1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012
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 * __sync Builtins:: Legacy built-in functions for atomic memory access.
83 * __atomic Builtins:: Atomic built-in functions with memory model.
84 * Object Size Checking:: Built-in functions for limited buffer overflow
86 * Other Builtins:: Other built-in functions.
87 * Target Builtins:: Built-in functions specific to particular targets.
88 * Target Format Checks:: Format checks specific to particular targets.
89 * Pragmas:: Pragmas accepted by GCC.
90 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
91 * Thread-Local:: Per-thread variables.
92 * Binary constants:: Binary constants using the @samp{0b} prefix.
96 @section Statements and Declarations in Expressions
97 @cindex statements inside expressions
98 @cindex declarations inside expressions
99 @cindex expressions containing statements
100 @cindex macros, statements in expressions
102 @c the above section title wrapped and causes an underfull hbox.. i
103 @c changed it from "within" to "in". --mew 4feb93
104 A compound statement enclosed in parentheses may appear as an expression
105 in GNU C@. This allows you to use loops, switches, and local variables
106 within an expression.
108 Recall that a compound statement is a sequence of statements surrounded
109 by braces; in this construct, parentheses go around the braces. For
113 (@{ int y = foo (); int z;
120 is a valid (though slightly more complex than necessary) expression
121 for the absolute value of @code{foo ()}.
123 The last thing in the compound statement should be an expression
124 followed by a semicolon; the value of this subexpression serves as the
125 value of the entire construct. (If you use some other kind of statement
126 last within the braces, the construct has type @code{void}, and thus
127 effectively no value.)
129 This feature is especially useful in making macro definitions ``safe'' (so
130 that they evaluate each operand exactly once). For example, the
131 ``maximum'' function is commonly defined as a macro in standard C as
135 #define max(a,b) ((a) > (b) ? (a) : (b))
139 @cindex side effects, macro argument
140 But this definition computes either @var{a} or @var{b} twice, with bad
141 results if the operand has side effects. In GNU C, if you know the
142 type of the operands (here taken as @code{int}), you can define
143 the macro safely as follows:
146 #define maxint(a,b) \
147 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
150 Embedded statements are not allowed in constant expressions, such as
151 the value of an enumeration constant, the width of a bit-field, or
152 the initial value of a static variable.
154 If you don't know the type of the operand, you can still do this, but you
155 must use @code{typeof} (@pxref{Typeof}).
157 In G++, the result value of a statement expression undergoes array and
158 function pointer decay, and is returned by value to the enclosing
159 expression. For instance, if @code{A} is a class, then
168 will construct a temporary @code{A} object to hold the result of the
169 statement expression, and that will be used to invoke @code{Foo}.
170 Therefore the @code{this} pointer observed by @code{Foo} will not be the
173 Any temporaries created within a statement within a statement expression
174 will be destroyed at the statement's end. This makes statement
175 expressions inside macros slightly different from function calls. In
176 the latter case temporaries introduced during argument evaluation will
177 be destroyed at the end of the statement that includes the function
178 call. In the statement expression case they will be destroyed during
179 the statement expression. For instance,
182 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
183 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
193 will have different places where temporaries are destroyed. For the
194 @code{macro} case, the temporary @code{X} will be destroyed just after
195 the initialization of @code{b}. In the @code{function} case that
196 temporary will be destroyed when the function returns.
198 These considerations mean that it is probably a bad idea to use
199 statement-expressions of this form in header files that are designed to
200 work with C++. (Note that some versions of the GNU C Library contained
201 header files using statement-expression that lead to precisely this
204 Jumping into a statement expression with @code{goto} or using a
205 @code{switch} statement outside the statement expression with a
206 @code{case} or @code{default} label inside the statement expression is
207 not permitted. Jumping into a statement expression with a computed
208 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
209 Jumping out of a statement expression is permitted, but if the
210 statement expression is part of a larger expression then it is
211 unspecified which other subexpressions of that expression have been
212 evaluated except where the language definition requires certain
213 subexpressions to be evaluated before or after the statement
214 expression. In any case, as with a function call the evaluation of a
215 statement expression is not interleaved with the evaluation of other
216 parts of the containing expression. For example,
219 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
223 will call @code{foo} and @code{bar1} and will not call @code{baz} but
224 may or may not call @code{bar2}. If @code{bar2} is called, it will be
225 called after @code{foo} and before @code{bar1}
228 @section Locally Declared Labels
230 @cindex macros, local labels
232 GCC allows you to declare @dfn{local labels} in any nested block
233 scope. A local label is just like an ordinary label, but you can
234 only reference it (with a @code{goto} statement, or by taking its
235 address) within the block in which it was declared.
237 A local label declaration looks like this:
240 __label__ @var{label};
247 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
250 Local label declarations must come at the beginning of the block,
251 before any ordinary declarations or statements.
253 The label declaration defines the label @emph{name}, but does not define
254 the label itself. You must do this in the usual way, with
255 @code{@var{label}:}, within the statements of the statement expression.
257 The local label feature is useful for complex macros. If a macro
258 contains nested loops, a @code{goto} can be useful for breaking out of
259 them. However, an ordinary label whose scope is the whole function
260 cannot be used: if the macro can be expanded several times in one
261 function, the label will be multiply defined in that function. A
262 local label avoids this problem. For example:
265 #define SEARCH(value, array, target) \
268 typeof (target) _SEARCH_target = (target); \
269 typeof (*(array)) *_SEARCH_array = (array); \
272 for (i = 0; i < max; i++) \
273 for (j = 0; j < max; j++) \
274 if (_SEARCH_array[i][j] == _SEARCH_target) \
275 @{ (value) = i; goto found; @} \
281 This could also be written using a statement-expression:
284 #define SEARCH(array, target) \
287 typeof (target) _SEARCH_target = (target); \
288 typeof (*(array)) *_SEARCH_array = (array); \
291 for (i = 0; i < max; i++) \
292 for (j = 0; j < max; j++) \
293 if (_SEARCH_array[i][j] == _SEARCH_target) \
294 @{ value = i; goto found; @} \
301 Local label declarations also make the labels they declare visible to
302 nested functions, if there are any. @xref{Nested Functions}, for details.
304 @node Labels as Values
305 @section Labels as Values
306 @cindex labels as values
307 @cindex computed gotos
308 @cindex goto with computed label
309 @cindex address of a label
311 You can get the address of a label defined in the current function
312 (or a containing function) with the unary operator @samp{&&}. The
313 value has type @code{void *}. This value is a constant and can be used
314 wherever a constant of that type is valid. For example:
322 To use these values, you need to be able to jump to one. This is done
323 with the computed goto statement@footnote{The analogous feature in
324 Fortran is called an assigned goto, but that name seems inappropriate in
325 C, where one can do more than simply store label addresses in label
326 variables.}, @code{goto *@var{exp};}. For example,
333 Any expression of type @code{void *} is allowed.
335 One way of using these constants is in initializing a static array that
336 will serve as a jump table:
339 static void *array[] = @{ &&foo, &&bar, &&hack @};
342 Then you can select a label with indexing, like this:
349 Note that this does not check whether the subscript is in bounds---array
350 indexing in C never does that.
352 Such an array of label values serves a purpose much like that of the
353 @code{switch} statement. The @code{switch} statement is cleaner, so
354 use that rather than an array unless the problem does not fit a
355 @code{switch} statement very well.
357 Another use of label values is in an interpreter for threaded code.
358 The labels within the interpreter function can be stored in the
359 threaded code for super-fast dispatching.
361 You may not use this mechanism to jump to code in a different function.
362 If you do that, totally unpredictable things will happen. The best way to
363 avoid this is to store the label address only in automatic variables and
364 never pass it as an argument.
366 An alternate way to write the above example is
369 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
371 goto *(&&foo + array[i]);
375 This is more friendly to code living in shared libraries, as it reduces
376 the number of dynamic relocations that are needed, and by consequence,
377 allows the data to be read-only.
379 The @code{&&foo} expressions for the same label might have different
380 values if the containing function is inlined or cloned. If a program
381 relies on them being always the same,
382 @code{__attribute__((__noinline__,__noclone__))} should be used to
383 prevent inlining and cloning. If @code{&&foo} is used in a static
384 variable initializer, inlining and cloning is forbidden.
386 @node Nested Functions
387 @section Nested Functions
388 @cindex nested functions
389 @cindex downward funargs
392 A @dfn{nested function} is a function defined inside another function.
393 (Nested functions are not supported for GNU C++.) The nested function's
394 name is local to the block where it is defined. For example, here we
395 define a nested function named @code{square}, and call it twice:
399 foo (double a, double b)
401 double square (double z) @{ return z * z; @}
403 return square (a) + square (b);
408 The nested function can access all the variables of the containing
409 function that are visible at the point of its definition. This is
410 called @dfn{lexical scoping}. For example, here we show a nested
411 function which uses an inherited variable named @code{offset}:
415 bar (int *array, int offset, int size)
417 int access (int *array, int index)
418 @{ return array[index + offset]; @}
421 for (i = 0; i < size; i++)
422 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
427 Nested function definitions are permitted within functions in the places
428 where variable definitions are allowed; that is, in any block, mixed
429 with the other declarations and statements in the block.
431 It is possible to call the nested function from outside the scope of its
432 name by storing its address or passing the address to another function:
435 hack (int *array, int size)
437 void store (int index, int value)
438 @{ array[index] = value; @}
440 intermediate (store, size);
444 Here, the function @code{intermediate} receives the address of
445 @code{store} as an argument. If @code{intermediate} calls @code{store},
446 the arguments given to @code{store} are used to store into @code{array}.
447 But this technique works only so long as the containing function
448 (@code{hack}, in this example) does not exit.
450 If you try to call the nested function through its address after the
451 containing function has exited, all hell will break loose. If you try
452 to call it after a containing scope level has exited, and if it refers
453 to some of the variables that are no longer in scope, you may be lucky,
454 but it's not wise to take the risk. If, however, the nested function
455 does not refer to anything that has gone out of scope, you should be
458 GCC implements taking the address of a nested function using a technique
459 called @dfn{trampolines}. This technique was described in
460 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
461 C++ Conference Proceedings, October 17-21, 1988).
463 A nested function can jump to a label inherited from a containing
464 function, provided the label was explicitly declared in the containing
465 function (@pxref{Local Labels}). Such a jump returns instantly to the
466 containing function, exiting the nested function which did the
467 @code{goto} and any intermediate functions as well. Here is an example:
471 bar (int *array, int offset, int size)
474 int access (int *array, int index)
478 return array[index + offset];
482 for (i = 0; i < size; i++)
483 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
487 /* @r{Control comes here from @code{access}
488 if it detects an error.} */
495 A nested function always has no linkage. Declaring one with
496 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
497 before its definition, use @code{auto} (which is otherwise meaningless
498 for function declarations).
501 bar (int *array, int offset, int size)
504 auto int access (int *, int);
506 int access (int *array, int index)
510 return array[index + offset];
516 @node Constructing Calls
517 @section Constructing Function Calls
518 @cindex constructing calls
519 @cindex forwarding calls
521 Using the built-in functions described below, you can record
522 the arguments a function received, and call another function
523 with the same arguments, without knowing the number or types
526 You can also record the return value of that function call,
527 and later return that value, without knowing what data type
528 the function tried to return (as long as your caller expects
531 However, these built-in functions may interact badly with some
532 sophisticated features or other extensions of the language. It
533 is, therefore, not recommended to use them outside very simple
534 functions acting as mere forwarders for their arguments.
536 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
537 This built-in function returns a pointer to data
538 describing how to perform a call with the same arguments as were passed
539 to the current function.
541 The function saves the arg pointer register, structure value address,
542 and all registers that might be used to pass arguments to a function
543 into a block of memory allocated on the stack. Then it returns the
544 address of that block.
547 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
548 This built-in function invokes @var{function}
549 with a copy of the parameters described by @var{arguments}
552 The value of @var{arguments} should be the value returned by
553 @code{__builtin_apply_args}. The argument @var{size} specifies the size
554 of the stack argument data, in bytes.
556 This function returns a pointer to data describing
557 how to return whatever value was returned by @var{function}. The data
558 is saved in a block of memory allocated on the stack.
560 It is not always simple to compute the proper value for @var{size}. The
561 value is used by @code{__builtin_apply} to compute the amount of data
562 that should be pushed on the stack and copied from the incoming argument
566 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
567 This built-in function returns the value described by @var{result} from
568 the containing function. You should specify, for @var{result}, a value
569 returned by @code{__builtin_apply}.
572 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
573 This built-in function represents all anonymous arguments of an inline
574 function. It can be used only in inline functions which will be always
575 inlined, never compiled as a separate function, such as those using
576 @code{__attribute__ ((__always_inline__))} or
577 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
578 It must be only passed as last argument to some other function
579 with variable arguments. This is useful for writing small wrapper
580 inlines for variable argument functions, when using preprocessor
581 macros is undesirable. For example:
583 extern int myprintf (FILE *f, const char *format, ...);
584 extern inline __attribute__ ((__gnu_inline__)) int
585 myprintf (FILE *f, const char *format, ...)
587 int r = fprintf (f, "myprintf: ");
590 int s = fprintf (f, format, __builtin_va_arg_pack ());
598 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
599 This built-in function returns the number of anonymous arguments of
600 an inline function. It can be used only in inline functions which
601 will be always inlined, never compiled as a separate function, such
602 as those using @code{__attribute__ ((__always_inline__))} or
603 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
604 For example following will do link or runtime checking of open
605 arguments for optimized code:
608 extern inline __attribute__((__gnu_inline__)) int
609 myopen (const char *path, int oflag, ...)
611 if (__builtin_va_arg_pack_len () > 1)
612 warn_open_too_many_arguments ();
614 if (__builtin_constant_p (oflag))
616 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
618 warn_open_missing_mode ();
619 return __open_2 (path, oflag);
621 return open (path, oflag, __builtin_va_arg_pack ());
624 if (__builtin_va_arg_pack_len () < 1)
625 return __open_2 (path, oflag);
627 return open (path, oflag, __builtin_va_arg_pack ());
634 @section Referring to a Type with @code{typeof}
637 @cindex macros, types of arguments
639 Another way to refer to the type of an expression is with @code{typeof}.
640 The syntax of using of this keyword looks like @code{sizeof}, but the
641 construct acts semantically like a type name defined with @code{typedef}.
643 There are two ways of writing the argument to @code{typeof}: with an
644 expression or with a type. Here is an example with an expression:
651 This assumes that @code{x} is an array of pointers to functions;
652 the type described is that of the values of the functions.
654 Here is an example with a typename as the argument:
661 Here the type described is that of pointers to @code{int}.
663 If you are writing a header file that must work when included in ISO C
664 programs, write @code{__typeof__} instead of @code{typeof}.
665 @xref{Alternate Keywords}.
667 A @code{typeof}-construct can be used anywhere a typedef name could be
668 used. For example, you can use it in a declaration, in a cast, or inside
669 of @code{sizeof} or @code{typeof}.
671 The operand of @code{typeof} is evaluated for its side effects if and
672 only if it is an expression of variably modified type or the name of
675 @code{typeof} is often useful in conjunction with the
676 statements-within-expressions feature. Here is how the two together can
677 be used to define a safe ``maximum'' macro that operates on any
678 arithmetic type and evaluates each of its arguments exactly once:
682 (@{ typeof (a) _a = (a); \
683 typeof (b) _b = (b); \
684 _a > _b ? _a : _b; @})
687 @cindex underscores in variables in macros
688 @cindex @samp{_} in variables in macros
689 @cindex local variables in macros
690 @cindex variables, local, in macros
691 @cindex macros, local variables in
693 The reason for using names that start with underscores for the local
694 variables is to avoid conflicts with variable names that occur within the
695 expressions that are substituted for @code{a} and @code{b}. Eventually we
696 hope to design a new form of declaration syntax that allows you to declare
697 variables whose scopes start only after their initializers; this will be a
698 more reliable way to prevent such conflicts.
701 Some more examples of the use of @code{typeof}:
705 This declares @code{y} with the type of what @code{x} points to.
712 This declares @code{y} as an array of such values.
719 This declares @code{y} as an array of pointers to characters:
722 typeof (typeof (char *)[4]) y;
726 It is equivalent to the following traditional C declaration:
732 To see the meaning of the declaration using @code{typeof}, and why it
733 might be a useful way to write, rewrite it with these macros:
736 #define pointer(T) typeof(T *)
737 #define array(T, N) typeof(T [N])
741 Now the declaration can be rewritten this way:
744 array (pointer (char), 4) y;
748 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
749 pointers to @code{char}.
752 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
753 a more limited extension which permitted one to write
756 typedef @var{T} = @var{expr};
760 with the effect of declaring @var{T} to have the type of the expression
761 @var{expr}. This extension does not work with GCC 3 (versions between
762 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
763 relies on it should be rewritten to use @code{typeof}:
766 typedef typeof(@var{expr}) @var{T};
770 This will work with all versions of GCC@.
773 @section Conditionals with Omitted Operands
774 @cindex conditional expressions, extensions
775 @cindex omitted middle-operands
776 @cindex middle-operands, omitted
777 @cindex extensions, @code{?:}
778 @cindex @code{?:} extensions
780 The middle operand in a conditional expression may be omitted. Then
781 if the first operand is nonzero, its value is the value of the conditional
784 Therefore, the expression
791 has the value of @code{x} if that is nonzero; otherwise, the value of
794 This example is perfectly equivalent to
800 @cindex side effect in @code{?:}
801 @cindex @code{?:} side effect
803 In this simple case, the ability to omit the middle operand is not
804 especially useful. When it becomes useful is when the first operand does,
805 or may (if it is a macro argument), contain a side effect. Then repeating
806 the operand in the middle would perform the side effect twice. Omitting
807 the middle operand uses the value already computed without the undesirable
808 effects of recomputing it.
811 @section 128-bits integers
812 @cindex @code{__int128} data types
814 As an extension the integer scalar type @code{__int128} is supported for
815 targets having an integer mode wide enough to hold 128-bit.
816 Simply write @code{__int128} for a signed 128-bit integer, or
817 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
818 support in GCC to express an integer constant of type @code{__int128}
819 for targets having @code{long long} integer with less then 128 bit width.
822 @section Double-Word Integers
823 @cindex @code{long long} data types
824 @cindex double-word arithmetic
825 @cindex multiprecision arithmetic
826 @cindex @code{LL} integer suffix
827 @cindex @code{ULL} integer suffix
829 ISO C99 supports data types for integers that are at least 64 bits wide,
830 and as an extension GCC supports them in C90 mode and in C++.
831 Simply write @code{long long int} for a signed integer, or
832 @code{unsigned long long int} for an unsigned integer. To make an
833 integer constant of type @code{long long int}, add the suffix @samp{LL}
834 to the integer. To make an integer constant of type @code{unsigned long
835 long int}, add the suffix @samp{ULL} to the integer.
837 You can use these types in arithmetic like any other integer types.
838 Addition, subtraction, and bitwise boolean operations on these types
839 are open-coded on all types of machines. Multiplication is open-coded
840 if the machine supports fullword-to-doubleword a widening multiply
841 instruction. Division and shifts are open-coded only on machines that
842 provide special support. The operations that are not open-coded use
843 special library routines that come with GCC@.
845 There may be pitfalls when you use @code{long long} types for function
846 arguments, unless you declare function prototypes. If a function
847 expects type @code{int} for its argument, and you pass a value of type
848 @code{long long int}, confusion will result because the caller and the
849 subroutine will disagree about the number of bytes for the argument.
850 Likewise, if the function expects @code{long long int} and you pass
851 @code{int}. The best way to avoid such problems is to use prototypes.
854 @section Complex Numbers
855 @cindex complex numbers
856 @cindex @code{_Complex} keyword
857 @cindex @code{__complex__} keyword
859 ISO C99 supports complex floating data types, and as an extension GCC
860 supports them in C90 mode and in C++, and supports complex integer data
861 types which are not part of ISO C99. You can declare complex types
862 using the keyword @code{_Complex}. As an extension, the older GNU
863 keyword @code{__complex__} is also supported.
865 For example, @samp{_Complex double x;} declares @code{x} as a
866 variable whose real part and imaginary part are both of type
867 @code{double}. @samp{_Complex short int y;} declares @code{y} to
868 have real and imaginary parts of type @code{short int}; this is not
869 likely to be useful, but it shows that the set of complex types is
872 To write a constant with a complex data type, use the suffix @samp{i} or
873 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
874 has type @code{_Complex float} and @code{3i} has type
875 @code{_Complex int}. Such a constant always has a pure imaginary
876 value, but you can form any complex value you like by adding one to a
877 real constant. This is a GNU extension; if you have an ISO C99
878 conforming C library (such as GNU libc), and want to construct complex
879 constants of floating type, you should include @code{<complex.h>} and
880 use the macros @code{I} or @code{_Complex_I} instead.
882 @cindex @code{__real__} keyword
883 @cindex @code{__imag__} keyword
884 To extract the real part of a complex-valued expression @var{exp}, write
885 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
886 extract the imaginary part. This is a GNU extension; for values of
887 floating type, you should use the ISO C99 functions @code{crealf},
888 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
889 @code{cimagl}, declared in @code{<complex.h>} and also provided as
890 built-in functions by GCC@.
892 @cindex complex conjugation
893 The operator @samp{~} performs complex conjugation when used on a value
894 with a complex type. This is a GNU extension; for values of
895 floating type, you should use the ISO C99 functions @code{conjf},
896 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
897 provided as built-in functions by GCC@.
899 GCC can allocate complex automatic variables in a noncontiguous
900 fashion; it's even possible for the real part to be in a register while
901 the imaginary part is on the stack (or vice-versa). Only the DWARF2
902 debug info format can represent this, so use of DWARF2 is recommended.
903 If you are using the stabs debug info format, GCC describes a noncontiguous
904 complex variable as if it were two separate variables of noncomplex type.
905 If the variable's actual name is @code{foo}, the two fictitious
906 variables are named @code{foo$real} and @code{foo$imag}. You can
907 examine and set these two fictitious variables with your debugger.
910 @section Additional Floating Types
911 @cindex additional floating types
912 @cindex @code{__float80} data type
913 @cindex @code{__float128} data type
914 @cindex @code{w} floating point suffix
915 @cindex @code{q} floating point suffix
916 @cindex @code{W} floating point suffix
917 @cindex @code{Q} floating point suffix
919 As an extension, the GNU C compiler supports additional floating
920 types, @code{__float80} and @code{__float128} to support 80bit
921 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
922 Support for additional types includes the arithmetic operators:
923 add, subtract, multiply, divide; unary arithmetic operators;
924 relational operators; equality operators; and conversions to and from
925 integer and other floating types. Use a suffix @samp{w} or @samp{W}
926 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
927 for @code{_float128}. You can declare complex types using the
928 corresponding internal complex type, @code{XCmode} for @code{__float80}
929 type and @code{TCmode} for @code{__float128} type:
932 typedef _Complex float __attribute__((mode(TC))) _Complex128;
933 typedef _Complex float __attribute__((mode(XC))) _Complex80;
936 Not all targets support additional floating point types. @code{__float80}
937 and @code{__float128} types are supported on i386, x86_64 and ia64 targets.
938 The @code{__float128} type is supported on hppa HP-UX targets.
941 @section Half-Precision Floating Point
942 @cindex half-precision floating point
943 @cindex @code{__fp16} data type
945 On ARM targets, GCC supports half-precision (16-bit) floating point via
946 the @code{__fp16} type. You must enable this type explicitly
947 with the @option{-mfp16-format} command-line option in order to use it.
949 ARM supports two incompatible representations for half-precision
950 floating-point values. You must choose one of the representations and
951 use it consistently in your program.
953 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
954 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
955 There are 11 bits of significand precision, approximately 3
958 Specifying @option{-mfp16-format=alternative} selects the ARM
959 alternative format. This representation is similar to the IEEE
960 format, but does not support infinities or NaNs. Instead, the range
961 of exponents is extended, so that this format can represent normalized
962 values in the range of @math{2^{-14}} to 131008.
964 The @code{__fp16} type is a storage format only. For purposes
965 of arithmetic and other operations, @code{__fp16} values in C or C++
966 expressions are automatically promoted to @code{float}. In addition,
967 you cannot declare a function with a return value or parameters
968 of type @code{__fp16}.
970 Note that conversions from @code{double} to @code{__fp16}
971 involve an intermediate conversion to @code{float}. Because
972 of rounding, this can sometimes produce a different result than a
975 ARM provides hardware support for conversions between
976 @code{__fp16} and @code{float} values
977 as an extension to VFP and NEON (Advanced SIMD). GCC generates
978 code using these hardware instructions if you compile with
979 options to select an FPU that provides them;
980 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
981 in addition to the @option{-mfp16-format} option to select
982 a half-precision format.
984 Language-level support for the @code{__fp16} data type is
985 independent of whether GCC generates code using hardware floating-point
986 instructions. In cases where hardware support is not specified, GCC
987 implements conversions between @code{__fp16} and @code{float} values
991 @section Decimal Floating Types
992 @cindex decimal floating types
993 @cindex @code{_Decimal32} data type
994 @cindex @code{_Decimal64} data type
995 @cindex @code{_Decimal128} data type
996 @cindex @code{df} integer suffix
997 @cindex @code{dd} integer suffix
998 @cindex @code{dl} integer suffix
999 @cindex @code{DF} integer suffix
1000 @cindex @code{DD} integer suffix
1001 @cindex @code{DL} integer suffix
1003 As an extension, the GNU C compiler supports decimal floating types as
1004 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1005 floating types in GCC will evolve as the draft technical report changes.
1006 Calling conventions for any target might also change. Not all targets
1007 support decimal floating types.
1009 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1010 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1011 @code{float}, @code{double}, and @code{long double} whose radix is not
1012 specified by the C standard but is usually two.
1014 Support for decimal floating types includes the arithmetic operators
1015 add, subtract, multiply, divide; unary arithmetic operators;
1016 relational operators; equality operators; and conversions to and from
1017 integer and other floating types. Use a suffix @samp{df} or
1018 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1019 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1022 GCC support of decimal float as specified by the draft technical report
1027 When the value of a decimal floating type cannot be represented in the
1028 integer type to which it is being converted, the result is undefined
1029 rather than the result value specified by the draft technical report.
1032 GCC does not provide the C library functionality associated with
1033 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1034 @file{wchar.h}, which must come from a separate C library implementation.
1035 Because of this the GNU C compiler does not define macro
1036 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1037 the technical report.
1040 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1041 are supported by the DWARF2 debug information format.
1047 ISO C99 supports floating-point numbers written not only in the usual
1048 decimal notation, such as @code{1.55e1}, but also numbers such as
1049 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1050 supports this in C90 mode (except in some cases when strictly
1051 conforming) and in C++. In that format the
1052 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1053 mandatory. The exponent is a decimal number that indicates the power of
1054 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1061 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1062 is the same as @code{1.55e1}.
1064 Unlike for floating-point numbers in the decimal notation the exponent
1065 is always required in the hexadecimal notation. Otherwise the compiler
1066 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1067 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1068 extension for floating-point constants of type @code{float}.
1071 @section Fixed-Point Types
1072 @cindex fixed-point types
1073 @cindex @code{_Fract} data type
1074 @cindex @code{_Accum} data type
1075 @cindex @code{_Sat} data type
1076 @cindex @code{hr} fixed-suffix
1077 @cindex @code{r} fixed-suffix
1078 @cindex @code{lr} fixed-suffix
1079 @cindex @code{llr} fixed-suffix
1080 @cindex @code{uhr} fixed-suffix
1081 @cindex @code{ur} fixed-suffix
1082 @cindex @code{ulr} fixed-suffix
1083 @cindex @code{ullr} fixed-suffix
1084 @cindex @code{hk} fixed-suffix
1085 @cindex @code{k} fixed-suffix
1086 @cindex @code{lk} fixed-suffix
1087 @cindex @code{llk} fixed-suffix
1088 @cindex @code{uhk} fixed-suffix
1089 @cindex @code{uk} fixed-suffix
1090 @cindex @code{ulk} fixed-suffix
1091 @cindex @code{ullk} fixed-suffix
1092 @cindex @code{HR} fixed-suffix
1093 @cindex @code{R} fixed-suffix
1094 @cindex @code{LR} fixed-suffix
1095 @cindex @code{LLR} fixed-suffix
1096 @cindex @code{UHR} fixed-suffix
1097 @cindex @code{UR} fixed-suffix
1098 @cindex @code{ULR} fixed-suffix
1099 @cindex @code{ULLR} fixed-suffix
1100 @cindex @code{HK} fixed-suffix
1101 @cindex @code{K} fixed-suffix
1102 @cindex @code{LK} fixed-suffix
1103 @cindex @code{LLK} fixed-suffix
1104 @cindex @code{UHK} fixed-suffix
1105 @cindex @code{UK} fixed-suffix
1106 @cindex @code{ULK} fixed-suffix
1107 @cindex @code{ULLK} fixed-suffix
1109 As an extension, the GNU C compiler supports fixed-point types as
1110 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1111 types in GCC will evolve as the draft technical report changes.
1112 Calling conventions for any target might also change. Not all targets
1113 support fixed-point types.
1115 The fixed-point types are
1116 @code{short _Fract},
1119 @code{long long _Fract},
1120 @code{unsigned short _Fract},
1121 @code{unsigned _Fract},
1122 @code{unsigned long _Fract},
1123 @code{unsigned long long _Fract},
1124 @code{_Sat short _Fract},
1126 @code{_Sat long _Fract},
1127 @code{_Sat long long _Fract},
1128 @code{_Sat unsigned short _Fract},
1129 @code{_Sat unsigned _Fract},
1130 @code{_Sat unsigned long _Fract},
1131 @code{_Sat unsigned long long _Fract},
1132 @code{short _Accum},
1135 @code{long long _Accum},
1136 @code{unsigned short _Accum},
1137 @code{unsigned _Accum},
1138 @code{unsigned long _Accum},
1139 @code{unsigned long long _Accum},
1140 @code{_Sat short _Accum},
1142 @code{_Sat long _Accum},
1143 @code{_Sat long long _Accum},
1144 @code{_Sat unsigned short _Accum},
1145 @code{_Sat unsigned _Accum},
1146 @code{_Sat unsigned long _Accum},
1147 @code{_Sat unsigned long long _Accum}.
1149 Fixed-point data values contain fractional and optional integral parts.
1150 The format of fixed-point data varies and depends on the target machine.
1152 Support for fixed-point types includes:
1155 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1157 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1159 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1161 binary shift operators (@code{<<}, @code{>>})
1163 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1165 equality operators (@code{==}, @code{!=})
1167 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1168 @code{<<=}, @code{>>=})
1170 conversions to and from integer, floating-point, or fixed-point types
1173 Use a suffix in a fixed-point literal constant:
1175 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1176 @code{_Sat short _Fract}
1177 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1178 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1179 @code{_Sat long _Fract}
1180 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1181 @code{_Sat long long _Fract}
1182 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1183 @code{_Sat unsigned short _Fract}
1184 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1185 @code{_Sat unsigned _Fract}
1186 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1187 @code{_Sat unsigned long _Fract}
1188 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1189 and @code{_Sat unsigned long long _Fract}
1190 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1191 @code{_Sat short _Accum}
1192 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1193 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1194 @code{_Sat long _Accum}
1195 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1196 @code{_Sat long long _Accum}
1197 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1198 @code{_Sat unsigned short _Accum}
1199 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1200 @code{_Sat unsigned _Accum}
1201 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1202 @code{_Sat unsigned long _Accum}
1203 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1204 and @code{_Sat unsigned long long _Accum}
1207 GCC support of fixed-point types as specified by the draft technical report
1212 Pragmas to control overflow and rounding behaviors are not implemented.
1215 Fixed-point types are supported by the DWARF2 debug information format.
1217 @node Named Address Spaces
1218 @section Named Address Spaces
1219 @cindex Named Address Spaces
1221 As an extension, the GNU C compiler supports named address spaces as
1222 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1223 address spaces in GCC will evolve as the draft technical report
1224 changes. Calling conventions for any target might also change. At
1225 present, only the AVR, SPU, M32C, and RL78 targets support address
1226 spaces other than the generic address space.
1228 Address space identifiers may be used exactly like any other C type
1229 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1230 document for more details.
1232 @anchor{AVR Named Address Spaces}
1233 @subsection AVR Named Address Spaces
1235 On the AVR target, there are several address spaces that can be used
1236 in order to put read-only data into the flash memory and access that
1237 data by means of the special instructions @code{LPM} or @code{ELPM}
1238 needed to read from flash.
1240 Per default, any data including read-only data is located in RAM
1241 (the generic address space) so that non-generic address spaces are
1242 needed to locate read-only data in flash memory
1243 @emph{and} to generate the right instructions to access this data
1244 without using (inline) assembler code.
1248 @cindex @code{__flash} AVR Named Address Spaces
1249 The @code{__flash} qualifier will locate data in the
1250 @code{.progmem.data} section. Data will be read using the @code{LPM}
1251 instruction. Pointers to this address space are 16 bits wide.
1258 @cindex @code{__flash1} AVR Named Address Spaces
1259 @cindex @code{__flash2} AVR Named Address Spaces
1260 @cindex @code{__flash3} AVR Named Address Spaces
1261 @cindex @code{__flash4} AVR Named Address Spaces
1262 @cindex @code{__flash5} AVR Named Address Spaces
1263 These are 16-bit address spaces locating data in section
1264 @code{.progmem@var{N}.data} where @var{N} refers to
1265 address space @code{__flash@var{N}}.
1266 The compiler will set the @code{RAMPZ} segment register approptiately
1267 before reading data by means of the @code{ELPM} instruction.
1270 @cindex @code{__memx} AVR Named Address Spaces
1271 This is a 24-bit address space that linearizes flash and RAM:
1272 If the high bit of the address is set, data is read from
1273 RAM using the lower two bytes as RAM address.
1274 If the high bit of the address is clear, data is read from flash
1275 with @code{RAMPZ} set according to the high byte of the address.
1277 Objects in this address space will be located in @code{.progmemx.data}.
1283 char my_read (const __flash char ** p)
1285 /* p is a pointer to RAM that points to a pointer to flash.
1286 The first indirection of p will read that flash pointer
1287 from RAM and the second indirection reads a char from this
1293 /* Locate array[] in flash memory */
1294 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1300 /* Return 17 by reading from flash memory */
1301 return array[array[i]];
1305 For each named address space supported by avr-gcc there is an equally
1306 named but uppercase built-in macro defined.
1307 The purpose is to facilitate testing if respective address space
1308 support is available or not:
1312 const __flash int var = 1;
1319 #include <avr/pgmspace.h> /* From AVR-LibC */
1321 const int var PROGMEM = 1;
1325 return (int) pgm_read_word (&var);
1327 #endif /* __FLASH */
1330 Notice that attribute @ref{AVR Variable Attributes,@code{progmem}}
1331 locates data in flash but
1332 accesses to these data will read from generic address space, i.e.@:
1334 so that you need special accessors like @code{pgm_read_byte}
1335 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1336 together with attribute @code{progmem}.
1338 @b{Limitations and caveats}
1342 Reading across the 64@tie{}KiB section boundary of
1343 the @code{__flash} or @code{__flash@var{N}} address spaces
1344 will show undefined behaviour. The only address space that
1345 supports reading across the 64@tie{}KiB flash segment boundaries is
1349 If you use one of the @code{__flash@var{N}} address spaces
1350 you will have to arrange your linker skript to locate the
1351 @code{.progmem@var{N}.data} sections according to your needs.
1354 Any data or pointers to the non-generic address spaces must
1355 be qualified as @code{const}, i.e.@: as read-only data.
1356 This still applies if the data in one of these address
1357 spaces like software version number or calibration lookup table are intended to
1358 be changed after load time by, say, a boot loader. In this case
1359 the right qualification is @code{const} @code{volatile} so that the compiler
1360 must not optimize away known values or insert them
1361 as immediates into operands of instructions.
1364 Code like the following is not yet supported because of missing
1365 support in avr-binutils,
1366 see @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1368 extern const __memx char foo;
1369 const __memx void *pfoo = &foo;
1371 The code will throw an assembler warning and the high byte of
1372 @code{pfoo} will be initialized with@tie{}@code{0}, i.e.@: the
1373 initialization will be as if @code{foo} was located in the first
1374 64@tie{}KiB chunk of flash.
1378 @subsection M32C Named Address Spaces
1379 @cindex @code{__far} M32C Named Address Spaces
1381 On the M32C target, with the R8C and M16C cpu variants, variables
1382 qualified with @code{__far} are accessed using 32-bit addresses in
1383 order to access memory beyond the first 64@tie{}Ki bytes. If
1384 @code{__far} is used with the M32CM or M32C cpu variants, it has no
1387 @subsection RL78 Named Address Spaces
1388 @cindex @code{__far} RL78 Named Address Spaces
1390 On the RL78 target, variables qualified with @code{__far} are accessed
1391 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1392 addresses. Non-far variables are assumed to appear in the topmost
1393 64@tie{}KiB of the address space.
1395 @subsection SPU Named Address Spaces
1396 @cindex @code{__ea} SPU Named Address Spaces
1398 On the SPU target variables may be declared as
1399 belonging to another address space by qualifying the type with the
1400 @code{__ea} address space identifier:
1406 When the variable @code{i} is accessed, the compiler will generate
1407 special code to access this variable. It may use runtime library
1408 support, or generate special machine instructions to access that address
1412 @section Arrays of Length Zero
1413 @cindex arrays of length zero
1414 @cindex zero-length arrays
1415 @cindex length-zero arrays
1416 @cindex flexible array members
1418 Zero-length arrays are allowed in GNU C@. They are very useful as the
1419 last element of a structure which is really a header for a variable-length
1428 struct line *thisline = (struct line *)
1429 malloc (sizeof (struct line) + this_length);
1430 thisline->length = this_length;
1433 In ISO C90, you would have to give @code{contents} a length of 1, which
1434 means either you waste space or complicate the argument to @code{malloc}.
1436 In ISO C99, you would use a @dfn{flexible array member}, which is
1437 slightly different in syntax and semantics:
1441 Flexible array members are written as @code{contents[]} without
1445 Flexible array members have incomplete type, and so the @code{sizeof}
1446 operator may not be applied. As a quirk of the original implementation
1447 of zero-length arrays, @code{sizeof} evaluates to zero.
1450 Flexible array members may only appear as the last member of a
1451 @code{struct} that is otherwise non-empty.
1454 A structure containing a flexible array member, or a union containing
1455 such a structure (possibly recursively), may not be a member of a
1456 structure or an element of an array. (However, these uses are
1457 permitted by GCC as extensions.)
1460 GCC versions before 3.0 allowed zero-length arrays to be statically
1461 initialized, as if they were flexible arrays. In addition to those
1462 cases that were useful, it also allowed initializations in situations
1463 that would corrupt later data. Non-empty initialization of zero-length
1464 arrays is now treated like any case where there are more initializer
1465 elements than the array holds, in that a suitable warning about "excess
1466 elements in array" is given, and the excess elements (all of them, in
1467 this case) are ignored.
1469 Instead GCC allows static initialization of flexible array members.
1470 This is equivalent to defining a new structure containing the original
1471 structure followed by an array of sufficient size to contain the data.
1472 I.e.@: in the following, @code{f1} is constructed as if it were declared
1478 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1481 struct f1 f1; int data[3];
1482 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1486 The convenience of this extension is that @code{f1} has the desired
1487 type, eliminating the need to consistently refer to @code{f2.f1}.
1489 This has symmetry with normal static arrays, in that an array of
1490 unknown size is also written with @code{[]}.
1492 Of course, this extension only makes sense if the extra data comes at
1493 the end of a top-level object, as otherwise we would be overwriting
1494 data at subsequent offsets. To avoid undue complication and confusion
1495 with initialization of deeply nested arrays, we simply disallow any
1496 non-empty initialization except when the structure is the top-level
1497 object. For example:
1500 struct foo @{ int x; int y[]; @};
1501 struct bar @{ struct foo z; @};
1503 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1504 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1505 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1506 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1509 @node Empty Structures
1510 @section Structures With No Members
1511 @cindex empty structures
1512 @cindex zero-size structures
1514 GCC permits a C structure to have no members:
1521 The structure will have size zero. In C++, empty structures are part
1522 of the language. G++ treats empty structures as if they had a single
1523 member of type @code{char}.
1525 @node Variable Length
1526 @section Arrays of Variable Length
1527 @cindex variable-length arrays
1528 @cindex arrays of variable length
1531 Variable-length automatic arrays are allowed in ISO C99, and as an
1532 extension GCC accepts them in C90 mode and in C++. These arrays are
1533 declared like any other automatic arrays, but with a length that is not
1534 a constant expression. The storage is allocated at the point of
1535 declaration and deallocated when the brace-level is exited. For
1540 concat_fopen (char *s1, char *s2, char *mode)
1542 char str[strlen (s1) + strlen (s2) + 1];
1545 return fopen (str, mode);
1549 @cindex scope of a variable length array
1550 @cindex variable-length array scope
1551 @cindex deallocating variable length arrays
1552 Jumping or breaking out of the scope of the array name deallocates the
1553 storage. Jumping into the scope is not allowed; you get an error
1556 @cindex @code{alloca} vs variable-length arrays
1557 You can use the function @code{alloca} to get an effect much like
1558 variable-length arrays. The function @code{alloca} is available in
1559 many other C implementations (but not in all). On the other hand,
1560 variable-length arrays are more elegant.
1562 There are other differences between these two methods. Space allocated
1563 with @code{alloca} exists until the containing @emph{function} returns.
1564 The space for a variable-length array is deallocated as soon as the array
1565 name's scope ends. (If you use both variable-length arrays and
1566 @code{alloca} in the same function, deallocation of a variable-length array
1567 will also deallocate anything more recently allocated with @code{alloca}.)
1569 You can also use variable-length arrays as arguments to functions:
1573 tester (int len, char data[len][len])
1579 The length of an array is computed once when the storage is allocated
1580 and is remembered for the scope of the array in case you access it with
1583 If you want to pass the array first and the length afterward, you can
1584 use a forward declaration in the parameter list---another GNU extension.
1588 tester (int len; char data[len][len], int len)
1594 @cindex parameter forward declaration
1595 The @samp{int len} before the semicolon is a @dfn{parameter forward
1596 declaration}, and it serves the purpose of making the name @code{len}
1597 known when the declaration of @code{data} is parsed.
1599 You can write any number of such parameter forward declarations in the
1600 parameter list. They can be separated by commas or semicolons, but the
1601 last one must end with a semicolon, which is followed by the ``real''
1602 parameter declarations. Each forward declaration must match a ``real''
1603 declaration in parameter name and data type. ISO C99 does not support
1604 parameter forward declarations.
1606 @node Variadic Macros
1607 @section Macros with a Variable Number of Arguments.
1608 @cindex variable number of arguments
1609 @cindex macro with variable arguments
1610 @cindex rest argument (in macro)
1611 @cindex variadic macros
1613 In the ISO C standard of 1999, a macro can be declared to accept a
1614 variable number of arguments much as a function can. The syntax for
1615 defining the macro is similar to that of a function. Here is an
1619 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1622 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1623 such a macro, it represents the zero or more tokens until the closing
1624 parenthesis that ends the invocation, including any commas. This set of
1625 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1626 wherever it appears. See the CPP manual for more information.
1628 GCC has long supported variadic macros, and used a different syntax that
1629 allowed you to give a name to the variable arguments just like any other
1630 argument. Here is an example:
1633 #define debug(format, args...) fprintf (stderr, format, args)
1636 This is in all ways equivalent to the ISO C example above, but arguably
1637 more readable and descriptive.
1639 GNU CPP has two further variadic macro extensions, and permits them to
1640 be used with either of the above forms of macro definition.
1642 In standard C, you are not allowed to leave the variable argument out
1643 entirely; but you are allowed to pass an empty argument. For example,
1644 this invocation is invalid in ISO C, because there is no comma after
1651 GNU CPP permits you to completely omit the variable arguments in this
1652 way. In the above examples, the compiler would complain, though since
1653 the expansion of the macro still has the extra comma after the format
1656 To help solve this problem, CPP behaves specially for variable arguments
1657 used with the token paste operator, @samp{##}. If instead you write
1660 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1663 and if the variable arguments are omitted or empty, the @samp{##}
1664 operator causes the preprocessor to remove the comma before it. If you
1665 do provide some variable arguments in your macro invocation, GNU CPP
1666 does not complain about the paste operation and instead places the
1667 variable arguments after the comma. Just like any other pasted macro
1668 argument, these arguments are not macro expanded.
1670 @node Escaped Newlines
1671 @section Slightly Looser Rules for Escaped Newlines
1672 @cindex escaped newlines
1673 @cindex newlines (escaped)
1675 Recently, the preprocessor has relaxed its treatment of escaped
1676 newlines. Previously, the newline had to immediately follow a
1677 backslash. The current implementation allows whitespace in the form
1678 of spaces, horizontal and vertical tabs, and form feeds between the
1679 backslash and the subsequent newline. The preprocessor issues a
1680 warning, but treats it as a valid escaped newline and combines the two
1681 lines to form a single logical line. This works within comments and
1682 tokens, as well as between tokens. Comments are @emph{not} treated as
1683 whitespace for the purposes of this relaxation, since they have not
1684 yet been replaced with spaces.
1687 @section Non-Lvalue Arrays May Have Subscripts
1688 @cindex subscripting
1689 @cindex arrays, non-lvalue
1691 @cindex subscripting and function values
1692 In ISO C99, arrays that are not lvalues still decay to pointers, and
1693 may be subscripted, although they may not be modified or used after
1694 the next sequence point and the unary @samp{&} operator may not be
1695 applied to them. As an extension, GCC allows such arrays to be
1696 subscripted in C90 mode, though otherwise they do not decay to
1697 pointers outside C99 mode. For example,
1698 this is valid in GNU C though not valid in C90:
1702 struct foo @{int a[4];@};
1708 return f().a[index];
1714 @section Arithmetic on @code{void}- and Function-Pointers
1715 @cindex void pointers, arithmetic
1716 @cindex void, size of pointer to
1717 @cindex function pointers, arithmetic
1718 @cindex function, size of pointer to
1720 In GNU C, addition and subtraction operations are supported on pointers to
1721 @code{void} and on pointers to functions. This is done by treating the
1722 size of a @code{void} or of a function as 1.
1724 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1725 and on function types, and returns 1.
1727 @opindex Wpointer-arith
1728 The option @option{-Wpointer-arith} requests a warning if these extensions
1732 @section Non-Constant Initializers
1733 @cindex initializers, non-constant
1734 @cindex non-constant initializers
1736 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1737 automatic variable are not required to be constant expressions in GNU C@.
1738 Here is an example of an initializer with run-time varying elements:
1741 foo (float f, float g)
1743 float beat_freqs[2] = @{ f-g, f+g @};
1748 @node Compound Literals
1749 @section Compound Literals
1750 @cindex constructor expressions
1751 @cindex initializations in expressions
1752 @cindex structures, constructor expression
1753 @cindex expressions, constructor
1754 @cindex compound literals
1755 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1757 ISO C99 supports compound literals. A compound literal looks like
1758 a cast containing an initializer. Its value is an object of the
1759 type specified in the cast, containing the elements specified in
1760 the initializer; it is an lvalue. As an extension, GCC supports
1761 compound literals in C90 mode and in C++, though the semantics are
1762 somewhat different in C++.
1764 Usually, the specified type is a structure. Assume that
1765 @code{struct foo} and @code{structure} are declared as shown:
1768 struct foo @{int a; char b[2];@} structure;
1772 Here is an example of constructing a @code{struct foo} with a compound literal:
1775 structure = ((struct foo) @{x + y, 'a', 0@});
1779 This is equivalent to writing the following:
1783 struct foo temp = @{x + y, 'a', 0@};
1788 You can also construct an array, though this is dangerous in C++, as
1789 explained below. If all the elements of the compound literal are
1790 (made up of) simple constant expressions, suitable for use in
1791 initializers of objects of static storage duration, then the compound
1792 literal can be coerced to a pointer to its first element and used in
1793 such an initializer, as shown here:
1796 char **foo = (char *[]) @{ "x", "y", "z" @};
1799 Compound literals for scalar types and union types are
1800 also allowed, but then the compound literal is equivalent
1803 As a GNU extension, GCC allows initialization of objects with static storage
1804 duration by compound literals (which is not possible in ISO C99, because
1805 the initializer is not a constant).
1806 It is handled as if the object was initialized only with the bracket
1807 enclosed list if the types of the compound literal and the object match.
1808 The initializer list of the compound literal must be constant.
1809 If the object being initialized has array type of unknown size, the size is
1810 determined by compound literal size.
1813 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1814 static int y[] = (int []) @{1, 2, 3@};
1815 static int z[] = (int [3]) @{1@};
1819 The above lines are equivalent to the following:
1821 static struct foo x = @{1, 'a', 'b'@};
1822 static int y[] = @{1, 2, 3@};
1823 static int z[] = @{1, 0, 0@};
1826 In C, a compound literal designates an unnamed object with static or
1827 automatic storage duration. In C++, a compound literal designates a
1828 temporary object, which only lives until the end of its
1829 full-expression. As a result, well-defined C code that takes the
1830 address of a subobject of a compound literal can be undefined in C++.
1831 For instance, if the array compound literal example above appeared
1832 inside a function, any subsequent use of @samp{foo} in C++ has
1833 undefined behavior because the lifetime of the array ends after the
1834 declaration of @samp{foo}. As a result, the C++ compiler now rejects
1835 the conversion of a temporary array to a pointer.
1837 As an optimization, the C++ compiler sometimes gives array compound
1838 literals longer lifetimes: when the array either appears outside a
1839 function or has const-qualified type. If @samp{foo} and its
1840 initializer had elements of @samp{char *const} type rather than
1841 @samp{char *}, or if @samp{foo} were a global variable, the array
1842 would have static storage duration. But it is probably safest just to
1843 avoid the use of array compound literals in code compiled as C++.
1845 @node Designated Inits
1846 @section Designated Initializers
1847 @cindex initializers with labeled elements
1848 @cindex labeled elements in initializers
1849 @cindex case labels in initializers
1850 @cindex designated initializers
1852 Standard C90 requires the elements of an initializer to appear in a fixed
1853 order, the same as the order of the elements in the array or structure
1856 In ISO C99 you can give the elements in any order, specifying the array
1857 indices or structure field names they apply to, and GNU C allows this as
1858 an extension in C90 mode as well. This extension is not
1859 implemented in GNU C++.
1861 To specify an array index, write
1862 @samp{[@var{index}] =} before the element value. For example,
1865 int a[6] = @{ [4] = 29, [2] = 15 @};
1872 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1876 The index values must be constant expressions, even if the array being
1877 initialized is automatic.
1879 An alternative syntax for this which has been obsolete since GCC 2.5 but
1880 GCC still accepts is to write @samp{[@var{index}]} before the element
1881 value, with no @samp{=}.
1883 To initialize a range of elements to the same value, write
1884 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1885 extension. For example,
1888 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1892 If the value in it has side-effects, the side-effects will happen only once,
1893 not for each initialized field by the range initializer.
1896 Note that the length of the array is the highest value specified
1899 In a structure initializer, specify the name of a field to initialize
1900 with @samp{.@var{fieldname} =} before the element value. For example,
1901 given the following structure,
1904 struct point @{ int x, y; @};
1908 the following initialization
1911 struct point p = @{ .y = yvalue, .x = xvalue @};
1918 struct point p = @{ xvalue, yvalue @};
1921 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1922 @samp{@var{fieldname}:}, as shown here:
1925 struct point p = @{ y: yvalue, x: xvalue @};
1929 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1930 @dfn{designator}. You can also use a designator (or the obsolete colon
1931 syntax) when initializing a union, to specify which element of the union
1932 should be used. For example,
1935 union foo @{ int i; double d; @};
1937 union foo f = @{ .d = 4 @};
1941 will convert 4 to a @code{double} to store it in the union using
1942 the second element. By contrast, casting 4 to type @code{union foo}
1943 would store it into the union as the integer @code{i}, since it is
1944 an integer. (@xref{Cast to Union}.)
1946 You can combine this technique of naming elements with ordinary C
1947 initialization of successive elements. Each initializer element that
1948 does not have a designator applies to the next consecutive element of the
1949 array or structure. For example,
1952 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1959 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1962 Labeling the elements of an array initializer is especially useful
1963 when the indices are characters or belong to an @code{enum} type.
1968 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1969 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1972 @cindex designator lists
1973 You can also write a series of @samp{.@var{fieldname}} and
1974 @samp{[@var{index}]} designators before an @samp{=} to specify a
1975 nested subobject to initialize; the list is taken relative to the
1976 subobject corresponding to the closest surrounding brace pair. For
1977 example, with the @samp{struct point} declaration above:
1980 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1984 If the same field is initialized multiple times, it will have value from
1985 the last initialization. If any such overridden initialization has
1986 side-effect, it is unspecified whether the side-effect happens or not.
1987 Currently, GCC will discard them and issue a warning.
1990 @section Case Ranges
1992 @cindex ranges in case statements
1994 You can specify a range of consecutive values in a single @code{case} label,
1998 case @var{low} ... @var{high}:
2002 This has the same effect as the proper number of individual @code{case}
2003 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2005 This feature is especially useful for ranges of ASCII character codes:
2011 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2012 it may be parsed wrong when you use it with integer values. For example,
2027 @section Cast to a Union Type
2028 @cindex cast to a union
2029 @cindex union, casting to a
2031 A cast to union type is similar to other casts, except that the type
2032 specified is a union type. You can specify the type either with
2033 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2034 a constructor though, not a cast, and hence does not yield an lvalue like
2035 normal casts. (@xref{Compound Literals}.)
2037 The types that may be cast to the union type are those of the members
2038 of the union. Thus, given the following union and variables:
2041 union foo @{ int i; double d; @};
2047 both @code{x} and @code{y} can be cast to type @code{union foo}.
2049 Using the cast as the right-hand side of an assignment to a variable of
2050 union type is equivalent to storing in a member of the union:
2055 u = (union foo) x @equiv{} u.i = x
2056 u = (union foo) y @equiv{} u.d = y
2059 You can also use the union cast as a function argument:
2062 void hack (union foo);
2064 hack ((union foo) x);
2067 @node Mixed Declarations
2068 @section Mixed Declarations and Code
2069 @cindex mixed declarations and code
2070 @cindex declarations, mixed with code
2071 @cindex code, mixed with declarations
2073 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2074 within compound statements. As an extension, GCC also allows this in
2075 C90 mode. For example, you could do:
2084 Each identifier is visible from where it is declared until the end of
2085 the enclosing block.
2087 @node Function Attributes
2088 @section Declaring Attributes of Functions
2089 @cindex function attributes
2090 @cindex declaring attributes of functions
2091 @cindex functions that never return
2092 @cindex functions that return more than once
2093 @cindex functions that have no side effects
2094 @cindex functions in arbitrary sections
2095 @cindex functions that behave like malloc
2096 @cindex @code{volatile} applied to function
2097 @cindex @code{const} applied to function
2098 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2099 @cindex functions with non-null pointer arguments
2100 @cindex functions that are passed arguments in registers on the 386
2101 @cindex functions that pop the argument stack on the 386
2102 @cindex functions that do not pop the argument stack on the 386
2103 @cindex functions that have different compilation options on the 386
2104 @cindex functions that have different optimization options
2105 @cindex functions that are dynamically resolved
2107 In GNU C, you declare certain things about functions called in your program
2108 which help the compiler optimize function calls and check your code more
2111 The keyword @code{__attribute__} allows you to specify special
2112 attributes when making a declaration. This keyword is followed by an
2113 attribute specification inside double parentheses. The following
2114 attributes are currently defined for functions on all targets:
2115 @code{aligned}, @code{alloc_size}, @code{noreturn},
2116 @code{returns_twice}, @code{noinline}, @code{noclone},
2117 @code{always_inline}, @code{flatten}, @code{pure}, @code{const},
2118 @code{nothrow}, @code{sentinel}, @code{format}, @code{format_arg},
2119 @code{no_instrument_function}, @code{no_split_stack},
2120 @code{section}, @code{constructor},
2121 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
2122 @code{weak}, @code{malloc}, @code{alias}, @code{ifunc},
2123 @code{warn_unused_result}, @code{nonnull}, @code{gnu_inline},
2124 @code{externally_visible}, @code{hot}, @code{cold}, @code{artificial},
2125 @code{error} and @code{warning}. Several other attributes are defined
2126 for functions on particular target systems. Other attributes,
2127 including @code{section} are supported for variables declarations
2128 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
2130 GCC plugins may provide their own attributes.
2132 You may also specify attributes with @samp{__} preceding and following
2133 each keyword. This allows you to use them in header files without
2134 being concerned about a possible macro of the same name. For example,
2135 you may use @code{__noreturn__} instead of @code{noreturn}.
2137 @xref{Attribute Syntax}, for details of the exact syntax for using
2141 @c Keep this table alphabetized by attribute name. Treat _ as space.
2143 @item alias ("@var{target}")
2144 @cindex @code{alias} attribute
2145 The @code{alias} attribute causes the declaration to be emitted as an
2146 alias for another symbol, which must be specified. For instance,
2149 void __f () @{ /* @r{Do something.} */; @}
2150 void f () __attribute__ ((weak, alias ("__f")));
2153 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2154 mangled name for the target must be used. It is an error if @samp{__f}
2155 is not defined in the same translation unit.
2157 Not all target machines support this attribute.
2159 @item aligned (@var{alignment})
2160 @cindex @code{aligned} attribute
2161 This attribute specifies a minimum alignment for the function,
2164 You cannot use this attribute to decrease the alignment of a function,
2165 only to increase it. However, when you explicitly specify a function
2166 alignment this will override the effect of the
2167 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2170 Note that the effectiveness of @code{aligned} attributes may be
2171 limited by inherent limitations in your linker. On many systems, the
2172 linker is only able to arrange for functions to be aligned up to a
2173 certain maximum alignment. (For some linkers, the maximum supported
2174 alignment may be very very small.) See your linker documentation for
2175 further information.
2177 The @code{aligned} attribute can also be used for variables and fields
2178 (@pxref{Variable Attributes}.)
2181 @cindex @code{alloc_size} attribute
2182 The @code{alloc_size} attribute is used to tell the compiler that the
2183 function return value points to memory, where the size is given by
2184 one or two of the functions parameters. GCC uses this
2185 information to improve the correctness of @code{__builtin_object_size}.
2187 The function parameter(s) denoting the allocated size are specified by
2188 one or two integer arguments supplied to the attribute. The allocated size
2189 is either the value of the single function argument specified or the product
2190 of the two function arguments specified. Argument numbering starts at
2196 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2197 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2200 declares that my_calloc will return memory of the size given by
2201 the product of parameter 1 and 2 and that my_realloc will return memory
2202 of the size given by parameter 2.
2205 @cindex @code{always_inline} function attribute
2206 Generally, functions are not inlined unless optimization is specified.
2207 For functions declared inline, this attribute inlines the function even
2208 if no optimization level was specified.
2211 @cindex @code{gnu_inline} function attribute
2212 This attribute should be used with a function which is also declared
2213 with the @code{inline} keyword. It directs GCC to treat the function
2214 as if it were defined in gnu90 mode even when compiling in C99 or
2217 If the function is declared @code{extern}, then this definition of the
2218 function is used only for inlining. In no case is the function
2219 compiled as a standalone function, not even if you take its address
2220 explicitly. Such an address becomes an external reference, as if you
2221 had only declared the function, and had not defined it. This has
2222 almost the effect of a macro. The way to use this is to put a
2223 function definition in a header file with this attribute, and put
2224 another copy of the function, without @code{extern}, in a library
2225 file. The definition in the header file will cause most calls to the
2226 function to be inlined. If any uses of the function remain, they will
2227 refer to the single copy in the library. Note that the two
2228 definitions of the functions need not be precisely the same, although
2229 if they do not have the same effect your program may behave oddly.
2231 In C, if the function is neither @code{extern} nor @code{static}, then
2232 the function is compiled as a standalone function, as well as being
2233 inlined where possible.
2235 This is how GCC traditionally handled functions declared
2236 @code{inline}. Since ISO C99 specifies a different semantics for
2237 @code{inline}, this function attribute is provided as a transition
2238 measure and as a useful feature in its own right. This attribute is
2239 available in GCC 4.1.3 and later. It is available if either of the
2240 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2241 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2242 Function is As Fast As a Macro}.
2244 In C++, this attribute does not depend on @code{extern} in any way,
2245 but it still requires the @code{inline} keyword to enable its special
2249 @cindex @code{artificial} function attribute
2250 This attribute is useful for small inline wrappers which if possible
2251 should appear during debugging as a unit, depending on the debug
2252 info format it will either mean marking the function as artificial
2253 or using the caller location for all instructions within the inlined
2257 @cindex interrupt handler functions
2258 When added to an interrupt handler with the M32C port, causes the
2259 prologue and epilogue to use bank switching to preserve the registers
2260 rather than saving them on the stack.
2263 @cindex @code{flatten} function attribute
2264 Generally, inlining into a function is limited. For a function marked with
2265 this attribute, every call inside this function will be inlined, if possible.
2266 Whether the function itself is considered for inlining depends on its size and
2267 the current inlining parameters.
2269 @item error ("@var{message}")
2270 @cindex @code{error} function attribute
2271 If this attribute is used on a function declaration and a call to such a function
2272 is not eliminated through dead code elimination or other optimizations, an error
2273 which will include @var{message} will be diagnosed. This is useful
2274 for compile time checking, especially together with @code{__builtin_constant_p}
2275 and inline functions where checking the inline function arguments is not
2276 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2277 While it is possible to leave the function undefined and thus invoke
2278 a link failure, when using this attribute the problem will be diagnosed
2279 earlier and with exact location of the call even in presence of inline
2280 functions or when not emitting debugging information.
2282 @item warning ("@var{message}")
2283 @cindex @code{warning} function attribute
2284 If this attribute is used on a function declaration and a call to such a function
2285 is not eliminated through dead code elimination or other optimizations, a warning
2286 which will include @var{message} will be diagnosed. This is useful
2287 for compile time checking, especially together with @code{__builtin_constant_p}
2288 and inline functions. While it is possible to define the function with
2289 a message in @code{.gnu.warning*} section, when using this attribute the problem
2290 will be diagnosed earlier and with exact location of the call even in presence
2291 of inline functions or when not emitting debugging information.
2294 @cindex functions that do pop the argument stack on the 386
2296 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2297 assume that the calling function will pop off the stack space used to
2298 pass arguments. This is
2299 useful to override the effects of the @option{-mrtd} switch.
2302 @cindex @code{const} function attribute
2303 Many functions do not examine any values except their arguments, and
2304 have no effects except the return value. Basically this is just slightly
2305 more strict class than the @code{pure} attribute below, since function is not
2306 allowed to read global memory.
2308 @cindex pointer arguments
2309 Note that a function that has pointer arguments and examines the data
2310 pointed to must @emph{not} be declared @code{const}. Likewise, a
2311 function that calls a non-@code{const} function usually must not be
2312 @code{const}. It does not make sense for a @code{const} function to
2315 The attribute @code{const} is not implemented in GCC versions earlier
2316 than 2.5. An alternative way to declare that a function has no side
2317 effects, which works in the current version and in some older versions,
2321 typedef int intfn ();
2323 extern const intfn square;
2326 This approach does not work in GNU C++ from 2.6.0 on, since the language
2327 specifies that the @samp{const} must be attached to the return value.
2331 @itemx constructor (@var{priority})
2332 @itemx destructor (@var{priority})
2333 @cindex @code{constructor} function attribute
2334 @cindex @code{destructor} function attribute
2335 The @code{constructor} attribute causes the function to be called
2336 automatically before execution enters @code{main ()}. Similarly, the
2337 @code{destructor} attribute causes the function to be called
2338 automatically after @code{main ()} has completed or @code{exit ()} has
2339 been called. Functions with these attributes are useful for
2340 initializing data that will be used implicitly during the execution of
2343 You may provide an optional integer priority to control the order in
2344 which constructor and destructor functions are run. A constructor
2345 with a smaller priority number runs before a constructor with a larger
2346 priority number; the opposite relationship holds for destructors. So,
2347 if you have a constructor that allocates a resource and a destructor
2348 that deallocates the same resource, both functions typically have the
2349 same priority. The priorities for constructor and destructor
2350 functions are the same as those specified for namespace-scope C++
2351 objects (@pxref{C++ Attributes}).
2353 These attributes are not currently implemented for Objective-C@.
2356 @itemx deprecated (@var{msg})
2357 @cindex @code{deprecated} attribute.
2358 The @code{deprecated} attribute results in a warning if the function
2359 is used anywhere in the source file. This is useful when identifying
2360 functions that are expected to be removed in a future version of a
2361 program. The warning also includes the location of the declaration
2362 of the deprecated function, to enable users to easily find further
2363 information about why the function is deprecated, or what they should
2364 do instead. Note that the warnings only occurs for uses:
2367 int old_fn () __attribute__ ((deprecated));
2369 int (*fn_ptr)() = old_fn;
2372 results in a warning on line 3 but not line 2. The optional msg
2373 argument, which must be a string, will be printed in the warning if
2376 The @code{deprecated} attribute can also be used for variables and
2377 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2380 @cindex @code{disinterrupt} attribute
2381 On Epiphany and MeP targets, this attribute causes the compiler to emit
2382 instructions to disable interrupts for the duration of the given
2386 @cindex @code{__declspec(dllexport)}
2387 On Microsoft Windows targets and Symbian OS targets the
2388 @code{dllexport} attribute causes the compiler to provide a global
2389 pointer to a pointer in a DLL, so that it can be referenced with the
2390 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2391 name is formed by combining @code{_imp__} and the function or variable
2394 You can use @code{__declspec(dllexport)} as a synonym for
2395 @code{__attribute__ ((dllexport))} for compatibility with other
2398 On systems that support the @code{visibility} attribute, this
2399 attribute also implies ``default'' visibility. It is an error to
2400 explicitly specify any other visibility.
2402 In previous versions of GCC, the @code{dllexport} attribute was ignored
2403 for inlined functions, unless the @option{-fkeep-inline-functions} flag
2404 had been used. The default behaviour now is to emit all dllexported
2405 inline functions; however, this can cause object file-size bloat, in
2406 which case the old behaviour can be restored by using
2407 @option{-fno-keep-inline-dllexport}.
2409 The attribute is also ignored for undefined symbols.
2411 When applied to C++ classes, the attribute marks defined non-inlined
2412 member functions and static data members as exports. Static consts
2413 initialized in-class are not marked unless they are also defined
2416 For Microsoft Windows targets there are alternative methods for
2417 including the symbol in the DLL's export table such as using a
2418 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2419 the @option{--export-all} linker flag.
2422 @cindex @code{__declspec(dllimport)}
2423 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2424 attribute causes the compiler to reference a function or variable via
2425 a global pointer to a pointer that is set up by the DLL exporting the
2426 symbol. The attribute implies @code{extern}. On Microsoft Windows
2427 targets, the pointer name is formed by combining @code{_imp__} and the
2428 function or variable name.
2430 You can use @code{__declspec(dllimport)} as a synonym for
2431 @code{__attribute__ ((dllimport))} for compatibility with other
2434 On systems that support the @code{visibility} attribute, this
2435 attribute also implies ``default'' visibility. It is an error to
2436 explicitly specify any other visibility.
2438 Currently, the attribute is ignored for inlined functions. If the
2439 attribute is applied to a symbol @emph{definition}, an error is reported.
2440 If a symbol previously declared @code{dllimport} is later defined, the
2441 attribute is ignored in subsequent references, and a warning is emitted.
2442 The attribute is also overridden by a subsequent declaration as
2445 When applied to C++ classes, the attribute marks non-inlined
2446 member functions and static data members as imports. However, the
2447 attribute is ignored for virtual methods to allow creation of vtables
2450 On the SH Symbian OS target the @code{dllimport} attribute also has
2451 another affect---it can cause the vtable and run-time type information
2452 for a class to be exported. This happens when the class has a
2453 dllimport'ed constructor or a non-inline, non-pure virtual function
2454 and, for either of those two conditions, the class also has an inline
2455 constructor or destructor and has a key function that is defined in
2456 the current translation unit.
2458 For Microsoft Windows based targets the use of the @code{dllimport}
2459 attribute on functions is not necessary, but provides a small
2460 performance benefit by eliminating a thunk in the DLL@. The use of the
2461 @code{dllimport} attribute on imported variables was required on older
2462 versions of the GNU linker, but can now be avoided by passing the
2463 @option{--enable-auto-import} switch to the GNU linker. As with
2464 functions, using the attribute for a variable eliminates a thunk in
2467 One drawback to using this attribute is that a pointer to a
2468 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2469 address. However, a pointer to a @emph{function} with the
2470 @code{dllimport} attribute can be used as a constant initializer; in
2471 this case, the address of a stub function in the import lib is
2472 referenced. On Microsoft Windows targets, the attribute can be disabled
2473 for functions by setting the @option{-mnop-fun-dllimport} flag.
2476 @cindex eight bit data on the H8/300, H8/300H, and H8S
2477 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2478 variable should be placed into the eight bit data section.
2479 The compiler will generate more efficient code for certain operations
2480 on data in the eight bit data area. Note the eight bit data area is limited to
2483 You must use GAS and GLD from GNU binutils version 2.7 or later for
2484 this attribute to work correctly.
2486 @item exception_handler
2487 @cindex exception handler functions on the Blackfin processor
2488 Use this attribute on the Blackfin to indicate that the specified function
2489 is an exception handler. The compiler will generate function entry and
2490 exit sequences suitable for use in an exception handler when this
2491 attribute is present.
2493 @item externally_visible
2494 @cindex @code{externally_visible} attribute.
2495 This attribute, attached to a global variable or function, nullifies
2496 the effect of the @option{-fwhole-program} command-line option, so the
2497 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.
2500 @cindex functions which handle memory bank switching
2501 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2502 use a calling convention that takes care of switching memory banks when
2503 entering and leaving a function. This calling convention is also the
2504 default when using the @option{-mlong-calls} option.
2506 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2507 to call and return from a function.
2509 On 68HC11 the compiler will generate a sequence of instructions
2510 to invoke a board-specific routine to switch the memory bank and call the
2511 real function. The board-specific routine simulates a @code{call}.
2512 At the end of a function, it will jump to a board-specific routine
2513 instead of using @code{rts}. The board-specific return routine simulates
2516 On MeP targets this causes the compiler to use a calling convention
2517 which assumes the called function is too far away for the built-in
2520 @item fast_interrupt
2521 @cindex interrupt handler functions
2522 Use this attribute on the M32C and RX ports to indicate that the specified
2523 function is a fast interrupt handler. This is just like the
2524 @code{interrupt} attribute, except that @code{freit} is used to return
2525 instead of @code{reit}.
2528 @cindex functions that pop the argument stack on the 386
2529 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2530 pass the first argument (if of integral type) in the register ECX and
2531 the second argument (if of integral type) in the register EDX@. Subsequent
2532 and other typed arguments are passed on the stack. The called function will
2533 pop the arguments off the stack. If the number of arguments is variable all
2534 arguments are pushed on the stack.
2537 @cindex functions that pop the argument stack on the 386
2538 On the Intel 386, the @code{thiscall} attribute causes the compiler to
2539 pass the first argument (if of integral type) in the register ECX.
2540 Subsequent and other typed arguments are passed on the stack. The called
2541 function will pop the arguments off the stack.
2542 If the number of arguments is variable all arguments are pushed on the
2544 The @code{thiscall} attribute is intended for C++ non-static member functions.
2545 As gcc extension this calling convention can be used for C-functions
2546 and for static member methods.
2548 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2549 @cindex @code{format} function attribute
2551 The @code{format} attribute specifies that a function takes @code{printf},
2552 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2553 should be type-checked against a format string. For example, the
2558 my_printf (void *my_object, const char *my_format, ...)
2559 __attribute__ ((format (printf, 2, 3)));
2563 causes the compiler to check the arguments in calls to @code{my_printf}
2564 for consistency with the @code{printf} style format string argument
2567 The parameter @var{archetype} determines how the format string is
2568 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2569 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2570 @code{strfmon}. (You can also use @code{__printf__},
2571 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2572 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2573 @code{ms_strftime} are also present.
2574 @var{archtype} values such as @code{printf} refer to the formats accepted
2575 by the system's C run-time library, while @code{gnu_} values always refer
2576 to the formats accepted by the GNU C Library. On Microsoft Windows
2577 targets, @code{ms_} values refer to the formats accepted by the
2578 @file{msvcrt.dll} library.
2579 The parameter @var{string-index}
2580 specifies which argument is the format string argument (starting
2581 from 1), while @var{first-to-check} is the number of the first
2582 argument to check against the format string. For functions
2583 where the arguments are not available to be checked (such as
2584 @code{vprintf}), specify the third parameter as zero. In this case the
2585 compiler only checks the format string for consistency. For
2586 @code{strftime} formats, the third parameter is required to be zero.
2587 Since non-static C++ methods have an implicit @code{this} argument, the
2588 arguments of such methods should be counted from two, not one, when
2589 giving values for @var{string-index} and @var{first-to-check}.
2591 In the example above, the format string (@code{my_format}) is the second
2592 argument of the function @code{my_print}, and the arguments to check
2593 start with the third argument, so the correct parameters for the format
2594 attribute are 2 and 3.
2596 @opindex ffreestanding
2597 @opindex fno-builtin
2598 The @code{format} attribute allows you to identify your own functions
2599 which take format strings as arguments, so that GCC can check the
2600 calls to these functions for errors. The compiler always (unless
2601 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2602 for the standard library functions @code{printf}, @code{fprintf},
2603 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2604 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2605 warnings are requested (using @option{-Wformat}), so there is no need to
2606 modify the header file @file{stdio.h}. In C99 mode, the functions
2607 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2608 @code{vsscanf} are also checked. Except in strictly conforming C
2609 standard modes, the X/Open function @code{strfmon} is also checked as
2610 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2611 @xref{C Dialect Options,,Options Controlling C Dialect}.
2613 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2614 recognized in the same context. Declarations including these format attributes
2615 will be parsed for correct syntax, however the result of checking of such format
2616 strings is not yet defined, and will not be carried out by this version of the
2619 The target may also provide additional types of format checks.
2620 @xref{Target Format Checks,,Format Checks Specific to Particular
2623 @item format_arg (@var{string-index})
2624 @cindex @code{format_arg} function attribute
2625 @opindex Wformat-nonliteral
2626 The @code{format_arg} attribute specifies that a function takes a format
2627 string for a @code{printf}, @code{scanf}, @code{strftime} or
2628 @code{strfmon} style function and modifies it (for example, to translate
2629 it into another language), so the result can be passed to a
2630 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2631 function (with the remaining arguments to the format function the same
2632 as they would have been for the unmodified string). For example, the
2637 my_dgettext (char *my_domain, const char *my_format)
2638 __attribute__ ((format_arg (2)));
2642 causes the compiler to check the arguments in calls to a @code{printf},
2643 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2644 format string argument is a call to the @code{my_dgettext} function, for
2645 consistency with the format string argument @code{my_format}. If the
2646 @code{format_arg} attribute had not been specified, all the compiler
2647 could tell in such calls to format functions would be that the format
2648 string argument is not constant; this would generate a warning when
2649 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2650 without the attribute.
2652 The parameter @var{string-index} specifies which argument is the format
2653 string argument (starting from one). Since non-static C++ methods have
2654 an implicit @code{this} argument, the arguments of such methods should
2655 be counted from two.
2657 The @code{format-arg} attribute allows you to identify your own
2658 functions which modify format strings, so that GCC can check the
2659 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2660 type function whose operands are a call to one of your own function.
2661 The compiler always treats @code{gettext}, @code{dgettext}, and
2662 @code{dcgettext} in this manner except when strict ISO C support is
2663 requested by @option{-ansi} or an appropriate @option{-std} option, or
2664 @option{-ffreestanding} or @option{-fno-builtin}
2665 is used. @xref{C Dialect Options,,Options
2666 Controlling C Dialect}.
2668 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2669 @code{NSString} reference for compatibility with the @code{format} attribute
2672 The target may also allow additional types in @code{format-arg} attributes.
2673 @xref{Target Format Checks,,Format Checks Specific to Particular
2676 @item function_vector
2677 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2678 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2679 function should be called through the function vector. Calling a
2680 function through the function vector will reduce code size, however;
2681 the function vector has a limited size (maximum 128 entries on the H8/300
2682 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2684 In SH2A target, this attribute declares a function to be called using the
2685 TBR relative addressing mode. The argument to this attribute is the entry
2686 number of the same function in a vector table containing all the TBR
2687 relative addressable functions. For the successful jump, register TBR
2688 should contain the start address of this TBR relative vector table.
2689 In the startup routine of the user application, user needs to care of this
2690 TBR register initialization. The TBR relative vector table can have at
2691 max 256 function entries. The jumps to these functions will be generated
2692 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2693 You must use GAS and GLD from GNU binutils version 2.7 or later for
2694 this attribute to work correctly.
2696 Please refer the example of M16C target, to see the use of this
2697 attribute while declaring a function,
2699 In an application, for a function being called once, this attribute will
2700 save at least 8 bytes of code; and if other successive calls are being
2701 made to the same function, it will save 2 bytes of code per each of these
2704 On M16C/M32C targets, the @code{function_vector} attribute declares a
2705 special page subroutine call function. Use of this attribute reduces
2706 the code size by 2 bytes for each call generated to the
2707 subroutine. The argument to the attribute is the vector number entry
2708 from the special page vector table which contains the 16 low-order
2709 bits of the subroutine's entry address. Each vector table has special
2710 page number (18 to 255) which are used in @code{jsrs} instruction.
2711 Jump addresses of the routines are generated by adding 0x0F0000 (in
2712 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2713 byte addresses set in the vector table. Therefore you need to ensure
2714 that all the special page vector routines should get mapped within the
2715 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2718 In the following example 2 bytes will be saved for each call to
2719 function @code{foo}.
2722 void foo (void) __attribute__((function_vector(0x18)));
2733 If functions are defined in one file and are called in another file,
2734 then be sure to write this declaration in both files.
2736 This attribute is ignored for R8C target.
2738 @item ifunc ("@var{resolver}")
2739 @cindex @code{ifunc} attribute
2740 The @code{ifunc} attribute is used to mark a function as an indirect
2741 function using the STT_GNU_IFUNC symbol type extension to the ELF
2742 standard. This allows the resolution of the symbol value to be
2743 determined dynamically at load time, and an optimized version of the
2744 routine can be selected for the particular processor or other system
2745 characteristics determined then. To use this attribute, first define
2746 the implementation functions available, and a resolver function that
2747 returns a pointer to the selected implementation function. The
2748 implementation functions' declarations must match the API of the
2749 function being implemented, the resolver's declaration is be a
2750 function returning pointer to void function returning void:
2753 void *my_memcpy (void *dst, const void *src, size_t len)
2758 static void (*resolve_memcpy (void)) (void)
2760 return my_memcpy; // we'll just always select this routine
2764 The exported header file declaring the function the user calls would
2768 extern void *memcpy (void *, const void *, size_t);
2771 allowing the user to call this as a regular function, unaware of the
2772 implementation. Finally, the indirect function needs to be defined in
2773 the same translation unit as the resolver function:
2776 void *memcpy (void *, const void *, size_t)
2777 __attribute__ ((ifunc ("resolve_memcpy")));
2780 Indirect functions cannot be weak, and require a recent binutils (at
2781 least version 2.20.1), and GNU C library (at least version 2.11.1).
2784 @cindex interrupt handler functions
2785 Use this attribute on the ARM, AVR, CR16, Epiphany, M32C, M32R/D, m68k, MeP, MIPS,
2786 RL78, RX and Xstormy16 ports to indicate that the specified function is an
2787 interrupt handler. The compiler will generate function entry and exit
2788 sequences suitable for use in an interrupt handler when this attribute
2789 is present. With Epiphany targets it may also generate a special section with
2790 code to initialize the interrupt vector table.
2792 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, MicroBlaze,
2793 and SH processors can be specified via the @code{interrupt_handler} attribute.
2795 Note, on the AVR, the hardware globally disables interrupts when an
2796 interrupt is executed. The first instruction of an interrupt handler
2797 declared with this attribute will be a @code{SEI} instruction to
2798 re-enable interrupts. See also the @code{signal} function attribute
2799 that does not insert a @code{SEI} instuction. If both @code{signal} and
2800 @code{interrupt} are specified for the same function, @code{signal}
2801 will be silently ignored.
2803 Note, for the ARM, you can specify the kind of interrupt to be handled by
2804 adding an optional parameter to the interrupt attribute like this:
2807 void f () __attribute__ ((interrupt ("IRQ")));
2810 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2812 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2813 may be called with a word aligned stack pointer.
2815 On Epiphany targets one or more optional parameters can be added like this:
2818 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
2821 Permissible values for these parameters are: @w{@code{reset}},
2822 @w{@code{software_exception}}, @w{@code{page_miss}},
2823 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
2824 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
2825 Multiple parameters indicate that multiple entries in the interrupt
2826 vector table should be initialized for this function, i.e. for each
2827 parameter @w{@var{name}}, a jump to the function will be emitted in
2828 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
2829 entirely, in which case no interrupt vector table entry will be provided.
2831 Note, on Epiphany targets, interrupts are enabled inside the function
2832 unless the @code{disinterrupt} attribute is also specified.
2834 On Epiphany targets, you can also use the following attribute to
2835 modify the behavior of an interrupt handler:
2837 @item forwarder_section
2838 @cindex @code{forwarder_section} attribute
2839 The interrupt handler may be in external memory which cannot be
2840 reached by a branch instruction, so generate a local memory trampoline
2841 to transfer control. The single parameter identifies the section where
2842 the trampoline will be placed.
2845 The following examples are all valid uses of these attributes on
2848 void __attribute__ ((interrupt)) universal_handler ();
2849 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
2850 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
2851 void __attribute__ ((interrupt ("timer0"), disinterrupt))
2852 fast_timer_handler ();
2853 void __attribute__ ((interrupt ("dma0, dma1"), forwarder_section ("tramp")))
2854 external_dma_handler ();
2857 On MIPS targets, you can use the following attributes to modify the behavior
2858 of an interrupt handler:
2860 @item use_shadow_register_set
2861 @cindex @code{use_shadow_register_set} attribute
2862 Assume that the handler uses a shadow register set, instead of
2863 the main general-purpose registers.
2865 @item keep_interrupts_masked
2866 @cindex @code{keep_interrupts_masked} attribute
2867 Keep interrupts masked for the whole function. Without this attribute,
2868 GCC tries to reenable interrupts for as much of the function as it can.
2870 @item use_debug_exception_return
2871 @cindex @code{use_debug_exception_return} attribute
2872 Return using the @code{deret} instruction. Interrupt handlers that don't
2873 have this attribute return using @code{eret} instead.
2876 You can use any combination of these attributes, as shown below:
2878 void __attribute__ ((interrupt)) v0 ();
2879 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
2880 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
2881 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
2882 void __attribute__ ((interrupt, use_shadow_register_set,
2883 keep_interrupts_masked)) v4 ();
2884 void __attribute__ ((interrupt, use_shadow_register_set,
2885 use_debug_exception_return)) v5 ();
2886 void __attribute__ ((interrupt, keep_interrupts_masked,
2887 use_debug_exception_return)) v6 ();
2888 void __attribute__ ((interrupt, use_shadow_register_set,
2889 keep_interrupts_masked,
2890 use_debug_exception_return)) v7 ();
2893 On RL78, use @code{brk_interrupt} instead of @code{interrupt} for
2894 handlers intended to be used with the @code{BRK} opcode (i.e. those
2895 that must end with @code{RETB} instead of @code{RETI}).
2897 @item interrupt_handler
2898 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2899 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2900 indicate that the specified function is an interrupt handler. The compiler
2901 will generate function entry and exit sequences suitable for use in an
2902 interrupt handler when this attribute is present.
2904 @item interrupt_thread
2905 @cindex interrupt thread functions on fido
2906 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2907 that the specified function is an interrupt handler that is designed
2908 to run as a thread. The compiler omits generate prologue/epilogue
2909 sequences and replaces the return instruction with a @code{sleep}
2910 instruction. This attribute is available only on fido.
2913 @cindex interrupt service routines on ARM
2914 Use this attribute on ARM to write Interrupt Service Routines. This is an
2915 alias to the @code{interrupt} attribute above.
2918 @cindex User stack pointer in interrupts on the Blackfin
2919 When used together with @code{interrupt_handler}, @code{exception_handler}
2920 or @code{nmi_handler}, code will be generated to load the stack pointer
2921 from the USP register in the function prologue.
2924 @cindex @code{l1_text} function attribute
2925 This attribute specifies a function to be placed into L1 Instruction
2926 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2927 With @option{-mfdpic}, function calls with a such function as the callee
2928 or caller will use inlined PLT.
2931 @cindex @code{l2} function attribute
2932 On the Blackfin, this attribute specifies a function to be placed into L2
2933 SRAM. The function will be put into a specific section named
2934 @code{.l1.text}. With @option{-mfdpic}, callers of such functions will use
2938 @cindex @code{leaf} function attribute
2939 Calls to external functions with this attribute must return to the current
2940 compilation unit only by return or by exception handling. In particular, leaf
2941 functions are not allowed to call callback function passed to it from the current
2942 compilation unit or directly call functions exported by the unit or longjmp
2943 into the unit. Leaf function might still call functions from other compilation
2944 units and thus they are not necessarily leaf in the sense that they contain no
2945 function calls at all.
2947 The attribute is intended for library functions to improve dataflow analysis.
2948 The compiler takes the hint that any data not escaping the current compilation unit can
2949 not be used or modified by the leaf function. For example, the @code{sin} function
2950 is a leaf function, but @code{qsort} is not.
2952 Note that leaf functions might invoke signals and signal handlers might be
2953 defined in the current compilation unit and use static variables. The only
2954 compliant way to write such a signal handler is to declare such variables
2957 The attribute has no effect on functions defined within the current compilation
2958 unit. This is to allow easy merging of multiple compilation units into one,
2959 for example, by using the link time optimization. For this reason the
2960 attribute is not allowed on types to annotate indirect calls.
2962 @item long_call/short_call
2963 @cindex indirect calls on ARM
2964 This attribute specifies how a particular function is called on
2965 ARM and Epiphany. Both attributes override the
2966 @option{-mlong-calls} (@pxref{ARM Options})
2967 command-line switch and @code{#pragma long_calls} settings. The
2968 @code{long_call} attribute indicates that the function might be far
2969 away from the call site and require a different (more expensive)
2970 calling sequence. The @code{short_call} attribute always places
2971 the offset to the function from the call site into the @samp{BL}
2972 instruction directly.
2974 @item longcall/shortcall
2975 @cindex functions called via pointer on the RS/6000 and PowerPC
2976 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2977 indicates that the function might be far away from the call site and
2978 require a different (more expensive) calling sequence. The
2979 @code{shortcall} attribute indicates that the function is always close
2980 enough for the shorter calling sequence to be used. These attributes
2981 override both the @option{-mlongcall} switch and, on the RS/6000 and
2982 PowerPC, the @code{#pragma longcall} setting.
2984 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2985 calls are necessary.
2987 @item long_call/near/far
2988 @cindex indirect calls on MIPS
2989 These attributes specify how a particular function is called on MIPS@.
2990 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2991 command-line switch. The @code{long_call} and @code{far} attributes are
2992 synonyms, and cause the compiler to always call
2993 the function by first loading its address into a register, and then using
2994 the contents of that register. The @code{near} attribute has the opposite
2995 effect; it specifies that non-PIC calls should be made using the more
2996 efficient @code{jal} instruction.
2999 @cindex @code{malloc} attribute
3000 The @code{malloc} attribute is used to tell the compiler that a function
3001 may be treated as if any non-@code{NULL} pointer it returns cannot
3002 alias any other pointer valid when the function returns and that the memory
3003 has undefined content.
3004 This will often improve optimization.
3005 Standard functions with this property include @code{malloc} and
3006 @code{calloc}. @code{realloc}-like functions do not have this
3007 property as the memory pointed to does not have undefined content.
3009 @item mips16/nomips16
3010 @cindex @code{mips16} attribute
3011 @cindex @code{nomips16} attribute
3013 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
3014 function attributes to locally select or turn off MIPS16 code generation.
3015 A function with the @code{mips16} attribute is emitted as MIPS16 code,
3016 while MIPS16 code generation is disabled for functions with the
3017 @code{nomips16} attribute. These attributes override the
3018 @option{-mips16} and @option{-mno-mips16} options on the command line
3019 (@pxref{MIPS Options}).
3021 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
3022 preprocessor symbol @code{__mips16} reflects the setting on the command line,
3023 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
3024 may interact badly with some GCC extensions such as @code{__builtin_apply}
3025 (@pxref{Constructing Calls}).
3027 @item model (@var{model-name})
3028 @cindex function addressability on the M32R/D
3029 @cindex variable addressability on the IA-64
3031 On the M32R/D, use this attribute to set the addressability of an
3032 object, and of the code generated for a function. The identifier
3033 @var{model-name} is one of @code{small}, @code{medium}, or
3034 @code{large}, representing each of the code models.
3036 Small model objects live in the lower 16MB of memory (so that their
3037 addresses can be loaded with the @code{ld24} instruction), and are
3038 callable with the @code{bl} instruction.
3040 Medium model objects may live anywhere in the 32-bit address space (the
3041 compiler will generate @code{seth/add3} instructions to load their addresses),
3042 and are callable with the @code{bl} instruction.
3044 Large model objects may live anywhere in the 32-bit address space (the
3045 compiler will generate @code{seth/add3} instructions to load their addresses),
3046 and may not be reachable with the @code{bl} instruction (the compiler will
3047 generate the much slower @code{seth/add3/jl} instruction sequence).
3049 On IA-64, use this attribute to set the addressability of an object.
3050 At present, the only supported identifier for @var{model-name} is
3051 @code{small}, indicating addressability via ``small'' (22-bit)
3052 addresses (so that their addresses can be loaded with the @code{addl}
3053 instruction). Caveat: such addressing is by definition not position
3054 independent and hence this attribute must not be used for objects
3055 defined by shared libraries.
3057 @item ms_abi/sysv_abi
3058 @cindex @code{ms_abi} attribute
3059 @cindex @code{sysv_abi} attribute
3061 On 32-bit and 64-bit (i?86|x86_64)-*-* targets, you can use an ABI attribute
3062 to indicate which calling convention should be used for a function. The
3063 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
3064 while the @code{sysv_abi} attribute tells the compiler to use the ABI
3065 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
3066 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
3068 Note, the @code{ms_abi} attribute for Windows 64-bit targets currently
3069 requires the @option{-maccumulate-outgoing-args} option.
3071 @item callee_pop_aggregate_return (@var{number})
3072 @cindex @code{callee_pop_aggregate_return} attribute
3074 On 32-bit i?86-*-* targets, you can control by those attribute for
3075 aggregate return in memory, if the caller is responsible to pop the hidden
3076 pointer together with the rest of the arguments - @var{number} equal to
3077 zero -, or if the callee is responsible to pop hidden pointer - @var{number}
3078 equal to one. The default i386 ABI assumes that the callee pops the
3079 stack for hidden pointer.
3081 Note, that on 32-bit i386 Windows targets the compiler assumes that the
3082 caller pops the stack for hidden pointer.
3084 @item ms_hook_prologue
3085 @cindex @code{ms_hook_prologue} attribute
3087 On 32 bit i[34567]86-*-* targets and 64 bit x86_64-*-* targets, you can use
3088 this function attribute to make gcc generate the "hot-patching" function
3089 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
3093 @cindex function without a prologue/epilogue code
3094 Use this attribute on the ARM, AVR, MCORE, RX and SPU ports to indicate that
3095 the specified function does not need prologue/epilogue sequences generated by
3096 the compiler. It is up to the programmer to provide these sequences. The
3097 only statements that can be safely included in naked functions are
3098 @code{asm} statements that do not have operands. All other statements,
3099 including declarations of local variables, @code{if} statements, and so
3100 forth, should be avoided. Naked functions should be used to implement the
3101 body of an assembly function, while allowing the compiler to construct
3102 the requisite function declaration for the assembler.
3105 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
3106 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
3107 use the normal calling convention based on @code{jsr} and @code{rts}.
3108 This attribute can be used to cancel the effect of the @option{-mlong-calls}
3111 On MeP targets this attribute causes the compiler to assume the called
3112 function is close enough to use the normal calling convention,
3113 overriding the @code{-mtf} command line option.
3116 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
3117 Use this attribute together with @code{interrupt_handler},
3118 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3119 entry code should enable nested interrupts or exceptions.
3122 @cindex NMI handler functions on the Blackfin processor
3123 Use this attribute on the Blackfin to indicate that the specified function
3124 is an NMI handler. The compiler will generate function entry and
3125 exit sequences suitable for use in an NMI handler when this
3126 attribute is present.
3128 @item no_instrument_function
3129 @cindex @code{no_instrument_function} function attribute
3130 @opindex finstrument-functions
3131 If @option{-finstrument-functions} is given, profiling function calls will
3132 be generated at entry and exit of most user-compiled functions.
3133 Functions with this attribute will not be so instrumented.
3135 @item no_split_stack
3136 @cindex @code{no_split_stack} function attribute
3137 @opindex fsplit-stack
3138 If @option{-fsplit-stack} is given, functions will have a small
3139 prologue which decides whether to split the stack. Functions with the
3140 @code{no_split_stack} attribute will not have that prologue, and thus
3141 may run with only a small amount of stack space available.
3144 @cindex @code{noinline} function attribute
3145 This function attribute prevents a function from being considered for
3147 @c Don't enumerate the optimizations by name here; we try to be
3148 @c future-compatible with this mechanism.
3149 If the function does not have side-effects, there are optimizations
3150 other than inlining that causes function calls to be optimized away,
3151 although the function call is live. To keep such calls from being
3156 (@pxref{Extended Asm}) in the called function, to serve as a special
3160 @cindex @code{noclone} function attribute
3161 This function attribute prevents a function from being considered for
3162 cloning - a mechanism which produces specialized copies of functions
3163 and which is (currently) performed by interprocedural constant
3166 @item nonnull (@var{arg-index}, @dots{})
3167 @cindex @code{nonnull} function attribute
3168 The @code{nonnull} attribute specifies that some function parameters should
3169 be non-null pointers. For instance, the declaration:
3173 my_memcpy (void *dest, const void *src, size_t len)
3174 __attribute__((nonnull (1, 2)));
3178 causes the compiler to check that, in calls to @code{my_memcpy},
3179 arguments @var{dest} and @var{src} are non-null. If the compiler
3180 determines that a null pointer is passed in an argument slot marked
3181 as non-null, and the @option{-Wnonnull} option is enabled, a warning
3182 is issued. The compiler may also choose to make optimizations based
3183 on the knowledge that certain function arguments will not be null.
3185 If no argument index list is given to the @code{nonnull} attribute,
3186 all pointer arguments are marked as non-null. To illustrate, the
3187 following declaration is equivalent to the previous example:
3191 my_memcpy (void *dest, const void *src, size_t len)
3192 __attribute__((nonnull));
3196 @cindex @code{noreturn} function attribute
3197 A few standard library functions, such as @code{abort} and @code{exit},
3198 cannot return. GCC knows this automatically. Some programs define
3199 their own functions that never return. You can declare them
3200 @code{noreturn} to tell the compiler this fact. For example,
3204 void fatal () __attribute__ ((noreturn));
3207 fatal (/* @r{@dots{}} */)
3209 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3215 The @code{noreturn} keyword tells the compiler to assume that
3216 @code{fatal} cannot return. It can then optimize without regard to what
3217 would happen if @code{fatal} ever did return. This makes slightly
3218 better code. More importantly, it helps avoid spurious warnings of
3219 uninitialized variables.
3221 The @code{noreturn} keyword does not affect the exceptional path when that
3222 applies: a @code{noreturn}-marked function may still return to the caller
3223 by throwing an exception or calling @code{longjmp}.
3225 Do not assume that registers saved by the calling function are
3226 restored before calling the @code{noreturn} function.
3228 It does not make sense for a @code{noreturn} function to have a return
3229 type other than @code{void}.
3231 The attribute @code{noreturn} is not implemented in GCC versions
3232 earlier than 2.5. An alternative way to declare that a function does
3233 not return, which works in the current version and in some older
3234 versions, is as follows:
3237 typedef void voidfn ();
3239 volatile voidfn fatal;
3242 This approach does not work in GNU C++.
3245 @cindex @code{nothrow} function attribute
3246 The @code{nothrow} attribute is used to inform the compiler that a
3247 function cannot throw an exception. For example, most functions in
3248 the standard C library can be guaranteed not to throw an exception
3249 with the notable exceptions of @code{qsort} and @code{bsearch} that
3250 take function pointer arguments. The @code{nothrow} attribute is not
3251 implemented in GCC versions earlier than 3.3.
3254 @cindex @code{optimize} function attribute
3255 The @code{optimize} attribute is used to specify that a function is to
3256 be compiled with different optimization options than specified on the
3257 command line. Arguments can either be numbers or strings. Numbers
3258 are assumed to be an optimization level. Strings that begin with
3259 @code{O} are assumed to be an optimization option, while other options
3260 are assumed to be used with a @code{-f} prefix. You can also use the
3261 @samp{#pragma GCC optimize} pragma to set the optimization options
3262 that affect more than one function.
3263 @xref{Function Specific Option Pragmas}, for details about the
3264 @samp{#pragma GCC optimize} pragma.
3266 This can be used for instance to have frequently executed functions
3267 compiled with more aggressive optimization options that produce faster
3268 and larger code, while other functions can be called with less
3271 @item OS_main/OS_task
3272 @cindex @code{OS_main} AVR function attribute
3273 @cindex @code{OS_task} AVR function attribute
3274 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3275 do not save/restore any call-saved register in their prologue/epilogue.
3277 The @code{OS_main} attribute can be used when there @emph{is
3278 guarantee} that interrupts are disabled at the time when the function
3279 is entered. This will save resources when the stack pointer has to be
3280 changed to set up a frame for local variables.
3282 The @code{OS_task} attribute can be used when there is @emph{no
3283 guarantee} that interrupts are disabled at that time when the function
3284 is entered like for, e@.g@. task functions in a multi-threading operating
3285 system. In that case, changing the stack pointer register will be
3286 guarded by save/clear/restore of the global interrupt enable flag.
3288 The differences to the @code{naked} function attribute are:
3290 @item @code{naked} functions do not have a return instruction whereas
3291 @code{OS_main} and @code{OS_task} functions will have a @code{RET} or
3292 @code{RETI} return instruction.
3293 @item @code{naked} functions do not set up a frame for local variables
3294 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3299 @cindex @code{pcs} function attribute
3301 The @code{pcs} attribute can be used to control the calling convention
3302 used for a function on ARM. The attribute takes an argument that specifies
3303 the calling convention to use.
3305 When compiling using the AAPCS ABI (or a variant of that) then valid
3306 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3307 order to use a variant other than @code{"aapcs"} then the compiler must
3308 be permitted to use the appropriate co-processor registers (i.e., the
3309 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3313 /* Argument passed in r0, and result returned in r0+r1. */
3314 double f2d (float) __attribute__((pcs("aapcs")));
3317 Variadic functions always use the @code{"aapcs"} calling convention and
3318 the compiler will reject attempts to specify an alternative.
3321 @cindex @code{pure} function attribute
3322 Many functions have no effects except the return value and their
3323 return value depends only on the parameters and/or global variables.
3324 Such a function can be subject
3325 to common subexpression elimination and loop optimization just as an
3326 arithmetic operator would be. These functions should be declared
3327 with the attribute @code{pure}. For example,
3330 int square (int) __attribute__ ((pure));
3334 says that the hypothetical function @code{square} is safe to call
3335 fewer times than the program says.
3337 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3338 Interesting non-pure functions are functions with infinite loops or those
3339 depending on volatile memory or other system resource, that may change between
3340 two consecutive calls (such as @code{feof} in a multithreading environment).
3342 The attribute @code{pure} is not implemented in GCC versions earlier
3346 @cindex @code{hot} function attribute
3347 The @code{hot} attribute is used to inform the compiler that a function is a
3348 hot spot of the compiled program. The function is optimized more aggressively
3349 and on many target it is placed into special subsection of the text section so
3350 all hot functions appears close together improving locality.
3352 When profile feedback is available, via @option{-fprofile-use}, hot functions
3353 are automatically detected and this attribute is ignored.
3355 The @code{hot} attribute is not implemented in GCC versions earlier
3359 @cindex @code{cold} function attribute
3360 The @code{cold} attribute is used to inform the compiler that a function is
3361 unlikely executed. The function is optimized for size rather than speed and on
3362 many targets it is placed into special subsection of the text section so all
3363 cold functions appears close together improving code locality of non-cold parts
3364 of program. The paths leading to call of cold functions within code are marked
3365 as unlikely by the branch prediction mechanism. It is thus useful to mark
3366 functions used to handle unlikely conditions, such as @code{perror}, as cold to
3367 improve optimization of hot functions that do call marked functions in rare
3370 When profile feedback is available, via @option{-fprofile-use}, hot functions
3371 are automatically detected and this attribute is ignored.
3373 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
3375 @item regparm (@var{number})
3376 @cindex @code{regparm} attribute
3377 @cindex functions that are passed arguments in registers on the 386
3378 On the Intel 386, the @code{regparm} attribute causes the compiler to
3379 pass arguments number one to @var{number} if they are of integral type
3380 in registers EAX, EDX, and ECX instead of on the stack. Functions that
3381 take a variable number of arguments will continue to be passed all of their
3382 arguments on the stack.
3384 Beware that on some ELF systems this attribute is unsuitable for
3385 global functions in shared libraries with lazy binding (which is the
3386 default). Lazy binding will send the first call via resolving code in
3387 the loader, which might assume EAX, EDX and ECX can be clobbered, as
3388 per the standard calling conventions. Solaris 8 is affected by this.
3389 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
3390 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
3391 disabled with the linker or the loader if desired, to avoid the
3395 @cindex @code{sseregparm} attribute
3396 On the Intel 386 with SSE support, the @code{sseregparm} attribute
3397 causes the compiler to pass up to 3 floating point arguments in
3398 SSE registers instead of on the stack. Functions that take a
3399 variable number of arguments will continue to pass all of their
3400 floating point arguments on the stack.
3402 @item force_align_arg_pointer
3403 @cindex @code{force_align_arg_pointer} attribute
3404 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
3405 applied to individual function definitions, generating an alternate
3406 prologue and epilogue that realigns the runtime stack if necessary.
3407 This supports mixing legacy codes that run with a 4-byte aligned stack
3408 with modern codes that keep a 16-byte stack for SSE compatibility.
3411 @cindex @code{resbank} attribute
3412 On the SH2A target, this attribute enables the high-speed register
3413 saving and restoration using a register bank for @code{interrupt_handler}
3414 routines. Saving to the bank is performed automatically after the CPU
3415 accepts an interrupt that uses a register bank.
3417 The nineteen 32-bit registers comprising general register R0 to R14,
3418 control register GBR, and system registers MACH, MACL, and PR and the
3419 vector table address offset are saved into a register bank. Register
3420 banks are stacked in first-in last-out (FILO) sequence. Restoration
3421 from the bank is executed by issuing a RESBANK instruction.
3424 @cindex @code{returns_twice} attribute
3425 The @code{returns_twice} attribute tells the compiler that a function may
3426 return more than one time. The compiler will ensure that all registers
3427 are dead before calling such a function and will emit a warning about
3428 the variables that may be clobbered after the second return from the
3429 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3430 The @code{longjmp}-like counterpart of such function, if any, might need
3431 to be marked with the @code{noreturn} attribute.
3434 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
3435 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
3436 all registers except the stack pointer should be saved in the prologue
3437 regardless of whether they are used or not.
3439 @item save_volatiles
3440 @cindex save volatile registers on the MicroBlaze
3441 Use this attribute on the MicroBlaze to indicate that the function is
3442 an interrupt handler. All volatile registers (in addition to non-volatile
3443 registers) will be saved in the function prologue. If the function is a leaf
3444 function, only volatiles used by the function are saved. A normal function
3445 return is generated instead of a return from interrupt.
3447 @item section ("@var{section-name}")
3448 @cindex @code{section} function attribute
3449 Normally, the compiler places the code it generates in the @code{text} section.
3450 Sometimes, however, you need additional sections, or you need certain
3451 particular functions to appear in special sections. The @code{section}
3452 attribute specifies that a function lives in a particular section.
3453 For example, the declaration:
3456 extern void foobar (void) __attribute__ ((section ("bar")));
3460 puts the function @code{foobar} in the @code{bar} section.
3462 Some file formats do not support arbitrary sections so the @code{section}
3463 attribute is not available on all platforms.
3464 If you need to map the entire contents of a module to a particular
3465 section, consider using the facilities of the linker instead.
3468 @cindex @code{sentinel} function attribute
3469 This function attribute ensures that a parameter in a function call is
3470 an explicit @code{NULL}. The attribute is only valid on variadic
3471 functions. By default, the sentinel is located at position zero, the
3472 last parameter of the function call. If an optional integer position
3473 argument P is supplied to the attribute, the sentinel must be located at
3474 position P counting backwards from the end of the argument list.
3477 __attribute__ ((sentinel))
3479 __attribute__ ((sentinel(0)))
3482 The attribute is automatically set with a position of 0 for the built-in
3483 functions @code{execl} and @code{execlp}. The built-in function
3484 @code{execle} has the attribute set with a position of 1.
3486 A valid @code{NULL} in this context is defined as zero with any pointer
3487 type. If your system defines the @code{NULL} macro with an integer type
3488 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3489 with a copy that redefines NULL appropriately.
3491 The warnings for missing or incorrect sentinels are enabled with
3495 See long_call/short_call.
3498 See longcall/shortcall.
3501 @cindex interrupt handler functions on the AVR processors
3502 Use this attribute on the AVR to indicate that the specified
3503 function is an interrupt handler. The compiler will generate function
3504 entry and exit sequences suitable for use in an interrupt handler when this
3505 attribute is present.
3507 See also the @code{interrupt} function attribute.
3509 The AVR hardware globally disables interrupts when an interrupt is executed.
3510 Interrupt handler functions defined with the @code{signal} attribute
3511 do not re-enable interrupts. It is save to enable interrupts in a
3512 @code{signal} handler. This ``save'' only applies to the code
3513 generated by the compiler and not to the IRQ-layout of the
3514 application which is responsibility of the application.
3516 If both @code{signal} and @code{interrupt} are specified for the same
3517 function, @code{signal} will be silently ignored.
3520 Use this attribute on the SH to indicate an @code{interrupt_handler}
3521 function should switch to an alternate stack. It expects a string
3522 argument that names a global variable holding the address of the
3527 void f () __attribute__ ((interrupt_handler,
3528 sp_switch ("alt_stack")));
3532 @cindex functions that pop the argument stack on the 386
3533 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3534 assume that the called function will pop off the stack space used to
3535 pass arguments, unless it takes a variable number of arguments.
3537 @item syscall_linkage
3538 @cindex @code{syscall_linkage} attribute
3539 This attribute is used to modify the IA64 calling convention by marking
3540 all input registers as live at all function exits. This makes it possible
3541 to restart a system call after an interrupt without having to save/restore
3542 the input registers. This also prevents kernel data from leaking into
3546 @cindex @code{target} function attribute
3547 The @code{target} attribute is used to specify that a function is to
3548 be compiled with different target options than specified on the
3549 command line. This can be used for instance to have functions
3550 compiled with a different ISA (instruction set architecture) than the
3551 default. You can also use the @samp{#pragma GCC target} pragma to set
3552 more than one function to be compiled with specific target options.
3553 @xref{Function Specific Option Pragmas}, for details about the
3554 @samp{#pragma GCC target} pragma.
3556 For instance on a 386, you could compile one function with
3557 @code{target("sse4.1,arch=core2")} and another with
3558 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
3559 compiling the first function with @option{-msse4.1} and
3560 @option{-march=core2} options, and the second function with
3561 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
3562 user to make sure that a function is only invoked on a machine that
3563 supports the particular ISA it was compiled for (for example by using
3564 @code{cpuid} on 386 to determine what feature bits and architecture
3568 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3569 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3572 On the 386, the following options are allowed:
3577 @cindex @code{target("abm")} attribute
3578 Enable/disable the generation of the advanced bit instructions.
3582 @cindex @code{target("aes")} attribute
3583 Enable/disable the generation of the AES instructions.
3587 @cindex @code{target("mmx")} attribute
3588 Enable/disable the generation of the MMX instructions.
3592 @cindex @code{target("pclmul")} attribute
3593 Enable/disable the generation of the PCLMUL instructions.
3597 @cindex @code{target("popcnt")} attribute
3598 Enable/disable the generation of the POPCNT instruction.
3602 @cindex @code{target("sse")} attribute
3603 Enable/disable the generation of the SSE instructions.
3607 @cindex @code{target("sse2")} attribute
3608 Enable/disable the generation of the SSE2 instructions.
3612 @cindex @code{target("sse3")} attribute
3613 Enable/disable the generation of the SSE3 instructions.
3617 @cindex @code{target("sse4")} attribute
3618 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3623 @cindex @code{target("sse4.1")} attribute
3624 Enable/disable the generation of the sse4.1 instructions.
3628 @cindex @code{target("sse4.2")} attribute
3629 Enable/disable the generation of the sse4.2 instructions.
3633 @cindex @code{target("sse4a")} attribute
3634 Enable/disable the generation of the SSE4A instructions.
3638 @cindex @code{target("fma4")} attribute
3639 Enable/disable the generation of the FMA4 instructions.
3643 @cindex @code{target("xop")} attribute
3644 Enable/disable the generation of the XOP instructions.
3648 @cindex @code{target("lwp")} attribute
3649 Enable/disable the generation of the LWP instructions.
3653 @cindex @code{target("ssse3")} attribute
3654 Enable/disable the generation of the SSSE3 instructions.
3658 @cindex @code{target("cld")} attribute
3659 Enable/disable the generation of the CLD before string moves.
3661 @item fancy-math-387
3662 @itemx no-fancy-math-387
3663 @cindex @code{target("fancy-math-387")} attribute
3664 Enable/disable the generation of the @code{sin}, @code{cos}, and
3665 @code{sqrt} instructions on the 387 floating point unit.
3668 @itemx no-fused-madd
3669 @cindex @code{target("fused-madd")} attribute
3670 Enable/disable the generation of the fused multiply/add instructions.
3674 @cindex @code{target("ieee-fp")} attribute
3675 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3677 @item inline-all-stringops
3678 @itemx no-inline-all-stringops
3679 @cindex @code{target("inline-all-stringops")} attribute
3680 Enable/disable inlining of string operations.
3682 @item inline-stringops-dynamically
3683 @itemx no-inline-stringops-dynamically
3684 @cindex @code{target("inline-stringops-dynamically")} attribute
3685 Enable/disable the generation of the inline code to do small string
3686 operations and calling the library routines for large operations.
3688 @item align-stringops
3689 @itemx no-align-stringops
3690 @cindex @code{target("align-stringops")} attribute
3691 Do/do not align destination of inlined string operations.
3695 @cindex @code{target("recip")} attribute
3696 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3697 instructions followed an additional Newton-Raphson step instead of
3698 doing a floating point division.
3700 @item arch=@var{ARCH}
3701 @cindex @code{target("arch=@var{ARCH}")} attribute
3702 Specify the architecture to generate code for in compiling the function.
3704 @item tune=@var{TUNE}
3705 @cindex @code{target("tune=@var{TUNE}")} attribute
3706 Specify the architecture to tune for in compiling the function.
3708 @item fpmath=@var{FPMATH}
3709 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3710 Specify which floating point unit to use. The
3711 @code{target("fpmath=sse,387")} option must be specified as
3712 @code{target("fpmath=sse+387")} because the comma would separate
3716 On the PowerPC, the following options are allowed:
3721 @cindex @code{target("altivec")} attribute
3722 Generate code that uses (does not use) AltiVec instructions. In
3723 32-bit code, you cannot enable Altivec instructions unless
3724 @option{-mabi=altivec} was used on the command line.
3728 @cindex @code{target("cmpb")} attribute
3729 Generate code that uses (does not use) the compare bytes instruction
3730 implemented on the POWER6 processor and other processors that support
3731 the PowerPC V2.05 architecture.
3735 @cindex @code{target("dlmzb")} attribute
3736 Generate code that uses (does not use) the string-search @samp{dlmzb}
3737 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
3738 generated by default when targetting those processors.
3742 @cindex @code{target("fprnd")} attribute
3743 Generate code that uses (does not use) the FP round to integer
3744 instructions implemented on the POWER5+ processor and other processors
3745 that support the PowerPC V2.03 architecture.
3749 @cindex @code{target("hard-dfp")} attribute
3750 Generate code that uses (does not use) the decimal floating point
3751 instructions implemented on some POWER processors.
3755 @cindex @code{target("isel")} attribute
3756 Generate code that uses (does not use) ISEL instruction.
3760 @cindex @code{target("mfcrf")} attribute
3761 Generate code that uses (does not use) the move from condition
3762 register field instruction implemented on the POWER4 processor and
3763 other processors that support the PowerPC V2.01 architecture.
3767 @cindex @code{target("mfpgpr")} attribute
3768 Generate code that uses (does not use) the FP move to/from general
3769 purpose register instructions implemented on the POWER6X processor and
3770 other processors that support the extended PowerPC V2.05 architecture.
3774 @cindex @code{target("mulhw")} attribute
3775 Generate code that uses (does not use) the half-word multiply and
3776 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
3777 These instructions are generated by default when targetting those
3782 @cindex @code{target("multiple")} attribute
3783 Generate code that uses (does not use) the load multiple word
3784 instructions and the store multiple word instructions.
3788 @cindex @code{target("update")} attribute
3789 Generate code that uses (does not use) the load or store instructions
3790 that update the base register to the address of the calculated memory
3795 @cindex @code{target("popcntb")} attribute
3796 Generate code that uses (does not use) the popcount and double
3797 precision FP reciprocal estimate instruction implemented on the POWER5
3798 processor and other processors that support the PowerPC V2.02
3803 @cindex @code{target("popcntd")} attribute
3804 Generate code that uses (does not use) the popcount instruction
3805 implemented on the POWER7 processor and other processors that support
3806 the PowerPC V2.06 architecture.
3808 @item powerpc-gfxopt
3809 @itemx no-powerpc-gfxopt
3810 @cindex @code{target("powerpc-gfxopt")} attribute
3811 Generate code that uses (does not use) the optional PowerPC
3812 architecture instructions in the Graphics group, including
3813 floating-point select.
3816 @itemx no-powerpc-gpopt
3817 @cindex @code{target("powerpc-gpopt")} attribute
3818 Generate code that uses (does not use) the optional PowerPC
3819 architecture instructions in the General Purpose group, including
3820 floating-point square root.
3822 @item recip-precision
3823 @itemx no-recip-precision
3824 @cindex @code{target("recip-precision")} attribute
3825 Assume (do not assume) that the reciprocal estimate instructions
3826 provide higher precision estimates than is mandated by the powerpc
3831 @cindex @code{target("string")} attribute
3832 Generate code that uses (does not use) the load string instructions
3833 and the store string word instructions to save multiple registers and
3834 do small block moves.
3838 @cindex @code{target("vsx")} attribute
3839 Generate code that uses (does not use) vector/scalar (VSX)
3840 instructions, and also enable the use of built-in functions that allow
3841 more direct access to the VSX instruction set. In 32-bit code, you
3842 cannot enable VSX or Altivec instructions unless
3843 @option{-mabi=altivec} was used on the command line.
3847 @cindex @code{target("friz")} attribute
3848 Generate (do not generate) the @code{friz} instruction when the
3849 @option{-funsafe-math-optimizations} option is used to optimize
3850 rounding a floating point value to 64-bit integer and back to floating
3851 point. The @code{friz} instruction does not return the same value if
3852 the floating point number is too large to fit in an integer.
3854 @item avoid-indexed-addresses
3855 @itemx no-avoid-indexed-addresses
3856 @cindex @code{target("avoid-indexed-addresses")} attribute
3857 Generate code that tries to avoid (not avoid) the use of indexed load
3858 or store instructions.
3862 @cindex @code{target("paired")} attribute
3863 Generate code that uses (does not use) the generation of PAIRED simd
3868 @cindex @code{target("longcall")} attribute
3869 Generate code that assumes (does not assume) that all calls are far
3870 away so that a longer more expensive calling sequence is required.
3873 @cindex @code{target("cpu=@var{CPU}")} attribute
3874 Specify the architecture to generate code for when compiling the
3875 function. If you select the @code{target("cpu=power7")} attribute when
3876 generating 32-bit code, VSX and Altivec instructions are not generated
3877 unless you use the @option{-mabi=altivec} option on the command line.
3879 @item tune=@var{TUNE}
3880 @cindex @code{target("tune=@var{TUNE}")} attribute
3881 Specify the architecture to tune for when compiling the function. If
3882 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
3883 you do specify the @code{target("cpu=@var{CPU}")} attribute,
3884 compilation will tune for the @var{CPU} architecture, and not the
3885 default tuning specified on the command line.
3888 On the 386/x86_64 and PowerPC backends, you can use either multiple
3889 strings to specify multiple options, or you can separate the option
3890 with a comma (@code{,}).
3892 On the 386/x86_64 and PowerPC backends, the inliner will not inline a
3893 function that has different target options than the caller, unless the
3894 callee has a subset of the target options of the caller. For example
3895 a function declared with @code{target("sse3")} can inline a function
3896 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
3898 The @code{target} attribute is not implemented in GCC versions earlier
3899 than 4.4 for the i386/x86_64 and 4.6 for the PowerPC backends. It is
3900 not currently implemented for other backends.
3903 @cindex tiny data section on the H8/300H and H8S
3904 Use this attribute on the H8/300H and H8S to indicate that the specified
3905 variable should be placed into the tiny data section.
3906 The compiler will generate more efficient code for loads and stores
3907 on data in the tiny data section. Note the tiny data area is limited to
3908 slightly under 32kbytes of data.
3911 Use this attribute on the SH for an @code{interrupt_handler} to return using
3912 @code{trapa} instead of @code{rte}. This attribute expects an integer
3913 argument specifying the trap number to be used.
3916 @cindex @code{unused} attribute.
3917 This attribute, attached to a function, means that the function is meant
3918 to be possibly unused. GCC will not produce a warning for this
3922 @cindex @code{used} attribute.
3923 This attribute, attached to a function, means that code must be emitted
3924 for the function even if it appears that the function is not referenced.
3925 This is useful, for example, when the function is referenced only in
3928 When applied to a member function of a C++ class template, the
3929 attribute also means that the function will be instantiated if the
3930 class itself is instantiated.
3933 @cindex @code{version_id} attribute
3934 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3935 symbol to contain a version string, thus allowing for function level
3936 versioning. HP-UX system header files may use version level functioning
3937 for some system calls.
3940 extern int foo () __attribute__((version_id ("20040821")));
3943 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3945 @item visibility ("@var{visibility_type}")
3946 @cindex @code{visibility} attribute
3947 This attribute affects the linkage of the declaration to which it is attached.
3948 There are four supported @var{visibility_type} values: default,
3949 hidden, protected or internal visibility.
3952 void __attribute__ ((visibility ("protected")))
3953 f () @{ /* @r{Do something.} */; @}
3954 int i __attribute__ ((visibility ("hidden")));
3957 The possible values of @var{visibility_type} correspond to the
3958 visibility settings in the ELF gABI.
3961 @c keep this list of visibilities in alphabetical order.
3964 Default visibility is the normal case for the object file format.
3965 This value is available for the visibility attribute to override other
3966 options that may change the assumed visibility of entities.
3968 On ELF, default visibility means that the declaration is visible to other
3969 modules and, in shared libraries, means that the declared entity may be
3972 On Darwin, default visibility means that the declaration is visible to
3975 Default visibility corresponds to ``external linkage'' in the language.
3978 Hidden visibility indicates that the entity declared will have a new
3979 form of linkage, which we'll call ``hidden linkage''. Two
3980 declarations of an object with hidden linkage refer to the same object
3981 if they are in the same shared object.
3984 Internal visibility is like hidden visibility, but with additional
3985 processor specific semantics. Unless otherwise specified by the
3986 psABI, GCC defines internal visibility to mean that a function is
3987 @emph{never} called from another module. Compare this with hidden
3988 functions which, while they cannot be referenced directly by other
3989 modules, can be referenced indirectly via function pointers. By
3990 indicating that a function cannot be called from outside the module,
3991 GCC may for instance omit the load of a PIC register since it is known
3992 that the calling function loaded the correct value.
3995 Protected visibility is like default visibility except that it
3996 indicates that references within the defining module will bind to the
3997 definition in that module. That is, the declared entity cannot be
3998 overridden by another module.
4002 All visibilities are supported on many, but not all, ELF targets
4003 (supported when the assembler supports the @samp{.visibility}
4004 pseudo-op). Default visibility is supported everywhere. Hidden
4005 visibility is supported on Darwin targets.
4007 The visibility attribute should be applied only to declarations which
4008 would otherwise have external linkage. The attribute should be applied
4009 consistently, so that the same entity should not be declared with
4010 different settings of the attribute.
4012 In C++, the visibility attribute applies to types as well as functions
4013 and objects, because in C++ types have linkage. A class must not have
4014 greater visibility than its non-static data member types and bases,
4015 and class members default to the visibility of their class. Also, a
4016 declaration without explicit visibility is limited to the visibility
4019 In C++, you can mark member functions and static member variables of a
4020 class with the visibility attribute. This is useful if you know a
4021 particular method or static member variable should only be used from
4022 one shared object; then you can mark it hidden while the rest of the
4023 class has default visibility. Care must be taken to avoid breaking
4024 the One Definition Rule; for example, it is usually not useful to mark
4025 an inline method as hidden without marking the whole class as hidden.
4027 A C++ namespace declaration can also have the visibility attribute.
4028 This attribute applies only to the particular namespace body, not to
4029 other definitions of the same namespace; it is equivalent to using
4030 @samp{#pragma GCC visibility} before and after the namespace
4031 definition (@pxref{Visibility Pragmas}).
4033 In C++, if a template argument has limited visibility, this
4034 restriction is implicitly propagated to the template instantiation.
4035 Otherwise, template instantiations and specializations default to the
4036 visibility of their template.
4038 If both the template and enclosing class have explicit visibility, the
4039 visibility from the template is used.
4042 @cindex @code{vliw} attribute
4043 On MeP, the @code{vliw} attribute tells the compiler to emit
4044 instructions in VLIW mode instead of core mode. Note that this
4045 attribute is not allowed unless a VLIW coprocessor has been configured
4046 and enabled through command line options.
4048 @item warn_unused_result
4049 @cindex @code{warn_unused_result} attribute
4050 The @code{warn_unused_result} attribute causes a warning to be emitted
4051 if a caller of the function with this attribute does not use its
4052 return value. This is useful for functions where not checking
4053 the result is either a security problem or always a bug, such as
4057 int fn () __attribute__ ((warn_unused_result));
4060 if (fn () < 0) return -1;
4066 results in warning on line 5.
4069 @cindex @code{weak} attribute
4070 The @code{weak} attribute causes the declaration to be emitted as a weak
4071 symbol rather than a global. This is primarily useful in defining
4072 library functions which can be overridden in user code, though it can
4073 also be used with non-function declarations. Weak symbols are supported
4074 for ELF targets, and also for a.out targets when using the GNU assembler
4078 @itemx weakref ("@var{target}")
4079 @cindex @code{weakref} attribute
4080 The @code{weakref} attribute marks a declaration as a weak reference.
4081 Without arguments, it should be accompanied by an @code{alias} attribute
4082 naming the target symbol. Optionally, the @var{target} may be given as
4083 an argument to @code{weakref} itself. In either case, @code{weakref}
4084 implicitly marks the declaration as @code{weak}. Without a
4085 @var{target}, given as an argument to @code{weakref} or to @code{alias},
4086 @code{weakref} is equivalent to @code{weak}.
4089 static int x() __attribute__ ((weakref ("y")));
4090 /* is equivalent to... */
4091 static int x() __attribute__ ((weak, weakref, alias ("y")));
4093 static int x() __attribute__ ((weakref));
4094 static int x() __attribute__ ((alias ("y")));
4097 A weak reference is an alias that does not by itself require a
4098 definition to be given for the target symbol. If the target symbol is
4099 only referenced through weak references, then it becomes a @code{weak}
4100 undefined symbol. If it is directly referenced, however, then such
4101 strong references prevail, and a definition will be required for the
4102 symbol, not necessarily in the same translation unit.
4104 The effect is equivalent to moving all references to the alias to a
4105 separate translation unit, renaming the alias to the aliased symbol,
4106 declaring it as weak, compiling the two separate translation units and
4107 performing a reloadable link on them.
4109 At present, a declaration to which @code{weakref} is attached can
4110 only be @code{static}.
4114 You can specify multiple attributes in a declaration by separating them
4115 by commas within the double parentheses or by immediately following an
4116 attribute declaration with another attribute declaration.
4118 @cindex @code{#pragma}, reason for not using
4119 @cindex pragma, reason for not using
4120 Some people object to the @code{__attribute__} feature, suggesting that
4121 ISO C's @code{#pragma} should be used instead. At the time
4122 @code{__attribute__} was designed, there were two reasons for not doing
4127 It is impossible to generate @code{#pragma} commands from a macro.
4130 There is no telling what the same @code{#pragma} might mean in another
4134 These two reasons applied to almost any application that might have been
4135 proposed for @code{#pragma}. It was basically a mistake to use
4136 @code{#pragma} for @emph{anything}.
4138 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
4139 to be generated from macros. In addition, a @code{#pragma GCC}
4140 namespace is now in use for GCC-specific pragmas. However, it has been
4141 found convenient to use @code{__attribute__} to achieve a natural
4142 attachment of attributes to their corresponding declarations, whereas
4143 @code{#pragma GCC} is of use for constructs that do not naturally form
4144 part of the grammar. @xref{Other Directives,,Miscellaneous
4145 Preprocessing Directives, cpp, The GNU C Preprocessor}.
4147 @node Attribute Syntax
4148 @section Attribute Syntax
4149 @cindex attribute syntax
4151 This section describes the syntax with which @code{__attribute__} may be
4152 used, and the constructs to which attribute specifiers bind, for the C
4153 language. Some details may vary for C++ and Objective-C@. Because of
4154 infelicities in the grammar for attributes, some forms described here
4155 may not be successfully parsed in all cases.
4157 There are some problems with the semantics of attributes in C++. For
4158 example, there are no manglings for attributes, although they may affect
4159 code generation, so problems may arise when attributed types are used in
4160 conjunction with templates or overloading. Similarly, @code{typeid}
4161 does not distinguish between types with different attributes. Support
4162 for attributes in C++ may be restricted in future to attributes on
4163 declarations only, but not on nested declarators.
4165 @xref{Function Attributes}, for details of the semantics of attributes
4166 applying to functions. @xref{Variable Attributes}, for details of the
4167 semantics of attributes applying to variables. @xref{Type Attributes},
4168 for details of the semantics of attributes applying to structure, union
4169 and enumerated types.
4171 An @dfn{attribute specifier} is of the form
4172 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
4173 is a possibly empty comma-separated sequence of @dfn{attributes}, where
4174 each attribute is one of the following:
4178 Empty. Empty attributes are ignored.
4181 A word (which may be an identifier such as @code{unused}, or a reserved
4182 word such as @code{const}).
4185 A word, followed by, in parentheses, parameters for the attribute.
4186 These parameters take one of the following forms:
4190 An identifier. For example, @code{mode} attributes use this form.
4193 An identifier followed by a comma and a non-empty comma-separated list
4194 of expressions. For example, @code{format} attributes use this form.
4197 A possibly empty comma-separated list of expressions. For example,
4198 @code{format_arg} attributes use this form with the list being a single
4199 integer constant expression, and @code{alias} attributes use this form
4200 with the list being a single string constant.
4204 An @dfn{attribute specifier list} is a sequence of one or more attribute
4205 specifiers, not separated by any other tokens.
4207 In GNU C, an attribute specifier list may appear after the colon following a
4208 label, other than a @code{case} or @code{default} label. The only
4209 attribute it makes sense to use after a label is @code{unused}. This
4210 feature is intended for code generated by programs which contains labels
4211 that may be unused but which is compiled with @option{-Wall}. It would
4212 not normally be appropriate to use in it human-written code, though it
4213 could be useful in cases where the code that jumps to the label is
4214 contained within an @code{#ifdef} conditional. GNU C++ only permits
4215 attributes on labels if the attribute specifier is immediately
4216 followed by a semicolon (i.e., the label applies to an empty
4217 statement). If the semicolon is missing, C++ label attributes are
4218 ambiguous, as it is permissible for a declaration, which could begin
4219 with an attribute list, to be labelled in C++. Declarations cannot be
4220 labelled in C90 or C99, so the ambiguity does not arise there.
4222 An attribute specifier list may appear as part of a @code{struct},
4223 @code{union} or @code{enum} specifier. It may go either immediately
4224 after the @code{struct}, @code{union} or @code{enum} keyword, or after
4225 the closing brace. The former syntax is preferred.
4226 Where attribute specifiers follow the closing brace, they are considered
4227 to relate to the structure, union or enumerated type defined, not to any
4228 enclosing declaration the type specifier appears in, and the type
4229 defined is not complete until after the attribute specifiers.
4230 @c Otherwise, there would be the following problems: a shift/reduce
4231 @c conflict between attributes binding the struct/union/enum and
4232 @c binding to the list of specifiers/qualifiers; and "aligned"
4233 @c attributes could use sizeof for the structure, but the size could be
4234 @c changed later by "packed" attributes.
4236 Otherwise, an attribute specifier appears as part of a declaration,
4237 counting declarations of unnamed parameters and type names, and relates
4238 to that declaration (which may be nested in another declaration, for
4239 example in the case of a parameter declaration), or to a particular declarator
4240 within a declaration. Where an
4241 attribute specifier is applied to a parameter declared as a function or
4242 an array, it should apply to the function or array rather than the
4243 pointer to which the parameter is implicitly converted, but this is not
4244 yet correctly implemented.
4246 Any list of specifiers and qualifiers at the start of a declaration may
4247 contain attribute specifiers, whether or not such a list may in that
4248 context contain storage class specifiers. (Some attributes, however,
4249 are essentially in the nature of storage class specifiers, and only make
4250 sense where storage class specifiers may be used; for example,
4251 @code{section}.) There is one necessary limitation to this syntax: the
4252 first old-style parameter declaration in a function definition cannot
4253 begin with an attribute specifier, because such an attribute applies to
4254 the function instead by syntax described below (which, however, is not
4255 yet implemented in this case). In some other cases, attribute
4256 specifiers are permitted by this grammar but not yet supported by the
4257 compiler. All attribute specifiers in this place relate to the
4258 declaration as a whole. In the obsolescent usage where a type of
4259 @code{int} is implied by the absence of type specifiers, such a list of
4260 specifiers and qualifiers may be an attribute specifier list with no
4261 other specifiers or qualifiers.
4263 At present, the first parameter in a function prototype must have some
4264 type specifier which is not an attribute specifier; this resolves an
4265 ambiguity in the interpretation of @code{void f(int
4266 (__attribute__((foo)) x))}, but is subject to change. At present, if
4267 the parentheses of a function declarator contain only attributes then
4268 those attributes are ignored, rather than yielding an error or warning
4269 or implying a single parameter of type int, but this is subject to
4272 An attribute specifier list may appear immediately before a declarator
4273 (other than the first) in a comma-separated list of declarators in a
4274 declaration of more than one identifier using a single list of
4275 specifiers and qualifiers. Such attribute specifiers apply
4276 only to the identifier before whose declarator they appear. For
4280 __attribute__((noreturn)) void d0 (void),
4281 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
4286 the @code{noreturn} attribute applies to all the functions
4287 declared; the @code{format} attribute only applies to @code{d1}.
4289 An attribute specifier list may appear immediately before the comma,
4290 @code{=} or semicolon terminating the declaration of an identifier other
4291 than a function definition. Such attribute specifiers apply
4292 to the declared object or function. Where an
4293 assembler name for an object or function is specified (@pxref{Asm
4294 Labels}), the attribute must follow the @code{asm}
4297 An attribute specifier list may, in future, be permitted to appear after
4298 the declarator in a function definition (before any old-style parameter
4299 declarations or the function body).
4301 Attribute specifiers may be mixed with type qualifiers appearing inside
4302 the @code{[]} of a parameter array declarator, in the C99 construct by
4303 which such qualifiers are applied to the pointer to which the array is
4304 implicitly converted. Such attribute specifiers apply to the pointer,
4305 not to the array, but at present this is not implemented and they are
4308 An attribute specifier list may appear at the start of a nested
4309 declarator. At present, there are some limitations in this usage: the
4310 attributes correctly apply to the declarator, but for most individual
4311 attributes the semantics this implies are not implemented.
4312 When attribute specifiers follow the @code{*} of a pointer
4313 declarator, they may be mixed with any type qualifiers present.
4314 The following describes the formal semantics of this syntax. It will make the
4315 most sense if you are familiar with the formal specification of
4316 declarators in the ISO C standard.
4318 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
4319 D1}, where @code{T} contains declaration specifiers that specify a type
4320 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
4321 contains an identifier @var{ident}. The type specified for @var{ident}
4322 for derived declarators whose type does not include an attribute
4323 specifier is as in the ISO C standard.
4325 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
4326 and the declaration @code{T D} specifies the type
4327 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4328 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4329 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
4331 If @code{D1} has the form @code{*
4332 @var{type-qualifier-and-attribute-specifier-list} D}, and the
4333 declaration @code{T D} specifies the type
4334 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4335 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4336 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
4342 void (__attribute__((noreturn)) ****f) (void);
4346 specifies the type ``pointer to pointer to pointer to pointer to
4347 non-returning function returning @code{void}''. As another example,
4350 char *__attribute__((aligned(8))) *f;
4354 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
4355 Note again that this does not work with most attributes; for example,
4356 the usage of @samp{aligned} and @samp{noreturn} attributes given above
4357 is not yet supported.
4359 For compatibility with existing code written for compiler versions that
4360 did not implement attributes on nested declarators, some laxity is
4361 allowed in the placing of attributes. If an attribute that only applies
4362 to types is applied to a declaration, it will be treated as applying to
4363 the type of that declaration. If an attribute that only applies to
4364 declarations is applied to the type of a declaration, it will be treated
4365 as applying to that declaration; and, for compatibility with code
4366 placing the attributes immediately before the identifier declared, such
4367 an attribute applied to a function return type will be treated as
4368 applying to the function type, and such an attribute applied to an array
4369 element type will be treated as applying to the array type. If an
4370 attribute that only applies to function types is applied to a
4371 pointer-to-function type, it will be treated as applying to the pointer
4372 target type; if such an attribute is applied to a function return type
4373 that is not a pointer-to-function type, it will be treated as applying
4374 to the function type.
4376 @node Function Prototypes
4377 @section Prototypes and Old-Style Function Definitions
4378 @cindex function prototype declarations
4379 @cindex old-style function definitions
4380 @cindex promotion of formal parameters
4382 GNU C extends ISO C to allow a function prototype to override a later
4383 old-style non-prototype definition. Consider the following example:
4386 /* @r{Use prototypes unless the compiler is old-fashioned.} */
4393 /* @r{Prototype function declaration.} */
4394 int isroot P((uid_t));
4396 /* @r{Old-style function definition.} */
4398 isroot (x) /* @r{??? lossage here ???} */
4405 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
4406 not allow this example, because subword arguments in old-style
4407 non-prototype definitions are promoted. Therefore in this example the
4408 function definition's argument is really an @code{int}, which does not
4409 match the prototype argument type of @code{short}.
4411 This restriction of ISO C makes it hard to write code that is portable
4412 to traditional C compilers, because the programmer does not know
4413 whether the @code{uid_t} type is @code{short}, @code{int}, or
4414 @code{long}. Therefore, in cases like these GNU C allows a prototype
4415 to override a later old-style definition. More precisely, in GNU C, a
4416 function prototype argument type overrides the argument type specified
4417 by a later old-style definition if the former type is the same as the
4418 latter type before promotion. Thus in GNU C the above example is
4419 equivalent to the following:
4432 GNU C++ does not support old-style function definitions, so this
4433 extension is irrelevant.
4436 @section C++ Style Comments
4438 @cindex C++ comments
4439 @cindex comments, C++ style
4441 In GNU C, you may use C++ style comments, which start with @samp{//} and
4442 continue until the end of the line. Many other C implementations allow
4443 such comments, and they are included in the 1999 C standard. However,
4444 C++ style comments are not recognized if you specify an @option{-std}
4445 option specifying a version of ISO C before C99, or @option{-ansi}
4446 (equivalent to @option{-std=c90}).
4449 @section Dollar Signs in Identifier Names
4451 @cindex dollar signs in identifier names
4452 @cindex identifier names, dollar signs in
4454 In GNU C, you may normally use dollar signs in identifier names.
4455 This is because many traditional C implementations allow such identifiers.
4456 However, dollar signs in identifiers are not supported on a few target
4457 machines, typically because the target assembler does not allow them.
4459 @node Character Escapes
4460 @section The Character @key{ESC} in Constants
4462 You can use the sequence @samp{\e} in a string or character constant to
4463 stand for the ASCII character @key{ESC}.
4465 @node Variable Attributes
4466 @section Specifying Attributes of Variables
4467 @cindex attribute of variables
4468 @cindex variable attributes
4470 The keyword @code{__attribute__} allows you to specify special
4471 attributes of variables or structure fields. This keyword is followed
4472 by an attribute specification inside double parentheses. Some
4473 attributes are currently defined generically for variables.
4474 Other attributes are defined for variables on particular target
4475 systems. Other attributes are available for functions
4476 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
4477 Other front ends might define more attributes
4478 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
4480 You may also specify attributes with @samp{__} preceding and following
4481 each keyword. This allows you to use them in header files without
4482 being concerned about a possible macro of the same name. For example,
4483 you may use @code{__aligned__} instead of @code{aligned}.
4485 @xref{Attribute Syntax}, for details of the exact syntax for using
4489 @cindex @code{aligned} attribute
4490 @item aligned (@var{alignment})
4491 This attribute specifies a minimum alignment for the variable or
4492 structure field, measured in bytes. For example, the declaration:
4495 int x __attribute__ ((aligned (16))) = 0;
4499 causes the compiler to allocate the global variable @code{x} on a
4500 16-byte boundary. On a 68040, this could be used in conjunction with
4501 an @code{asm} expression to access the @code{move16} instruction which
4502 requires 16-byte aligned operands.
4504 You can also specify the alignment of structure fields. For example, to
4505 create a double-word aligned @code{int} pair, you could write:
4508 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
4512 This is an alternative to creating a union with a @code{double} member
4513 that forces the union to be double-word aligned.
4515 As in the preceding examples, you can explicitly specify the alignment
4516 (in bytes) that you wish the compiler to use for a given variable or
4517 structure field. Alternatively, you can leave out the alignment factor
4518 and just ask the compiler to align a variable or field to the
4519 default alignment for the target architecture you are compiling for.
4520 The default alignment is sufficient for all scalar types, but may not be
4521 enough for all vector types on a target which supports vector operations.
4522 The default alignment is fixed for a particular target ABI.
4524 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
4525 which is the largest alignment ever used for any data type on the
4526 target machine you are compiling for. For example, you could write:
4529 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
4532 The compiler automatically sets the alignment for the declared
4533 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
4534 often make copy operations more efficient, because the compiler can
4535 use whatever instructions copy the biggest chunks of memory when
4536 performing copies to or from the variables or fields that you have
4537 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
4538 may change depending on command line options.
4540 When used on a struct, or struct member, the @code{aligned} attribute can
4541 only increase the alignment; in order to decrease it, the @code{packed}
4542 attribute must be specified as well. When used as part of a typedef, the
4543 @code{aligned} attribute can both increase and decrease alignment, and
4544 specifying the @code{packed} attribute will generate a warning.
4546 Note that the effectiveness of @code{aligned} attributes may be limited
4547 by inherent limitations in your linker. On many systems, the linker is
4548 only able to arrange for variables to be aligned up to a certain maximum
4549 alignment. (For some linkers, the maximum supported alignment may
4550 be very very small.) If your linker is only able to align variables
4551 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4552 in an @code{__attribute__} will still only provide you with 8 byte
4553 alignment. See your linker documentation for further information.
4555 The @code{aligned} attribute can also be used for functions
4556 (@pxref{Function Attributes}.)
4558 @item cleanup (@var{cleanup_function})
4559 @cindex @code{cleanup} attribute
4560 The @code{cleanup} attribute runs a function when the variable goes
4561 out of scope. This attribute can only be applied to auto function
4562 scope variables; it may not be applied to parameters or variables
4563 with static storage duration. The function must take one parameter,
4564 a pointer to a type compatible with the variable. The return value
4565 of the function (if any) is ignored.
4567 If @option{-fexceptions} is enabled, then @var{cleanup_function}
4568 will be run during the stack unwinding that happens during the
4569 processing of the exception. Note that the @code{cleanup} attribute
4570 does not allow the exception to be caught, only to perform an action.
4571 It is undefined what happens if @var{cleanup_function} does not
4576 @cindex @code{common} attribute
4577 @cindex @code{nocommon} attribute
4580 The @code{common} attribute requests GCC to place a variable in
4581 ``common'' storage. The @code{nocommon} attribute requests the
4582 opposite---to allocate space for it directly.
4584 These attributes override the default chosen by the
4585 @option{-fno-common} and @option{-fcommon} flags respectively.
4588 @itemx deprecated (@var{msg})
4589 @cindex @code{deprecated} attribute
4590 The @code{deprecated} attribute results in a warning if the variable
4591 is used anywhere in the source file. This is useful when identifying
4592 variables that are expected to be removed in a future version of a
4593 program. The warning also includes the location of the declaration
4594 of the deprecated variable, to enable users to easily find further
4595 information about why the variable is deprecated, or what they should
4596 do instead. Note that the warning only occurs for uses:
4599 extern int old_var __attribute__ ((deprecated));
4601 int new_fn () @{ return old_var; @}
4604 results in a warning on line 3 but not line 2. The optional msg
4605 argument, which must be a string, will be printed in the warning if
4608 The @code{deprecated} attribute can also be used for functions and
4609 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
4611 @item mode (@var{mode})
4612 @cindex @code{mode} attribute
4613 This attribute specifies the data type for the declaration---whichever
4614 type corresponds to the mode @var{mode}. This in effect lets you
4615 request an integer or floating point type according to its width.
4617 You may also specify a mode of @samp{byte} or @samp{__byte__} to
4618 indicate the mode corresponding to a one-byte integer, @samp{word} or
4619 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
4620 or @samp{__pointer__} for the mode used to represent pointers.
4623 @cindex @code{packed} attribute
4624 The @code{packed} attribute specifies that a variable or structure field
4625 should have the smallest possible alignment---one byte for a variable,
4626 and one bit for a field, unless you specify a larger value with the
4627 @code{aligned} attribute.
4629 Here is a structure in which the field @code{x} is packed, so that it
4630 immediately follows @code{a}:
4636 int x[2] __attribute__ ((packed));
4640 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
4641 @code{packed} attribute on bit-fields of type @code{char}. This has
4642 been fixed in GCC 4.4 but the change can lead to differences in the
4643 structure layout. See the documentation of
4644 @option{-Wpacked-bitfield-compat} for more information.
4646 @item section ("@var{section-name}")
4647 @cindex @code{section} variable attribute
4648 Normally, the compiler places the objects it generates in sections like
4649 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
4650 or you need certain particular variables to appear in special sections,
4651 for example to map to special hardware. The @code{section}
4652 attribute specifies that a variable (or function) lives in a particular
4653 section. For example, this small program uses several specific section names:
4656 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
4657 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
4658 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
4659 int init_data __attribute__ ((section ("INITDATA")));
4663 /* @r{Initialize stack pointer} */
4664 init_sp (stack + sizeof (stack));
4666 /* @r{Initialize initialized data} */
4667 memcpy (&init_data, &data, &edata - &data);
4669 /* @r{Turn on the serial ports} */
4676 Use the @code{section} attribute with
4677 @emph{global} variables and not @emph{local} variables,
4678 as shown in the example.
4680 You may use the @code{section} attribute with initialized or
4681 uninitialized global variables but the linker requires
4682 each object be defined once, with the exception that uninitialized
4683 variables tentatively go in the @code{common} (or @code{bss}) section
4684 and can be multiply ``defined''. Using the @code{section} attribute
4685 will change what section the variable goes into and may cause the
4686 linker to issue an error if an uninitialized variable has multiple
4687 definitions. You can force a variable to be initialized with the
4688 @option{-fno-common} flag or the @code{nocommon} attribute.
4690 Some file formats do not support arbitrary sections so the @code{section}
4691 attribute is not available on all platforms.
4692 If you need to map the entire contents of a module to a particular
4693 section, consider using the facilities of the linker instead.
4696 @cindex @code{shared} variable attribute
4697 On Microsoft Windows, in addition to putting variable definitions in a named
4698 section, the section can also be shared among all running copies of an
4699 executable or DLL@. For example, this small program defines shared data
4700 by putting it in a named section @code{shared} and marking the section
4704 int foo __attribute__((section ("shared"), shared)) = 0;
4709 /* @r{Read and write foo. All running
4710 copies see the same value.} */
4716 You may only use the @code{shared} attribute along with @code{section}
4717 attribute with a fully initialized global definition because of the way
4718 linkers work. See @code{section} attribute for more information.
4720 The @code{shared} attribute is only available on Microsoft Windows@.
4722 @item tls_model ("@var{tls_model}")
4723 @cindex @code{tls_model} attribute
4724 The @code{tls_model} attribute sets thread-local storage model
4725 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
4726 overriding @option{-ftls-model=} command-line switch on a per-variable
4728 The @var{tls_model} argument should be one of @code{global-dynamic},
4729 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
4731 Not all targets support this attribute.
4734 This attribute, attached to a variable, means that the variable is meant
4735 to be possibly unused. GCC will not produce a warning for this
4739 This attribute, attached to a variable, means that the variable must be
4740 emitted even if it appears that the variable is not referenced.
4742 When applied to a static data member of a C++ class template, the
4743 attribute also means that the member will be instantiated if the
4744 class itself is instantiated.
4746 @item vector_size (@var{bytes})
4747 This attribute specifies the vector size for the variable, measured in
4748 bytes. For example, the declaration:
4751 int foo __attribute__ ((vector_size (16)));
4755 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
4756 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
4757 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
4759 This attribute is only applicable to integral and float scalars,
4760 although arrays, pointers, and function return values are allowed in
4761 conjunction with this construct.
4763 Aggregates with this attribute are invalid, even if they are of the same
4764 size as a corresponding scalar. For example, the declaration:
4767 struct S @{ int a; @};
4768 struct S __attribute__ ((vector_size (16))) foo;
4772 is invalid even if the size of the structure is the same as the size of
4776 The @code{selectany} attribute causes an initialized global variable to
4777 have link-once semantics. When multiple definitions of the variable are
4778 encountered by the linker, the first is selected and the remainder are
4779 discarded. Following usage by the Microsoft compiler, the linker is told
4780 @emph{not} to warn about size or content differences of the multiple
4783 Although the primary usage of this attribute is for POD types, the
4784 attribute can also be applied to global C++ objects that are initialized
4785 by a constructor. In this case, the static initialization and destruction
4786 code for the object is emitted in each translation defining the object,
4787 but the calls to the constructor and destructor are protected by a
4788 link-once guard variable.
4790 The @code{selectany} attribute is only available on Microsoft Windows
4791 targets. You can use @code{__declspec (selectany)} as a synonym for
4792 @code{__attribute__ ((selectany))} for compatibility with other
4796 The @code{weak} attribute is described in @ref{Function Attributes}.
4799 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4802 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4806 @anchor{AVR Variable Attributes}
4807 @subsection AVR Variable Attributes
4811 @cindex @code{progmem} AVR variable attribute
4812 The @code{progmem} attribute is used on the AVR to place read-only
4813 data in the non-volatile program memory (flash). The @code{progmem}
4814 attribute accomplishes this by putting respective variables into a
4815 section whose name starts with @code{.progmem}.
4817 This attribute works similar to the @code{section} attribute
4818 but adds additional checking. Notice that just like the
4819 @code{section} attribute, @code{progmem} affects the location
4820 of the data but not how this data is accessed.
4822 In order to read data located with the @code{progmem} attribute
4823 (inline) assembler must be used.
4825 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
4826 #include <avr/pgmspace.h>
4828 /* Locate var in flash memory */
4829 const int var[2] PROGMEM = @{ 1, 2 @};
4831 int read_var (int i)
4833 /* Access var[] by accessor macro from avr/pgmspace.h */
4834 return (int) pgm_read_word (& var[i]);
4838 AVR is a Harvard architecture processor and data and read-only data
4839 normally resides in the data memory (RAM).
4841 See also the @ref{AVR Named Address Spaces} section for
4842 an alternate way to locate and access data in flash memory.
4845 @subsection Blackfin Variable Attributes
4847 Three attributes are currently defined for the Blackfin.
4853 @cindex @code{l1_data} variable attribute
4854 @cindex @code{l1_data_A} variable attribute
4855 @cindex @code{l1_data_B} variable attribute
4856 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4857 Variables with @code{l1_data} attribute will be put into the specific section
4858 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4859 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4860 attribute will be put into the specific section named @code{.l1.data.B}.
4863 @cindex @code{l2} variable attribute
4864 Use this attribute on the Blackfin to place the variable into L2 SRAM.
4865 Variables with @code{l2} attribute will be put into the specific section
4866 named @code{.l2.data}.
4869 @subsection M32R/D Variable Attributes
4871 One attribute is currently defined for the M32R/D@.
4874 @item model (@var{model-name})
4875 @cindex variable addressability on the M32R/D
4876 Use this attribute on the M32R/D to set the addressability of an object.
4877 The identifier @var{model-name} is one of @code{small}, @code{medium},
4878 or @code{large}, representing each of the code models.
4880 Small model objects live in the lower 16MB of memory (so that their
4881 addresses can be loaded with the @code{ld24} instruction).
4883 Medium and large model objects may live anywhere in the 32-bit address space
4884 (the compiler will generate @code{seth/add3} instructions to load their
4888 @anchor{MeP Variable Attributes}
4889 @subsection MeP Variable Attributes
4891 The MeP target has a number of addressing modes and busses. The
4892 @code{near} space spans the standard memory space's first 16 megabytes
4893 (24 bits). The @code{far} space spans the entire 32-bit memory space.
4894 The @code{based} space is a 128 byte region in the memory space which
4895 is addressed relative to the @code{$tp} register. The @code{tiny}
4896 space is a 65536 byte region relative to the @code{$gp} register. In
4897 addition to these memory regions, the MeP target has a separate 16-bit
4898 control bus which is specified with @code{cb} attributes.
4903 Any variable with the @code{based} attribute will be assigned to the
4904 @code{.based} section, and will be accessed with relative to the
4905 @code{$tp} register.
4908 Likewise, the @code{tiny} attribute assigned variables to the
4909 @code{.tiny} section, relative to the @code{$gp} register.
4912 Variables with the @code{near} attribute are assumed to have addresses
4913 that fit in a 24-bit addressing mode. This is the default for large
4914 variables (@code{-mtiny=4} is the default) but this attribute can
4915 override @code{-mtiny=} for small variables, or override @code{-ml}.
4918 Variables with the @code{far} attribute are addressed using a full
4919 32-bit address. Since this covers the entire memory space, this
4920 allows modules to make no assumptions about where variables might be
4924 @itemx io (@var{addr})
4925 Variables with the @code{io} attribute are used to address
4926 memory-mapped peripherals. If an address is specified, the variable
4927 is assigned that address, else it is not assigned an address (it is
4928 assumed some other module will assign an address). Example:
4931 int timer_count __attribute__((io(0x123)));
4935 @itemx cb (@var{addr})
4936 Variables with the @code{cb} attribute are used to access the control
4937 bus, using special instructions. @code{addr} indicates the control bus
4941 int cpu_clock __attribute__((cb(0x123)));
4946 @anchor{i386 Variable Attributes}
4947 @subsection i386 Variable Attributes
4949 Two attributes are currently defined for i386 configurations:
4950 @code{ms_struct} and @code{gcc_struct}
4955 @cindex @code{ms_struct} attribute
4956 @cindex @code{gcc_struct} attribute
4958 If @code{packed} is used on a structure, or if bit-fields are used
4959 it may be that the Microsoft ABI packs them differently
4960 than GCC would normally pack them. Particularly when moving packed
4961 data between functions compiled with GCC and the native Microsoft compiler
4962 (either via function call or as data in a file), it may be necessary to access
4965 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4966 compilers to match the native Microsoft compiler.
4968 The Microsoft structure layout algorithm is fairly simple with the exception
4969 of the bitfield packing:
4971 The padding and alignment of members of structures and whether a bit field
4972 can straddle a storage-unit boundary
4975 @item Structure members are stored sequentially in the order in which they are
4976 declared: the first member has the lowest memory address and the last member
4979 @item Every data object has an alignment-requirement. The alignment-requirement
4980 for all data except structures, unions, and arrays is either the size of the
4981 object or the current packing size (specified with either the aligned attribute
4982 or the pack pragma), whichever is less. For structures, unions, and arrays,
4983 the alignment-requirement is the largest alignment-requirement of its members.
4984 Every object is allocated an offset so that:
4986 offset % alignment-requirement == 0
4988 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4989 unit if the integral types are the same size and if the next bit field fits
4990 into the current allocation unit without crossing the boundary imposed by the
4991 common alignment requirements of the bit fields.
4994 Handling of zero-length bitfields:
4996 MSVC interprets zero-length bitfields in the following ways:
4999 @item If a zero-length bitfield is inserted between two bitfields that would
5000 normally be coalesced, the bitfields will not be coalesced.
5007 unsigned long bf_1 : 12;
5009 unsigned long bf_2 : 12;
5013 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
5014 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
5016 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
5017 alignment of the zero-length bitfield is greater than the member that follows it,
5018 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
5038 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
5039 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
5040 bitfield will not affect the alignment of @code{bar} or, as a result, the size
5043 Taking this into account, it is important to note the following:
5046 @item If a zero-length bitfield follows a normal bitfield, the type of the
5047 zero-length bitfield may affect the alignment of the structure as whole. For
5048 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
5049 normal bitfield, and is of type short.
5051 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
5052 still affect the alignment of the structure:
5062 Here, @code{t4} will take up 4 bytes.
5065 @item Zero-length bitfields following non-bitfield members are ignored:
5076 Here, @code{t5} will take up 2 bytes.
5080 @subsection PowerPC Variable Attributes
5082 Three attributes currently are defined for PowerPC configurations:
5083 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5085 For full documentation of the struct attributes please see the
5086 documentation in @ref{i386 Variable Attributes}.
5088 For documentation of @code{altivec} attribute please see the
5089 documentation in @ref{PowerPC Type Attributes}.
5091 @subsection SPU Variable Attributes
5093 The SPU supports the @code{spu_vector} attribute for variables. For
5094 documentation of this attribute please see the documentation in
5095 @ref{SPU Type Attributes}.
5097 @subsection Xstormy16 Variable Attributes
5099 One attribute is currently defined for xstormy16 configurations:
5104 @cindex @code{below100} attribute
5106 If a variable has the @code{below100} attribute (@code{BELOW100} is
5107 allowed also), GCC will place the variable in the first 0x100 bytes of
5108 memory and use special opcodes to access it. Such variables will be
5109 placed in either the @code{.bss_below100} section or the
5110 @code{.data_below100} section.
5114 @node Type Attributes
5115 @section Specifying Attributes of Types
5116 @cindex attribute of types
5117 @cindex type attributes
5119 The keyword @code{__attribute__} allows you to specify special
5120 attributes of @code{struct} and @code{union} types when you define
5121 such types. This keyword is followed by an attribute specification
5122 inside double parentheses. Seven attributes are currently defined for
5123 types: @code{aligned}, @code{packed}, @code{transparent_union},
5124 @code{unused}, @code{deprecated}, @code{visibility}, and
5125 @code{may_alias}. Other attributes are defined for functions
5126 (@pxref{Function Attributes}) and for variables (@pxref{Variable
5129 You may also specify any one of these attributes with @samp{__}
5130 preceding and following its keyword. This allows you to use these
5131 attributes in header files without being concerned about a possible
5132 macro of the same name. For example, you may use @code{__aligned__}
5133 instead of @code{aligned}.
5135 You may specify type attributes in an enum, struct or union type
5136 declaration or definition, or for other types in a @code{typedef}
5139 For an enum, struct or union type, you may specify attributes either
5140 between the enum, struct or union tag and the name of the type, or
5141 just past the closing curly brace of the @emph{definition}. The
5142 former syntax is preferred.
5144 @xref{Attribute Syntax}, for details of the exact syntax for using
5148 @cindex @code{aligned} attribute
5149 @item aligned (@var{alignment})
5150 This attribute specifies a minimum alignment (in bytes) for variables
5151 of the specified type. For example, the declarations:
5154 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
5155 typedef int more_aligned_int __attribute__ ((aligned (8)));
5159 force the compiler to insure (as far as it can) that each variable whose
5160 type is @code{struct S} or @code{more_aligned_int} will be allocated and
5161 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
5162 variables of type @code{struct S} aligned to 8-byte boundaries allows
5163 the compiler to use the @code{ldd} and @code{std} (doubleword load and
5164 store) instructions when copying one variable of type @code{struct S} to
5165 another, thus improving run-time efficiency.
5167 Note that the alignment of any given @code{struct} or @code{union} type
5168 is required by the ISO C standard to be at least a perfect multiple of
5169 the lowest common multiple of the alignments of all of the members of
5170 the @code{struct} or @code{union} in question. This means that you @emph{can}
5171 effectively adjust the alignment of a @code{struct} or @code{union}
5172 type by attaching an @code{aligned} attribute to any one of the members
5173 of such a type, but the notation illustrated in the example above is a
5174 more obvious, intuitive, and readable way to request the compiler to
5175 adjust the alignment of an entire @code{struct} or @code{union} type.
5177 As in the preceding example, you can explicitly specify the alignment
5178 (in bytes) that you wish the compiler to use for a given @code{struct}
5179 or @code{union} type. Alternatively, you can leave out the alignment factor
5180 and just ask the compiler to align a type to the maximum
5181 useful alignment for the target machine you are compiling for. For
5182 example, you could write:
5185 struct S @{ short f[3]; @} __attribute__ ((aligned));
5188 Whenever you leave out the alignment factor in an @code{aligned}
5189 attribute specification, the compiler automatically sets the alignment
5190 for the type to the largest alignment which is ever used for any data
5191 type on the target machine you are compiling for. Doing this can often
5192 make copy operations more efficient, because the compiler can use
5193 whatever instructions copy the biggest chunks of memory when performing
5194 copies to or from the variables which have types that you have aligned
5197 In the example above, if the size of each @code{short} is 2 bytes, then
5198 the size of the entire @code{struct S} type is 6 bytes. The smallest
5199 power of two which is greater than or equal to that is 8, so the
5200 compiler sets the alignment for the entire @code{struct S} type to 8
5203 Note that although you can ask the compiler to select a time-efficient
5204 alignment for a given type and then declare only individual stand-alone
5205 objects of that type, the compiler's ability to select a time-efficient
5206 alignment is primarily useful only when you plan to create arrays of
5207 variables having the relevant (efficiently aligned) type. If you
5208 declare or use arrays of variables of an efficiently-aligned type, then
5209 it is likely that your program will also be doing pointer arithmetic (or
5210 subscripting, which amounts to the same thing) on pointers to the
5211 relevant type, and the code that the compiler generates for these
5212 pointer arithmetic operations will often be more efficient for
5213 efficiently-aligned types than for other types.
5215 The @code{aligned} attribute can only increase the alignment; but you
5216 can decrease it by specifying @code{packed} as well. See below.
5218 Note that the effectiveness of @code{aligned} attributes may be limited
5219 by inherent limitations in your linker. On many systems, the linker is
5220 only able to arrange for variables to be aligned up to a certain maximum
5221 alignment. (For some linkers, the maximum supported alignment may
5222 be very very small.) If your linker is only able to align variables
5223 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
5224 in an @code{__attribute__} will still only provide you with 8 byte
5225 alignment. See your linker documentation for further information.
5228 This attribute, attached to @code{struct} or @code{union} type
5229 definition, specifies that each member (other than zero-width bitfields)
5230 of the structure or union is placed to minimize the memory required. When
5231 attached to an @code{enum} definition, it indicates that the smallest
5232 integral type should be used.
5234 @opindex fshort-enums
5235 Specifying this attribute for @code{struct} and @code{union} types is
5236 equivalent to specifying the @code{packed} attribute on each of the
5237 structure or union members. Specifying the @option{-fshort-enums}
5238 flag on the line is equivalent to specifying the @code{packed}
5239 attribute on all @code{enum} definitions.
5241 In the following example @code{struct my_packed_struct}'s members are
5242 packed closely together, but the internal layout of its @code{s} member
5243 is not packed---to do that, @code{struct my_unpacked_struct} would need to
5247 struct my_unpacked_struct
5253 struct __attribute__ ((__packed__)) my_packed_struct
5257 struct my_unpacked_struct s;
5261 You may only specify this attribute on the definition of an @code{enum},
5262 @code{struct} or @code{union}, not on a @code{typedef} which does not
5263 also define the enumerated type, structure or union.
5265 @item transparent_union
5266 This attribute, attached to a @code{union} type definition, indicates
5267 that any function parameter having that union type causes calls to that
5268 function to be treated in a special way.
5270 First, the argument corresponding to a transparent union type can be of
5271 any type in the union; no cast is required. Also, if the union contains
5272 a pointer type, the corresponding argument can be a null pointer
5273 constant or a void pointer expression; and if the union contains a void
5274 pointer type, the corresponding argument can be any pointer expression.
5275 If the union member type is a pointer, qualifiers like @code{const} on
5276 the referenced type must be respected, just as with normal pointer
5279 Second, the argument is passed to the function using the calling
5280 conventions of the first member of the transparent union, not the calling
5281 conventions of the union itself. All members of the union must have the
5282 same machine representation; this is necessary for this argument passing
5285 Transparent unions are designed for library functions that have multiple
5286 interfaces for compatibility reasons. For example, suppose the
5287 @code{wait} function must accept either a value of type @code{int *} to
5288 comply with Posix, or a value of type @code{union wait *} to comply with
5289 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
5290 @code{wait} would accept both kinds of arguments, but it would also
5291 accept any other pointer type and this would make argument type checking
5292 less useful. Instead, @code{<sys/wait.h>} might define the interface
5296 typedef union __attribute__ ((__transparent_union__))
5300 @} wait_status_ptr_t;
5302 pid_t wait (wait_status_ptr_t);
5305 This interface allows either @code{int *} or @code{union wait *}
5306 arguments to be passed, using the @code{int *} calling convention.
5307 The program can call @code{wait} with arguments of either type:
5310 int w1 () @{ int w; return wait (&w); @}
5311 int w2 () @{ union wait w; return wait (&w); @}
5314 With this interface, @code{wait}'s implementation might look like this:
5317 pid_t wait (wait_status_ptr_t p)
5319 return waitpid (-1, p.__ip, 0);
5324 When attached to a type (including a @code{union} or a @code{struct}),
5325 this attribute means that variables of that type are meant to appear
5326 possibly unused. GCC will not produce a warning for any variables of
5327 that type, even if the variable appears to do nothing. This is often
5328 the case with lock or thread classes, which are usually defined and then
5329 not referenced, but contain constructors and destructors that have
5330 nontrivial bookkeeping functions.
5333 @itemx deprecated (@var{msg})
5334 The @code{deprecated} attribute results in a warning if the type
5335 is used anywhere in the source file. This is useful when identifying
5336 types that are expected to be removed in a future version of a program.
5337 If possible, the warning also includes the location of the declaration
5338 of the deprecated type, to enable users to easily find further
5339 information about why the type is deprecated, or what they should do
5340 instead. Note that the warnings only occur for uses and then only
5341 if the type is being applied to an identifier that itself is not being
5342 declared as deprecated.
5345 typedef int T1 __attribute__ ((deprecated));
5349 typedef T1 T3 __attribute__ ((deprecated));
5350 T3 z __attribute__ ((deprecated));
5353 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
5354 warning is issued for line 4 because T2 is not explicitly
5355 deprecated. Line 5 has no warning because T3 is explicitly
5356 deprecated. Similarly for line 6. The optional msg
5357 argument, which must be a string, will be printed in the warning if
5360 The @code{deprecated} attribute can also be used for functions and
5361 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
5364 Accesses through pointers to types with this attribute are not subject
5365 to type-based alias analysis, but are instead assumed to be able to alias
5366 any other type of objects. In the context of 6.5/7 an lvalue expression
5367 dereferencing such a pointer is treated like having a character type.
5368 See @option{-fstrict-aliasing} for more information on aliasing issues.
5369 This extension exists to support some vector APIs, in which pointers to
5370 one vector type are permitted to alias pointers to a different vector type.
5372 Note that an object of a type with this attribute does not have any
5378 typedef short __attribute__((__may_alias__)) short_a;
5384 short_a *b = (short_a *) &a;
5388 if (a == 0x12345678)
5395 If you replaced @code{short_a} with @code{short} in the variable
5396 declaration, the above program would abort when compiled with
5397 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
5398 above in recent GCC versions.
5401 In C++, attribute visibility (@pxref{Function Attributes}) can also be
5402 applied to class, struct, union and enum types. Unlike other type
5403 attributes, the attribute must appear between the initial keyword and
5404 the name of the type; it cannot appear after the body of the type.
5406 Note that the type visibility is applied to vague linkage entities
5407 associated with the class (vtable, typeinfo node, etc.). In
5408 particular, if a class is thrown as an exception in one shared object
5409 and caught in another, the class must have default visibility.
5410 Otherwise the two shared objects will be unable to use the same
5411 typeinfo node and exception handling will break.
5415 @subsection ARM Type Attributes
5417 On those ARM targets that support @code{dllimport} (such as Symbian
5418 OS), you can use the @code{notshared} attribute to indicate that the
5419 virtual table and other similar data for a class should not be
5420 exported from a DLL@. For example:
5423 class __declspec(notshared) C @{
5425 __declspec(dllimport) C();
5429 __declspec(dllexport)
5433 In this code, @code{C::C} is exported from the current DLL, but the
5434 virtual table for @code{C} is not exported. (You can use
5435 @code{__attribute__} instead of @code{__declspec} if you prefer, but
5436 most Symbian OS code uses @code{__declspec}.)
5438 @anchor{MeP Type Attributes}
5439 @subsection MeP Type Attributes
5441 Many of the MeP variable attributes may be applied to types as well.
5442 Specifically, the @code{based}, @code{tiny}, @code{near}, and
5443 @code{far} attributes may be applied to either. The @code{io} and
5444 @code{cb} attributes may not be applied to types.
5446 @anchor{i386 Type Attributes}
5447 @subsection i386 Type Attributes
5449 Two attributes are currently defined for i386 configurations:
5450 @code{ms_struct} and @code{gcc_struct}.
5456 @cindex @code{ms_struct}
5457 @cindex @code{gcc_struct}
5459 If @code{packed} is used on a structure, or if bit-fields are used
5460 it may be that the Microsoft ABI packs them differently
5461 than GCC would normally pack them. Particularly when moving packed
5462 data between functions compiled with GCC and the native Microsoft compiler
5463 (either via function call or as data in a file), it may be necessary to access
5466 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
5467 compilers to match the native Microsoft compiler.
5470 To specify multiple attributes, separate them by commas within the
5471 double parentheses: for example, @samp{__attribute__ ((aligned (16),
5474 @anchor{PowerPC Type Attributes}
5475 @subsection PowerPC Type Attributes
5477 Three attributes currently are defined for PowerPC configurations:
5478 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5480 For full documentation of the @code{ms_struct} and @code{gcc_struct}
5481 attributes please see the documentation in @ref{i386 Type Attributes}.
5483 The @code{altivec} attribute allows one to declare AltiVec vector data
5484 types supported by the AltiVec Programming Interface Manual. The
5485 attribute requires an argument to specify one of three vector types:
5486 @code{vector__}, @code{pixel__} (always followed by unsigned short),
5487 and @code{bool__} (always followed by unsigned).
5490 __attribute__((altivec(vector__)))
5491 __attribute__((altivec(pixel__))) unsigned short
5492 __attribute__((altivec(bool__))) unsigned
5495 These attributes mainly are intended to support the @code{__vector},
5496 @code{__pixel}, and @code{__bool} AltiVec keywords.
5498 @anchor{SPU Type Attributes}
5499 @subsection SPU Type Attributes
5501 The SPU supports the @code{spu_vector} attribute for types. This attribute
5502 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
5503 Language Extensions Specification. It is intended to support the
5504 @code{__vector} keyword.
5507 @section Inquiring on Alignment of Types or Variables
5509 @cindex type alignment
5510 @cindex variable alignment
5512 The keyword @code{__alignof__} allows you to inquire about how an object
5513 is aligned, or the minimum alignment usually required by a type. Its
5514 syntax is just like @code{sizeof}.
5516 For example, if the target machine requires a @code{double} value to be
5517 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
5518 This is true on many RISC machines. On more traditional machine
5519 designs, @code{__alignof__ (double)} is 4 or even 2.
5521 Some machines never actually require alignment; they allow reference to any
5522 data type even at an odd address. For these machines, @code{__alignof__}
5523 reports the smallest alignment that GCC will give the data type, usually as
5524 mandated by the target ABI.
5526 If the operand of @code{__alignof__} is an lvalue rather than a type,
5527 its value is the required alignment for its type, taking into account
5528 any minimum alignment specified with GCC's @code{__attribute__}
5529 extension (@pxref{Variable Attributes}). For example, after this
5533 struct foo @{ int x; char y; @} foo1;
5537 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
5538 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
5540 It is an error to ask for the alignment of an incomplete type.
5544 @section An Inline Function is As Fast As a Macro
5545 @cindex inline functions
5546 @cindex integrating function code
5548 @cindex macros, inline alternative
5550 By declaring a function inline, you can direct GCC to make
5551 calls to that function faster. One way GCC can achieve this is to
5552 integrate that function's code into the code for its callers. This
5553 makes execution faster by eliminating the function-call overhead; in
5554 addition, if any of the actual argument values are constant, their
5555 known values may permit simplifications at compile time so that not
5556 all of the inline function's code needs to be included. The effect on
5557 code size is less predictable; object code may be larger or smaller
5558 with function inlining, depending on the particular case. You can
5559 also direct GCC to try to integrate all ``simple enough'' functions
5560 into their callers with the option @option{-finline-functions}.
5562 GCC implements three different semantics of declaring a function
5563 inline. One is available with @option{-std=gnu89} or
5564 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
5565 on all inline declarations, another when
5566 @option{-std=c99}, @option{-std=c11},
5567 @option{-std=gnu99} or @option{-std=gnu11}
5568 (without @option{-fgnu89-inline}), and the third
5569 is used when compiling C++.
5571 To declare a function inline, use the @code{inline} keyword in its
5572 declaration, like this:
5582 If you are writing a header file to be included in ISO C90 programs, write
5583 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
5585 The three types of inlining behave similarly in two important cases:
5586 when the @code{inline} keyword is used on a @code{static} function,
5587 like the example above, and when a function is first declared without
5588 using the @code{inline} keyword and then is defined with
5589 @code{inline}, like this:
5592 extern int inc (int *a);
5600 In both of these common cases, the program behaves the same as if you
5601 had not used the @code{inline} keyword, except for its speed.
5603 @cindex inline functions, omission of
5604 @opindex fkeep-inline-functions
5605 When a function is both inline and @code{static}, if all calls to the
5606 function are integrated into the caller, and the function's address is
5607 never used, then the function's own assembler code is never referenced.
5608 In this case, GCC does not actually output assembler code for the
5609 function, unless you specify the option @option{-fkeep-inline-functions}.
5610 Some calls cannot be integrated for various reasons (in particular,
5611 calls that precede the function's definition cannot be integrated, and
5612 neither can recursive calls within the definition). If there is a
5613 nonintegrated call, then the function is compiled to assembler code as
5614 usual. The function must also be compiled as usual if the program
5615 refers to its address, because that can't be inlined.
5618 Note that certain usages in a function definition can make it unsuitable
5619 for inline substitution. Among these usages are: use of varargs, use of
5620 alloca, use of variable sized data types (@pxref{Variable Length}),
5621 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
5622 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
5623 will warn when a function marked @code{inline} could not be substituted,
5624 and will give the reason for the failure.
5626 @cindex automatic @code{inline} for C++ member fns
5627 @cindex @code{inline} automatic for C++ member fns
5628 @cindex member fns, automatically @code{inline}
5629 @cindex C++ member fns, automatically @code{inline}
5630 @opindex fno-default-inline
5631 As required by ISO C++, GCC considers member functions defined within
5632 the body of a class to be marked inline even if they are
5633 not explicitly declared with the @code{inline} keyword. You can
5634 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
5635 Options,,Options Controlling C++ Dialect}.
5637 GCC does not inline any functions when not optimizing unless you specify
5638 the @samp{always_inline} attribute for the function, like this:
5641 /* @r{Prototype.} */
5642 inline void foo (const char) __attribute__((always_inline));
5645 The remainder of this section is specific to GNU C90 inlining.
5647 @cindex non-static inline function
5648 When an inline function is not @code{static}, then the compiler must assume
5649 that there may be calls from other source files; since a global symbol can
5650 be defined only once in any program, the function must not be defined in
5651 the other source files, so the calls therein cannot be integrated.
5652 Therefore, a non-@code{static} inline function is always compiled on its
5653 own in the usual fashion.
5655 If you specify both @code{inline} and @code{extern} in the function
5656 definition, then the definition is used only for inlining. In no case
5657 is the function compiled on its own, not even if you refer to its
5658 address explicitly. Such an address becomes an external reference, as
5659 if you had only declared the function, and had not defined it.
5661 This combination of @code{inline} and @code{extern} has almost the
5662 effect of a macro. The way to use it is to put a function definition in
5663 a header file with these keywords, and put another copy of the
5664 definition (lacking @code{inline} and @code{extern}) in a library file.
5665 The definition in the header file will cause most calls to the function
5666 to be inlined. If any uses of the function remain, they will refer to
5667 the single copy in the library.
5670 @section When is a Volatile Object Accessed?
5671 @cindex accessing volatiles
5672 @cindex volatile read
5673 @cindex volatile write
5674 @cindex volatile access
5676 C has the concept of volatile objects. These are normally accessed by
5677 pointers and used for accessing hardware or inter-thread
5678 communication. The standard encourages compilers to refrain from
5679 optimizations concerning accesses to volatile objects, but leaves it
5680 implementation defined as to what constitutes a volatile access. The
5681 minimum requirement is that at a sequence point all previous accesses
5682 to volatile objects have stabilized and no subsequent accesses have
5683 occurred. Thus an implementation is free to reorder and combine
5684 volatile accesses which occur between sequence points, but cannot do
5685 so for accesses across a sequence point. The use of volatile does
5686 not allow you to violate the restriction on updating objects multiple
5687 times between two sequence points.
5689 Accesses to non-volatile objects are not ordered with respect to
5690 volatile accesses. You cannot use a volatile object as a memory
5691 barrier to order a sequence of writes to non-volatile memory. For
5695 int *ptr = @var{something};
5697 *ptr = @var{something};
5701 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
5702 that the write to @var{*ptr} will have occurred by the time the update
5703 of @var{vobj} has happened. If you need this guarantee, you must use
5704 a stronger memory barrier such as:
5707 int *ptr = @var{something};
5709 *ptr = @var{something};
5710 asm volatile ("" : : : "memory");
5714 A scalar volatile object is read when it is accessed in a void context:
5717 volatile int *src = @var{somevalue};
5721 Such expressions are rvalues, and GCC implements this as a
5722 read of the volatile object being pointed to.
5724 Assignments are also expressions and have an rvalue. However when
5725 assigning to a scalar volatile, the volatile object is not reread,
5726 regardless of whether the assignment expression's rvalue is used or
5727 not. If the assignment's rvalue is used, the value is that assigned
5728 to the volatile object. For instance, there is no read of @var{vobj}
5729 in all the following cases:
5734 vobj = @var{something};
5735 obj = vobj = @var{something};
5736 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
5737 obj = (@var{something}, vobj = @var{anotherthing});
5740 If you need to read the volatile object after an assignment has
5741 occurred, you must use a separate expression with an intervening
5744 As bitfields are not individually addressable, volatile bitfields may
5745 be implicitly read when written to, or when adjacent bitfields are
5746 accessed. Bitfield operations may be optimized such that adjacent
5747 bitfields are only partially accessed, if they straddle a storage unit
5748 boundary. For these reasons it is unwise to use volatile bitfields to
5752 @section Assembler Instructions with C Expression Operands
5753 @cindex extended @code{asm}
5754 @cindex @code{asm} expressions
5755 @cindex assembler instructions
5758 In an assembler instruction using @code{asm}, you can specify the
5759 operands of the instruction using C expressions. This means you need not
5760 guess which registers or memory locations will contain the data you want
5763 You must specify an assembler instruction template much like what
5764 appears in a machine description, plus an operand constraint string for
5767 For example, here is how to use the 68881's @code{fsinx} instruction:
5770 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
5774 Here @code{angle} is the C expression for the input operand while
5775 @code{result} is that of the output operand. Each has @samp{"f"} as its
5776 operand constraint, saying that a floating point register is required.
5777 The @samp{=} in @samp{=f} indicates that the operand is an output; all
5778 output operands' constraints must use @samp{=}. The constraints use the
5779 same language used in the machine description (@pxref{Constraints}).
5781 Each operand is described by an operand-constraint string followed by
5782 the C expression in parentheses. A colon separates the assembler
5783 template from the first output operand and another separates the last
5784 output operand from the first input, if any. Commas separate the
5785 operands within each group. The total number of operands is currently
5786 limited to 30; this limitation may be lifted in some future version of
5789 If there are no output operands but there are input operands, you must
5790 place two consecutive colons surrounding the place where the output
5793 As of GCC version 3.1, it is also possible to specify input and output
5794 operands using symbolic names which can be referenced within the
5795 assembler code. These names are specified inside square brackets
5796 preceding the constraint string, and can be referenced inside the
5797 assembler code using @code{%[@var{name}]} instead of a percentage sign
5798 followed by the operand number. Using named operands the above example
5802 asm ("fsinx %[angle],%[output]"
5803 : [output] "=f" (result)
5804 : [angle] "f" (angle));
5808 Note that the symbolic operand names have no relation whatsoever to
5809 other C identifiers. You may use any name you like, even those of
5810 existing C symbols, but you must ensure that no two operands within the same
5811 assembler construct use the same symbolic name.
5813 Output operand expressions must be lvalues; the compiler can check this.
5814 The input operands need not be lvalues. The compiler cannot check
5815 whether the operands have data types that are reasonable for the
5816 instruction being executed. It does not parse the assembler instruction
5817 template and does not know what it means or even whether it is valid
5818 assembler input. The extended @code{asm} feature is most often used for
5819 machine instructions the compiler itself does not know exist. If
5820 the output expression cannot be directly addressed (for example, it is a
5821 bit-field), your constraint must allow a register. In that case, GCC
5822 will use the register as the output of the @code{asm}, and then store
5823 that register into the output.
5825 The ordinary output operands must be write-only; GCC will assume that
5826 the values in these operands before the instruction are dead and need
5827 not be generated. Extended asm supports input-output or read-write
5828 operands. Use the constraint character @samp{+} to indicate such an
5829 operand and list it with the output operands. You should only use
5830 read-write operands when the constraints for the operand (or the
5831 operand in which only some of the bits are to be changed) allow a
5834 You may, as an alternative, logically split its function into two
5835 separate operands, one input operand and one write-only output
5836 operand. The connection between them is expressed by constraints
5837 which say they need to be in the same location when the instruction
5838 executes. You can use the same C expression for both operands, or
5839 different expressions. For example, here we write the (fictitious)
5840 @samp{combine} instruction with @code{bar} as its read-only source
5841 operand and @code{foo} as its read-write destination:
5844 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
5848 The constraint @samp{"0"} for operand 1 says that it must occupy the
5849 same location as operand 0. A number in constraint is allowed only in
5850 an input operand and it must refer to an output operand.
5852 Only a number in the constraint can guarantee that one operand will be in
5853 the same place as another. The mere fact that @code{foo} is the value
5854 of both operands is not enough to guarantee that they will be in the
5855 same place in the generated assembler code. The following would not
5859 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
5862 Various optimizations or reloading could cause operands 0 and 1 to be in
5863 different registers; GCC knows no reason not to do so. For example, the
5864 compiler might find a copy of the value of @code{foo} in one register and
5865 use it for operand 1, but generate the output operand 0 in a different
5866 register (copying it afterward to @code{foo}'s own address). Of course,
5867 since the register for operand 1 is not even mentioned in the assembler
5868 code, the result will not work, but GCC can't tell that.
5870 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
5871 the operand number for a matching constraint. For example:
5874 asm ("cmoveq %1,%2,%[result]"
5875 : [result] "=r"(result)
5876 : "r" (test), "r"(new), "[result]"(old));
5879 Sometimes you need to make an @code{asm} operand be a specific register,
5880 but there's no matching constraint letter for that register @emph{by
5881 itself}. To force the operand into that register, use a local variable
5882 for the operand and specify the register in the variable declaration.
5883 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
5884 register constraint letter that matches the register:
5887 register int *p1 asm ("r0") = @dots{};
5888 register int *p2 asm ("r1") = @dots{};
5889 register int *result asm ("r0");
5890 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5893 @anchor{Example of asm with clobbered asm reg}
5894 In the above example, beware that a register that is call-clobbered by
5895 the target ABI will be overwritten by any function call in the
5896 assignment, including library calls for arithmetic operators.
5897 Also a register may be clobbered when generating some operations,
5898 like variable shift, memory copy or memory move on x86.
5899 Assuming it is a call-clobbered register, this may happen to @code{r0}
5900 above by the assignment to @code{p2}. If you have to use such a
5901 register, use temporary variables for expressions between the register
5906 register int *p1 asm ("r0") = @dots{};
5907 register int *p2 asm ("r1") = t1;
5908 register int *result asm ("r0");
5909 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5912 Some instructions clobber specific hard registers. To describe this,
5913 write a third colon after the input operands, followed by the names of
5914 the clobbered hard registers (given as strings). Here is a realistic
5915 example for the VAX:
5918 asm volatile ("movc3 %0,%1,%2"
5919 : /* @r{no outputs} */
5920 : "g" (from), "g" (to), "g" (count)
5921 : "r0", "r1", "r2", "r3", "r4", "r5");
5924 You may not write a clobber description in a way that overlaps with an
5925 input or output operand. For example, you may not have an operand
5926 describing a register class with one member if you mention that register
5927 in the clobber list. Variables declared to live in specific registers
5928 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
5929 have no part mentioned in the clobber description.
5930 There is no way for you to specify that an input
5931 operand is modified without also specifying it as an output
5932 operand. Note that if all the output operands you specify are for this
5933 purpose (and hence unused), you will then also need to specify
5934 @code{volatile} for the @code{asm} construct, as described below, to
5935 prevent GCC from deleting the @code{asm} statement as unused.
5937 If you refer to a particular hardware register from the assembler code,
5938 you will probably have to list the register after the third colon to
5939 tell the compiler the register's value is modified. In some assemblers,
5940 the register names begin with @samp{%}; to produce one @samp{%} in the
5941 assembler code, you must write @samp{%%} in the input.
5943 If your assembler instruction can alter the condition code register, add
5944 @samp{cc} to the list of clobbered registers. GCC on some machines
5945 represents the condition codes as a specific hardware register;
5946 @samp{cc} serves to name this register. On other machines, the
5947 condition code is handled differently, and specifying @samp{cc} has no
5948 effect. But it is valid no matter what the machine.
5950 If your assembler instructions access memory in an unpredictable
5951 fashion, add @samp{memory} to the list of clobbered registers. This
5952 will cause GCC to not keep memory values cached in registers across the
5953 assembler instruction and not optimize stores or loads to that memory.
5954 You will also want to add the @code{volatile} keyword if the memory
5955 affected is not listed in the inputs or outputs of the @code{asm}, as
5956 the @samp{memory} clobber does not count as a side-effect of the
5957 @code{asm}. If you know how large the accessed memory is, you can add
5958 it as input or output but if this is not known, you should add
5959 @samp{memory}. As an example, if you access ten bytes of a string, you
5960 can use a memory input like:
5963 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
5966 Note that in the following example the memory input is necessary,
5967 otherwise GCC might optimize the store to @code{x} away:
5974 asm ("magic stuff accessing an 'int' pointed to by '%1'"
5975 "=&d" (r) : "a" (y), "m" (*y));
5980 You can put multiple assembler instructions together in a single
5981 @code{asm} template, separated by the characters normally used in assembly
5982 code for the system. A combination that works in most places is a newline
5983 to break the line, plus a tab character to move to the instruction field
5984 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
5985 assembler allows semicolons as a line-breaking character. Note that some
5986 assembler dialects use semicolons to start a comment.
5987 The input operands are guaranteed not to use any of the clobbered
5988 registers, and neither will the output operands' addresses, so you can
5989 read and write the clobbered registers as many times as you like. Here
5990 is an example of multiple instructions in a template; it assumes the
5991 subroutine @code{_foo} accepts arguments in registers 9 and 10:
5994 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
5996 : "g" (from), "g" (to)
6000 Unless an output operand has the @samp{&} constraint modifier, GCC
6001 may allocate it in the same register as an unrelated input operand, on
6002 the assumption the inputs are consumed before the outputs are produced.
6003 This assumption may be false if the assembler code actually consists of
6004 more than one instruction. In such a case, use @samp{&} for each output
6005 operand that may not overlap an input. @xref{Modifiers}.
6007 If you want to test the condition code produced by an assembler
6008 instruction, you must include a branch and a label in the @code{asm}
6009 construct, as follows:
6012 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
6018 This assumes your assembler supports local labels, as the GNU assembler
6019 and most Unix assemblers do.
6021 Speaking of labels, jumps from one @code{asm} to another are not
6022 supported. The compiler's optimizers do not know about these jumps, and
6023 therefore they cannot take account of them when deciding how to
6024 optimize. @xref{Extended asm with goto}.
6026 @cindex macros containing @code{asm}
6027 Usually the most convenient way to use these @code{asm} instructions is to
6028 encapsulate them in macros that look like functions. For example,
6032 (@{ double __value, __arg = (x); \
6033 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
6038 Here the variable @code{__arg} is used to make sure that the instruction
6039 operates on a proper @code{double} value, and to accept only those
6040 arguments @code{x} which can convert automatically to a @code{double}.
6042 Another way to make sure the instruction operates on the correct data
6043 type is to use a cast in the @code{asm}. This is different from using a
6044 variable @code{__arg} in that it converts more different types. For
6045 example, if the desired type were @code{int}, casting the argument to
6046 @code{int} would accept a pointer with no complaint, while assigning the
6047 argument to an @code{int} variable named @code{__arg} would warn about
6048 using a pointer unless the caller explicitly casts it.
6050 If an @code{asm} has output operands, GCC assumes for optimization
6051 purposes the instruction has no side effects except to change the output
6052 operands. This does not mean instructions with a side effect cannot be
6053 used, but you must be careful, because the compiler may eliminate them
6054 if the output operands aren't used, or move them out of loops, or
6055 replace two with one if they constitute a common subexpression. Also,
6056 if your instruction does have a side effect on a variable that otherwise
6057 appears not to change, the old value of the variable may be reused later
6058 if it happens to be found in a register.
6060 You can prevent an @code{asm} instruction from being deleted
6061 by writing the keyword @code{volatile} after
6062 the @code{asm}. For example:
6065 #define get_and_set_priority(new) \
6067 asm volatile ("get_and_set_priority %0, %1" \
6068 : "=g" (__old) : "g" (new)); \
6073 The @code{volatile} keyword indicates that the instruction has
6074 important side-effects. GCC will not delete a volatile @code{asm} if
6075 it is reachable. (The instruction can still be deleted if GCC can
6076 prove that control-flow will never reach the location of the
6077 instruction.) Note that even a volatile @code{asm} instruction
6078 can be moved relative to other code, including across jump
6079 instructions. For example, on many targets there is a system
6080 register which can be set to control the rounding mode of
6081 floating point operations. You might try
6082 setting it with a volatile @code{asm}, like this PowerPC example:
6085 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
6090 This will not work reliably, as the compiler may move the addition back
6091 before the volatile @code{asm}. To make it work you need to add an
6092 artificial dependency to the @code{asm} referencing a variable in the code
6093 you don't want moved, for example:
6096 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
6100 Similarly, you can't expect a
6101 sequence of volatile @code{asm} instructions to remain perfectly
6102 consecutive. If you want consecutive output, use a single @code{asm}.
6103 Also, GCC will perform some optimizations across a volatile @code{asm}
6104 instruction; GCC does not ``forget everything'' when it encounters
6105 a volatile @code{asm} instruction the way some other compilers do.
6107 An @code{asm} instruction without any output operands will be treated
6108 identically to a volatile @code{asm} instruction.
6110 It is a natural idea to look for a way to give access to the condition
6111 code left by the assembler instruction. However, when we attempted to
6112 implement this, we found no way to make it work reliably. The problem
6113 is that output operands might need reloading, which would result in
6114 additional following ``store'' instructions. On most machines, these
6115 instructions would alter the condition code before there was time to
6116 test it. This problem doesn't arise for ordinary ``test'' and
6117 ``compare'' instructions because they don't have any output operands.
6119 For reasons similar to those described above, it is not possible to give
6120 an assembler instruction access to the condition code left by previous
6123 @anchor{Extended asm with goto}
6124 As of GCC version 4.5, @code{asm goto} may be used to have the assembly
6125 jump to one or more C labels. In this form, a fifth section after the
6126 clobber list contains a list of all C labels to which the assembly may jump.
6127 Each label operand is implicitly self-named. The @code{asm} is also assumed
6128 to fall through to the next statement.
6130 This form of @code{asm} is restricted to not have outputs. This is due
6131 to a internal restriction in the compiler that control transfer instructions
6132 cannot have outputs. This restriction on @code{asm goto} may be lifted
6133 in some future version of the compiler. In the mean time, @code{asm goto}
6134 may include a memory clobber, and so leave outputs in memory.
6140 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
6141 : : "r"(x), "r"(&y) : "r5", "memory" : error);
6148 In this (inefficient) example, the @code{frob} instruction sets the
6149 carry bit to indicate an error. The @code{jc} instruction detects
6150 this and branches to the @code{error} label. Finally, the output
6151 of the @code{frob} instruction (@code{%r5}) is stored into the memory
6152 for variable @code{y}, which is later read by the @code{return} statement.
6158 asm goto ("mfsr %%r1, 123; jmp %%r1;"
6159 ".pushsection doit_table;"
6160 ".long %l0, %l1, %l2, %l3;"
6162 : : : "r1" : label1, label2, label3, label4);
6163 __builtin_unreachable ();
6178 In this (also inefficient) example, the @code{mfsr} instruction reads
6179 an address from some out-of-band machine register, and the following
6180 @code{jmp} instruction branches to that address. The address read by
6181 the @code{mfsr} instruction is assumed to have been previously set via
6182 some application-specific mechanism to be one of the four values stored
6183 in the @code{doit_table} section. Finally, the @code{asm} is followed
6184 by a call to @code{__builtin_unreachable} to indicate that the @code{asm}
6185 does not in fact fall through.
6188 #define TRACE1(NUM) \
6190 asm goto ("0: nop;" \
6191 ".pushsection trace_table;" \
6194 : : : : trace#NUM); \
6195 if (0) @{ trace#NUM: trace(); @} \
6197 #define TRACE TRACE1(__COUNTER__)
6200 In this example (which in fact inspired the @code{asm goto} feature)
6201 we want on rare occasions to call the @code{trace} function; on other
6202 occasions we'd like to keep the overhead to the absolute minimum.
6203 The normal code path consists of a single @code{nop} instruction.
6204 However, we record the address of this @code{nop} together with the
6205 address of a label that calls the @code{trace} function. This allows
6206 the @code{nop} instruction to be patched at runtime to be an
6207 unconditional branch to the stored label. It is assumed that an
6208 optimizing compiler will move the labeled block out of line, to
6209 optimize the fall through path from the @code{asm}.
6211 If you are writing a header file that should be includable in ISO C
6212 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
6215 @subsection Size of an @code{asm}
6217 Some targets require that GCC track the size of each instruction used in
6218 order to generate correct code. Because the final length of an
6219 @code{asm} is only known by the assembler, GCC must make an estimate as
6220 to how big it will be. The estimate is formed by counting the number of
6221 statements in the pattern of the @code{asm} and multiplying that by the
6222 length of the longest instruction on that processor. Statements in the
6223 @code{asm} are identified by newline characters and whatever statement
6224 separator characters are supported by the assembler; on most processors
6225 this is the `@code{;}' character.
6227 Normally, GCC's estimate is perfectly adequate to ensure that correct
6228 code is generated, but it is possible to confuse the compiler if you use
6229 pseudo instructions or assembler macros that expand into multiple real
6230 instructions or if you use assembler directives that expand to more
6231 space in the object file than would be needed for a single instruction.
6232 If this happens then the assembler will produce a diagnostic saying that
6233 a label is unreachable.
6235 @subsection i386 floating point asm operands
6237 There are several rules on the usage of stack-like regs in
6238 asm_operands insns. These rules apply only to the operands that are
6243 Given a set of input regs that die in an asm_operands, it is
6244 necessary to know which are implicitly popped by the asm, and
6245 which must be explicitly popped by gcc.
6247 An input reg that is implicitly popped by the asm must be
6248 explicitly clobbered, unless it is constrained to match an
6252 For any input reg that is implicitly popped by an asm, it is
6253 necessary to know how to adjust the stack to compensate for the pop.
6254 If any non-popped input is closer to the top of the reg-stack than
6255 the implicitly popped reg, it would not be possible to know what the
6256 stack looked like---it's not clear how the rest of the stack ``slides
6259 All implicitly popped input regs must be closer to the top of
6260 the reg-stack than any input that is not implicitly popped.
6262 It is possible that if an input dies in an insn, reload might
6263 use the input reg for an output reload. Consider this example:
6266 asm ("foo" : "=t" (a) : "f" (b));
6269 This asm says that input B is not popped by the asm, and that
6270 the asm pushes a result onto the reg-stack, i.e., the stack is one
6271 deeper after the asm than it was before. But, it is possible that
6272 reload will think that it can use the same reg for both the input and
6273 the output, if input B dies in this insn.
6275 If any input operand uses the @code{f} constraint, all output reg
6276 constraints must use the @code{&} earlyclobber.
6278 The asm above would be written as
6281 asm ("foo" : "=&t" (a) : "f" (b));
6285 Some operands need to be in particular places on the stack. All
6286 output operands fall in this category---there is no other way to
6287 know which regs the outputs appear in unless the user indicates
6288 this in the constraints.
6290 Output operands must specifically indicate which reg an output
6291 appears in after an asm. @code{=f} is not allowed: the operand
6292 constraints must select a class with a single reg.
6295 Output operands may not be ``inserted'' between existing stack regs.
6296 Since no 387 opcode uses a read/write operand, all output operands
6297 are dead before the asm_operands, and are pushed by the asm_operands.
6298 It makes no sense to push anywhere but the top of the reg-stack.
6300 Output operands must start at the top of the reg-stack: output
6301 operands may not ``skip'' a reg.
6304 Some asm statements may need extra stack space for internal
6305 calculations. This can be guaranteed by clobbering stack registers
6306 unrelated to the inputs and outputs.
6310 Here are a couple of reasonable asms to want to write. This asm
6311 takes one input, which is internally popped, and produces two outputs.
6314 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
6317 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
6318 and replaces them with one output. The user must code the @code{st(1)}
6319 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
6322 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
6328 @section Controlling Names Used in Assembler Code
6329 @cindex assembler names for identifiers
6330 @cindex names used in assembler code
6331 @cindex identifiers, names in assembler code
6333 You can specify the name to be used in the assembler code for a C
6334 function or variable by writing the @code{asm} (or @code{__asm__})
6335 keyword after the declarator as follows:
6338 int foo asm ("myfoo") = 2;
6342 This specifies that the name to be used for the variable @code{foo} in
6343 the assembler code should be @samp{myfoo} rather than the usual
6346 On systems where an underscore is normally prepended to the name of a C
6347 function or variable, this feature allows you to define names for the
6348 linker that do not start with an underscore.
6350 It does not make sense to use this feature with a non-static local
6351 variable since such variables do not have assembler names. If you are
6352 trying to put the variable in a particular register, see @ref{Explicit
6353 Reg Vars}. GCC presently accepts such code with a warning, but will
6354 probably be changed to issue an error, rather than a warning, in the
6357 You cannot use @code{asm} in this way in a function @emph{definition}; but
6358 you can get the same effect by writing a declaration for the function
6359 before its definition and putting @code{asm} there, like this:
6362 extern func () asm ("FUNC");
6369 It is up to you to make sure that the assembler names you choose do not
6370 conflict with any other assembler symbols. Also, you must not use a
6371 register name; that would produce completely invalid assembler code. GCC
6372 does not as yet have the ability to store static variables in registers.
6373 Perhaps that will be added.
6375 @node Explicit Reg Vars
6376 @section Variables in Specified Registers
6377 @cindex explicit register variables
6378 @cindex variables in specified registers
6379 @cindex specified registers
6380 @cindex registers, global allocation
6382 GNU C allows you to put a few global variables into specified hardware
6383 registers. You can also specify the register in which an ordinary
6384 register variable should be allocated.
6388 Global register variables reserve registers throughout the program.
6389 This may be useful in programs such as programming language
6390 interpreters which have a couple of global variables that are accessed
6394 Local register variables in specific registers do not reserve the
6395 registers, except at the point where they are used as input or output
6396 operands in an @code{asm} statement and the @code{asm} statement itself is
6397 not deleted. The compiler's data flow analysis is capable of determining
6398 where the specified registers contain live values, and where they are
6399 available for other uses. Stores into local register variables may be deleted
6400 when they appear to be dead according to dataflow analysis. References
6401 to local register variables may be deleted or moved or simplified.
6403 These local variables are sometimes convenient for use with the extended
6404 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
6405 output of the assembler instruction directly into a particular register.
6406 (This will work provided the register you specify fits the constraints
6407 specified for that operand in the @code{asm}.)
6415 @node Global Reg Vars
6416 @subsection Defining Global Register Variables
6417 @cindex global register variables
6418 @cindex registers, global variables in
6420 You can define a global register variable in GNU C like this:
6423 register int *foo asm ("a5");
6427 Here @code{a5} is the name of the register which should be used. Choose a
6428 register which is normally saved and restored by function calls on your
6429 machine, so that library routines will not clobber it.
6431 Naturally the register name is cpu-dependent, so you would need to
6432 conditionalize your program according to cpu type. The register
6433 @code{a5} would be a good choice on a 68000 for a variable of pointer
6434 type. On machines with register windows, be sure to choose a ``global''
6435 register that is not affected magically by the function call mechanism.
6437 In addition, operating systems on one type of cpu may differ in how they
6438 name the registers; then you would need additional conditionals. For
6439 example, some 68000 operating systems call this register @code{%a5}.
6441 Eventually there may be a way of asking the compiler to choose a register
6442 automatically, but first we need to figure out how it should choose and
6443 how to enable you to guide the choice. No solution is evident.
6445 Defining a global register variable in a certain register reserves that
6446 register entirely for this use, at least within the current compilation.
6447 The register will not be allocated for any other purpose in the functions
6448 in the current compilation. The register will not be saved and restored by
6449 these functions. Stores into this register are never deleted even if they
6450 would appear to be dead, but references may be deleted or moved or
6453 It is not safe to access the global register variables from signal
6454 handlers, or from more than one thread of control, because the system
6455 library routines may temporarily use the register for other things (unless
6456 you recompile them specially for the task at hand).
6458 @cindex @code{qsort}, and global register variables
6459 It is not safe for one function that uses a global register variable to
6460 call another such function @code{foo} by way of a third function
6461 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
6462 different source file in which the variable wasn't declared). This is
6463 because @code{lose} might save the register and put some other value there.
6464 For example, you can't expect a global register variable to be available in
6465 the comparison-function that you pass to @code{qsort}, since @code{qsort}
6466 might have put something else in that register. (If you are prepared to
6467 recompile @code{qsort} with the same global register variable, you can
6468 solve this problem.)
6470 If you want to recompile @code{qsort} or other source files which do not
6471 actually use your global register variable, so that they will not use that
6472 register for any other purpose, then it suffices to specify the compiler
6473 option @option{-ffixed-@var{reg}}. You need not actually add a global
6474 register declaration to their source code.
6476 A function which can alter the value of a global register variable cannot
6477 safely be called from a function compiled without this variable, because it
6478 could clobber the value the caller expects to find there on return.
6479 Therefore, the function which is the entry point into the part of the
6480 program that uses the global register variable must explicitly save and
6481 restore the value which belongs to its caller.
6483 @cindex register variable after @code{longjmp}
6484 @cindex global register after @code{longjmp}
6485 @cindex value after @code{longjmp}
6488 On most machines, @code{longjmp} will restore to each global register
6489 variable the value it had at the time of the @code{setjmp}. On some
6490 machines, however, @code{longjmp} will not change the value of global
6491 register variables. To be portable, the function that called @code{setjmp}
6492 should make other arrangements to save the values of the global register
6493 variables, and to restore them in a @code{longjmp}. This way, the same
6494 thing will happen regardless of what @code{longjmp} does.
6496 All global register variable declarations must precede all function
6497 definitions. If such a declaration could appear after function
6498 definitions, the declaration would be too late to prevent the register from
6499 being used for other purposes in the preceding functions.
6501 Global register variables may not have initial values, because an
6502 executable file has no means to supply initial contents for a register.
6504 On the SPARC, there are reports that g3 @dots{} g7 are suitable
6505 registers, but certain library functions, such as @code{getwd}, as well
6506 as the subroutines for division and remainder, modify g3 and g4. g1 and
6507 g2 are local temporaries.
6509 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
6510 Of course, it will not do to use more than a few of those.
6512 @node Local Reg Vars
6513 @subsection Specifying Registers for Local Variables
6514 @cindex local variables, specifying registers
6515 @cindex specifying registers for local variables
6516 @cindex registers for local variables
6518 You can define a local register variable with a specified register
6522 register int *foo asm ("a5");
6526 Here @code{a5} is the name of the register which should be used. Note
6527 that this is the same syntax used for defining global register
6528 variables, but for a local variable it would appear within a function.
6530 Naturally the register name is cpu-dependent, but this is not a
6531 problem, since specific registers are most often useful with explicit
6532 assembler instructions (@pxref{Extended Asm}). Both of these things
6533 generally require that you conditionalize your program according to
6536 In addition, operating systems on one type of cpu may differ in how they
6537 name the registers; then you would need additional conditionals. For
6538 example, some 68000 operating systems call this register @code{%a5}.
6540 Defining such a register variable does not reserve the register; it
6541 remains available for other uses in places where flow control determines
6542 the variable's value is not live.
6544 This option does not guarantee that GCC will generate code that has
6545 this variable in the register you specify at all times. You may not
6546 code an explicit reference to this register in the @emph{assembler
6547 instruction template} part of an @code{asm} statement and assume it will
6548 always refer to this variable. However, using the variable as an
6549 @code{asm} @emph{operand} guarantees that the specified register is used
6552 Stores into local register variables may be deleted when they appear to be dead
6553 according to dataflow analysis. References to local register variables may
6554 be deleted or moved or simplified.
6556 As for global register variables, it's recommended that you choose a
6557 register which is normally saved and restored by function calls on
6558 your machine, so that library routines will not clobber it. A common
6559 pitfall is to initialize multiple call-clobbered registers with
6560 arbitrary expressions, where a function call or library call for an
6561 arithmetic operator will overwrite a register value from a previous
6562 assignment, for example @code{r0} below:
6564 register int *p1 asm ("r0") = @dots{};
6565 register int *p2 asm ("r1") = @dots{};
6567 In those cases, a solution is to use a temporary variable for
6568 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
6570 @node Alternate Keywords
6571 @section Alternate Keywords
6572 @cindex alternate keywords
6573 @cindex keywords, alternate
6575 @option{-ansi} and the various @option{-std} options disable certain
6576 keywords. This causes trouble when you want to use GNU C extensions, or
6577 a general-purpose header file that should be usable by all programs,
6578 including ISO C programs. The keywords @code{asm}, @code{typeof} and
6579 @code{inline} are not available in programs compiled with
6580 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
6581 program compiled with @option{-std=c99} or @option{-std=c11}). The
6583 @code{restrict} is only available when @option{-std=gnu99} (which will
6584 eventually be the default) or @option{-std=c99} (or the equivalent
6585 @option{-std=iso9899:1999}), or an option for a later standard
6588 The way to solve these problems is to put @samp{__} at the beginning and
6589 end of each problematical keyword. For example, use @code{__asm__}
6590 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
6592 Other C compilers won't accept these alternative keywords; if you want to
6593 compile with another compiler, you can define the alternate keywords as
6594 macros to replace them with the customary keywords. It looks like this:
6602 @findex __extension__
6604 @option{-pedantic} and other options cause warnings for many GNU C extensions.
6606 prevent such warnings within one expression by writing
6607 @code{__extension__} before the expression. @code{__extension__} has no
6608 effect aside from this.
6610 @node Incomplete Enums
6611 @section Incomplete @code{enum} Types
6613 You can define an @code{enum} tag without specifying its possible values.
6614 This results in an incomplete type, much like what you get if you write
6615 @code{struct foo} without describing the elements. A later declaration
6616 which does specify the possible values completes the type.
6618 You can't allocate variables or storage using the type while it is
6619 incomplete. However, you can work with pointers to that type.
6621 This extension may not be very useful, but it makes the handling of
6622 @code{enum} more consistent with the way @code{struct} and @code{union}
6625 This extension is not supported by GNU C++.
6627 @node Function Names
6628 @section Function Names as Strings
6629 @cindex @code{__func__} identifier
6630 @cindex @code{__FUNCTION__} identifier
6631 @cindex @code{__PRETTY_FUNCTION__} identifier
6633 GCC provides three magic variables which hold the name of the current
6634 function, as a string. The first of these is @code{__func__}, which
6635 is part of the C99 standard:
6637 The identifier @code{__func__} is implicitly declared by the translator
6638 as if, immediately following the opening brace of each function
6639 definition, the declaration
6642 static const char __func__[] = "function-name";
6646 appeared, where function-name is the name of the lexically-enclosing
6647 function. This name is the unadorned name of the function.
6649 @code{__FUNCTION__} is another name for @code{__func__}. Older
6650 versions of GCC recognize only this name. However, it is not
6651 standardized. For maximum portability, we recommend you use
6652 @code{__func__}, but provide a fallback definition with the
6656 #if __STDC_VERSION__ < 199901L
6658 # define __func__ __FUNCTION__
6660 # define __func__ "<unknown>"
6665 In C, @code{__PRETTY_FUNCTION__} is yet another name for
6666 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
6667 the type signature of the function as well as its bare name. For
6668 example, this program:
6672 extern int printf (char *, ...);
6679 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
6680 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
6698 __PRETTY_FUNCTION__ = void a::sub(int)
6701 These identifiers are not preprocessor macros. In GCC 3.3 and
6702 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
6703 were treated as string literals; they could be used to initialize
6704 @code{char} arrays, and they could be concatenated with other string
6705 literals. GCC 3.4 and later treat them as variables, like
6706 @code{__func__}. In C++, @code{__FUNCTION__} and
6707 @code{__PRETTY_FUNCTION__} have always been variables.
6709 @node Return Address
6710 @section Getting the Return or Frame Address of a Function
6712 These functions may be used to get information about the callers of a
6715 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
6716 This function returns the return address of the current function, or of
6717 one of its callers. The @var{level} argument is number of frames to
6718 scan up the call stack. A value of @code{0} yields the return address
6719 of the current function, a value of @code{1} yields the return address
6720 of the caller of the current function, and so forth. When inlining
6721 the expected behavior is that the function will return the address of
6722 the function that will be returned to. To work around this behavior use
6723 the @code{noinline} function attribute.
6725 The @var{level} argument must be a constant integer.
6727 On some machines it may be impossible to determine the return address of
6728 any function other than the current one; in such cases, or when the top
6729 of the stack has been reached, this function will return @code{0} or a
6730 random value. In addition, @code{__builtin_frame_address} may be used
6731 to determine if the top of the stack has been reached.
6733 Additional post-processing of the returned value may be needed, see
6734 @code{__builtin_extract_return_address}.
6736 This function should only be used with a nonzero argument for debugging
6740 @deftypefn {Built-in Function} {void *} __builtin_extract_return_address (void *@var{addr})
6741 The address as returned by @code{__builtin_return_address} may have to be fed
6742 through this function to get the actual encoded address. For example, on the
6743 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
6744 platforms an offset has to be added for the true next instruction to be
6747 If no fixup is needed, this function simply passes through @var{addr}.
6750 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
6751 This function does the reverse of @code{__builtin_extract_return_address}.
6754 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
6755 This function is similar to @code{__builtin_return_address}, but it
6756 returns the address of the function frame rather than the return address
6757 of the function. Calling @code{__builtin_frame_address} with a value of
6758 @code{0} yields the frame address of the current function, a value of
6759 @code{1} yields the frame address of the caller of the current function,
6762 The frame is the area on the stack which holds local variables and saved
6763 registers. The frame address is normally the address of the first word
6764 pushed on to the stack by the function. However, the exact definition
6765 depends upon the processor and the calling convention. If the processor
6766 has a dedicated frame pointer register, and the function has a frame,
6767 then @code{__builtin_frame_address} will return the value of the frame
6770 On some machines it may be impossible to determine the frame address of
6771 any function other than the current one; in such cases, or when the top
6772 of the stack has been reached, this function will return @code{0} if
6773 the first frame pointer is properly initialized by the startup code.
6775 This function should only be used with a nonzero argument for debugging
6779 @node Vector Extensions
6780 @section Using vector instructions through built-in functions
6782 On some targets, the instruction set contains SIMD vector instructions that
6783 operate on multiple values contained in one large register at the same time.
6784 For example, on the i386 the MMX, 3DNow!@: and SSE extensions can be used
6787 The first step in using these extensions is to provide the necessary data
6788 types. This should be done using an appropriate @code{typedef}:
6791 typedef int v4si __attribute__ ((vector_size (16)));
6794 The @code{int} type specifies the base type, while the attribute specifies
6795 the vector size for the variable, measured in bytes. For example, the
6796 declaration above causes the compiler to set the mode for the @code{v4si}
6797 type to be 16 bytes wide and divided into @code{int} sized units. For
6798 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
6799 corresponding mode of @code{foo} will be @acronym{V4SI}.
6801 The @code{vector_size} attribute is only applicable to integral and
6802 float scalars, although arrays, pointers, and function return values
6803 are allowed in conjunction with this construct.
6805 All the basic integer types can be used as base types, both as signed
6806 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
6807 @code{long long}. In addition, @code{float} and @code{double} can be
6808 used to build floating-point vector types.
6810 Specifying a combination that is not valid for the current architecture
6811 will cause GCC to synthesize the instructions using a narrower mode.
6812 For example, if you specify a variable of type @code{V4SI} and your
6813 architecture does not allow for this specific SIMD type, GCC will
6814 produce code that uses 4 @code{SIs}.
6816 The types defined in this manner can be used with a subset of normal C
6817 operations. Currently, GCC will allow using the following operators
6818 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
6820 The operations behave like C++ @code{valarrays}. Addition is defined as
6821 the addition of the corresponding elements of the operands. For
6822 example, in the code below, each of the 4 elements in @var{a} will be
6823 added to the corresponding 4 elements in @var{b} and the resulting
6824 vector will be stored in @var{c}.
6827 typedef int v4si __attribute__ ((vector_size (16)));
6834 Subtraction, multiplication, division, and the logical operations
6835 operate in a similar manner. Likewise, the result of using the unary
6836 minus or complement operators on a vector type is a vector whose
6837 elements are the negative or complemented values of the corresponding
6838 elements in the operand.
6840 In C it is possible to use shifting operators @code{<<}, @code{>>} on
6841 integer-type vectors. The operation is defined as following: @code{@{a0,
6842 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
6843 @dots{}, an >> bn@}}@. Vector operands must have the same number of
6846 For the convenience in C it is allowed to use a binary vector operation
6847 where one operand is a scalar. In that case the compiler will transform
6848 the scalar operand into a vector where each element is the scalar from
6849 the operation. The transformation will happen only if the scalar could be
6850 safely converted to the vector-element type.
6851 Consider the following code.
6854 typedef int v4si __attribute__ ((vector_size (16)));
6859 a = b + 1; /* a = b + @{1,1,1,1@}; */
6860 a = 2 * b; /* a = @{2,2,2,2@} * b; */
6862 a = l + a; /* Error, cannot convert long to int. */
6865 In C vectors can be subscripted as if the vector were an array with
6866 the same number of elements and base type. Out of bound accesses
6867 invoke undefined behavior at runtime. Warnings for out of bound
6868 accesses for vector subscription can be enabled with
6869 @option{-Warray-bounds}.
6871 In GNU C vector comparison is supported within standard comparison
6872 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
6873 vector expressions of integer-type or real-type. Comparison between
6874 integer-type vectors and real-type vectors are not supported. The
6875 result of the comparison is a vector of the same width and number of
6876 elements as the comparison operands with a signed integral element
6879 Vectors are compared element-wise producing 0 when comparison is false
6880 and -1 (constant of the appropriate type where all bits are set)
6881 otherwise. Consider the following example.
6884 typedef int v4si __attribute__ ((vector_size (16)));
6886 v4si a = @{1,2,3,4@};
6887 v4si b = @{3,2,1,4@};
6890 c = a > b; /* The result would be @{0, 0,-1, 0@} */
6891 c = a == b; /* The result would be @{0,-1, 0,-1@} */
6894 Vector shuffling is available using functions
6895 @code{__builtin_shuffle (vec, mask)} and
6896 @code{__builtin_shuffle (vec0, vec1, mask)}.
6897 Both functions construct a permutation of elements from one or two
6898 vectors and return a vector of the same type as the input vector(s).
6899 The @var{mask} is an integral vector with the same width (@var{W})
6900 and element count (@var{N}) as the output vector.
6902 The elements of the input vectors are numbered in memory ordering of
6903 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
6904 elements of @var{mask} are considered modulo @var{N} in the single-operand
6905 case and modulo @math{2*@var{N}} in the two-operand case.
6907 Consider the following example,
6910 typedef int v4si __attribute__ ((vector_size (16)));
6912 v4si a = @{1,2,3,4@};
6913 v4si b = @{5,6,7,8@};
6914 v4si mask1 = @{0,1,1,3@};
6915 v4si mask2 = @{0,4,2,5@};
6918 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
6919 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
6922 Note that @code{__builtin_shuffle} is intentionally semantically
6923 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
6925 You can declare variables and use them in function calls and returns, as
6926 well as in assignments and some casts. You can specify a vector type as
6927 a return type for a function. Vector types can also be used as function
6928 arguments. It is possible to cast from one vector type to another,
6929 provided they are of the same size (in fact, you can also cast vectors
6930 to and from other datatypes of the same size).
6932 You cannot operate between vectors of different lengths or different
6933 signedness without a cast.
6937 @findex __builtin_offsetof
6939 GCC implements for both C and C++ a syntactic extension to implement
6940 the @code{offsetof} macro.
6944 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
6946 offsetof_member_designator:
6948 | offsetof_member_designator "." @code{identifier}
6949 | offsetof_member_designator "[" @code{expr} "]"
6952 This extension is sufficient such that
6955 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
6958 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
6959 may be dependent. In either case, @var{member} may consist of a single
6960 identifier, or a sequence of member accesses and array references.
6962 @node __sync Builtins
6963 @section Legacy __sync built-in functions for atomic memory access
6965 The following builtins are intended to be compatible with those described
6966 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
6967 section 7.4. As such, they depart from the normal GCC practice of using
6968 the ``__builtin_'' prefix, and further that they are overloaded such that
6969 they work on multiple types.
6971 The definition given in the Intel documentation allows only for the use of
6972 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
6973 counterparts. GCC will allow any integral scalar or pointer type that is
6974 1, 2, 4 or 8 bytes in length.
6976 Not all operations are supported by all target processors. If a particular
6977 operation cannot be implemented on the target processor, a warning will be
6978 generated and a call an external function will be generated. The external
6979 function will carry the same name as the builtin, with an additional suffix
6980 @samp{_@var{n}} where @var{n} is the size of the data type.
6982 @c ??? Should we have a mechanism to suppress this warning? This is almost
6983 @c useful for implementing the operation under the control of an external
6986 In most cases, these builtins are considered a @dfn{full barrier}. That is,
6987 no memory operand will be moved across the operation, either forward or
6988 backward. Further, instructions will be issued as necessary to prevent the
6989 processor from speculating loads across the operation and from queuing stores
6990 after the operation.
6992 All of the routines are described in the Intel documentation to take
6993 ``an optional list of variables protected by the memory barrier''. It's
6994 not clear what is meant by that; it could mean that @emph{only} the
6995 following variables are protected, or it could mean that these variables
6996 should in addition be protected. At present GCC ignores this list and
6997 protects all variables which are globally accessible. If in the future
6998 we make some use of this list, an empty list will continue to mean all
6999 globally accessible variables.
7002 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
7003 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
7004 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
7005 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
7006 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
7007 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
7008 @findex __sync_fetch_and_add
7009 @findex __sync_fetch_and_sub
7010 @findex __sync_fetch_and_or
7011 @findex __sync_fetch_and_and
7012 @findex __sync_fetch_and_xor
7013 @findex __sync_fetch_and_nand
7014 These builtins perform the operation suggested by the name, and
7015 returns the value that had previously been in memory. That is,
7018 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
7019 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
7022 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
7023 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
7025 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
7026 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
7027 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
7028 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
7029 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
7030 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
7031 @findex __sync_add_and_fetch
7032 @findex __sync_sub_and_fetch
7033 @findex __sync_or_and_fetch
7034 @findex __sync_and_and_fetch
7035 @findex __sync_xor_and_fetch
7036 @findex __sync_nand_and_fetch
7037 These builtins perform the operation suggested by the name, and
7038 return the new value. That is,
7041 @{ *ptr @var{op}= value; return *ptr; @}
7042 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
7045 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
7046 builtin as @code{*ptr = ~(*ptr & value)} instead of
7047 @code{*ptr = ~*ptr & value}.
7049 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
7050 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
7051 @findex __sync_bool_compare_and_swap
7052 @findex __sync_val_compare_and_swap
7053 These builtins perform an atomic compare and swap. That is, if the current
7054 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
7057 The ``bool'' version returns true if the comparison is successful and
7058 @var{newval} was written. The ``val'' version returns the contents
7059 of @code{*@var{ptr}} before the operation.
7061 @item __sync_synchronize (...)
7062 @findex __sync_synchronize
7063 This builtin issues a full memory barrier.
7065 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
7066 @findex __sync_lock_test_and_set
7067 This builtin, as described by Intel, is not a traditional test-and-set
7068 operation, but rather an atomic exchange operation. It writes @var{value}
7069 into @code{*@var{ptr}}, and returns the previous contents of
7072 Many targets have only minimal support for such locks, and do not support
7073 a full exchange operation. In this case, a target may support reduced
7074 functionality here by which the @emph{only} valid value to store is the
7075 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
7076 is implementation defined.
7078 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
7079 This means that references after the builtin cannot move to (or be
7080 speculated to) before the builtin, but previous memory stores may not
7081 be globally visible yet, and previous memory loads may not yet be
7084 @item void __sync_lock_release (@var{type} *ptr, ...)
7085 @findex __sync_lock_release
7086 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
7087 Normally this means writing the constant 0 to @code{*@var{ptr}}.
7089 This builtin is not a full barrier, but rather a @dfn{release barrier}.
7090 This means that all previous memory stores are globally visible, and all
7091 previous memory loads have been satisfied, but following memory reads
7092 are not prevented from being speculated to before the barrier.
7095 @node __atomic Builtins
7096 @section Built-in functions for memory model aware atomic operations
7098 The following built-in functions approximately match the requirements for
7099 C++11 memory model. Many are similar to the @samp{__sync} prefixed built-in
7100 functions, but all also have a memory model parameter. These are all
7101 identified by being prefixed with @samp{__atomic}, and most are overloaded
7102 such that they work with multiple types.
7104 GCC will allow any integral scalar or pointer type that is 1, 2, 4, or 8
7105 bytes in length. 16-byte integral types are also allowed if
7106 @samp{__int128} (@pxref{__int128}) is supported by the architecture.
7108 Target architectures are encouraged to provide their own patterns for
7109 each of these built-in functions. If no target is provided, the original
7110 non-memory model set of @samp{__sync} atomic built-in functions will be
7111 utilized, along with any required synchronization fences surrounding it in
7112 order to achieve the proper behaviour. Execution in this case is subject
7113 to the same restrictions as those built-in functions.
7115 If there is no pattern or mechanism to provide a lock free instruction
7116 sequence, a call is made to an external routine with the same parameters
7117 to be resolved at runtime.
7119 The four non-arithmetic functions (load, store, exchange, and
7120 compare_exchange) all have a generic version as well. This generic
7121 version will work on any data type. If the data type size maps to one
7122 of the integral sizes which may have lock free support, the generic
7123 version will utilize the lock free built-in function. Otherwise an
7124 external call is left to be resolved at runtime. This external call will
7125 be the same format with the addition of a @samp{size_t} parameter inserted
7126 as the first parameter indicating the size of the object being pointed to.
7127 All objects must be the same size.
7129 There are 6 different memory models which can be specified. These map
7130 to the same names in the C++11 standard. Refer there or to the
7131 @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki on
7132 atomic synchronization} for more detailed definitions. These memory
7133 models integrate both barriers to code motion as well as synchronization
7134 requirements with other threads. These are listed in approximately
7135 ascending order of strength.
7138 @item __ATOMIC_RELAXED
7139 No barriers or synchronization.
7140 @item __ATOMIC_CONSUME
7141 Data dependency only for both barrier and synchronization with another
7143 @item __ATOMIC_ACQUIRE
7144 Barrier to hoisting of code and synchronizes with release (or stronger)
7145 semantic stores from another thread.
7146 @item __ATOMIC_RELEASE
7147 Barrier to sinking of code and synchronizes with acquire (or stronger)
7148 semantic loads from another thread.
7149 @item __ATOMIC_ACQ_REL
7150 Full barrier in both directions and synchronizes with acquire loads and
7151 release stores in another thread.
7152 @item __ATOMIC_SEQ_CST
7153 Full barrier in both directions and synchronizes with acquire loads and
7154 release stores in all threads.
7157 When implementing patterns for these built-in functions , the memory model
7158 parameter can be ignored as long as the pattern implements the most
7159 restrictive @code{__ATOMIC_SEQ_CST} model. Any of the other memory models
7160 will execute correctly with this memory model but they may not execute as
7161 efficiently as they could with a more appropriate implemention of the
7162 relaxed requirements.
7164 Note that the C++11 standard allows for the memory model parameter to be
7165 determined at runtime rather than at compile time. These built-in
7166 functions will map any runtime value to @code{__ATOMIC_SEQ_CST} rather
7167 than invoke a runtime library call or inline a switch statement. This is
7168 standard compliant, safe, and the simplest approach for now.
7170 The memory model parameter is a signed int, but only the lower 8 bits are
7171 reserved for the memory model. The remainder of the signed int is reserved
7172 for future use and should be 0. Use of the predefined atomic values will
7173 ensure proper usage.
7175 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memmodel)
7176 This built-in function implements an atomic load operation. It returns the
7177 contents of @code{*@var{ptr}}.
7179 The valid memory model variants are
7180 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
7181 and @code{__ATOMIC_CONSUME}.
7185 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memmodel)
7186 This is the generic version of an atomic load. It will return the
7187 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
7191 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memmodel)
7192 This built-in function implements an atomic store operation. It writes
7193 @code{@var{val}} into @code{*@var{ptr}}.
7195 The valid memory model variants are
7196 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
7200 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memmodel)
7201 This is the generic version of an atomic store. It will store the value
7202 of @code{*@var{val}} into @code{*@var{ptr}}.
7206 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memmodel)
7207 This built-in function implements an atomic exchange operation. It writes
7208 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
7211 The valid memory model variants are
7212 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
7213 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
7217 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memmodel)
7218 This is the generic version of an atomic exchange. It will store the
7219 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
7220 of @code{*@var{ptr}} will be copied into @code{*@var{ret}}.
7224 @deftypefn {Built-in Function} bool __atomic_compare_exchange_n (@var{type} *ptr, @var{type} *expected, @var{type} desired, bool weak, int success_memmodel, int failure_memmodel)
7225 This built-in function implements an atomic compare and exchange operation.
7226 This compares the contents of @code{*@var{ptr}} with the contents of
7227 @code{*@var{expected}} and if equal, writes @var{desired} into
7228 @code{*@var{ptr}}. If they are not equal, the current contents of
7229 @code{*@var{ptr}} is written into @code{*@var{expected}}. @var{weak} is true
7230 for weak compare_exchange, and false for the strong variation. Many targets
7231 only offer the strong variation and ignore the parameter. When in doubt, use
7232 the strong variation.
7234 True is returned if @var{desired} is written into
7235 @code{*@var{ptr}} and the execution is considered to conform to the
7236 memory model specified by @var{success_memmodel}. There are no
7237 restrictions on what memory model can be used here.
7239 False is returned otherwise, and the execution is considered to conform
7240 to @var{failure_memmodel}. This memory model cannot be
7241 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
7242 stronger model than that specified by @var{success_memmodel}.
7246 @deftypefn {Built-in Function} bool __atomic_compare_exchange (@var{type} *ptr, @var{type} *expected, @var{type} *desired, bool weak, int success_memmodel, int failure_memmodel)
7247 This built-in function implements the generic version of
7248 @code{__atomic_compare_exchange}. The function is virtually identical to
7249 @code{__atomic_compare_exchange_n}, except the desired value is also a
7254 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7255 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7256 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7257 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7258 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7259 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7260 These built-in functions perform the operation suggested by the name, and
7261 return the result of the operation. That is,
7264 @{ *ptr @var{op}= val; return *ptr; @}
7267 All memory models are valid.
7271 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memmodel)
7272 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memmodel)
7273 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memmodel)
7274 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memmodel)
7275 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memmodel)
7276 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memmodel)
7277 These built-in functions perform the operation suggested by the name, and
7278 return the value that had previously been in @code{*@var{ptr}}. That is,
7281 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
7284 All memory models are valid.
7288 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memmodel)
7290 This built-in function performs an atomic test-and-set operation on
7291 the byte at @code{*@var{ptr}}. The byte is set to some implementation
7292 defined non-zero "set" value and the return value is @code{true} if and only
7293 if the previous contents were "set".
7295 All memory models are valid.
7299 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memmodel)
7301 This built-in function performs an atomic clear operation on
7302 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} will contain 0.
7304 The valid memory model variants are
7305 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
7306 @code{__ATOMIC_RELEASE}.
7310 @deftypefn {Built-in Function} void __atomic_thread_fence (int memmodel)
7312 This built-in function acts as a synchronization fence between threads
7313 based on the specified memory model.
7315 All memory orders are valid.
7319 @deftypefn {Built-in Function} void __atomic_signal_fence (int memmodel)
7321 This built-in function acts as a synchronization fence between a thread
7322 and signal handlers based in the same thread.
7324 All memory orders are valid.
7328 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
7330 This built-in function returns true if objects of @var{size} bytes will always
7331 generate lock free atomic instructions for the target architecture.
7332 @var{size} must resolve to a compile time constant and the result also resolves to compile time constant.
7334 @var{ptr} is an optional pointer to the object which may be used to determine
7335 alignment. A value of 0 indicates typical alignment should be used. The
7336 compiler may also ignore this parameter.
7339 if (_atomic_always_lock_free (sizeof (long long), 0))
7344 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
7346 This built-in function returns true if objects of @var{size} bytes will always
7347 generate lock free atomic instructions for the target architecture. If
7348 it is not known to be lock free a call is made to a runtime routine named
7349 @code{__atomic_is_lock_free}.
7351 @var{ptr} is an optional pointer to the object which may be used to determine
7352 alignment. A value of 0 indicates typical alignment should be used. The
7353 compiler may also ignore this parameter.
7356 @node Object Size Checking
7357 @section Object Size Checking Builtins
7358 @findex __builtin_object_size
7359 @findex __builtin___memcpy_chk
7360 @findex __builtin___mempcpy_chk
7361 @findex __builtin___memmove_chk
7362 @findex __builtin___memset_chk
7363 @findex __builtin___strcpy_chk
7364 @findex __builtin___stpcpy_chk
7365 @findex __builtin___strncpy_chk
7366 @findex __builtin___strcat_chk
7367 @findex __builtin___strncat_chk
7368 @findex __builtin___sprintf_chk
7369 @findex __builtin___snprintf_chk
7370 @findex __builtin___vsprintf_chk
7371 @findex __builtin___vsnprintf_chk
7372 @findex __builtin___printf_chk
7373 @findex __builtin___vprintf_chk
7374 @findex __builtin___fprintf_chk
7375 @findex __builtin___vfprintf_chk
7377 GCC implements a limited buffer overflow protection mechanism
7378 that can prevent some buffer overflow attacks.
7380 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
7381 is a built-in construct that returns a constant number of bytes from
7382 @var{ptr} to the end of the object @var{ptr} pointer points to
7383 (if known at compile time). @code{__builtin_object_size} never evaluates
7384 its arguments for side-effects. If there are any side-effects in them, it
7385 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
7386 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
7387 point to and all of them are known at compile time, the returned number
7388 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
7389 0 and minimum if nonzero. If it is not possible to determine which objects
7390 @var{ptr} points to at compile time, @code{__builtin_object_size} should
7391 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
7392 for @var{type} 2 or 3.
7394 @var{type} is an integer constant from 0 to 3. If the least significant
7395 bit is clear, objects are whole variables, if it is set, a closest
7396 surrounding subobject is considered the object a pointer points to.
7397 The second bit determines if maximum or minimum of remaining bytes
7401 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
7402 char *p = &var.buf1[1], *q = &var.b;
7404 /* Here the object p points to is var. */
7405 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
7406 /* The subobject p points to is var.buf1. */
7407 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
7408 /* The object q points to is var. */
7409 assert (__builtin_object_size (q, 0)
7410 == (char *) (&var + 1) - (char *) &var.b);
7411 /* The subobject q points to is var.b. */
7412 assert (__builtin_object_size (q, 1) == sizeof (var.b));
7416 There are built-in functions added for many common string operation
7417 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
7418 built-in is provided. This built-in has an additional last argument,
7419 which is the number of bytes remaining in object the @var{dest}
7420 argument points to or @code{(size_t) -1} if the size is not known.
7422 The built-in functions are optimized into the normal string functions
7423 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
7424 it is known at compile time that the destination object will not
7425 be overflown. If the compiler can determine at compile time the
7426 object will be always overflown, it issues a warning.
7428 The intended use can be e.g.
7432 #define bos0(dest) __builtin_object_size (dest, 0)
7433 #define memcpy(dest, src, n) \
7434 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
7438 /* It is unknown what object p points to, so this is optimized
7439 into plain memcpy - no checking is possible. */
7440 memcpy (p, "abcde", n);
7441 /* Destination is known and length too. It is known at compile
7442 time there will be no overflow. */
7443 memcpy (&buf[5], "abcde", 5);
7444 /* Destination is known, but the length is not known at compile time.
7445 This will result in __memcpy_chk call that can check for overflow
7447 memcpy (&buf[5], "abcde", n);
7448 /* Destination is known and it is known at compile time there will
7449 be overflow. There will be a warning and __memcpy_chk call that
7450 will abort the program at runtime. */
7451 memcpy (&buf[6], "abcde", 5);
7454 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
7455 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
7456 @code{strcat} and @code{strncat}.
7458 There are also checking built-in functions for formatted output functions.
7460 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
7461 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
7462 const char *fmt, ...);
7463 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
7465 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
7466 const char *fmt, va_list ap);
7469 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
7470 etc.@: functions and can contain implementation specific flags on what
7471 additional security measures the checking function might take, such as
7472 handling @code{%n} differently.
7474 The @var{os} argument is the object size @var{s} points to, like in the
7475 other built-in functions. There is a small difference in the behavior
7476 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
7477 optimized into the non-checking functions only if @var{flag} is 0, otherwise
7478 the checking function is called with @var{os} argument set to
7481 In addition to this, there are checking built-in functions
7482 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
7483 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
7484 These have just one additional argument, @var{flag}, right before
7485 format string @var{fmt}. If the compiler is able to optimize them to
7486 @code{fputc} etc.@: functions, it will, otherwise the checking function
7487 should be called and the @var{flag} argument passed to it.
7489 @node Other Builtins
7490 @section Other built-in functions provided by GCC
7491 @cindex built-in functions
7492 @findex __builtin_fpclassify
7493 @findex __builtin_isfinite
7494 @findex __builtin_isnormal
7495 @findex __builtin_isgreater
7496 @findex __builtin_isgreaterequal
7497 @findex __builtin_isinf_sign
7498 @findex __builtin_isless
7499 @findex __builtin_islessequal
7500 @findex __builtin_islessgreater
7501 @findex __builtin_isunordered
7502 @findex __builtin_powi
7503 @findex __builtin_powif
7504 @findex __builtin_powil
7662 @findex fprintf_unlocked
7664 @findex fputs_unlocked
7781 @findex printf_unlocked
7813 @findex significandf
7814 @findex significandl
7885 GCC provides a large number of built-in functions other than the ones
7886 mentioned above. Some of these are for internal use in the processing
7887 of exceptions or variable-length argument lists and will not be
7888 documented here because they may change from time to time; we do not
7889 recommend general use of these functions.
7891 The remaining functions are provided for optimization purposes.
7893 @opindex fno-builtin
7894 GCC includes built-in versions of many of the functions in the standard
7895 C library. The versions prefixed with @code{__builtin_} will always be
7896 treated as having the same meaning as the C library function even if you
7897 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
7898 Many of these functions are only optimized in certain cases; if they are
7899 not optimized in a particular case, a call to the library function will
7904 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
7905 @option{-std=c99} or @option{-std=c11}), the functions
7906 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
7907 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
7908 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
7909 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
7910 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
7911 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
7912 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
7913 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
7914 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
7915 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
7916 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
7917 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
7918 @code{signbitd64}, @code{signbitd128}, @code{significandf},
7919 @code{significandl}, @code{significand}, @code{sincosf},
7920 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
7921 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
7922 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
7923 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
7925 may be handled as built-in functions.
7926 All these functions have corresponding versions
7927 prefixed with @code{__builtin_}, which may be used even in strict C90
7930 The ISO C99 functions
7931 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
7932 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
7933 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
7934 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
7935 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
7936 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
7937 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
7938 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
7939 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
7940 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
7941 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
7942 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
7943 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
7944 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
7945 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
7946 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
7947 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
7948 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
7949 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
7950 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
7951 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
7952 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
7953 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
7954 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
7955 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
7956 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
7957 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
7958 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
7959 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
7960 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
7961 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
7962 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
7963 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
7964 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
7965 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
7966 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
7967 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
7968 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
7969 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
7970 are handled as built-in functions
7971 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7973 There are also built-in versions of the ISO C99 functions
7974 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
7975 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
7976 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
7977 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
7978 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
7979 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
7980 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
7981 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
7982 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
7983 that are recognized in any mode since ISO C90 reserves these names for
7984 the purpose to which ISO C99 puts them. All these functions have
7985 corresponding versions prefixed with @code{__builtin_}.
7987 The ISO C94 functions
7988 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
7989 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
7990 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
7992 are handled as built-in functions
7993 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7995 The ISO C90 functions
7996 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
7997 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
7998 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
7999 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
8000 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
8001 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
8002 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
8003 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
8004 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
8005 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
8006 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
8007 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
8008 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
8009 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
8010 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
8011 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
8012 are all recognized as built-in functions unless
8013 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
8014 is specified for an individual function). All of these functions have
8015 corresponding versions prefixed with @code{__builtin_}.
8017 GCC provides built-in versions of the ISO C99 floating point comparison
8018 macros that avoid raising exceptions for unordered operands. They have
8019 the same names as the standard macros ( @code{isgreater},
8020 @code{isgreaterequal}, @code{isless}, @code{islessequal},
8021 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
8022 prefixed. We intend for a library implementor to be able to simply
8023 @code{#define} each standard macro to its built-in equivalent.
8024 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
8025 @code{isinf_sign} and @code{isnormal} built-ins used with
8026 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
8027 builtins appear both with and without the @code{__builtin_} prefix.
8029 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
8031 You can use the built-in function @code{__builtin_types_compatible_p} to
8032 determine whether two types are the same.
8034 This built-in function returns 1 if the unqualified versions of the
8035 types @var{type1} and @var{type2} (which are types, not expressions) are
8036 compatible, 0 otherwise. The result of this built-in function can be
8037 used in integer constant expressions.
8039 This built-in function ignores top level qualifiers (e.g., @code{const},
8040 @code{volatile}). For example, @code{int} is equivalent to @code{const
8043 The type @code{int[]} and @code{int[5]} are compatible. On the other
8044 hand, @code{int} and @code{char *} are not compatible, even if the size
8045 of their types, on the particular architecture are the same. Also, the
8046 amount of pointer indirection is taken into account when determining
8047 similarity. Consequently, @code{short *} is not similar to
8048 @code{short **}. Furthermore, two types that are typedefed are
8049 considered compatible if their underlying types are compatible.
8051 An @code{enum} type is not considered to be compatible with another
8052 @code{enum} type even if both are compatible with the same integer
8053 type; this is what the C standard specifies.
8054 For example, @code{enum @{foo, bar@}} is not similar to
8055 @code{enum @{hot, dog@}}.
8057 You would typically use this function in code whose execution varies
8058 depending on the arguments' types. For example:
8063 typeof (x) tmp = (x); \
8064 if (__builtin_types_compatible_p (typeof (x), long double)) \
8065 tmp = foo_long_double (tmp); \
8066 else if (__builtin_types_compatible_p (typeof (x), double)) \
8067 tmp = foo_double (tmp); \
8068 else if (__builtin_types_compatible_p (typeof (x), float)) \
8069 tmp = foo_float (tmp); \
8076 @emph{Note:} This construct is only available for C@.
8080 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
8082 You can use the built-in function @code{__builtin_choose_expr} to
8083 evaluate code depending on the value of a constant expression. This
8084 built-in function returns @var{exp1} if @var{const_exp}, which is an
8085 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
8087 This built-in function is analogous to the @samp{? :} operator in C,
8088 except that the expression returned has its type unaltered by promotion
8089 rules. Also, the built-in function does not evaluate the expression
8090 that was not chosen. For example, if @var{const_exp} evaluates to true,
8091 @var{exp2} is not evaluated even if it has side-effects.
8093 This built-in function can return an lvalue if the chosen argument is an
8096 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
8097 type. Similarly, if @var{exp2} is returned, its return type is the same
8104 __builtin_choose_expr ( \
8105 __builtin_types_compatible_p (typeof (x), double), \
8107 __builtin_choose_expr ( \
8108 __builtin_types_compatible_p (typeof (x), float), \
8110 /* @r{The void expression results in a compile-time error} \
8111 @r{when assigning the result to something.} */ \
8115 @emph{Note:} This construct is only available for C@. Furthermore, the
8116 unused expression (@var{exp1} or @var{exp2} depending on the value of
8117 @var{const_exp}) may still generate syntax errors. This may change in
8122 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
8124 The built-in function @code{__builtin_complex} is provided for use in
8125 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
8126 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
8127 real binary floating-point type, and the result has the corresponding
8128 complex type with real and imaginary parts @var{real} and @var{imag}.
8129 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
8130 infinities, NaNs and negative zeros are involved.
8134 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
8135 You can use the built-in function @code{__builtin_constant_p} to
8136 determine if a value is known to be constant at compile-time and hence
8137 that GCC can perform constant-folding on expressions involving that
8138 value. The argument of the function is the value to test. The function
8139 returns the integer 1 if the argument is known to be a compile-time
8140 constant and 0 if it is not known to be a compile-time constant. A
8141 return of 0 does not indicate that the value is @emph{not} a constant,
8142 but merely that GCC cannot prove it is a constant with the specified
8143 value of the @option{-O} option.
8145 You would typically use this function in an embedded application where
8146 memory was a critical resource. If you have some complex calculation,
8147 you may want it to be folded if it involves constants, but need to call
8148 a function if it does not. For example:
8151 #define Scale_Value(X) \
8152 (__builtin_constant_p (X) \
8153 ? ((X) * SCALE + OFFSET) : Scale (X))
8156 You may use this built-in function in either a macro or an inline
8157 function. However, if you use it in an inlined function and pass an
8158 argument of the function as the argument to the built-in, GCC will
8159 never return 1 when you call the inline function with a string constant
8160 or compound literal (@pxref{Compound Literals}) and will not return 1
8161 when you pass a constant numeric value to the inline function unless you
8162 specify the @option{-O} option.
8164 You may also use @code{__builtin_constant_p} in initializers for static
8165 data. For instance, you can write
8168 static const int table[] = @{
8169 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
8175 This is an acceptable initializer even if @var{EXPRESSION} is not a
8176 constant expression, including the case where
8177 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
8178 folded to a constant but @var{EXPRESSION} contains operands that would
8179 not otherwise be permitted in a static initializer (for example,
8180 @code{0 && foo ()}). GCC must be more conservative about evaluating the
8181 built-in in this case, because it has no opportunity to perform
8184 Previous versions of GCC did not accept this built-in in data
8185 initializers. The earliest version where it is completely safe is
8189 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
8190 @opindex fprofile-arcs
8191 You may use @code{__builtin_expect} to provide the compiler with
8192 branch prediction information. In general, you should prefer to
8193 use actual profile feedback for this (@option{-fprofile-arcs}), as
8194 programmers are notoriously bad at predicting how their programs
8195 actually perform. However, there are applications in which this
8196 data is hard to collect.
8198 The return value is the value of @var{exp}, which should be an integral
8199 expression. The semantics of the built-in are that it is expected that
8200 @var{exp} == @var{c}. For example:
8203 if (__builtin_expect (x, 0))
8208 would indicate that we do not expect to call @code{foo}, since
8209 we expect @code{x} to be zero. Since you are limited to integral
8210 expressions for @var{exp}, you should use constructions such as
8213 if (__builtin_expect (ptr != NULL, 1))
8218 when testing pointer or floating-point values.
8221 @deftypefn {Built-in Function} void __builtin_trap (void)
8222 This function causes the program to exit abnormally. GCC implements
8223 this function by using a target-dependent mechanism (such as
8224 intentionally executing an illegal instruction) or by calling
8225 @code{abort}. The mechanism used may vary from release to release so
8226 you should not rely on any particular implementation.
8229 @deftypefn {Built-in Function} void __builtin_unreachable (void)
8230 If control flow reaches the point of the @code{__builtin_unreachable},
8231 the program is undefined. It is useful in situations where the
8232 compiler cannot deduce the unreachability of the code.
8234 One such case is immediately following an @code{asm} statement that
8235 will either never terminate, or one that transfers control elsewhere
8236 and never returns. In this example, without the
8237 @code{__builtin_unreachable}, GCC would issue a warning that control
8238 reaches the end of a non-void function. It would also generate code
8239 to return after the @code{asm}.
8242 int f (int c, int v)
8250 asm("jmp error_handler");
8251 __builtin_unreachable ();
8256 Because the @code{asm} statement unconditionally transfers control out
8257 of the function, control will never reach the end of the function
8258 body. The @code{__builtin_unreachable} is in fact unreachable and
8259 communicates this fact to the compiler.
8261 Another use for @code{__builtin_unreachable} is following a call a
8262 function that never returns but that is not declared
8263 @code{__attribute__((noreturn))}, as in this example:
8266 void function_that_never_returns (void);
8276 function_that_never_returns ();
8277 __builtin_unreachable ();
8284 @deftypefn {Built-in Function} void *__builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
8285 This function returns its first argument, and allows the compiler
8286 to assume that the returned pointer is at least @var{align} bytes
8287 aligned. This built-in can have either two or three arguments,
8288 if it has three, the third argument should have integer type, and
8289 if it is non-zero means misalignment offset. For example:
8292 void *x = __builtin_assume_aligned (arg, 16);
8295 means that the compiler can assume x, set to arg, is at least
8296 16 byte aligned, while:
8299 void *x = __builtin_assume_aligned (arg, 32, 8);
8302 means that the compiler can assume for x, set to arg, that
8303 (char *) x - 8 is 32 byte aligned.
8306 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
8307 This function is used to flush the processor's instruction cache for
8308 the region of memory between @var{begin} inclusive and @var{end}
8309 exclusive. Some targets require that the instruction cache be
8310 flushed, after modifying memory containing code, in order to obtain
8311 deterministic behavior.
8313 If the target does not require instruction cache flushes,
8314 @code{__builtin___clear_cache} has no effect. Otherwise either
8315 instructions are emitted in-line to clear the instruction cache or a
8316 call to the @code{__clear_cache} function in libgcc is made.
8319 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
8320 This function is used to minimize cache-miss latency by moving data into
8321 a cache before it is accessed.
8322 You can insert calls to @code{__builtin_prefetch} into code for which
8323 you know addresses of data in memory that is likely to be accessed soon.
8324 If the target supports them, data prefetch instructions will be generated.
8325 If the prefetch is done early enough before the access then the data will
8326 be in the cache by the time it is accessed.
8328 The value of @var{addr} is the address of the memory to prefetch.
8329 There are two optional arguments, @var{rw} and @var{locality}.
8330 The value of @var{rw} is a compile-time constant one or zero; one
8331 means that the prefetch is preparing for a write to the memory address
8332 and zero, the default, means that the prefetch is preparing for a read.
8333 The value @var{locality} must be a compile-time constant integer between
8334 zero and three. A value of zero means that the data has no temporal
8335 locality, so it need not be left in the cache after the access. A value
8336 of three means that the data has a high degree of temporal locality and
8337 should be left in all levels of cache possible. Values of one and two
8338 mean, respectively, a low or moderate degree of temporal locality. The
8342 for (i = 0; i < n; i++)
8345 __builtin_prefetch (&a[i+j], 1, 1);
8346 __builtin_prefetch (&b[i+j], 0, 1);
8351 Data prefetch does not generate faults if @var{addr} is invalid, but
8352 the address expression itself must be valid. For example, a prefetch
8353 of @code{p->next} will not fault if @code{p->next} is not a valid
8354 address, but evaluation will fault if @code{p} is not a valid address.
8356 If the target does not support data prefetch, the address expression
8357 is evaluated if it includes side effects but no other code is generated
8358 and GCC does not issue a warning.
8361 @deftypefn {Built-in Function} double __builtin_huge_val (void)
8362 Returns a positive infinity, if supported by the floating-point format,
8363 else @code{DBL_MAX}. This function is suitable for implementing the
8364 ISO C macro @code{HUGE_VAL}.
8367 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
8368 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
8371 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
8372 Similar to @code{__builtin_huge_val}, except the return
8373 type is @code{long double}.
8376 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
8377 This built-in implements the C99 fpclassify functionality. The first
8378 five int arguments should be the target library's notion of the
8379 possible FP classes and are used for return values. They must be
8380 constant values and they must appear in this order: @code{FP_NAN},
8381 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
8382 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
8383 to classify. GCC treats the last argument as type-generic, which
8384 means it does not do default promotion from float to double.
8387 @deftypefn {Built-in Function} double __builtin_inf (void)
8388 Similar to @code{__builtin_huge_val}, except a warning is generated
8389 if the target floating-point format does not support infinities.
8392 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
8393 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
8396 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
8397 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
8400 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
8401 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
8404 @deftypefn {Built-in Function} float __builtin_inff (void)
8405 Similar to @code{__builtin_inf}, except the return type is @code{float}.
8406 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
8409 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
8410 Similar to @code{__builtin_inf}, except the return
8411 type is @code{long double}.
8414 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
8415 Similar to @code{isinf}, except the return value will be negative for
8416 an argument of @code{-Inf}. Note while the parameter list is an
8417 ellipsis, this function only accepts exactly one floating point
8418 argument. GCC treats this parameter as type-generic, which means it
8419 does not do default promotion from float to double.
8422 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
8423 This is an implementation of the ISO C99 function @code{nan}.
8425 Since ISO C99 defines this function in terms of @code{strtod}, which we
8426 do not implement, a description of the parsing is in order. The string
8427 is parsed as by @code{strtol}; that is, the base is recognized by
8428 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
8429 in the significand such that the least significant bit of the number
8430 is at the least significant bit of the significand. The number is
8431 truncated to fit the significand field provided. The significand is
8432 forced to be a quiet NaN@.
8434 This function, if given a string literal all of which would have been
8435 consumed by strtol, is evaluated early enough that it is considered a
8436 compile-time constant.
8439 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
8440 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
8443 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
8444 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
8447 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
8448 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
8451 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
8452 Similar to @code{__builtin_nan}, except the return type is @code{float}.
8455 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
8456 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
8459 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
8460 Similar to @code{__builtin_nan}, except the significand is forced
8461 to be a signaling NaN@. The @code{nans} function is proposed by
8462 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
8465 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
8466 Similar to @code{__builtin_nans}, except the return type is @code{float}.
8469 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
8470 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
8473 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
8474 Returns one plus the index of the least significant 1-bit of @var{x}, or
8475 if @var{x} is zero, returns zero.
8478 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
8479 Returns the number of leading 0-bits in @var{x}, starting at the most
8480 significant bit position. If @var{x} is 0, the result is undefined.
8483 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
8484 Returns the number of trailing 0-bits in @var{x}, starting at the least
8485 significant bit position. If @var{x} is 0, the result is undefined.
8488 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
8489 Returns the number of leading redundant sign bits in @var{x}, i.e. the
8490 number of bits following the most significant bit which are identical
8491 to it. There are no special cases for 0 or other values.
8494 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
8495 Returns the number of 1-bits in @var{x}.
8498 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
8499 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
8503 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
8504 Similar to @code{__builtin_ffs}, except the argument type is
8505 @code{unsigned long}.
8508 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
8509 Similar to @code{__builtin_clz}, except the argument type is
8510 @code{unsigned long}.
8513 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
8514 Similar to @code{__builtin_ctz}, except the argument type is
8515 @code{unsigned long}.
8518 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
8519 Similar to @code{__builtin_clrsb}, except the argument type is
8523 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
8524 Similar to @code{__builtin_popcount}, except the argument type is
8525 @code{unsigned long}.
8528 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
8529 Similar to @code{__builtin_parity}, except the argument type is
8530 @code{unsigned long}.
8533 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
8534 Similar to @code{__builtin_ffs}, except the argument type is
8535 @code{unsigned long long}.
8538 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
8539 Similar to @code{__builtin_clz}, except the argument type is
8540 @code{unsigned long long}.
8543 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
8544 Similar to @code{__builtin_ctz}, except the argument type is
8545 @code{unsigned long long}.
8548 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
8549 Similar to @code{__builtin_clrsb}, except the argument type is
8553 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
8554 Similar to @code{__builtin_popcount}, except the argument type is
8555 @code{unsigned long long}.
8558 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
8559 Similar to @code{__builtin_parity}, except the argument type is
8560 @code{unsigned long long}.
8563 @deftypefn {Built-in Function} double __builtin_powi (double, int)
8564 Returns the first argument raised to the power of the second. Unlike the
8565 @code{pow} function no guarantees about precision and rounding are made.
8568 @deftypefn {Built-in Function} float __builtin_powif (float, int)
8569 Similar to @code{__builtin_powi}, except the argument and return types
8573 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
8574 Similar to @code{__builtin_powi}, except the argument and return types
8575 are @code{long double}.
8578 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
8579 Returns @var{x} with the order of the bytes reversed; for example,
8580 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
8584 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
8585 Similar to @code{__builtin_bswap32}, except the argument and return types
8589 @node Target Builtins
8590 @section Built-in Functions Specific to Particular Target Machines
8592 On some target machines, GCC supports many built-in functions specific
8593 to those machines. Generally these generate calls to specific machine
8594 instructions, but allow the compiler to schedule those calls.
8597 * Alpha Built-in Functions::
8598 * ARM iWMMXt Built-in Functions::
8599 * ARM NEON Intrinsics::
8600 * AVR Built-in Functions::
8601 * Blackfin Built-in Functions::
8602 * FR-V Built-in Functions::
8603 * X86 Built-in Functions::
8604 * MIPS DSP Built-in Functions::
8605 * MIPS Paired-Single Support::
8606 * MIPS Loongson Built-in Functions::
8607 * Other MIPS Built-in Functions::
8608 * picoChip Built-in Functions::
8609 * PowerPC AltiVec/VSX Built-in Functions::
8610 * RX Built-in Functions::
8611 * SPARC VIS Built-in Functions::
8612 * SPU Built-in Functions::
8613 * TI C6X Built-in Functions::
8614 * TILE-Gx Built-in Functions::
8615 * TILEPro Built-in Functions::
8618 @node Alpha Built-in Functions
8619 @subsection Alpha Built-in Functions
8621 These built-in functions are available for the Alpha family of
8622 processors, depending on the command-line switches used.
8624 The following built-in functions are always available. They
8625 all generate the machine instruction that is part of the name.
8628 long __builtin_alpha_implver (void)
8629 long __builtin_alpha_rpcc (void)
8630 long __builtin_alpha_amask (long)
8631 long __builtin_alpha_cmpbge (long, long)
8632 long __builtin_alpha_extbl (long, long)
8633 long __builtin_alpha_extwl (long, long)
8634 long __builtin_alpha_extll (long, long)
8635 long __builtin_alpha_extql (long, long)
8636 long __builtin_alpha_extwh (long, long)
8637 long __builtin_alpha_extlh (long, long)
8638 long __builtin_alpha_extqh (long, long)
8639 long __builtin_alpha_insbl (long, long)
8640 long __builtin_alpha_inswl (long, long)
8641 long __builtin_alpha_insll (long, long)
8642 long __builtin_alpha_insql (long, long)
8643 long __builtin_alpha_inswh (long, long)
8644 long __builtin_alpha_inslh (long, long)
8645 long __builtin_alpha_insqh (long, long)
8646 long __builtin_alpha_mskbl (long, long)
8647 long __builtin_alpha_mskwl (long, long)
8648 long __builtin_alpha_mskll (long, long)
8649 long __builtin_alpha_mskql (long, long)
8650 long __builtin_alpha_mskwh (long, long)
8651 long __builtin_alpha_msklh (long, long)
8652 long __builtin_alpha_mskqh (long, long)
8653 long __builtin_alpha_umulh (long, long)
8654 long __builtin_alpha_zap (long, long)
8655 long __builtin_alpha_zapnot (long, long)
8658 The following built-in functions are always with @option{-mmax}
8659 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
8660 later. They all generate the machine instruction that is part
8664 long __builtin_alpha_pklb (long)
8665 long __builtin_alpha_pkwb (long)
8666 long __builtin_alpha_unpkbl (long)
8667 long __builtin_alpha_unpkbw (long)
8668 long __builtin_alpha_minub8 (long, long)
8669 long __builtin_alpha_minsb8 (long, long)
8670 long __builtin_alpha_minuw4 (long, long)
8671 long __builtin_alpha_minsw4 (long, long)
8672 long __builtin_alpha_maxub8 (long, long)
8673 long __builtin_alpha_maxsb8 (long, long)
8674 long __builtin_alpha_maxuw4 (long, long)
8675 long __builtin_alpha_maxsw4 (long, long)
8676 long __builtin_alpha_perr (long, long)
8679 The following built-in functions are always with @option{-mcix}
8680 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
8681 later. They all generate the machine instruction that is part
8685 long __builtin_alpha_cttz (long)
8686 long __builtin_alpha_ctlz (long)
8687 long __builtin_alpha_ctpop (long)
8690 The following builtins are available on systems that use the OSF/1
8691 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
8692 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
8693 @code{rdval} and @code{wrval}.
8696 void *__builtin_thread_pointer (void)
8697 void __builtin_set_thread_pointer (void *)
8700 @node ARM iWMMXt Built-in Functions
8701 @subsection ARM iWMMXt Built-in Functions
8703 These built-in functions are available for the ARM family of
8704 processors when the @option{-mcpu=iwmmxt} switch is used:
8707 typedef int v2si __attribute__ ((vector_size (8)));
8708 typedef short v4hi __attribute__ ((vector_size (8)));
8709 typedef char v8qi __attribute__ ((vector_size (8)));
8711 int __builtin_arm_getwcx (int)
8712 void __builtin_arm_setwcx (int, int)
8713 int __builtin_arm_textrmsb (v8qi, int)
8714 int __builtin_arm_textrmsh (v4hi, int)
8715 int __builtin_arm_textrmsw (v2si, int)
8716 int __builtin_arm_textrmub (v8qi, int)
8717 int __builtin_arm_textrmuh (v4hi, int)
8718 int __builtin_arm_textrmuw (v2si, int)
8719 v8qi __builtin_arm_tinsrb (v8qi, int)
8720 v4hi __builtin_arm_tinsrh (v4hi, int)
8721 v2si __builtin_arm_tinsrw (v2si, int)
8722 long long __builtin_arm_tmia (long long, int, int)
8723 long long __builtin_arm_tmiabb (long long, int, int)
8724 long long __builtin_arm_tmiabt (long long, int, int)
8725 long long __builtin_arm_tmiaph (long long, int, int)
8726 long long __builtin_arm_tmiatb (long long, int, int)
8727 long long __builtin_arm_tmiatt (long long, int, int)
8728 int __builtin_arm_tmovmskb (v8qi)
8729 int __builtin_arm_tmovmskh (v4hi)
8730 int __builtin_arm_tmovmskw (v2si)
8731 long long __builtin_arm_waccb (v8qi)
8732 long long __builtin_arm_wacch (v4hi)
8733 long long __builtin_arm_waccw (v2si)
8734 v8qi __builtin_arm_waddb (v8qi, v8qi)
8735 v8qi __builtin_arm_waddbss (v8qi, v8qi)
8736 v8qi __builtin_arm_waddbus (v8qi, v8qi)
8737 v4hi __builtin_arm_waddh (v4hi, v4hi)
8738 v4hi __builtin_arm_waddhss (v4hi, v4hi)
8739 v4hi __builtin_arm_waddhus (v4hi, v4hi)
8740 v2si __builtin_arm_waddw (v2si, v2si)
8741 v2si __builtin_arm_waddwss (v2si, v2si)
8742 v2si __builtin_arm_waddwus (v2si, v2si)
8743 v8qi __builtin_arm_walign (v8qi, v8qi, int)
8744 long long __builtin_arm_wand(long long, long long)
8745 long long __builtin_arm_wandn (long long, long long)
8746 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
8747 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
8748 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
8749 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
8750 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
8751 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
8752 v2si __builtin_arm_wcmpeqw (v2si, v2si)
8753 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
8754 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
8755 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
8756 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
8757 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
8758 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
8759 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
8760 long long __builtin_arm_wmacsz (v4hi, v4hi)
8761 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
8762 long long __builtin_arm_wmacuz (v4hi, v4hi)
8763 v4hi __builtin_arm_wmadds (v4hi, v4hi)
8764 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
8765 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
8766 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
8767 v2si __builtin_arm_wmaxsw (v2si, v2si)
8768 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
8769 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
8770 v2si __builtin_arm_wmaxuw (v2si, v2si)
8771 v8qi __builtin_arm_wminsb (v8qi, v8qi)
8772 v4hi __builtin_arm_wminsh (v4hi, v4hi)
8773 v2si __builtin_arm_wminsw (v2si, v2si)
8774 v8qi __builtin_arm_wminub (v8qi, v8qi)
8775 v4hi __builtin_arm_wminuh (v4hi, v4hi)
8776 v2si __builtin_arm_wminuw (v2si, v2si)
8777 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
8778 v4hi __builtin_arm_wmulul (v4hi, v4hi)
8779 v4hi __builtin_arm_wmulum (v4hi, v4hi)
8780 long long __builtin_arm_wor (long long, long long)
8781 v2si __builtin_arm_wpackdss (long long, long long)
8782 v2si __builtin_arm_wpackdus (long long, long long)
8783 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
8784 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
8785 v4hi __builtin_arm_wpackwss (v2si, v2si)
8786 v4hi __builtin_arm_wpackwus (v2si, v2si)
8787 long long __builtin_arm_wrord (long long, long long)
8788 long long __builtin_arm_wrordi (long long, int)
8789 v4hi __builtin_arm_wrorh (v4hi, long long)
8790 v4hi __builtin_arm_wrorhi (v4hi, int)
8791 v2si __builtin_arm_wrorw (v2si, long long)
8792 v2si __builtin_arm_wrorwi (v2si, int)
8793 v2si __builtin_arm_wsadb (v8qi, v8qi)
8794 v2si __builtin_arm_wsadbz (v8qi, v8qi)
8795 v2si __builtin_arm_wsadh (v4hi, v4hi)
8796 v2si __builtin_arm_wsadhz (v4hi, v4hi)
8797 v4hi __builtin_arm_wshufh (v4hi, int)
8798 long long __builtin_arm_wslld (long long, long long)
8799 long long __builtin_arm_wslldi (long long, int)
8800 v4hi __builtin_arm_wsllh (v4hi, long long)
8801 v4hi __builtin_arm_wsllhi (v4hi, int)
8802 v2si __builtin_arm_wsllw (v2si, long long)
8803 v2si __builtin_arm_wsllwi (v2si, int)
8804 long long __builtin_arm_wsrad (long long, long long)
8805 long long __builtin_arm_wsradi (long long, int)
8806 v4hi __builtin_arm_wsrah (v4hi, long long)
8807 v4hi __builtin_arm_wsrahi (v4hi, int)
8808 v2si __builtin_arm_wsraw (v2si, long long)
8809 v2si __builtin_arm_wsrawi (v2si, int)
8810 long long __builtin_arm_wsrld (long long, long long)
8811 long long __builtin_arm_wsrldi (long long, int)
8812 v4hi __builtin_arm_wsrlh (v4hi, long long)
8813 v4hi __builtin_arm_wsrlhi (v4hi, int)
8814 v2si __builtin_arm_wsrlw (v2si, long long)
8815 v2si __builtin_arm_wsrlwi (v2si, int)
8816 v8qi __builtin_arm_wsubb (v8qi, v8qi)
8817 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
8818 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
8819 v4hi __builtin_arm_wsubh (v4hi, v4hi)
8820 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
8821 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
8822 v2si __builtin_arm_wsubw (v2si, v2si)
8823 v2si __builtin_arm_wsubwss (v2si, v2si)
8824 v2si __builtin_arm_wsubwus (v2si, v2si)
8825 v4hi __builtin_arm_wunpckehsb (v8qi)
8826 v2si __builtin_arm_wunpckehsh (v4hi)
8827 long long __builtin_arm_wunpckehsw (v2si)
8828 v4hi __builtin_arm_wunpckehub (v8qi)
8829 v2si __builtin_arm_wunpckehuh (v4hi)
8830 long long __builtin_arm_wunpckehuw (v2si)
8831 v4hi __builtin_arm_wunpckelsb (v8qi)
8832 v2si __builtin_arm_wunpckelsh (v4hi)
8833 long long __builtin_arm_wunpckelsw (v2si)
8834 v4hi __builtin_arm_wunpckelub (v8qi)
8835 v2si __builtin_arm_wunpckeluh (v4hi)
8836 long long __builtin_arm_wunpckeluw (v2si)
8837 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
8838 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
8839 v2si __builtin_arm_wunpckihw (v2si, v2si)
8840 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
8841 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
8842 v2si __builtin_arm_wunpckilw (v2si, v2si)
8843 long long __builtin_arm_wxor (long long, long long)
8844 long long __builtin_arm_wzero ()
8847 @node ARM NEON Intrinsics
8848 @subsection ARM NEON Intrinsics
8850 These built-in intrinsics for the ARM Advanced SIMD extension are available
8851 when the @option{-mfpu=neon} switch is used:
8853 @include arm-neon-intrinsics.texi
8855 @node AVR Built-in Functions
8856 @subsection AVR Built-in Functions
8858 For each built-in function for AVR, there is an equally named,
8859 uppercase built-in macro defined. That way users can easily query if
8860 or if not a specific built-in is implemented or not. For example, if
8861 @code{__builtin_avr_nop} is available the macro
8862 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
8864 The following built-in functions map to the respective machine
8865 instruction, i.e. @code{nop}, @code{sei}, @code{cli}, @code{sleep},
8866 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
8867 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
8868 as library call if no hardware multiplier is available.
8871 void __builtin_avr_nop (void)
8872 void __builtin_avr_sei (void)
8873 void __builtin_avr_cli (void)
8874 void __builtin_avr_sleep (void)
8875 void __builtin_avr_wdr (void)
8876 unsigned char __builtin_avr_swap (unsigned char)
8877 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
8878 int __builtin_avr_fmuls (char, char)
8879 int __builtin_avr_fmulsu (char, unsigned char)
8882 In order to delay execution for a specific number of cycles, GCC
8885 void __builtin_avr_delay_cycles (unsigned long ticks)
8889 @code{ticks} is the number of ticks to delay execution. Note that this
8890 built-in does not take into account the effect of interrupts which
8891 might increase delay time. @code{ticks} must be a compile time
8892 integer constant; delays with a variable number of cycles are not supported.
8895 char __builtin_avr_flash_segment (const __memx void*)
8899 This built-in takes a byte address to the 24-bit
8900 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
8901 the number of the flash segment (the 64 KiB chunk) where the address
8902 points to. Counting starts at @code{0}.
8903 If the address does not point to flash memory, return @code{-1}.
8906 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
8910 Insert bits from @var{bits} into @var{val} and return the resulting
8911 value. The nibbles of @var{map} determine how the insertion is
8912 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
8914 @item If @var{X} is @code{0xf},
8915 then the @var{n}-th bit of @var{val} is returned unaltered.
8917 @item If X is in the range 0@dots{}7,
8918 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
8920 @item If X is in the range 8@dots{}@code{0xe},
8921 then the @var{n}-th result bit is undefined.
8925 One typical use case for this built-in is adjusting input and
8926 output values to non-contiguous port layouts. Some examples:
8929 // same as val, bits is unused
8930 __builtin_avr_insert_bits (0xffffffff, bits, val)
8934 // same as bits, val is unused
8935 __builtin_avr_insert_bits (0x76543210, bits, val)
8939 // same as rotating bits by 4
8940 __builtin_avr_insert_bits (0x32107654, bits, 0)
8944 // high-nibble of result is the high-nibble of val
8945 // low-nibble of result is the low-nibble of bits
8946 __builtin_avr_insert_bits (0xffff3210, bits, val)
8950 // reverse the bit order of bits
8951 __builtin_avr_insert_bits (0x01234567, bits, 0)
8954 @node Blackfin Built-in Functions
8955 @subsection Blackfin Built-in Functions
8957 Currently, there are two Blackfin-specific built-in functions. These are
8958 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
8959 using inline assembly; by using these built-in functions the compiler can
8960 automatically add workarounds for hardware errata involving these
8961 instructions. These functions are named as follows:
8964 void __builtin_bfin_csync (void)
8965 void __builtin_bfin_ssync (void)
8968 @node FR-V Built-in Functions
8969 @subsection FR-V Built-in Functions
8971 GCC provides many FR-V-specific built-in functions. In general,
8972 these functions are intended to be compatible with those described
8973 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
8974 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
8975 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
8976 pointer rather than by value.
8978 Most of the functions are named after specific FR-V instructions.
8979 Such functions are said to be ``directly mapped'' and are summarized
8980 here in tabular form.
8984 * Directly-mapped Integer Functions::
8985 * Directly-mapped Media Functions::
8986 * Raw read/write Functions::
8987 * Other Built-in Functions::
8990 @node Argument Types
8991 @subsubsection Argument Types
8993 The arguments to the built-in functions can be divided into three groups:
8994 register numbers, compile-time constants and run-time values. In order
8995 to make this classification clear at a glance, the arguments and return
8996 values are given the following pseudo types:
8998 @multitable @columnfractions .20 .30 .15 .35
8999 @item Pseudo type @tab Real C type @tab Constant? @tab Description
9000 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
9001 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
9002 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
9003 @item @code{uw2} @tab @code{unsigned long long} @tab No
9004 @tab an unsigned doubleword
9005 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
9006 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
9007 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
9008 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
9011 These pseudo types are not defined by GCC, they are simply a notational
9012 convenience used in this manual.
9014 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
9015 and @code{sw2} are evaluated at run time. They correspond to
9016 register operands in the underlying FR-V instructions.
9018 @code{const} arguments represent immediate operands in the underlying
9019 FR-V instructions. They must be compile-time constants.
9021 @code{acc} arguments are evaluated at compile time and specify the number
9022 of an accumulator register. For example, an @code{acc} argument of 2
9023 will select the ACC2 register.
9025 @code{iacc} arguments are similar to @code{acc} arguments but specify the
9026 number of an IACC register. See @pxref{Other Built-in Functions}
9029 @node Directly-mapped Integer Functions
9030 @subsubsection Directly-mapped Integer Functions
9032 The functions listed below map directly to FR-V I-type instructions.
9034 @multitable @columnfractions .45 .32 .23
9035 @item Function prototype @tab Example usage @tab Assembly output
9036 @item @code{sw1 __ADDSS (sw1, sw1)}
9037 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
9038 @tab @code{ADDSS @var{a},@var{b},@var{c}}
9039 @item @code{sw1 __SCAN (sw1, sw1)}
9040 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
9041 @tab @code{SCAN @var{a},@var{b},@var{c}}
9042 @item @code{sw1 __SCUTSS (sw1)}
9043 @tab @code{@var{b} = __SCUTSS (@var{a})}
9044 @tab @code{SCUTSS @var{a},@var{b}}
9045 @item @code{sw1 __SLASS (sw1, sw1)}
9046 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
9047 @tab @code{SLASS @var{a},@var{b},@var{c}}
9048 @item @code{void __SMASS (sw1, sw1)}
9049 @tab @code{__SMASS (@var{a}, @var{b})}
9050 @tab @code{SMASS @var{a},@var{b}}
9051 @item @code{void __SMSSS (sw1, sw1)}
9052 @tab @code{__SMSSS (@var{a}, @var{b})}
9053 @tab @code{SMSSS @var{a},@var{b}}
9054 @item @code{void __SMU (sw1, sw1)}
9055 @tab @code{__SMU (@var{a}, @var{b})}
9056 @tab @code{SMU @var{a},@var{b}}
9057 @item @code{sw2 __SMUL (sw1, sw1)}
9058 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
9059 @tab @code{SMUL @var{a},@var{b},@var{c}}
9060 @item @code{sw1 __SUBSS (sw1, sw1)}
9061 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
9062 @tab @code{SUBSS @var{a},@var{b},@var{c}}
9063 @item @code{uw2 __UMUL (uw1, uw1)}
9064 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
9065 @tab @code{UMUL @var{a},@var{b},@var{c}}
9068 @node Directly-mapped Media Functions
9069 @subsubsection Directly-mapped Media Functions
9071 The functions listed below map directly to FR-V M-type instructions.
9073 @multitable @columnfractions .45 .32 .23
9074 @item Function prototype @tab Example usage @tab Assembly output
9075 @item @code{uw1 __MABSHS (sw1)}
9076 @tab @code{@var{b} = __MABSHS (@var{a})}
9077 @tab @code{MABSHS @var{a},@var{b}}
9078 @item @code{void __MADDACCS (acc, acc)}
9079 @tab @code{__MADDACCS (@var{b}, @var{a})}
9080 @tab @code{MADDACCS @var{a},@var{b}}
9081 @item @code{sw1 __MADDHSS (sw1, sw1)}
9082 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
9083 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
9084 @item @code{uw1 __MADDHUS (uw1, uw1)}
9085 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
9086 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
9087 @item @code{uw1 __MAND (uw1, uw1)}
9088 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
9089 @tab @code{MAND @var{a},@var{b},@var{c}}
9090 @item @code{void __MASACCS (acc, acc)}
9091 @tab @code{__MASACCS (@var{b}, @var{a})}
9092 @tab @code{MASACCS @var{a},@var{b}}
9093 @item @code{uw1 __MAVEH (uw1, uw1)}
9094 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
9095 @tab @code{MAVEH @var{a},@var{b},@var{c}}
9096 @item @code{uw2 __MBTOH (uw1)}
9097 @tab @code{@var{b} = __MBTOH (@var{a})}
9098 @tab @code{MBTOH @var{a},@var{b}}
9099 @item @code{void __MBTOHE (uw1 *, uw1)}
9100 @tab @code{__MBTOHE (&@var{b}, @var{a})}
9101 @tab @code{MBTOHE @var{a},@var{b}}
9102 @item @code{void __MCLRACC (acc)}
9103 @tab @code{__MCLRACC (@var{a})}
9104 @tab @code{MCLRACC @var{a}}
9105 @item @code{void __MCLRACCA (void)}
9106 @tab @code{__MCLRACCA ()}
9107 @tab @code{MCLRACCA}
9108 @item @code{uw1 __Mcop1 (uw1, uw1)}
9109 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
9110 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
9111 @item @code{uw1 __Mcop2 (uw1, uw1)}
9112 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
9113 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
9114 @item @code{uw1 __MCPLHI (uw2, const)}
9115 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
9116 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
9117 @item @code{uw1 __MCPLI (uw2, const)}
9118 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
9119 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
9120 @item @code{void __MCPXIS (acc, sw1, sw1)}
9121 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
9122 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
9123 @item @code{void __MCPXIU (acc, uw1, uw1)}
9124 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
9125 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
9126 @item @code{void __MCPXRS (acc, sw1, sw1)}
9127 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
9128 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
9129 @item @code{void __MCPXRU (acc, uw1, uw1)}
9130 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
9131 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
9132 @item @code{uw1 __MCUT (acc, uw1)}
9133 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
9134 @tab @code{MCUT @var{a},@var{b},@var{c}}
9135 @item @code{uw1 __MCUTSS (acc, sw1)}
9136 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
9137 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
9138 @item @code{void __MDADDACCS (acc, acc)}
9139 @tab @code{__MDADDACCS (@var{b}, @var{a})}
9140 @tab @code{MDADDACCS @var{a},@var{b}}
9141 @item @code{void __MDASACCS (acc, acc)}
9142 @tab @code{__MDASACCS (@var{b}, @var{a})}
9143 @tab @code{MDASACCS @var{a},@var{b}}
9144 @item @code{uw2 __MDCUTSSI (acc, const)}
9145 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
9146 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
9147 @item @code{uw2 __MDPACKH (uw2, uw2)}
9148 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
9149 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
9150 @item @code{uw2 __MDROTLI (uw2, const)}
9151 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
9152 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
9153 @item @code{void __MDSUBACCS (acc, acc)}
9154 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
9155 @tab @code{MDSUBACCS @var{a},@var{b}}
9156 @item @code{void __MDUNPACKH (uw1 *, uw2)}
9157 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
9158 @tab @code{MDUNPACKH @var{a},@var{b}}
9159 @item @code{uw2 __MEXPDHD (uw1, const)}
9160 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
9161 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
9162 @item @code{uw1 __MEXPDHW (uw1, const)}
9163 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
9164 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
9165 @item @code{uw1 __MHDSETH (uw1, const)}
9166 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
9167 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
9168 @item @code{sw1 __MHDSETS (const)}
9169 @tab @code{@var{b} = __MHDSETS (@var{a})}
9170 @tab @code{MHDSETS #@var{a},@var{b}}
9171 @item @code{uw1 __MHSETHIH (uw1, const)}
9172 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
9173 @tab @code{MHSETHIH #@var{a},@var{b}}
9174 @item @code{sw1 __MHSETHIS (sw1, const)}
9175 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
9176 @tab @code{MHSETHIS #@var{a},@var{b}}
9177 @item @code{uw1 __MHSETLOH (uw1, const)}
9178 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
9179 @tab @code{MHSETLOH #@var{a},@var{b}}
9180 @item @code{sw1 __MHSETLOS (sw1, const)}
9181 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
9182 @tab @code{MHSETLOS #@var{a},@var{b}}
9183 @item @code{uw1 __MHTOB (uw2)}
9184 @tab @code{@var{b} = __MHTOB (@var{a})}
9185 @tab @code{MHTOB @var{a},@var{b}}
9186 @item @code{void __MMACHS (acc, sw1, sw1)}
9187 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
9188 @tab @code{MMACHS @var{a},@var{b},@var{c}}
9189 @item @code{void __MMACHU (acc, uw1, uw1)}
9190 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
9191 @tab @code{MMACHU @var{a},@var{b},@var{c}}
9192 @item @code{void __MMRDHS (acc, sw1, sw1)}
9193 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
9194 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
9195 @item @code{void __MMRDHU (acc, uw1, uw1)}
9196 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
9197 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
9198 @item @code{void __MMULHS (acc, sw1, sw1)}
9199 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
9200 @tab @code{MMULHS @var{a},@var{b},@var{c}}
9201 @item @code{void __MMULHU (acc, uw1, uw1)}
9202 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
9203 @tab @code{MMULHU @var{a},@var{b},@var{c}}
9204 @item @code{void __MMULXHS (acc, sw1, sw1)}
9205 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
9206 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
9207 @item @code{void __MMULXHU (acc, uw1, uw1)}
9208 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
9209 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
9210 @item @code{uw1 __MNOT (uw1)}
9211 @tab @code{@var{b} = __MNOT (@var{a})}
9212 @tab @code{MNOT @var{a},@var{b}}
9213 @item @code{uw1 __MOR (uw1, uw1)}
9214 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
9215 @tab @code{MOR @var{a},@var{b},@var{c}}
9216 @item @code{uw1 __MPACKH (uh, uh)}
9217 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
9218 @tab @code{MPACKH @var{a},@var{b},@var{c}}
9219 @item @code{sw2 __MQADDHSS (sw2, sw2)}
9220 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
9221 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
9222 @item @code{uw2 __MQADDHUS (uw2, uw2)}
9223 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
9224 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
9225 @item @code{void __MQCPXIS (acc, sw2, sw2)}
9226 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
9227 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
9228 @item @code{void __MQCPXIU (acc, uw2, uw2)}
9229 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
9230 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
9231 @item @code{void __MQCPXRS (acc, sw2, sw2)}
9232 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
9233 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
9234 @item @code{void __MQCPXRU (acc, uw2, uw2)}
9235 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
9236 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
9237 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
9238 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
9239 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
9240 @item @code{sw2 __MQLMTHS (sw2, sw2)}
9241 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
9242 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
9243 @item @code{void __MQMACHS (acc, sw2, sw2)}
9244 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
9245 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
9246 @item @code{void __MQMACHU (acc, uw2, uw2)}
9247 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
9248 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
9249 @item @code{void __MQMACXHS (acc, sw2, sw2)}
9250 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
9251 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
9252 @item @code{void __MQMULHS (acc, sw2, sw2)}
9253 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
9254 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
9255 @item @code{void __MQMULHU (acc, uw2, uw2)}
9256 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
9257 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
9258 @item @code{void __MQMULXHS (acc, sw2, sw2)}
9259 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
9260 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
9261 @item @code{void __MQMULXHU (acc, uw2, uw2)}
9262 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
9263 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
9264 @item @code{sw2 __MQSATHS (sw2, sw2)}
9265 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
9266 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
9267 @item @code{uw2 __MQSLLHI (uw2, int)}
9268 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
9269 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
9270 @item @code{sw2 __MQSRAHI (sw2, int)}
9271 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
9272 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
9273 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
9274 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
9275 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
9276 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
9277 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
9278 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
9279 @item @code{void __MQXMACHS (acc, sw2, sw2)}
9280 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
9281 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
9282 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
9283 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
9284 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
9285 @item @code{uw1 __MRDACC (acc)}
9286 @tab @code{@var{b} = __MRDACC (@var{a})}
9287 @tab @code{MRDACC @var{a},@var{b}}
9288 @item @code{uw1 __MRDACCG (acc)}
9289 @tab @code{@var{b} = __MRDACCG (@var{a})}
9290 @tab @code{MRDACCG @var{a},@var{b}}
9291 @item @code{uw1 __MROTLI (uw1, const)}
9292 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
9293 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
9294 @item @code{uw1 __MROTRI (uw1, const)}
9295 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
9296 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
9297 @item @code{sw1 __MSATHS (sw1, sw1)}
9298 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
9299 @tab @code{MSATHS @var{a},@var{b},@var{c}}
9300 @item @code{uw1 __MSATHU (uw1, uw1)}
9301 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
9302 @tab @code{MSATHU @var{a},@var{b},@var{c}}
9303 @item @code{uw1 __MSLLHI (uw1, const)}
9304 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
9305 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
9306 @item @code{sw1 __MSRAHI (sw1, const)}
9307 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
9308 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
9309 @item @code{uw1 __MSRLHI (uw1, const)}
9310 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
9311 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
9312 @item @code{void __MSUBACCS (acc, acc)}
9313 @tab @code{__MSUBACCS (@var{b}, @var{a})}
9314 @tab @code{MSUBACCS @var{a},@var{b}}
9315 @item @code{sw1 __MSUBHSS (sw1, sw1)}
9316 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
9317 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
9318 @item @code{uw1 __MSUBHUS (uw1, uw1)}
9319 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
9320 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
9321 @item @code{void __MTRAP (void)}
9322 @tab @code{__MTRAP ()}
9324 @item @code{uw2 __MUNPACKH (uw1)}
9325 @tab @code{@var{b} = __MUNPACKH (@var{a})}
9326 @tab @code{MUNPACKH @var{a},@var{b}}
9327 @item @code{uw1 __MWCUT (uw2, uw1)}
9328 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
9329 @tab @code{MWCUT @var{a},@var{b},@var{c}}
9330 @item @code{void __MWTACC (acc, uw1)}
9331 @tab @code{__MWTACC (@var{b}, @var{a})}
9332 @tab @code{MWTACC @var{a},@var{b}}
9333 @item @code{void __MWTACCG (acc, uw1)}
9334 @tab @code{__MWTACCG (@var{b}, @var{a})}
9335 @tab @code{MWTACCG @var{a},@var{b}}
9336 @item @code{uw1 __MXOR (uw1, uw1)}
9337 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
9338 @tab @code{MXOR @var{a},@var{b},@var{c}}
9341 @node Raw read/write Functions
9342 @subsubsection Raw read/write Functions
9344 This sections describes built-in functions related to read and write
9345 instructions to access memory. These functions generate
9346 @code{membar} instructions to flush the I/O load and stores where
9347 appropriate, as described in Fujitsu's manual described above.
9351 @item unsigned char __builtin_read8 (void *@var{data})
9352 @item unsigned short __builtin_read16 (void *@var{data})
9353 @item unsigned long __builtin_read32 (void *@var{data})
9354 @item unsigned long long __builtin_read64 (void *@var{data})
9356 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
9357 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
9358 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
9359 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
9362 @node Other Built-in Functions
9363 @subsubsection Other Built-in Functions
9365 This section describes built-in functions that are not named after
9366 a specific FR-V instruction.
9369 @item sw2 __IACCreadll (iacc @var{reg})
9370 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
9371 for future expansion and must be 0.
9373 @item sw1 __IACCreadl (iacc @var{reg})
9374 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
9375 Other values of @var{reg} are rejected as invalid.
9377 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
9378 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
9379 is reserved for future expansion and must be 0.
9381 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
9382 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
9383 is 1. Other values of @var{reg} are rejected as invalid.
9385 @item void __data_prefetch0 (const void *@var{x})
9386 Use the @code{dcpl} instruction to load the contents of address @var{x}
9387 into the data cache.
9389 @item void __data_prefetch (const void *@var{x})
9390 Use the @code{nldub} instruction to load the contents of address @var{x}
9391 into the data cache. The instruction will be issued in slot I1@.
9394 @node X86 Built-in Functions
9395 @subsection X86 Built-in Functions
9397 These built-in functions are available for the i386 and x86-64 family
9398 of computers, depending on the command-line switches used.
9400 Note that, if you specify command-line switches such as @option{-msse},
9401 the compiler could use the extended instruction sets even if the built-ins
9402 are not used explicitly in the program. For this reason, applications
9403 which perform runtime CPU detection must compile separate files for each
9404 supported architecture, using the appropriate flags. In particular,
9405 the file containing the CPU detection code should be compiled without
9408 The following machine modes are available for use with MMX built-in functions
9409 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
9410 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
9411 vector of eight 8-bit integers. Some of the built-in functions operate on
9412 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
9414 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
9415 of two 32-bit floating point values.
9417 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
9418 floating point values. Some instructions use a vector of four 32-bit
9419 integers, these use @code{V4SI}. Finally, some instructions operate on an
9420 entire vector register, interpreting it as a 128-bit integer, these use mode
9423 In 64-bit mode, the x86-64 family of processors uses additional built-in
9424 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
9425 floating point and @code{TC} 128-bit complex floating point values.
9427 The following floating point built-in functions are available in 64-bit
9428 mode. All of them implement the function that is part of the name.
9431 __float128 __builtin_fabsq (__float128)
9432 __float128 __builtin_copysignq (__float128, __float128)
9435 The following built-in function is always available.
9438 @item void __builtin_ia32_pause (void)
9439 Generates the @code{pause} machine instruction with a compiler memory
9443 The following floating point built-in functions are made available in the
9447 @item __float128 __builtin_infq (void)
9448 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
9449 @findex __builtin_infq
9451 @item __float128 __builtin_huge_valq (void)
9452 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
9453 @findex __builtin_huge_valq
9456 The following built-in functions are made available by @option{-mmmx}.
9457 All of them generate the machine instruction that is part of the name.
9460 v8qi __builtin_ia32_paddb (v8qi, v8qi)
9461 v4hi __builtin_ia32_paddw (v4hi, v4hi)
9462 v2si __builtin_ia32_paddd (v2si, v2si)
9463 v8qi __builtin_ia32_psubb (v8qi, v8qi)
9464 v4hi __builtin_ia32_psubw (v4hi, v4hi)
9465 v2si __builtin_ia32_psubd (v2si, v2si)
9466 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
9467 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
9468 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
9469 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
9470 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
9471 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
9472 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
9473 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
9474 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
9475 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
9476 di __builtin_ia32_pand (di, di)
9477 di __builtin_ia32_pandn (di,di)
9478 di __builtin_ia32_por (di, di)
9479 di __builtin_ia32_pxor (di, di)
9480 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
9481 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
9482 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
9483 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
9484 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
9485 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
9486 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
9487 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
9488 v2si __builtin_ia32_punpckhdq (v2si, v2si)
9489 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
9490 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
9491 v2si __builtin_ia32_punpckldq (v2si, v2si)
9492 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
9493 v4hi __builtin_ia32_packssdw (v2si, v2si)
9494 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
9496 v4hi __builtin_ia32_psllw (v4hi, v4hi)
9497 v2si __builtin_ia32_pslld (v2si, v2si)
9498 v1di __builtin_ia32_psllq (v1di, v1di)
9499 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
9500 v2si __builtin_ia32_psrld (v2si, v2si)
9501 v1di __builtin_ia32_psrlq (v1di, v1di)
9502 v4hi __builtin_ia32_psraw (v4hi, v4hi)
9503 v2si __builtin_ia32_psrad (v2si, v2si)
9504 v4hi __builtin_ia32_psllwi (v4hi, int)
9505 v2si __builtin_ia32_pslldi (v2si, int)
9506 v1di __builtin_ia32_psllqi (v1di, int)
9507 v4hi __builtin_ia32_psrlwi (v4hi, int)
9508 v2si __builtin_ia32_psrldi (v2si, int)
9509 v1di __builtin_ia32_psrlqi (v1di, int)
9510 v4hi __builtin_ia32_psrawi (v4hi, int)
9511 v2si __builtin_ia32_psradi (v2si, int)
9515 The following built-in functions are made available either with
9516 @option{-msse}, or with a combination of @option{-m3dnow} and
9517 @option{-march=athlon}. All of them generate the machine
9518 instruction that is part of the name.
9521 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
9522 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
9523 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
9524 v1di __builtin_ia32_psadbw (v8qi, v8qi)
9525 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
9526 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
9527 v8qi __builtin_ia32_pminub (v8qi, v8qi)
9528 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
9529 int __builtin_ia32_pextrw (v4hi, int)
9530 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
9531 int __builtin_ia32_pmovmskb (v8qi)
9532 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
9533 void __builtin_ia32_movntq (di *, di)
9534 void __builtin_ia32_sfence (void)
9537 The following built-in functions are available when @option{-msse} is used.
9538 All of them generate the machine instruction that is part of the name.
9541 int __builtin_ia32_comieq (v4sf, v4sf)
9542 int __builtin_ia32_comineq (v4sf, v4sf)
9543 int __builtin_ia32_comilt (v4sf, v4sf)
9544 int __builtin_ia32_comile (v4sf, v4sf)
9545 int __builtin_ia32_comigt (v4sf, v4sf)
9546 int __builtin_ia32_comige (v4sf, v4sf)
9547 int __builtin_ia32_ucomieq (v4sf, v4sf)
9548 int __builtin_ia32_ucomineq (v4sf, v4sf)
9549 int __builtin_ia32_ucomilt (v4sf, v4sf)
9550 int __builtin_ia32_ucomile (v4sf, v4sf)
9551 int __builtin_ia32_ucomigt (v4sf, v4sf)
9552 int __builtin_ia32_ucomige (v4sf, v4sf)
9553 v4sf __builtin_ia32_addps (v4sf, v4sf)
9554 v4sf __builtin_ia32_subps (v4sf, v4sf)
9555 v4sf __builtin_ia32_mulps (v4sf, v4sf)
9556 v4sf __builtin_ia32_divps (v4sf, v4sf)
9557 v4sf __builtin_ia32_addss (v4sf, v4sf)
9558 v4sf __builtin_ia32_subss (v4sf, v4sf)
9559 v4sf __builtin_ia32_mulss (v4sf, v4sf)
9560 v4sf __builtin_ia32_divss (v4sf, v4sf)
9561 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
9562 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
9563 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
9564 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
9565 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
9566 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
9567 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
9568 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
9569 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
9570 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
9571 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
9572 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
9573 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
9574 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
9575 v4si __builtin_ia32_cmpless (v4sf, v4sf)
9576 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
9577 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
9578 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
9579 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
9580 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
9581 v4sf __builtin_ia32_maxps (v4sf, v4sf)
9582 v4sf __builtin_ia32_maxss (v4sf, v4sf)
9583 v4sf __builtin_ia32_minps (v4sf, v4sf)
9584 v4sf __builtin_ia32_minss (v4sf, v4sf)
9585 v4sf __builtin_ia32_andps (v4sf, v4sf)
9586 v4sf __builtin_ia32_andnps (v4sf, v4sf)
9587 v4sf __builtin_ia32_orps (v4sf, v4sf)
9588 v4sf __builtin_ia32_xorps (v4sf, v4sf)
9589 v4sf __builtin_ia32_movss (v4sf, v4sf)
9590 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
9591 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
9592 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
9593 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
9594 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
9595 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
9596 v2si __builtin_ia32_cvtps2pi (v4sf)
9597 int __builtin_ia32_cvtss2si (v4sf)
9598 v2si __builtin_ia32_cvttps2pi (v4sf)
9599 int __builtin_ia32_cvttss2si (v4sf)
9600 v4sf __builtin_ia32_rcpps (v4sf)
9601 v4sf __builtin_ia32_rsqrtps (v4sf)
9602 v4sf __builtin_ia32_sqrtps (v4sf)
9603 v4sf __builtin_ia32_rcpss (v4sf)
9604 v4sf __builtin_ia32_rsqrtss (v4sf)
9605 v4sf __builtin_ia32_sqrtss (v4sf)
9606 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
9607 void __builtin_ia32_movntps (float *, v4sf)
9608 int __builtin_ia32_movmskps (v4sf)
9611 The following built-in functions are available when @option{-msse} is used.
9614 @item v4sf __builtin_ia32_loadaps (float *)
9615 Generates the @code{movaps} machine instruction as a load from memory.
9616 @item void __builtin_ia32_storeaps (float *, v4sf)
9617 Generates the @code{movaps} machine instruction as a store to memory.
9618 @item v4sf __builtin_ia32_loadups (float *)
9619 Generates the @code{movups} machine instruction as a load from memory.
9620 @item void __builtin_ia32_storeups (float *, v4sf)
9621 Generates the @code{movups} machine instruction as a store to memory.
9622 @item v4sf __builtin_ia32_loadsss (float *)
9623 Generates the @code{movss} machine instruction as a load from memory.
9624 @item void __builtin_ia32_storess (float *, v4sf)
9625 Generates the @code{movss} machine instruction as a store to memory.
9626 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
9627 Generates the @code{movhps} machine instruction as a load from memory.
9628 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
9629 Generates the @code{movlps} machine instruction as a load from memory
9630 @item void __builtin_ia32_storehps (v2sf *, v4sf)
9631 Generates the @code{movhps} machine instruction as a store to memory.
9632 @item void __builtin_ia32_storelps (v2sf *, v4sf)
9633 Generates the @code{movlps} machine instruction as a store to memory.
9636 The following built-in functions are available when @option{-msse2} is used.
9637 All of them generate the machine instruction that is part of the name.
9640 int __builtin_ia32_comisdeq (v2df, v2df)
9641 int __builtin_ia32_comisdlt (v2df, v2df)
9642 int __builtin_ia32_comisdle (v2df, v2df)
9643 int __builtin_ia32_comisdgt (v2df, v2df)
9644 int __builtin_ia32_comisdge (v2df, v2df)
9645 int __builtin_ia32_comisdneq (v2df, v2df)
9646 int __builtin_ia32_ucomisdeq (v2df, v2df)
9647 int __builtin_ia32_ucomisdlt (v2df, v2df)
9648 int __builtin_ia32_ucomisdle (v2df, v2df)
9649 int __builtin_ia32_ucomisdgt (v2df, v2df)
9650 int __builtin_ia32_ucomisdge (v2df, v2df)
9651 int __builtin_ia32_ucomisdneq (v2df, v2df)
9652 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
9653 v2df __builtin_ia32_cmpltpd (v2df, v2df)
9654 v2df __builtin_ia32_cmplepd (v2df, v2df)
9655 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
9656 v2df __builtin_ia32_cmpgepd (v2df, v2df)
9657 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
9658 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
9659 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
9660 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
9661 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
9662 v2df __builtin_ia32_cmpngepd (v2df, v2df)
9663 v2df __builtin_ia32_cmpordpd (v2df, v2df)
9664 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
9665 v2df __builtin_ia32_cmpltsd (v2df, v2df)
9666 v2df __builtin_ia32_cmplesd (v2df, v2df)
9667 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
9668 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
9669 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
9670 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
9671 v2df __builtin_ia32_cmpordsd (v2df, v2df)
9672 v2di __builtin_ia32_paddq (v2di, v2di)
9673 v2di __builtin_ia32_psubq (v2di, v2di)
9674 v2df __builtin_ia32_addpd (v2df, v2df)
9675 v2df __builtin_ia32_subpd (v2df, v2df)
9676 v2df __builtin_ia32_mulpd (v2df, v2df)
9677 v2df __builtin_ia32_divpd (v2df, v2df)
9678 v2df __builtin_ia32_addsd (v2df, v2df)
9679 v2df __builtin_ia32_subsd (v2df, v2df)
9680 v2df __builtin_ia32_mulsd (v2df, v2df)
9681 v2df __builtin_ia32_divsd (v2df, v2df)
9682 v2df __builtin_ia32_minpd (v2df, v2df)
9683 v2df __builtin_ia32_maxpd (v2df, v2df)
9684 v2df __builtin_ia32_minsd (v2df, v2df)
9685 v2df __builtin_ia32_maxsd (v2df, v2df)
9686 v2df __builtin_ia32_andpd (v2df, v2df)
9687 v2df __builtin_ia32_andnpd (v2df, v2df)
9688 v2df __builtin_ia32_orpd (v2df, v2df)
9689 v2df __builtin_ia32_xorpd (v2df, v2df)
9690 v2df __builtin_ia32_movsd (v2df, v2df)
9691 v2df __builtin_ia32_unpckhpd (v2df, v2df)
9692 v2df __builtin_ia32_unpcklpd (v2df, v2df)
9693 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
9694 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
9695 v4si __builtin_ia32_paddd128 (v4si, v4si)
9696 v2di __builtin_ia32_paddq128 (v2di, v2di)
9697 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
9698 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
9699 v4si __builtin_ia32_psubd128 (v4si, v4si)
9700 v2di __builtin_ia32_psubq128 (v2di, v2di)
9701 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
9702 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
9703 v2di __builtin_ia32_pand128 (v2di, v2di)
9704 v2di __builtin_ia32_pandn128 (v2di, v2di)
9705 v2di __builtin_ia32_por128 (v2di, v2di)
9706 v2di __builtin_ia32_pxor128 (v2di, v2di)
9707 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
9708 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
9709 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
9710 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
9711 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
9712 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
9713 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
9714 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
9715 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
9716 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
9717 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
9718 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
9719 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
9720 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
9721 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
9722 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
9723 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
9724 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
9725 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
9726 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
9727 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
9728 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
9729 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
9730 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
9731 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
9732 v2df __builtin_ia32_loadupd (double *)
9733 void __builtin_ia32_storeupd (double *, v2df)
9734 v2df __builtin_ia32_loadhpd (v2df, double const *)
9735 v2df __builtin_ia32_loadlpd (v2df, double const *)
9736 int __builtin_ia32_movmskpd (v2df)
9737 int __builtin_ia32_pmovmskb128 (v16qi)
9738 void __builtin_ia32_movnti (int *, int)
9739 void __builtin_ia32_movnti64 (long long int *, long long int)
9740 void __builtin_ia32_movntpd (double *, v2df)
9741 void __builtin_ia32_movntdq (v2df *, v2df)
9742 v4si __builtin_ia32_pshufd (v4si, int)
9743 v8hi __builtin_ia32_pshuflw (v8hi, int)
9744 v8hi __builtin_ia32_pshufhw (v8hi, int)
9745 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
9746 v2df __builtin_ia32_sqrtpd (v2df)
9747 v2df __builtin_ia32_sqrtsd (v2df)
9748 v2df __builtin_ia32_shufpd (v2df, v2df, int)
9749 v2df __builtin_ia32_cvtdq2pd (v4si)
9750 v4sf __builtin_ia32_cvtdq2ps (v4si)
9751 v4si __builtin_ia32_cvtpd2dq (v2df)
9752 v2si __builtin_ia32_cvtpd2pi (v2df)
9753 v4sf __builtin_ia32_cvtpd2ps (v2df)
9754 v4si __builtin_ia32_cvttpd2dq (v2df)
9755 v2si __builtin_ia32_cvttpd2pi (v2df)
9756 v2df __builtin_ia32_cvtpi2pd (v2si)
9757 int __builtin_ia32_cvtsd2si (v2df)
9758 int __builtin_ia32_cvttsd2si (v2df)
9759 long long __builtin_ia32_cvtsd2si64 (v2df)
9760 long long __builtin_ia32_cvttsd2si64 (v2df)
9761 v4si __builtin_ia32_cvtps2dq (v4sf)
9762 v2df __builtin_ia32_cvtps2pd (v4sf)
9763 v4si __builtin_ia32_cvttps2dq (v4sf)
9764 v2df __builtin_ia32_cvtsi2sd (v2df, int)
9765 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
9766 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
9767 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
9768 void __builtin_ia32_clflush (const void *)
9769 void __builtin_ia32_lfence (void)
9770 void __builtin_ia32_mfence (void)
9771 v16qi __builtin_ia32_loaddqu (const char *)
9772 void __builtin_ia32_storedqu (char *, v16qi)
9773 v1di __builtin_ia32_pmuludq (v2si, v2si)
9774 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
9775 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
9776 v4si __builtin_ia32_pslld128 (v4si, v4si)
9777 v2di __builtin_ia32_psllq128 (v2di, v2di)
9778 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
9779 v4si __builtin_ia32_psrld128 (v4si, v4si)
9780 v2di __builtin_ia32_psrlq128 (v2di, v2di)
9781 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
9782 v4si __builtin_ia32_psrad128 (v4si, v4si)
9783 v2di __builtin_ia32_pslldqi128 (v2di, int)
9784 v8hi __builtin_ia32_psllwi128 (v8hi, int)
9785 v4si __builtin_ia32_pslldi128 (v4si, int)
9786 v2di __builtin_ia32_psllqi128 (v2di, int)
9787 v2di __builtin_ia32_psrldqi128 (v2di, int)
9788 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
9789 v4si __builtin_ia32_psrldi128 (v4si, int)
9790 v2di __builtin_ia32_psrlqi128 (v2di, int)
9791 v8hi __builtin_ia32_psrawi128 (v8hi, int)
9792 v4si __builtin_ia32_psradi128 (v4si, int)
9793 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
9794 v2di __builtin_ia32_movq128 (v2di)
9797 The following built-in functions are available when @option{-msse3} is used.
9798 All of them generate the machine instruction that is part of the name.
9801 v2df __builtin_ia32_addsubpd (v2df, v2df)
9802 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
9803 v2df __builtin_ia32_haddpd (v2df, v2df)
9804 v4sf __builtin_ia32_haddps (v4sf, v4sf)
9805 v2df __builtin_ia32_hsubpd (v2df, v2df)
9806 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
9807 v16qi __builtin_ia32_lddqu (char const *)
9808 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
9809 v2df __builtin_ia32_movddup (v2df)
9810 v4sf __builtin_ia32_movshdup (v4sf)
9811 v4sf __builtin_ia32_movsldup (v4sf)
9812 void __builtin_ia32_mwait (unsigned int, unsigned int)
9815 The following built-in functions are available when @option{-msse3} is used.
9818 @item v2df __builtin_ia32_loadddup (double const *)
9819 Generates the @code{movddup} machine instruction as a load from memory.
9822 The following built-in functions are available when @option{-mssse3} is used.
9823 All of them generate the machine instruction that is part of the name
9827 v2si __builtin_ia32_phaddd (v2si, v2si)
9828 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
9829 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
9830 v2si __builtin_ia32_phsubd (v2si, v2si)
9831 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
9832 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
9833 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
9834 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
9835 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
9836 v8qi __builtin_ia32_psignb (v8qi, v8qi)
9837 v2si __builtin_ia32_psignd (v2si, v2si)
9838 v4hi __builtin_ia32_psignw (v4hi, v4hi)
9839 v1di __builtin_ia32_palignr (v1di, v1di, int)
9840 v8qi __builtin_ia32_pabsb (v8qi)
9841 v2si __builtin_ia32_pabsd (v2si)
9842 v4hi __builtin_ia32_pabsw (v4hi)
9845 The following built-in functions are available when @option{-mssse3} is used.
9846 All of them generate the machine instruction that is part of the name
9850 v4si __builtin_ia32_phaddd128 (v4si, v4si)
9851 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
9852 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
9853 v4si __builtin_ia32_phsubd128 (v4si, v4si)
9854 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
9855 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
9856 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
9857 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
9858 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
9859 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
9860 v4si __builtin_ia32_psignd128 (v4si, v4si)
9861 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
9862 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
9863 v16qi __builtin_ia32_pabsb128 (v16qi)
9864 v4si __builtin_ia32_pabsd128 (v4si)
9865 v8hi __builtin_ia32_pabsw128 (v8hi)
9868 The following built-in functions are available when @option{-msse4.1} is
9869 used. All of them generate the machine instruction that is part of the
9873 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
9874 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
9875 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
9876 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
9877 v2df __builtin_ia32_dppd (v2df, v2df, const int)
9878 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
9879 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
9880 v2di __builtin_ia32_movntdqa (v2di *);
9881 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
9882 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
9883 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
9884 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
9885 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
9886 v8hi __builtin_ia32_phminposuw128 (v8hi)
9887 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
9888 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
9889 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
9890 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
9891 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
9892 v4si __builtin_ia32_pminsd128 (v4si, v4si)
9893 v4si __builtin_ia32_pminud128 (v4si, v4si)
9894 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
9895 v4si __builtin_ia32_pmovsxbd128 (v16qi)
9896 v2di __builtin_ia32_pmovsxbq128 (v16qi)
9897 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
9898 v2di __builtin_ia32_pmovsxdq128 (v4si)
9899 v4si __builtin_ia32_pmovsxwd128 (v8hi)
9900 v2di __builtin_ia32_pmovsxwq128 (v8hi)
9901 v4si __builtin_ia32_pmovzxbd128 (v16qi)
9902 v2di __builtin_ia32_pmovzxbq128 (v16qi)
9903 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
9904 v2di __builtin_ia32_pmovzxdq128 (v4si)
9905 v4si __builtin_ia32_pmovzxwd128 (v8hi)
9906 v2di __builtin_ia32_pmovzxwq128 (v8hi)
9907 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
9908 v4si __builtin_ia32_pmulld128 (v4si, v4si)
9909 int __builtin_ia32_ptestc128 (v2di, v2di)
9910 int __builtin_ia32_ptestnzc128 (v2di, v2di)
9911 int __builtin_ia32_ptestz128 (v2di, v2di)
9912 v2df __builtin_ia32_roundpd (v2df, const int)
9913 v4sf __builtin_ia32_roundps (v4sf, const int)
9914 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
9915 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
9918 The following built-in functions are available when @option{-msse4.1} is
9922 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
9923 Generates the @code{insertps} machine instruction.
9924 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
9925 Generates the @code{pextrb} machine instruction.
9926 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
9927 Generates the @code{pinsrb} machine instruction.
9928 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
9929 Generates the @code{pinsrd} machine instruction.
9930 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
9931 Generates the @code{pinsrq} machine instruction in 64bit mode.
9934 The following built-in functions are changed to generate new SSE4.1
9935 instructions when @option{-msse4.1} is used.
9938 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
9939 Generates the @code{extractps} machine instruction.
9940 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
9941 Generates the @code{pextrd} machine instruction.
9942 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
9943 Generates the @code{pextrq} machine instruction in 64bit mode.
9946 The following built-in functions are available when @option{-msse4.2} is
9947 used. All of them generate the machine instruction that is part of the
9951 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
9952 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
9953 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
9954 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
9955 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
9956 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
9957 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
9958 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
9959 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
9960 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
9961 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
9962 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
9963 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
9964 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
9965 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
9968 The following built-in functions are available when @option{-msse4.2} is
9972 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
9973 Generates the @code{crc32b} machine instruction.
9974 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
9975 Generates the @code{crc32w} machine instruction.
9976 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
9977 Generates the @code{crc32l} machine instruction.
9978 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
9979 Generates the @code{crc32q} machine instruction.
9982 The following built-in functions are changed to generate new SSE4.2
9983 instructions when @option{-msse4.2} is used.
9986 @item int __builtin_popcount (unsigned int)
9987 Generates the @code{popcntl} machine instruction.
9988 @item int __builtin_popcountl (unsigned long)
9989 Generates the @code{popcntl} or @code{popcntq} machine instruction,
9990 depending on the size of @code{unsigned long}.
9991 @item int __builtin_popcountll (unsigned long long)
9992 Generates the @code{popcntq} machine instruction.
9995 The following built-in functions are available when @option{-mavx} is
9996 used. All of them generate the machine instruction that is part of the
10000 v4df __builtin_ia32_addpd256 (v4df,v4df)
10001 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
10002 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
10003 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
10004 v4df __builtin_ia32_andnpd256 (v4df,v4df)
10005 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
10006 v4df __builtin_ia32_andpd256 (v4df,v4df)
10007 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
10008 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
10009 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
10010 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
10011 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
10012 v2df __builtin_ia32_cmppd (v2df,v2df,int)
10013 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
10014 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
10015 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
10016 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
10017 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
10018 v4df __builtin_ia32_cvtdq2pd256 (v4si)
10019 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
10020 v4si __builtin_ia32_cvtpd2dq256 (v4df)
10021 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
10022 v8si __builtin_ia32_cvtps2dq256 (v8sf)
10023 v4df __builtin_ia32_cvtps2pd256 (v4sf)
10024 v4si __builtin_ia32_cvttpd2dq256 (v4df)
10025 v8si __builtin_ia32_cvttps2dq256 (v8sf)
10026 v4df __builtin_ia32_divpd256 (v4df,v4df)
10027 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
10028 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
10029 v4df __builtin_ia32_haddpd256 (v4df,v4df)
10030 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
10031 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
10032 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
10033 v32qi __builtin_ia32_lddqu256 (pcchar)
10034 v32qi __builtin_ia32_loaddqu256 (pcchar)
10035 v4df __builtin_ia32_loadupd256 (pcdouble)
10036 v8sf __builtin_ia32_loadups256 (pcfloat)
10037 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
10038 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
10039 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
10040 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
10041 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
10042 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
10043 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
10044 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
10045 v4df __builtin_ia32_maxpd256 (v4df,v4df)
10046 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
10047 v4df __builtin_ia32_minpd256 (v4df,v4df)
10048 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
10049 v4df __builtin_ia32_movddup256 (v4df)
10050 int __builtin_ia32_movmskpd256 (v4df)
10051 int __builtin_ia32_movmskps256 (v8sf)
10052 v8sf __builtin_ia32_movshdup256 (v8sf)
10053 v8sf __builtin_ia32_movsldup256 (v8sf)
10054 v4df __builtin_ia32_mulpd256 (v4df,v4df)
10055 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
10056 v4df __builtin_ia32_orpd256 (v4df,v4df)
10057 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
10058 v2df __builtin_ia32_pd_pd256 (v4df)
10059 v4df __builtin_ia32_pd256_pd (v2df)
10060 v4sf __builtin_ia32_ps_ps256 (v8sf)
10061 v8sf __builtin_ia32_ps256_ps (v4sf)
10062 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
10063 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
10064 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
10065 v8sf __builtin_ia32_rcpps256 (v8sf)
10066 v4df __builtin_ia32_roundpd256 (v4df,int)
10067 v8sf __builtin_ia32_roundps256 (v8sf,int)
10068 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
10069 v8sf __builtin_ia32_rsqrtps256 (v8sf)
10070 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
10071 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
10072 v4si __builtin_ia32_si_si256 (v8si)
10073 v8si __builtin_ia32_si256_si (v4si)
10074 v4df __builtin_ia32_sqrtpd256 (v4df)
10075 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
10076 v8sf __builtin_ia32_sqrtps256 (v8sf)
10077 void __builtin_ia32_storedqu256 (pchar,v32qi)
10078 void __builtin_ia32_storeupd256 (pdouble,v4df)
10079 void __builtin_ia32_storeups256 (pfloat,v8sf)
10080 v4df __builtin_ia32_subpd256 (v4df,v4df)
10081 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
10082 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
10083 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
10084 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
10085 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
10086 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
10087 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
10088 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
10089 v4sf __builtin_ia32_vbroadcastss (pcfloat)
10090 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
10091 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
10092 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
10093 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
10094 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
10095 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
10096 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
10097 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
10098 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
10099 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
10100 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
10101 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
10102 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
10103 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
10104 v2df __builtin_ia32_vpermilpd (v2df,int)
10105 v4df __builtin_ia32_vpermilpd256 (v4df,int)
10106 v4sf __builtin_ia32_vpermilps (v4sf,int)
10107 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
10108 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
10109 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
10110 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
10111 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
10112 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
10113 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
10114 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
10115 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
10116 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
10117 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
10118 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
10119 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
10120 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
10121 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
10122 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
10123 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
10124 void __builtin_ia32_vzeroall (void)
10125 void __builtin_ia32_vzeroupper (void)
10126 v4df __builtin_ia32_xorpd256 (v4df,v4df)
10127 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
10130 The following built-in functions are available when @option{-mavx2} is
10131 used. All of them generate the machine instruction that is part of the
10135 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,v32qi,int)
10136 v32qi __builtin_ia32_pabsb256 (v32qi)
10137 v16hi __builtin_ia32_pabsw256 (v16hi)
10138 v8si __builtin_ia32_pabsd256 (v8si)
10139 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
10140 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
10141 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
10142 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
10143 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
10144 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
10145 v8si __builtin_ia32_paddd256 (v8si,v8si)
10146 v4di __builtin_ia32_paddq256 (v4di,v4di)
10147 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
10148 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
10149 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
10150 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
10151 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
10152 v4di __builtin_ia32_andsi256 (v4di,v4di)
10153 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
10154 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
10155 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
10156 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
10157 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
10158 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
10159 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
10160 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
10161 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
10162 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
10163 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
10164 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
10165 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
10166 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
10167 v8si __builtin_ia32_phaddd256 (v8si,v8si)
10168 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
10169 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
10170 v8si __builtin_ia32_phsubd256 (v8si,v8si)
10171 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
10172 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
10173 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
10174 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
10175 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
10176 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
10177 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
10178 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
10179 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
10180 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
10181 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
10182 v8si __builtin_ia32_pminsd256 (v8si,v8si)
10183 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
10184 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
10185 v8si __builtin_ia32_pminud256 (v8si,v8si)
10186 int __builtin_ia32_pmovmskb256 (v32qi)
10187 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
10188 v8si __builtin_ia32_pmovsxbd256 (v16qi)
10189 v4di __builtin_ia32_pmovsxbq256 (v16qi)
10190 v8si __builtin_ia32_pmovsxwd256 (v8hi)
10191 v4di __builtin_ia32_pmovsxwq256 (v8hi)
10192 v4di __builtin_ia32_pmovsxdq256 (v4si)
10193 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
10194 v8si __builtin_ia32_pmovzxbd256 (v16qi)
10195 v4di __builtin_ia32_pmovzxbq256 (v16qi)
10196 v8si __builtin_ia32_pmovzxwd256 (v8hi)
10197 v4di __builtin_ia32_pmovzxwq256 (v8hi)
10198 v4di __builtin_ia32_pmovzxdq256 (v4si)
10199 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
10200 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
10201 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
10202 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
10203 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
10204 v8si __builtin_ia32_pmulld256 (v8si,v8si)
10205 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
10206 v4di __builtin_ia32_por256 (v4di,v4di)
10207 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
10208 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
10209 v8si __builtin_ia32_pshufd256 (v8si,int)
10210 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
10211 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
10212 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
10213 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
10214 v8si __builtin_ia32_psignd256 (v8si,v8si)
10215 v4di __builtin_ia32_pslldqi256 (v4di,int)
10216 v16hi __builtin_ia32_psllwi256 (16hi,int)
10217 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
10218 v8si __builtin_ia32_pslldi256 (v8si,int)
10219 v8si __builtin_ia32_pslld256(v8si,v4si)
10220 v4di __builtin_ia32_psllqi256 (v4di,int)
10221 v4di __builtin_ia32_psllq256(v4di,v2di)
10222 v16hi __builtin_ia32_psrawi256 (v16hi,int)
10223 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
10224 v8si __builtin_ia32_psradi256 (v8si,int)
10225 v8si __builtin_ia32_psrad256 (v8si,v4si)
10226 v4di __builtin_ia32_psrldqi256 (v4di, int)
10227 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
10228 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
10229 v8si __builtin_ia32_psrldi256 (v8si,int)
10230 v8si __builtin_ia32_psrld256 (v8si,v4si)
10231 v4di __builtin_ia32_psrlqi256 (v4di,int)
10232 v4di __builtin_ia32_psrlq256(v4di,v2di)
10233 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
10234 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
10235 v8si __builtin_ia32_psubd256 (v8si,v8si)
10236 v4di __builtin_ia32_psubq256 (v4di,v4di)
10237 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
10238 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
10239 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
10240 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
10241 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
10242 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
10243 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
10244 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
10245 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
10246 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
10247 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
10248 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
10249 v4di __builtin_ia32_pxor256 (v4di,v4di)
10250 v4di __builtin_ia32_movntdqa256 (pv4di)
10251 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
10252 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
10253 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
10254 v4di __builtin_ia32_vbroadcastsi256 (v2di)
10255 v4si __builtin_ia32_pblendd128 (v4si,v4si)
10256 v8si __builtin_ia32_pblendd256 (v8si,v8si)
10257 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
10258 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
10259 v8si __builtin_ia32_pbroadcastd256 (v4si)
10260 v4di __builtin_ia32_pbroadcastq256 (v2di)
10261 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
10262 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
10263 v4si __builtin_ia32_pbroadcastd128 (v4si)
10264 v2di __builtin_ia32_pbroadcastq128 (v2di)
10265 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
10266 v4df __builtin_ia32_permdf256 (v4df,int)
10267 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
10268 v4di __builtin_ia32_permdi256 (v4di,int)
10269 v4di __builtin_ia32_permti256 (v4di,v4di,int)
10270 v4di __builtin_ia32_extract128i256 (v4di,int)
10271 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
10272 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
10273 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
10274 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
10275 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
10276 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
10277 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
10278 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
10279 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
10280 v8si __builtin_ia32_psllv8si (v8si,v8si)
10281 v4si __builtin_ia32_psllv4si (v4si,v4si)
10282 v4di __builtin_ia32_psllv4di (v4di,v4di)
10283 v2di __builtin_ia32_psllv2di (v2di,v2di)
10284 v8si __builtin_ia32_psrav8si (v8si,v8si)
10285 v4si __builtin_ia32_psrav4si (v4si,v4si)
10286 v8si __builtin_ia32_psrlv8si (v8si,v8si)
10287 v4si __builtin_ia32_psrlv4si (v4si,v4si)
10288 v4di __builtin_ia32_psrlv4di (v4di,v4di)
10289 v2di __builtin_ia32_psrlv2di (v2di,v2di)
10290 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
10291 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
10292 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
10293 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
10294 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
10295 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
10296 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
10297 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
10298 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
10299 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
10300 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
10301 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
10302 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
10303 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
10304 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
10305 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
10308 The following built-in functions are available when @option{-maes} is
10309 used. All of them generate the machine instruction that is part of the
10313 v2di __builtin_ia32_aesenc128 (v2di, v2di)
10314 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
10315 v2di __builtin_ia32_aesdec128 (v2di, v2di)
10316 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
10317 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
10318 v2di __builtin_ia32_aesimc128 (v2di)
10321 The following built-in function is available when @option{-mpclmul} is
10325 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
10326 Generates the @code{pclmulqdq} machine instruction.
10329 The following built-in function is available when @option{-mfsgsbase} is
10330 used. All of them generate the machine instruction that is part of the
10334 unsigned int __builtin_ia32_rdfsbase32 (void)
10335 unsigned long long __builtin_ia32_rdfsbase64 (void)
10336 unsigned int __builtin_ia32_rdgsbase32 (void)
10337 unsigned long long __builtin_ia32_rdgsbase64 (void)
10338 void _writefsbase_u32 (unsigned int)
10339 void _writefsbase_u64 (unsigned long long)
10340 void _writegsbase_u32 (unsigned int)
10341 void _writegsbase_u64 (unsigned long long)
10344 The following built-in function is available when @option{-mrdrnd} is
10345 used. All of them generate the machine instruction that is part of the
10349 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
10350 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
10351 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
10354 The following built-in functions are available when @option{-msse4a} is used.
10355 All of them generate the machine instruction that is part of the name.
10358 void __builtin_ia32_movntsd (double *, v2df)
10359 void __builtin_ia32_movntss (float *, v4sf)
10360 v2di __builtin_ia32_extrq (v2di, v16qi)
10361 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
10362 v2di __builtin_ia32_insertq (v2di, v2di)
10363 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
10366 The following built-in functions are available when @option{-mxop} is used.
10368 v2df __builtin_ia32_vfrczpd (v2df)
10369 v4sf __builtin_ia32_vfrczps (v4sf)
10370 v2df __builtin_ia32_vfrczsd (v2df, v2df)
10371 v4sf __builtin_ia32_vfrczss (v4sf, v4sf)
10372 v4df __builtin_ia32_vfrczpd256 (v4df)
10373 v8sf __builtin_ia32_vfrczps256 (v8sf)
10374 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
10375 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
10376 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
10377 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
10378 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
10379 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
10380 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
10381 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
10382 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
10383 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
10384 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
10385 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
10386 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
10387 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
10388 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
10389 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
10390 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
10391 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
10392 v4si __builtin_ia32_vpcomequd (v4si, v4si)
10393 v2di __builtin_ia32_vpcomequq (v2di, v2di)
10394 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
10395 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
10396 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
10397 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
10398 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
10399 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
10400 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
10401 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
10402 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
10403 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
10404 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
10405 v4si __builtin_ia32_vpcomged (v4si, v4si)
10406 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
10407 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
10408 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
10409 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
10410 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
10411 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
10412 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
10413 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
10414 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
10415 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
10416 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
10417 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
10418 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
10419 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
10420 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
10421 v4si __builtin_ia32_vpcomled (v4si, v4si)
10422 v2di __builtin_ia32_vpcomleq (v2di, v2di)
10423 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
10424 v4si __builtin_ia32_vpcomleud (v4si, v4si)
10425 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
10426 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
10427 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
10428 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
10429 v4si __builtin_ia32_vpcomltd (v4si, v4si)
10430 v2di __builtin_ia32_vpcomltq (v2di, v2di)
10431 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
10432 v4si __builtin_ia32_vpcomltud (v4si, v4si)
10433 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
10434 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
10435 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
10436 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
10437 v4si __builtin_ia32_vpcomned (v4si, v4si)
10438 v2di __builtin_ia32_vpcomneq (v2di, v2di)
10439 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
10440 v4si __builtin_ia32_vpcomneud (v4si, v4si)
10441 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
10442 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
10443 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
10444 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
10445 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
10446 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
10447 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
10448 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
10449 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
10450 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
10451 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
10452 v4si __builtin_ia32_vphaddbd (v16qi)
10453 v2di __builtin_ia32_vphaddbq (v16qi)
10454 v8hi __builtin_ia32_vphaddbw (v16qi)
10455 v2di __builtin_ia32_vphadddq (v4si)
10456 v4si __builtin_ia32_vphaddubd (v16qi)
10457 v2di __builtin_ia32_vphaddubq (v16qi)
10458 v8hi __builtin_ia32_vphaddubw (v16qi)
10459 v2di __builtin_ia32_vphaddudq (v4si)
10460 v4si __builtin_ia32_vphadduwd (v8hi)
10461 v2di __builtin_ia32_vphadduwq (v8hi)
10462 v4si __builtin_ia32_vphaddwd (v8hi)
10463 v2di __builtin_ia32_vphaddwq (v8hi)
10464 v8hi __builtin_ia32_vphsubbw (v16qi)
10465 v2di __builtin_ia32_vphsubdq (v4si)
10466 v4si __builtin_ia32_vphsubwd (v8hi)
10467 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
10468 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
10469 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
10470 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
10471 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
10472 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
10473 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
10474 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
10475 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
10476 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
10477 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
10478 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
10479 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
10480 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
10481 v4si __builtin_ia32_vprotd (v4si, v4si)
10482 v2di __builtin_ia32_vprotq (v2di, v2di)
10483 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
10484 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
10485 v4si __builtin_ia32_vpshad (v4si, v4si)
10486 v2di __builtin_ia32_vpshaq (v2di, v2di)
10487 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
10488 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
10489 v4si __builtin_ia32_vpshld (v4si, v4si)
10490 v2di __builtin_ia32_vpshlq (v2di, v2di)
10491 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
10494 The following built-in functions are available when @option{-mfma4} is used.
10495 All of them generate the machine instruction that is part of the name
10496 with MMX registers.
10499 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
10500 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
10501 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
10502 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
10503 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
10504 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
10505 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
10506 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
10507 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
10508 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
10509 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
10510 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
10511 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
10512 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
10513 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
10514 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
10515 v2df __builtin_ia32_fmaddsubpd (v2df, v2df, v2df)
10516 v4sf __builtin_ia32_fmaddsubps (v4sf, v4sf, v4sf)
10517 v2df __builtin_ia32_fmsubaddpd (v2df, v2df, v2df)
10518 v4sf __builtin_ia32_fmsubaddps (v4sf, v4sf, v4sf)
10519 v4df __builtin_ia32_fmaddpd256 (v4df, v4df, v4df)
10520 v8sf __builtin_ia32_fmaddps256 (v8sf, v8sf, v8sf)
10521 v4df __builtin_ia32_fmsubpd256 (v4df, v4df, v4df)
10522 v8sf __builtin_ia32_fmsubps256 (v8sf, v8sf, v8sf)
10523 v4df __builtin_ia32_fnmaddpd256 (v4df, v4df, v4df)
10524 v8sf __builtin_ia32_fnmaddps256 (v8sf, v8sf, v8sf)
10525 v4df __builtin_ia32_fnmsubpd256 (v4df, v4df, v4df)
10526 v8sf __builtin_ia32_fnmsubps256 (v8sf, v8sf, v8sf)
10527 v4df __builtin_ia32_fmaddsubpd256 (v4df, v4df, v4df)
10528 v8sf __builtin_ia32_fmaddsubps256 (v8sf, v8sf, v8sf)
10529 v4df __builtin_ia32_fmsubaddpd256 (v4df, v4df, v4df)
10530 v8sf __builtin_ia32_fmsubaddps256 (v8sf, v8sf, v8sf)
10534 The following built-in functions are available when @option{-mlwp} is used.
10537 void __builtin_ia32_llwpcb16 (void *);
10538 void __builtin_ia32_llwpcb32 (void *);
10539 void __builtin_ia32_llwpcb64 (void *);
10540 void * __builtin_ia32_llwpcb16 (void);
10541 void * __builtin_ia32_llwpcb32 (void);
10542 void * __builtin_ia32_llwpcb64 (void);
10543 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
10544 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
10545 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
10546 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
10547 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
10548 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
10551 The following built-in functions are available when @option{-mbmi} is used.
10552 All of them generate the machine instruction that is part of the name.
10554 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
10555 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
10558 The following built-in functions are available when @option{-mbmi2} is used.
10559 All of them generate the machine instruction that is part of the name.
10561 unsigned int _bzhi_u32 (unsigned int, unsigned int)
10562 unsigned int _pdep_u32 (unsigned int, unsigned int)
10563 unsigned int _pext_u32 (unsigned int, unsigned int)
10564 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
10565 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
10566 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
10569 The following built-in functions are available when @option{-mlzcnt} is used.
10570 All of them generate the machine instruction that is part of the name.
10572 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
10573 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
10574 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
10577 The following built-in functions are available when @option{-mtbm} is used.
10578 Both of them generate the immediate form of the bextr machine instruction.
10580 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
10581 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
10585 The following built-in functions are available when @option{-m3dnow} is used.
10586 All of them generate the machine instruction that is part of the name.
10589 void __builtin_ia32_femms (void)
10590 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
10591 v2si __builtin_ia32_pf2id (v2sf)
10592 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
10593 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
10594 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
10595 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
10596 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
10597 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
10598 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
10599 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
10600 v2sf __builtin_ia32_pfrcp (v2sf)
10601 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
10602 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
10603 v2sf __builtin_ia32_pfrsqrt (v2sf)
10604 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
10605 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
10606 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
10607 v2sf __builtin_ia32_pi2fd (v2si)
10608 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
10611 The following built-in functions are available when both @option{-m3dnow}
10612 and @option{-march=athlon} are used. All of them generate the machine
10613 instruction that is part of the name.
10616 v2si __builtin_ia32_pf2iw (v2sf)
10617 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
10618 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
10619 v2sf __builtin_ia32_pi2fw (v2si)
10620 v2sf __builtin_ia32_pswapdsf (v2sf)
10621 v2si __builtin_ia32_pswapdsi (v2si)
10624 @node MIPS DSP Built-in Functions
10625 @subsection MIPS DSP Built-in Functions
10627 The MIPS DSP Application-Specific Extension (ASE) includes new
10628 instructions that are designed to improve the performance of DSP and
10629 media applications. It provides instructions that operate on packed
10630 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
10632 GCC supports MIPS DSP operations using both the generic
10633 vector extensions (@pxref{Vector Extensions}) and a collection of
10634 MIPS-specific built-in functions. Both kinds of support are
10635 enabled by the @option{-mdsp} command-line option.
10637 Revision 2 of the ASE was introduced in the second half of 2006.
10638 This revision adds extra instructions to the original ASE, but is
10639 otherwise backwards-compatible with it. You can select revision 2
10640 using the command-line option @option{-mdspr2}; this option implies
10643 The SCOUNT and POS bits of the DSP control register are global. The
10644 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
10645 POS bits. During optimization, the compiler will not delete these
10646 instructions and it will not delete calls to functions containing
10647 these instructions.
10649 At present, GCC only provides support for operations on 32-bit
10650 vectors. The vector type associated with 8-bit integer data is
10651 usually called @code{v4i8}, the vector type associated with Q7
10652 is usually called @code{v4q7}, the vector type associated with 16-bit
10653 integer data is usually called @code{v2i16}, and the vector type
10654 associated with Q15 is usually called @code{v2q15}. They can be
10655 defined in C as follows:
10658 typedef signed char v4i8 __attribute__ ((vector_size(4)));
10659 typedef signed char v4q7 __attribute__ ((vector_size(4)));
10660 typedef short v2i16 __attribute__ ((vector_size(4)));
10661 typedef short v2q15 __attribute__ ((vector_size(4)));
10664 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
10665 initialized in the same way as aggregates. For example:
10668 v4i8 a = @{1, 2, 3, 4@};
10670 b = (v4i8) @{5, 6, 7, 8@};
10672 v2q15 c = @{0x0fcb, 0x3a75@};
10674 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
10677 @emph{Note:} The CPU's endianness determines the order in which values
10678 are packed. On little-endian targets, the first value is the least
10679 significant and the last value is the most significant. The opposite
10680 order applies to big-endian targets. For example, the code above will
10681 set the lowest byte of @code{a} to @code{1} on little-endian targets
10682 and @code{4} on big-endian targets.
10684 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
10685 representation. As shown in this example, the integer representation
10686 of a Q7 value can be obtained by multiplying the fractional value by
10687 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
10688 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
10691 The table below lists the @code{v4i8} and @code{v2q15} operations for which
10692 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
10693 and @code{c} and @code{d} are @code{v2q15} values.
10695 @multitable @columnfractions .50 .50
10696 @item C code @tab MIPS instruction
10697 @item @code{a + b} @tab @code{addu.qb}
10698 @item @code{c + d} @tab @code{addq.ph}
10699 @item @code{a - b} @tab @code{subu.qb}
10700 @item @code{c - d} @tab @code{subq.ph}
10703 The table below lists the @code{v2i16} operation for which
10704 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
10705 @code{v2i16} values.
10707 @multitable @columnfractions .50 .50
10708 @item C code @tab MIPS instruction
10709 @item @code{e * f} @tab @code{mul.ph}
10712 It is easier to describe the DSP built-in functions if we first define
10713 the following types:
10718 typedef unsigned int ui32;
10719 typedef long long a64;
10722 @code{q31} and @code{i32} are actually the same as @code{int}, but we
10723 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
10724 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
10725 @code{long long}, but we use @code{a64} to indicate values that will
10726 be placed in one of the four DSP accumulators (@code{$ac0},
10727 @code{$ac1}, @code{$ac2} or @code{$ac3}).
10729 Also, some built-in functions prefer or require immediate numbers as
10730 parameters, because the corresponding DSP instructions accept both immediate
10731 numbers and register operands, or accept immediate numbers only. The
10732 immediate parameters are listed as follows.
10740 imm0_255: 0 to 255.
10741 imm_n32_31: -32 to 31.
10742 imm_n512_511: -512 to 511.
10745 The following built-in functions map directly to a particular MIPS DSP
10746 instruction. Please refer to the architecture specification
10747 for details on what each instruction does.
10750 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
10751 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
10752 q31 __builtin_mips_addq_s_w (q31, q31)
10753 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
10754 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
10755 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
10756 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
10757 q31 __builtin_mips_subq_s_w (q31, q31)
10758 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
10759 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
10760 i32 __builtin_mips_addsc (i32, i32)
10761 i32 __builtin_mips_addwc (i32, i32)
10762 i32 __builtin_mips_modsub (i32, i32)
10763 i32 __builtin_mips_raddu_w_qb (v4i8)
10764 v2q15 __builtin_mips_absq_s_ph (v2q15)
10765 q31 __builtin_mips_absq_s_w (q31)
10766 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
10767 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
10768 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
10769 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
10770 q31 __builtin_mips_preceq_w_phl (v2q15)
10771 q31 __builtin_mips_preceq_w_phr (v2q15)
10772 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
10773 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
10774 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
10775 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
10776 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
10777 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
10778 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
10779 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
10780 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
10781 v4i8 __builtin_mips_shll_qb (v4i8, i32)
10782 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
10783 v2q15 __builtin_mips_shll_ph (v2q15, i32)
10784 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
10785 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
10786 q31 __builtin_mips_shll_s_w (q31, imm0_31)
10787 q31 __builtin_mips_shll_s_w (q31, i32)
10788 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
10789 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
10790 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
10791 v2q15 __builtin_mips_shra_ph (v2q15, i32)
10792 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
10793 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
10794 q31 __builtin_mips_shra_r_w (q31, imm0_31)
10795 q31 __builtin_mips_shra_r_w (q31, i32)
10796 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
10797 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
10798 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
10799 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
10800 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
10801 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
10802 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
10803 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
10804 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
10805 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
10806 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
10807 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
10808 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
10809 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
10810 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
10811 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
10812 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
10813 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
10814 i32 __builtin_mips_bitrev (i32)
10815 i32 __builtin_mips_insv (i32, i32)
10816 v4i8 __builtin_mips_repl_qb (imm0_255)
10817 v4i8 __builtin_mips_repl_qb (i32)
10818 v2q15 __builtin_mips_repl_ph (imm_n512_511)
10819 v2q15 __builtin_mips_repl_ph (i32)
10820 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
10821 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
10822 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
10823 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
10824 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
10825 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
10826 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
10827 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
10828 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
10829 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
10830 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
10831 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
10832 i32 __builtin_mips_extr_w (a64, imm0_31)
10833 i32 __builtin_mips_extr_w (a64, i32)
10834 i32 __builtin_mips_extr_r_w (a64, imm0_31)
10835 i32 __builtin_mips_extr_s_h (a64, i32)
10836 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
10837 i32 __builtin_mips_extr_rs_w (a64, i32)
10838 i32 __builtin_mips_extr_s_h (a64, imm0_31)
10839 i32 __builtin_mips_extr_r_w (a64, i32)
10840 i32 __builtin_mips_extp (a64, imm0_31)
10841 i32 __builtin_mips_extp (a64, i32)
10842 i32 __builtin_mips_extpdp (a64, imm0_31)
10843 i32 __builtin_mips_extpdp (a64, i32)
10844 a64 __builtin_mips_shilo (a64, imm_n32_31)
10845 a64 __builtin_mips_shilo (a64, i32)
10846 a64 __builtin_mips_mthlip (a64, i32)
10847 void __builtin_mips_wrdsp (i32, imm0_63)
10848 i32 __builtin_mips_rddsp (imm0_63)
10849 i32 __builtin_mips_lbux (void *, i32)
10850 i32 __builtin_mips_lhx (void *, i32)
10851 i32 __builtin_mips_lwx (void *, i32)
10852 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
10853 i32 __builtin_mips_bposge32 (void)
10854 a64 __builtin_mips_madd (a64, i32, i32);
10855 a64 __builtin_mips_maddu (a64, ui32, ui32);
10856 a64 __builtin_mips_msub (a64, i32, i32);
10857 a64 __builtin_mips_msubu (a64, ui32, ui32);
10858 a64 __builtin_mips_mult (i32, i32);
10859 a64 __builtin_mips_multu (ui32, ui32);
10862 The following built-in functions map directly to a particular MIPS DSP REV 2
10863 instruction. Please refer to the architecture specification
10864 for details on what each instruction does.
10867 v4q7 __builtin_mips_absq_s_qb (v4q7);
10868 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
10869 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
10870 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
10871 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
10872 i32 __builtin_mips_append (i32, i32, imm0_31);
10873 i32 __builtin_mips_balign (i32, i32, imm0_3);
10874 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
10875 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
10876 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
10877 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
10878 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
10879 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
10880 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
10881 q31 __builtin_mips_mulq_rs_w (q31, q31);
10882 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
10883 q31 __builtin_mips_mulq_s_w (q31, q31);
10884 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
10885 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
10886 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
10887 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
10888 i32 __builtin_mips_prepend (i32, i32, imm0_31);
10889 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
10890 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
10891 v4i8 __builtin_mips_shra_qb (v4i8, i32);
10892 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
10893 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
10894 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
10895 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
10896 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
10897 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
10898 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
10899 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
10900 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
10901 q31 __builtin_mips_addqh_w (q31, q31);
10902 q31 __builtin_mips_addqh_r_w (q31, q31);
10903 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
10904 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
10905 q31 __builtin_mips_subqh_w (q31, q31);
10906 q31 __builtin_mips_subqh_r_w (q31, q31);
10907 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
10908 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
10909 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
10910 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
10911 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
10912 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
10916 @node MIPS Paired-Single Support
10917 @subsection MIPS Paired-Single Support
10919 The MIPS64 architecture includes a number of instructions that
10920 operate on pairs of single-precision floating-point values.
10921 Each pair is packed into a 64-bit floating-point register,
10922 with one element being designated the ``upper half'' and
10923 the other being designated the ``lower half''.
10925 GCC supports paired-single operations using both the generic
10926 vector extensions (@pxref{Vector Extensions}) and a collection of
10927 MIPS-specific built-in functions. Both kinds of support are
10928 enabled by the @option{-mpaired-single} command-line option.
10930 The vector type associated with paired-single values is usually
10931 called @code{v2sf}. It can be defined in C as follows:
10934 typedef float v2sf __attribute__ ((vector_size (8)));
10937 @code{v2sf} values are initialized in the same way as aggregates.
10941 v2sf a = @{1.5, 9.1@};
10944 b = (v2sf) @{e, f@};
10947 @emph{Note:} The CPU's endianness determines which value is stored in
10948 the upper half of a register and which value is stored in the lower half.
10949 On little-endian targets, the first value is the lower one and the second
10950 value is the upper one. The opposite order applies to big-endian targets.
10951 For example, the code above will set the lower half of @code{a} to
10952 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
10954 @node MIPS Loongson Built-in Functions
10955 @subsection MIPS Loongson Built-in Functions
10957 GCC provides intrinsics to access the SIMD instructions provided by the
10958 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
10959 available after inclusion of the @code{loongson.h} header file,
10960 operate on the following 64-bit vector types:
10963 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
10964 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
10965 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
10966 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
10967 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
10968 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
10971 The intrinsics provided are listed below; each is named after the
10972 machine instruction to which it corresponds, with suffixes added as
10973 appropriate to distinguish intrinsics that expand to the same machine
10974 instruction yet have different argument types. Refer to the architecture
10975 documentation for a description of the functionality of each
10979 int16x4_t packsswh (int32x2_t s, int32x2_t t);
10980 int8x8_t packsshb (int16x4_t s, int16x4_t t);
10981 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
10982 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
10983 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
10984 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
10985 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
10986 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
10987 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
10988 uint64_t paddd_u (uint64_t s, uint64_t t);
10989 int64_t paddd_s (int64_t s, int64_t t);
10990 int16x4_t paddsh (int16x4_t s, int16x4_t t);
10991 int8x8_t paddsb (int8x8_t s, int8x8_t t);
10992 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
10993 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
10994 uint64_t pandn_ud (uint64_t s, uint64_t t);
10995 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
10996 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
10997 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
10998 int64_t pandn_sd (int64_t s, int64_t t);
10999 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
11000 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
11001 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
11002 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
11003 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
11004 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
11005 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
11006 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
11007 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
11008 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
11009 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
11010 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
11011 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
11012 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
11013 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
11014 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
11015 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
11016 uint16x4_t pextrh_u (uint16x4_t s, int field);
11017 int16x4_t pextrh_s (int16x4_t s, int field);
11018 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
11019 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
11020 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
11021 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
11022 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
11023 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
11024 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
11025 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
11026 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
11027 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
11028 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
11029 int16x4_t pminsh (int16x4_t s, int16x4_t t);
11030 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
11031 uint8x8_t pmovmskb_u (uint8x8_t s);
11032 int8x8_t pmovmskb_s (int8x8_t s);
11033 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
11034 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
11035 int16x4_t pmullh (int16x4_t s, int16x4_t t);
11036 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
11037 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
11038 uint16x4_t biadd (uint8x8_t s);
11039 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
11040 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
11041 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
11042 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
11043 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
11044 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
11045 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
11046 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
11047 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
11048 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
11049 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
11050 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
11051 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
11052 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
11053 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
11054 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
11055 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
11056 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
11057 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
11058 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
11059 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
11060 uint64_t psubd_u (uint64_t s, uint64_t t);
11061 int64_t psubd_s (int64_t s, int64_t t);
11062 int16x4_t psubsh (int16x4_t s, int16x4_t t);
11063 int8x8_t psubsb (int8x8_t s, int8x8_t t);
11064 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
11065 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
11066 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
11067 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
11068 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
11069 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
11070 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
11071 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
11072 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
11073 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
11074 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
11075 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
11076 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
11077 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
11081 * Paired-Single Arithmetic::
11082 * Paired-Single Built-in Functions::
11083 * MIPS-3D Built-in Functions::
11086 @node Paired-Single Arithmetic
11087 @subsubsection Paired-Single Arithmetic
11089 The table below lists the @code{v2sf} operations for which hardware
11090 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
11091 values and @code{x} is an integral value.
11093 @multitable @columnfractions .50 .50
11094 @item C code @tab MIPS instruction
11095 @item @code{a + b} @tab @code{add.ps}
11096 @item @code{a - b} @tab @code{sub.ps}
11097 @item @code{-a} @tab @code{neg.ps}
11098 @item @code{a * b} @tab @code{mul.ps}
11099 @item @code{a * b + c} @tab @code{madd.ps}
11100 @item @code{a * b - c} @tab @code{msub.ps}
11101 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
11102 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
11103 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
11106 Note that the multiply-accumulate instructions can be disabled
11107 using the command-line option @code{-mno-fused-madd}.
11109 @node Paired-Single Built-in Functions
11110 @subsubsection Paired-Single Built-in Functions
11112 The following paired-single functions map directly to a particular
11113 MIPS instruction. Please refer to the architecture specification
11114 for details on what each instruction does.
11117 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
11118 Pair lower lower (@code{pll.ps}).
11120 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
11121 Pair upper lower (@code{pul.ps}).
11123 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
11124 Pair lower upper (@code{plu.ps}).
11126 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
11127 Pair upper upper (@code{puu.ps}).
11129 @item v2sf __builtin_mips_cvt_ps_s (float, float)
11130 Convert pair to paired single (@code{cvt.ps.s}).
11132 @item float __builtin_mips_cvt_s_pl (v2sf)
11133 Convert pair lower to single (@code{cvt.s.pl}).
11135 @item float __builtin_mips_cvt_s_pu (v2sf)
11136 Convert pair upper to single (@code{cvt.s.pu}).
11138 @item v2sf __builtin_mips_abs_ps (v2sf)
11139 Absolute value (@code{abs.ps}).
11141 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
11142 Align variable (@code{alnv.ps}).
11144 @emph{Note:} The value of the third parameter must be 0 or 4
11145 modulo 8, otherwise the result will be unpredictable. Please read the
11146 instruction description for details.
11149 The following multi-instruction functions are also available.
11150 In each case, @var{cond} can be any of the 16 floating-point conditions:
11151 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
11152 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
11153 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
11156 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11157 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11158 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
11159 @code{movt.ps}/@code{movf.ps}).
11161 The @code{movt} functions return the value @var{x} computed by:
11164 c.@var{cond}.ps @var{cc},@var{a},@var{b}
11165 mov.ps @var{x},@var{c}
11166 movt.ps @var{x},@var{d},@var{cc}
11169 The @code{movf} functions are similar but use @code{movf.ps} instead
11172 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11173 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11174 Comparison of two paired-single values (@code{c.@var{cond}.ps},
11175 @code{bc1t}/@code{bc1f}).
11177 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
11178 and return either the upper or lower half of the result. For example:
11182 if (__builtin_mips_upper_c_eq_ps (a, b))
11183 upper_halves_are_equal ();
11185 upper_halves_are_unequal ();
11187 if (__builtin_mips_lower_c_eq_ps (a, b))
11188 lower_halves_are_equal ();
11190 lower_halves_are_unequal ();
11194 @node MIPS-3D Built-in Functions
11195 @subsubsection MIPS-3D Built-in Functions
11197 The MIPS-3D Application-Specific Extension (ASE) includes additional
11198 paired-single instructions that are designed to improve the performance
11199 of 3D graphics operations. Support for these instructions is controlled
11200 by the @option{-mips3d} command-line option.
11202 The functions listed below map directly to a particular MIPS-3D
11203 instruction. Please refer to the architecture specification for
11204 more details on what each instruction does.
11207 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
11208 Reduction add (@code{addr.ps}).
11210 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
11211 Reduction multiply (@code{mulr.ps}).
11213 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
11214 Convert paired single to paired word (@code{cvt.pw.ps}).
11216 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
11217 Convert paired word to paired single (@code{cvt.ps.pw}).
11219 @item float __builtin_mips_recip1_s (float)
11220 @itemx double __builtin_mips_recip1_d (double)
11221 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
11222 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
11224 @item float __builtin_mips_recip2_s (float, float)
11225 @itemx double __builtin_mips_recip2_d (double, double)
11226 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
11227 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
11229 @item float __builtin_mips_rsqrt1_s (float)
11230 @itemx double __builtin_mips_rsqrt1_d (double)
11231 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
11232 Reduced precision reciprocal square root (sequence step 1)
11233 (@code{rsqrt1.@var{fmt}}).
11235 @item float __builtin_mips_rsqrt2_s (float, float)
11236 @itemx double __builtin_mips_rsqrt2_d (double, double)
11237 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
11238 Reduced precision reciprocal square root (sequence step 2)
11239 (@code{rsqrt2.@var{fmt}}).
11242 The following multi-instruction functions are also available.
11243 In each case, @var{cond} can be any of the 16 floating-point conditions:
11244 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
11245 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
11246 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
11249 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
11250 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
11251 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
11252 @code{bc1t}/@code{bc1f}).
11254 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
11255 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
11260 if (__builtin_mips_cabs_eq_s (a, b))
11266 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11267 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11268 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
11269 @code{bc1t}/@code{bc1f}).
11271 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
11272 and return either the upper or lower half of the result. For example:
11276 if (__builtin_mips_upper_cabs_eq_ps (a, b))
11277 upper_halves_are_equal ();
11279 upper_halves_are_unequal ();
11281 if (__builtin_mips_lower_cabs_eq_ps (a, b))
11282 lower_halves_are_equal ();
11284 lower_halves_are_unequal ();
11287 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11288 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11289 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
11290 @code{movt.ps}/@code{movf.ps}).
11292 The @code{movt} functions return the value @var{x} computed by:
11295 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
11296 mov.ps @var{x},@var{c}
11297 movt.ps @var{x},@var{d},@var{cc}
11300 The @code{movf} functions are similar but use @code{movf.ps} instead
11303 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11304 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11305 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11306 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11307 Comparison of two paired-single values
11308 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
11309 @code{bc1any2t}/@code{bc1any2f}).
11311 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
11312 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
11313 result is true and the @code{all} forms return true if both results are true.
11318 if (__builtin_mips_any_c_eq_ps (a, b))
11323 if (__builtin_mips_all_c_eq_ps (a, b))
11329 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11330 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11331 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11332 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11333 Comparison of four paired-single values
11334 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
11335 @code{bc1any4t}/@code{bc1any4f}).
11337 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
11338 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
11339 The @code{any} forms return true if any of the four results are true
11340 and the @code{all} forms return true if all four results are true.
11345 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
11350 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
11357 @node picoChip Built-in Functions
11358 @subsection picoChip Built-in Functions
11360 GCC provides an interface to selected machine instructions from the
11361 picoChip instruction set.
11364 @item int __builtin_sbc (int @var{value})
11365 Sign bit count. Return the number of consecutive bits in @var{value}
11366 which have the same value as the sign-bit. The result is the number of
11367 leading sign bits minus one, giving the number of redundant sign bits in
11370 @item int __builtin_byteswap (int @var{value})
11371 Byte swap. Return the result of swapping the upper and lower bytes of
11374 @item int __builtin_brev (int @var{value})
11375 Bit reversal. Return the result of reversing the bits in
11376 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
11379 @item int __builtin_adds (int @var{x}, int @var{y})
11380 Saturating addition. Return the result of adding @var{x} and @var{y},
11381 storing the value 32767 if the result overflows.
11383 @item int __builtin_subs (int @var{x}, int @var{y})
11384 Saturating subtraction. Return the result of subtracting @var{y} from
11385 @var{x}, storing the value @minus{}32768 if the result overflows.
11387 @item void __builtin_halt (void)
11388 Halt. The processor will stop execution. This built-in is useful for
11389 implementing assertions.
11393 @node Other MIPS Built-in Functions
11394 @subsection Other MIPS Built-in Functions
11396 GCC provides other MIPS-specific built-in functions:
11399 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
11400 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
11401 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
11402 when this function is available.
11405 @node PowerPC AltiVec/VSX Built-in Functions
11406 @subsection PowerPC AltiVec Built-in Functions
11408 GCC provides an interface for the PowerPC family of processors to access
11409 the AltiVec operations described in Motorola's AltiVec Programming
11410 Interface Manual. The interface is made available by including
11411 @code{<altivec.h>} and using @option{-maltivec} and
11412 @option{-mabi=altivec}. The interface supports the following vector
11416 vector unsigned char
11420 vector unsigned short
11421 vector signed short
11425 vector unsigned int
11431 If @option{-mvsx} is used the following additional vector types are
11435 vector unsigned long
11440 The long types are only implemented for 64-bit code generation, and
11441 the long type is only used in the floating point/integer conversion
11444 GCC's implementation of the high-level language interface available from
11445 C and C++ code differs from Motorola's documentation in several ways.
11450 A vector constant is a list of constant expressions within curly braces.
11453 A vector initializer requires no cast if the vector constant is of the
11454 same type as the variable it is initializing.
11457 If @code{signed} or @code{unsigned} is omitted, the signedness of the
11458 vector type is the default signedness of the base type. The default
11459 varies depending on the operating system, so a portable program should
11460 always specify the signedness.
11463 Compiling with @option{-maltivec} adds keywords @code{__vector},
11464 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
11465 @code{bool}. When compiling ISO C, the context-sensitive substitution
11466 of the keywords @code{vector}, @code{pixel} and @code{bool} is
11467 disabled. To use them, you must include @code{<altivec.h>} instead.
11470 GCC allows using a @code{typedef} name as the type specifier for a
11474 For C, overloaded functions are implemented with macros so the following
11478 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
11481 Since @code{vec_add} is a macro, the vector constant in the example
11482 is treated as four separate arguments. Wrap the entire argument in
11483 parentheses for this to work.
11486 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
11487 Internally, GCC uses built-in functions to achieve the functionality in
11488 the aforementioned header file, but they are not supported and are
11489 subject to change without notice.
11491 The following interfaces are supported for the generic and specific
11492 AltiVec operations and the AltiVec predicates. In cases where there
11493 is a direct mapping between generic and specific operations, only the
11494 generic names are shown here, although the specific operations can also
11497 Arguments that are documented as @code{const int} require literal
11498 integral values within the range required for that operation.
11501 vector signed char vec_abs (vector signed char);
11502 vector signed short vec_abs (vector signed short);
11503 vector signed int vec_abs (vector signed int);
11504 vector float vec_abs (vector float);
11506 vector signed char vec_abss (vector signed char);
11507 vector signed short vec_abss (vector signed short);
11508 vector signed int vec_abss (vector signed int);
11510 vector signed char vec_add (vector bool char, vector signed char);
11511 vector signed char vec_add (vector signed char, vector bool char);
11512 vector signed char vec_add (vector signed char, vector signed char);
11513 vector unsigned char vec_add (vector bool char, vector unsigned char);
11514 vector unsigned char vec_add (vector unsigned char, vector bool char);
11515 vector unsigned char vec_add (vector unsigned char,
11516 vector unsigned char);
11517 vector signed short vec_add (vector bool short, vector signed short);
11518 vector signed short vec_add (vector signed short, vector bool short);
11519 vector signed short vec_add (vector signed short, vector signed short);
11520 vector unsigned short vec_add (vector bool short,
11521 vector unsigned short);
11522 vector unsigned short vec_add (vector unsigned short,
11523 vector bool short);
11524 vector unsigned short vec_add (vector unsigned short,
11525 vector unsigned short);
11526 vector signed int vec_add (vector bool int, vector signed int);
11527 vector signed int vec_add (vector signed int, vector bool int);
11528 vector signed int vec_add (vector signed int, vector signed int);
11529 vector unsigned int vec_add (vector bool int, vector unsigned int);
11530 vector unsigned int vec_add (vector unsigned int, vector bool int);
11531 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
11532 vector float vec_add (vector float, vector float);
11534 vector float vec_vaddfp (vector float, vector float);
11536 vector signed int vec_vadduwm (vector bool int, vector signed int);
11537 vector signed int vec_vadduwm (vector signed int, vector bool int);
11538 vector signed int vec_vadduwm (vector signed int, vector signed int);
11539 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
11540 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
11541 vector unsigned int vec_vadduwm (vector unsigned int,
11542 vector unsigned int);
11544 vector signed short vec_vadduhm (vector bool short,
11545 vector signed short);
11546 vector signed short vec_vadduhm (vector signed short,
11547 vector bool short);
11548 vector signed short vec_vadduhm (vector signed short,
11549 vector signed short);
11550 vector unsigned short vec_vadduhm (vector bool short,
11551 vector unsigned short);
11552 vector unsigned short vec_vadduhm (vector unsigned short,
11553 vector bool short);
11554 vector unsigned short vec_vadduhm (vector unsigned short,
11555 vector unsigned short);
11557 vector signed char vec_vaddubm (vector bool char, vector signed char);
11558 vector signed char vec_vaddubm (vector signed char, vector bool char);
11559 vector signed char vec_vaddubm (vector signed char, vector signed char);
11560 vector unsigned char vec_vaddubm (vector bool char,
11561 vector unsigned char);
11562 vector unsigned char vec_vaddubm (vector unsigned char,
11564 vector unsigned char vec_vaddubm (vector unsigned char,
11565 vector unsigned char);
11567 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
11569 vector unsigned char vec_adds (vector bool char, vector unsigned char);
11570 vector unsigned char vec_adds (vector unsigned char, vector bool char);
11571 vector unsigned char vec_adds (vector unsigned char,
11572 vector unsigned char);
11573 vector signed char vec_adds (vector bool char, vector signed char);
11574 vector signed char vec_adds (vector signed char, vector bool char);
11575 vector signed char vec_adds (vector signed char, vector signed char);
11576 vector unsigned short vec_adds (vector bool short,
11577 vector unsigned short);
11578 vector unsigned short vec_adds (vector unsigned short,
11579 vector bool short);
11580 vector unsigned short vec_adds (vector unsigned short,
11581 vector unsigned short);
11582 vector signed short vec_adds (vector bool short, vector signed short);
11583 vector signed short vec_adds (vector signed short, vector bool short);
11584 vector signed short vec_adds (vector signed short, vector signed short);
11585 vector unsigned int vec_adds (vector bool int, vector unsigned int);
11586 vector unsigned int vec_adds (vector unsigned int, vector bool int);
11587 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
11588 vector signed int vec_adds (vector bool int, vector signed int);
11589 vector signed int vec_adds (vector signed int, vector bool int);
11590 vector signed int vec_adds (vector signed int, vector signed int);
11592 vector signed int vec_vaddsws (vector bool int, vector signed int);
11593 vector signed int vec_vaddsws (vector signed int, vector bool int);
11594 vector signed int vec_vaddsws (vector signed int, vector signed int);
11596 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
11597 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
11598 vector unsigned int vec_vadduws (vector unsigned int,
11599 vector unsigned int);
11601 vector signed short vec_vaddshs (vector bool short,
11602 vector signed short);
11603 vector signed short vec_vaddshs (vector signed short,
11604 vector bool short);
11605 vector signed short vec_vaddshs (vector signed short,
11606 vector signed short);
11608 vector unsigned short vec_vadduhs (vector bool short,
11609 vector unsigned short);
11610 vector unsigned short vec_vadduhs (vector unsigned short,
11611 vector bool short);
11612 vector unsigned short vec_vadduhs (vector unsigned short,
11613 vector unsigned short);
11615 vector signed char vec_vaddsbs (vector bool char, vector signed char);
11616 vector signed char vec_vaddsbs (vector signed char, vector bool char);
11617 vector signed char vec_vaddsbs (vector signed char, vector signed char);
11619 vector unsigned char vec_vaddubs (vector bool char,
11620 vector unsigned char);
11621 vector unsigned char vec_vaddubs (vector unsigned char,
11623 vector unsigned char vec_vaddubs (vector unsigned char,
11624 vector unsigned char);
11626 vector float vec_and (vector float, vector float);
11627 vector float vec_and (vector float, vector bool int);
11628 vector float vec_and (vector bool int, vector float);
11629 vector bool int vec_and (vector bool int, vector bool int);
11630 vector signed int vec_and (vector bool int, vector signed int);
11631 vector signed int vec_and (vector signed int, vector bool int);
11632 vector signed int vec_and (vector signed int, vector signed int);
11633 vector unsigned int vec_and (vector bool int, vector unsigned int);
11634 vector unsigned int vec_and (vector unsigned int, vector bool int);
11635 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
11636 vector bool short vec_and (vector bool short, vector bool short);
11637 vector signed short vec_and (vector bool short, vector signed short);
11638 vector signed short vec_and (vector signed short, vector bool short);
11639 vector signed short vec_and (vector signed short, vector signed short);
11640 vector unsigned short vec_and (vector bool short,
11641 vector unsigned short);
11642 vector unsigned short vec_and (vector unsigned short,
11643 vector bool short);
11644 vector unsigned short vec_and (vector unsigned short,
11645 vector unsigned short);
11646 vector signed char vec_and (vector bool char, vector signed char);
11647 vector bool char vec_and (vector bool char, vector bool char);
11648 vector signed char vec_and (vector signed char, vector bool char);
11649 vector signed char vec_and (vector signed char, vector signed char);
11650 vector unsigned char vec_and (vector bool char, vector unsigned char);
11651 vector unsigned char vec_and (vector unsigned char, vector bool char);
11652 vector unsigned char vec_and (vector unsigned char,
11653 vector unsigned char);
11655 vector float vec_andc (vector float, vector float);
11656 vector float vec_andc (vector float, vector bool int);
11657 vector float vec_andc (vector bool int, vector float);
11658 vector bool int vec_andc (vector bool int, vector bool int);
11659 vector signed int vec_andc (vector bool int, vector signed int);
11660 vector signed int vec_andc (vector signed int, vector bool int);
11661 vector signed int vec_andc (vector signed int, vector signed int);
11662 vector unsigned int vec_andc (vector bool int, vector unsigned int);
11663 vector unsigned int vec_andc (vector unsigned int, vector bool int);
11664 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
11665 vector bool short vec_andc (vector bool short, vector bool short);
11666 vector signed short vec_andc (vector bool short, vector signed short);
11667 vector signed short vec_andc (vector signed short, vector bool short);
11668 vector signed short vec_andc (vector signed short, vector signed short);
11669 vector unsigned short vec_andc (vector bool short,
11670 vector unsigned short);
11671 vector unsigned short vec_andc (vector unsigned short,
11672 vector bool short);
11673 vector unsigned short vec_andc (vector unsigned short,
11674 vector unsigned short);
11675 vector signed char vec_andc (vector bool char, vector signed char);
11676 vector bool char vec_andc (vector bool char, vector bool char);
11677 vector signed char vec_andc (vector signed char, vector bool char);
11678 vector signed char vec_andc (vector signed char, vector signed char);
11679 vector unsigned char vec_andc (vector bool char, vector unsigned char);
11680 vector unsigned char vec_andc (vector unsigned char, vector bool char);
11681 vector unsigned char vec_andc (vector unsigned char,
11682 vector unsigned char);
11684 vector unsigned char vec_avg (vector unsigned char,
11685 vector unsigned char);
11686 vector signed char vec_avg (vector signed char, vector signed char);
11687 vector unsigned short vec_avg (vector unsigned short,
11688 vector unsigned short);
11689 vector signed short vec_avg (vector signed short, vector signed short);
11690 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
11691 vector signed int vec_avg (vector signed int, vector signed int);
11693 vector signed int vec_vavgsw (vector signed int, vector signed int);
11695 vector unsigned int vec_vavguw (vector unsigned int,
11696 vector unsigned int);
11698 vector signed short vec_vavgsh (vector signed short,
11699 vector signed short);
11701 vector unsigned short vec_vavguh (vector unsigned short,
11702 vector unsigned short);
11704 vector signed char vec_vavgsb (vector signed char, vector signed char);
11706 vector unsigned char vec_vavgub (vector unsigned char,
11707 vector unsigned char);
11709 vector float vec_copysign (vector float);
11711 vector float vec_ceil (vector float);
11713 vector signed int vec_cmpb (vector float, vector float);
11715 vector bool char vec_cmpeq (vector signed char, vector signed char);
11716 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
11717 vector bool short vec_cmpeq (vector signed short, vector signed short);
11718 vector bool short vec_cmpeq (vector unsigned short,
11719 vector unsigned short);
11720 vector bool int vec_cmpeq (vector signed int, vector signed int);
11721 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
11722 vector bool int vec_cmpeq (vector float, vector float);
11724 vector bool int vec_vcmpeqfp (vector float, vector float);
11726 vector bool int vec_vcmpequw (vector signed int, vector signed int);
11727 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
11729 vector bool short vec_vcmpequh (vector signed short,
11730 vector signed short);
11731 vector bool short vec_vcmpequh (vector unsigned short,
11732 vector unsigned short);
11734 vector bool char vec_vcmpequb (vector signed char, vector signed char);
11735 vector bool char vec_vcmpequb (vector unsigned char,
11736 vector unsigned char);
11738 vector bool int vec_cmpge (vector float, vector float);
11740 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
11741 vector bool char vec_cmpgt (vector signed char, vector signed char);
11742 vector bool short vec_cmpgt (vector unsigned short,
11743 vector unsigned short);
11744 vector bool short vec_cmpgt (vector signed short, vector signed short);
11745 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
11746 vector bool int vec_cmpgt (vector signed int, vector signed int);
11747 vector bool int vec_cmpgt (vector float, vector float);
11749 vector bool int vec_vcmpgtfp (vector float, vector float);
11751 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
11753 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
11755 vector bool short vec_vcmpgtsh (vector signed short,
11756 vector signed short);
11758 vector bool short vec_vcmpgtuh (vector unsigned short,
11759 vector unsigned short);
11761 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
11763 vector bool char vec_vcmpgtub (vector unsigned char,
11764 vector unsigned char);
11766 vector bool int vec_cmple (vector float, vector float);
11768 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
11769 vector bool char vec_cmplt (vector signed char, vector signed char);
11770 vector bool short vec_cmplt (vector unsigned short,
11771 vector unsigned short);
11772 vector bool short vec_cmplt (vector signed short, vector signed short);
11773 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
11774 vector bool int vec_cmplt (vector signed int, vector signed int);
11775 vector bool int vec_cmplt (vector float, vector float);
11777 vector float vec_ctf (vector unsigned int, const int);
11778 vector float vec_ctf (vector signed int, const int);
11780 vector float vec_vcfsx (vector signed int, const int);
11782 vector float vec_vcfux (vector unsigned int, const int);
11784 vector signed int vec_cts (vector float, const int);
11786 vector unsigned int vec_ctu (vector float, const int);
11788 void vec_dss (const int);
11790 void vec_dssall (void);
11792 void vec_dst (const vector unsigned char *, int, const int);
11793 void vec_dst (const vector signed char *, int, const int);
11794 void vec_dst (const vector bool char *, int, const int);
11795 void vec_dst (const vector unsigned short *, int, const int);
11796 void vec_dst (const vector signed short *, int, const int);
11797 void vec_dst (const vector bool short *, int, const int);
11798 void vec_dst (const vector pixel *, int, const int);
11799 void vec_dst (const vector unsigned int *, int, const int);
11800 void vec_dst (const vector signed int *, int, const int);
11801 void vec_dst (const vector bool int *, int, const int);
11802 void vec_dst (const vector float *, int, const int);
11803 void vec_dst (const unsigned char *, int, const int);
11804 void vec_dst (const signed char *, int, const int);
11805 void vec_dst (const unsigned short *, int, const int);
11806 void vec_dst (const short *, int, const int);
11807 void vec_dst (const unsigned int *, int, const int);
11808 void vec_dst (const int *, int, const int);
11809 void vec_dst (const unsigned long *, int, const int);
11810 void vec_dst (const long *, int, const int);
11811 void vec_dst (const float *, int, const int);
11813 void vec_dstst (const vector unsigned char *, int, const int);
11814 void vec_dstst (const vector signed char *, int, const int);
11815 void vec_dstst (const vector bool char *, int, const int);
11816 void vec_dstst (const vector unsigned short *, int, const int);
11817 void vec_dstst (const vector signed short *, int, const int);
11818 void vec_dstst (const vector bool short *, int, const int);
11819 void vec_dstst (const vector pixel *, int, const int);
11820 void vec_dstst (const vector unsigned int *, int, const int);
11821 void vec_dstst (const vector signed int *, int, const int);
11822 void vec_dstst (const vector bool int *, int, const int);
11823 void vec_dstst (const vector float *, int, const int);
11824 void vec_dstst (const unsigned char *, int, const int);
11825 void vec_dstst (const signed char *, int, const int);
11826 void vec_dstst (const unsigned short *, int, const int);
11827 void vec_dstst (const short *, int, const int);
11828 void vec_dstst (const unsigned int *, int, const int);
11829 void vec_dstst (const int *, int, const int);
11830 void vec_dstst (const unsigned long *, int, const int);
11831 void vec_dstst (const long *, int, const int);
11832 void vec_dstst (const float *, int, const int);
11834 void vec_dststt (const vector unsigned char *, int, const int);
11835 void vec_dststt (const vector signed char *, int, const int);
11836 void vec_dststt (const vector bool char *, int, const int);
11837 void vec_dststt (const vector unsigned short *, int, const int);
11838 void vec_dststt (const vector signed short *, int, const int);
11839 void vec_dststt (const vector bool short *, int, const int);
11840 void vec_dststt (const vector pixel *, int, const int);
11841 void vec_dststt (const vector unsigned int *, int, const int);
11842 void vec_dststt (const vector signed int *, int, const int);
11843 void vec_dststt (const vector bool int *, int, const int);
11844 void vec_dststt (const vector float *, int, const int);
11845 void vec_dststt (const unsigned char *, int, const int);
11846 void vec_dststt (const signed char *, int, const int);
11847 void vec_dststt (const unsigned short *, int, const int);
11848 void vec_dststt (const short *, int, const int);
11849 void vec_dststt (const unsigned int *, int, const int);
11850 void vec_dststt (const int *, int, const int);
11851 void vec_dststt (const unsigned long *, int, const int);
11852 void vec_dststt (const long *, int, const int);
11853 void vec_dststt (const float *, int, const int);
11855 void vec_dstt (const vector unsigned char *, int, const int);
11856 void vec_dstt (const vector signed char *, int, const int);
11857 void vec_dstt (const vector bool char *, int, const int);
11858 void vec_dstt (const vector unsigned short *, int, const int);
11859 void vec_dstt (const vector signed short *, int, const int);
11860 void vec_dstt (const vector bool short *, int, const int);
11861 void vec_dstt (const vector pixel *, int, const int);
11862 void vec_dstt (const vector unsigned int *, int, const int);
11863 void vec_dstt (const vector signed int *, int, const int);
11864 void vec_dstt (const vector bool int *, int, const int);
11865 void vec_dstt (const vector float *, int, const int);
11866 void vec_dstt (const unsigned char *, int, const int);
11867 void vec_dstt (const signed char *, int, const int);
11868 void vec_dstt (const unsigned short *, int, const int);
11869 void vec_dstt (const short *, int, const int);
11870 void vec_dstt (const unsigned int *, int, const int);
11871 void vec_dstt (const int *, int, const int);
11872 void vec_dstt (const unsigned long *, int, const int);
11873 void vec_dstt (const long *, int, const int);
11874 void vec_dstt (const float *, int, const int);
11876 vector float vec_expte (vector float);
11878 vector float vec_floor (vector float);
11880 vector float vec_ld (int, const vector float *);
11881 vector float vec_ld (int, const float *);
11882 vector bool int vec_ld (int, const vector bool int *);
11883 vector signed int vec_ld (int, const vector signed int *);
11884 vector signed int vec_ld (int, const int *);
11885 vector signed int vec_ld (int, const long *);
11886 vector unsigned int vec_ld (int, const vector unsigned int *);
11887 vector unsigned int vec_ld (int, const unsigned int *);
11888 vector unsigned int vec_ld (int, const unsigned long *);
11889 vector bool short vec_ld (int, const vector bool short *);
11890 vector pixel vec_ld (int, const vector pixel *);
11891 vector signed short vec_ld (int, const vector signed short *);
11892 vector signed short vec_ld (int, const short *);
11893 vector unsigned short vec_ld (int, const vector unsigned short *);
11894 vector unsigned short vec_ld (int, const unsigned short *);
11895 vector bool char vec_ld (int, const vector bool char *);
11896 vector signed char vec_ld (int, const vector signed char *);
11897 vector signed char vec_ld (int, const signed char *);
11898 vector unsigned char vec_ld (int, const vector unsigned char *);
11899 vector unsigned char vec_ld (int, const unsigned char *);
11901 vector signed char vec_lde (int, const signed char *);
11902 vector unsigned char vec_lde (int, const unsigned char *);
11903 vector signed short vec_lde (int, const short *);
11904 vector unsigned short vec_lde (int, const unsigned short *);
11905 vector float vec_lde (int, const float *);
11906 vector signed int vec_lde (int, const int *);
11907 vector unsigned int vec_lde (int, const unsigned int *);
11908 vector signed int vec_lde (int, const long *);
11909 vector unsigned int vec_lde (int, const unsigned long *);
11911 vector float vec_lvewx (int, float *);
11912 vector signed int vec_lvewx (int, int *);
11913 vector unsigned int vec_lvewx (int, unsigned int *);
11914 vector signed int vec_lvewx (int, long *);
11915 vector unsigned int vec_lvewx (int, unsigned long *);
11917 vector signed short vec_lvehx (int, short *);
11918 vector unsigned short vec_lvehx (int, unsigned short *);
11920 vector signed char vec_lvebx (int, char *);
11921 vector unsigned char vec_lvebx (int, unsigned char *);
11923 vector float vec_ldl (int, const vector float *);
11924 vector float vec_ldl (int, const float *);
11925 vector bool int vec_ldl (int, const vector bool int *);
11926 vector signed int vec_ldl (int, const vector signed int *);
11927 vector signed int vec_ldl (int, const int *);
11928 vector signed int vec_ldl (int, const long *);
11929 vector unsigned int vec_ldl (int, const vector unsigned int *);
11930 vector unsigned int vec_ldl (int, const unsigned int *);
11931 vector unsigned int vec_ldl (int, const unsigned long *);
11932 vector bool short vec_ldl (int, const vector bool short *);
11933 vector pixel vec_ldl (int, const vector pixel *);
11934 vector signed short vec_ldl (int, const vector signed short *);
11935 vector signed short vec_ldl (int, const short *);
11936 vector unsigned short vec_ldl (int, const vector unsigned short *);
11937 vector unsigned short vec_ldl (int, const unsigned short *);
11938 vector bool char vec_ldl (int, const vector bool char *);
11939 vector signed char vec_ldl (int, const vector signed char *);
11940 vector signed char vec_ldl (int, const signed char *);
11941 vector unsigned char vec_ldl (int, const vector unsigned char *);
11942 vector unsigned char vec_ldl (int, const unsigned char *);
11944 vector float vec_loge (vector float);
11946 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
11947 vector unsigned char vec_lvsl (int, const volatile signed char *);
11948 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
11949 vector unsigned char vec_lvsl (int, const volatile short *);
11950 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
11951 vector unsigned char vec_lvsl (int, const volatile int *);
11952 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
11953 vector unsigned char vec_lvsl (int, const volatile long *);
11954 vector unsigned char vec_lvsl (int, const volatile float *);
11956 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
11957 vector unsigned char vec_lvsr (int, const volatile signed char *);
11958 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
11959 vector unsigned char vec_lvsr (int, const volatile short *);
11960 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
11961 vector unsigned char vec_lvsr (int, const volatile int *);
11962 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
11963 vector unsigned char vec_lvsr (int, const volatile long *);
11964 vector unsigned char vec_lvsr (int, const volatile float *);
11966 vector float vec_madd (vector float, vector float, vector float);
11968 vector signed short vec_madds (vector signed short,
11969 vector signed short,
11970 vector signed short);
11972 vector unsigned char vec_max (vector bool char, vector unsigned char);
11973 vector unsigned char vec_max (vector unsigned char, vector bool char);
11974 vector unsigned char vec_max (vector unsigned char,
11975 vector unsigned char);
11976 vector signed char vec_max (vector bool char, vector signed char);
11977 vector signed char vec_max (vector signed char, vector bool char);
11978 vector signed char vec_max (vector signed char, vector signed char);
11979 vector unsigned short vec_max (vector bool short,
11980 vector unsigned short);
11981 vector unsigned short vec_max (vector unsigned short,
11982 vector bool short);
11983 vector unsigned short vec_max (vector unsigned short,
11984 vector unsigned short);
11985 vector signed short vec_max (vector bool short, vector signed short);
11986 vector signed short vec_max (vector signed short, vector bool short);
11987 vector signed short vec_max (vector signed short, vector signed short);
11988 vector unsigned int vec_max (vector bool int, vector unsigned int);
11989 vector unsigned int vec_max (vector unsigned int, vector bool int);
11990 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
11991 vector signed int vec_max (vector bool int, vector signed int);
11992 vector signed int vec_max (vector signed int, vector bool int);
11993 vector signed int vec_max (vector signed int, vector signed int);
11994 vector float vec_max (vector float, vector float);
11996 vector float vec_vmaxfp (vector float, vector float);
11998 vector signed int vec_vmaxsw (vector bool int, vector signed int);
11999 vector signed int vec_vmaxsw (vector signed int, vector bool int);
12000 vector signed int vec_vmaxsw (vector signed int, vector signed int);
12002 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
12003 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
12004 vector unsigned int vec_vmaxuw (vector unsigned int,
12005 vector unsigned int);
12007 vector signed short vec_vmaxsh (vector bool short, vector signed short);
12008 vector signed short vec_vmaxsh (vector signed short, vector bool short);
12009 vector signed short vec_vmaxsh (vector signed short,
12010 vector signed short);
12012 vector unsigned short vec_vmaxuh (vector bool short,
12013 vector unsigned short);
12014 vector unsigned short vec_vmaxuh (vector unsigned short,
12015 vector bool short);
12016 vector unsigned short vec_vmaxuh (vector unsigned short,
12017 vector unsigned short);
12019 vector signed char vec_vmaxsb (vector bool char, vector signed char);
12020 vector signed char vec_vmaxsb (vector signed char, vector bool char);
12021 vector signed char vec_vmaxsb (vector signed char, vector signed char);
12023 vector unsigned char vec_vmaxub (vector bool char,
12024 vector unsigned char);
12025 vector unsigned char vec_vmaxub (vector unsigned char,
12027 vector unsigned char vec_vmaxub (vector unsigned char,
12028 vector unsigned char);
12030 vector bool char vec_mergeh (vector bool char, vector bool char);
12031 vector signed char vec_mergeh (vector signed char, vector signed char);
12032 vector unsigned char vec_mergeh (vector unsigned char,
12033 vector unsigned char);
12034 vector bool short vec_mergeh (vector bool short, vector bool short);
12035 vector pixel vec_mergeh (vector pixel, vector pixel);
12036 vector signed short vec_mergeh (vector signed short,
12037 vector signed short);
12038 vector unsigned short vec_mergeh (vector unsigned short,
12039 vector unsigned short);
12040 vector float vec_mergeh (vector float, vector float);
12041 vector bool int vec_mergeh (vector bool int, vector bool int);
12042 vector signed int vec_mergeh (vector signed int, vector signed int);
12043 vector unsigned int vec_mergeh (vector unsigned int,
12044 vector unsigned int);
12046 vector float vec_vmrghw (vector float, vector float);
12047 vector bool int vec_vmrghw (vector bool int, vector bool int);
12048 vector signed int vec_vmrghw (vector signed int, vector signed int);
12049 vector unsigned int vec_vmrghw (vector unsigned int,
12050 vector unsigned int);
12052 vector bool short vec_vmrghh (vector bool short, vector bool short);
12053 vector signed short vec_vmrghh (vector signed short,
12054 vector signed short);
12055 vector unsigned short vec_vmrghh (vector unsigned short,
12056 vector unsigned short);
12057 vector pixel vec_vmrghh (vector pixel, vector pixel);
12059 vector bool char vec_vmrghb (vector bool char, vector bool char);
12060 vector signed char vec_vmrghb (vector signed char, vector signed char);
12061 vector unsigned char vec_vmrghb (vector unsigned char,
12062 vector unsigned char);
12064 vector bool char vec_mergel (vector bool char, vector bool char);
12065 vector signed char vec_mergel (vector signed char, vector signed char);
12066 vector unsigned char vec_mergel (vector unsigned char,
12067 vector unsigned char);
12068 vector bool short vec_mergel (vector bool short, vector bool short);
12069 vector pixel vec_mergel (vector pixel, vector pixel);
12070 vector signed short vec_mergel (vector signed short,
12071 vector signed short);
12072 vector unsigned short vec_mergel (vector unsigned short,
12073 vector unsigned short);
12074 vector float vec_mergel (vector float, vector float);
12075 vector bool int vec_mergel (vector bool int, vector bool int);
12076 vector signed int vec_mergel (vector signed int, vector signed int);
12077 vector unsigned int vec_mergel (vector unsigned int,
12078 vector unsigned int);
12080 vector float vec_vmrglw (vector float, vector float);
12081 vector signed int vec_vmrglw (vector signed int, vector signed int);
12082 vector unsigned int vec_vmrglw (vector unsigned int,
12083 vector unsigned int);
12084 vector bool int vec_vmrglw (vector bool int, vector bool int);
12086 vector bool short vec_vmrglh (vector bool short, vector bool short);
12087 vector signed short vec_vmrglh (vector signed short,
12088 vector signed short);
12089 vector unsigned short vec_vmrglh (vector unsigned short,
12090 vector unsigned short);
12091 vector pixel vec_vmrglh (vector pixel, vector pixel);
12093 vector bool char vec_vmrglb (vector bool char, vector bool char);
12094 vector signed char vec_vmrglb (vector signed char, vector signed char);
12095 vector unsigned char vec_vmrglb (vector unsigned char,
12096 vector unsigned char);
12098 vector unsigned short vec_mfvscr (void);
12100 vector unsigned char vec_min (vector bool char, vector unsigned char);
12101 vector unsigned char vec_min (vector unsigned char, vector bool char);
12102 vector unsigned char vec_min (vector unsigned char,
12103 vector unsigned char);
12104 vector signed char vec_min (vector bool char, vector signed char);
12105 vector signed char vec_min (vector signed char, vector bool char);
12106 vector signed char vec_min (vector signed char, vector signed char);
12107 vector unsigned short vec_min (vector bool short,
12108 vector unsigned short);
12109 vector unsigned short vec_min (vector unsigned short,
12110 vector bool short);
12111 vector unsigned short vec_min (vector unsigned short,
12112 vector unsigned short);
12113 vector signed short vec_min (vector bool short, vector signed short);
12114 vector signed short vec_min (vector signed short, vector bool short);
12115 vector signed short vec_min (vector signed short, vector signed short);
12116 vector unsigned int vec_min (vector bool int, vector unsigned int);
12117 vector unsigned int vec_min (vector unsigned int, vector bool int);
12118 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
12119 vector signed int vec_min (vector bool int, vector signed int);
12120 vector signed int vec_min (vector signed int, vector bool int);
12121 vector signed int vec_min (vector signed int, vector signed int);
12122 vector float vec_min (vector float, vector float);
12124 vector float vec_vminfp (vector float, vector float);
12126 vector signed int vec_vminsw (vector bool int, vector signed int);
12127 vector signed int vec_vminsw (vector signed int, vector bool int);
12128 vector signed int vec_vminsw (vector signed int, vector signed int);
12130 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
12131 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
12132 vector unsigned int vec_vminuw (vector unsigned int,
12133 vector unsigned int);
12135 vector signed short vec_vminsh (vector bool short, vector signed short);
12136 vector signed short vec_vminsh (vector signed short, vector bool short);
12137 vector signed short vec_vminsh (vector signed short,
12138 vector signed short);
12140 vector unsigned short vec_vminuh (vector bool short,
12141 vector unsigned short);
12142 vector unsigned short vec_vminuh (vector unsigned short,
12143 vector bool short);
12144 vector unsigned short vec_vminuh (vector unsigned short,
12145 vector unsigned short);
12147 vector signed char vec_vminsb (vector bool char, vector signed char);
12148 vector signed char vec_vminsb (vector signed char, vector bool char);
12149 vector signed char vec_vminsb (vector signed char, vector signed char);
12151 vector unsigned char vec_vminub (vector bool char,
12152 vector unsigned char);
12153 vector unsigned char vec_vminub (vector unsigned char,
12155 vector unsigned char vec_vminub (vector unsigned char,
12156 vector unsigned char);
12158 vector signed short vec_mladd (vector signed short,
12159 vector signed short,
12160 vector signed short);
12161 vector signed short vec_mladd (vector signed short,
12162 vector unsigned short,
12163 vector unsigned short);
12164 vector signed short vec_mladd (vector unsigned short,
12165 vector signed short,
12166 vector signed short);
12167 vector unsigned short vec_mladd (vector unsigned short,
12168 vector unsigned short,
12169 vector unsigned short);
12171 vector signed short vec_mradds (vector signed short,
12172 vector signed short,
12173 vector signed short);
12175 vector unsigned int vec_msum (vector unsigned char,
12176 vector unsigned char,
12177 vector unsigned int);
12178 vector signed int vec_msum (vector signed char,
12179 vector unsigned char,
12180 vector signed int);
12181 vector unsigned int vec_msum (vector unsigned short,
12182 vector unsigned short,
12183 vector unsigned int);
12184 vector signed int vec_msum (vector signed short,
12185 vector signed short,
12186 vector signed int);
12188 vector signed int vec_vmsumshm (vector signed short,
12189 vector signed short,
12190 vector signed int);
12192 vector unsigned int vec_vmsumuhm (vector unsigned short,
12193 vector unsigned short,
12194 vector unsigned int);
12196 vector signed int vec_vmsummbm (vector signed char,
12197 vector unsigned char,
12198 vector signed int);
12200 vector unsigned int vec_vmsumubm (vector unsigned char,
12201 vector unsigned char,
12202 vector unsigned int);
12204 vector unsigned int vec_msums (vector unsigned short,
12205 vector unsigned short,
12206 vector unsigned int);
12207 vector signed int vec_msums (vector signed short,
12208 vector signed short,
12209 vector signed int);
12211 vector signed int vec_vmsumshs (vector signed short,
12212 vector signed short,
12213 vector signed int);
12215 vector unsigned int vec_vmsumuhs (vector unsigned short,
12216 vector unsigned short,
12217 vector unsigned int);
12219 void vec_mtvscr (vector signed int);
12220 void vec_mtvscr (vector unsigned int);
12221 void vec_mtvscr (vector bool int);
12222 void vec_mtvscr (vector signed short);
12223 void vec_mtvscr (vector unsigned short);
12224 void vec_mtvscr (vector bool short);
12225 void vec_mtvscr (vector pixel);
12226 void vec_mtvscr (vector signed char);
12227 void vec_mtvscr (vector unsigned char);
12228 void vec_mtvscr (vector bool char);
12230 vector unsigned short vec_mule (vector unsigned char,
12231 vector unsigned char);
12232 vector signed short vec_mule (vector signed char,
12233 vector signed char);
12234 vector unsigned int vec_mule (vector unsigned short,
12235 vector unsigned short);
12236 vector signed int vec_mule (vector signed short, vector signed short);
12238 vector signed int vec_vmulesh (vector signed short,
12239 vector signed short);
12241 vector unsigned int vec_vmuleuh (vector unsigned short,
12242 vector unsigned short);
12244 vector signed short vec_vmulesb (vector signed char,
12245 vector signed char);
12247 vector unsigned short vec_vmuleub (vector unsigned char,
12248 vector unsigned char);
12250 vector unsigned short vec_mulo (vector unsigned char,
12251 vector unsigned char);
12252 vector signed short vec_mulo (vector signed char, vector signed char);
12253 vector unsigned int vec_mulo (vector unsigned short,
12254 vector unsigned short);
12255 vector signed int vec_mulo (vector signed short, vector signed short);
12257 vector signed int vec_vmulosh (vector signed short,
12258 vector signed short);
12260 vector unsigned int vec_vmulouh (vector unsigned short,
12261 vector unsigned short);
12263 vector signed short vec_vmulosb (vector signed char,
12264 vector signed char);
12266 vector unsigned short vec_vmuloub (vector unsigned char,
12267 vector unsigned char);
12269 vector float vec_nmsub (vector float, vector float, vector float);
12271 vector float vec_nor (vector float, vector float);
12272 vector signed int vec_nor (vector signed int, vector signed int);
12273 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
12274 vector bool int vec_nor (vector bool int, vector bool int);
12275 vector signed short vec_nor (vector signed short, vector signed short);
12276 vector unsigned short vec_nor (vector unsigned short,
12277 vector unsigned short);
12278 vector bool short vec_nor (vector bool short, vector bool short);
12279 vector signed char vec_nor (vector signed char, vector signed char);
12280 vector unsigned char vec_nor (vector unsigned char,
12281 vector unsigned char);
12282 vector bool char vec_nor (vector bool char, vector bool char);
12284 vector float vec_or (vector float, vector float);
12285 vector float vec_or (vector float, vector bool int);
12286 vector float vec_or (vector bool int, vector float);
12287 vector bool int vec_or (vector bool int, vector bool int);
12288 vector signed int vec_or (vector bool int, vector signed int);
12289 vector signed int vec_or (vector signed int, vector bool int);
12290 vector signed int vec_or (vector signed int, vector signed int);
12291 vector unsigned int vec_or (vector bool int, vector unsigned int);
12292 vector unsigned int vec_or (vector unsigned int, vector bool int);
12293 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
12294 vector bool short vec_or (vector bool short, vector bool short);
12295 vector signed short vec_or (vector bool short, vector signed short);
12296 vector signed short vec_or (vector signed short, vector bool short);
12297 vector signed short vec_or (vector signed short, vector signed short);
12298 vector unsigned short vec_or (vector bool short, vector unsigned short);
12299 vector unsigned short vec_or (vector unsigned short, vector bool short);
12300 vector unsigned short vec_or (vector unsigned short,
12301 vector unsigned short);
12302 vector signed char vec_or (vector bool char, vector signed char);
12303 vector bool char vec_or (vector bool char, vector bool char);
12304 vector signed char vec_or (vector signed char, vector bool char);
12305 vector signed char vec_or (vector signed char, vector signed char);
12306 vector unsigned char vec_or (vector bool char, vector unsigned char);
12307 vector unsigned char vec_or (vector unsigned char, vector bool char);
12308 vector unsigned char vec_or (vector unsigned char,
12309 vector unsigned char);
12311 vector signed char vec_pack (vector signed short, vector signed short);
12312 vector unsigned char vec_pack (vector unsigned short,
12313 vector unsigned short);
12314 vector bool char vec_pack (vector bool short, vector bool short);
12315 vector signed short vec_pack (vector signed int, vector signed int);
12316 vector unsigned short vec_pack (vector unsigned int,
12317 vector unsigned int);
12318 vector bool short vec_pack (vector bool int, vector bool int);
12320 vector bool short vec_vpkuwum (vector bool int, vector bool int);
12321 vector signed short vec_vpkuwum (vector signed int, vector signed int);
12322 vector unsigned short vec_vpkuwum (vector unsigned int,
12323 vector unsigned int);
12325 vector bool char vec_vpkuhum (vector bool short, vector bool short);
12326 vector signed char vec_vpkuhum (vector signed short,
12327 vector signed short);
12328 vector unsigned char vec_vpkuhum (vector unsigned short,
12329 vector unsigned short);
12331 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
12333 vector unsigned char vec_packs (vector unsigned short,
12334 vector unsigned short);
12335 vector signed char vec_packs (vector signed short, vector signed short);
12336 vector unsigned short vec_packs (vector unsigned int,
12337 vector unsigned int);
12338 vector signed short vec_packs (vector signed int, vector signed int);
12340 vector signed short vec_vpkswss (vector signed int, vector signed int);
12342 vector unsigned short vec_vpkuwus (vector unsigned int,
12343 vector unsigned int);
12345 vector signed char vec_vpkshss (vector signed short,
12346 vector signed short);
12348 vector unsigned char vec_vpkuhus (vector unsigned short,
12349 vector unsigned short);
12351 vector unsigned char vec_packsu (vector unsigned short,
12352 vector unsigned short);
12353 vector unsigned char vec_packsu (vector signed short,
12354 vector signed short);
12355 vector unsigned short vec_packsu (vector unsigned int,
12356 vector unsigned int);
12357 vector unsigned short vec_packsu (vector signed int, vector signed int);
12359 vector unsigned short vec_vpkswus (vector signed int,
12360 vector signed int);
12362 vector unsigned char vec_vpkshus (vector signed short,
12363 vector signed short);
12365 vector float vec_perm (vector float,
12367 vector unsigned char);
12368 vector signed int vec_perm (vector signed int,
12370 vector unsigned char);
12371 vector unsigned int vec_perm (vector unsigned int,
12372 vector unsigned int,
12373 vector unsigned char);
12374 vector bool int vec_perm (vector bool int,
12376 vector unsigned char);
12377 vector signed short vec_perm (vector signed short,
12378 vector signed short,
12379 vector unsigned char);
12380 vector unsigned short vec_perm (vector unsigned short,
12381 vector unsigned short,
12382 vector unsigned char);
12383 vector bool short vec_perm (vector bool short,
12385 vector unsigned char);
12386 vector pixel vec_perm (vector pixel,
12388 vector unsigned char);
12389 vector signed char vec_perm (vector signed char,
12390 vector signed char,
12391 vector unsigned char);
12392 vector unsigned char vec_perm (vector unsigned char,
12393 vector unsigned char,
12394 vector unsigned char);
12395 vector bool char vec_perm (vector bool char,
12397 vector unsigned char);
12399 vector float vec_re (vector float);
12401 vector signed char vec_rl (vector signed char,
12402 vector unsigned char);
12403 vector unsigned char vec_rl (vector unsigned char,
12404 vector unsigned char);
12405 vector signed short vec_rl (vector signed short, vector unsigned short);
12406 vector unsigned short vec_rl (vector unsigned short,
12407 vector unsigned short);
12408 vector signed int vec_rl (vector signed int, vector unsigned int);
12409 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
12411 vector signed int vec_vrlw (vector signed int, vector unsigned int);
12412 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
12414 vector signed short vec_vrlh (vector signed short,
12415 vector unsigned short);
12416 vector unsigned short vec_vrlh (vector unsigned short,
12417 vector unsigned short);
12419 vector signed char vec_vrlb (vector signed char, vector unsigned char);
12420 vector unsigned char vec_vrlb (vector unsigned char,
12421 vector unsigned char);
12423 vector float vec_round (vector float);
12425 vector float vec_recip (vector float, vector float);
12427 vector float vec_rsqrt (vector float);
12429 vector float vec_rsqrte (vector float);
12431 vector float vec_sel (vector float, vector float, vector bool int);
12432 vector float vec_sel (vector float, vector float, vector unsigned int);
12433 vector signed int vec_sel (vector signed int,
12436 vector signed int vec_sel (vector signed int,
12438 vector unsigned int);
12439 vector unsigned int vec_sel (vector unsigned int,
12440 vector unsigned int,
12442 vector unsigned int vec_sel (vector unsigned int,
12443 vector unsigned int,
12444 vector unsigned int);
12445 vector bool int vec_sel (vector bool int,
12448 vector bool int vec_sel (vector bool int,
12450 vector unsigned int);
12451 vector signed short vec_sel (vector signed short,
12452 vector signed short,
12453 vector bool short);
12454 vector signed short vec_sel (vector signed short,
12455 vector signed short,
12456 vector unsigned short);
12457 vector unsigned short vec_sel (vector unsigned short,
12458 vector unsigned short,
12459 vector bool short);
12460 vector unsigned short vec_sel (vector unsigned short,
12461 vector unsigned short,
12462 vector unsigned short);
12463 vector bool short vec_sel (vector bool short,
12465 vector bool short);
12466 vector bool short vec_sel (vector bool short,
12468 vector unsigned short);
12469 vector signed char vec_sel (vector signed char,
12470 vector signed char,
12472 vector signed char vec_sel (vector signed char,
12473 vector signed char,
12474 vector unsigned char);
12475 vector unsigned char vec_sel (vector unsigned char,
12476 vector unsigned char,
12478 vector unsigned char vec_sel (vector unsigned char,
12479 vector unsigned char,
12480 vector unsigned char);
12481 vector bool char vec_sel (vector bool char,
12484 vector bool char vec_sel (vector bool char,
12486 vector unsigned char);
12488 vector signed char vec_sl (vector signed char,
12489 vector unsigned char);
12490 vector unsigned char vec_sl (vector unsigned char,
12491 vector unsigned char);
12492 vector signed short vec_sl (vector signed short, vector unsigned short);
12493 vector unsigned short vec_sl (vector unsigned short,
12494 vector unsigned short);
12495 vector signed int vec_sl (vector signed int, vector unsigned int);
12496 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
12498 vector signed int vec_vslw (vector signed int, vector unsigned int);
12499 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
12501 vector signed short vec_vslh (vector signed short,
12502 vector unsigned short);
12503 vector unsigned short vec_vslh (vector unsigned short,
12504 vector unsigned short);
12506 vector signed char vec_vslb (vector signed char, vector unsigned char);
12507 vector unsigned char vec_vslb (vector unsigned char,
12508 vector unsigned char);
12510 vector float vec_sld (vector float, vector float, const int);
12511 vector signed int vec_sld (vector signed int,
12514 vector unsigned int vec_sld (vector unsigned int,
12515 vector unsigned int,
12517 vector bool int vec_sld (vector bool int,
12520 vector signed short vec_sld (vector signed short,
12521 vector signed short,
12523 vector unsigned short vec_sld (vector unsigned short,
12524 vector unsigned short,
12526 vector bool short vec_sld (vector bool short,
12529 vector pixel vec_sld (vector pixel,
12532 vector signed char vec_sld (vector signed char,
12533 vector signed char,
12535 vector unsigned char vec_sld (vector unsigned char,
12536 vector unsigned char,
12538 vector bool char vec_sld (vector bool char,
12542 vector signed int vec_sll (vector signed int,
12543 vector unsigned int);
12544 vector signed int vec_sll (vector signed int,
12545 vector unsigned short);
12546 vector signed int vec_sll (vector signed int,
12547 vector unsigned char);
12548 vector unsigned int vec_sll (vector unsigned int,
12549 vector unsigned int);
12550 vector unsigned int vec_sll (vector unsigned int,
12551 vector unsigned short);
12552 vector unsigned int vec_sll (vector unsigned int,
12553 vector unsigned char);
12554 vector bool int vec_sll (vector bool int,
12555 vector unsigned int);
12556 vector bool int vec_sll (vector bool int,
12557 vector unsigned short);
12558 vector bool int vec_sll (vector bool int,
12559 vector unsigned char);
12560 vector signed short vec_sll (vector signed short,
12561 vector unsigned int);
12562 vector signed short vec_sll (vector signed short,
12563 vector unsigned short);
12564 vector signed short vec_sll (vector signed short,
12565 vector unsigned char);
12566 vector unsigned short vec_sll (vector unsigned short,
12567 vector unsigned int);
12568 vector unsigned short vec_sll (vector unsigned short,
12569 vector unsigned short);
12570 vector unsigned short vec_sll (vector unsigned short,
12571 vector unsigned char);
12572 vector bool short vec_sll (vector bool short, vector unsigned int);
12573 vector bool short vec_sll (vector bool short, vector unsigned short);
12574 vector bool short vec_sll (vector bool short, vector unsigned char);
12575 vector pixel vec_sll (vector pixel, vector unsigned int);
12576 vector pixel vec_sll (vector pixel, vector unsigned short);
12577 vector pixel vec_sll (vector pixel, vector unsigned char);
12578 vector signed char vec_sll (vector signed char, vector unsigned int);
12579 vector signed char vec_sll (vector signed char, vector unsigned short);
12580 vector signed char vec_sll (vector signed char, vector unsigned char);
12581 vector unsigned char vec_sll (vector unsigned char,
12582 vector unsigned int);
12583 vector unsigned char vec_sll (vector unsigned char,
12584 vector unsigned short);
12585 vector unsigned char vec_sll (vector unsigned char,
12586 vector unsigned char);
12587 vector bool char vec_sll (vector bool char, vector unsigned int);
12588 vector bool char vec_sll (vector bool char, vector unsigned short);
12589 vector bool char vec_sll (vector bool char, vector unsigned char);
12591 vector float vec_slo (vector float, vector signed char);
12592 vector float vec_slo (vector float, vector unsigned char);
12593 vector signed int vec_slo (vector signed int, vector signed char);
12594 vector signed int vec_slo (vector signed int, vector unsigned char);
12595 vector unsigned int vec_slo (vector unsigned int, vector signed char);
12596 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
12597 vector signed short vec_slo (vector signed short, vector signed char);
12598 vector signed short vec_slo (vector signed short, vector unsigned char);
12599 vector unsigned short vec_slo (vector unsigned short,
12600 vector signed char);
12601 vector unsigned short vec_slo (vector unsigned short,
12602 vector unsigned char);
12603 vector pixel vec_slo (vector pixel, vector signed char);
12604 vector pixel vec_slo (vector pixel, vector unsigned char);
12605 vector signed char vec_slo (vector signed char, vector signed char);
12606 vector signed char vec_slo (vector signed char, vector unsigned char);
12607 vector unsigned char vec_slo (vector unsigned char, vector signed char);
12608 vector unsigned char vec_slo (vector unsigned char,
12609 vector unsigned char);
12611 vector signed char vec_splat (vector signed char, const int);
12612 vector unsigned char vec_splat (vector unsigned char, const int);
12613 vector bool char vec_splat (vector bool char, const int);
12614 vector signed short vec_splat (vector signed short, const int);
12615 vector unsigned short vec_splat (vector unsigned short, const int);
12616 vector bool short vec_splat (vector bool short, const int);
12617 vector pixel vec_splat (vector pixel, const int);
12618 vector float vec_splat (vector float, const int);
12619 vector signed int vec_splat (vector signed int, const int);
12620 vector unsigned int vec_splat (vector unsigned int, const int);
12621 vector bool int vec_splat (vector bool int, const int);
12623 vector float vec_vspltw (vector float, const int);
12624 vector signed int vec_vspltw (vector signed int, const int);
12625 vector unsigned int vec_vspltw (vector unsigned int, const int);
12626 vector bool int vec_vspltw (vector bool int, const int);
12628 vector bool short vec_vsplth (vector bool short, const int);
12629 vector signed short vec_vsplth (vector signed short, const int);
12630 vector unsigned short vec_vsplth (vector unsigned short, const int);
12631 vector pixel vec_vsplth (vector pixel, const int);
12633 vector signed char vec_vspltb (vector signed char, const int);
12634 vector unsigned char vec_vspltb (vector unsigned char, const int);
12635 vector bool char vec_vspltb (vector bool char, const int);
12637 vector signed char vec_splat_s8 (const int);
12639 vector signed short vec_splat_s16 (const int);
12641 vector signed int vec_splat_s32 (const int);
12643 vector unsigned char vec_splat_u8 (const int);
12645 vector unsigned short vec_splat_u16 (const int);
12647 vector unsigned int vec_splat_u32 (const int);
12649 vector signed char vec_sr (vector signed char, vector unsigned char);
12650 vector unsigned char vec_sr (vector unsigned char,
12651 vector unsigned char);
12652 vector signed short vec_sr (vector signed short,
12653 vector unsigned short);
12654 vector unsigned short vec_sr (vector unsigned short,
12655 vector unsigned short);
12656 vector signed int vec_sr (vector signed int, vector unsigned int);
12657 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
12659 vector signed int vec_vsrw (vector signed int, vector unsigned int);
12660 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
12662 vector signed short vec_vsrh (vector signed short,
12663 vector unsigned short);
12664 vector unsigned short vec_vsrh (vector unsigned short,
12665 vector unsigned short);
12667 vector signed char vec_vsrb (vector signed char, vector unsigned char);
12668 vector unsigned char vec_vsrb (vector unsigned char,
12669 vector unsigned char);
12671 vector signed char vec_sra (vector signed char, vector unsigned char);
12672 vector unsigned char vec_sra (vector unsigned char,
12673 vector unsigned char);
12674 vector signed short vec_sra (vector signed short,
12675 vector unsigned short);
12676 vector unsigned short vec_sra (vector unsigned short,
12677 vector unsigned short);
12678 vector signed int vec_sra (vector signed int, vector unsigned int);
12679 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
12681 vector signed int vec_vsraw (vector signed int, vector unsigned int);
12682 vector unsigned int vec_vsraw (vector unsigned int,
12683 vector unsigned int);
12685 vector signed short vec_vsrah (vector signed short,
12686 vector unsigned short);
12687 vector unsigned short vec_vsrah (vector unsigned short,
12688 vector unsigned short);
12690 vector signed char vec_vsrab (vector signed char, vector unsigned char);
12691 vector unsigned char vec_vsrab (vector unsigned char,
12692 vector unsigned char);
12694 vector signed int vec_srl (vector signed int, vector unsigned int);
12695 vector signed int vec_srl (vector signed int, vector unsigned short);
12696 vector signed int vec_srl (vector signed int, vector unsigned char);
12697 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
12698 vector unsigned int vec_srl (vector unsigned int,
12699 vector unsigned short);
12700 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
12701 vector bool int vec_srl (vector bool int, vector unsigned int);
12702 vector bool int vec_srl (vector bool int, vector unsigned short);
12703 vector bool int vec_srl (vector bool int, vector unsigned char);
12704 vector signed short vec_srl (vector signed short, vector unsigned int);
12705 vector signed short vec_srl (vector signed short,
12706 vector unsigned short);
12707 vector signed short vec_srl (vector signed short, vector unsigned char);
12708 vector unsigned short vec_srl (vector unsigned short,
12709 vector unsigned int);
12710 vector unsigned short vec_srl (vector unsigned short,
12711 vector unsigned short);
12712 vector unsigned short vec_srl (vector unsigned short,
12713 vector unsigned char);
12714 vector bool short vec_srl (vector bool short, vector unsigned int);
12715 vector bool short vec_srl (vector bool short, vector unsigned short);
12716 vector bool short vec_srl (vector bool short, vector unsigned char);
12717 vector pixel vec_srl (vector pixel, vector unsigned int);
12718 vector pixel vec_srl (vector pixel, vector unsigned short);
12719 vector pixel vec_srl (vector pixel, vector unsigned char);
12720 vector signed char vec_srl (vector signed char, vector unsigned int);
12721 vector signed char vec_srl (vector signed char, vector unsigned short);
12722 vector signed char vec_srl (vector signed char, vector unsigned char);
12723 vector unsigned char vec_srl (vector unsigned char,
12724 vector unsigned int);
12725 vector unsigned char vec_srl (vector unsigned char,
12726 vector unsigned short);
12727 vector unsigned char vec_srl (vector unsigned char,
12728 vector unsigned char);
12729 vector bool char vec_srl (vector bool char, vector unsigned int);
12730 vector bool char vec_srl (vector bool char, vector unsigned short);
12731 vector bool char vec_srl (vector bool char, vector unsigned char);
12733 vector float vec_sro (vector float, vector signed char);
12734 vector float vec_sro (vector float, vector unsigned char);
12735 vector signed int vec_sro (vector signed int, vector signed char);
12736 vector signed int vec_sro (vector signed int, vector unsigned char);
12737 vector unsigned int vec_sro (vector unsigned int, vector signed char);
12738 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
12739 vector signed short vec_sro (vector signed short, vector signed char);
12740 vector signed short vec_sro (vector signed short, vector unsigned char);
12741 vector unsigned short vec_sro (vector unsigned short,
12742 vector signed char);
12743 vector unsigned short vec_sro (vector unsigned short,
12744 vector unsigned char);
12745 vector pixel vec_sro (vector pixel, vector signed char);
12746 vector pixel vec_sro (vector pixel, vector unsigned char);
12747 vector signed char vec_sro (vector signed char, vector signed char);
12748 vector signed char vec_sro (vector signed char, vector unsigned char);
12749 vector unsigned char vec_sro (vector unsigned char, vector signed char);
12750 vector unsigned char vec_sro (vector unsigned char,
12751 vector unsigned char);
12753 void vec_st (vector float, int, vector float *);
12754 void vec_st (vector float, int, float *);
12755 void vec_st (vector signed int, int, vector signed int *);
12756 void vec_st (vector signed int, int, int *);
12757 void vec_st (vector unsigned int, int, vector unsigned int *);
12758 void vec_st (vector unsigned int, int, unsigned int *);
12759 void vec_st (vector bool int, int, vector bool int *);
12760 void vec_st (vector bool int, int, unsigned int *);
12761 void vec_st (vector bool int, int, int *);
12762 void vec_st (vector signed short, int, vector signed short *);
12763 void vec_st (vector signed short, int, short *);
12764 void vec_st (vector unsigned short, int, vector unsigned short *);
12765 void vec_st (vector unsigned short, int, unsigned short *);
12766 void vec_st (vector bool short, int, vector bool short *);
12767 void vec_st (vector bool short, int, unsigned short *);
12768 void vec_st (vector pixel, int, vector pixel *);
12769 void vec_st (vector pixel, int, unsigned short *);
12770 void vec_st (vector pixel, int, short *);
12771 void vec_st (vector bool short, int, short *);
12772 void vec_st (vector signed char, int, vector signed char *);
12773 void vec_st (vector signed char, int, signed char *);
12774 void vec_st (vector unsigned char, int, vector unsigned char *);
12775 void vec_st (vector unsigned char, int, unsigned char *);
12776 void vec_st (vector bool char, int, vector bool char *);
12777 void vec_st (vector bool char, int, unsigned char *);
12778 void vec_st (vector bool char, int, signed char *);
12780 void vec_ste (vector signed char, int, signed char *);
12781 void vec_ste (vector unsigned char, int, unsigned char *);
12782 void vec_ste (vector bool char, int, signed char *);
12783 void vec_ste (vector bool char, int, unsigned char *);
12784 void vec_ste (vector signed short, int, short *);
12785 void vec_ste (vector unsigned short, int, unsigned short *);
12786 void vec_ste (vector bool short, int, short *);
12787 void vec_ste (vector bool short, int, unsigned short *);
12788 void vec_ste (vector pixel, int, short *);
12789 void vec_ste (vector pixel, int, unsigned short *);
12790 void vec_ste (vector float, int, float *);
12791 void vec_ste (vector signed int, int, int *);
12792 void vec_ste (vector unsigned int, int, unsigned int *);
12793 void vec_ste (vector bool int, int, int *);
12794 void vec_ste (vector bool int, int, unsigned int *);
12796 void vec_stvewx (vector float, int, float *);
12797 void vec_stvewx (vector signed int, int, int *);
12798 void vec_stvewx (vector unsigned int, int, unsigned int *);
12799 void vec_stvewx (vector bool int, int, int *);
12800 void vec_stvewx (vector bool int, int, unsigned int *);
12802 void vec_stvehx (vector signed short, int, short *);
12803 void vec_stvehx (vector unsigned short, int, unsigned short *);
12804 void vec_stvehx (vector bool short, int, short *);
12805 void vec_stvehx (vector bool short, int, unsigned short *);
12806 void vec_stvehx (vector pixel, int, short *);
12807 void vec_stvehx (vector pixel, int, unsigned short *);
12809 void vec_stvebx (vector signed char, int, signed char *);
12810 void vec_stvebx (vector unsigned char, int, unsigned char *);
12811 void vec_stvebx (vector bool char, int, signed char *);
12812 void vec_stvebx (vector bool char, int, unsigned char *);
12814 void vec_stl (vector float, int, vector float *);
12815 void vec_stl (vector float, int, float *);
12816 void vec_stl (vector signed int, int, vector signed int *);
12817 void vec_stl (vector signed int, int, int *);
12818 void vec_stl (vector unsigned int, int, vector unsigned int *);
12819 void vec_stl (vector unsigned int, int, unsigned int *);
12820 void vec_stl (vector bool int, int, vector bool int *);
12821 void vec_stl (vector bool int, int, unsigned int *);
12822 void vec_stl (vector bool int, int, int *);
12823 void vec_stl (vector signed short, int, vector signed short *);
12824 void vec_stl (vector signed short, int, short *);
12825 void vec_stl (vector unsigned short, int, vector unsigned short *);
12826 void vec_stl (vector unsigned short, int, unsigned short *);
12827 void vec_stl (vector bool short, int, vector bool short *);
12828 void vec_stl (vector bool short, int, unsigned short *);
12829 void vec_stl (vector bool short, int, short *);
12830 void vec_stl (vector pixel, int, vector pixel *);
12831 void vec_stl (vector pixel, int, unsigned short *);
12832 void vec_stl (vector pixel, int, short *);
12833 void vec_stl (vector signed char, int, vector signed char *);
12834 void vec_stl (vector signed char, int, signed char *);
12835 void vec_stl (vector unsigned char, int, vector unsigned char *);
12836 void vec_stl (vector unsigned char, int, unsigned char *);
12837 void vec_stl (vector bool char, int, vector bool char *);
12838 void vec_stl (vector bool char, int, unsigned char *);
12839 void vec_stl (vector bool char, int, signed char *);
12841 vector signed char vec_sub (vector bool char, vector signed char);
12842 vector signed char vec_sub (vector signed char, vector bool char);
12843 vector signed char vec_sub (vector signed char, vector signed char);
12844 vector unsigned char vec_sub (vector bool char, vector unsigned char);
12845 vector unsigned char vec_sub (vector unsigned char, vector bool char);
12846 vector unsigned char vec_sub (vector unsigned char,
12847 vector unsigned char);
12848 vector signed short vec_sub (vector bool short, vector signed short);
12849 vector signed short vec_sub (vector signed short, vector bool short);
12850 vector signed short vec_sub (vector signed short, vector signed short);
12851 vector unsigned short vec_sub (vector bool short,
12852 vector unsigned short);
12853 vector unsigned short vec_sub (vector unsigned short,
12854 vector bool short);
12855 vector unsigned short vec_sub (vector unsigned short,
12856 vector unsigned short);
12857 vector signed int vec_sub (vector bool int, vector signed int);
12858 vector signed int vec_sub (vector signed int, vector bool int);
12859 vector signed int vec_sub (vector signed int, vector signed int);
12860 vector unsigned int vec_sub (vector bool int, vector unsigned int);
12861 vector unsigned int vec_sub (vector unsigned int, vector bool int);
12862 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
12863 vector float vec_sub (vector float, vector float);
12865 vector float vec_vsubfp (vector float, vector float);
12867 vector signed int vec_vsubuwm (vector bool int, vector signed int);
12868 vector signed int vec_vsubuwm (vector signed int, vector bool int);
12869 vector signed int vec_vsubuwm (vector signed int, vector signed int);
12870 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
12871 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
12872 vector unsigned int vec_vsubuwm (vector unsigned int,
12873 vector unsigned int);
12875 vector signed short vec_vsubuhm (vector bool short,
12876 vector signed short);
12877 vector signed short vec_vsubuhm (vector signed short,
12878 vector bool short);
12879 vector signed short vec_vsubuhm (vector signed short,
12880 vector signed short);
12881 vector unsigned short vec_vsubuhm (vector bool short,
12882 vector unsigned short);
12883 vector unsigned short vec_vsubuhm (vector unsigned short,
12884 vector bool short);
12885 vector unsigned short vec_vsubuhm (vector unsigned short,
12886 vector unsigned short);
12888 vector signed char vec_vsububm (vector bool char, vector signed char);
12889 vector signed char vec_vsububm (vector signed char, vector bool char);
12890 vector signed char vec_vsububm (vector signed char, vector signed char);
12891 vector unsigned char vec_vsububm (vector bool char,
12892 vector unsigned char);
12893 vector unsigned char vec_vsububm (vector unsigned char,
12895 vector unsigned char vec_vsububm (vector unsigned char,
12896 vector unsigned char);
12898 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
12900 vector unsigned char vec_subs (vector bool char, vector unsigned char);
12901 vector unsigned char vec_subs (vector unsigned char, vector bool char);
12902 vector unsigned char vec_subs (vector unsigned char,
12903 vector unsigned char);
12904 vector signed char vec_subs (vector bool char, vector signed char);
12905 vector signed char vec_subs (vector signed char, vector bool char);
12906 vector signed char vec_subs (vector signed char, vector signed char);
12907 vector unsigned short vec_subs (vector bool short,
12908 vector unsigned short);
12909 vector unsigned short vec_subs (vector unsigned short,
12910 vector bool short);
12911 vector unsigned short vec_subs (vector unsigned short,
12912 vector unsigned short);
12913 vector signed short vec_subs (vector bool short, vector signed short);
12914 vector signed short vec_subs (vector signed short, vector bool short);
12915 vector signed short vec_subs (vector signed short, vector signed short);
12916 vector unsigned int vec_subs (vector bool int, vector unsigned int);
12917 vector unsigned int vec_subs (vector unsigned int, vector bool int);
12918 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
12919 vector signed int vec_subs (vector bool int, vector signed int);
12920 vector signed int vec_subs (vector signed int, vector bool int);
12921 vector signed int vec_subs (vector signed int, vector signed int);
12923 vector signed int vec_vsubsws (vector bool int, vector signed int);
12924 vector signed int vec_vsubsws (vector signed int, vector bool int);
12925 vector signed int vec_vsubsws (vector signed int, vector signed int);
12927 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
12928 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
12929 vector unsigned int vec_vsubuws (vector unsigned int,
12930 vector unsigned int);
12932 vector signed short vec_vsubshs (vector bool short,
12933 vector signed short);
12934 vector signed short vec_vsubshs (vector signed short,
12935 vector bool short);
12936 vector signed short vec_vsubshs (vector signed short,
12937 vector signed short);
12939 vector unsigned short vec_vsubuhs (vector bool short,
12940 vector unsigned short);
12941 vector unsigned short vec_vsubuhs (vector unsigned short,
12942 vector bool short);
12943 vector unsigned short vec_vsubuhs (vector unsigned short,
12944 vector unsigned short);
12946 vector signed char vec_vsubsbs (vector bool char, vector signed char);
12947 vector signed char vec_vsubsbs (vector signed char, vector bool char);
12948 vector signed char vec_vsubsbs (vector signed char, vector signed char);
12950 vector unsigned char vec_vsububs (vector bool char,
12951 vector unsigned char);
12952 vector unsigned char vec_vsububs (vector unsigned char,
12954 vector unsigned char vec_vsububs (vector unsigned char,
12955 vector unsigned char);
12957 vector unsigned int vec_sum4s (vector unsigned char,
12958 vector unsigned int);
12959 vector signed int vec_sum4s (vector signed char, vector signed int);
12960 vector signed int vec_sum4s (vector signed short, vector signed int);
12962 vector signed int vec_vsum4shs (vector signed short, vector signed int);
12964 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
12966 vector unsigned int vec_vsum4ubs (vector unsigned char,
12967 vector unsigned int);
12969 vector signed int vec_sum2s (vector signed int, vector signed int);
12971 vector signed int vec_sums (vector signed int, vector signed int);
12973 vector float vec_trunc (vector float);
12975 vector signed short vec_unpackh (vector signed char);
12976 vector bool short vec_unpackh (vector bool char);
12977 vector signed int vec_unpackh (vector signed short);
12978 vector bool int vec_unpackh (vector bool short);
12979 vector unsigned int vec_unpackh (vector pixel);
12981 vector bool int vec_vupkhsh (vector bool short);
12982 vector signed int vec_vupkhsh (vector signed short);
12984 vector unsigned int vec_vupkhpx (vector pixel);
12986 vector bool short vec_vupkhsb (vector bool char);
12987 vector signed short vec_vupkhsb (vector signed char);
12989 vector signed short vec_unpackl (vector signed char);
12990 vector bool short vec_unpackl (vector bool char);
12991 vector unsigned int vec_unpackl (vector pixel);
12992 vector signed int vec_unpackl (vector signed short);
12993 vector bool int vec_unpackl (vector bool short);
12995 vector unsigned int vec_vupklpx (vector pixel);
12997 vector bool int vec_vupklsh (vector bool short);
12998 vector signed int vec_vupklsh (vector signed short);
13000 vector bool short vec_vupklsb (vector bool char);
13001 vector signed short vec_vupklsb (vector signed char);
13003 vector float vec_xor (vector float, vector float);
13004 vector float vec_xor (vector float, vector bool int);
13005 vector float vec_xor (vector bool int, vector float);
13006 vector bool int vec_xor (vector bool int, vector bool int);
13007 vector signed int vec_xor (vector bool int, vector signed int);
13008 vector signed int vec_xor (vector signed int, vector bool int);
13009 vector signed int vec_xor (vector signed int, vector signed int);
13010 vector unsigned int vec_xor (vector bool int, vector unsigned int);
13011 vector unsigned int vec_xor (vector unsigned int, vector bool int);
13012 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
13013 vector bool short vec_xor (vector bool short, vector bool short);
13014 vector signed short vec_xor (vector bool short, vector signed short);
13015 vector signed short vec_xor (vector signed short, vector bool short);
13016 vector signed short vec_xor (vector signed short, vector signed short);
13017 vector unsigned short vec_xor (vector bool short,
13018 vector unsigned short);
13019 vector unsigned short vec_xor (vector unsigned short,
13020 vector bool short);
13021 vector unsigned short vec_xor (vector unsigned short,
13022 vector unsigned short);
13023 vector signed char vec_xor (vector bool char, vector signed char);
13024 vector bool char vec_xor (vector bool char, vector bool char);
13025 vector signed char vec_xor (vector signed char, vector bool char);
13026 vector signed char vec_xor (vector signed char, vector signed char);
13027 vector unsigned char vec_xor (vector bool char, vector unsigned char);
13028 vector unsigned char vec_xor (vector unsigned char, vector bool char);
13029 vector unsigned char vec_xor (vector unsigned char,
13030 vector unsigned char);
13032 int vec_all_eq (vector signed char, vector bool char);
13033 int vec_all_eq (vector signed char, vector signed char);
13034 int vec_all_eq (vector unsigned char, vector bool char);
13035 int vec_all_eq (vector unsigned char, vector unsigned char);
13036 int vec_all_eq (vector bool char, vector bool char);
13037 int vec_all_eq (vector bool char, vector unsigned char);
13038 int vec_all_eq (vector bool char, vector signed char);
13039 int vec_all_eq (vector signed short, vector bool short);
13040 int vec_all_eq (vector signed short, vector signed short);
13041 int vec_all_eq (vector unsigned short, vector bool short);
13042 int vec_all_eq (vector unsigned short, vector unsigned short);
13043 int vec_all_eq (vector bool short, vector bool short);
13044 int vec_all_eq (vector bool short, vector unsigned short);
13045 int vec_all_eq (vector bool short, vector signed short);
13046 int vec_all_eq (vector pixel, vector pixel);
13047 int vec_all_eq (vector signed int, vector bool int);
13048 int vec_all_eq (vector signed int, vector signed int);
13049 int vec_all_eq (vector unsigned int, vector bool int);
13050 int vec_all_eq (vector unsigned int, vector unsigned int);
13051 int vec_all_eq (vector bool int, vector bool int);
13052 int vec_all_eq (vector bool int, vector unsigned int);
13053 int vec_all_eq (vector bool int, vector signed int);
13054 int vec_all_eq (vector float, vector float);
13056 int vec_all_ge (vector bool char, vector unsigned char);
13057 int vec_all_ge (vector unsigned char, vector bool char);
13058 int vec_all_ge (vector unsigned char, vector unsigned char);
13059 int vec_all_ge (vector bool char, vector signed char);
13060 int vec_all_ge (vector signed char, vector bool char);
13061 int vec_all_ge (vector signed char, vector signed char);
13062 int vec_all_ge (vector bool short, vector unsigned short);
13063 int vec_all_ge (vector unsigned short, vector bool short);
13064 int vec_all_ge (vector unsigned short, vector unsigned short);
13065 int vec_all_ge (vector signed short, vector signed short);
13066 int vec_all_ge (vector bool short, vector signed short);
13067 int vec_all_ge (vector signed short, vector bool short);
13068 int vec_all_ge (vector bool int, vector unsigned int);
13069 int vec_all_ge (vector unsigned int, vector bool int);
13070 int vec_all_ge (vector unsigned int, vector unsigned int);
13071 int vec_all_ge (vector bool int, vector signed int);
13072 int vec_all_ge (vector signed int, vector bool int);
13073 int vec_all_ge (vector signed int, vector signed int);
13074 int vec_all_ge (vector float, vector float);
13076 int vec_all_gt (vector bool char, vector unsigned char);
13077 int vec_all_gt (vector unsigned char, vector bool char);
13078 int vec_all_gt (vector unsigned char, vector unsigned char);
13079 int vec_all_gt (vector bool char, vector signed char);
13080 int vec_all_gt (vector signed char, vector bool char);
13081 int vec_all_gt (vector signed char, vector signed char);
13082 int vec_all_gt (vector bool short, vector unsigned short);
13083 int vec_all_gt (vector unsigned short, vector bool short);
13084 int vec_all_gt (vector unsigned short, vector unsigned short);
13085 int vec_all_gt (vector bool short, vector signed short);
13086 int vec_all_gt (vector signed short, vector bool short);
13087 int vec_all_gt (vector signed short, vector signed short);
13088 int vec_all_gt (vector bool int, vector unsigned int);
13089 int vec_all_gt (vector unsigned int, vector bool int);
13090 int vec_all_gt (vector unsigned int, vector unsigned int);
13091 int vec_all_gt (vector bool int, vector signed int);
13092 int vec_all_gt (vector signed int, vector bool int);
13093 int vec_all_gt (vector signed int, vector signed int);
13094 int vec_all_gt (vector float, vector float);
13096 int vec_all_in (vector float, vector float);
13098 int vec_all_le (vector bool char, vector unsigned char);
13099 int vec_all_le (vector unsigned char, vector bool char);
13100 int vec_all_le (vector unsigned char, vector unsigned char);
13101 int vec_all_le (vector bool char, vector signed char);
13102 int vec_all_le (vector signed char, vector bool char);
13103 int vec_all_le (vector signed char, vector signed char);
13104 int vec_all_le (vector bool short, vector unsigned short);
13105 int vec_all_le (vector unsigned short, vector bool short);
13106 int vec_all_le (vector unsigned short, vector unsigned short);
13107 int vec_all_le (vector bool short, vector signed short);
13108 int vec_all_le (vector signed short, vector bool short);
13109 int vec_all_le (vector signed short, vector signed short);
13110 int vec_all_le (vector bool int, vector unsigned int);
13111 int vec_all_le (vector unsigned int, vector bool int);
13112 int vec_all_le (vector unsigned int, vector unsigned int);
13113 int vec_all_le (vector bool int, vector signed int);
13114 int vec_all_le (vector signed int, vector bool int);
13115 int vec_all_le (vector signed int, vector signed int);
13116 int vec_all_le (vector float, vector float);
13118 int vec_all_lt (vector bool char, vector unsigned char);
13119 int vec_all_lt (vector unsigned char, vector bool char);
13120 int vec_all_lt (vector unsigned char, vector unsigned char);
13121 int vec_all_lt (vector bool char, vector signed char);
13122 int vec_all_lt (vector signed char, vector bool char);
13123 int vec_all_lt (vector signed char, vector signed char);
13124 int vec_all_lt (vector bool short, vector unsigned short);
13125 int vec_all_lt (vector unsigned short, vector bool short);
13126 int vec_all_lt (vector unsigned short, vector unsigned short);
13127 int vec_all_lt (vector bool short, vector signed short);
13128 int vec_all_lt (vector signed short, vector bool short);
13129 int vec_all_lt (vector signed short, vector signed short);
13130 int vec_all_lt (vector bool int, vector unsigned int);
13131 int vec_all_lt (vector unsigned int, vector bool int);
13132 int vec_all_lt (vector unsigned int, vector unsigned int);
13133 int vec_all_lt (vector bool int, vector signed int);
13134 int vec_all_lt (vector signed int, vector bool int);
13135 int vec_all_lt (vector signed int, vector signed int);
13136 int vec_all_lt (vector float, vector float);
13138 int vec_all_nan (vector float);
13140 int vec_all_ne (vector signed char, vector bool char);
13141 int vec_all_ne (vector signed char, vector signed char);
13142 int vec_all_ne (vector unsigned char, vector bool char);
13143 int vec_all_ne (vector unsigned char, vector unsigned char);
13144 int vec_all_ne (vector bool char, vector bool char);
13145 int vec_all_ne (vector bool char, vector unsigned char);
13146 int vec_all_ne (vector bool char, vector signed char);
13147 int vec_all_ne (vector signed short, vector bool short);
13148 int vec_all_ne (vector signed short, vector signed short);
13149 int vec_all_ne (vector unsigned short, vector bool short);
13150 int vec_all_ne (vector unsigned short, vector unsigned short);
13151 int vec_all_ne (vector bool short, vector bool short);
13152 int vec_all_ne (vector bool short, vector unsigned short);
13153 int vec_all_ne (vector bool short, vector signed short);
13154 int vec_all_ne (vector pixel, vector pixel);
13155 int vec_all_ne (vector signed int, vector bool int);
13156 int vec_all_ne (vector signed int, vector signed int);
13157 int vec_all_ne (vector unsigned int, vector bool int);
13158 int vec_all_ne (vector unsigned int, vector unsigned int);
13159 int vec_all_ne (vector bool int, vector bool int);
13160 int vec_all_ne (vector bool int, vector unsigned int);
13161 int vec_all_ne (vector bool int, vector signed int);
13162 int vec_all_ne (vector float, vector float);
13164 int vec_all_nge (vector float, vector float);
13166 int vec_all_ngt (vector float, vector float);
13168 int vec_all_nle (vector float, vector float);
13170 int vec_all_nlt (vector float, vector float);
13172 int vec_all_numeric (vector float);
13174 int vec_any_eq (vector signed char, vector bool char);
13175 int vec_any_eq (vector signed char, vector signed char);
13176 int vec_any_eq (vector unsigned char, vector bool char);
13177 int vec_any_eq (vector unsigned char, vector unsigned char);
13178 int vec_any_eq (vector bool char, vector bool char);
13179 int vec_any_eq (vector bool char, vector unsigned char);
13180 int vec_any_eq (vector bool char, vector signed char);
13181 int vec_any_eq (vector signed short, vector bool short);
13182 int vec_any_eq (vector signed short, vector signed short);
13183 int vec_any_eq (vector unsigned short, vector bool short);
13184 int vec_any_eq (vector unsigned short, vector unsigned short);
13185 int vec_any_eq (vector bool short, vector bool short);
13186 int vec_any_eq (vector bool short, vector unsigned short);
13187 int vec_any_eq (vector bool short, vector signed short);
13188 int vec_any_eq (vector pixel, vector pixel);
13189 int vec_any_eq (vector signed int, vector bool int);
13190 int vec_any_eq (vector signed int, vector signed int);
13191 int vec_any_eq (vector unsigned int, vector bool int);
13192 int vec_any_eq (vector unsigned int, vector unsigned int);
13193 int vec_any_eq (vector bool int, vector bool int);
13194 int vec_any_eq (vector bool int, vector unsigned int);
13195 int vec_any_eq (vector bool int, vector signed int);
13196 int vec_any_eq (vector float, vector float);
13198 int vec_any_ge (vector signed char, vector bool char);
13199 int vec_any_ge (vector unsigned char, vector bool char);
13200 int vec_any_ge (vector unsigned char, vector unsigned char);
13201 int vec_any_ge (vector signed char, vector signed char);
13202 int vec_any_ge (vector bool char, vector unsigned char);
13203 int vec_any_ge (vector bool char, vector signed char);
13204 int vec_any_ge (vector unsigned short, vector bool short);
13205 int vec_any_ge (vector unsigned short, vector unsigned short);
13206 int vec_any_ge (vector signed short, vector signed short);
13207 int vec_any_ge (vector signed short, vector bool short);
13208 int vec_any_ge (vector bool short, vector unsigned short);
13209 int vec_any_ge (vector bool short, vector signed short);
13210 int vec_any_ge (vector signed int, vector bool int);
13211 int vec_any_ge (vector unsigned int, vector bool int);
13212 int vec_any_ge (vector unsigned int, vector unsigned int);
13213 int vec_any_ge (vector signed int, vector signed int);
13214 int vec_any_ge (vector bool int, vector unsigned int);
13215 int vec_any_ge (vector bool int, vector signed int);
13216 int vec_any_ge (vector float, vector float);
13218 int vec_any_gt (vector bool char, vector unsigned char);
13219 int vec_any_gt (vector unsigned char, vector bool char);
13220 int vec_any_gt (vector unsigned char, vector unsigned char);
13221 int vec_any_gt (vector bool char, vector signed char);
13222 int vec_any_gt (vector signed char, vector bool char);
13223 int vec_any_gt (vector signed char, vector signed char);
13224 int vec_any_gt (vector bool short, vector unsigned short);
13225 int vec_any_gt (vector unsigned short, vector bool short);
13226 int vec_any_gt (vector unsigned short, vector unsigned short);
13227 int vec_any_gt (vector bool short, vector signed short);
13228 int vec_any_gt (vector signed short, vector bool short);
13229 int vec_any_gt (vector signed short, vector signed short);
13230 int vec_any_gt (vector bool int, vector unsigned int);
13231 int vec_any_gt (vector unsigned int, vector bool int);
13232 int vec_any_gt (vector unsigned int, vector unsigned int);
13233 int vec_any_gt (vector bool int, vector signed int);
13234 int vec_any_gt (vector signed int, vector bool int);
13235 int vec_any_gt (vector signed int, vector signed int);
13236 int vec_any_gt (vector float, vector float);
13238 int vec_any_le (vector bool char, vector unsigned char);
13239 int vec_any_le (vector unsigned char, vector bool char);
13240 int vec_any_le (vector unsigned char, vector unsigned char);
13241 int vec_any_le (vector bool char, vector signed char);
13242 int vec_any_le (vector signed char, vector bool char);
13243 int vec_any_le (vector signed char, vector signed char);
13244 int vec_any_le (vector bool short, vector unsigned short);
13245 int vec_any_le (vector unsigned short, vector bool short);
13246 int vec_any_le (vector unsigned short, vector unsigned short);
13247 int vec_any_le (vector bool short, vector signed short);
13248 int vec_any_le (vector signed short, vector bool short);
13249 int vec_any_le (vector signed short, vector signed short);
13250 int vec_any_le (vector bool int, vector unsigned int);
13251 int vec_any_le (vector unsigned int, vector bool int);
13252 int vec_any_le (vector unsigned int, vector unsigned int);
13253 int vec_any_le (vector bool int, vector signed int);
13254 int vec_any_le (vector signed int, vector bool int);
13255 int vec_any_le (vector signed int, vector signed int);
13256 int vec_any_le (vector float, vector float);
13258 int vec_any_lt (vector bool char, vector unsigned char);
13259 int vec_any_lt (vector unsigned char, vector bool char);
13260 int vec_any_lt (vector unsigned char, vector unsigned char);
13261 int vec_any_lt (vector bool char, vector signed char);
13262 int vec_any_lt (vector signed char, vector bool char);
13263 int vec_any_lt (vector signed char, vector signed char);
13264 int vec_any_lt (vector bool short, vector unsigned short);
13265 int vec_any_lt (vector unsigned short, vector bool short);
13266 int vec_any_lt (vector unsigned short, vector unsigned short);
13267 int vec_any_lt (vector bool short, vector signed short);
13268 int vec_any_lt (vector signed short, vector bool short);
13269 int vec_any_lt (vector signed short, vector signed short);
13270 int vec_any_lt (vector bool int, vector unsigned int);
13271 int vec_any_lt (vector unsigned int, vector bool int);
13272 int vec_any_lt (vector unsigned int, vector unsigned int);
13273 int vec_any_lt (vector bool int, vector signed int);
13274 int vec_any_lt (vector signed int, vector bool int);
13275 int vec_any_lt (vector signed int, vector signed int);
13276 int vec_any_lt (vector float, vector float);
13278 int vec_any_nan (vector float);
13280 int vec_any_ne (vector signed char, vector bool char);
13281 int vec_any_ne (vector signed char, vector signed char);
13282 int vec_any_ne (vector unsigned char, vector bool char);
13283 int vec_any_ne (vector unsigned char, vector unsigned char);
13284 int vec_any_ne (vector bool char, vector bool char);
13285 int vec_any_ne (vector bool char, vector unsigned char);
13286 int vec_any_ne (vector bool char, vector signed char);
13287 int vec_any_ne (vector signed short, vector bool short);
13288 int vec_any_ne (vector signed short, vector signed short);
13289 int vec_any_ne (vector unsigned short, vector bool short);
13290 int vec_any_ne (vector unsigned short, vector unsigned short);
13291 int vec_any_ne (vector bool short, vector bool short);
13292 int vec_any_ne (vector bool short, vector unsigned short);
13293 int vec_any_ne (vector bool short, vector signed short);
13294 int vec_any_ne (vector pixel, vector pixel);
13295 int vec_any_ne (vector signed int, vector bool int);
13296 int vec_any_ne (vector signed int, vector signed int);
13297 int vec_any_ne (vector unsigned int, vector bool int);
13298 int vec_any_ne (vector unsigned int, vector unsigned int);
13299 int vec_any_ne (vector bool int, vector bool int);
13300 int vec_any_ne (vector bool int, vector unsigned int);
13301 int vec_any_ne (vector bool int, vector signed int);
13302 int vec_any_ne (vector float, vector float);
13304 int vec_any_nge (vector float, vector float);
13306 int vec_any_ngt (vector float, vector float);
13308 int vec_any_nle (vector float, vector float);
13310 int vec_any_nlt (vector float, vector float);
13312 int vec_any_numeric (vector float);
13314 int vec_any_out (vector float, vector float);
13317 If the vector/scalar (VSX) instruction set is available, the following
13318 additional functions are available:
13321 vector double vec_abs (vector double);
13322 vector double vec_add (vector double, vector double);
13323 vector double vec_and (vector double, vector double);
13324 vector double vec_and (vector double, vector bool long);
13325 vector double vec_and (vector bool long, vector double);
13326 vector double vec_andc (vector double, vector double);
13327 vector double vec_andc (vector double, vector bool long);
13328 vector double vec_andc (vector bool long, vector double);
13329 vector double vec_ceil (vector double);
13330 vector bool long vec_cmpeq (vector double, vector double);
13331 vector bool long vec_cmpge (vector double, vector double);
13332 vector bool long vec_cmpgt (vector double, vector double);
13333 vector bool long vec_cmple (vector double, vector double);
13334 vector bool long vec_cmplt (vector double, vector double);
13335 vector float vec_div (vector float, vector float);
13336 vector double vec_div (vector double, vector double);
13337 vector double vec_floor (vector double);
13338 vector double vec_ld (int, const vector double *);
13339 vector double vec_ld (int, const double *);
13340 vector double vec_ldl (int, const vector double *);
13341 vector double vec_ldl (int, const double *);
13342 vector unsigned char vec_lvsl (int, const volatile double *);
13343 vector unsigned char vec_lvsr (int, const volatile double *);
13344 vector double vec_madd (vector double, vector double, vector double);
13345 vector double vec_max (vector double, vector double);
13346 vector double vec_min (vector double, vector double);
13347 vector float vec_msub (vector float, vector float, vector float);
13348 vector double vec_msub (vector double, vector double, vector double);
13349 vector float vec_mul (vector float, vector float);
13350 vector double vec_mul (vector double, vector double);
13351 vector float vec_nearbyint (vector float);
13352 vector double vec_nearbyint (vector double);
13353 vector float vec_nmadd (vector float, vector float, vector float);
13354 vector double vec_nmadd (vector double, vector double, vector double);
13355 vector double vec_nmsub (vector double, vector double, vector double);
13356 vector double vec_nor (vector double, vector double);
13357 vector double vec_or (vector double, vector double);
13358 vector double vec_or (vector double, vector bool long);
13359 vector double vec_or (vector bool long, vector double);
13360 vector double vec_perm (vector double,
13362 vector unsigned char);
13363 vector double vec_rint (vector double);
13364 vector double vec_recip (vector double, vector double);
13365 vector double vec_rsqrt (vector double);
13366 vector double vec_rsqrte (vector double);
13367 vector double vec_sel (vector double, vector double, vector bool long);
13368 vector double vec_sel (vector double, vector double, vector unsigned long);
13369 vector double vec_sub (vector double, vector double);
13370 vector float vec_sqrt (vector float);
13371 vector double vec_sqrt (vector double);
13372 void vec_st (vector double, int, vector double *);
13373 void vec_st (vector double, int, double *);
13374 vector double vec_trunc (vector double);
13375 vector double vec_xor (vector double, vector double);
13376 vector double vec_xor (vector double, vector bool long);
13377 vector double vec_xor (vector bool long, vector double);
13378 int vec_all_eq (vector double, vector double);
13379 int vec_all_ge (vector double, vector double);
13380 int vec_all_gt (vector double, vector double);
13381 int vec_all_le (vector double, vector double);
13382 int vec_all_lt (vector double, vector double);
13383 int vec_all_nan (vector double);
13384 int vec_all_ne (vector double, vector double);
13385 int vec_all_nge (vector double, vector double);
13386 int vec_all_ngt (vector double, vector double);
13387 int vec_all_nle (vector double, vector double);
13388 int vec_all_nlt (vector double, vector double);
13389 int vec_all_numeric (vector double);
13390 int vec_any_eq (vector double, vector double);
13391 int vec_any_ge (vector double, vector double);
13392 int vec_any_gt (vector double, vector double);
13393 int vec_any_le (vector double, vector double);
13394 int vec_any_lt (vector double, vector double);
13395 int vec_any_nan (vector double);
13396 int vec_any_ne (vector double, vector double);
13397 int vec_any_nge (vector double, vector double);
13398 int vec_any_ngt (vector double, vector double);
13399 int vec_any_nle (vector double, vector double);
13400 int vec_any_nlt (vector double, vector double);
13401 int vec_any_numeric (vector double);
13403 vector double vec_vsx_ld (int, const vector double *);
13404 vector double vec_vsx_ld (int, const double *);
13405 vector float vec_vsx_ld (int, const vector float *);
13406 vector float vec_vsx_ld (int, const float *);
13407 vector bool int vec_vsx_ld (int, const vector bool int *);
13408 vector signed int vec_vsx_ld (int, const vector signed int *);
13409 vector signed int vec_vsx_ld (int, const int *);
13410 vector signed int vec_vsx_ld (int, const long *);
13411 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
13412 vector unsigned int vec_vsx_ld (int, const unsigned int *);
13413 vector unsigned int vec_vsx_ld (int, const unsigned long *);
13414 vector bool short vec_vsx_ld (int, const vector bool short *);
13415 vector pixel vec_vsx_ld (int, const vector pixel *);
13416 vector signed short vec_vsx_ld (int, const vector signed short *);
13417 vector signed short vec_vsx_ld (int, const short *);
13418 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
13419 vector unsigned short vec_vsx_ld (int, const unsigned short *);
13420 vector bool char vec_vsx_ld (int, const vector bool char *);
13421 vector signed char vec_vsx_ld (int, const vector signed char *);
13422 vector signed char vec_vsx_ld (int, const signed char *);
13423 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
13424 vector unsigned char vec_vsx_ld (int, const unsigned char *);
13426 void vec_vsx_st (vector double, int, vector double *);
13427 void vec_vsx_st (vector double, int, double *);
13428 void vec_vsx_st (vector float, int, vector float *);
13429 void vec_vsx_st (vector float, int, float *);
13430 void vec_vsx_st (vector signed int, int, vector signed int *);
13431 void vec_vsx_st (vector signed int, int, int *);
13432 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
13433 void vec_vsx_st (vector unsigned int, int, unsigned int *);
13434 void vec_vsx_st (vector bool int, int, vector bool int *);
13435 void vec_vsx_st (vector bool int, int, unsigned int *);
13436 void vec_vsx_st (vector bool int, int, int *);
13437 void vec_vsx_st (vector signed short, int, vector signed short *);
13438 void vec_vsx_st (vector signed short, int, short *);
13439 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
13440 void vec_vsx_st (vector unsigned short, int, unsigned short *);
13441 void vec_vsx_st (vector bool short, int, vector bool short *);
13442 void vec_vsx_st (vector bool short, int, unsigned short *);
13443 void vec_vsx_st (vector pixel, int, vector pixel *);
13444 void vec_vsx_st (vector pixel, int, unsigned short *);
13445 void vec_vsx_st (vector pixel, int, short *);
13446 void vec_vsx_st (vector bool short, int, short *);
13447 void vec_vsx_st (vector signed char, int, vector signed char *);
13448 void vec_vsx_st (vector signed char, int, signed char *);
13449 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
13450 void vec_vsx_st (vector unsigned char, int, unsigned char *);
13451 void vec_vsx_st (vector bool char, int, vector bool char *);
13452 void vec_vsx_st (vector bool char, int, unsigned char *);
13453 void vec_vsx_st (vector bool char, int, signed char *);
13456 Note that the @samp{vec_ld} and @samp{vec_st} builtins will always
13457 generate the Altivec @samp{LVX} and @samp{STVX} instructions even
13458 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
13459 @samp{vec_vsx_st} builtins will always generate the VSX @samp{LXVD2X},
13460 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
13462 GCC provides a few other builtins on Powerpc to access certain instructions:
13464 float __builtin_recipdivf (float, float);
13465 float __builtin_rsqrtf (float);
13466 double __builtin_recipdiv (double, double);
13467 double __builtin_rsqrt (double);
13468 long __builtin_bpermd (long, long);
13469 int __builtin_bswap16 (int);
13472 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13473 @code{__builtin_rsqrtf} functions generate multiple instructions to
13474 implement the reciprocal sqrt functionality using reciprocal sqrt
13475 estimate instructions.
13477 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13478 functions generate multiple instructions to implement division using
13479 the reciprocal estimate instructions.
13481 @node RX Built-in Functions
13482 @subsection RX Built-in Functions
13483 GCC supports some of the RX instructions which cannot be expressed in
13484 the C programming language via the use of built-in functions. The
13485 following functions are supported:
13487 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
13488 Generates the @code{brk} machine instruction.
13491 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
13492 Generates the @code{clrpsw} machine instruction to clear the specified
13493 bit in the processor status word.
13496 @deftypefn {Built-in Function} void __builtin_rx_int (int)
13497 Generates the @code{int} machine instruction to generate an interrupt
13498 with the specified value.
13501 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
13502 Generates the @code{machi} machine instruction to add the result of
13503 multiplying the top 16-bits of the two arguments into the
13507 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
13508 Generates the @code{maclo} machine instruction to add the result of
13509 multiplying the bottom 16-bits of the two arguments into the
13513 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
13514 Generates the @code{mulhi} machine instruction to place the result of
13515 multiplying the top 16-bits of the two arguments into the
13519 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
13520 Generates the @code{mullo} machine instruction to place the result of
13521 multiplying the bottom 16-bits of the two arguments into the
13525 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
13526 Generates the @code{mvfachi} machine instruction to read the top
13527 32-bits of the accumulator.
13530 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
13531 Generates the @code{mvfacmi} machine instruction to read the middle
13532 32-bits of the accumulator.
13535 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
13536 Generates the @code{mvfc} machine instruction which reads the control
13537 register specified in its argument and returns its value.
13540 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
13541 Generates the @code{mvtachi} machine instruction to set the top
13542 32-bits of the accumulator.
13545 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
13546 Generates the @code{mvtaclo} machine instruction to set the bottom
13547 32-bits of the accumulator.
13550 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
13551 Generates the @code{mvtc} machine instruction which sets control
13552 register number @code{reg} to @code{val}.
13555 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
13556 Generates the @code{mvtipl} machine instruction set the interrupt
13560 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
13561 Generates the @code{racw} machine instruction to round the accumulator
13562 according to the specified mode.
13565 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
13566 Generates the @code{revw} machine instruction which swaps the bytes in
13567 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
13568 and also bits 16--23 occupy bits 24--31 and vice versa.
13571 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
13572 Generates the @code{rmpa} machine instruction which initiates a
13573 repeated multiply and accumulate sequence.
13576 @deftypefn {Built-in Function} void __builtin_rx_round (float)
13577 Generates the @code{round} machine instruction which returns the
13578 floating point argument rounded according to the current rounding mode
13579 set in the floating point status word register.
13582 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
13583 Generates the @code{sat} machine instruction which returns the
13584 saturated value of the argument.
13587 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
13588 Generates the @code{setpsw} machine instruction to set the specified
13589 bit in the processor status word.
13592 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
13593 Generates the @code{wait} machine instruction.
13596 @node SPARC VIS Built-in Functions
13597 @subsection SPARC VIS Built-in Functions
13599 GCC supports SIMD operations on the SPARC using both the generic vector
13600 extensions (@pxref{Vector Extensions}) as well as built-in functions for
13601 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
13602 switch, the VIS extension is exposed as the following built-in functions:
13605 typedef int v1si __attribute__ ((vector_size (4)));
13606 typedef int v2si __attribute__ ((vector_size (8)));
13607 typedef short v4hi __attribute__ ((vector_size (8)));
13608 typedef short v2hi __attribute__ ((vector_size (4)));
13609 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
13610 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
13612 void __builtin_vis_write_gsr (int64_t);
13613 int64_t __builtin_vis_read_gsr (void);
13615 void * __builtin_vis_alignaddr (void *, long);
13616 void * __builtin_vis_alignaddrl (void *, long);
13617 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
13618 v2si __builtin_vis_faligndatav2si (v2si, v2si);
13619 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
13620 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
13622 v4hi __builtin_vis_fexpand (v4qi);
13624 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
13625 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
13626 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
13627 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
13628 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
13629 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
13630 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
13632 v4qi __builtin_vis_fpack16 (v4hi);
13633 v8qi __builtin_vis_fpack32 (v2si, v8qi);
13634 v2hi __builtin_vis_fpackfix (v2si);
13635 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
13637 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
13639 long __builtin_vis_edge8 (void *, void *);
13640 long __builtin_vis_edge8l (void *, void *);
13641 long __builtin_vis_edge16 (void *, void *);
13642 long __builtin_vis_edge16l (void *, void *);
13643 long __builtin_vis_edge32 (void *, void *);
13644 long __builtin_vis_edge32l (void *, void *);
13646 long __builtin_vis_fcmple16 (v4hi, v4hi);
13647 long __builtin_vis_fcmple32 (v2si, v2si);
13648 long __builtin_vis_fcmpne16 (v4hi, v4hi);
13649 long __builtin_vis_fcmpne32 (v2si, v2si);
13650 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
13651 long __builtin_vis_fcmpgt32 (v2si, v2si);
13652 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
13653 long __builtin_vis_fcmpeq32 (v2si, v2si);
13655 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
13656 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
13657 v2si __builtin_vis_fpadd32 (v2si, v2si);
13658 v1si __builtin_vis_fpadd32s (v1si, v1si);
13659 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
13660 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
13661 v2si __builtin_vis_fpsub32 (v2si, v2si);
13662 v1si __builtin_vis_fpsub32s (v1si, v1si);
13664 long __builtin_vis_array8 (long, long);
13665 long __builtin_vis_array16 (long, long);
13666 long __builtin_vis_array32 (long, long);
13669 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
13670 functions also become available:
13673 long __builtin_vis_bmask (long, long);
13674 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
13675 v2si __builtin_vis_bshufflev2si (v2si, v2si);
13676 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
13677 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
13679 long __builtin_vis_edge8n (void *, void *);
13680 long __builtin_vis_edge8ln (void *, void *);
13681 long __builtin_vis_edge16n (void *, void *);
13682 long __builtin_vis_edge16ln (void *, void *);
13683 long __builtin_vis_edge32n (void *, void *);
13684 long __builtin_vis_edge32ln (void *, void *);
13687 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
13688 functions also become available:
13691 void __builtin_vis_cmask8 (long);
13692 void __builtin_vis_cmask16 (long);
13693 void __builtin_vis_cmask32 (long);
13695 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
13697 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
13698 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
13699 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
13700 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
13701 v2si __builtin_vis_fsll16 (v2si, v2si);
13702 v2si __builtin_vis_fslas16 (v2si, v2si);
13703 v2si __builtin_vis_fsrl16 (v2si, v2si);
13704 v2si __builtin_vis_fsra16 (v2si, v2si);
13706 long __builtin_vis_pdistn (v8qi, v8qi);
13708 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
13710 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
13711 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
13713 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
13714 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
13715 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
13716 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
13717 v2si __builtin_vis_fpadds32 (v2si, v2si);
13718 v1si __builtin_vis_fpadds32s (v1si, v1si);
13719 v2si __builtin_vis_fpsubs32 (v2si, v2si);
13720 v1si __builtin_vis_fpsubs32s (v1si, v1si);
13722 long __builtin_vis_fucmple8 (v8qi, v8qi);
13723 long __builtin_vis_fucmpne8 (v8qi, v8qi);
13724 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
13725 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
13727 float __builtin_vis_fhadds (float, float);
13728 double __builtin_vis_fhaddd (double, double);
13729 float __builtin_vis_fhsubs (float, float);
13730 double __builtin_vis_fhsubd (double, double);
13731 float __builtin_vis_fnhadds (float, float);
13732 double __builtin_vis_fnhaddd (double, double);
13734 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
13735 int64_t __builtin_vis_xmulx (int64_t, int64_t);
13736 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
13739 @node SPU Built-in Functions
13740 @subsection SPU Built-in Functions
13742 GCC provides extensions for the SPU processor as described in the
13743 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
13744 found at @uref{http://cell.scei.co.jp/} or
13745 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
13746 implementation differs in several ways.
13751 The optional extension of specifying vector constants in parentheses is
13755 A vector initializer requires no cast if the vector constant is of the
13756 same type as the variable it is initializing.
13759 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13760 vector type is the default signedness of the base type. The default
13761 varies depending on the operating system, so a portable program should
13762 always specify the signedness.
13765 By default, the keyword @code{__vector} is added. The macro
13766 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
13770 GCC allows using a @code{typedef} name as the type specifier for a
13774 For C, overloaded functions are implemented with macros so the following
13778 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13781 Since @code{spu_add} is a macro, the vector constant in the example
13782 is treated as four separate arguments. Wrap the entire argument in
13783 parentheses for this to work.
13786 The extended version of @code{__builtin_expect} is not supported.
13790 @emph{Note:} Only the interface described in the aforementioned
13791 specification is supported. Internally, GCC uses built-in functions to
13792 implement the required functionality, but these are not supported and
13793 are subject to change without notice.
13795 @node TI C6X Built-in Functions
13796 @subsection TI C6X Built-in Functions
13798 GCC provides intrinsics to access certain instructions of the TI C6X
13799 processors. These intrinsics, listed below, are available after
13800 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
13801 to C6X instructions.
13805 int _sadd (int, int)
13806 int _ssub (int, int)
13807 int _sadd2 (int, int)
13808 int _ssub2 (int, int)
13809 long long _mpy2 (int, int)
13810 long long _smpy2 (int, int)
13811 int _add4 (int, int)
13812 int _sub4 (int, int)
13813 int _saddu4 (int, int)
13815 int _smpy (int, int)
13816 int _smpyh (int, int)
13817 int _smpyhl (int, int)
13818 int _smpylh (int, int)
13820 int _sshl (int, int)
13821 int _subc (int, int)
13823 int _avg2 (int, int)
13824 int _avgu4 (int, int)
13826 int _clrr (int, int)
13827 int _extr (int, int)
13828 int _extru (int, int)
13834 @node TILE-Gx Built-in Functions
13835 @subsection TILE-Gx Built-in Functions
13837 GCC provides intrinsics to access every instruction of the TILE-Gx
13838 processor. The intrinsics are of the form:
13842 unsigned long long __insn_@var{op} (...)
13846 Where @var{op} is the name of the instruction. Refer to the ISA manual
13847 for the complete list of instructions.
13849 GCC also provides intrinsics to directly access the network registers.
13850 The intrinsics are:
13854 unsigned long long __tile_idn0_receive (void)
13855 unsigned long long __tile_idn1_receive (void)
13856 unsigned long long __tile_udn0_receive (void)
13857 unsigned long long __tile_udn1_receive (void)
13858 unsigned long long __tile_udn2_receive (void)
13859 unsigned long long __tile_udn3_receive (void)
13860 void __tile_idn_send (unsigned long long)
13861 void __tile_udn_send (unsigned long long)
13865 The intrinsic @code{void __tile_network_barrier (void)} is used to
13866 guarantee that no network operatons before it will be reordered with
13869 @node TILEPro Built-in Functions
13870 @subsection TILEPro Built-in Functions
13872 GCC provides intrinsics to access every instruction of the TILEPro
13873 processor. The intrinsics are of the form:
13877 unsigned __insn_@var{op} (...)
13881 Where @var{op} is the name of the instruction. Refer to the ISA manual
13882 for the complete list of instructions.
13884 GCC also provides intrinsics to directly access the network registers.
13885 The intrinsics are:
13889 unsigned __tile_idn0_receive (void)
13890 unsigned __tile_idn1_receive (void)
13891 unsigned __tile_sn_receive (void)
13892 unsigned __tile_udn0_receive (void)
13893 unsigned __tile_udn1_receive (void)
13894 unsigned __tile_udn2_receive (void)
13895 unsigned __tile_udn3_receive (void)
13896 void __tile_idn_send (unsigned)
13897 void __tile_sn_send (unsigned)
13898 void __tile_udn_send (unsigned)
13902 The intrinsic @code{void __tile_network_barrier (void)} is used to
13903 guarantee that no network operatons before it will be reordered with
13906 @node Target Format Checks
13907 @section Format Checks Specific to Particular Target Machines
13909 For some target machines, GCC supports additional options to the
13911 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
13914 * Solaris Format Checks::
13915 * Darwin Format Checks::
13918 @node Solaris Format Checks
13919 @subsection Solaris Format Checks
13921 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
13922 check. @code{cmn_err} accepts a subset of the standard @code{printf}
13923 conversions, and the two-argument @code{%b} conversion for displaying
13924 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
13926 @node Darwin Format Checks
13927 @subsection Darwin Format Checks
13929 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
13930 attribute context. Declarations made with such attribution will be parsed for correct syntax
13931 and format argument types. However, parsing of the format string itself is currently undefined
13932 and will not be carried out by this version of the compiler.
13934 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
13935 also be used as format arguments. Note that the relevant headers are only likely to be
13936 available on Darwin (OSX) installations. On such installations, the XCode and system
13937 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
13938 associated functions.
13941 @section Pragmas Accepted by GCC
13943 @cindex @code{#pragma}
13945 GCC supports several types of pragmas, primarily in order to compile
13946 code originally written for other compilers. Note that in general
13947 we do not recommend the use of pragmas; @xref{Function Attributes},
13948 for further explanation.
13954 * RS/6000 and PowerPC Pragmas::
13956 * Solaris Pragmas::
13957 * Symbol-Renaming Pragmas::
13958 * Structure-Packing Pragmas::
13960 * Diagnostic Pragmas::
13961 * Visibility Pragmas::
13962 * Push/Pop Macro Pragmas::
13963 * Function Specific Option Pragmas::
13967 @subsection ARM Pragmas
13969 The ARM target defines pragmas for controlling the default addition of
13970 @code{long_call} and @code{short_call} attributes to functions.
13971 @xref{Function Attributes}, for information about the effects of these
13976 @cindex pragma, long_calls
13977 Set all subsequent functions to have the @code{long_call} attribute.
13979 @item no_long_calls
13980 @cindex pragma, no_long_calls
13981 Set all subsequent functions to have the @code{short_call} attribute.
13983 @item long_calls_off
13984 @cindex pragma, long_calls_off
13985 Do not affect the @code{long_call} or @code{short_call} attributes of
13986 subsequent functions.
13990 @subsection M32C Pragmas
13993 @item GCC memregs @var{number}
13994 @cindex pragma, memregs
13995 Overrides the command-line option @code{-memregs=} for the current
13996 file. Use with care! This pragma must be before any function in the
13997 file, and mixing different memregs values in different objects may
13998 make them incompatible. This pragma is useful when a
13999 performance-critical function uses a memreg for temporary values,
14000 as it may allow you to reduce the number of memregs used.
14002 @item ADDRESS @var{name} @var{address}
14003 @cindex pragma, address
14004 For any declared symbols matching @var{name}, this does three things
14005 to that symbol: it forces the symbol to be located at the given
14006 address (a number), it forces the symbol to be volatile, and it
14007 changes the symbol's scope to be static. This pragma exists for
14008 compatibility with other compilers, but note that the common
14009 @code{1234H} numeric syntax is not supported (use @code{0x1234}
14013 #pragma ADDRESS port3 0x103
14020 @subsection MeP Pragmas
14024 @item custom io_volatile (on|off)
14025 @cindex pragma, custom io_volatile
14026 Overrides the command line option @code{-mio-volatile} for the current
14027 file. Note that for compatibility with future GCC releases, this
14028 option should only be used once before any @code{io} variables in each
14031 @item GCC coprocessor available @var{registers}
14032 @cindex pragma, coprocessor available
14033 Specifies which coprocessor registers are available to the register
14034 allocator. @var{registers} may be a single register, register range
14035 separated by ellipses, or comma-separated list of those. Example:
14038 #pragma GCC coprocessor available $c0...$c10, $c28
14041 @item GCC coprocessor call_saved @var{registers}
14042 @cindex pragma, coprocessor call_saved
14043 Specifies which coprocessor registers are to be saved and restored by
14044 any function using them. @var{registers} may be a single register,
14045 register range separated by ellipses, or comma-separated list of
14049 #pragma GCC coprocessor call_saved $c4...$c6, $c31
14052 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
14053 @cindex pragma, coprocessor subclass
14054 Creates and defines a register class. These register classes can be
14055 used by inline @code{asm} constructs. @var{registers} may be a single
14056 register, register range separated by ellipses, or comma-separated
14057 list of those. Example:
14060 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
14062 asm ("cpfoo %0" : "=B" (x));
14065 @item GCC disinterrupt @var{name} , @var{name} @dots{}
14066 @cindex pragma, disinterrupt
14067 For the named functions, the compiler adds code to disable interrupts
14068 for the duration of those functions. Any functions so named, which
14069 are not encountered in the source, cause a warning that the pragma was
14070 not used. Examples:
14073 #pragma disinterrupt foo
14074 #pragma disinterrupt bar, grill
14075 int foo () @{ @dots{} @}
14078 @item GCC call @var{name} , @var{name} @dots{}
14079 @cindex pragma, call
14080 For the named functions, the compiler always uses a register-indirect
14081 call model when calling the named functions. Examples:
14090 @node RS/6000 and PowerPC Pragmas
14091 @subsection RS/6000 and PowerPC Pragmas
14093 The RS/6000 and PowerPC targets define one pragma for controlling
14094 whether or not the @code{longcall} attribute is added to function
14095 declarations by default. This pragma overrides the @option{-mlongcall}
14096 option, but not the @code{longcall} and @code{shortcall} attributes.
14097 @xref{RS/6000 and PowerPC Options}, for more information about when long
14098 calls are and are not necessary.
14102 @cindex pragma, longcall
14103 Apply the @code{longcall} attribute to all subsequent function
14107 Do not apply the @code{longcall} attribute to subsequent function
14111 @c Describe h8300 pragmas here.
14112 @c Describe sh pragmas here.
14113 @c Describe v850 pragmas here.
14115 @node Darwin Pragmas
14116 @subsection Darwin Pragmas
14118 The following pragmas are available for all architectures running the
14119 Darwin operating system. These are useful for compatibility with other
14123 @item mark @var{tokens}@dots{}
14124 @cindex pragma, mark
14125 This pragma is accepted, but has no effect.
14127 @item options align=@var{alignment}
14128 @cindex pragma, options align
14129 This pragma sets the alignment of fields in structures. The values of
14130 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
14131 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
14132 properly; to restore the previous setting, use @code{reset} for the
14135 @item segment @var{tokens}@dots{}
14136 @cindex pragma, segment
14137 This pragma is accepted, but has no effect.
14139 @item unused (@var{var} [, @var{var}]@dots{})
14140 @cindex pragma, unused
14141 This pragma declares variables to be possibly unused. GCC will not
14142 produce warnings for the listed variables. The effect is similar to
14143 that of the @code{unused} attribute, except that this pragma may appear
14144 anywhere within the variables' scopes.
14147 @node Solaris Pragmas
14148 @subsection Solaris Pragmas
14150 The Solaris target supports @code{#pragma redefine_extname}
14151 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
14152 @code{#pragma} directives for compatibility with the system compiler.
14155 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
14156 @cindex pragma, align
14158 Increase the minimum alignment of each @var{variable} to @var{alignment}.
14159 This is the same as GCC's @code{aligned} attribute @pxref{Variable
14160 Attributes}). Macro expansion occurs on the arguments to this pragma
14161 when compiling C and Objective-C@. It does not currently occur when
14162 compiling C++, but this is a bug which may be fixed in a future
14165 @item fini (@var{function} [, @var{function}]...)
14166 @cindex pragma, fini
14168 This pragma causes each listed @var{function} to be called after
14169 main, or during shared module unloading, by adding a call to the
14170 @code{.fini} section.
14172 @item init (@var{function} [, @var{function}]...)
14173 @cindex pragma, init
14175 This pragma causes each listed @var{function} to be called during
14176 initialization (before @code{main}) or during shared module loading, by
14177 adding a call to the @code{.init} section.
14181 @node Symbol-Renaming Pragmas
14182 @subsection Symbol-Renaming Pragmas
14184 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
14185 supports two @code{#pragma} directives which change the name used in
14186 assembly for a given declaration. @code{#pragma extern_prefix} is only
14187 available on platforms whose system headers need it. To get this effect
14188 on all platforms supported by GCC, use the asm labels extension (@pxref{Asm
14192 @item redefine_extname @var{oldname} @var{newname}
14193 @cindex pragma, redefine_extname
14195 This pragma gives the C function @var{oldname} the assembly symbol
14196 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
14197 will be defined if this pragma is available (currently on all platforms).
14199 @item extern_prefix @var{string}
14200 @cindex pragma, extern_prefix
14202 This pragma causes all subsequent external function and variable
14203 declarations to have @var{string} prepended to their assembly symbols.
14204 This effect may be terminated with another @code{extern_prefix} pragma
14205 whose argument is an empty string. The preprocessor macro
14206 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
14207 available (currently only on Tru64 UNIX)@.
14210 These pragmas and the asm labels extension interact in a complicated
14211 manner. Here are some corner cases you may want to be aware of.
14214 @item Both pragmas silently apply only to declarations with external
14215 linkage. Asm labels do not have this restriction.
14217 @item In C++, both pragmas silently apply only to declarations with
14218 ``C'' linkage. Again, asm labels do not have this restriction.
14220 @item If any of the three ways of changing the assembly name of a
14221 declaration is applied to a declaration whose assembly name has
14222 already been determined (either by a previous use of one of these
14223 features, or because the compiler needed the assembly name in order to
14224 generate code), and the new name is different, a warning issues and
14225 the name does not change.
14227 @item The @var{oldname} used by @code{#pragma redefine_extname} is
14228 always the C-language name.
14230 @item If @code{#pragma extern_prefix} is in effect, and a declaration
14231 occurs with an asm label attached, the prefix is silently ignored for
14234 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
14235 apply to the same declaration, whichever triggered first wins, and a
14236 warning issues if they contradict each other. (We would like to have
14237 @code{#pragma redefine_extname} always win, for consistency with asm
14238 labels, but if @code{#pragma extern_prefix} triggers first we have no
14239 way of knowing that that happened.)
14242 @node Structure-Packing Pragmas
14243 @subsection Structure-Packing Pragmas
14245 For compatibility with Microsoft Windows compilers, GCC supports a
14246 set of @code{#pragma} directives which change the maximum alignment of
14247 members of structures (other than zero-width bitfields), unions, and
14248 classes subsequently defined. The @var{n} value below always is required
14249 to be a small power of two and specifies the new alignment in bytes.
14252 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
14253 @item @code{#pragma pack()} sets the alignment to the one that was in
14254 effect when compilation started (see also command-line option
14255 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
14256 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
14257 setting on an internal stack and then optionally sets the new alignment.
14258 @item @code{#pragma pack(pop)} restores the alignment setting to the one
14259 saved at the top of the internal stack (and removes that stack entry).
14260 Note that @code{#pragma pack([@var{n}])} does not influence this internal
14261 stack; thus it is possible to have @code{#pragma pack(push)} followed by
14262 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
14263 @code{#pragma pack(pop)}.
14266 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
14267 @code{#pragma} which lays out a structure as the documented
14268 @code{__attribute__ ((ms_struct))}.
14270 @item @code{#pragma ms_struct on} turns on the layout for structures
14272 @item @code{#pragma ms_struct off} turns off the layout for structures
14274 @item @code{#pragma ms_struct reset} goes back to the default layout.
14278 @subsection Weak Pragmas
14280 For compatibility with SVR4, GCC supports a set of @code{#pragma}
14281 directives for declaring symbols to be weak, and defining weak
14285 @item #pragma weak @var{symbol}
14286 @cindex pragma, weak
14287 This pragma declares @var{symbol} to be weak, as if the declaration
14288 had the attribute of the same name. The pragma may appear before
14289 or after the declaration of @var{symbol}. It is not an error for
14290 @var{symbol} to never be defined at all.
14292 @item #pragma weak @var{symbol1} = @var{symbol2}
14293 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
14294 It is an error if @var{symbol2} is not defined in the current
14298 @node Diagnostic Pragmas
14299 @subsection Diagnostic Pragmas
14301 GCC allows the user to selectively enable or disable certain types of
14302 diagnostics, and change the kind of the diagnostic. For example, a
14303 project's policy might require that all sources compile with
14304 @option{-Werror} but certain files might have exceptions allowing
14305 specific types of warnings. Or, a project might selectively enable
14306 diagnostics and treat them as errors depending on which preprocessor
14307 macros are defined.
14310 @item #pragma GCC diagnostic @var{kind} @var{option}
14311 @cindex pragma, diagnostic
14313 Modifies the disposition of a diagnostic. Note that not all
14314 diagnostics are modifiable; at the moment only warnings (normally
14315 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
14316 Use @option{-fdiagnostics-show-option} to determine which diagnostics
14317 are controllable and which option controls them.
14319 @var{kind} is @samp{error} to treat this diagnostic as an error,
14320 @samp{warning} to treat it like a warning (even if @option{-Werror} is
14321 in effect), or @samp{ignored} if the diagnostic is to be ignored.
14322 @var{option} is a double quoted string which matches the command-line
14326 #pragma GCC diagnostic warning "-Wformat"
14327 #pragma GCC diagnostic error "-Wformat"
14328 #pragma GCC diagnostic ignored "-Wformat"
14331 Note that these pragmas override any command-line options. GCC keeps
14332 track of the location of each pragma, and issues diagnostics according
14333 to the state as of that point in the source file. Thus, pragmas occurring
14334 after a line do not affect diagnostics caused by that line.
14336 @item #pragma GCC diagnostic push
14337 @itemx #pragma GCC diagnostic pop
14339 Causes GCC to remember the state of the diagnostics as of each
14340 @code{push}, and restore to that point at each @code{pop}. If a
14341 @code{pop} has no matching @code{push}, the command line options are
14345 #pragma GCC diagnostic error "-Wuninitialized"
14346 foo(a); /* error is given for this one */
14347 #pragma GCC diagnostic push
14348 #pragma GCC diagnostic ignored "-Wuninitialized"
14349 foo(b); /* no diagnostic for this one */
14350 #pragma GCC diagnostic pop
14351 foo(c); /* error is given for this one */
14352 #pragma GCC diagnostic pop
14353 foo(d); /* depends on command line options */
14358 GCC also offers a simple mechanism for printing messages during
14362 @item #pragma message @var{string}
14363 @cindex pragma, diagnostic
14365 Prints @var{string} as a compiler message on compilation. The message
14366 is informational only, and is neither a compilation warning nor an error.
14369 #pragma message "Compiling " __FILE__ "..."
14372 @var{string} may be parenthesized, and is printed with location
14373 information. For example,
14376 #define DO_PRAGMA(x) _Pragma (#x)
14377 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
14379 TODO(Remember to fix this)
14382 prints @samp{/tmp/file.c:4: note: #pragma message:
14383 TODO - Remember to fix this}.
14387 @node Visibility Pragmas
14388 @subsection Visibility Pragmas
14391 @item #pragma GCC visibility push(@var{visibility})
14392 @itemx #pragma GCC visibility pop
14393 @cindex pragma, visibility
14395 This pragma allows the user to set the visibility for multiple
14396 declarations without having to give each a visibility attribute
14397 @xref{Function Attributes}, for more information about visibility and
14398 the attribute syntax.
14400 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
14401 declarations. Class members and template specializations are not
14402 affected; if you want to override the visibility for a particular
14403 member or instantiation, you must use an attribute.
14408 @node Push/Pop Macro Pragmas
14409 @subsection Push/Pop Macro Pragmas
14411 For compatibility with Microsoft Windows compilers, GCC supports
14412 @samp{#pragma push_macro(@var{"macro_name"})}
14413 and @samp{#pragma pop_macro(@var{"macro_name"})}.
14416 @item #pragma push_macro(@var{"macro_name"})
14417 @cindex pragma, push_macro
14418 This pragma saves the value of the macro named as @var{macro_name} to
14419 the top of the stack for this macro.
14421 @item #pragma pop_macro(@var{"macro_name"})
14422 @cindex pragma, pop_macro
14423 This pragma sets the value of the macro named as @var{macro_name} to
14424 the value on top of the stack for this macro. If the stack for
14425 @var{macro_name} is empty, the value of the macro remains unchanged.
14432 #pragma push_macro("X")
14435 #pragma pop_macro("X")
14439 In this example, the definition of X as 1 is saved by @code{#pragma
14440 push_macro} and restored by @code{#pragma pop_macro}.
14442 @node Function Specific Option Pragmas
14443 @subsection Function Specific Option Pragmas
14446 @item #pragma GCC target (@var{"string"}...)
14447 @cindex pragma GCC target
14449 This pragma allows you to set target specific options for functions
14450 defined later in the source file. One or more strings can be
14451 specified. Each function that is defined after this point will be as
14452 if @code{attribute((target("STRING")))} was specified for that
14453 function. The parenthesis around the options is optional.
14454 @xref{Function Attributes}, for more information about the
14455 @code{target} attribute and the attribute syntax.
14457 The @code{#pragma GCC target} attribute is not implemented in GCC versions earlier
14458 than 4.4 for the i386/x86_64 and 4.6 for the PowerPC backends. At
14459 present, it is not implemented for other backends.
14463 @item #pragma GCC optimize (@var{"string"}...)
14464 @cindex pragma GCC optimize
14466 This pragma allows you to set global optimization options for functions
14467 defined later in the source file. One or more strings can be
14468 specified. Each function that is defined after this point will be as
14469 if @code{attribute((optimize("STRING")))} was specified for that
14470 function. The parenthesis around the options is optional.
14471 @xref{Function Attributes}, for more information about the
14472 @code{optimize} attribute and the attribute syntax.
14474 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
14475 versions earlier than 4.4.
14479 @item #pragma GCC push_options
14480 @itemx #pragma GCC pop_options
14481 @cindex pragma GCC push_options
14482 @cindex pragma GCC pop_options
14484 These pragmas maintain a stack of the current target and optimization
14485 options. It is intended for include files where you temporarily want
14486 to switch to using a different @samp{#pragma GCC target} or
14487 @samp{#pragma GCC optimize} and then to pop back to the previous
14490 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
14491 pragmas are not implemented in GCC versions earlier than 4.4.
14495 @item #pragma GCC reset_options
14496 @cindex pragma GCC reset_options
14498 This pragma clears the current @code{#pragma GCC target} and
14499 @code{#pragma GCC optimize} to use the default switches as specified
14500 on the command line.
14502 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
14503 versions earlier than 4.4.
14506 @node Unnamed Fields
14507 @section Unnamed struct/union fields within structs/unions
14508 @cindex @code{struct}
14509 @cindex @code{union}
14511 As permitted by ISO C11 and for compatibility with other compilers,
14512 GCC allows you to define
14513 a structure or union that contains, as fields, structures and unions
14514 without names. For example:
14527 In this example, the user would be able to access members of the unnamed
14528 union with code like @samp{foo.b}. Note that only unnamed structs and
14529 unions are allowed, you may not have, for example, an unnamed
14532 You must never create such structures that cause ambiguous field definitions.
14533 For example, this structure:
14544 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
14545 The compiler gives errors for such constructs.
14547 @opindex fms-extensions
14548 Unless @option{-fms-extensions} is used, the unnamed field must be a
14549 structure or union definition without a tag (for example, @samp{struct
14550 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
14551 also be a definition with a tag such as @samp{struct foo @{ int a;
14552 @};}, a reference to a previously defined structure or union such as
14553 @samp{struct foo;}, or a reference to a @code{typedef} name for a
14554 previously defined structure or union type.
14556 @opindex fplan9-extensions
14557 The option @option{-fplan9-extensions} enables
14558 @option{-fms-extensions} as well as two other extensions. First, a
14559 pointer to a structure is automatically converted to a pointer to an
14560 anonymous field for assignments and function calls. For example:
14563 struct s1 @{ int a; @};
14564 struct s2 @{ struct s1; @};
14565 extern void f1 (struct s1 *);
14566 void f2 (struct s2 *p) @{ f1 (p); @}
14569 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
14570 converted into a pointer to the anonymous field.
14572 Second, when the type of an anonymous field is a @code{typedef} for a
14573 @code{struct} or @code{union}, code may refer to the field using the
14574 name of the @code{typedef}.
14577 typedef struct @{ int a; @} s1;
14578 struct s2 @{ s1; @};
14579 s1 f1 (struct s2 *p) @{ return p->s1; @}
14582 These usages are only permitted when they are not ambiguous.
14585 @section Thread-Local Storage
14586 @cindex Thread-Local Storage
14587 @cindex @acronym{TLS}
14588 @cindex @code{__thread}
14590 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
14591 are allocated such that there is one instance of the variable per extant
14592 thread. The run-time model GCC uses to implement this originates
14593 in the IA-64 processor-specific ABI, but has since been migrated
14594 to other processors as well. It requires significant support from
14595 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
14596 system libraries (@file{libc.so} and @file{libpthread.so}), so it
14597 is not available everywhere.
14599 At the user level, the extension is visible with a new storage
14600 class keyword: @code{__thread}. For example:
14604 extern __thread struct state s;
14605 static __thread char *p;
14608 The @code{__thread} specifier may be used alone, with the @code{extern}
14609 or @code{static} specifiers, but with no other storage class specifier.
14610 When used with @code{extern} or @code{static}, @code{__thread} must appear
14611 immediately after the other storage class specifier.
14613 The @code{__thread} specifier may be applied to any global, file-scoped
14614 static, function-scoped static, or static data member of a class. It may
14615 not be applied to block-scoped automatic or non-static data member.
14617 When the address-of operator is applied to a thread-local variable, it is
14618 evaluated at run-time and returns the address of the current thread's
14619 instance of that variable. An address so obtained may be used by any
14620 thread. When a thread terminates, any pointers to thread-local variables
14621 in that thread become invalid.
14623 No static initialization may refer to the address of a thread-local variable.
14625 In C++, if an initializer is present for a thread-local variable, it must
14626 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
14629 See @uref{http://www.akkadia.org/drepper/tls.pdf,
14630 ELF Handling For Thread-Local Storage} for a detailed explanation of
14631 the four thread-local storage addressing models, and how the run-time
14632 is expected to function.
14635 * C99 Thread-Local Edits::
14636 * C++98 Thread-Local Edits::
14639 @node C99 Thread-Local Edits
14640 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
14642 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
14643 that document the exact semantics of the language extension.
14647 @cite{5.1.2 Execution environments}
14649 Add new text after paragraph 1
14652 Within either execution environment, a @dfn{thread} is a flow of
14653 control within a program. It is implementation defined whether
14654 or not there may be more than one thread associated with a program.
14655 It is implementation defined how threads beyond the first are
14656 created, the name and type of the function called at thread
14657 startup, and how threads may be terminated. However, objects
14658 with thread storage duration shall be initialized before thread
14663 @cite{6.2.4 Storage durations of objects}
14665 Add new text before paragraph 3
14668 An object whose identifier is declared with the storage-class
14669 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
14670 Its lifetime is the entire execution of the thread, and its
14671 stored value is initialized only once, prior to thread startup.
14675 @cite{6.4.1 Keywords}
14677 Add @code{__thread}.
14680 @cite{6.7.1 Storage-class specifiers}
14682 Add @code{__thread} to the list of storage class specifiers in
14685 Change paragraph 2 to
14688 With the exception of @code{__thread}, at most one storage-class
14689 specifier may be given [@dots{}]. The @code{__thread} specifier may
14690 be used alone, or immediately following @code{extern} or
14694 Add new text after paragraph 6
14697 The declaration of an identifier for a variable that has
14698 block scope that specifies @code{__thread} shall also
14699 specify either @code{extern} or @code{static}.
14701 The @code{__thread} specifier shall be used only with
14706 @node C++98 Thread-Local Edits
14707 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
14709 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
14710 that document the exact semantics of the language extension.
14714 @b{[intro.execution]}
14716 New text after paragraph 4
14719 A @dfn{thread} is a flow of control within the abstract machine.
14720 It is implementation defined whether or not there may be more than
14724 New text after paragraph 7
14727 It is unspecified whether additional action must be taken to
14728 ensure when and whether side effects are visible to other threads.
14734 Add @code{__thread}.
14737 @b{[basic.start.main]}
14739 Add after paragraph 5
14742 The thread that begins execution at the @code{main} function is called
14743 the @dfn{main thread}. It is implementation defined how functions
14744 beginning threads other than the main thread are designated or typed.
14745 A function so designated, as well as the @code{main} function, is called
14746 a @dfn{thread startup function}. It is implementation defined what
14747 happens if a thread startup function returns. It is implementation
14748 defined what happens to other threads when any thread calls @code{exit}.
14752 @b{[basic.start.init]}
14754 Add after paragraph 4
14757 The storage for an object of thread storage duration shall be
14758 statically initialized before the first statement of the thread startup
14759 function. An object of thread storage duration shall not require
14760 dynamic initialization.
14764 @b{[basic.start.term]}
14766 Add after paragraph 3
14769 The type of an object with thread storage duration shall not have a
14770 non-trivial destructor, nor shall it be an array type whose elements
14771 (directly or indirectly) have non-trivial destructors.
14777 Add ``thread storage duration'' to the list in paragraph 1.
14782 Thread, static, and automatic storage durations are associated with
14783 objects introduced by declarations [@dots{}].
14786 Add @code{__thread} to the list of specifiers in paragraph 3.
14789 @b{[basic.stc.thread]}
14791 New section before @b{[basic.stc.static]}
14794 The keyword @code{__thread} applied to a non-local object gives the
14795 object thread storage duration.
14797 A local variable or class data member declared both @code{static}
14798 and @code{__thread} gives the variable or member thread storage
14803 @b{[basic.stc.static]}
14808 All objects which have neither thread storage duration, dynamic
14809 storage duration nor are local [@dots{}].
14815 Add @code{__thread} to the list in paragraph 1.
14820 With the exception of @code{__thread}, at most one
14821 @var{storage-class-specifier} shall appear in a given
14822 @var{decl-specifier-seq}. The @code{__thread} specifier may
14823 be used alone, or immediately following the @code{extern} or
14824 @code{static} specifiers. [@dots{}]
14827 Add after paragraph 5
14830 The @code{__thread} specifier can be applied only to the names of objects
14831 and to anonymous unions.
14837 Add after paragraph 6
14840 Non-@code{static} members shall not be @code{__thread}.
14844 @node Binary constants
14845 @section Binary constants using the @samp{0b} prefix
14846 @cindex Binary constants using the @samp{0b} prefix
14848 Integer constants can be written as binary constants, consisting of a
14849 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
14850 @samp{0B}. This is particularly useful in environments that operate a
14851 lot on the bit-level (like microcontrollers).
14853 The following statements are identical:
14862 The type of these constants follows the same rules as for octal or
14863 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
14866 @node C++ Extensions
14867 @chapter Extensions to the C++ Language
14868 @cindex extensions, C++ language
14869 @cindex C++ language extensions
14871 The GNU compiler provides these extensions to the C++ language (and you
14872 can also use most of the C language extensions in your C++ programs). If you
14873 want to write code that checks whether these features are available, you can
14874 test for the GNU compiler the same way as for C programs: check for a
14875 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
14876 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
14877 Predefined Macros,cpp,The GNU C Preprocessor}).
14880 * C++ Volatiles:: What constitutes an access to a volatile object.
14881 * Restricted Pointers:: C99 restricted pointers and references.
14882 * Vague Linkage:: Where G++ puts inlines, vtables and such.
14883 * C++ Interface:: You can use a single C++ header file for both
14884 declarations and definitions.
14885 * Template Instantiation:: Methods for ensuring that exactly one copy of
14886 each needed template instantiation is emitted.
14887 * Bound member functions:: You can extract a function pointer to the
14888 method denoted by a @samp{->*} or @samp{.*} expression.
14889 * C++ Attributes:: Variable, function, and type attributes for C++ only.
14890 * Namespace Association:: Strong using-directives for namespace association.
14891 * Type Traits:: Compiler support for type traits
14892 * Java Exceptions:: Tweaking exception handling to work with Java.
14893 * Deprecated Features:: Things will disappear from g++.
14894 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
14897 @node C++ Volatiles
14898 @section When is a Volatile C++ Object Accessed?
14899 @cindex accessing volatiles
14900 @cindex volatile read
14901 @cindex volatile write
14902 @cindex volatile access
14904 The C++ standard differs from the C standard in its treatment of
14905 volatile objects. It fails to specify what constitutes a volatile
14906 access, except to say that C++ should behave in a similar manner to C
14907 with respect to volatiles, where possible. However, the different
14908 lvalueness of expressions between C and C++ complicate the behavior.
14909 G++ behaves the same as GCC for volatile access, @xref{C
14910 Extensions,,Volatiles}, for a description of GCC's behavior.
14912 The C and C++ language specifications differ when an object is
14913 accessed in a void context:
14916 volatile int *src = @var{somevalue};
14920 The C++ standard specifies that such expressions do not undergo lvalue
14921 to rvalue conversion, and that the type of the dereferenced object may
14922 be incomplete. The C++ standard does not specify explicitly that it
14923 is lvalue to rvalue conversion which is responsible for causing an
14924 access. There is reason to believe that it is, because otherwise
14925 certain simple expressions become undefined. However, because it
14926 would surprise most programmers, G++ treats dereferencing a pointer to
14927 volatile object of complete type as GCC would do for an equivalent
14928 type in C@. When the object has incomplete type, G++ issues a
14929 warning; if you wish to force an error, you must force a conversion to
14930 rvalue with, for instance, a static cast.
14932 When using a reference to volatile, G++ does not treat equivalent
14933 expressions as accesses to volatiles, but instead issues a warning that
14934 no volatile is accessed. The rationale for this is that otherwise it
14935 becomes difficult to determine where volatile access occur, and not
14936 possible to ignore the return value from functions returning volatile
14937 references. Again, if you wish to force a read, cast the reference to
14940 G++ implements the same behavior as GCC does when assigning to a
14941 volatile object -- there is no reread of the assigned-to object, the
14942 assigned rvalue is reused. Note that in C++ assignment expressions
14943 are lvalues, and if used as an lvalue, the volatile object will be
14944 referred to. For instance, @var{vref} will refer to @var{vobj}, as
14945 expected, in the following example:
14949 volatile int &vref = vobj = @var{something};
14952 @node Restricted Pointers
14953 @section Restricting Pointer Aliasing
14954 @cindex restricted pointers
14955 @cindex restricted references
14956 @cindex restricted this pointer
14958 As with the C front end, G++ understands the C99 feature of restricted pointers,
14959 specified with the @code{__restrict__}, or @code{__restrict} type
14960 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
14961 language flag, @code{restrict} is not a keyword in C++.
14963 In addition to allowing restricted pointers, you can specify restricted
14964 references, which indicate that the reference is not aliased in the local
14968 void fn (int *__restrict__ rptr, int &__restrict__ rref)
14975 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
14976 @var{rref} refers to a (different) unaliased integer.
14978 You may also specify whether a member function's @var{this} pointer is
14979 unaliased by using @code{__restrict__} as a member function qualifier.
14982 void T::fn () __restrict__
14989 Within the body of @code{T::fn}, @var{this} will have the effective
14990 definition @code{T *__restrict__ const this}. Notice that the
14991 interpretation of a @code{__restrict__} member function qualifier is
14992 different to that of @code{const} or @code{volatile} qualifier, in that it
14993 is applied to the pointer rather than the object. This is consistent with
14994 other compilers which implement restricted pointers.
14996 As with all outermost parameter qualifiers, @code{__restrict__} is
14997 ignored in function definition matching. This means you only need to
14998 specify @code{__restrict__} in a function definition, rather than
14999 in a function prototype as well.
15001 @node Vague Linkage
15002 @section Vague Linkage
15003 @cindex vague linkage
15005 There are several constructs in C++ which require space in the object
15006 file but are not clearly tied to a single translation unit. We say that
15007 these constructs have ``vague linkage''. Typically such constructs are
15008 emitted wherever they are needed, though sometimes we can be more
15012 @item Inline Functions
15013 Inline functions are typically defined in a header file which can be
15014 included in many different compilations. Hopefully they can usually be
15015 inlined, but sometimes an out-of-line copy is necessary, if the address
15016 of the function is taken or if inlining fails. In general, we emit an
15017 out-of-line copy in all translation units where one is needed. As an
15018 exception, we only emit inline virtual functions with the vtable, since
15019 it will always require a copy.
15021 Local static variables and string constants used in an inline function
15022 are also considered to have vague linkage, since they must be shared
15023 between all inlined and out-of-line instances of the function.
15027 C++ virtual functions are implemented in most compilers using a lookup
15028 table, known as a vtable. The vtable contains pointers to the virtual
15029 functions provided by a class, and each object of the class contains a
15030 pointer to its vtable (or vtables, in some multiple-inheritance
15031 situations). If the class declares any non-inline, non-pure virtual
15032 functions, the first one is chosen as the ``key method'' for the class,
15033 and the vtable is only emitted in the translation unit where the key
15036 @emph{Note:} If the chosen key method is later defined as inline, the
15037 vtable will still be emitted in every translation unit which defines it.
15038 Make sure that any inline virtuals are declared inline in the class
15039 body, even if they are not defined there.
15041 @item @code{type_info} objects
15042 @cindex @code{type_info}
15044 C++ requires information about types to be written out in order to
15045 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
15046 For polymorphic classes (classes with virtual functions), the @samp{type_info}
15047 object is written out along with the vtable so that @samp{dynamic_cast}
15048 can determine the dynamic type of a class object at runtime. For all
15049 other types, we write out the @samp{type_info} object when it is used: when
15050 applying @samp{typeid} to an expression, throwing an object, or
15051 referring to a type in a catch clause or exception specification.
15053 @item Template Instantiations
15054 Most everything in this section also applies to template instantiations,
15055 but there are other options as well.
15056 @xref{Template Instantiation,,Where's the Template?}.
15060 When used with GNU ld version 2.8 or later on an ELF system such as
15061 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
15062 these constructs will be discarded at link time. This is known as
15065 On targets that don't support COMDAT, but do support weak symbols, GCC
15066 will use them. This way one copy will override all the others, but
15067 the unused copies will still take up space in the executable.
15069 For targets which do not support either COMDAT or weak symbols,
15070 most entities with vague linkage will be emitted as local symbols to
15071 avoid duplicate definition errors from the linker. This will not happen
15072 for local statics in inlines, however, as having multiple copies will
15073 almost certainly break things.
15075 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
15076 another way to control placement of these constructs.
15078 @node C++ Interface
15079 @section #pragma interface and implementation
15081 @cindex interface and implementation headers, C++
15082 @cindex C++ interface and implementation headers
15083 @cindex pragmas, interface and implementation
15085 @code{#pragma interface} and @code{#pragma implementation} provide the
15086 user with a way of explicitly directing the compiler to emit entities
15087 with vague linkage (and debugging information) in a particular
15090 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
15091 most cases, because of COMDAT support and the ``key method'' heuristic
15092 mentioned in @ref{Vague Linkage}. Using them can actually cause your
15093 program to grow due to unnecessary out-of-line copies of inline
15094 functions. Currently (3.4) the only benefit of these
15095 @code{#pragma}s is reduced duplication of debugging information, and
15096 that should be addressed soon on DWARF 2 targets with the use of
15100 @item #pragma interface
15101 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
15102 @kindex #pragma interface
15103 Use this directive in @emph{header files} that define object classes, to save
15104 space in most of the object files that use those classes. Normally,
15105 local copies of certain information (backup copies of inline member
15106 functions, debugging information, and the internal tables that implement
15107 virtual functions) must be kept in each object file that includes class
15108 definitions. You can use this pragma to avoid such duplication. When a
15109 header file containing @samp{#pragma interface} is included in a
15110 compilation, this auxiliary information will not be generated (unless
15111 the main input source file itself uses @samp{#pragma implementation}).
15112 Instead, the object files will contain references to be resolved at link
15115 The second form of this directive is useful for the case where you have
15116 multiple headers with the same name in different directories. If you
15117 use this form, you must specify the same string to @samp{#pragma
15120 @item #pragma implementation
15121 @itemx #pragma implementation "@var{objects}.h"
15122 @kindex #pragma implementation
15123 Use this pragma in a @emph{main input file}, when you want full output from
15124 included header files to be generated (and made globally visible). The
15125 included header file, in turn, should use @samp{#pragma interface}.
15126 Backup copies of inline member functions, debugging information, and the
15127 internal tables used to implement virtual functions are all generated in
15128 implementation files.
15130 @cindex implied @code{#pragma implementation}
15131 @cindex @code{#pragma implementation}, implied
15132 @cindex naming convention, implementation headers
15133 If you use @samp{#pragma implementation} with no argument, it applies to
15134 an include file with the same basename@footnote{A file's @dfn{basename}
15135 was the name stripped of all leading path information and of trailing
15136 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
15137 file. For example, in @file{allclass.cc}, giving just
15138 @samp{#pragma implementation}
15139 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
15141 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
15142 an implementation file whenever you would include it from
15143 @file{allclass.cc} even if you never specified @samp{#pragma
15144 implementation}. This was deemed to be more trouble than it was worth,
15145 however, and disabled.
15147 Use the string argument if you want a single implementation file to
15148 include code from multiple header files. (You must also use
15149 @samp{#include} to include the header file; @samp{#pragma
15150 implementation} only specifies how to use the file---it doesn't actually
15153 There is no way to split up the contents of a single header file into
15154 multiple implementation files.
15157 @cindex inlining and C++ pragmas
15158 @cindex C++ pragmas, effect on inlining
15159 @cindex pragmas in C++, effect on inlining
15160 @samp{#pragma implementation} and @samp{#pragma interface} also have an
15161 effect on function inlining.
15163 If you define a class in a header file marked with @samp{#pragma
15164 interface}, the effect on an inline function defined in that class is
15165 similar to an explicit @code{extern} declaration---the compiler emits
15166 no code at all to define an independent version of the function. Its
15167 definition is used only for inlining with its callers.
15169 @opindex fno-implement-inlines
15170 Conversely, when you include the same header file in a main source file
15171 that declares it as @samp{#pragma implementation}, the compiler emits
15172 code for the function itself; this defines a version of the function
15173 that can be found via pointers (or by callers compiled without
15174 inlining). If all calls to the function can be inlined, you can avoid
15175 emitting the function by compiling with @option{-fno-implement-inlines}.
15176 If any calls were not inlined, you will get linker errors.
15178 @node Template Instantiation
15179 @section Where's the Template?
15180 @cindex template instantiation
15182 C++ templates are the first language feature to require more
15183 intelligence from the environment than one usually finds on a UNIX
15184 system. Somehow the compiler and linker have to make sure that each
15185 template instance occurs exactly once in the executable if it is needed,
15186 and not at all otherwise. There are two basic approaches to this
15187 problem, which are referred to as the Borland model and the Cfront model.
15190 @item Borland model
15191 Borland C++ solved the template instantiation problem by adding the code
15192 equivalent of common blocks to their linker; the compiler emits template
15193 instances in each translation unit that uses them, and the linker
15194 collapses them together. The advantage of this model is that the linker
15195 only has to consider the object files themselves; there is no external
15196 complexity to worry about. This disadvantage is that compilation time
15197 is increased because the template code is being compiled repeatedly.
15198 Code written for this model tends to include definitions of all
15199 templates in the header file, since they must be seen to be
15203 The AT&T C++ translator, Cfront, solved the template instantiation
15204 problem by creating the notion of a template repository, an
15205 automatically maintained place where template instances are stored. A
15206 more modern version of the repository works as follows: As individual
15207 object files are built, the compiler places any template definitions and
15208 instantiations encountered in the repository. At link time, the link
15209 wrapper adds in the objects in the repository and compiles any needed
15210 instances that were not previously emitted. The advantages of this
15211 model are more optimal compilation speed and the ability to use the
15212 system linker; to implement the Borland model a compiler vendor also
15213 needs to replace the linker. The disadvantages are vastly increased
15214 complexity, and thus potential for error; for some code this can be
15215 just as transparent, but in practice it can been very difficult to build
15216 multiple programs in one directory and one program in multiple
15217 directories. Code written for this model tends to separate definitions
15218 of non-inline member templates into a separate file, which should be
15219 compiled separately.
15222 When used with GNU ld version 2.8 or later on an ELF system such as
15223 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
15224 Borland model. On other systems, G++ implements neither automatic
15227 A future version of G++ will support a hybrid model whereby the compiler
15228 will emit any instantiations for which the template definition is
15229 included in the compile, and store template definitions and
15230 instantiation context information into the object file for the rest.
15231 The link wrapper will extract that information as necessary and invoke
15232 the compiler to produce the remaining instantiations. The linker will
15233 then combine duplicate instantiations.
15235 In the mean time, you have the following options for dealing with
15236 template instantiations:
15241 Compile your template-using code with @option{-frepo}. The compiler will
15242 generate files with the extension @samp{.rpo} listing all of the
15243 template instantiations used in the corresponding object files which
15244 could be instantiated there; the link wrapper, @samp{collect2}, will
15245 then update the @samp{.rpo} files to tell the compiler where to place
15246 those instantiations and rebuild any affected object files. The
15247 link-time overhead is negligible after the first pass, as the compiler
15248 will continue to place the instantiations in the same files.
15250 This is your best option for application code written for the Borland
15251 model, as it will just work. Code written for the Cfront model will
15252 need to be modified so that the template definitions are available at
15253 one or more points of instantiation; usually this is as simple as adding
15254 @code{#include <tmethods.cc>} to the end of each template header.
15256 For library code, if you want the library to provide all of the template
15257 instantiations it needs, just try to link all of its object files
15258 together; the link will fail, but cause the instantiations to be
15259 generated as a side effect. Be warned, however, that this may cause
15260 conflicts if multiple libraries try to provide the same instantiations.
15261 For greater control, use explicit instantiation as described in the next
15265 @opindex fno-implicit-templates
15266 Compile your code with @option{-fno-implicit-templates} to disable the
15267 implicit generation of template instances, and explicitly instantiate
15268 all the ones you use. This approach requires more knowledge of exactly
15269 which instances you need than do the others, but it's less
15270 mysterious and allows greater control. You can scatter the explicit
15271 instantiations throughout your program, perhaps putting them in the
15272 translation units where the instances are used or the translation units
15273 that define the templates themselves; you can put all of the explicit
15274 instantiations you need into one big file; or you can create small files
15281 template class Foo<int>;
15282 template ostream& operator <<
15283 (ostream&, const Foo<int>&);
15286 for each of the instances you need, and create a template instantiation
15287 library from those.
15289 If you are using Cfront-model code, you can probably get away with not
15290 using @option{-fno-implicit-templates} when compiling files that don't
15291 @samp{#include} the member template definitions.
15293 If you use one big file to do the instantiations, you may want to
15294 compile it without @option{-fno-implicit-templates} so you get all of the
15295 instances required by your explicit instantiations (but not by any
15296 other files) without having to specify them as well.
15298 G++ has extended the template instantiation syntax given in the ISO
15299 standard to allow forward declaration of explicit instantiations
15300 (with @code{extern}), instantiation of the compiler support data for a
15301 template class (i.e.@: the vtable) without instantiating any of its
15302 members (with @code{inline}), and instantiation of only the static data
15303 members of a template class, without the support data or member
15304 functions (with (@code{static}):
15307 extern template int max (int, int);
15308 inline template class Foo<int>;
15309 static template class Foo<int>;
15313 Do nothing. Pretend G++ does implement automatic instantiation
15314 management. Code written for the Borland model will work fine, but
15315 each translation unit will contain instances of each of the templates it
15316 uses. In a large program, this can lead to an unacceptable amount of code
15320 @node Bound member functions
15321 @section Extracting the function pointer from a bound pointer to member function
15323 @cindex pointer to member function
15324 @cindex bound pointer to member function
15326 In C++, pointer to member functions (PMFs) are implemented using a wide
15327 pointer of sorts to handle all the possible call mechanisms; the PMF
15328 needs to store information about how to adjust the @samp{this} pointer,
15329 and if the function pointed to is virtual, where to find the vtable, and
15330 where in the vtable to look for the member function. If you are using
15331 PMFs in an inner loop, you should really reconsider that decision. If
15332 that is not an option, you can extract the pointer to the function that
15333 would be called for a given object/PMF pair and call it directly inside
15334 the inner loop, to save a bit of time.
15336 Note that you will still be paying the penalty for the call through a
15337 function pointer; on most modern architectures, such a call defeats the
15338 branch prediction features of the CPU@. This is also true of normal
15339 virtual function calls.
15341 The syntax for this extension is
15345 extern int (A::*fp)();
15346 typedef int (*fptr)(A *);
15348 fptr p = (fptr)(a.*fp);
15351 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
15352 no object is needed to obtain the address of the function. They can be
15353 converted to function pointers directly:
15356 fptr p1 = (fptr)(&A::foo);
15359 @opindex Wno-pmf-conversions
15360 You must specify @option{-Wno-pmf-conversions} to use this extension.
15362 @node C++ Attributes
15363 @section C++-Specific Variable, Function, and Type Attributes
15365 Some attributes only make sense for C++ programs.
15368 @item init_priority (@var{priority})
15369 @cindex @code{init_priority} attribute
15372 In Standard C++, objects defined at namespace scope are guaranteed to be
15373 initialized in an order in strict accordance with that of their definitions
15374 @emph{in a given translation unit}. No guarantee is made for initializations
15375 across translation units. However, GNU C++ allows users to control the
15376 order of initialization of objects defined at namespace scope with the
15377 @code{init_priority} attribute by specifying a relative @var{priority},
15378 a constant integral expression currently bounded between 101 and 65535
15379 inclusive. Lower numbers indicate a higher priority.
15381 In the following example, @code{A} would normally be created before
15382 @code{B}, but the @code{init_priority} attribute has reversed that order:
15385 Some_Class A __attribute__ ((init_priority (2000)));
15386 Some_Class B __attribute__ ((init_priority (543)));
15390 Note that the particular values of @var{priority} do not matter; only their
15393 @item java_interface
15394 @cindex @code{java_interface} attribute
15396 This type attribute informs C++ that the class is a Java interface. It may
15397 only be applied to classes declared within an @code{extern "Java"} block.
15398 Calls to methods declared in this interface will be dispatched using GCJ's
15399 interface table mechanism, instead of regular virtual table dispatch.
15403 See also @ref{Namespace Association}.
15405 @node Namespace Association
15406 @section Namespace Association
15408 @strong{Caution:} The semantics of this extension are not fully
15409 defined. Users should refrain from using this extension as its
15410 semantics may change subtly over time. It is possible that this
15411 extension will be removed in future versions of G++.
15413 A using-directive with @code{__attribute ((strong))} is stronger
15414 than a normal using-directive in two ways:
15418 Templates from the used namespace can be specialized and explicitly
15419 instantiated as though they were members of the using namespace.
15422 The using namespace is considered an associated namespace of all
15423 templates in the used namespace for purposes of argument-dependent
15427 The used namespace must be nested within the using namespace so that
15428 normal unqualified lookup works properly.
15430 This is useful for composing a namespace transparently from
15431 implementation namespaces. For example:
15436 template <class T> struct A @{ @};
15438 using namespace debug __attribute ((__strong__));
15439 template <> struct A<int> @{ @}; // @r{ok to specialize}
15441 template <class T> void f (A<T>);
15446 f (std::A<float>()); // @r{lookup finds} std::f
15452 @section Type Traits
15454 The C++ front-end implements syntactic extensions that allow to
15455 determine at compile time various characteristics of a type (or of a
15459 @item __has_nothrow_assign (type)
15460 If @code{type} is const qualified or is a reference type then the trait is
15461 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
15462 is true, else if @code{type} is a cv class or union type with copy assignment
15463 operators that are known not to throw an exception then the trait is true,
15464 else it is false. Requires: @code{type} shall be a complete type,
15465 (possibly cv-qualified) @code{void}, or an array of unknown bound.
15467 @item __has_nothrow_copy (type)
15468 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
15469 @code{type} is a cv class or union type with copy constructors that
15470 are known not to throw an exception then the trait is true, else it is false.
15471 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
15472 @code{void}, or an array of unknown bound.
15474 @item __has_nothrow_constructor (type)
15475 If @code{__has_trivial_constructor (type)} is true then the trait is
15476 true, else if @code{type} is a cv class or union type (or array
15477 thereof) with a default constructor that is known not to throw an
15478 exception then the trait is true, else it is false. Requires:
15479 @code{type} shall be a complete type, (possibly cv-qualified)
15480 @code{void}, or an array of unknown bound.
15482 @item __has_trivial_assign (type)
15483 If @code{type} is const qualified or is a reference type then the trait is
15484 false. Otherwise if @code{__is_pod (type)} is true then the trait is
15485 true, else if @code{type} is a cv class or union type with a trivial
15486 copy assignment ([class.copy]) then the trait is true, else it is
15487 false. Requires: @code{type} shall be a complete type, (possibly
15488 cv-qualified) @code{void}, or an array of unknown bound.
15490 @item __has_trivial_copy (type)
15491 If @code{__is_pod (type)} is true or @code{type} is a reference type
15492 then the trait is true, else if @code{type} is a cv class or union type
15493 with a trivial copy constructor ([class.copy]) then the trait
15494 is true, else it is false. Requires: @code{type} shall be a complete
15495 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15497 @item __has_trivial_constructor (type)
15498 If @code{__is_pod (type)} is true then the trait is true, else if
15499 @code{type} is a cv class or union type (or array thereof) with a
15500 trivial default constructor ([class.ctor]) then the trait is true,
15501 else it is false. Requires: @code{type} shall be a complete
15502 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15504 @item __has_trivial_destructor (type)
15505 If @code{__is_pod (type)} is true or @code{type} is a reference type then
15506 the trait is true, else if @code{type} is a cv class or union type (or
15507 array thereof) with a trivial destructor ([class.dtor]) then the trait
15508 is true, else it is false. Requires: @code{type} shall be a complete
15509 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15511 @item __has_virtual_destructor (type)
15512 If @code{type} is a class type with a virtual destructor
15513 ([class.dtor]) then the trait is true, else it is false. Requires:
15514 @code{type} shall be a complete type, (possibly cv-qualified)
15515 @code{void}, or an array of unknown bound.
15517 @item __is_abstract (type)
15518 If @code{type} is an abstract class ([class.abstract]) then the trait
15519 is true, else it is false. Requires: @code{type} shall be a complete
15520 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15522 @item __is_base_of (base_type, derived_type)
15523 If @code{base_type} is a base class of @code{derived_type}
15524 ([class.derived]) then the trait is true, otherwise it is false.
15525 Top-level cv qualifications of @code{base_type} and
15526 @code{derived_type} are ignored. For the purposes of this trait, a
15527 class type is considered is own base. Requires: if @code{__is_class
15528 (base_type)} and @code{__is_class (derived_type)} are true and
15529 @code{base_type} and @code{derived_type} are not the same type
15530 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
15531 type. Diagnostic is produced if this requirement is not met.
15533 @item __is_class (type)
15534 If @code{type} is a cv class type, and not a union type
15535 ([basic.compound]) the trait is true, else it is false.
15537 @item __is_empty (type)
15538 If @code{__is_class (type)} is false then the trait is false.
15539 Otherwise @code{type} is considered empty if and only if: @code{type}
15540 has no non-static data members, or all non-static data members, if
15541 any, are bit-fields of length 0, and @code{type} has no virtual
15542 members, and @code{type} has no virtual base classes, and @code{type}
15543 has no base classes @code{base_type} for which
15544 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
15545 be a complete type, (possibly cv-qualified) @code{void}, or an array
15548 @item __is_enum (type)
15549 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
15550 true, else it is false.
15552 @item __is_literal_type (type)
15553 If @code{type} is a literal type ([basic.types]) the trait is
15554 true, else it is false. Requires: @code{type} shall be a complete type,
15555 (possibly cv-qualified) @code{void}, or an array of unknown bound.
15557 @item __is_pod (type)
15558 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
15559 else it is false. Requires: @code{type} shall be a complete type,
15560 (possibly cv-qualified) @code{void}, or an array of unknown bound.
15562 @item __is_polymorphic (type)
15563 If @code{type} is a polymorphic class ([class.virtual]) then the trait
15564 is true, else it is false. Requires: @code{type} shall be a complete
15565 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15567 @item __is_standard_layout (type)
15568 If @code{type} is a standard-layout type ([basic.types]) the trait is
15569 true, else it is false. Requires: @code{type} shall be a complete
15570 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15572 @item __is_trivial (type)
15573 If @code{type} is a trivial type ([basic.types]) the trait is
15574 true, else it is false. Requires: @code{type} shall be a complete
15575 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15577 @item __is_union (type)
15578 If @code{type} is a cv union type ([basic.compound]) the trait is
15579 true, else it is false.
15581 @item __underlying_type (type)
15582 The underlying type of @code{type}. Requires: @code{type} shall be
15583 an enumeration type ([dcl.enum]).
15587 @node Java Exceptions
15588 @section Java Exceptions
15590 The Java language uses a slightly different exception handling model
15591 from C++. Normally, GNU C++ will automatically detect when you are
15592 writing C++ code that uses Java exceptions, and handle them
15593 appropriately. However, if C++ code only needs to execute destructors
15594 when Java exceptions are thrown through it, GCC will guess incorrectly.
15595 Sample problematic code is:
15598 struct S @{ ~S(); @};
15599 extern void bar(); // @r{is written in Java, and may throw exceptions}
15608 The usual effect of an incorrect guess is a link failure, complaining of
15609 a missing routine called @samp{__gxx_personality_v0}.
15611 You can inform the compiler that Java exceptions are to be used in a
15612 translation unit, irrespective of what it might think, by writing
15613 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
15614 @samp{#pragma} must appear before any functions that throw or catch
15615 exceptions, or run destructors when exceptions are thrown through them.
15617 You cannot mix Java and C++ exceptions in the same translation unit. It
15618 is believed to be safe to throw a C++ exception from one file through
15619 another file compiled for the Java exception model, or vice versa, but
15620 there may be bugs in this area.
15622 @node Deprecated Features
15623 @section Deprecated Features
15625 In the past, the GNU C++ compiler was extended to experiment with new
15626 features, at a time when the C++ language was still evolving. Now that
15627 the C++ standard is complete, some of those features are superseded by
15628 superior alternatives. Using the old features might cause a warning in
15629 some cases that the feature will be dropped in the future. In other
15630 cases, the feature might be gone already.
15632 While the list below is not exhaustive, it documents some of the options
15633 that are now deprecated:
15636 @item -fexternal-templates
15637 @itemx -falt-external-templates
15638 These are two of the many ways for G++ to implement template
15639 instantiation. @xref{Template Instantiation}. The C++ standard clearly
15640 defines how template definitions have to be organized across
15641 implementation units. G++ has an implicit instantiation mechanism that
15642 should work just fine for standard-conforming code.
15644 @item -fstrict-prototype
15645 @itemx -fno-strict-prototype
15646 Previously it was possible to use an empty prototype parameter list to
15647 indicate an unspecified number of parameters (like C), rather than no
15648 parameters, as C++ demands. This feature has been removed, except where
15649 it is required for backwards compatibility. @xref{Backwards Compatibility}.
15652 G++ allows a virtual function returning @samp{void *} to be overridden
15653 by one returning a different pointer type. This extension to the
15654 covariant return type rules is now deprecated and will be removed from a
15657 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
15658 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
15659 and are now removed from G++. Code using these operators should be
15660 modified to use @code{std::min} and @code{std::max} instead.
15662 The named return value extension has been deprecated, and is now
15665 The use of initializer lists with new expressions has been deprecated,
15666 and is now removed from G++.
15668 Floating and complex non-type template parameters have been deprecated,
15669 and are now removed from G++.
15671 The implicit typename extension has been deprecated and is now
15674 The use of default arguments in function pointers, function typedefs
15675 and other places where they are not permitted by the standard is
15676 deprecated and will be removed from a future version of G++.
15678 G++ allows floating-point literals to appear in integral constant expressions,
15679 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
15680 This extension is deprecated and will be removed from a future version.
15682 G++ allows static data members of const floating-point type to be declared
15683 with an initializer in a class definition. The standard only allows
15684 initializers for static members of const integral types and const
15685 enumeration types so this extension has been deprecated and will be removed
15686 from a future version.
15688 @node Backwards Compatibility
15689 @section Backwards Compatibility
15690 @cindex Backwards Compatibility
15691 @cindex ARM [Annotated C++ Reference Manual]
15693 Now that there is a definitive ISO standard C++, G++ has a specification
15694 to adhere to. The C++ language evolved over time, and features that
15695 used to be acceptable in previous drafts of the standard, such as the ARM
15696 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
15697 compilation of C++ written to such drafts, G++ contains some backwards
15698 compatibilities. @emph{All such backwards compatibility features are
15699 liable to disappear in future versions of G++.} They should be considered
15700 deprecated. @xref{Deprecated Features}.
15704 If a variable is declared at for scope, it used to remain in scope until
15705 the end of the scope which contained the for statement (rather than just
15706 within the for scope). G++ retains this, but issues a warning, if such a
15707 variable is accessed outside the for scope.
15709 @item Implicit C language
15710 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
15711 scope to set the language. On such systems, all header files are
15712 implicitly scoped inside a C language scope. Also, an empty prototype
15713 @code{()} will be treated as an unspecified number of arguments, rather
15714 than no arguments, as C++ demands.