code{pll.ps}).
9880 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
9881 Pair upper lower (@code{pul.ps}).
9883 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
9884 Pair lower upper (@code{plu.ps}).
9886 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
9887 Pair upper upper (@code{puu.ps}).
9889 @item v2sf __builtin_mips_cvt_ps_s (float, float)
9890 Convert pair to paired single (@code{cvt.ps.s}).
9892 @item float __builtin_mips_cvt_s_pl (v2sf)
9893 Convert pair lower to single (@code{cvt.s.pl}).
9895 @item float __builtin_mips_cvt_s_pu (v2sf)
9896 Convert pair upper to single (@code{cvt.s.pu}).
9898 @item v2sf __builtin_mips_abs_ps (v2sf)
9899 Absolute value (@code{abs.ps}).
9901 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
9902 Align variable (@code{alnv.ps}).
9904 @emph{Note:} The value of the third parameter must be 0 or 4
9905 modulo 8, otherwise the result will be unpredictable. Please read the
9906 instruction description for details.
9909 The following multi-instruction functions are also available.
9910 In each case, @var{cond} can be any of the 16 floating-point conditions:
9911 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9912 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
9913 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9916 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9917 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9918 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
9919 @code{movt.ps}/@code{movf.ps}).
9921 The @code{movt} functions return the value @var{x} computed by:
9924 c.@var{cond}.ps @var{cc},@var{a},@var{b}
9925 mov.ps @var{x},@var{c}
9926 movt.ps @var{x},@var{d},@var{cc}
9929 The @code{movf} functions are similar but use @code{movf.ps} instead
9932 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9933 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9934 Comparison of two paired-single values (@code{c.@var{cond}.ps},
9935 @code{bc1t}/@code{bc1f}).
9937 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9938 and return either the upper or lower half of the result. For example:
9942 if (__builtin_mips_upper_c_eq_ps (a, b))
9943 upper_halves_are_equal ();
9945 upper_halves_are_unequal ();
9947 if (__builtin_mips_lower_c_eq_ps (a, b))
9948 lower_halves_are_equal ();
9950 lower_halves_are_unequal ();
9954 @node MIPS-3D Built-in Functions
9955 @subsubsection MIPS-3D Built-in Functions
9957 The MIPS-3D Application-Specific Extension (ASE) includes additional
9958 paired-single instructions that are designed to improve the performance
9959 of 3D graphics operations. Support for these instructions is controlled
9960 by the @option{-mips3d} command-line option.
9962 The functions listed below map directly to a particular MIPS-3D
9963 instruction. Please refer to the architecture specification for
9964 more details on what each instruction does.
9967 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
9968 Reduction add (@code{addr.ps}).
9970 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
9971 Reduction multiply (@code{mulr.ps}).
9973 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
9974 Convert paired single to paired word (@code{cvt.pw.ps}).
9976 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
9977 Convert paired word to paired single (@code{cvt.ps.pw}).
9979 @item float __builtin_mips_recip1_s (float)
9980 @itemx double __builtin_mips_recip1_d (double)
9981 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
9982 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
9984 @item float __builtin_mips_recip2_s (float, float)
9985 @itemx double __builtin_mips_recip2_d (double, double)
9986 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
9987 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
9989 @item float __builtin_mips_rsqrt1_s (float)
9990 @itemx double __builtin_mips_rsqrt1_d (double)
9991 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
9992 Reduced precision reciprocal square root (sequence step 1)
9993 (@code{rsqrt1.@var{fmt}}).
9995 @item float __builtin_mips_rsqrt2_s (float, float)
9996 @itemx double __builtin_mips_rsqrt2_d (double, double)
9997 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
9998 Reduced precision reciprocal square root (sequence step 2)
9999 (@code{rsqrt2.@var{fmt}}).
10002 The following multi-instruction functions are also available.
10003 In each case, @var{cond} can be any of the 16 floating-point conditions:
10004 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
10005 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
10006 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
10009 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
10010 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
10011 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
10012 @code{bc1t}/@code{bc1f}).
10014 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
10015 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
10020 if (__builtin_mips_cabs_eq_s (a, b))
10026 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10027 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10028 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
10029 @code{bc1t}/@code{bc1f}).
10031 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
10032 and return either the upper or lower half of the result. For example:
10036 if (__builtin_mips_upper_cabs_eq_ps (a, b))
10037 upper_halves_are_equal ();
10039 upper_halves_are_unequal ();
10041 if (__builtin_mips_lower_cabs_eq_ps (a, b))
10042 lower_halves_are_equal ();
10044 lower_halves_are_unequal ();
10047 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10048 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10049 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
10050 @code{movt.ps}/@code{movf.ps}).
10052 The @code{movt} functions return the value @var{x} computed by:
10055 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
10056 mov.ps @var{x},@var{c}
10057 movt.ps @var{x},@var{d},@var{cc}
10060 The @code{movf} functions are similar but use @code{movf.ps} instead
10063 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10064 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10065 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10066 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
10067 Comparison of two paired-single values
10068 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
10069 @code{bc1any2t}/@code{bc1any2f}).
10071 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
10072 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
10073 result is true and the @code{all} forms return true if both results are true.
10078 if (__builtin_mips_any_c_eq_ps (a, b))
10083 if (__builtin_mips_all_c_eq_ps (a, b))
10089 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10090 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10091 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10092 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
10093 Comparison of four paired-single values
10094 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
10095 @code{bc1any4t}/@code{bc1any4f}).
10097 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
10098 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
10099 The @code{any} forms return true if any of the four results are true
10100 and the @code{all} forms return true if all four results are true.
10105 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
10110 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
10117 @node picoChip Built-in Functions
10118 @subsection picoChip Built-in Functions
10120 GCC provides an interface to selected machine instructions from the
10121 picoChip instruction set.
10124 @item int __builtin_sbc (int @var{value})
10125 Sign bit count. Return the number of consecutive bits in @var{value}
10126 which have the same value as the sign-bit. The result is the number of
10127 leading sign bits minus one, giving the number of redundant sign bits in
10130 @item int __builtin_byteswap (int @var{value})
10131 Byte swap. Return the result of swapping the upper and lower bytes of
10134 @item int __builtin_brev (int @var{value})
10135 Bit reversal. Return the result of reversing the bits in
10136 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
10139 @item int __builtin_adds (int @var{x}, int @var{y})
10140 Saturating addition. Return the result of adding @var{x} and @var{y},
10141 storing the value 32767 if the result overflows.
10143 @item int __builtin_subs (int @var{x}, int @var{y})
10144 Saturating subtraction. Return the result of subtracting @var{y} from
10145 @var{x}, storing the value @minus{}32768 if the result overflows.
10147 @item void __builtin_halt (void)
10148 Halt. The processor will stop execution. This built-in is useful for
10149 implementing assertions.
10153 @node Other MIPS Built-in Functions
10154 @subsection Other MIPS Built-in Functions
10156 GCC provides other MIPS-specific built-in functions:
10159 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
10160 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
10161 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
10162 when this function is available.
10165 @node PowerPC AltiVec/VSX Built-in Functions
10166 @subsection PowerPC AltiVec Built-in Functions
10168 GCC provides an interface for the PowerPC family of processors to access
10169 the AltiVec operations described in Motorola's AltiVec Programming
10170 Interface Manual. The interface is made available by including
10171 @code{<altivec.h>} and using @option{-maltivec} and
10172 @option{-mabi=altivec}. The interface supports the following vector
10176 vector unsigned char
10180 vector unsigned short
10181 vector signed short
10185 vector unsigned int
10191 If @option{-mvsx} is used the following additional vector types are
10195 vector unsigned long
10200 The long types are only implemented for 64-bit code generation, and
10201 the long type is only used in the floating point/integer conversion
10204 GCC's implementation of the high-level language interface available from
10205 C and C++ code differs from Motorola's documentation in several ways.
10210 A vector constant is a list of constant expressions within curly braces.
10213 A vector initializer requires no cast if the vector constant is of the
10214 same type as the variable it is initializing.
10217 If @code{signed} or @code{unsigned} is omitted, the signedness of the
10218 vector type is the default signedness of the base type. The default
10219 varies depending on the operating system, so a portable program should
10220 always specify the signedness.
10223 Compiling with @option{-maltivec} adds keywords @code{__vector},
10224 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
10225 @code{bool}. When compiling ISO C, the context-sensitive substitution
10226 of the keywords @code{vector}, @code{pixel} and @code{bool} is
10227 disabled. To use them, you must include @code{<altivec.h>} instead.
10230 GCC allows using a @code{typedef} name as the type specifier for a
10234 For C, overloaded functions are implemented with macros so the following
10238 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
10241 Since @code{vec_add} is a macro, the vector constant in the example
10242 is treated as four separate arguments. Wrap the entire argument in
10243 parentheses for this to work.
10246 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
10247 Internally, GCC uses built-in functions to achieve the functionality in
10248 the aforementioned header file, but they are not supported and are
10249 subject to change without notice.
10251 The following interfaces are supported for the generic and specific
10252 AltiVec operations and the AltiVec predicates. In cases where there
10253 is a direct mapping between generic and specific operations, only the
10254 generic names are shown here, although the specific operations can also
10257 Arguments that are documented as @code{const int} require literal
10258 integral values within the range required for that operation.
10261 vector signed char vec_abs (vector signed char);
10262 vector signed short vec_abs (vector signed short);
10263 vector signed int vec_abs (vector signed int);
10264 vector float vec_abs (vector float);
10266 vector signed char vec_abss (vector signed char);
10267 vector signed short vec_abss (vector signed short);
10268 vector signed int vec_abss (vector signed int);
10270 vector signed char vec_add (vector bool char, vector signed char);
10271 vector signed char vec_add (vector signed char, vector bool char);
10272 vector signed char vec_add (vector signed char, vector signed char);
10273 vector unsigned char vec_add (vector bool char, vector unsigned char);
10274 vector unsigned char vec_add (vector unsigned char, vector bool char);
10275 vector unsigned char vec_add (vector unsigned char,
10276 vector unsigned char);
10277 vector signed short vec_add (vector bool short, vector signed short);
10278 vector signed short vec_add (vector signed short, vector bool short);
10279 vector signed short vec_add (vector signed short, vector signed short);
10280 vector unsigned short vec_add (vector bool short,
10281 vector unsigned short);
10282 vector unsigned short vec_add (vector unsigned short,
10283 vector bool short);
10284 vector unsigned short vec_add (vector unsigned short,
10285 vector unsigned short);
10286 vector signed int vec_add (vector bool int, vector signed int);
10287 vector signed int vec_add (vector signed int, vector bool int);
10288 vector signed int vec_add (vector signed int, vector signed int);
10289 vector unsigned int vec_add (vector bool int, vector unsigned int);
10290 vector unsigned int vec_add (vector unsigned int, vector bool int);
10291 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
10292 vector float vec_add (vector float, vector float);
10294 vector float vec_vaddfp (vector float, vector float);
10296 vector signed int vec_vadduwm (vector bool int, vector signed int);
10297 vector signed int vec_vadduwm (vector signed int, vector bool int);
10298 vector signed int vec_vadduwm (vector signed int, vector signed int);
10299 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
10300 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
10301 vector unsigned int vec_vadduwm (vector unsigned int,
10302 vector unsigned int);
10304 vector signed short vec_vadduhm (vector bool short,
10305 vector signed short);
10306 vector signed short vec_vadduhm (vector signed short,
10307 vector bool short);
10308 vector signed short vec_vadduhm (vector signed short,
10309 vector signed short);
10310 vector unsigned short vec_vadduhm (vector bool short,
10311 vector unsigned short);
10312 vector unsigned short vec_vadduhm (vector unsigned short,
10313 vector bool short);
10314 vector unsigned short vec_vadduhm (vector unsigned short,
10315 vector unsigned short);
10317 vector signed char vec_vaddubm (vector bool char, vector signed char);
10318 vector signed char vec_vaddubm (vector signed char, vector bool char);
10319 vector signed char vec_vaddubm (vector signed char, vector signed char);
10320 vector unsigned char vec_vaddubm (vector bool char,
10321 vector unsigned char);
10322 vector unsigned char vec_vaddubm (vector unsigned char,
10324 vector unsigned char vec_vaddubm (vector unsigned char,
10325 vector unsigned char);
10327 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
10329 vector unsigned char vec_adds (vector bool char, vector unsigned char);
10330 vector unsigned char vec_adds (vector unsigned char, vector bool char);
10331 vector unsigned char vec_adds (vector unsigned char,
10332 vector unsigned char);
10333 vector signed char vec_adds (vector bool char, vector signed char);
10334 vector signed char vec_adds (vector signed char, vector bool char);
10335 vector signed char vec_adds (vector signed char, vector signed char);
10336 vector unsigned short vec_adds (vector bool short,
10337 vector unsigned short);
10338 vector unsigned short vec_adds (vector unsigned short,
10339 vector bool short);
10340 vector unsigned short vec_adds (vector unsigned short,
10341 vector unsigned short);
10342 vector signed short vec_adds (vector bool short, vector signed short);
10343 vector signed short vec_adds (vector signed short, vector bool short);
10344 vector signed short vec_adds (vector signed short, vector signed short);
10345 vector unsigned int vec_adds (vector bool int, vector unsigned int);
10346 vector unsigned int vec_adds (vector unsigned int, vector bool int);
10347 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
10348 vector signed int vec_adds (vector bool int, vector signed int);
10349 vector signed int vec_adds (vector signed int, vector bool int);
10350 vector signed int vec_adds (vector signed int, vector signed int);
10352 vector signed int vec_vaddsws (vector bool int, vector signed int);
10353 vector signed int vec_vaddsws (vector signed int, vector bool int);
10354 vector signed int vec_vaddsws (vector signed int, vector signed int);
10356 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
10357 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
10358 vector unsigned int vec_vadduws (vector unsigned int,
10359 vector unsigned int);
10361 vector signed short vec_vaddshs (vector bool short,
10362 vector signed short);
10363 vector signed short vec_vaddshs (vector signed short,
10364 vector bool short);
10365 vector signed short vec_vaddshs (vector signed short,
10366 vector signed short);
10368 vector unsigned short vec_vadduhs (vector bool short,
10369 vector unsigned short);
10370 vector unsigned short vec_vadduhs (vector unsigned short,
10371 vector bool short);
10372 vector unsigned short vec_vadduhs (vector unsigned short,
10373 vector unsigned short);
10375 vector signed char vec_vaddsbs (vector bool char, vector signed char);
10376 vector signed char vec_vaddsbs (vector signed char, vector bool char);
10377 vector signed char vec_vaddsbs (vector signed char, vector signed char);
10379 vector unsigned char vec_vaddubs (vector bool char,
10380 vector unsigned char);
10381 vector unsigned char vec_vaddubs (vector unsigned char,
10383 vector unsigned char vec_vaddubs (vector unsigned char,
10384 vector unsigned char);
10386 vector float vec_and (vector float, vector float);
10387 vector float vec_and (vector float, vector bool int);
10388 vector float vec_and (vector bool int, vector float);
10389 vector bool int vec_and (vector bool int, vector bool int);
10390 vector signed int vec_and (vector bool int, vector signed int);
10391 vector signed int vec_and (vector signed int, vector bool int);
10392 vector signed int vec_and (vector signed int, vector signed int);
10393 vector unsigned int vec_and (vector bool int, vector unsigned int);
10394 vector unsigned int vec_and (vector unsigned int, vector bool int);
10395 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
10396 vector bool short vec_and (vector bool short, vector bool short);
10397 vector signed short vec_and (vector bool short, vector signed short);
10398 vector signed short vec_and (vector signed short, vector bool short);
10399 vector signed short vec_and (vector signed short, vector signed short);
10400 vector unsigned short vec_and (vector bool short,
10401 vector unsigned short);
10402 vector unsigned short vec_and (vector unsigned short,
10403 vector bool short);
10404 vector unsigned short vec_and (vector unsigned short,
10405 vector unsigned short);
10406 vector signed char vec_and (vector bool char, vector signed char);
10407 vector bool char vec_and (vector bool char, vector bool char);
10408 vector signed char vec_and (vector signed char, vector bool char);
10409 vector signed char vec_and (vector signed char, vector signed char);
10410 vector unsigned char vec_and (vector bool char, vector unsigned char);
10411 vector unsigned char vec_and (vector unsigned char, vector bool char);
10412 vector unsigned char vec_and (vector unsigned char,
10413 vector unsigned char);
10415 vector float vec_andc (vector float, vector float);
10416 vector float vec_andc (vector float, vector bool int);
10417 vector float vec_andc (vector bool int, vector float);
10418 vector bool int vec_andc (vector bool int, vector bool int);
10419 vector signed int vec_andc (vector bool int, vector signed int);
10420 vector signed int vec_andc (vector signed int, vector bool int);
10421 vector signed int vec_andc (vector signed int, vector signed int);
10422 vector unsigned int vec_andc (vector bool int, vector unsigned int);
10423 vector unsigned int vec_andc (vector unsigned int, vector bool int);
10424 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
10425 vector bool short vec_andc (vector bool short, vector bool short);
10426 vector signed short vec_andc (vector bool short, vector signed short);
10427 vector signed short vec_andc (vector signed short, vector bool short);
10428 vector signed short vec_andc (vector signed short, vector signed short);
10429 vector unsigned short vec_andc (vector bool short,
10430 vector unsigned short);
10431 vector unsigned short vec_andc (vector unsigned short,
10432 vector bool short);
10433 vector unsigned short vec_andc (vector unsigned short,
10434 vector unsigned short);
10435 vector signed char vec_andc (vector bool char, vector signed char);
10436 vector bool char vec_andc (vector bool char, vector bool char);
10437 vector signed char vec_andc (vector signed char, vector bool char);
10438 vector signed char vec_andc (vector signed char, vector signed char);
10439 vector unsigned char vec_andc (vector bool char, vector unsigned char);
10440 vector unsigned char vec_andc (vector unsigned char, vector bool char);
10441 vector unsigned char vec_andc (vector unsigned char,
10442 vector unsigned char);
10444 vector unsigned char vec_avg (vector unsigned char,
10445 vector unsigned char);
10446 vector signed char vec_avg (vector signed char, vector signed char);
10447 vector unsigned short vec_avg (vector unsigned short,
10448 vector unsigned short);
10449 vector signed short vec_avg (vector signed short, vector signed short);
10450 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
10451 vector signed int vec_avg (vector signed int, vector signed int);
10453 vector signed int vec_vavgsw (vector signed int, vector signed int);
10455 vector unsigned int vec_vavguw (vector unsigned int,
10456 vector unsigned int);
10458 vector signed short vec_vavgsh (vector signed short,
10459 vector signed short);
10461 vector unsigned short vec_vavguh (vector unsigned short,
10462 vector unsigned short);
10464 vector signed char vec_vavgsb (vector signed char, vector signed char);
10466 vector unsigned char vec_vavgub (vector unsigned char,
10467 vector unsigned char);
10469 vector float vec_copysign (vector float);
10471 vector float vec_ceil (vector float);
10473 vector signed int vec_cmpb (vector float, vector float);
10475 vector bool char vec_cmpeq (vector signed char, vector signed char);
10476 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
10477 vector bool short vec_cmpeq (vector signed short, vector signed short);
10478 vector bool short vec_cmpeq (vector unsigned short,
10479 vector unsigned short);
10480 vector bool int vec_cmpeq (vector signed int, vector signed int);
10481 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
10482 vector bool int vec_cmpeq (vector float, vector float);
10484 vector bool int vec_vcmpeqfp (vector float, vector float);
10486 vector bool int vec_vcmpequw (vector signed int, vector signed int);
10487 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
10489 vector bool short vec_vcmpequh (vector signed short,
10490 vector signed short);
10491 vector bool short vec_vcmpequh (vector unsigned short,
10492 vector unsigned short);
10494 vector bool char vec_vcmpequb (vector signed char, vector signed char);
10495 vector bool char vec_vcmpequb (vector unsigned char,
10496 vector unsigned char);
10498 vector bool int vec_cmpge (vector float, vector float);
10500 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
10501 vector bool char vec_cmpgt (vector signed char, vector signed char);
10502 vector bool short vec_cmpgt (vector unsigned short,
10503 vector unsigned short);
10504 vector bool short vec_cmpgt (vector signed short, vector signed short);
10505 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
10506 vector bool int vec_cmpgt (vector signed int, vector signed int);
10507 vector bool int vec_cmpgt (vector float, vector float);
10509 vector bool int vec_vcmpgtfp (vector float, vector float);
10511 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
10513 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
10515 vector bool short vec_vcmpgtsh (vector signed short,
10516 vector signed short);
10518 vector bool short vec_vcmpgtuh (vector unsigned short,
10519 vector unsigned short);
10521 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
10523 vector bool char vec_vcmpgtub (vector unsigned char,
10524 vector unsigned char);
10526 vector bool int vec_cmple (vector float, vector float);
10528 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
10529 vector bool char vec_cmplt (vector signed char, vector signed char);
10530 vector bool short vec_cmplt (vector unsigned short,
10531 vector unsigned short);
10532 vector bool short vec_cmplt (vector signed short, vector signed short);
10533 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
10534 vector bool int vec_cmplt (vector signed int, vector signed int);
10535 vector bool int vec_cmplt (vector float, vector float);
10537 vector float vec_ctf (vector unsigned int, const int);
10538 vector float vec_ctf (vector signed int, const int);
10540 vector float vec_vcfsx (vector signed int, const int);
10542 vector float vec_vcfux (vector unsigned int, const int);
10544 vector signed int vec_cts (vector float, const int);
10546 vector unsigned int vec_ctu (vector float, const int);
10548 void vec_dss (const int);
10550 void vec_dssall (void);
10552 void vec_dst (const vector unsigned char *, int, const int);
10553 void vec_dst (const vector signed char *, int, const int);
10554 void vec_dst (const vector bool char *, int, const int);
10555 void vec_dst (const vector unsigned short *, int, const int);
10556 void vec_dst (const vector signed short *, int, const int);
10557 void vec_dst (const vector bool short *, int, const int);
10558 void vec_dst (const vector pixel *, int, const int);
10559 void vec_dst (const vector unsigned int *, int, const int);
10560 void vec_dst (const vector signed int *, int, const int);
10561 void vec_dst (const vector bool int *, int, const int);
10562 void vec_dst (const vector float *, int, const int);
10563 void vec_dst (const unsigned char *, int, const int);
10564 void vec_dst (const signed char *, int, const int);
10565 void vec_dst (const unsigned short *, int, const int);
10566 void vec_dst (const short *, int, const int);
10567 void vec_dst (const unsigned int *, int, const int);
10568 void vec_dst (const int *, int, const int);
10569 void vec_dst (const unsigned long *, int, const int);
10570 void vec_dst (const long *, int, const int);
10571 void vec_dst (const float *, int, const int);
10573 void vec_dstst (const vector unsigned char *, int, const int);
10574 void vec_dstst (const vector signed char *, int, const int);
10575 void vec_dstst (const vector bool char *, int, const int);
10576 void vec_dstst (const vector unsigned short *, int, const int);
10577 void vec_dstst (const vector signed short *, int, const int);
10578 void vec_dstst (const vector bool short *, int, const int);
10579 void vec_dstst (const vector pixel *, int, const int);
10580 void vec_dstst (const vector unsigned int *, int, const int);
10581 void vec_dstst (const vector signed int *, int, const int);
10582 void vec_dstst (const vector bool int *, int, const int);
10583 void vec_dstst (const vector float *, int, const int);
10584 void vec_dstst (const unsigned char *, int, const int);
10585 void vec_dstst (const signed char *, int, const int);
10586 void vec_dstst (const unsigned short *, int, const int);
10587 void vec_dstst (const short *, int, const int);
10588 void vec_dstst (const unsigned int *, int, const int);
10589 void vec_dstst (const int *, int, const int);
10590 void vec_dstst (const unsigned long *, int, const int);
10591 void vec_dstst (const long *, int, const int);
10592 void vec_dstst (const float *, int, const int);
10594 void vec_dststt (const vector unsigned char *, int, const int);
10595 void vec_dststt (const vector signed char *, int, const int);
10596 void vec_dststt (const vector bool char *, int, const int);
10597 void vec_dststt (const vector unsigned short *, int, const int);
10598 void vec_dststt (const vector signed short *, int, const int);
10599 void vec_dststt (const vector bool short *, int, const int);
10600 void vec_dststt (const vector pixel *, int, const int);
10601 void vec_dststt (const vector unsigned int *, int, const int);
10602 void vec_dststt (const vector signed int *, int, const int);
10603 void vec_dststt (const vector bool int *, int, const int);
10604 void vec_dststt (const vector float *, int, const int);
10605 void vec_dststt (const unsigned char *, int, const int);
10606 void vec_dststt (const signed char *, int, const int);
10607 void vec_dststt (const unsigned short *, int, const int);
10608 void vec_dststt (const short *, int, const int);
10609 void vec_dststt (const unsigned int *, int, const int);
10610 void vec_dststt (const int *, int, const int);
10611 void vec_dststt (const unsigned long *, int, const int);
10612 void vec_dststt (const long *, int, const int);
10613 void vec_dststt (const float *, int, const int);
10615 void vec_dstt (const vector unsigned char *, int, const int);
10616 void vec_dstt (const vector signed char *, int, const int);
10617 void vec_dstt (const vector bool char *, int, const int);
10618 void vec_dstt (const vector unsigned short *, int, const int);
10619 void vec_dstt (const vector signed short *, int, const int);
10620 void vec_dstt (const vector bool short *, int, const int);
10621 void vec_dstt (const vector pixel *, int, const int);
10622 void vec_dstt (const vector unsigned int *, int, const int);
10623 void vec_dstt (const vector signed int *, int, const int);
10624 void vec_dstt (const vector bool int *, int, const int);
10625 void vec_dstt (const vector float *, int, const int);
10626 void vec_dstt (const unsigned char *, int, const int);
10627 void vec_dstt (const signed char *, int, const int);
10628 void vec_dstt (const unsigned short *, int, const int);
10629 void vec_dstt (const short *, int, const int);
10630 void vec_dstt (const unsigned int *, int, const int);
10631 void vec_dstt (const int *, int, const int);
10632 void vec_dstt (const unsigned long *, int, const int);
10633 void vec_dstt (const long *, int, const int);
10634 void vec_dstt (const float *, int, const int);
10636 vector float vec_expte (vector float);
10638 vector float vec_floor (vector float);
10640 vector float vec_ld (int, const vector float *);
10641 vector float vec_ld (int, const float *);
10642 vector bool int vec_ld (int, const vector bool int *);
10643 vector signed int vec_ld (int, const vector signed int *);
10644 vector signed int vec_ld (int, const int *);
10645 vector signed int vec_ld (int, const long *);
10646 vector unsigned int vec_ld (int, const vector unsigned int *);
10647 vector unsigned int vec_ld (int, const unsigned int *);
10648 vector unsigned int vec_ld (int, const unsigned long *);
10649 vector bool short vec_ld (int, const vector bool short *);
10650 vector pixel vec_ld (int, const vector pixel *);
10651 vector signed short vec_ld (int, const vector signed short *);
10652 vector signed short vec_ld (int, const short *);
10653 vector unsigned short vec_ld (int, const vector unsigned short *);
10654 vector unsigned short vec_ld (int, const unsigned short *);
10655 vector bool char vec_ld (int, const vector bool char *);
10656 vector signed char vec_ld (int, const vector signed char *);
10657 vector signed char vec_ld (int, const signed char *);
10658 vector unsigned char vec_ld (int, const vector unsigned char *);
10659 vector unsigned char vec_ld (int, const unsigned char *);
10661 vector signed char vec_lde (int, const signed char *);
10662 vector unsigned char vec_lde (int, const unsigned char *);
10663 vector signed short vec_lde (int, const short *);
10664 vector unsigned short vec_lde (int, const unsigned short *);
10665 vector float vec_lde (int, const float *);
10666 vector signed int vec_lde (int, const int *);
10667 vector unsigned int vec_lde (int, const unsigned int *);
10668 vector signed int vec_lde (int, const long *);
10669 vector unsigned int vec_lde (int, const unsigned long *);
10671 vector float vec_lvewx (int, float *);
10672 vector signed int vec_lvewx (int, int *);
10673 vector unsigned int vec_lvewx (int, unsigned int *);
10674 vector signed int vec_lvewx (int, long *);
10675 vector unsigned int vec_lvewx (int, unsigned long *);
10677 vector signed short vec_lvehx (int, short *);
10678 vector unsigned short vec_lvehx (int, unsigned short *);
10680 vector signed char vec_lvebx (int, char *);
10681 vector unsigned char vec_lvebx (int, unsigned char *);
10683 vector float vec_ldl (int, const vector float *);
10684 vector float vec_ldl (int, const float *);
10685 vector bool int vec_ldl (int, const vector bool int *);
10686 vector signed int vec_ldl (int, const vector signed int *);
10687 vector signed int vec_ldl (int, const int *);
10688 vector signed int vec_ldl (int, const long *);
10689 vector unsigned int vec_ldl (int, const vector unsigned int *);
10690 vector unsigned int vec_ldl (int, const unsigned int *);
10691 vector unsigned int vec_ldl (int, const unsigned long *);
10692 vector bool short vec_ldl (int, const vector bool short *);
10693 vector pixel vec_ldl (int, const vector pixel *);
10694 vector signed short vec_ldl (int, const vector signed short *);
10695 vector signed short vec_ldl (int, const short *);
10696 vector unsigned short vec_ldl (int, const vector unsigned short *);
10697 vector unsigned short vec_ldl (int, const unsigned short *);
10698 vector bool char vec_ldl (int, const vector bool char *);
10699 vector signed char vec_ldl (int, const vector signed char *);
10700 vector signed char vec_ldl (int, const signed char *);
10701 vector unsigned char vec_ldl (int, const vector unsigned char *);
10702 vector unsigned char vec_ldl (int, const unsigned char *);
10704 vector float vec_loge (vector float);
10706 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
10707 vector unsigned char vec_lvsl (int, const volatile signed char *);
10708 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
10709 vector unsigned char vec_lvsl (int, const volatile short *);
10710 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
10711 vector unsigned char vec_lvsl (int, const volatile int *);
10712 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
10713 vector unsigned char vec_lvsl (int, const volatile long *);
10714 vector unsigned char vec_lvsl (int, const volatile float *);
10716 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
10717 vector unsigned char vec_lvsr (int, const volatile signed char *);
10718 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
10719 vector unsigned char vec_lvsr (int, const volatile short *);
10720 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
10721 vector unsigned char vec_lvsr (int, const volatile int *);
10722 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
10723 vector unsigned char vec_lvsr (int, const volatile long *);
10724 vector unsigned char vec_lvsr (int, const volatile float *);
10726 vector float vec_madd (vector float, vector float, vector float);
10728 vector signed short vec_madds (vector signed short,
10729 vector signed short,
10730 vector signed short);
10732 vector unsigned char vec_max (vector bool char, vector unsigned char);
10733 vector unsigned char vec_max (vector unsigned char, vector bool char);
10734 vector unsigned char vec_max (vector unsigned char,
10735 vector unsigned char);
10736 vector signed char vec_max (vector bool char, vector signed char);
10737 vector signed char vec_max (vector signed char, vector bool char);
10738 vector signed char vec_max (vector signed char, vector signed char);
10739 vector unsigned short vec_max (vector bool short,
10740 vector unsigned short);
10741 vector unsigned short vec_max (vector unsigned short,
10742 vector bool short);
10743 vector unsigned short vec_max (vector unsigned short,
10744 vector unsigned short);
10745 vector signed short vec_max (vector bool short, vector signed short);
10746 vector signed short vec_max (vector signed short, vector bool short);
10747 vector signed short vec_max (vector signed short, vector signed short);
10748 vector unsigned int vec_max (vector bool int, vector unsigned int);
10749 vector unsigned int vec_max (vector unsigned int, vector bool int);
10750 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
10751 vector signed int vec_max (vector bool int, vector signed int);
10752 vector signed int vec_max (vector signed int, vector bool int);
10753 vector signed int vec_max (vector signed int, vector signed int);
10754 vector float vec_max (vector float, vector float);
10756 vector float vec_vmaxfp (vector float, vector float);
10758 vector signed int vec_vmaxsw (vector bool int, vector signed int);
10759 vector signed int vec_vmaxsw (vector signed int, vector bool int);
10760 vector signed int vec_vmaxsw (vector signed int, vector signed int);
10762 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
10763 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
10764 vector unsigned int vec_vmaxuw (vector unsigned int,
10765 vector unsigned int);
10767 vector signed short vec_vmaxsh (vector bool short, vector signed short);
10768 vector signed short vec_vmaxsh (vector signed short, vector bool short);
10769 vector signed short vec_vmaxsh (vector signed short,
10770 vector signed short);
10772 vector unsigned short vec_vmaxuh (vector bool short,
10773 vector unsigned short);
10774 vector unsigned short vec_vmaxuh (vector unsigned short,
10775 vector bool short);
10776 vector unsigned short vec_vmaxuh (vector unsigned short,
10777 vector unsigned short);
10779 vector signed char vec_vmaxsb (vector bool char, vector signed char);
10780 vector signed char vec_vmaxsb (vector signed char, vector bool char);
10781 vector signed char vec_vmaxsb (vector signed char, vector signed char);
10783 vector unsigned char vec_vmaxub (vector bool char,
10784 vector unsigned char);
10785 vector unsigned char vec_vmaxub (vector unsigned char,
10787 vector unsigned char vec_vmaxub (vector unsigned char,
10788 vector unsigned char);
10790 vector bool char vec_mergeh (vector bool char, vector bool char);
10791 vector signed char vec_mergeh (vector signed char, vector signed char);
10792 vector unsigned char vec_mergeh (vector unsigned char,
10793 vector unsigned char);
10794 vector bool short vec_mergeh (vector bool short, vector bool short);
10795 vector pixel vec_mergeh (vector pixel, vector pixel);
10796 vector signed short vec_mergeh (vector signed short,
10797 vector signed short);
10798 vector unsigned short vec_mergeh (vector unsigned short,
10799 vector unsigned short);
10800 vector float vec_mergeh (vector float, vector float);
10801 vector bool int vec_mergeh (vector bool int, vector bool int);
10802 vector signed int vec_mergeh (vector signed int, vector signed int);
10803 vector unsigned int vec_mergeh (vector unsigned int,
10804 vector unsigned int);
10806 vector float vec_vmrghw (vector float, vector float);
10807 vector bool int vec_vmrghw (vector bool int, vector bool int);
10808 vector signed int vec_vmrghw (vector signed int, vector signed int);
10809 vector unsigned int vec_vmrghw (vector unsigned int,
10810 vector unsigned int);
10812 vector bool short vec_vmrghh (vector bool short, vector bool short);
10813 vector signed short vec_vmrghh (vector signed short,
10814 vector signed short);
10815 vector unsigned short vec_vmrghh (vector unsigned short,
10816 vector unsigned short);
10817 vector pixel vec_vmrghh (vector pixel, vector pixel);
10819 vector bool char vec_vmrghb (vector bool char, vector bool char);
10820 vector signed char vec_vmrghb (vector signed char, vector signed char);
10821 vector unsigned char vec_vmrghb (vector unsigned char,
10822 vector unsigned char);
10824 vector bool char vec_mergel (vector bool char, vector bool char);
10825 vector signed char vec_mergel (vector signed char, vector signed char);
10826 vector unsigned char vec_mergel (vector unsigned char,
10827 vector unsigned char);
10828 vector bool short vec_mergel (vector bool short, vector bool short);
10829 vector pixel vec_mergel (vector pixel, vector pixel);
10830 vector signed short vec_mergel (vector signed short,
10831 vector signed short);
10832 vector unsigned short vec_mergel (vector unsigned short,
10833 vector unsigned short);
10834 vector float vec_mergel (vector float, vector float);
10835 vector bool int vec_mergel (vector bool int, vector bool int);
10836 vector signed int vec_mergel (vector signed int, vector signed int);
10837 vector unsigned int vec_mergel (vector unsigned int,
10838 vector unsigned int);
10840 vector float vec_vmrglw (vector float, vector float);
10841 vector signed int vec_vmrglw (vector signed int, vector signed int);
10842 vector unsigned int vec_vmrglw (vector unsigned int,
10843 vector unsigned int);
10844 vector bool int vec_vmrglw (vector bool int, vector bool int);
10846 vector bool short vec_vmrglh (vector bool short, vector bool short);
10847 vector signed short vec_vmrglh (vector signed short,
10848 vector signed short);
10849 vector unsigned short vec_vmrglh (vector unsigned short,
10850 vector unsigned short);
10851 vector pixel vec_vmrglh (vector pixel, vector pixel);
10853 vector bool char vec_vmrglb (vector bool char, vector bool char);
10854 vector signed char vec_vmrglb (vector signed char, vector signed char);
10855 vector unsigned char vec_vmrglb (vector unsigned char,
10856 vector unsigned char);
10858 vector unsigned short vec_mfvscr (void);
10860 vector unsigned char vec_min (vector bool char, vector unsigned char);
10861 vector unsigned char vec_min (vector unsigned char, vector bool char);
10862 vector unsigned char vec_min (vector unsigned char,
10863 vector unsigned char);
10864 vector signed char vec_min (vector bool char, vector signed char);
10865 vector signed char vec_min (vector signed char, vector bool char);
10866 vector signed char vec_min (vector signed char, vector signed char);
10867 vector unsigned short vec_min (vector bool short,
10868 vector unsigned short);
10869 vector unsigned short vec_min (vector unsigned short,
10870 vector bool short);
10871 vector unsigned short vec_min (vector unsigned short,
10872 vector unsigned short);
10873 vector signed short vec_min (vector bool short, vector signed short);
10874 vector signed short vec_min (vector signed short, vector bool short);
10875 vector signed short vec_min (vector signed short, vector signed short);
10876 vector unsigned int vec_min (vector bool int, vector unsigned int);
10877 vector unsigned int vec_min (vector unsigned int, vector bool int);
10878 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
10879 vector signed int vec_min (vector bool int, vector signed int);
10880 vector signed int vec_min (vector signed int, vector bool int);
10881 vector signed int vec_min (vector signed int, vector signed int);
10882 vector float vec_min (vector float, vector float);
10884 vector float vec_vminfp (vector float, vector float);
10886 vector signed int vec_vminsw (vector bool int, vector signed int);
10887 vector signed int vec_vminsw (vector signed int, vector bool int);
10888 vector signed int vec_vminsw (vector signed int, vector signed int);
10890 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
10891 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
10892 vector unsigned int vec_vminuw (vector unsigned int,
10893 vector unsigned int);
10895 vector signed short vec_vminsh (vector bool short, vector signed short);
10896 vector signed short vec_vminsh (vector signed short, vector bool short);
10897 vector signed short vec_vminsh (vector signed short,
10898 vector signed short);
10900 vector unsigned short vec_vminuh (vector bool short,
10901 vector unsigned short);
10902 vector unsigned short vec_vminuh (vector unsigned short,
10903 vector bool short);
10904 vector unsigned short vec_vminuh (vector unsigned short,
10905 vector unsigned short);
10907 vector signed char vec_vminsb (vector bool char, vector signed char);
10908 vector signed char vec_vminsb (vector signed char, vector bool char);
10909 vector signed char vec_vminsb (vector signed char, vector signed char);
10911 vector unsigned char vec_vminub (vector bool char,
10912 vector unsigned char);
10913 vector unsigned char vec_vminub (vector unsigned char,
10915 vector unsigned char vec_vminub (vector unsigned char,
10916 vector unsigned char);
10918 vector signed short vec_mladd (vector signed short,
10919 vector signed short,
10920 vector signed short);
10921 vector signed short vec_mladd (vector signed short,
10922 vector unsigned short,
10923 vector unsigned short);
10924 vector signed short vec_mladd (vector unsigned short,
10925 vector signed short,
10926 vector signed short);
10927 vector unsigned short vec_mladd (vector unsigned short,
10928 vector unsigned short,
10929 vector unsigned short);
10931 vector signed short vec_mradds (vector signed short,
10932 vector signed short,
10933 vector signed short);
10935 vector unsigned int vec_msum (vector unsigned char,
10936 vector unsigned char,
10937 vector unsigned int);
10938 vector signed int vec_msum (vector signed char,
10939 vector unsigned char,
10940 vector signed int);
10941 vector unsigned int vec_msum (vector unsigned short,
10942 vector unsigned short,
10943 vector unsigned int);
10944 vector signed int vec_msum (vector signed short,
10945 vector signed short,
10946 vector signed int);
10948 vector signed int vec_vmsumshm (vector signed short,
10949 vector signed short,
10950 vector signed int);
10952 vector unsigned int vec_vmsumuhm (vector unsigned short,
10953 vector unsigned short,
10954 vector unsigned int);
10956 vector signed int vec_vmsummbm (vector signed char,
10957 vector unsigned char,
10958 vector signed int);
10960 vector unsigned int vec_vmsumubm (vector unsigned char,
10961 vector unsigned char,
10962 vector unsigned int);
10964 vector unsigned int vec_msums (vector unsigned short,
10965 vector unsigned short,
10966 vector unsigned int);
10967 vector signed int vec_msums (vector signed short,
10968 vector signed short,
10969 vector signed int);
10971 vector signed int vec_vmsumshs (vector signed short,
10972 vector signed short,
10973 vector signed int);
10975 vector unsigned int vec_vmsumuhs (vector unsigned short,
10976 vector unsigned short,
10977 vector unsigned int);
10979 void vec_mtvscr (vector signed int);
10980 void vec_mtvscr (vector unsigned int);
10981 void vec_mtvscr (vector bool int);
10982 void vec_mtvscr (vector signed short);
10983 void vec_mtvscr (vector unsigned short);
10984 void vec_mtvscr (vector bool short);
10985 void vec_mtvscr (vector pixel);
10986 void vec_mtvscr (vector signed char);
10987 void vec_mtvscr (vector unsigned char);
10988 void vec_mtvscr (vector bool char);
10990 vector unsigned short vec_mule (vector unsigned char,
10991 vector unsigned char);
10992 vector signed short vec_mule (vector signed char,
10993 vector signed char);
10994 vector unsigned int vec_mule (vector unsigned short,
10995 vector unsigned short);
10996 vector signed int vec_mule (vector signed short, vector signed short);
10998 vector signed int vec_vmulesh (vector signed short,
10999 vector signed short);
11001 vector unsigned int vec_vmuleuh (vector unsigned short,
11002 vector unsigned short);
11004 vector signed short vec_vmulesb (vector signed char,
11005 vector signed char);
11007 vector unsigned short vec_vmuleub (vector unsigned char,
11008 vector unsigned char);
11010 vector unsigned short vec_mulo (vector unsigned char,
11011 vector unsigned char);
11012 vector signed short vec_mulo (vector signed char, vector signed char);
11013 vector unsigned int vec_mulo (vector unsigned short,
11014 vector unsigned short);
11015 vector signed int vec_mulo (vector signed short, vector signed short);
11017 vector signed int vec_vmulosh (vector signed short,
11018 vector signed short);
11020 vector unsigned int vec_vmulouh (vector unsigned short,
11021 vector unsigned short);
11023 vector signed short vec_vmulosb (vector signed char,
11024 vector signed char);
11026 vector unsigned short vec_vmuloub (vector unsigned char,
11027 vector unsigned char);
11029 vector float vec_nmsub (vector float, vector float, vector float);
11031 vector float vec_nor (vector float, vector float);
11032 vector signed int vec_nor (vector signed int, vector signed int);
11033 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
11034 vector bool int vec_nor (vector bool int, vector bool int);
11035 vector signed short vec_nor (vector signed short, vector signed short);
11036 vector unsigned short vec_nor (vector unsigned short,
11037 vector unsigned short);
11038 vector bool short vec_nor (vector bool short, vector bool short);
11039 vector signed char vec_nor (vector signed char, vector signed char);
11040 vector unsigned char vec_nor (vector unsigned char,
11041 vector unsigned char);
11042 vector bool char vec_nor (vector bool char, vector bool char);
11044 vector float vec_or (vector float, vector float);
11045 vector float vec_or (vector float, vector bool int);
11046 vector float vec_or (vector bool int, vector float);
11047 vector bool int vec_or (vector bool int, vector bool int);
11048 vector signed int vec_or (vector bool int, vector signed int);
11049 vector signed int vec_or (vector signed int, vector bool int);
11050 vector signed int vec_or (vector signed int, vector signed int);
11051 vector unsigned int vec_or (vector bool int, vector unsigned int);
11052 vector unsigned int vec_or (vector unsigned int, vector bool int);
11053 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
11054 vector bool short vec_or (vector bool short, vector bool short);
11055 vector signed short vec_or (vector bool short, vector signed short);
11056 vector signed short vec_or (vector signed short, vector bool short);
11057 vector signed short vec_or (vector signed short, vector signed short);
11058 vector unsigned short vec_or (vector bool short, vector unsigned short);
11059 vector unsigned short vec_or (vector unsigned short, vector bool short);
11060 vector unsigned short vec_or (vector unsigned short,
11061 vector unsigned short);
11062 vector signed char vec_or (vector bool char, vector signed char);
11063 vector bool char vec_or (vector bool char, vector bool char);
11064 vector signed char vec_or (vector signed char, vector bool char);
11065 vector signed char vec_or (vector signed char, vector signed char);
11066 vector unsigned char vec_or (vector bool char, vector unsigned char);
11067 vector unsigned char vec_or (vector unsigned char, vector bool char);
11068 vector unsigned char vec_or (vector unsigned char,
11069 vector unsigned char);
11071 vector signed char vec_pack (vector signed short, vector signed short);
11072 vector unsigned char vec_pack (vector unsigned short,
11073 vector unsigned short);
11074 vector bool char vec_pack (vector bool short, vector bool short);
11075 vector signed short vec_pack (vector signed int, vector signed int);
11076 vector unsigned short vec_pack (vector unsigned int,
11077 vector unsigned int);
11078 vector bool short vec_pack (vector bool int, vector bool int);
11080 vector bool short vec_vpkuwum (vector bool int, vector bool int);
11081 vector signed short vec_vpkuwum (vector signed int, vector signed int);
11082 vector unsigned short vec_vpkuwum (vector unsigned int,
11083 vector unsigned int);
11085 vector bool char vec_vpkuhum (vector bool short, vector bool short);
11086 vector signed char vec_vpkuhum (vector signed short,
11087 vector signed short);
11088 vector unsigned char vec_vpkuhum (vector unsigned short,
11089 vector unsigned short);
11091 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
11093 vector unsigned char vec_packs (vector unsigned short,
11094 vector unsigned short);
11095 vector signed char vec_packs (vector signed short, vector signed short);
11096 vector unsigned short vec_packs (vector unsigned int,
11097 vector unsigned int);
11098 vector signed short vec_packs (vector signed int, vector signed int);
11100 vector signed short vec_vpkswss (vector signed int, vector signed int);
11102 vector unsigned short vec_vpkuwus (vector unsigned int,
11103 vector unsigned int);
11105 vector signed char vec_vpkshss (vector signed short,
11106 vector signed short);
11108 vector unsigned char vec_vpkuhus (vector unsigned short,
11109 vector unsigned short);
11111 vector unsigned char vec_packsu (vector unsigned short,
11112 vector unsigned short);
11113 vector unsigned char vec_packsu (vector signed short,
11114 vector signed short);
11115 vector unsigned short vec_packsu (vector unsigned int,
11116 vector unsigned int);
11117 vector unsigned short vec_packsu (vector signed int, vector signed int);
11119 vector unsigned short vec_vpkswus (vector signed int,
11120 vector signed int);
11122 vector unsigned char vec_vpkshus (vector signed short,
11123 vector signed short);
11125 vector float vec_perm (vector float,
11127 vector unsigned char);
11128 vector signed int vec_perm (vector signed int,
11130 vector unsigned char);
11131 vector unsigned int vec_perm (vector unsigned int,
11132 vector unsigned int,
11133 vector unsigned char);
11134 vector bool int vec_perm (vector bool int,
11136 vector unsigned char);
11137 vector signed short vec_perm (vector signed short,
11138 vector signed short,
11139 vector unsigned char);
11140 vector unsigned short vec_perm (vector unsigned short,
11141 vector unsigned short,
11142 vector unsigned char);
11143 vector bool short vec_perm (vector bool short,
11145 vector unsigned char);
11146 vector pixel vec_perm (vector pixel,
11148 vector unsigned char);
11149 vector signed char vec_perm (vector signed char,
11150 vector signed char,
11151 vector unsigned char);
11152 vector unsigned char vec_perm (vector unsigned char,
11153 vector unsigned char,
11154 vector unsigned char);
11155 vector bool char vec_perm (vector bool char,
11157 vector unsigned char);
11159 vector float vec_re (vector float);
11161 vector signed char vec_rl (vector signed char,
11162 vector unsigned char);
11163 vector unsigned char vec_rl (vector unsigned char,
11164 vector unsigned char);
11165 vector signed short vec_rl (vector signed short, vector unsigned short);
11166 vector unsigned short vec_rl (vector unsigned short,
11167 vector unsigned short);
11168 vector signed int vec_rl (vector signed int, vector unsigned int);
11169 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
11171 vector signed int vec_vrlw (vector signed int, vector unsigned int);
11172 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
11174 vector signed short vec_vrlh (vector signed short,
11175 vector unsigned short);
11176 vector unsigned short vec_vrlh (vector unsigned short,
11177 vector unsigned short);
11179 vector signed char vec_vrlb (vector signed char, vector unsigned char);
11180 vector unsigned char vec_vrlb (vector unsigned char,
11181 vector unsigned char);
11183 vector float vec_round (vector float);
11185 vector float vec_recip (vector float, vector float);
11187 vector float vec_rsqrt (vector float);
11189 vector float vec_rsqrte (vector float);
11191 vector float vec_sel (vector float, vector float, vector bool int);
11192 vector float vec_sel (vector float, vector float, vector unsigned int);
11193 vector signed int vec_sel (vector signed int,
11196 vector signed int vec_sel (vector signed int,
11198 vector unsigned int);
11199 vector unsigned int vec_sel (vector unsigned int,
11200 vector unsigned int,
11202 vector unsigned int vec_sel (vector unsigned int,
11203 vector unsigned int,
11204 vector unsigned int);
11205 vector bool int vec_sel (vector bool int,
11208 vector bool int vec_sel (vector bool int,
11210 vector unsigned int);
11211 vector signed short vec_sel (vector signed short,
11212 vector signed short,
11213 vector bool short);
11214 vector signed short vec_sel (vector signed short,
11215 vector signed short,
11216 vector unsigned short);
11217 vector unsigned short vec_sel (vector unsigned short,
11218 vector unsigned short,
11219 vector bool short);
11220 vector unsigned short vec_sel (vector unsigned short,
11221 vector unsigned short,
11222 vector unsigned short);
11223 vector bool short vec_sel (vector bool short,
11225 vector bool short);
11226 vector bool short vec_sel (vector bool short,
11228 vector unsigned short);
11229 vector signed char vec_sel (vector signed char,
11230 vector signed char,
11232 vector signed char vec_sel (vector signed char,
11233 vector signed char,
11234 vector unsigned char);
11235 vector unsigned char vec_sel (vector unsigned char,
11236 vector unsigned char,
11238 vector unsigned char vec_sel (vector unsigned char,
11239 vector unsigned char,
11240 vector unsigned char);
11241 vector bool char vec_sel (vector bool char,
11244 vector bool char vec_sel (vector bool char,
11246 vector unsigned char);
11248 vector signed char vec_sl (vector signed char,
11249 vector unsigned char);
11250 vector unsigned char vec_sl (vector unsigned char,
11251 vector unsigned char);
11252 vector signed short vec_sl (vector signed short, vector unsigned short);
11253 vector unsigned short vec_sl (vector unsigned short,
11254 vector unsigned short);
11255 vector signed int vec_sl (vector signed int, vector unsigned int);
11256 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
11258 vector signed int vec_vslw (vector signed int, vector unsigned int);
11259 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
11261 vector signed short vec_vslh (vector signed short,
11262 vector unsigned short);
11263 vector unsigned short vec_vslh (vector unsigned short,
11264 vector unsigned short);
11266 vector signed char vec_vslb (vector signed char, vector unsigned char);
11267 vector unsigned char vec_vslb (vector unsigned char,
11268 vector unsigned char);
11270 vector float vec_sld (vector float, vector float, const int);
11271 vector signed int vec_sld (vector signed int,
11274 vector unsigned int vec_sld (vector unsigned int,
11275 vector unsigned int,
11277 vector bool int vec_sld (vector bool int,
11280 vector signed short vec_sld (vector signed short,
11281 vector signed short,
11283 vector unsigned short vec_sld (vector unsigned short,
11284 vector unsigned short,
11286 vector bool short vec_sld (vector bool short,
11289 vector pixel vec_sld (vector pixel,
11292 vector signed char vec_sld (vector signed char,
11293 vector signed char,
11295 vector unsigned char vec_sld (vector unsigned char,
11296 vector unsigned char,
11298 vector bool char vec_sld (vector bool char,
11302 vector signed int vec_sll (vector signed int,
11303 vector unsigned int);
11304 vector signed int vec_sll (vector signed int,
11305 vector unsigned short);
11306 vector signed int vec_sll (vector signed int,
11307 vector unsigned char);
11308 vector unsigned int vec_sll (vector unsigned int,
11309 vector unsigned int);
11310 vector unsigned int vec_sll (vector unsigned int,
11311 vector unsigned short);
11312 vector unsigned int vec_sll (vector unsigned int,
11313 vector unsigned char);
11314 vector bool int vec_sll (vector bool int,
11315 vector unsigned int);
11316 vector bool int vec_sll (vector bool int,
11317 vector unsigned short);
11318 vector bool int vec_sll (vector bool int,
11319 vector unsigned char);
11320 vector signed short vec_sll (vector signed short,
11321 vector unsigned int);
11322 vector signed short vec_sll (vector signed short,
11323 vector unsigned short);
11324 vector signed short vec_sll (vector signed short,
11325 vector unsigned char);
11326 vector unsigned short vec_sll (vector unsigned short,
11327 vector unsigned int);
11328 vector unsigned short vec_sll (vector unsigned short,
11329 vector unsigned short);
11330 vector unsigned short vec_sll (vector unsigned short,
11331 vector unsigned char);
11332 vector bool short vec_sll (vector bool short, vector unsigned int);
11333 vector bool short vec_sll (vector bool short, vector unsigned short);
11334 vector bool short vec_sll (vector bool short, vector unsigned char);
11335 vector pixel vec_sll (vector pixel, vector unsigned int);
11336 vector pixel vec_sll (vector pixel, vector unsigned short);
11337 vector pixel vec_sll (vector pixel, vector unsigned char);
11338 vector signed char vec_sll (vector signed char, vector unsigned int);
11339 vector signed char vec_sll (vector signed char, vector unsigned short);
11340 vector signed char vec_sll (vector signed char, vector unsigned char);
11341 vector unsigned char vec_sll (vector unsigned char,
11342 vector unsigned int);
11343 vector unsigned char vec_sll (vector unsigned char,
11344 vector unsigned short);
11345 vector unsigned char vec_sll (vector unsigned char,
11346 vector unsigned char);
11347 vector bool char vec_sll (vector bool char, vector unsigned int);
11348 vector bool char vec_sll (vector bool char, vector unsigned short);
11349 vector bool char vec_sll (vector bool char, vector unsigned char);
11351 vector float vec_slo (vector float, vector signed char);
11352 vector float vec_slo (vector float, vector unsigned char);
11353 vector signed int vec_slo (vector signed int, vector signed char);
11354 vector signed int vec_slo (vector signed int, vector unsigned char);
11355 vector unsigned int vec_slo (vector unsigned int, vector signed char);
11356 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
11357 vector signed short vec_slo (vector signed short, vector signed char);
11358 vector signed short vec_slo (vector signed short, vector unsigned char);
11359 vector unsigned short vec_slo (vector unsigned short,
11360 vector signed char);
11361 vector unsigned short vec_slo (vector unsigned short,
11362 vector unsigned char);
11363 vector pixel vec_slo (vector pixel, vector signed char);
11364 vector pixel vec_slo (vector pixel, vector unsigned char);
11365 vector signed char vec_slo (vector signed char, vector signed char);
11366 vector signed char vec_slo (vector signed char, vector unsigned char);
11367 vector unsigned char vec_slo (vector unsigned char, vector signed char);
11368 vector unsigned char vec_slo (vector unsigned char,
11369 vector unsigned char);
11371 vector signed char vec_splat (vector signed char, const int);
11372 vector unsigned char vec_splat (vector unsigned char, const int);
11373 vector bool char vec_splat (vector bool char, const int);
11374 vector signed short vec_splat (vector signed short, const int);
11375 vector unsigned short vec_splat (vector unsigned short, const int);
11376 vector bool short vec_splat (vector bool short, const int);
11377 vector pixel vec_splat (vector pixel, const int);
11378 vector float vec_splat (vector float, const int);
11379 vector signed int vec_splat (vector signed int, const int);
11380 vector unsigned int vec_splat (vector unsigned int, const int);
11381 vector bool int vec_splat (vector bool int, const int);
11383 vector float vec_vspltw (vector float, const int);
11384 vector signed int vec_vspltw (vector signed int, const int);
11385 vector unsigned int vec_vspltw (vector unsigned int, const int);
11386 vector bool int vec_vspltw (vector bool int, const int);
11388 vector bool short vec_vsplth (vector bool short, const int);
11389 vector signed short vec_vsplth (vector signed short, const int);
11390 vector unsigned short vec_vsplth (vector unsigned short, const int);
11391 vector pixel vec_vsplth (vector pixel, const int);
11393 vector signed char vec_vspltb (vector signed char, const int);
11394 vector unsigned char vec_vspltb (vector unsigned char, const int);
11395 vector bool char vec_vspltb (vector bool char, const int);
11397 vector signed char vec_splat_s8 (const int);
11399 vector signed short vec_splat_s16 (const int);
11401 vector signed int vec_splat_s32 (const int);
11403 vector unsigned char vec_splat_u8 (const int);
11405 vector unsigned short vec_splat_u16 (const int);
11407 vector unsigned int vec_splat_u32 (const int);
11409 vector signed char vec_sr (vector signed char, vector unsigned char);
11410 vector unsigned char vec_sr (vector unsigned char,
11411 vector unsigned char);
11412 vector signed short vec_sr (vector signed short,
11413 vector unsigned short);
11414 vector unsigned short vec_sr (vector unsigned short,
11415 vector unsigned short);
11416 vector signed int vec_sr (vector signed int, vector unsigned int);
11417 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
11419 vector signed int vec_vsrw (vector signed int, vector unsigned int);
11420 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
11422 vector signed short vec_vsrh (vector signed short,
11423 vector unsigned short);
11424 vector unsigned short vec_vsrh (vector unsigned short,
11425 vector unsigned short);
11427 vector signed char vec_vsrb (vector signed char, vector unsigned char);
11428 vector unsigned char vec_vsrb (vector unsigned char,
11429 vector unsigned char);
11431 vector signed char vec_sra (vector signed char, vector unsigned char);
11432 vector unsigned char vec_sra (vector unsigned char,
11433 vector unsigned char);
11434 vector signed short vec_sra (vector signed short,
11435 vector unsigned short);
11436 vector unsigned short vec_sra (vector unsigned short,
11437 vector unsigned short);
11438 vector signed int vec_sra (vector signed int, vector unsigned int);
11439 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
11441 vector signed int vec_vsraw (vector signed int, vector unsigned int);
11442 vector unsigned int vec_vsraw (vector unsigned int,
11443 vector unsigned int);
11445 vector signed short vec_vsrah (vector signed short,
11446 vector unsigned short);
11447 vector unsigned short vec_vsrah (vector unsigned short,
11448 vector unsigned short);
11450 vector signed char vec_vsrab (vector signed char, vector unsigned char);
11451 vector unsigned char vec_vsrab (vector unsigned char,
11452 vector unsigned char);
11454 vector signed int vec_srl (vector signed int, vector unsigned int);
11455 vector signed int vec_srl (vector signed int, vector unsigned short);
11456 vector signed int vec_srl (vector signed int, vector unsigned char);
11457 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
11458 vector unsigned int vec_srl (vector unsigned int,
11459 vector unsigned short);
11460 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
11461 vector bool int vec_srl (vector bool int, vector unsigned int);
11462 vector bool int vec_srl (vector bool int, vector unsigned short);
11463 vector bool int vec_srl (vector bool int, vector unsigned char);
11464 vector signed short vec_srl (vector signed short, vector unsigned int);
11465 vector signed short vec_srl (vector signed short,
11466 vector unsigned short);
11467 vector signed short vec_srl (vector signed short, vector unsigned char);
11468 vector unsigned short vec_srl (vector unsigned short,
11469 vector unsigned int);
11470 vector unsigned short vec_srl (vector unsigned short,
11471 vector unsigned short);
11472 vector unsigned short vec_srl (vector unsigned short,
11473 vector unsigned char);
11474 vector bool short vec_srl (vector bool short, vector unsigned int);
11475 vector bool short vec_srl (vector bool short, vector unsigned short);
11476 vector bool short vec_srl (vector bool short, vector unsigned char);
11477 vector pixel vec_srl (vector pixel, vector unsigned int);
11478 vector pixel vec_srl (vector pixel, vector unsigned short);
11479 vector pixel vec_srl (vector pixel, vector unsigned char);
11480 vector signed char vec_srl (vector signed char, vector unsigned int);
11481 vector signed char vec_srl (vector signed char, vector unsigned short);
11482 vector signed char vec_srl (vector signed char, vector unsigned char);
11483 vector unsigned char vec_srl (vector unsigned char,
11484 vector unsigned int);
11485 vector unsigned char vec_srl (vector unsigned char,
11486 vector unsigned short);
11487 vector unsigned char vec_srl (vector unsigned char,
11488 vector unsigned char);
11489 vector bool char vec_srl (vector bool char, vector unsigned int);
11490 vector bool char vec_srl (vector bool char, vector unsigned short);
11491 vector bool char vec_srl (vector bool char, vector unsigned char);
11493 vector float vec_sro (vector float, vector signed char);
11494 vector float vec_sro (vector float, vector unsigned char);
11495 vector signed int vec_sro (vector signed int, vector signed char);
11496 vector signed int vec_sro (vector signed int, vector unsigned char);
11497 vector unsigned int vec_sro (vector unsigned int, vector signed char);
11498 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
11499 vector signed short vec_sro (vector signed short, vector signed char);
11500 vector signed short vec_sro (vector signed short, vector unsigned char);
11501 vector unsigned short vec_sro (vector unsigned short,
11502 vector signed char);
11503 vector unsigned short vec_sro (vector unsigned short,
11504 vector unsigned char);
11505 vector pixel vec_sro (vector pixel, vector signed char);
11506 vector pixel vec_sro (vector pixel, vector unsigned char);
11507 vector signed char vec_sro (vector signed char, vector signed char);
11508 vector signed char vec_sro (vector signed char, vector unsigned char);
11509 vector unsigned char vec_sro (vector unsigned char, vector signed char);
11510 vector unsigned char vec_sro (vector unsigned char,
11511 vector unsigned char);
11513 void vec_st (vector float, int, vector float *);
11514 void vec_st (vector float, int, float *);
11515 void vec_st (vector signed int, int, vector signed int *);
11516 void vec_st (vector signed int, int, int *);
11517 void vec_st (vector unsigned int, int, vector unsigned int *);
11518 void vec_st (vector unsigned int, int, unsigned int *);
11519 void vec_st (vector bool int, int, vector bool int *);
11520 void vec_st (vector bool int, int, unsigned int *);
11521 void vec_st (vector bool int, int, int *);
11522 void vec_st (vector signed short, int, vector signed short *);
11523 void vec_st (vector signed short, int, short *);
11524 void vec_st (vector unsigned short, int, vector unsigned short *);
11525 void vec_st (vector unsigned short, int, unsigned short *);
11526 void vec_st (vector bool short, int, vector bool short *);
11527 void vec_st (vector bool short, int, unsigned short *);
11528 void vec_st (vector pixel, int, vector pixel *);
11529 void vec_st (vector pixel, int, unsigned short *);
11530 void vec_st (vector pixel, int, short *);
11531 void vec_st (vector bool short, int, short *);
11532 void vec_st (vector signed char, int, vector signed char *);
11533 void vec_st (vector signed char, int, signed char *);
11534 void vec_st (vector unsigned char, int, vector unsigned char *);
11535 void vec_st (vector unsigned char, int, unsigned char *);
11536 void vec_st (vector bool char, int, vector bool char *);
11537 void vec_st (vector bool char, int, unsigned char *);
11538 void vec_st (vector bool char, int, signed char *);
11540 void vec_ste (vector signed char, int, signed char *);
11541 void vec_ste (vector unsigned char, int, unsigned char *);
11542 void vec_ste (vector bool char, int, signed char *);
11543 void vec_ste (vector bool char, int, unsigned char *);
11544 void vec_ste (vector signed short, int, short *);
11545 void vec_ste (vector unsigned short, int, unsigned short *);
11546 void vec_ste (vector bool short, int, short *);
11547 void vec_ste (vector bool short, int, unsigned short *);
11548 void vec_ste (vector pixel, int, short *);
11549 void vec_ste (vector pixel, int, unsigned short *);
11550 void vec_ste (vector float, int, float *);
11551 void vec_ste (vector signed int, int, int *);
11552 void vec_ste (vector unsigned int, int, unsigned int *);
11553 void vec_ste (vector bool int, int, int *);
11554 void vec_ste (vector bool int, int, unsigned int *);
11556 void vec_stvewx (vector float, int, float *);
11557 void vec_stvewx (vector signed int, int, int *);
11558 void vec_stvewx (vector unsigned int, int, unsigned int *);
11559 void vec_stvewx (vector bool int, int, int *);
11560 void vec_stvewx (vector bool int, int, unsigned int *);
11562 void vec_stvehx (vector signed short, int, short *);
11563 void vec_stvehx (vector unsigned short, int, unsigned short *);
11564 void vec_stvehx (vector bool short, int, short *);
11565 void vec_stvehx (vector bool short, int, unsigned short *);
11566 void vec_stvehx (vector pixel, int, short *);
11567 void vec_stvehx (vector pixel, int, unsigned short *);
11569 void vec_stvebx (vector signed char, int, signed char *);
11570 void vec_stvebx (vector unsigned char, int, unsigned char *);
11571 void vec_stvebx (vector bool char, int, signed char *);
11572 void vec_stvebx (vector bool char, int, unsigned char *);
11574 void vec_stl (vector float, int, vector float *);
11575 void vec_stl (vector float, int, float *);
11576 void vec_stl (vector signed int, int, vector signed int *);
11577 void vec_stl (vector signed int, int, int *);
11578 void vec_stl (vector unsigned int, int, vector unsigned int *);
11579 void vec_stl (vector unsigned int, int, unsigned int *);
11580 void vec_stl (vector bool int, int, vector bool int *);
11581 void vec_stl (vector bool int, int, unsigned int *);
11582 void vec_stl (vector bool int, int, int *);
11583 void vec_stl (vector signed short, int, vector signed short *);
11584 void vec_stl (vector signed short, int, short *);
11585 void vec_stl (vector unsigned short, int, vector unsigned short *);
11586 void vec_stl (vector unsigned short, int, unsigned short *);
11587 void vec_stl (vector bool short, int, vector bool short *);
11588 void vec_stl (vector bool short, int, unsigned short *);
11589 void vec_stl (vector bool short, int, short *);
11590 void vec_stl (vector pixel, int, vector pixel *);
11591 void vec_stl (vector pixel, int, unsigned short *);
11592 void vec_stl (vector pixel, int, short *);
11593 void vec_stl (vector signed char, int, vector signed char *);
11594 void vec_stl (vector signed char, int, signed char *);
11595 void vec_stl (vector unsigned char, int, vector unsigned char *);
11596 void vec_stl (vector unsigned char, int, unsigned char *);
11597 void vec_stl (vector bool char, int, vector bool char *);
11598 void vec_stl (vector bool char, int, unsigned char *);
11599 void vec_stl (vector bool char, int, signed char *);
11601 vector signed char vec_sub (vector bool char, vector signed char);
11602 vector signed char vec_sub (vector signed char, vector bool char);
11603 vector signed char vec_sub (vector signed char, vector signed char);
11604 vector unsigned char vec_sub (vector bool char, vector unsigned char);
11605 vector unsigned char vec_sub (vector unsigned char, vector bool char);
11606 vector unsigned char vec_sub (vector unsigned char,
11607 vector unsigned char);
11608 vector signed short vec_sub (vector bool short, vector signed short);
11609 vector signed short vec_sub (vector signed short, vector bool short);
11610 vector signed short vec_sub (vector signed short, vector signed short);
11611 vector unsigned short vec_sub (vector bool short,
11612 vector unsigned short);
11613 vector unsigned short vec_sub (vector unsigned short,
11614 vector bool short);
11615 vector unsigned short vec_sub (vector unsigned short,
11616 vector unsigned short);
11617 vector signed int vec_sub (vector bool int, vector signed int);
11618 vector signed int vec_sub (vector signed int, vector bool int);
11619 vector signed int vec_sub (vector signed int, vector signed int);
11620 vector unsigned int vec_sub (vector bool int, vector unsigned int);
11621 vector unsigned int vec_sub (vector unsigned int, vector bool int);
11622 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
11623 vector float vec_sub (vector float, vector float);
11625 vector float vec_vsubfp (vector float, vector float);
11627 vector signed int vec_vsubuwm (vector bool int, vector signed int);
11628 vector signed int vec_vsubuwm (vector signed int, vector bool int);
11629 vector signed int vec_vsubuwm (vector signed int, vector signed int);
11630 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
11631 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
11632 vector unsigned int vec_vsubuwm (vector unsigned int,
11633 vector unsigned int);
11635 vector signed short vec_vsubuhm (vector bool short,
11636 vector signed short);
11637 vector signed short vec_vsubuhm (vector signed short,
11638 vector bool short);
11639 vector signed short vec_vsubuhm (vector signed short,
11640 vector signed short);
11641 vector unsigned short vec_vsubuhm (vector bool short,
11642 vector unsigned short);
11643 vector unsigned short vec_vsubuhm (vector unsigned short,
11644 vector bool short);
11645 vector unsigned short vec_vsubuhm (vector unsigned short,
11646 vector unsigned short);
11648 vector signed char vec_vsububm (vector bool char, vector signed char);
11649 vector signed char vec_vsububm (vector signed char, vector bool char);
11650 vector signed char vec_vsububm (vector signed char, vector signed char);
11651 vector unsigned char vec_vsububm (vector bool char,
11652 vector unsigned char);
11653 vector unsigned char vec_vsububm (vector unsigned char,
11655 vector unsigned char vec_vsububm (vector unsigned char,
11656 vector unsigned char);
11658 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
11660 vector unsigned char vec_subs (vector bool char, vector unsigned char);
11661 vector unsigned char vec_subs (vector unsigned char, vector bool char);
11662 vector unsigned char vec_subs (vector unsigned char,
11663 vector unsigned char);
11664 vector signed char vec_subs (vector bool char, vector signed char);
11665 vector signed char vec_subs (vector signed char, vector bool char);
11666 vector signed char vec_subs (vector signed char, vector signed char);
11667 vector unsigned short vec_subs (vector bool short,
11668 vector unsigned short);
11669 vector unsigned short vec_subs (vector unsigned short,
11670 vector bool short);
11671 vector unsigned short vec_subs (vector unsigned short,
11672 vector unsigned short);
11673 vector signed short vec_subs (vector bool short, vector signed short);
11674 vector signed short vec_subs (vector signed short, vector bool short);
11675 vector signed short vec_subs (vector signed short, vector signed short);
11676 vector unsigned int vec_subs (vector bool int, vector unsigned int);
11677 vector unsigned int vec_subs (vector unsigned int, vector bool int);
11678 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
11679 vector signed int vec_subs (vector bool int, vector signed int);
11680 vector signed int vec_subs (vector signed int, vector bool int);
11681 vector signed int vec_subs (vector signed int, vector signed int);
11683 vector signed int vec_vsubsws (vector bool int, vector signed int);
11684 vector signed int vec_vsubsws (vector signed int, vector bool int);
11685 vector signed int vec_vsubsws (vector signed int, vector signed int);
11687 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
11688 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
11689 vector unsigned int vec_vsubuws (vector unsigned int,
11690 vector unsigned int);
11692 vector signed short vec_vsubshs (vector bool short,
11693 vector signed short);
11694 vector signed short vec_vsubshs (vector signed short,
11695 vector bool short);
11696 vector signed short vec_vsubshs (vector signed short,
11697 vector signed short);
11699 vector unsigned short vec_vsubuhs (vector bool short,
11700 vector unsigned short);
11701 vector unsigned short vec_vsubuhs (vector unsigned short,
11702 vector bool short);
11703 vector unsigned short vec_vsubuhs (vector unsigned short,
11704 vector unsigned short);
11706 vector signed char vec_vsubsbs (vector bool char, vector signed char);
11707 vector signed char vec_vsubsbs (vector signed char, vector bool char);
11708 vector signed char vec_vsubsbs (vector signed char, vector signed char);
11710 vector unsigned char vec_vsububs (vector bool char,
11711 vector unsigned char);
11712 vector unsigned char vec_vsububs (vector unsigned char,
11714 vector unsigned char vec_vsububs (vector unsigned char,
11715 vector unsigned char);
11717 vector unsigned int vec_sum4s (vector unsigned char,
11718 vector unsigned int);
11719 vector signed int vec_sum4s (vector signed char, vector signed int);
11720 vector signed int vec_sum4s (vector signed short, vector signed int);
11722 vector signed int vec_vsum4shs (vector signed short, vector signed int);
11724 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
11726 vector unsigned int vec_vsum4ubs (vector unsigned char,
11727 vector unsigned int);
11729 vector signed int vec_sum2s (vector signed int, vector signed int);
11731 vector signed int vec_sums (vector signed int, vector signed int);
11733 vector float vec_trunc (vector float);
11735 vector signed short vec_unpackh (vector signed char);
11736 vector bool short vec_unpackh (vector bool char);
11737 vector signed int vec_unpackh (vector signed short);
11738 vector bool int vec_unpackh (vector bool short);
11739 vector unsigned int vec_unpackh (vector pixel);
11741 vector bool int vec_vupkhsh (vector bool short);
11742 vector signed int vec_vupkhsh (vector signed short);
11744 vector unsigned int vec_vupkhpx (vector pixel);
11746 vector bool short vec_vupkhsb (vector bool char);
11747 vector signed short vec_vupkhsb (vector signed char);
11749 vector signed short vec_unpackl (vector signed char);
11750 vector bool short vec_unpackl (vector bool char);
11751 vector unsigned int vec_unpackl (vector pixel);
11752 vector signed int vec_unpackl (vector signed short);
11753 vector bool int vec_unpackl (vector bool short);
11755 vector unsigned int vec_vupklpx (vector pixel);
11757 vector bool int vec_vupklsh (vector bool short);
11758 vector signed int vec_vupklsh (vector signed short);
11760 vector bool short vec_vupklsb (vector bool char);
11761 vector signed short vec_vupklsb (vector signed char);
11763 vector float vec_xor (vector float, vector float);
11764 vector float vec_xor (vector float, vector bool int);
11765 vector float vec_xor (vector bool int, vector float);
11766 vector bool int vec_xor (vector bool int, vector bool int);
11767 vector signed int vec_xor (vector bool int, vector signed int);
11768 vector signed int vec_xor (vector signed int, vector bool int);
11769 vector signed int vec_xor (vector signed int, vector signed int);
11770 vector unsigned int vec_xor (vector bool int, vector unsigned int);
11771 vector unsigned int vec_xor (vector unsigned int, vector bool int);
11772 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
11773 vector bool short vec_xor (vector bool short, vector bool short);
11774 vector signed short vec_xor (vector bool short, vector signed short);
11775 vector signed short vec_xor (vector signed short, vector bool short);
11776 vector signed short vec_xor (vector signed short, vector signed short);
11777 vector unsigned short vec_xor (vector bool short,
11778 vector unsigned short);
11779 vector unsigned short vec_xor (vector unsigned short,
11780 vector bool short);
11781 vector unsigned short vec_xor (vector unsigned short,
11782 vector unsigned short);
11783 vector signed char vec_xor (vector bool char, vector signed char);
11784 vector bool char vec_xor (vector bool char, vector bool char);
11785 vector signed char vec_xor (vector signed char, vector bool char);
11786 vector signed char vec_xor (vector signed char, vector signed char);
11787 vector unsigned char vec_xor (vector bool char, vector unsigned char);
11788 vector unsigned char vec_xor (vector unsigned char, vector bool char);
11789 vector unsigned char vec_xor (vector unsigned char,
11790 vector unsigned char);
11792 int vec_all_eq (vector signed char, vector bool char);
11793 int vec_all_eq (vector signed char, vector signed char);
11794 int vec_all_eq (vector unsigned char, vector bool char);
11795 int vec_all_eq (vector unsigned char, vector unsigned char);
11796 int vec_all_eq (vector bool char, vector bool char);
11797 int vec_all_eq (vector bool char, vector unsigned char);
11798 int vec_all_eq (vector bool char, vector signed char);
11799 int vec_all_eq (vector signed short, vector bool short);
11800 int vec_all_eq (vector signed short, vector signed short);
11801 int vec_all_eq (vector unsigned short, vector bool short);
11802 int vec_all_eq (vector unsigned short, vector unsigned short);
11803 int vec_all_eq (vector bool short, vector bool short);
11804 int vec_all_eq (vector bool short, vector unsigned short);
11805 int vec_all_eq (vector bool short, vector signed short);
11806 int vec_all_eq (vector pixel, vector pixel);
11807 int vec_all_eq (vector signed int, vector bool int);
11808 int vec_all_eq (vector signed int, vector signed int);
11809 int vec_all_eq (vector unsigned int, vector bool int);
11810 int vec_all_eq (vector unsigned int, vector unsigned int);
11811 int vec_all_eq (vector bool int, vector bool int);
11812 int vec_all_eq (vector bool int, vector unsigned int);
11813 int vec_all_eq (vector bool int, vector signed int);
11814 int vec_all_eq (vector float, vector float);
11816 int vec_all_ge (vector bool char, vector unsigned char);
11817 int vec_all_ge (vector unsigned char, vector bool char);
11818 int vec_all_ge (vector unsigned char, vector unsigned char);
11819 int vec_all_ge (vector bool char, vector signed char);
11820 int vec_all_ge (vector signed char, vector bool char);
11821 int vec_all_ge (vector signed char, vector signed char);
11822 int vec_all_ge (vector bool short, vector unsigned short);
11823 int vec_all_ge (vector unsigned short, vector bool short);
11824 int vec_all_ge (vector unsigned short, vector unsigned short);
11825 int vec_all_ge (vector signed short, vector signed short);
11826 int vec_all_ge (vector bool short, vector signed short);
11827 int vec_all_ge (vector signed short, vector bool short);
11828 int vec_all_ge (vector bool int, vector unsigned int);
11829 int vec_all_ge (vector unsigned int, vector bool int);
11830 int vec_all_ge (vector unsigned int, vector unsigned int);
11831 int vec_all_ge (vector bool int, vector signed int);
11832 int vec_all_ge (vector signed int, vector bool int);
11833 int vec_all_ge (vector signed int, vector signed int);
11834 int vec_all_ge (vector float, vector float);
11836 int vec_all_gt (vector bool char, vector unsigned char);
11837 int vec_all_gt (vector unsigned char, vector bool char);
11838 int vec_all_gt (vector unsigned char, vector unsigned char);
11839 int vec_all_gt (vector bool char, vector signed char);
11840 int vec_all_gt (vector signed char, vector bool char);
11841 int vec_all_gt (vector signed char, vector signed char);
11842 int vec_all_gt (vector bool short, vector unsigned short);
11843 int vec_all_gt (vector unsigned short, vector bool short);
11844 int vec_all_gt (vector unsigned short, vector unsigned short);
11845 int vec_all_gt (vector bool short, vector signed short);
11846 int vec_all_gt (vector signed short, vector bool short);
11847 int vec_all_gt (vector signed short, vector signed short);
11848 int vec_all_gt (vector bool int, vector unsigned int);
11849 int vec_all_gt (vector unsigned int, vector bool int);
11850 int vec_all_gt (vector unsigned int, vector unsigned int);
11851 int vec_all_gt (vector bool int, vector signed int);
11852 int vec_all_gt (vector signed int, vector bool int);
11853 int vec_all_gt (vector signed int, vector signed int);
11854 int vec_all_gt (vector float, vector float);
11856 int vec_all_in (vector float, vector float);
11858 int vec_all_le (vector bool char, vector unsigned char);
11859 int vec_all_le (vector unsigned char, vector bool char);
11860 int vec_all_le (vector unsigned char, vector unsigned char);
11861 int vec_all_le (vector bool char, vector signed char);
11862 int vec_all_le (vector signed char, vector bool char);
11863 int vec_all_le (vector signed char, vector signed char);
11864 int vec_all_le (vector bool short, vector unsigned short);
11865 int vec_all_le (vector unsigned short, vector bool short);
11866 int vec_all_le (vector unsigned short, vector unsigned short);
11867 int vec_all_le (vector bool short, vector signed short);
11868 int vec_all_le (vector signed short, vector bool short);
11869 int vec_all_le (vector signed short, vector signed short);
11870 int vec_all_le (vector bool int, vector unsigned int);
11871 int vec_all_le (vector unsigned int, vector bool int);
11872 int vec_all_le (vector unsigned int, vector unsigned int);
11873 int vec_all_le (vector bool int, vector signed int);
11874 int vec_all_le (vector signed int, vector bool int);
11875 int vec_all_le (vector signed int, vector signed int);
11876 int vec_all_le (vector float, vector float);
11878 int vec_all_lt (vector bool char, vector unsigned char);
11879 int vec_all_lt (vector unsigned char, vector bool char);
11880 int vec_all_lt (vector unsigned char, vector unsigned char);
11881 int vec_all_lt (vector bool char, vector signed char);
11882 int vec_all_lt (vector signed char, vector bool char);
11883 int vec_all_lt (vector signed char, vector signed char);
11884 int vec_all_lt (vector bool short, vector unsigned short);
11885 int vec_all_lt (vector unsigned short, vector bool short);
11886 int vec_all_lt (vector unsigned short, vector unsigned short);
11887 int vec_all_lt (vector bool short, vector signed short);
11888 int vec_all_lt (vector signed short, vector bool short);
11889 int vec_all_lt (vector signed short, vector signed short);
11890 int vec_all_lt (vector bool int, vector unsigned int);
11891 int vec_all_lt (vector unsigned int, vector bool int);
11892 int vec_all_lt (vector unsigned int, vector unsigned int);
11893 int vec_all_lt (vector bool int, vector signed int);
11894 int vec_all_lt (vector signed int, vector bool int);
11895 int vec_all_lt (vector signed int, vector signed int);
11896 int vec_all_lt (vector float, vector float);
11898 int vec_all_nan (vector float);
11900 int vec_all_ne (vector signed char, vector bool char);
11901 int vec_all_ne (vector signed char, vector signed char);
11902 int vec_all_ne (vector unsigned char, vector bool char);
11903 int vec_all_ne (vector unsigned char, vector unsigned char);
11904 int vec_all_ne (vector bool char, vector bool char);
11905 int vec_all_ne (vector bool char, vector unsigned char);
11906 int vec_all_ne (vector bool char, vector signed char);
11907 int vec_all_ne (vector signed short, vector bool short);
11908 int vec_all_ne (vector signed short, vector signed short);
11909 int vec_all_ne (vector unsigned short, vector bool short);
11910 int vec_all_ne (vector unsigned short, vector unsigned short);
11911 int vec_all_ne (vector bool short, vector bool short);
11912 int vec_all_ne (vector bool short, vector unsigned short);
11913 int vec_all_ne (vector bool short, vector signed short);
11914 int vec_all_ne (vector pixel, vector pixel);
11915 int vec_all_ne (vector signed int, vector bool int);
11916 int vec_all_ne (vector signed int, vector signed int);
11917 int vec_all_ne (vector unsigned int, vector bool int);
11918 int vec_all_ne (vector unsigned int, vector unsigned int);
11919 int vec_all_ne (vector bool int, vector bool int);
11920 int vec_all_ne (vector bool int, vector unsigned int);
11921 int vec_all_ne (vector bool int, vector signed int);
11922 int vec_all_ne (vector float, vector float);
11924 int vec_all_nge (vector float, vector float);
11926 int vec_all_ngt (vector float, vector float);
11928 int vec_all_nle (vector float, vector float);
11930 int vec_all_nlt (vector float, vector float);
11932 int vec_all_numeric (vector float);
11934 int vec_any_eq (vector signed char, vector bool char);
11935 int vec_any_eq (vector signed char, vector signed char);
11936 int vec_any_eq (vector unsigned char, vector bool char);
11937 int vec_any_eq (vector unsigned char, vector unsigned char);
11938 int vec_any_eq (vector bool char, vector bool char);
11939 int vec_any_eq (vector bool char, vector unsigned char);
11940 int vec_any_eq (vector bool char, vector signed char);
11941 int vec_any_eq (vector signed short, vector bool short);
11942 int vec_any_eq (vector signed short, vector signed short);
11943 int vec_any_eq (vector unsigned short, vector bool short);
11944 int vec_any_eq (vector unsigned short, vector unsigned short);
11945 int vec_any_eq (vector bool short, vector bool short);
11946 int vec_any_eq (vector bool short, vector unsigned short);
11947 int vec_any_eq (vector bool short, vector signed short);
11948 int vec_any_eq (vector pixel, vector pixel);
11949 int vec_any_eq (vector signed int, vector bool int);
11950 int vec_any_eq (vector signed int, vector signed int);
11951 int vec_any_eq (vector unsigned int, vector bool int);
11952 int vec_any_eq (vector unsigned int, vector unsigned int);
11953 int vec_any_eq (vector bool int, vector bool int);
11954 int vec_any_eq (vector bool int, vector unsigned int);
11955 int vec_any_eq (vector bool int, vector signed int);
11956 int vec_any_eq (vector float, vector float);
11958 int vec_any_ge (vector signed char, vector bool char);
11959 int vec_any_ge (vector unsigned char, vector bool char);
11960 int vec_any_ge (vector unsigned char, vector unsigned char);
11961 int vec_any_ge (vector signed char, vector signed char);
11962 int vec_any_ge (vector bool char, vector unsigned char);
11963 int vec_any_ge (vector bool char, vector signed char);
11964 int vec_any_ge (vector unsigned short, vector bool short);
11965 int vec_any_ge (vector unsigned short, vector unsigned short);
11966 int vec_any_ge (vector signed short, vector signed short);
11967 int vec_any_ge (vector signed short, vector bool short);
11968 int vec_any_ge (vector bool short, vector unsigned short);
11969 int vec_any_ge (vector bool short, vector signed short);
11970 int vec_any_ge (vector signed int, vector bool int);
11971 int vec_any_ge (vector unsigned int, vector bool int);
11972 int vec_any_ge (vector unsigned int, vector unsigned int);
11973 int vec_any_ge (vector signed int, vector signed int);
11974 int vec_any_ge (vector bool int, vector unsigned int);
11975 int vec_any_ge (vector bool int, vector signed int);
11976 int vec_any_ge (vector float, vector float);
11978 int vec_any_gt (vector bool char, vector unsigned char);
11979 int vec_any_gt (vector unsigned char, vector bool char);
11980 int vec_any_gt (vector unsigned char, vector unsigned char);
11981 int vec_any_gt (vector bool char, vector signed char);
11982 int vec_any_gt (vector signed char, vector bool char);
11983 int vec_any_gt (vector signed char, vector signed char);
11984 int vec_any_gt (vector bool short, vector unsigned short);
11985 int vec_any_gt (vector unsigned short, vector bool short);
11986 int vec_any_gt (vector unsigned short, vector unsigned short);
11987 int vec_any_gt (vector bool short, vector signed short);
11988 int vec_any_gt (vector signed short, vector bool short);
11989 int vec_any_gt (vector signed short, vector signed short);
11990 int vec_any_gt (vector bool int, vector unsigned int);
11991 int vec_any_gt (vector unsigned int, vector bool int);
11992 int vec_any_gt (vector unsigned int, vector unsigned int);
11993 int vec_any_gt (vector bool int, vector signed int);
11994 int vec_any_gt (vector signed int, vector bool int);
11995 int vec_any_gt (vector signed int, vector signed int);
11996 int vec_any_gt (vector float, vector float);
11998 int vec_any_le (vector bool char, vector unsigned char);
11999 int vec_any_le (vector unsigned char, vector bool char);
12000 int vec_any_le (vector unsigned char, vector unsigned char);
12001 int vec_any_le (vector bool char, vector signed char);
12002 int vec_any_le (vector signed char, vector bool char);
12003 int vec_any_le (vector signed char, vector signed char);
12004 int vec_any_le (vector bool short, vector unsigned short);
12005 int vec_any_le (vector unsigned short, vector bool short);
12006 int vec_any_le (vector unsigned short, vector unsigned short);
12007 int vec_any_le (vector bool short, vector signed short);
12008 int vec_any_le (vector signed short, vector bool short);
12009 int vec_any_le (vector signed short, vector signed short);
12010 int vec_any_le (vector bool int, vector unsigned int);
12011 int vec_any_le (vector unsigned int, vector bool int);
12012 int vec_any_le (vector unsigned int, vector unsigned int);
12013 int vec_any_le (vector bool int, vector signed int);
12014 int vec_any_le (vector signed int, vector bool int);
12015 int vec_any_le (vector signed int, vector signed int);
12016 int vec_any_le (vector float, vector float);
12018 int vec_any_lt (vector bool char, vector unsigned char);
12019 int vec_any_lt (vector unsigned char, vector bool char);
12020 int vec_any_lt (vector unsigned char, vector unsigned char);
12021 int vec_any_lt (vector bool char, vector signed char);
12022 int vec_any_lt (vector signed char, vector bool char);
12023 int vec_any_lt (vector signed char, vector signed char);
12024 int vec_any_lt (vector bool short, vector unsigned short);
12025 int vec_any_lt (vector unsigned short, vector bool short);
12026 int vec_any_lt (vector unsigned short, vector unsigned short);
12027 int vec_any_lt (vector bool short, vector signed short);
12028 int vec_any_lt (vector signed short, vector bool short);
12029 int vec_any_lt (vector signed short, vector signed short);
12030 int vec_any_lt (vector bool int, vector unsigned int);
12031 int vec_any_lt (vector unsigned int, vector bool int);
12032 int vec_any_lt (vector unsigned int, vector unsigned int);
12033 int vec_any_lt (vector bool int, vector signed int);
12034 int vec_any_lt (vector signed int, vector bool int);
12035 int vec_any_lt (vector signed int, vector signed int);
12036 int vec_any_lt (vector float, vector float);
12038 int vec_any_nan (vector float);
12040 int vec_any_ne (vector signed char, vector bool char);
12041 int vec_any_ne (vector signed char, vector signed char);
12042 int vec_any_ne (vector unsigned char, vector bool char);
12043 int vec_any_ne (vector unsigned char, vector unsigned char);
12044 int vec_any_ne (vector bool char, vector bool char);
12045 int vec_any_ne (vector bool char, vector unsigned char);
12046 int vec_any_ne (vector bool char, vector signed char);
12047 int vec_any_ne (vector signed short, vector bool short);
12048 int vec_any_ne (vector signed short, vector signed short);
12049 int vec_any_ne (vector unsigned short, vector bool short);
12050 int vec_any_ne (vector unsigned short, vector unsigned short);
12051 int vec_any_ne (vector bool short, vector bool short);
12052 int vec_any_ne (vector bool short, vector unsigned short);
12053 int vec_any_ne (vector bool short, vector signed short);
12054 int vec_any_ne (vector pixel, vector pixel);
12055 int vec_any_ne (vector signed int, vector bool int);
12056 int vec_any_ne (vector signed int, vector signed int);
12057 int vec_any_ne (vector unsigned int, vector bool int);
12058 int vec_any_ne (vector unsigned int, vector unsigned int);
12059 int vec_any_ne (vector bool int, vector bool int);
12060 int vec_any_ne (vector bool int, vector unsigned int);
12061 int vec_any_ne (vector bool int, vector signed int);
12062 int vec_any_ne (vector float, vector float);
12064 int vec_any_nge (vector float, vector float);
12066 int vec_any_ngt (vector float, vector float);
12068 int vec_any_nle (vector float, vector float);
12070 int vec_any_nlt (vector float, vector float);
12072 int vec_any_numeric (vector float);
12074 int vec_any_out (vector float, vector float);
12077 If the vector/scalar (VSX) instruction set is available, the following
12078 additional functions are available:
12081 vector double vec_abs (vector double);
12082 vector double vec_add (vector double, vector double);
12083 vector double vec_and (vector double, vector double);
12084 vector double vec_and (vector double, vector bool long);
12085 vector double vec_and (vector bool long, vector double);
12086 vector double vec_andc (vector double, vector double);
12087 vector double vec_andc (vector double, vector bool long);
12088 vector double vec_andc (vector bool long, vector double);
12089 vector double vec_ceil (vector double);
12090 vector bool long vec_cmpeq (vector double, vector double);
12091 vector bool long vec_cmpge (vector double, vector double);
12092 vector bool long vec_cmpgt (vector double, vector double);
12093 vector bool long vec_cmple (vector double, vector double);
12094 vector bool long vec_cmplt (vector double, vector double);
12095 vector float vec_div (vector float, vector float);
12096 vector double vec_div (vector double, vector double);
12097 vector double vec_floor (vector double);
12098 vector double vec_madd (vector double, vector double, vector double);
12099 vector double vec_max (vector double, vector double);
12100 vector double vec_min (vector double, vector double);
12101 vector float vec_msub (vector float, vector float, vector float);
12102 vector double vec_msub (vector double, vector double, vector double);
12103 vector float vec_mul (vector float, vector float);
12104 vector double vec_mul (vector double, vector double);
12105 vector float vec_nearbyint (vector float);
12106 vector double vec_nearbyint (vector double);
12107 vector float vec_nmadd (vector float, vector float, vector float);
12108 vector double vec_nmadd (vector double, vector double, vector double);
12109 vector double vec_nmsub (vector double, vector double, vector double);
12110 vector double vec_nor (vector double, vector double);
12111 vector double vec_or (vector double, vector double);
12112 vector double vec_or (vector double, vector bool long);
12113 vector double vec_or (vector bool long, vector double);
12114 vector double vec_perm (vector double,
12116 vector unsigned char);
12117 vector double vec_rint (vector double);
12118 vector double vec_recip (vector double, vector double);
12119 vector double vec_rsqrt (vector double);
12120 vector double vec_rsqrte (vector double);
12121 vector double vec_sel (vector double, vector double, vector bool long);
12122 vector double vec_sel (vector double, vector double, vector unsigned long);
12123 vector double vec_sub (vector double, vector double);
12124 vector float vec_sqrt (vector float);
12125 vector double vec_sqrt (vector double);
12126 vector double vec_trunc (vector double);
12127 vector double vec_xor (vector double, vector double);
12128 vector double vec_xor (vector double, vector bool long);
12129 vector double vec_xor (vector bool long, vector double);
12130 int vec_all_eq (vector double, vector double);
12131 int vec_all_ge (vector double, vector double);
12132 int vec_all_gt (vector double, vector double);
12133 int vec_all_le (vector double, vector double);
12134 int vec_all_lt (vector double, vector double);
12135 int vec_all_nan (vector double);
12136 int vec_all_ne (vector double, vector double);
12137 int vec_all_nge (vector double, vector double);
12138 int vec_all_ngt (vector double, vector double);
12139 int vec_all_nle (vector double, vector double);
12140 int vec_all_nlt (vector double, vector double);
12141 int vec_all_numeric (vector double);
12142 int vec_any_eq (vector double, vector double);
12143 int vec_any_ge (vector double, vector double);
12144 int vec_any_gt (vector double, vector double);
12145 int vec_any_le (vector double, vector double);
12146 int vec_any_lt (vector double, vector double);
12147 int vec_any_nan (vector double);
12148 int vec_any_ne (vector double, vector double);
12149 int vec_any_nge (vector double, vector double);
12150 int vec_any_ngt (vector double, vector double);
12151 int vec_any_nle (vector double, vector double);
12152 int vec_any_nlt (vector double, vector double);
12153 int vec_any_numeric (vector double);
12156 GCC provides a few other builtins on Powerpc to access certain instructions:
12158 float __builtin_recipdivf (float, float);
12159 float __builtin_rsqrtf (float);
12160 double __builtin_recipdiv (double, double);
12161 double __builtin_rsqrt (double);
12162 long __builtin_bpermd (long, long);
12163 int __builtin_bswap16 (int);
12166 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
12167 @code{__builtin_rsqrtf} functions generate multiple instructions to
12168 implement the reciprocal sqrt functionality using reciprocal sqrt
12169 estimate instructions.
12171 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
12172 functions generate multiple instructions to implement division using
12173 the reciprocal estimate instructions.
12175 @node RX Built-in Functions
12176 @subsection RX Built-in Functions
12177 GCC supports some of the RX instructions which cannot be expressed in
12178 the C programming language via the use of built-in functions. The
12179 following functions are supported:
12181 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
12182 Generates the @code{brk} machine instruction.
12185 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
12186 Generates the @code{clrpsw} machine instruction to clear the specified
12187 bit in the processor status word.
12190 @deftypefn {Built-in Function} void __builtin_rx_int (int)
12191 Generates the @code{int} machine instruction to generate an interrupt
12192 with the specified value.
12195 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
12196 Generates the @code{machi} machine instruction to add the result of
12197 multiplying the top 16-bits of the two arguments into the
12201 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
12202 Generates the @code{maclo} machine instruction to add the result of
12203 multiplying the bottom 16-bits of the two arguments into the
12207 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
12208 Generates the @code{mulhi} machine instruction to place the result of
12209 multiplying the top 16-bits of the two arguments into the
12213 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
12214 Generates the @code{mullo} machine instruction to place the result of
12215 multiplying the bottom 16-bits of the two arguments into the
12219 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
12220 Generates the @code{mvfachi} machine instruction to read the top
12221 32-bits of the accumulator.
12224 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
12225 Generates the @code{mvfacmi} machine instruction to read the middle
12226 32-bits of the accumulator.
12229 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
12230 Generates the @code{mvfc} machine instruction which reads the control
12231 register specified in its argument and returns its value.
12234 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
12235 Generates the @code{mvtachi} machine instruction to set the top
12236 32-bits of the accumulator.
12239 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
12240 Generates the @code{mvtaclo} machine instruction to set the bottom
12241 32-bits of the accumulator.
12244 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
12245 Generates the @code{mvtc} machine instruction which sets control
12246 register number @code{reg} to @code{val}.
12249 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
12250 Generates the @code{mvtipl} machine instruction set the interrupt
12254 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
12255 Generates the @code{racw} machine instruction to round the accumulator
12256 according to the specified mode.
12259 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
12260 Generates the @code{revw} machine instruction which swaps the bytes in
12261 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
12262 and also bits 16--23 occupy bits 24--31 and vice versa.
12265 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
12266 Generates the @code{rmpa} machine instruction which initiates a
12267 repeated multiply and accumulate sequence.
12270 @deftypefn {Built-in Function} void __builtin_rx_round (float)
12271 Generates the @code{round} machine instruction which returns the
12272 floating point argument rounded according to the current rounding mode
12273 set in the floating point status word register.
12276 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
12277 Generates the @code{sat} machine instruction which returns the
12278 saturated value of the argument.
12281 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
12282 Generates the @code{setpsw} machine instruction to set the specified
12283 bit in the processor status word.
12286 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
12287 Generates the @code{wait} machine instruction.
12290 @node SPARC VIS Built-in Functions
12291 @subsection SPARC VIS Built-in Functions
12293 GCC supports SIMD operations on the SPARC using both the generic vector
12294 extensions (@pxref{Vector Extensions}) as well as built-in functions for
12295 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
12296 switch, the VIS extension is exposed as the following built-in functions:
12299 typedef int v2si __attribute__ ((vector_size (8)));
12300 typedef short v4hi __attribute__ ((vector_size (8)));
12301 typedef short v2hi __attribute__ ((vector_size (4)));
12302 typedef char v8qi __attribute__ ((vector_size (8)));
12303 typedef char v4qi __attribute__ ((vector_size (4)));
12305 void * __builtin_vis_alignaddr (void *, long);
12306 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
12307 v2si __builtin_vis_faligndatav2si (v2si, v2si);
12308 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
12309 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
12311 v4hi __builtin_vis_fexpand (v4qi);
12313 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
12314 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
12315 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
12316 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
12317 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
12318 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
12319 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
12321 v4qi __builtin_vis_fpack16 (v4hi);
12322 v8qi __builtin_vis_fpack32 (v2si, v2si);
12323 v2hi __builtin_vis_fpackfix (v2si);
12324 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
12326 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
12329 @node SPU Built-in Functions
12330 @subsection SPU Built-in Functions
12332 GCC provides extensions for the SPU processor as described in the
12333 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
12334 found at @uref{http://cell.scei.co.jp/} or
12335 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
12336 implementation differs in several ways.
12341 The optional extension of specifying vector constants in parentheses is
12345 A vector initializer requires no cast if the vector constant is of the
12346 same type as the variable it is initializing.
12349 If @code{signed} or @code{unsigned} is omitted, the signedness of the
12350 vector type is the default signedness of the base type. The default
12351 varies depending on the operating system, so a portable program should
12352 always specify the signedness.
12355 By default, the keyword @code{__vector} is added. The macro
12356 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
12360 GCC allows using a @code{typedef} name as the type specifier for a
12364 For C, overloaded functions are implemented with macros so the following
12368 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
12371 Since @code{spu_add} is a macro, the vector constant in the example
12372 is treated as four separate arguments. Wrap the entire argument in
12373 parentheses for this to work.
12376 The extended version of @code{__builtin_expect} is not supported.
12380 @emph{Note:} Only the interface described in the aforementioned
12381 specification is supported. Internally, GCC uses built-in functions to
12382 implement the required functionality, but these are not supported and
12383 are subject to change without notice.
12385 @node Target Format Checks
12386 @section Format Checks Specific to Particular Target Machines
12388 For some target machines, GCC supports additional options to the
12390 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
12393 * Solaris Format Checks::
12396 @node Solaris Format Checks
12397 @subsection Solaris Format Checks
12399 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
12400 check. @code{cmn_err} accepts a subset of the standard @code{printf}
12401 conversions, and the two-argument @code{%b} conversion for displaying
12402 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
12405 @section Pragmas Accepted by GCC
12407 @cindex @code{#pragma}
12409 GCC supports several types of pragmas, primarily in order to compile
12410 code originally written for other compilers. Note that in general
12411 we do not recommend the use of pragmas; @xref{Function Attributes},
12412 for further explanation.
12418 * RS/6000 and PowerPC Pragmas::
12420 * Solaris Pragmas::
12421 * Symbol-Renaming Pragmas::
12422 * Structure-Packing Pragmas::
12424 * Diagnostic Pragmas::
12425 * Visibility Pragmas::
12426 * Push/Pop Macro Pragmas::
12427 * Function Specific Option Pragmas::
12431 @subsection ARM Pragmas
12433 The ARM target defines pragmas for controlling the default addition of
12434 @code{long_call} and @code{short_call} attributes to functions.
12435 @xref{Function Attributes}, for information about the effects of these
12440 @cindex pragma, long_calls
12441 Set all subsequent functions to have the @code{long_call} attribute.
12443 @item no_long_calls
12444 @cindex pragma, no_long_calls
12445 Set all subsequent functions to have the @code{short_call} attribute.
12447 @item long_calls_off
12448 @cindex pragma, long_calls_off
12449 Do not affect the @code{long_call} or @code{short_call} attributes of
12450 subsequent functions.
12454 @subsection M32C Pragmas
12457 @item GCC memregs @var{number}
12458 @cindex pragma, memregs
12459 Overrides the command-line option @code{-memregs=} for the current
12460 file. Use with care! This pragma must be before any function in the
12461 file, and mixing different memregs values in different objects may
12462 make them incompatible. This pragma is useful when a
12463 performance-critical function uses a memreg for temporary values,
12464 as it may allow you to reduce the number of memregs used.
12466 @item ADDRESS @var{name} @var{address}
12467 @cindex pragma, address
12468 For any declared symbols matching @var{name}, this does three things
12469 to that symbol: it forces the symbol to be located at the given
12470 address (a number), it forces the symbol to be volatile, and it
12471 changes the symbol's scope to be static. This pragma exists for
12472 compatibility with other compilers, but note that the common
12473 @code{1234H} numeric syntax is not supported (use @code{0x1234}
12477 #pragma ADDRESS port3 0x103
12484 @subsection MeP Pragmas
12488 @item custom io_volatile (on|off)
12489 @cindex pragma, custom io_volatile
12490 Overrides the command line option @code{-mio-volatile} for the current
12491 file. Note that for compatibility with future GCC releases, this
12492 option should only be used once before any @code{io} variables in each
12495 @item GCC coprocessor available @var{registers}
12496 @cindex pragma, coprocessor available
12497 Specifies which coprocessor registers are available to the register
12498 allocator. @var{registers} may be a single register, register range
12499 separated by ellipses, or comma-separated list of those. Example:
12502 #pragma GCC coprocessor available $c0...$c10, $c28
12505 @item GCC coprocessor call_saved @var{registers}
12506 @cindex pragma, coprocessor call_saved
12507 Specifies which coprocessor registers are to be saved and restored by
12508 any function using them. @var{registers} may be a single register,
12509 register range separated by ellipses, or comma-separated list of
12513 #pragma GCC coprocessor call_saved $c4...$c6, $c31
12516 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
12517 @cindex pragma, coprocessor subclass
12518 Creates and defines a register class. These register classes can be
12519 used by inline @code{asm} constructs. @var{registers} may be a single
12520 register, register range separated by ellipses, or comma-separated
12521 list of those. Example:
12524 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
12526 asm ("cpfoo %0" : "=B" (x));
12529 @item GCC disinterrupt @var{name} , @var{name} @dots{}
12530 @cindex pragma, disinterrupt
12531 For the named functions, the compiler adds code to disable interrupts
12532 for the duration of those functions. Any functions so named, which
12533 are not encountered in the source, cause a warning that the pragma was
12534 not used. Examples:
12537 #pragma disinterrupt foo
12538 #pragma disinterrupt bar, grill
12539 int foo () @{ @dots{} @}
12542 @item GCC call @var{name} , @var{name} @dots{}
12543 @cindex pragma, call
12544 For the named functions, the compiler always uses a register-indirect
12545 call model when calling the named functions. Examples:
12554 @node RS/6000 and PowerPC Pragmas
12555 @subsection RS/6000 and PowerPC Pragmas
12557 The RS/6000 and PowerPC targets define one pragma for controlling
12558 whether or not the @code{longcall} attribute is added to function
12559 declarations by default. This pragma overrides the @option{-mlongcall}
12560 option, but not the @code{longcall} and @code{shortcall} attributes.
12561 @xref{RS/6000 and PowerPC Options}, for more information about when long
12562 calls are and are not necessary.
12566 @cindex pragma, longcall
12567 Apply the @code{longcall} attribute to all subsequent function
12571 Do not apply the @code{longcall} attribute to subsequent function
12575 @c Describe h8300 pragmas here.
12576 @c Describe sh pragmas here.
12577 @c Describe v850 pragmas here.
12579 @node Darwin Pragmas
12580 @subsection Darwin Pragmas
12582 The following pragmas are available for all architectures running the
12583 Darwin operating system. These are useful for compatibility with other
12587 @item mark @var{tokens}@dots{}
12588 @cindex pragma, mark
12589 This pragma is accepted, but has no effect.
12591 @item options align=@var{alignment}
12592 @cindex pragma, options align
12593 This pragma sets the alignment of fields in structures. The values of
12594 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
12595 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
12596 properly; to restore the previous setting, use @code{reset} for the
12599 @item segment @var{tokens}@dots{}
12600 @cindex pragma, segment
12601 This pragma is accepted, but has no effect.
12603 @item unused (@var{var} [, @var{var}]@dots{})
12604 @cindex pragma, unused
12605 This pragma declares variables to be possibly unused. GCC will not
12606 produce warnings for the listed variables. The effect is similar to
12607 that of the @code{unused} attribute, except that this pragma may appear
12608 anywhere within the variables' scopes.
12611 @node Solaris Pragmas
12612 @subsection Solaris Pragmas
12614 The Solaris target supports @code{#pragma redefine_extname}
12615 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
12616 @code{#pragma} directives for compatibility with the system compiler.
12619 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
12620 @cindex pragma, align
12622 Increase the minimum alignment of each @var{variable} to @var{alignment}.
12623 This is the same as GCC's @code{aligned} attribute @pxref{Variable
12624 Attributes}). Macro expansion occurs on the arguments to this pragma
12625 when compiling C and Objective-C@. It does not currently occur when
12626 compiling C++, but this is a bug which may be fixed in a future
12629 @item fini (@var{function} [, @var{function}]...)
12630 @cindex pragma, fini
12632 This pragma causes each listed @var{function} to be called after
12633 main, or during shared module unloading, by adding a call to the
12634 @code{.fini} section.
12636 @item init (@var{function} [, @var{function}]...)
12637 @cindex pragma, init
12639 This pragma causes each listed @var{function} to be called during
12640 initialization (before @code{main}) or during shared module loading, by
12641 adding a call to the @code{.init} section.
12645 @node Symbol-Renaming Pragmas
12646 @subsection Symbol-Renaming Pragmas
12648 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
12649 supports two @code{#pragma} directives which change the name used in
12650 assembly for a given declaration. @code{#pragma extern_prefix} is only
12651 available on platforms whose system headers need it. To get this effect
12652 on all platforms supported by GCC, use the asm labels extension (@pxref{Asm
12656 @item redefine_extname @var{oldname} @var{newname}
12657 @cindex pragma, redefine_extname
12659 This pragma gives the C function @var{oldname} the assembly symbol
12660 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
12661 will be defined if this pragma is available (currently on all platforms).
12663 @item extern_prefix @var{string}
12664 @cindex pragma, extern_prefix
12666 This pragma causes all subsequent external function and variable
12667 declarations to have @var{string} prepended to their assembly symbols.
12668 This effect may be terminated with another @code{extern_prefix} pragma
12669 whose argument is an empty string. The preprocessor macro
12670 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
12671 available (currently only on Tru64 UNIX)@.
12674 These pragmas and the asm labels extension interact in a complicated
12675 manner. Here are some corner cases you may want to be aware of.
12678 @item Both pragmas silently apply only to declarations with external
12679 linkage. Asm labels do not have this restriction.
12681 @item In C++, both pragmas silently apply only to declarations with
12682 ``C'' linkage. Again, asm labels do not have this restriction.
12684 @item If any of the three ways of changing the assembly name of a
12685 declaration is applied to a declaration whose assembly name has
12686 already been determined (either by a previous use of one of these
12687 features, or because the compiler needed the assembly name in order to
12688 generate code), and the new name is different, a warning issues and
12689 the name does not change.
12691 @item The @var{oldname} used by @code{#pragma redefine_extname} is
12692 always the C-language name.
12694 @item If @code{#pragma extern_prefix} is in effect, and a declaration
12695 occurs with an asm label attached, the prefix is silently ignored for
12698 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
12699 apply to the same declaration, whichever triggered first wins, and a
12700 warning issues if they contradict each other. (We would like to have
12701 @code{#pragma redefine_extname} always win, for consistency with asm
12702 labels, but if @code{#pragma extern_prefix} triggers first we have no
12703 way of knowing that that happened.)
12706 @node Structure-Packing Pragmas
12707 @subsection Structure-Packing Pragmas
12709 For compatibility with Microsoft Windows compilers, GCC supports a
12710 set of @code{#pragma} directives which change the maximum alignment of
12711 members of structures (other than zero-width bitfields), unions, and
12712 classes subsequently defined. The @var{n} value below always is required
12713 to be a small power of two and specifies the new alignment in bytes.
12716 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
12717 @item @code{#pragma pack()} sets the alignment to the one that was in
12718 effect when compilation started (see also command-line option
12719 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
12720 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
12721 setting on an internal stack and then optionally sets the new alignment.
12722 @item @code{#pragma pack(pop)} restores the alignment setting to the one
12723 saved at the top of the internal stack (and removes that stack entry).
12724 Note that @code{#pragma pack([@var{n}])} does not influence this internal
12725 stack; thus it is possible to have @code{#pragma pack(push)} followed by
12726 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
12727 @code{#pragma pack(pop)}.
12730 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
12731 @code{#pragma} which lays out a structure as the documented
12732 @code{__attribute__ ((ms_struct))}.
12734 @item @code{#pragma ms_struct on} turns on the layout for structures
12736 @item @code{#pragma ms_struct off} turns off the layout for structures
12738 @item @code{#pragma ms_struct reset} goes back to the default layout.
12742 @subsection Weak Pragmas
12744 For compatibility with SVR4, GCC supports a set of @code{#pragma}
12745 directives for declaring symbols to be weak, and defining weak
12749 @item #pragma weak @var{symbol}
12750 @cindex pragma, weak
12751 This pragma declares @var{symbol} to be weak, as if the declaration
12752 had the attribute of the same name. The pragma may appear before
12753 or after the declaration of @var{symbol}, but must appear before
12754 either its first use or its definition. It is not an error for
12755 @var{symbol} to never be defined at all.
12757 @item #pragma weak @var{symbol1} = @var{symbol2}
12758 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
12759 It is an error if @var{symbol2} is not defined in the current
12763 @node Diagnostic Pragmas
12764 @subsection Diagnostic Pragmas
12766 GCC allows the user to selectively enable or disable certain types of
12767 diagnostics, and change the kind of the diagnostic. For example, a
12768 project's policy might require that all sources compile with
12769 @option{-Werror} but certain files might have exceptions allowing
12770 specific types of warnings. Or, a project might selectively enable
12771 diagnostics and treat them as errors depending on which preprocessor
12772 macros are defined.
12775 @item #pragma GCC diagnostic @var{kind} @var{option}
12776 @cindex pragma, diagnostic
12778 Modifies the disposition of a diagnostic. Note that not all
12779 diagnostics are modifiable; at the moment only warnings (normally
12780 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
12781 Use @option{-fdiagnostics-show-option} to determine which diagnostics
12782 are controllable and which option controls them.
12784 @var{kind} is @samp{error} to treat this diagnostic as an error,
12785 @samp{warning} to treat it like a warning (even if @option{-Werror} is
12786 in effect), or @samp{ignored} if the diagnostic is to be ignored.
12787 @var{option} is a double quoted string which matches the command-line
12791 #pragma GCC diagnostic warning "-Wformat"
12792 #pragma GCC diagnostic error "-Wformat"
12793 #pragma GCC diagnostic ignored "-Wformat"
12796 Note that these pragmas override any command-line options. GCC keeps
12797 track of the location of each pragma, and issues diagnostics according
12798 to the state as of that point in the source file. Thus, pragmas occurring
12799 after a line do not affect diagnostics caused by that line.
12801 @item #pragma GCC diagnostic push
12802 @itemx #pragma GCC diagnostic pop
12804 Causes GCC to remember the state of the diagnostics as of each
12805 @code{push}, and restore to that point at each @code{pop}. If a
12806 @code{pop} has no matching @code{push}, the command line options are
12810 #pragma GCC diagnostic error "-Wuninitialized"
12811 foo(a); /* error is given for this one */
12812 #pragma GCC diagnostic push
12813 #pragma GCC diagnostic ignored "-Wuninitialized"
12814 foo(b); /* no diagnostic for this one */
12815 #pragma GCC diagnostic pop
12816 foo(c); /* error is given for this one */
12817 #pragma GCC diagnostic pop
12818 foo(d); /* depends on command line options */
12823 GCC also offers a simple mechanism for printing messages during
12827 @item #pragma message @var{string}
12828 @cindex pragma, diagnostic
12830 Prints @var{string} as a compiler message on compilation. The message
12831 is informational only, and is neither a compilation warning nor an error.
12834 #pragma message "Compiling " __FILE__ "..."
12837 @var{string} may be parenthesized, and is printed with location
12838 information. For example,
12841 #define DO_PRAGMA(x) _Pragma (#x)
12842 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
12844 TODO(Remember to fix this)
12847 prints @samp{/tmp/file.c:4: note: #pragma message:
12848 TODO - Remember to fix this}.
12852 @node Visibility Pragmas
12853 @subsection Visibility Pragmas
12856 @item #pragma GCC visibility push(@var{visibility})
12857 @itemx #pragma GCC visibility pop
12858 @cindex pragma, visibility
12860 This pragma allows the user to set the visibility for multiple
12861 declarations without having to give each a visibility attribute
12862 @xref{Function Attributes}, for more information about visibility and
12863 the attribute syntax.
12865 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
12866 declarations. Class members and template specializations are not
12867 affected; if you want to override the visibility for a particular
12868 member or instantiation, you must use an attribute.
12873 @node Push/Pop Macro Pragmas
12874 @subsection Push/Pop Macro Pragmas
12876 For compatibility with Microsoft Windows compilers, GCC supports
12877 @samp{#pragma push_macro(@var{"macro_name"})}
12878 and @samp{#pragma pop_macro(@var{"macro_name"})}.
12881 @item #pragma push_macro(@var{"macro_name"})
12882 @cindex pragma, push_macro
12883 This pragma saves the value of the macro named as @var{macro_name} to
12884 the top of the stack for this macro.
12886 @item #pragma pop_macro(@var{"macro_name"})
12887 @cindex pragma, pop_macro
12888 This pragma sets the value of the macro named as @var{macro_name} to
12889 the value on top of the stack for this macro. If the stack for
12890 @var{macro_name} is empty, the value of the macro remains unchanged.
12897 #pragma push_macro("X")
12900 #pragma pop_macro("X")
12904 In this example, the definition of X as 1 is saved by @code{#pragma
12905 push_macro} and restored by @code{#pragma pop_macro}.
12907 @node Function Specific Option Pragmas
12908 @subsection Function Specific Option Pragmas
12911 @item #pragma GCC target (@var{"string"}...)
12912 @cindex pragma GCC target
12914 This pragma allows you to set target specific options for functions
12915 defined later in the source file. One or more strings can be
12916 specified. Each function that is defined after this point will be as
12917 if @code{attribute((target("STRING")))} was specified for that
12918 function. The parenthesis around the options is optional.
12919 @xref{Function Attributes}, for more information about the
12920 @code{target} attribute and the attribute syntax.
12922 The @samp{#pragma GCC target} pragma is not implemented in GCC
12923 versions earlier than 4.4, and is currently only implemented for the
12924 386 and x86_64 backends.
12928 @item #pragma GCC optimize (@var{"string"}...)
12929 @cindex pragma GCC optimize
12931 This pragma allows you to set global optimization options for functions
12932 defined later in the source file. One or more strings can be
12933 specified. Each function that is defined after this point will be as
12934 if @code{attribute((optimize("STRING")))} was specified for that
12935 function. The parenthesis around the options is optional.
12936 @xref{Function Attributes}, for more information about the
12937 @code{optimize} attribute and the attribute syntax.
12939 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
12940 versions earlier than 4.4.
12944 @item #pragma GCC push_options
12945 @itemx #pragma GCC pop_options
12946 @cindex pragma GCC push_options
12947 @cindex pragma GCC pop_options
12949 These pragmas maintain a stack of the current target and optimization
12950 options. It is intended for include files where you temporarily want
12951 to switch to using a different @samp{#pragma GCC target} or
12952 @samp{#pragma GCC optimize} and then to pop back to the previous
12955 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
12956 pragmas are not implemented in GCC versions earlier than 4.4.
12960 @item #pragma GCC reset_options
12961 @cindex pragma GCC reset_options
12963 This pragma clears the current @code{#pragma GCC target} and
12964 @code{#pragma GCC optimize} to use the default switches as specified
12965 on the command line.
12967 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
12968 versions earlier than 4.4.
12971 @node Unnamed Fields
12972 @section Unnamed struct/union fields within structs/unions
12973 @cindex @code{struct}
12974 @cindex @code{union}
12976 As permitted by ISO C1X and for compatibility with other compilers,
12977 GCC allows you to define
12978 a structure or union that contains, as fields, structures and unions
12979 without names. For example:
12992 In this example, the user would be able to access members of the unnamed
12993 union with code like @samp{foo.b}. Note that only unnamed structs and
12994 unions are allowed, you may not have, for example, an unnamed
12997 You must never create such structures that cause ambiguous field definitions.
12998 For example, this structure:
13009 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
13010 The compiler gives errors for such constructs.
13012 @opindex fms-extensions
13013 Unless @option{-fms-extensions} is used, the unnamed field must be a
13014 structure or union definition without a tag (for example, @samp{struct
13015 @{ int a; @};}), or a @code{typedef} name for such a structure or
13016 union. If @option{-fms-extensions} is used, the field may
13017 also be a definition with a tag such as @samp{struct foo @{ int a;
13018 @};}, a reference to a previously defined structure or union such as
13019 @samp{struct foo;}, or a reference to a @code{typedef} name for a
13020 previously defined structure or union type with a tag.
13023 @section Thread-Local Storage
13024 @cindex Thread-Local Storage
13025 @cindex @acronym{TLS}
13026 @cindex @code{__thread}
13028 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
13029 are allocated such that there is one instance of the variable per extant
13030 thread. The run-time model GCC uses to implement this originates
13031 in the IA-64 processor-specific ABI, but has since been migrated
13032 to other processors as well. It requires significant support from
13033 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
13034 system libraries (@file{libc.so} and @file{libpthread.so}), so it
13035 is not available everywhere.
13037 At the user level, the extension is visible with a new storage
13038 class keyword: @code{__thread}. For example:
13042 extern __thread struct state s;
13043 static __thread char *p;
13046 The @code{__thread} specifier may be used alone, with the @code{extern}
13047 or @code{static} specifiers, but with no other storage class specifier.
13048 When used with @code{extern} or @code{static}, @code{__thread} must appear
13049 immediately after the other storage class specifier.
13051 The @code{__thread} specifier may be applied to any global, file-scoped
13052 static, function-scoped static, or static data member of a class. It may
13053 not be applied to block-scoped automatic or non-static data member.
13055 When the address-of operator is applied to a thread-local variable, it is
13056 evaluated at run-time and returns the address of the current thread's
13057 instance of that variable. An address so obtained may be used by any
13058 thread. When a thread terminates, any pointers to thread-local variables
13059 in that thread become invalid.
13061 No static initialization may refer to the address of a thread-local variable.
13063 In C++, if an initializer is present for a thread-local variable, it must
13064 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
13067 See @uref{http://people.redhat.com/drepper/tls.pdf,
13068 ELF Handling For Thread-Local Storage} for a detailed explanation of
13069 the four thread-local storage addressing models, and how the run-time
13070 is expected to function.
13073 * C99 Thread-Local Edits::
13074 * C++98 Thread-Local Edits::
13077 @node C99 Thread-Local Edits
13078 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
13080 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
13081 that document the exact semantics of the language extension.
13085 @cite{5.1.2 Execution environments}
13087 Add new text after paragraph 1
13090 Within either execution environment, a @dfn{thread} is a flow of
13091 control within a program. It is implementation defined whether
13092 or not there may be more than one thread associated with a program.
13093 It is implementation defined how threads beyond the first are
13094 created, the name and type of the function called at thread
13095 startup, and how threads may be terminated. However, objects
13096 with thread storage duration shall be initialized before thread
13101 @cite{6.2.4 Storage durations of objects}
13103 Add new text before paragraph 3
13106 An object whose identifier is declared with the storage-class
13107 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
13108 Its lifetime is the entire execution of the thread, and its
13109 stored value is initialized only once, prior to thread startup.
13113 @cite{6.4.1 Keywords}
13115 Add @code{__thread}.
13118 @cite{6.7.1 Storage-class specifiers}
13120 Add @code{__thread} to the list of storage class specifiers in
13123 Change paragraph 2 to
13126 With the exception of @code{__thread}, at most one storage-class
13127 specifier may be given [@dots{}]. The @code{__thread} specifier may
13128 be used alone, or immediately following @code{extern} or
13132 Add new text after paragraph 6
13135 The declaration of an identifier for a variable that has
13136 block scope that specifies @code{__thread} shall also
13137 specify either @code{extern} or @code{static}.
13139 The @code{__thread} specifier shall be used only with
13144 @node C++98 Thread-Local Edits
13145 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
13147 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
13148 that document the exact semantics of the language extension.
13152 @b{[intro.execution]}
13154 New text after paragraph 4
13157 A @dfn{thread} is a flow of control within the abstract machine.
13158 It is implementation defined whether or not there may be more than
13162 New text after paragraph 7
13165 It is unspecified whether additional action must be taken to
13166 ensure when and whether side effects are visible to other threads.
13172 Add @code{__thread}.
13175 @b{[basic.start.main]}
13177 Add after paragraph 5
13180 The thread that begins execution at the @code{main} function is called
13181 the @dfn{main thread}. It is implementation defined how functions
13182 beginning threads other than the main thread are designated or typed.
13183 A function so designated, as well as the @code{main} function, is called
13184 a @dfn{thread startup function}. It is implementation defined what
13185 happens if a thread startup function returns. It is implementation
13186 defined what happens to other threads when any thread calls @code{exit}.
13190 @b{[basic.start.init]}
13192 Add after paragraph 4
13195 The storage for an object of thread storage duration shall be
13196 statically initialized before the first statement of the thread startup
13197 function. An object of thread storage duration shall not require
13198 dynamic initialization.
13202 @b{[basic.start.term]}
13204 Add after paragraph 3
13207 The type of an object with thread storage duration shall not have a
13208 non-trivial destructor, nor shall it be an array type whose elements
13209 (directly or indirectly) have non-trivial destructors.
13215 Add ``thread storage duration'' to the list in paragraph 1.
13220 Thread, static, and automatic storage durations are associated with
13221 objects introduced by declarations [@dots{}].
13224 Add @code{__thread} to the list of specifiers in paragraph 3.
13227 @b{[basic.stc.thread]}
13229 New section before @b{[basic.stc.static]}
13232 The keyword @code{__thread} applied to a non-local object gives the
13233 object thread storage duration.
13235 A local variable or class data member declared both @code{static}
13236 and @code{__thread} gives the variable or member thread storage
13241 @b{[basic.stc.static]}
13246 All objects which have neither thread storage duration, dynamic
13247 storage duration nor are local [@dots{}].
13253 Add @code{__thread} to the list in paragraph 1.
13258 With the exception of @code{__thread}, at most one
13259 @var{storage-class-specifier} shall appear in a given
13260 @var{decl-specifier-seq}. The @code{__thread} specifier may
13261 be used alone, or immediately following the @code{extern} or
13262 @code{static} specifiers. [@dots{}]
13265 Add after paragraph 5
13268 The @code{__thread} specifier can be applied only to the names of objects
13269 and to anonymous unions.
13275 Add after paragraph 6
13278 Non-@code{static} members shall not be @code{__thread}.
13282 @node Binary constants
13283 @section Binary constants using the @samp{0b} prefix
13284 @cindex Binary constants using the @samp{0b} prefix
13286 Integer constants can be written as binary constants, consisting of a
13287 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
13288 @samp{0B}. This is particularly useful in environments that operate a
13289 lot on the bit-level (like microcontrollers).
13291 The following statements are identical:
13300 The type of these constants follows the same rules as for octal or
13301 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
13304 @node C++ Extensions
13305 @chapter Extensions to the C++ Language
13306 @cindex extensions, C++ language
13307 @cindex C++ language extensions
13309 The GNU compiler provides these extensions to the C++ language (and you
13310 can also use most of the C language extensions in your C++ programs). If you
13311 want to write code that checks whether these features are available, you can
13312 test for the GNU compiler the same way as for C programs: check for a
13313 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
13314 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
13315 Predefined Macros,cpp,The GNU C Preprocessor}).
13318 * C++ Volatiles:: What constitutes an access to a volatile object.
13319 * Restricted Pointers:: C99 restricted pointers and references.
13320 * Vague Linkage:: Where G++ puts inlines, vtables and such.
13321 * C++ Interface:: You can use a single C++ header file for both
13322 declarations and definitions.
13323 * Template Instantiation:: Methods for ensuring that exactly one copy of
13324 each needed template instantiation is emitted.
13325 * Bound member functions:: You can extract a function pointer to the
13326 method denoted by a @samp{->*} or @samp{.*} expression.
13327 * C++ Attributes:: Variable, function, and type attributes for C++ only.
13328 * Namespace Association:: Strong using-directives for namespace association.
13329 * Type Traits:: Compiler support for type traits
13330 * Java Exceptions:: Tweaking exception handling to work with Java.
13331 * Deprecated Features:: Things will disappear from g++.
13332 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
13335 @node C++ Volatiles
13336 @section When is a Volatile C++ Object Accessed?
13337 @cindex accessing volatiles
13338 @cindex volatile read
13339 @cindex volatile write
13340 @cindex volatile access
13342 The C++ standard differs from the C standard in its treatment of
13343 volatile objects. It fails to specify what constitutes a volatile
13344 access, except to say that C++ should behave in a similar manner to C
13345 with respect to volatiles, where possible. However, the different
13346 lvalueness of expressions between C and C++ complicate the behaviour.
13347 G++ behaves the same as GCC for volatile access, @xref{C
13348 Extensions,,Volatiles}, for a description of GCC's behaviour.
13350 The C and C++ language specifications differ when an object is
13351 accessed in a void context:
13354 volatile int *src = @var{somevalue};
13358 The C++ standard specifies that such expressions do not undergo lvalue
13359 to rvalue conversion, and that the type of the dereferenced object may
13360 be incomplete. The C++ standard does not specify explicitly that it
13361 is lvalue to rvalue conversion which is responsible for causing an
13362 access. There is reason to believe that it is, because otherwise
13363 certain simple expressions become undefined. However, because it
13364 would surprise most programmers, G++ treats dereferencing a pointer to
13365 volatile object of complete type as GCC would do for an equivalent
13366 type in C@. When the object has incomplete type, G++ issues a
13367 warning; if you wish to force an error, you must force a conversion to
13368 rvalue with, for instance, a static cast.
13370 When using a reference to volatile, G++ does not treat equivalent
13371 expressions as accesses to volatiles, but instead issues a warning that
13372 no volatile is accessed. The rationale for this is that otherwise it
13373 becomes difficult to determine where volatile access occur, and not
13374 possible to ignore the return value from functions returning volatile
13375 references. Again, if you wish to force a read, cast the reference to
13378 G++ implements the same behaviour as GCC does when assigning to a
13379 volatile object -- there is no reread of the assigned-to object, the
13380 assigned rvalue is reused. Note that in C++ assignment expressions
13381 are lvalues, and if used as an lvalue, the volatile object will be
13382 referred to. For instance, @var{vref} will refer to @var{vobj}, as
13383 expected, in the following example:
13387 volatile int &vref = vobj = @var{something};
13390 @node Restricted Pointers
13391 @section Restricting Pointer Aliasing
13392 @cindex restricted pointers
13393 @cindex restricted references
13394 @cindex restricted this pointer
13396 As with the C front end, G++ understands the C99 feature of restricted pointers,
13397 specified with the @code{__restrict__}, or @code{__restrict} type
13398 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
13399 language flag, @code{restrict} is not a keyword in C++.
13401 In addition to allowing restricted pointers, you can specify restricted
13402 references, which indicate that the reference is not aliased in the local
13406 void fn (int *__restrict__ rptr, int &__restrict__ rref)
13413 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
13414 @var{rref} refers to a (different) unaliased integer.
13416 You may also specify whether a member function's @var{this} pointer is
13417 unaliased by using @code{__restrict__} as a member function qualifier.
13420 void T::fn () __restrict__
13427 Within the body of @code{T::fn}, @var{this} will have the effective
13428 definition @code{T *__restrict__ const this}. Notice that the
13429 interpretation of a @code{__restrict__} member function qualifier is
13430 different to that of @code{const} or @code{volatile} qualifier, in that it
13431 is applied to the pointer rather than the object. This is consistent with
13432 other compilers which implement restricted pointers.
13434 As with all outermost parameter qualifiers, @code{__restrict__} is
13435 ignored in function definition matching. This means you only need to
13436 specify @code{__restrict__} in a function definition, rather than
13437 in a function prototype as well.
13439 @node Vague Linkage
13440 @section Vague Linkage
13441 @cindex vague linkage
13443 There are several constructs in C++ which require space in the object
13444 file but are not clearly tied to a single translation unit. We say that
13445 these constructs have ``vague linkage''. Typically such constructs are
13446 emitted wherever they are needed, though sometimes we can be more
13450 @item Inline Functions
13451 Inline functions are typically defined in a header file which can be
13452 included in many different compilations. Hopefully they can usually be
13453 inlined, but sometimes an out-of-line copy is necessary, if the address
13454 of the function is taken or if inlining fails. In general, we emit an
13455 out-of-line copy in all translation units where one is needed. As an
13456 exception, we only emit inline virtual functions with the vtable, since
13457 it will always require a copy.
13459 Local static variables and string constants used in an inline function
13460 are also considered to have vague linkage, since they must be shared
13461 between all inlined and out-of-line instances of the function.
13465 C++ virtual functions are implemented in most compilers using a lookup
13466 table, known as a vtable. The vtable contains pointers to the virtual
13467 functions provided by a class, and each object of the class contains a
13468 pointer to its vtable (or vtables, in some multiple-inheritance
13469 situations). If the class declares any non-inline, non-pure virtual
13470 functions, the first one is chosen as the ``key method'' for the class,
13471 and the vtable is only emitted in the translation unit where the key
13474 @emph{Note:} If the chosen key method is later defined as inline, the
13475 vtable will still be emitted in every translation unit which defines it.
13476 Make sure that any inline virtuals are declared inline in the class
13477 body, even if they are not defined there.
13479 @item @code{type_info} objects
13480 @cindex @code{type_info}
13482 C++ requires information about types to be written out in order to
13483 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
13484 For polymorphic classes (classes with virtual functions), the @samp{type_info}
13485 object is written out along with the vtable so that @samp{dynamic_cast}
13486 can determine the dynamic type of a class object at runtime. For all
13487 other types, we write out the @samp{type_info} object when it is used: when
13488 applying @samp{typeid} to an expression, throwing an object, or
13489 referring to a type in a catch clause or exception specification.
13491 @item Template Instantiations
13492 Most everything in this section also applies to template instantiations,
13493 but there are other options as well.
13494 @xref{Template Instantiation,,Where's the Template?}.
13498 When used with GNU ld version 2.8 or later on an ELF system such as
13499 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
13500 these constructs will be discarded at link time. This is known as
13503 On targets that don't support COMDAT, but do support weak symbols, GCC
13504 will use them. This way one copy will override all the others, but
13505 the unused copies will still take up space in the executable.
13507 For targets which do not support either COMDAT or weak symbols,
13508 most entities with vague linkage will be emitted as local symbols to
13509 avoid duplicate definition errors from the linker. This will not happen
13510 for local statics in inlines, however, as having multiple copies will
13511 almost certainly break things.
13513 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
13514 another way to control placement of these constructs.
13516 @node C++ Interface
13517 @section #pragma interface and implementation
13519 @cindex interface and implementation headers, C++
13520 @cindex C++ interface and implementation headers
13521 @cindex pragmas, interface and implementation
13523 @code{#pragma interface} and @code{#pragma implementation} provide the
13524 user with a way of explicitly directing the compiler to emit entities
13525 with vague linkage (and debugging information) in a particular
13528 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
13529 most cases, because of COMDAT support and the ``key method'' heuristic
13530 mentioned in @ref{Vague Linkage}. Using them can actually cause your
13531 program to grow due to unnecessary out-of-line copies of inline
13532 functions. Currently (3.4) the only benefit of these
13533 @code{#pragma}s is reduced duplication of debugging information, and
13534 that should be addressed soon on DWARF 2 targets with the use of
13538 @item #pragma interface
13539 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
13540 @kindex #pragma interface
13541 Use this directive in @emph{header files} that define object classes, to save
13542 space in most of the object files that use those classes. Normally,
13543 local copies of certain information (backup copies of inline member
13544 functions, debugging information, and the internal tables that implement
13545 virtual functions) must be kept in each object file that includes class
13546 definitions. You can use this pragma to avoid such duplication. When a
13547 header file containing @samp{#pragma interface} is included in a
13548 compilation, this auxiliary information will not be generated (unless
13549 the main input source file itself uses @samp{#pragma implementation}).
13550 Instead, the object files will contain references to be resolved at link
13553 The second form of this directive is useful for the case where you have
13554 multiple headers with the same name in different directories. If you
13555 use this form, you must specify the same string to @samp{#pragma
13558 @item #pragma implementation
13559 @itemx #pragma implementation "@var{objects}.h"
13560 @kindex #pragma implementation
13561 Use this pragma in a @emph{main input file}, when you want full output from
13562 included header files to be generated (and made globally visible). The
13563 included header file, in turn, should use @samp{#pragma interface}.
13564 Backup copies of inline member functions, debugging information, and the
13565 internal tables used to implement virtual functions are all generated in
13566 implementation files.
13568 @cindex implied @code{#pragma implementation}
13569 @cindex @code{#pragma implementation}, implied
13570 @cindex naming convention, implementation headers
13571 If you use @samp{#pragma implementation} with no argument, it applies to
13572 an include file with the same basename@footnote{A file's @dfn{basename}
13573 was the name stripped of all leading path information and of trailing
13574 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
13575 file. For example, in @file{allclass.cc}, giving just
13576 @samp{#pragma implementation}
13577 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
13579 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
13580 an implementation file whenever you would include it from
13581 @file{allclass.cc} even if you never specified @samp{#pragma
13582 implementation}. This was deemed to be more trouble than it was worth,
13583 however, and disabled.
13585 Use the string argument if you want a single implementation file to
13586 include code from multiple header files. (You must also use
13587 @samp{#include} to include the header file; @samp{#pragma
13588 implementation} only specifies how to use the file---it doesn't actually
13591 There is no way to split up the contents of a single header file into
13592 multiple implementation files.
13595 @cindex inlining and C++ pragmas
13596 @cindex C++ pragmas, effect on inlining
13597 @cindex pragmas in C++, effect on inlining
13598 @samp{#pragma implementation} and @samp{#pragma interface} also have an
13599 effect on function inlining.
13601 If you define a class in a header file marked with @samp{#pragma
13602 interface}, the effect on an inline function defined in that class is
13603 similar to an explicit @code{extern} declaration---the compiler emits
13604 no code at all to define an independent version of the function. Its
13605 definition is used only for inlining with its callers.
13607 @opindex fno-implement-inlines
13608 Conversely, when you include the same header file in a main source file
13609 that declares it as @samp{#pragma implementation}, the compiler emits
13610 code for the function itself; this defines a version of the function
13611 that can be found via pointers (or by callers compiled without
13612 inlining). If all calls to the function can be inlined, you can avoid
13613 emitting the function by compiling with @option{-fno-implement-inlines}.
13614 If any calls were not inlined, you will get linker errors.
13616 @node Template Instantiation
13617 @section Where's the Template?
13618 @cindex template instantiation
13620 C++ templates are the first language feature to require more
13621 intelligence from the environment than one usually finds on a UNIX
13622 system. Somehow the compiler and linker have to make sure that each
13623 template instance occurs exactly once in the executable if it is needed,
13624 and not at all otherwise. There are two basic approaches to this
13625 problem, which are referred to as the Borland model and the Cfront model.
13628 @item Borland model
13629 Borland C++ solved the template instantiation problem by adding the code
13630 equivalent of common blocks to their linker; the compiler emits template
13631 instances in each translation unit that uses them, and the linker
13632 collapses them together. The advantage of this model is that the linker
13633 only has to consider the object files themselves; there is no external
13634 complexity to worry about. This disadvantage is that compilation time
13635 is increased because the template code is being compiled repeatedly.
13636 Code written for this model tends to include definitions of all
13637 templates in the header file, since they must be seen to be
13641 The AT&T C++ translator, Cfront, solved the template instantiation
13642 problem by creating the notion of a template repository, an
13643 automatically maintained place where template instances are stored. A
13644 more modern version of the repository works as follows: As individual
13645 object files are built, the compiler places any template definitions and
13646 instantiations encountered in the repository. At link time, the link
13647 wrapper adds in the objects in the repository and compiles any needed
13648 instances that were not previously emitted. The advantages of this
13649 model are more optimal compilation speed and the ability to use the
13650 system linker; to implement the Borland model a compiler vendor also
13651 needs to replace the linker. The disadvantages are vastly increased
13652 complexity, and thus potential for error; for some code this can be
13653 just as transparent, but in practice it can been very difficult to build
13654 multiple programs in one directory and one program in multiple
13655 directories. Code written for this model tends to separate definitions
13656 of non-inline member templates into a separate file, which should be
13657 compiled separately.
13660 When used with GNU ld version 2.8 or later on an ELF system such as
13661 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
13662 Borland model. On other systems, G++ implements neither automatic
13665 A future version of G++ will support a hybrid model whereby the compiler
13666 will emit any instantiations for which the template definition is
13667 included in the compile, and store template definitions and
13668 instantiation context information into the object file for the rest.
13669 The link wrapper will extract that information as necessary and invoke
13670 the compiler to produce the remaining instantiations. The linker will
13671 then combine duplicate instantiations.
13673 In the mean time, you have the following options for dealing with
13674 template instantiations:
13679 Compile your template-using code with @option{-frepo}. The compiler will
13680 generate files with the extension @samp{.rpo} listing all of the
13681 template instantiations used in the corresponding object files which
13682 could be instantiated there; the link wrapper, @samp{collect2}, will
13683 then update the @samp{.rpo} files to tell the compiler where to place
13684 those instantiations and rebuild any affected object files. The
13685 link-time overhead is negligible after the first pass, as the compiler
13686 will continue to place the instantiations in the same files.
13688 This is your best option for application code written for the Borland
13689 model, as it will just work. Code written for the Cfront model will
13690 need to be modified so that the template definitions are available at
13691 one or more points of instantiation; usually this is as simple as adding
13692 @code{#include <tmethods.cc>} to the end of each template header.
13694 For library code, if you want the library to provide all of the template
13695 instantiations it needs, just try to link all of its object files
13696 together; the link will fail, but cause the instantiations to be
13697 generated as a side effect. Be warned, however, that this may cause
13698 conflicts if multiple libraries try to provide the same instantiations.
13699 For greater control, use explicit instantiation as described in the next
13703 @opindex fno-implicit-templates
13704 Compile your code with @option{-fno-implicit-templates} to disable the
13705 implicit generation of template instances, and explicitly instantiate
13706 all the ones you use. This approach requires more knowledge of exactly
13707 which instances you need than do the others, but it's less
13708 mysterious and allows greater control. You can scatter the explicit
13709 instantiations throughout your program, perhaps putting them in the
13710 translation units where the instances are used or the translation units
13711 that define the templates themselves; you can put all of the explicit
13712 instantiations you need into one big file; or you can create small files
13719 template class Foo<int>;
13720 template ostream& operator <<
13721 (ostream&, const Foo<int>&);
13724 for each of the instances you need, and create a template instantiation
13725 library from those.
13727 If you are using Cfront-model code, you can probably get away with not
13728 using @option{-fno-implicit-templates} when compiling files that don't
13729 @samp{#include} the member template definitions.
13731 If you use one big file to do the instantiations, you may want to
13732 compile it without @option{-fno-implicit-templates} so you get all of the
13733 instances required by your explicit instantiations (but not by any
13734 other files) without having to specify them as well.
13736 G++ has extended the template instantiation syntax given in the ISO
13737 standard to allow forward declaration of explicit instantiations
13738 (with @code{extern}), instantiation of the compiler support data for a
13739 template class (i.e.@: the vtable) without instantiating any of its
13740 members (with @code{inline}), and instantiation of only the static data
13741 members of a template class, without the support data or member
13742 functions (with (@code{static}):
13745 extern template int max (int, int);
13746 inline template class Foo<int>;
13747 static template class Foo<int>;
13751 Do nothing. Pretend G++ does implement automatic instantiation
13752 management. Code written for the Borland model will work fine, but
13753 each translation unit will contain instances of each of the templates it
13754 uses. In a large program, this can lead to an unacceptable amount of code
13758 @node Bound member functions
13759 @section Extracting the function pointer from a bound pointer to member function
13761 @cindex pointer to member function
13762 @cindex bound pointer to member function
13764 In C++, pointer to member functions (PMFs) are implemented using a wide
13765 pointer of sorts to handle all the possible call mechanisms; the PMF
13766 needs to store information about how to adjust the @samp{this} pointer,
13767 and if the function pointed to is virtual, where to find the vtable, and
13768 where in the vtable to look for the member function. If you are using
13769 PMFs in an inner loop, you should really reconsider that decision. If
13770 that is not an option, you can extract the pointer to the function that
13771 would be called for a given object/PMF pair and call it directly inside
13772 the inner loop, to save a bit of time.
13774 Note that you will still be paying the penalty for the call through a
13775 function pointer; on most modern architectures, such a call defeats the
13776 branch prediction features of the CPU@. This is also true of normal
13777 virtual function calls.
13779 The syntax for this extension is
13783 extern int (A::*fp)();
13784 typedef int (*fptr)(A *);
13786 fptr p = (fptr)(a.*fp);
13789 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
13790 no object is needed to obtain the address of the function. They can be
13791 converted to function pointers directly:
13794 fptr p1 = (fptr)(&A::foo);
13797 @opindex Wno-pmf-conversions
13798 You must specify @option{-Wno-pmf-conversions} to use this extension.
13800 @node C++ Attributes
13801 @section C++-Specific Variable, Function, and Type Attributes
13803 Some attributes only make sense for C++ programs.
13806 @item init_priority (@var{priority})
13807 @cindex @code{init_priority} attribute
13810 In Standard C++, objects defined at namespace scope are guaranteed to be
13811 initialized in an order in strict accordance with that of their definitions
13812 @emph{in a given translation unit}. No guarantee is made for initializations
13813 across translation units. However, GNU C++ allows users to control the
13814 order of initialization of objects defined at namespace scope with the
13815 @code{init_priority} attribute by specifying a relative @var{priority},
13816 a constant integral expression currently bounded between 101 and 65535
13817 inclusive. Lower numbers indicate a higher priority.
13819 In the following example, @code{A} would normally be created before
13820 @code{B}, but the @code{init_priority} attribute has reversed that order:
13823 Some_Class A __attribute__ ((init_priority (2000)));
13824 Some_Class B __attribute__ ((init_priority (543)));
13828 Note that the particular values of @var{priority} do not matter; only their
13831 @item java_interface
13832 @cindex @code{java_interface} attribute
13834 This type attribute informs C++ that the class is a Java interface. It may
13835 only be applied to classes declared within an @code{extern "Java"} block.
13836 Calls to methods declared in this interface will be dispatched using GCJ's
13837 interface table mechanism, instead of regular virtual table dispatch.
13841 See also @ref{Namespace Association}.
13843 @node Namespace Association
13844 @section Namespace Association
13846 @strong{Caution:} The semantics of this extension are not fully
13847 defined. Users should refrain from using this extension as its
13848 semantics may change subtly over time. It is possible that this
13849 extension will be removed in future versions of G++.
13851 A using-directive with @code{__attribute ((strong))} is stronger
13852 than a normal using-directive in two ways:
13856 Templates from the used namespace can be specialized and explicitly
13857 instantiated as though they were members of the using namespace.
13860 The using namespace is considered an associated namespace of all
13861 templates in the used namespace for purposes of argument-dependent
13865 The used namespace must be nested within the using namespace so that
13866 normal unqualified lookup works properly.
13868 This is useful for composing a namespace transparently from
13869 implementation namespaces. For example:
13874 template <class T> struct A @{ @};
13876 using namespace debug __attribute ((__strong__));
13877 template <> struct A<int> @{ @}; // @r{ok to specialize}
13879 template <class T> void f (A<T>);
13884 f (std::A<float>()); // @r{lookup finds} std::f
13890 @section Type Traits
13892 The C++ front-end implements syntactic extensions that allow to
13893 determine at compile time various characteristics of a type (or of a
13897 @item __has_nothrow_assign (type)
13898 If @code{type} is const qualified or is a reference type then the trait is
13899 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
13900 is true, else if @code{type} is a cv class or union type with copy assignment
13901 operators that are known not to throw an exception then the trait is true,
13902 else it is false. Requires: @code{type} shall be a complete type, an array
13903 type of unknown bound, or is a @code{void} type.
13905 @item __has_nothrow_copy (type)
13906 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
13907 @code{type} is a cv class or union type with copy constructors that
13908 are known not to throw an exception then the trait is true, else it is false.
13909 Requires: @code{type} shall be a complete type, an array type of
13910 unknown bound, or is a @code{void} type.
13912 @item __has_nothrow_constructor (type)
13913 If @code{__has_trivial_constructor (type)} is true then the trait is
13914 true, else if @code{type} is a cv class or union type (or array
13915 thereof) with a default constructor that is known not to throw an
13916 exception then the trait is true, else it is false. Requires:
13917 @code{type} shall be a complete type, an array type of unknown bound,
13918 or is a @code{void} type.
13920 @item __has_trivial_assign (type)
13921 If @code{type} is const qualified or is a reference type then the trait is
13922 false. Otherwise if @code{__is_pod (type)} is true then the trait is
13923 true, else if @code{type} is a cv class or union type with a trivial
13924 copy assignment ([class.copy]) then the trait is true, else it is
13925 false. Requires: @code{type} shall be a complete type, an array type
13926 of unknown bound, or is a @code{void} type.
13928 @item __has_trivial_copy (type)
13929 If @code{__is_pod (type)} is true or @code{type} is a reference type
13930 then the trait is true, else if @code{type} is a cv class or union type
13931 with a trivial copy constructor ([class.copy]) then the trait
13932 is true, else it is false. Requires: @code{type} shall be a complete
13933 type, an array type of unknown bound, or is a @code{void} type.
13935 @item __has_trivial_constructor (type)
13936 If @code{__is_pod (type)} is true then the trait is true, else if
13937 @code{type} is a cv class or union type (or array thereof) with a
13938 trivial default constructor ([class.ctor]) then the trait is true,
13939 else it is false. Requires: @code{type} shall be a complete type, an
13940 array type of unknown bound, or is a @code{void} type.
13942 @item __has_trivial_destructor (type)
13943 If @code{__is_pod (type)} is true or @code{type} is a reference type then
13944 the trait is true, else if @code{type} is a cv class or union type (or
13945 array thereof) with a trivial destructor ([class.dtor]) then the trait
13946 is true, else it is false. Requires: @code{type} shall be a complete
13947 type, an array type of unknown bound, or is a @code{void} type.
13949 @item __has_virtual_destructor (type)
13950 If @code{type} is a class type with a virtual destructor
13951 ([class.dtor]) then the trait is true, else it is false. Requires:
13952 @code{type} shall be a complete type, an array type of unknown bound,
13953 or is a @code{void} type.
13955 @item __is_abstract (type)
13956 If @code{type} is an abstract class ([class.abstract]) then the trait
13957 is true, else it is false. Requires: @code{type} shall be a complete
13958 type, an array type of unknown bound, or is a @code{void} type.
13960 @item __is_base_of (base_type, derived_type)
13961 If @code{base_type} is a base class of @code{derived_type}
13962 ([class.derived]) then the trait is true, otherwise it is false.
13963 Top-level cv qualifications of @code{base_type} and
13964 @code{derived_type} are ignored. For the purposes of this trait, a
13965 class type is considered is own base. Requires: if @code{__is_class
13966 (base_type)} and @code{__is_class (derived_type)} are true and
13967 @code{base_type} and @code{derived_type} are not the same type
13968 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
13969 type. Diagnostic is produced if this requirement is not met.
13971 @item __is_class (type)
13972 If @code{type} is a cv class type, and not a union type
13973 ([basic.compound]) the trait is true, else it is false.
13975 @item __is_empty (type)
13976 If @code{__is_class (type)} is false then the trait is false.
13977 Otherwise @code{type} is considered empty if and only if: @code{type}
13978 has no non-static data members, or all non-static data members, if
13979 any, are bit-fields of length 0, and @code{type} has no virtual
13980 members, and @code{type} has no virtual base classes, and @code{type}
13981 has no base classes @code{base_type} for which
13982 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
13983 be a complete type, an array type of unknown bound, or is a
13986 @item __is_enum (type)
13987 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
13988 true, else it is false.
13990 @item __is_pod (type)
13991 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
13992 else it is false. Requires: @code{type} shall be a complete type,
13993 an array type of unknown bound, or is a @code{void} type.
13995 @item __is_polymorphic (type)
13996 If @code{type} is a polymorphic class ([class.virtual]) then the trait
13997 is true, else it is false. Requires: @code{type} shall be a complete
13998 type, an array type of unknown bound, or is a @code{void} type.
14000 @item __is_union (type)
14001 If @code{type} is a cv union type ([basic.compound]) the trait is
14002 true, else it is false.
14006 @node Java Exceptions
14007 @section Java Exceptions
14009 The Java language uses a slightly different exception handling model
14010 from C++. Normally, GNU C++ will automatically detect when you are
14011 writing C++ code that uses Java exceptions, and handle them
14012 appropriately. However, if C++ code only needs to execute destructors
14013 when Java exceptions are thrown through it, GCC will guess incorrectly.
14014 Sample problematic code is:
14017 struct S @{ ~S(); @};
14018 extern void bar(); // @r{is written in Java, and may throw exceptions}
14027 The usual effect of an incorrect guess is a link failure, complaining of
14028 a missing routine called @samp{__gxx_personality_v0}.
14030 You can inform the compiler that Java exceptions are to be used in a
14031 translation unit, irrespective of what it might think, by writing
14032 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
14033 @samp{#pragma} must appear before any functions that throw or catch
14034 exceptions, or run destructors when exceptions are thrown through them.
14036 You cannot mix Java and C++ exceptions in the same translation unit. It
14037 is believed to be safe to throw a C++ exception from one file through
14038 another file compiled for the Java exception model, or vice versa, but
14039 there may be bugs in this area.
14041 @node Deprecated Features
14042 @section Deprecated Features
14044 In the past, the GNU C++ compiler was extended to experiment with new
14045 features, at a time when the C++ language was still evolving. Now that
14046 the C++ standard is complete, some of those features are superseded by
14047 superior alternatives. Using the old features might cause a warning in
14048 some cases that the feature will be dropped in the future. In other
14049 cases, the feature might be gone already.
14051 While the list below is not exhaustive, it documents some of the options
14052 that are now deprecated:
14055 @item -fexternal-templates
14056 @itemx -falt-external-templates
14057 These are two of the many ways for G++ to implement template
14058 instantiation. @xref{Template Instantiation}. The C++ standard clearly
14059 defines how template definitions have to be organized across
14060 implementation units. G++ has an implicit instantiation mechanism that
14061 should work just fine for standard-conforming code.
14063 @item -fstrict-prototype
14064 @itemx -fno-strict-prototype
14065 Previously it was possible to use an empty prototype parameter list to
14066 indicate an unspecified number of parameters (like C), rather than no
14067 parameters, as C++ demands. This feature has been removed, except where
14068 it is required for backwards compatibility. @xref{Backwards Compatibility}.
14071 G++ allows a virtual function returning @samp{void *} to be overridden
14072 by one returning a different pointer type. This extension to the
14073 covariant return type rules is now deprecated and will be removed from a
14076 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
14077 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
14078 and are now removed from G++. Code using these operators should be
14079 modified to use @code{std::min} and @code{std::max} instead.
14081 The named return value extension has been deprecated, and is now
14084 The use of initializer lists with new expressions has been deprecated,
14085 and is now removed from G++.
14087 Floating and complex non-type template parameters have been deprecated,
14088 and are now removed from G++.
14090 The implicit typename extension has been deprecated and is now
14093 The use of default arguments in function pointers, function typedefs
14094 and other places where they are not permitted by the standard is
14095 deprecated and will be removed from a future version of G++.
14097 G++ allows floating-point literals to appear in integral constant expressions,
14098 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
14099 This extension is deprecated and will be removed from a future version.
14101 G++ allows static data members of const floating-point type to be declared
14102 with an initializer in a class definition. The standard only allows
14103 initializers for static members of const integral types and const
14104 enumeration types so this extension has been deprecated and will be removed
14105 from a future version.
14107 @node Backwards Compatibility
14108 @section Backwards Compatibility
14109 @cindex Backwards Compatibility
14110 @cindex ARM [Annotated C++ Reference Manual]
14112 Now that there is a definitive ISO standard C++, G++ has a specification
14113 to adhere to. The C++ language evolved over time, and features that
14114 used to be acceptable in previous drafts of the standard, such as the ARM
14115 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
14116 compilation of C++ written to such drafts, G++ contains some backwards
14117 compatibilities. @emph{All such backwards compatibility features are
14118 liable to disappear in future versions of G++.} They should be considered
14119 deprecated. @xref{Deprecated Features}.
14123 If a variable is declared at for scope, it used to remain in scope until
14124 the end of the scope which contained the for statement (rather than just
14125 within the for scope). G++ retains this, but issues a warning, if such a
14126 variable is accessed outside the for scope.
14128 @item Implicit C language
14129 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
14130 scope to set the language. On such systems, all header files are
14131 implicitly scoped inside a C language scope. Also, an empty prototype
14132 @code{()} will be treated as an unspecified number of arguments, rather
14133 than no arguments, as C++ demands.