1 c Copyright (C) 1988-2017 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Extensions to the C Language Family
8 @cindex extensions, C language
9 @cindex C language extensions
12 GNU C provides several language features not found in ISO standard C@.
13 (The @option{-pedantic} option directs GCC to print a warning message if
14 any of these features is used.) To test for the availability of these
15 features in conditional compilation, check for a predefined macro
16 @code{__GNUC__}, which is always defined under GCC@.
18 These extensions are available in C and Objective-C@. Most of them are
19 also available in C++. @xref{C++ Extensions,,Extensions to the
20 C++ Language}, for extensions that apply @emph{only} to C++.
22 Some features that are in ISO C99 but not C90 or C++ are also, as
23 extensions, accepted by GCC in C90 mode and in C++.
26 * Statement Exprs:: Putting statements and declarations inside expressions.
27 * Local Labels:: Labels local to a block.
28 * Labels as Values:: Getting pointers to labels, and computed gotos.
29 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
30 * Constructing Calls:: Dispatching a call to another function.
31 * Typeof:: @code{typeof}: referring to the type of an expression.
32 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
33 * __int128:: 128-bit integers---@code{__int128}.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Floating Types:: Additional Floating Types.
37 * Half-Precision:: Half-Precision Floating Point.
38 * Decimal Float:: Decimal Floating Types.
39 * Hex Floats:: Hexadecimal floating-point constants.
40 * Fixed-Point:: Fixed-Point Types.
41 * Named Address Spaces::Named address spaces.
42 * Zero Length:: Zero-length arrays.
43 * Empty Structures:: Structures with no members.
44 * Variable Length:: Arrays whose length is computed at run time.
45 * Variadic Macros:: Macros with a variable number of arguments.
46 * Escaped Newlines:: Slightly looser rules for escaped newlines.
47 * Subscripting:: Any array can be subscripted, even if not an lvalue.
48 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
49 * Pointers to Arrays:: Pointers to arrays with qualifiers work as expected.
50 * Initializers:: Non-constant initializers.
51 * Compound Literals:: Compound literals give structures, unions
53 * Designated Inits:: Labeling elements of initializers.
54 * Case Ranges:: `case 1 ... 9' and such.
55 * Cast to Union:: Casting to union type from any member of the union.
56 * Mixed Declarations:: Mixing declarations and code.
57 * Function Attributes:: Declaring that functions have no side effects,
58 or that they can never return.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Label Attributes:: Specifying attributes on labels.
62 * Enumerator Attributes:: Specifying attributes on enumerators.
63 * Statement Attributes:: Specifying attributes on statements.
64 * Attribute Syntax:: Formal syntax for attributes.
65 * Function Prototypes:: Prototype declarations and old-style definitions.
66 * C++ Comments:: C++ comments are recognized.
67 * Dollar Signs:: Dollar sign is allowed in identifiers.
68 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
69 * Alignment:: Inquiring about the alignment of a type or variable.
70 * Inline:: Defining inline functions (as fast as macros).
71 * Volatiles:: What constitutes an access to a volatile object.
72 * Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
73 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
74 * Incomplete Enums:: @code{enum foo;}, with details to follow.
75 * Function Names:: Printable strings which are the name of the current
77 * Return Address:: Getting the return or frame address of a function.
78 * Vector Extensions:: Using vector instructions through built-in functions.
79 * Offsetof:: Special syntax for implementing @code{offsetof}.
80 * __sync Builtins:: Legacy built-in functions for atomic memory access.
81 * __atomic Builtins:: Atomic built-in functions with memory model.
82 * Integer Overflow Builtins:: Built-in functions to perform arithmetics and
83 arithmetic overflow checking.
84 * x86 specific memory model extensions for transactional memory:: x86 memory models.
85 * Object Size Checking:: Built-in functions for limited buffer overflow
87 * Pointer Bounds Checker builtins:: Built-in functions for Pointer Bounds Checker.
88 * Cilk Plus Builtins:: Built-in functions for the Cilk Plus language extension.
89 * Other Builtins:: Other built-in functions.
90 * Target Builtins:: Built-in functions specific to particular targets.
91 * Target Format Checks:: Format checks specific to particular targets.
92 * Pragmas:: Pragmas accepted by GCC.
93 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
94 * Thread-Local:: Per-thread variables.
95 * Binary constants:: Binary constants using the @samp{0b} prefix.
99 @section Statements and Declarations in Expressions
100 @cindex statements inside expressions
101 @cindex declarations inside expressions
102 @cindex expressions containing statements
103 @cindex macros, statements in expressions
105 @c the above section title wrapped and causes an underfull hbox.. i
106 @c changed it from "within" to "in". --mew 4feb93
107 A compound statement enclosed in parentheses may appear as an expression
108 in GNU C@. This allows you to use loops, switches, and local variables
109 within an expression.
111 Recall that a compound statement is a sequence of statements surrounded
112 by braces; in this construct, parentheses go around the braces. For
116 (@{ int y = foo (); int z;
123 is a valid (though slightly more complex than necessary) expression
124 for the absolute value of @code{foo ()}.
126 The last thing in the compound statement should be an expression
127 followed by a semicolon; the value of this subexpression serves as the
128 value of the entire construct. (If you use some other kind of statement
129 last within the braces, the construct has type @code{void}, and thus
130 effectively no value.)
132 This feature is especially useful in making macro definitions ``safe'' (so
133 that they evaluate each operand exactly once). For example, the
134 ``maximum'' function is commonly defined as a macro in standard C as
138 #define max(a,b) ((a) > (b) ? (a) : (b))
142 @cindex side effects, macro argument
143 But this definition computes either @var{a} or @var{b} twice, with bad
144 results if the operand has side effects. In GNU C, if you know the
145 type of the operands (here taken as @code{int}), you can define
146 the macro safely as follows:
149 #define maxint(a,b) \
150 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
153 Embedded statements are not allowed in constant expressions, such as
154 the value of an enumeration constant, the width of a bit-field, or
155 the initial value of a static variable.
157 If you don't know the type of the operand, you can still do this, but you
158 must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).
160 In G++, the result value of a statement expression undergoes array and
161 function pointer decay, and is returned by value to the enclosing
162 expression. For instance, if @code{A} is a class, then
171 constructs a temporary @code{A} object to hold the result of the
172 statement expression, and that is used to invoke @code{Foo}.
173 Therefore the @code{this} pointer observed by @code{Foo} is not the
176 In a statement expression, any temporaries created within a statement
177 are destroyed at that statement's end. This makes statement
178 expressions inside macros slightly different from function calls. In
179 the latter case temporaries introduced during argument evaluation are
180 destroyed at the end of the statement that includes the function
181 call. In the statement expression case they are destroyed during
182 the statement expression. For instance,
185 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
186 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
196 has different places where temporaries are destroyed. For the
197 @code{macro} case, the temporary @code{X} is destroyed just after
198 the initialization of @code{b}. In the @code{function} case that
199 temporary is destroyed when the function returns.
201 These considerations mean that it is probably a bad idea to use
202 statement expressions of this form in header files that are designed to
203 work with C++. (Note that some versions of the GNU C Library contained
204 header files using statement expressions that lead to precisely this
207 Jumping into a statement expression with @code{goto} or using a
208 @code{switch} statement outside the statement expression with a
209 @code{case} or @code{default} label inside the statement expression is
210 not permitted. Jumping into a statement expression with a computed
211 @code{goto} (@pxref{Labels as Values}) has undefined behavior.
212 Jumping out of a statement expression is permitted, but if the
213 statement expression is part of a larger expression then it is
214 unspecified which other subexpressions of that expression have been
215 evaluated except where the language definition requires certain
216 subexpressions to be evaluated before or after the statement
217 expression. In any case, as with a function call, the evaluation of a
218 statement expression is not interleaved with the evaluation of other
219 parts of the containing expression. For example,
222 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
226 calls @code{foo} and @code{bar1} and does not call @code{baz} but
227 may or may not call @code{bar2}. If @code{bar2} is called, it is
228 called after @code{foo} and before @code{bar1}.
231 @section Locally Declared Labels
233 @cindex macros, local labels
235 GCC allows you to declare @dfn{local labels} in any nested block
236 scope. A local label is just like an ordinary label, but you can
237 only reference it (with a @code{goto} statement, or by taking its
238 address) within the block in which it is declared.
240 A local label declaration looks like this:
243 __label__ @var{label};
250 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
253 Local label declarations must come at the beginning of the block,
254 before any ordinary declarations or statements.
256 The label declaration defines the label @emph{name}, but does not define
257 the label itself. You must do this in the usual way, with
258 @code{@var{label}:}, within the statements of the statement expression.
260 The local label feature is useful for complex macros. If a macro
261 contains nested loops, a @code{goto} can be useful for breaking out of
262 them. However, an ordinary label whose scope is the whole function
263 cannot be used: if the macro can be expanded several times in one
264 function, the label is multiply defined in that function. A
265 local label avoids this problem. For example:
268 #define SEARCH(value, array, target) \
271 typeof (target) _SEARCH_target = (target); \
272 typeof (*(array)) *_SEARCH_array = (array); \
275 for (i = 0; i < max; i++) \
276 for (j = 0; j < max; j++) \
277 if (_SEARCH_array[i][j] == _SEARCH_target) \
278 @{ (value) = i; goto found; @} \
284 This could also be written using a statement expression:
287 #define SEARCH(array, target) \
290 typeof (target) _SEARCH_target = (target); \
291 typeof (*(array)) *_SEARCH_array = (array); \
294 for (i = 0; i < max; i++) \
295 for (j = 0; j < max; j++) \
296 if (_SEARCH_array[i][j] == _SEARCH_target) \
297 @{ value = i; goto found; @} \
304 Local label declarations also make the labels they declare visible to
305 nested functions, if there are any. @xref{Nested Functions}, for details.
307 @node Labels as Values
308 @section Labels as Values
309 @cindex labels as values
310 @cindex computed gotos
311 @cindex goto with computed label
312 @cindex address of a label
314 You can get the address of a label defined in the current function
315 (or a containing function) with the unary operator @samp{&&}. The
316 value has type @code{void *}. This value is a constant and can be used
317 wherever a constant of that type is valid. For example:
325 To use these values, you need to be able to jump to one. This is done
326 with the computed goto statement@footnote{The analogous feature in
327 Fortran is called an assigned goto, but that name seems inappropriate in
328 C, where one can do more than simply store label addresses in label
329 variables.}, @code{goto *@var{exp};}. For example,
336 Any expression of type @code{void *} is allowed.
338 One way of using these constants is in initializing a static array that
339 serves as a jump table:
342 static void *array[] = @{ &&foo, &&bar, &&hack @};
346 Then you can select a label with indexing, like this:
353 Note that this does not check whether the subscript is in bounds---array
354 indexing in C never does that.
356 Such an array of label values serves a purpose much like that of the
357 @code{switch} statement. The @code{switch} statement is cleaner, so
358 use that rather than an array unless the problem does not fit a
359 @code{switch} statement very well.
361 Another use of label values is in an interpreter for threaded code.
362 The labels within the interpreter function can be stored in the
363 threaded code for super-fast dispatching.
365 You may not use this mechanism to jump to code in a different function.
366 If you do that, totally unpredictable things happen. The best way to
367 avoid this is to store the label address only in automatic variables and
368 never pass it as an argument.
370 An alternate way to write the above example is
373 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
375 goto *(&&foo + array[i]);
379 This is more friendly to code living in shared libraries, as it reduces
380 the number of dynamic relocations that are needed, and by consequence,
381 allows the data to be read-only.
382 This alternative with label differences is not supported for the AVR target,
383 please use the first approach for AVR programs.
385 The @code{&&foo} expressions for the same label might have different
386 values if the containing function is inlined or cloned. If a program
387 relies on them being always the same,
388 @code{__attribute__((__noinline__,__noclone__))} should be used to
389 prevent inlining and cloning. If @code{&&foo} is used in a static
390 variable initializer, inlining and cloning is forbidden.
392 @node Nested Functions
393 @section Nested Functions
394 @cindex nested functions
395 @cindex downward funargs
398 A @dfn{nested function} is a function defined inside another function.
399 Nested functions are supported as an extension in GNU C, but are not
400 supported by GNU C++.
402 The nested function's name is local to the block where it is defined.
403 For example, here we define a nested function named @code{square}, and
408 foo (double a, double b)
410 double square (double z) @{ return z * z; @}
412 return square (a) + square (b);
417 The nested function can access all the variables of the containing
418 function that are visible at the point of its definition. This is
419 called @dfn{lexical scoping}. For example, here we show a nested
420 function which uses an inherited variable named @code{offset}:
424 bar (int *array, int offset, int size)
426 int access (int *array, int index)
427 @{ return array[index + offset]; @}
430 for (i = 0; i < size; i++)
431 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
436 Nested function definitions are permitted within functions in the places
437 where variable definitions are allowed; that is, in any block, mixed
438 with the other declarations and statements in the block.
440 It is possible to call the nested function from outside the scope of its
441 name by storing its address or passing the address to another function:
444 hack (int *array, int size)
446 void store (int index, int value)
447 @{ array[index] = value; @}
449 intermediate (store, size);
453 Here, the function @code{intermediate} receives the address of
454 @code{store} as an argument. If @code{intermediate} calls @code{store},
455 the arguments given to @code{store} are used to store into @code{array}.
456 But this technique works only so long as the containing function
457 (@code{hack}, in this example) does not exit.
459 If you try to call the nested function through its address after the
460 containing function exits, all hell breaks loose. If you try
461 to call it after a containing scope level exits, and if it refers
462 to some of the variables that are no longer in scope, you may be lucky,
463 but it's not wise to take the risk. If, however, the nested function
464 does not refer to anything that has gone out of scope, you should be
467 GCC implements taking the address of a nested function using a technique
468 called @dfn{trampolines}. This technique was described in
469 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
470 C++ Conference Proceedings, October 17-21, 1988).
472 A nested function can jump to a label inherited from a containing
473 function, provided the label is explicitly declared in the containing
474 function (@pxref{Local Labels}). Such a jump returns instantly to the
475 containing function, exiting the nested function that did the
476 @code{goto} and any intermediate functions as well. Here is an example:
480 bar (int *array, int offset, int size)
483 int access (int *array, int index)
487 return array[index + offset];
491 for (i = 0; i < size; i++)
492 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
496 /* @r{Control comes here from @code{access}
497 if it detects an error.} */
504 A nested function always has no linkage. Declaring one with
505 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
506 before its definition, use @code{auto} (which is otherwise meaningless
507 for function declarations).
510 bar (int *array, int offset, int size)
513 auto int access (int *, int);
515 int access (int *array, int index)
519 return array[index + offset];
525 @node Constructing Calls
526 @section Constructing Function Calls
527 @cindex constructing calls
528 @cindex forwarding calls
530 Using the built-in functions described below, you can record
531 the arguments a function received, and call another function
532 with the same arguments, without knowing the number or types
535 You can also record the return value of that function call,
536 and later return that value, without knowing what data type
537 the function tried to return (as long as your caller expects
540 However, these built-in functions may interact badly with some
541 sophisticated features or other extensions of the language. It
542 is, therefore, not recommended to use them outside very simple
543 functions acting as mere forwarders for their arguments.
545 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
546 This built-in function returns a pointer to data
547 describing how to perform a call with the same arguments as are passed
548 to the current function.
550 The function saves the arg pointer register, structure value address,
551 and all registers that might be used to pass arguments to a function
552 into a block of memory allocated on the stack. Then it returns the
553 address of that block.
556 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
557 This built-in function invokes @var{function}
558 with a copy of the parameters described by @var{arguments}
561 The value of @var{arguments} should be the value returned by
562 @code{__builtin_apply_args}. The argument @var{size} specifies the size
563 of the stack argument data, in bytes.
565 This function returns a pointer to data describing
566 how to return whatever value is returned by @var{function}. The data
567 is saved in a block of memory allocated on the stack.
569 It is not always simple to compute the proper value for @var{size}. The
570 value is used by @code{__builtin_apply} to compute the amount of data
571 that should be pushed on the stack and copied from the incoming argument
575 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
576 This built-in function returns the value described by @var{result} from
577 the containing function. You should specify, for @var{result}, a value
578 returned by @code{__builtin_apply}.
581 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
582 This built-in function represents all anonymous arguments of an inline
583 function. It can be used only in inline functions that are always
584 inlined, never compiled as a separate function, such as those using
585 @code{__attribute__ ((__always_inline__))} or
586 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
587 It must be only passed as last argument to some other function
588 with variable arguments. This is useful for writing small wrapper
589 inlines for variable argument functions, when using preprocessor
590 macros is undesirable. For example:
592 extern int myprintf (FILE *f, const char *format, ...);
593 extern inline __attribute__ ((__gnu_inline__)) int
594 myprintf (FILE *f, const char *format, ...)
596 int r = fprintf (f, "myprintf: ");
599 int s = fprintf (f, format, __builtin_va_arg_pack ());
607 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
608 This built-in function returns the number of anonymous arguments of
609 an inline function. It can be used only in inline functions that
610 are always inlined, never compiled as a separate function, such
611 as those using @code{__attribute__ ((__always_inline__))} or
612 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
613 For example following does link- or run-time checking of open
614 arguments for optimized code:
617 extern inline __attribute__((__gnu_inline__)) int
618 myopen (const char *path, int oflag, ...)
620 if (__builtin_va_arg_pack_len () > 1)
621 warn_open_too_many_arguments ();
623 if (__builtin_constant_p (oflag))
625 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
627 warn_open_missing_mode ();
628 return __open_2 (path, oflag);
630 return open (path, oflag, __builtin_va_arg_pack ());
633 if (__builtin_va_arg_pack_len () < 1)
634 return __open_2 (path, oflag);
636 return open (path, oflag, __builtin_va_arg_pack ());
643 @section Referring to a Type with @code{typeof}
646 @cindex macros, types of arguments
648 Another way to refer to the type of an expression is with @code{typeof}.
649 The syntax of using of this keyword looks like @code{sizeof}, but the
650 construct acts semantically like a type name defined with @code{typedef}.
652 There are two ways of writing the argument to @code{typeof}: with an
653 expression or with a type. Here is an example with an expression:
660 This assumes that @code{x} is an array of pointers to functions;
661 the type described is that of the values of the functions.
663 Here is an example with a typename as the argument:
670 Here the type described is that of pointers to @code{int}.
672 If you are writing a header file that must work when included in ISO C
673 programs, write @code{__typeof__} instead of @code{typeof}.
674 @xref{Alternate Keywords}.
676 A @code{typeof} construct can be used anywhere a typedef name can be
677 used. For example, you can use it in a declaration, in a cast, or inside
678 of @code{sizeof} or @code{typeof}.
680 The operand of @code{typeof} is evaluated for its side effects if and
681 only if it is an expression of variably modified type or the name of
684 @code{typeof} is often useful in conjunction with
685 statement expressions (@pxref{Statement Exprs}).
686 Here is how the two together can
687 be used to define a safe ``maximum'' macro which operates on any
688 arithmetic type and evaluates each of its arguments exactly once:
692 (@{ typeof (a) _a = (a); \
693 typeof (b) _b = (b); \
694 _a > _b ? _a : _b; @})
697 @cindex underscores in variables in macros
698 @cindex @samp{_} in variables in macros
699 @cindex local variables in macros
700 @cindex variables, local, in macros
701 @cindex macros, local variables in
703 The reason for using names that start with underscores for the local
704 variables is to avoid conflicts with variable names that occur within the
705 expressions that are substituted for @code{a} and @code{b}. Eventually we
706 hope to design a new form of declaration syntax that allows you to declare
707 variables whose scopes start only after their initializers; this will be a
708 more reliable way to prevent such conflicts.
711 Some more examples of the use of @code{typeof}:
715 This declares @code{y} with the type of what @code{x} points to.
722 This declares @code{y} as an array of such values.
729 This declares @code{y} as an array of pointers to characters:
732 typeof (typeof (char *)[4]) y;
736 It is equivalent to the following traditional C declaration:
742 To see the meaning of the declaration using @code{typeof}, and why it
743 might be a useful way to write, rewrite it with these macros:
746 #define pointer(T) typeof(T *)
747 #define array(T, N) typeof(T [N])
751 Now the declaration can be rewritten this way:
754 array (pointer (char), 4) y;
758 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
759 pointers to @code{char}.
762 In GNU C, but not GNU C++, you may also declare the type of a variable
763 as @code{__auto_type}. In that case, the declaration must declare
764 only one variable, whose declarator must just be an identifier, the
765 declaration must be initialized, and the type of the variable is
766 determined by the initializer; the name of the variable is not in
767 scope until after the initializer. (In C++, you should use C++11
768 @code{auto} for this purpose.) Using @code{__auto_type}, the
769 ``maximum'' macro above could be written as:
773 (@{ __auto_type _a = (a); \
774 __auto_type _b = (b); \
775 _a > _b ? _a : _b; @})
778 Using @code{__auto_type} instead of @code{typeof} has two advantages:
781 @item Each argument to the macro appears only once in the expansion of
782 the macro. This prevents the size of the macro expansion growing
783 exponentially when calls to such macros are nested inside arguments of
786 @item If the argument to the macro has variably modified type, it is
787 evaluated only once when using @code{__auto_type}, but twice if
788 @code{typeof} is used.
792 @section Conditionals with Omitted Operands
793 @cindex conditional expressions, extensions
794 @cindex omitted middle-operands
795 @cindex middle-operands, omitted
796 @cindex extensions, @code{?:}
797 @cindex @code{?:} extensions
799 The middle operand in a conditional expression may be omitted. Then
800 if the first operand is nonzero, its value is the value of the conditional
803 Therefore, the expression
810 has the value of @code{x} if that is nonzero; otherwise, the value of
813 This example is perfectly equivalent to
819 @cindex side effect in @code{?:}
820 @cindex @code{?:} side effect
822 In this simple case, the ability to omit the middle operand is not
823 especially useful. When it becomes useful is when the first operand does,
824 or may (if it is a macro argument), contain a side effect. Then repeating
825 the operand in the middle would perform the side effect twice. Omitting
826 the middle operand uses the value already computed without the undesirable
827 effects of recomputing it.
830 @section 128-bit Integers
831 @cindex @code{__int128} data types
833 As an extension the integer scalar type @code{__int128} is supported for
834 targets which have an integer mode wide enough to hold 128 bits.
835 Simply write @code{__int128} for a signed 128-bit integer, or
836 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
837 support in GCC for expressing an integer constant of type @code{__int128}
838 for targets with @code{long long} integer less than 128 bits wide.
841 @section Double-Word Integers
842 @cindex @code{long long} data types
843 @cindex double-word arithmetic
844 @cindex multiprecision arithmetic
845 @cindex @code{LL} integer suffix
846 @cindex @code{ULL} integer suffix
848 ISO C99 supports data types for integers that are at least 64 bits wide,
849 and as an extension GCC supports them in C90 mode and in C++.
850 Simply write @code{long long int} for a signed integer, or
851 @code{unsigned long long int} for an unsigned integer. To make an
852 integer constant of type @code{long long int}, add the suffix @samp{LL}
853 to the integer. To make an integer constant of type @code{unsigned long
854 long int}, add the suffix @samp{ULL} to the integer.
856 You can use these types in arithmetic like any other integer types.
857 Addition, subtraction, and bitwise boolean operations on these types
858 are open-coded on all types of machines. Multiplication is open-coded
859 if the machine supports a fullword-to-doubleword widening multiply
860 instruction. Division and shifts are open-coded only on machines that
861 provide special support. The operations that are not open-coded use
862 special library routines that come with GCC@.
864 There may be pitfalls when you use @code{long long} types for function
865 arguments without function prototypes. If a function
866 expects type @code{int} for its argument, and you pass a value of type
867 @code{long long int}, confusion results because the caller and the
868 subroutine disagree about the number of bytes for the argument.
869 Likewise, if the function expects @code{long long int} and you pass
870 @code{int}. The best way to avoid such problems is to use prototypes.
873 @section Complex Numbers
874 @cindex complex numbers
875 @cindex @code{_Complex} keyword
876 @cindex @code{__complex__} keyword
878 ISO C99 supports complex floating data types, and as an extension GCC
879 supports them in C90 mode and in C++. GCC also supports complex integer data
880 types which are not part of ISO C99. You can declare complex types
881 using the keyword @code{_Complex}. As an extension, the older GNU
882 keyword @code{__complex__} is also supported.
884 For example, @samp{_Complex double x;} declares @code{x} as a
885 variable whose real part and imaginary part are both of type
886 @code{double}. @samp{_Complex short int y;} declares @code{y} to
887 have real and imaginary parts of type @code{short int}; this is not
888 likely to be useful, but it shows that the set of complex types is
891 To write a constant with a complex data type, use the suffix @samp{i} or
892 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
893 has type @code{_Complex float} and @code{3i} has type
894 @code{_Complex int}. Such a constant always has a pure imaginary
895 value, but you can form any complex value you like by adding one to a
896 real constant. This is a GNU extension; if you have an ISO C99
897 conforming C library (such as the GNU C Library), and want to construct complex
898 constants of floating type, you should include @code{<complex.h>} and
899 use the macros @code{I} or @code{_Complex_I} instead.
901 @cindex @code{__real__} keyword
902 @cindex @code{__imag__} keyword
903 To extract the real part of a complex-valued expression @var{exp}, write
904 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
905 extract the imaginary part. This is a GNU extension; for values of
906 floating type, you should use the ISO C99 functions @code{crealf},
907 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
908 @code{cimagl}, declared in @code{<complex.h>} and also provided as
909 built-in functions by GCC@.
911 @cindex complex conjugation
912 The operator @samp{~} performs complex conjugation when used on a value
913 with a complex type. This is a GNU extension; for values of
914 floating type, you should use the ISO C99 functions @code{conjf},
915 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
916 provided as built-in functions by GCC@.
918 GCC can allocate complex automatic variables in a noncontiguous
919 fashion; it's even possible for the real part to be in a register while
920 the imaginary part is on the stack (or vice versa). Only the DWARF
921 debug info format can represent this, so use of DWARF is recommended.
922 If you are using the stabs debug info format, GCC describes a noncontiguous
923 complex variable as if it were two separate variables of noncomplex type.
924 If the variable's actual name is @code{foo}, the two fictitious
925 variables are named @code{foo$real} and @code{foo$imag}. You can
926 examine and set these two fictitious variables with your debugger.
929 @section Additional Floating Types
930 @cindex additional floating types
931 @cindex @code{_Float@var{n}} data types
932 @cindex @code{_Float@var{n}x} data types
933 @cindex @code{__float80} data type
934 @cindex @code{__float128} data type
935 @cindex @code{__ibm128} data type
936 @cindex @code{w} floating point suffix
937 @cindex @code{q} floating point suffix
938 @cindex @code{W} floating point suffix
939 @cindex @code{Q} floating point suffix
941 ISO/IEC TS 18661-3:2015 defines C support for additional floating
942 types @code{_Float@var{n}} and @code{_Float@var{n}x}, and GCC supports
943 these type names; the set of types supported depends on the target
944 architecture. These types are not supported when compiling C++.
945 Constants with these types use suffixes @code{f@var{n}} or
946 @code{F@var{n}} and @code{f@var{n}x} or @code{F@var{n}x}. These type
947 names can be used together with @code{_Complex} to declare complex
950 As an extension, GNU C and GNU C++ support additional floating
951 types, which are not supported by all targets.
953 @item @code{__float128} is available on i386, x86_64, IA-64, and
954 hppa HP-UX, as well as on PowerPC GNU/Linux targets that enable
955 the vector scalar (VSX) instruction set. @code{__float128} supports
956 the 128-bit floating type. On i386, x86_64, PowerPC, and IA-64
957 other than HP-UX, @code{__float128} is an alias for @code{_Float128}.
958 On hppa and IA-64 HP-UX, @code{__float128} is an alias for @code{long
961 @item @code{__float80} is available on the i386, x86_64, and IA-64
962 targets, and supports the 80-bit (@code{XFmode}) floating type. It is
963 an alias for the type name @code{_Float64x} on these targets.
965 @item @code{__ibm128} is available on PowerPC targets, and provides
966 access to the IBM extended double format which is the current format
967 used for @code{long double}. When @code{long double} transitions to
968 @code{__float128} on PowerPC in the future, @code{__ibm128} will remain
969 for use in conversions between the two types.
972 Support for these additional types includes the arithmetic operators:
973 add, subtract, multiply, divide; unary arithmetic operators;
974 relational operators; equality operators; and conversions to and from
975 integer and other floating types. Use a suffix @samp{w} or @samp{W}
976 in a literal constant of type @code{__float80} or type
977 @code{__ibm128}. Use a suffix @samp{q} or @samp{Q} for @code{_float128}.
979 In order to use @code{_Float128}, @code{__float128}, and @code{__ibm128}
980 on PowerPC Linux systems, you must use the @option{-mfloat128} option. It is
981 expected in future versions of GCC that @code{_Float128} and @code{__float128}
982 will be enabled automatically.
984 The @code{_Float128} type is supported on all systems where
985 @code{__float128} is supported or where @code{long double} has the
986 IEEE binary128 format. The @code{_Float64x} type is supported on all
987 systems where @code{__float128} is supported. The @code{_Float32}
988 type is supported on all systems supporting IEEE binary32; the
989 @code{_Float64} and @code{_Float32x} types are supported on all systems
990 supporting IEEE binary64. The @code{_Float16} type is supported on AArch64
991 systems by default, and on ARM systems when the IEEE format for 16-bit
992 floating-point types is selected with @option{-mfp16-format=ieee}.
993 GCC does not currently support @code{_Float128x} on any systems.
995 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
996 types using the corresponding internal complex type, @code{XCmode} for
997 @code{__float80} type and @code{TCmode} for @code{__float128} type:
1000 typedef _Complex float __attribute__((mode(TC))) _Complex128;
1001 typedef _Complex float __attribute__((mode(XC))) _Complex80;
1004 On the PowerPC Linux VSX targets, you can declare complex types using
1005 the corresponding internal complex type, @code{KCmode} for
1006 @code{__float128} type and @code{ICmode} for @code{__ibm128} type:
1009 typedef _Complex float __attribute__((mode(KC))) _Complex_float128;
1010 typedef _Complex float __attribute__((mode(IC))) _Complex_ibm128;
1013 @node Half-Precision
1014 @section Half-Precision Floating Point
1015 @cindex half-precision floating point
1016 @cindex @code{__fp16} data type
1018 On ARM and AArch64 targets, GCC supports half-precision (16-bit) floating
1019 point via the @code{__fp16} type defined in the ARM C Language Extensions.
1020 On ARM systems, you must enable this type explicitly with the
1021 @option{-mfp16-format} command-line option in order to use it.
1023 ARM targets support two incompatible representations for half-precision
1024 floating-point values. You must choose one of the representations and
1025 use it consistently in your program.
1027 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
1028 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
1029 There are 11 bits of significand precision, approximately 3
1032 Specifying @option{-mfp16-format=alternative} selects the ARM
1033 alternative format. This representation is similar to the IEEE
1034 format, but does not support infinities or NaNs. Instead, the range
1035 of exponents is extended, so that this format can represent normalized
1036 values in the range of @math{2^{-14}} to 131008.
1038 The GCC port for AArch64 only supports the IEEE 754-2008 format, and does
1039 not require use of the @option{-mfp16-format} command-line option.
1041 The @code{__fp16} type may only be used as an argument to intrinsics defined
1042 in @code{<arm_fp16.h>}, or as a storage format. For purposes of
1043 arithmetic and other operations, @code{__fp16} values in C or C++
1044 expressions are automatically promoted to @code{float}.
1046 The ARM target provides hardware support for conversions between
1047 @code{__fp16} and @code{float} values
1048 as an extension to VFP and NEON (Advanced SIMD), and from ARMv8 provides
1049 hardware support for conversions between @code{__fp16} and @code{double}
1050 values. GCC generates code using these hardware instructions if you
1051 compile with options to select an FPU that provides them;
1052 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1053 in addition to the @option{-mfp16-format} option to select
1054 a half-precision format.
1056 Language-level support for the @code{__fp16} data type is
1057 independent of whether GCC generates code using hardware floating-point
1058 instructions. In cases where hardware support is not specified, GCC
1059 implements conversions between @code{__fp16} and other types as library
1062 It is recommended that portable code use the @code{_Float16} type defined
1063 by ISO/IEC TS 18661-3:2015. @xref{Floating Types}.
1066 @section Decimal Floating Types
1067 @cindex decimal floating types
1068 @cindex @code{_Decimal32} data type
1069 @cindex @code{_Decimal64} data type
1070 @cindex @code{_Decimal128} data type
1071 @cindex @code{df} integer suffix
1072 @cindex @code{dd} integer suffix
1073 @cindex @code{dl} integer suffix
1074 @cindex @code{DF} integer suffix
1075 @cindex @code{DD} integer suffix
1076 @cindex @code{DL} integer suffix
1078 As an extension, GNU C supports decimal floating types as
1079 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1080 floating types in GCC will evolve as the draft technical report changes.
1081 Calling conventions for any target might also change. Not all targets
1082 support decimal floating types.
1084 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1085 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1086 @code{float}, @code{double}, and @code{long double} whose radix is not
1087 specified by the C standard but is usually two.
1089 Support for decimal floating types includes the arithmetic operators
1090 add, subtract, multiply, divide; unary arithmetic operators;
1091 relational operators; equality operators; and conversions to and from
1092 integer and other floating types. Use a suffix @samp{df} or
1093 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1094 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1097 GCC support of decimal float as specified by the draft technical report
1102 When the value of a decimal floating type cannot be represented in the
1103 integer type to which it is being converted, the result is undefined
1104 rather than the result value specified by the draft technical report.
1107 GCC does not provide the C library functionality associated with
1108 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1109 @file{wchar.h}, which must come from a separate C library implementation.
1110 Because of this the GNU C compiler does not define macro
1111 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1112 the technical report.
1115 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1116 are supported by the DWARF debug information format.
1122 ISO C99 supports floating-point numbers written not only in the usual
1123 decimal notation, such as @code{1.55e1}, but also numbers such as
1124 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1125 supports this in C90 mode (except in some cases when strictly
1126 conforming) and in C++. In that format the
1127 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1128 mandatory. The exponent is a decimal number that indicates the power of
1129 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1136 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1137 is the same as @code{1.55e1}.
1139 Unlike for floating-point numbers in the decimal notation the exponent
1140 is always required in the hexadecimal notation. Otherwise the compiler
1141 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1142 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1143 extension for floating-point constants of type @code{float}.
1146 @section Fixed-Point Types
1147 @cindex fixed-point types
1148 @cindex @code{_Fract} data type
1149 @cindex @code{_Accum} data type
1150 @cindex @code{_Sat} data type
1151 @cindex @code{hr} fixed-suffix
1152 @cindex @code{r} fixed-suffix
1153 @cindex @code{lr} fixed-suffix
1154 @cindex @code{llr} fixed-suffix
1155 @cindex @code{uhr} fixed-suffix
1156 @cindex @code{ur} fixed-suffix
1157 @cindex @code{ulr} fixed-suffix
1158 @cindex @code{ullr} fixed-suffix
1159 @cindex @code{hk} fixed-suffix
1160 @cindex @code{k} fixed-suffix
1161 @cindex @code{lk} fixed-suffix
1162 @cindex @code{llk} fixed-suffix
1163 @cindex @code{uhk} fixed-suffix
1164 @cindex @code{uk} fixed-suffix
1165 @cindex @code{ulk} fixed-suffix
1166 @cindex @code{ullk} fixed-suffix
1167 @cindex @code{HR} fixed-suffix
1168 @cindex @code{R} fixed-suffix
1169 @cindex @code{LR} fixed-suffix
1170 @cindex @code{LLR} fixed-suffix
1171 @cindex @code{UHR} fixed-suffix
1172 @cindex @code{UR} fixed-suffix
1173 @cindex @code{ULR} fixed-suffix
1174 @cindex @code{ULLR} fixed-suffix
1175 @cindex @code{HK} fixed-suffix
1176 @cindex @code{K} fixed-suffix
1177 @cindex @code{LK} fixed-suffix
1178 @cindex @code{LLK} fixed-suffix
1179 @cindex @code{UHK} fixed-suffix
1180 @cindex @code{UK} fixed-suffix
1181 @cindex @code{ULK} fixed-suffix
1182 @cindex @code{ULLK} fixed-suffix
1184 As an extension, GNU C supports fixed-point types as
1185 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1186 types in GCC will evolve as the draft technical report changes.
1187 Calling conventions for any target might also change. Not all targets
1188 support fixed-point types.
1190 The fixed-point types are
1191 @code{short _Fract},
1194 @code{long long _Fract},
1195 @code{unsigned short _Fract},
1196 @code{unsigned _Fract},
1197 @code{unsigned long _Fract},
1198 @code{unsigned long long _Fract},
1199 @code{_Sat short _Fract},
1201 @code{_Sat long _Fract},
1202 @code{_Sat long long _Fract},
1203 @code{_Sat unsigned short _Fract},
1204 @code{_Sat unsigned _Fract},
1205 @code{_Sat unsigned long _Fract},
1206 @code{_Sat unsigned long long _Fract},
1207 @code{short _Accum},
1210 @code{long long _Accum},
1211 @code{unsigned short _Accum},
1212 @code{unsigned _Accum},
1213 @code{unsigned long _Accum},
1214 @code{unsigned long long _Accum},
1215 @code{_Sat short _Accum},
1217 @code{_Sat long _Accum},
1218 @code{_Sat long long _Accum},
1219 @code{_Sat unsigned short _Accum},
1220 @code{_Sat unsigned _Accum},
1221 @code{_Sat unsigned long _Accum},
1222 @code{_Sat unsigned long long _Accum}.
1224 Fixed-point data values contain fractional and optional integral parts.
1225 The format of fixed-point data varies and depends on the target machine.
1227 Support for fixed-point types includes:
1230 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1232 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1234 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1236 binary shift operators (@code{<<}, @code{>>})
1238 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1240 equality operators (@code{==}, @code{!=})
1242 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1243 @code{<<=}, @code{>>=})
1245 conversions to and from integer, floating-point, or fixed-point types
1248 Use a suffix in a fixed-point literal constant:
1250 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1251 @code{_Sat short _Fract}
1252 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1253 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1254 @code{_Sat long _Fract}
1255 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1256 @code{_Sat long long _Fract}
1257 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1258 @code{_Sat unsigned short _Fract}
1259 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1260 @code{_Sat unsigned _Fract}
1261 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1262 @code{_Sat unsigned long _Fract}
1263 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1264 and @code{_Sat unsigned long long _Fract}
1265 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1266 @code{_Sat short _Accum}
1267 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1268 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1269 @code{_Sat long _Accum}
1270 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1271 @code{_Sat long long _Accum}
1272 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1273 @code{_Sat unsigned short _Accum}
1274 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1275 @code{_Sat unsigned _Accum}
1276 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1277 @code{_Sat unsigned long _Accum}
1278 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1279 and @code{_Sat unsigned long long _Accum}
1282 GCC support of fixed-point types as specified by the draft technical report
1287 Pragmas to control overflow and rounding behaviors are not implemented.
1290 Fixed-point types are supported by the DWARF debug information format.
1292 @node Named Address Spaces
1293 @section Named Address Spaces
1294 @cindex Named Address Spaces
1296 As an extension, GNU C supports named address spaces as
1297 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1298 address spaces in GCC will evolve as the draft technical report
1299 changes. Calling conventions for any target might also change. At
1300 present, only the AVR, SPU, M32C, RL78, and x86 targets support
1301 address spaces other than the generic address space.
1303 Address space identifiers may be used exactly like any other C type
1304 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1305 document for more details.
1307 @anchor{AVR Named Address Spaces}
1308 @subsection AVR Named Address Spaces
1310 On the AVR target, there are several address spaces that can be used
1311 in order to put read-only data into the flash memory and access that
1312 data by means of the special instructions @code{LPM} or @code{ELPM}
1313 needed to read from flash.
1315 Devices belonging to @code{avrtiny} and @code{avrxmega3} can access
1316 flash memory by means of @code{LD*} instructions because the flash
1317 memory is mapped into the RAM address space. There is @emph{no need}
1318 for language extensions like @code{__flash} or attribute
1319 @ref{AVR Variable Attributes,,@code{progmem}}.
1320 The default linker description files for these devices cater for that
1321 feature and @code{.rodata} stays in flash: The compiler just generates
1322 @code{LD*} instructions, and the linker script adds core specific
1323 offsets to all @code{.rodata} symbols: @code{0x4000} in the case of
1324 @code{avrtiny} and @code{0x8000} in the case of @code{avrxmega3}.
1325 See @ref{AVR Options} for a list of respective devices.
1327 For devices not in @code{avrtiny} or @code{avrxmega3},
1328 any data including read-only data is located in RAM (the generic
1329 address space) because flash memory is not visible in the RAM address
1330 space. In order to locate read-only data in flash memory @emph{and}
1331 to generate the right instructions to access this data without
1332 using (inline) assembler code, special address spaces are needed.
1336 @cindex @code{__flash} AVR Named Address Spaces
1337 The @code{__flash} qualifier locates data in the
1338 @code{.progmem.data} section. Data is read using the @code{LPM}
1339 instruction. Pointers to this address space are 16 bits wide.
1346 @cindex @code{__flash1} AVR Named Address Spaces
1347 @cindex @code{__flash2} AVR Named Address Spaces
1348 @cindex @code{__flash3} AVR Named Address Spaces
1349 @cindex @code{__flash4} AVR Named Address Spaces
1350 @cindex @code{__flash5} AVR Named Address Spaces
1351 These are 16-bit address spaces locating data in section
1352 @code{.progmem@var{N}.data} where @var{N} refers to
1353 address space @code{__flash@var{N}}.
1354 The compiler sets the @code{RAMPZ} segment register appropriately
1355 before reading data by means of the @code{ELPM} instruction.
1358 @cindex @code{__memx} AVR Named Address Spaces
1359 This is a 24-bit address space that linearizes flash and RAM:
1360 If the high bit of the address is set, data is read from
1361 RAM using the lower two bytes as RAM address.
1362 If the high bit of the address is clear, data is read from flash
1363 with @code{RAMPZ} set according to the high byte of the address.
1364 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1366 Objects in this address space are located in @code{.progmemx.data}.
1372 char my_read (const __flash char ** p)
1374 /* p is a pointer to RAM that points to a pointer to flash.
1375 The first indirection of p reads that flash pointer
1376 from RAM and the second indirection reads a char from this
1382 /* Locate array[] in flash memory */
1383 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1389 /* Return 17 by reading from flash memory */
1390 return array[array[i]];
1395 For each named address space supported by avr-gcc there is an equally
1396 named but uppercase built-in macro defined.
1397 The purpose is to facilitate testing if respective address space
1398 support is available or not:
1402 const __flash int var = 1;
1409 #include <avr/pgmspace.h> /* From AVR-LibC */
1411 const int var PROGMEM = 1;
1415 return (int) pgm_read_word (&var);
1417 #endif /* __FLASH */
1421 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1422 locates data in flash but
1423 accesses to these data read from generic address space, i.e.@:
1425 so that you need special accessors like @code{pgm_read_byte}
1426 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1427 together with attribute @code{progmem}.
1430 @b{Limitations and caveats}
1434 Reading across the 64@tie{}KiB section boundary of
1435 the @code{__flash} or @code{__flash@var{N}} address spaces
1436 shows undefined behavior. The only address space that
1437 supports reading across the 64@tie{}KiB flash segment boundaries is
1441 If you use one of the @code{__flash@var{N}} address spaces
1442 you must arrange your linker script to locate the
1443 @code{.progmem@var{N}.data} sections according to your needs.
1446 Any data or pointers to the non-generic address spaces must
1447 be qualified as @code{const}, i.e.@: as read-only data.
1448 This still applies if the data in one of these address
1449 spaces like software version number or calibration lookup table are intended to
1450 be changed after load time by, say, a boot loader. In this case
1451 the right qualification is @code{const} @code{volatile} so that the compiler
1452 must not optimize away known values or insert them
1453 as immediates into operands of instructions.
1456 The following code initializes a variable @code{pfoo}
1457 located in static storage with a 24-bit address:
1459 extern const __memx char foo;
1460 const __memx void *pfoo = &foo;
1464 On the reduced Tiny devices like ATtiny40, no address spaces are supported.
1465 Just use vanilla C / C++ code without overhead as outlined above.
1466 Attribute @code{progmem} is supported but works differently,
1467 see @ref{AVR Variable Attributes}.
1471 @subsection M32C Named Address Spaces
1472 @cindex @code{__far} M32C Named Address Spaces
1474 On the M32C target, with the R8C and M16C CPU variants, variables
1475 qualified with @code{__far} are accessed using 32-bit addresses in
1476 order to access memory beyond the first 64@tie{}Ki bytes. If
1477 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1480 @subsection RL78 Named Address Spaces
1481 @cindex @code{__far} RL78 Named Address Spaces
1483 On the RL78 target, variables qualified with @code{__far} are accessed
1484 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1485 addresses. Non-far variables are assumed to appear in the topmost
1486 64@tie{}KiB of the address space.
1488 @subsection SPU Named Address Spaces
1489 @cindex @code{__ea} SPU Named Address Spaces
1491 On the SPU target variables may be declared as
1492 belonging to another address space by qualifying the type with the
1493 @code{__ea} address space identifier:
1500 The compiler generates special code to access the variable @code{i}.
1501 It may use runtime library
1502 support, or generate special machine instructions to access that address
1505 @subsection x86 Named Address Spaces
1506 @cindex x86 named address spaces
1508 On the x86 target, variables may be declared as being relative
1509 to the @code{%fs} or @code{%gs} segments.
1514 @cindex @code{__seg_fs} x86 named address space
1515 @cindex @code{__seg_gs} x86 named address space
1516 The object is accessed with the respective segment override prefix.
1518 The respective segment base must be set via some method specific to
1519 the operating system. Rather than require an expensive system call
1520 to retrieve the segment base, these address spaces are not considered
1521 to be subspaces of the generic (flat) address space. This means that
1522 explicit casts are required to convert pointers between these address
1523 spaces and the generic address space. In practice the application
1524 should cast to @code{uintptr_t} and apply the segment base offset
1525 that it installed previously.
1527 The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
1528 defined when these address spaces are supported.
1532 @section Arrays of Length Zero
1533 @cindex arrays of length zero
1534 @cindex zero-length arrays
1535 @cindex length-zero arrays
1536 @cindex flexible array members
1538 Zero-length arrays are allowed in GNU C@. They are very useful as the
1539 last element of a structure that is really a header for a variable-length
1548 struct line *thisline = (struct line *)
1549 malloc (sizeof (struct line) + this_length);
1550 thisline->length = this_length;
1553 In ISO C90, you would have to give @code{contents} a length of 1, which
1554 means either you waste space or complicate the argument to @code{malloc}.
1556 In ISO C99, you would use a @dfn{flexible array member}, which is
1557 slightly different in syntax and semantics:
1561 Flexible array members are written as @code{contents[]} without
1565 Flexible array members have incomplete type, and so the @code{sizeof}
1566 operator may not be applied. As a quirk of the original implementation
1567 of zero-length arrays, @code{sizeof} evaluates to zero.
1570 Flexible array members may only appear as the last member of a
1571 @code{struct} that is otherwise non-empty.
1574 A structure containing a flexible array member, or a union containing
1575 such a structure (possibly recursively), may not be a member of a
1576 structure or an element of an array. (However, these uses are
1577 permitted by GCC as extensions.)
1580 Non-empty initialization of zero-length
1581 arrays is treated like any case where there are more initializer
1582 elements than the array holds, in that a suitable warning about ``excess
1583 elements in array'' is given, and the excess elements (all of them, in
1584 this case) are ignored.
1586 GCC allows static initialization of flexible array members.
1587 This is equivalent to defining a new structure containing the original
1588 structure followed by an array of sufficient size to contain the data.
1589 E.g.@: in the following, @code{f1} is constructed as if it were declared
1595 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1598 struct f1 f1; int data[3];
1599 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1603 The convenience of this extension is that @code{f1} has the desired
1604 type, eliminating the need to consistently refer to @code{f2.f1}.
1606 This has symmetry with normal static arrays, in that an array of
1607 unknown size is also written with @code{[]}.
1609 Of course, this extension only makes sense if the extra data comes at
1610 the end of a top-level object, as otherwise we would be overwriting
1611 data at subsequent offsets. To avoid undue complication and confusion
1612 with initialization of deeply nested arrays, we simply disallow any
1613 non-empty initialization except when the structure is the top-level
1614 object. For example:
1617 struct foo @{ int x; int y[]; @};
1618 struct bar @{ struct foo z; @};
1620 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1621 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1622 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1623 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1626 @node Empty Structures
1627 @section Structures with No Members
1628 @cindex empty structures
1629 @cindex zero-size structures
1631 GCC permits a C structure to have no members:
1638 The structure has size zero. In C++, empty structures are part
1639 of the language. G++ treats empty structures as if they had a single
1640 member of type @code{char}.
1642 @node Variable Length
1643 @section Arrays of Variable Length
1644 @cindex variable-length arrays
1645 @cindex arrays of variable length
1648 Variable-length automatic arrays are allowed in ISO C99, and as an
1649 extension GCC accepts them in C90 mode and in C++. These arrays are
1650 declared like any other automatic arrays, but with a length that is not
1651 a constant expression. The storage is allocated at the point of
1652 declaration and deallocated when the block scope containing the declaration
1658 concat_fopen (char *s1, char *s2, char *mode)
1660 char str[strlen (s1) + strlen (s2) + 1];
1663 return fopen (str, mode);
1667 @cindex scope of a variable length array
1668 @cindex variable-length array scope
1669 @cindex deallocating variable length arrays
1670 Jumping or breaking out of the scope of the array name deallocates the
1671 storage. Jumping into the scope is not allowed; you get an error
1674 @cindex variable-length array in a structure
1675 As an extension, GCC accepts variable-length arrays as a member of
1676 a structure or a union. For example:
1682 struct S @{ int x[n]; @};
1686 @cindex @code{alloca} vs variable-length arrays
1687 You can use the function @code{alloca} to get an effect much like
1688 variable-length arrays. The function @code{alloca} is available in
1689 many other C implementations (but not in all). On the other hand,
1690 variable-length arrays are more elegant.
1692 There are other differences between these two methods. Space allocated
1693 with @code{alloca} exists until the containing @emph{function} returns.
1694 The space for a variable-length array is deallocated as soon as the array
1695 name's scope ends, unless you also use @code{alloca} in this scope.
1697 You can also use variable-length arrays as arguments to functions:
1701 tester (int len, char data[len][len])
1707 The length of an array is computed once when the storage is allocated
1708 and is remembered for the scope of the array in case you access it with
1711 If you want to pass the array first and the length afterward, you can
1712 use a forward declaration in the parameter list---another GNU extension.
1716 tester (int len; char data[len][len], int len)
1722 @cindex parameter forward declaration
1723 The @samp{int len} before the semicolon is a @dfn{parameter forward
1724 declaration}, and it serves the purpose of making the name @code{len}
1725 known when the declaration of @code{data} is parsed.
1727 You can write any number of such parameter forward declarations in the
1728 parameter list. They can be separated by commas or semicolons, but the
1729 last one must end with a semicolon, which is followed by the ``real''
1730 parameter declarations. Each forward declaration must match a ``real''
1731 declaration in parameter name and data type. ISO C99 does not support
1732 parameter forward declarations.
1734 @node Variadic Macros
1735 @section Macros with a Variable Number of Arguments.
1736 @cindex variable number of arguments
1737 @cindex macro with variable arguments
1738 @cindex rest argument (in macro)
1739 @cindex variadic macros
1741 In the ISO C standard of 1999, a macro can be declared to accept a
1742 variable number of arguments much as a function can. The syntax for
1743 defining the macro is similar to that of a function. Here is an
1747 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1751 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1752 such a macro, it represents the zero or more tokens until the closing
1753 parenthesis that ends the invocation, including any commas. This set of
1754 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1755 wherever it appears. See the CPP manual for more information.
1757 GCC has long supported variadic macros, and used a different syntax that
1758 allowed you to give a name to the variable arguments just like any other
1759 argument. Here is an example:
1762 #define debug(format, args...) fprintf (stderr, format, args)
1766 This is in all ways equivalent to the ISO C example above, but arguably
1767 more readable and descriptive.
1769 GNU CPP has two further variadic macro extensions, and permits them to
1770 be used with either of the above forms of macro definition.
1772 In standard C, you are not allowed to leave the variable argument out
1773 entirely; but you are allowed to pass an empty argument. For example,
1774 this invocation is invalid in ISO C, because there is no comma after
1781 GNU CPP permits you to completely omit the variable arguments in this
1782 way. In the above examples, the compiler would complain, though since
1783 the expansion of the macro still has the extra comma after the format
1786 To help solve this problem, CPP behaves specially for variable arguments
1787 used with the token paste operator, @samp{##}. If instead you write
1790 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1794 and if the variable arguments are omitted or empty, the @samp{##}
1795 operator causes the preprocessor to remove the comma before it. If you
1796 do provide some variable arguments in your macro invocation, GNU CPP
1797 does not complain about the paste operation and instead places the
1798 variable arguments after the comma. Just like any other pasted macro
1799 argument, these arguments are not macro expanded.
1801 @node Escaped Newlines
1802 @section Slightly Looser Rules for Escaped Newlines
1803 @cindex escaped newlines
1804 @cindex newlines (escaped)
1806 The preprocessor treatment of escaped newlines is more relaxed
1807 than that specified by the C90 standard, which requires the newline
1808 to immediately follow a backslash.
1809 GCC's implementation allows whitespace in the form
1810 of spaces, horizontal and vertical tabs, and form feeds between the
1811 backslash and the subsequent newline. The preprocessor issues a
1812 warning, but treats it as a valid escaped newline and combines the two
1813 lines to form a single logical line. This works within comments and
1814 tokens, as well as between tokens. Comments are @emph{not} treated as
1815 whitespace for the purposes of this relaxation, since they have not
1816 yet been replaced with spaces.
1819 @section Non-Lvalue Arrays May Have Subscripts
1820 @cindex subscripting
1821 @cindex arrays, non-lvalue
1823 @cindex subscripting and function values
1824 In ISO C99, arrays that are not lvalues still decay to pointers, and
1825 may be subscripted, although they may not be modified or used after
1826 the next sequence point and the unary @samp{&} operator may not be
1827 applied to them. As an extension, GNU C allows such arrays to be
1828 subscripted in C90 mode, though otherwise they do not decay to
1829 pointers outside C99 mode. For example,
1830 this is valid in GNU C though not valid in C90:
1834 struct foo @{int a[4];@};
1840 return f().a[index];
1846 @section Arithmetic on @code{void}- and Function-Pointers
1847 @cindex void pointers, arithmetic
1848 @cindex void, size of pointer to
1849 @cindex function pointers, arithmetic
1850 @cindex function, size of pointer to
1852 In GNU C, addition and subtraction operations are supported on pointers to
1853 @code{void} and on pointers to functions. This is done by treating the
1854 size of a @code{void} or of a function as 1.
1856 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1857 and on function types, and returns 1.
1859 @opindex Wpointer-arith
1860 The option @option{-Wpointer-arith} requests a warning if these extensions
1863 @node Pointers to Arrays
1864 @section Pointers to Arrays with Qualifiers Work as Expected
1865 @cindex pointers to arrays
1866 @cindex const qualifier
1868 In GNU C, pointers to arrays with qualifiers work similar to pointers
1869 to other qualified types. For example, a value of type @code{int (*)[5]}
1870 can be used to initialize a variable of type @code{const int (*)[5]}.
1871 These types are incompatible in ISO C because the @code{const} qualifier
1872 is formally attached to the element type of the array and not the
1877 transpose (int N, int M, double out[M][N], const double in[N][M]);
1881 transpose(3, 2, y, x);
1885 @section Non-Constant Initializers
1886 @cindex initializers, non-constant
1887 @cindex non-constant initializers
1889 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1890 automatic variable are not required to be constant expressions in GNU C@.
1891 Here is an example of an initializer with run-time varying elements:
1894 foo (float f, float g)
1896 float beat_freqs[2] = @{ f-g, f+g @};
1901 @node Compound Literals
1902 @section Compound Literals
1903 @cindex constructor expressions
1904 @cindex initializations in expressions
1905 @cindex structures, constructor expression
1906 @cindex expressions, constructor
1907 @cindex compound literals
1908 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1910 A compound literal looks like a cast of a brace-enclosed aggregate
1911 initializer list. Its value is an object of the type specified in
1912 the cast, containing the elements specified in the initializer.
1913 Unlike the result of a cast, a compound literal is an lvalue. ISO
1914 C99 and later support compound literals. As an extension, GCC
1915 supports compound literals also in C90 mode and in C++, although
1916 as explained below, the C++ semantics are somewhat different.
1918 Usually, the specified type of a compound literal is a structure. Assume
1919 that @code{struct foo} and @code{structure} are declared as shown:
1922 struct foo @{int a; char b[2];@} structure;
1926 Here is an example of constructing a @code{struct foo} with a compound literal:
1929 structure = ((struct foo) @{x + y, 'a', 0@});
1933 This is equivalent to writing the following:
1937 struct foo temp = @{x + y, 'a', 0@};
1942 You can also construct an array, though this is dangerous in C++, as
1943 explained below. If all the elements of the compound literal are
1944 (made up of) simple constant expressions suitable for use in
1945 initializers of objects of static storage duration, then the compound
1946 literal can be coerced to a pointer to its first element and used in
1947 such an initializer, as shown here:
1950 char **foo = (char *[]) @{ "x", "y", "z" @};
1953 Compound literals for scalar types and union types are also allowed. In
1954 the following example the variable @code{i} is initialized to the value
1955 @code{2}, the result of incrementing the unnamed object created by
1956 the compound literal.
1959 int i = ++(int) @{ 1 @};
1962 As a GNU extension, GCC allows initialization of objects with static storage
1963 duration by compound literals (which is not possible in ISO C99 because
1964 the initializer is not a constant).
1965 It is handled as if the object were initialized only with the brace-enclosed
1966 list if the types of the compound literal and the object match.
1967 The elements of the compound literal must be constant.
1968 If the object being initialized has array type of unknown size, the size is
1969 determined by the size of the compound literal.
1972 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1973 static int y[] = (int []) @{1, 2, 3@};
1974 static int z[] = (int [3]) @{1@};
1978 The above lines are equivalent to the following:
1980 static struct foo x = @{1, 'a', 'b'@};
1981 static int y[] = @{1, 2, 3@};
1982 static int z[] = @{1, 0, 0@};
1985 In C, a compound literal designates an unnamed object with static or
1986 automatic storage duration. In C++, a compound literal designates a
1987 temporary object that only lives until the end of its full-expression.
1988 As a result, well-defined C code that takes the address of a subobject
1989 of a compound literal can be undefined in C++, so G++ rejects
1990 the conversion of a temporary array to a pointer. For instance, if
1991 the array compound literal example above appeared inside a function,
1992 any subsequent use of @code{foo} in C++ would have undefined behavior
1993 because the lifetime of the array ends after the declaration of @code{foo}.
1995 As an optimization, G++ sometimes gives array compound literals longer
1996 lifetimes: when the array either appears outside a function or has
1997 a @code{const}-qualified type. If @code{foo} and its initializer had
1998 elements of type @code{char *const} rather than @code{char *}, or if
1999 @code{foo} were a global variable, the array would have static storage
2000 duration. But it is probably safest just to avoid the use of array
2001 compound literals in C++ code.
2003 @node Designated Inits
2004 @section Designated Initializers
2005 @cindex initializers with labeled elements
2006 @cindex labeled elements in initializers
2007 @cindex case labels in initializers
2008 @cindex designated initializers
2010 Standard C90 requires the elements of an initializer to appear in a fixed
2011 order, the same as the order of the elements in the array or structure
2014 In ISO C99 you can give the elements in any order, specifying the array
2015 indices or structure field names they apply to, and GNU C allows this as
2016 an extension in C90 mode as well. This extension is not
2017 implemented in GNU C++.
2019 To specify an array index, write
2020 @samp{[@var{index}] =} before the element value. For example,
2023 int a[6] = @{ [4] = 29, [2] = 15 @};
2030 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
2034 The index values must be constant expressions, even if the array being
2035 initialized is automatic.
2037 An alternative syntax for this that has been obsolete since GCC 2.5 but
2038 GCC still accepts is to write @samp{[@var{index}]} before the element
2039 value, with no @samp{=}.
2041 To initialize a range of elements to the same value, write
2042 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
2043 extension. For example,
2046 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2050 If the value in it has side-effects, the side-effects happen only once,
2051 not for each initialized field by the range initializer.
2054 Note that the length of the array is the highest value specified
2057 In a structure initializer, specify the name of a field to initialize
2058 with @samp{.@var{fieldname} =} before the element value. For example,
2059 given the following structure,
2062 struct point @{ int x, y; @};
2066 the following initialization
2069 struct point p = @{ .y = yvalue, .x = xvalue @};
2076 struct point p = @{ xvalue, yvalue @};
2079 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2080 @samp{@var{fieldname}:}, as shown here:
2083 struct point p = @{ y: yvalue, x: xvalue @};
2086 Omitted field members are implicitly initialized the same as objects
2087 that have static storage duration.
2090 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2091 @dfn{designator}. You can also use a designator (or the obsolete colon
2092 syntax) when initializing a union, to specify which element of the union
2093 should be used. For example,
2096 union foo @{ int i; double d; @};
2098 union foo f = @{ .d = 4 @};
2102 converts 4 to a @code{double} to store it in the union using
2103 the second element. By contrast, casting 4 to type @code{union foo}
2104 stores it into the union as the integer @code{i}, since it is
2105 an integer. @xref{Cast to Union}.
2107 You can combine this technique of naming elements with ordinary C
2108 initialization of successive elements. Each initializer element that
2109 does not have a designator applies to the next consecutive element of the
2110 array or structure. For example,
2113 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2120 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2123 Labeling the elements of an array initializer is especially useful
2124 when the indices are characters or belong to an @code{enum} type.
2129 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2130 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2133 @cindex designator lists
2134 You can also write a series of @samp{.@var{fieldname}} and
2135 @samp{[@var{index}]} designators before an @samp{=} to specify a
2136 nested subobject to initialize; the list is taken relative to the
2137 subobject corresponding to the closest surrounding brace pair. For
2138 example, with the @samp{struct point} declaration above:
2141 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2145 If the same field is initialized multiple times, it has the value from
2146 the last initialization. If any such overridden initialization has
2147 side-effect, it is unspecified whether the side-effect happens or not.
2148 Currently, GCC discards them and issues a warning.
2151 @section Case Ranges
2153 @cindex ranges in case statements
2155 You can specify a range of consecutive values in a single @code{case} label,
2159 case @var{low} ... @var{high}:
2163 This has the same effect as the proper number of individual @code{case}
2164 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2166 This feature is especially useful for ranges of ASCII character codes:
2172 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2173 it may be parsed wrong when you use it with integer values. For example,
2188 @section Cast to a Union Type
2189 @cindex cast to a union
2190 @cindex union, casting to a
2192 A cast to union type looks similar to other casts, except that the type
2193 specified is a union type. You can specify the type either with the
2194 @code{union} keyword or with a @code{typedef} name that refers to
2195 a union. A cast to a union actually creates a compound literal and
2196 yields an lvalue, not an rvalue like true casts do.
2197 @xref{Compound Literals}.
2199 The types that may be cast to the union type are those of the members
2200 of the union. Thus, given the following union and variables:
2203 union foo @{ int i; double d; @};
2209 both @code{x} and @code{y} can be cast to type @code{union foo}.
2211 Using the cast as the right-hand side of an assignment to a variable of
2212 union type is equivalent to storing in a member of the union:
2217 u = (union foo) x @equiv{} u.i = x
2218 u = (union foo) y @equiv{} u.d = y
2221 You can also use the union cast as a function argument:
2224 void hack (union foo);
2226 hack ((union foo) x);
2229 @node Mixed Declarations
2230 @section Mixed Declarations and Code
2231 @cindex mixed declarations and code
2232 @cindex declarations, mixed with code
2233 @cindex code, mixed with declarations
2235 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2236 within compound statements. As an extension, GNU C also allows this in
2237 C90 mode. For example, you could do:
2246 Each identifier is visible from where it is declared until the end of
2247 the enclosing block.
2249 @node Function Attributes
2250 @section Declaring Attributes of Functions
2251 @cindex function attributes
2252 @cindex declaring attributes of functions
2253 @cindex @code{volatile} applied to function
2254 @cindex @code{const} applied to function
2256 In GNU C, you can use function attributes to declare certain things
2257 about functions called in your program which help the compiler
2258 optimize calls and check your code more carefully. For example, you
2259 can use attributes to declare that a function never returns
2260 (@code{noreturn}), returns a value depending only on its arguments
2261 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2263 You can also use attributes to control memory placement, code
2264 generation options or call/return conventions within the function
2265 being annotated. Many of these attributes are target-specific. For
2266 example, many targets support attributes for defining interrupt
2267 handler functions, which typically must follow special register usage
2268 and return conventions.
2270 Function attributes are introduced by the @code{__attribute__} keyword
2271 on a declaration, followed by an attribute specification inside double
2272 parentheses. You can specify multiple attributes in a declaration by
2273 separating them by commas within the double parentheses or by
2274 immediately following an attribute declaration with another attribute
2275 declaration. @xref{Attribute Syntax}, for the exact rules on
2276 attribute syntax and placement.
2278 GCC also supports attributes on
2279 variable declarations (@pxref{Variable Attributes}),
2280 labels (@pxref{Label Attributes}),
2281 enumerators (@pxref{Enumerator Attributes}),
2282 statements (@pxref{Statement Attributes}),
2283 and types (@pxref{Type Attributes}).
2285 There is some overlap between the purposes of attributes and pragmas
2286 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2287 found convenient to use @code{__attribute__} to achieve a natural
2288 attachment of attributes to their corresponding declarations, whereas
2289 @code{#pragma} is of use for compatibility with other compilers
2290 or constructs that do not naturally form part of the grammar.
2292 In addition to the attributes documented here,
2293 GCC plugins may provide their own attributes.
2296 * Common Function Attributes::
2297 * AArch64 Function Attributes::
2298 * ARC Function Attributes::
2299 * ARM Function Attributes::
2300 * AVR Function Attributes::
2301 * Blackfin Function Attributes::
2302 * CR16 Function Attributes::
2303 * Epiphany Function Attributes::
2304 * H8/300 Function Attributes::
2305 * IA-64 Function Attributes::
2306 * M32C Function Attributes::
2307 * M32R/D Function Attributes::
2308 * m68k Function Attributes::
2309 * MCORE Function Attributes::
2310 * MeP Function Attributes::
2311 * MicroBlaze Function Attributes::
2312 * Microsoft Windows Function Attributes::
2313 * MIPS Function Attributes::
2314 * MSP430 Function Attributes::
2315 * NDS32 Function Attributes::
2316 * Nios II Function Attributes::
2317 * Nvidia PTX Function Attributes::
2318 * PowerPC Function Attributes::
2319 * RL78 Function Attributes::
2320 * RX Function Attributes::
2321 * S/390 Function Attributes::
2322 * SH Function Attributes::
2323 * SPU Function Attributes::
2324 * Symbian OS Function Attributes::
2325 * V850 Function Attributes::
2326 * Visium Function Attributes::
2327 * x86 Function Attributes::
2328 * Xstormy16 Function Attributes::
2331 @node Common Function Attributes
2332 @subsection Common Function Attributes
2334 The following attributes are supported on most targets.
2337 @c Keep this table alphabetized by attribute name. Treat _ as space.
2339 @item alias ("@var{target}")
2340 @cindex @code{alias} function attribute
2341 The @code{alias} attribute causes the declaration to be emitted as an
2342 alias for another symbol, which must be specified. For instance,
2345 void __f () @{ /* @r{Do something.} */; @}
2346 void f () __attribute__ ((weak, alias ("__f")));
2350 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2351 mangled name for the target must be used. It is an error if @samp{__f}
2352 is not defined in the same translation unit.
2354 This attribute requires assembler and object file support,
2355 and may not be available on all targets.
2357 @item aligned (@var{alignment})
2358 @cindex @code{aligned} function attribute
2359 This attribute specifies a minimum alignment for the function,
2362 You cannot use this attribute to decrease the alignment of a function,
2363 only to increase it. However, when you explicitly specify a function
2364 alignment this overrides the effect of the
2365 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2368 Note that the effectiveness of @code{aligned} attributes may be
2369 limited by inherent limitations in your linker. On many systems, the
2370 linker is only able to arrange for functions to be aligned up to a
2371 certain maximum alignment. (For some linkers, the maximum supported
2372 alignment may be very very small.) See your linker documentation for
2373 further information.
2375 The @code{aligned} attribute can also be used for variables and fields
2376 (@pxref{Variable Attributes}.)
2379 @cindex @code{alloc_align} function attribute
2380 The @code{alloc_align} attribute is used to tell the compiler that the
2381 function return value points to memory, where the returned pointer minimum
2382 alignment is given by one of the functions parameters. GCC uses this
2383 information to improve pointer alignment analysis.
2385 The function parameter denoting the allocated alignment is specified by
2386 one integer argument, whose number is the argument of the attribute.
2387 Argument numbering starts at one.
2392 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2396 declares that @code{my_memalign} returns memory with minimum alignment
2397 given by parameter 1.
2400 @cindex @code{alloc_size} function attribute
2401 The @code{alloc_size} attribute is used to tell the compiler that the
2402 function return value points to memory, where the size is given by
2403 one or two of the functions parameters. GCC uses this
2404 information to improve the correctness of @code{__builtin_object_size}.
2406 The function parameter(s) denoting the allocated size are specified by
2407 one or two integer arguments supplied to the attribute. The allocated size
2408 is either the value of the single function argument specified or the product
2409 of the two function arguments specified. Argument numbering starts at
2415 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2416 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2420 declares that @code{my_calloc} returns memory of the size given by
2421 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2422 of the size given by parameter 2.
2425 @cindex @code{always_inline} function attribute
2426 Generally, functions are not inlined unless optimization is specified.
2427 For functions declared inline, this attribute inlines the function
2428 independent of any restrictions that otherwise apply to inlining.
2429 Failure to inline such a function is diagnosed as an error.
2430 Note that if such a function is called indirectly the compiler may
2431 or may not inline it depending on optimization level and a failure
2432 to inline an indirect call may or may not be diagnosed.
2435 @cindex @code{artificial} function attribute
2436 This attribute is useful for small inline wrappers that if possible
2437 should appear during debugging as a unit. Depending on the debug
2438 info format it either means marking the function as artificial
2439 or using the caller location for all instructions within the inlined
2442 @item assume_aligned
2443 @cindex @code{assume_aligned} function attribute
2444 The @code{assume_aligned} attribute is used to tell the compiler that the
2445 function return value points to memory, where the returned pointer minimum
2446 alignment is given by the first argument.
2447 If the attribute has two arguments, the second argument is misalignment offset.
2452 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2453 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2457 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2458 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2461 @item bnd_instrument
2462 @cindex @code{bnd_instrument} function attribute
2463 The @code{bnd_instrument} attribute on functions is used to inform the
2464 compiler that the function should be instrumented when compiled
2465 with the @option{-fchkp-instrument-marked-only} option.
2468 @cindex @code{bnd_legacy} function attribute
2469 @cindex Pointer Bounds Checker attributes
2470 The @code{bnd_legacy} attribute on functions is used to inform the
2471 compiler that the function should not be instrumented when compiled
2472 with the @option{-fcheck-pointer-bounds} option.
2475 @cindex @code{cold} function attribute
2476 The @code{cold} attribute on functions is used to inform the compiler that
2477 the function is unlikely to be executed. The function is optimized for
2478 size rather than speed and on many targets it is placed into a special
2479 subsection of the text section so all cold functions appear close together,
2480 improving code locality of non-cold parts of program. The paths leading
2481 to calls of cold functions within code are marked as unlikely by the branch
2482 prediction mechanism. It is thus useful to mark functions used to handle
2483 unlikely conditions, such as @code{perror}, as cold to improve optimization
2484 of hot functions that do call marked functions in rare occasions.
2486 When profile feedback is available, via @option{-fprofile-use}, cold functions
2487 are automatically detected and this attribute is ignored.
2490 @cindex @code{const} function attribute
2491 @cindex functions that have no side effects
2492 Many functions do not examine any values except their arguments, and
2493 have no effects except the return value. Basically this is just slightly
2494 more strict class than the @code{pure} attribute below, since function is not
2495 allowed to read global memory.
2497 @cindex pointer arguments
2498 Note that a function that has pointer arguments and examines the data
2499 pointed to must @emph{not} be declared @code{const}. Likewise, a
2500 function that calls a non-@code{const} function usually must not be
2501 @code{const}. It does not make sense for a @code{const} function to
2506 @itemx constructor (@var{priority})
2507 @itemx destructor (@var{priority})
2508 @cindex @code{constructor} function attribute
2509 @cindex @code{destructor} function attribute
2510 The @code{constructor} attribute causes the function to be called
2511 automatically before execution enters @code{main ()}. Similarly, the
2512 @code{destructor} attribute causes the function to be called
2513 automatically after @code{main ()} completes or @code{exit ()} is
2514 called. Functions with these attributes are useful for
2515 initializing data that is used implicitly during the execution of
2518 You may provide an optional integer priority to control the order in
2519 which constructor and destructor functions are run. A constructor
2520 with a smaller priority number runs before a constructor with a larger
2521 priority number; the opposite relationship holds for destructors. So,
2522 if you have a constructor that allocates a resource and a destructor
2523 that deallocates the same resource, both functions typically have the
2524 same priority. The priorities for constructor and destructor
2525 functions are the same as those specified for namespace-scope C++
2526 objects (@pxref{C++ Attributes}). However, at present, the order in which
2527 constructors for C++ objects with static storage duration and functions
2528 decorated with attribute @code{constructor} are invoked is unspecified.
2529 In mixed declarations, attribute @code{init_priority} can be used to
2530 impose a specific ordering.
2533 @itemx deprecated (@var{msg})
2534 @cindex @code{deprecated} function attribute
2535 The @code{deprecated} attribute results in a warning if the function
2536 is used anywhere in the source file. This is useful when identifying
2537 functions that are expected to be removed in a future version of a
2538 program. The warning also includes the location of the declaration
2539 of the deprecated function, to enable users to easily find further
2540 information about why the function is deprecated, or what they should
2541 do instead. Note that the warnings only occurs for uses:
2544 int old_fn () __attribute__ ((deprecated));
2546 int (*fn_ptr)() = old_fn;
2550 results in a warning on line 3 but not line 2. The optional @var{msg}
2551 argument, which must be a string, is printed in the warning if
2554 The @code{deprecated} attribute can also be used for variables and
2555 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2557 @item error ("@var{message}")
2558 @itemx warning ("@var{message}")
2559 @cindex @code{error} function attribute
2560 @cindex @code{warning} function attribute
2561 If the @code{error} or @code{warning} attribute
2562 is used on a function declaration and a call to such a function
2563 is not eliminated through dead code elimination or other optimizations,
2564 an error or warning (respectively) that includes @var{message} is diagnosed.
2566 for compile-time checking, especially together with @code{__builtin_constant_p}
2567 and inline functions where checking the inline function arguments is not
2568 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2570 While it is possible to leave the function undefined and thus invoke
2571 a link failure (to define the function with
2572 a message in @code{.gnu.warning*} section),
2573 when using these attributes the problem is diagnosed
2574 earlier and with exact location of the call even in presence of inline
2575 functions or when not emitting debugging information.
2577 @item externally_visible
2578 @cindex @code{externally_visible} function attribute
2579 This attribute, attached to a global variable or function, nullifies
2580 the effect of the @option{-fwhole-program} command-line option, so the
2581 object remains visible outside the current compilation unit.
2583 If @option{-fwhole-program} is used together with @option{-flto} and
2584 @command{gold} is used as the linker plugin,
2585 @code{externally_visible} attributes are automatically added to functions
2586 (not variable yet due to a current @command{gold} issue)
2587 that are accessed outside of LTO objects according to resolution file
2588 produced by @command{gold}.
2589 For other linkers that cannot generate resolution file,
2590 explicit @code{externally_visible} attributes are still necessary.
2593 @cindex @code{flatten} function attribute
2594 Generally, inlining into a function is limited. For a function marked with
2595 this attribute, every call inside this function is inlined, if possible.
2596 Whether the function itself is considered for inlining depends on its size and
2597 the current inlining parameters.
2599 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2600 @cindex @code{format} function attribute
2601 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2603 The @code{format} attribute specifies that a function takes @code{printf},
2604 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2605 should be type-checked against a format string. For example, the
2610 my_printf (void *my_object, const char *my_format, ...)
2611 __attribute__ ((format (printf, 2, 3)));
2615 causes the compiler to check the arguments in calls to @code{my_printf}
2616 for consistency with the @code{printf} style format string argument
2619 The parameter @var{archetype} determines how the format string is
2620 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2621 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2622 @code{strfmon}. (You can also use @code{__printf__},
2623 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2624 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2625 @code{ms_strftime} are also present.
2626 @var{archetype} values such as @code{printf} refer to the formats accepted
2627 by the system's C runtime library,
2628 while values prefixed with @samp{gnu_} always refer
2629 to the formats accepted by the GNU C Library. On Microsoft Windows
2630 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2631 @file{msvcrt.dll} library.
2632 The parameter @var{string-index}
2633 specifies which argument is the format string argument (starting
2634 from 1), while @var{first-to-check} is the number of the first
2635 argument to check against the format string. For functions
2636 where the arguments are not available to be checked (such as
2637 @code{vprintf}), specify the third parameter as zero. In this case the
2638 compiler only checks the format string for consistency. For
2639 @code{strftime} formats, the third parameter is required to be zero.
2640 Since non-static C++ methods have an implicit @code{this} argument, the
2641 arguments of such methods should be counted from two, not one, when
2642 giving values for @var{string-index} and @var{first-to-check}.
2644 In the example above, the format string (@code{my_format}) is the second
2645 argument of the function @code{my_print}, and the arguments to check
2646 start with the third argument, so the correct parameters for the format
2647 attribute are 2 and 3.
2649 @opindex ffreestanding
2650 @opindex fno-builtin
2651 The @code{format} attribute allows you to identify your own functions
2652 that take format strings as arguments, so that GCC can check the
2653 calls to these functions for errors. The compiler always (unless
2654 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2655 for the standard library functions @code{printf}, @code{fprintf},
2656 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2657 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2658 warnings are requested (using @option{-Wformat}), so there is no need to
2659 modify the header file @file{stdio.h}. In C99 mode, the functions
2660 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2661 @code{vsscanf} are also checked. Except in strictly conforming C
2662 standard modes, the X/Open function @code{strfmon} is also checked as
2663 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2664 @xref{C Dialect Options,,Options Controlling C Dialect}.
2666 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2667 recognized in the same context. Declarations including these format attributes
2668 are parsed for correct syntax, however the result of checking of such format
2669 strings is not yet defined, and is not carried out by this version of the
2672 The target may also provide additional types of format checks.
2673 @xref{Target Format Checks,,Format Checks Specific to Particular
2676 @item format_arg (@var{string-index})
2677 @cindex @code{format_arg} function attribute
2678 @opindex Wformat-nonliteral
2679 The @code{format_arg} attribute specifies that a function takes a format
2680 string for a @code{printf}, @code{scanf}, @code{strftime} or
2681 @code{strfmon} style function and modifies it (for example, to translate
2682 it into another language), so the result can be passed to a
2683 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2684 function (with the remaining arguments to the format function the same
2685 as they would have been for the unmodified string). For example, the
2690 my_dgettext (char *my_domain, const char *my_format)
2691 __attribute__ ((format_arg (2)));
2695 causes the compiler to check the arguments in calls to a @code{printf},
2696 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2697 format string argument is a call to the @code{my_dgettext} function, for
2698 consistency with the format string argument @code{my_format}. If the
2699 @code{format_arg} attribute had not been specified, all the compiler
2700 could tell in such calls to format functions would be that the format
2701 string argument is not constant; this would generate a warning when
2702 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2703 without the attribute.
2705 The parameter @var{string-index} specifies which argument is the format
2706 string argument (starting from one). Since non-static C++ methods have
2707 an implicit @code{this} argument, the arguments of such methods should
2708 be counted from two.
2710 The @code{format_arg} attribute allows you to identify your own
2711 functions that modify format strings, so that GCC can check the
2712 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2713 type function whose operands are a call to one of your own function.
2714 The compiler always treats @code{gettext}, @code{dgettext}, and
2715 @code{dcgettext} in this manner except when strict ISO C support is
2716 requested by @option{-ansi} or an appropriate @option{-std} option, or
2717 @option{-ffreestanding} or @option{-fno-builtin}
2718 is used. @xref{C Dialect Options,,Options
2719 Controlling C Dialect}.
2721 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2722 @code{NSString} reference for compatibility with the @code{format} attribute
2725 The target may also allow additional types in @code{format-arg} attributes.
2726 @xref{Target Format Checks,,Format Checks Specific to Particular
2730 @cindex @code{gnu_inline} function attribute
2731 This attribute should be used with a function that is also declared
2732 with the @code{inline} keyword. It directs GCC to treat the function
2733 as if it were defined in gnu90 mode even when compiling in C99 or
2736 If the function is declared @code{extern}, then this definition of the
2737 function is used only for inlining. In no case is the function
2738 compiled as a standalone function, not even if you take its address
2739 explicitly. Such an address becomes an external reference, as if you
2740 had only declared the function, and had not defined it. This has
2741 almost the effect of a macro. The way to use this is to put a
2742 function definition in a header file with this attribute, and put
2743 another copy of the function, without @code{extern}, in a library
2744 file. The definition in the header file causes most calls to the
2745 function to be inlined. If any uses of the function remain, they
2746 refer to the single copy in the library. Note that the two
2747 definitions of the functions need not be precisely the same, although
2748 if they do not have the same effect your program may behave oddly.
2750 In C, if the function is neither @code{extern} nor @code{static}, then
2751 the function is compiled as a standalone function, as well as being
2752 inlined where possible.
2754 This is how GCC traditionally handled functions declared
2755 @code{inline}. Since ISO C99 specifies a different semantics for
2756 @code{inline}, this function attribute is provided as a transition
2757 measure and as a useful feature in its own right. This attribute is
2758 available in GCC 4.1.3 and later. It is available if either of the
2759 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2760 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2761 Function is As Fast As a Macro}.
2763 In C++, this attribute does not depend on @code{extern} in any way,
2764 but it still requires the @code{inline} keyword to enable its special
2768 @cindex @code{hot} function attribute
2769 The @code{hot} attribute on a function is used to inform the compiler that
2770 the function is a hot spot of the compiled program. The function is
2771 optimized more aggressively and on many targets it is placed into a special
2772 subsection of the text section so all hot functions appear close together,
2775 When profile feedback is available, via @option{-fprofile-use}, hot functions
2776 are automatically detected and this attribute is ignored.
2778 @item ifunc ("@var{resolver}")
2779 @cindex @code{ifunc} function attribute
2780 @cindex indirect functions
2781 @cindex functions that are dynamically resolved
2782 The @code{ifunc} attribute is used to mark a function as an indirect
2783 function using the STT_GNU_IFUNC symbol type extension to the ELF
2784 standard. This allows the resolution of the symbol value to be
2785 determined dynamically at load time, and an optimized version of the
2786 routine can be selected for the particular processor or other system
2787 characteristics determined then. To use this attribute, first define
2788 the implementation functions available, and a resolver function that
2789 returns a pointer to the selected implementation function. The
2790 implementation functions' declarations must match the API of the
2791 function being implemented, the resolver's declaration is be a
2792 function returning pointer to void function returning void:
2795 void *my_memcpy (void *dst, const void *src, size_t len)
2800 static void (*resolve_memcpy (void)) (void)
2802 return my_memcpy; // we'll just always select this routine
2807 The exported header file declaring the function the user calls would
2811 extern void *memcpy (void *, const void *, size_t);
2815 allowing the user to call this as a regular function, unaware of the
2816 implementation. Finally, the indirect function needs to be defined in
2817 the same translation unit as the resolver function:
2820 void *memcpy (void *, const void *, size_t)
2821 __attribute__ ((ifunc ("resolve_memcpy")));
2824 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2825 and GNU C Library version 2.11.1 are required to use this feature.
2828 @itemx interrupt_handler
2829 Many GCC back ends support attributes to indicate that a function is
2830 an interrupt handler, which tells the compiler to generate function
2831 entry and exit sequences that differ from those from regular
2832 functions. The exact syntax and behavior are target-specific;
2833 refer to the following subsections for details.
2836 @cindex @code{leaf} function attribute
2837 Calls to external functions with this attribute must return to the
2838 current compilation unit only by return or by exception handling. In
2839 particular, a leaf function is not allowed to invoke callback functions
2840 passed to it from the current compilation unit, directly call functions
2841 exported by the unit, or @code{longjmp} into the unit. Leaf functions
2842 might still call functions from other compilation units and thus they
2843 are not necessarily leaf in the sense that they contain no function
2846 The attribute is intended for library functions to improve dataflow
2847 analysis. The compiler takes the hint that any data not escaping the
2848 current compilation unit cannot be used or modified by the leaf
2849 function. For example, the @code{sin} function is a leaf function, but
2850 @code{qsort} is not.
2852 Note that leaf functions might indirectly run a signal handler defined
2853 in the current compilation unit that uses static variables. Similarly,
2854 when lazy symbol resolution is in effect, leaf functions might invoke
2855 indirect functions whose resolver function or implementation function is
2856 defined in the current compilation unit and uses static variables. There
2857 is no standard-compliant way to write such a signal handler, resolver
2858 function, or implementation function, and the best that you can do is to
2859 remove the @code{leaf} attribute or mark all such static variables
2860 @code{volatile}. Lastly, for ELF-based systems that support symbol
2861 interposition, care should be taken that functions defined in the
2862 current compilation unit do not unexpectedly interpose other symbols
2863 based on the defined standards mode and defined feature test macros;
2864 otherwise an inadvertent callback would be added.
2866 The attribute has no effect on functions defined within the current
2867 compilation unit. This is to allow easy merging of multiple compilation
2868 units into one, for example, by using the link-time optimization. For
2869 this reason the attribute is not allowed on types to annotate indirect
2873 @cindex @code{malloc} function attribute
2874 @cindex functions that behave like malloc
2875 This tells the compiler that a function is @code{malloc}-like, i.e.,
2876 that the pointer @var{P} returned by the function cannot alias any
2877 other pointer valid when the function returns, and moreover no
2878 pointers to valid objects occur in any storage addressed by @var{P}.
2880 Using this attribute can improve optimization. Functions like
2881 @code{malloc} and @code{calloc} have this property because they return
2882 a pointer to uninitialized or zeroed-out storage. However, functions
2883 like @code{realloc} do not have this property, as they can return a
2884 pointer to storage containing pointers.
2887 @cindex @code{no_icf} function attribute
2888 This function attribute prevents a functions from being merged with another
2889 semantically equivalent function.
2891 @item no_instrument_function
2892 @cindex @code{no_instrument_function} function attribute
2893 @opindex finstrument-functions
2894 If @option{-finstrument-functions} is given, profiling function calls are
2895 generated at entry and exit of most user-compiled functions.
2896 Functions with this attribute are not so instrumented.
2898 @item no_profile_instrument_function
2899 @cindex @code{no_profile_instrument_function} function attribute
2900 The @code{no_profile_instrument_function} attribute on functions is used
2901 to inform the compiler that it should not process any profile feedback based
2902 optimization code instrumentation.
2905 @cindex @code{no_reorder} function attribute
2906 Do not reorder functions or variables marked @code{no_reorder}
2907 against each other or top level assembler statements the executable.
2908 The actual order in the program will depend on the linker command
2909 line. Static variables marked like this are also not removed.
2910 This has a similar effect
2911 as the @option{-fno-toplevel-reorder} option, but only applies to the
2914 @item no_sanitize ("@var{sanitize_option}")
2915 @cindex @code{no_sanitize} function attribute
2916 The @code{no_sanitize} attribute on functions is used
2917 to inform the compiler that it should not do sanitization of all options
2918 mentioned in @var{sanitize_option}. A list of values acceptable by
2919 @option{-fsanitize} option can be provided.
2922 void __attribute__ ((no_sanitize ("alignment", "object-size")))
2923 f () @{ /* @r{Do something.} */; @}
2926 @item no_sanitize_address
2927 @itemx no_address_safety_analysis
2928 @cindex @code{no_sanitize_address} function attribute
2929 The @code{no_sanitize_address} attribute on functions is used
2930 to inform the compiler that it should not instrument memory accesses
2931 in the function when compiling with the @option{-fsanitize=address} option.
2932 The @code{no_address_safety_analysis} is a deprecated alias of the
2933 @code{no_sanitize_address} attribute, new code should use
2934 @code{no_sanitize_address}.
2936 @item no_sanitize_thread
2937 @cindex @code{no_sanitize_thread} function attribute
2938 The @code{no_sanitize_thread} attribute on functions is used
2939 to inform the compiler that it should not instrument memory accesses
2940 in the function when compiling with the @option{-fsanitize=thread} option.
2942 @item no_sanitize_undefined
2943 @cindex @code{no_sanitize_undefined} function attribute
2944 The @code{no_sanitize_undefined} attribute on functions is used
2945 to inform the compiler that it should not check for undefined behavior
2946 in the function when compiling with the @option{-fsanitize=undefined} option.
2948 @item no_split_stack
2949 @cindex @code{no_split_stack} function attribute
2950 @opindex fsplit-stack
2951 If @option{-fsplit-stack} is given, functions have a small
2952 prologue which decides whether to split the stack. Functions with the
2953 @code{no_split_stack} attribute do not have that prologue, and thus
2954 may run with only a small amount of stack space available.
2956 @item no_stack_limit
2957 @cindex @code{no_stack_limit} function attribute
2958 This attribute locally overrides the @option{-fstack-limit-register}
2959 and @option{-fstack-limit-symbol} command-line options; it has the effect
2960 of disabling stack limit checking in the function it applies to.
2963 @cindex @code{noclone} function attribute
2964 This function attribute prevents a function from being considered for
2965 cloning---a mechanism that produces specialized copies of functions
2966 and which is (currently) performed by interprocedural constant
2970 @cindex @code{noinline} function attribute
2971 This function attribute prevents a function from being considered for
2973 @c Don't enumerate the optimizations by name here; we try to be
2974 @c future-compatible with this mechanism.
2975 If the function does not have side-effects, there are optimizations
2976 other than inlining that cause function calls to be optimized away,
2977 although the function call is live. To keep such calls from being
2984 (@pxref{Extended Asm}) in the called function, to serve as a special
2987 @item nonnull (@var{arg-index}, @dots{})
2988 @cindex @code{nonnull} function attribute
2989 @cindex functions with non-null pointer arguments
2990 The @code{nonnull} attribute specifies that some function parameters should
2991 be non-null pointers. For instance, the declaration:
2995 my_memcpy (void *dest, const void *src, size_t len)
2996 __attribute__((nonnull (1, 2)));
3000 causes the compiler to check that, in calls to @code{my_memcpy},
3001 arguments @var{dest} and @var{src} are non-null. If the compiler
3002 determines that a null pointer is passed in an argument slot marked
3003 as non-null, and the @option{-Wnonnull} option is enabled, a warning
3004 is issued. The compiler may also choose to make optimizations based
3005 on the knowledge that certain function arguments will never be null.
3007 If no argument index list is given to the @code{nonnull} attribute,
3008 all pointer arguments are marked as non-null. To illustrate, the
3009 following declaration is equivalent to the previous example:
3013 my_memcpy (void *dest, const void *src, size_t len)
3014 __attribute__((nonnull));
3018 @cindex @code{noplt} function attribute
3019 The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
3020 Calls to functions marked with this attribute in position-independent code
3025 /* Externally defined function foo. */
3026 int foo () __attribute__ ((noplt));
3029 main (/* @r{@dots{}} */)
3038 The @code{noplt} attribute on function @code{foo}
3039 tells the compiler to assume that
3040 the function @code{foo} is externally defined and that the call to
3041 @code{foo} must avoid the PLT
3042 in position-independent code.
3044 In position-dependent code, a few targets also convert calls to
3045 functions that are marked to not use the PLT to use the GOT instead.
3048 @cindex @code{noreturn} function attribute
3049 @cindex functions that never return
3050 A few standard library functions, such as @code{abort} and @code{exit},
3051 cannot return. GCC knows this automatically. Some programs define
3052 their own functions that never return. You can declare them
3053 @code{noreturn} to tell the compiler this fact. For example,
3057 void fatal () __attribute__ ((noreturn));
3060 fatal (/* @r{@dots{}} */)
3062 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3068 The @code{noreturn} keyword tells the compiler to assume that
3069 @code{fatal} cannot return. It can then optimize without regard to what
3070 would happen if @code{fatal} ever did return. This makes slightly
3071 better code. More importantly, it helps avoid spurious warnings of
3072 uninitialized variables.
3074 The @code{noreturn} keyword does not affect the exceptional path when that
3075 applies: a @code{noreturn}-marked function may still return to the caller
3076 by throwing an exception or calling @code{longjmp}.
3078 Do not assume that registers saved by the calling function are
3079 restored before calling the @code{noreturn} function.
3081 It does not make sense for a @code{noreturn} function to have a return
3082 type other than @code{void}.
3085 @cindex @code{nothrow} function attribute
3086 The @code{nothrow} attribute is used to inform the compiler that a
3087 function cannot throw an exception. For example, most functions in
3088 the standard C library can be guaranteed not to throw an exception
3089 with the notable exceptions of @code{qsort} and @code{bsearch} that
3090 take function pointer arguments.
3093 @cindex @code{optimize} function attribute
3094 The @code{optimize} attribute is used to specify that a function is to
3095 be compiled with different optimization options than specified on the
3096 command line. Arguments can either be numbers or strings. Numbers
3097 are assumed to be an optimization level. Strings that begin with
3098 @code{O} are assumed to be an optimization option, while other options
3099 are assumed to be used with a @code{-f} prefix. You can also use the
3100 @samp{#pragma GCC optimize} pragma to set the optimization options
3101 that affect more than one function.
3102 @xref{Function Specific Option Pragmas}, for details about the
3103 @samp{#pragma GCC optimize} pragma.
3105 This attribute should be used for debugging purposes only. It is not
3106 suitable in production code.
3109 @cindex @code{pure} function attribute
3110 @cindex functions that have no side effects
3111 Many functions have no effects except the return value and their
3112 return value depends only on the parameters and/or global variables.
3113 Such a function can be subject
3114 to common subexpression elimination and loop optimization just as an
3115 arithmetic operator would be. These functions should be declared
3116 with the attribute @code{pure}. For example,
3119 int square (int) __attribute__ ((pure));
3123 says that the hypothetical function @code{square} is safe to call
3124 fewer times than the program says.
3126 Some common examples of pure functions are @code{strlen} or @code{memcmp}.
3127 Interesting non-pure functions are functions with infinite loops or those
3128 depending on volatile memory or other system resource, that may change between
3129 two consecutive calls (such as @code{feof} in a multithreading environment).
3131 @item returns_nonnull
3132 @cindex @code{returns_nonnull} function attribute
3133 The @code{returns_nonnull} attribute specifies that the function
3134 return value should be a non-null pointer. For instance, the declaration:
3138 mymalloc (size_t len) __attribute__((returns_nonnull));
3142 lets the compiler optimize callers based on the knowledge
3143 that the return value will never be null.
3146 @cindex @code{returns_twice} function attribute
3147 @cindex functions that return more than once
3148 The @code{returns_twice} attribute tells the compiler that a function may
3149 return more than one time. The compiler ensures that all registers
3150 are dead before calling such a function and emits a warning about
3151 the variables that may be clobbered after the second return from the
3152 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3153 The @code{longjmp}-like counterpart of such function, if any, might need
3154 to be marked with the @code{noreturn} attribute.
3156 @item section ("@var{section-name}")
3157 @cindex @code{section} function attribute
3158 @cindex functions in arbitrary sections
3159 Normally, the compiler places the code it generates in the @code{text} section.
3160 Sometimes, however, you need additional sections, or you need certain
3161 particular functions to appear in special sections. The @code{section}
3162 attribute specifies that a function lives in a particular section.
3163 For example, the declaration:
3166 extern void foobar (void) __attribute__ ((section ("bar")));
3170 puts the function @code{foobar} in the @code{bar} section.
3172 Some file formats do not support arbitrary sections so the @code{section}
3173 attribute is not available on all platforms.
3174 If you need to map the entire contents of a module to a particular
3175 section, consider using the facilities of the linker instead.
3178 @cindex @code{sentinel} function attribute
3179 This function attribute ensures that a parameter in a function call is
3180 an explicit @code{NULL}. The attribute is only valid on variadic
3181 functions. By default, the sentinel is located at position zero, the
3182 last parameter of the function call. If an optional integer position
3183 argument P is supplied to the attribute, the sentinel must be located at
3184 position P counting backwards from the end of the argument list.
3187 __attribute__ ((sentinel))
3189 __attribute__ ((sentinel(0)))
3192 The attribute is automatically set with a position of 0 for the built-in
3193 functions @code{execl} and @code{execlp}. The built-in function
3194 @code{execle} has the attribute set with a position of 1.
3196 A valid @code{NULL} in this context is defined as zero with any pointer
3197 type. If your system defines the @code{NULL} macro with an integer type
3198 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3199 with a copy that redefines NULL appropriately.
3201 The warnings for missing or incorrect sentinels are enabled with
3205 @itemx simd("@var{mask}")
3206 @cindex @code{simd} function attribute
3207 This attribute enables creation of one or more function versions that
3208 can process multiple arguments using SIMD instructions from a
3209 single invocation. Specifying this attribute allows compiler to
3210 assume that such versions are available at link time (provided
3211 in the same or another translation unit). Generated versions are
3212 target-dependent and described in the corresponding Vector ABI document. For
3213 x86_64 target this document can be found
3214 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3216 The optional argument @var{mask} may have the value
3217 @code{notinbranch} or @code{inbranch},
3218 and instructs the compiler to generate non-masked or masked
3219 clones correspondingly. By default, all clones are generated.
3221 The attribute should not be used together with Cilk Plus @code{vector}
3222 attribute on the same function.
3224 If the attribute is specified and @code{#pragma omp declare simd} is
3225 present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
3226 switch is specified, then the attribute is ignored.
3229 @cindex @code{stack_protect} function attribute
3230 This attribute adds stack protection code to the function if
3231 flags @option{-fstack-protector}, @option{-fstack-protector-strong}
3232 or @option{-fstack-protector-explicit} are set.
3234 @item target (@var{options})
3235 @cindex @code{target} function attribute
3236 Multiple target back ends implement the @code{target} attribute
3237 to specify that a function is to
3238 be compiled with different target options than specified on the
3239 command line. This can be used for instance to have functions
3240 compiled with a different ISA (instruction set architecture) than the
3241 default. You can also use the @samp{#pragma GCC target} pragma to set
3242 more than one function to be compiled with specific target options.
3243 @xref{Function Specific Option Pragmas}, for details about the
3244 @samp{#pragma GCC target} pragma.
3246 For instance, on an x86, you could declare one function with the
3247 @code{target("sse4.1,arch=core2")} attribute and another with
3248 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3249 compiling the first function with @option{-msse4.1} and
3250 @option{-march=core2} options, and the second function with
3251 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3252 to make sure that a function is only invoked on a machine that
3253 supports the particular ISA it is compiled for (for example by using
3254 @code{cpuid} on x86 to determine what feature bits and architecture
3258 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3259 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3262 You can either use multiple
3263 strings separated by commas to specify multiple options,
3264 or separate the options with a comma (@samp{,}) within a single string.
3266 The options supported are specific to each target; refer to @ref{x86
3267 Function Attributes}, @ref{PowerPC Function Attributes},
3268 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3271 @item target_clones (@var{options})
3272 @cindex @code{target_clones} function attribute
3273 The @code{target_clones} attribute is used to specify that a function
3274 be cloned into multiple versions compiled with different target options
3275 than specified on the command line. The supported options and restrictions
3276 are the same as for @code{target} attribute.
3278 For instance, on an x86, you could compile a function with
3279 @code{target_clones("sse4.1,avx")}. GCC creates two function clones,
3280 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3282 On a PowerPC, you can compile a function with
3283 @code{target_clones("cpu=power9,default")}. GCC will create two
3284 function clones, one compiled with @option{-mcpu=power9} and another
3285 with the default options. GCC must be configured to use GLIBC 2.23 or
3286 newer in order to use the @code{target_clones} attribute.
3288 It also creates a resolver function (see
3289 the @code{ifunc} attribute above) that dynamically selects a clone
3290 suitable for current architecture. The resolver is created only if there
3291 is a usage of a function with @code{target_clones} attribute.
3294 @cindex @code{unused} function attribute
3295 This attribute, attached to a function, means that the function is meant
3296 to be possibly unused. GCC does not produce a warning for this
3300 @cindex @code{used} function attribute
3301 This attribute, attached to a function, means that code must be emitted
3302 for the function even if it appears that the function is not referenced.
3303 This is useful, for example, when the function is referenced only in
3306 When applied to a member function of a C++ class template, the
3307 attribute also means that the function is instantiated if the
3308 class itself is instantiated.
3310 @item visibility ("@var{visibility_type}")
3311 @cindex @code{visibility} function attribute
3312 This attribute affects the linkage of the declaration to which it is attached.
3313 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3314 (@pxref{Common Type Attributes}) as well as functions.
3316 There are four supported @var{visibility_type} values: default,
3317 hidden, protected or internal visibility.
3320 void __attribute__ ((visibility ("protected")))
3321 f () @{ /* @r{Do something.} */; @}
3322 int i __attribute__ ((visibility ("hidden")));
3325 The possible values of @var{visibility_type} correspond to the
3326 visibility settings in the ELF gABI.
3329 @c keep this list of visibilities in alphabetical order.
3332 Default visibility is the normal case for the object file format.
3333 This value is available for the visibility attribute to override other
3334 options that may change the assumed visibility of entities.
3336 On ELF, default visibility means that the declaration is visible to other
3337 modules and, in shared libraries, means that the declared entity may be
3340 On Darwin, default visibility means that the declaration is visible to
3343 Default visibility corresponds to ``external linkage'' in the language.
3346 Hidden visibility indicates that the entity declared has a new
3347 form of linkage, which we call ``hidden linkage''. Two
3348 declarations of an object with hidden linkage refer to the same object
3349 if they are in the same shared object.
3352 Internal visibility is like hidden visibility, but with additional
3353 processor specific semantics. Unless otherwise specified by the
3354 psABI, GCC defines internal visibility to mean that a function is
3355 @emph{never} called from another module. Compare this with hidden
3356 functions which, while they cannot be referenced directly by other
3357 modules, can be referenced indirectly via function pointers. By
3358 indicating that a function cannot be called from outside the module,
3359 GCC may for instance omit the load of a PIC register since it is known
3360 that the calling function loaded the correct value.
3363 Protected visibility is like default visibility except that it
3364 indicates that references within the defining module bind to the
3365 definition in that module. That is, the declared entity cannot be
3366 overridden by another module.
3370 All visibilities are supported on many, but not all, ELF targets
3371 (supported when the assembler supports the @samp{.visibility}
3372 pseudo-op). Default visibility is supported everywhere. Hidden
3373 visibility is supported on Darwin targets.
3375 The visibility attribute should be applied only to declarations that
3376 would otherwise have external linkage. The attribute should be applied
3377 consistently, so that the same entity should not be declared with
3378 different settings of the attribute.
3380 In C++, the visibility attribute applies to types as well as functions
3381 and objects, because in C++ types have linkage. A class must not have
3382 greater visibility than its non-static data member types and bases,
3383 and class members default to the visibility of their class. Also, a
3384 declaration without explicit visibility is limited to the visibility
3387 In C++, you can mark member functions and static member variables of a
3388 class with the visibility attribute. This is useful if you know a
3389 particular method or static member variable should only be used from
3390 one shared object; then you can mark it hidden while the rest of the
3391 class has default visibility. Care must be taken to avoid breaking
3392 the One Definition Rule; for example, it is usually not useful to mark
3393 an inline method as hidden without marking the whole class as hidden.
3395 A C++ namespace declaration can also have the visibility attribute.
3398 namespace nspace1 __attribute__ ((visibility ("protected")))
3399 @{ /* @r{Do something.} */; @}
3402 This attribute applies only to the particular namespace body, not to
3403 other definitions of the same namespace; it is equivalent to using
3404 @samp{#pragma GCC visibility} before and after the namespace
3405 definition (@pxref{Visibility Pragmas}).
3407 In C++, if a template argument has limited visibility, this
3408 restriction is implicitly propagated to the template instantiation.
3409 Otherwise, template instantiations and specializations default to the
3410 visibility of their template.
3412 If both the template and enclosing class have explicit visibility, the
3413 visibility from the template is used.
3415 @item warn_unused_result
3416 @cindex @code{warn_unused_result} function attribute
3417 The @code{warn_unused_result} attribute causes a warning to be emitted
3418 if a caller of the function with this attribute does not use its
3419 return value. This is useful for functions where not checking
3420 the result is either a security problem or always a bug, such as
3424 int fn () __attribute__ ((warn_unused_result));
3427 if (fn () < 0) return -1;
3434 results in warning on line 5.
3437 @cindex @code{weak} function attribute
3438 The @code{weak} attribute causes the declaration to be emitted as a weak
3439 symbol rather than a global. This is primarily useful in defining
3440 library functions that can be overridden in user code, though it can
3441 also be used with non-function declarations. Weak symbols are supported
3442 for ELF targets, and also for a.out targets when using the GNU assembler
3446 @itemx weakref ("@var{target}")
3447 @cindex @code{weakref} function attribute
3448 The @code{weakref} attribute marks a declaration as a weak reference.
3449 Without arguments, it should be accompanied by an @code{alias} attribute
3450 naming the target symbol. Optionally, the @var{target} may be given as
3451 an argument to @code{weakref} itself. In either case, @code{weakref}
3452 implicitly marks the declaration as @code{weak}. Without a
3453 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3454 @code{weakref} is equivalent to @code{weak}.
3457 static int x() __attribute__ ((weakref ("y")));
3458 /* is equivalent to... */
3459 static int x() __attribute__ ((weak, weakref, alias ("y")));
3461 static int x() __attribute__ ((weakref));
3462 static int x() __attribute__ ((alias ("y")));
3465 A weak reference is an alias that does not by itself require a
3466 definition to be given for the target symbol. If the target symbol is
3467 only referenced through weak references, then it becomes a @code{weak}
3468 undefined symbol. If it is directly referenced, however, then such
3469 strong references prevail, and a definition is required for the
3470 symbol, not necessarily in the same translation unit.
3472 The effect is equivalent to moving all references to the alias to a
3473 separate translation unit, renaming the alias to the aliased symbol,
3474 declaring it as weak, compiling the two separate translation units and
3475 performing a reloadable link on them.
3477 At present, a declaration to which @code{weakref} is attached can
3478 only be @code{static}.
3483 @c This is the end of the target-independent attribute table
3485 @node AArch64 Function Attributes
3486 @subsection AArch64 Function Attributes
3488 The following target-specific function attributes are available for the
3489 AArch64 target. For the most part, these options mirror the behavior of
3490 similar command-line options (@pxref{AArch64 Options}), but on a
3494 @item general-regs-only
3495 @cindex @code{general-regs-only} function attribute, AArch64
3496 Indicates that no floating-point or Advanced SIMD registers should be
3497 used when generating code for this function. If the function explicitly
3498 uses floating-point code, then the compiler gives an error. This is
3499 the same behavior as that of the command-line option
3500 @option{-mgeneral-regs-only}.
3502 @item fix-cortex-a53-835769
3503 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3504 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3505 applied to this function. To explicitly disable the workaround for this
3506 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3507 This corresponds to the behavior of the command line options
3508 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3511 @cindex @code{cmodel=} function attribute, AArch64
3512 Indicates that code should be generated for a particular code model for
3513 this function. The behavior and permissible arguments are the same as
3514 for the command line option @option{-mcmodel=}.
3517 @cindex @code{strict-align} function attribute, AArch64
3518 Indicates that the compiler should not assume that unaligned memory references
3519 are handled by the system. The behavior is the same as for the command-line
3520 option @option{-mstrict-align}.
3522 @item omit-leaf-frame-pointer
3523 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3524 Indicates that the frame pointer should be omitted for a leaf function call.
3525 To keep the frame pointer, the inverse attribute
3526 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3527 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3528 and @option{-mno-omit-leaf-frame-pointer}.
3531 @cindex @code{tls-dialect=} function attribute, AArch64
3532 Specifies the TLS dialect to use for this function. The behavior and
3533 permissible arguments are the same as for the command-line option
3534 @option{-mtls-dialect=}.
3537 @cindex @code{arch=} function attribute, AArch64
3538 Specifies the architecture version and architectural extensions to use
3539 for this function. The behavior and permissible arguments are the same as
3540 for the @option{-march=} command-line option.
3543 @cindex @code{tune=} function attribute, AArch64
3544 Specifies the core for which to tune the performance of this function.
3545 The behavior and permissible arguments are the same as for the @option{-mtune=}
3546 command-line option.
3549 @cindex @code{cpu=} function attribute, AArch64
3550 Specifies the core for which to tune the performance of this function and also
3551 whose architectural features to use. The behavior and valid arguments are the
3552 same as for the @option{-mcpu=} command-line option.
3554 @item sign-return-address
3555 @cindex @code{sign-return-address} function attribute, AArch64
3556 Select the function scope on which return address signing will be applied. The
3557 behavior and permissible arguments are the same as for the command-line option
3558 @option{-msign-return-address=}. The default value is @code{none}.
3562 The above target attributes can be specified as follows:
3565 __attribute__((target("@var{attr-string}")))
3573 where @code{@var{attr-string}} is one of the attribute strings specified above.
3575 Additionally, the architectural extension string may be specified on its
3576 own. This can be used to turn on and off particular architectural extensions
3577 without having to specify a particular architecture version or core. Example:
3580 __attribute__((target("+crc+nocrypto")))
3588 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3589 extension and disables the @code{crypto} extension for the function @code{foo}
3590 without modifying an existing @option{-march=} or @option{-mcpu} option.
3592 Multiple target function attributes can be specified by separating them with
3593 a comma. For example:
3595 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3603 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3604 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3606 @subsubsection Inlining rules
3607 Specifying target attributes on individual functions or performing link-time
3608 optimization across translation units compiled with different target options
3609 can affect function inlining rules:
3611 In particular, a caller function can inline a callee function only if the
3612 architectural features available to the callee are a subset of the features
3613 available to the caller.
3614 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3615 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3616 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3617 because the all the architectural features that function @code{bar} requires
3618 are available to function @code{foo}. Conversely, function @code{bar} cannot
3619 inline function @code{foo}.
3621 Additionally inlining a function compiled with @option{-mstrict-align} into a
3622 function compiled without @code{-mstrict-align} is not allowed.
3623 However, inlining a function compiled without @option{-mstrict-align} into a
3624 function compiled with @option{-mstrict-align} is allowed.
3626 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3627 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3628 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3629 architectural feature rules specified above.
3631 @node ARC Function Attributes
3632 @subsection ARC Function Attributes
3634 These function attributes are supported by the ARC back end:
3638 @cindex @code{interrupt} function attribute, ARC
3639 Use this attribute to indicate
3640 that the specified function is an interrupt handler. The compiler generates
3641 function entry and exit sequences suitable for use in an interrupt handler
3642 when this attribute is present.
3644 On the ARC, you must specify the kind of interrupt to be handled
3645 in a parameter to the interrupt attribute like this:
3648 void f () __attribute__ ((interrupt ("ilink1")));
3651 Permissible values for this parameter are: @w{@code{ilink1}} and
3657 @cindex @code{long_call} function attribute, ARC
3658 @cindex @code{medium_call} function attribute, ARC
3659 @cindex @code{short_call} function attribute, ARC
3660 @cindex indirect calls, ARC
3661 These attributes specify how a particular function is called.
3662 These attributes override the
3663 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3664 command-line switches and @code{#pragma long_calls} settings.
3666 For ARC, a function marked with the @code{long_call} attribute is
3667 always called using register-indirect jump-and-link instructions,
3668 thereby enabling the called function to be placed anywhere within the
3669 32-bit address space. A function marked with the @code{medium_call}
3670 attribute will always be close enough to be called with an unconditional
3671 branch-and-link instruction, which has a 25-bit offset from
3672 the call site. A function marked with the @code{short_call}
3673 attribute will always be close enough to be called with a conditional
3674 branch-and-link instruction, which has a 21-bit offset from
3678 @node ARM Function Attributes
3679 @subsection ARM Function Attributes
3681 These function attributes are supported for ARM targets:
3685 @cindex @code{interrupt} function attribute, ARM
3686 Use this attribute to indicate
3687 that the specified function is an interrupt handler. The compiler generates
3688 function entry and exit sequences suitable for use in an interrupt handler
3689 when this attribute is present.
3691 You can specify the kind of interrupt to be handled by
3692 adding an optional parameter to the interrupt attribute like this:
3695 void f () __attribute__ ((interrupt ("IRQ")));
3699 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3700 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3702 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3703 may be called with a word-aligned stack pointer.
3706 @cindex @code{isr} function attribute, ARM
3707 Use this attribute on ARM to write Interrupt Service Routines. This is an
3708 alias to the @code{interrupt} attribute above.
3712 @cindex @code{long_call} function attribute, ARM
3713 @cindex @code{short_call} function attribute, ARM
3714 @cindex indirect calls, ARM
3715 These attributes specify how a particular function is called.
3716 These attributes override the
3717 @option{-mlong-calls} (@pxref{ARM Options})
3718 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3719 @code{long_call} attribute indicates that the function might be far
3720 away from the call site and require a different (more expensive)
3721 calling sequence. The @code{short_call} attribute always places
3722 the offset to the function from the call site into the @samp{BL}
3723 instruction directly.
3726 @cindex @code{naked} function attribute, ARM
3727 This attribute allows the compiler to construct the
3728 requisite function declaration, while allowing the body of the
3729 function to be assembly code. The specified function will not have
3730 prologue/epilogue sequences generated by the compiler. Only basic
3731 @code{asm} statements can safely be included in naked functions
3732 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3733 basic @code{asm} and C code may appear to work, they cannot be
3734 depended upon to work reliably and are not supported.
3737 @cindex @code{pcs} function attribute, ARM
3739 The @code{pcs} attribute can be used to control the calling convention
3740 used for a function on ARM. The attribute takes an argument that specifies
3741 the calling convention to use.
3743 When compiling using the AAPCS ABI (or a variant of it) then valid
3744 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3745 order to use a variant other than @code{"aapcs"} then the compiler must
3746 be permitted to use the appropriate co-processor registers (i.e., the
3747 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3751 /* Argument passed in r0, and result returned in r0+r1. */
3752 double f2d (float) __attribute__((pcs("aapcs")));
3755 Variadic functions always use the @code{"aapcs"} calling convention and
3756 the compiler rejects attempts to specify an alternative.
3758 @item target (@var{options})
3759 @cindex @code{target} function attribute
3760 As discussed in @ref{Common Function Attributes}, this attribute
3761 allows specification of target-specific compilation options.
3763 On ARM, the following options are allowed:
3767 @cindex @code{target("thumb")} function attribute, ARM
3768 Force code generation in the Thumb (T16/T32) ISA, depending on the
3772 @cindex @code{target("arm")} function attribute, ARM
3773 Force code generation in the ARM (A32) ISA.
3775 Functions from different modes can be inlined in the caller's mode.
3778 @cindex @code{target("fpu=")} function attribute, ARM
3779 Specifies the fpu for which to tune the performance of this function.
3780 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3781 command-line option.
3787 @node AVR Function Attributes
3788 @subsection AVR Function Attributes
3790 These function attributes are supported by the AVR back end:
3794 @cindex @code{interrupt} function attribute, AVR
3795 Use this attribute to indicate
3796 that the specified function is an interrupt handler. The compiler generates
3797 function entry and exit sequences suitable for use in an interrupt handler
3798 when this attribute is present.
3800 On the AVR, the hardware globally disables interrupts when an
3801 interrupt is executed. The first instruction of an interrupt handler
3802 declared with this attribute is a @code{SEI} instruction to
3803 re-enable interrupts. See also the @code{signal} function attribute
3804 that does not insert a @code{SEI} instruction. If both @code{signal} and
3805 @code{interrupt} are specified for the same function, @code{signal}
3806 is silently ignored.
3809 @cindex @code{naked} function attribute, AVR
3810 This attribute allows the compiler to construct the
3811 requisite function declaration, while allowing the body of the
3812 function to be assembly code. The specified function will not have
3813 prologue/epilogue sequences generated by the compiler. Only basic
3814 @code{asm} statements can safely be included in naked functions
3815 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3816 basic @code{asm} and C code may appear to work, they cannot be
3817 depended upon to work reliably and are not supported.
3820 @cindex @code{no_gccisr} function attribute, AVR
3821 Do not use @code{__gcc_isr} pseudo instructions in a function with
3822 the @code{interrupt} or @code{signal} attribute aka. interrupt
3823 service routine (ISR).
3824 Use this attribute if the preamble of the ISR prologue should always read
3828 in __tmp_reg__, __SREG__
3832 and accordingly for the postamble of the epilogue --- no matter whether
3833 the mentioned registers are actually used in the ISR or not.
3834 Situations where you might want to use this attribute include:
3837 Code that (effectively) clobbers bits of @code{SREG} other than the
3838 @code{I}-flag by writing to the memory location of @code{SREG}.
3840 Code that uses inline assembler to jump to a different function which
3841 expects (parts of) the prologue code as outlined above to be present.
3843 To disable @code{__gcc_isr} generation for the whole compilation unit,
3844 there is option @option{-mno-gas-isr-prologues}, @pxref{AVR Options}.
3848 @cindex @code{OS_main} function attribute, AVR
3849 @cindex @code{OS_task} function attribute, AVR
3850 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3851 do not save/restore any call-saved register in their prologue/epilogue.
3853 The @code{OS_main} attribute can be used when there @emph{is
3854 guarantee} that interrupts are disabled at the time when the function
3855 is entered. This saves resources when the stack pointer has to be
3856 changed to set up a frame for local variables.
3858 The @code{OS_task} attribute can be used when there is @emph{no
3859 guarantee} that interrupts are disabled at that time when the function
3860 is entered like for, e@.g@. task functions in a multi-threading operating
3861 system. In that case, changing the stack pointer register is
3862 guarded by save/clear/restore of the global interrupt enable flag.
3864 The differences to the @code{naked} function attribute are:
3866 @item @code{naked} functions do not have a return instruction whereas
3867 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3868 @code{RETI} return instruction.
3869 @item @code{naked} functions do not set up a frame for local variables
3870 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3875 @cindex @code{signal} function attribute, AVR
3876 Use this attribute on the AVR to indicate that the specified
3877 function is an interrupt handler. The compiler generates function
3878 entry and exit sequences suitable for use in an interrupt handler when this
3879 attribute is present.
3881 See also the @code{interrupt} function attribute.
3883 The AVR hardware globally disables interrupts when an interrupt is executed.
3884 Interrupt handler functions defined with the @code{signal} attribute
3885 do not re-enable interrupts. It is save to enable interrupts in a
3886 @code{signal} handler. This ``save'' only applies to the code
3887 generated by the compiler and not to the IRQ layout of the
3888 application which is responsibility of the application.
3890 If both @code{signal} and @code{interrupt} are specified for the same
3891 function, @code{signal} is silently ignored.
3894 @node Blackfin Function Attributes
3895 @subsection Blackfin Function Attributes
3897 These function attributes are supported by the Blackfin back end:
3901 @item exception_handler
3902 @cindex @code{exception_handler} function attribute
3903 @cindex exception handler functions, Blackfin
3904 Use this attribute on the Blackfin to indicate that the specified function
3905 is an exception handler. The compiler generates function entry and
3906 exit sequences suitable for use in an exception handler when this
3907 attribute is present.
3909 @item interrupt_handler
3910 @cindex @code{interrupt_handler} function attribute, Blackfin
3911 Use this attribute to
3912 indicate that the specified function is an interrupt handler. The compiler
3913 generates function entry and exit sequences suitable for use in an
3914 interrupt handler when this attribute is present.
3917 @cindex @code{kspisusp} function attribute, Blackfin
3918 @cindex User stack pointer in interrupts on the Blackfin
3919 When used together with @code{interrupt_handler}, @code{exception_handler}
3920 or @code{nmi_handler}, code is generated to load the stack pointer
3921 from the USP register in the function prologue.
3924 @cindex @code{l1_text} function attribute, Blackfin
3925 This attribute specifies a function to be placed into L1 Instruction
3926 SRAM@. The function is put into a specific section named @code{.l1.text}.
3927 With @option{-mfdpic}, function calls with a such function as the callee
3928 or caller uses inlined PLT.
3931 @cindex @code{l2} function attribute, Blackfin
3932 This attribute specifies a function to be placed into L2
3933 SRAM. The function is put into a specific section named
3934 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3939 @cindex indirect calls, Blackfin
3940 @cindex @code{longcall} function attribute, Blackfin
3941 @cindex @code{shortcall} function attribute, Blackfin
3942 The @code{longcall} attribute
3943 indicates that the function might be far away from the call site and
3944 require a different (more expensive) calling sequence. The
3945 @code{shortcall} attribute indicates that the function is always close
3946 enough for the shorter calling sequence to be used. These attributes
3947 override the @option{-mlongcall} switch.
3950 @cindex @code{nesting} function attribute, Blackfin
3951 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3952 Use this attribute together with @code{interrupt_handler},
3953 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3954 entry code should enable nested interrupts or exceptions.
3957 @cindex @code{nmi_handler} function attribute, Blackfin
3958 @cindex NMI handler functions on the Blackfin processor
3959 Use this attribute on the Blackfin to indicate that the specified function
3960 is an NMI handler. The compiler generates function entry and
3961 exit sequences suitable for use in an NMI handler when this
3962 attribute is present.
3965 @cindex @code{saveall} function attribute, Blackfin
3966 @cindex save all registers on the Blackfin
3967 Use this attribute to indicate that
3968 all registers except the stack pointer should be saved in the prologue
3969 regardless of whether they are used or not.
3972 @node CR16 Function Attributes
3973 @subsection CR16 Function Attributes
3975 These function attributes are supported by the CR16 back end:
3979 @cindex @code{interrupt} function attribute, CR16
3980 Use this attribute to indicate
3981 that the specified function is an interrupt handler. The compiler generates
3982 function entry and exit sequences suitable for use in an interrupt handler
3983 when this attribute is present.
3986 @node Epiphany Function Attributes
3987 @subsection Epiphany Function Attributes
3989 These function attributes are supported by the Epiphany back end:
3993 @cindex @code{disinterrupt} function attribute, Epiphany
3994 This attribute causes the compiler to emit
3995 instructions to disable interrupts for the duration of the given
3998 @item forwarder_section
3999 @cindex @code{forwarder_section} function attribute, Epiphany
4000 This attribute modifies the behavior of an interrupt handler.
4001 The interrupt handler may be in external memory which cannot be
4002 reached by a branch instruction, so generate a local memory trampoline
4003 to transfer control. The single parameter identifies the section where
4004 the trampoline is placed.
4007 @cindex @code{interrupt} function attribute, Epiphany
4008 Use this attribute to indicate
4009 that the specified function is an interrupt handler. The compiler generates
4010 function entry and exit sequences suitable for use in an interrupt handler
4011 when this attribute is present. It may also generate
4012 a special section with code to initialize the interrupt vector table.
4014 On Epiphany targets one or more optional parameters can be added like this:
4017 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
4020 Permissible values for these parameters are: @w{@code{reset}},
4021 @w{@code{software_exception}}, @w{@code{page_miss}},
4022 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
4023 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
4024 Multiple parameters indicate that multiple entries in the interrupt
4025 vector table should be initialized for this function, i.e.@: for each
4026 parameter @w{@var{name}}, a jump to the function is emitted in
4027 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
4028 entirely, in which case no interrupt vector table entry is provided.
4030 Note that interrupts are enabled inside the function
4031 unless the @code{disinterrupt} attribute is also specified.
4033 The following examples are all valid uses of these attributes on
4036 void __attribute__ ((interrupt)) universal_handler ();
4037 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
4038 void __attribute__ ((interrupt ("dma0, dma1")))
4039 universal_dma_handler ();
4040 void __attribute__ ((interrupt ("timer0"), disinterrupt))
4041 fast_timer_handler ();
4042 void __attribute__ ((interrupt ("dma0, dma1"),
4043 forwarder_section ("tramp")))
4044 external_dma_handler ();
4049 @cindex @code{long_call} function attribute, Epiphany
4050 @cindex @code{short_call} function attribute, Epiphany
4051 @cindex indirect calls, Epiphany
4052 These attributes specify how a particular function is called.
4053 These attributes override the
4054 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
4055 command-line switch and @code{#pragma long_calls} settings.
4059 @node H8/300 Function Attributes
4060 @subsection H8/300 Function Attributes
4062 These function attributes are available for H8/300 targets:
4065 @item function_vector
4066 @cindex @code{function_vector} function attribute, H8/300
4067 Use this attribute on the H8/300, H8/300H, and H8S to indicate
4068 that the specified function should be called through the function vector.
4069 Calling a function through the function vector reduces code size; however,
4070 the function vector has a limited size (maximum 128 entries on the H8/300
4071 and 64 entries on the H8/300H and H8S)
4072 and shares space with the interrupt vector.
4074 @item interrupt_handler
4075 @cindex @code{interrupt_handler} function attribute, H8/300
4076 Use this attribute on the H8/300, H8/300H, and H8S to
4077 indicate that the specified function is an interrupt handler. The compiler
4078 generates function entry and exit sequences suitable for use in an
4079 interrupt handler when this attribute is present.
4082 @cindex @code{saveall} function attribute, H8/300
4083 @cindex save all registers on the H8/300, H8/300H, and H8S
4084 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
4085 all registers except the stack pointer should be saved in the prologue
4086 regardless of whether they are used or not.
4089 @node IA-64 Function Attributes
4090 @subsection IA-64 Function Attributes
4092 These function attributes are supported on IA-64 targets:
4095 @item syscall_linkage
4096 @cindex @code{syscall_linkage} function attribute, IA-64
4097 This attribute is used to modify the IA-64 calling convention by marking
4098 all input registers as live at all function exits. This makes it possible
4099 to restart a system call after an interrupt without having to save/restore
4100 the input registers. This also prevents kernel data from leaking into
4104 @cindex @code{version_id} function attribute, IA-64
4105 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4106 symbol to contain a version string, thus allowing for function level
4107 versioning. HP-UX system header files may use function level versioning
4108 for some system calls.
4111 extern int foo () __attribute__((version_id ("20040821")));
4115 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4118 @node M32C Function Attributes
4119 @subsection M32C Function Attributes
4121 These function attributes are supported by the M32C back end:
4125 @cindex @code{bank_switch} function attribute, M32C
4126 When added to an interrupt handler with the M32C port, causes the
4127 prologue and epilogue to use bank switching to preserve the registers
4128 rather than saving them on the stack.
4130 @item fast_interrupt
4131 @cindex @code{fast_interrupt} function attribute, M32C
4132 Use this attribute on the M32C port to indicate that the specified
4133 function is a fast interrupt handler. This is just like the
4134 @code{interrupt} attribute, except that @code{freit} is used to return
4135 instead of @code{reit}.
4137 @item function_vector
4138 @cindex @code{function_vector} function attribute, M16C/M32C
4139 On M16C/M32C targets, the @code{function_vector} attribute declares a
4140 special page subroutine call function. Use of this attribute reduces
4141 the code size by 2 bytes for each call generated to the
4142 subroutine. The argument to the attribute is the vector number entry
4143 from the special page vector table which contains the 16 low-order
4144 bits of the subroutine's entry address. Each vector table has special
4145 page number (18 to 255) that is used in @code{jsrs} instructions.
4146 Jump addresses of the routines are generated by adding 0x0F0000 (in
4147 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4148 2-byte addresses set in the vector table. Therefore you need to ensure
4149 that all the special page vector routines should get mapped within the
4150 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4153 In the following example 2 bytes are saved for each call to
4154 function @code{foo}.
4157 void foo (void) __attribute__((function_vector(0x18)));
4168 If functions are defined in one file and are called in another file,
4169 then be sure to write this declaration in both files.
4171 This attribute is ignored for R8C target.
4174 @cindex @code{interrupt} function attribute, M32C
4175 Use this attribute to indicate
4176 that the specified function is an interrupt handler. The compiler generates
4177 function entry and exit sequences suitable for use in an interrupt handler
4178 when this attribute is present.
4181 @node M32R/D Function Attributes
4182 @subsection M32R/D Function Attributes
4184 These function attributes are supported by the M32R/D back end:
4188 @cindex @code{interrupt} function attribute, M32R/D
4189 Use this attribute to indicate
4190 that the specified function is an interrupt handler. The compiler generates
4191 function entry and exit sequences suitable for use in an interrupt handler
4192 when this attribute is present.
4194 @item model (@var{model-name})
4195 @cindex @code{model} function attribute, M32R/D
4196 @cindex function addressability on the M32R/D
4198 On the M32R/D, use this attribute to set the addressability of an
4199 object, and of the code generated for a function. The identifier
4200 @var{model-name} is one of @code{small}, @code{medium}, or
4201 @code{large}, representing each of the code models.
4203 Small model objects live in the lower 16MB of memory (so that their
4204 addresses can be loaded with the @code{ld24} instruction), and are
4205 callable with the @code{bl} instruction.
4207 Medium model objects may live anywhere in the 32-bit address space (the
4208 compiler generates @code{seth/add3} instructions to load their addresses),
4209 and are callable with the @code{bl} instruction.
4211 Large model objects may live anywhere in the 32-bit address space (the
4212 compiler generates @code{seth/add3} instructions to load their addresses),
4213 and may not be reachable with the @code{bl} instruction (the compiler
4214 generates the much slower @code{seth/add3/jl} instruction sequence).
4217 @node m68k Function Attributes
4218 @subsection m68k Function Attributes
4220 These function attributes are supported by the m68k back end:
4224 @itemx interrupt_handler
4225 @cindex @code{interrupt} function attribute, m68k
4226 @cindex @code{interrupt_handler} function attribute, m68k
4227 Use this attribute to
4228 indicate that the specified function is an interrupt handler. The compiler
4229 generates function entry and exit sequences suitable for use in an
4230 interrupt handler when this attribute is present. Either name may be used.
4232 @item interrupt_thread
4233 @cindex @code{interrupt_thread} function attribute, fido
4234 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4235 that the specified function is an interrupt handler that is designed
4236 to run as a thread. The compiler omits generate prologue/epilogue
4237 sequences and replaces the return instruction with a @code{sleep}
4238 instruction. This attribute is available only on fido.
4241 @node MCORE Function Attributes
4242 @subsection MCORE Function Attributes
4244 These function attributes are supported by the MCORE back end:
4248 @cindex @code{naked} function attribute, MCORE
4249 This attribute allows the compiler to construct the
4250 requisite function declaration, while allowing the body of the
4251 function to be assembly code. The specified function will not have
4252 prologue/epilogue sequences generated by the compiler. Only basic
4253 @code{asm} statements can safely be included in naked functions
4254 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4255 basic @code{asm} and C code may appear to work, they cannot be
4256 depended upon to work reliably and are not supported.
4259 @node MeP Function Attributes
4260 @subsection MeP Function Attributes
4262 These function attributes are supported by the MeP back end:
4266 @cindex @code{disinterrupt} function attribute, MeP
4267 On MeP targets, this attribute causes the compiler to emit
4268 instructions to disable interrupts for the duration of the given
4272 @cindex @code{interrupt} function attribute, MeP
4273 Use this attribute to indicate
4274 that the specified function is an interrupt handler. The compiler generates
4275 function entry and exit sequences suitable for use in an interrupt handler
4276 when this attribute is present.
4279 @cindex @code{near} function attribute, MeP
4280 This attribute causes the compiler to assume the called
4281 function is close enough to use the normal calling convention,
4282 overriding the @option{-mtf} command-line option.
4285 @cindex @code{far} function attribute, MeP
4286 On MeP targets this causes the compiler to use a calling convention
4287 that assumes the called function is too far away for the built-in
4291 @cindex @code{vliw} function attribute, MeP
4292 The @code{vliw} attribute tells the compiler to emit
4293 instructions in VLIW mode instead of core mode. Note that this
4294 attribute is not allowed unless a VLIW coprocessor has been configured
4295 and enabled through command-line options.
4298 @node MicroBlaze Function Attributes
4299 @subsection MicroBlaze Function Attributes
4301 These function attributes are supported on MicroBlaze targets:
4304 @item save_volatiles
4305 @cindex @code{save_volatiles} function attribute, MicroBlaze
4306 Use this attribute to indicate that the function is
4307 an interrupt handler. All volatile registers (in addition to non-volatile
4308 registers) are saved in the function prologue. If the function is a leaf
4309 function, only volatiles used by the function are saved. A normal function
4310 return is generated instead of a return from interrupt.
4313 @cindex @code{break_handler} function attribute, MicroBlaze
4314 @cindex break handler functions
4315 Use this attribute to indicate that
4316 the specified function is a break handler. The compiler generates function
4317 entry and exit sequences suitable for use in an break handler when this
4318 attribute is present. The return from @code{break_handler} is done through
4319 the @code{rtbd} instead of @code{rtsd}.
4322 void f () __attribute__ ((break_handler));
4325 @item interrupt_handler
4326 @itemx fast_interrupt
4327 @cindex @code{interrupt_handler} function attribute, MicroBlaze
4328 @cindex @code{fast_interrupt} function attribute, MicroBlaze
4329 These attributes indicate that the specified function is an interrupt
4330 handler. Use the @code{fast_interrupt} attribute to indicate handlers
4331 used in low-latency interrupt mode, and @code{interrupt_handler} for
4332 interrupts that do not use low-latency handlers. In both cases, GCC
4333 emits appropriate prologue code and generates a return from the handler
4334 using @code{rtid} instead of @code{rtsd}.
4337 @node Microsoft Windows Function Attributes
4338 @subsection Microsoft Windows Function Attributes
4340 The following attributes are available on Microsoft Windows and Symbian OS
4345 @cindex @code{dllexport} function attribute
4346 @cindex @code{__declspec(dllexport)}
4347 On Microsoft Windows targets and Symbian OS targets the
4348 @code{dllexport} attribute causes the compiler to provide a global
4349 pointer to a pointer in a DLL, so that it can be referenced with the
4350 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4351 name is formed by combining @code{_imp__} and the function or variable
4354 You can use @code{__declspec(dllexport)} as a synonym for
4355 @code{__attribute__ ((dllexport))} for compatibility with other
4358 On systems that support the @code{visibility} attribute, this
4359 attribute also implies ``default'' visibility. It is an error to
4360 explicitly specify any other visibility.
4362 GCC's default behavior is to emit all inline functions with the
4363 @code{dllexport} attribute. Since this can cause object file-size bloat,
4364 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4365 ignore the attribute for inlined functions unless the
4366 @option{-fkeep-inline-functions} flag is used instead.
4368 The attribute is ignored for undefined symbols.
4370 When applied to C++ classes, the attribute marks defined non-inlined
4371 member functions and static data members as exports. Static consts
4372 initialized in-class are not marked unless they are also defined
4375 For Microsoft Windows targets there are alternative methods for
4376 including the symbol in the DLL's export table such as using a
4377 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4378 the @option{--export-all} linker flag.
4381 @cindex @code{dllimport} function attribute
4382 @cindex @code{__declspec(dllimport)}
4383 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4384 attribute causes the compiler to reference a function or variable via
4385 a global pointer to a pointer that is set up by the DLL exporting the
4386 symbol. The attribute implies @code{extern}. On Microsoft Windows
4387 targets, the pointer name is formed by combining @code{_imp__} and the
4388 function or variable name.
4390 You can use @code{__declspec(dllimport)} as a synonym for
4391 @code{__attribute__ ((dllimport))} for compatibility with other
4394 On systems that support the @code{visibility} attribute, this
4395 attribute also implies ``default'' visibility. It is an error to
4396 explicitly specify any other visibility.
4398 Currently, the attribute is ignored for inlined functions. If the
4399 attribute is applied to a symbol @emph{definition}, an error is reported.
4400 If a symbol previously declared @code{dllimport} is later defined, the
4401 attribute is ignored in subsequent references, and a warning is emitted.
4402 The attribute is also overridden by a subsequent declaration as
4405 When applied to C++ classes, the attribute marks non-inlined
4406 member functions and static data members as imports. However, the
4407 attribute is ignored for virtual methods to allow creation of vtables
4410 On the SH Symbian OS target the @code{dllimport} attribute also has
4411 another affect---it can cause the vtable and run-time type information
4412 for a class to be exported. This happens when the class has a
4413 dllimported constructor or a non-inline, non-pure virtual function
4414 and, for either of those two conditions, the class also has an inline
4415 constructor or destructor and has a key function that is defined in
4416 the current translation unit.
4418 For Microsoft Windows targets the use of the @code{dllimport}
4419 attribute on functions is not necessary, but provides a small
4420 performance benefit by eliminating a thunk in the DLL@. The use of the
4421 @code{dllimport} attribute on imported variables can be avoided by passing the
4422 @option{--enable-auto-import} switch to the GNU linker. As with
4423 functions, using the attribute for a variable eliminates a thunk in
4426 One drawback to using this attribute is that a pointer to a
4427 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4428 address. However, a pointer to a @emph{function} with the
4429 @code{dllimport} attribute can be used as a constant initializer; in
4430 this case, the address of a stub function in the import lib is
4431 referenced. On Microsoft Windows targets, the attribute can be disabled
4432 for functions by setting the @option{-mnop-fun-dllimport} flag.
4435 @node MIPS Function Attributes
4436 @subsection MIPS Function Attributes
4438 These function attributes are supported by the MIPS back end:
4442 @cindex @code{interrupt} function attribute, MIPS
4443 Use this attribute to indicate that the specified function is an interrupt
4444 handler. The compiler generates function entry and exit sequences suitable
4445 for use in an interrupt handler when this attribute is present.
4446 An optional argument is supported for the interrupt attribute which allows
4447 the interrupt mode to be described. By default GCC assumes the external
4448 interrupt controller (EIC) mode is in use, this can be explicitly set using
4449 @code{eic}. When interrupts are non-masked then the requested Interrupt
4450 Priority Level (IPL) is copied to the current IPL which has the effect of only
4451 enabling higher priority interrupts. To use vectored interrupt mode use
4452 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4453 the behavior of the non-masked interrupt support and GCC will arrange to mask
4454 all interrupts from sw0 up to and including the specified interrupt vector.
4456 You can use the following attributes to modify the behavior
4457 of an interrupt handler:
4459 @item use_shadow_register_set
4460 @cindex @code{use_shadow_register_set} function attribute, MIPS
4461 Assume that the handler uses a shadow register set, instead of
4462 the main general-purpose registers. An optional argument @code{intstack} is
4463 supported to indicate that the shadow register set contains a valid stack
4466 @item keep_interrupts_masked
4467 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4468 Keep interrupts masked for the whole function. Without this attribute,
4469 GCC tries to reenable interrupts for as much of the function as it can.
4471 @item use_debug_exception_return
4472 @cindex @code{use_debug_exception_return} function attribute, MIPS
4473 Return using the @code{deret} instruction. Interrupt handlers that don't
4474 have this attribute return using @code{eret} instead.
4477 You can use any combination of these attributes, as shown below:
4479 void __attribute__ ((interrupt)) v0 ();
4480 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4481 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4482 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4483 void __attribute__ ((interrupt, use_shadow_register_set,
4484 keep_interrupts_masked)) v4 ();
4485 void __attribute__ ((interrupt, use_shadow_register_set,
4486 use_debug_exception_return)) v5 ();
4487 void __attribute__ ((interrupt, keep_interrupts_masked,
4488 use_debug_exception_return)) v6 ();
4489 void __attribute__ ((interrupt, use_shadow_register_set,
4490 keep_interrupts_masked,
4491 use_debug_exception_return)) v7 ();
4492 void __attribute__ ((interrupt("eic"))) v8 ();
4493 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4499 @cindex indirect calls, MIPS
4500 @cindex @code{long_call} function attribute, MIPS
4501 @cindex @code{near} function attribute, MIPS
4502 @cindex @code{far} function attribute, MIPS
4503 These attributes specify how a particular function is called on MIPS@.
4504 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4505 command-line switch. The @code{long_call} and @code{far} attributes are
4506 synonyms, and cause the compiler to always call
4507 the function by first loading its address into a register, and then using
4508 the contents of that register. The @code{near} attribute has the opposite
4509 effect; it specifies that non-PIC calls should be made using the more
4510 efficient @code{jal} instruction.
4514 @cindex @code{mips16} function attribute, MIPS
4515 @cindex @code{nomips16} function attribute, MIPS
4517 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4518 function attributes to locally select or turn off MIPS16 code generation.
4519 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4520 while MIPS16 code generation is disabled for functions with the
4521 @code{nomips16} attribute. These attributes override the
4522 @option{-mips16} and @option{-mno-mips16} options on the command line
4523 (@pxref{MIPS Options}).
4525 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4526 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4527 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4528 may interact badly with some GCC extensions such as @code{__builtin_apply}
4529 (@pxref{Constructing Calls}).
4531 @item micromips, MIPS
4532 @itemx nomicromips, MIPS
4533 @cindex @code{micromips} function attribute
4534 @cindex @code{nomicromips} function attribute
4536 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4537 function attributes to locally select or turn off microMIPS code generation.
4538 A function with the @code{micromips} attribute is emitted as microMIPS code,
4539 while microMIPS code generation is disabled for functions with the
4540 @code{nomicromips} attribute. These attributes override the
4541 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4542 (@pxref{MIPS Options}).
4544 When compiling files containing mixed microMIPS and non-microMIPS code, the
4545 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4547 not that within individual functions. Mixed microMIPS and non-microMIPS code
4548 may interact badly with some GCC extensions such as @code{__builtin_apply}
4549 (@pxref{Constructing Calls}).
4552 @cindex @code{nocompression} function attribute, MIPS
4553 On MIPS targets, you can use the @code{nocompression} function attribute
4554 to locally turn off MIPS16 and microMIPS code generation. This attribute
4555 overrides the @option{-mips16} and @option{-mmicromips} options on the
4556 command line (@pxref{MIPS Options}).
4559 @node MSP430 Function Attributes
4560 @subsection MSP430 Function Attributes
4562 These function attributes are supported by the MSP430 back end:
4566 @cindex @code{critical} function attribute, MSP430
4567 Critical functions disable interrupts upon entry and restore the
4568 previous interrupt state upon exit. Critical functions cannot also
4569 have the @code{naked} or @code{reentrant} attributes. They can have
4570 the @code{interrupt} attribute.
4573 @cindex @code{interrupt} function attribute, MSP430
4574 Use this attribute to indicate
4575 that the specified function is an interrupt handler. The compiler generates
4576 function entry and exit sequences suitable for use in an interrupt handler
4577 when this attribute is present.
4579 You can provide an argument to the interrupt
4580 attribute which specifies a name or number. If the argument is a
4581 number it indicates the slot in the interrupt vector table (0 - 31) to
4582 which this handler should be assigned. If the argument is a name it
4583 is treated as a symbolic name for the vector slot. These names should
4584 match up with appropriate entries in the linker script. By default
4585 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4586 @code{reset} for vector 31 are recognized.
4589 @cindex @code{naked} function attribute, MSP430
4590 This attribute allows the compiler to construct the
4591 requisite function declaration, while allowing the body of the
4592 function to be assembly code. The specified function will not have
4593 prologue/epilogue sequences generated by the compiler. Only basic
4594 @code{asm} statements can safely be included in naked functions
4595 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4596 basic @code{asm} and C code may appear to work, they cannot be
4597 depended upon to work reliably and are not supported.
4600 @cindex @code{reentrant} function attribute, MSP430
4601 Reentrant functions disable interrupts upon entry and enable them
4602 upon exit. Reentrant functions cannot also have the @code{naked}
4603 or @code{critical} attributes. They can have the @code{interrupt}
4607 @cindex @code{wakeup} function attribute, MSP430
4608 This attribute only applies to interrupt functions. It is silently
4609 ignored if applied to a non-interrupt function. A wakeup interrupt
4610 function will rouse the processor from any low-power state that it
4611 might be in when the function exits.
4616 @cindex @code{lower} function attribute, MSP430
4617 @cindex @code{upper} function attribute, MSP430
4618 @cindex @code{either} function attribute, MSP430
4619 On the MSP430 target these attributes can be used to specify whether
4620 the function or variable should be placed into low memory, high
4621 memory, or the placement should be left to the linker to decide. The
4622 attributes are only significant if compiling for the MSP430X
4625 The attributes work in conjunction with a linker script that has been
4626 augmented to specify where to place sections with a @code{.lower} and
4627 a @code{.upper} prefix. So, for example, as well as placing the
4628 @code{.data} section, the script also specifies the placement of a
4629 @code{.lower.data} and a @code{.upper.data} section. The intention
4630 is that @code{lower} sections are placed into a small but easier to
4631 access memory region and the upper sections are placed into a larger, but
4632 slower to access, region.
4634 The @code{either} attribute is special. It tells the linker to place
4635 the object into the corresponding @code{lower} section if there is
4636 room for it. If there is insufficient room then the object is placed
4637 into the corresponding @code{upper} section instead. Note that the
4638 placement algorithm is not very sophisticated. It does not attempt to
4639 find an optimal packing of the @code{lower} sections. It just makes
4640 one pass over the objects and does the best that it can. Using the
4641 @option{-ffunction-sections} and @option{-fdata-sections} command-line
4642 options can help the packing, however, since they produce smaller,
4643 easier to pack regions.
4646 @node NDS32 Function Attributes
4647 @subsection NDS32 Function Attributes
4649 These function attributes are supported by the NDS32 back end:
4653 @cindex @code{exception} function attribute
4654 @cindex exception handler functions, NDS32
4655 Use this attribute on the NDS32 target to indicate that the specified function
4656 is an exception handler. The compiler will generate corresponding sections
4657 for use in an exception handler.
4660 @cindex @code{interrupt} function attribute, NDS32
4661 On NDS32 target, this attribute indicates that the specified function
4662 is an interrupt handler. The compiler generates corresponding sections
4663 for use in an interrupt handler. You can use the following attributes
4664 to modify the behavior:
4667 @cindex @code{nested} function attribute, NDS32
4668 This interrupt service routine is interruptible.
4670 @cindex @code{not_nested} function attribute, NDS32
4671 This interrupt service routine is not interruptible.
4673 @cindex @code{nested_ready} function attribute, NDS32
4674 This interrupt service routine is interruptible after @code{PSW.GIE}
4675 (global interrupt enable) is set. This allows interrupt service routine to
4676 finish some short critical code before enabling interrupts.
4678 @cindex @code{save_all} function attribute, NDS32
4679 The system will help save all registers into stack before entering
4682 @cindex @code{partial_save} function attribute, NDS32
4683 The system will help save caller registers into stack before entering
4688 @cindex @code{naked} function attribute, NDS32
4689 This attribute allows the compiler to construct the
4690 requisite function declaration, while allowing the body of the
4691 function to be assembly code. The specified function will not have
4692 prologue/epilogue sequences generated by the compiler. Only basic
4693 @code{asm} statements can safely be included in naked functions
4694 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4695 basic @code{asm} and C code may appear to work, they cannot be
4696 depended upon to work reliably and are not supported.
4699 @cindex @code{reset} function attribute, NDS32
4700 @cindex reset handler functions
4701 Use this attribute on the NDS32 target to indicate that the specified function
4702 is a reset handler. The compiler will generate corresponding sections
4703 for use in a reset handler. You can use the following attributes
4704 to provide extra exception handling:
4707 @cindex @code{nmi} function attribute, NDS32
4708 Provide a user-defined function to handle NMI exception.
4710 @cindex @code{warm} function attribute, NDS32
4711 Provide a user-defined function to handle warm reset exception.
4715 @node Nios II Function Attributes
4716 @subsection Nios II Function Attributes
4718 These function attributes are supported by the Nios II back end:
4721 @item target (@var{options})
4722 @cindex @code{target} function attribute
4723 As discussed in @ref{Common Function Attributes}, this attribute
4724 allows specification of target-specific compilation options.
4726 When compiling for Nios II, the following options are allowed:
4729 @item custom-@var{insn}=@var{N}
4730 @itemx no-custom-@var{insn}
4731 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4732 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4733 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4734 custom instruction with encoding @var{N} when generating code that uses
4735 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4736 the custom instruction @var{insn}.
4737 These target attributes correspond to the
4738 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4739 command-line options, and support the same set of @var{insn} keywords.
4740 @xref{Nios II Options}, for more information.
4742 @item custom-fpu-cfg=@var{name}
4743 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4744 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4745 command-line option, to select a predefined set of custom instructions
4747 @xref{Nios II Options}, for more information.
4751 @node Nvidia PTX Function Attributes
4752 @subsection Nvidia PTX Function Attributes
4754 These function attributes are supported by the Nvidia PTX back end:
4758 @cindex @code{kernel} attribute, Nvidia PTX
4759 This attribute indicates that the corresponding function should be compiled
4760 as a kernel function, which can be invoked from the host via the CUDA RT
4762 By default functions are only callable only from other PTX functions.
4764 Kernel functions must have @code{void} return type.
4767 @node PowerPC Function Attributes
4768 @subsection PowerPC Function Attributes
4770 These function attributes are supported by the PowerPC back end:
4775 @cindex indirect calls, PowerPC
4776 @cindex @code{longcall} function attribute, PowerPC
4777 @cindex @code{shortcall} function attribute, PowerPC
4778 The @code{longcall} attribute
4779 indicates that the function might be far away from the call site and
4780 require a different (more expensive) calling sequence. The
4781 @code{shortcall} attribute indicates that the function is always close
4782 enough for the shorter calling sequence to be used. These attributes
4783 override both the @option{-mlongcall} switch and
4784 the @code{#pragma longcall} setting.
4786 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4787 calls are necessary.
4789 @item target (@var{options})
4790 @cindex @code{target} function attribute
4791 As discussed in @ref{Common Function Attributes}, this attribute
4792 allows specification of target-specific compilation options.
4794 On the PowerPC, the following options are allowed:
4799 @cindex @code{target("altivec")} function attribute, PowerPC
4800 Generate code that uses (does not use) AltiVec instructions. In
4801 32-bit code, you cannot enable AltiVec instructions unless
4802 @option{-mabi=altivec} is used on the command line.
4806 @cindex @code{target("cmpb")} function attribute, PowerPC
4807 Generate code that uses (does not use) the compare bytes instruction
4808 implemented on the POWER6 processor and other processors that support
4809 the PowerPC V2.05 architecture.
4813 @cindex @code{target("dlmzb")} function attribute, PowerPC
4814 Generate code that uses (does not use) the string-search @samp{dlmzb}
4815 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4816 generated by default when targeting those processors.
4820 @cindex @code{target("fprnd")} function attribute, PowerPC
4821 Generate code that uses (does not use) the FP round to integer
4822 instructions implemented on the POWER5+ processor and other processors
4823 that support the PowerPC V2.03 architecture.
4827 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4828 Generate code that uses (does not use) the decimal floating-point
4829 instructions implemented on some POWER processors.
4833 @cindex @code{target("isel")} function attribute, PowerPC
4834 Generate code that uses (does not use) ISEL instruction.
4838 @cindex @code{target("mfcrf")} function attribute, PowerPC
4839 Generate code that uses (does not use) the move from condition
4840 register field instruction implemented on the POWER4 processor and
4841 other processors that support the PowerPC V2.01 architecture.
4845 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4846 Generate code that uses (does not use) the FP move to/from general
4847 purpose register instructions implemented on the POWER6X processor and
4848 other processors that support the extended PowerPC V2.05 architecture.
4852 @cindex @code{target("mulhw")} function attribute, PowerPC
4853 Generate code that uses (does not use) the half-word multiply and
4854 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4855 These instructions are generated by default when targeting those
4860 @cindex @code{target("multiple")} function attribute, PowerPC
4861 Generate code that uses (does not use) the load multiple word
4862 instructions and the store multiple word instructions.
4866 @cindex @code{target("update")} function attribute, PowerPC
4867 Generate code that uses (does not use) the load or store instructions
4868 that update the base register to the address of the calculated memory
4873 @cindex @code{target("popcntb")} function attribute, PowerPC
4874 Generate code that uses (does not use) the popcount and double-precision
4875 FP reciprocal estimate instruction implemented on the POWER5
4876 processor and other processors that support the PowerPC V2.02
4881 @cindex @code{target("popcntd")} function attribute, PowerPC
4882 Generate code that uses (does not use) the popcount instruction
4883 implemented on the POWER7 processor and other processors that support
4884 the PowerPC V2.06 architecture.
4886 @item powerpc-gfxopt
4887 @itemx no-powerpc-gfxopt
4888 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4889 Generate code that uses (does not use) the optional PowerPC
4890 architecture instructions in the Graphics group, including
4891 floating-point select.
4894 @itemx no-powerpc-gpopt
4895 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4896 Generate code that uses (does not use) the optional PowerPC
4897 architecture instructions in the General Purpose group, including
4898 floating-point square root.
4900 @item recip-precision
4901 @itemx no-recip-precision
4902 @cindex @code{target("recip-precision")} function attribute, PowerPC
4903 Assume (do not assume) that the reciprocal estimate instructions
4904 provide higher-precision estimates than is mandated by the PowerPC
4909 @cindex @code{target("string")} function attribute, PowerPC
4910 Generate code that uses (does not use) the load string instructions
4911 and the store string word instructions to save multiple registers and
4912 do small block moves.
4916 @cindex @code{target("vsx")} function attribute, PowerPC
4917 Generate code that uses (does not use) vector/scalar (VSX)
4918 instructions, and also enable the use of built-in functions that allow
4919 more direct access to the VSX instruction set. In 32-bit code, you
4920 cannot enable VSX or AltiVec instructions unless
4921 @option{-mabi=altivec} is used on the command line.
4925 @cindex @code{target("friz")} function attribute, PowerPC
4926 Generate (do not generate) the @code{friz} instruction when the
4927 @option{-funsafe-math-optimizations} option is used to optimize
4928 rounding a floating-point value to 64-bit integer and back to floating
4929 point. The @code{friz} instruction does not return the same value if
4930 the floating-point number is too large to fit in an integer.
4932 @item avoid-indexed-addresses
4933 @itemx no-avoid-indexed-addresses
4934 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4935 Generate code that tries to avoid (not avoid) the use of indexed load
4936 or store instructions.
4940 @cindex @code{target("paired")} function attribute, PowerPC
4941 Generate code that uses (does not use) the generation of PAIRED simd
4946 @cindex @code{target("longcall")} function attribute, PowerPC
4947 Generate code that assumes (does not assume) that all calls are far
4948 away so that a longer more expensive calling sequence is required.
4951 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4952 Specify the architecture to generate code for when compiling the
4953 function. If you select the @code{target("cpu=power7")} attribute when
4954 generating 32-bit code, VSX and AltiVec instructions are not generated
4955 unless you use the @option{-mabi=altivec} option on the command line.
4957 @item tune=@var{TUNE}
4958 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4959 Specify the architecture to tune for when compiling the function. If
4960 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4961 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4962 compilation tunes for the @var{CPU} architecture, and not the
4963 default tuning specified on the command line.
4966 On the PowerPC, the inliner does not inline a
4967 function that has different target options than the caller, unless the
4968 callee has a subset of the target options of the caller.
4971 @node RL78 Function Attributes
4972 @subsection RL78 Function Attributes
4974 These function attributes are supported by the RL78 back end:
4978 @itemx brk_interrupt
4979 @cindex @code{interrupt} function attribute, RL78
4980 @cindex @code{brk_interrupt} function attribute, RL78
4981 These attributes indicate
4982 that the specified function is an interrupt handler. The compiler generates
4983 function entry and exit sequences suitable for use in an interrupt handler
4984 when this attribute is present.
4986 Use @code{brk_interrupt} instead of @code{interrupt} for
4987 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4988 that must end with @code{RETB} instead of @code{RETI}).
4991 @cindex @code{naked} function attribute, RL78
4992 This attribute allows the compiler to construct the
4993 requisite function declaration, while allowing the body of the
4994 function to be assembly code. The specified function will not have
4995 prologue/epilogue sequences generated by the compiler. Only basic
4996 @code{asm} statements can safely be included in naked functions
4997 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4998 basic @code{asm} and C code may appear to work, they cannot be
4999 depended upon to work reliably and are not supported.
5002 @node RX Function Attributes
5003 @subsection RX Function Attributes
5005 These function attributes are supported by the RX back end:
5008 @item fast_interrupt
5009 @cindex @code{fast_interrupt} function attribute, RX
5010 Use this attribute on the RX port to indicate that the specified
5011 function is a fast interrupt handler. This is just like the
5012 @code{interrupt} attribute, except that @code{freit} is used to return
5013 instead of @code{reit}.
5016 @cindex @code{interrupt} function attribute, RX
5017 Use this attribute to indicate
5018 that the specified function is an interrupt handler. The compiler generates
5019 function entry and exit sequences suitable for use in an interrupt handler
5020 when this attribute is present.
5022 On RX targets, you may specify one or more vector numbers as arguments
5023 to the attribute, as well as naming an alternate table name.
5024 Parameters are handled sequentially, so one handler can be assigned to
5025 multiple entries in multiple tables. One may also pass the magic
5026 string @code{"$default"} which causes the function to be used for any
5027 unfilled slots in the current table.
5029 This example shows a simple assignment of a function to one vector in
5030 the default table (note that preprocessor macros may be used for
5031 chip-specific symbolic vector names):
5033 void __attribute__ ((interrupt (5))) txd1_handler ();
5036 This example assigns a function to two slots in the default table
5037 (using preprocessor macros defined elsewhere) and makes it the default
5038 for the @code{dct} table:
5040 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
5045 @cindex @code{naked} function attribute, RX
5046 This attribute allows the compiler to construct the
5047 requisite function declaration, while allowing the body of the
5048 function to be assembly code. The specified function will not have
5049 prologue/epilogue sequences generated by the compiler. Only basic
5050 @code{asm} statements can safely be included in naked functions
5051 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5052 basic @code{asm} and C code may appear to work, they cannot be
5053 depended upon to work reliably and are not supported.
5056 @cindex @code{vector} function attribute, RX
5057 This RX attribute is similar to the @code{interrupt} attribute, including its
5058 parameters, but does not make the function an interrupt-handler type
5059 function (i.e. it retains the normal C function calling ABI). See the
5060 @code{interrupt} attribute for a description of its arguments.
5063 @node S/390 Function Attributes
5064 @subsection S/390 Function Attributes
5066 These function attributes are supported on the S/390:
5069 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
5070 @cindex @code{hotpatch} function attribute, S/390
5072 On S/390 System z targets, you can use this function attribute to
5073 make GCC generate a ``hot-patching'' function prologue. If the
5074 @option{-mhotpatch=} command-line option is used at the same time,
5075 the @code{hotpatch} attribute takes precedence. The first of the
5076 two arguments specifies the number of halfwords to be added before
5077 the function label. A second argument can be used to specify the
5078 number of halfwords to be added after the function label. For
5079 both arguments the maximum allowed value is 1000000.
5081 If both arguments are zero, hotpatching is disabled.
5083 @item target (@var{options})
5084 @cindex @code{target} function attribute
5085 As discussed in @ref{Common Function Attributes}, this attribute
5086 allows specification of target-specific compilation options.
5088 On S/390, the following options are supported:
5096 @item warn-framesize=
5108 @itemx no-packed-stack
5110 @itemx no-small-exec
5113 @item warn-dynamicstack
5114 @itemx no-warn-dynamicstack
5117 The options work exactly like the S/390 specific command line
5118 options (without the prefix @option{-m}) except that they do not
5119 change any feature macros. For example,
5122 @code{target("no-vx")}
5125 does not undefine the @code{__VEC__} macro.
5128 @node SH Function Attributes
5129 @subsection SH Function Attributes
5131 These function attributes are supported on the SH family of processors:
5134 @item function_vector
5135 @cindex @code{function_vector} function attribute, SH
5136 @cindex calling functions through the function vector on SH2A
5137 On SH2A targets, this attribute declares a function to be called using the
5138 TBR relative addressing mode. The argument to this attribute is the entry
5139 number of the same function in a vector table containing all the TBR
5140 relative addressable functions. For correct operation the TBR must be setup
5141 accordingly to point to the start of the vector table before any functions with
5142 this attribute are invoked. Usually a good place to do the initialization is
5143 the startup routine. The TBR relative vector table can have at max 256 function
5144 entries. The jumps to these functions are generated using a SH2A specific,
5145 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5146 from GNU binutils version 2.7 or later for this attribute to work correctly.
5148 In an application, for a function being called once, this attribute
5149 saves at least 8 bytes of code; and if other successive calls are being
5150 made to the same function, it saves 2 bytes of code per each of these
5153 @item interrupt_handler
5154 @cindex @code{interrupt_handler} function attribute, SH
5155 Use this attribute to
5156 indicate that the specified function is an interrupt handler. The compiler
5157 generates function entry and exit sequences suitable for use in an
5158 interrupt handler when this attribute is present.
5160 @item nosave_low_regs
5161 @cindex @code{nosave_low_regs} function attribute, SH
5162 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5163 function should not save and restore registers R0..R7. This can be used on SH3*
5164 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5168 @cindex @code{renesas} function attribute, SH
5169 On SH targets this attribute specifies that the function or struct follows the
5173 @cindex @code{resbank} function attribute, SH
5174 On the SH2A target, this attribute enables the high-speed register
5175 saving and restoration using a register bank for @code{interrupt_handler}
5176 routines. Saving to the bank is performed automatically after the CPU
5177 accepts an interrupt that uses a register bank.
5179 The nineteen 32-bit registers comprising general register R0 to R14,
5180 control register GBR, and system registers MACH, MACL, and PR and the
5181 vector table address offset are saved into a register bank. Register
5182 banks are stacked in first-in last-out (FILO) sequence. Restoration
5183 from the bank is executed by issuing a RESBANK instruction.
5186 @cindex @code{sp_switch} function attribute, SH
5187 Use this attribute on the SH to indicate an @code{interrupt_handler}
5188 function should switch to an alternate stack. It expects a string
5189 argument that names a global variable holding the address of the
5194 void f () __attribute__ ((interrupt_handler,
5195 sp_switch ("alt_stack")));
5199 @cindex @code{trap_exit} function attribute, SH
5200 Use this attribute on the SH for an @code{interrupt_handler} to return using
5201 @code{trapa} instead of @code{rte}. This attribute expects an integer
5202 argument specifying the trap number to be used.
5205 @cindex @code{trapa_handler} function attribute, SH
5206 On SH targets this function attribute is similar to @code{interrupt_handler}
5207 but it does not save and restore all registers.
5210 @node SPU Function Attributes
5211 @subsection SPU Function Attributes
5213 These function attributes are supported by the SPU back end:
5217 @cindex @code{naked} function attribute, SPU
5218 This attribute allows the compiler to construct the
5219 requisite function declaration, while allowing the body of the
5220 function to be assembly code. The specified function will not have
5221 prologue/epilogue sequences generated by the compiler. Only basic
5222 @code{asm} statements can safely be included in naked functions
5223 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5224 basic @code{asm} and C code may appear to work, they cannot be
5225 depended upon to work reliably and are not supported.
5228 @node Symbian OS Function Attributes
5229 @subsection Symbian OS Function Attributes
5231 @xref{Microsoft Windows Function Attributes}, for discussion of the
5232 @code{dllexport} and @code{dllimport} attributes.
5234 @node V850 Function Attributes
5235 @subsection V850 Function Attributes
5237 The V850 back end supports these function attributes:
5241 @itemx interrupt_handler
5242 @cindex @code{interrupt} function attribute, V850
5243 @cindex @code{interrupt_handler} function attribute, V850
5244 Use these attributes to indicate
5245 that the specified function is an interrupt handler. The compiler generates
5246 function entry and exit sequences suitable for use in an interrupt handler
5247 when either attribute is present.
5250 @node Visium Function Attributes
5251 @subsection Visium Function Attributes
5253 These function attributes are supported by the Visium back end:
5257 @cindex @code{interrupt} function attribute, Visium
5258 Use this attribute to indicate
5259 that the specified function is an interrupt handler. The compiler generates
5260 function entry and exit sequences suitable for use in an interrupt handler
5261 when this attribute is present.
5264 @node x86 Function Attributes
5265 @subsection x86 Function Attributes
5267 These function attributes are supported by the x86 back end:
5271 @cindex @code{cdecl} function attribute, x86-32
5272 @cindex functions that pop the argument stack on x86-32
5274 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5275 assume that the calling function pops off the stack space used to
5276 pass arguments. This is
5277 useful to override the effects of the @option{-mrtd} switch.
5280 @cindex @code{fastcall} function attribute, x86-32
5281 @cindex functions that pop the argument stack on x86-32
5282 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5283 pass the first argument (if of integral type) in the register ECX and
5284 the second argument (if of integral type) in the register EDX@. Subsequent
5285 and other typed arguments are passed on the stack. The called function
5286 pops the arguments off the stack. If the number of arguments is variable all
5287 arguments are pushed on the stack.
5290 @cindex @code{thiscall} function attribute, x86-32
5291 @cindex functions that pop the argument stack on x86-32
5292 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5293 pass the first argument (if of integral type) in the register ECX.
5294 Subsequent and other typed arguments are passed on the stack. The called
5295 function pops the arguments off the stack.
5296 If the number of arguments is variable all arguments are pushed on the
5298 The @code{thiscall} attribute is intended for C++ non-static member functions.
5299 As a GCC extension, this calling convention can be used for C functions
5300 and for static member methods.
5304 @cindex @code{ms_abi} function attribute, x86
5305 @cindex @code{sysv_abi} function attribute, x86
5307 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5308 to indicate which calling convention should be used for a function. The
5309 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5310 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5311 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5312 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5314 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5315 requires the @option{-maccumulate-outgoing-args} option.
5317 @item callee_pop_aggregate_return (@var{number})
5318 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5320 On x86-32 targets, you can use this attribute to control how
5321 aggregates are returned in memory. If the caller is responsible for
5322 popping the hidden pointer together with the rest of the arguments, specify
5323 @var{number} equal to zero. If callee is responsible for popping the
5324 hidden pointer, specify @var{number} equal to one.
5326 The default x86-32 ABI assumes that the callee pops the
5327 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5328 the compiler assumes that the
5329 caller pops the stack for hidden pointer.
5331 @item ms_hook_prologue
5332 @cindex @code{ms_hook_prologue} function attribute, x86
5334 On 32-bit and 64-bit x86 targets, you can use
5335 this function attribute to make GCC generate the ``hot-patching'' function
5336 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5339 @item regparm (@var{number})
5340 @cindex @code{regparm} function attribute, x86
5341 @cindex functions that are passed arguments in registers on x86-32
5342 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5343 pass arguments number one to @var{number} if they are of integral type
5344 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5345 take a variable number of arguments continue to be passed all of their
5346 arguments on the stack.
5348 Beware that on some ELF systems this attribute is unsuitable for
5349 global functions in shared libraries with lazy binding (which is the
5350 default). Lazy binding sends the first call via resolving code in
5351 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5352 per the standard calling conventions. Solaris 8 is affected by this.
5353 Systems with the GNU C Library version 2.1 or higher
5354 and FreeBSD are believed to be
5355 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5356 disabled with the linker or the loader if desired, to avoid the
5360 @cindex @code{sseregparm} function attribute, x86
5361 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5362 causes the compiler to pass up to 3 floating-point arguments in
5363 SSE registers instead of on the stack. Functions that take a
5364 variable number of arguments continue to pass all of their
5365 floating-point arguments on the stack.
5367 @item force_align_arg_pointer
5368 @cindex @code{force_align_arg_pointer} function attribute, x86
5369 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5370 applied to individual function definitions, generating an alternate
5371 prologue and epilogue that realigns the run-time stack if necessary.
5372 This supports mixing legacy codes that run with a 4-byte aligned stack
5373 with modern codes that keep a 16-byte stack for SSE compatibility.
5376 @cindex @code{stdcall} function attribute, x86-32
5377 @cindex functions that pop the argument stack on x86-32
5378 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5379 assume that the called function pops off the stack space used to
5380 pass arguments, unless it takes a variable number of arguments.
5382 @item no_caller_saved_registers
5383 @cindex @code{no_caller_saved_registers} function attribute, x86
5384 Use this attribute to indicate that the specified function has no
5385 caller-saved registers. That is, all registers are callee-saved. For
5386 example, this attribute can be used for a function called from an
5387 interrupt handler. The compiler generates proper function entry and
5388 exit sequences to save and restore any modified registers, except for
5389 the EFLAGS register. Since GCC doesn't preserve MPX, SSE, MMX nor x87
5390 states, the GCC option @option{-mgeneral-regs-only} should be used to
5391 compile functions with @code{no_caller_saved_registers} attribute.
5394 @cindex @code{interrupt} function attribute, x86
5395 Use this attribute to indicate that the specified function is an
5396 interrupt handler or an exception handler (depending on parameters passed
5397 to the function, explained further). The compiler generates function
5398 entry and exit sequences suitable for use in an interrupt handler when
5399 this attribute is present. The @code{IRET} instruction, instead of the
5400 @code{RET} instruction, is used to return from interrupt handlers. All
5401 registers, except for the EFLAGS register which is restored by the
5402 @code{IRET} instruction, are preserved by the compiler. Since GCC
5403 doesn't preserve MPX, SSE, MMX nor x87 states, the GCC option
5404 @option{-mgeneral-regs-only} should be used to compile interrupt and
5407 Any interruptible-without-stack-switch code must be compiled with
5408 @option{-mno-red-zone} since interrupt handlers can and will, because
5409 of the hardware design, touch the red zone.
5411 An interrupt handler must be declared with a mandatory pointer
5415 struct interrupt_frame;
5417 __attribute__ ((interrupt))
5419 f (struct interrupt_frame *frame)
5425 and you must define @code{struct interrupt_frame} as described in the
5428 Exception handlers differ from interrupt handlers because the system
5429 pushes an error code on the stack. An exception handler declaration is
5430 similar to that for an interrupt handler, but with a different mandatory
5431 function signature. The compiler arranges to pop the error code off the
5432 stack before the @code{IRET} instruction.
5436 typedef unsigned long long int uword_t;
5438 typedef unsigned int uword_t;
5441 struct interrupt_frame;
5443 __attribute__ ((interrupt))
5445 f (struct interrupt_frame *frame, uword_t error_code)
5451 Exception handlers should only be used for exceptions that push an error
5452 code; you should use an interrupt handler in other cases. The system
5453 will crash if the wrong kind of handler is used.
5455 @item target (@var{options})
5456 @cindex @code{target} function attribute
5457 As discussed in @ref{Common Function Attributes}, this attribute
5458 allows specification of target-specific compilation options.
5460 On the x86, the following options are allowed:
5464 @cindex @code{target("abm")} function attribute, x86
5465 Enable/disable the generation of the advanced bit instructions.
5469 @cindex @code{target("aes")} function attribute, x86
5470 Enable/disable the generation of the AES instructions.
5473 @cindex @code{target("default")} function attribute, x86
5474 @xref{Function Multiversioning}, where it is used to specify the
5475 default function version.
5479 @cindex @code{target("mmx")} function attribute, x86
5480 Enable/disable the generation of the MMX instructions.
5484 @cindex @code{target("pclmul")} function attribute, x86
5485 Enable/disable the generation of the PCLMUL instructions.
5489 @cindex @code{target("popcnt")} function attribute, x86
5490 Enable/disable the generation of the POPCNT instruction.
5494 @cindex @code{target("sse")} function attribute, x86
5495 Enable/disable the generation of the SSE instructions.
5499 @cindex @code{target("sse2")} function attribute, x86
5500 Enable/disable the generation of the SSE2 instructions.
5504 @cindex @code{target("sse3")} function attribute, x86
5505 Enable/disable the generation of the SSE3 instructions.
5509 @cindex @code{target("sse4")} function attribute, x86
5510 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5515 @cindex @code{target("sse4.1")} function attribute, x86
5516 Enable/disable the generation of the sse4.1 instructions.
5520 @cindex @code{target("sse4.2")} function attribute, x86
5521 Enable/disable the generation of the sse4.2 instructions.
5525 @cindex @code{target("sse4a")} function attribute, x86
5526 Enable/disable the generation of the SSE4A instructions.
5530 @cindex @code{target("fma4")} function attribute, x86
5531 Enable/disable the generation of the FMA4 instructions.
5535 @cindex @code{target("xop")} function attribute, x86
5536 Enable/disable the generation of the XOP instructions.
5540 @cindex @code{target("lwp")} function attribute, x86
5541 Enable/disable the generation of the LWP instructions.
5545 @cindex @code{target("ssse3")} function attribute, x86
5546 Enable/disable the generation of the SSSE3 instructions.
5550 @cindex @code{target("cld")} function attribute, x86
5551 Enable/disable the generation of the CLD before string moves.
5553 @item fancy-math-387
5554 @itemx no-fancy-math-387
5555 @cindex @code{target("fancy-math-387")} function attribute, x86
5556 Enable/disable the generation of the @code{sin}, @code{cos}, and
5557 @code{sqrt} instructions on the 387 floating-point unit.
5561 @cindex @code{target("ieee-fp")} function attribute, x86
5562 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5564 @item inline-all-stringops
5565 @itemx no-inline-all-stringops
5566 @cindex @code{target("inline-all-stringops")} function attribute, x86
5567 Enable/disable inlining of string operations.
5569 @item inline-stringops-dynamically
5570 @itemx no-inline-stringops-dynamically
5571 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5572 Enable/disable the generation of the inline code to do small string
5573 operations and calling the library routines for large operations.
5575 @item align-stringops
5576 @itemx no-align-stringops
5577 @cindex @code{target("align-stringops")} function attribute, x86
5578 Do/do not align destination of inlined string operations.
5582 @cindex @code{target("recip")} function attribute, x86
5583 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5584 instructions followed an additional Newton-Raphson step instead of
5585 doing a floating-point division.
5587 @item arch=@var{ARCH}
5588 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5589 Specify the architecture to generate code for in compiling the function.
5591 @item tune=@var{TUNE}
5592 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5593 Specify the architecture to tune for in compiling the function.
5595 @item fpmath=@var{FPMATH}
5596 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5597 Specify which floating-point unit to use. You must specify the
5598 @code{target("fpmath=sse,387")} option as
5599 @code{target("fpmath=sse+387")} because the comma would separate
5603 On the x86, the inliner does not inline a
5604 function that has different target options than the caller, unless the
5605 callee has a subset of the target options of the caller. For example
5606 a function declared with @code{target("sse3")} can inline a function
5607 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5610 @node Xstormy16 Function Attributes
5611 @subsection Xstormy16 Function Attributes
5613 These function attributes are supported by the Xstormy16 back end:
5617 @cindex @code{interrupt} function attribute, Xstormy16
5618 Use this attribute to indicate
5619 that the specified function is an interrupt handler. The compiler generates
5620 function entry and exit sequences suitable for use in an interrupt handler
5621 when this attribute is present.
5624 @node Variable Attributes
5625 @section Specifying Attributes of Variables
5626 @cindex attribute of variables
5627 @cindex variable attributes
5629 The keyword @code{__attribute__} allows you to specify special
5630 attributes of variables or structure fields. This keyword is followed
5631 by an attribute specification inside double parentheses. Some
5632 attributes are currently defined generically for variables.
5633 Other attributes are defined for variables on particular target
5634 systems. Other attributes are available for functions
5635 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5636 enumerators (@pxref{Enumerator Attributes}), statements
5637 (@pxref{Statement Attributes}), and for types (@pxref{Type Attributes}).
5638 Other front ends might define more attributes
5639 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5641 @xref{Attribute Syntax}, for details of the exact syntax for using
5645 * Common Variable Attributes::
5646 * AVR Variable Attributes::
5647 * Blackfin Variable Attributes::
5648 * H8/300 Variable Attributes::
5649 * IA-64 Variable Attributes::
5650 * M32R/D Variable Attributes::
5651 * MeP Variable Attributes::
5652 * Microsoft Windows Variable Attributes::
5653 * MSP430 Variable Attributes::
5654 * Nvidia PTX Variable Attributes::
5655 * PowerPC Variable Attributes::
5656 * RL78 Variable Attributes::
5657 * SPU Variable Attributes::
5658 * V850 Variable Attributes::
5659 * x86 Variable Attributes::
5660 * Xstormy16 Variable Attributes::
5663 @node Common Variable Attributes
5664 @subsection Common Variable Attributes
5666 The following attributes are supported on most targets.
5669 @cindex @code{aligned} variable attribute
5670 @item aligned (@var{alignment})
5671 This attribute specifies a minimum alignment for the variable or
5672 structure field, measured in bytes. For example, the declaration:
5675 int x __attribute__ ((aligned (16))) = 0;
5679 causes the compiler to allocate the global variable @code{x} on a
5680 16-byte boundary. On a 68040, this could be used in conjunction with
5681 an @code{asm} expression to access the @code{move16} instruction which
5682 requires 16-byte aligned operands.
5684 You can also specify the alignment of structure fields. For example, to
5685 create a double-word aligned @code{int} pair, you could write:
5688 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5692 This is an alternative to creating a union with a @code{double} member,
5693 which forces the union to be double-word aligned.
5695 As in the preceding examples, you can explicitly specify the alignment
5696 (in bytes) that you wish the compiler to use for a given variable or
5697 structure field. Alternatively, you can leave out the alignment factor
5698 and just ask the compiler to align a variable or field to the
5699 default alignment for the target architecture you are compiling for.
5700 The default alignment is sufficient for all scalar types, but may not be
5701 enough for all vector types on a target that supports vector operations.
5702 The default alignment is fixed for a particular target ABI.
5704 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5705 which is the largest alignment ever used for any data type on the
5706 target machine you are compiling for. For example, you could write:
5709 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5712 The compiler automatically sets the alignment for the declared
5713 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5714 often make copy operations more efficient, because the compiler can
5715 use whatever instructions copy the biggest chunks of memory when
5716 performing copies to or from the variables or fields that you have
5717 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5718 may change depending on command-line options.
5720 When used on a struct, or struct member, the @code{aligned} attribute can
5721 only increase the alignment; in order to decrease it, the @code{packed}
5722 attribute must be specified as well. When used as part of a typedef, the
5723 @code{aligned} attribute can both increase and decrease alignment, and
5724 specifying the @code{packed} attribute generates a warning.
5726 Note that the effectiveness of @code{aligned} attributes may be limited
5727 by inherent limitations in your linker. On many systems, the linker is
5728 only able to arrange for variables to be aligned up to a certain maximum
5729 alignment. (For some linkers, the maximum supported alignment may
5730 be very very small.) If your linker is only able to align variables
5731 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5732 in an @code{__attribute__} still only provides you with 8-byte
5733 alignment. See your linker documentation for further information.
5735 The @code{aligned} attribute can also be used for functions
5736 (@pxref{Common Function Attributes}.)
5738 @item cleanup (@var{cleanup_function})
5739 @cindex @code{cleanup} variable attribute
5740 The @code{cleanup} attribute runs a function when the variable goes
5741 out of scope. This attribute can only be applied to auto function
5742 scope variables; it may not be applied to parameters or variables
5743 with static storage duration. The function must take one parameter,
5744 a pointer to a type compatible with the variable. The return value
5745 of the function (if any) is ignored.
5747 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5748 is run during the stack unwinding that happens during the
5749 processing of the exception. Note that the @code{cleanup} attribute
5750 does not allow the exception to be caught, only to perform an action.
5751 It is undefined what happens if @var{cleanup_function} does not
5756 @cindex @code{common} variable attribute
5757 @cindex @code{nocommon} variable attribute
5760 The @code{common} attribute requests GCC to place a variable in
5761 ``common'' storage. The @code{nocommon} attribute requests the
5762 opposite---to allocate space for it directly.
5764 These attributes override the default chosen by the
5765 @option{-fno-common} and @option{-fcommon} flags respectively.
5768 @itemx deprecated (@var{msg})
5769 @cindex @code{deprecated} variable attribute
5770 The @code{deprecated} attribute results in a warning if the variable
5771 is used anywhere in the source file. This is useful when identifying
5772 variables that are expected to be removed in a future version of a
5773 program. The warning also includes the location of the declaration
5774 of the deprecated variable, to enable users to easily find further
5775 information about why the variable is deprecated, or what they should
5776 do instead. Note that the warning only occurs for uses:
5779 extern int old_var __attribute__ ((deprecated));
5781 int new_fn () @{ return old_var; @}
5785 results in a warning on line 3 but not line 2. The optional @var{msg}
5786 argument, which must be a string, is printed in the warning if
5789 The @code{deprecated} attribute can also be used for functions and
5790 types (@pxref{Common Function Attributes},
5791 @pxref{Common Type Attributes}).
5793 @item mode (@var{mode})
5794 @cindex @code{mode} variable attribute
5795 This attribute specifies the data type for the declaration---whichever
5796 type corresponds to the mode @var{mode}. This in effect lets you
5797 request an integer or floating-point type according to its width.
5799 @xref{Machine Modes,,, gccint, GNU Compiler Collection (GCC) Internals},
5800 for a list of the possible keywords for @var{mode}.
5801 You may also specify a mode of @code{byte} or @code{__byte__} to
5802 indicate the mode corresponding to a one-byte integer, @code{word} or
5803 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5804 or @code{__pointer__} for the mode used to represent pointers.
5807 @cindex @code{packed} variable attribute
5808 The @code{packed} attribute specifies that a variable or structure field
5809 should have the smallest possible alignment---one byte for a variable,
5810 and one bit for a field, unless you specify a larger value with the
5811 @code{aligned} attribute.
5813 Here is a structure in which the field @code{x} is packed, so that it
5814 immediately follows @code{a}:
5820 int x[2] __attribute__ ((packed));
5824 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5825 @code{packed} attribute on bit-fields of type @code{char}. This has
5826 been fixed in GCC 4.4 but the change can lead to differences in the
5827 structure layout. See the documentation of
5828 @option{-Wpacked-bitfield-compat} for more information.
5830 @item section ("@var{section-name}")
5831 @cindex @code{section} variable attribute
5832 Normally, the compiler places the objects it generates in sections like
5833 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5834 or you need certain particular variables to appear in special sections,
5835 for example to map to special hardware. The @code{section}
5836 attribute specifies that a variable (or function) lives in a particular
5837 section. For example, this small program uses several specific section names:
5840 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5841 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5842 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5843 int init_data __attribute__ ((section ("INITDATA")));
5847 /* @r{Initialize stack pointer} */
5848 init_sp (stack + sizeof (stack));
5850 /* @r{Initialize initialized data} */
5851 memcpy (&init_data, &data, &edata - &data);
5853 /* @r{Turn on the serial ports} */
5860 Use the @code{section} attribute with
5861 @emph{global} variables and not @emph{local} variables,
5862 as shown in the example.
5864 You may use the @code{section} attribute with initialized or
5865 uninitialized global variables but the linker requires
5866 each object be defined once, with the exception that uninitialized
5867 variables tentatively go in the @code{common} (or @code{bss}) section
5868 and can be multiply ``defined''. Using the @code{section} attribute
5869 changes what section the variable goes into and may cause the
5870 linker to issue an error if an uninitialized variable has multiple
5871 definitions. You can force a variable to be initialized with the
5872 @option{-fno-common} flag or the @code{nocommon} attribute.
5874 Some file formats do not support arbitrary sections so the @code{section}
5875 attribute is not available on all platforms.
5876 If you need to map the entire contents of a module to a particular
5877 section, consider using the facilities of the linker instead.
5879 @item tls_model ("@var{tls_model}")
5880 @cindex @code{tls_model} variable attribute
5881 The @code{tls_model} attribute sets thread-local storage model
5882 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5883 overriding @option{-ftls-model=} command-line switch on a per-variable
5885 The @var{tls_model} argument should be one of @code{global-dynamic},
5886 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5888 Not all targets support this attribute.
5891 @cindex @code{unused} variable attribute
5892 This attribute, attached to a variable, means that the variable is meant
5893 to be possibly unused. GCC does not produce a warning for this
5897 @cindex @code{used} variable attribute
5898 This attribute, attached to a variable with static storage, means that
5899 the variable must be emitted even if it appears that the variable is not
5902 When applied to a static data member of a C++ class template, the
5903 attribute also means that the member is instantiated if the
5904 class itself is instantiated.
5906 @item vector_size (@var{bytes})
5907 @cindex @code{vector_size} variable attribute
5908 This attribute specifies the vector size for the variable, measured in
5909 bytes. For example, the declaration:
5912 int foo __attribute__ ((vector_size (16)));
5916 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5917 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5918 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5920 This attribute is only applicable to integral and float scalars,
5921 although arrays, pointers, and function return values are allowed in
5922 conjunction with this construct.
5924 Aggregates with this attribute are invalid, even if they are of the same
5925 size as a corresponding scalar. For example, the declaration:
5928 struct S @{ int a; @};
5929 struct S __attribute__ ((vector_size (16))) foo;
5933 is invalid even if the size of the structure is the same as the size of
5936 @item visibility ("@var{visibility_type}")
5937 @cindex @code{visibility} variable attribute
5938 This attribute affects the linkage of the declaration to which it is attached.
5939 The @code{visibility} attribute is described in
5940 @ref{Common Function Attributes}.
5943 @cindex @code{weak} variable attribute
5944 The @code{weak} attribute is described in
5945 @ref{Common Function Attributes}.
5949 @node AVR Variable Attributes
5950 @subsection AVR Variable Attributes
5954 @cindex @code{progmem} variable attribute, AVR
5955 The @code{progmem} attribute is used on the AVR to place read-only
5956 data in the non-volatile program memory (flash). The @code{progmem}
5957 attribute accomplishes this by putting respective variables into a
5958 section whose name starts with @code{.progmem}.
5960 This attribute works similar to the @code{section} attribute
5961 but adds additional checking.
5964 @item @bullet{}@tie{} Ordinary AVR cores with 32 general purpose registers:
5965 @code{progmem} affects the location
5966 of the data but not how this data is accessed.
5967 In order to read data located with the @code{progmem} attribute
5968 (inline) assembler must be used.
5970 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5971 #include <avr/pgmspace.h>
5973 /* Locate var in flash memory */
5974 const int var[2] PROGMEM = @{ 1, 2 @};
5976 int read_var (int i)
5978 /* Access var[] by accessor macro from avr/pgmspace.h */
5979 return (int) pgm_read_word (& var[i]);
5983 AVR is a Harvard architecture processor and data and read-only data
5984 normally resides in the data memory (RAM).
5986 See also the @ref{AVR Named Address Spaces} section for
5987 an alternate way to locate and access data in flash memory.
5989 @item @bullet{}@tie{} AVR cores with flash memory visible in the RAM address range:
5990 On such devices, there is no need for attribute @code{progmem} or
5991 @ref{AVR Named Address Spaces,,@code{__flash}} qualifier at all.
5992 Just use standard C / C++. The compiler will generate @code{LD*}
5993 instructions. As flash memory is visible in the RAM address range,
5994 and the default linker script does @emph{not} locate @code{.rodata} in
5995 RAM, no special features are needed in order not to waste RAM for
5996 read-only data or to read from flash. You might even get slightly better
5998 avoiding @code{progmem} and @code{__flash}. This applies to devices from
5999 families @code{avrtiny} and @code{avrxmega3}, see @ref{AVR Options} for
6002 @item @bullet{}@tie{}Reduced AVR Tiny cores like ATtiny40:
6003 The compiler adds @code{0x4000}
6004 to the addresses of objects and declarations in @code{progmem} and locates
6005 the objects in flash memory, namely in section @code{.progmem.data}.
6006 The offset is needed because the flash memory is visible in the RAM
6007 address space starting at address @code{0x4000}.
6009 Data in @code{progmem} can be accessed by means of ordinary C@tie{}code,
6010 no special functions or macros are needed.
6013 /* var is located in flash memory */
6014 extern const int var[2] __attribute__((progmem));
6016 int read_var (int i)
6022 Please notice that on these devices, there is no need for @code{progmem}
6028 @itemx io (@var{addr})
6029 @cindex @code{io} variable attribute, AVR
6030 Variables with the @code{io} attribute are used to address
6031 memory-mapped peripherals in the io address range.
6032 If an address is specified, the variable
6033 is assigned that address, and the value is interpreted as an
6034 address in the data address space.
6038 volatile int porta __attribute__((io (0x22)));
6041 The address specified in the address in the data address range.
6043 Otherwise, the variable it is not assigned an address, but the
6044 compiler will still use in/out instructions where applicable,
6045 assuming some other module assigns an address in the io address range.
6049 extern volatile int porta __attribute__((io));
6053 @itemx io_low (@var{addr})
6054 @cindex @code{io_low} variable attribute, AVR
6055 This is like the @code{io} attribute, but additionally it informs the
6056 compiler that the object lies in the lower half of the I/O area,
6057 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
6061 @itemx address (@var{addr})
6062 @cindex @code{address} variable attribute, AVR
6063 Variables with the @code{address} attribute are used to address
6064 memory-mapped peripherals that may lie outside the io address range.
6067 volatile int porta __attribute__((address (0x600)));
6071 @cindex @code{absdata} variable attribute, AVR
6072 Variables in static storage and with the @code{absdata} attribute can
6073 be accessed by the @code{LDS} and @code{STS} instructions which take
6078 This attribute is only supported for the reduced AVR Tiny core
6082 You must make sure that respective data is located in the
6083 address range @code{0x40}@dots{}@code{0xbf} accessible by
6084 @code{LDS} and @code{STS}. One way to achieve this as an
6085 appropriate linker description file.
6088 If the location does not fit the address range of @code{LDS}
6089 and @code{STS}, there is currently (Binutils 2.26) just an unspecific
6092 @code{module.c:(.text+0x1c): warning: internal error: out of range error}
6097 See also the @option{-mabsdata} @ref{AVR Options,command-line option}.
6101 @node Blackfin Variable Attributes
6102 @subsection Blackfin Variable Attributes
6104 Three attributes are currently defined for the Blackfin.
6110 @cindex @code{l1_data} variable attribute, Blackfin
6111 @cindex @code{l1_data_A} variable attribute, Blackfin
6112 @cindex @code{l1_data_B} variable attribute, Blackfin
6113 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
6114 Variables with @code{l1_data} attribute are put into the specific section
6115 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
6116 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
6117 attribute are put into the specific section named @code{.l1.data.B}.
6120 @cindex @code{l2} variable attribute, Blackfin
6121 Use this attribute on the Blackfin to place the variable into L2 SRAM.
6122 Variables with @code{l2} attribute are put into the specific section
6123 named @code{.l2.data}.
6126 @node H8/300 Variable Attributes
6127 @subsection H8/300 Variable Attributes
6129 These variable attributes are available for H8/300 targets:
6133 @cindex @code{eightbit_data} variable attribute, H8/300
6134 @cindex eight-bit data on the H8/300, H8/300H, and H8S
6135 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
6136 variable should be placed into the eight-bit data section.
6137 The compiler generates more efficient code for certain operations
6138 on data in the eight-bit data area. Note the eight-bit data area is limited to
6141 You must use GAS and GLD from GNU binutils version 2.7 or later for
6142 this attribute to work correctly.
6145 @cindex @code{tiny_data} variable attribute, H8/300
6146 @cindex tiny data section on the H8/300H and H8S
6147 Use this attribute on the H8/300H and H8S to indicate that the specified
6148 variable should be placed into the tiny data section.
6149 The compiler generates more efficient code for loads and stores
6150 on data in the tiny data section. Note the tiny data area is limited to
6151 slightly under 32KB of data.
6155 @node IA-64 Variable Attributes
6156 @subsection IA-64 Variable Attributes
6158 The IA-64 back end supports the following variable attribute:
6161 @item model (@var{model-name})
6162 @cindex @code{model} variable attribute, IA-64
6164 On IA-64, use this attribute to set the addressability of an object.
6165 At present, the only supported identifier for @var{model-name} is
6166 @code{small}, indicating addressability via ``small'' (22-bit)
6167 addresses (so that their addresses can be loaded with the @code{addl}
6168 instruction). Caveat: such addressing is by definition not position
6169 independent and hence this attribute must not be used for objects
6170 defined by shared libraries.
6174 @node M32R/D Variable Attributes
6175 @subsection M32R/D Variable Attributes
6177 One attribute is currently defined for the M32R/D@.
6180 @item model (@var{model-name})
6181 @cindex @code{model-name} variable attribute, M32R/D
6182 @cindex variable addressability on the M32R/D
6183 Use this attribute on the M32R/D to set the addressability of an object.
6184 The identifier @var{model-name} is one of @code{small}, @code{medium},
6185 or @code{large}, representing each of the code models.
6187 Small model objects live in the lower 16MB of memory (so that their
6188 addresses can be loaded with the @code{ld24} instruction).
6190 Medium and large model objects may live anywhere in the 32-bit address space
6191 (the compiler generates @code{seth/add3} instructions to load their
6195 @node MeP Variable Attributes
6196 @subsection MeP Variable Attributes
6198 The MeP target has a number of addressing modes and busses. The
6199 @code{near} space spans the standard memory space's first 16 megabytes
6200 (24 bits). The @code{far} space spans the entire 32-bit memory space.
6201 The @code{based} space is a 128-byte region in the memory space that
6202 is addressed relative to the @code{$tp} register. The @code{tiny}
6203 space is a 65536-byte region relative to the @code{$gp} register. In
6204 addition to these memory regions, the MeP target has a separate 16-bit
6205 control bus which is specified with @code{cb} attributes.
6210 @cindex @code{based} variable attribute, MeP
6211 Any variable with the @code{based} attribute is assigned to the
6212 @code{.based} section, and is accessed with relative to the
6213 @code{$tp} register.
6216 @cindex @code{tiny} variable attribute, MeP
6217 Likewise, the @code{tiny} attribute assigned variables to the
6218 @code{.tiny} section, relative to the @code{$gp} register.
6221 @cindex @code{near} variable attribute, MeP
6222 Variables with the @code{near} attribute are assumed to have addresses
6223 that fit in a 24-bit addressing mode. This is the default for large
6224 variables (@code{-mtiny=4} is the default) but this attribute can
6225 override @code{-mtiny=} for small variables, or override @code{-ml}.
6228 @cindex @code{far} variable attribute, MeP
6229 Variables with the @code{far} attribute are addressed using a full
6230 32-bit address. Since this covers the entire memory space, this
6231 allows modules to make no assumptions about where variables might be
6235 @cindex @code{io} variable attribute, MeP
6236 @itemx io (@var{addr})
6237 Variables with the @code{io} attribute are used to address
6238 memory-mapped peripherals. If an address is specified, the variable
6239 is assigned that address, else it is not assigned an address (it is
6240 assumed some other module assigns an address). Example:
6243 int timer_count __attribute__((io(0x123)));
6247 @itemx cb (@var{addr})
6248 @cindex @code{cb} variable attribute, MeP
6249 Variables with the @code{cb} attribute are used to access the control
6250 bus, using special instructions. @code{addr} indicates the control bus
6254 int cpu_clock __attribute__((cb(0x123)));
6259 @node Microsoft Windows Variable Attributes
6260 @subsection Microsoft Windows Variable Attributes
6262 You can use these attributes on Microsoft Windows targets.
6263 @ref{x86 Variable Attributes} for additional Windows compatibility
6264 attributes available on all x86 targets.
6269 @cindex @code{dllimport} variable attribute
6270 @cindex @code{dllexport} variable attribute
6271 The @code{dllimport} and @code{dllexport} attributes are described in
6272 @ref{Microsoft Windows Function Attributes}.
6275 @cindex @code{selectany} variable attribute
6276 The @code{selectany} attribute causes an initialized global variable to
6277 have link-once semantics. When multiple definitions of the variable are
6278 encountered by the linker, the first is selected and the remainder are
6279 discarded. Following usage by the Microsoft compiler, the linker is told
6280 @emph{not} to warn about size or content differences of the multiple
6283 Although the primary usage of this attribute is for POD types, the
6284 attribute can also be applied to global C++ objects that are initialized
6285 by a constructor. In this case, the static initialization and destruction
6286 code for the object is emitted in each translation defining the object,
6287 but the calls to the constructor and destructor are protected by a
6288 link-once guard variable.
6290 The @code{selectany} attribute is only available on Microsoft Windows
6291 targets. You can use @code{__declspec (selectany)} as a synonym for
6292 @code{__attribute__ ((selectany))} for compatibility with other
6296 @cindex @code{shared} variable attribute
6297 On Microsoft Windows, in addition to putting variable definitions in a named
6298 section, the section can also be shared among all running copies of an
6299 executable or DLL@. For example, this small program defines shared data
6300 by putting it in a named section @code{shared} and marking the section
6304 int foo __attribute__((section ("shared"), shared)) = 0;
6309 /* @r{Read and write foo. All running
6310 copies see the same value.} */
6316 You may only use the @code{shared} attribute along with @code{section}
6317 attribute with a fully-initialized global definition because of the way
6318 linkers work. See @code{section} attribute for more information.
6320 The @code{shared} attribute is only available on Microsoft Windows@.
6324 @node MSP430 Variable Attributes
6325 @subsection MSP430 Variable Attributes
6329 @cindex @code{noinit} variable attribute, MSP430
6330 Any data with the @code{noinit} attribute will not be initialised by
6331 the C runtime startup code, or the program loader. Not initialising
6332 data in this way can reduce program startup times.
6335 @cindex @code{persistent} variable attribute, MSP430
6336 Any variable with the @code{persistent} attribute will not be
6337 initialised by the C runtime startup code. Instead its value will be
6338 set once, when the application is loaded, and then never initialised
6339 again, even if the processor is reset or the program restarts.
6340 Persistent data is intended to be placed into FLASH RAM, where its
6341 value will be retained across resets. The linker script being used to
6342 create the application should ensure that persistent data is correctly
6348 @cindex @code{lower} variable attribute, MSP430
6349 @cindex @code{upper} variable attribute, MSP430
6350 @cindex @code{either} variable attribute, MSP430
6351 These attributes are the same as the MSP430 function attributes of the
6352 same name (@pxref{MSP430 Function Attributes}).
6353 These attributes can be applied to both functions and variables.
6356 @node Nvidia PTX Variable Attributes
6357 @subsection Nvidia PTX Variable Attributes
6359 These variable attributes are supported by the Nvidia PTX back end:
6363 @cindex @code{shared} attribute, Nvidia PTX
6364 Use this attribute to place a variable in the @code{.shared} memory space.
6365 This memory space is private to each cooperative thread array; only threads
6366 within one thread block refer to the same instance of the variable.
6367 The runtime does not initialize variables in this memory space.
6370 @node PowerPC Variable Attributes
6371 @subsection PowerPC Variable Attributes
6373 Three attributes currently are defined for PowerPC configurations:
6374 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6376 @cindex @code{ms_struct} variable attribute, PowerPC
6377 @cindex @code{gcc_struct} variable attribute, PowerPC
6378 For full documentation of the struct attributes please see the
6379 documentation in @ref{x86 Variable Attributes}.
6381 @cindex @code{altivec} variable attribute, PowerPC
6382 For documentation of @code{altivec} attribute please see the
6383 documentation in @ref{PowerPC Type Attributes}.
6385 @node RL78 Variable Attributes
6386 @subsection RL78 Variable Attributes
6388 @cindex @code{saddr} variable attribute, RL78
6389 The RL78 back end supports the @code{saddr} variable attribute. This
6390 specifies placement of the corresponding variable in the SADDR area,
6391 which can be accessed more efficiently than the default memory region.
6393 @node SPU Variable Attributes
6394 @subsection SPU Variable Attributes
6396 @cindex @code{spu_vector} variable attribute, SPU
6397 The SPU supports the @code{spu_vector} attribute for variables. For
6398 documentation of this attribute please see the documentation in
6399 @ref{SPU Type Attributes}.
6401 @node V850 Variable Attributes
6402 @subsection V850 Variable Attributes
6404 These variable attributes are supported by the V850 back end:
6409 @cindex @code{sda} variable attribute, V850
6410 Use this attribute to explicitly place a variable in the small data area,
6411 which can hold up to 64 kilobytes.
6414 @cindex @code{tda} variable attribute, V850
6415 Use this attribute to explicitly place a variable in the tiny data area,
6416 which can hold up to 256 bytes in total.
6419 @cindex @code{zda} variable attribute, V850
6420 Use this attribute to explicitly place a variable in the first 32 kilobytes
6424 @node x86 Variable Attributes
6425 @subsection x86 Variable Attributes
6427 Two attributes are currently defined for x86 configurations:
6428 @code{ms_struct} and @code{gcc_struct}.
6433 @cindex @code{ms_struct} variable attribute, x86
6434 @cindex @code{gcc_struct} variable attribute, x86
6436 If @code{packed} is used on a structure, or if bit-fields are used,
6437 it may be that the Microsoft ABI lays out the structure differently
6438 than the way GCC normally does. Particularly when moving packed
6439 data between functions compiled with GCC and the native Microsoft compiler
6440 (either via function call or as data in a file), it may be necessary to access
6443 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6444 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6445 command-line options, respectively;
6446 see @ref{x86 Options}, for details of how structure layout is affected.
6447 @xref{x86 Type Attributes}, for information about the corresponding
6448 attributes on types.
6452 @node Xstormy16 Variable Attributes
6453 @subsection Xstormy16 Variable Attributes
6455 One attribute is currently defined for xstormy16 configurations:
6460 @cindex @code{below100} variable attribute, Xstormy16
6462 If a variable has the @code{below100} attribute (@code{BELOW100} is
6463 allowed also), GCC places the variable in the first 0x100 bytes of
6464 memory and use special opcodes to access it. Such variables are
6465 placed in either the @code{.bss_below100} section or the
6466 @code{.data_below100} section.
6470 @node Type Attributes
6471 @section Specifying Attributes of Types
6472 @cindex attribute of types
6473 @cindex type attributes
6475 The keyword @code{__attribute__} allows you to specify special
6476 attributes of types. Some type attributes apply only to @code{struct}
6477 and @code{union} types, while others can apply to any type defined
6478 via a @code{typedef} declaration. Other attributes are defined for
6479 functions (@pxref{Function Attributes}), labels (@pxref{Label
6480 Attributes}), enumerators (@pxref{Enumerator Attributes}),
6481 statements (@pxref{Statement Attributes}), and for
6482 variables (@pxref{Variable Attributes}).
6484 The @code{__attribute__} keyword is followed by an attribute specification
6485 inside double parentheses.
6487 You may specify type attributes in an enum, struct or union type
6488 declaration or definition by placing them immediately after the
6489 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6490 syntax is to place them just past the closing curly brace of the
6493 You can also include type attributes in a @code{typedef} declaration.
6494 @xref{Attribute Syntax}, for details of the exact syntax for using
6498 * Common Type Attributes::
6499 * ARM Type Attributes::
6500 * MeP Type Attributes::
6501 * PowerPC Type Attributes::
6502 * SPU Type Attributes::
6503 * x86 Type Attributes::
6506 @node Common Type Attributes
6507 @subsection Common Type Attributes
6509 The following type attributes are supported on most targets.
6512 @cindex @code{aligned} type attribute
6513 @item aligned (@var{alignment})
6514 This attribute specifies a minimum alignment (in bytes) for variables
6515 of the specified type. For example, the declarations:
6518 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6519 typedef int more_aligned_int __attribute__ ((aligned (8)));
6523 force the compiler to ensure (as far as it can) that each variable whose
6524 type is @code{struct S} or @code{more_aligned_int} is allocated and
6525 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6526 variables of type @code{struct S} aligned to 8-byte boundaries allows
6527 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6528 store) instructions when copying one variable of type @code{struct S} to
6529 another, thus improving run-time efficiency.
6531 Note that the alignment of any given @code{struct} or @code{union} type
6532 is required by the ISO C standard to be at least a perfect multiple of
6533 the lowest common multiple of the alignments of all of the members of
6534 the @code{struct} or @code{union} in question. This means that you @emph{can}
6535 effectively adjust the alignment of a @code{struct} or @code{union}
6536 type by attaching an @code{aligned} attribute to any one of the members
6537 of such a type, but the notation illustrated in the example above is a
6538 more obvious, intuitive, and readable way to request the compiler to
6539 adjust the alignment of an entire @code{struct} or @code{union} type.
6541 As in the preceding example, you can explicitly specify the alignment
6542 (in bytes) that you wish the compiler to use for a given @code{struct}
6543 or @code{union} type. Alternatively, you can leave out the alignment factor
6544 and just ask the compiler to align a type to the maximum
6545 useful alignment for the target machine you are compiling for. For
6546 example, you could write:
6549 struct S @{ short f[3]; @} __attribute__ ((aligned));
6552 Whenever you leave out the alignment factor in an @code{aligned}
6553 attribute specification, the compiler automatically sets the alignment
6554 for the type to the largest alignment that is ever used for any data
6555 type on the target machine you are compiling for. Doing this can often
6556 make copy operations more efficient, because the compiler can use
6557 whatever instructions copy the biggest chunks of memory when performing
6558 copies to or from the variables that have types that you have aligned
6561 In the example above, if the size of each @code{short} is 2 bytes, then
6562 the size of the entire @code{struct S} type is 6 bytes. The smallest
6563 power of two that is greater than or equal to that is 8, so the
6564 compiler sets the alignment for the entire @code{struct S} type to 8
6567 Note that although you can ask the compiler to select a time-efficient
6568 alignment for a given type and then declare only individual stand-alone
6569 objects of that type, the compiler's ability to select a time-efficient
6570 alignment is primarily useful only when you plan to create arrays of
6571 variables having the relevant (efficiently aligned) type. If you
6572 declare or use arrays of variables of an efficiently-aligned type, then
6573 it is likely that your program also does pointer arithmetic (or
6574 subscripting, which amounts to the same thing) on pointers to the
6575 relevant type, and the code that the compiler generates for these
6576 pointer arithmetic operations is often more efficient for
6577 efficiently-aligned types than for other types.
6579 Note that the effectiveness of @code{aligned} attributes may be limited
6580 by inherent limitations in your linker. On many systems, the linker is
6581 only able to arrange for variables to be aligned up to a certain maximum
6582 alignment. (For some linkers, the maximum supported alignment may
6583 be very very small.) If your linker is only able to align variables
6584 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6585 in an @code{__attribute__} still only provides you with 8-byte
6586 alignment. See your linker documentation for further information.
6588 The @code{aligned} attribute can only increase alignment. Alignment
6589 can be decreased by specifying the @code{packed} attribute. See below.
6591 @item bnd_variable_size
6592 @cindex @code{bnd_variable_size} type attribute
6593 @cindex Pointer Bounds Checker attributes
6594 When applied to a structure field, this attribute tells Pointer
6595 Bounds Checker that the size of this field should not be computed
6596 using static type information. It may be used to mark variably-sized
6597 static array fields placed at the end of a structure.
6605 S *p = (S *)malloc (sizeof(S) + 100);
6606 p->data[10] = 0; //Bounds violation
6610 By using an attribute for the field we may avoid unwanted bound
6617 char data[1] __attribute__((bnd_variable_size));
6619 S *p = (S *)malloc (sizeof(S) + 100);
6620 p->data[10] = 0; //OK
6624 @itemx deprecated (@var{msg})
6625 @cindex @code{deprecated} type attribute
6626 The @code{deprecated} attribute results in a warning if the type
6627 is used anywhere in the source file. This is useful when identifying
6628 types that are expected to be removed in a future version of a program.
6629 If possible, the warning also includes the location of the declaration
6630 of the deprecated type, to enable users to easily find further
6631 information about why the type is deprecated, or what they should do
6632 instead. Note that the warnings only occur for uses and then only
6633 if the type is being applied to an identifier that itself is not being
6634 declared as deprecated.
6637 typedef int T1 __attribute__ ((deprecated));
6641 typedef T1 T3 __attribute__ ((deprecated));
6642 T3 z __attribute__ ((deprecated));
6646 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6647 warning is issued for line 4 because T2 is not explicitly
6648 deprecated. Line 5 has no warning because T3 is explicitly
6649 deprecated. Similarly for line 6. The optional @var{msg}
6650 argument, which must be a string, is printed in the warning if
6653 The @code{deprecated} attribute can also be used for functions and
6654 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6656 @item designated_init
6657 @cindex @code{designated_init} type attribute
6658 This attribute may only be applied to structure types. It indicates
6659 that any initialization of an object of this type must use designated
6660 initializers rather than positional initializers. The intent of this
6661 attribute is to allow the programmer to indicate that a structure's
6662 layout may change, and that therefore relying on positional
6663 initialization will result in future breakage.
6665 GCC emits warnings based on this attribute by default; use
6666 @option{-Wno-designated-init} to suppress them.
6669 @cindex @code{may_alias} type attribute
6670 Accesses through pointers to types with this attribute are not subject
6671 to type-based alias analysis, but are instead assumed to be able to alias
6672 any other type of objects.
6673 In the context of section 6.5 paragraph 7 of the C99 standard,
6674 an lvalue expression
6675 dereferencing such a pointer is treated like having a character type.
6676 See @option{-fstrict-aliasing} for more information on aliasing issues.
6677 This extension exists to support some vector APIs, in which pointers to
6678 one vector type are permitted to alias pointers to a different vector type.
6680 Note that an object of a type with this attribute does not have any
6686 typedef short __attribute__((__may_alias__)) short_a;
6692 short_a *b = (short_a *) &a;
6696 if (a == 0x12345678)
6704 If you replaced @code{short_a} with @code{short} in the variable
6705 declaration, the above program would abort when compiled with
6706 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6710 @cindex @code{packed} type attribute
6711 This attribute, attached to @code{struct} or @code{union} type
6712 definition, specifies that each member (other than zero-width bit-fields)
6713 of the structure or union is placed to minimize the memory required. When
6714 attached to an @code{enum} definition, it indicates that the smallest
6715 integral type should be used.
6717 @opindex fshort-enums
6718 Specifying the @code{packed} attribute for @code{struct} and @code{union}
6719 types is equivalent to specifying the @code{packed} attribute on each
6720 of the structure or union members. Specifying the @option{-fshort-enums}
6721 flag on the command line is equivalent to specifying the @code{packed}
6722 attribute on all @code{enum} definitions.
6724 In the following example @code{struct my_packed_struct}'s members are
6725 packed closely together, but the internal layout of its @code{s} member
6726 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6730 struct my_unpacked_struct
6736 struct __attribute__ ((__packed__)) my_packed_struct
6740 struct my_unpacked_struct s;
6744 You may only specify the @code{packed} attribute attribute on the definition
6745 of an @code{enum}, @code{struct} or @code{union}, not on a @code{typedef}
6746 that does not also define the enumerated type, structure or union.
6748 @item scalar_storage_order ("@var{endianness}")
6749 @cindex @code{scalar_storage_order} type attribute
6750 When attached to a @code{union} or a @code{struct}, this attribute sets
6751 the storage order, aka endianness, of the scalar fields of the type, as
6752 well as the array fields whose component is scalar. The supported
6753 endiannesses are @code{big-endian} and @code{little-endian}. The attribute
6754 has no effects on fields which are themselves a @code{union}, a @code{struct}
6755 or an array whose component is a @code{union} or a @code{struct}, and it is
6756 possible for these fields to have a different scalar storage order than the
6759 This attribute is supported only for targets that use a uniform default
6760 scalar storage order (fortunately, most of them), i.e. targets that store
6761 the scalars either all in big-endian or all in little-endian.
6763 Additional restrictions are enforced for types with the reverse scalar
6764 storage order with regard to the scalar storage order of the target:
6767 @item Taking the address of a scalar field of a @code{union} or a
6768 @code{struct} with reverse scalar storage order is not permitted and yields
6770 @item Taking the address of an array field, whose component is scalar, of
6771 a @code{union} or a @code{struct} with reverse scalar storage order is
6772 permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
6774 @item Taking the address of a @code{union} or a @code{struct} with reverse
6775 scalar storage order is permitted.
6778 These restrictions exist because the storage order attribute is lost when
6779 the address of a scalar or the address of an array with scalar component is
6780 taken, so storing indirectly through this address generally does not work.
6781 The second case is nevertheless allowed to be able to perform a block copy
6782 from or to the array.
6784 Moreover, the use of type punning or aliasing to toggle the storage order
6785 is not supported; that is to say, a given scalar object cannot be accessed
6786 through distinct types that assign a different storage order to it.
6788 @item transparent_union
6789 @cindex @code{transparent_union} type attribute
6791 This attribute, attached to a @code{union} type definition, indicates
6792 that any function parameter having that union type causes calls to that
6793 function to be treated in a special way.
6795 First, the argument corresponding to a transparent union type can be of
6796 any type in the union; no cast is required. Also, if the union contains
6797 a pointer type, the corresponding argument can be a null pointer
6798 constant or a void pointer expression; and if the union contains a void
6799 pointer type, the corresponding argument can be any pointer expression.
6800 If the union member type is a pointer, qualifiers like @code{const} on
6801 the referenced type must be respected, just as with normal pointer
6804 Second, the argument is passed to the function using the calling
6805 conventions of the first member of the transparent union, not the calling
6806 conventions of the union itself. All members of the union must have the
6807 same machine representation; this is necessary for this argument passing
6810 Transparent unions are designed for library functions that have multiple
6811 interfaces for compatibility reasons. For example, suppose the
6812 @code{wait} function must accept either a value of type @code{int *} to
6813 comply with POSIX, or a value of type @code{union wait *} to comply with
6814 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6815 @code{wait} would accept both kinds of arguments, but it would also
6816 accept any other pointer type and this would make argument type checking
6817 less useful. Instead, @code{<sys/wait.h>} might define the interface
6821 typedef union __attribute__ ((__transparent_union__))
6825 @} wait_status_ptr_t;
6827 pid_t wait (wait_status_ptr_t);
6831 This interface allows either @code{int *} or @code{union wait *}
6832 arguments to be passed, using the @code{int *} calling convention.
6833 The program can call @code{wait} with arguments of either type:
6836 int w1 () @{ int w; return wait (&w); @}
6837 int w2 () @{ union wait w; return wait (&w); @}
6841 With this interface, @code{wait}'s implementation might look like this:
6844 pid_t wait (wait_status_ptr_t p)
6846 return waitpid (-1, p.__ip, 0);
6851 @cindex @code{unused} type attribute
6852 When attached to a type (including a @code{union} or a @code{struct}),
6853 this attribute means that variables of that type are meant to appear
6854 possibly unused. GCC does not produce a warning for any variables of
6855 that type, even if the variable appears to do nothing. This is often
6856 the case with lock or thread classes, which are usually defined and then
6857 not referenced, but contain constructors and destructors that have
6858 nontrivial bookkeeping functions.
6861 @cindex @code{visibility} type attribute
6862 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6863 applied to class, struct, union and enum types. Unlike other type
6864 attributes, the attribute must appear between the initial keyword and
6865 the name of the type; it cannot appear after the body of the type.
6867 Note that the type visibility is applied to vague linkage entities
6868 associated with the class (vtable, typeinfo node, etc.). In
6869 particular, if a class is thrown as an exception in one shared object
6870 and caught in another, the class must have default visibility.
6871 Otherwise the two shared objects are unable to use the same
6872 typeinfo node and exception handling will break.
6876 To specify multiple attributes, separate them by commas within the
6877 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6880 @node ARM Type Attributes
6881 @subsection ARM Type Attributes
6883 @cindex @code{notshared} type attribute, ARM
6884 On those ARM targets that support @code{dllimport} (such as Symbian
6885 OS), you can use the @code{notshared} attribute to indicate that the
6886 virtual table and other similar data for a class should not be
6887 exported from a DLL@. For example:
6890 class __declspec(notshared) C @{
6892 __declspec(dllimport) C();
6896 __declspec(dllexport)
6901 In this code, @code{C::C} is exported from the current DLL, but the
6902 virtual table for @code{C} is not exported. (You can use
6903 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6904 most Symbian OS code uses @code{__declspec}.)
6906 @node MeP Type Attributes
6907 @subsection MeP Type Attributes
6909 @cindex @code{based} type attribute, MeP
6910 @cindex @code{tiny} type attribute, MeP
6911 @cindex @code{near} type attribute, MeP
6912 @cindex @code{far} type attribute, MeP
6913 Many of the MeP variable attributes may be applied to types as well.
6914 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6915 @code{far} attributes may be applied to either. The @code{io} and
6916 @code{cb} attributes may not be applied to types.
6918 @node PowerPC Type Attributes
6919 @subsection PowerPC Type Attributes
6921 Three attributes currently are defined for PowerPC configurations:
6922 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6924 @cindex @code{ms_struct} type attribute, PowerPC
6925 @cindex @code{gcc_struct} type attribute, PowerPC
6926 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6927 attributes please see the documentation in @ref{x86 Type Attributes}.
6929 @cindex @code{altivec} type attribute, PowerPC
6930 The @code{altivec} attribute allows one to declare AltiVec vector data
6931 types supported by the AltiVec Programming Interface Manual. The
6932 attribute requires an argument to specify one of three vector types:
6933 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6934 and @code{bool__} (always followed by unsigned).
6937 __attribute__((altivec(vector__)))
6938 __attribute__((altivec(pixel__))) unsigned short
6939 __attribute__((altivec(bool__))) unsigned
6942 These attributes mainly are intended to support the @code{__vector},
6943 @code{__pixel}, and @code{__bool} AltiVec keywords.
6945 @node SPU Type Attributes
6946 @subsection SPU Type Attributes
6948 @cindex @code{spu_vector} type attribute, SPU
6949 The SPU supports the @code{spu_vector} attribute for types. This attribute
6950 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6951 Language Extensions Specification. It is intended to support the
6952 @code{__vector} keyword.
6954 @node x86 Type Attributes
6955 @subsection x86 Type Attributes
6957 Two attributes are currently defined for x86 configurations:
6958 @code{ms_struct} and @code{gcc_struct}.
6964 @cindex @code{ms_struct} type attribute, x86
6965 @cindex @code{gcc_struct} type attribute, x86
6967 If @code{packed} is used on a structure, or if bit-fields are used
6968 it may be that the Microsoft ABI packs them differently
6969 than GCC normally packs them. Particularly when moving packed
6970 data between functions compiled with GCC and the native Microsoft compiler
6971 (either via function call or as data in a file), it may be necessary to access
6974 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6975 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6976 command-line options, respectively;
6977 see @ref{x86 Options}, for details of how structure layout is affected.
6978 @xref{x86 Variable Attributes}, for information about the corresponding
6979 attributes on variables.
6983 @node Label Attributes
6984 @section Label Attributes
6985 @cindex Label Attributes
6987 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6988 details of the exact syntax for using attributes. Other attributes are
6989 available for functions (@pxref{Function Attributes}), variables
6990 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6991 statements (@pxref{Statement Attributes}), and for types
6992 (@pxref{Type Attributes}).
6994 This example uses the @code{cold} label attribute to indicate the
6995 @code{ErrorHandling} branch is unlikely to be taken and that the
6996 @code{ErrorHandling} label is unused:
7000 asm goto ("some asm" : : : : NoError);
7002 /* This branch (the fall-through from the asm) is less commonly used */
7004 __attribute__((cold, unused)); /* Semi-colon is required here */
7009 printf("no error\n");
7015 @cindex @code{unused} label attribute
7016 This feature is intended for program-generated code that may contain
7017 unused labels, but which is compiled with @option{-Wall}. It is
7018 not normally appropriate to use in it human-written code, though it
7019 could be useful in cases where the code that jumps to the label is
7020 contained within an @code{#ifdef} conditional.
7023 @cindex @code{hot} label attribute
7024 The @code{hot} attribute on a label is used to inform the compiler that
7025 the path following the label is more likely than paths that are not so
7026 annotated. This attribute is used in cases where @code{__builtin_expect}
7027 cannot be used, for instance with computed goto or @code{asm goto}.
7030 @cindex @code{cold} label attribute
7031 The @code{cold} attribute on labels is used to inform the compiler that
7032 the path following the label is unlikely to be executed. This attribute
7033 is used in cases where @code{__builtin_expect} cannot be used, for instance
7034 with computed goto or @code{asm goto}.
7038 @node Enumerator Attributes
7039 @section Enumerator Attributes
7040 @cindex Enumerator Attributes
7042 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
7043 details of the exact syntax for using attributes. Other attributes are
7044 available for functions (@pxref{Function Attributes}), variables
7045 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}), statements
7046 (@pxref{Statement Attributes}), and for types (@pxref{Type Attributes}).
7048 This example uses the @code{deprecated} enumerator attribute to indicate the
7049 @code{oldval} enumerator is deprecated:
7053 oldval __attribute__((deprecated)),
7066 @cindex @code{deprecated} enumerator attribute
7067 The @code{deprecated} attribute results in a warning if the enumerator
7068 is used anywhere in the source file. This is useful when identifying
7069 enumerators that are expected to be removed in a future version of a
7070 program. The warning also includes the location of the declaration
7071 of the deprecated enumerator, to enable users to easily find further
7072 information about why the enumerator is deprecated, or what they should
7073 do instead. Note that the warnings only occurs for uses.
7077 @node Statement Attributes
7078 @section Statement Attributes
7079 @cindex Statement Attributes
7081 GCC allows attributes to be set on null statements. @xref{Attribute Syntax},
7082 for details of the exact syntax for using attributes. Other attributes are
7083 available for functions (@pxref{Function Attributes}), variables
7084 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}), enumerators
7085 (@pxref{Enumerator Attributes}), and for types (@pxref{Type Attributes}).
7087 This example uses the @code{fallthrough} statement attribute to indicate that
7088 the @option{-Wimplicit-fallthrough} warning should not be emitted:
7095 __attribute__((fallthrough));
7103 @cindex @code{fallthrough} statement attribute
7104 The @code{fallthrough} attribute with a null statement serves as a
7105 fallthrough statement. It hints to the compiler that a statement
7106 that falls through to another case label, or user-defined label
7107 in a switch statement is intentional and thus the
7108 @option{-Wimplicit-fallthrough} warning must not trigger. The
7109 fallthrough attribute may appear at most once in each attribute
7110 list, and may not be mixed with other attributes. It can only
7111 be used in a switch statement (the compiler will issue an error
7112 otherwise), after a preceding statement and before a logically
7113 succeeding case label, or user-defined label.
7117 @node Attribute Syntax
7118 @section Attribute Syntax
7119 @cindex attribute syntax
7121 This section describes the syntax with which @code{__attribute__} may be
7122 used, and the constructs to which attribute specifiers bind, for the C
7123 language. Some details may vary for C++ and Objective-C@. Because of
7124 infelicities in the grammar for attributes, some forms described here
7125 may not be successfully parsed in all cases.
7127 There are some problems with the semantics of attributes in C++. For
7128 example, there are no manglings for attributes, although they may affect
7129 code generation, so problems may arise when attributed types are used in
7130 conjunction with templates or overloading. Similarly, @code{typeid}
7131 does not distinguish between types with different attributes. Support
7132 for attributes in C++ may be restricted in future to attributes on
7133 declarations only, but not on nested declarators.
7135 @xref{Function Attributes}, for details of the semantics of attributes
7136 applying to functions. @xref{Variable Attributes}, for details of the
7137 semantics of attributes applying to variables. @xref{Type Attributes},
7138 for details of the semantics of attributes applying to structure, union
7139 and enumerated types.
7140 @xref{Label Attributes}, for details of the semantics of attributes
7142 @xref{Enumerator Attributes}, for details of the semantics of attributes
7143 applying to enumerators.
7144 @xref{Statement Attributes}, for details of the semantics of attributes
7145 applying to statements.
7147 An @dfn{attribute specifier} is of the form
7148 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
7149 is a possibly empty comma-separated sequence of @dfn{attributes}, where
7150 each attribute is one of the following:
7154 Empty. Empty attributes are ignored.
7158 (which may be an identifier such as @code{unused}, or a reserved
7159 word such as @code{const}).
7162 An attribute name followed by a parenthesized list of
7163 parameters for the attribute.
7164 These parameters take one of the following forms:
7168 An identifier. For example, @code{mode} attributes use this form.
7171 An identifier followed by a comma and a non-empty comma-separated list
7172 of expressions. For example, @code{format} attributes use this form.
7175 A possibly empty comma-separated list of expressions. For example,
7176 @code{format_arg} attributes use this form with the list being a single
7177 integer constant expression, and @code{alias} attributes use this form
7178 with the list being a single string constant.
7182 An @dfn{attribute specifier list} is a sequence of one or more attribute
7183 specifiers, not separated by any other tokens.
7185 You may optionally specify attribute names with @samp{__}
7186 preceding and following the name.
7187 This allows you to use them in header files without
7188 being concerned about a possible macro of the same name. For example,
7189 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
7192 @subsubheading Label Attributes
7194 In GNU C, an attribute specifier list may appear after the colon following a
7195 label, other than a @code{case} or @code{default} label. GNU C++ only permits
7196 attributes on labels if the attribute specifier is immediately
7197 followed by a semicolon (i.e., the label applies to an empty
7198 statement). If the semicolon is missing, C++ label attributes are
7199 ambiguous, as it is permissible for a declaration, which could begin
7200 with an attribute list, to be labelled in C++. Declarations cannot be
7201 labelled in C90 or C99, so the ambiguity does not arise there.
7203 @subsubheading Enumerator Attributes
7205 In GNU C, an attribute specifier list may appear as part of an enumerator.
7206 The attribute goes after the enumeration constant, before @code{=}, if
7207 present. The optional attribute in the enumerator appertains to the
7208 enumeration constant. It is not possible to place the attribute after
7209 the constant expression, if present.
7211 @subsubheading Statement Attributes
7212 In GNU C, an attribute specifier list may appear as part of a null
7213 statement. The attribute goes before the semicolon.
7215 @subsubheading Type Attributes
7217 An attribute specifier list may appear as part of a @code{struct},
7218 @code{union} or @code{enum} specifier. It may go either immediately
7219 after the @code{struct}, @code{union} or @code{enum} keyword, or after
7220 the closing brace. The former syntax is preferred.
7221 Where attribute specifiers follow the closing brace, they are considered
7222 to relate to the structure, union or enumerated type defined, not to any
7223 enclosing declaration the type specifier appears in, and the type
7224 defined is not complete until after the attribute specifiers.
7225 @c Otherwise, there would be the following problems: a shift/reduce
7226 @c conflict between attributes binding the struct/union/enum and
7227 @c binding to the list of specifiers/qualifiers; and "aligned"
7228 @c attributes could use sizeof for the structure, but the size could be
7229 @c changed later by "packed" attributes.
7232 @subsubheading All other attributes
7234 Otherwise, an attribute specifier appears as part of a declaration,
7235 counting declarations of unnamed parameters and type names, and relates
7236 to that declaration (which may be nested in another declaration, for
7237 example in the case of a parameter declaration), or to a particular declarator
7238 within a declaration. Where an
7239 attribute specifier is applied to a parameter declared as a function or
7240 an array, it should apply to the function or array rather than the
7241 pointer to which the parameter is implicitly converted, but this is not
7242 yet correctly implemented.
7244 Any list of specifiers and qualifiers at the start of a declaration may
7245 contain attribute specifiers, whether or not such a list may in that
7246 context contain storage class specifiers. (Some attributes, however,
7247 are essentially in the nature of storage class specifiers, and only make
7248 sense where storage class specifiers may be used; for example,
7249 @code{section}.) There is one necessary limitation to this syntax: the
7250 first old-style parameter declaration in a function definition cannot
7251 begin with an attribute specifier, because such an attribute applies to
7252 the function instead by syntax described below (which, however, is not
7253 yet implemented in this case). In some other cases, attribute
7254 specifiers are permitted by this grammar but not yet supported by the
7255 compiler. All attribute specifiers in this place relate to the
7256 declaration as a whole. In the obsolescent usage where a type of
7257 @code{int} is implied by the absence of type specifiers, such a list of
7258 specifiers and qualifiers may be an attribute specifier list with no
7259 other specifiers or qualifiers.
7261 At present, the first parameter in a function prototype must have some
7262 type specifier that is not an attribute specifier; this resolves an
7263 ambiguity in the interpretation of @code{void f(int
7264 (__attribute__((foo)) x))}, but is subject to change. At present, if
7265 the parentheses of a function declarator contain only attributes then
7266 those attributes are ignored, rather than yielding an error or warning
7267 or implying a single parameter of type int, but this is subject to
7270 An attribute specifier list may appear immediately before a declarator
7271 (other than the first) in a comma-separated list of declarators in a
7272 declaration of more than one identifier using a single list of
7273 specifiers and qualifiers. Such attribute specifiers apply
7274 only to the identifier before whose declarator they appear. For
7278 __attribute__((noreturn)) void d0 (void),
7279 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
7284 the @code{noreturn} attribute applies to all the functions
7285 declared; the @code{format} attribute only applies to @code{d1}.
7287 An attribute specifier list may appear immediately before the comma,
7288 @code{=} or semicolon terminating the declaration of an identifier other
7289 than a function definition. Such attribute specifiers apply
7290 to the declared object or function. Where an
7291 assembler name for an object or function is specified (@pxref{Asm
7292 Labels}), the attribute must follow the @code{asm}
7295 An attribute specifier list may, in future, be permitted to appear after
7296 the declarator in a function definition (before any old-style parameter
7297 declarations or the function body).
7299 Attribute specifiers may be mixed with type qualifiers appearing inside
7300 the @code{[]} of a parameter array declarator, in the C99 construct by
7301 which such qualifiers are applied to the pointer to which the array is
7302 implicitly converted. Such attribute specifiers apply to the pointer,
7303 not to the array, but at present this is not implemented and they are
7306 An attribute specifier list may appear at the start of a nested
7307 declarator. At present, there are some limitations in this usage: the
7308 attributes correctly apply to the declarator, but for most individual
7309 attributes the semantics this implies are not implemented.
7310 When attribute specifiers follow the @code{*} of a pointer
7311 declarator, they may be mixed with any type qualifiers present.
7312 The following describes the formal semantics of this syntax. It makes the
7313 most sense if you are familiar with the formal specification of
7314 declarators in the ISO C standard.
7316 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
7317 D1}, where @code{T} contains declaration specifiers that specify a type
7318 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
7319 contains an identifier @var{ident}. The type specified for @var{ident}
7320 for derived declarators whose type does not include an attribute
7321 specifier is as in the ISO C standard.
7323 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
7324 and the declaration @code{T D} specifies the type
7325 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7326 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7327 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
7329 If @code{D1} has the form @code{*
7330 @var{type-qualifier-and-attribute-specifier-list} D}, and the
7331 declaration @code{T D} specifies the type
7332 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7333 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7334 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7340 void (__attribute__((noreturn)) ****f) (void);
7344 specifies the type ``pointer to pointer to pointer to pointer to
7345 non-returning function returning @code{void}''. As another example,
7348 char *__attribute__((aligned(8))) *f;
7352 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7353 Note again that this does not work with most attributes; for example,
7354 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7355 is not yet supported.
7357 For compatibility with existing code written for compiler versions that
7358 did not implement attributes on nested declarators, some laxity is
7359 allowed in the placing of attributes. If an attribute that only applies
7360 to types is applied to a declaration, it is treated as applying to
7361 the type of that declaration. If an attribute that only applies to
7362 declarations is applied to the type of a declaration, it is treated
7363 as applying to that declaration; and, for compatibility with code
7364 placing the attributes immediately before the identifier declared, such
7365 an attribute applied to a function return type is treated as
7366 applying to the function type, and such an attribute applied to an array
7367 element type is treated as applying to the array type. If an
7368 attribute that only applies to function types is applied to a
7369 pointer-to-function type, it is treated as applying to the pointer
7370 target type; if such an attribute is applied to a function return type
7371 that is not a pointer-to-function type, it is treated as applying
7372 to the function type.
7374 @node Function Prototypes
7375 @section Prototypes and Old-Style Function Definitions
7376 @cindex function prototype declarations
7377 @cindex old-style function definitions
7378 @cindex promotion of formal parameters
7380 GNU C extends ISO C to allow a function prototype to override a later
7381 old-style non-prototype definition. Consider the following example:
7384 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7391 /* @r{Prototype function declaration.} */
7392 int isroot P((uid_t));
7394 /* @r{Old-style function definition.} */
7396 isroot (x) /* @r{??? lossage here ???} */
7403 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7404 not allow this example, because subword arguments in old-style
7405 non-prototype definitions are promoted. Therefore in this example the
7406 function definition's argument is really an @code{int}, which does not
7407 match the prototype argument type of @code{short}.
7409 This restriction of ISO C makes it hard to write code that is portable
7410 to traditional C compilers, because the programmer does not know
7411 whether the @code{uid_t} type is @code{short}, @code{int}, or
7412 @code{long}. Therefore, in cases like these GNU C allows a prototype
7413 to override a later old-style definition. More precisely, in GNU C, a
7414 function prototype argument type overrides the argument type specified
7415 by a later old-style definition if the former type is the same as the
7416 latter type before promotion. Thus in GNU C the above example is
7417 equivalent to the following:
7430 GNU C++ does not support old-style function definitions, so this
7431 extension is irrelevant.
7434 @section C++ Style Comments
7436 @cindex C++ comments
7437 @cindex comments, C++ style
7439 In GNU C, you may use C++ style comments, which start with @samp{//} and
7440 continue until the end of the line. Many other C implementations allow
7441 such comments, and they are included in the 1999 C standard. However,
7442 C++ style comments are not recognized if you specify an @option{-std}
7443 option specifying a version of ISO C before C99, or @option{-ansi}
7444 (equivalent to @option{-std=c90}).
7447 @section Dollar Signs in Identifier Names
7449 @cindex dollar signs in identifier names
7450 @cindex identifier names, dollar signs in
7452 In GNU C, you may normally use dollar signs in identifier names.
7453 This is because many traditional C implementations allow such identifiers.
7454 However, dollar signs in identifiers are not supported on a few target
7455 machines, typically because the target assembler does not allow them.
7457 @node Character Escapes
7458 @section The Character @key{ESC} in Constants
7460 You can use the sequence @samp{\e} in a string or character constant to
7461 stand for the ASCII character @key{ESC}.
7464 @section Inquiring on Alignment of Types or Variables
7466 @cindex type alignment
7467 @cindex variable alignment
7469 The keyword @code{__alignof__} allows you to inquire about how an object
7470 is aligned, or the minimum alignment usually required by a type. Its
7471 syntax is just like @code{sizeof}.
7473 For example, if the target machine requires a @code{double} value to be
7474 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7475 This is true on many RISC machines. On more traditional machine
7476 designs, @code{__alignof__ (double)} is 4 or even 2.
7478 Some machines never actually require alignment; they allow reference to any
7479 data type even at an odd address. For these machines, @code{__alignof__}
7480 reports the smallest alignment that GCC gives the data type, usually as
7481 mandated by the target ABI.
7483 If the operand of @code{__alignof__} is an lvalue rather than a type,
7484 its value is the required alignment for its type, taking into account
7485 any minimum alignment specified with GCC's @code{__attribute__}
7486 extension (@pxref{Variable Attributes}). For example, after this
7490 struct foo @{ int x; char y; @} foo1;
7494 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7495 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7497 It is an error to ask for the alignment of an incomplete type.
7501 @section An Inline Function is As Fast As a Macro
7502 @cindex inline functions
7503 @cindex integrating function code
7505 @cindex macros, inline alternative
7507 By declaring a function inline, you can direct GCC to make
7508 calls to that function faster. One way GCC can achieve this is to
7509 integrate that function's code into the code for its callers. This
7510 makes execution faster by eliminating the function-call overhead; in
7511 addition, if any of the actual argument values are constant, their
7512 known values may permit simplifications at compile time so that not
7513 all of the inline function's code needs to be included. The effect on
7514 code size is less predictable; object code may be larger or smaller
7515 with function inlining, depending on the particular case. You can
7516 also direct GCC to try to integrate all ``simple enough'' functions
7517 into their callers with the option @option{-finline-functions}.
7519 GCC implements three different semantics of declaring a function
7520 inline. One is available with @option{-std=gnu89} or
7521 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7522 on all inline declarations, another when
7523 @option{-std=c99}, @option{-std=c11},
7524 @option{-std=gnu99} or @option{-std=gnu11}
7525 (without @option{-fgnu89-inline}), and the third
7526 is used when compiling C++.
7528 To declare a function inline, use the @code{inline} keyword in its
7529 declaration, like this:
7539 If you are writing a header file to be included in ISO C90 programs, write
7540 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7542 The three types of inlining behave similarly in two important cases:
7543 when the @code{inline} keyword is used on a @code{static} function,
7544 like the example above, and when a function is first declared without
7545 using the @code{inline} keyword and then is defined with
7546 @code{inline}, like this:
7549 extern int inc (int *a);
7557 In both of these common cases, the program behaves the same as if you
7558 had not used the @code{inline} keyword, except for its speed.
7560 @cindex inline functions, omission of
7561 @opindex fkeep-inline-functions
7562 When a function is both inline and @code{static}, if all calls to the
7563 function are integrated into the caller, and the function's address is
7564 never used, then the function's own assembler code is never referenced.
7565 In this case, GCC does not actually output assembler code for the
7566 function, unless you specify the option @option{-fkeep-inline-functions}.
7567 If there is a nonintegrated call, then the function is compiled to
7568 assembler code as usual. The function must also be compiled as usual if
7569 the program refers to its address, because that cannot be inlined.
7572 Note that certain usages in a function definition can make it unsuitable
7573 for inline substitution. Among these usages are: variadic functions,
7574 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7575 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7576 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7577 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7578 function marked @code{inline} could not be substituted, and gives the
7579 reason for the failure.
7581 @cindex automatic @code{inline} for C++ member fns
7582 @cindex @code{inline} automatic for C++ member fns
7583 @cindex member fns, automatically @code{inline}
7584 @cindex C++ member fns, automatically @code{inline}
7585 @opindex fno-default-inline
7586 As required by ISO C++, GCC considers member functions defined within
7587 the body of a class to be marked inline even if they are
7588 not explicitly declared with the @code{inline} keyword. You can
7589 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7590 Options,,Options Controlling C++ Dialect}.
7592 GCC does not inline any functions when not optimizing unless you specify
7593 the @samp{always_inline} attribute for the function, like this:
7596 /* @r{Prototype.} */
7597 inline void foo (const char) __attribute__((always_inline));
7600 The remainder of this section is specific to GNU C90 inlining.
7602 @cindex non-static inline function
7603 When an inline function is not @code{static}, then the compiler must assume
7604 that there may be calls from other source files; since a global symbol can
7605 be defined only once in any program, the function must not be defined in
7606 the other source files, so the calls therein cannot be integrated.
7607 Therefore, a non-@code{static} inline function is always compiled on its
7608 own in the usual fashion.
7610 If you specify both @code{inline} and @code{extern} in the function
7611 definition, then the definition is used only for inlining. In no case
7612 is the function compiled on its own, not even if you refer to its
7613 address explicitly. Such an address becomes an external reference, as
7614 if you had only declared the function, and had not defined it.
7616 This combination of @code{inline} and @code{extern} has almost the
7617 effect of a macro. The way to use it is to put a function definition in
7618 a header file with these keywords, and put another copy of the
7619 definition (lacking @code{inline} and @code{extern}) in a library file.
7620 The definition in the header file causes most calls to the function
7621 to be inlined. If any uses of the function remain, they refer to
7622 the single copy in the library.
7625 @section When is a Volatile Object Accessed?
7626 @cindex accessing volatiles
7627 @cindex volatile read
7628 @cindex volatile write
7629 @cindex volatile access
7631 C has the concept of volatile objects. These are normally accessed by
7632 pointers and used for accessing hardware or inter-thread
7633 communication. The standard encourages compilers to refrain from
7634 optimizations concerning accesses to volatile objects, but leaves it
7635 implementation defined as to what constitutes a volatile access. The
7636 minimum requirement is that at a sequence point all previous accesses
7637 to volatile objects have stabilized and no subsequent accesses have
7638 occurred. Thus an implementation is free to reorder and combine
7639 volatile accesses that occur between sequence points, but cannot do
7640 so for accesses across a sequence point. The use of volatile does
7641 not allow you to violate the restriction on updating objects multiple
7642 times between two sequence points.
7644 Accesses to non-volatile objects are not ordered with respect to
7645 volatile accesses. You cannot use a volatile object as a memory
7646 barrier to order a sequence of writes to non-volatile memory. For
7650 int *ptr = @var{something};
7652 *ptr = @var{something};
7657 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7658 that the write to @var{*ptr} occurs by the time the update
7659 of @var{vobj} happens. If you need this guarantee, you must use
7660 a stronger memory barrier such as:
7663 int *ptr = @var{something};
7665 *ptr = @var{something};
7666 asm volatile ("" : : : "memory");
7670 A scalar volatile object is read when it is accessed in a void context:
7673 volatile int *src = @var{somevalue};
7677 Such expressions are rvalues, and GCC implements this as a
7678 read of the volatile object being pointed to.
7680 Assignments are also expressions and have an rvalue. However when
7681 assigning to a scalar volatile, the volatile object is not reread,
7682 regardless of whether the assignment expression's rvalue is used or
7683 not. If the assignment's rvalue is used, the value is that assigned
7684 to the volatile object. For instance, there is no read of @var{vobj}
7685 in all the following cases:
7690 vobj = @var{something};
7691 obj = vobj = @var{something};
7692 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7693 obj = (@var{something}, vobj = @var{anotherthing});
7696 If you need to read the volatile object after an assignment has
7697 occurred, you must use a separate expression with an intervening
7700 As bit-fields are not individually addressable, volatile bit-fields may
7701 be implicitly read when written to, or when adjacent bit-fields are
7702 accessed. Bit-field operations may be optimized such that adjacent
7703 bit-fields are only partially accessed, if they straddle a storage unit
7704 boundary. For these reasons it is unwise to use volatile bit-fields to
7707 @node Using Assembly Language with C
7708 @section How to Use Inline Assembly Language in C Code
7709 @cindex @code{asm} keyword
7710 @cindex assembly language in C
7711 @cindex inline assembly language
7712 @cindex mixing assembly language and C
7714 The @code{asm} keyword allows you to embed assembler instructions
7715 within C code. GCC provides two forms of inline @code{asm}
7716 statements. A @dfn{basic @code{asm}} statement is one with no
7717 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7718 statement (@pxref{Extended Asm}) includes one or more operands.
7719 The extended form is preferred for mixing C and assembly language
7720 within a function, but to include assembly language at
7721 top level you must use basic @code{asm}.
7723 You can also use the @code{asm} keyword to override the assembler name
7724 for a C symbol, or to place a C variable in a specific register.
7727 * Basic Asm:: Inline assembler without operands.
7728 * Extended Asm:: Inline assembler with operands.
7729 * Constraints:: Constraints for @code{asm} operands
7730 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7731 * Explicit Register Variables:: Defining variables residing in specified
7733 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7737 @subsection Basic Asm --- Assembler Instructions Without Operands
7738 @cindex basic @code{asm}
7739 @cindex assembly language in C, basic
7741 A basic @code{asm} statement has the following syntax:
7744 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7747 The @code{asm} keyword is a GNU extension.
7748 When writing code that can be compiled with @option{-ansi} and the
7749 various @option{-std} options, use @code{__asm__} instead of
7750 @code{asm} (@pxref{Alternate Keywords}).
7752 @subsubheading Qualifiers
7755 The optional @code{volatile} qualifier has no effect.
7756 All basic @code{asm} blocks are implicitly volatile.
7759 @subsubheading Parameters
7762 @item AssemblerInstructions
7763 This is a literal string that specifies the assembler code. The string can
7764 contain any instructions recognized by the assembler, including directives.
7765 GCC does not parse the assembler instructions themselves and
7766 does not know what they mean or even whether they are valid assembler input.
7768 You may place multiple assembler instructions together in a single @code{asm}
7769 string, separated by the characters normally used in assembly code for the
7770 system. A combination that works in most places is a newline to break the
7771 line, plus a tab character (written as @samp{\n\t}).
7772 Some assemblers allow semicolons as a line separator. However,
7773 note that some assembler dialects use semicolons to start a comment.
7776 @subsubheading Remarks
7777 Using extended @code{asm} (@pxref{Extended Asm}) typically produces
7778 smaller, safer, and more efficient code, and in most cases it is a
7779 better solution than basic @code{asm}. However, there are two
7780 situations where only basic @code{asm} can be used:
7784 Extended @code{asm} statements have to be inside a C
7785 function, so to write inline assembly language at file scope (``top-level''),
7786 outside of C functions, you must use basic @code{asm}.
7787 You can use this technique to emit assembler directives,
7788 define assembly language macros that can be invoked elsewhere in the file,
7789 or write entire functions in assembly language.
7793 with the @code{naked} attribute also require basic @code{asm}
7794 (@pxref{Function Attributes}).
7797 Safely accessing C data and calling functions from basic @code{asm} is more
7798 complex than it may appear. To access C data, it is better to use extended
7801 Do not expect a sequence of @code{asm} statements to remain perfectly
7802 consecutive after compilation. If certain instructions need to remain
7803 consecutive in the output, put them in a single multi-instruction @code{asm}
7804 statement. Note that GCC's optimizers can move @code{asm} statements
7805 relative to other code, including across jumps.
7807 @code{asm} statements may not perform jumps into other @code{asm} statements.
7808 GCC does not know about these jumps, and therefore cannot take
7809 account of them when deciding how to optimize. Jumps from @code{asm} to C
7810 labels are only supported in extended @code{asm}.
7812 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7813 assembly code when optimizing. This can lead to unexpected duplicate
7814 symbol errors during compilation if your assembly code defines symbols or
7817 @strong{Warning:} The C standards do not specify semantics for @code{asm},
7818 making it a potential source of incompatibilities between compilers. These
7819 incompatibilities may not produce compiler warnings/errors.
7821 GCC does not parse basic @code{asm}'s @var{AssemblerInstructions}, which
7822 means there is no way to communicate to the compiler what is happening
7823 inside them. GCC has no visibility of symbols in the @code{asm} and may
7824 discard them as unreferenced. It also does not know about side effects of
7825 the assembler code, such as modifications to memory or registers. Unlike
7826 some compilers, GCC assumes that no changes to general purpose registers
7827 occur. This assumption may change in a future release.
7829 To avoid complications from future changes to the semantics and the
7830 compatibility issues between compilers, consider replacing basic @code{asm}
7831 with extended @code{asm}. See
7832 @uref{https://gcc.gnu.org/wiki/ConvertBasicAsmToExtended, How to convert
7833 from basic asm to extended asm} for information about how to perform this
7836 The compiler copies the assembler instructions in a basic @code{asm}
7837 verbatim to the assembly language output file, without
7838 processing dialects or any of the @samp{%} operators that are available with
7839 extended @code{asm}. This results in minor differences between basic
7840 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7841 registers you might use @samp{%eax} in basic @code{asm} and
7842 @samp{%%eax} in extended @code{asm}.
7844 On targets such as x86 that support multiple assembler dialects,
7845 all basic @code{asm} blocks use the assembler dialect specified by the
7846 @option{-masm} command-line option (@pxref{x86 Options}).
7847 Basic @code{asm} provides no
7848 mechanism to provide different assembler strings for different dialects.
7850 For basic @code{asm} with non-empty assembler string GCC assumes
7851 the assembler block does not change any general purpose registers,
7852 but it may read or write any globally accessible variable.
7854 Here is an example of basic @code{asm} for i386:
7857 /* Note that this code will not compile with -masm=intel */
7858 #define DebugBreak() asm("int $3")
7862 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7863 @cindex extended @code{asm}
7864 @cindex assembly language in C, extended
7866 With extended @code{asm} you can read and write C variables from
7867 assembler and perform jumps from assembler code to C labels.
7868 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7869 the operand parameters after the assembler template:
7872 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7873 : @var{OutputOperands}
7874 @r{[} : @var{InputOperands}
7875 @r{[} : @var{Clobbers} @r{]} @r{]})
7877 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7879 : @var{InputOperands}
7884 The @code{asm} keyword is a GNU extension.
7885 When writing code that can be compiled with @option{-ansi} and the
7886 various @option{-std} options, use @code{__asm__} instead of
7887 @code{asm} (@pxref{Alternate Keywords}).
7889 @subsubheading Qualifiers
7893 The typical use of extended @code{asm} statements is to manipulate input
7894 values to produce output values. However, your @code{asm} statements may
7895 also produce side effects. If so, you may need to use the @code{volatile}
7896 qualifier to disable certain optimizations. @xref{Volatile}.
7899 This qualifier informs the compiler that the @code{asm} statement may
7900 perform a jump to one of the labels listed in the @var{GotoLabels}.
7904 @subsubheading Parameters
7906 @item AssemblerTemplate
7907 This is a literal string that is the template for the assembler code. It is a
7908 combination of fixed text and tokens that refer to the input, output,
7909 and goto parameters. @xref{AssemblerTemplate}.
7911 @item OutputOperands
7912 A comma-separated list of the C variables modified by the instructions in the
7913 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7916 A comma-separated list of C expressions read by the instructions in the
7917 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7920 A comma-separated list of registers or other values changed by the
7921 @var{AssemblerTemplate}, beyond those listed as outputs.
7922 An empty list is permitted. @xref{Clobbers}.
7925 When you are using the @code{goto} form of @code{asm}, this section contains
7926 the list of all C labels to which the code in the
7927 @var{AssemblerTemplate} may jump.
7930 @code{asm} statements may not perform jumps into other @code{asm} statements,
7931 only to the listed @var{GotoLabels}.
7932 GCC's optimizers do not know about other jumps; therefore they cannot take
7933 account of them when deciding how to optimize.
7936 The total number of input + output + goto operands is limited to 30.
7938 @subsubheading Remarks
7939 The @code{asm} statement allows you to include assembly instructions directly
7940 within C code. This may help you to maximize performance in time-sensitive
7941 code or to access assembly instructions that are not readily available to C
7944 Note that extended @code{asm} statements must be inside a function. Only
7945 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7946 Functions declared with the @code{naked} attribute also require basic
7947 @code{asm} (@pxref{Function Attributes}).
7949 While the uses of @code{asm} are many and varied, it may help to think of an
7950 @code{asm} statement as a series of low-level instructions that convert input
7951 parameters to output parameters. So a simple (if not particularly useful)
7952 example for i386 using @code{asm} might look like this:
7958 asm ("mov %1, %0\n\t"
7963 printf("%d\n", dst);
7966 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7969 @subsubsection Volatile
7970 @cindex volatile @code{asm}
7971 @cindex @code{asm} volatile
7973 GCC's optimizers sometimes discard @code{asm} statements if they determine
7974 there is no need for the output variables. Also, the optimizers may move
7975 code out of loops if they believe that the code will always return the same
7976 result (i.e. none of its input values change between calls). Using the
7977 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7978 that have no output operands, including @code{asm goto} statements,
7979 are implicitly volatile.
7981 This i386 code demonstrates a case that does not use (or require) the
7982 @code{volatile} qualifier. If it is performing assertion checking, this code
7983 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7984 unreferenced by any code. As a result, the optimizers can discard the
7985 @code{asm} statement, which in turn removes the need for the entire
7986 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7987 isn't needed you allow the optimizers to produce the most efficient code
7991 void DoCheck(uint32_t dwSomeValue)
7995 // Assumes dwSomeValue is not zero.
8005 The next example shows a case where the optimizers can recognize that the input
8006 (@code{dwSomeValue}) never changes during the execution of the function and can
8007 therefore move the @code{asm} outside the loop to produce more efficient code.
8008 Again, using @code{volatile} disables this type of optimization.
8011 void do_print(uint32_t dwSomeValue)
8015 for (uint32_t x=0; x < 5; x++)
8017 // Assumes dwSomeValue is not zero.
8023 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
8028 The following example demonstrates a case where you need to use the
8029 @code{volatile} qualifier.
8030 It uses the x86 @code{rdtsc} instruction, which reads
8031 the computer's time-stamp counter. Without the @code{volatile} qualifier,
8032 the optimizers might assume that the @code{asm} block will always return the
8033 same value and therefore optimize away the second call.
8038 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
8039 "shl $32, %%rdx\n\t" // Shift the upper bits left.
8040 "or %%rdx, %0" // 'Or' in the lower bits.
8045 printf("msr: %llx\n", msr);
8049 // Reprint the timestamp
8050 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
8051 "shl $32, %%rdx\n\t" // Shift the upper bits left.
8052 "or %%rdx, %0" // 'Or' in the lower bits.
8057 printf("msr: %llx\n", msr);
8060 GCC's optimizers do not treat this code like the non-volatile code in the
8061 earlier examples. They do not move it out of loops or omit it on the
8062 assumption that the result from a previous call is still valid.
8064 Note that the compiler can move even volatile @code{asm} instructions relative
8065 to other code, including across jump instructions. For example, on many
8066 targets there is a system register that controls the rounding mode of
8067 floating-point operations. Setting it with a volatile @code{asm}, as in the
8068 following PowerPC example, does not work reliably.
8071 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
8075 The compiler may move the addition back before the volatile @code{asm}. To
8076 make it work as expected, add an artificial dependency to the @code{asm} by
8077 referencing a variable in the subsequent code, for example:
8080 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
8084 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
8085 assembly code when optimizing. This can lead to unexpected duplicate symbol
8086 errors during compilation if your asm code defines symbols or labels.
8088 (@pxref{AssemblerTemplate}) may help resolve this problem.
8090 @anchor{AssemblerTemplate}
8091 @subsubsection Assembler Template
8092 @cindex @code{asm} assembler template
8094 An assembler template is a literal string containing assembler instructions.
8095 The compiler replaces tokens in the template that refer
8096 to inputs, outputs, and goto labels,
8097 and then outputs the resulting string to the assembler. The
8098 string can contain any instructions recognized by the assembler, including
8099 directives. GCC does not parse the assembler instructions
8100 themselves and does not know what they mean or even whether they are valid
8101 assembler input. However, it does count the statements
8102 (@pxref{Size of an asm}).
8104 You may place multiple assembler instructions together in a single @code{asm}
8105 string, separated by the characters normally used in assembly code for the
8106 system. A combination that works in most places is a newline to break the
8107 line, plus a tab character to move to the instruction field (written as
8109 Some assemblers allow semicolons as a line separator. However, note
8110 that some assembler dialects use semicolons to start a comment.
8112 Do not expect a sequence of @code{asm} statements to remain perfectly
8113 consecutive after compilation, even when you are using the @code{volatile}
8114 qualifier. If certain instructions need to remain consecutive in the output,
8115 put them in a single multi-instruction asm statement.
8117 Accessing data from C programs without using input/output operands (such as
8118 by using global symbols directly from the assembler template) may not work as
8119 expected. Similarly, calling functions directly from an assembler template
8120 requires a detailed understanding of the target assembler and ABI.
8122 Since GCC does not parse the assembler template,
8123 it has no visibility of any
8124 symbols it references. This may result in GCC discarding those symbols as
8125 unreferenced unless they are also listed as input, output, or goto operands.
8127 @subsubheading Special format strings
8129 In addition to the tokens described by the input, output, and goto operands,
8130 these tokens have special meanings in the assembler template:
8134 Outputs a single @samp{%} into the assembler code.
8137 Outputs a number that is unique to each instance of the @code{asm}
8138 statement in the entire compilation. This option is useful when creating local
8139 labels and referring to them multiple times in a single template that
8140 generates multiple assembler instructions.
8145 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
8146 into the assembler code. When unescaped, these characters have special
8147 meaning to indicate multiple assembler dialects, as described below.
8150 @subsubheading Multiple assembler dialects in @code{asm} templates
8152 On targets such as x86, GCC supports multiple assembler dialects.
8153 The @option{-masm} option controls which dialect GCC uses as its
8154 default for inline assembler. The target-specific documentation for the
8155 @option{-masm} option contains the list of supported dialects, as well as the
8156 default dialect if the option is not specified. This information may be
8157 important to understand, since assembler code that works correctly when
8158 compiled using one dialect will likely fail if compiled using another.
8161 If your code needs to support multiple assembler dialects (for example, if
8162 you are writing public headers that need to support a variety of compilation
8163 options), use constructs of this form:
8166 @{ dialect0 | dialect1 | dialect2... @}
8169 This construct outputs @code{dialect0}
8170 when using dialect #0 to compile the code,
8171 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
8172 braces than the number of dialects the compiler supports, the construct
8175 For example, if an x86 compiler supports two dialects
8176 (@samp{att}, @samp{intel}), an
8177 assembler template such as this:
8180 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
8184 is equivalent to one of
8187 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
8188 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
8191 Using that same compiler, this code:
8194 "xchg@{l@}\t@{%%@}ebx, %1"
8198 corresponds to either
8201 "xchgl\t%%ebx, %1" @r{/* att dialect */}
8202 "xchg\tebx, %1" @r{/* intel dialect */}
8205 There is no support for nesting dialect alternatives.
8207 @anchor{OutputOperands}
8208 @subsubsection Output Operands
8209 @cindex @code{asm} output operands
8211 An @code{asm} statement has zero or more output operands indicating the names
8212 of C variables modified by the assembler code.
8214 In this i386 example, @code{old} (referred to in the template string as
8215 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
8216 (@code{%2}) is an input:
8221 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
8222 "sbb %0,%0" // Use the CF to calculate old.
8223 : "=r" (old), "+rm" (*Base)
8230 Operands are separated by commas. Each operand has this format:
8233 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
8237 @item asmSymbolicName
8238 Specifies a symbolic name for the operand.
8239 Reference the name in the assembler template
8240 by enclosing it in square brackets
8241 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8242 that contains the definition. Any valid C variable name is acceptable,
8243 including names already defined in the surrounding code. No two operands
8244 within the same @code{asm} statement can use the same symbolic name.
8246 When not using an @var{asmSymbolicName}, use the (zero-based) position
8248 in the list of operands in the assembler template. For example if there are
8249 three output operands, use @samp{%0} in the template to refer to the first,
8250 @samp{%1} for the second, and @samp{%2} for the third.
8253 A string constant specifying constraints on the placement of the operand;
8254 @xref{Constraints}, for details.
8256 Output constraints must begin with either @samp{=} (a variable overwriting an
8257 existing value) or @samp{+} (when reading and writing). When using
8258 @samp{=}, do not assume the location contains the existing value
8259 on entry to the @code{asm}, except
8260 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
8262 After the prefix, there must be one or more additional constraints
8263 (@pxref{Constraints}) that describe where the value resides. Common
8264 constraints include @samp{r} for register and @samp{m} for memory.
8265 When you list more than one possible location (for example, @code{"=rm"}),
8266 the compiler chooses the most efficient one based on the current context.
8267 If you list as many alternates as the @code{asm} statement allows, you permit
8268 the optimizers to produce the best possible code.
8269 If you must use a specific register, but your Machine Constraints do not
8270 provide sufficient control to select the specific register you want,
8271 local register variables may provide a solution (@pxref{Local Register
8275 Specifies a C lvalue expression to hold the output, typically a variable name.
8276 The enclosing parentheses are a required part of the syntax.
8280 When the compiler selects the registers to use to
8281 represent the output operands, it does not use any of the clobbered registers
8284 Output operand expressions must be lvalues. The compiler cannot check whether
8285 the operands have data types that are reasonable for the instruction being
8286 executed. For output expressions that are not directly addressable (for
8287 example a bit-field), the constraint must allow a register. In that case, GCC
8288 uses the register as the output of the @code{asm}, and then stores that
8289 register into the output.
8291 Operands using the @samp{+} constraint modifier count as two operands
8292 (that is, both as input and output) towards the total maximum of 30 operands
8293 per @code{asm} statement.
8295 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
8296 operands that must not overlap an input. Otherwise,
8297 GCC may allocate the output operand in the same register as an unrelated
8298 input operand, on the assumption that the assembler code consumes its
8299 inputs before producing outputs. This assumption may be false if the assembler
8300 code actually consists of more than one instruction.
8302 The same problem can occur if one output parameter (@var{a}) allows a register
8303 constraint and another output parameter (@var{b}) allows a memory constraint.
8304 The code generated by GCC to access the memory address in @var{b} can contain
8305 registers which @emph{might} be shared by @var{a}, and GCC considers those
8306 registers to be inputs to the asm. As above, GCC assumes that such input
8307 registers are consumed before any outputs are written. This assumption may
8308 result in incorrect behavior if the asm writes to @var{a} before using
8309 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
8310 ensures that modifying @var{a} does not affect the address referenced by
8311 @var{b}. Otherwise, the location of @var{b}
8312 is undefined if @var{a} is modified before using @var{b}.
8314 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8315 instead of simply @samp{%2}). Typically these qualifiers are hardware
8316 dependent. The list of supported modifiers for x86 is found at
8317 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8319 If the C code that follows the @code{asm} makes no use of any of the output
8320 operands, use @code{volatile} for the @code{asm} statement to prevent the
8321 optimizers from discarding the @code{asm} statement as unneeded
8322 (see @ref{Volatile}).
8324 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
8325 references the first output operand as @code{%0} (were there a second, it
8326 would be @code{%1}, etc). The number of the first input operand is one greater
8327 than that of the last output operand. In this i386 example, that makes
8328 @code{Mask} referenced as @code{%1}:
8331 uint32_t Mask = 1234;
8340 That code overwrites the variable @code{Index} (@samp{=}),
8341 placing the value in a register (@samp{r}).
8342 Using the generic @samp{r} constraint instead of a constraint for a specific
8343 register allows the compiler to pick the register to use, which can result
8344 in more efficient code. This may not be possible if an assembler instruction
8345 requires a specific register.
8347 The following i386 example uses the @var{asmSymbolicName} syntax.
8349 same result as the code above, but some may consider it more readable or more
8350 maintainable since reordering index numbers is not necessary when adding or
8351 removing operands. The names @code{aIndex} and @code{aMask}
8352 are only used in this example to emphasize which
8353 names get used where.
8354 It is acceptable to reuse the names @code{Index} and @code{Mask}.
8357 uint32_t Mask = 1234;
8360 asm ("bsfl %[aMask], %[aIndex]"
8361 : [aIndex] "=r" (Index)
8362 : [aMask] "r" (Mask)
8366 Here are some more examples of output operands.
8373 asm ("mov %[e], %[d]"
8378 Here, @code{d} may either be in a register or in memory. Since the compiler
8379 might already have the current value of the @code{uint32_t} location
8380 pointed to by @code{e}
8381 in a register, you can enable it to choose the best location
8382 for @code{d} by specifying both constraints.
8384 @anchor{FlagOutputOperands}
8385 @subsubsection Flag Output Operands
8386 @cindex @code{asm} flag output operands
8388 Some targets have a special register that holds the ``flags'' for the
8389 result of an operation or comparison. Normally, the contents of that
8390 register are either unmodifed by the asm, or the asm is considered to
8391 clobber the contents.
8393 On some targets, a special form of output operand exists by which
8394 conditions in the flags register may be outputs of the asm. The set of
8395 conditions supported are target specific, but the general rule is that
8396 the output variable must be a scalar integer, and the value is boolean.
8397 When supported, the target defines the preprocessor symbol
8398 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8400 Because of the special nature of the flag output operands, the constraint
8401 may not include alternatives.
8403 Most often, the target has only one flags register, and thus is an implied
8404 operand of many instructions. In this case, the operand should not be
8405 referenced within the assembler template via @code{%0} etc, as there's
8406 no corresponding text in the assembly language.
8410 The flag output constraints for the x86 family are of the form
8411 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8412 conditions defined in the ISA manual for @code{j@var{cc}} or
8417 ``above'' or unsigned greater than
8419 ``above or equal'' or unsigned greater than or equal
8421 ``below'' or unsigned less than
8423 ``below or equal'' or unsigned less than or equal
8428 ``equal'' or zero flag set
8432 signed greater than or equal
8436 signed less than or equal
8457 ``not'' @var{flag}, or inverted versions of those above
8462 @anchor{InputOperands}
8463 @subsubsection Input Operands
8464 @cindex @code{asm} input operands
8465 @cindex @code{asm} expressions
8467 Input operands make values from C variables and expressions available to the
8470 Operands are separated by commas. Each operand has this format:
8473 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8477 @item asmSymbolicName
8478 Specifies a symbolic name for the operand.
8479 Reference the name in the assembler template
8480 by enclosing it in square brackets
8481 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8482 that contains the definition. Any valid C variable name is acceptable,
8483 including names already defined in the surrounding code. No two operands
8484 within the same @code{asm} statement can use the same symbolic name.
8486 When not using an @var{asmSymbolicName}, use the (zero-based) position
8488 in the list of operands in the assembler template. For example if there are
8489 two output operands and three inputs,
8490 use @samp{%2} in the template to refer to the first input operand,
8491 @samp{%3} for the second, and @samp{%4} for the third.
8494 A string constant specifying constraints on the placement of the operand;
8495 @xref{Constraints}, for details.
8497 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8498 When you list more than one possible location (for example, @samp{"irm"}),
8499 the compiler chooses the most efficient one based on the current context.
8500 If you must use a specific register, but your Machine Constraints do not
8501 provide sufficient control to select the specific register you want,
8502 local register variables may provide a solution (@pxref{Local Register
8505 Input constraints can also be digits (for example, @code{"0"}). This indicates
8506 that the specified input must be in the same place as the output constraint
8507 at the (zero-based) index in the output constraint list.
8508 When using @var{asmSymbolicName} syntax for the output operands,
8509 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8512 This is the C variable or expression being passed to the @code{asm} statement
8513 as input. The enclosing parentheses are a required part of the syntax.
8517 When the compiler selects the registers to use to represent the input
8518 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8520 If there are no output operands but there are input operands, place two
8521 consecutive colons where the output operands would go:
8524 __asm__ ("some instructions"
8526 : "r" (Offset / 8));
8529 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8530 (except for inputs tied to outputs). The compiler assumes that on exit from
8531 the @code{asm} statement these operands contain the same values as they
8532 had before executing the statement.
8533 It is @emph{not} possible to use clobbers
8534 to inform the compiler that the values in these inputs are changing. One
8535 common work-around is to tie the changing input variable to an output variable
8536 that never gets used. Note, however, that if the code that follows the
8537 @code{asm} statement makes no use of any of the output operands, the GCC
8538 optimizers may discard the @code{asm} statement as unneeded
8539 (see @ref{Volatile}).
8541 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8542 instead of simply @samp{%2}). Typically these qualifiers are hardware
8543 dependent. The list of supported modifiers for x86 is found at
8544 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8546 In this example using the fictitious @code{combine} instruction, the
8547 constraint @code{"0"} for input operand 1 says that it must occupy the same
8548 location as output operand 0. Only input operands may use numbers in
8549 constraints, and they must each refer to an output operand. Only a number (or
8550 the symbolic assembler name) in the constraint can guarantee that one operand
8551 is in the same place as another. The mere fact that @code{foo} is the value of
8552 both operands is not enough to guarantee that they are in the same place in
8553 the generated assembler code.
8556 asm ("combine %2, %0"
8558 : "0" (foo), "g" (bar));
8561 Here is an example using symbolic names.
8564 asm ("cmoveq %1, %2, %[result]"
8565 : [result] "=r"(result)
8566 : "r" (test), "r" (new), "[result]" (old));
8570 @subsubsection Clobbers
8571 @cindex @code{asm} clobbers
8573 While the compiler is aware of changes to entries listed in the output
8574 operands, the inline @code{asm} code may modify more than just the outputs. For
8575 example, calculations may require additional registers, or the processor may
8576 overwrite a register as a side effect of a particular assembler instruction.
8577 In order to inform the compiler of these changes, list them in the clobber
8578 list. Clobber list items are either register names or the special clobbers
8579 (listed below). Each clobber list item is a string constant
8580 enclosed in double quotes and separated by commas.
8582 Clobber descriptions may not in any way overlap with an input or output
8583 operand. For example, you may not have an operand describing a register class
8584 with one member when listing that register in the clobber list. Variables
8585 declared to live in specific registers (@pxref{Explicit Register
8586 Variables}) and used
8587 as @code{asm} input or output operands must have no part mentioned in the
8588 clobber description. In particular, there is no way to specify that input
8589 operands get modified without also specifying them as output operands.
8591 When the compiler selects which registers to use to represent input and output
8592 operands, it does not use any of the clobbered registers. As a result,
8593 clobbered registers are available for any use in the assembler code.
8595 Here is a realistic example for the VAX showing the use of clobbered
8599 asm volatile ("movc3 %0, %1, %2"
8601 : "g" (from), "g" (to), "g" (count)
8602 : "r0", "r1", "r2", "r3", "r4", "r5");
8605 Also, there are two special clobber arguments:
8609 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8610 register. On some machines, GCC represents the condition codes as a specific
8611 hardware register; @code{"cc"} serves to name this register.
8612 On other machines, condition code handling is different,
8613 and specifying @code{"cc"} has no effect. But
8614 it is valid no matter what the target.
8617 The @code{"memory"} clobber tells the compiler that the assembly code
8619 reads or writes to items other than those listed in the input and output
8620 operands (for example, accessing the memory pointed to by one of the input
8621 parameters). To ensure memory contains correct values, GCC may need to flush
8622 specific register values to memory before executing the @code{asm}. Further,
8623 the compiler does not assume that any values read from memory before an
8624 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8626 Using the @code{"memory"} clobber effectively forms a read/write
8627 memory barrier for the compiler.
8629 Note that this clobber does not prevent the @emph{processor} from doing
8630 speculative reads past the @code{asm} statement. To prevent that, you need
8631 processor-specific fence instructions.
8633 Flushing registers to memory has performance implications and may be an issue
8634 for time-sensitive code. You can use a trick to avoid this if the size of
8635 the memory being accessed is known at compile time. For example, if accessing
8636 ten bytes of a string, use a memory input like:
8638 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8643 @subsubsection Goto Labels
8644 @cindex @code{asm} goto labels
8646 @code{asm goto} allows assembly code to jump to one or more C labels. The
8647 @var{GotoLabels} section in an @code{asm goto} statement contains
8649 list of all C labels to which the assembler code may jump. GCC assumes that
8650 @code{asm} execution falls through to the next statement (if this is not the
8651 case, consider using the @code{__builtin_unreachable} intrinsic after the
8652 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8653 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8656 An @code{asm goto} statement cannot have outputs.
8657 This is due to an internal restriction of
8658 the compiler: control transfer instructions cannot have outputs.
8659 If the assembler code does modify anything, use the @code{"memory"} clobber
8661 optimizers to flush all register values to memory and reload them if
8662 necessary after the @code{asm} statement.
8664 Also note that an @code{asm goto} statement is always implicitly
8665 considered volatile.
8667 To reference a label in the assembler template,
8668 prefix it with @samp{%l} (lowercase @samp{L}) followed
8669 by its (zero-based) position in @var{GotoLabels} plus the number of input
8670 operands. For example, if the @code{asm} has three inputs and references two
8671 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8673 Alternately, you can reference labels using the actual C label name enclosed
8674 in brackets. For example, to reference a label named @code{carry}, you can
8675 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8676 section when using this approach.
8678 Here is an example of @code{asm goto} for i386:
8685 : "r" (p1), "r" (p2)
8695 The following example shows an @code{asm goto} that uses a memory clobber.
8701 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8712 @anchor{x86Operandmodifiers}
8713 @subsubsection x86 Operand Modifiers
8715 References to input, output, and goto operands in the assembler template
8716 of extended @code{asm} statements can use
8717 modifiers to affect the way the operands are formatted in
8718 the code output to the assembler. For example, the
8719 following code uses the @samp{h} and @samp{b} modifiers for x86:
8723 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8727 These modifiers generate this assembler code:
8733 The rest of this discussion uses the following code for illustrative purposes.
8742 asm volatile goto ("some assembler instructions here"
8744 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8745 : /* No clobbers. */
8750 With no modifiers, this is what the output from the operands would be for the
8751 @samp{att} and @samp{intel} dialects of assembler:
8753 @multitable {Operand} {$.L2} {OFFSET FLAT:.L2}
8754 @headitem Operand @tab @samp{att} @tab @samp{intel}
8763 @tab @code{OFFSET FLAT:.L2}
8766 The table below shows the list of supported modifiers and their effects.
8768 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {@samp{att}} {@samp{intel}}
8769 @headitem Modifier @tab Description @tab Operand @tab @samp{att} @tab @samp{intel}
8771 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8776 @tab Print the QImode name of the register.
8781 @tab Print the QImode name for a ``high'' register.
8786 @tab Print the HImode name of the register.
8791 @tab Print the SImode name of the register.
8796 @tab Print the DImode name of the register.
8801 @tab Print the label name with no punctuation.
8806 @tab Require a constant operand and print the constant expression with no punctuation.
8812 @anchor{x86floatingpointasmoperands}
8813 @subsubsection x86 Floating-Point @code{asm} Operands
8815 On x86 targets, there are several rules on the usage of stack-like registers
8816 in the operands of an @code{asm}. These rules apply only to the operands
8817 that are stack-like registers:
8821 Given a set of input registers that die in an @code{asm}, it is
8822 necessary to know which are implicitly popped by the @code{asm}, and
8823 which must be explicitly popped by GCC@.
8825 An input register that is implicitly popped by the @code{asm} must be
8826 explicitly clobbered, unless it is constrained to match an
8830 For any input register that is implicitly popped by an @code{asm}, it is
8831 necessary to know how to adjust the stack to compensate for the pop.
8832 If any non-popped input is closer to the top of the reg-stack than
8833 the implicitly popped register, it would not be possible to know what the
8834 stack looked like---it's not clear how the rest of the stack ``slides
8837 All implicitly popped input registers must be closer to the top of
8838 the reg-stack than any input that is not implicitly popped.
8840 It is possible that if an input dies in an @code{asm}, the compiler might
8841 use the input register for an output reload. Consider this example:
8844 asm ("foo" : "=t" (a) : "f" (b));
8848 This code says that input @code{b} is not popped by the @code{asm}, and that
8849 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8850 deeper after the @code{asm} than it was before. But, it is possible that
8851 reload may think that it can use the same register for both the input and
8854 To prevent this from happening,
8855 if any input operand uses the @samp{f} constraint, all output register
8856 constraints must use the @samp{&} early-clobber modifier.
8858 The example above is correctly written as:
8861 asm ("foo" : "=&t" (a) : "f" (b));
8865 Some operands need to be in particular places on the stack. All
8866 output operands fall in this category---GCC has no other way to
8867 know which registers the outputs appear in unless you indicate
8868 this in the constraints.
8870 Output operands must specifically indicate which register an output
8871 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8872 constraints must select a class with a single register.
8875 Output operands may not be ``inserted'' between existing stack registers.
8876 Since no 387 opcode uses a read/write operand, all output operands
8877 are dead before the @code{asm}, and are pushed by the @code{asm}.
8878 It makes no sense to push anywhere but the top of the reg-stack.
8880 Output operands must start at the top of the reg-stack: output
8881 operands may not ``skip'' a register.
8884 Some @code{asm} statements may need extra stack space for internal
8885 calculations. This can be guaranteed by clobbering stack registers
8886 unrelated to the inputs and outputs.
8891 takes one input, which is internally popped, and produces two outputs.
8894 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8898 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8899 and replaces them with one output. The @code{st(1)} clobber is necessary
8900 for the compiler to know that @code{fyl2xp1} pops both inputs.
8903 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8911 @subsection Controlling Names Used in Assembler Code
8912 @cindex assembler names for identifiers
8913 @cindex names used in assembler code
8914 @cindex identifiers, names in assembler code
8916 You can specify the name to be used in the assembler code for a C
8917 function or variable by writing the @code{asm} (or @code{__asm__})
8918 keyword after the declarator.
8919 It is up to you to make sure that the assembler names you choose do not
8920 conflict with any other assembler symbols, or reference registers.
8922 @subsubheading Assembler names for data:
8924 This sample shows how to specify the assembler name for data:
8927 int foo asm ("myfoo") = 2;
8931 This specifies that the name to be used for the variable @code{foo} in
8932 the assembler code should be @samp{myfoo} rather than the usual
8935 On systems where an underscore is normally prepended to the name of a C
8936 variable, this feature allows you to define names for the
8937 linker that do not start with an underscore.
8939 GCC does not support using this feature with a non-static local variable
8940 since such variables do not have assembler names. If you are
8941 trying to put the variable in a particular register, see
8942 @ref{Explicit Register Variables}.
8944 @subsubheading Assembler names for functions:
8946 To specify the assembler name for functions, write a declaration for the
8947 function before its definition and put @code{asm} there, like this:
8950 int func (int x, int y) asm ("MYFUNC");
8952 int func (int x, int y)
8958 This specifies that the name to be used for the function @code{func} in
8959 the assembler code should be @code{MYFUNC}.
8961 @node Explicit Register Variables
8962 @subsection Variables in Specified Registers
8963 @anchor{Explicit Reg Vars}
8964 @cindex explicit register variables
8965 @cindex variables in specified registers
8966 @cindex specified registers
8968 GNU C allows you to associate specific hardware registers with C
8969 variables. In almost all cases, allowing the compiler to assign
8970 registers produces the best code. However under certain unusual
8971 circumstances, more precise control over the variable storage is
8974 Both global and local variables can be associated with a register. The
8975 consequences of performing this association are very different between
8976 the two, as explained in the sections below.
8979 * Global Register Variables:: Variables declared at global scope.
8980 * Local Register Variables:: Variables declared within a function.
8983 @node Global Register Variables
8984 @subsubsection Defining Global Register Variables
8985 @anchor{Global Reg Vars}
8986 @cindex global register variables
8987 @cindex registers, global variables in
8988 @cindex registers, global allocation
8990 You can define a global register variable and associate it with a specified
8994 register int *foo asm ("r12");
8998 Here @code{r12} is the name of the register that should be used. Note that
8999 this is the same syntax used for defining local register variables, but for
9000 a global variable the declaration appears outside a function. The
9001 @code{register} keyword is required, and cannot be combined with
9002 @code{static}. The register name must be a valid register name for the
9005 Registers are a scarce resource on most systems and allowing the
9006 compiler to manage their usage usually results in the best code. However,
9007 under special circumstances it can make sense to reserve some globally.
9008 For example this may be useful in programs such as programming language
9009 interpreters that have a couple of global variables that are accessed
9012 After defining a global register variable, for the current compilation
9016 @item The register is reserved entirely for this use, and will not be
9017 allocated for any other purpose.
9018 @item The register is not saved and restored by any functions.
9019 @item Stores into this register are never deleted even if they appear to be
9020 dead, but references may be deleted, moved or simplified.
9023 Note that these points @emph{only} apply to code that is compiled with the
9024 definition. The behavior of code that is merely linked in (for example
9025 code from libraries) is not affected.
9027 If you want to recompile source files that do not actually use your global
9028 register variable so they do not use the specified register for any other
9029 purpose, you need not actually add the global register declaration to
9030 their source code. It suffices to specify the compiler option
9031 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
9034 @subsubheading Declaring the variable
9036 Global register variables can not have initial values, because an
9037 executable file has no means to supply initial contents for a register.
9039 When selecting a register, choose one that is normally saved and
9040 restored by function calls on your machine. This ensures that code
9041 which is unaware of this reservation (such as library routines) will
9042 restore it before returning.
9044 On machines with register windows, be sure to choose a global
9045 register that is not affected magically by the function call mechanism.
9047 @subsubheading Using the variable
9049 @cindex @code{qsort}, and global register variables
9050 When calling routines that are not aware of the reservation, be
9051 cautious if those routines call back into code which uses them. As an
9052 example, if you call the system library version of @code{qsort}, it may
9053 clobber your registers during execution, but (if you have selected
9054 appropriate registers) it will restore them before returning. However
9055 it will @emph{not} restore them before calling @code{qsort}'s comparison
9056 function. As a result, global values will not reliably be available to
9057 the comparison function unless the @code{qsort} function itself is rebuilt.
9059 Similarly, it is not safe to access the global register variables from signal
9060 handlers or from more than one thread of control. Unless you recompile
9061 them specially for the task at hand, the system library routines may
9062 temporarily use the register for other things.
9064 @cindex register variable after @code{longjmp}
9065 @cindex global register after @code{longjmp}
9066 @cindex value after @code{longjmp}
9069 On most machines, @code{longjmp} restores to each global register
9070 variable the value it had at the time of the @code{setjmp}. On some
9071 machines, however, @code{longjmp} does not change the value of global
9072 register variables. To be portable, the function that called @code{setjmp}
9073 should make other arrangements to save the values of the global register
9074 variables, and to restore them in a @code{longjmp}. This way, the same
9075 thing happens regardless of what @code{longjmp} does.
9077 Eventually there may be a way of asking the compiler to choose a register
9078 automatically, but first we need to figure out how it should choose and
9079 how to enable you to guide the choice. No solution is evident.
9081 @node Local Register Variables
9082 @subsubsection Specifying Registers for Local Variables
9083 @anchor{Local Reg Vars}
9084 @cindex local variables, specifying registers
9085 @cindex specifying registers for local variables
9086 @cindex registers for local variables
9088 You can define a local register variable and associate it with a specified
9092 register int *foo asm ("r12");
9096 Here @code{r12} is the name of the register that should be used. Note
9097 that this is the same syntax used for defining global register variables,
9098 but for a local variable the declaration appears within a function. The
9099 @code{register} keyword is required, and cannot be combined with
9100 @code{static}. The register name must be a valid register name for the
9103 As with global register variables, it is recommended that you choose
9104 a register that is normally saved and restored by function calls on your
9105 machine, so that calls to library routines will not clobber it.
9107 The only supported use for this feature is to specify registers
9108 for input and output operands when calling Extended @code{asm}
9109 (@pxref{Extended Asm}). This may be necessary if the constraints for a
9110 particular machine don't provide sufficient control to select the desired
9111 register. To force an operand into a register, create a local variable
9112 and specify the register name after the variable's declaration. Then use
9113 the local variable for the @code{asm} operand and specify any constraint
9114 letter that matches the register:
9117 register int *p1 asm ("r0") = @dots{};
9118 register int *p2 asm ("r1") = @dots{};
9119 register int *result asm ("r0");
9120 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
9123 @emph{Warning:} In the above example, be aware that a register (for example
9124 @code{r0}) can be call-clobbered by subsequent code, including function
9125 calls and library calls for arithmetic operators on other variables (for
9126 example the initialization of @code{p2}). In this case, use temporary
9127 variables for expressions between the register assignments:
9131 register int *p1 asm ("r0") = @dots{};
9132 register int *p2 asm ("r1") = t1;
9133 register int *result asm ("r0");
9134 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
9137 Defining a register variable does not reserve the register. Other than
9138 when invoking the Extended @code{asm}, the contents of the specified
9139 register are not guaranteed. For this reason, the following uses
9140 are explicitly @emph{not} supported. If they appear to work, it is only
9141 happenstance, and may stop working as intended due to (seemingly)
9142 unrelated changes in surrounding code, or even minor changes in the
9143 optimization of a future version of gcc:
9146 @item Passing parameters to or from Basic @code{asm}
9147 @item Passing parameters to or from Extended @code{asm} without using input
9149 @item Passing parameters to or from routines written in assembler (or
9150 other languages) using non-standard calling conventions.
9153 Some developers use Local Register Variables in an attempt to improve
9154 gcc's allocation of registers, especially in large functions. In this
9155 case the register name is essentially a hint to the register allocator.
9156 While in some instances this can generate better code, improvements are
9157 subject to the whims of the allocator/optimizers. Since there are no
9158 guarantees that your improvements won't be lost, this usage of Local
9159 Register Variables is discouraged.
9161 On the MIPS platform, there is related use for local register variables
9162 with slightly different characteristics (@pxref{MIPS Coprocessors,,
9163 Defining coprocessor specifics for MIPS targets, gccint,
9164 GNU Compiler Collection (GCC) Internals}).
9166 @node Size of an asm
9167 @subsection Size of an @code{asm}
9169 Some targets require that GCC track the size of each instruction used
9170 in order to generate correct code. Because the final length of the
9171 code produced by an @code{asm} statement is only known by the
9172 assembler, GCC must make an estimate as to how big it will be. It
9173 does this by counting the number of instructions in the pattern of the
9174 @code{asm} and multiplying that by the length of the longest
9175 instruction supported by that processor. (When working out the number
9176 of instructions, it assumes that any occurrence of a newline or of
9177 whatever statement separator character is supported by the assembler --
9178 typically @samp{;} --- indicates the end of an instruction.)
9180 Normally, GCC's estimate is adequate to ensure that correct
9181 code is generated, but it is possible to confuse the compiler if you use
9182 pseudo instructions or assembler macros that expand into multiple real
9183 instructions, or if you use assembler directives that expand to more
9184 space in the object file than is needed for a single instruction.
9185 If this happens then the assembler may produce a diagnostic saying that
9186 a label is unreachable.
9188 @node Alternate Keywords
9189 @section Alternate Keywords
9190 @cindex alternate keywords
9191 @cindex keywords, alternate
9193 @option{-ansi} and the various @option{-std} options disable certain
9194 keywords. This causes trouble when you want to use GNU C extensions, or
9195 a general-purpose header file that should be usable by all programs,
9196 including ISO C programs. The keywords @code{asm}, @code{typeof} and
9197 @code{inline} are not available in programs compiled with
9198 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
9199 program compiled with @option{-std=c99} or @option{-std=c11}). The
9201 @code{restrict} is only available when @option{-std=gnu99} (which will
9202 eventually be the default) or @option{-std=c99} (or the equivalent
9203 @option{-std=iso9899:1999}), or an option for a later standard
9206 The way to solve these problems is to put @samp{__} at the beginning and
9207 end of each problematical keyword. For example, use @code{__asm__}
9208 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
9210 Other C compilers won't accept these alternative keywords; if you want to
9211 compile with another compiler, you can define the alternate keywords as
9212 macros to replace them with the customary keywords. It looks like this:
9220 @findex __extension__
9222 @option{-pedantic} and other options cause warnings for many GNU C extensions.
9224 prevent such warnings within one expression by writing
9225 @code{__extension__} before the expression. @code{__extension__} has no
9226 effect aside from this.
9228 @node Incomplete Enums
9229 @section Incomplete @code{enum} Types
9231 You can define an @code{enum} tag without specifying its possible values.
9232 This results in an incomplete type, much like what you get if you write
9233 @code{struct foo} without describing the elements. A later declaration
9234 that does specify the possible values completes the type.
9236 You cannot allocate variables or storage using the type while it is
9237 incomplete. However, you can work with pointers to that type.
9239 This extension may not be very useful, but it makes the handling of
9240 @code{enum} more consistent with the way @code{struct} and @code{union}
9243 This extension is not supported by GNU C++.
9245 @node Function Names
9246 @section Function Names as Strings
9247 @cindex @code{__func__} identifier
9248 @cindex @code{__FUNCTION__} identifier
9249 @cindex @code{__PRETTY_FUNCTION__} identifier
9251 GCC provides three magic constants that hold the name of the current
9252 function as a string. In C++11 and later modes, all three are treated
9253 as constant expressions and can be used in @code{constexpr} constexts.
9254 The first of these constants is @code{__func__}, which is part of
9257 The identifier @code{__func__} is implicitly declared by the translator
9258 as if, immediately following the opening brace of each function
9259 definition, the declaration
9262 static const char __func__[] = "function-name";
9266 appeared, where function-name is the name of the lexically-enclosing
9267 function. This name is the unadorned name of the function. As an
9268 extension, at file (or, in C++, namespace scope), @code{__func__}
9269 evaluates to the empty string.
9271 @code{__FUNCTION__} is another name for @code{__func__}, provided for
9272 backward compatibility with old versions of GCC.
9274 In C, @code{__PRETTY_FUNCTION__} is yet another name for
9275 @code{__func__}, except that at file (or, in C++, namespace scope),
9276 it evaluates to the string @code{"top level"}. In addition, in C++,
9277 @code{__PRETTY_FUNCTION__} contains the signature of the function as
9278 well as its bare name. For example, this program:
9281 extern "C" int printf (const char *, ...);
9287 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
9288 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
9306 __PRETTY_FUNCTION__ = void a::sub(int)
9309 These identifiers are variables, not preprocessor macros, and may not
9310 be used to initialize @code{char} arrays or be concatenated with string
9313 @node Return Address
9314 @section Getting the Return or Frame Address of a Function
9316 These functions may be used to get information about the callers of a
9319 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
9320 This function returns the return address of the current function, or of
9321 one of its callers. The @var{level} argument is number of frames to
9322 scan up the call stack. A value of @code{0} yields the return address
9323 of the current function, a value of @code{1} yields the return address
9324 of the caller of the current function, and so forth. When inlining
9325 the expected behavior is that the function returns the address of
9326 the function that is returned to. To work around this behavior use
9327 the @code{noinline} function attribute.
9329 The @var{level} argument must be a constant integer.
9331 On some machines it may be impossible to determine the return address of
9332 any function other than the current one; in such cases, or when the top
9333 of the stack has been reached, this function returns @code{0} or a
9334 random value. In addition, @code{__builtin_frame_address} may be used
9335 to determine if the top of the stack has been reached.
9337 Additional post-processing of the returned value may be needed, see
9338 @code{__builtin_extract_return_addr}.
9340 Calling this function with a nonzero argument can have unpredictable
9341 effects, including crashing the calling program. As a result, calls
9342 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9343 option is in effect. Such calls should only be made in debugging
9347 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
9348 The address as returned by @code{__builtin_return_address} may have to be fed
9349 through this function to get the actual encoded address. For example, on the
9350 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
9351 platforms an offset has to be added for the true next instruction to be
9354 If no fixup is needed, this function simply passes through @var{addr}.
9357 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
9358 This function does the reverse of @code{__builtin_extract_return_addr}.
9361 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9362 This function is similar to @code{__builtin_return_address}, but it
9363 returns the address of the function frame rather than the return address
9364 of the function. Calling @code{__builtin_frame_address} with a value of
9365 @code{0} yields the frame address of the current function, a value of
9366 @code{1} yields the frame address of the caller of the current function,
9369 The frame is the area on the stack that holds local variables and saved
9370 registers. The frame address is normally the address of the first word
9371 pushed on to the stack by the function. However, the exact definition
9372 depends upon the processor and the calling convention. If the processor
9373 has a dedicated frame pointer register, and the function has a frame,
9374 then @code{__builtin_frame_address} returns the value of the frame
9377 On some machines it may be impossible to determine the frame address of
9378 any function other than the current one; in such cases, or when the top
9379 of the stack has been reached, this function returns @code{0} if
9380 the first frame pointer is properly initialized by the startup code.
9382 Calling this function with a nonzero argument can have unpredictable
9383 effects, including crashing the calling program. As a result, calls
9384 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9385 option is in effect. Such calls should only be made in debugging
9389 @node Vector Extensions
9390 @section Using Vector Instructions through Built-in Functions
9392 On some targets, the instruction set contains SIMD vector instructions which
9393 operate on multiple values contained in one large register at the same time.
9394 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9397 The first step in using these extensions is to provide the necessary data
9398 types. This should be done using an appropriate @code{typedef}:
9401 typedef int v4si __attribute__ ((vector_size (16)));
9405 The @code{int} type specifies the base type, while the attribute specifies
9406 the vector size for the variable, measured in bytes. For example, the
9407 declaration above causes the compiler to set the mode for the @code{v4si}
9408 type to be 16 bytes wide and divided into @code{int} sized units. For
9409 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9410 corresponding mode of @code{foo} is @acronym{V4SI}.
9412 The @code{vector_size} attribute is only applicable to integral and
9413 float scalars, although arrays, pointers, and function return values
9414 are allowed in conjunction with this construct. Only sizes that are
9415 a power of two are currently allowed.
9417 All the basic integer types can be used as base types, both as signed
9418 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9419 @code{long long}. In addition, @code{float} and @code{double} can be
9420 used to build floating-point vector types.
9422 Specifying a combination that is not valid for the current architecture
9423 causes GCC to synthesize the instructions using a narrower mode.
9424 For example, if you specify a variable of type @code{V4SI} and your
9425 architecture does not allow for this specific SIMD type, GCC
9426 produces code that uses 4 @code{SIs}.
9428 The types defined in this manner can be used with a subset of normal C
9429 operations. Currently, GCC allows using the following operators
9430 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9432 The operations behave like C++ @code{valarrays}. Addition is defined as
9433 the addition of the corresponding elements of the operands. For
9434 example, in the code below, each of the 4 elements in @var{a} is
9435 added to the corresponding 4 elements in @var{b} and the resulting
9436 vector is stored in @var{c}.
9439 typedef int v4si __attribute__ ((vector_size (16)));
9446 Subtraction, multiplication, division, and the logical operations
9447 operate in a similar manner. Likewise, the result of using the unary
9448 minus or complement operators on a vector type is a vector whose
9449 elements are the negative or complemented values of the corresponding
9450 elements in the operand.
9452 It is possible to use shifting operators @code{<<}, @code{>>} on
9453 integer-type vectors. The operation is defined as following: @code{@{a0,
9454 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9455 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9458 For convenience, it is allowed to use a binary vector operation
9459 where one operand is a scalar. In that case the compiler transforms
9460 the scalar operand into a vector where each element is the scalar from
9461 the operation. The transformation happens only if the scalar could be
9462 safely converted to the vector-element type.
9463 Consider the following code.
9466 typedef int v4si __attribute__ ((vector_size (16)));
9471 a = b + 1; /* a = b + @{1,1,1,1@}; */
9472 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9474 a = l + a; /* Error, cannot convert long to int. */
9477 Vectors can be subscripted as if the vector were an array with
9478 the same number of elements and base type. Out of bound accesses
9479 invoke undefined behavior at run time. Warnings for out of bound
9480 accesses for vector subscription can be enabled with
9481 @option{-Warray-bounds}.
9483 Vector comparison is supported with standard comparison
9484 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9485 vector expressions of integer-type or real-type. Comparison between
9486 integer-type vectors and real-type vectors are not supported. The
9487 result of the comparison is a vector of the same width and number of
9488 elements as the comparison operands with a signed integral element
9491 Vectors are compared element-wise producing 0 when comparison is false
9492 and -1 (constant of the appropriate type where all bits are set)
9493 otherwise. Consider the following example.
9496 typedef int v4si __attribute__ ((vector_size (16)));
9498 v4si a = @{1,2,3,4@};
9499 v4si b = @{3,2,1,4@};
9502 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9503 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9506 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9507 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9508 integer vector with the same number of elements of the same size as @code{b}
9509 and @code{c}, computes all three arguments and creates a vector
9510 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9511 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9512 As in the case of binary operations, this syntax is also accepted when
9513 one of @code{b} or @code{c} is a scalar that is then transformed into a
9514 vector. If both @code{b} and @code{c} are scalars and the type of
9515 @code{true?b:c} has the same size as the element type of @code{a}, then
9516 @code{b} and @code{c} are converted to a vector type whose elements have
9517 this type and with the same number of elements as @code{a}.
9519 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9520 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9521 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9522 For mixed operations between a scalar @code{s} and a vector @code{v},
9523 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9524 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9526 Vector shuffling is available using functions
9527 @code{__builtin_shuffle (vec, mask)} and
9528 @code{__builtin_shuffle (vec0, vec1, mask)}.
9529 Both functions construct a permutation of elements from one or two
9530 vectors and return a vector of the same type as the input vector(s).
9531 The @var{mask} is an integral vector with the same width (@var{W})
9532 and element count (@var{N}) as the output vector.
9534 The elements of the input vectors are numbered in memory ordering of
9535 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9536 elements of @var{mask} are considered modulo @var{N} in the single-operand
9537 case and modulo @math{2*@var{N}} in the two-operand case.
9539 Consider the following example,
9542 typedef int v4si __attribute__ ((vector_size (16)));
9544 v4si a = @{1,2,3,4@};
9545 v4si b = @{5,6,7,8@};
9546 v4si mask1 = @{0,1,1,3@};
9547 v4si mask2 = @{0,4,2,5@};
9550 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9551 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9554 Note that @code{__builtin_shuffle} is intentionally semantically
9555 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9557 You can declare variables and use them in function calls and returns, as
9558 well as in assignments and some casts. You can specify a vector type as
9559 a return type for a function. Vector types can also be used as function
9560 arguments. It is possible to cast from one vector type to another,
9561 provided they are of the same size (in fact, you can also cast vectors
9562 to and from other datatypes of the same size).
9564 You cannot operate between vectors of different lengths or different
9565 signedness without a cast.
9568 @section Support for @code{offsetof}
9569 @findex __builtin_offsetof
9571 GCC implements for both C and C++ a syntactic extension to implement
9572 the @code{offsetof} macro.
9576 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9578 offsetof_member_designator:
9580 | offsetof_member_designator "." @code{identifier}
9581 | offsetof_member_designator "[" @code{expr} "]"
9584 This extension is sufficient such that
9587 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9591 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9592 may be dependent. In either case, @var{member} may consist of a single
9593 identifier, or a sequence of member accesses and array references.
9595 @node __sync Builtins
9596 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9598 The following built-in functions
9599 are intended to be compatible with those described
9600 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9601 section 7.4. As such, they depart from normal GCC practice by not using
9602 the @samp{__builtin_} prefix and also by being overloaded so that they
9603 work on multiple types.
9605 The definition given in the Intel documentation allows only for the use of
9606 the types @code{int}, @code{long}, @code{long long} or their unsigned
9607 counterparts. GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
9608 size other than the C type @code{_Bool} or the C++ type @code{bool}.
9609 Operations on pointer arguments are performed as if the operands were
9610 of the @code{uintptr_t} type. That is, they are not scaled by the size
9611 of the type to which the pointer points.
9613 These functions are implemented in terms of the @samp{__atomic}
9614 builtins (@pxref{__atomic Builtins}). They should not be used for new
9615 code which should use the @samp{__atomic} builtins instead.
9617 Not all operations are supported by all target processors. If a particular
9618 operation cannot be implemented on the target processor, a warning is
9619 generated and a call to an external function is generated. The external
9620 function carries the same name as the built-in version,
9621 with an additional suffix
9622 @samp{_@var{n}} where @var{n} is the size of the data type.
9624 @c ??? Should we have a mechanism to suppress this warning? This is almost
9625 @c useful for implementing the operation under the control of an external
9628 In most cases, these built-in functions are considered a @dfn{full barrier}.
9630 no memory operand is moved across the operation, either forward or
9631 backward. Further, instructions are issued as necessary to prevent the
9632 processor from speculating loads across the operation and from queuing stores
9633 after the operation.
9635 All of the routines are described in the Intel documentation to take
9636 ``an optional list of variables protected by the memory barrier''. It's
9637 not clear what is meant by that; it could mean that @emph{only} the
9638 listed variables are protected, or it could mean a list of additional
9639 variables to be protected. The list is ignored by GCC which treats it as
9640 empty. GCC interprets an empty list as meaning that all globally
9641 accessible variables should be protected.
9644 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9645 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9646 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9647 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9648 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9649 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9650 @findex __sync_fetch_and_add
9651 @findex __sync_fetch_and_sub
9652 @findex __sync_fetch_and_or
9653 @findex __sync_fetch_and_and
9654 @findex __sync_fetch_and_xor
9655 @findex __sync_fetch_and_nand
9656 These built-in functions perform the operation suggested by the name, and
9657 returns the value that had previously been in memory. That is, operations
9658 on integer operands have the following semantics. Operations on pointer
9659 arguments are performed as if the operands were of the @code{uintptr_t}
9660 type. That is, they are not scaled by the size of the type to which
9664 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9665 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9668 The object pointed to by the first argument must be of integer or pointer
9669 type. It must not be a boolean type.
9671 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9672 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9674 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9675 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9676 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9677 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9678 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9679 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9680 @findex __sync_add_and_fetch
9681 @findex __sync_sub_and_fetch
9682 @findex __sync_or_and_fetch
9683 @findex __sync_and_and_fetch
9684 @findex __sync_xor_and_fetch
9685 @findex __sync_nand_and_fetch
9686 These built-in functions perform the operation suggested by the name, and
9687 return the new value. That is, operations on integer operands have
9688 the following semantics. Operations on pointer operands are performed as
9689 if the operand's type were @code{uintptr_t}.
9692 @{ *ptr @var{op}= value; return *ptr; @}
9693 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9696 The same constraints on arguments apply as for the corresponding
9697 @code{__sync_op_and_fetch} built-in functions.
9699 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9700 as @code{*ptr = ~(*ptr & value)} instead of
9701 @code{*ptr = ~*ptr & value}.
9703 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9704 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9705 @findex __sync_bool_compare_and_swap
9706 @findex __sync_val_compare_and_swap
9707 These built-in functions perform an atomic compare and swap.
9708 That is, if the current
9709 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9712 The ``bool'' version returns true if the comparison is successful and
9713 @var{newval} is written. The ``val'' version returns the contents
9714 of @code{*@var{ptr}} before the operation.
9716 @item __sync_synchronize (...)
9717 @findex __sync_synchronize
9718 This built-in function issues a full memory barrier.
9720 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9721 @findex __sync_lock_test_and_set
9722 This built-in function, as described by Intel, is not a traditional test-and-set
9723 operation, but rather an atomic exchange operation. It writes @var{value}
9724 into @code{*@var{ptr}}, and returns the previous contents of
9727 Many targets have only minimal support for such locks, and do not support
9728 a full exchange operation. In this case, a target may support reduced
9729 functionality here by which the @emph{only} valid value to store is the
9730 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9731 is implementation defined.
9733 This built-in function is not a full barrier,
9734 but rather an @dfn{acquire barrier}.
9735 This means that references after the operation cannot move to (or be
9736 speculated to) before the operation, but previous memory stores may not
9737 be globally visible yet, and previous memory loads may not yet be
9740 @item void __sync_lock_release (@var{type} *ptr, ...)
9741 @findex __sync_lock_release
9742 This built-in function releases the lock acquired by
9743 @code{__sync_lock_test_and_set}.
9744 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9746 This built-in function is not a full barrier,
9747 but rather a @dfn{release barrier}.
9748 This means that all previous memory stores are globally visible, and all
9749 previous memory loads have been satisfied, but following memory reads
9750 are not prevented from being speculated to before the barrier.
9753 @node __atomic Builtins
9754 @section Built-in Functions for Memory Model Aware Atomic Operations
9756 The following built-in functions approximately match the requirements
9757 for the C++11 memory model. They are all
9758 identified by being prefixed with @samp{__atomic} and most are
9759 overloaded so that they work with multiple types.
9761 These functions are intended to replace the legacy @samp{__sync}
9762 builtins. The main difference is that the memory order that is requested
9763 is a parameter to the functions. New code should always use the
9764 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9766 Note that the @samp{__atomic} builtins assume that programs will
9767 conform to the C++11 memory model. In particular, they assume
9768 that programs are free of data races. See the C++11 standard for
9769 detailed requirements.
9771 The @samp{__atomic} builtins can be used with any integral scalar or
9772 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9773 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9774 supported by the architecture.
9776 The four non-arithmetic functions (load, store, exchange, and
9777 compare_exchange) all have a generic version as well. This generic
9778 version works on any data type. It uses the lock-free built-in function
9779 if the specific data type size makes that possible; otherwise, an
9780 external call is left to be resolved at run time. This external call is
9781 the same format with the addition of a @samp{size_t} parameter inserted
9782 as the first parameter indicating the size of the object being pointed to.
9783 All objects must be the same size.
9785 There are 6 different memory orders that can be specified. These map
9786 to the C++11 memory orders with the same names, see the C++11 standard
9787 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9788 on atomic synchronization} for detailed definitions. Individual
9789 targets may also support additional memory orders for use on specific
9790 architectures. Refer to the target documentation for details of
9793 An atomic operation can both constrain code motion and
9794 be mapped to hardware instructions for synchronization between threads
9795 (e.g., a fence). To which extent this happens is controlled by the
9796 memory orders, which are listed here in approximately ascending order of
9797 strength. The description of each memory order is only meant to roughly
9798 illustrate the effects and is not a specification; see the C++11
9799 memory model for precise semantics.
9802 @item __ATOMIC_RELAXED
9803 Implies no inter-thread ordering constraints.
9804 @item __ATOMIC_CONSUME
9805 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9806 memory order because of a deficiency in C++11's semantics for
9807 @code{memory_order_consume}.
9808 @item __ATOMIC_ACQUIRE
9809 Creates an inter-thread happens-before constraint from the release (or
9810 stronger) semantic store to this acquire load. Can prevent hoisting
9811 of code to before the operation.
9812 @item __ATOMIC_RELEASE
9813 Creates an inter-thread happens-before constraint to acquire (or stronger)
9814 semantic loads that read from this release store. Can prevent sinking
9815 of code to after the operation.
9816 @item __ATOMIC_ACQ_REL
9817 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9818 @code{__ATOMIC_RELEASE}.
9819 @item __ATOMIC_SEQ_CST
9820 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9823 Note that in the C++11 memory model, @emph{fences} (e.g.,
9824 @samp{__atomic_thread_fence}) take effect in combination with other
9825 atomic operations on specific memory locations (e.g., atomic loads);
9826 operations on specific memory locations do not necessarily affect other
9827 operations in the same way.
9829 Target architectures are encouraged to provide their own patterns for
9830 each of the atomic built-in functions. If no target is provided, the original
9831 non-memory model set of @samp{__sync} atomic built-in functions are
9832 used, along with any required synchronization fences surrounding it in
9833 order to achieve the proper behavior. Execution in this case is subject
9834 to the same restrictions as those built-in functions.
9836 If there is no pattern or mechanism to provide a lock-free instruction
9837 sequence, a call is made to an external routine with the same parameters
9838 to be resolved at run time.
9840 When implementing patterns for these built-in functions, the memory order
9841 parameter can be ignored as long as the pattern implements the most
9842 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9843 orders execute correctly with this memory order but they may not execute as
9844 efficiently as they could with a more appropriate implementation of the
9845 relaxed requirements.
9847 Note that the C++11 standard allows for the memory order parameter to be
9848 determined at run time rather than at compile time. These built-in
9849 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9850 than invoke a runtime library call or inline a switch statement. This is
9851 standard compliant, safe, and the simplest approach for now.
9853 The memory order parameter is a signed int, but only the lower 16 bits are
9854 reserved for the memory order. The remainder of the signed int is reserved
9855 for target use and should be 0. Use of the predefined atomic values
9856 ensures proper usage.
9858 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9859 This built-in function implements an atomic load operation. It returns the
9860 contents of @code{*@var{ptr}}.
9862 The valid memory order variants are
9863 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9864 and @code{__ATOMIC_CONSUME}.
9868 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9869 This is the generic version of an atomic load. It returns the
9870 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9874 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9875 This built-in function implements an atomic store operation. It writes
9876 @code{@var{val}} into @code{*@var{ptr}}.
9878 The valid memory order variants are
9879 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9883 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9884 This is the generic version of an atomic store. It stores the value
9885 of @code{*@var{val}} into @code{*@var{ptr}}.
9889 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9890 This built-in function implements an atomic exchange operation. It writes
9891 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9894 The valid memory order variants are
9895 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9896 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9900 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9901 This is the generic version of an atomic exchange. It stores the
9902 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9903 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9907 @deftypefn {Built-in Function} bool __atomic_compare_exchange_n (@var{type} *ptr, @var{type} *expected, @var{type} desired, bool weak, int success_memorder, int failure_memorder)
9908 This built-in function implements an atomic compare and exchange operation.
9909 This compares the contents of @code{*@var{ptr}} with the contents of
9910 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9911 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9912 equal, the operation is a @emph{read} and the current contents of
9913 @code{*@var{ptr}} are written into @code{*@var{expected}}. @var{weak} is true
9914 for weak compare_exchange, which may fail spuriously, and false for
9915 the strong variation, which never fails spuriously. Many targets
9916 only offer the strong variation and ignore the parameter. When in doubt, use
9917 the strong variation.
9919 If @var{desired} is written into @code{*@var{ptr}} then true is returned
9920 and memory is affected according to the
9921 memory order specified by @var{success_memorder}. There are no
9922 restrictions on what memory order can be used here.
9924 Otherwise, false is returned and memory is affected according
9925 to @var{failure_memorder}. This memory order cannot be
9926 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9927 stronger order than that specified by @var{success_memorder}.
9931 @deftypefn {Built-in Function} bool __atomic_compare_exchange (@var{type} *ptr, @var{type} *expected, @var{type} *desired, bool weak, int success_memorder, int failure_memorder)
9932 This built-in function implements the generic version of
9933 @code{__atomic_compare_exchange}. The function is virtually identical to
9934 @code{__atomic_compare_exchange_n}, except the desired value is also a
9939 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9940 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9941 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9942 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9943 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9944 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9945 These built-in functions perform the operation suggested by the name, and
9946 return the result of the operation. Operations on pointer arguments are
9947 performed as if the operands were of the @code{uintptr_t} type. That is,
9948 they are not scaled by the size of the type to which the pointer points.
9951 @{ *ptr @var{op}= val; return *ptr; @}
9954 The object pointed to by the first argument must be of integer or pointer
9955 type. It must not be a boolean type. All memory orders are valid.
9959 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9960 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9961 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9962 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9963 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9964 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9965 These built-in functions perform the operation suggested by the name, and
9966 return the value that had previously been in @code{*@var{ptr}}. Operations
9967 on pointer arguments are performed as if the operands were of
9968 the @code{uintptr_t} type. That is, they are not scaled by the size of
9969 the type to which the pointer points.
9972 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9975 The same constraints on arguments apply as for the corresponding
9976 @code{__atomic_op_fetch} built-in functions. All memory orders are valid.
9980 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9982 This built-in function performs an atomic test-and-set operation on
9983 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9984 defined nonzero ``set'' value and the return value is @code{true} if and only
9985 if the previous contents were ``set''.
9986 It should be only used for operands of type @code{bool} or @code{char}. For
9987 other types only part of the value may be set.
9989 All memory orders are valid.
9993 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9995 This built-in function performs an atomic clear operation on
9996 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9997 It should be only used for operands of type @code{bool} or @code{char} and
9998 in conjunction with @code{__atomic_test_and_set}.
9999 For other types it may only clear partially. If the type is not @code{bool}
10000 prefer using @code{__atomic_store}.
10002 The valid memory order variants are
10003 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
10004 @code{__ATOMIC_RELEASE}.
10008 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
10010 This built-in function acts as a synchronization fence between threads
10011 based on the specified memory order.
10013 All memory orders are valid.
10017 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
10019 This built-in function acts as a synchronization fence between a thread
10020 and signal handlers based in the same thread.
10022 All memory orders are valid.
10026 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
10028 This built-in function returns true if objects of @var{size} bytes always
10029 generate lock-free atomic instructions for the target architecture.
10030 @var{size} must resolve to a compile-time constant and the result also
10031 resolves to a compile-time constant.
10033 @var{ptr} is an optional pointer to the object that may be used to determine
10034 alignment. A value of 0 indicates typical alignment should be used. The
10035 compiler may also ignore this parameter.
10038 if (__atomic_always_lock_free (sizeof (long long), 0))
10043 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
10045 This built-in function returns true if objects of @var{size} bytes always
10046 generate lock-free atomic instructions for the target architecture. If
10047 the built-in function is not known to be lock-free, a call is made to a
10048 runtime routine named @code{__atomic_is_lock_free}.
10050 @var{ptr} is an optional pointer to the object that may be used to determine
10051 alignment. A value of 0 indicates typical alignment should be used. The
10052 compiler may also ignore this parameter.
10055 @node Integer Overflow Builtins
10056 @section Built-in Functions to Perform Arithmetic with Overflow Checking
10058 The following built-in functions allow performing simple arithmetic operations
10059 together with checking whether the operations overflowed.
10061 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
10062 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
10063 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
10064 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long long int *res)
10065 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
10066 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
10067 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)
10069 These built-in functions promote the first two operands into infinite precision signed
10070 type and perform addition on those promoted operands. The result is then
10071 cast to the type the third pointer argument points to and stored there.
10072 If the stored result is equal to the infinite precision result, the built-in
10073 functions return false, otherwise they return true. As the addition is
10074 performed in infinite signed precision, these built-in functions have fully defined
10075 behavior for all argument values.
10077 The first built-in function allows arbitrary integral types for operands and
10078 the result type must be pointer to some integral type other than enumerated or
10079 boolean type, the rest of the built-in functions have explicit integer types.
10081 The compiler will attempt to use hardware instructions to implement
10082 these built-in functions where possible, like conditional jump on overflow
10083 after addition, conditional jump on carry etc.
10087 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
10088 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
10089 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
10090 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long long int *res)
10091 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
10092 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
10093 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)
10095 These built-in functions are similar to the add overflow checking built-in
10096 functions above, except they perform subtraction, subtract the second argument
10097 from the first one, instead of addition.
10101 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
10102 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
10103 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
10104 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long long int *res)
10105 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
10106 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
10107 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)
10109 These built-in functions are similar to the add overflow checking built-in
10110 functions above, except they perform multiplication, instead of addition.
10114 The following built-in functions allow checking if simple arithmetic operation
10117 @deftypefn {Built-in Function} bool __builtin_add_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
10118 @deftypefnx {Built-in Function} bool __builtin_sub_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
10119 @deftypefnx {Built-in Function} bool __builtin_mul_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
10121 These built-in functions are similar to @code{__builtin_add_overflow},
10122 @code{__builtin_sub_overflow}, or @code{__builtin_mul_overflow}, except that
10123 they don't store the result of the arithmetic operation anywhere and the
10124 last argument is not a pointer, but some expression with integral type other
10125 than enumerated or boolean type.
10127 The built-in functions promote the first two operands into infinite precision signed type
10128 and perform addition on those promoted operands. The result is then
10129 cast to the type of the third argument. If the cast result is equal to the infinite
10130 precision result, the built-in functions return false, otherwise they return true.
10131 The value of the third argument is ignored, just the side-effects in the third argument
10132 are evaluated, and no integral argument promotions are performed on the last argument.
10133 If the third argument is a bit-field, the type used for the result cast has the
10134 precision and signedness of the given bit-field, rather than precision and signedness
10135 of the underlying type.
10137 For example, the following macro can be used to portably check, at
10138 compile-time, whether or not adding two constant integers will overflow,
10139 and perform the addition only when it is known to be safe and not to trigger
10140 a @option{-Woverflow} warning.
10143 #define INT_ADD_OVERFLOW_P(a, b) \
10144 __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0)
10147 A = INT_MAX, B = 3,
10148 C = INT_ADD_OVERFLOW_P (A, B) ? 0 : A + B,
10149 D = __builtin_add_overflow_p (1, SCHAR_MAX, (signed char) 0)
10153 The compiler will attempt to use hardware instructions to implement
10154 these built-in functions where possible, like conditional jump on overflow
10155 after addition, conditional jump on carry etc.
10159 @node x86 specific memory model extensions for transactional memory
10160 @section x86-Specific Memory Model Extensions for Transactional Memory
10162 The x86 architecture supports additional memory ordering flags
10163 to mark critical sections for hardware lock elision.
10164 These must be specified in addition to an existing memory order to
10168 @item __ATOMIC_HLE_ACQUIRE
10169 Start lock elision on a lock variable.
10170 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
10171 @item __ATOMIC_HLE_RELEASE
10172 End lock elision on a lock variable.
10173 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
10176 When a lock acquire fails, it is required for good performance to abort
10177 the transaction quickly. This can be done with a @code{_mm_pause}.
10180 #include <immintrin.h> // For _mm_pause
10184 /* Acquire lock with lock elision */
10185 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
10186 _mm_pause(); /* Abort failed transaction */
10188 /* Free lock with lock elision */
10189 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
10192 @node Object Size Checking
10193 @section Object Size Checking Built-in Functions
10194 @findex __builtin_object_size
10195 @findex __builtin___memcpy_chk
10196 @findex __builtin___mempcpy_chk
10197 @findex __builtin___memmove_chk
10198 @findex __builtin___memset_chk
10199 @findex __builtin___strcpy_chk
10200 @findex __builtin___stpcpy_chk
10201 @findex __builtin___strncpy_chk
10202 @findex __builtin___strcat_chk
10203 @findex __builtin___strncat_chk
10204 @findex __builtin___sprintf_chk
10205 @findex __builtin___snprintf_chk
10206 @findex __builtin___vsprintf_chk
10207 @findex __builtin___vsnprintf_chk
10208 @findex __builtin___printf_chk
10209 @findex __builtin___vprintf_chk
10210 @findex __builtin___fprintf_chk
10211 @findex __builtin___vfprintf_chk
10213 GCC implements a limited buffer overflow protection mechanism that can
10214 prevent some buffer overflow attacks by determining the sizes of objects
10215 into which data is about to be written and preventing the writes when
10216 the size isn't sufficient. The built-in functions described below yield
10217 the best results when used together and when optimization is enabled.
10218 For example, to detect object sizes across function boundaries or to
10219 follow pointer assignments through non-trivial control flow they rely
10220 on various optimization passes enabled with @option{-O2}. However, to
10221 a limited extent, they can be used without optimization as well.
10223 @deftypefn {Built-in Function} {size_t} __builtin_object_size (const void * @var{ptr}, int @var{type})
10224 is a built-in construct that returns a constant number of bytes from
10225 @var{ptr} to the end of the object @var{ptr} pointer points to
10226 (if known at compile time). @code{__builtin_object_size} never evaluates
10227 its arguments for side-effects. If there are any side-effects in them, it
10228 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10229 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
10230 point to and all of them are known at compile time, the returned number
10231 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
10232 0 and minimum if nonzero. If it is not possible to determine which objects
10233 @var{ptr} points to at compile time, @code{__builtin_object_size} should
10234 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10235 for @var{type} 2 or 3.
10237 @var{type} is an integer constant from 0 to 3. If the least significant
10238 bit is clear, objects are whole variables, if it is set, a closest
10239 surrounding subobject is considered the object a pointer points to.
10240 The second bit determines if maximum or minimum of remaining bytes
10244 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
10245 char *p = &var.buf1[1], *q = &var.b;
10247 /* Here the object p points to is var. */
10248 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
10249 /* The subobject p points to is var.buf1. */
10250 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
10251 /* The object q points to is var. */
10252 assert (__builtin_object_size (q, 0)
10253 == (char *) (&var + 1) - (char *) &var.b);
10254 /* The subobject q points to is var.b. */
10255 assert (__builtin_object_size (q, 1) == sizeof (var.b));
10259 There are built-in functions added for many common string operation
10260 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
10261 built-in is provided. This built-in has an additional last argument,
10262 which is the number of bytes remaining in the object the @var{dest}
10263 argument points to or @code{(size_t) -1} if the size is not known.
10265 The built-in functions are optimized into the normal string functions
10266 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
10267 it is known at compile time that the destination object will not
10268 be overflowed. If the compiler can determine at compile time that the
10269 object will always be overflowed, it issues a warning.
10271 The intended use can be e.g.@:
10275 #define bos0(dest) __builtin_object_size (dest, 0)
10276 #define memcpy(dest, src, n) \
10277 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
10281 /* It is unknown what object p points to, so this is optimized
10282 into plain memcpy - no checking is possible. */
10283 memcpy (p, "abcde", n);
10284 /* Destination is known and length too. It is known at compile
10285 time there will be no overflow. */
10286 memcpy (&buf[5], "abcde", 5);
10287 /* Destination is known, but the length is not known at compile time.
10288 This will result in __memcpy_chk call that can check for overflow
10290 memcpy (&buf[5], "abcde", n);
10291 /* Destination is known and it is known at compile time there will
10292 be overflow. There will be a warning and __memcpy_chk call that
10293 will abort the program at run time. */
10294 memcpy (&buf[6], "abcde", 5);
10297 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
10298 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
10299 @code{strcat} and @code{strncat}.
10301 There are also checking built-in functions for formatted output functions.
10303 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
10304 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10305 const char *fmt, ...);
10306 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
10308 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10309 const char *fmt, va_list ap);
10312 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
10313 etc.@: functions and can contain implementation specific flags on what
10314 additional security measures the checking function might take, such as
10315 handling @code{%n} differently.
10317 The @var{os} argument is the object size @var{s} points to, like in the
10318 other built-in functions. There is a small difference in the behavior
10319 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
10320 optimized into the non-checking functions only if @var{flag} is 0, otherwise
10321 the checking function is called with @var{os} argument set to
10322 @code{(size_t) -1}.
10324 In addition to this, there are checking built-in functions
10325 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
10326 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
10327 These have just one additional argument, @var{flag}, right before
10328 format string @var{fmt}. If the compiler is able to optimize them to
10329 @code{fputc} etc.@: functions, it does, otherwise the checking function
10330 is called and the @var{flag} argument passed to it.
10332 @node Pointer Bounds Checker builtins
10333 @section Pointer Bounds Checker Built-in Functions
10334 @cindex Pointer Bounds Checker builtins
10335 @findex __builtin___bnd_set_ptr_bounds
10336 @findex __builtin___bnd_narrow_ptr_bounds
10337 @findex __builtin___bnd_copy_ptr_bounds
10338 @findex __builtin___bnd_init_ptr_bounds
10339 @findex __builtin___bnd_null_ptr_bounds
10340 @findex __builtin___bnd_store_ptr_bounds
10341 @findex __builtin___bnd_chk_ptr_lbounds
10342 @findex __builtin___bnd_chk_ptr_ubounds
10343 @findex __builtin___bnd_chk_ptr_bounds
10344 @findex __builtin___bnd_get_ptr_lbound
10345 @findex __builtin___bnd_get_ptr_ubound
10347 GCC provides a set of built-in functions to control Pointer Bounds Checker
10348 instrumentation. Note that all Pointer Bounds Checker builtins can be used
10349 even if you compile with Pointer Bounds Checker off
10350 (@option{-fno-check-pointer-bounds}).
10351 The behavior may differ in such case as documented below.
10353 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
10355 This built-in function returns a new pointer with the value of @var{q}, and
10356 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
10357 Bounds Checker off, the built-in function just returns the first argument.
10360 extern void *__wrap_malloc (size_t n)
10362 void *p = (void *)__real_malloc (n);
10363 if (!p) return __builtin___bnd_null_ptr_bounds (p);
10364 return __builtin___bnd_set_ptr_bounds (p, n);
10370 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
10372 This built-in function returns a new pointer with the value of @var{p}
10373 and associates it with the narrowed bounds formed by the intersection
10374 of bounds associated with @var{q} and the bounds
10375 [@var{p}, @var{p} + @var{size} - 1].
10376 With Pointer Bounds Checker off, the built-in function just returns the first
10380 void init_objects (object *objs, size_t size)
10383 /* Initialize objects one-by-one passing pointers with bounds of
10384 an object, not the full array of objects. */
10385 for (i = 0; i < size; i++)
10386 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
10393 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
10395 This built-in function returns a new pointer with the value of @var{q},
10396 and associates it with the bounds already associated with pointer @var{r}.
10397 With Pointer Bounds Checker off, the built-in function just returns the first
10401 /* Here is a way to get pointer to object's field but
10402 still with the full object's bounds. */
10403 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
10409 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
10411 This built-in function returns a new pointer with the value of @var{q}, and
10412 associates it with INIT (allowing full memory access) bounds. With Pointer
10413 Bounds Checker off, the built-in function just returns the first argument.
10417 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
10419 This built-in function returns a new pointer with the value of @var{q}, and
10420 associates it with NULL (allowing no memory access) bounds. With Pointer
10421 Bounds Checker off, the built-in function just returns the first argument.
10425 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
10427 This built-in function stores the bounds associated with pointer @var{ptr_val}
10428 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
10429 bounds from legacy code without touching the associated pointer's memory when
10430 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
10431 function call is ignored.
10435 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10437 This built-in function checks if the pointer @var{q} is within the lower
10438 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10439 function call is ignored.
10442 extern void *__wrap_memset (void *dst, int c, size_t len)
10446 __builtin___bnd_chk_ptr_lbounds (dst);
10447 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10448 __real_memset (dst, c, len);
10456 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10458 This built-in function checks if the pointer @var{q} is within the upper
10459 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10460 function call is ignored.
10464 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10466 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10467 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10468 off, the built-in function call is ignored.
10471 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10475 __bnd_chk_ptr_bounds (dst, n);
10476 __bnd_chk_ptr_bounds (src, n);
10477 __real_memcpy (dst, src, n);
10485 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10487 This built-in function returns the lower bound associated
10488 with the pointer @var{q}, as a pointer value.
10489 This is useful for debugging using @code{printf}.
10490 With Pointer Bounds Checker off, the built-in function returns 0.
10493 void *lb = __builtin___bnd_get_ptr_lbound (q);
10494 void *ub = __builtin___bnd_get_ptr_ubound (q);
10495 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10500 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10502 This built-in function returns the upper bound (which is a pointer) associated
10503 with the pointer @var{q}. With Pointer Bounds Checker off,
10504 the built-in function returns -1.
10508 @node Cilk Plus Builtins
10509 @section Cilk Plus C/C++ Language Extension Built-in Functions
10511 GCC provides support for the following built-in reduction functions if Cilk Plus
10512 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10515 @item @code{__sec_implicit_index}
10516 @item @code{__sec_reduce}
10517 @item @code{__sec_reduce_add}
10518 @item @code{__sec_reduce_all_nonzero}
10519 @item @code{__sec_reduce_all_zero}
10520 @item @code{__sec_reduce_any_nonzero}
10521 @item @code{__sec_reduce_any_zero}
10522 @item @code{__sec_reduce_max}
10523 @item @code{__sec_reduce_min}
10524 @item @code{__sec_reduce_max_ind}
10525 @item @code{__sec_reduce_min_ind}
10526 @item @code{__sec_reduce_mul}
10527 @item @code{__sec_reduce_mutating}
10530 Further details and examples about these built-in functions are described
10531 in the Cilk Plus language manual which can be found at
10532 @uref{https://www.cilkplus.org}.
10534 @node Other Builtins
10535 @section Other Built-in Functions Provided by GCC
10536 @cindex built-in functions
10537 @findex __builtin_alloca
10538 @findex __builtin_alloca_with_align
10539 @findex __builtin_call_with_static_chain
10540 @findex __builtin_fpclassify
10541 @findex __builtin_isfinite
10542 @findex __builtin_isnormal
10543 @findex __builtin_isgreater
10544 @findex __builtin_isgreaterequal
10545 @findex __builtin_isinf_sign
10546 @findex __builtin_isless
10547 @findex __builtin_islessequal
10548 @findex __builtin_islessgreater
10549 @findex __builtin_isunordered
10550 @findex __builtin_powi
10551 @findex __builtin_powif
10552 @findex __builtin_powil
10713 @findex fprintf_unlocked
10715 @findex fputs_unlocked
10823 @findex nexttowardf
10824 @findex nexttowardl
10832 @findex printf_unlocked
10862 @findex signbitd128
10863 @findex significand
10864 @findex significandf
10865 @findex significandl
10893 @findex strncasecmp
10936 GCC provides a large number of built-in functions other than the ones
10937 mentioned above. Some of these are for internal use in the processing
10938 of exceptions or variable-length argument lists and are not
10939 documented here because they may change from time to time; we do not
10940 recommend general use of these functions.
10942 The remaining functions are provided for optimization purposes.
10944 With the exception of built-ins that have library equivalents such as
10945 the standard C library functions discussed below, or that expand to
10946 library calls, GCC built-in functions are always expanded inline and
10947 thus do not have corresponding entry points and their address cannot
10948 be obtained. Attempting to use them in an expression other than
10949 a function call results in a compile-time error.
10951 @opindex fno-builtin
10952 GCC includes built-in versions of many of the functions in the standard
10953 C library. These functions come in two forms: one whose names start with
10954 the @code{__builtin_} prefix, and the other without. Both forms have the
10955 same type (including prototype), the same address (when their address is
10956 taken), and the same meaning as the C library functions even if you specify
10957 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10958 functions are only optimized in certain cases; if they are not optimized in
10959 a particular case, a call to the library function is emitted.
10963 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10964 @option{-std=c99} or @option{-std=c11}), the functions
10965 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10966 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10967 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10968 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10969 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10970 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10971 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10972 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10973 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10974 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10975 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10976 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10977 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10978 @code{significandl}, @code{significand}, @code{sincosf},
10979 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10980 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10981 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10982 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10984 may be handled as built-in functions.
10985 All these functions have corresponding versions
10986 prefixed with @code{__builtin_}, which may be used even in strict C90
10989 The ISO C99 functions
10990 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10991 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10992 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10993 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10994 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10995 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10996 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10997 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10998 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10999 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
11000 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
11001 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
11002 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
11003 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
11004 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
11005 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
11006 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
11007 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
11008 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
11009 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
11010 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
11011 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
11012 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
11013 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
11014 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
11015 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
11016 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
11017 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
11018 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
11019 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
11020 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
11021 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
11022 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
11023 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
11024 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
11025 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
11026 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
11027 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
11028 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
11029 are handled as built-in functions
11030 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
11032 There are also built-in versions of the ISO C99 functions
11033 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
11034 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
11035 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
11036 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
11037 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
11038 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
11039 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
11040 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
11041 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
11042 that are recognized in any mode since ISO C90 reserves these names for
11043 the purpose to which ISO C99 puts them. All these functions have
11044 corresponding versions prefixed with @code{__builtin_}.
11046 There are also built-in functions @code{__builtin_fabsf@var{n}},
11047 @code{__builtin_fabsf@var{n}x}, @code{__builtin_copysignf@var{n}} and
11048 @code{__builtin_copysignf@var{n}x}, corresponding to the TS 18661-3
11049 functions @code{fabsf@var{n}}, @code{fabsf@var{n}x},
11050 @code{copysignf@var{n}} and @code{copysignf@var{n}x}, for supported
11051 types @code{_Float@var{n}} and @code{_Float@var{n}x}.
11053 There are also GNU extension functions @code{clog10}, @code{clog10f} and
11054 @code{clog10l} which names are reserved by ISO C99 for future use.
11055 All these functions have versions prefixed with @code{__builtin_}.
11057 The ISO C94 functions
11058 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
11059 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
11060 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
11062 are handled as built-in functions
11063 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
11065 The ISO C90 functions
11066 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
11067 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
11068 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
11069 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
11070 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
11071 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
11072 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
11073 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
11074 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
11075 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
11076 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
11077 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
11078 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
11079 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
11080 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
11081 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
11082 are all recognized as built-in functions unless
11083 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
11084 is specified for an individual function). All of these functions have
11085 corresponding versions prefixed with @code{__builtin_}.
11087 GCC provides built-in versions of the ISO C99 floating-point comparison
11088 macros that avoid raising exceptions for unordered operands. They have
11089 the same names as the standard macros ( @code{isgreater},
11090 @code{isgreaterequal}, @code{isless}, @code{islessequal},
11091 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
11092 prefixed. We intend for a library implementor to be able to simply
11093 @code{#define} each standard macro to its built-in equivalent.
11094 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
11095 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
11096 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
11097 built-in functions appear both with and without the @code{__builtin_} prefix.
11099 @deftypefn {Built-in Function} void *__builtin_alloca (size_t size)
11100 The @code{__builtin_alloca} function must be called at block scope.
11101 The function allocates an object @var{size} bytes large on the stack
11102 of the calling function. The object is aligned on the default stack
11103 alignment boundary for the target determined by the
11104 @code{__BIGGEST_ALIGNMENT__} macro. The @code{__builtin_alloca}
11105 function returns a pointer to the first byte of the allocated object.
11106 The lifetime of the allocated object ends just before the calling
11107 function returns to its caller. This is so even when
11108 @code{__builtin_alloca} is called within a nested block.
11110 For example, the following function allocates eight objects of @code{n}
11111 bytes each on the stack, storing a pointer to each in consecutive elements
11112 of the array @code{a}. It then passes the array to function @code{g}
11113 which can safely use the storage pointed to by each of the array elements.
11116 void f (unsigned n)
11119 for (int i = 0; i != 8; ++i)
11120 a [i] = __builtin_alloca (n);
11122 g (a, n); // @r{safe}
11126 Since the @code{__builtin_alloca} function doesn't validate its argument
11127 it is the responsibility of its caller to make sure the argument doesn't
11128 cause it to exceed the stack size limit.
11129 The @code{__builtin_alloca} function is provided to make it possible to
11130 allocate on the stack arrays of bytes with an upper bound that may be
11131 computed at run time. Since C99 Variable Length Arrays offer
11132 similar functionality under a portable, more convenient, and safer
11133 interface they are recommended instead, in both C99 and C++ programs
11134 where GCC provides them as an extension.
11135 @xref{Variable Length}, for details.
11139 @deftypefn {Built-in Function} void *__builtin_alloca_with_align (size_t size, size_t alignment)
11140 The @code{__builtin_alloca_with_align} function must be called at block
11141 scope. The function allocates an object @var{size} bytes large on
11142 the stack of the calling function. The allocated object is aligned on
11143 the boundary specified by the argument @var{alignment} whose unit is given
11144 in bits (not bytes). The @var{size} argument must be positive and not
11145 exceed the stack size limit. The @var{alignment} argument must be a constant
11146 integer expression that evaluates to a power of 2 greater than or equal to
11147 @code{CHAR_BIT} and less than some unspecified maximum. Invocations
11148 with other values are rejected with an error indicating the valid bounds.
11149 The function returns a pointer to the first byte of the allocated object.
11150 The lifetime of the allocated object ends at the end of the block in which
11151 the function was called. The allocated storage is released no later than
11152 just before the calling function returns to its caller, but may be released
11153 at the end of the block in which the function was called.
11155 For example, in the following function the call to @code{g} is unsafe
11156 because when @code{overalign} is non-zero, the space allocated by
11157 @code{__builtin_alloca_with_align} may have been released at the end
11158 of the @code{if} statement in which it was called.
11161 void f (unsigned n, bool overalign)
11165 p = __builtin_alloca_with_align (n, 64 /* bits */);
11167 p = __builtin_alloc (n);
11169 g (p, n); // @r{unsafe}
11173 Since the @code{__builtin_alloca_with_align} function doesn't validate its
11174 @var{size} argument it is the responsibility of its caller to make sure
11175 the argument doesn't cause it to exceed the stack size limit.
11176 The @code{__builtin_alloca_with_align} function is provided to make
11177 it possible to allocate on the stack overaligned arrays of bytes with
11178 an upper bound that may be computed at run time. Since C99
11179 Variable Length Arrays offer the same functionality under
11180 a portable, more convenient, and safer interface they are recommended
11181 instead, in both C99 and C++ programs where GCC provides them as
11182 an extension. @xref{Variable Length}, for details.
11186 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
11188 You can use the built-in function @code{__builtin_types_compatible_p} to
11189 determine whether two types are the same.
11191 This built-in function returns 1 if the unqualified versions of the
11192 types @var{type1} and @var{type2} (which are types, not expressions) are
11193 compatible, 0 otherwise. The result of this built-in function can be
11194 used in integer constant expressions.
11196 This built-in function ignores top level qualifiers (e.g., @code{const},
11197 @code{volatile}). For example, @code{int} is equivalent to @code{const
11200 The type @code{int[]} and @code{int[5]} are compatible. On the other
11201 hand, @code{int} and @code{char *} are not compatible, even if the size
11202 of their types, on the particular architecture are the same. Also, the
11203 amount of pointer indirection is taken into account when determining
11204 similarity. Consequently, @code{short *} is not similar to
11205 @code{short **}. Furthermore, two types that are typedefed are
11206 considered compatible if their underlying types are compatible.
11208 An @code{enum} type is not considered to be compatible with another
11209 @code{enum} type even if both are compatible with the same integer
11210 type; this is what the C standard specifies.
11211 For example, @code{enum @{foo, bar@}} is not similar to
11212 @code{enum @{hot, dog@}}.
11214 You typically use this function in code whose execution varies
11215 depending on the arguments' types. For example:
11220 typeof (x) tmp = (x); \
11221 if (__builtin_types_compatible_p (typeof (x), long double)) \
11222 tmp = foo_long_double (tmp); \
11223 else if (__builtin_types_compatible_p (typeof (x), double)) \
11224 tmp = foo_double (tmp); \
11225 else if (__builtin_types_compatible_p (typeof (x), float)) \
11226 tmp = foo_float (tmp); \
11233 @emph{Note:} This construct is only available for C@.
11237 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
11239 The @var{call_exp} expression must be a function call, and the
11240 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
11241 is passed to the function call in the target's static chain location.
11242 The result of builtin is the result of the function call.
11244 @emph{Note:} This builtin is only available for C@.
11245 This builtin can be used to call Go closures from C.
11249 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
11251 You can use the built-in function @code{__builtin_choose_expr} to
11252 evaluate code depending on the value of a constant expression. This
11253 built-in function returns @var{exp1} if @var{const_exp}, which is an
11254 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
11256 This built-in function is analogous to the @samp{? :} operator in C,
11257 except that the expression returned has its type unaltered by promotion
11258 rules. Also, the built-in function does not evaluate the expression
11259 that is not chosen. For example, if @var{const_exp} evaluates to true,
11260 @var{exp2} is not evaluated even if it has side-effects.
11262 This built-in function can return an lvalue if the chosen argument is an
11265 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
11266 type. Similarly, if @var{exp2} is returned, its return type is the same
11273 __builtin_choose_expr ( \
11274 __builtin_types_compatible_p (typeof (x), double), \
11276 __builtin_choose_expr ( \
11277 __builtin_types_compatible_p (typeof (x), float), \
11279 /* @r{The void expression results in a compile-time error} \
11280 @r{when assigning the result to something.} */ \
11284 @emph{Note:} This construct is only available for C@. Furthermore, the
11285 unused expression (@var{exp1} or @var{exp2} depending on the value of
11286 @var{const_exp}) may still generate syntax errors. This may change in
11291 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
11293 The built-in function @code{__builtin_complex} is provided for use in
11294 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
11295 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
11296 real binary floating-point type, and the result has the corresponding
11297 complex type with real and imaginary parts @var{real} and @var{imag}.
11298 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
11299 infinities, NaNs and negative zeros are involved.
11303 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
11304 You can use the built-in function @code{__builtin_constant_p} to
11305 determine if a value is known to be constant at compile time and hence
11306 that GCC can perform constant-folding on expressions involving that
11307 value. The argument of the function is the value to test. The function
11308 returns the integer 1 if the argument is known to be a compile-time
11309 constant and 0 if it is not known to be a compile-time constant. A
11310 return of 0 does not indicate that the value is @emph{not} a constant,
11311 but merely that GCC cannot prove it is a constant with the specified
11312 value of the @option{-O} option.
11314 You typically use this function in an embedded application where
11315 memory is a critical resource. If you have some complex calculation,
11316 you may want it to be folded if it involves constants, but need to call
11317 a function if it does not. For example:
11320 #define Scale_Value(X) \
11321 (__builtin_constant_p (X) \
11322 ? ((X) * SCALE + OFFSET) : Scale (X))
11325 You may use this built-in function in either a macro or an inline
11326 function. However, if you use it in an inlined function and pass an
11327 argument of the function as the argument to the built-in, GCC
11328 never returns 1 when you call the inline function with a string constant
11329 or compound literal (@pxref{Compound Literals}) and does not return 1
11330 when you pass a constant numeric value to the inline function unless you
11331 specify the @option{-O} option.
11333 You may also use @code{__builtin_constant_p} in initializers for static
11334 data. For instance, you can write
11337 static const int table[] = @{
11338 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
11344 This is an acceptable initializer even if @var{EXPRESSION} is not a
11345 constant expression, including the case where
11346 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
11347 folded to a constant but @var{EXPRESSION} contains operands that are
11348 not otherwise permitted in a static initializer (for example,
11349 @code{0 && foo ()}). GCC must be more conservative about evaluating the
11350 built-in in this case, because it has no opportunity to perform
11354 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
11355 @opindex fprofile-arcs
11356 You may use @code{__builtin_expect} to provide the compiler with
11357 branch prediction information. In general, you should prefer to
11358 use actual profile feedback for this (@option{-fprofile-arcs}), as
11359 programmers are notoriously bad at predicting how their programs
11360 actually perform. However, there are applications in which this
11361 data is hard to collect.
11363 The return value is the value of @var{exp}, which should be an integral
11364 expression. The semantics of the built-in are that it is expected that
11365 @var{exp} == @var{c}. For example:
11368 if (__builtin_expect (x, 0))
11373 indicates that we do not expect to call @code{foo}, since
11374 we expect @code{x} to be zero. Since you are limited to integral
11375 expressions for @var{exp}, you should use constructions such as
11378 if (__builtin_expect (ptr != NULL, 1))
11383 when testing pointer or floating-point values.
11386 @deftypefn {Built-in Function} void __builtin_trap (void)
11387 This function causes the program to exit abnormally. GCC implements
11388 this function by using a target-dependent mechanism (such as
11389 intentionally executing an illegal instruction) or by calling
11390 @code{abort}. The mechanism used may vary from release to release so
11391 you should not rely on any particular implementation.
11394 @deftypefn {Built-in Function} void __builtin_unreachable (void)
11395 If control flow reaches the point of the @code{__builtin_unreachable},
11396 the program is undefined. It is useful in situations where the
11397 compiler cannot deduce the unreachability of the code.
11399 One such case is immediately following an @code{asm} statement that
11400 either never terminates, or one that transfers control elsewhere
11401 and never returns. In this example, without the
11402 @code{__builtin_unreachable}, GCC issues a warning that control
11403 reaches the end of a non-void function. It also generates code
11404 to return after the @code{asm}.
11407 int f (int c, int v)
11415 asm("jmp error_handler");
11416 __builtin_unreachable ();
11422 Because the @code{asm} statement unconditionally transfers control out
11423 of the function, control never reaches the end of the function
11424 body. The @code{__builtin_unreachable} is in fact unreachable and
11425 communicates this fact to the compiler.
11427 Another use for @code{__builtin_unreachable} is following a call a
11428 function that never returns but that is not declared
11429 @code{__attribute__((noreturn))}, as in this example:
11432 void function_that_never_returns (void);
11442 function_that_never_returns ();
11443 __builtin_unreachable ();
11450 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
11451 This function returns its first argument, and allows the compiler
11452 to assume that the returned pointer is at least @var{align} bytes
11453 aligned. This built-in can have either two or three arguments,
11454 if it has three, the third argument should have integer type, and
11455 if it is nonzero means misalignment offset. For example:
11458 void *x = __builtin_assume_aligned (arg, 16);
11462 means that the compiler can assume @code{x}, set to @code{arg}, is at least
11463 16-byte aligned, while:
11466 void *x = __builtin_assume_aligned (arg, 32, 8);
11470 means that the compiler can assume for @code{x}, set to @code{arg}, that
11471 @code{(char *) x - 8} is 32-byte aligned.
11474 @deftypefn {Built-in Function} int __builtin_LINE ()
11475 This function is the equivalent of the preprocessor @code{__LINE__}
11476 macro and returns a constant integer expression that evaluates to
11477 the line number of the invocation of the built-in. When used as a C++
11478 default argument for a function @var{F}, it returns the line number
11479 of the call to @var{F}.
11482 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
11483 This function is the equivalent of the @code{__FUNCTION__} symbol
11484 and returns an address constant pointing to the name of the function
11485 from which the built-in was invoked, or the empty string if
11486 the invocation is not at function scope. When used as a C++ default
11487 argument for a function @var{F}, it returns the name of @var{F}'s
11488 caller or the empty string if the call was not made at function
11492 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
11493 This function is the equivalent of the preprocessor @code{__FILE__}
11494 macro and returns an address constant pointing to the file name
11495 containing the invocation of the built-in, or the empty string if
11496 the invocation is not at function scope. When used as a C++ default
11497 argument for a function @var{F}, it returns the file name of the call
11498 to @var{F} or the empty string if the call was not made at function
11501 For example, in the following, each call to function @code{foo} will
11502 print a line similar to @code{"file.c:123: foo: message"} with the name
11503 of the file and the line number of the @code{printf} call, the name of
11504 the function @code{foo}, followed by the word @code{message}.
11508 function (const char *func = __builtin_FUNCTION ())
11515 printf ("%s:%i: %s: message\n", file (), line (), function ());
11521 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
11522 This function is used to flush the processor's instruction cache for
11523 the region of memory between @var{begin} inclusive and @var{end}
11524 exclusive. Some targets require that the instruction cache be
11525 flushed, after modifying memory containing code, in order to obtain
11526 deterministic behavior.
11528 If the target does not require instruction cache flushes,
11529 @code{__builtin___clear_cache} has no effect. Otherwise either
11530 instructions are emitted in-line to clear the instruction cache or a
11531 call to the @code{__clear_cache} function in libgcc is made.
11534 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
11535 This function is used to minimize cache-miss latency by moving data into
11536 a cache before it is accessed.
11537 You can insert calls to @code{__builtin_prefetch} into code for which
11538 you know addresses of data in memory that is likely to be accessed soon.
11539 If the target supports them, data prefetch instructions are generated.
11540 If the prefetch is done early enough before the access then the data will
11541 be in the cache by the time it is accessed.
11543 The value of @var{addr} is the address of the memory to prefetch.
11544 There are two optional arguments, @var{rw} and @var{locality}.
11545 The value of @var{rw} is a compile-time constant one or zero; one
11546 means that the prefetch is preparing for a write to the memory address
11547 and zero, the default, means that the prefetch is preparing for a read.
11548 The value @var{locality} must be a compile-time constant integer between
11549 zero and three. A value of zero means that the data has no temporal
11550 locality, so it need not be left in the cache after the access. A value
11551 of three means that the data has a high degree of temporal locality and
11552 should be left in all levels of cache possible. Values of one and two
11553 mean, respectively, a low or moderate degree of temporal locality. The
11557 for (i = 0; i < n; i++)
11559 a[i] = a[i] + b[i];
11560 __builtin_prefetch (&a[i+j], 1, 1);
11561 __builtin_prefetch (&b[i+j], 0, 1);
11566 Data prefetch does not generate faults if @var{addr} is invalid, but
11567 the address expression itself must be valid. For example, a prefetch
11568 of @code{p->next} does not fault if @code{p->next} is not a valid
11569 address, but evaluation faults if @code{p} is not a valid address.
11571 If the target does not support data prefetch, the address expression
11572 is evaluated if it includes side effects but no other code is generated
11573 and GCC does not issue a warning.
11576 @deftypefn {Built-in Function} double __builtin_huge_val (void)
11577 Returns a positive infinity, if supported by the floating-point format,
11578 else @code{DBL_MAX}. This function is suitable for implementing the
11579 ISO C macro @code{HUGE_VAL}.
11582 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
11583 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11586 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11587 Similar to @code{__builtin_huge_val}, except the return
11588 type is @code{long double}.
11591 @deftypefn {Built-in Function} _Float@var{n} __builtin_huge_valf@var{n} (void)
11592 Similar to @code{__builtin_huge_val}, except the return type is
11593 @code{_Float@var{n}}.
11596 @deftypefn {Built-in Function} _Float@var{n}x __builtin_huge_valf@var{n}x (void)
11597 Similar to @code{__builtin_huge_val}, except the return type is
11598 @code{_Float@var{n}x}.
11601 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11602 This built-in implements the C99 fpclassify functionality. The first
11603 five int arguments should be the target library's notion of the
11604 possible FP classes and are used for return values. They must be
11605 constant values and they must appear in this order: @code{FP_NAN},
11606 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11607 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11608 to classify. GCC treats the last argument as type-generic, which
11609 means it does not do default promotion from float to double.
11612 @deftypefn {Built-in Function} double __builtin_inf (void)
11613 Similar to @code{__builtin_huge_val}, except a warning is generated
11614 if the target floating-point format does not support infinities.
11617 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11618 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11621 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11622 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11625 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11626 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11629 @deftypefn {Built-in Function} float __builtin_inff (void)
11630 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11631 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11634 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11635 Similar to @code{__builtin_inf}, except the return
11636 type is @code{long double}.
11639 @deftypefn {Built-in Function} _Float@var{n} __builtin_inff@var{n} (void)
11640 Similar to @code{__builtin_inf}, except the return
11641 type is @code{_Float@var{n}}.
11644 @deftypefn {Built-in Function} _Float@var{n} __builtin_inff@var{n}x (void)
11645 Similar to @code{__builtin_inf}, except the return
11646 type is @code{_Float@var{n}x}.
11649 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11650 Similar to @code{isinf}, except the return value is -1 for
11651 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11652 Note while the parameter list is an
11653 ellipsis, this function only accepts exactly one floating-point
11654 argument. GCC treats this parameter as type-generic, which means it
11655 does not do default promotion from float to double.
11658 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11659 This is an implementation of the ISO C99 function @code{nan}.
11661 Since ISO C99 defines this function in terms of @code{strtod}, which we
11662 do not implement, a description of the parsing is in order. The string
11663 is parsed as by @code{strtol}; that is, the base is recognized by
11664 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11665 in the significand such that the least significant bit of the number
11666 is at the least significant bit of the significand. The number is
11667 truncated to fit the significand field provided. The significand is
11668 forced to be a quiet NaN@.
11670 This function, if given a string literal all of which would have been
11671 consumed by @code{strtol}, is evaluated early enough that it is considered a
11672 compile-time constant.
11675 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11676 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11679 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11680 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11683 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11684 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11687 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11688 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11691 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11692 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11695 @deftypefn {Built-in Function} _Float@var{n} __builtin_nanf@var{n} (const char *str)
11696 Similar to @code{__builtin_nan}, except the return type is
11697 @code{_Float@var{n}}.
11700 @deftypefn {Built-in Function} _Float@var{n}x __builtin_nanf@var{n}x (const char *str)
11701 Similar to @code{__builtin_nan}, except the return type is
11702 @code{_Float@var{n}x}.
11705 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11706 Similar to @code{__builtin_nan}, except the significand is forced
11707 to be a signaling NaN@. The @code{nans} function is proposed by
11708 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11711 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11712 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11715 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11716 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11719 @deftypefn {Built-in Function} _Float@var{n} __builtin_nansf@var{n} (const char *str)
11720 Similar to @code{__builtin_nans}, except the return type is
11721 @code{_Float@var{n}}.
11724 @deftypefn {Built-in Function} _Float@var{n}x __builtin_nansf@var{n}x (const char *str)
11725 Similar to @code{__builtin_nans}, except the return type is
11726 @code{_Float@var{n}x}.
11729 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11730 Returns one plus the index of the least significant 1-bit of @var{x}, or
11731 if @var{x} is zero, returns zero.
11734 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11735 Returns the number of leading 0-bits in @var{x}, starting at the most
11736 significant bit position. If @var{x} is 0, the result is undefined.
11739 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11740 Returns the number of trailing 0-bits in @var{x}, starting at the least
11741 significant bit position. If @var{x} is 0, the result is undefined.
11744 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11745 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11746 number of bits following the most significant bit that are identical
11747 to it. There are no special cases for 0 or other values.
11750 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11751 Returns the number of 1-bits in @var{x}.
11754 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11755 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11759 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11760 Similar to @code{__builtin_ffs}, except the argument type is
11764 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11765 Similar to @code{__builtin_clz}, except the argument type is
11766 @code{unsigned long}.
11769 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11770 Similar to @code{__builtin_ctz}, except the argument type is
11771 @code{unsigned long}.
11774 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11775 Similar to @code{__builtin_clrsb}, except the argument type is
11779 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11780 Similar to @code{__builtin_popcount}, except the argument type is
11781 @code{unsigned long}.
11784 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11785 Similar to @code{__builtin_parity}, except the argument type is
11786 @code{unsigned long}.
11789 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11790 Similar to @code{__builtin_ffs}, except the argument type is
11794 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11795 Similar to @code{__builtin_clz}, except the argument type is
11796 @code{unsigned long long}.
11799 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11800 Similar to @code{__builtin_ctz}, except the argument type is
11801 @code{unsigned long long}.
11804 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11805 Similar to @code{__builtin_clrsb}, except the argument type is
11809 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11810 Similar to @code{__builtin_popcount}, except the argument type is
11811 @code{unsigned long long}.
11814 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11815 Similar to @code{__builtin_parity}, except the argument type is
11816 @code{unsigned long long}.
11819 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11820 Returns the first argument raised to the power of the second. Unlike the
11821 @code{pow} function no guarantees about precision and rounding are made.
11824 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11825 Similar to @code{__builtin_powi}, except the argument and return types
11829 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11830 Similar to @code{__builtin_powi}, except the argument and return types
11831 are @code{long double}.
11834 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11835 Returns @var{x} with the order of the bytes reversed; for example,
11836 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11840 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11841 Similar to @code{__builtin_bswap16}, except the argument and return types
11845 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11846 Similar to @code{__builtin_bswap32}, except the argument and return types
11850 @node Target Builtins
11851 @section Built-in Functions Specific to Particular Target Machines
11853 On some target machines, GCC supports many built-in functions specific
11854 to those machines. Generally these generate calls to specific machine
11855 instructions, but allow the compiler to schedule those calls.
11858 * AArch64 Built-in Functions::
11859 * Alpha Built-in Functions::
11860 * Altera Nios II Built-in Functions::
11861 * ARC Built-in Functions::
11862 * ARC SIMD Built-in Functions::
11863 * ARM iWMMXt Built-in Functions::
11864 * ARM C Language Extensions (ACLE)::
11865 * ARM Floating Point Status and Control Intrinsics::
11866 * ARM ARMv8-M Security Extensions::
11867 * AVR Built-in Functions::
11868 * Blackfin Built-in Functions::
11869 * FR-V Built-in Functions::
11870 * MIPS DSP Built-in Functions::
11871 * MIPS Paired-Single Support::
11872 * MIPS Loongson Built-in Functions::
11873 * MIPS SIMD Architecture (MSA) Support::
11874 * Other MIPS Built-in Functions::
11875 * MSP430 Built-in Functions::
11876 * NDS32 Built-in Functions::
11877 * picoChip Built-in Functions::
11878 * PowerPC Built-in Functions::
11879 * PowerPC AltiVec/VSX Built-in Functions::
11880 * PowerPC Hardware Transactional Memory Built-in Functions::
11881 * RX Built-in Functions::
11882 * S/390 System z Built-in Functions::
11883 * SH Built-in Functions::
11884 * SPARC VIS Built-in Functions::
11885 * SPU Built-in Functions::
11886 * TI C6X Built-in Functions::
11887 * TILE-Gx Built-in Functions::
11888 * TILEPro Built-in Functions::
11889 * x86 Built-in Functions::
11890 * x86 transactional memory intrinsics::
11893 @node AArch64 Built-in Functions
11894 @subsection AArch64 Built-in Functions
11896 These built-in functions are available for the AArch64 family of
11899 unsigned int __builtin_aarch64_get_fpcr ()
11900 void __builtin_aarch64_set_fpcr (unsigned int)
11901 unsigned int __builtin_aarch64_get_fpsr ()
11902 void __builtin_aarch64_set_fpsr (unsigned int)
11905 @node Alpha Built-in Functions
11906 @subsection Alpha Built-in Functions
11908 These built-in functions are available for the Alpha family of
11909 processors, depending on the command-line switches used.
11911 The following built-in functions are always available. They
11912 all generate the machine instruction that is part of the name.
11915 long __builtin_alpha_implver (void)
11916 long __builtin_alpha_rpcc (void)
11917 long __builtin_alpha_amask (long)
11918 long __builtin_alpha_cmpbge (long, long)
11919 long __builtin_alpha_extbl (long, long)
11920 long __builtin_alpha_extwl (long, long)
11921 long __builtin_alpha_extll (long, long)
11922 long __builtin_alpha_extql (long, long)
11923 long __builtin_alpha_extwh (long, long)
11924 long __builtin_alpha_extlh (long, long)
11925 long __builtin_alpha_extqh (long, long)
11926 long __builtin_alpha_insbl (long, long)
11927 long __builtin_alpha_inswl (long, long)
11928 long __builtin_alpha_insll (long, long)
11929 long __builtin_alpha_insql (long, long)
11930 long __builtin_alpha_inswh (long, long)
11931 long __builtin_alpha_inslh (long, long)
11932 long __builtin_alpha_insqh (long, long)
11933 long __builtin_alpha_mskbl (long, long)
11934 long __builtin_alpha_mskwl (long, long)
11935 long __builtin_alpha_mskll (long, long)
11936 long __builtin_alpha_mskql (long, long)
11937 long __builtin_alpha_mskwh (long, long)
11938 long __builtin_alpha_msklh (long, long)
11939 long __builtin_alpha_mskqh (long, long)
11940 long __builtin_alpha_umulh (long, long)
11941 long __builtin_alpha_zap (long, long)
11942 long __builtin_alpha_zapnot (long, long)
11945 The following built-in functions are always with @option{-mmax}
11946 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11947 later. They all generate the machine instruction that is part
11951 long __builtin_alpha_pklb (long)
11952 long __builtin_alpha_pkwb (long)
11953 long __builtin_alpha_unpkbl (long)
11954 long __builtin_alpha_unpkbw (long)
11955 long __builtin_alpha_minub8 (long, long)
11956 long __builtin_alpha_minsb8 (long, long)
11957 long __builtin_alpha_minuw4 (long, long)
11958 long __builtin_alpha_minsw4 (long, long)
11959 long __builtin_alpha_maxub8 (long, long)
11960 long __builtin_alpha_maxsb8 (long, long)
11961 long __builtin_alpha_maxuw4 (long, long)
11962 long __builtin_alpha_maxsw4 (long, long)
11963 long __builtin_alpha_perr (long, long)
11966 The following built-in functions are always with @option{-mcix}
11967 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11968 later. They all generate the machine instruction that is part
11972 long __builtin_alpha_cttz (long)
11973 long __builtin_alpha_ctlz (long)
11974 long __builtin_alpha_ctpop (long)
11977 The following built-in functions are available on systems that use the OSF/1
11978 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11979 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11980 @code{rdval} and @code{wrval}.
11983 void *__builtin_thread_pointer (void)
11984 void __builtin_set_thread_pointer (void *)
11987 @node Altera Nios II Built-in Functions
11988 @subsection Altera Nios II Built-in Functions
11990 These built-in functions are available for the Altera Nios II
11991 family of processors.
11993 The following built-in functions are always available. They
11994 all generate the machine instruction that is part of the name.
11997 int __builtin_ldbio (volatile const void *)
11998 int __builtin_ldbuio (volatile const void *)
11999 int __builtin_ldhio (volatile const void *)
12000 int __builtin_ldhuio (volatile const void *)
12001 int __builtin_ldwio (volatile const void *)
12002 void __builtin_stbio (volatile void *, int)
12003 void __builtin_sthio (volatile void *, int)
12004 void __builtin_stwio (volatile void *, int)
12005 void __builtin_sync (void)
12006 int __builtin_rdctl (int)
12007 int __builtin_rdprs (int, int)
12008 void __builtin_wrctl (int, int)
12009 void __builtin_flushd (volatile void *)
12010 void __builtin_flushda (volatile void *)
12011 int __builtin_wrpie (int);
12012 void __builtin_eni (int);
12013 int __builtin_ldex (volatile const void *)
12014 int __builtin_stex (volatile void *, int)
12015 int __builtin_ldsex (volatile const void *)
12016 int __builtin_stsex (volatile void *, int)
12019 The following built-in functions are always available. They
12020 all generate a Nios II Custom Instruction. The name of the
12021 function represents the types that the function takes and
12022 returns. The letter before the @code{n} is the return type
12023 or void if absent. The @code{n} represents the first parameter
12024 to all the custom instructions, the custom instruction number.
12025 The two letters after the @code{n} represent the up to two
12026 parameters to the function.
12028 The letters represent the following data types:
12031 @code{void} for return type and no parameter for parameter types.
12034 @code{int} for return type and parameter type
12037 @code{float} for return type and parameter type
12040 @code{void *} for return type and parameter type
12044 And the function names are:
12046 void __builtin_custom_n (void)
12047 void __builtin_custom_ni (int)
12048 void __builtin_custom_nf (float)
12049 void __builtin_custom_np (void *)
12050 void __builtin_custom_nii (int, int)
12051 void __builtin_custom_nif (int, float)
12052 void __builtin_custom_nip (int, void *)
12053 void __builtin_custom_nfi (float, int)
12054 void __builtin_custom_nff (float, float)
12055 void __builtin_custom_nfp (float, void *)
12056 void __builtin_custom_npi (void *, int)
12057 void __builtin_custom_npf (void *, float)
12058 void __builtin_custom_npp (void *, void *)
12059 int __builtin_custom_in (void)
12060 int __builtin_custom_ini (int)
12061 int __builtin_custom_inf (float)
12062 int __builtin_custom_inp (void *)
12063 int __builtin_custom_inii (int, int)
12064 int __builtin_custom_inif (int, float)
12065 int __builtin_custom_inip (int, void *)
12066 int __builtin_custom_infi (float, int)
12067 int __builtin_custom_inff (float, float)
12068 int __builtin_custom_infp (float, void *)
12069 int __builtin_custom_inpi (void *, int)
12070 int __builtin_custom_inpf (void *, float)
12071 int __builtin_custom_inpp (void *, void *)
12072 float __builtin_custom_fn (void)
12073 float __builtin_custom_fni (int)
12074 float __builtin_custom_fnf (float)
12075 float __builtin_custom_fnp (void *)
12076 float __builtin_custom_fnii (int, int)
12077 float __builtin_custom_fnif (int, float)
12078 float __builtin_custom_fnip (int, void *)
12079 float __builtin_custom_fnfi (float, int)
12080 float __builtin_custom_fnff (float, float)
12081 float __builtin_custom_fnfp (float, void *)
12082 float __builtin_custom_fnpi (void *, int)
12083 float __builtin_custom_fnpf (void *, float)
12084 float __builtin_custom_fnpp (void *, void *)
12085 void * __builtin_custom_pn (void)
12086 void * __builtin_custom_pni (int)
12087 void * __builtin_custom_pnf (float)
12088 void * __builtin_custom_pnp (void *)
12089 void * __builtin_custom_pnii (int, int)
12090 void * __builtin_custom_pnif (int, float)
12091 void * __builtin_custom_pnip (int, void *)
12092 void * __builtin_custom_pnfi (float, int)
12093 void * __builtin_custom_pnff (float, float)
12094 void * __builtin_custom_pnfp (float, void *)
12095 void * __builtin_custom_pnpi (void *, int)
12096 void * __builtin_custom_pnpf (void *, float)
12097 void * __builtin_custom_pnpp (void *, void *)
12100 @node ARC Built-in Functions
12101 @subsection ARC Built-in Functions
12103 The following built-in functions are provided for ARC targets. The
12104 built-ins generate the corresponding assembly instructions. In the
12105 examples given below, the generated code often requires an operand or
12106 result to be in a register. Where necessary further code will be
12107 generated to ensure this is true, but for brevity this is not
12108 described in each case.
12110 @emph{Note:} Using a built-in to generate an instruction not supported
12111 by a target may cause problems. At present the compiler is not
12112 guaranteed to detect such misuse, and as a result an internal compiler
12113 error may be generated.
12115 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
12116 Return 1 if @var{val} is known to have the byte alignment given
12117 by @var{alignval}, otherwise return 0.
12118 Note that this is different from
12120 __alignof__(*(char *)@var{val}) >= alignval
12122 because __alignof__ sees only the type of the dereference, whereas
12123 __builtin_arc_align uses alignment information from the pointer
12124 as well as from the pointed-to type.
12125 The information available will depend on optimization level.
12128 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
12135 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
12136 The operand is the number of a register to be read. Generates:
12138 mov @var{dest}, r@var{regno}
12140 where the value in @var{dest} will be the result returned from the
12144 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
12145 The first operand is the number of a register to be written, the
12146 second operand is a compile time constant to write into that
12147 register. Generates:
12149 mov r@var{regno}, @var{val}
12153 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
12154 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
12157 divaw @var{dest}, @var{a}, @var{b}
12159 where the value in @var{dest} will be the result returned from the
12163 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
12170 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
12171 The operand, @var{auxv}, is the address of an auxiliary register and
12172 must be a compile time constant. Generates:
12174 lr @var{dest}, [@var{auxr}]
12176 Where the value in @var{dest} will be the result returned from the
12180 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
12181 Only available with @option{-mmul64}. Generates:
12183 mul64 @var{a}, @var{b}
12187 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
12188 Only available with @option{-mmul64}. Generates:
12190 mulu64 @var{a}, @var{b}
12194 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
12201 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
12202 Only valid if the @samp{norm} instruction is available through the
12203 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
12206 norm @var{dest}, @var{src}
12208 Where the value in @var{dest} will be the result returned from the
12212 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
12213 Only valid if the @samp{normw} instruction is available through the
12214 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
12217 normw @var{dest}, @var{src}
12219 Where the value in @var{dest} will be the result returned from the
12223 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
12230 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
12237 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
12238 The first argument, @var{auxv}, is the address of an auxiliary
12239 register, the second argument, @var{val}, is a compile time constant
12240 to be written to the register. Generates:
12242 sr @var{auxr}, [@var{val}]
12246 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
12247 Only valid with @option{-mswap}. Generates:
12249 swap @var{dest}, @var{src}
12251 Where the value in @var{dest} will be the result returned from the
12255 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
12262 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
12263 Only available with @option{-mcpu=ARC700}. Generates:
12269 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
12270 Only available with @option{-mcpu=ARC700}. Generates:
12276 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
12277 Only available with @option{-mcpu=ARC700}. Generates:
12283 The instructions generated by the following builtins are not
12284 considered as candidates for scheduling. They are not moved around by
12285 the compiler during scheduling, and thus can be expected to appear
12286 where they are put in the C code:
12288 __builtin_arc_brk()
12289 __builtin_arc_core_read()
12290 __builtin_arc_core_write()
12291 __builtin_arc_flag()
12293 __builtin_arc_sleep()
12295 __builtin_arc_swi()
12298 @node ARC SIMD Built-in Functions
12299 @subsection ARC SIMD Built-in Functions
12301 SIMD builtins provided by the compiler can be used to generate the
12302 vector instructions. This section describes the available builtins
12303 and their usage in programs. With the @option{-msimd} option, the
12304 compiler provides 128-bit vector types, which can be specified using
12305 the @code{vector_size} attribute. The header file @file{arc-simd.h}
12306 can be included to use the following predefined types:
12308 typedef int __v4si __attribute__((vector_size(16)));
12309 typedef short __v8hi __attribute__((vector_size(16)));
12312 These types can be used to define 128-bit variables. The built-in
12313 functions listed in the following section can be used on these
12314 variables to generate the vector operations.
12316 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
12317 @file{arc-simd.h} also provides equivalent macros called
12318 @code{_@var{someinsn}} that can be used for programming ease and
12319 improved readability. The following macros for DMA control are also
12322 #define _setup_dma_in_channel_reg _vdiwr
12323 #define _setup_dma_out_channel_reg _vdowr
12326 The following is a complete list of all the SIMD built-ins provided
12327 for ARC, grouped by calling signature.
12329 The following take two @code{__v8hi} arguments and return a
12330 @code{__v8hi} result:
12332 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
12333 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
12334 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
12335 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
12336 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
12337 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
12338 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
12339 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
12340 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
12341 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
12342 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
12343 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
12344 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
12345 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
12346 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
12347 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
12348 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
12349 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
12350 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
12351 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
12352 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
12353 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
12354 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
12355 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
12356 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
12357 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
12358 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
12359 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
12360 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
12361 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
12362 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
12363 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
12364 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
12365 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
12366 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
12367 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
12368 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
12369 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
12370 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
12371 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
12372 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
12373 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
12374 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
12375 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
12376 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
12377 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
12378 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
12379 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
12382 The following take one @code{__v8hi} and one @code{int} argument and return a
12383 @code{__v8hi} result:
12386 __v8hi __builtin_arc_vbaddw (__v8hi, int)
12387 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
12388 __v8hi __builtin_arc_vbminw (__v8hi, int)
12389 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
12390 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
12391 __v8hi __builtin_arc_vbmulw (__v8hi, int)
12392 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
12393 __v8hi __builtin_arc_vbsubw (__v8hi, int)
12396 The following take one @code{__v8hi} argument and one @code{int} argument which
12397 must be a 3-bit compile time constant indicating a register number
12398 I0-I7. They return a @code{__v8hi} result.
12400 __v8hi __builtin_arc_vasrw (__v8hi, const int)
12401 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
12402 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
12405 The following take one @code{__v8hi} argument and one @code{int}
12406 argument which must be a 6-bit compile time constant. They return a
12407 @code{__v8hi} result.
12409 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
12410 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
12411 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
12412 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
12413 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
12414 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
12415 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
12418 The following take one @code{__v8hi} argument and one @code{int} argument which
12419 must be a 8-bit compile time constant. They return a @code{__v8hi}
12422 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
12423 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
12424 __v8hi __builtin_arc_vmvw (__v8hi, const int)
12425 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
12428 The following take two @code{int} arguments, the second of which which
12429 must be a 8-bit compile time constant. They return a @code{__v8hi}
12432 __v8hi __builtin_arc_vmovaw (int, const int)
12433 __v8hi __builtin_arc_vmovw (int, const int)
12434 __v8hi __builtin_arc_vmovzw (int, const int)
12437 The following take a single @code{__v8hi} argument and return a
12438 @code{__v8hi} result:
12440 __v8hi __builtin_arc_vabsaw (__v8hi)
12441 __v8hi __builtin_arc_vabsw (__v8hi)
12442 __v8hi __builtin_arc_vaddsuw (__v8hi)
12443 __v8hi __builtin_arc_vexch1 (__v8hi)
12444 __v8hi __builtin_arc_vexch2 (__v8hi)
12445 __v8hi __builtin_arc_vexch4 (__v8hi)
12446 __v8hi __builtin_arc_vsignw (__v8hi)
12447 __v8hi __builtin_arc_vupbaw (__v8hi)
12448 __v8hi __builtin_arc_vupbw (__v8hi)
12449 __v8hi __builtin_arc_vupsbaw (__v8hi)
12450 __v8hi __builtin_arc_vupsbw (__v8hi)
12453 The following take two @code{int} arguments and return no result:
12455 void __builtin_arc_vdirun (int, int)
12456 void __builtin_arc_vdorun (int, int)
12459 The following take two @code{int} arguments and return no result. The
12460 first argument must a 3-bit compile time constant indicating one of
12461 the DR0-DR7 DMA setup channels:
12463 void __builtin_arc_vdiwr (const int, int)
12464 void __builtin_arc_vdowr (const int, int)
12467 The following take an @code{int} argument and return no result:
12469 void __builtin_arc_vendrec (int)
12470 void __builtin_arc_vrec (int)
12471 void __builtin_arc_vrecrun (int)
12472 void __builtin_arc_vrun (int)
12475 The following take a @code{__v8hi} argument and two @code{int}
12476 arguments and return a @code{__v8hi} result. The second argument must
12477 be a 3-bit compile time constants, indicating one the registers I0-I7,
12478 and the third argument must be an 8-bit compile time constant.
12480 @emph{Note:} Although the equivalent hardware instructions do not take
12481 an SIMD register as an operand, these builtins overwrite the relevant
12482 bits of the @code{__v8hi} register provided as the first argument with
12483 the value loaded from the @code{[Ib, u8]} location in the SDM.
12486 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
12487 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
12488 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
12489 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
12492 The following take two @code{int} arguments and return a @code{__v8hi}
12493 result. The first argument must be a 3-bit compile time constants,
12494 indicating one the registers I0-I7, and the second argument must be an
12495 8-bit compile time constant.
12498 __v8hi __builtin_arc_vld128 (const int, const int)
12499 __v8hi __builtin_arc_vld64w (const int, const int)
12502 The following take a @code{__v8hi} argument and two @code{int}
12503 arguments and return no result. The second argument must be a 3-bit
12504 compile time constants, indicating one the registers I0-I7, and the
12505 third argument must be an 8-bit compile time constant.
12508 void __builtin_arc_vst128 (__v8hi, const int, const int)
12509 void __builtin_arc_vst64 (__v8hi, const int, const int)
12512 The following take a @code{__v8hi} argument and three @code{int}
12513 arguments and return no result. The second argument must be a 3-bit
12514 compile-time constant, identifying the 16-bit sub-register to be
12515 stored, the third argument must be a 3-bit compile time constants,
12516 indicating one the registers I0-I7, and the fourth argument must be an
12517 8-bit compile time constant.
12520 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
12521 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
12524 @node ARM iWMMXt Built-in Functions
12525 @subsection ARM iWMMXt Built-in Functions
12527 These built-in functions are available for the ARM family of
12528 processors when the @option{-mcpu=iwmmxt} switch is used:
12531 typedef int v2si __attribute__ ((vector_size (8)));
12532 typedef short v4hi __attribute__ ((vector_size (8)));
12533 typedef char v8qi __attribute__ ((vector_size (8)));
12535 int __builtin_arm_getwcgr0 (void)
12536 void __builtin_arm_setwcgr0 (int)
12537 int __builtin_arm_getwcgr1 (void)
12538 void __builtin_arm_setwcgr1 (int)
12539 int __builtin_arm_getwcgr2 (void)
12540 void __builtin_arm_setwcgr2 (int)
12541 int __builtin_arm_getwcgr3 (void)
12542 void __builtin_arm_setwcgr3 (int)
12543 int __builtin_arm_textrmsb (v8qi, int)
12544 int __builtin_arm_textrmsh (v4hi, int)
12545 int __builtin_arm_textrmsw (v2si, int)
12546 int __builtin_arm_textrmub (v8qi, int)
12547 int __builtin_arm_textrmuh (v4hi, int)
12548 int __builtin_arm_textrmuw (v2si, int)
12549 v8qi __builtin_arm_tinsrb (v8qi, int, int)
12550 v4hi __builtin_arm_tinsrh (v4hi, int, int)
12551 v2si __builtin_arm_tinsrw (v2si, int, int)
12552 long long __builtin_arm_tmia (long long, int, int)
12553 long long __builtin_arm_tmiabb (long long, int, int)
12554 long long __builtin_arm_tmiabt (long long, int, int)
12555 long long __builtin_arm_tmiaph (long long, int, int)
12556 long long __builtin_arm_tmiatb (long long, int, int)
12557 long long __builtin_arm_tmiatt (long long, int, int)
12558 int __builtin_arm_tmovmskb (v8qi)
12559 int __builtin_arm_tmovmskh (v4hi)
12560 int __builtin_arm_tmovmskw (v2si)
12561 long long __builtin_arm_waccb (v8qi)
12562 long long __builtin_arm_wacch (v4hi)
12563 long long __builtin_arm_waccw (v2si)
12564 v8qi __builtin_arm_waddb (v8qi, v8qi)
12565 v8qi __builtin_arm_waddbss (v8qi, v8qi)
12566 v8qi __builtin_arm_waddbus (v8qi, v8qi)
12567 v4hi __builtin_arm_waddh (v4hi, v4hi)
12568 v4hi __builtin_arm_waddhss (v4hi, v4hi)
12569 v4hi __builtin_arm_waddhus (v4hi, v4hi)
12570 v2si __builtin_arm_waddw (v2si, v2si)
12571 v2si __builtin_arm_waddwss (v2si, v2si)
12572 v2si __builtin_arm_waddwus (v2si, v2si)
12573 v8qi __builtin_arm_walign (v8qi, v8qi, int)
12574 long long __builtin_arm_wand(long long, long long)
12575 long long __builtin_arm_wandn (long long, long long)
12576 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
12577 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
12578 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
12579 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
12580 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
12581 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
12582 v2si __builtin_arm_wcmpeqw (v2si, v2si)
12583 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
12584 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
12585 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
12586 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
12587 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
12588 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
12589 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
12590 long long __builtin_arm_wmacsz (v4hi, v4hi)
12591 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
12592 long long __builtin_arm_wmacuz (v4hi, v4hi)
12593 v4hi __builtin_arm_wmadds (v4hi, v4hi)
12594 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
12595 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
12596 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
12597 v2si __builtin_arm_wmaxsw (v2si, v2si)
12598 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
12599 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
12600 v2si __builtin_arm_wmaxuw (v2si, v2si)
12601 v8qi __builtin_arm_wminsb (v8qi, v8qi)
12602 v4hi __builtin_arm_wminsh (v4hi, v4hi)
12603 v2si __builtin_arm_wminsw (v2si, v2si)
12604 v8qi __builtin_arm_wminub (v8qi, v8qi)
12605 v4hi __builtin_arm_wminuh (v4hi, v4hi)
12606 v2si __builtin_arm_wminuw (v2si, v2si)
12607 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
12608 v4hi __builtin_arm_wmulul (v4hi, v4hi)
12609 v4hi __builtin_arm_wmulum (v4hi, v4hi)
12610 long long __builtin_arm_wor (long long, long long)
12611 v2si __builtin_arm_wpackdss (long long, long long)
12612 v2si __builtin_arm_wpackdus (long long, long long)
12613 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
12614 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
12615 v4hi __builtin_arm_wpackwss (v2si, v2si)
12616 v4hi __builtin_arm_wpackwus (v2si, v2si)
12617 long long __builtin_arm_wrord (long long, long long)
12618 long long __builtin_arm_wrordi (long long, int)
12619 v4hi __builtin_arm_wrorh (v4hi, long long)
12620 v4hi __builtin_arm_wrorhi (v4hi, int)
12621 v2si __builtin_arm_wrorw (v2si, long long)
12622 v2si __builtin_arm_wrorwi (v2si, int)
12623 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
12624 v2si __builtin_arm_wsadbz (v8qi, v8qi)
12625 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12626 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12627 v4hi __builtin_arm_wshufh (v4hi, int)
12628 long long __builtin_arm_wslld (long long, long long)
12629 long long __builtin_arm_wslldi (long long, int)
12630 v4hi __builtin_arm_wsllh (v4hi, long long)
12631 v4hi __builtin_arm_wsllhi (v4hi, int)
12632 v2si __builtin_arm_wsllw (v2si, long long)
12633 v2si __builtin_arm_wsllwi (v2si, int)
12634 long long __builtin_arm_wsrad (long long, long long)
12635 long long __builtin_arm_wsradi (long long, int)
12636 v4hi __builtin_arm_wsrah (v4hi, long long)
12637 v4hi __builtin_arm_wsrahi (v4hi, int)
12638 v2si __builtin_arm_wsraw (v2si, long long)
12639 v2si __builtin_arm_wsrawi (v2si, int)
12640 long long __builtin_arm_wsrld (long long, long long)
12641 long long __builtin_arm_wsrldi (long long, int)
12642 v4hi __builtin_arm_wsrlh (v4hi, long long)
12643 v4hi __builtin_arm_wsrlhi (v4hi, int)
12644 v2si __builtin_arm_wsrlw (v2si, long long)
12645 v2si __builtin_arm_wsrlwi (v2si, int)
12646 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12647 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12648 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12649 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12650 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12651 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12652 v2si __builtin_arm_wsubw (v2si, v2si)
12653 v2si __builtin_arm_wsubwss (v2si, v2si)
12654 v2si __builtin_arm_wsubwus (v2si, v2si)
12655 v4hi __builtin_arm_wunpckehsb (v8qi)
12656 v2si __builtin_arm_wunpckehsh (v4hi)
12657 long long __builtin_arm_wunpckehsw (v2si)
12658 v4hi __builtin_arm_wunpckehub (v8qi)
12659 v2si __builtin_arm_wunpckehuh (v4hi)
12660 long long __builtin_arm_wunpckehuw (v2si)
12661 v4hi __builtin_arm_wunpckelsb (v8qi)
12662 v2si __builtin_arm_wunpckelsh (v4hi)
12663 long long __builtin_arm_wunpckelsw (v2si)
12664 v4hi __builtin_arm_wunpckelub (v8qi)
12665 v2si __builtin_arm_wunpckeluh (v4hi)
12666 long long __builtin_arm_wunpckeluw (v2si)
12667 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12668 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12669 v2si __builtin_arm_wunpckihw (v2si, v2si)
12670 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12671 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12672 v2si __builtin_arm_wunpckilw (v2si, v2si)
12673 long long __builtin_arm_wxor (long long, long long)
12674 long long __builtin_arm_wzero ()
12678 @node ARM C Language Extensions (ACLE)
12679 @subsection ARM C Language Extensions (ACLE)
12681 GCC implements extensions for C as described in the ARM C Language
12682 Extensions (ACLE) specification, which can be found at
12683 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12685 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12686 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12687 intrinsics can be found at
12688 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12689 The built-in intrinsics for the Advanced SIMD extension are available when
12692 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12693 back ends support CRC32 intrinsics and the ARM back end supports the
12694 Coprocessor intrinsics, all from @file{arm_acle.h}. The ARM back end's 16-bit
12695 floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12696 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12699 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12700 availability of extensions.
12702 @node ARM Floating Point Status and Control Intrinsics
12703 @subsection ARM Floating Point Status and Control Intrinsics
12705 These built-in functions are available for the ARM family of
12706 processors with floating-point unit.
12709 unsigned int __builtin_arm_get_fpscr ()
12710 void __builtin_arm_set_fpscr (unsigned int)
12713 @node ARM ARMv8-M Security Extensions
12714 @subsection ARM ARMv8-M Security Extensions
12716 GCC implements the ARMv8-M Security Extensions as described in the ARMv8-M
12717 Security Extensions: Requirements on Development Tools Engineering
12718 Specification, which can be found at
12719 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
12721 As part of the Security Extensions GCC implements two new function attributes:
12722 @code{cmse_nonsecure_entry} and @code{cmse_nonsecure_call}.
12724 As part of the Security Extensions GCC implements the intrinsics below. FPTR
12725 is used here to mean any function pointer type.
12728 cmse_address_info_t cmse_TT (void *)
12729 cmse_address_info_t cmse_TT_fptr (FPTR)
12730 cmse_address_info_t cmse_TTT (void *)
12731 cmse_address_info_t cmse_TTT_fptr (FPTR)
12732 cmse_address_info_t cmse_TTA (void *)
12733 cmse_address_info_t cmse_TTA_fptr (FPTR)
12734 cmse_address_info_t cmse_TTAT (void *)
12735 cmse_address_info_t cmse_TTAT_fptr (FPTR)
12736 void * cmse_check_address_range (void *, size_t, int)
12737 typeof(p) cmse_nsfptr_create (FPTR p)
12738 intptr_t cmse_is_nsfptr (FPTR)
12739 int cmse_nonsecure_caller (void)
12742 @node AVR Built-in Functions
12743 @subsection AVR Built-in Functions
12745 For each built-in function for AVR, there is an equally named,
12746 uppercase built-in macro defined. That way users can easily query if
12747 or if not a specific built-in is implemented or not. For example, if
12748 @code{__builtin_avr_nop} is available the macro
12749 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12751 The following built-in functions map to the respective machine
12752 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12753 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12754 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12755 as library call if no hardware multiplier is available.
12758 void __builtin_avr_nop (void)
12759 void __builtin_avr_sei (void)
12760 void __builtin_avr_cli (void)
12761 void __builtin_avr_sleep (void)
12762 void __builtin_avr_wdr (void)
12763 unsigned char __builtin_avr_swap (unsigned char)
12764 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12765 int __builtin_avr_fmuls (char, char)
12766 int __builtin_avr_fmulsu (char, unsigned char)
12769 In order to delay execution for a specific number of cycles, GCC
12772 void __builtin_avr_delay_cycles (unsigned long ticks)
12776 @code{ticks} is the number of ticks to delay execution. Note that this
12777 built-in does not take into account the effect of interrupts that
12778 might increase delay time. @code{ticks} must be a compile-time
12779 integer constant; delays with a variable number of cycles are not supported.
12782 char __builtin_avr_flash_segment (const __memx void*)
12786 This built-in takes a byte address to the 24-bit
12787 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12788 the number of the flash segment (the 64 KiB chunk) where the address
12789 points to. Counting starts at @code{0}.
12790 If the address does not point to flash memory, return @code{-1}.
12793 unsigned char __builtin_avr_insert_bits (unsigned long map,
12794 unsigned char bits,
12799 Insert bits from @var{bits} into @var{val} and return the resulting
12800 value. The nibbles of @var{map} determine how the insertion is
12801 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12803 @item If @var{X} is @code{0xf},
12804 then the @var{n}-th bit of @var{val} is returned unaltered.
12806 @item If X is in the range 0@dots{}7,
12807 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12809 @item If X is in the range 8@dots{}@code{0xe},
12810 then the @var{n}-th result bit is undefined.
12814 One typical use case for this built-in is adjusting input and
12815 output values to non-contiguous port layouts. Some examples:
12818 // same as val, bits is unused
12819 __builtin_avr_insert_bits (0xffffffff, bits, val)
12823 // same as bits, val is unused
12824 __builtin_avr_insert_bits (0x76543210, bits, val)
12828 // same as rotating bits by 4
12829 __builtin_avr_insert_bits (0x32107654, bits, 0)
12833 // high nibble of result is the high nibble of val
12834 // low nibble of result is the low nibble of bits
12835 __builtin_avr_insert_bits (0xffff3210, bits, val)
12839 // reverse the bit order of bits
12840 __builtin_avr_insert_bits (0x01234567, bits, 0)
12844 void __builtin_avr_nops (unsigned count)
12848 Insert @code{count} @code{NOP} instructions.
12849 The number of instructions must be a compile-time integer constant.
12851 @node Blackfin Built-in Functions
12852 @subsection Blackfin Built-in Functions
12854 Currently, there are two Blackfin-specific built-in functions. These are
12855 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12856 using inline assembly; by using these built-in functions the compiler can
12857 automatically add workarounds for hardware errata involving these
12858 instructions. These functions are named as follows:
12861 void __builtin_bfin_csync (void)
12862 void __builtin_bfin_ssync (void)
12865 @node FR-V Built-in Functions
12866 @subsection FR-V Built-in Functions
12868 GCC provides many FR-V-specific built-in functions. In general,
12869 these functions are intended to be compatible with those described
12870 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12871 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12872 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12873 pointer rather than by value.
12875 Most of the functions are named after specific FR-V instructions.
12876 Such functions are said to be ``directly mapped'' and are summarized
12877 here in tabular form.
12881 * Directly-mapped Integer Functions::
12882 * Directly-mapped Media Functions::
12883 * Raw read/write Functions::
12884 * Other Built-in Functions::
12887 @node Argument Types
12888 @subsubsection Argument Types
12890 The arguments to the built-in functions can be divided into three groups:
12891 register numbers, compile-time constants and run-time values. In order
12892 to make this classification clear at a glance, the arguments and return
12893 values are given the following pseudo types:
12895 @multitable @columnfractions .20 .30 .15 .35
12896 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12897 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12898 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12899 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12900 @item @code{uw2} @tab @code{unsigned long long} @tab No
12901 @tab an unsigned doubleword
12902 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12903 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12904 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12905 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12908 These pseudo types are not defined by GCC, they are simply a notational
12909 convenience used in this manual.
12911 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12912 and @code{sw2} are evaluated at run time. They correspond to
12913 register operands in the underlying FR-V instructions.
12915 @code{const} arguments represent immediate operands in the underlying
12916 FR-V instructions. They must be compile-time constants.
12918 @code{acc} arguments are evaluated at compile time and specify the number
12919 of an accumulator register. For example, an @code{acc} argument of 2
12920 selects the ACC2 register.
12922 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12923 number of an IACC register. See @pxref{Other Built-in Functions}
12926 @node Directly-mapped Integer Functions
12927 @subsubsection Directly-Mapped Integer Functions
12929 The functions listed below map directly to FR-V I-type instructions.
12931 @multitable @columnfractions .45 .32 .23
12932 @item Function prototype @tab Example usage @tab Assembly output
12933 @item @code{sw1 __ADDSS (sw1, sw1)}
12934 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12935 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12936 @item @code{sw1 __SCAN (sw1, sw1)}
12937 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12938 @tab @code{SCAN @var{a},@var{b},@var{c}}
12939 @item @code{sw1 __SCUTSS (sw1)}
12940 @tab @code{@var{b} = __SCUTSS (@var{a})}
12941 @tab @code{SCUTSS @var{a},@var{b}}
12942 @item @code{sw1 __SLASS (sw1, sw1)}
12943 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12944 @tab @code{SLASS @var{a},@var{b},@var{c}}
12945 @item @code{void __SMASS (sw1, sw1)}
12946 @tab @code{__SMASS (@var{a}, @var{b})}
12947 @tab @code{SMASS @var{a},@var{b}}
12948 @item @code{void __SMSSS (sw1, sw1)}
12949 @tab @code{__SMSSS (@var{a}, @var{b})}
12950 @tab @code{SMSSS @var{a},@var{b}}
12951 @item @code{void __SMU (sw1, sw1)}
12952 @tab @code{__SMU (@var{a}, @var{b})}
12953 @tab @code{SMU @var{a},@var{b}}
12954 @item @code{sw2 __SMUL (sw1, sw1)}
12955 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12956 @tab @code{SMUL @var{a},@var{b},@var{c}}
12957 @item @code{sw1 __SUBSS (sw1, sw1)}
12958 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12959 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12960 @item @code{uw2 __UMUL (uw1, uw1)}
12961 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12962 @tab @code{UMUL @var{a},@var{b},@var{c}}
12965 @node Directly-mapped Media Functions
12966 @subsubsection Directly-Mapped Media Functions
12968 The functions listed below map directly to FR-V M-type instructions.
12970 @multitable @columnfractions .45 .32 .23
12971 @item Function prototype @tab Example usage @tab Assembly output
12972 @item @code{uw1 __MABSHS (sw1)}
12973 @tab @code{@var{b} = __MABSHS (@var{a})}
12974 @tab @code{MABSHS @var{a},@var{b}}
12975 @item @code{void __MADDACCS (acc, acc)}
12976 @tab @code{__MADDACCS (@var{b}, @var{a})}
12977 @tab @code{MADDACCS @var{a},@var{b}}
12978 @item @code{sw1 __MADDHSS (sw1, sw1)}
12979 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12980 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12981 @item @code{uw1 __MADDHUS (uw1, uw1)}
12982 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12983 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12984 @item @code{uw1 __MAND (uw1, uw1)}
12985 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12986 @tab @code{MAND @var{a},@var{b},@var{c}}
12987 @item @code{void __MASACCS (acc, acc)}
12988 @tab @code{__MASACCS (@var{b}, @var{a})}
12989 @tab @code{MASACCS @var{a},@var{b}}
12990 @item @code{uw1 __MAVEH (uw1, uw1)}
12991 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12992 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12993 @item @code{uw2 __MBTOH (uw1)}
12994 @tab @code{@var{b} = __MBTOH (@var{a})}
12995 @tab @code{MBTOH @var{a},@var{b}}
12996 @item @code{void __MBTOHE (uw1 *, uw1)}
12997 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12998 @tab @code{MBTOHE @var{a},@var{b}}
12999 @item @code{void __MCLRACC (acc)}
13000 @tab @code{__MCLRACC (@var{a})}
13001 @tab @code{MCLRACC @var{a}}
13002 @item @code{void __MCLRACCA (void)}
13003 @tab @code{__MCLRACCA ()}
13004 @tab @code{MCLRACCA}
13005 @item @code{uw1 __Mcop1 (uw1, uw1)}
13006 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
13007 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
13008 @item @code{uw1 __Mcop2 (uw1, uw1)}
13009 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
13010 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
13011 @item @code{uw1 __MCPLHI (uw2, const)}
13012 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
13013 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
13014 @item @code{uw1 __MCPLI (uw2, const)}
13015 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
13016 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
13017 @item @code{void __MCPXIS (acc, sw1, sw1)}
13018 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
13019 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
13020 @item @code{void __MCPXIU (acc, uw1, uw1)}
13021 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
13022 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
13023 @item @code{void __MCPXRS (acc, sw1, sw1)}
13024 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
13025 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
13026 @item @code{void __MCPXRU (acc, uw1, uw1)}
13027 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
13028 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
13029 @item @code{uw1 __MCUT (acc, uw1)}
13030 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
13031 @tab @code{MCUT @var{a},@var{b},@var{c}}
13032 @item @code{uw1 __MCUTSS (acc, sw1)}
13033 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
13034 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
13035 @item @code{void __MDADDACCS (acc, acc)}
13036 @tab @code{__MDADDACCS (@var{b}, @var{a})}
13037 @tab @code{MDADDACCS @var{a},@var{b}}
13038 @item @code{void __MDASACCS (acc, acc)}
13039 @tab @code{__MDASACCS (@var{b}, @var{a})}
13040 @tab @code{MDASACCS @var{a},@var{b}}
13041 @item @code{uw2 __MDCUTSSI (acc, const)}
13042 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
13043 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
13044 @item @code{uw2 __MDPACKH (uw2, uw2)}
13045 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
13046 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
13047 @item @code{uw2 __MDROTLI (uw2, const)}
13048 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
13049 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
13050 @item @code{void __MDSUBACCS (acc, acc)}
13051 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
13052 @tab @code{MDSUBACCS @var{a},@var{b}}
13053 @item @code{void __MDUNPACKH (uw1 *, uw2)}
13054 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
13055 @tab @code{MDUNPACKH @var{a},@var{b}}
13056 @item @code{uw2 __MEXPDHD (uw1, const)}
13057 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
13058 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
13059 @item @code{uw1 __MEXPDHW (uw1, const)}
13060 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
13061 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
13062 @item @code{uw1 __MHDSETH (uw1, const)}
13063 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
13064 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
13065 @item @code{sw1 __MHDSETS (const)}
13066 @tab @code{@var{b} = __MHDSETS (@var{a})}
13067 @tab @code{MHDSETS #@var{a},@var{b}}
13068 @item @code{uw1 __MHSETHIH (uw1, const)}
13069 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
13070 @tab @code{MHSETHIH #@var{a},@var{b}}
13071 @item @code{sw1 __MHSETHIS (sw1, const)}
13072 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
13073 @tab @code{MHSETHIS #@var{a},@var{b}}
13074 @item @code{uw1 __MHSETLOH (uw1, const)}
13075 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
13076 @tab @code{MHSETLOH #@var{a},@var{b}}
13077 @item @code{sw1 __MHSETLOS (sw1, const)}
13078 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
13079 @tab @code{MHSETLOS #@var{a},@var{b}}
13080 @item @code{uw1 __MHTOB (uw2)}
13081 @tab @code{@var{b} = __MHTOB (@var{a})}
13082 @tab @code{MHTOB @var{a},@var{b}}
13083 @item @code{void __MMACHS (acc, sw1, sw1)}
13084 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
13085 @tab @code{MMACHS @var{a},@var{b},@var{c}}
13086 @item @code{void __MMACHU (acc, uw1, uw1)}
13087 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
13088 @tab @code{MMACHU @var{a},@var{b},@var{c}}
13089 @item @code{void __MMRDHS (acc, sw1, sw1)}
13090 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
13091 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
13092 @item @code{void __MMRDHU (acc, uw1, uw1)}
13093 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
13094 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
13095 @item @code{void __MMULHS (acc, sw1, sw1)}
13096 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
13097 @tab @code{MMULHS @var{a},@var{b},@var{c}}
13098 @item @code{void __MMULHU (acc, uw1, uw1)}
13099 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
13100 @tab @code{MMULHU @var{a},@var{b},@var{c}}
13101 @item @code{void __MMULXHS (acc, sw1, sw1)}
13102 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
13103 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
13104 @item @code{void __MMULXHU (acc, uw1, uw1)}
13105 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
13106 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
13107 @item @code{uw1 __MNOT (uw1)}
13108 @tab @code{@var{b} = __MNOT (@var{a})}
13109 @tab @code{MNOT @var{a},@var{b}}
13110 @item @code{uw1 __MOR (uw1, uw1)}
13111 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
13112 @tab @code{MOR @var{a},@var{b},@var{c}}
13113 @item @code{uw1 __MPACKH (uh, uh)}
13114 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
13115 @tab @code{MPACKH @var{a},@var{b},@var{c}}
13116 @item @code{sw2 __MQADDHSS (sw2, sw2)}
13117 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
13118 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
13119 @item @code{uw2 __MQADDHUS (uw2, uw2)}
13120 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
13121 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
13122 @item @code{void __MQCPXIS (acc, sw2, sw2)}
13123 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
13124 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
13125 @item @code{void __MQCPXIU (acc, uw2, uw2)}
13126 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
13127 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
13128 @item @code{void __MQCPXRS (acc, sw2, sw2)}
13129 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
13130 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
13131 @item @code{void __MQCPXRU (acc, uw2, uw2)}
13132 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
13133 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
13134 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
13135 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
13136 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
13137 @item @code{sw2 __MQLMTHS (sw2, sw2)}
13138 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
13139 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
13140 @item @code{void __MQMACHS (acc, sw2, sw2)}
13141 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
13142 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
13143 @item @code{void __MQMACHU (acc, uw2, uw2)}
13144 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
13145 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
13146 @item @code{void __MQMACXHS (acc, sw2, sw2)}
13147 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
13148 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
13149 @item @code{void __MQMULHS (acc, sw2, sw2)}
13150 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
13151 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
13152 @item @code{void __MQMULHU (acc, uw2, uw2)}
13153 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
13154 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
13155 @item @code{void __MQMULXHS (acc, sw2, sw2)}
13156 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
13157 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
13158 @item @code{void __MQMULXHU (acc, uw2, uw2)}
13159 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
13160 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
13161 @item @code{sw2 __MQSATHS (sw2, sw2)}
13162 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
13163 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
13164 @item @code{uw2 __MQSLLHI (uw2, int)}
13165 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
13166 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
13167 @item @code{sw2 __MQSRAHI (sw2, int)}
13168 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
13169 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
13170 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
13171 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
13172 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
13173 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
13174 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
13175 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
13176 @item @code{void __MQXMACHS (acc, sw2, sw2)}
13177 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
13178 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
13179 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
13180 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
13181 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
13182 @item @code{uw1 __MRDACC (acc)}
13183 @tab @code{@var{b} = __MRDACC (@var{a})}
13184 @tab @code{MRDACC @var{a},@var{b}}
13185 @item @code{uw1 __MRDACCG (acc)}
13186 @tab @code{@var{b} = __MRDACCG (@var{a})}
13187 @tab @code{MRDACCG @var{a},@var{b}}
13188 @item @code{uw1 __MROTLI (uw1, const)}
13189 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
13190 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
13191 @item @code{uw1 __MROTRI (uw1, const)}
13192 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
13193 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
13194 @item @code{sw1 __MSATHS (sw1, sw1)}
13195 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
13196 @tab @code{MSATHS @var{a},@var{b},@var{c}}
13197 @item @code{uw1 __MSATHU (uw1, uw1)}
13198 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
13199 @tab @code{MSATHU @var{a},@var{b},@var{c}}
13200 @item @code{uw1 __MSLLHI (uw1, const)}
13201 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
13202 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
13203 @item @code{sw1 __MSRAHI (sw1, const)}
13204 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
13205 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
13206 @item @code{uw1 __MSRLHI (uw1, const)}
13207 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
13208 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
13209 @item @code{void __MSUBACCS (acc, acc)}
13210 @tab @code{__MSUBACCS (@var{b}, @var{a})}
13211 @tab @code{MSUBACCS @var{a},@var{b}}
13212 @item @code{sw1 __MSUBHSS (sw1, sw1)}
13213 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
13214 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
13215 @item @code{uw1 __MSUBHUS (uw1, uw1)}
13216 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
13217 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
13218 @item @code{void __MTRAP (void)}
13219 @tab @code{__MTRAP ()}
13221 @item @code{uw2 __MUNPACKH (uw1)}
13222 @tab @code{@var{b} = __MUNPACKH (@var{a})}
13223 @tab @code{MUNPACKH @var{a},@var{b}}
13224 @item @code{uw1 __MWCUT (uw2, uw1)}
13225 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
13226 @tab @code{MWCUT @var{a},@var{b},@var{c}}
13227 @item @code{void __MWTACC (acc, uw1)}
13228 @tab @code{__MWTACC (@var{b}, @var{a})}
13229 @tab @code{MWTACC @var{a},@var{b}}
13230 @item @code{void __MWTACCG (acc, uw1)}
13231 @tab @code{__MWTACCG (@var{b}, @var{a})}
13232 @tab @code{MWTACCG @var{a},@var{b}}
13233 @item @code{uw1 __MXOR (uw1, uw1)}
13234 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
13235 @tab @code{MXOR @var{a},@var{b},@var{c}}
13238 @node Raw read/write Functions
13239 @subsubsection Raw Read/Write Functions
13241 This sections describes built-in functions related to read and write
13242 instructions to access memory. These functions generate
13243 @code{membar} instructions to flush the I/O load and stores where
13244 appropriate, as described in Fujitsu's manual described above.
13248 @item unsigned char __builtin_read8 (void *@var{data})
13249 @item unsigned short __builtin_read16 (void *@var{data})
13250 @item unsigned long __builtin_read32 (void *@var{data})
13251 @item unsigned long long __builtin_read64 (void *@var{data})
13253 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
13254 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
13255 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
13256 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
13259 @node Other Built-in Functions
13260 @subsubsection Other Built-in Functions
13262 This section describes built-in functions that are not named after
13263 a specific FR-V instruction.
13266 @item sw2 __IACCreadll (iacc @var{reg})
13267 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
13268 for future expansion and must be 0.
13270 @item sw1 __IACCreadl (iacc @var{reg})
13271 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
13272 Other values of @var{reg} are rejected as invalid.
13274 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
13275 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
13276 is reserved for future expansion and must be 0.
13278 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
13279 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
13280 is 1. Other values of @var{reg} are rejected as invalid.
13282 @item void __data_prefetch0 (const void *@var{x})
13283 Use the @code{dcpl} instruction to load the contents of address @var{x}
13284 into the data cache.
13286 @item void __data_prefetch (const void *@var{x})
13287 Use the @code{nldub} instruction to load the contents of address @var{x}
13288 into the data cache. The instruction is issued in slot I1@.
13291 @node MIPS DSP Built-in Functions
13292 @subsection MIPS DSP Built-in Functions
13294 The MIPS DSP Application-Specific Extension (ASE) includes new
13295 instructions that are designed to improve the performance of DSP and
13296 media applications. It provides instructions that operate on packed
13297 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
13299 GCC supports MIPS DSP operations using both the generic
13300 vector extensions (@pxref{Vector Extensions}) and a collection of
13301 MIPS-specific built-in functions. Both kinds of support are
13302 enabled by the @option{-mdsp} command-line option.
13304 Revision 2 of the ASE was introduced in the second half of 2006.
13305 This revision adds extra instructions to the original ASE, but is
13306 otherwise backwards-compatible with it. You can select revision 2
13307 using the command-line option @option{-mdspr2}; this option implies
13310 The SCOUNT and POS bits of the DSP control register are global. The
13311 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
13312 POS bits. During optimization, the compiler does not delete these
13313 instructions and it does not delete calls to functions containing
13314 these instructions.
13316 At present, GCC only provides support for operations on 32-bit
13317 vectors. The vector type associated with 8-bit integer data is
13318 usually called @code{v4i8}, the vector type associated with Q7
13319 is usually called @code{v4q7}, the vector type associated with 16-bit
13320 integer data is usually called @code{v2i16}, and the vector type
13321 associated with Q15 is usually called @code{v2q15}. They can be
13322 defined in C as follows:
13325 typedef signed char v4i8 __attribute__ ((vector_size(4)));
13326 typedef signed char v4q7 __attribute__ ((vector_size(4)));
13327 typedef short v2i16 __attribute__ ((vector_size(4)));
13328 typedef short v2q15 __attribute__ ((vector_size(4)));
13331 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
13332 initialized in the same way as aggregates. For example:
13335 v4i8 a = @{1, 2, 3, 4@};
13337 b = (v4i8) @{5, 6, 7, 8@};
13339 v2q15 c = @{0x0fcb, 0x3a75@};
13341 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
13344 @emph{Note:} The CPU's endianness determines the order in which values
13345 are packed. On little-endian targets, the first value is the least
13346 significant and the last value is the most significant. The opposite
13347 order applies to big-endian targets. For example, the code above
13348 sets the lowest byte of @code{a} to @code{1} on little-endian targets
13349 and @code{4} on big-endian targets.
13351 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
13352 representation. As shown in this example, the integer representation
13353 of a Q7 value can be obtained by multiplying the fractional value by
13354 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
13355 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
13358 The table below lists the @code{v4i8} and @code{v2q15} operations for which
13359 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
13360 and @code{c} and @code{d} are @code{v2q15} values.
13362 @multitable @columnfractions .50 .50
13363 @item C code @tab MIPS instruction
13364 @item @code{a + b} @tab @code{addu.qb}
13365 @item @code{c + d} @tab @code{addq.ph}
13366 @item @code{a - b} @tab @code{subu.qb}
13367 @item @code{c - d} @tab @code{subq.ph}
13370 The table below lists the @code{v2i16} operation for which
13371 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
13372 @code{v2i16} values.
13374 @multitable @columnfractions .50 .50
13375 @item C code @tab MIPS instruction
13376 @item @code{e * f} @tab @code{mul.ph}
13379 It is easier to describe the DSP built-in functions if we first define
13380 the following types:
13385 typedef unsigned int ui32;
13386 typedef long long a64;
13389 @code{q31} and @code{i32} are actually the same as @code{int}, but we
13390 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
13391 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
13392 @code{long long}, but we use @code{a64} to indicate values that are
13393 placed in one of the four DSP accumulators (@code{$ac0},
13394 @code{$ac1}, @code{$ac2} or @code{$ac3}).
13396 Also, some built-in functions prefer or require immediate numbers as
13397 parameters, because the corresponding DSP instructions accept both immediate
13398 numbers and register operands, or accept immediate numbers only. The
13399 immediate parameters are listed as follows.
13407 imm0_255: 0 to 255.
13408 imm_n32_31: -32 to 31.
13409 imm_n512_511: -512 to 511.
13412 The following built-in functions map directly to a particular MIPS DSP
13413 instruction. Please refer to the architecture specification
13414 for details on what each instruction does.
13417 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
13418 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
13419 q31 __builtin_mips_addq_s_w (q31, q31)
13420 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
13421 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
13422 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
13423 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
13424 q31 __builtin_mips_subq_s_w (q31, q31)
13425 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
13426 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
13427 i32 __builtin_mips_addsc (i32, i32)
13428 i32 __builtin_mips_addwc (i32, i32)
13429 i32 __builtin_mips_modsub (i32, i32)
13430 i32 __builtin_mips_raddu_w_qb (v4i8)
13431 v2q15 __builtin_mips_absq_s_ph (v2q15)
13432 q31 __builtin_mips_absq_s_w (q31)
13433 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
13434 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
13435 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
13436 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
13437 q31 __builtin_mips_preceq_w_phl (v2q15)
13438 q31 __builtin_mips_preceq_w_phr (v2q15)
13439 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
13440 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
13441 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
13442 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
13443 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
13444 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
13445 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
13446 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
13447 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
13448 v4i8 __builtin_mips_shll_qb (v4i8, i32)
13449 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
13450 v2q15 __builtin_mips_shll_ph (v2q15, i32)
13451 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
13452 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
13453 q31 __builtin_mips_shll_s_w (q31, imm0_31)
13454 q31 __builtin_mips_shll_s_w (q31, i32)
13455 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
13456 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
13457 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
13458 v2q15 __builtin_mips_shra_ph (v2q15, i32)
13459 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
13460 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
13461 q31 __builtin_mips_shra_r_w (q31, imm0_31)
13462 q31 __builtin_mips_shra_r_w (q31, i32)
13463 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
13464 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
13465 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
13466 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
13467 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
13468 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
13469 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
13470 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
13471 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
13472 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
13473 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
13474 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
13475 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
13476 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
13477 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
13478 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
13479 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
13480 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
13481 i32 __builtin_mips_bitrev (i32)
13482 i32 __builtin_mips_insv (i32, i32)
13483 v4i8 __builtin_mips_repl_qb (imm0_255)
13484 v4i8 __builtin_mips_repl_qb (i32)
13485 v2q15 __builtin_mips_repl_ph (imm_n512_511)
13486 v2q15 __builtin_mips_repl_ph (i32)
13487 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
13488 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
13489 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
13490 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
13491 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
13492 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
13493 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
13494 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
13495 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
13496 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
13497 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
13498 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
13499 i32 __builtin_mips_extr_w (a64, imm0_31)
13500 i32 __builtin_mips_extr_w (a64, i32)
13501 i32 __builtin_mips_extr_r_w (a64, imm0_31)
13502 i32 __builtin_mips_extr_s_h (a64, i32)
13503 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
13504 i32 __builtin_mips_extr_rs_w (a64, i32)
13505 i32 __builtin_mips_extr_s_h (a64, imm0_31)
13506 i32 __builtin_mips_extr_r_w (a64, i32)
13507 i32 __builtin_mips_extp (a64, imm0_31)
13508 i32 __builtin_mips_extp (a64, i32)
13509 i32 __builtin_mips_extpdp (a64, imm0_31)
13510 i32 __builtin_mips_extpdp (a64, i32)
13511 a64 __builtin_mips_shilo (a64, imm_n32_31)
13512 a64 __builtin_mips_shilo (a64, i32)
13513 a64 __builtin_mips_mthlip (a64, i32)
13514 void __builtin_mips_wrdsp (i32, imm0_63)
13515 i32 __builtin_mips_rddsp (imm0_63)
13516 i32 __builtin_mips_lbux (void *, i32)
13517 i32 __builtin_mips_lhx (void *, i32)
13518 i32 __builtin_mips_lwx (void *, i32)
13519 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
13520 i32 __builtin_mips_bposge32 (void)
13521 a64 __builtin_mips_madd (a64, i32, i32);
13522 a64 __builtin_mips_maddu (a64, ui32, ui32);
13523 a64 __builtin_mips_msub (a64, i32, i32);
13524 a64 __builtin_mips_msubu (a64, ui32, ui32);
13525 a64 __builtin_mips_mult (i32, i32);
13526 a64 __builtin_mips_multu (ui32, ui32);
13529 The following built-in functions map directly to a particular MIPS DSP REV 2
13530 instruction. Please refer to the architecture specification
13531 for details on what each instruction does.
13534 v4q7 __builtin_mips_absq_s_qb (v4q7);
13535 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
13536 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
13537 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
13538 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
13539 i32 __builtin_mips_append (i32, i32, imm0_31);
13540 i32 __builtin_mips_balign (i32, i32, imm0_3);
13541 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
13542 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
13543 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
13544 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
13545 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
13546 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
13547 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
13548 q31 __builtin_mips_mulq_rs_w (q31, q31);
13549 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
13550 q31 __builtin_mips_mulq_s_w (q31, q31);
13551 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
13552 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
13553 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
13554 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
13555 i32 __builtin_mips_prepend (i32, i32, imm0_31);
13556 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
13557 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
13558 v4i8 __builtin_mips_shra_qb (v4i8, i32);
13559 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
13560 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
13561 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
13562 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
13563 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
13564 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
13565 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
13566 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
13567 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
13568 q31 __builtin_mips_addqh_w (q31, q31);
13569 q31 __builtin_mips_addqh_r_w (q31, q31);
13570 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
13571 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
13572 q31 __builtin_mips_subqh_w (q31, q31);
13573 q31 __builtin_mips_subqh_r_w (q31, q31);
13574 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
13575 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
13576 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
13577 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
13578 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
13579 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
13583 @node MIPS Paired-Single Support
13584 @subsection MIPS Paired-Single Support
13586 The MIPS64 architecture includes a number of instructions that
13587 operate on pairs of single-precision floating-point values.
13588 Each pair is packed into a 64-bit floating-point register,
13589 with one element being designated the ``upper half'' and
13590 the other being designated the ``lower half''.
13592 GCC supports paired-single operations using both the generic
13593 vector extensions (@pxref{Vector Extensions}) and a collection of
13594 MIPS-specific built-in functions. Both kinds of support are
13595 enabled by the @option{-mpaired-single} command-line option.
13597 The vector type associated with paired-single values is usually
13598 called @code{v2sf}. It can be defined in C as follows:
13601 typedef float v2sf __attribute__ ((vector_size (8)));
13604 @code{v2sf} values are initialized in the same way as aggregates.
13608 v2sf a = @{1.5, 9.1@};
13611 b = (v2sf) @{e, f@};
13614 @emph{Note:} The CPU's endianness determines which value is stored in
13615 the upper half of a register and which value is stored in the lower half.
13616 On little-endian targets, the first value is the lower one and the second
13617 value is the upper one. The opposite order applies to big-endian targets.
13618 For example, the code above sets the lower half of @code{a} to
13619 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
13621 @node MIPS Loongson Built-in Functions
13622 @subsection MIPS Loongson Built-in Functions
13624 GCC provides intrinsics to access the SIMD instructions provided by the
13625 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
13626 available after inclusion of the @code{loongson.h} header file,
13627 operate on the following 64-bit vector types:
13630 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
13631 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
13632 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
13633 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
13634 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
13635 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
13638 The intrinsics provided are listed below; each is named after the
13639 machine instruction to which it corresponds, with suffixes added as
13640 appropriate to distinguish intrinsics that expand to the same machine
13641 instruction yet have different argument types. Refer to the architecture
13642 documentation for a description of the functionality of each
13646 int16x4_t packsswh (int32x2_t s, int32x2_t t);
13647 int8x8_t packsshb (int16x4_t s, int16x4_t t);
13648 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
13649 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
13650 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
13651 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
13652 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
13653 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
13654 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
13655 uint64_t paddd_u (uint64_t s, uint64_t t);
13656 int64_t paddd_s (int64_t s, int64_t t);
13657 int16x4_t paddsh (int16x4_t s, int16x4_t t);
13658 int8x8_t paddsb (int8x8_t s, int8x8_t t);
13659 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
13660 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
13661 uint64_t pandn_ud (uint64_t s, uint64_t t);
13662 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
13663 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
13664 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
13665 int64_t pandn_sd (int64_t s, int64_t t);
13666 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13667 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13668 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13669 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13670 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13671 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13672 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13673 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13674 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13675 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13676 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13677 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13678 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13679 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13680 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13681 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13682 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13683 uint16x4_t pextrh_u (uint16x4_t s, int field);
13684 int16x4_t pextrh_s (int16x4_t s, int field);
13685 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13686 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13687 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13688 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13689 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13690 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13691 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13692 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13693 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13694 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13695 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13696 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13697 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13698 uint8x8_t pmovmskb_u (uint8x8_t s);
13699 int8x8_t pmovmskb_s (int8x8_t s);
13700 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13701 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13702 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13703 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13704 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13705 uint16x4_t biadd (uint8x8_t s);
13706 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13707 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13708 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13709 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13710 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13711 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13712 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13713 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13714 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13715 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13716 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13717 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13718 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13719 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13720 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13721 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13722 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13723 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13724 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13725 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13726 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13727 uint64_t psubd_u (uint64_t s, uint64_t t);
13728 int64_t psubd_s (int64_t s, int64_t t);
13729 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13730 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13731 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13732 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13733 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13734 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13735 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13736 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13737 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13738 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13739 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13740 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13741 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13742 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13743 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13744 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13748 * Paired-Single Arithmetic::
13749 * Paired-Single Built-in Functions::
13750 * MIPS-3D Built-in Functions::
13753 @node Paired-Single Arithmetic
13754 @subsubsection Paired-Single Arithmetic
13756 The table below lists the @code{v2sf} operations for which hardware
13757 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13758 values and @code{x} is an integral value.
13760 @multitable @columnfractions .50 .50
13761 @item C code @tab MIPS instruction
13762 @item @code{a + b} @tab @code{add.ps}
13763 @item @code{a - b} @tab @code{sub.ps}
13764 @item @code{-a} @tab @code{neg.ps}
13765 @item @code{a * b} @tab @code{mul.ps}
13766 @item @code{a * b + c} @tab @code{madd.ps}
13767 @item @code{a * b - c} @tab @code{msub.ps}
13768 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13769 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13770 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13773 Note that the multiply-accumulate instructions can be disabled
13774 using the command-line option @code{-mno-fused-madd}.
13776 @node Paired-Single Built-in Functions
13777 @subsubsection Paired-Single Built-in Functions
13779 The following paired-single functions map directly to a particular
13780 MIPS instruction. Please refer to the architecture specification
13781 for details on what each instruction does.
13784 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13785 Pair lower lower (@code{pll.ps}).
13787 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13788 Pair upper lower (@code{pul.ps}).
13790 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13791 Pair lower upper (@code{plu.ps}).
13793 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13794 Pair upper upper (@code{puu.ps}).
13796 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13797 Convert pair to paired single (@code{cvt.ps.s}).
13799 @item float __builtin_mips_cvt_s_pl (v2sf)
13800 Convert pair lower to single (@code{cvt.s.pl}).
13802 @item float __builtin_mips_cvt_s_pu (v2sf)
13803 Convert pair upper to single (@code{cvt.s.pu}).
13805 @item v2sf __builtin_mips_abs_ps (v2sf)
13806 Absolute value (@code{abs.ps}).
13808 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13809 Align variable (@code{alnv.ps}).
13811 @emph{Note:} The value of the third parameter must be 0 or 4
13812 modulo 8, otherwise the result is unpredictable. Please read the
13813 instruction description for details.
13816 The following multi-instruction functions are also available.
13817 In each case, @var{cond} can be any of the 16 floating-point conditions:
13818 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13819 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13820 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13823 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13824 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13825 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13826 @code{movt.ps}/@code{movf.ps}).
13828 The @code{movt} functions return the value @var{x} computed by:
13831 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13832 mov.ps @var{x},@var{c}
13833 movt.ps @var{x},@var{d},@var{cc}
13836 The @code{movf} functions are similar but use @code{movf.ps} instead
13839 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13840 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13841 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13842 @code{bc1t}/@code{bc1f}).
13844 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13845 and return either the upper or lower half of the result. For example:
13849 if (__builtin_mips_upper_c_eq_ps (a, b))
13850 upper_halves_are_equal ();
13852 upper_halves_are_unequal ();
13854 if (__builtin_mips_lower_c_eq_ps (a, b))
13855 lower_halves_are_equal ();
13857 lower_halves_are_unequal ();
13861 @node MIPS-3D Built-in Functions
13862 @subsubsection MIPS-3D Built-in Functions
13864 The MIPS-3D Application-Specific Extension (ASE) includes additional
13865 paired-single instructions that are designed to improve the performance
13866 of 3D graphics operations. Support for these instructions is controlled
13867 by the @option{-mips3d} command-line option.
13869 The functions listed below map directly to a particular MIPS-3D
13870 instruction. Please refer to the architecture specification for
13871 more details on what each instruction does.
13874 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13875 Reduction add (@code{addr.ps}).
13877 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13878 Reduction multiply (@code{mulr.ps}).
13880 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13881 Convert paired single to paired word (@code{cvt.pw.ps}).
13883 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13884 Convert paired word to paired single (@code{cvt.ps.pw}).
13886 @item float __builtin_mips_recip1_s (float)
13887 @itemx double __builtin_mips_recip1_d (double)
13888 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13889 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13891 @item float __builtin_mips_recip2_s (float, float)
13892 @itemx double __builtin_mips_recip2_d (double, double)
13893 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13894 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13896 @item float __builtin_mips_rsqrt1_s (float)
13897 @itemx double __builtin_mips_rsqrt1_d (double)
13898 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13899 Reduced-precision reciprocal square root (sequence step 1)
13900 (@code{rsqrt1.@var{fmt}}).
13902 @item float __builtin_mips_rsqrt2_s (float, float)
13903 @itemx double __builtin_mips_rsqrt2_d (double, double)
13904 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13905 Reduced-precision reciprocal square root (sequence step 2)
13906 (@code{rsqrt2.@var{fmt}}).
13909 The following multi-instruction functions are also available.
13910 In each case, @var{cond} can be any of the 16 floating-point conditions:
13911 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13912 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13913 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13916 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13917 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13918 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13919 @code{bc1t}/@code{bc1f}).
13921 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13922 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13927 if (__builtin_mips_cabs_eq_s (a, b))
13933 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13934 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13935 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13936 @code{bc1t}/@code{bc1f}).
13938 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13939 and return either the upper or lower half of the result. For example:
13943 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13944 upper_halves_are_equal ();
13946 upper_halves_are_unequal ();
13948 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13949 lower_halves_are_equal ();
13951 lower_halves_are_unequal ();
13954 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13955 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13956 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13957 @code{movt.ps}/@code{movf.ps}).
13959 The @code{movt} functions return the value @var{x} computed by:
13962 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13963 mov.ps @var{x},@var{c}
13964 movt.ps @var{x},@var{d},@var{cc}
13967 The @code{movf} functions are similar but use @code{movf.ps} instead
13970 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13971 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13972 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13973 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13974 Comparison of two paired-single values
13975 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13976 @code{bc1any2t}/@code{bc1any2f}).
13978 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13979 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13980 result is true and the @code{all} forms return true if both results are true.
13985 if (__builtin_mips_any_c_eq_ps (a, b))
13990 if (__builtin_mips_all_c_eq_ps (a, b))
13996 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13997 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13998 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13999 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14000 Comparison of four paired-single values
14001 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
14002 @code{bc1any4t}/@code{bc1any4f}).
14004 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
14005 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
14006 The @code{any} forms return true if any of the four results are true
14007 and the @code{all} forms return true if all four results are true.
14012 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
14017 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
14024 @node MIPS SIMD Architecture (MSA) Support
14025 @subsection MIPS SIMD Architecture (MSA) Support
14028 * MIPS SIMD Architecture Built-in Functions::
14031 GCC provides intrinsics to access the SIMD instructions provided by the
14032 MSA MIPS SIMD Architecture. The interface is made available by including
14033 @code{<msa.h>} and using @option{-mmsa -mhard-float -mfp64 -mnan=2008}.
14034 For each @code{__builtin_msa_*}, there is a shortened name of the intrinsic,
14037 MSA implements 128-bit wide vector registers, operating on 8-, 16-, 32- and
14038 64-bit integer, 16- and 32-bit fixed-point, or 32- and 64-bit floating point
14039 data elements. The following vectors typedefs are included in @code{msa.h}:
14041 @item @code{v16i8}, a vector of sixteen signed 8-bit integers;
14042 @item @code{v16u8}, a vector of sixteen unsigned 8-bit integers;
14043 @item @code{v8i16}, a vector of eight signed 16-bit integers;
14044 @item @code{v8u16}, a vector of eight unsigned 16-bit integers;
14045 @item @code{v4i32}, a vector of four signed 32-bit integers;
14046 @item @code{v4u32}, a vector of four unsigned 32-bit integers;
14047 @item @code{v2i64}, a vector of two signed 64-bit integers;
14048 @item @code{v2u64}, a vector of two unsigned 64-bit integers;
14049 @item @code{v4f32}, a vector of four 32-bit floats;
14050 @item @code{v2f64}, a vector of two 64-bit doubles.
14053 Instructions and corresponding built-ins may have additional restrictions and/or
14054 input/output values manipulated:
14056 @item @code{imm0_1}, an integer literal in range 0 to 1;
14057 @item @code{imm0_3}, an integer literal in range 0 to 3;
14058 @item @code{imm0_7}, an integer literal in range 0 to 7;
14059 @item @code{imm0_15}, an integer literal in range 0 to 15;
14060 @item @code{imm0_31}, an integer literal in range 0 to 31;
14061 @item @code{imm0_63}, an integer literal in range 0 to 63;
14062 @item @code{imm0_255}, an integer literal in range 0 to 255;
14063 @item @code{imm_n16_15}, an integer literal in range -16 to 15;
14064 @item @code{imm_n512_511}, an integer literal in range -512 to 511;
14065 @item @code{imm_n1024_1022}, an integer literal in range -512 to 511 left
14066 shifted by 1 bit, i.e., -1024, -1022, @dots{}, 1020, 1022;
14067 @item @code{imm_n2048_2044}, an integer literal in range -512 to 511 left
14068 shifted by 2 bits, i.e., -2048, -2044, @dots{}, 2040, 2044;
14069 @item @code{imm_n4096_4088}, an integer literal in range -512 to 511 left
14070 shifted by 3 bits, i.e., -4096, -4088, @dots{}, 4080, 4088;
14071 @item @code{imm1_4}, an integer literal in range 1 to 4;
14072 @item @code{i32, i64, u32, u64, f32, f64}, defined as follows:
14078 #if __LONG_MAX__ == __LONG_LONG_MAX__
14081 typedef long long i64;
14084 typedef unsigned int u32;
14085 #if __LONG_MAX__ == __LONG_LONG_MAX__
14086 typedef unsigned long u64;
14088 typedef unsigned long long u64;
14091 typedef double f64;
14096 @node MIPS SIMD Architecture Built-in Functions
14097 @subsubsection MIPS SIMD Architecture Built-in Functions
14099 The intrinsics provided are listed below; each is named after the
14100 machine instruction.
14103 v16i8 __builtin_msa_add_a_b (v16i8, v16i8);
14104 v8i16 __builtin_msa_add_a_h (v8i16, v8i16);
14105 v4i32 __builtin_msa_add_a_w (v4i32, v4i32);
14106 v2i64 __builtin_msa_add_a_d (v2i64, v2i64);
14108 v16i8 __builtin_msa_adds_a_b (v16i8, v16i8);
14109 v8i16 __builtin_msa_adds_a_h (v8i16, v8i16);
14110 v4i32 __builtin_msa_adds_a_w (v4i32, v4i32);
14111 v2i64 __builtin_msa_adds_a_d (v2i64, v2i64);
14113 v16i8 __builtin_msa_adds_s_b (v16i8, v16i8);
14114 v8i16 __builtin_msa_adds_s_h (v8i16, v8i16);
14115 v4i32 __builtin_msa_adds_s_w (v4i32, v4i32);
14116 v2i64 __builtin_msa_adds_s_d (v2i64, v2i64);
14118 v16u8 __builtin_msa_adds_u_b (v16u8, v16u8);
14119 v8u16 __builtin_msa_adds_u_h (v8u16, v8u16);
14120 v4u32 __builtin_msa_adds_u_w (v4u32, v4u32);
14121 v2u64 __builtin_msa_adds_u_d (v2u64, v2u64);
14123 v16i8 __builtin_msa_addv_b (v16i8, v16i8);
14124 v8i16 __builtin_msa_addv_h (v8i16, v8i16);
14125 v4i32 __builtin_msa_addv_w (v4i32, v4i32);
14126 v2i64 __builtin_msa_addv_d (v2i64, v2i64);
14128 v16i8 __builtin_msa_addvi_b (v16i8, imm0_31);
14129 v8i16 __builtin_msa_addvi_h (v8i16, imm0_31);
14130 v4i32 __builtin_msa_addvi_w (v4i32, imm0_31);
14131 v2i64 __builtin_msa_addvi_d (v2i64, imm0_31);
14133 v16u8 __builtin_msa_and_v (v16u8, v16u8);
14135 v16u8 __builtin_msa_andi_b (v16u8, imm0_255);
14137 v16i8 __builtin_msa_asub_s_b (v16i8, v16i8);
14138 v8i16 __builtin_msa_asub_s_h (v8i16, v8i16);
14139 v4i32 __builtin_msa_asub_s_w (v4i32, v4i32);
14140 v2i64 __builtin_msa_asub_s_d (v2i64, v2i64);
14142 v16u8 __builtin_msa_asub_u_b (v16u8, v16u8);
14143 v8u16 __builtin_msa_asub_u_h (v8u16, v8u16);
14144 v4u32 __builtin_msa_asub_u_w (v4u32, v4u32);
14145 v2u64 __builtin_msa_asub_u_d (v2u64, v2u64);
14147 v16i8 __builtin_msa_ave_s_b (v16i8, v16i8);
14148 v8i16 __builtin_msa_ave_s_h (v8i16, v8i16);
14149 v4i32 __builtin_msa_ave_s_w (v4i32, v4i32);
14150 v2i64 __builtin_msa_ave_s_d (v2i64, v2i64);
14152 v16u8 __builtin_msa_ave_u_b (v16u8, v16u8);
14153 v8u16 __builtin_msa_ave_u_h (v8u16, v8u16);
14154 v4u32 __builtin_msa_ave_u_w (v4u32, v4u32);
14155 v2u64 __builtin_msa_ave_u_d (v2u64, v2u64);
14157 v16i8 __builtin_msa_aver_s_b (v16i8, v16i8);
14158 v8i16 __builtin_msa_aver_s_h (v8i16, v8i16);
14159 v4i32 __builtin_msa_aver_s_w (v4i32, v4i32);
14160 v2i64 __builtin_msa_aver_s_d (v2i64, v2i64);
14162 v16u8 __builtin_msa_aver_u_b (v16u8, v16u8);
14163 v8u16 __builtin_msa_aver_u_h (v8u16, v8u16);
14164 v4u32 __builtin_msa_aver_u_w (v4u32, v4u32);
14165 v2u64 __builtin_msa_aver_u_d (v2u64, v2u64);
14167 v16u8 __builtin_msa_bclr_b (v16u8, v16u8);
14168 v8u16 __builtin_msa_bclr_h (v8u16, v8u16);
14169 v4u32 __builtin_msa_bclr_w (v4u32, v4u32);
14170 v2u64 __builtin_msa_bclr_d (v2u64, v2u64);
14172 v16u8 __builtin_msa_bclri_b (v16u8, imm0_7);
14173 v8u16 __builtin_msa_bclri_h (v8u16, imm0_15);
14174 v4u32 __builtin_msa_bclri_w (v4u32, imm0_31);
14175 v2u64 __builtin_msa_bclri_d (v2u64, imm0_63);
14177 v16u8 __builtin_msa_binsl_b (v16u8, v16u8, v16u8);
14178 v8u16 __builtin_msa_binsl_h (v8u16, v8u16, v8u16);
14179 v4u32 __builtin_msa_binsl_w (v4u32, v4u32, v4u32);
14180 v2u64 __builtin_msa_binsl_d (v2u64, v2u64, v2u64);
14182 v16u8 __builtin_msa_binsli_b (v16u8, v16u8, imm0_7);
14183 v8u16 __builtin_msa_binsli_h (v8u16, v8u16, imm0_15);
14184 v4u32 __builtin_msa_binsli_w (v4u32, v4u32, imm0_31);
14185 v2u64 __builtin_msa_binsli_d (v2u64, v2u64, imm0_63);
14187 v16u8 __builtin_msa_binsr_b (v16u8, v16u8, v16u8);
14188 v8u16 __builtin_msa_binsr_h (v8u16, v8u16, v8u16);
14189 v4u32 __builtin_msa_binsr_w (v4u32, v4u32, v4u32);
14190 v2u64 __builtin_msa_binsr_d (v2u64, v2u64, v2u64);
14192 v16u8 __builtin_msa_binsri_b (v16u8, v16u8, imm0_7);
14193 v8u16 __builtin_msa_binsri_h (v8u16, v8u16, imm0_15);
14194 v4u32 __builtin_msa_binsri_w (v4u32, v4u32, imm0_31);
14195 v2u64 __builtin_msa_binsri_d (v2u64, v2u64, imm0_63);
14197 v16u8 __builtin_msa_bmnz_v (v16u8, v16u8, v16u8);
14199 v16u8 __builtin_msa_bmnzi_b (v16u8, v16u8, imm0_255);
14201 v16u8 __builtin_msa_bmz_v (v16u8, v16u8, v16u8);
14203 v16u8 __builtin_msa_bmzi_b (v16u8, v16u8, imm0_255);
14205 v16u8 __builtin_msa_bneg_b (v16u8, v16u8);
14206 v8u16 __builtin_msa_bneg_h (v8u16, v8u16);
14207 v4u32 __builtin_msa_bneg_w (v4u32, v4u32);
14208 v2u64 __builtin_msa_bneg_d (v2u64, v2u64);
14210 v16u8 __builtin_msa_bnegi_b (v16u8, imm0_7);
14211 v8u16 __builtin_msa_bnegi_h (v8u16, imm0_15);
14212 v4u32 __builtin_msa_bnegi_w (v4u32, imm0_31);
14213 v2u64 __builtin_msa_bnegi_d (v2u64, imm0_63);
14215 i32 __builtin_msa_bnz_b (v16u8);
14216 i32 __builtin_msa_bnz_h (v8u16);
14217 i32 __builtin_msa_bnz_w (v4u32);
14218 i32 __builtin_msa_bnz_d (v2u64);
14220 i32 __builtin_msa_bnz_v (v16u8);
14222 v16u8 __builtin_msa_bsel_v (v16u8, v16u8, v16u8);
14224 v16u8 __builtin_msa_bseli_b (v16u8, v16u8, imm0_255);
14226 v16u8 __builtin_msa_bset_b (v16u8, v16u8);
14227 v8u16 __builtin_msa_bset_h (v8u16, v8u16);
14228 v4u32 __builtin_msa_bset_w (v4u32, v4u32);
14229 v2u64 __builtin_msa_bset_d (v2u64, v2u64);
14231 v16u8 __builtin_msa_bseti_b (v16u8, imm0_7);
14232 v8u16 __builtin_msa_bseti_h (v8u16, imm0_15);
14233 v4u32 __builtin_msa_bseti_w (v4u32, imm0_31);
14234 v2u64 __builtin_msa_bseti_d (v2u64, imm0_63);
14236 i32 __builtin_msa_bz_b (v16u8);
14237 i32 __builtin_msa_bz_h (v8u16);
14238 i32 __builtin_msa_bz_w (v4u32);
14239 i32 __builtin_msa_bz_d (v2u64);
14241 i32 __builtin_msa_bz_v (v16u8);
14243 v16i8 __builtin_msa_ceq_b (v16i8, v16i8);
14244 v8i16 __builtin_msa_ceq_h (v8i16, v8i16);
14245 v4i32 __builtin_msa_ceq_w (v4i32, v4i32);
14246 v2i64 __builtin_msa_ceq_d (v2i64, v2i64);
14248 v16i8 __builtin_msa_ceqi_b (v16i8, imm_n16_15);
14249 v8i16 __builtin_msa_ceqi_h (v8i16, imm_n16_15);
14250 v4i32 __builtin_msa_ceqi_w (v4i32, imm_n16_15);
14251 v2i64 __builtin_msa_ceqi_d (v2i64, imm_n16_15);
14253 i32 __builtin_msa_cfcmsa (imm0_31);
14255 v16i8 __builtin_msa_cle_s_b (v16i8, v16i8);
14256 v8i16 __builtin_msa_cle_s_h (v8i16, v8i16);
14257 v4i32 __builtin_msa_cle_s_w (v4i32, v4i32);
14258 v2i64 __builtin_msa_cle_s_d (v2i64, v2i64);
14260 v16i8 __builtin_msa_cle_u_b (v16u8, v16u8);
14261 v8i16 __builtin_msa_cle_u_h (v8u16, v8u16);
14262 v4i32 __builtin_msa_cle_u_w (v4u32, v4u32);
14263 v2i64 __builtin_msa_cle_u_d (v2u64, v2u64);
14265 v16i8 __builtin_msa_clei_s_b (v16i8, imm_n16_15);
14266 v8i16 __builtin_msa_clei_s_h (v8i16, imm_n16_15);
14267 v4i32 __builtin_msa_clei_s_w (v4i32, imm_n16_15);
14268 v2i64 __builtin_msa_clei_s_d (v2i64, imm_n16_15);
14270 v16i8 __builtin_msa_clei_u_b (v16u8, imm0_31);
14271 v8i16 __builtin_msa_clei_u_h (v8u16, imm0_31);
14272 v4i32 __builtin_msa_clei_u_w (v4u32, imm0_31);
14273 v2i64 __builtin_msa_clei_u_d (v2u64, imm0_31);
14275 v16i8 __builtin_msa_clt_s_b (v16i8, v16i8);
14276 v8i16 __builtin_msa_clt_s_h (v8i16, v8i16);
14277 v4i32 __builtin_msa_clt_s_w (v4i32, v4i32);
14278 v2i64 __builtin_msa_clt_s_d (v2i64, v2i64);
14280 v16i8 __builtin_msa_clt_u_b (v16u8, v16u8);
14281 v8i16 __builtin_msa_clt_u_h (v8u16, v8u16);
14282 v4i32 __builtin_msa_clt_u_w (v4u32, v4u32);
14283 v2i64 __builtin_msa_clt_u_d (v2u64, v2u64);
14285 v16i8 __builtin_msa_clti_s_b (v16i8, imm_n16_15);
14286 v8i16 __builtin_msa_clti_s_h (v8i16, imm_n16_15);
14287 v4i32 __builtin_msa_clti_s_w (v4i32, imm_n16_15);
14288 v2i64 __builtin_msa_clti_s_d (v2i64, imm_n16_15);
14290 v16i8 __builtin_msa_clti_u_b (v16u8, imm0_31);
14291 v8i16 __builtin_msa_clti_u_h (v8u16, imm0_31);
14292 v4i32 __builtin_msa_clti_u_w (v4u32, imm0_31);
14293 v2i64 __builtin_msa_clti_u_d (v2u64, imm0_31);
14295 i32 __builtin_msa_copy_s_b (v16i8, imm0_15);
14296 i32 __builtin_msa_copy_s_h (v8i16, imm0_7);
14297 i32 __builtin_msa_copy_s_w (v4i32, imm0_3);
14298 i64 __builtin_msa_copy_s_d (v2i64, imm0_1);
14300 u32 __builtin_msa_copy_u_b (v16i8, imm0_15);
14301 u32 __builtin_msa_copy_u_h (v8i16, imm0_7);
14302 u32 __builtin_msa_copy_u_w (v4i32, imm0_3);
14303 u64 __builtin_msa_copy_u_d (v2i64, imm0_1);
14305 void __builtin_msa_ctcmsa (imm0_31, i32);
14307 v16i8 __builtin_msa_div_s_b (v16i8, v16i8);
14308 v8i16 __builtin_msa_div_s_h (v8i16, v8i16);
14309 v4i32 __builtin_msa_div_s_w (v4i32, v4i32);
14310 v2i64 __builtin_msa_div_s_d (v2i64, v2i64);
14312 v16u8 __builtin_msa_div_u_b (v16u8, v16u8);
14313 v8u16 __builtin_msa_div_u_h (v8u16, v8u16);
14314 v4u32 __builtin_msa_div_u_w (v4u32, v4u32);
14315 v2u64 __builtin_msa_div_u_d (v2u64, v2u64);
14317 v8i16 __builtin_msa_dotp_s_h (v16i8, v16i8);
14318 v4i32 __builtin_msa_dotp_s_w (v8i16, v8i16);
14319 v2i64 __builtin_msa_dotp_s_d (v4i32, v4i32);
14321 v8u16 __builtin_msa_dotp_u_h (v16u8, v16u8);
14322 v4u32 __builtin_msa_dotp_u_w (v8u16, v8u16);
14323 v2u64 __builtin_msa_dotp_u_d (v4u32, v4u32);
14325 v8i16 __builtin_msa_dpadd_s_h (v8i16, v16i8, v16i8);
14326 v4i32 __builtin_msa_dpadd_s_w (v4i32, v8i16, v8i16);
14327 v2i64 __builtin_msa_dpadd_s_d (v2i64, v4i32, v4i32);
14329 v8u16 __builtin_msa_dpadd_u_h (v8u16, v16u8, v16u8);
14330 v4u32 __builtin_msa_dpadd_u_w (v4u32, v8u16, v8u16);
14331 v2u64 __builtin_msa_dpadd_u_d (v2u64, v4u32, v4u32);
14333 v8i16 __builtin_msa_dpsub_s_h (v8i16, v16i8, v16i8);
14334 v4i32 __builtin_msa_dpsub_s_w (v4i32, v8i16, v8i16);
14335 v2i64 __builtin_msa_dpsub_s_d (v2i64, v4i32, v4i32);
14337 v8i16 __builtin_msa_dpsub_u_h (v8i16, v16u8, v16u8);
14338 v4i32 __builtin_msa_dpsub_u_w (v4i32, v8u16, v8u16);
14339 v2i64 __builtin_msa_dpsub_u_d (v2i64, v4u32, v4u32);
14341 v4f32 __builtin_msa_fadd_w (v4f32, v4f32);
14342 v2f64 __builtin_msa_fadd_d (v2f64, v2f64);
14344 v4i32 __builtin_msa_fcaf_w (v4f32, v4f32);
14345 v2i64 __builtin_msa_fcaf_d (v2f64, v2f64);
14347 v4i32 __builtin_msa_fceq_w (v4f32, v4f32);
14348 v2i64 __builtin_msa_fceq_d (v2f64, v2f64);
14350 v4i32 __builtin_msa_fclass_w (v4f32);
14351 v2i64 __builtin_msa_fclass_d (v2f64);
14353 v4i32 __builtin_msa_fcle_w (v4f32, v4f32);
14354 v2i64 __builtin_msa_fcle_d (v2f64, v2f64);
14356 v4i32 __builtin_msa_fclt_w (v4f32, v4f32);
14357 v2i64 __builtin_msa_fclt_d (v2f64, v2f64);
14359 v4i32 __builtin_msa_fcne_w (v4f32, v4f32);
14360 v2i64 __builtin_msa_fcne_d (v2f64, v2f64);
14362 v4i32 __builtin_msa_fcor_w (v4f32, v4f32);
14363 v2i64 __builtin_msa_fcor_d (v2f64, v2f64);
14365 v4i32 __builtin_msa_fcueq_w (v4f32, v4f32);
14366 v2i64 __builtin_msa_fcueq_d (v2f64, v2f64);
14368 v4i32 __builtin_msa_fcule_w (v4f32, v4f32);
14369 v2i64 __builtin_msa_fcule_d (v2f64, v2f64);
14371 v4i32 __builtin_msa_fcult_w (v4f32, v4f32);
14372 v2i64 __builtin_msa_fcult_d (v2f64, v2f64);
14374 v4i32 __builtin_msa_fcun_w (v4f32, v4f32);
14375 v2i64 __builtin_msa_fcun_d (v2f64, v2f64);
14377 v4i32 __builtin_msa_fcune_w (v4f32, v4f32);
14378 v2i64 __builtin_msa_fcune_d (v2f64, v2f64);
14380 v4f32 __builtin_msa_fdiv_w (v4f32, v4f32);
14381 v2f64 __builtin_msa_fdiv_d (v2f64, v2f64);
14383 v8i16 __builtin_msa_fexdo_h (v4f32, v4f32);
14384 v4f32 __builtin_msa_fexdo_w (v2f64, v2f64);
14386 v4f32 __builtin_msa_fexp2_w (v4f32, v4i32);
14387 v2f64 __builtin_msa_fexp2_d (v2f64, v2i64);
14389 v4f32 __builtin_msa_fexupl_w (v8i16);
14390 v2f64 __builtin_msa_fexupl_d (v4f32);
14392 v4f32 __builtin_msa_fexupr_w (v8i16);
14393 v2f64 __builtin_msa_fexupr_d (v4f32);
14395 v4f32 __builtin_msa_ffint_s_w (v4i32);
14396 v2f64 __builtin_msa_ffint_s_d (v2i64);
14398 v4f32 __builtin_msa_ffint_u_w (v4u32);
14399 v2f64 __builtin_msa_ffint_u_d (v2u64);
14401 v4f32 __builtin_msa_ffql_w (v8i16);
14402 v2f64 __builtin_msa_ffql_d (v4i32);
14404 v4f32 __builtin_msa_ffqr_w (v8i16);
14405 v2f64 __builtin_msa_ffqr_d (v4i32);
14407 v16i8 __builtin_msa_fill_b (i32);
14408 v8i16 __builtin_msa_fill_h (i32);
14409 v4i32 __builtin_msa_fill_w (i32);
14410 v2i64 __builtin_msa_fill_d (i64);
14412 v4f32 __builtin_msa_flog2_w (v4f32);
14413 v2f64 __builtin_msa_flog2_d (v2f64);
14415 v4f32 __builtin_msa_fmadd_w (v4f32, v4f32, v4f32);
14416 v2f64 __builtin_msa_fmadd_d (v2f64, v2f64, v2f64);
14418 v4f32 __builtin_msa_fmax_w (v4f32, v4f32);
14419 v2f64 __builtin_msa_fmax_d (v2f64, v2f64);
14421 v4f32 __builtin_msa_fmax_a_w (v4f32, v4f32);
14422 v2f64 __builtin_msa_fmax_a_d (v2f64, v2f64);
14424 v4f32 __builtin_msa_fmin_w (v4f32, v4f32);
14425 v2f64 __builtin_msa_fmin_d (v2f64, v2f64);
14427 v4f32 __builtin_msa_fmin_a_w (v4f32, v4f32);
14428 v2f64 __builtin_msa_fmin_a_d (v2f64, v2f64);
14430 v4f32 __builtin_msa_fmsub_w (v4f32, v4f32, v4f32);
14431 v2f64 __builtin_msa_fmsub_d (v2f64, v2f64, v2f64);
14433 v4f32 __builtin_msa_fmul_w (v4f32, v4f32);
14434 v2f64 __builtin_msa_fmul_d (v2f64, v2f64);
14436 v4f32 __builtin_msa_frint_w (v4f32);
14437 v2f64 __builtin_msa_frint_d (v2f64);
14439 v4f32 __builtin_msa_frcp_w (v4f32);
14440 v2f64 __builtin_msa_frcp_d (v2f64);
14442 v4f32 __builtin_msa_frsqrt_w (v4f32);
14443 v2f64 __builtin_msa_frsqrt_d (v2f64);
14445 v4i32 __builtin_msa_fsaf_w (v4f32, v4f32);
14446 v2i64 __builtin_msa_fsaf_d (v2f64, v2f64);
14448 v4i32 __builtin_msa_fseq_w (v4f32, v4f32);
14449 v2i64 __builtin_msa_fseq_d (v2f64, v2f64);
14451 v4i32 __builtin_msa_fsle_w (v4f32, v4f32);
14452 v2i64 __builtin_msa_fsle_d (v2f64, v2f64);
14454 v4i32 __builtin_msa_fslt_w (v4f32, v4f32);
14455 v2i64 __builtin_msa_fslt_d (v2f64, v2f64);
14457 v4i32 __builtin_msa_fsne_w (v4f32, v4f32);
14458 v2i64 __builtin_msa_fsne_d (v2f64, v2f64);
14460 v4i32 __builtin_msa_fsor_w (v4f32, v4f32);
14461 v2i64 __builtin_msa_fsor_d (v2f64, v2f64);
14463 v4f32 __builtin_msa_fsqrt_w (v4f32);
14464 v2f64 __builtin_msa_fsqrt_d (v2f64);
14466 v4f32 __builtin_msa_fsub_w (v4f32, v4f32);
14467 v2f64 __builtin_msa_fsub_d (v2f64, v2f64);
14469 v4i32 __builtin_msa_fsueq_w (v4f32, v4f32);
14470 v2i64 __builtin_msa_fsueq_d (v2f64, v2f64);
14472 v4i32 __builtin_msa_fsule_w (v4f32, v4f32);
14473 v2i64 __builtin_msa_fsule_d (v2f64, v2f64);
14475 v4i32 __builtin_msa_fsult_w (v4f32, v4f32);
14476 v2i64 __builtin_msa_fsult_d (v2f64, v2f64);
14478 v4i32 __builtin_msa_fsun_w (v4f32, v4f32);
14479 v2i64 __builtin_msa_fsun_d (v2f64, v2f64);
14481 v4i32 __builtin_msa_fsune_w (v4f32, v4f32);
14482 v2i64 __builtin_msa_fsune_d (v2f64, v2f64);
14484 v4i32 __builtin_msa_ftint_s_w (v4f32);
14485 v2i64 __builtin_msa_ftint_s_d (v2f64);
14487 v4u32 __builtin_msa_ftint_u_w (v4f32);
14488 v2u64 __builtin_msa_ftint_u_d (v2f64);
14490 v8i16 __builtin_msa_ftq_h (v4f32, v4f32);
14491 v4i32 __builtin_msa_ftq_w (v2f64, v2f64);
14493 v4i32 __builtin_msa_ftrunc_s_w (v4f32);
14494 v2i64 __builtin_msa_ftrunc_s_d (v2f64);
14496 v4u32 __builtin_msa_ftrunc_u_w (v4f32);
14497 v2u64 __builtin_msa_ftrunc_u_d (v2f64);
14499 v8i16 __builtin_msa_hadd_s_h (v16i8, v16i8);
14500 v4i32 __builtin_msa_hadd_s_w (v8i16, v8i16);
14501 v2i64 __builtin_msa_hadd_s_d (v4i32, v4i32);
14503 v8u16 __builtin_msa_hadd_u_h (v16u8, v16u8);
14504 v4u32 __builtin_msa_hadd_u_w (v8u16, v8u16);
14505 v2u64 __builtin_msa_hadd_u_d (v4u32, v4u32);
14507 v8i16 __builtin_msa_hsub_s_h (v16i8, v16i8);
14508 v4i32 __builtin_msa_hsub_s_w (v8i16, v8i16);
14509 v2i64 __builtin_msa_hsub_s_d (v4i32, v4i32);
14511 v8i16 __builtin_msa_hsub_u_h (v16u8, v16u8);
14512 v4i32 __builtin_msa_hsub_u_w (v8u16, v8u16);
14513 v2i64 __builtin_msa_hsub_u_d (v4u32, v4u32);
14515 v16i8 __builtin_msa_ilvev_b (v16i8, v16i8);
14516 v8i16 __builtin_msa_ilvev_h (v8i16, v8i16);
14517 v4i32 __builtin_msa_ilvev_w (v4i32, v4i32);
14518 v2i64 __builtin_msa_ilvev_d (v2i64, v2i64);
14520 v16i8 __builtin_msa_ilvl_b (v16i8, v16i8);
14521 v8i16 __builtin_msa_ilvl_h (v8i16, v8i16);
14522 v4i32 __builtin_msa_ilvl_w (v4i32, v4i32);
14523 v2i64 __builtin_msa_ilvl_d (v2i64, v2i64);
14525 v16i8 __builtin_msa_ilvod_b (v16i8, v16i8);
14526 v8i16 __builtin_msa_ilvod_h (v8i16, v8i16);
14527 v4i32 __builtin_msa_ilvod_w (v4i32, v4i32);
14528 v2i64 __builtin_msa_ilvod_d (v2i64, v2i64);
14530 v16i8 __builtin_msa_ilvr_b (v16i8, v16i8);
14531 v8i16 __builtin_msa_ilvr_h (v8i16, v8i16);
14532 v4i32 __builtin_msa_ilvr_w (v4i32, v4i32);
14533 v2i64 __builtin_msa_ilvr_d (v2i64, v2i64);
14535 v16i8 __builtin_msa_insert_b (v16i8, imm0_15, i32);
14536 v8i16 __builtin_msa_insert_h (v8i16, imm0_7, i32);
14537 v4i32 __builtin_msa_insert_w (v4i32, imm0_3, i32);
14538 v2i64 __builtin_msa_insert_d (v2i64, imm0_1, i64);
14540 v16i8 __builtin_msa_insve_b (v16i8, imm0_15, v16i8);
14541 v8i16 __builtin_msa_insve_h (v8i16, imm0_7, v8i16);
14542 v4i32 __builtin_msa_insve_w (v4i32, imm0_3, v4i32);
14543 v2i64 __builtin_msa_insve_d (v2i64, imm0_1, v2i64);
14545 v16i8 __builtin_msa_ld_b (void *, imm_n512_511);
14546 v8i16 __builtin_msa_ld_h (void *, imm_n1024_1022);
14547 v4i32 __builtin_msa_ld_w (void *, imm_n2048_2044);
14548 v2i64 __builtin_msa_ld_d (void *, imm_n4096_4088);
14550 v16i8 __builtin_msa_ldi_b (imm_n512_511);
14551 v8i16 __builtin_msa_ldi_h (imm_n512_511);
14552 v4i32 __builtin_msa_ldi_w (imm_n512_511);
14553 v2i64 __builtin_msa_ldi_d (imm_n512_511);
14555 v8i16 __builtin_msa_madd_q_h (v8i16, v8i16, v8i16);
14556 v4i32 __builtin_msa_madd_q_w (v4i32, v4i32, v4i32);
14558 v8i16 __builtin_msa_maddr_q_h (v8i16, v8i16, v8i16);
14559 v4i32 __builtin_msa_maddr_q_w (v4i32, v4i32, v4i32);
14561 v16i8 __builtin_msa_maddv_b (v16i8, v16i8, v16i8);
14562 v8i16 __builtin_msa_maddv_h (v8i16, v8i16, v8i16);
14563 v4i32 __builtin_msa_maddv_w (v4i32, v4i32, v4i32);
14564 v2i64 __builtin_msa_maddv_d (v2i64, v2i64, v2i64);
14566 v16i8 __builtin_msa_max_a_b (v16i8, v16i8);
14567 v8i16 __builtin_msa_max_a_h (v8i16, v8i16);
14568 v4i32 __builtin_msa_max_a_w (v4i32, v4i32);
14569 v2i64 __builtin_msa_max_a_d (v2i64, v2i64);
14571 v16i8 __builtin_msa_max_s_b (v16i8, v16i8);
14572 v8i16 __builtin_msa_max_s_h (v8i16, v8i16);
14573 v4i32 __builtin_msa_max_s_w (v4i32, v4i32);
14574 v2i64 __builtin_msa_max_s_d (v2i64, v2i64);
14576 v16u8 __builtin_msa_max_u_b (v16u8, v16u8);
14577 v8u16 __builtin_msa_max_u_h (v8u16, v8u16);
14578 v4u32 __builtin_msa_max_u_w (v4u32, v4u32);
14579 v2u64 __builtin_msa_max_u_d (v2u64, v2u64);
14581 v16i8 __builtin_msa_maxi_s_b (v16i8, imm_n16_15);
14582 v8i16 __builtin_msa_maxi_s_h (v8i16, imm_n16_15);
14583 v4i32 __builtin_msa_maxi_s_w (v4i32, imm_n16_15);
14584 v2i64 __builtin_msa_maxi_s_d (v2i64, imm_n16_15);
14586 v16u8 __builtin_msa_maxi_u_b (v16u8, imm0_31);
14587 v8u16 __builtin_msa_maxi_u_h (v8u16, imm0_31);
14588 v4u32 __builtin_msa_maxi_u_w (v4u32, imm0_31);
14589 v2u64 __builtin_msa_maxi_u_d (v2u64, imm0_31);
14591 v16i8 __builtin_msa_min_a_b (v16i8, v16i8);
14592 v8i16 __builtin_msa_min_a_h (v8i16, v8i16);
14593 v4i32 __builtin_msa_min_a_w (v4i32, v4i32);
14594 v2i64 __builtin_msa_min_a_d (v2i64, v2i64);
14596 v16i8 __builtin_msa_min_s_b (v16i8, v16i8);
14597 v8i16 __builtin_msa_min_s_h (v8i16, v8i16);
14598 v4i32 __builtin_msa_min_s_w (v4i32, v4i32);
14599 v2i64 __builtin_msa_min_s_d (v2i64, v2i64);
14601 v16u8 __builtin_msa_min_u_b (v16u8, v16u8);
14602 v8u16 __builtin_msa_min_u_h (v8u16, v8u16);
14603 v4u32 __builtin_msa_min_u_w (v4u32, v4u32);
14604 v2u64 __builtin_msa_min_u_d (v2u64, v2u64);
14606 v16i8 __builtin_msa_mini_s_b (v16i8, imm_n16_15);
14607 v8i16 __builtin_msa_mini_s_h (v8i16, imm_n16_15);
14608 v4i32 __builtin_msa_mini_s_w (v4i32, imm_n16_15);
14609 v2i64 __builtin_msa_mini_s_d (v2i64, imm_n16_15);
14611 v16u8 __builtin_msa_mini_u_b (v16u8, imm0_31);
14612 v8u16 __builtin_msa_mini_u_h (v8u16, imm0_31);
14613 v4u32 __builtin_msa_mini_u_w (v4u32, imm0_31);
14614 v2u64 __builtin_msa_mini_u_d (v2u64, imm0_31);
14616 v16i8 __builtin_msa_mod_s_b (v16i8, v16i8);
14617 v8i16 __builtin_msa_mod_s_h (v8i16, v8i16);
14618 v4i32 __builtin_msa_mod_s_w (v4i32, v4i32);
14619 v2i64 __builtin_msa_mod_s_d (v2i64, v2i64);
14621 v16u8 __builtin_msa_mod_u_b (v16u8, v16u8);
14622 v8u16 __builtin_msa_mod_u_h (v8u16, v8u16);
14623 v4u32 __builtin_msa_mod_u_w (v4u32, v4u32);
14624 v2u64 __builtin_msa_mod_u_d (v2u64, v2u64);
14626 v16i8 __builtin_msa_move_v (v16i8);
14628 v8i16 __builtin_msa_msub_q_h (v8i16, v8i16, v8i16);
14629 v4i32 __builtin_msa_msub_q_w (v4i32, v4i32, v4i32);
14631 v8i16 __builtin_msa_msubr_q_h (v8i16, v8i16, v8i16);
14632 v4i32 __builtin_msa_msubr_q_w (v4i32, v4i32, v4i32);
14634 v16i8 __builtin_msa_msubv_b (v16i8, v16i8, v16i8);
14635 v8i16 __builtin_msa_msubv_h (v8i16, v8i16, v8i16);
14636 v4i32 __builtin_msa_msubv_w (v4i32, v4i32, v4i32);
14637 v2i64 __builtin_msa_msubv_d (v2i64, v2i64, v2i64);
14639 v8i16 __builtin_msa_mul_q_h (v8i16, v8i16);
14640 v4i32 __builtin_msa_mul_q_w (v4i32, v4i32);
14642 v8i16 __builtin_msa_mulr_q_h (v8i16, v8i16);
14643 v4i32 __builtin_msa_mulr_q_w (v4i32, v4i32);
14645 v16i8 __builtin_msa_mulv_b (v16i8, v16i8);
14646 v8i16 __builtin_msa_mulv_h (v8i16, v8i16);
14647 v4i32 __builtin_msa_mulv_w (v4i32, v4i32);
14648 v2i64 __builtin_msa_mulv_d (v2i64, v2i64);
14650 v16i8 __builtin_msa_nloc_b (v16i8);
14651 v8i16 __builtin_msa_nloc_h (v8i16);
14652 v4i32 __builtin_msa_nloc_w (v4i32);
14653 v2i64 __builtin_msa_nloc_d (v2i64);
14655 v16i8 __builtin_msa_nlzc_b (v16i8);
14656 v8i16 __builtin_msa_nlzc_h (v8i16);
14657 v4i32 __builtin_msa_nlzc_w (v4i32);
14658 v2i64 __builtin_msa_nlzc_d (v2i64);
14660 v16u8 __builtin_msa_nor_v (v16u8, v16u8);
14662 v16u8 __builtin_msa_nori_b (v16u8, imm0_255);
14664 v16u8 __builtin_msa_or_v (v16u8, v16u8);
14666 v16u8 __builtin_msa_ori_b (v16u8, imm0_255);
14668 v16i8 __builtin_msa_pckev_b (v16i8, v16i8);
14669 v8i16 __builtin_msa_pckev_h (v8i16, v8i16);
14670 v4i32 __builtin_msa_pckev_w (v4i32, v4i32);
14671 v2i64 __builtin_msa_pckev_d (v2i64, v2i64);
14673 v16i8 __builtin_msa_pckod_b (v16i8, v16i8);
14674 v8i16 __builtin_msa_pckod_h (v8i16, v8i16);
14675 v4i32 __builtin_msa_pckod_w (v4i32, v4i32);
14676 v2i64 __builtin_msa_pckod_d (v2i64, v2i64);
14678 v16i8 __builtin_msa_pcnt_b (v16i8);
14679 v8i16 __builtin_msa_pcnt_h (v8i16);
14680 v4i32 __builtin_msa_pcnt_w (v4i32);
14681 v2i64 __builtin_msa_pcnt_d (v2i64);
14683 v16i8 __builtin_msa_sat_s_b (v16i8, imm0_7);
14684 v8i16 __builtin_msa_sat_s_h (v8i16, imm0_15);
14685 v4i32 __builtin_msa_sat_s_w (v4i32, imm0_31);
14686 v2i64 __builtin_msa_sat_s_d (v2i64, imm0_63);
14688 v16u8 __builtin_msa_sat_u_b (v16u8, imm0_7);
14689 v8u16 __builtin_msa_sat_u_h (v8u16, imm0_15);
14690 v4u32 __builtin_msa_sat_u_w (v4u32, imm0_31);
14691 v2u64 __builtin_msa_sat_u_d (v2u64, imm0_63);
14693 v16i8 __builtin_msa_shf_b (v16i8, imm0_255);
14694 v8i16 __builtin_msa_shf_h (v8i16, imm0_255);
14695 v4i32 __builtin_msa_shf_w (v4i32, imm0_255);
14697 v16i8 __builtin_msa_sld_b (v16i8, v16i8, i32);
14698 v8i16 __builtin_msa_sld_h (v8i16, v8i16, i32);
14699 v4i32 __builtin_msa_sld_w (v4i32, v4i32, i32);
14700 v2i64 __builtin_msa_sld_d (v2i64, v2i64, i32);
14702 v16i8 __builtin_msa_sldi_b (v16i8, v16i8, imm0_15);
14703 v8i16 __builtin_msa_sldi_h (v8i16, v8i16, imm0_7);
14704 v4i32 __builtin_msa_sldi_w (v4i32, v4i32, imm0_3);
14705 v2i64 __builtin_msa_sldi_d (v2i64, v2i64, imm0_1);
14707 v16i8 __builtin_msa_sll_b (v16i8, v16i8);
14708 v8i16 __builtin_msa_sll_h (v8i16, v8i16);
14709 v4i32 __builtin_msa_sll_w (v4i32, v4i32);
14710 v2i64 __builtin_msa_sll_d (v2i64, v2i64);
14712 v16i8 __builtin_msa_slli_b (v16i8, imm0_7);
14713 v8i16 __builtin_msa_slli_h (v8i16, imm0_15);
14714 v4i32 __builtin_msa_slli_w (v4i32, imm0_31);
14715 v2i64 __builtin_msa_slli_d (v2i64, imm0_63);
14717 v16i8 __builtin_msa_splat_b (v16i8, i32);
14718 v8i16 __builtin_msa_splat_h (v8i16, i32);
14719 v4i32 __builtin_msa_splat_w (v4i32, i32);
14720 v2i64 __builtin_msa_splat_d (v2i64, i32);
14722 v16i8 __builtin_msa_splati_b (v16i8, imm0_15);
14723 v8i16 __builtin_msa_splati_h (v8i16, imm0_7);
14724 v4i32 __builtin_msa_splati_w (v4i32, imm0_3);
14725 v2i64 __builtin_msa_splati_d (v2i64, imm0_1);
14727 v16i8 __builtin_msa_sra_b (v16i8, v16i8);
14728 v8i16 __builtin_msa_sra_h (v8i16, v8i16);
14729 v4i32 __builtin_msa_sra_w (v4i32, v4i32);
14730 v2i64 __builtin_msa_sra_d (v2i64, v2i64);
14732 v16i8 __builtin_msa_srai_b (v16i8, imm0_7);
14733 v8i16 __builtin_msa_srai_h (v8i16, imm0_15);
14734 v4i32 __builtin_msa_srai_w (v4i32, imm0_31);
14735 v2i64 __builtin_msa_srai_d (v2i64, imm0_63);
14737 v16i8 __builtin_msa_srar_b (v16i8, v16i8);
14738 v8i16 __builtin_msa_srar_h (v8i16, v8i16);
14739 v4i32 __builtin_msa_srar_w (v4i32, v4i32);
14740 v2i64 __builtin_msa_srar_d (v2i64, v2i64);
14742 v16i8 __builtin_msa_srari_b (v16i8, imm0_7);
14743 v8i16 __builtin_msa_srari_h (v8i16, imm0_15);
14744 v4i32 __builtin_msa_srari_w (v4i32, imm0_31);
14745 v2i64 __builtin_msa_srari_d (v2i64, imm0_63);
14747 v16i8 __builtin_msa_srl_b (v16i8, v16i8);
14748 v8i16 __builtin_msa_srl_h (v8i16, v8i16);
14749 v4i32 __builtin_msa_srl_w (v4i32, v4i32);
14750 v2i64 __builtin_msa_srl_d (v2i64, v2i64);
14752 v16i8 __builtin_msa_srli_b (v16i8, imm0_7);
14753 v8i16 __builtin_msa_srli_h (v8i16, imm0_15);
14754 v4i32 __builtin_msa_srli_w (v4i32, imm0_31);
14755 v2i64 __builtin_msa_srli_d (v2i64, imm0_63);
14757 v16i8 __builtin_msa_srlr_b (v16i8, v16i8);
14758 v8i16 __builtin_msa_srlr_h (v8i16, v8i16);
14759 v4i32 __builtin_msa_srlr_w (v4i32, v4i32);
14760 v2i64 __builtin_msa_srlr_d (v2i64, v2i64);
14762 v16i8 __builtin_msa_srlri_b (v16i8, imm0_7);
14763 v8i16 __builtin_msa_srlri_h (v8i16, imm0_15);
14764 v4i32 __builtin_msa_srlri_w (v4i32, imm0_31);
14765 v2i64 __builtin_msa_srlri_d (v2i64, imm0_63);
14767 void __builtin_msa_st_b (v16i8, void *, imm_n512_511);
14768 void __builtin_msa_st_h (v8i16, void *, imm_n1024_1022);
14769 void __builtin_msa_st_w (v4i32, void *, imm_n2048_2044);
14770 void __builtin_msa_st_d (v2i64, void *, imm_n4096_4088);
14772 v16i8 __builtin_msa_subs_s_b (v16i8, v16i8);
14773 v8i16 __builtin_msa_subs_s_h (v8i16, v8i16);
14774 v4i32 __builtin_msa_subs_s_w (v4i32, v4i32);
14775 v2i64 __builtin_msa_subs_s_d (v2i64, v2i64);
14777 v16u8 __builtin_msa_subs_u_b (v16u8, v16u8);
14778 v8u16 __builtin_msa_subs_u_h (v8u16, v8u16);
14779 v4u32 __builtin_msa_subs_u_w (v4u32, v4u32);
14780 v2u64 __builtin_msa_subs_u_d (v2u64, v2u64);
14782 v16u8 __builtin_msa_subsus_u_b (v16u8, v16i8);
14783 v8u16 __builtin_msa_subsus_u_h (v8u16, v8i16);
14784 v4u32 __builtin_msa_subsus_u_w (v4u32, v4i32);
14785 v2u64 __builtin_msa_subsus_u_d (v2u64, v2i64);
14787 v16i8 __builtin_msa_subsuu_s_b (v16u8, v16u8);
14788 v8i16 __builtin_msa_subsuu_s_h (v8u16, v8u16);
14789 v4i32 __builtin_msa_subsuu_s_w (v4u32, v4u32);
14790 v2i64 __builtin_msa_subsuu_s_d (v2u64, v2u64);
14792 v16i8 __builtin_msa_subv_b (v16i8, v16i8);
14793 v8i16 __builtin_msa_subv_h (v8i16, v8i16);
14794 v4i32 __builtin_msa_subv_w (v4i32, v4i32);
14795 v2i64 __builtin_msa_subv_d (v2i64, v2i64);
14797 v16i8 __builtin_msa_subvi_b (v16i8, imm0_31);
14798 v8i16 __builtin_msa_subvi_h (v8i16, imm0_31);
14799 v4i32 __builtin_msa_subvi_w (v4i32, imm0_31);
14800 v2i64 __builtin_msa_subvi_d (v2i64, imm0_31);
14802 v16i8 __builtin_msa_vshf_b (v16i8, v16i8, v16i8);
14803 v8i16 __builtin_msa_vshf_h (v8i16, v8i16, v8i16);
14804 v4i32 __builtin_msa_vshf_w (v4i32, v4i32, v4i32);
14805 v2i64 __builtin_msa_vshf_d (v2i64, v2i64, v2i64);
14807 v16u8 __builtin_msa_xor_v (v16u8, v16u8);
14809 v16u8 __builtin_msa_xori_b (v16u8, imm0_255);
14812 @node Other MIPS Built-in Functions
14813 @subsection Other MIPS Built-in Functions
14815 GCC provides other MIPS-specific built-in functions:
14818 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
14819 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
14820 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
14821 when this function is available.
14823 @item unsigned int __builtin_mips_get_fcsr (void)
14824 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
14825 Get and set the contents of the floating-point control and status register
14826 (FPU control register 31). These functions are only available in hard-float
14827 code but can be called in both MIPS16 and non-MIPS16 contexts.
14829 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
14830 register except the condition codes, which GCC assumes are preserved.
14833 @node MSP430 Built-in Functions
14834 @subsection MSP430 Built-in Functions
14836 GCC provides a couple of special builtin functions to aid in the
14837 writing of interrupt handlers in C.
14840 @item __bic_SR_register_on_exit (int @var{mask})
14841 This clears the indicated bits in the saved copy of the status register
14842 currently residing on the stack. This only works inside interrupt
14843 handlers and the changes to the status register will only take affect
14844 once the handler returns.
14846 @item __bis_SR_register_on_exit (int @var{mask})
14847 This sets the indicated bits in the saved copy of the status register
14848 currently residing on the stack. This only works inside interrupt
14849 handlers and the changes to the status register will only take affect
14850 once the handler returns.
14852 @item __delay_cycles (long long @var{cycles})
14853 This inserts an instruction sequence that takes exactly @var{cycles}
14854 cycles (between 0 and about 17E9) to complete. The inserted sequence
14855 may use jumps, loops, or no-ops, and does not interfere with any other
14856 instructions. Note that @var{cycles} must be a compile-time constant
14857 integer - that is, you must pass a number, not a variable that may be
14858 optimized to a constant later. The number of cycles delayed by this
14862 @node NDS32 Built-in Functions
14863 @subsection NDS32 Built-in Functions
14865 These built-in functions are available for the NDS32 target:
14867 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
14868 Insert an ISYNC instruction into the instruction stream where
14869 @var{addr} is an instruction address for serialization.
14872 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
14873 Insert an ISB instruction into the instruction stream.
14876 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
14877 Return the content of a system register which is mapped by @var{sr}.
14880 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
14881 Return the content of a user space register which is mapped by @var{usr}.
14884 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
14885 Move the @var{value} to a system register which is mapped by @var{sr}.
14888 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
14889 Move the @var{value} to a user space register which is mapped by @var{usr}.
14892 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
14893 Enable global interrupt.
14896 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
14897 Disable global interrupt.
14900 @node picoChip Built-in Functions
14901 @subsection picoChip Built-in Functions
14903 GCC provides an interface to selected machine instructions from the
14904 picoChip instruction set.
14907 @item int __builtin_sbc (int @var{value})
14908 Sign bit count. Return the number of consecutive bits in @var{value}
14909 that have the same value as the sign bit. The result is the number of
14910 leading sign bits minus one, giving the number of redundant sign bits in
14913 @item int __builtin_byteswap (int @var{value})
14914 Byte swap. Return the result of swapping the upper and lower bytes of
14917 @item int __builtin_brev (int @var{value})
14918 Bit reversal. Return the result of reversing the bits in
14919 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
14922 @item int __builtin_adds (int @var{x}, int @var{y})
14923 Saturating addition. Return the result of adding @var{x} and @var{y},
14924 storing the value 32767 if the result overflows.
14926 @item int __builtin_subs (int @var{x}, int @var{y})
14927 Saturating subtraction. Return the result of subtracting @var{y} from
14928 @var{x}, storing the value @minus{}32768 if the result overflows.
14930 @item void __builtin_halt (void)
14931 Halt. The processor stops execution. This built-in is useful for
14932 implementing assertions.
14936 @node PowerPC Built-in Functions
14937 @subsection PowerPC Built-in Functions
14939 The following built-in functions are always available and can be used to
14940 check the PowerPC target platform type:
14942 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
14943 This function is a @code{nop} on the PowerPC platform and is included solely
14944 to maintain API compatibility with the x86 builtins.
14947 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
14948 This function returns a value of @code{1} if the run-time CPU is of type
14949 @var{cpuname} and returns @code{0} otherwise
14951 The @code{__builtin_cpu_is} function requires GLIBC 2.23 or newer
14952 which exports the hardware capability bits. GCC defines the macro
14953 @code{__BUILTIN_CPU_SUPPORTS__} if the @code{__builtin_cpu_supports}
14954 built-in function is fully supported.
14956 If GCC was configured to use a GLIBC before 2.23, the built-in
14957 function @code{__builtin_cpu_is} always returns a 0 and the compiler
14960 The following CPU names can be detected:
14964 IBM POWER9 Server CPU.
14966 IBM POWER8 Server CPU.
14968 IBM POWER7 Server CPU.
14970 IBM POWER6 Server CPU (RAW mode).
14972 IBM POWER6 Server CPU (Architected mode).
14974 IBM POWER5+ Server CPU.
14976 IBM POWER5 Server CPU.
14978 IBM 970 Server CPU (ie, Apple G5).
14980 IBM POWER4 Server CPU.
14982 IBM A2 64-bit Embedded CPU
14984 IBM PowerPC 476FP 32-bit Embedded CPU.
14986 IBM PowerPC 464 32-bit Embedded CPU.
14988 PowerPC 440 32-bit Embedded CPU.
14990 PowerPC 405 32-bit Embedded CPU.
14992 IBM PowerPC Cell Broadband Engine Architecture CPU.
14995 Here is an example:
14997 #ifdef __BUILTIN_CPU_SUPPORTS__
14998 if (__builtin_cpu_is ("power8"))
15000 do_power8 (); // POWER8 specific implementation.
15005 do_generic (); // Generic implementation.
15010 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
15011 This function returns a value of @code{1} if the run-time CPU supports the HWCAP
15012 feature @var{feature} and returns @code{0} otherwise.
15014 The @code{__builtin_cpu_supports} function requires GLIBC 2.23 or
15015 newer which exports the hardware capability bits. GCC defines the
15016 macro @code{__BUILTIN_CPU_SUPPORTS__} if the
15017 @code{__builtin_cpu_supports} built-in function is fully supported.
15019 If GCC was configured to use a GLIBC before 2.23, the built-in
15020 function @code{__builtin_cpu_suports} always returns a 0 and the
15021 compiler issues a warning.
15023 The following features can be
15028 4xx CPU has a Multiply Accumulator.
15030 CPU has a SIMD/Vector Unit.
15032 CPU supports ISA 2.05 (eg, POWER6)
15034 CPU supports ISA 2.06 (eg, POWER7)
15036 CPU supports ISA 2.07 (eg, POWER8)
15038 CPU supports ISA 3.0 (eg, POWER9)
15040 CPU supports the set of compatible performance monitoring events.
15042 CPU supports the Embedded ISA category.
15044 CPU has a CELL broadband engine.
15046 CPU has a decimal floating point unit.
15048 CPU supports the data stream control register.
15050 CPU supports event base branching.
15052 CPU has a SPE double precision floating point unit.
15054 CPU has a SPE single precision floating point unit.
15056 CPU has a floating point unit.
15058 CPU has hardware transaction memory instructions.
15060 Kernel aborts hardware transactions when a syscall is made.
15062 CPU supports icache snooping capabilities.
15064 CPU supports 128-bit IEEE binary floating point instructions.
15066 CPU supports the integer select instruction.
15068 CPU has a memory management unit.
15070 CPU does not have a timebase (eg, 601 and 403gx).
15072 CPU supports the PA Semi 6T CORE ISA.
15074 CPU supports ISA 2.00 (eg, POWER4)
15076 CPU supports ISA 2.02 (eg, POWER5)
15078 CPU supports ISA 2.03 (eg, POWER5+)
15080 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
15082 CPU supports 32-bit mode execution.
15084 CPU supports the old POWER ISA (eg, 601)
15086 CPU supports 64-bit mode execution.
15088 CPU supports a little-endian mode that uses address swizzling.
15090 CPU support simultaneous multi-threading.
15092 CPU has a signal processing extension unit.
15094 CPU supports the target address register.
15096 CPU supports true little-endian mode.
15098 CPU has unified I/D cache.
15100 CPU supports the vector cryptography instructions.
15102 CPU supports the vector-scalar extension.
15105 Here is an example:
15107 #ifdef __BUILTIN_CPU_SUPPORTS__
15108 if (__builtin_cpu_supports ("fpu"))
15110 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
15115 dst = __fadd (src1, src2); // Software FP addition function.
15120 These built-in functions are available for the PowerPC family of
15123 float __builtin_recipdivf (float, float);
15124 float __builtin_rsqrtf (float);
15125 double __builtin_recipdiv (double, double);
15126 double __builtin_rsqrt (double);
15127 uint64_t __builtin_ppc_get_timebase ();
15128 unsigned long __builtin_ppc_mftb ();
15129 double __builtin_unpack_longdouble (long double, int);
15130 long double __builtin_pack_longdouble (double, double);
15133 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
15134 @code{__builtin_rsqrtf} functions generate multiple instructions to
15135 implement the reciprocal sqrt functionality using reciprocal sqrt
15136 estimate instructions.
15138 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
15139 functions generate multiple instructions to implement division using
15140 the reciprocal estimate instructions.
15142 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
15143 functions generate instructions to read the Time Base Register. The
15144 @code{__builtin_ppc_get_timebase} function may generate multiple
15145 instructions and always returns the 64 bits of the Time Base Register.
15146 The @code{__builtin_ppc_mftb} function always generates one instruction and
15147 returns the Time Base Register value as an unsigned long, throwing away
15148 the most significant word on 32-bit environments.
15150 Additional built-in functions are available for the 64-bit PowerPC
15151 family of processors, for efficient use of 128-bit floating point
15152 (@code{__float128}) values.
15154 The following floating-point built-in functions are available with
15155 @code{-mfloat128} and Altivec support. All of them implement the
15156 function that is part of the name.
15159 __float128 __builtin_fabsq (__float128)
15160 __float128 __builtin_copysignq (__float128, __float128)
15163 The following built-in functions are available with @code{-mfloat128}
15164 and Altivec support.
15167 @item __float128 __builtin_infq (void)
15168 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
15169 @findex __builtin_infq
15171 @item __float128 __builtin_huge_valq (void)
15172 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
15173 @findex __builtin_huge_valq
15175 @item __float128 __builtin_nanq (void)
15176 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
15177 @findex __builtin_nanq
15179 @item __float128 __builtin_nansq (void)
15180 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
15181 @findex __builtin_nansq
15184 The following built-in functions are available for the PowerPC family
15185 of processors, starting with ISA 2.05 or later (@option{-mcpu=power6}
15186 or @option{-mcmpb}):
15188 unsigned long long __builtin_cmpb (unsigned long long int, unsigned long long int);
15189 unsigned int __builtin_cmpb (unsigned int, unsigned int);
15192 The @code{__builtin_cmpb} function
15193 performs a byte-wise compare on the contents of its two arguments,
15194 returning the result of the byte-wise comparison as the returned
15195 value. For each byte comparison, the corresponding byte of the return
15196 value holds 0xff if the input bytes are equal and 0 if the input bytes
15197 are not equal. If either of the arguments to this built-in function
15198 is wider than 32 bits, the function call expands into the form that
15199 expects @code{unsigned long long int} arguments
15200 which is only available on 64-bit targets.
15202 The following built-in functions are available for the PowerPC family
15203 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
15204 or @option{-mpopcntd}):
15206 long __builtin_bpermd (long, long);
15207 int __builtin_divwe (int, int);
15208 int __builtin_divweo (int, int);
15209 unsigned int __builtin_divweu (unsigned int, unsigned int);
15210 unsigned int __builtin_divweuo (unsigned int, unsigned int);
15211 long __builtin_divde (long, long);
15212 long __builtin_divdeo (long, long);
15213 unsigned long __builtin_divdeu (unsigned long, unsigned long);
15214 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
15215 unsigned int cdtbcd (unsigned int);
15216 unsigned int cbcdtd (unsigned int);
15217 unsigned int addg6s (unsigned int, unsigned int);
15220 The @code{__builtin_divde}, @code{__builtin_divdeo},
15221 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
15222 64-bit environment support ISA 2.06 or later.
15224 The following built-in functions are available for the PowerPC family
15225 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
15227 long long __builtin_darn (void);
15228 long long __builtin_darn_raw (void);
15229 int __builtin_darn_32 (void);
15231 unsigned int scalar_extract_exp (double source);
15232 unsigned long long int scalar_extract_exp (__ieee128 source);
15234 unsigned long long int scalar_extract_sig (double source);
15235 unsigned __int128 scalar_extract_sig (__ieee128 source);
15238 scalar_insert_exp (unsigned long long int significand, unsigned long long int exponent);
15240 scalar_insert_exp (double significand, unsigned long long int exponent);
15243 scalar_insert_exp (unsigned __int128 significand, unsigned long long int exponent);
15245 scalar_insert_exp (ieee_128 significand, unsigned long long int exponent);
15247 int scalar_cmp_exp_gt (double arg1, double arg2);
15248 int scalar_cmp_exp_lt (double arg1, double arg2);
15249 int scalar_cmp_exp_eq (double arg1, double arg2);
15250 int scalar_cmp_exp_unordered (double arg1, double arg2);
15252 bool scalar_test_data_class (float source, const int condition);
15253 bool scalar_test_data_class (double source, const int condition);
15254 bool scalar_test_data_class (__ieee128 source, const int condition);
15256 bool scalar_test_neg (float source);
15257 bool scalar_test_neg (double source);
15258 bool scalar_test_neg (__ieee128 source);
15260 int __builtin_byte_in_set (unsigned char u, unsigned long long set);
15261 int __builtin_byte_in_range (unsigned char u, unsigned int range);
15262 int __builtin_byte_in_either_range (unsigned char u, unsigned int ranges);
15264 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal64 value);
15265 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal128 value);
15266 int __builtin_dfp_dtstsfi_lt_dd (unsigned int comparison, _Decimal64 value);
15267 int __builtin_dfp_dtstsfi_lt_td (unsigned int comparison, _Decimal128 value);
15269 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal64 value);
15270 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal128 value);
15271 int __builtin_dfp_dtstsfi_gt_dd (unsigned int comparison, _Decimal64 value);
15272 int __builtin_dfp_dtstsfi_gt_td (unsigned int comparison, _Decimal128 value);
15274 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal64 value);
15275 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal128 value);
15276 int __builtin_dfp_dtstsfi_eq_dd (unsigned int comparison, _Decimal64 value);
15277 int __builtin_dfp_dtstsfi_eq_td (unsigned int comparison, _Decimal128 value);
15279 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal64 value);
15280 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal128 value);
15281 int __builtin_dfp_dtstsfi_ov_dd (unsigned int comparison, _Decimal64 value);
15282 int __builtin_dfp_dtstsfi_ov_td (unsigned int comparison, _Decimal128 value);
15285 The @code{__builtin_darn} and @code{__builtin_darn_raw}
15286 functions require a
15287 64-bit environment supporting ISA 3.0 or later.
15288 The @code{__builtin_darn} function provides a 64-bit conditioned
15289 random number. The @code{__builtin_darn_raw} function provides a
15290 64-bit raw random number. The @code{__builtin_darn_32} function
15291 provides a 32-bit random number.
15293 The @code{scalar_extract_exp} and @code{scalar_extract_sig}
15294 functions require a 64-bit environment supporting ISA 3.0 or later.
15295 The @code{scalar_extract_exp} and @code{scalar_extract_sig} built-in
15296 functions return the significand and the biased exponent value
15297 respectively of their @code{source} arguments.
15298 When supplied with a 64-bit @code{source} argument, the
15299 result returned by @code{scalar_extract_sig} has
15300 the @code{0x0010000000000000} bit set if the
15301 function's @code{source} argument is in normalized form.
15302 Otherwise, this bit is set to 0.
15303 When supplied with a 128-bit @code{source} argument, the
15304 @code{0x00010000000000000000000000000000} bit of the result is
15306 Note that the sign of the significand is not represented in the result
15307 returned from the @code{scalar_extract_sig} function. Use the
15308 @code{scalar_test_neg} function to test the sign of its @code{double}
15311 The @code{scalar_insert_exp}
15312 functions require a 64-bit environment supporting ISA 3.0 or later.
15313 When supplied with a 64-bit first argument, the
15314 @code{scalar_insert_exp} built-in function returns a double-precision
15315 floating point value that is constructed by assembling the values of its
15316 @code{significand} and @code{exponent} arguments. The sign of the
15317 result is copied from the most significant bit of the
15318 @code{significand} argument. The significand and exponent components
15319 of the result are composed of the least significant 11 bits of the
15320 @code{exponent} argument and the least significant 52 bits of the
15321 @code{significand} argument respectively.
15323 When supplied with a 128-bit first argument, the
15324 @code{scalar_insert_exp} built-in function returns a quad-precision
15325 ieee floating point value. The sign bit of the result is copied from
15326 the most significant bit of the @code{significand} argument.
15327 The significand and exponent components of the result are composed of
15328 the least significant 15 bits of the @code{exponent} argument and the
15329 least significant 112 bits of the @code{significand} argument respectively.
15331 The @code{scalar_cmp_exp_gt}, @code{scalar_cmp_exp_lt},
15332 @code{scalar_cmp_exp_eq}, and @code{scalar_cmp_exp_unordered} built-in
15333 functions return a non-zero value if @code{arg1} is greater than, less
15334 than, equal to, or not comparable to @code{arg2} respectively. The
15335 arguments are not comparable if one or the other equals NaN (not a
15338 The @code{scalar_test_data_class} built-in function returns 1
15339 if any of the condition tests enabled by the value of the
15340 @code{condition} variable are true, and 0 otherwise. The
15341 @code{condition} argument must be a compile-time constant integer with
15342 value not exceeding 127. The
15343 @code{condition} argument is encoded as a bitmask with each bit
15344 enabling the testing of a different condition, as characterized by the
15348 0x20 Test for +Infinity
15349 0x10 Test for -Infinity
15350 0x08 Test for +Zero
15351 0x04 Test for -Zero
15352 0x02 Test for +Denormal
15353 0x01 Test for -Denormal
15356 The @code{scalar_test_neg} built-in function returns 1 if its
15357 @code{source} argument holds a negative value, 0 otherwise.
15359 The @code{__builtin_byte_in_set} function requires a
15360 64-bit environment supporting ISA 3.0 or later. This function returns
15361 a non-zero value if and only if its @code{u} argument exactly equals one of
15362 the eight bytes contained within its 64-bit @code{set} argument.
15364 The @code{__builtin_byte_in_range} and
15365 @code{__builtin_byte_in_either_range} require an environment
15366 supporting ISA 3.0 or later. For these two functions, the
15367 @code{range} argument is encoded as 4 bytes, organized as
15368 @code{hi_1:lo_1:hi_2:lo_2}.
15369 The @code{__builtin_byte_in_range} function returns a
15370 non-zero value if and only if its @code{u} argument is within the
15371 range bounded between @code{lo_2} and @code{hi_2} inclusive.
15372 The @code{__builtin_byte_in_either_range} function returns non-zero if
15373 and only if its @code{u} argument is within either the range bounded
15374 between @code{lo_1} and @code{hi_1} inclusive or the range bounded
15375 between @code{lo_2} and @code{hi_2} inclusive.
15377 The @code{__builtin_dfp_dtstsfi_lt} function returns a non-zero value
15378 if and only if the number of signficant digits of its @code{value} argument
15379 is less than its @code{comparison} argument. The
15380 @code{__builtin_dfp_dtstsfi_lt_dd} and
15381 @code{__builtin_dfp_dtstsfi_lt_td} functions behave similarly, but
15382 require that the type of the @code{value} argument be
15383 @code{__Decimal64} and @code{__Decimal128} respectively.
15385 The @code{__builtin_dfp_dtstsfi_gt} function returns a non-zero value
15386 if and only if the number of signficant digits of its @code{value} argument
15387 is greater than its @code{comparison} argument. The
15388 @code{__builtin_dfp_dtstsfi_gt_dd} and
15389 @code{__builtin_dfp_dtstsfi_gt_td} functions behave similarly, but
15390 require that the type of the @code{value} argument be
15391 @code{__Decimal64} and @code{__Decimal128} respectively.
15393 The @code{__builtin_dfp_dtstsfi_eq} function returns a non-zero value
15394 if and only if the number of signficant digits of its @code{value} argument
15395 equals its @code{comparison} argument. The
15396 @code{__builtin_dfp_dtstsfi_eq_dd} and
15397 @code{__builtin_dfp_dtstsfi_eq_td} functions behave similarly, but
15398 require that the type of the @code{value} argument be
15399 @code{__Decimal64} and @code{__Decimal128} respectively.
15401 The @code{__builtin_dfp_dtstsfi_ov} function returns a non-zero value
15402 if and only if its @code{value} argument has an undefined number of
15403 significant digits, such as when @code{value} is an encoding of @code{NaN}.
15404 The @code{__builtin_dfp_dtstsfi_ov_dd} and
15405 @code{__builtin_dfp_dtstsfi_ov_td} functions behave similarly, but
15406 require that the type of the @code{value} argument be
15407 @code{__Decimal64} and @code{__Decimal128} respectively.
15409 The following built-in functions are also available for the PowerPC family
15410 of processors, starting with ISA 3.0 or later
15411 (@option{-mcpu=power9}). These string functions are described
15412 separately in order to group the descriptions closer to the function
15415 int vec_all_nez (vector signed char, vector signed char);
15416 int vec_all_nez (vector unsigned char, vector unsigned char);
15417 int vec_all_nez (vector signed short, vector signed short);
15418 int vec_all_nez (vector unsigned short, vector unsigned short);
15419 int vec_all_nez (vector signed int, vector signed int);
15420 int vec_all_nez (vector unsigned int, vector unsigned int);
15422 int vec_any_eqz (vector signed char, vector signed char);
15423 int vec_any_eqz (vector unsigned char, vector unsigned char);
15424 int vec_any_eqz (vector signed short, vector signed short);
15425 int vec_any_eqz (vector unsigned short, vector unsigned short);
15426 int vec_any_eqz (vector signed int, vector signed int);
15427 int vec_any_eqz (vector unsigned int, vector unsigned int);
15429 vector bool char vec_cmpnez (vector signed char arg1, vector signed char arg2);
15430 vector bool char vec_cmpnez (vector unsigned char arg1, vector unsigned char arg2);
15431 vector bool short vec_cmpnez (vector signed short arg1, vector signed short arg2);
15432 vector bool short vec_cmpnez (vector unsigned short arg1, vector unsigned short arg2);
15433 vector bool int vec_cmpnez (vector signed int arg1, vector signed int arg2);
15434 vector bool int vec_cmpnez (vector unsigned int, vector unsigned int);
15436 signed int vec_cntlz_lsbb (vector signed char);
15437 signed int vec_cntlz_lsbb (vector unsigned char);
15439 signed int vec_cnttz_lsbb (vector signed char);
15440 signed int vec_cnttz_lsbb (vector unsigned char);
15442 vector unsigned short vec_pack_to_short_fp32 (vector float, vector float);
15444 vector signed char vec_xl_len (signed char *addr, size_t len);
15445 vector unsigned char vec_xl_len (unsigned char *addr, size_t len);
15446 vector signed int vec_xl_len (signed int *addr, size_t len);
15447 vector unsigned int vec_xl_len (unsigned int *addr, size_t len);
15448 vector signed __int128 vec_xl_len (signed __int128 *addr, size_t len);
15449 vector unsigned __int128 vec_xl_len (unsigned __int128 *addr, size_t len);
15450 vector signed long long vec_xl_len (signed long long *addr, size_t len);
15451 vector unsigned long long vec_xl_len (unsigned long long *addr, size_t len);
15452 vector signed short vec_xl_len (signed short *addr, size_t len);
15453 vector unsigned short vec_xl_len (unsigned short *addr, size_t len);
15454 vector double vec_xl_len (double *addr, size_t len);
15455 vector float vec_xl_len (float *addr, size_t len);
15457 void vec_xst_len (vector signed char data, signed char *addr, size_t len);
15458 void vec_xst_len (vector unsigned char data, unsigned char *addr, size_t len);
15459 void vec_xst_len (vector signed int data, signed int *addr, size_t len);
15460 void vec_xst_len (vector unsigned int data, unsigned int *addr, size_t len);
15461 void vec_xst_len (vector unsigned __int128 data, unsigned __int128 *addr, size_t len);
15462 void vec_xst_len (vector signed long long data, signed long long *addr, size_t len);
15463 void vec_xst_len (vector unsigned long long data, unsigned long long *addr, size_t len);
15464 void vec_xst_len (vector signed short data, signed short *addr, size_t len);
15465 void vec_xst_len (vector unsigned short data, unsigned short *addr, size_t len);
15466 void vec_xst_len (vector signed __int128 data, signed __int128 *addr, size_t len);
15467 void vec_xst_len (vector double data, double *addr, size_t len);
15468 void vec_xst_len (vector float data, float *addr, size_t len);
15470 signed char vec_xlx (unsigned int index, vector signed char data);
15471 unsigned char vec_xlx (unsigned int index, vector unsigned char data);
15472 signed short vec_xlx (unsigned int index, vector signed short data);
15473 unsigned short vec_xlx (unsigned int index, vector unsigned short data);
15474 signed int vec_xlx (unsigned int index, vector signed int data);
15475 unsigned int vec_xlx (unsigned int index, vector unsigned int data);
15476 float vec_xlx (unsigned int index, vector float data);
15478 signed char vec_xrx (unsigned int index, vector signed char data);
15479 unsigned char vec_xrx (unsigned int index, vector unsigned char data);
15480 signed short vec_xrx (unsigned int index, vector signed short data);
15481 unsigned short vec_xrx (unsigned int index, vector unsigned short data);
15482 signed int vec_xrx (unsigned int index, vector signed int data);
15483 unsigned int vec_xrx (unsigned int index, vector unsigned int data);
15484 float vec_xrx (unsigned int index, vector float data);
15487 The @code{vec_all_nez}, @code{vec_any_eqz}, and @code{vec_cmpnez}
15488 perform pairwise comparisons between the elements at the same
15489 positions within their two vector arguments.
15490 The @code{vec_all_nez} function returns a
15491 non-zero value if and only if all pairwise comparisons are not
15492 equal and no element of either vector argument contains a zero.
15493 The @code{vec_any_eqz} function returns a
15494 non-zero value if and only if at least one pairwise comparison is equal
15495 or if at least one element of either vector argument contains a zero.
15496 The @code{vec_cmpnez} function returns a vector of the same type as
15497 its two arguments, within which each element consists of all ones to
15498 denote that either the corresponding elements of the incoming arguments are
15499 not equal or that at least one of the corresponding elements contains
15500 zero. Otherwise, the element of the returned vector contains all zeros.
15502 The @code{vec_cntlz_lsbb} function returns the count of the number of
15503 consecutive leading byte elements (starting from position 0 within the
15504 supplied vector argument) for which the least-significant bit
15505 equals zero. The @code{vec_cnttz_lsbb} function returns the count of
15506 the number of consecutive trailing byte elements (starting from
15507 position 15 and counting backwards within the supplied vector
15508 argument) for which the least-significant bit equals zero.
15510 The @code{vec_xl_len} and @code{vec_xst_len} functions require a
15511 64-bit environment supporting ISA 3.0 or later. The @code{vec_xl_len}
15512 function loads a variable length vector from memory. The
15513 @code{vec_xst_len} function stores a variable length vector to memory.
15514 With both the @code{vec_xl_len} and @code{vec_xst_len} functions, the
15515 @code{addr} argument represents the memory address to or from which
15516 data will be transferred, and the
15517 @code{len} argument represents the number of bytes to be
15518 transferred, as computed by the C expression @code{min((len & 0xff), 16)}.
15519 If this expression's value is not a multiple of the vector element's
15520 size, the behavior of this function is undefined.
15521 In the case that the underlying computer is configured to run in
15522 big-endian mode, the data transfer moves bytes 0 to @code{(len - 1)} of
15523 the corresponding vector. In little-endian mode, the data transfer
15524 moves bytes @code{(16 - len)} to @code{15} of the corresponding
15525 vector. For the load function, any bytes of the result vector that
15526 are not loaded from memory are set to zero.
15527 The value of the @code{addr} argument need not be aligned on a
15528 multiple of the vector's element size.
15530 The @code{vec_xlx} and @code{vec_xrx} functions extract the single
15531 element selected by the @code{index} argument from the vector
15532 represented by the @code{data} argument. The @code{index} argument
15533 always specifies a byte offset, regardless of the size of the vector
15534 element. With @code{vec_xlx}, @code{index} is the offset of the first
15535 byte of the element to be extracted. With @code{vec_xrx}, @code{index}
15536 represents the last byte of the element to be extracted, measured
15537 from the right end of the vector. In other words, the last byte of
15538 the element to be extracted is found at position @code{(15 - index)}.
15539 There is no requirement that @code{index} be a multiple of the vector
15540 element size. However, if the size of the vector element added to
15541 @code{index} is greater than 15, the content of the returned value is
15544 The following built-in functions are available for the PowerPC family
15545 of processors when hardware decimal floating point
15546 (@option{-mhard-dfp}) is available:
15548 long long __builtin_dxex (_Decimal64);
15549 long long __builtin_dxexq (_Decimal128);
15550 _Decimal64 __builtin_ddedpd (int, _Decimal64);
15551 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
15552 _Decimal64 __builtin_denbcd (int, _Decimal64);
15553 _Decimal128 __builtin_denbcdq (int, _Decimal128);
15554 _Decimal64 __builtin_diex (long long, _Decimal64);
15555 _Decimal128 _builtin_diexq (long long, _Decimal128);
15556 _Decimal64 __builtin_dscli (_Decimal64, int);
15557 _Decimal128 __builtin_dscliq (_Decimal128, int);
15558 _Decimal64 __builtin_dscri (_Decimal64, int);
15559 _Decimal128 __builtin_dscriq (_Decimal128, int);
15560 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
15561 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
15564 The following built-in functions are available for the PowerPC family
15565 of processors when the Vector Scalar (vsx) instruction set is
15568 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
15569 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
15570 unsigned long long);
15573 @node PowerPC AltiVec/VSX Built-in Functions
15574 @subsection PowerPC AltiVec Built-in Functions
15576 GCC provides an interface for the PowerPC family of processors to access
15577 the AltiVec operations described in Motorola's AltiVec Programming
15578 Interface Manual. The interface is made available by including
15579 @code{<altivec.h>} and using @option{-maltivec} and
15580 @option{-mabi=altivec}. The interface supports the following vector
15584 vector unsigned char
15588 vector unsigned short
15589 vector signed short
15593 vector unsigned int
15599 If @option{-mvsx} is used the following additional vector types are
15603 vector unsigned long
15608 The long types are only implemented for 64-bit code generation, and
15609 the long type is only used in the floating point/integer conversion
15612 GCC's implementation of the high-level language interface available from
15613 C and C++ code differs from Motorola's documentation in several ways.
15618 A vector constant is a list of constant expressions within curly braces.
15621 A vector initializer requires no cast if the vector constant is of the
15622 same type as the variable it is initializing.
15625 If @code{signed} or @code{unsigned} is omitted, the signedness of the
15626 vector type is the default signedness of the base type. The default
15627 varies depending on the operating system, so a portable program should
15628 always specify the signedness.
15631 Compiling with @option{-maltivec} adds keywords @code{__vector},
15632 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
15633 @code{bool}. When compiling ISO C, the context-sensitive substitution
15634 of the keywords @code{vector}, @code{pixel} and @code{bool} is
15635 disabled. To use them, you must include @code{<altivec.h>} instead.
15638 GCC allows using a @code{typedef} name as the type specifier for a
15642 For C, overloaded functions are implemented with macros so the following
15646 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
15650 Since @code{vec_add} is a macro, the vector constant in the example
15651 is treated as four separate arguments. Wrap the entire argument in
15652 parentheses for this to work.
15655 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
15656 Internally, GCC uses built-in functions to achieve the functionality in
15657 the aforementioned header file, but they are not supported and are
15658 subject to change without notice.
15660 GCC complies with the OpenPOWER 64-Bit ELF V2 ABI Specification,
15661 which may be found at
15662 @uref{http://openpowerfoundation.org/wp-content/uploads/resources/leabi-prd/content/index.html}.
15663 Appendix A of this document lists the vector API interfaces that must be
15664 provided by compliant compilers. Programmers should preferentially use
15665 the interfaces described therein. However, historically GCC has provided
15666 additional interfaces for access to vector instructions. These are
15667 briefly described below.
15669 The following interfaces are supported for the generic and specific
15670 AltiVec operations and the AltiVec predicates. In cases where there
15671 is a direct mapping between generic and specific operations, only the
15672 generic names are shown here, although the specific operations can also
15675 Arguments that are documented as @code{const int} require literal
15676 integral values within the range required for that operation.
15679 vector signed char vec_abs (vector signed char);
15680 vector signed short vec_abs (vector signed short);
15681 vector signed int vec_abs (vector signed int);
15682 vector float vec_abs (vector float);
15684 vector signed char vec_abss (vector signed char);
15685 vector signed short vec_abss (vector signed short);
15686 vector signed int vec_abss (vector signed int);
15688 vector signed char vec_add (vector bool char, vector signed char);
15689 vector signed char vec_add (vector signed char, vector bool char);
15690 vector signed char vec_add (vector signed char, vector signed char);
15691 vector unsigned char vec_add (vector bool char, vector unsigned char);
15692 vector unsigned char vec_add (vector unsigned char, vector bool char);
15693 vector unsigned char vec_add (vector unsigned char,
15694 vector unsigned char);
15695 vector signed short vec_add (vector bool short, vector signed short);
15696 vector signed short vec_add (vector signed short, vector bool short);
15697 vector signed short vec_add (vector signed short, vector signed short);
15698 vector unsigned short vec_add (vector bool short,
15699 vector unsigned short);
15700 vector unsigned short vec_add (vector unsigned short,
15701 vector bool short);
15702 vector unsigned short vec_add (vector unsigned short,
15703 vector unsigned short);
15704 vector signed int vec_add (vector bool int, vector signed int);
15705 vector signed int vec_add (vector signed int, vector bool int);
15706 vector signed int vec_add (vector signed int, vector signed int);
15707 vector unsigned int vec_add (vector bool int, vector unsigned int);
15708 vector unsigned int vec_add (vector unsigned int, vector bool int);
15709 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
15710 vector float vec_add (vector float, vector float);
15712 vector float vec_vaddfp (vector float, vector float);
15714 vector signed int vec_vadduwm (vector bool int, vector signed int);
15715 vector signed int vec_vadduwm (vector signed int, vector bool int);
15716 vector signed int vec_vadduwm (vector signed int, vector signed int);
15717 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
15718 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
15719 vector unsigned int vec_vadduwm (vector unsigned int,
15720 vector unsigned int);
15722 vector signed short vec_vadduhm (vector bool short,
15723 vector signed short);
15724 vector signed short vec_vadduhm (vector signed short,
15725 vector bool short);
15726 vector signed short vec_vadduhm (vector signed short,
15727 vector signed short);
15728 vector unsigned short vec_vadduhm (vector bool short,
15729 vector unsigned short);
15730 vector unsigned short vec_vadduhm (vector unsigned short,
15731 vector bool short);
15732 vector unsigned short vec_vadduhm (vector unsigned short,
15733 vector unsigned short);
15735 vector signed char vec_vaddubm (vector bool char, vector signed char);
15736 vector signed char vec_vaddubm (vector signed char, vector bool char);
15737 vector signed char vec_vaddubm (vector signed char, vector signed char);
15738 vector unsigned char vec_vaddubm (vector bool char,
15739 vector unsigned char);
15740 vector unsigned char vec_vaddubm (vector unsigned char,
15742 vector unsigned char vec_vaddubm (vector unsigned char,
15743 vector unsigned char);
15745 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
15747 vector unsigned char vec_adds (vector bool char, vector unsigned char);
15748 vector unsigned char vec_adds (vector unsigned char, vector bool char);
15749 vector unsigned char vec_adds (vector unsigned char,
15750 vector unsigned char);
15751 vector signed char vec_adds (vector bool char, vector signed char);
15752 vector signed char vec_adds (vector signed char, vector bool char);
15753 vector signed char vec_adds (vector signed char, vector signed char);
15754 vector unsigned short vec_adds (vector bool short,
15755 vector unsigned short);
15756 vector unsigned short vec_adds (vector unsigned short,
15757 vector bool short);
15758 vector unsigned short vec_adds (vector unsigned short,
15759 vector unsigned short);
15760 vector signed short vec_adds (vector bool short, vector signed short);
15761 vector signed short vec_adds (vector signed short, vector bool short);
15762 vector signed short vec_adds (vector signed short, vector signed short);
15763 vector unsigned int vec_adds (vector bool int, vector unsigned int);
15764 vector unsigned int vec_adds (vector unsigned int, vector bool int);
15765 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
15766 vector signed int vec_adds (vector bool int, vector signed int);
15767 vector signed int vec_adds (vector signed int, vector bool int);
15768 vector signed int vec_adds (vector signed int, vector signed int);
15770 vector signed int vec_vaddsws (vector bool int, vector signed int);
15771 vector signed int vec_vaddsws (vector signed int, vector bool int);
15772 vector signed int vec_vaddsws (vector signed int, vector signed int);
15774 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
15775 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
15776 vector unsigned int vec_vadduws (vector unsigned int,
15777 vector unsigned int);
15779 vector signed short vec_vaddshs (vector bool short,
15780 vector signed short);
15781 vector signed short vec_vaddshs (vector signed short,
15782 vector bool short);
15783 vector signed short vec_vaddshs (vector signed short,
15784 vector signed short);
15786 vector unsigned short vec_vadduhs (vector bool short,
15787 vector unsigned short);
15788 vector unsigned short vec_vadduhs (vector unsigned short,
15789 vector bool short);
15790 vector unsigned short vec_vadduhs (vector unsigned short,
15791 vector unsigned short);
15793 vector signed char vec_vaddsbs (vector bool char, vector signed char);
15794 vector signed char vec_vaddsbs (vector signed char, vector bool char);
15795 vector signed char vec_vaddsbs (vector signed char, vector signed char);
15797 vector unsigned char vec_vaddubs (vector bool char,
15798 vector unsigned char);
15799 vector unsigned char vec_vaddubs (vector unsigned char,
15801 vector unsigned char vec_vaddubs (vector unsigned char,
15802 vector unsigned char);
15804 vector float vec_and (vector float, vector float);
15805 vector float vec_and (vector float, vector bool int);
15806 vector float vec_and (vector bool int, vector float);
15807 vector bool int vec_and (vector bool int, vector bool int);
15808 vector signed int vec_and (vector bool int, vector signed int);
15809 vector signed int vec_and (vector signed int, vector bool int);
15810 vector signed int vec_and (vector signed int, vector signed int);
15811 vector unsigned int vec_and (vector bool int, vector unsigned int);
15812 vector unsigned int vec_and (vector unsigned int, vector bool int);
15813 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
15814 vector bool short vec_and (vector bool short, vector bool short);
15815 vector signed short vec_and (vector bool short, vector signed short);
15816 vector signed short vec_and (vector signed short, vector bool short);
15817 vector signed short vec_and (vector signed short, vector signed short);
15818 vector unsigned short vec_and (vector bool short,
15819 vector unsigned short);
15820 vector unsigned short vec_and (vector unsigned short,
15821 vector bool short);
15822 vector unsigned short vec_and (vector unsigned short,
15823 vector unsigned short);
15824 vector signed char vec_and (vector bool char, vector signed char);
15825 vector bool char vec_and (vector bool char, vector bool char);
15826 vector signed char vec_and (vector signed char, vector bool char);
15827 vector signed char vec_and (vector signed char, vector signed char);
15828 vector unsigned char vec_and (vector bool char, vector unsigned char);
15829 vector unsigned char vec_and (vector unsigned char, vector bool char);
15830 vector unsigned char vec_and (vector unsigned char,
15831 vector unsigned char);
15833 vector float vec_andc (vector float, vector float);
15834 vector float vec_andc (vector float, vector bool int);
15835 vector float vec_andc (vector bool int, vector float);
15836 vector bool int vec_andc (vector bool int, vector bool int);
15837 vector signed int vec_andc (vector bool int, vector signed int);
15838 vector signed int vec_andc (vector signed int, vector bool int);
15839 vector signed int vec_andc (vector signed int, vector signed int);
15840 vector unsigned int vec_andc (vector bool int, vector unsigned int);
15841 vector unsigned int vec_andc (vector unsigned int, vector bool int);
15842 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
15843 vector bool short vec_andc (vector bool short, vector bool short);
15844 vector signed short vec_andc (vector bool short, vector signed short);
15845 vector signed short vec_andc (vector signed short, vector bool short);
15846 vector signed short vec_andc (vector signed short, vector signed short);
15847 vector unsigned short vec_andc (vector bool short,
15848 vector unsigned short);
15849 vector unsigned short vec_andc (vector unsigned short,
15850 vector bool short);
15851 vector unsigned short vec_andc (vector unsigned short,
15852 vector unsigned short);
15853 vector signed char vec_andc (vector bool char, vector signed char);
15854 vector bool char vec_andc (vector bool char, vector bool char);
15855 vector signed char vec_andc (vector signed char, vector bool char);
15856 vector signed char vec_andc (vector signed char, vector signed char);
15857 vector unsigned char vec_andc (vector bool char, vector unsigned char);
15858 vector unsigned char vec_andc (vector unsigned char, vector bool char);
15859 vector unsigned char vec_andc (vector unsigned char,
15860 vector unsigned char);
15862 vector unsigned char vec_avg (vector unsigned char,
15863 vector unsigned char);
15864 vector signed char vec_avg (vector signed char, vector signed char);
15865 vector unsigned short vec_avg (vector unsigned short,
15866 vector unsigned short);
15867 vector signed short vec_avg (vector signed short, vector signed short);
15868 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
15869 vector signed int vec_avg (vector signed int, vector signed int);
15871 vector signed int vec_vavgsw (vector signed int, vector signed int);
15873 vector unsigned int vec_vavguw (vector unsigned int,
15874 vector unsigned int);
15876 vector signed short vec_vavgsh (vector signed short,
15877 vector signed short);
15879 vector unsigned short vec_vavguh (vector unsigned short,
15880 vector unsigned short);
15882 vector signed char vec_vavgsb (vector signed char, vector signed char);
15884 vector unsigned char vec_vavgub (vector unsigned char,
15885 vector unsigned char);
15887 vector float vec_copysign (vector float);
15889 vector float vec_ceil (vector float);
15891 vector signed int vec_cmpb (vector float, vector float);
15893 vector bool char vec_cmpeq (vector bool char, vector bool char);
15894 vector bool short vec_cmpeq (vector bool short, vector bool short);
15895 vector bool int vec_cmpeq (vector bool int, vector bool int);
15896 vector bool char vec_cmpeq (vector signed char, vector signed char);
15897 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
15898 vector bool short vec_cmpeq (vector signed short, vector signed short);
15899 vector bool short vec_cmpeq (vector unsigned short,
15900 vector unsigned short);
15901 vector bool int vec_cmpeq (vector signed int, vector signed int);
15902 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
15903 vector bool int vec_cmpeq (vector float, vector float);
15905 vector bool int vec_vcmpeqfp (vector float, vector float);
15907 vector bool int vec_vcmpequw (vector signed int, vector signed int);
15908 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
15910 vector bool short vec_vcmpequh (vector signed short,
15911 vector signed short);
15912 vector bool short vec_vcmpequh (vector unsigned short,
15913 vector unsigned short);
15915 vector bool char vec_vcmpequb (vector signed char, vector signed char);
15916 vector bool char vec_vcmpequb (vector unsigned char,
15917 vector unsigned char);
15919 vector bool int vec_cmpge (vector float, vector float);
15921 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
15922 vector bool char vec_cmpgt (vector signed char, vector signed char);
15923 vector bool short vec_cmpgt (vector unsigned short,
15924 vector unsigned short);
15925 vector bool short vec_cmpgt (vector signed short, vector signed short);
15926 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
15927 vector bool int vec_cmpgt (vector signed int, vector signed int);
15928 vector bool int vec_cmpgt (vector float, vector float);
15930 vector bool int vec_vcmpgtfp (vector float, vector float);
15932 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
15934 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
15936 vector bool short vec_vcmpgtsh (vector signed short,
15937 vector signed short);
15939 vector bool short vec_vcmpgtuh (vector unsigned short,
15940 vector unsigned short);
15942 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
15944 vector bool char vec_vcmpgtub (vector unsigned char,
15945 vector unsigned char);
15947 vector bool int vec_cmple (vector float, vector float);
15949 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
15950 vector bool char vec_cmplt (vector signed char, vector signed char);
15951 vector bool short vec_cmplt (vector unsigned short,
15952 vector unsigned short);
15953 vector bool short vec_cmplt (vector signed short, vector signed short);
15954 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
15955 vector bool int vec_cmplt (vector signed int, vector signed int);
15956 vector bool int vec_cmplt (vector float, vector float);
15958 vector float vec_cpsgn (vector float, vector float);
15960 vector float vec_ctf (vector unsigned int, const int);
15961 vector float vec_ctf (vector signed int, const int);
15962 vector double vec_ctf (vector unsigned long, const int);
15963 vector double vec_ctf (vector signed long, const int);
15965 vector float vec_vcfsx (vector signed int, const int);
15967 vector float vec_vcfux (vector unsigned int, const int);
15969 vector signed int vec_cts (vector float, const int);
15970 vector signed long vec_cts (vector double, const int);
15972 vector unsigned int vec_ctu (vector float, const int);
15973 vector unsigned long vec_ctu (vector double, const int);
15975 vector double vec_doublee (vector float);
15976 vector double vec_doublee (vector signed int);
15977 vector double vec_doublee (vector unsigned int);
15979 vector double vec_doubleo (vector float);
15980 vector double vec_doubleo (vector signed int);
15981 vector double vec_doubleo (vector unsigned int);
15983 vector double vec_doubleh (vector float);
15984 vector double vec_doubleh (vector signed int);
15985 vector double vec_doubleh (vector unsigned int);
15987 vector double vec_doublel (vector float);
15988 vector double vec_doublel (vector signed int);
15989 vector double vec_doublel (vector unsigned int);
15991 void vec_dss (const int);
15993 void vec_dssall (void);
15995 void vec_dst (const vector unsigned char *, int, const int);
15996 void vec_dst (const vector signed char *, int, const int);
15997 void vec_dst (const vector bool char *, int, const int);
15998 void vec_dst (const vector unsigned short *, int, const int);
15999 void vec_dst (const vector signed short *, int, const int);
16000 void vec_dst (const vector bool short *, int, const int);
16001 void vec_dst (const vector pixel *, int, const int);
16002 void vec_dst (const vector unsigned int *, int, const int);
16003 void vec_dst (const vector signed int *, int, const int);
16004 void vec_dst (const vector bool int *, int, const int);
16005 void vec_dst (const vector float *, int, const int);
16006 void vec_dst (const unsigned char *, int, const int);
16007 void vec_dst (const signed char *, int, const int);
16008 void vec_dst (const unsigned short *, int, const int);
16009 void vec_dst (const short *, int, const int);
16010 void vec_dst (const unsigned int *, int, const int);
16011 void vec_dst (const int *, int, const int);
16012 void vec_dst (const unsigned long *, int, const int);
16013 void vec_dst (const long *, int, const int);
16014 void vec_dst (const float *, int, const int);
16016 void vec_dstst (const vector unsigned char *, int, const int);
16017 void vec_dstst (const vector signed char *, int, const int);
16018 void vec_dstst (const vector bool char *, int, const int);
16019 void vec_dstst (const vector unsigned short *, int, const int);
16020 void vec_dstst (const vector signed short *, int, const int);
16021 void vec_dstst (const vector bool short *, int, const int);
16022 void vec_dstst (const vector pixel *, int, const int);
16023 void vec_dstst (const vector unsigned int *, int, const int);
16024 void vec_dstst (const vector signed int *, int, const int);
16025 void vec_dstst (const vector bool int *, int, const int);
16026 void vec_dstst (const vector float *, int, const int);
16027 void vec_dstst (const unsigned char *, int, const int);
16028 void vec_dstst (const signed char *, int, const int);
16029 void vec_dstst (const unsigned short *, int, const int);
16030 void vec_dstst (const short *, int, const int);
16031 void vec_dstst (const unsigned int *, int, const int);
16032 void vec_dstst (const int *, int, const int);
16033 void vec_dstst (const unsigned long *, int, const int);
16034 void vec_dstst (const long *, int, const int);
16035 void vec_dstst (const float *, int, const int);
16037 void vec_dststt (const vector unsigned char *, int, const int);
16038 void vec_dststt (const vector signed char *, int, const int);
16039 void vec_dststt (const vector bool char *, int, const int);
16040 void vec_dststt (const vector unsigned short *, int, const int);
16041 void vec_dststt (const vector signed short *, int, const int);
16042 void vec_dststt (const vector bool short *, int, const int);
16043 void vec_dststt (const vector pixel *, int, const int);
16044 void vec_dststt (const vector unsigned int *, int, const int);
16045 void vec_dststt (const vector signed int *, int, const int);
16046 void vec_dststt (const vector bool int *, int, const int);
16047 void vec_dststt (const vector float *, int, const int);
16048 void vec_dststt (const unsigned char *, int, const int);
16049 void vec_dststt (const signed char *, int, const int);
16050 void vec_dststt (const unsigned short *, int, const int);
16051 void vec_dststt (const short *, int, const int);
16052 void vec_dststt (const unsigned int *, int, const int);
16053 void vec_dststt (const int *, int, const int);
16054 void vec_dststt (const unsigned long *, int, const int);
16055 void vec_dststt (const long *, int, const int);
16056 void vec_dststt (const float *, int, const int);
16058 void vec_dstt (const vector unsigned char *, int, const int);
16059 void vec_dstt (const vector signed char *, int, const int);
16060 void vec_dstt (const vector bool char *, int, const int);
16061 void vec_dstt (const vector unsigned short *, int, const int);
16062 void vec_dstt (const vector signed short *, int, const int);
16063 void vec_dstt (const vector bool short *, int, const int);
16064 void vec_dstt (const vector pixel *, int, const int);
16065 void vec_dstt (const vector unsigned int *, int, const int);
16066 void vec_dstt (const vector signed int *, int, const int);
16067 void vec_dstt (const vector bool int *, int, const int);
16068 void vec_dstt (const vector float *, int, const int);
16069 void vec_dstt (const unsigned char *, int, const int);
16070 void vec_dstt (const signed char *, int, const int);
16071 void vec_dstt (const unsigned short *, int, const int);
16072 void vec_dstt (const short *, int, const int);
16073 void vec_dstt (const unsigned int *, int, const int);
16074 void vec_dstt (const int *, int, const int);
16075 void vec_dstt (const unsigned long *, int, const int);
16076 void vec_dstt (const long *, int, const int);
16077 void vec_dstt (const float *, int, const int);
16079 vector float vec_expte (vector float);
16081 vector float vec_floor (vector float);
16083 vector float vec_float (vector signed int);
16084 vector float vec_float (vector unsigned int);
16086 vector float vec_float2 (vector signed long long, vector signed long long);
16087 vector float vec_float2 (vector unsigned long long, vector signed long long);
16089 vector float vec_floate (vector double);
16090 vector float vec_floate (vector signed long long);
16091 vector float vec_floate (vector unsigned long long);
16093 vector float vec_floato (vector double);
16094 vector float vec_floato (vector signed long long);
16095 vector float vec_floato (vector unsigned long long);
16097 vector float vec_ld (int, const vector float *);
16098 vector float vec_ld (int, const float *);
16099 vector bool int vec_ld (int, const vector bool int *);
16100 vector signed int vec_ld (int, const vector signed int *);
16101 vector signed int vec_ld (int, const int *);
16102 vector signed int vec_ld (int, const long *);
16103 vector unsigned int vec_ld (int, const vector unsigned int *);
16104 vector unsigned int vec_ld (int, const unsigned int *);
16105 vector unsigned int vec_ld (int, const unsigned long *);
16106 vector bool short vec_ld (int, const vector bool short *);
16107 vector pixel vec_ld (int, const vector pixel *);
16108 vector signed short vec_ld (int, const vector signed short *);
16109 vector signed short vec_ld (int, const short *);
16110 vector unsigned short vec_ld (int, const vector unsigned short *);
16111 vector unsigned short vec_ld (int, const unsigned short *);
16112 vector bool char vec_ld (int, const vector bool char *);
16113 vector signed char vec_ld (int, const vector signed char *);
16114 vector signed char vec_ld (int, const signed char *);
16115 vector unsigned char vec_ld (int, const vector unsigned char *);
16116 vector unsigned char vec_ld (int, const unsigned char *);
16118 vector signed char vec_lde (int, const signed char *);
16119 vector unsigned char vec_lde (int, const unsigned char *);
16120 vector signed short vec_lde (int, const short *);
16121 vector unsigned short vec_lde (int, const unsigned short *);
16122 vector float vec_lde (int, const float *);
16123 vector signed int vec_lde (int, const int *);
16124 vector unsigned int vec_lde (int, const unsigned int *);
16125 vector signed int vec_lde (int, const long *);
16126 vector unsigned int vec_lde (int, const unsigned long *);
16128 vector float vec_lvewx (int, float *);
16129 vector signed int vec_lvewx (int, int *);
16130 vector unsigned int vec_lvewx (int, unsigned int *);
16131 vector signed int vec_lvewx (int, long *);
16132 vector unsigned int vec_lvewx (int, unsigned long *);
16134 vector signed short vec_lvehx (int, short *);
16135 vector unsigned short vec_lvehx (int, unsigned short *);
16137 vector signed char vec_lvebx (int, char *);
16138 vector unsigned char vec_lvebx (int, unsigned char *);
16140 vector float vec_ldl (int, const vector float *);
16141 vector float vec_ldl (int, const float *);
16142 vector bool int vec_ldl (int, const vector bool int *);
16143 vector signed int vec_ldl (int, const vector signed int *);
16144 vector signed int vec_ldl (int, const int *);
16145 vector signed int vec_ldl (int, const long *);
16146 vector unsigned int vec_ldl (int, const vector unsigned int *);
16147 vector unsigned int vec_ldl (int, const unsigned int *);
16148 vector unsigned int vec_ldl (int, const unsigned long *);
16149 vector bool short vec_ldl (int, const vector bool short *);
16150 vector pixel vec_ldl (int, const vector pixel *);
16151 vector signed short vec_ldl (int, const vector signed short *);
16152 vector signed short vec_ldl (int, const short *);
16153 vector unsigned short vec_ldl (int, const vector unsigned short *);
16154 vector unsigned short vec_ldl (int, const unsigned short *);
16155 vector bool char vec_ldl (int, const vector bool char *);
16156 vector signed char vec_ldl (int, const vector signed char *);
16157 vector signed char vec_ldl (int, const signed char *);
16158 vector unsigned char vec_ldl (int, const vector unsigned char *);
16159 vector unsigned char vec_ldl (int, const unsigned char *);
16161 vector float vec_loge (vector float);
16163 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
16164 vector unsigned char vec_lvsl (int, const volatile signed char *);
16165 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
16166 vector unsigned char vec_lvsl (int, const volatile short *);
16167 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
16168 vector unsigned char vec_lvsl (int, const volatile int *);
16169 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
16170 vector unsigned char vec_lvsl (int, const volatile long *);
16171 vector unsigned char vec_lvsl (int, const volatile float *);
16173 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
16174 vector unsigned char vec_lvsr (int, const volatile signed char *);
16175 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
16176 vector unsigned char vec_lvsr (int, const volatile short *);
16177 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
16178 vector unsigned char vec_lvsr (int, const volatile int *);
16179 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
16180 vector unsigned char vec_lvsr (int, const volatile long *);
16181 vector unsigned char vec_lvsr (int, const volatile float *);
16183 vector float vec_madd (vector float, vector float, vector float);
16185 vector signed short vec_madds (vector signed short,
16186 vector signed short,
16187 vector signed short);
16189 vector unsigned char vec_max (vector bool char, vector unsigned char);
16190 vector unsigned char vec_max (vector unsigned char, vector bool char);
16191 vector unsigned char vec_max (vector unsigned char,
16192 vector unsigned char);
16193 vector signed char vec_max (vector bool char, vector signed char);
16194 vector signed char vec_max (vector signed char, vector bool char);
16195 vector signed char vec_max (vector signed char, vector signed char);
16196 vector unsigned short vec_max (vector bool short,
16197 vector unsigned short);
16198 vector unsigned short vec_max (vector unsigned short,
16199 vector bool short);
16200 vector unsigned short vec_max (vector unsigned short,
16201 vector unsigned short);
16202 vector signed short vec_max (vector bool short, vector signed short);
16203 vector signed short vec_max (vector signed short, vector bool short);
16204 vector signed short vec_max (vector signed short, vector signed short);
16205 vector unsigned int vec_max (vector bool int, vector unsigned int);
16206 vector unsigned int vec_max (vector unsigned int, vector bool int);
16207 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
16208 vector signed int vec_max (vector bool int, vector signed int);
16209 vector signed int vec_max (vector signed int, vector bool int);
16210 vector signed int vec_max (vector signed int, vector signed int);
16211 vector float vec_max (vector float, vector float);
16213 vector float vec_vmaxfp (vector float, vector float);
16215 vector signed int vec_vmaxsw (vector bool int, vector signed int);
16216 vector signed int vec_vmaxsw (vector signed int, vector bool int);
16217 vector signed int vec_vmaxsw (vector signed int, vector signed int);
16219 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
16220 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
16221 vector unsigned int vec_vmaxuw (vector unsigned int,
16222 vector unsigned int);
16224 vector signed short vec_vmaxsh (vector bool short, vector signed short);
16225 vector signed short vec_vmaxsh (vector signed short, vector bool short);
16226 vector signed short vec_vmaxsh (vector signed short,
16227 vector signed short);
16229 vector unsigned short vec_vmaxuh (vector bool short,
16230 vector unsigned short);
16231 vector unsigned short vec_vmaxuh (vector unsigned short,
16232 vector bool short);
16233 vector unsigned short vec_vmaxuh (vector unsigned short,
16234 vector unsigned short);
16236 vector signed char vec_vmaxsb (vector bool char, vector signed char);
16237 vector signed char vec_vmaxsb (vector signed char, vector bool char);
16238 vector signed char vec_vmaxsb (vector signed char, vector signed char);
16240 vector unsigned char vec_vmaxub (vector bool char,
16241 vector unsigned char);
16242 vector unsigned char vec_vmaxub (vector unsigned char,
16244 vector unsigned char vec_vmaxub (vector unsigned char,
16245 vector unsigned char);
16247 vector bool char vec_mergeh (vector bool char, vector bool char);
16248 vector signed char vec_mergeh (vector signed char, vector signed char);
16249 vector unsigned char vec_mergeh (vector unsigned char,
16250 vector unsigned char);
16251 vector bool short vec_mergeh (vector bool short, vector bool short);
16252 vector pixel vec_mergeh (vector pixel, vector pixel);
16253 vector signed short vec_mergeh (vector signed short,
16254 vector signed short);
16255 vector unsigned short vec_mergeh (vector unsigned short,
16256 vector unsigned short);
16257 vector float vec_mergeh (vector float, vector float);
16258 vector bool int vec_mergeh (vector bool int, vector bool int);
16259 vector signed int vec_mergeh (vector signed int, vector signed int);
16260 vector unsigned int vec_mergeh (vector unsigned int,
16261 vector unsigned int);
16263 vector float vec_vmrghw (vector float, vector float);
16264 vector bool int vec_vmrghw (vector bool int, vector bool int);
16265 vector signed int vec_vmrghw (vector signed int, vector signed int);
16266 vector unsigned int vec_vmrghw (vector unsigned int,
16267 vector unsigned int);
16269 vector bool short vec_vmrghh (vector bool short, vector bool short);
16270 vector signed short vec_vmrghh (vector signed short,
16271 vector signed short);
16272 vector unsigned short vec_vmrghh (vector unsigned short,
16273 vector unsigned short);
16274 vector pixel vec_vmrghh (vector pixel, vector pixel);
16276 vector bool char vec_vmrghb (vector bool char, vector bool char);
16277 vector signed char vec_vmrghb (vector signed char, vector signed char);
16278 vector unsigned char vec_vmrghb (vector unsigned char,
16279 vector unsigned char);
16281 vector bool char vec_mergel (vector bool char, vector bool char);
16282 vector signed char vec_mergel (vector signed char, vector signed char);
16283 vector unsigned char vec_mergel (vector unsigned char,
16284 vector unsigned char);
16285 vector bool short vec_mergel (vector bool short, vector bool short);
16286 vector pixel vec_mergel (vector pixel, vector pixel);
16287 vector signed short vec_mergel (vector signed short,
16288 vector signed short);
16289 vector unsigned short vec_mergel (vector unsigned short,
16290 vector unsigned short);
16291 vector float vec_mergel (vector float, vector float);
16292 vector bool int vec_mergel (vector bool int, vector bool int);
16293 vector signed int vec_mergel (vector signed int, vector signed int);
16294 vector unsigned int vec_mergel (vector unsigned int,
16295 vector unsigned int);
16297 vector float vec_vmrglw (vector float, vector float);
16298 vector signed int vec_vmrglw (vector signed int, vector signed int);
16299 vector unsigned int vec_vmrglw (vector unsigned int,
16300 vector unsigned int);
16301 vector bool int vec_vmrglw (vector bool int, vector bool int);
16303 vector bool short vec_vmrglh (vector bool short, vector bool short);
16304 vector signed short vec_vmrglh (vector signed short,
16305 vector signed short);
16306 vector unsigned short vec_vmrglh (vector unsigned short,
16307 vector unsigned short);
16308 vector pixel vec_vmrglh (vector pixel, vector pixel);
16310 vector bool char vec_vmrglb (vector bool char, vector bool char);
16311 vector signed char vec_vmrglb (vector signed char, vector signed char);
16312 vector unsigned char vec_vmrglb (vector unsigned char,
16313 vector unsigned char);
16315 vector unsigned short vec_mfvscr (void);
16317 vector unsigned char vec_min (vector bool char, vector unsigned char);
16318 vector unsigned char vec_min (vector unsigned char, vector bool char);
16319 vector unsigned char vec_min (vector unsigned char,
16320 vector unsigned char);
16321 vector signed char vec_min (vector bool char, vector signed char);
16322 vector signed char vec_min (vector signed char, vector bool char);
16323 vector signed char vec_min (vector signed char, vector signed char);
16324 vector unsigned short vec_min (vector bool short,
16325 vector unsigned short);
16326 vector unsigned short vec_min (vector unsigned short,
16327 vector bool short);
16328 vector unsigned short vec_min (vector unsigned short,
16329 vector unsigned short);
16330 vector signed short vec_min (vector bool short, vector signed short);
16331 vector signed short vec_min (vector signed short, vector bool short);
16332 vector signed short vec_min (vector signed short, vector signed short);
16333 vector unsigned int vec_min (vector bool int, vector unsigned int);
16334 vector unsigned int vec_min (vector unsigned int, vector bool int);
16335 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
16336 vector signed int vec_min (vector bool int, vector signed int);
16337 vector signed int vec_min (vector signed int, vector bool int);
16338 vector signed int vec_min (vector signed int, vector signed int);
16339 vector float vec_min (vector float, vector float);
16341 vector float vec_vminfp (vector float, vector float);
16343 vector signed int vec_vminsw (vector bool int, vector signed int);
16344 vector signed int vec_vminsw (vector signed int, vector bool int);
16345 vector signed int vec_vminsw (vector signed int, vector signed int);
16347 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
16348 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
16349 vector unsigned int vec_vminuw (vector unsigned int,
16350 vector unsigned int);
16352 vector signed short vec_vminsh (vector bool short, vector signed short);
16353 vector signed short vec_vminsh (vector signed short, vector bool short);
16354 vector signed short vec_vminsh (vector signed short,
16355 vector signed short);
16357 vector unsigned short vec_vminuh (vector bool short,
16358 vector unsigned short);
16359 vector unsigned short vec_vminuh (vector unsigned short,
16360 vector bool short);
16361 vector unsigned short vec_vminuh (vector unsigned short,
16362 vector unsigned short);
16364 vector signed char vec_vminsb (vector bool char, vector signed char);
16365 vector signed char vec_vminsb (vector signed char, vector bool char);
16366 vector signed char vec_vminsb (vector signed char, vector signed char);
16368 vector unsigned char vec_vminub (vector bool char,
16369 vector unsigned char);
16370 vector unsigned char vec_vminub (vector unsigned char,
16372 vector unsigned char vec_vminub (vector unsigned char,
16373 vector unsigned char);
16375 vector signed short vec_mladd (vector signed short,
16376 vector signed short,
16377 vector signed short);
16378 vector signed short vec_mladd (vector signed short,
16379 vector unsigned short,
16380 vector unsigned short);
16381 vector signed short vec_mladd (vector unsigned short,
16382 vector signed short,
16383 vector signed short);
16384 vector unsigned short vec_mladd (vector unsigned short,
16385 vector unsigned short,
16386 vector unsigned short);
16388 vector signed short vec_mradds (vector signed short,
16389 vector signed short,
16390 vector signed short);
16392 vector unsigned int vec_msum (vector unsigned char,
16393 vector unsigned char,
16394 vector unsigned int);
16395 vector signed int vec_msum (vector signed char,
16396 vector unsigned char,
16397 vector signed int);
16398 vector unsigned int vec_msum (vector unsigned short,
16399 vector unsigned short,
16400 vector unsigned int);
16401 vector signed int vec_msum (vector signed short,
16402 vector signed short,
16403 vector signed int);
16405 vector signed int vec_vmsumshm (vector signed short,
16406 vector signed short,
16407 vector signed int);
16409 vector unsigned int vec_vmsumuhm (vector unsigned short,
16410 vector unsigned short,
16411 vector unsigned int);
16413 vector signed int vec_vmsummbm (vector signed char,
16414 vector unsigned char,
16415 vector signed int);
16417 vector unsigned int vec_vmsumubm (vector unsigned char,
16418 vector unsigned char,
16419 vector unsigned int);
16421 vector unsigned int vec_msums (vector unsigned short,
16422 vector unsigned short,
16423 vector unsigned int);
16424 vector signed int vec_msums (vector signed short,
16425 vector signed short,
16426 vector signed int);
16428 vector signed int vec_vmsumshs (vector signed short,
16429 vector signed short,
16430 vector signed int);
16432 vector unsigned int vec_vmsumuhs (vector unsigned short,
16433 vector unsigned short,
16434 vector unsigned int);
16436 void vec_mtvscr (vector signed int);
16437 void vec_mtvscr (vector unsigned int);
16438 void vec_mtvscr (vector bool int);
16439 void vec_mtvscr (vector signed short);
16440 void vec_mtvscr (vector unsigned short);
16441 void vec_mtvscr (vector bool short);
16442 void vec_mtvscr (vector pixel);
16443 void vec_mtvscr (vector signed char);
16444 void vec_mtvscr (vector unsigned char);
16445 void vec_mtvscr (vector bool char);
16447 vector unsigned short vec_mule (vector unsigned char,
16448 vector unsigned char);
16449 vector signed short vec_mule (vector signed char,
16450 vector signed char);
16451 vector unsigned int vec_mule (vector unsigned short,
16452 vector unsigned short);
16453 vector signed int vec_mule (vector signed short, vector signed short);
16454 vector unsigned long long vec_mule (vector unsigned int,
16455 vector unsigned int);
16456 vector signed long long vec_mule (vector signed int,
16457 vector signed int);
16459 vector signed int vec_vmulesh (vector signed short,
16460 vector signed short);
16462 vector unsigned int vec_vmuleuh (vector unsigned short,
16463 vector unsigned short);
16465 vector signed short vec_vmulesb (vector signed char,
16466 vector signed char);
16468 vector unsigned short vec_vmuleub (vector unsigned char,
16469 vector unsigned char);
16471 vector unsigned short vec_mulo (vector unsigned char,
16472 vector unsigned char);
16473 vector signed short vec_mulo (vector signed char, vector signed char);
16474 vector unsigned int vec_mulo (vector unsigned short,
16475 vector unsigned short);
16476 vector signed int vec_mulo (vector signed short, vector signed short);
16477 vector unsigned long long vec_mulo (vector unsigned int,
16478 vector unsigned int);
16479 vector signed long long vec_mulo (vector signed int,
16480 vector signed int);
16482 vector signed int vec_vmulosh (vector signed short,
16483 vector signed short);
16485 vector unsigned int vec_vmulouh (vector unsigned short,
16486 vector unsigned short);
16488 vector signed short vec_vmulosb (vector signed char,
16489 vector signed char);
16491 vector unsigned short vec_vmuloub (vector unsigned char,
16492 vector unsigned char);
16494 vector float vec_nmsub (vector float, vector float, vector float);
16496 vector signed char vec_nabs (vector signed char);
16497 vector signed short vec_nabs (vector signed short);
16498 vector signed int vec_nabs (vector signed int);
16499 vector float vec_nabs (vector float);
16500 vector double vec_nabs (vector double);
16502 vector signed char vec_neg (vector signed char);
16503 vector signed short vec_neg (vector signed short);
16504 vector signed int vec_neg (vector signed int);
16505 vector signed long long vec_neg (vector signed long long);
16506 vector float char vec_neg (vector float);
16507 vector double vec_neg (vector double);
16509 vector float vec_nor (vector float, vector float);
16510 vector signed int vec_nor (vector signed int, vector signed int);
16511 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
16512 vector bool int vec_nor (vector bool int, vector bool int);
16513 vector signed short vec_nor (vector signed short, vector signed short);
16514 vector unsigned short vec_nor (vector unsigned short,
16515 vector unsigned short);
16516 vector bool short vec_nor (vector bool short, vector bool short);
16517 vector signed char vec_nor (vector signed char, vector signed char);
16518 vector unsigned char vec_nor (vector unsigned char,
16519 vector unsigned char);
16520 vector bool char vec_nor (vector bool char, vector bool char);
16522 vector float vec_or (vector float, vector float);
16523 vector float vec_or (vector float, vector bool int);
16524 vector float vec_or (vector bool int, vector float);
16525 vector bool int vec_or (vector bool int, vector bool int);
16526 vector signed int vec_or (vector bool int, vector signed int);
16527 vector signed int vec_or (vector signed int, vector bool int);
16528 vector signed int vec_or (vector signed int, vector signed int);
16529 vector unsigned int vec_or (vector bool int, vector unsigned int);
16530 vector unsigned int vec_or (vector unsigned int, vector bool int);
16531 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
16532 vector bool short vec_or (vector bool short, vector bool short);
16533 vector signed short vec_or (vector bool short, vector signed short);
16534 vector signed short vec_or (vector signed short, vector bool short);
16535 vector signed short vec_or (vector signed short, vector signed short);
16536 vector unsigned short vec_or (vector bool short, vector unsigned short);
16537 vector unsigned short vec_or (vector unsigned short, vector bool short);
16538 vector unsigned short vec_or (vector unsigned short,
16539 vector unsigned short);
16540 vector signed char vec_or (vector bool char, vector signed char);
16541 vector bool char vec_or (vector bool char, vector bool char);
16542 vector signed char vec_or (vector signed char, vector bool char);
16543 vector signed char vec_or (vector signed char, vector signed char);
16544 vector unsigned char vec_or (vector bool char, vector unsigned char);
16545 vector unsigned char vec_or (vector unsigned char, vector bool char);
16546 vector unsigned char vec_or (vector unsigned char,
16547 vector unsigned char);
16549 vector signed char vec_pack (vector signed short, vector signed short);
16550 vector unsigned char vec_pack (vector unsigned short,
16551 vector unsigned short);
16552 vector bool char vec_pack (vector bool short, vector bool short);
16553 vector signed short vec_pack (vector signed int, vector signed int);
16554 vector unsigned short vec_pack (vector unsigned int,
16555 vector unsigned int);
16556 vector bool short vec_pack (vector bool int, vector bool int);
16558 vector bool short vec_vpkuwum (vector bool int, vector bool int);
16559 vector signed short vec_vpkuwum (vector signed int, vector signed int);
16560 vector unsigned short vec_vpkuwum (vector unsigned int,
16561 vector unsigned int);
16563 vector bool char vec_vpkuhum (vector bool short, vector bool short);
16564 vector signed char vec_vpkuhum (vector signed short,
16565 vector signed short);
16566 vector unsigned char vec_vpkuhum (vector unsigned short,
16567 vector unsigned short);
16569 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
16571 vector unsigned char vec_packs (vector unsigned short,
16572 vector unsigned short);
16573 vector signed char vec_packs (vector signed short, vector signed short);
16574 vector unsigned short vec_packs (vector unsigned int,
16575 vector unsigned int);
16576 vector signed short vec_packs (vector signed int, vector signed int);
16578 vector signed short vec_vpkswss (vector signed int, vector signed int);
16580 vector unsigned short vec_vpkuwus (vector unsigned int,
16581 vector unsigned int);
16583 vector signed char vec_vpkshss (vector signed short,
16584 vector signed short);
16586 vector unsigned char vec_vpkuhus (vector unsigned short,
16587 vector unsigned short);
16589 vector unsigned char vec_packsu (vector unsigned short,
16590 vector unsigned short);
16591 vector unsigned char vec_packsu (vector signed short,
16592 vector signed short);
16593 vector unsigned short vec_packsu (vector unsigned int,
16594 vector unsigned int);
16595 vector unsigned short vec_packsu (vector signed int, vector signed int);
16597 vector unsigned short vec_vpkswus (vector signed int,
16598 vector signed int);
16600 vector unsigned char vec_vpkshus (vector signed short,
16601 vector signed short);
16603 vector float vec_perm (vector float,
16605 vector unsigned char);
16606 vector signed int vec_perm (vector signed int,
16608 vector unsigned char);
16609 vector unsigned int vec_perm (vector unsigned int,
16610 vector unsigned int,
16611 vector unsigned char);
16612 vector bool int vec_perm (vector bool int,
16614 vector unsigned char);
16615 vector signed short vec_perm (vector signed short,
16616 vector signed short,
16617 vector unsigned char);
16618 vector unsigned short vec_perm (vector unsigned short,
16619 vector unsigned short,
16620 vector unsigned char);
16621 vector bool short vec_perm (vector bool short,
16623 vector unsigned char);
16624 vector pixel vec_perm (vector pixel,
16626 vector unsigned char);
16627 vector signed char vec_perm (vector signed char,
16628 vector signed char,
16629 vector unsigned char);
16630 vector unsigned char vec_perm (vector unsigned char,
16631 vector unsigned char,
16632 vector unsigned char);
16633 vector bool char vec_perm (vector bool char,
16635 vector unsigned char);
16637 vector float vec_re (vector float);
16639 vector bool char vec_reve (vector bool char);
16640 vector signed char vec_reve (vector signed char);
16641 vector unsigned char vec_reve (vector unsigned char);
16642 vector bool int vec_reve (vector bool int);
16643 vector signed int vec_reve (vector signed int);
16644 vector unsigned int vec_reve (vector unsigned int);
16645 vector bool long long vec_reve (vector bool long long);
16646 vector signed long long vec_reve (vector signed long long);
16647 vector unsigned long long vec_reve (vector unsigned long long);
16648 vector bool short vec_reve (vector bool short);
16649 vector signed short vec_reve (vector signed short);
16650 vector unsigned short vec_reve (vector unsigned short);
16652 vector signed char vec_rl (vector signed char,
16653 vector unsigned char);
16654 vector unsigned char vec_rl (vector unsigned char,
16655 vector unsigned char);
16656 vector signed short vec_rl (vector signed short, vector unsigned short);
16657 vector unsigned short vec_rl (vector unsigned short,
16658 vector unsigned short);
16659 vector signed int vec_rl (vector signed int, vector unsigned int);
16660 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
16662 vector signed int vec_vrlw (vector signed int, vector unsigned int);
16663 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
16665 vector signed short vec_vrlh (vector signed short,
16666 vector unsigned short);
16667 vector unsigned short vec_vrlh (vector unsigned short,
16668 vector unsigned short);
16670 vector signed char vec_vrlb (vector signed char, vector unsigned char);
16671 vector unsigned char vec_vrlb (vector unsigned char,
16672 vector unsigned char);
16674 vector float vec_round (vector float);
16676 vector float vec_recip (vector float, vector float);
16678 vector float vec_rsqrt (vector float);
16680 vector float vec_rsqrte (vector float);
16682 vector float vec_sel (vector float, vector float, vector bool int);
16683 vector float vec_sel (vector float, vector float, vector unsigned int);
16684 vector signed int vec_sel (vector signed int,
16687 vector signed int vec_sel (vector signed int,
16689 vector unsigned int);
16690 vector unsigned int vec_sel (vector unsigned int,
16691 vector unsigned int,
16693 vector unsigned int vec_sel (vector unsigned int,
16694 vector unsigned int,
16695 vector unsigned int);
16696 vector bool int vec_sel (vector bool int,
16699 vector bool int vec_sel (vector bool int,
16701 vector unsigned int);
16702 vector signed short vec_sel (vector signed short,
16703 vector signed short,
16704 vector bool short);
16705 vector signed short vec_sel (vector signed short,
16706 vector signed short,
16707 vector unsigned short);
16708 vector unsigned short vec_sel (vector unsigned short,
16709 vector unsigned short,
16710 vector bool short);
16711 vector unsigned short vec_sel (vector unsigned short,
16712 vector unsigned short,
16713 vector unsigned short);
16714 vector bool short vec_sel (vector bool short,
16716 vector bool short);
16717 vector bool short vec_sel (vector bool short,
16719 vector unsigned short);
16720 vector signed char vec_sel (vector signed char,
16721 vector signed char,
16723 vector signed char vec_sel (vector signed char,
16724 vector signed char,
16725 vector unsigned char);
16726 vector unsigned char vec_sel (vector unsigned char,
16727 vector unsigned char,
16729 vector unsigned char vec_sel (vector unsigned char,
16730 vector unsigned char,
16731 vector unsigned char);
16732 vector bool char vec_sel (vector bool char,
16735 vector bool char vec_sel (vector bool char,
16737 vector unsigned char);
16739 vector signed long long vec_signed (vector double);
16740 vector signed int vec_signed (vector float);
16742 vector signed int vec_signede (vector double);
16743 vector signed int vec_signedo (vector double);
16744 vector signed int vec_signed2 (vector double, vector double);
16746 vector signed char vec_sl (vector signed char,
16747 vector unsigned char);
16748 vector unsigned char vec_sl (vector unsigned char,
16749 vector unsigned char);
16750 vector signed short vec_sl (vector signed short, vector unsigned short);
16751 vector unsigned short vec_sl (vector unsigned short,
16752 vector unsigned short);
16753 vector signed int vec_sl (vector signed int, vector unsigned int);
16754 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
16756 vector signed int vec_vslw (vector signed int, vector unsigned int);
16757 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
16759 vector signed short vec_vslh (vector signed short,
16760 vector unsigned short);
16761 vector unsigned short vec_vslh (vector unsigned short,
16762 vector unsigned short);
16764 vector signed char vec_vslb (vector signed char, vector unsigned char);
16765 vector unsigned char vec_vslb (vector unsigned char,
16766 vector unsigned char);
16768 vector float vec_sld (vector float, vector float, const int);
16769 vector double vec_sld (vector double, vector double, const int);
16771 vector signed int vec_sld (vector signed int,
16774 vector unsigned int vec_sld (vector unsigned int,
16775 vector unsigned int,
16777 vector bool int vec_sld (vector bool int,
16780 vector signed short vec_sld (vector signed short,
16781 vector signed short,
16783 vector unsigned short vec_sld (vector unsigned short,
16784 vector unsigned short,
16786 vector bool short vec_sld (vector bool short,
16789 vector pixel vec_sld (vector pixel,
16792 vector signed char vec_sld (vector signed char,
16793 vector signed char,
16795 vector unsigned char vec_sld (vector unsigned char,
16796 vector unsigned char,
16798 vector bool char vec_sld (vector bool char,
16802 vector signed char vec_sldw (vector signed char,
16803 vector signed char,
16805 vector unsigned char vec_sldw (vector unsigned char,
16806 vector unsigned char,
16808 vector signed short vec_sldw (vector signed short,
16809 vector signed short,
16811 vector unsigned short vec_sldw (vector unsigned short,
16812 vector unsigned short,
16814 vector signed int vec_sldw (vector signed int,
16817 vector unsigned int vec_sldw (vector unsigned int,
16818 vector unsigned int,
16820 vector signed long long vec_sldw (vector signed long long,
16821 vector signed long long,
16823 vector unsigned long long vec_sldw (vector unsigned long long,
16824 vector unsigned long long,
16827 vector signed int vec_sll (vector signed int,
16828 vector unsigned int);
16829 vector signed int vec_sll (vector signed int,
16830 vector unsigned short);
16831 vector signed int vec_sll (vector signed int,
16832 vector unsigned char);
16833 vector unsigned int vec_sll (vector unsigned int,
16834 vector unsigned int);
16835 vector unsigned int vec_sll (vector unsigned int,
16836 vector unsigned short);
16837 vector unsigned int vec_sll (vector unsigned int,
16838 vector unsigned char);
16839 vector bool int vec_sll (vector bool int,
16840 vector unsigned int);
16841 vector bool int vec_sll (vector bool int,
16842 vector unsigned short);
16843 vector bool int vec_sll (vector bool int,
16844 vector unsigned char);
16845 vector signed short vec_sll (vector signed short,
16846 vector unsigned int);
16847 vector signed short vec_sll (vector signed short,
16848 vector unsigned short);
16849 vector signed short vec_sll (vector signed short,
16850 vector unsigned char);
16851 vector unsigned short vec_sll (vector unsigned short,
16852 vector unsigned int);
16853 vector unsigned short vec_sll (vector unsigned short,
16854 vector unsigned short);
16855 vector unsigned short vec_sll (vector unsigned short,
16856 vector unsigned char);
16857 vector bool short vec_sll (vector bool short, vector unsigned int);
16858 vector bool short vec_sll (vector bool short, vector unsigned short);
16859 vector bool short vec_sll (vector bool short, vector unsigned char);
16860 vector pixel vec_sll (vector pixel, vector unsigned int);
16861 vector pixel vec_sll (vector pixel, vector unsigned short);
16862 vector pixel vec_sll (vector pixel, vector unsigned char);
16863 vector signed char vec_sll (vector signed char, vector unsigned int);
16864 vector signed char vec_sll (vector signed char, vector unsigned short);
16865 vector signed char vec_sll (vector signed char, vector unsigned char);
16866 vector unsigned char vec_sll (vector unsigned char,
16867 vector unsigned int);
16868 vector unsigned char vec_sll (vector unsigned char,
16869 vector unsigned short);
16870 vector unsigned char vec_sll (vector unsigned char,
16871 vector unsigned char);
16872 vector bool char vec_sll (vector bool char, vector unsigned int);
16873 vector bool char vec_sll (vector bool char, vector unsigned short);
16874 vector bool char vec_sll (vector bool char, vector unsigned char);
16876 vector float vec_slo (vector float, vector signed char);
16877 vector float vec_slo (vector float, vector unsigned char);
16878 vector signed int vec_slo (vector signed int, vector signed char);
16879 vector signed int vec_slo (vector signed int, vector unsigned char);
16880 vector unsigned int vec_slo (vector unsigned int, vector signed char);
16881 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
16882 vector signed short vec_slo (vector signed short, vector signed char);
16883 vector signed short vec_slo (vector signed short, vector unsigned char);
16884 vector unsigned short vec_slo (vector unsigned short,
16885 vector signed char);
16886 vector unsigned short vec_slo (vector unsigned short,
16887 vector unsigned char);
16888 vector pixel vec_slo (vector pixel, vector signed char);
16889 vector pixel vec_slo (vector pixel, vector unsigned char);
16890 vector signed char vec_slo (vector signed char, vector signed char);
16891 vector signed char vec_slo (vector signed char, vector unsigned char);
16892 vector unsigned char vec_slo (vector unsigned char, vector signed char);
16893 vector unsigned char vec_slo (vector unsigned char,
16894 vector unsigned char);
16895 vector signed long long vec_slo (vector signed long long, vector signed char);
16896 vector signed long long vec_slo (vector signed long long, vector unsigned char);
16897 vector unsigned long long vec_slo (vector unsigned long long, vector signed char);
16898 vector unsigned long long vec_slo (vector unsigned long long, vector unsigned char);
16900 vector signed char vec_splat (vector signed char, const int);
16901 vector unsigned char vec_splat (vector unsigned char, const int);
16902 vector bool char vec_splat (vector bool char, const int);
16903 vector signed short vec_splat (vector signed short, const int);
16904 vector unsigned short vec_splat (vector unsigned short, const int);
16905 vector bool short vec_splat (vector bool short, const int);
16906 vector pixel vec_splat (vector pixel, const int);
16907 vector float vec_splat (vector float, const int);
16908 vector signed int vec_splat (vector signed int, const int);
16909 vector unsigned int vec_splat (vector unsigned int, const int);
16910 vector bool int vec_splat (vector bool int, const int);
16911 vector signed long vec_splat (vector signed long, const int);
16912 vector unsigned long vec_splat (vector unsigned long, const int);
16914 vector signed char vec_splats (signed char);
16915 vector unsigned char vec_splats (unsigned char);
16916 vector signed short vec_splats (signed short);
16917 vector unsigned short vec_splats (unsigned short);
16918 vector signed int vec_splats (signed int);
16919 vector unsigned int vec_splats (unsigned int);
16920 vector float vec_splats (float);
16922 vector float vec_vspltw (vector float, const int);
16923 vector signed int vec_vspltw (vector signed int, const int);
16924 vector unsigned int vec_vspltw (vector unsigned int, const int);
16925 vector bool int vec_vspltw (vector bool int, const int);
16927 vector bool short vec_vsplth (vector bool short, const int);
16928 vector signed short vec_vsplth (vector signed short, const int);
16929 vector unsigned short vec_vsplth (vector unsigned short, const int);
16930 vector pixel vec_vsplth (vector pixel, const int);
16932 vector signed char vec_vspltb (vector signed char, const int);
16933 vector unsigned char vec_vspltb (vector unsigned char, const int);
16934 vector bool char vec_vspltb (vector bool char, const int);
16936 vector signed char vec_splat_s8 (const int);
16938 vector signed short vec_splat_s16 (const int);
16940 vector signed int vec_splat_s32 (const int);
16942 vector unsigned char vec_splat_u8 (const int);
16944 vector unsigned short vec_splat_u16 (const int);
16946 vector unsigned int vec_splat_u32 (const int);
16948 vector signed char vec_sr (vector signed char, vector unsigned char);
16949 vector unsigned char vec_sr (vector unsigned char,
16950 vector unsigned char);
16951 vector signed short vec_sr (vector signed short,
16952 vector unsigned short);
16953 vector unsigned short vec_sr (vector unsigned short,
16954 vector unsigned short);
16955 vector signed int vec_sr (vector signed int, vector unsigned int);
16956 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
16958 vector signed int vec_vsrw (vector signed int, vector unsigned int);
16959 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
16961 vector signed short vec_vsrh (vector signed short,
16962 vector unsigned short);
16963 vector unsigned short vec_vsrh (vector unsigned short,
16964 vector unsigned short);
16966 vector signed char vec_vsrb (vector signed char, vector unsigned char);
16967 vector unsigned char vec_vsrb (vector unsigned char,
16968 vector unsigned char);
16970 vector signed char vec_sra (vector signed char, vector unsigned char);
16971 vector unsigned char vec_sra (vector unsigned char,
16972 vector unsigned char);
16973 vector signed short vec_sra (vector signed short,
16974 vector unsigned short);
16975 vector unsigned short vec_sra (vector unsigned short,
16976 vector unsigned short);
16977 vector signed int vec_sra (vector signed int, vector unsigned int);
16978 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
16980 vector signed int vec_vsraw (vector signed int, vector unsigned int);
16981 vector unsigned int vec_vsraw (vector unsigned int,
16982 vector unsigned int);
16984 vector signed short vec_vsrah (vector signed short,
16985 vector unsigned short);
16986 vector unsigned short vec_vsrah (vector unsigned short,
16987 vector unsigned short);
16989 vector signed char vec_vsrab (vector signed char, vector unsigned char);
16990 vector unsigned char vec_vsrab (vector unsigned char,
16991 vector unsigned char);
16993 vector signed int vec_srl (vector signed int, vector unsigned int);
16994 vector signed int vec_srl (vector signed int, vector unsigned short);
16995 vector signed int vec_srl (vector signed int, vector unsigned char);
16996 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
16997 vector unsigned int vec_srl (vector unsigned int,
16998 vector unsigned short);
16999 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
17000 vector bool int vec_srl (vector bool int, vector unsigned int);
17001 vector bool int vec_srl (vector bool int, vector unsigned short);
17002 vector bool int vec_srl (vector bool int, vector unsigned char);
17003 vector signed short vec_srl (vector signed short, vector unsigned int);
17004 vector signed short vec_srl (vector signed short,
17005 vector unsigned short);
17006 vector signed short vec_srl (vector signed short, vector unsigned char);
17007 vector unsigned short vec_srl (vector unsigned short,
17008 vector unsigned int);
17009 vector unsigned short vec_srl (vector unsigned short,
17010 vector unsigned short);
17011 vector unsigned short vec_srl (vector unsigned short,
17012 vector unsigned char);
17013 vector bool short vec_srl (vector bool short, vector unsigned int);
17014 vector bool short vec_srl (vector bool short, vector unsigned short);
17015 vector bool short vec_srl (vector bool short, vector unsigned char);
17016 vector pixel vec_srl (vector pixel, vector unsigned int);
17017 vector pixel vec_srl (vector pixel, vector unsigned short);
17018 vector pixel vec_srl (vector pixel, vector unsigned char);
17019 vector signed char vec_srl (vector signed char, vector unsigned int);
17020 vector signed char vec_srl (vector signed char, vector unsigned short);
17021 vector signed char vec_srl (vector signed char, vector unsigned char);
17022 vector unsigned char vec_srl (vector unsigned char,
17023 vector unsigned int);
17024 vector unsigned char vec_srl (vector unsigned char,
17025 vector unsigned short);
17026 vector unsigned char vec_srl (vector unsigned char,
17027 vector unsigned char);
17028 vector bool char vec_srl (vector bool char, vector unsigned int);
17029 vector bool char vec_srl (vector bool char, vector unsigned short);
17030 vector bool char vec_srl (vector bool char, vector unsigned char);
17032 vector float vec_sro (vector float, vector signed char);
17033 vector float vec_sro (vector float, vector unsigned char);
17034 vector signed int vec_sro (vector signed int, vector signed char);
17035 vector signed int vec_sro (vector signed int, vector unsigned char);
17036 vector unsigned int vec_sro (vector unsigned int, vector signed char);
17037 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
17038 vector signed short vec_sro (vector signed short, vector signed char);
17039 vector signed short vec_sro (vector signed short, vector unsigned char);
17040 vector unsigned short vec_sro (vector unsigned short,
17041 vector signed char);
17042 vector unsigned short vec_sro (vector unsigned short,
17043 vector unsigned char);
17044 vector pixel vec_sro (vector pixel, vector signed char);
17045 vector pixel vec_sro (vector pixel, vector unsigned char);
17046 vector signed char vec_sro (vector signed char, vector signed char);
17047 vector signed char vec_sro (vector signed char, vector unsigned char);
17048 vector unsigned char vec_sro (vector unsigned char, vector signed char);
17049 vector unsigned char vec_sro (vector unsigned char,
17050 vector unsigned char);
17052 void vec_st (vector float, int, vector float *);
17053 void vec_st (vector float, int, float *);
17054 void vec_st (vector signed int, int, vector signed int *);
17055 void vec_st (vector signed int, int, int *);
17056 void vec_st (vector unsigned int, int, vector unsigned int *);
17057 void vec_st (vector unsigned int, int, unsigned int *);
17058 void vec_st (vector bool int, int, vector bool int *);
17059 void vec_st (vector bool int, int, unsigned int *);
17060 void vec_st (vector bool int, int, int *);
17061 void vec_st (vector signed short, int, vector signed short *);
17062 void vec_st (vector signed short, int, short *);
17063 void vec_st (vector unsigned short, int, vector unsigned short *);
17064 void vec_st (vector unsigned short, int, unsigned short *);
17065 void vec_st (vector bool short, int, vector bool short *);
17066 void vec_st (vector bool short, int, unsigned short *);
17067 void vec_st (vector pixel, int, vector pixel *);
17068 void vec_st (vector pixel, int, unsigned short *);
17069 void vec_st (vector pixel, int, short *);
17070 void vec_st (vector bool short, int, short *);
17071 void vec_st (vector signed char, int, vector signed char *);
17072 void vec_st (vector signed char, int, signed char *);
17073 void vec_st (vector unsigned char, int, vector unsigned char *);
17074 void vec_st (vector unsigned char, int, unsigned char *);
17075 void vec_st (vector bool char, int, vector bool char *);
17076 void vec_st (vector bool char, int, unsigned char *);
17077 void vec_st (vector bool char, int, signed char *);
17079 void vec_ste (vector signed char, int, signed char *);
17080 void vec_ste (vector unsigned char, int, unsigned char *);
17081 void vec_ste (vector bool char, int, signed char *);
17082 void vec_ste (vector bool char, int, unsigned char *);
17083 void vec_ste (vector signed short, int, short *);
17084 void vec_ste (vector unsigned short, int, unsigned short *);
17085 void vec_ste (vector bool short, int, short *);
17086 void vec_ste (vector bool short, int, unsigned short *);
17087 void vec_ste (vector pixel, int, short *);
17088 void vec_ste (vector pixel, int, unsigned short *);
17089 void vec_ste (vector float, int, float *);
17090 void vec_ste (vector signed int, int, int *);
17091 void vec_ste (vector unsigned int, int, unsigned int *);
17092 void vec_ste (vector bool int, int, int *);
17093 void vec_ste (vector bool int, int, unsigned int *);
17095 void vec_stvewx (vector float, int, float *);
17096 void vec_stvewx (vector signed int, int, int *);
17097 void vec_stvewx (vector unsigned int, int, unsigned int *);
17098 void vec_stvewx (vector bool int, int, int *);
17099 void vec_stvewx (vector bool int, int, unsigned int *);
17101 void vec_stvehx (vector signed short, int, short *);
17102 void vec_stvehx (vector unsigned short, int, unsigned short *);
17103 void vec_stvehx (vector bool short, int, short *);
17104 void vec_stvehx (vector bool short, int, unsigned short *);
17105 void vec_stvehx (vector pixel, int, short *);
17106 void vec_stvehx (vector pixel, int, unsigned short *);
17108 void vec_stvebx (vector signed char, int, signed char *);
17109 void vec_stvebx (vector unsigned char, int, unsigned char *);
17110 void vec_stvebx (vector bool char, int, signed char *);
17111 void vec_stvebx (vector bool char, int, unsigned char *);
17113 void vec_stl (vector float, int, vector float *);
17114 void vec_stl (vector float, int, float *);
17115 void vec_stl (vector signed int, int, vector signed int *);
17116 void vec_stl (vector signed int, int, int *);
17117 void vec_stl (vector unsigned int, int, vector unsigned int *);
17118 void vec_stl (vector unsigned int, int, unsigned int *);
17119 void vec_stl (vector bool int, int, vector bool int *);
17120 void vec_stl (vector bool int, int, unsigned int *);
17121 void vec_stl (vector bool int, int, int *);
17122 void vec_stl (vector signed short, int, vector signed short *);
17123 void vec_stl (vector signed short, int, short *);
17124 void vec_stl (vector unsigned short, int, vector unsigned short *);
17125 void vec_stl (vector unsigned short, int, unsigned short *);
17126 void vec_stl (vector bool short, int, vector bool short *);
17127 void vec_stl (vector bool short, int, unsigned short *);
17128 void vec_stl (vector bool short, int, short *);
17129 void vec_stl (vector pixel, int, vector pixel *);
17130 void vec_stl (vector pixel, int, unsigned short *);
17131 void vec_stl (vector pixel, int, short *);
17132 void vec_stl (vector signed char, int, vector signed char *);
17133 void vec_stl (vector signed char, int, signed char *);
17134 void vec_stl (vector unsigned char, int, vector unsigned char *);
17135 void vec_stl (vector unsigned char, int, unsigned char *);
17136 void vec_stl (vector bool char, int, vector bool char *);
17137 void vec_stl (vector bool char, int, unsigned char *);
17138 void vec_stl (vector bool char, int, signed char *);
17140 vector signed char vec_sub (vector bool char, vector signed char);
17141 vector signed char vec_sub (vector signed char, vector bool char);
17142 vector signed char vec_sub (vector signed char, vector signed char);
17143 vector unsigned char vec_sub (vector bool char, vector unsigned char);
17144 vector unsigned char vec_sub (vector unsigned char, vector bool char);
17145 vector unsigned char vec_sub (vector unsigned char,
17146 vector unsigned char);
17147 vector signed short vec_sub (vector bool short, vector signed short);
17148 vector signed short vec_sub (vector signed short, vector bool short);
17149 vector signed short vec_sub (vector signed short, vector signed short);
17150 vector unsigned short vec_sub (vector bool short,
17151 vector unsigned short);
17152 vector unsigned short vec_sub (vector unsigned short,
17153 vector bool short);
17154 vector unsigned short vec_sub (vector unsigned short,
17155 vector unsigned short);
17156 vector signed int vec_sub (vector bool int, vector signed int);
17157 vector signed int vec_sub (vector signed int, vector bool int);
17158 vector signed int vec_sub (vector signed int, vector signed int);
17159 vector unsigned int vec_sub (vector bool int, vector unsigned int);
17160 vector unsigned int vec_sub (vector unsigned int, vector bool int);
17161 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
17162 vector float vec_sub (vector float, vector float);
17164 vector float vec_vsubfp (vector float, vector float);
17166 vector signed int vec_vsubuwm (vector bool int, vector signed int);
17167 vector signed int vec_vsubuwm (vector signed int, vector bool int);
17168 vector signed int vec_vsubuwm (vector signed int, vector signed int);
17169 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
17170 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
17171 vector unsigned int vec_vsubuwm (vector unsigned int,
17172 vector unsigned int);
17174 vector signed short vec_vsubuhm (vector bool short,
17175 vector signed short);
17176 vector signed short vec_vsubuhm (vector signed short,
17177 vector bool short);
17178 vector signed short vec_vsubuhm (vector signed short,
17179 vector signed short);
17180 vector unsigned short vec_vsubuhm (vector bool short,
17181 vector unsigned short);
17182 vector unsigned short vec_vsubuhm (vector unsigned short,
17183 vector bool short);
17184 vector unsigned short vec_vsubuhm (vector unsigned short,
17185 vector unsigned short);
17187 vector signed char vec_vsububm (vector bool char, vector signed char);
17188 vector signed char vec_vsububm (vector signed char, vector bool char);
17189 vector signed char vec_vsububm (vector signed char, vector signed char);
17190 vector unsigned char vec_vsububm (vector bool char,
17191 vector unsigned char);
17192 vector unsigned char vec_vsububm (vector unsigned char,
17194 vector unsigned char vec_vsububm (vector unsigned char,
17195 vector unsigned char);
17197 vector signed int vec_subc (vector signed int, vector signed int);
17198 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
17199 vector signed __int128 vec_subc (vector signed __int128,
17200 vector signed __int128);
17201 vector unsigned __int128 vec_subc (vector unsigned __int128,
17202 vector unsigned __int128);
17204 vector signed int vec_sube (vector signed int, vector signed int,
17205 vector signed int);
17206 vector unsigned int vec_sube (vector unsigned int, vector unsigned int,
17207 vector unsigned int);
17208 vector signed __int128 vec_sube (vector signed __int128,
17209 vector signed __int128,
17210 vector signed __int128);
17211 vector unsigned __int128 vec_sube (vector unsigned __int128,
17212 vector unsigned __int128,
17213 vector unsigned __int128);
17215 vector signed int vec_subec (vector signed int, vector signed int,
17216 vector signed int);
17217 vector unsigned int vec_subec (vector unsigned int, vector unsigned int,
17218 vector unsigned int);
17219 vector signed __int128 vec_subec (vector signed __int128,
17220 vector signed __int128,
17221 vector signed __int128);
17222 vector unsigned __int128 vec_subec (vector unsigned __int128,
17223 vector unsigned __int128,
17224 vector unsigned __int128);
17226 vector unsigned char vec_subs (vector bool char, vector unsigned char);
17227 vector unsigned char vec_subs (vector unsigned char, vector bool char);
17228 vector unsigned char vec_subs (vector unsigned char,
17229 vector unsigned char);
17230 vector signed char vec_subs (vector bool char, vector signed char);
17231 vector signed char vec_subs (vector signed char, vector bool char);
17232 vector signed char vec_subs (vector signed char, vector signed char);
17233 vector unsigned short vec_subs (vector bool short,
17234 vector unsigned short);
17235 vector unsigned short vec_subs (vector unsigned short,
17236 vector bool short);
17237 vector unsigned short vec_subs (vector unsigned short,
17238 vector unsigned short);
17239 vector signed short vec_subs (vector bool short, vector signed short);
17240 vector signed short vec_subs (vector signed short, vector bool short);
17241 vector signed short vec_subs (vector signed short, vector signed short);
17242 vector unsigned int vec_subs (vector bool int, vector unsigned int);
17243 vector unsigned int vec_subs (vector unsigned int, vector bool int);
17244 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
17245 vector signed int vec_subs (vector bool int, vector signed int);
17246 vector signed int vec_subs (vector signed int, vector bool int);
17247 vector signed int vec_subs (vector signed int, vector signed int);
17249 vector signed int vec_vsubsws (vector bool int, vector signed int);
17250 vector signed int vec_vsubsws (vector signed int, vector bool int);
17251 vector signed int vec_vsubsws (vector signed int, vector signed int);
17253 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
17254 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
17255 vector unsigned int vec_vsubuws (vector unsigned int,
17256 vector unsigned int);
17258 vector signed short vec_vsubshs (vector bool short,
17259 vector signed short);
17260 vector signed short vec_vsubshs (vector signed short,
17261 vector bool short);
17262 vector signed short vec_vsubshs (vector signed short,
17263 vector signed short);
17265 vector unsigned short vec_vsubuhs (vector bool short,
17266 vector unsigned short);
17267 vector unsigned short vec_vsubuhs (vector unsigned short,
17268 vector bool short);
17269 vector unsigned short vec_vsubuhs (vector unsigned short,
17270 vector unsigned short);
17272 vector signed char vec_vsubsbs (vector bool char, vector signed char);
17273 vector signed char vec_vsubsbs (vector signed char, vector bool char);
17274 vector signed char vec_vsubsbs (vector signed char, vector signed char);
17276 vector unsigned char vec_vsububs (vector bool char,
17277 vector unsigned char);
17278 vector unsigned char vec_vsububs (vector unsigned char,
17280 vector unsigned char vec_vsububs (vector unsigned char,
17281 vector unsigned char);
17283 vector unsigned int vec_sum4s (vector unsigned char,
17284 vector unsigned int);
17285 vector signed int vec_sum4s (vector signed char, vector signed int);
17286 vector signed int vec_sum4s (vector signed short, vector signed int);
17288 vector signed int vec_vsum4shs (vector signed short, vector signed int);
17290 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
17292 vector unsigned int vec_vsum4ubs (vector unsigned char,
17293 vector unsigned int);
17295 vector signed int vec_sum2s (vector signed int, vector signed int);
17297 vector signed int vec_sums (vector signed int, vector signed int);
17299 vector float vec_trunc (vector float);
17301 vector signed long long vec_unsigned (vector double);
17302 vector signed int vec_unsigned (vector float);
17304 vector signed int vec_unsignede (vector double);
17305 vector signed int vec_unsignedo (vector double);
17306 vector signed int vec_unsigned2 (vector double, vector double);
17308 vector signed short vec_unpackh (vector signed char);
17309 vector bool short vec_unpackh (vector bool char);
17310 vector signed int vec_unpackh (vector signed short);
17311 vector bool int vec_unpackh (vector bool short);
17312 vector unsigned int vec_unpackh (vector pixel);
17314 vector bool int vec_vupkhsh (vector bool short);
17315 vector signed int vec_vupkhsh (vector signed short);
17317 vector unsigned int vec_vupkhpx (vector pixel);
17319 vector bool short vec_vupkhsb (vector bool char);
17320 vector signed short vec_vupkhsb (vector signed char);
17322 vector signed short vec_unpackl (vector signed char);
17323 vector bool short vec_unpackl (vector bool char);
17324 vector unsigned int vec_unpackl (vector pixel);
17325 vector signed int vec_unpackl (vector signed short);
17326 vector bool int vec_unpackl (vector bool short);
17328 vector unsigned int vec_vupklpx (vector pixel);
17330 vector bool int vec_vupklsh (vector bool short);
17331 vector signed int vec_vupklsh (vector signed short);
17333 vector bool short vec_vupklsb (vector bool char);
17334 vector signed short vec_vupklsb (vector signed char);
17336 vector float vec_xor (vector float, vector float);
17337 vector float vec_xor (vector float, vector bool int);
17338 vector float vec_xor (vector bool int, vector float);
17339 vector bool int vec_xor (vector bool int, vector bool int);
17340 vector signed int vec_xor (vector bool int, vector signed int);
17341 vector signed int vec_xor (vector signed int, vector bool int);
17342 vector signed int vec_xor (vector signed int, vector signed int);
17343 vector unsigned int vec_xor (vector bool int, vector unsigned int);
17344 vector unsigned int vec_xor (vector unsigned int, vector bool int);
17345 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
17346 vector bool short vec_xor (vector bool short, vector bool short);
17347 vector signed short vec_xor (vector bool short, vector signed short);
17348 vector signed short vec_xor (vector signed short, vector bool short);
17349 vector signed short vec_xor (vector signed short, vector signed short);
17350 vector unsigned short vec_xor (vector bool short,
17351 vector unsigned short);
17352 vector unsigned short vec_xor (vector unsigned short,
17353 vector bool short);
17354 vector unsigned short vec_xor (vector unsigned short,
17355 vector unsigned short);
17356 vector signed char vec_xor (vector bool char, vector signed char);
17357 vector bool char vec_xor (vector bool char, vector bool char);
17358 vector signed char vec_xor (vector signed char, vector bool char);
17359 vector signed char vec_xor (vector signed char, vector signed char);
17360 vector unsigned char vec_xor (vector bool char, vector unsigned char);
17361 vector unsigned char vec_xor (vector unsigned char, vector bool char);
17362 vector unsigned char vec_xor (vector unsigned char,
17363 vector unsigned char);
17365 int vec_all_eq (vector signed char, vector bool char);
17366 int vec_all_eq (vector signed char, vector signed char);
17367 int vec_all_eq (vector unsigned char, vector bool char);
17368 int vec_all_eq (vector unsigned char, vector unsigned char);
17369 int vec_all_eq (vector bool char, vector bool char);
17370 int vec_all_eq (vector bool char, vector unsigned char);
17371 int vec_all_eq (vector bool char, vector signed char);
17372 int vec_all_eq (vector signed short, vector bool short);
17373 int vec_all_eq (vector signed short, vector signed short);
17374 int vec_all_eq (vector unsigned short, vector bool short);
17375 int vec_all_eq (vector unsigned short, vector unsigned short);
17376 int vec_all_eq (vector bool short, vector bool short);
17377 int vec_all_eq (vector bool short, vector unsigned short);
17378 int vec_all_eq (vector bool short, vector signed short);
17379 int vec_all_eq (vector pixel, vector pixel);
17380 int vec_all_eq (vector signed int, vector bool int);
17381 int vec_all_eq (vector signed int, vector signed int);
17382 int vec_all_eq (vector unsigned int, vector bool int);
17383 int vec_all_eq (vector unsigned int, vector unsigned int);
17384 int vec_all_eq (vector bool int, vector bool int);
17385 int vec_all_eq (vector bool int, vector unsigned int);
17386 int vec_all_eq (vector bool int, vector signed int);
17387 int vec_all_eq (vector float, vector float);
17389 int vec_all_ge (vector bool char, vector unsigned char);
17390 int vec_all_ge (vector unsigned char, vector bool char);
17391 int vec_all_ge (vector unsigned char, vector unsigned char);
17392 int vec_all_ge (vector bool char, vector signed char);
17393 int vec_all_ge (vector signed char, vector bool char);
17394 int vec_all_ge (vector signed char, vector signed char);
17395 int vec_all_ge (vector bool short, vector unsigned short);
17396 int vec_all_ge (vector unsigned short, vector bool short);
17397 int vec_all_ge (vector unsigned short, vector unsigned short);
17398 int vec_all_ge (vector signed short, vector signed short);
17399 int vec_all_ge (vector bool short, vector signed short);
17400 int vec_all_ge (vector signed short, vector bool short);
17401 int vec_all_ge (vector bool int, vector unsigned int);
17402 int vec_all_ge (vector unsigned int, vector bool int);
17403 int vec_all_ge (vector unsigned int, vector unsigned int);
17404 int vec_all_ge (vector bool int, vector signed int);
17405 int vec_all_ge (vector signed int, vector bool int);
17406 int vec_all_ge (vector signed int, vector signed int);
17407 int vec_all_ge (vector float, vector float);
17409 int vec_all_gt (vector bool char, vector unsigned char);
17410 int vec_all_gt (vector unsigned char, vector bool char);
17411 int vec_all_gt (vector unsigned char, vector unsigned char);
17412 int vec_all_gt (vector bool char, vector signed char);
17413 int vec_all_gt (vector signed char, vector bool char);
17414 int vec_all_gt (vector signed char, vector signed char);
17415 int vec_all_gt (vector bool short, vector unsigned short);
17416 int vec_all_gt (vector unsigned short, vector bool short);
17417 int vec_all_gt (vector unsigned short, vector unsigned short);
17418 int vec_all_gt (vector bool short, vector signed short);
17419 int vec_all_gt (vector signed short, vector bool short);
17420 int vec_all_gt (vector signed short, vector signed short);
17421 int vec_all_gt (vector bool int, vector unsigned int);
17422 int vec_all_gt (vector unsigned int, vector bool int);
17423 int vec_all_gt (vector unsigned int, vector unsigned int);
17424 int vec_all_gt (vector bool int, vector signed int);
17425 int vec_all_gt (vector signed int, vector bool int);
17426 int vec_all_gt (vector signed int, vector signed int);
17427 int vec_all_gt (vector float, vector float);
17429 int vec_all_in (vector float, vector float);
17431 int vec_all_le (vector bool char, vector unsigned char);
17432 int vec_all_le (vector unsigned char, vector bool char);
17433 int vec_all_le (vector unsigned char, vector unsigned char);
17434 int vec_all_le (vector bool char, vector signed char);
17435 int vec_all_le (vector signed char, vector bool char);
17436 int vec_all_le (vector signed char, vector signed char);
17437 int vec_all_le (vector bool short, vector unsigned short);
17438 int vec_all_le (vector unsigned short, vector bool short);
17439 int vec_all_le (vector unsigned short, vector unsigned short);
17440 int vec_all_le (vector bool short, vector signed short);
17441 int vec_all_le (vector signed short, vector bool short);
17442 int vec_all_le (vector signed short, vector signed short);
17443 int vec_all_le (vector bool int, vector unsigned int);
17444 int vec_all_le (vector unsigned int, vector bool int);
17445 int vec_all_le (vector unsigned int, vector unsigned int);
17446 int vec_all_le (vector bool int, vector signed int);
17447 int vec_all_le (vector signed int, vector bool int);
17448 int vec_all_le (vector signed int, vector signed int);
17449 int vec_all_le (vector float, vector float);
17451 int vec_all_lt (vector bool char, vector unsigned char);
17452 int vec_all_lt (vector unsigned char, vector bool char);
17453 int vec_all_lt (vector unsigned char, vector unsigned char);
17454 int vec_all_lt (vector bool char, vector signed char);
17455 int vec_all_lt (vector signed char, vector bool char);
17456 int vec_all_lt (vector signed char, vector signed char);
17457 int vec_all_lt (vector bool short, vector unsigned short);
17458 int vec_all_lt (vector unsigned short, vector bool short);
17459 int vec_all_lt (vector unsigned short, vector unsigned short);
17460 int vec_all_lt (vector bool short, vector signed short);
17461 int vec_all_lt (vector signed short, vector bool short);
17462 int vec_all_lt (vector signed short, vector signed short);
17463 int vec_all_lt (vector bool int, vector unsigned int);
17464 int vec_all_lt (vector unsigned int, vector bool int);
17465 int vec_all_lt (vector unsigned int, vector unsigned int);
17466 int vec_all_lt (vector bool int, vector signed int);
17467 int vec_all_lt (vector signed int, vector bool int);
17468 int vec_all_lt (vector signed int, vector signed int);
17469 int vec_all_lt (vector float, vector float);
17471 int vec_all_nan (vector float);
17473 int vec_all_ne (vector signed char, vector bool char);
17474 int vec_all_ne (vector signed char, vector signed char);
17475 int vec_all_ne (vector unsigned char, vector bool char);
17476 int vec_all_ne (vector unsigned char, vector unsigned char);
17477 int vec_all_ne (vector bool char, vector bool char);
17478 int vec_all_ne (vector bool char, vector unsigned char);
17479 int vec_all_ne (vector bool char, vector signed char);
17480 int vec_all_ne (vector signed short, vector bool short);
17481 int vec_all_ne (vector signed short, vector signed short);
17482 int vec_all_ne (vector unsigned short, vector bool short);
17483 int vec_all_ne (vector unsigned short, vector unsigned short);
17484 int vec_all_ne (vector bool short, vector bool short);
17485 int vec_all_ne (vector bool short, vector unsigned short);
17486 int vec_all_ne (vector bool short, vector signed short);
17487 int vec_all_ne (vector pixel, vector pixel);
17488 int vec_all_ne (vector signed int, vector bool int);
17489 int vec_all_ne (vector signed int, vector signed int);
17490 int vec_all_ne (vector unsigned int, vector bool int);
17491 int vec_all_ne (vector unsigned int, vector unsigned int);
17492 int vec_all_ne (vector bool int, vector bool int);
17493 int vec_all_ne (vector bool int, vector unsigned int);
17494 int vec_all_ne (vector bool int, vector signed int);
17495 int vec_all_ne (vector float, vector float);
17497 int vec_all_nge (vector float, vector float);
17499 int vec_all_ngt (vector float, vector float);
17501 int vec_all_nle (vector float, vector float);
17503 int vec_all_nlt (vector float, vector float);
17505 int vec_all_numeric (vector float);
17507 int vec_any_eq (vector signed char, vector bool char);
17508 int vec_any_eq (vector signed char, vector signed char);
17509 int vec_any_eq (vector unsigned char, vector bool char);
17510 int vec_any_eq (vector unsigned char, vector unsigned char);
17511 int vec_any_eq (vector bool char, vector bool char);
17512 int vec_any_eq (vector bool char, vector unsigned char);
17513 int vec_any_eq (vector bool char, vector signed char);
17514 int vec_any_eq (vector signed short, vector bool short);
17515 int vec_any_eq (vector signed short, vector signed short);
17516 int vec_any_eq (vector unsigned short, vector bool short);
17517 int vec_any_eq (vector unsigned short, vector unsigned short);
17518 int vec_any_eq (vector bool short, vector bool short);
17519 int vec_any_eq (vector bool short, vector unsigned short);
17520 int vec_any_eq (vector bool short, vector signed short);
17521 int vec_any_eq (vector pixel, vector pixel);
17522 int vec_any_eq (vector signed int, vector bool int);
17523 int vec_any_eq (vector signed int, vector signed int);
17524 int vec_any_eq (vector unsigned int, vector bool int);
17525 int vec_any_eq (vector unsigned int, vector unsigned int);
17526 int vec_any_eq (vector bool int, vector bool int);
17527 int vec_any_eq (vector bool int, vector unsigned int);
17528 int vec_any_eq (vector bool int, vector signed int);
17529 int vec_any_eq (vector float, vector float);
17531 int vec_any_ge (vector signed char, vector bool char);
17532 int vec_any_ge (vector unsigned char, vector bool char);
17533 int vec_any_ge (vector unsigned char, vector unsigned char);
17534 int vec_any_ge (vector signed char, vector signed char);
17535 int vec_any_ge (vector bool char, vector unsigned char);
17536 int vec_any_ge (vector bool char, vector signed char);
17537 int vec_any_ge (vector unsigned short, vector bool short);
17538 int vec_any_ge (vector unsigned short, vector unsigned short);
17539 int vec_any_ge (vector signed short, vector signed short);
17540 int vec_any_ge (vector signed short, vector bool short);
17541 int vec_any_ge (vector bool short, vector unsigned short);
17542 int vec_any_ge (vector bool short, vector signed short);
17543 int vec_any_ge (vector signed int, vector bool int);
17544 int vec_any_ge (vector unsigned int, vector bool int);
17545 int vec_any_ge (vector unsigned int, vector unsigned int);
17546 int vec_any_ge (vector signed int, vector signed int);
17547 int vec_any_ge (vector bool int, vector unsigned int);
17548 int vec_any_ge (vector bool int, vector signed int);
17549 int vec_any_ge (vector float, vector float);
17551 int vec_any_gt (vector bool char, vector unsigned char);
17552 int vec_any_gt (vector unsigned char, vector bool char);
17553 int vec_any_gt (vector unsigned char, vector unsigned char);
17554 int vec_any_gt (vector bool char, vector signed char);
17555 int vec_any_gt (vector signed char, vector bool char);
17556 int vec_any_gt (vector signed char, vector signed char);
17557 int vec_any_gt (vector bool short, vector unsigned short);
17558 int vec_any_gt (vector unsigned short, vector bool short);
17559 int vec_any_gt (vector unsigned short, vector unsigned short);
17560 int vec_any_gt (vector bool short, vector signed short);
17561 int vec_any_gt (vector signed short, vector bool short);
17562 int vec_any_gt (vector signed short, vector signed short);
17563 int vec_any_gt (vector bool int, vector unsigned int);
17564 int vec_any_gt (vector unsigned int, vector bool int);
17565 int vec_any_gt (vector unsigned int, vector unsigned int);
17566 int vec_any_gt (vector bool int, vector signed int);
17567 int vec_any_gt (vector signed int, vector bool int);
17568 int vec_any_gt (vector signed int, vector signed int);
17569 int vec_any_gt (vector float, vector float);
17571 int vec_any_le (vector bool char, vector unsigned char);
17572 int vec_any_le (vector unsigned char, vector bool char);
17573 int vec_any_le (vector unsigned char, vector unsigned char);
17574 int vec_any_le (vector bool char, vector signed char);
17575 int vec_any_le (vector signed char, vector bool char);
17576 int vec_any_le (vector signed char, vector signed char);
17577 int vec_any_le (vector bool short, vector unsigned short);
17578 int vec_any_le (vector unsigned short, vector bool short);
17579 int vec_any_le (vector unsigned short, vector unsigned short);
17580 int vec_any_le (vector bool short, vector signed short);
17581 int vec_any_le (vector signed short, vector bool short);
17582 int vec_any_le (vector signed short, vector signed short);
17583 int vec_any_le (vector bool int, vector unsigned int);
17584 int vec_any_le (vector unsigned int, vector bool int);
17585 int vec_any_le (vector unsigned int, vector unsigned int);
17586 int vec_any_le (vector bool int, vector signed int);
17587 int vec_any_le (vector signed int, vector bool int);
17588 int vec_any_le (vector signed int, vector signed int);
17589 int vec_any_le (vector float, vector float);
17591 int vec_any_lt (vector bool char, vector unsigned char);
17592 int vec_any_lt (vector unsigned char, vector bool char);
17593 int vec_any_lt (vector unsigned char, vector unsigned char);
17594 int vec_any_lt (vector bool char, vector signed char);
17595 int vec_any_lt (vector signed char, vector bool char);
17596 int vec_any_lt (vector signed char, vector signed char);
17597 int vec_any_lt (vector bool short, vector unsigned short);
17598 int vec_any_lt (vector unsigned short, vector bool short);
17599 int vec_any_lt (vector unsigned short, vector unsigned short);
17600 int vec_any_lt (vector bool short, vector signed short);
17601 int vec_any_lt (vector signed short, vector bool short);
17602 int vec_any_lt (vector signed short, vector signed short);
17603 int vec_any_lt (vector bool int, vector unsigned int);
17604 int vec_any_lt (vector unsigned int, vector bool int);
17605 int vec_any_lt (vector unsigned int, vector unsigned int);
17606 int vec_any_lt (vector bool int, vector signed int);
17607 int vec_any_lt (vector signed int, vector bool int);
17608 int vec_any_lt (vector signed int, vector signed int);
17609 int vec_any_lt (vector float, vector float);
17611 int vec_any_nan (vector float);
17613 int vec_any_ne (vector signed char, vector bool char);
17614 int vec_any_ne (vector signed char, vector signed char);
17615 int vec_any_ne (vector unsigned char, vector bool char);
17616 int vec_any_ne (vector unsigned char, vector unsigned char);
17617 int vec_any_ne (vector bool char, vector bool char);
17618 int vec_any_ne (vector bool char, vector unsigned char);
17619 int vec_any_ne (vector bool char, vector signed char);
17620 int vec_any_ne (vector signed short, vector bool short);
17621 int vec_any_ne (vector signed short, vector signed short);
17622 int vec_any_ne (vector unsigned short, vector bool short);
17623 int vec_any_ne (vector unsigned short, vector unsigned short);
17624 int vec_any_ne (vector bool short, vector bool short);
17625 int vec_any_ne (vector bool short, vector unsigned short);
17626 int vec_any_ne (vector bool short, vector signed short);
17627 int vec_any_ne (vector pixel, vector pixel);
17628 int vec_any_ne (vector signed int, vector bool int);
17629 int vec_any_ne (vector signed int, vector signed int);
17630 int vec_any_ne (vector unsigned int, vector bool int);
17631 int vec_any_ne (vector unsigned int, vector unsigned int);
17632 int vec_any_ne (vector bool int, vector bool int);
17633 int vec_any_ne (vector bool int, vector unsigned int);
17634 int vec_any_ne (vector bool int, vector signed int);
17635 int vec_any_ne (vector float, vector float);
17637 int vec_any_nge (vector float, vector float);
17639 int vec_any_ngt (vector float, vector float);
17641 int vec_any_nle (vector float, vector float);
17643 int vec_any_nlt (vector float, vector float);
17645 int vec_any_numeric (vector float);
17647 int vec_any_out (vector float, vector float);
17650 If the vector/scalar (VSX) instruction set is available, the following
17651 additional functions are available:
17654 vector double vec_abs (vector double);
17655 vector double vec_add (vector double, vector double);
17656 vector double vec_and (vector double, vector double);
17657 vector double vec_and (vector double, vector bool long);
17658 vector double vec_and (vector bool long, vector double);
17659 vector long vec_and (vector long, vector long);
17660 vector long vec_and (vector long, vector bool long);
17661 vector long vec_and (vector bool long, vector long);
17662 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
17663 vector unsigned long vec_and (vector unsigned long, vector bool long);
17664 vector unsigned long vec_and (vector bool long, vector unsigned long);
17665 vector double vec_andc (vector double, vector double);
17666 vector double vec_andc (vector double, vector bool long);
17667 vector double vec_andc (vector bool long, vector double);
17668 vector long vec_andc (vector long, vector long);
17669 vector long vec_andc (vector long, vector bool long);
17670 vector long vec_andc (vector bool long, vector long);
17671 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
17672 vector unsigned long vec_andc (vector unsigned long, vector bool long);
17673 vector unsigned long vec_andc (vector bool long, vector unsigned long);
17674 vector double vec_ceil (vector double);
17675 vector bool long vec_cmpeq (vector double, vector double);
17676 vector bool long vec_cmpge (vector double, vector double);
17677 vector bool long vec_cmpgt (vector double, vector double);
17678 vector bool long vec_cmple (vector double, vector double);
17679 vector bool long vec_cmplt (vector double, vector double);
17680 vector double vec_cpsgn (vector double, vector double);
17681 vector float vec_div (vector float, vector float);
17682 vector double vec_div (vector double, vector double);
17683 vector long vec_div (vector long, vector long);
17684 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
17685 vector double vec_floor (vector double);
17686 vector double vec_ld (int, const vector double *);
17687 vector double vec_ld (int, const double *);
17688 vector double vec_ldl (int, const vector double *);
17689 vector double vec_ldl (int, const double *);
17690 vector unsigned char vec_lvsl (int, const volatile double *);
17691 vector unsigned char vec_lvsr (int, const volatile double *);
17692 vector double vec_madd (vector double, vector double, vector double);
17693 vector double vec_max (vector double, vector double);
17694 vector signed long vec_mergeh (vector signed long, vector signed long);
17695 vector signed long vec_mergeh (vector signed long, vector bool long);
17696 vector signed long vec_mergeh (vector bool long, vector signed long);
17697 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
17698 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
17699 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
17700 vector signed long vec_mergel (vector signed long, vector signed long);
17701 vector signed long vec_mergel (vector signed long, vector bool long);
17702 vector signed long vec_mergel (vector bool long, vector signed long);
17703 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
17704 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
17705 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
17706 vector double vec_min (vector double, vector double);
17707 vector float vec_msub (vector float, vector float, vector float);
17708 vector double vec_msub (vector double, vector double, vector double);
17709 vector float vec_mul (vector float, vector float);
17710 vector double vec_mul (vector double, vector double);
17711 vector long vec_mul (vector long, vector long);
17712 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
17713 vector float vec_nearbyint (vector float);
17714 vector double vec_nearbyint (vector double);
17715 vector float vec_nmadd (vector float, vector float, vector float);
17716 vector double vec_nmadd (vector double, vector double, vector double);
17717 vector double vec_nmsub (vector double, vector double, vector double);
17718 vector double vec_nor (vector double, vector double);
17719 vector long vec_nor (vector long, vector long);
17720 vector long vec_nor (vector long, vector bool long);
17721 vector long vec_nor (vector bool long, vector long);
17722 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
17723 vector unsigned long vec_nor (vector unsigned long, vector bool long);
17724 vector unsigned long vec_nor (vector bool long, vector unsigned long);
17725 vector double vec_or (vector double, vector double);
17726 vector double vec_or (vector double, vector bool long);
17727 vector double vec_or (vector bool long, vector double);
17728 vector long vec_or (vector long, vector long);
17729 vector long vec_or (vector long, vector bool long);
17730 vector long vec_or (vector bool long, vector long);
17731 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
17732 vector unsigned long vec_or (vector unsigned long, vector bool long);
17733 vector unsigned long vec_or (vector bool long, vector unsigned long);
17734 vector double vec_perm (vector double, vector double, vector unsigned char);
17735 vector long vec_perm (vector long, vector long, vector unsigned char);
17736 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
17737 vector unsigned char);
17738 vector double vec_rint (vector double);
17739 vector double vec_recip (vector double, vector double);
17740 vector double vec_rsqrt (vector double);
17741 vector double vec_rsqrte (vector double);
17742 vector double vec_sel (vector double, vector double, vector bool long);
17743 vector double vec_sel (vector double, vector double, vector unsigned long);
17744 vector long vec_sel (vector long, vector long, vector long);
17745 vector long vec_sel (vector long, vector long, vector unsigned long);
17746 vector long vec_sel (vector long, vector long, vector bool long);
17747 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17749 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17750 vector unsigned long);
17751 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17753 vector double vec_splats (double);
17754 vector signed long vec_splats (signed long);
17755 vector unsigned long vec_splats (unsigned long);
17756 vector float vec_sqrt (vector float);
17757 vector double vec_sqrt (vector double);
17758 void vec_st (vector double, int, vector double *);
17759 void vec_st (vector double, int, double *);
17760 vector double vec_sub (vector double, vector double);
17761 vector double vec_trunc (vector double);
17762 vector double vec_xl (int, vector double *);
17763 vector double vec_xl (int, double *);
17764 vector long long vec_xl (int, vector long long *);
17765 vector long long vec_xl (int, long long *);
17766 vector unsigned long long vec_xl (int, vector unsigned long long *);
17767 vector unsigned long long vec_xl (int, unsigned long long *);
17768 vector float vec_xl (int, vector float *);
17769 vector float vec_xl (int, float *);
17770 vector int vec_xl (int, vector int *);
17771 vector int vec_xl (int, int *);
17772 vector unsigned int vec_xl (int, vector unsigned int *);
17773 vector unsigned int vec_xl (int, unsigned int *);
17774 vector double vec_xor (vector double, vector double);
17775 vector double vec_xor (vector double, vector bool long);
17776 vector double vec_xor (vector bool long, vector double);
17777 vector long vec_xor (vector long, vector long);
17778 vector long vec_xor (vector long, vector bool long);
17779 vector long vec_xor (vector bool long, vector long);
17780 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
17781 vector unsigned long vec_xor (vector unsigned long, vector bool long);
17782 vector unsigned long vec_xor (vector bool long, vector unsigned long);
17783 void vec_xst (vector double, int, vector double *);
17784 void vec_xst (vector double, int, double *);
17785 void vec_xst (vector long long, int, vector long long *);
17786 void vec_xst (vector long long, int, long long *);
17787 void vec_xst (vector unsigned long long, int, vector unsigned long long *);
17788 void vec_xst (vector unsigned long long, int, unsigned long long *);
17789 void vec_xst (vector float, int, vector float *);
17790 void vec_xst (vector float, int, float *);
17791 void vec_xst (vector int, int, vector int *);
17792 void vec_xst (vector int, int, int *);
17793 void vec_xst (vector unsigned int, int, vector unsigned int *);
17794 void vec_xst (vector unsigned int, int, unsigned int *);
17795 int vec_all_eq (vector double, vector double);
17796 int vec_all_ge (vector double, vector double);
17797 int vec_all_gt (vector double, vector double);
17798 int vec_all_le (vector double, vector double);
17799 int vec_all_lt (vector double, vector double);
17800 int vec_all_nan (vector double);
17801 int vec_all_ne (vector double, vector double);
17802 int vec_all_nge (vector double, vector double);
17803 int vec_all_ngt (vector double, vector double);
17804 int vec_all_nle (vector double, vector double);
17805 int vec_all_nlt (vector double, vector double);
17806 int vec_all_numeric (vector double);
17807 int vec_any_eq (vector double, vector double);
17808 int vec_any_ge (vector double, vector double);
17809 int vec_any_gt (vector double, vector double);
17810 int vec_any_le (vector double, vector double);
17811 int vec_any_lt (vector double, vector double);
17812 int vec_any_nan (vector double);
17813 int vec_any_ne (vector double, vector double);
17814 int vec_any_nge (vector double, vector double);
17815 int vec_any_ngt (vector double, vector double);
17816 int vec_any_nle (vector double, vector double);
17817 int vec_any_nlt (vector double, vector double);
17818 int vec_any_numeric (vector double);
17820 vector double vec_vsx_ld (int, const vector double *);
17821 vector double vec_vsx_ld (int, const double *);
17822 vector float vec_vsx_ld (int, const vector float *);
17823 vector float vec_vsx_ld (int, const float *);
17824 vector bool int vec_vsx_ld (int, const vector bool int *);
17825 vector signed int vec_vsx_ld (int, const vector signed int *);
17826 vector signed int vec_vsx_ld (int, const int *);
17827 vector signed int vec_vsx_ld (int, const long *);
17828 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
17829 vector unsigned int vec_vsx_ld (int, const unsigned int *);
17830 vector unsigned int vec_vsx_ld (int, const unsigned long *);
17831 vector bool short vec_vsx_ld (int, const vector bool short *);
17832 vector pixel vec_vsx_ld (int, const vector pixel *);
17833 vector signed short vec_vsx_ld (int, const vector signed short *);
17834 vector signed short vec_vsx_ld (int, const short *);
17835 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
17836 vector unsigned short vec_vsx_ld (int, const unsigned short *);
17837 vector bool char vec_vsx_ld (int, const vector bool char *);
17838 vector signed char vec_vsx_ld (int, const vector signed char *);
17839 vector signed char vec_vsx_ld (int, const signed char *);
17840 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
17841 vector unsigned char vec_vsx_ld (int, const unsigned char *);
17843 void vec_vsx_st (vector double, int, vector double *);
17844 void vec_vsx_st (vector double, int, double *);
17845 void vec_vsx_st (vector float, int, vector float *);
17846 void vec_vsx_st (vector float, int, float *);
17847 void vec_vsx_st (vector signed int, int, vector signed int *);
17848 void vec_vsx_st (vector signed int, int, int *);
17849 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
17850 void vec_vsx_st (vector unsigned int, int, unsigned int *);
17851 void vec_vsx_st (vector bool int, int, vector bool int *);
17852 void vec_vsx_st (vector bool int, int, unsigned int *);
17853 void vec_vsx_st (vector bool int, int, int *);
17854 void vec_vsx_st (vector signed short, int, vector signed short *);
17855 void vec_vsx_st (vector signed short, int, short *);
17856 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
17857 void vec_vsx_st (vector unsigned short, int, unsigned short *);
17858 void vec_vsx_st (vector bool short, int, vector bool short *);
17859 void vec_vsx_st (vector bool short, int, unsigned short *);
17860 void vec_vsx_st (vector pixel, int, vector pixel *);
17861 void vec_vsx_st (vector pixel, int, unsigned short *);
17862 void vec_vsx_st (vector pixel, int, short *);
17863 void vec_vsx_st (vector bool short, int, short *);
17864 void vec_vsx_st (vector signed char, int, vector signed char *);
17865 void vec_vsx_st (vector signed char, int, signed char *);
17866 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
17867 void vec_vsx_st (vector unsigned char, int, unsigned char *);
17868 void vec_vsx_st (vector bool char, int, vector bool char *);
17869 void vec_vsx_st (vector bool char, int, unsigned char *);
17870 void vec_vsx_st (vector bool char, int, signed char *);
17872 vector double vec_xxpermdi (vector double, vector double, const int);
17873 vector float vec_xxpermdi (vector float, vector float, const int);
17874 vector long long vec_xxpermdi (vector long long, vector long long, const int);
17875 vector unsigned long long vec_xxpermdi (vector unsigned long long,
17876 vector unsigned long long, const int);
17877 vector int vec_xxpermdi (vector int, vector int, const int);
17878 vector unsigned int vec_xxpermdi (vector unsigned int,
17879 vector unsigned int, const int);
17880 vector short vec_xxpermdi (vector short, vector short, const int);
17881 vector unsigned short vec_xxpermdi (vector unsigned short,
17882 vector unsigned short, const int);
17883 vector signed char vec_xxpermdi (vector signed char, vector signed char,
17885 vector unsigned char vec_xxpermdi (vector unsigned char,
17886 vector unsigned char, const int);
17888 vector double vec_xxsldi (vector double, vector double, int);
17889 vector float vec_xxsldi (vector float, vector float, int);
17890 vector long long vec_xxsldi (vector long long, vector long long, int);
17891 vector unsigned long long vec_xxsldi (vector unsigned long long,
17892 vector unsigned long long, int);
17893 vector int vec_xxsldi (vector int, vector int, int);
17894 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
17895 vector short vec_xxsldi (vector short, vector short, int);
17896 vector unsigned short vec_xxsldi (vector unsigned short,
17897 vector unsigned short, int);
17898 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
17899 vector unsigned char vec_xxsldi (vector unsigned char,
17900 vector unsigned char, int);
17903 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
17904 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
17905 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
17906 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
17907 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
17909 If the ISA 2.07 additions to the vector/scalar (power8-vector)
17910 instruction set are available, the following additional functions are
17911 available for both 32-bit and 64-bit targets. For 64-bit targets, you
17912 can use @var{vector long} instead of @var{vector long long},
17913 @var{vector bool long} instead of @var{vector bool long long}, and
17914 @var{vector unsigned long} instead of @var{vector unsigned long long}.
17917 vector long long vec_abs (vector long long);
17919 vector long long vec_add (vector long long, vector long long);
17920 vector unsigned long long vec_add (vector unsigned long long,
17921 vector unsigned long long);
17923 int vec_all_eq (vector long long, vector long long);
17924 int vec_all_eq (vector unsigned long long, vector unsigned long long);
17925 int vec_all_ge (vector long long, vector long long);
17926 int vec_all_ge (vector unsigned long long, vector unsigned long long);
17927 int vec_all_gt (vector long long, vector long long);
17928 int vec_all_gt (vector unsigned long long, vector unsigned long long);
17929 int vec_all_le (vector long long, vector long long);
17930 int vec_all_le (vector unsigned long long, vector unsigned long long);
17931 int vec_all_lt (vector long long, vector long long);
17932 int vec_all_lt (vector unsigned long long, vector unsigned long long);
17933 int vec_all_ne (vector long long, vector long long);
17934 int vec_all_ne (vector unsigned long long, vector unsigned long long);
17936 int vec_any_eq (vector long long, vector long long);
17937 int vec_any_eq (vector unsigned long long, vector unsigned long long);
17938 int vec_any_ge (vector long long, vector long long);
17939 int vec_any_ge (vector unsigned long long, vector unsigned long long);
17940 int vec_any_gt (vector long long, vector long long);
17941 int vec_any_gt (vector unsigned long long, vector unsigned long long);
17942 int vec_any_le (vector long long, vector long long);
17943 int vec_any_le (vector unsigned long long, vector unsigned long long);
17944 int vec_any_lt (vector long long, vector long long);
17945 int vec_any_lt (vector unsigned long long, vector unsigned long long);
17946 int vec_any_ne (vector long long, vector long long);
17947 int vec_any_ne (vector unsigned long long, vector unsigned long long);
17949 vector bool long long vec_cmpeq (vector bool long long, vector bool long long);
17951 vector long long vec_eqv (vector long long, vector long long);
17952 vector long long vec_eqv (vector bool long long, vector long long);
17953 vector long long vec_eqv (vector long long, vector bool long long);
17954 vector unsigned long long vec_eqv (vector unsigned long long,
17955 vector unsigned long long);
17956 vector unsigned long long vec_eqv (vector bool long long,
17957 vector unsigned long long);
17958 vector unsigned long long vec_eqv (vector unsigned long long,
17959 vector bool long long);
17960 vector int vec_eqv (vector int, vector int);
17961 vector int vec_eqv (vector bool int, vector int);
17962 vector int vec_eqv (vector int, vector bool int);
17963 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
17964 vector unsigned int vec_eqv (vector bool unsigned int,
17965 vector unsigned int);
17966 vector unsigned int vec_eqv (vector unsigned int,
17967 vector bool unsigned int);
17968 vector short vec_eqv (vector short, vector short);
17969 vector short vec_eqv (vector bool short, vector short);
17970 vector short vec_eqv (vector short, vector bool short);
17971 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
17972 vector unsigned short vec_eqv (vector bool unsigned short,
17973 vector unsigned short);
17974 vector unsigned short vec_eqv (vector unsigned short,
17975 vector bool unsigned short);
17976 vector signed char vec_eqv (vector signed char, vector signed char);
17977 vector signed char vec_eqv (vector bool signed char, vector signed char);
17978 vector signed char vec_eqv (vector signed char, vector bool signed char);
17979 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
17980 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
17981 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
17983 vector long long vec_max (vector long long, vector long long);
17984 vector unsigned long long vec_max (vector unsigned long long,
17985 vector unsigned long long);
17987 vector signed int vec_mergee (vector signed int, vector signed int);
17988 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
17989 vector bool int vec_mergee (vector bool int, vector bool int);
17991 vector signed int vec_mergeo (vector signed int, vector signed int);
17992 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
17993 vector bool int vec_mergeo (vector bool int, vector bool int);
17995 vector long long vec_min (vector long long, vector long long);
17996 vector unsigned long long vec_min (vector unsigned long long,
17997 vector unsigned long long);
17999 vector signed long long vec_nabs (vector signed long long);
18001 vector long long vec_nand (vector long long, vector long long);
18002 vector long long vec_nand (vector bool long long, vector long long);
18003 vector long long vec_nand (vector long long, vector bool long long);
18004 vector unsigned long long vec_nand (vector unsigned long long,
18005 vector unsigned long long);
18006 vector unsigned long long vec_nand (vector bool long long,
18007 vector unsigned long long);
18008 vector unsigned long long vec_nand (vector unsigned long long,
18009 vector bool long long);
18010 vector int vec_nand (vector int, vector int);
18011 vector int vec_nand (vector bool int, vector int);
18012 vector int vec_nand (vector int, vector bool int);
18013 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
18014 vector unsigned int vec_nand (vector bool unsigned int,
18015 vector unsigned int);
18016 vector unsigned int vec_nand (vector unsigned int,
18017 vector bool unsigned int);
18018 vector short vec_nand (vector short, vector short);
18019 vector short vec_nand (vector bool short, vector short);
18020 vector short vec_nand (vector short, vector bool short);
18021 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
18022 vector unsigned short vec_nand (vector bool unsigned short,
18023 vector unsigned short);
18024 vector unsigned short vec_nand (vector unsigned short,
18025 vector bool unsigned short);
18026 vector signed char vec_nand (vector signed char, vector signed char);
18027 vector signed char vec_nand (vector bool signed char, vector signed char);
18028 vector signed char vec_nand (vector signed char, vector bool signed char);
18029 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
18030 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
18031 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
18033 vector long long vec_orc (vector long long, vector long long);
18034 vector long long vec_orc (vector bool long long, vector long long);
18035 vector long long vec_orc (vector long long, vector bool long long);
18036 vector unsigned long long vec_orc (vector unsigned long long,
18037 vector unsigned long long);
18038 vector unsigned long long vec_orc (vector bool long long,
18039 vector unsigned long long);
18040 vector unsigned long long vec_orc (vector unsigned long long,
18041 vector bool long long);
18042 vector int vec_orc (vector int, vector int);
18043 vector int vec_orc (vector bool int, vector int);
18044 vector int vec_orc (vector int, vector bool int);
18045 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
18046 vector unsigned int vec_orc (vector bool unsigned int,
18047 vector unsigned int);
18048 vector unsigned int vec_orc (vector unsigned int,
18049 vector bool unsigned int);
18050 vector short vec_orc (vector short, vector short);
18051 vector short vec_orc (vector bool short, vector short);
18052 vector short vec_orc (vector short, vector bool short);
18053 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
18054 vector unsigned short vec_orc (vector bool unsigned short,
18055 vector unsigned short);
18056 vector unsigned short vec_orc (vector unsigned short,
18057 vector bool unsigned short);
18058 vector signed char vec_orc (vector signed char, vector signed char);
18059 vector signed char vec_orc (vector bool signed char, vector signed char);
18060 vector signed char vec_orc (vector signed char, vector bool signed char);
18061 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
18062 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
18063 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
18065 vector int vec_pack (vector long long, vector long long);
18066 vector unsigned int vec_pack (vector unsigned long long,
18067 vector unsigned long long);
18068 vector bool int vec_pack (vector bool long long, vector bool long long);
18069 vector float vec_pack (vector double, vector double);
18071 vector int vec_packs (vector long long, vector long long);
18072 vector unsigned int vec_packs (vector unsigned long long,
18073 vector unsigned long long);
18075 vector unsigned int vec_packsu (vector long long, vector long long);
18076 vector unsigned int vec_packsu (vector unsigned long long,
18077 vector unsigned long long);
18079 vector unsigned char vec_popcnt (vector signed char);
18080 vector unsigned char vec_popcnt (vector unsigned char);
18081 vector unsigned short vec_popcnt (vector signed short);
18082 vector unsigned short vec_popcnt (vector unsigned short);
18083 vector unsigned int vec_popcnt (vector signed int);
18084 vector unsigned int vec_popcnt (vector unsigned int);
18085 vector unsigned long long vec_popcnt (vector signed long long);
18086 vector unsigned long long vec_popcnt (vector unsigned long long);
18088 vector long long vec_rl (vector long long,
18089 vector unsigned long long);
18090 vector long long vec_rl (vector unsigned long long,
18091 vector unsigned long long);
18093 vector long long vec_sl (vector long long, vector unsigned long long);
18094 vector long long vec_sl (vector unsigned long long,
18095 vector unsigned long long);
18097 vector long long vec_sr (vector long long, vector unsigned long long);
18098 vector unsigned long long char vec_sr (vector unsigned long long,
18099 vector unsigned long long);
18101 vector long long vec_sra (vector long long, vector unsigned long long);
18102 vector unsigned long long vec_sra (vector unsigned long long,
18103 vector unsigned long long);
18105 vector long long vec_sub (vector long long, vector long long);
18106 vector unsigned long long vec_sub (vector unsigned long long,
18107 vector unsigned long long);
18109 vector long long vec_unpackh (vector int);
18110 vector unsigned long long vec_unpackh (vector unsigned int);
18112 vector long long vec_unpackl (vector int);
18113 vector unsigned long long vec_unpackl (vector unsigned int);
18115 vector long long vec_vaddudm (vector long long, vector long long);
18116 vector long long vec_vaddudm (vector bool long long, vector long long);
18117 vector long long vec_vaddudm (vector long long, vector bool long long);
18118 vector unsigned long long vec_vaddudm (vector unsigned long long,
18119 vector unsigned long long);
18120 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
18121 vector unsigned long long);
18122 vector unsigned long long vec_vaddudm (vector unsigned long long,
18123 vector bool unsigned long long);
18125 vector long long vec_vbpermq (vector signed char, vector signed char);
18126 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
18128 vector unsigned char vec_bperm (vector unsigned char, vector unsigned char);
18129 vector unsigned char vec_bperm (vector unsigned long long,
18130 vector unsigned char);
18131 vector unsigned long long vec_bperm (vector unsigned __int128,
18132 vector unsigned char);
18134 vector long long vec_cntlz (vector long long);
18135 vector unsigned long long vec_cntlz (vector unsigned long long);
18136 vector int vec_cntlz (vector int);
18137 vector unsigned int vec_cntlz (vector int);
18138 vector short vec_cntlz (vector short);
18139 vector unsigned short vec_cntlz (vector unsigned short);
18140 vector signed char vec_cntlz (vector signed char);
18141 vector unsigned char vec_cntlz (vector unsigned char);
18143 vector long long vec_vclz (vector long long);
18144 vector unsigned long long vec_vclz (vector unsigned long long);
18145 vector int vec_vclz (vector int);
18146 vector unsigned int vec_vclz (vector int);
18147 vector short vec_vclz (vector short);
18148 vector unsigned short vec_vclz (vector unsigned short);
18149 vector signed char vec_vclz (vector signed char);
18150 vector unsigned char vec_vclz (vector unsigned char);
18152 vector signed char vec_vclzb (vector signed char);
18153 vector unsigned char vec_vclzb (vector unsigned char);
18155 vector long long vec_vclzd (vector long long);
18156 vector unsigned long long vec_vclzd (vector unsigned long long);
18158 vector short vec_vclzh (vector short);
18159 vector unsigned short vec_vclzh (vector unsigned short);
18161 vector int vec_vclzw (vector int);
18162 vector unsigned int vec_vclzw (vector int);
18164 vector signed char vec_vgbbd (vector signed char);
18165 vector unsigned char vec_vgbbd (vector unsigned char);
18167 vector long long vec_vmaxsd (vector long long, vector long long);
18169 vector unsigned long long vec_vmaxud (vector unsigned long long,
18170 unsigned vector long long);
18172 vector long long vec_vminsd (vector long long, vector long long);
18174 vector unsigned long long vec_vminud (vector long long,
18177 vector int vec_vpksdss (vector long long, vector long long);
18178 vector unsigned int vec_vpksdss (vector long long, vector long long);
18180 vector unsigned int vec_vpkudus (vector unsigned long long,
18181 vector unsigned long long);
18183 vector int vec_vpkudum (vector long long, vector long long);
18184 vector unsigned int vec_vpkudum (vector unsigned long long,
18185 vector unsigned long long);
18186 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
18188 vector long long vec_vpopcnt (vector long long);
18189 vector unsigned long long vec_vpopcnt (vector unsigned long long);
18190 vector int vec_vpopcnt (vector int);
18191 vector unsigned int vec_vpopcnt (vector int);
18192 vector short vec_vpopcnt (vector short);
18193 vector unsigned short vec_vpopcnt (vector unsigned short);
18194 vector signed char vec_vpopcnt (vector signed char);
18195 vector unsigned char vec_vpopcnt (vector unsigned char);
18197 vector signed char vec_vpopcntb (vector signed char);
18198 vector unsigned char vec_vpopcntb (vector unsigned char);
18200 vector long long vec_vpopcntd (vector long long);
18201 vector unsigned long long vec_vpopcntd (vector unsigned long long);
18203 vector short vec_vpopcnth (vector short);
18204 vector unsigned short vec_vpopcnth (vector unsigned short);
18206 vector int vec_vpopcntw (vector int);
18207 vector unsigned int vec_vpopcntw (vector int);
18209 vector long long vec_vrld (vector long long, vector unsigned long long);
18210 vector unsigned long long vec_vrld (vector unsigned long long,
18211 vector unsigned long long);
18213 vector long long vec_vsld (vector long long, vector unsigned long long);
18214 vector long long vec_vsld (vector unsigned long long,
18215 vector unsigned long long);
18217 vector long long vec_vsrad (vector long long, vector unsigned long long);
18218 vector unsigned long long vec_vsrad (vector unsigned long long,
18219 vector unsigned long long);
18221 vector long long vec_vsrd (vector long long, vector unsigned long long);
18222 vector unsigned long long char vec_vsrd (vector unsigned long long,
18223 vector unsigned long long);
18225 vector long long vec_vsubudm (vector long long, vector long long);
18226 vector long long vec_vsubudm (vector bool long long, vector long long);
18227 vector long long vec_vsubudm (vector long long, vector bool long long);
18228 vector unsigned long long vec_vsubudm (vector unsigned long long,
18229 vector unsigned long long);
18230 vector unsigned long long vec_vsubudm (vector bool long long,
18231 vector unsigned long long);
18232 vector unsigned long long vec_vsubudm (vector unsigned long long,
18233 vector bool long long);
18235 vector long long vec_vupkhsw (vector int);
18236 vector unsigned long long vec_vupkhsw (vector unsigned int);
18238 vector long long vec_vupklsw (vector int);
18239 vector unsigned long long vec_vupklsw (vector int);
18242 If the ISA 2.07 additions to the vector/scalar (power8-vector)
18243 instruction set are available, the following additional functions are
18244 available for 64-bit targets. New vector types
18245 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
18246 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
18249 The normal vector extract, and set operations work on
18250 @var{vector __int128_t} and @var{vector __uint128_t} types,
18251 but the index value must be 0.
18254 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
18255 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
18257 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
18258 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
18260 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
18261 vector __int128_t);
18262 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
18263 vector __uint128_t);
18265 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
18266 vector __int128_t);
18267 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
18268 vector __uint128_t);
18270 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
18271 vector __int128_t);
18272 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
18273 vector __uint128_t);
18275 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
18276 vector __int128_t);
18277 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
18278 vector __uint128_t);
18280 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
18281 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
18283 __int128_t vec_vsubuqm (__int128_t, __int128_t);
18284 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
18286 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
18287 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
18288 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
18289 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
18290 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
18291 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
18292 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
18293 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
18294 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
18295 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
18298 If the ISA 3.0 instruction set additions (@option{-mcpu=power9})
18302 vector unsigned long long vec_bperm (vector unsigned long long,
18303 vector unsigned char);
18305 vector bool char vec_cmpne (vector bool char, vector bool char);
18306 vector bool short vec_cmpne (vector bool short, vector bool short);
18307 vector bool int vec_cmpne (vector bool int, vector bool int);
18308 vector bool long long vec_cmpne (vector bool long long, vector bool long long);
18310 vector long long vec_vctz (vector long long);
18311 vector unsigned long long vec_vctz (vector unsigned long long);
18312 vector int vec_vctz (vector int);
18313 vector unsigned int vec_vctz (vector int);
18314 vector short vec_vctz (vector short);
18315 vector unsigned short vec_vctz (vector unsigned short);
18316 vector signed char vec_vctz (vector signed char);
18317 vector unsigned char vec_vctz (vector unsigned char);
18319 vector signed char vec_vctzb (vector signed char);
18320 vector unsigned char vec_vctzb (vector unsigned char);
18322 vector long long vec_vctzd (vector long long);
18323 vector unsigned long long vec_vctzd (vector unsigned long long);
18325 vector short vec_vctzh (vector short);
18326 vector unsigned short vec_vctzh (vector unsigned short);
18328 vector int vec_vctzw (vector int);
18329 vector unsigned int vec_vctzw (vector int);
18331 long long vec_vextract4b (const vector signed char, const int);
18332 long long vec_vextract4b (const vector unsigned char, const int);
18334 vector signed char vec_insert4b (vector int, vector signed char, const int);
18335 vector unsigned char vec_insert4b (vector unsigned int, vector unsigned char,
18337 vector signed char vec_insert4b (long long, vector signed char, const int);
18338 vector unsigned char vec_insert4b (long long, vector unsigned char, const int);
18340 vector unsigned int vec_parity_lsbb (vector signed int);
18341 vector unsigned int vec_parity_lsbb (vector unsigned int);
18342 vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
18343 vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
18344 vector unsigned long long vec_parity_lsbb (vector signed long long);
18345 vector unsigned long long vec_parity_lsbb (vector unsigned long long);
18347 vector int vec_vprtyb (vector int);
18348 vector unsigned int vec_vprtyb (vector unsigned int);
18349 vector long long vec_vprtyb (vector long long);
18350 vector unsigned long long vec_vprtyb (vector unsigned long long);
18352 vector int vec_vprtybw (vector int);
18353 vector unsigned int vec_vprtybw (vector unsigned int);
18355 vector long long vec_vprtybd (vector long long);
18356 vector unsigned long long vec_vprtybd (vector unsigned long long);
18359 On 64-bit targets, if the ISA 3.0 additions (@option{-mcpu=power9})
18363 vector long vec_vprtyb (vector long);
18364 vector unsigned long vec_vprtyb (vector unsigned long);
18365 vector __int128_t vec_vprtyb (vector __int128_t);
18366 vector __uint128_t vec_vprtyb (vector __uint128_t);
18368 vector long vec_vprtybd (vector long);
18369 vector unsigned long vec_vprtybd (vector unsigned long);
18371 vector __int128_t vec_vprtybq (vector __int128_t);
18372 vector __uint128_t vec_vprtybd (vector __uint128_t);
18375 The following built-in vector functions are available for the PowerPC family
18376 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
18378 __vector unsigned char
18379 vec_slv (__vector unsigned char src, __vector unsigned char shift_distance);
18380 __vector unsigned char
18381 vec_srv (__vector unsigned char src, __vector unsigned char shift_distance);
18384 The @code{vec_slv} and @code{vec_srv} functions operate on
18385 all of the bytes of their @code{src} and @code{shift_distance}
18386 arguments in parallel. The behavior of the @code{vec_slv} is as if
18387 there existed a temporary array of 17 unsigned characters
18388 @code{slv_array} within which elements 0 through 15 are the same as
18389 the entries in the @code{src} array and element 16 equals 0. The
18390 result returned from the @code{vec_slv} function is a
18391 @code{__vector} of 16 unsigned characters within which element
18392 @code{i} is computed using the C expression
18393 @code{0xff & (*((unsigned short *)(slv_array + i)) << (0x07 &
18394 shift_distance[i]))},
18395 with this resulting value coerced to the @code{unsigned char} type.
18396 The behavior of the @code{vec_srv} is as if
18397 there existed a temporary array of 17 unsigned characters
18398 @code{srv_array} within which element 0 equals zero and
18399 elements 1 through 16 equal the elements 0 through 15 of
18400 the @code{src} array. The
18401 result returned from the @code{vec_srv} function is a
18402 @code{__vector} of 16 unsigned characters within which element
18403 @code{i} is computed using the C expression
18404 @code{0xff & (*((unsigned short *)(srv_array + i)) >>
18405 (0x07 & shift_distance[i]))},
18406 with this resulting value coerced to the @code{unsigned char} type.
18408 The following built-in functions are available for the PowerPC family
18409 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
18411 __vector unsigned char
18412 vec_absd (__vector unsigned char arg1, __vector unsigned char arg2);
18413 __vector unsigned short
18414 vec_absd (__vector unsigned short arg1, __vector unsigned short arg2);
18415 __vector unsigned int
18416 vec_absd (__vector unsigned int arg1, __vector unsigned int arg2);
18418 __vector unsigned char
18419 vec_absdb (__vector unsigned char arg1, __vector unsigned char arg2);
18420 __vector unsigned short
18421 vec_absdh (__vector unsigned short arg1, __vector unsigned short arg2);
18422 __vector unsigned int
18423 vec_absdw (__vector unsigned int arg1, __vector unsigned int arg2);
18426 The @code{vec_absd}, @code{vec_absdb}, @code{vec_absdh}, and
18427 @code{vec_absdw} built-in functions each computes the absolute
18428 differences of the pairs of vector elements supplied in its two vector
18429 arguments, placing the absolute differences into the corresponding
18430 elements of the vector result.
18432 The following built-in functions are available for the PowerPC family
18433 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
18435 __vector unsigned int
18436 vec_extract_exp (__vector float source);
18437 __vector unsigned long long int
18438 vec_extract_exp (__vector double source);
18440 __vector unsigned int
18441 vec_extract_sig (__vector float source);
18442 __vector unsigned long long int
18443 vec_extract_sig (__vector double source);
18446 vec_insert_exp (__vector unsigned int significands,
18447 __vector unsigned int exponents);
18449 vec_insert_exp (__vector unsigned float significands,
18450 __vector unsigned int exponents);
18452 vec_insert_exp (__vector unsigned long long int significands,
18453 __vector unsigned long long int exponents);
18455 vec_insert_exp (__vector unsigned double significands,
18456 __vector unsigned long long int exponents);
18458 __vector bool int vec_test_data_class (__vector float source,
18459 const int condition);
18460 __vector bool long long int vec_test_data_class (__vector double source,
18461 const int condition);
18464 The @code{vec_extract_sig} and @code{vec_extract_exp} built-in
18465 functions return vectors representing the significands and biased
18466 exponent values of their @code{source} arguments respectively.
18467 Within the result vector returned by @code{vec_extract_sig}, the
18468 @code{0x800000} bit of each vector element returned when the
18469 function's @code{source} argument is of type @code{float} is set to 1
18470 if the corresponding floating point value is in normalized form.
18471 Otherwise, this bit is set to 0. When the @code{source} argument is
18472 of type @code{double}, the @code{0x10000000000000} bit within each of
18473 the result vector's elements is set according to the same rules.
18474 Note that the sign of the significand is not represented in the result
18475 returned from the @code{vec_extract_sig} function. To extract the
18477 @code{vec_cpsgn} function, which returns a new vector within which all
18478 of the sign bits of its second argument vector are overwritten with the
18479 sign bits copied from the coresponding elements of its first argument
18480 vector, and all other (non-sign) bits of the second argument vector
18481 are copied unchanged into the result vector.
18483 The @code{vec_insert_exp} built-in functions return a vector of
18484 single- or double-precision floating
18485 point values constructed by assembling the values of their
18486 @code{significands} and @code{exponents} arguments into the
18487 corresponding elements of the returned vector.
18489 element of the result is copied from the most significant bit of the
18490 corresponding entry within the @code{significands} argument.
18491 Note that the relevant
18492 bits of the @code{significands} argument are the same, for both integer
18493 and floating point types.
18495 significand and exponent components of each element of the result are
18496 composed of the least significant bits of the corresponding
18497 @code{significands} element and the least significant bits of the
18498 corresponding @code{exponents} element.
18500 The @code{vec_test_data_class} built-in function returns a vector
18501 representing the results of testing the @code{source} vector for the
18502 condition selected by the @code{condition} argument. The
18503 @code{condition} argument must be a compile-time constant integer with
18504 value not exceeding 127. The
18505 @code{condition} argument is encoded as a bitmask with each bit
18506 enabling the testing of a different condition, as characterized by the
18510 0x20 Test for +Infinity
18511 0x10 Test for -Infinity
18512 0x08 Test for +Zero
18513 0x04 Test for -Zero
18514 0x02 Test for +Denormal
18515 0x01 Test for -Denormal
18518 If any of the enabled test conditions is true, the corresponding entry
18519 in the result vector is -1. Otherwise (all of the enabled test
18520 conditions are false), the corresponding entry of the result vector is 0.
18522 The following built-in functions are available for the PowerPC family
18523 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
18525 vector unsigned int vec_rlmi (vector unsigned int, vector unsigned int,
18526 vector unsigned int);
18527 vector unsigned long long vec_rlmi (vector unsigned long long,
18528 vector unsigned long long,
18529 vector unsigned long long);
18530 vector unsigned int vec_rlnm (vector unsigned int, vector unsigned int,
18531 vector unsigned int);
18532 vector unsigned long long vec_rlnm (vector unsigned long long,
18533 vector unsigned long long,
18534 vector unsigned long long);
18535 vector unsigned int vec_vrlnm (vector unsigned int, vector unsigned int);
18536 vector unsigned long long vec_vrlnm (vector unsigned long long,
18537 vector unsigned long long);
18540 The result of @code{vec_rlmi} is obtained by rotating each element of
18541 the first argument vector left and inserting it under mask into the
18542 second argument vector. The third argument vector contains the mask
18543 beginning in bits 11:15, the mask end in bits 19:23, and the shift
18544 count in bits 27:31, of each element.
18546 The result of @code{vec_rlnm} is obtained by rotating each element of
18547 the first argument vector left and ANDing it with a mask specified by
18548 the second and third argument vectors. The second argument vector
18549 contains the shift count for each element in the low-order byte. The
18550 third argument vector contains the mask end for each element in the
18551 low-order byte, with the mask begin in the next higher byte.
18553 The result of @code{vec_vrlnm} is obtained by rotating each element
18554 of the first argument vector left and ANDing it with a mask. The
18555 second argument vector contains the mask beginning in bits 11:15,
18556 the mask end in bits 19:23, and the shift count in bits 27:31,
18559 If the ISA 3.0 instruction set additions (@option{-mcpu=power9})
18562 vector signed bool char vec_revb (vector signed char);
18563 vector signed char vec_revb (vector signed char);
18564 vector unsigned char vec_revb (vector unsigned char);
18565 vector bool short vec_revb (vector bool short);
18566 vector short vec_revb (vector short);
18567 vector unsigned short vec_revb (vector unsigned short);
18568 vector bool int vec_revb (vector bool int);
18569 vector int vec_revb (vector int);
18570 vector unsigned int vec_revb (vector unsigned int);
18571 vector float vec_revb (vector float);
18572 vector bool long long vec_revb (vector bool long long);
18573 vector long long vec_revb (vector long long);
18574 vector unsigned long long vec_revb (vector unsigned long long);
18575 vector double vec_revb (vector double);
18578 On 64-bit targets, if the ISA 3.0 additions (@option{-mcpu=power9})
18581 vector long vec_revb (vector long);
18582 vector unsigned long vec_revb (vector unsigned long);
18583 vector __int128_t vec_revb (vector __int128_t);
18584 vector __uint128_t vec_revb (vector __uint128_t);
18587 The @code{vec_revb} built-in function reverses the bytes on an element
18588 by element basis. A vector of @code{vector unsigned char} or
18589 @code{vector signed char} reverses the bytes in the whole word.
18591 If the cryptographic instructions are enabled (@option{-mcrypto} or
18592 @option{-mcpu=power8}), the following builtins are enabled.
18595 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
18597 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
18598 vector unsigned long long);
18600 vector unsigned long long __builtin_crypto_vcipherlast
18601 (vector unsigned long long,
18602 vector unsigned long long);
18604 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
18605 vector unsigned long long);
18607 vector unsigned long long __builtin_crypto_vncipherlast
18608 (vector unsigned long long,
18609 vector unsigned long long);
18611 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
18612 vector unsigned char,
18613 vector unsigned char);
18615 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
18616 vector unsigned short,
18617 vector unsigned short);
18619 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
18620 vector unsigned int,
18621 vector unsigned int);
18623 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
18624 vector unsigned long long,
18625 vector unsigned long long);
18627 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
18628 vector unsigned char);
18630 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
18631 vector unsigned short);
18633 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
18634 vector unsigned int);
18636 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
18637 vector unsigned long long);
18639 vector unsigned long long __builtin_crypto_vshasigmad
18640 (vector unsigned long long, int, int);
18642 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
18646 The second argument to @var{__builtin_crypto_vshasigmad} and
18647 @var{__builtin_crypto_vshasigmaw} must be a constant
18648 integer that is 0 or 1. The third argument to these built-in functions
18649 must be a constant integer in the range of 0 to 15.
18651 If the ISA 3.0 instruction set additions
18652 are enabled (@option{-mcpu=power9}), the following additional
18653 functions are available for both 32-bit and 64-bit targets.
18655 vector short vec_xl (int, vector short *);
18656 vector short vec_xl (int, short *);
18657 vector unsigned short vec_xl (int, vector unsigned short *);
18658 vector unsigned short vec_xl (int, unsigned short *);
18659 vector char vec_xl (int, vector char *);
18660 vector char vec_xl (int, char *);
18661 vector unsigned char vec_xl (int, vector unsigned char *);
18662 vector unsigned char vec_xl (int, unsigned char *);
18664 void vec_xst (vector short, int, vector short *);
18665 void vec_xst (vector short, int, short *);
18666 void vec_xst (vector unsigned short, int, vector unsigned short *);
18667 void vec_xst (vector unsigned short, int, unsigned short *);
18668 void vec_xst (vector char, int, vector char *);
18669 void vec_xst (vector char, int, char *);
18670 void vec_xst (vector unsigned char, int, vector unsigned char *);
18671 void vec_xst (vector unsigned char, int, unsigned char *);
18673 @node PowerPC Hardware Transactional Memory Built-in Functions
18674 @subsection PowerPC Hardware Transactional Memory Built-in Functions
18675 GCC provides two interfaces for accessing the Hardware Transactional
18676 Memory (HTM) instructions available on some of the PowerPC family
18677 of processors (eg, POWER8). The two interfaces come in a low level
18678 interface, consisting of built-in functions specific to PowerPC and a
18679 higher level interface consisting of inline functions that are common
18680 between PowerPC and S/390.
18682 @subsubsection PowerPC HTM Low Level Built-in Functions
18684 The following low level built-in functions are available with
18685 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
18686 They all generate the machine instruction that is part of the name.
18688 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
18689 the full 4-bit condition register value set by their associated hardware
18690 instruction. The header file @code{htmintrin.h} defines some macros that can
18691 be used to decipher the return value. The @code{__builtin_tbegin} builtin
18692 returns a simple true or false value depending on whether a transaction was
18693 successfully started or not. The arguments of the builtins match exactly the
18694 type and order of the associated hardware instruction's operands, except for
18695 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
18696 Refer to the ISA manual for a description of each instruction's operands.
18699 unsigned int __builtin_tbegin (unsigned int)
18700 unsigned int __builtin_tend (unsigned int)
18702 unsigned int __builtin_tabort (unsigned int)
18703 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
18704 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
18705 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
18706 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
18708 unsigned int __builtin_tcheck (void)
18709 unsigned int __builtin_treclaim (unsigned int)
18710 unsigned int __builtin_trechkpt (void)
18711 unsigned int __builtin_tsr (unsigned int)
18714 In addition to the above HTM built-ins, we have added built-ins for
18715 some common extended mnemonics of the HTM instructions:
18718 unsigned int __builtin_tendall (void)
18719 unsigned int __builtin_tresume (void)
18720 unsigned int __builtin_tsuspend (void)
18723 Note that the semantics of the above HTM builtins are required to mimic
18724 the locking semantics used for critical sections. Builtins that are used
18725 to create a new transaction or restart a suspended transaction must have
18726 lock acquisition like semantics while those builtins that end or suspend a
18727 transaction must have lock release like semantics. Specifically, this must
18728 mimic lock semantics as specified by C++11, for example: Lock acquisition is
18729 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
18730 that returns 0, and lock release is as-if an execution of
18731 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
18732 implicit implementation-defined lock used for all transactions. The HTM
18733 instructions associated with with the builtins inherently provide the
18734 correct acquisition and release hardware barriers required. However,
18735 the compiler must also be prohibited from moving loads and stores across
18736 the builtins in a way that would violate their semantics. This has been
18737 accomplished by adding memory barriers to the associated HTM instructions
18738 (which is a conservative approach to provide acquire and release semantics).
18739 Earlier versions of the compiler did not treat the HTM instructions as
18740 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
18741 be used to determine whether the current compiler treats HTM instructions
18742 as memory barriers or not. This allows the user to explicitly add memory
18743 barriers to their code when using an older version of the compiler.
18745 The following set of built-in functions are available to gain access
18746 to the HTM specific special purpose registers.
18749 unsigned long __builtin_get_texasr (void)
18750 unsigned long __builtin_get_texasru (void)
18751 unsigned long __builtin_get_tfhar (void)
18752 unsigned long __builtin_get_tfiar (void)
18754 void __builtin_set_texasr (unsigned long);
18755 void __builtin_set_texasru (unsigned long);
18756 void __builtin_set_tfhar (unsigned long);
18757 void __builtin_set_tfiar (unsigned long);
18760 Example usage of these low level built-in functions may look like:
18763 #include <htmintrin.h>
18765 int num_retries = 10;
18769 if (__builtin_tbegin (0))
18771 /* Transaction State Initiated. */
18772 if (is_locked (lock))
18773 __builtin_tabort (0);
18774 ... transaction code...
18775 __builtin_tend (0);
18780 /* Transaction State Failed. Use locks if the transaction
18781 failure is "persistent" or we've tried too many times. */
18782 if (num_retries-- <= 0
18783 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
18785 acquire_lock (lock);
18786 ... non transactional fallback path...
18787 release_lock (lock);
18794 One final built-in function has been added that returns the value of
18795 the 2-bit Transaction State field of the Machine Status Register (MSR)
18796 as stored in @code{CR0}.
18799 unsigned long __builtin_ttest (void)
18802 This built-in can be used to determine the current transaction state
18803 using the following code example:
18806 #include <htmintrin.h>
18808 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
18810 if (tx_state == _HTM_TRANSACTIONAL)
18812 /* Code to use in transactional state. */
18814 else if (tx_state == _HTM_NONTRANSACTIONAL)
18816 /* Code to use in non-transactional state. */
18818 else if (tx_state == _HTM_SUSPENDED)
18820 /* Code to use in transaction suspended state. */
18824 @subsubsection PowerPC HTM High Level Inline Functions
18826 The following high level HTM interface is made available by including
18827 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
18828 where CPU is `power8' or later. This interface is common between PowerPC
18829 and S/390, allowing users to write one HTM source implementation that
18830 can be compiled and executed on either system.
18833 long __TM_simple_begin (void)
18834 long __TM_begin (void* const TM_buff)
18835 long __TM_end (void)
18836 void __TM_abort (void)
18837 void __TM_named_abort (unsigned char const code)
18838 void __TM_resume (void)
18839 void __TM_suspend (void)
18841 long __TM_is_user_abort (void* const TM_buff)
18842 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
18843 long __TM_is_illegal (void* const TM_buff)
18844 long __TM_is_footprint_exceeded (void* const TM_buff)
18845 long __TM_nesting_depth (void* const TM_buff)
18846 long __TM_is_nested_too_deep(void* const TM_buff)
18847 long __TM_is_conflict(void* const TM_buff)
18848 long __TM_is_failure_persistent(void* const TM_buff)
18849 long __TM_failure_address(void* const TM_buff)
18850 long long __TM_failure_code(void* const TM_buff)
18853 Using these common set of HTM inline functions, we can create
18854 a more portable version of the HTM example in the previous
18855 section that will work on either PowerPC or S/390:
18858 #include <htmxlintrin.h>
18860 int num_retries = 10;
18861 TM_buff_type TM_buff;
18865 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
18867 /* Transaction State Initiated. */
18868 if (is_locked (lock))
18870 ... transaction code...
18876 /* Transaction State Failed. Use locks if the transaction
18877 failure is "persistent" or we've tried too many times. */
18878 if (num_retries-- <= 0
18879 || __TM_is_failure_persistent (TM_buff))
18881 acquire_lock (lock);
18882 ... non transactional fallback path...
18883 release_lock (lock);
18890 @node RX Built-in Functions
18891 @subsection RX Built-in Functions
18892 GCC supports some of the RX instructions which cannot be expressed in
18893 the C programming language via the use of built-in functions. The
18894 following functions are supported:
18896 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
18897 Generates the @code{brk} machine instruction.
18900 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
18901 Generates the @code{clrpsw} machine instruction to clear the specified
18902 bit in the processor status word.
18905 @deftypefn {Built-in Function} void __builtin_rx_int (int)
18906 Generates the @code{int} machine instruction to generate an interrupt
18907 with the specified value.
18910 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
18911 Generates the @code{machi} machine instruction to add the result of
18912 multiplying the top 16 bits of the two arguments into the
18916 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
18917 Generates the @code{maclo} machine instruction to add the result of
18918 multiplying the bottom 16 bits of the two arguments into the
18922 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
18923 Generates the @code{mulhi} machine instruction to place the result of
18924 multiplying the top 16 bits of the two arguments into the
18928 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
18929 Generates the @code{mullo} machine instruction to place the result of
18930 multiplying the bottom 16 bits of the two arguments into the
18934 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
18935 Generates the @code{mvfachi} machine instruction to read the top
18936 32 bits of the accumulator.
18939 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
18940 Generates the @code{mvfacmi} machine instruction to read the middle
18941 32 bits of the accumulator.
18944 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
18945 Generates the @code{mvfc} machine instruction which reads the control
18946 register specified in its argument and returns its value.
18949 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
18950 Generates the @code{mvtachi} machine instruction to set the top
18951 32 bits of the accumulator.
18954 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
18955 Generates the @code{mvtaclo} machine instruction to set the bottom
18956 32 bits of the accumulator.
18959 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
18960 Generates the @code{mvtc} machine instruction which sets control
18961 register number @code{reg} to @code{val}.
18964 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
18965 Generates the @code{mvtipl} machine instruction set the interrupt
18969 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
18970 Generates the @code{racw} machine instruction to round the accumulator
18971 according to the specified mode.
18974 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
18975 Generates the @code{revw} machine instruction which swaps the bytes in
18976 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
18977 and also bits 16--23 occupy bits 24--31 and vice versa.
18980 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
18981 Generates the @code{rmpa} machine instruction which initiates a
18982 repeated multiply and accumulate sequence.
18985 @deftypefn {Built-in Function} void __builtin_rx_round (float)
18986 Generates the @code{round} machine instruction which returns the
18987 floating-point argument rounded according to the current rounding mode
18988 set in the floating-point status word register.
18991 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
18992 Generates the @code{sat} machine instruction which returns the
18993 saturated value of the argument.
18996 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
18997 Generates the @code{setpsw} machine instruction to set the specified
18998 bit in the processor status word.
19001 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
19002 Generates the @code{wait} machine instruction.
19005 @node S/390 System z Built-in Functions
19006 @subsection S/390 System z Built-in Functions
19007 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
19008 Generates the @code{tbegin} machine instruction starting a
19009 non-constrained hardware transaction. If the parameter is non-NULL the
19010 memory area is used to store the transaction diagnostic buffer and
19011 will be passed as first operand to @code{tbegin}. This buffer can be
19012 defined using the @code{struct __htm_tdb} C struct defined in
19013 @code{htmintrin.h} and must reside on a double-word boundary. The
19014 second tbegin operand is set to @code{0xff0c}. This enables
19015 save/restore of all GPRs and disables aborts for FPR and AR
19016 manipulations inside the transaction body. The condition code set by
19017 the tbegin instruction is returned as integer value. The tbegin
19018 instruction by definition overwrites the content of all FPRs. The
19019 compiler will generate code which saves and restores the FPRs. For
19020 soft-float code it is recommended to used the @code{*_nofloat}
19021 variant. In order to prevent a TDB from being written it is required
19022 to pass a constant zero value as parameter. Passing a zero value
19023 through a variable is not sufficient. Although modifications of
19024 access registers inside the transaction will not trigger an
19025 transaction abort it is not supported to actually modify them. Access
19026 registers do not get saved when entering a transaction. They will have
19027 undefined state when reaching the abort code.
19030 Macros for the possible return codes of tbegin are defined in the
19031 @code{htmintrin.h} header file:
19034 @item _HTM_TBEGIN_STARTED
19035 @code{tbegin} has been executed as part of normal processing. The
19036 transaction body is supposed to be executed.
19037 @item _HTM_TBEGIN_INDETERMINATE
19038 The transaction was aborted due to an indeterminate condition which
19039 might be persistent.
19040 @item _HTM_TBEGIN_TRANSIENT
19041 The transaction aborted due to a transient failure. The transaction
19042 should be re-executed in that case.
19043 @item _HTM_TBEGIN_PERSISTENT
19044 The transaction aborted due to a persistent failure. Re-execution
19045 under same circumstances will not be productive.
19048 @defmac _HTM_FIRST_USER_ABORT_CODE
19049 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
19050 specifies the first abort code which can be used for
19051 @code{__builtin_tabort}. Values below this threshold are reserved for
19055 @deftp {Data type} {struct __htm_tdb}
19056 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
19057 the structure of the transaction diagnostic block as specified in the
19058 Principles of Operation manual chapter 5-91.
19061 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
19062 Same as @code{__builtin_tbegin} but without FPR saves and restores.
19063 Using this variant in code making use of FPRs will leave the FPRs in
19064 undefined state when entering the transaction abort handler code.
19067 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
19068 In addition to @code{__builtin_tbegin} a loop for transient failures
19069 is generated. If tbegin returns a condition code of 2 the transaction
19070 will be retried as often as specified in the second argument. The
19071 perform processor assist instruction is used to tell the CPU about the
19072 number of fails so far.
19075 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
19076 Same as @code{__builtin_tbegin_retry} but without FPR saves and
19077 restores. Using this variant in code making use of FPRs will leave
19078 the FPRs in undefined state when entering the transaction abort
19082 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
19083 Generates the @code{tbeginc} machine instruction starting a constrained
19084 hardware transaction. The second operand is set to @code{0xff08}.
19087 @deftypefn {Built-in Function} int __builtin_tend (void)
19088 Generates the @code{tend} machine instruction finishing a transaction
19089 and making the changes visible to other threads. The condition code
19090 generated by tend is returned as integer value.
19093 @deftypefn {Built-in Function} void __builtin_tabort (int)
19094 Generates the @code{tabort} machine instruction with the specified
19095 abort code. Abort codes from 0 through 255 are reserved and will
19096 result in an error message.
19099 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
19100 Generates the @code{ppa rX,rY,1} machine instruction. Where the
19101 integer parameter is loaded into rX and a value of zero is loaded into
19102 rY. The integer parameter specifies the number of times the
19103 transaction repeatedly aborted.
19106 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
19107 Generates the @code{etnd} machine instruction. The current nesting
19108 depth is returned as integer value. For a nesting depth of 0 the code
19109 is not executed as part of an transaction.
19112 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
19114 Generates the @code{ntstg} machine instruction. The second argument
19115 is written to the first arguments location. The store operation will
19116 not be rolled-back in case of an transaction abort.
19119 @node SH Built-in Functions
19120 @subsection SH Built-in Functions
19121 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
19122 families of processors:
19124 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
19125 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
19126 used by system code that manages threads and execution contexts. The compiler
19127 normally does not generate code that modifies the contents of @samp{GBR} and
19128 thus the value is preserved across function calls. Changing the @samp{GBR}
19129 value in user code must be done with caution, since the compiler might use
19130 @samp{GBR} in order to access thread local variables.
19134 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
19135 Returns the value that is currently set in the @samp{GBR} register.
19136 Memory loads and stores that use the thread pointer as a base address are
19137 turned into @samp{GBR} based displacement loads and stores, if possible.
19145 int get_tcb_value (void)
19147 // Generate @samp{mov.l @@(8,gbr),r0} instruction
19148 return ((my_tcb*)__builtin_thread_pointer ())->c;
19154 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
19155 Returns the value that is currently set in the @samp{FPSCR} register.
19158 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
19159 Sets the @samp{FPSCR} register to the specified value @var{val}, while
19160 preserving the current values of the FR, SZ and PR bits.
19163 @node SPARC VIS Built-in Functions
19164 @subsection SPARC VIS Built-in Functions
19166 GCC supports SIMD operations on the SPARC using both the generic vector
19167 extensions (@pxref{Vector Extensions}) as well as built-in functions for
19168 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
19169 switch, the VIS extension is exposed as the following built-in functions:
19172 typedef int v1si __attribute__ ((vector_size (4)));
19173 typedef int v2si __attribute__ ((vector_size (8)));
19174 typedef short v4hi __attribute__ ((vector_size (8)));
19175 typedef short v2hi __attribute__ ((vector_size (4)));
19176 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
19177 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
19179 void __builtin_vis_write_gsr (int64_t);
19180 int64_t __builtin_vis_read_gsr (void);
19182 void * __builtin_vis_alignaddr (void *, long);
19183 void * __builtin_vis_alignaddrl (void *, long);
19184 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
19185 v2si __builtin_vis_faligndatav2si (v2si, v2si);
19186 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
19187 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
19189 v4hi __builtin_vis_fexpand (v4qi);
19191 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
19192 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
19193 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
19194 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
19195 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
19196 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
19197 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
19199 v4qi __builtin_vis_fpack16 (v4hi);
19200 v8qi __builtin_vis_fpack32 (v2si, v8qi);
19201 v2hi __builtin_vis_fpackfix (v2si);
19202 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
19204 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
19206 long __builtin_vis_edge8 (void *, void *);
19207 long __builtin_vis_edge8l (void *, void *);
19208 long __builtin_vis_edge16 (void *, void *);
19209 long __builtin_vis_edge16l (void *, void *);
19210 long __builtin_vis_edge32 (void *, void *);
19211 long __builtin_vis_edge32l (void *, void *);
19213 long __builtin_vis_fcmple16 (v4hi, v4hi);
19214 long __builtin_vis_fcmple32 (v2si, v2si);
19215 long __builtin_vis_fcmpne16 (v4hi, v4hi);
19216 long __builtin_vis_fcmpne32 (v2si, v2si);
19217 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
19218 long __builtin_vis_fcmpgt32 (v2si, v2si);
19219 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
19220 long __builtin_vis_fcmpeq32 (v2si, v2si);
19222 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
19223 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
19224 v2si __builtin_vis_fpadd32 (v2si, v2si);
19225 v1si __builtin_vis_fpadd32s (v1si, v1si);
19226 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
19227 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
19228 v2si __builtin_vis_fpsub32 (v2si, v2si);
19229 v1si __builtin_vis_fpsub32s (v1si, v1si);
19231 long __builtin_vis_array8 (long, long);
19232 long __builtin_vis_array16 (long, long);
19233 long __builtin_vis_array32 (long, long);
19236 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
19237 functions also become available:
19240 long __builtin_vis_bmask (long, long);
19241 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
19242 v2si __builtin_vis_bshufflev2si (v2si, v2si);
19243 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
19244 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
19246 long __builtin_vis_edge8n (void *, void *);
19247 long __builtin_vis_edge8ln (void *, void *);
19248 long __builtin_vis_edge16n (void *, void *);
19249 long __builtin_vis_edge16ln (void *, void *);
19250 long __builtin_vis_edge32n (void *, void *);
19251 long __builtin_vis_edge32ln (void *, void *);
19254 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
19255 functions also become available:
19258 void __builtin_vis_cmask8 (long);
19259 void __builtin_vis_cmask16 (long);
19260 void __builtin_vis_cmask32 (long);
19262 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
19264 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
19265 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
19266 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
19267 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
19268 v2si __builtin_vis_fsll16 (v2si, v2si);
19269 v2si __builtin_vis_fslas16 (v2si, v2si);
19270 v2si __builtin_vis_fsrl16 (v2si, v2si);
19271 v2si __builtin_vis_fsra16 (v2si, v2si);
19273 long __builtin_vis_pdistn (v8qi, v8qi);
19275 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
19277 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
19278 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
19280 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
19281 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
19282 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
19283 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
19284 v2si __builtin_vis_fpadds32 (v2si, v2si);
19285 v1si __builtin_vis_fpadds32s (v1si, v1si);
19286 v2si __builtin_vis_fpsubs32 (v2si, v2si);
19287 v1si __builtin_vis_fpsubs32s (v1si, v1si);
19289 long __builtin_vis_fucmple8 (v8qi, v8qi);
19290 long __builtin_vis_fucmpne8 (v8qi, v8qi);
19291 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
19292 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
19294 float __builtin_vis_fhadds (float, float);
19295 double __builtin_vis_fhaddd (double, double);
19296 float __builtin_vis_fhsubs (float, float);
19297 double __builtin_vis_fhsubd (double, double);
19298 float __builtin_vis_fnhadds (float, float);
19299 double __builtin_vis_fnhaddd (double, double);
19301 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
19302 int64_t __builtin_vis_xmulx (int64_t, int64_t);
19303 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
19306 When you use the @option{-mvis4} switch, the VIS version 4.0 built-in
19307 functions also become available:
19310 v8qi __builtin_vis_fpadd8 (v8qi, v8qi);
19311 v8qi __builtin_vis_fpadds8 (v8qi, v8qi);
19312 v8qi __builtin_vis_fpaddus8 (v8qi, v8qi);
19313 v4hi __builtin_vis_fpaddus16 (v4hi, v4hi);
19315 v8qi __builtin_vis_fpsub8 (v8qi, v8qi);
19316 v8qi __builtin_vis_fpsubs8 (v8qi, v8qi);
19317 v8qi __builtin_vis_fpsubus8 (v8qi, v8qi);
19318 v4hi __builtin_vis_fpsubus16 (v4hi, v4hi);
19320 long __builtin_vis_fpcmple8 (v8qi, v8qi);
19321 long __builtin_vis_fpcmpgt8 (v8qi, v8qi);
19322 long __builtin_vis_fpcmpule16 (v4hi, v4hi);
19323 long __builtin_vis_fpcmpugt16 (v4hi, v4hi);
19324 long __builtin_vis_fpcmpule32 (v2si, v2si);
19325 long __builtin_vis_fpcmpugt32 (v2si, v2si);
19327 v8qi __builtin_vis_fpmax8 (v8qi, v8qi);
19328 v4hi __builtin_vis_fpmax16 (v4hi, v4hi);
19329 v2si __builtin_vis_fpmax32 (v2si, v2si);
19331 v8qi __builtin_vis_fpmaxu8 (v8qi, v8qi);
19332 v4hi __builtin_vis_fpmaxu16 (v4hi, v4hi);
19333 v2si __builtin_vis_fpmaxu32 (v2si, v2si);
19336 v8qi __builtin_vis_fpmin8 (v8qi, v8qi);
19337 v4hi __builtin_vis_fpmin16 (v4hi, v4hi);
19338 v2si __builtin_vis_fpmin32 (v2si, v2si);
19340 v8qi __builtin_vis_fpminu8 (v8qi, v8qi);
19341 v4hi __builtin_vis_fpminu16 (v4hi, v4hi);
19342 v2si __builtin_vis_fpminu32 (v2si, v2si);
19345 When you use the @option{-mvis4b} switch, the VIS version 4.0B
19346 built-in functions also become available:
19349 v8qi __builtin_vis_dictunpack8 (double, int);
19350 v4hi __builtin_vis_dictunpack16 (double, int);
19351 v2si __builtin_vis_dictunpack32 (double, int);
19353 long __builtin_vis_fpcmple8shl (v8qi, v8qi, int);
19354 long __builtin_vis_fpcmpgt8shl (v8qi, v8qi, int);
19355 long __builtin_vis_fpcmpeq8shl (v8qi, v8qi, int);
19356 long __builtin_vis_fpcmpne8shl (v8qi, v8qi, int);
19358 long __builtin_vis_fpcmple16shl (v4hi, v4hi, int);
19359 long __builtin_vis_fpcmpgt16shl (v4hi, v4hi, int);
19360 long __builtin_vis_fpcmpeq16shl (v4hi, v4hi, int);
19361 long __builtin_vis_fpcmpne16shl (v4hi, v4hi, int);
19363 long __builtin_vis_fpcmple32shl (v2si, v2si, int);
19364 long __builtin_vis_fpcmpgt32shl (v2si, v2si, int);
19365 long __builtin_vis_fpcmpeq32shl (v2si, v2si, int);
19366 long __builtin_vis_fpcmpne32shl (v2si, v2si, int);
19368 long __builtin_vis_fpcmpule8shl (v8qi, v8qi, int);
19369 long __builtin_vis_fpcmpugt8shl (v8qi, v8qi, int);
19370 long __builtin_vis_fpcmpule16shl (v4hi, v4hi, int);
19371 long __builtin_vis_fpcmpugt16shl (v4hi, v4hi, int);
19372 long __builtin_vis_fpcmpule32shl (v2si, v2si, int);
19373 long __builtin_vis_fpcmpugt32shl (v2si, v2si, int);
19375 long __builtin_vis_fpcmpde8shl (v8qi, v8qi, int);
19376 long __builtin_vis_fpcmpde16shl (v4hi, v4hi, int);
19377 long __builtin_vis_fpcmpde32shl (v2si, v2si, int);
19379 long __builtin_vis_fpcmpur8shl (v8qi, v8qi, int);
19380 long __builtin_vis_fpcmpur16shl (v4hi, v4hi, int);
19381 long __builtin_vis_fpcmpur32shl (v2si, v2si, int);
19384 @node SPU Built-in Functions
19385 @subsection SPU Built-in Functions
19387 GCC provides extensions for the SPU processor as described in the
19388 Sony/Toshiba/IBM SPU Language Extensions Specification. GCC's
19389 implementation differs in several ways.
19394 The optional extension of specifying vector constants in parentheses is
19398 A vector initializer requires no cast if the vector constant is of the
19399 same type as the variable it is initializing.
19402 If @code{signed} or @code{unsigned} is omitted, the signedness of the
19403 vector type is the default signedness of the base type. The default
19404 varies depending on the operating system, so a portable program should
19405 always specify the signedness.
19408 By default, the keyword @code{__vector} is added. The macro
19409 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
19413 GCC allows using a @code{typedef} name as the type specifier for a
19417 For C, overloaded functions are implemented with macros so the following
19421 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
19425 Since @code{spu_add} is a macro, the vector constant in the example
19426 is treated as four separate arguments. Wrap the entire argument in
19427 parentheses for this to work.
19430 The extended version of @code{__builtin_expect} is not supported.
19434 @emph{Note:} Only the interface described in the aforementioned
19435 specification is supported. Internally, GCC uses built-in functions to
19436 implement the required functionality, but these are not supported and
19437 are subject to change without notice.
19439 @node TI C6X Built-in Functions
19440 @subsection TI C6X Built-in Functions
19442 GCC provides intrinsics to access certain instructions of the TI C6X
19443 processors. These intrinsics, listed below, are available after
19444 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
19445 to C6X instructions.
19449 int _sadd (int, int)
19450 int _ssub (int, int)
19451 int _sadd2 (int, int)
19452 int _ssub2 (int, int)
19453 long long _mpy2 (int, int)
19454 long long _smpy2 (int, int)
19455 int _add4 (int, int)
19456 int _sub4 (int, int)
19457 int _saddu4 (int, int)
19459 int _smpy (int, int)
19460 int _smpyh (int, int)
19461 int _smpyhl (int, int)
19462 int _smpylh (int, int)
19464 int _sshl (int, int)
19465 int _subc (int, int)
19467 int _avg2 (int, int)
19468 int _avgu4 (int, int)
19470 int _clrr (int, int)
19471 int _extr (int, int)
19472 int _extru (int, int)
19478 @node TILE-Gx Built-in Functions
19479 @subsection TILE-Gx Built-in Functions
19481 GCC provides intrinsics to access every instruction of the TILE-Gx
19482 processor. The intrinsics are of the form:
19486 unsigned long long __insn_@var{op} (...)
19490 Where @var{op} is the name of the instruction. Refer to the ISA manual
19491 for the complete list of instructions.
19493 GCC also provides intrinsics to directly access the network registers.
19494 The intrinsics are:
19498 unsigned long long __tile_idn0_receive (void)
19499 unsigned long long __tile_idn1_receive (void)
19500 unsigned long long __tile_udn0_receive (void)
19501 unsigned long long __tile_udn1_receive (void)
19502 unsigned long long __tile_udn2_receive (void)
19503 unsigned long long __tile_udn3_receive (void)
19504 void __tile_idn_send (unsigned long long)
19505 void __tile_udn_send (unsigned long long)
19509 The intrinsic @code{void __tile_network_barrier (void)} is used to
19510 guarantee that no network operations before it are reordered with
19513 @node TILEPro Built-in Functions
19514 @subsection TILEPro Built-in Functions
19516 GCC provides intrinsics to access every instruction of the TILEPro
19517 processor. The intrinsics are of the form:
19521 unsigned __insn_@var{op} (...)
19526 where @var{op} is the name of the instruction. Refer to the ISA manual
19527 for the complete list of instructions.
19529 GCC also provides intrinsics to directly access the network registers.
19530 The intrinsics are:
19534 unsigned __tile_idn0_receive (void)
19535 unsigned __tile_idn1_receive (void)
19536 unsigned __tile_sn_receive (void)
19537 unsigned __tile_udn0_receive (void)
19538 unsigned __tile_udn1_receive (void)
19539 unsigned __tile_udn2_receive (void)
19540 unsigned __tile_udn3_receive (void)
19541 void __tile_idn_send (unsigned)
19542 void __tile_sn_send (unsigned)
19543 void __tile_udn_send (unsigned)
19547 The intrinsic @code{void __tile_network_barrier (void)} is used to
19548 guarantee that no network operations before it are reordered with
19551 @node x86 Built-in Functions
19552 @subsection x86 Built-in Functions
19554 These built-in functions are available for the x86-32 and x86-64 family
19555 of computers, depending on the command-line switches used.
19557 If you specify command-line switches such as @option{-msse},
19558 the compiler could use the extended instruction sets even if the built-ins
19559 are not used explicitly in the program. For this reason, applications
19560 that perform run-time CPU detection must compile separate files for each
19561 supported architecture, using the appropriate flags. In particular,
19562 the file containing the CPU detection code should be compiled without
19565 The following machine modes are available for use with MMX built-in functions
19566 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
19567 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
19568 vector of eight 8-bit integers. Some of the built-in functions operate on
19569 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
19571 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
19572 of two 32-bit floating-point values.
19574 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
19575 floating-point values. Some instructions use a vector of four 32-bit
19576 integers, these use @code{V4SI}. Finally, some instructions operate on an
19577 entire vector register, interpreting it as a 128-bit integer, these use mode
19580 The x86-32 and x86-64 family of processors use additional built-in
19581 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
19582 floating point and @code{TC} 128-bit complex floating-point values.
19584 The following floating-point built-in functions are always available. All
19585 of them implement the function that is part of the name.
19588 __float128 __builtin_fabsq (__float128)
19589 __float128 __builtin_copysignq (__float128, __float128)
19592 The following built-in functions are always available.
19595 @item __float128 __builtin_infq (void)
19596 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
19597 @findex __builtin_infq
19599 @item __float128 __builtin_huge_valq (void)
19600 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
19601 @findex __builtin_huge_valq
19603 @item __float128 __builtin_nanq (void)
19604 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
19605 @findex __builtin_nanq
19607 @item __float128 __builtin_nansq (void)
19608 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
19609 @findex __builtin_nansq
19612 The following built-in function is always available.
19615 @item void __builtin_ia32_pause (void)
19616 Generates the @code{pause} machine instruction with a compiler memory
19620 The following built-in functions are always available and can be used to
19621 check the target platform type.
19623 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
19624 This function runs the CPU detection code to check the type of CPU and the
19625 features supported. This built-in function needs to be invoked along with the built-in functions
19626 to check CPU type and features, @code{__builtin_cpu_is} and
19627 @code{__builtin_cpu_supports}, only when used in a function that is
19628 executed before any constructors are called. The CPU detection code is
19629 automatically executed in a very high priority constructor.
19631 For example, this function has to be used in @code{ifunc} resolvers that
19632 check for CPU type using the built-in functions @code{__builtin_cpu_is}
19633 and @code{__builtin_cpu_supports}, or in constructors on targets that
19634 don't support constructor priority.
19637 static void (*resolve_memcpy (void)) (void)
19639 // ifunc resolvers fire before constructors, explicitly call the init
19641 __builtin_cpu_init ();
19642 if (__builtin_cpu_supports ("ssse3"))
19643 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
19645 return default_memcpy;
19648 void *memcpy (void *, const void *, size_t)
19649 __attribute__ ((ifunc ("resolve_memcpy")));
19654 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
19655 This function returns a positive integer if the run-time CPU
19656 is of type @var{cpuname}
19657 and returns @code{0} otherwise. The following CPU names can be detected:
19673 Intel Core i7 Nehalem CPU.
19676 Intel Core i7 Westmere CPU.
19679 Intel Core i7 Sandy Bridge CPU.
19685 AMD Family 10h CPU.
19688 AMD Family 10h Barcelona CPU.
19691 AMD Family 10h Shanghai CPU.
19694 AMD Family 10h Istanbul CPU.
19697 AMD Family 14h CPU.
19700 AMD Family 15h CPU.
19703 AMD Family 15h Bulldozer version 1.
19706 AMD Family 15h Bulldozer version 2.
19709 AMD Family 15h Bulldozer version 3.
19712 AMD Family 15h Bulldozer version 4.
19715 AMD Family 16h CPU.
19718 AMD Family 17h CPU.
19721 Here is an example:
19723 if (__builtin_cpu_is ("corei7"))
19725 do_corei7 (); // Core i7 specific implementation.
19729 do_generic (); // Generic implementation.
19734 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
19735 This function returns a positive integer if the run-time CPU
19736 supports @var{feature}
19737 and returns @code{0} otherwise. The following features can be detected:
19745 POPCNT instruction.
19753 SSSE3 instructions.
19755 SSE4.1 instructions.
19757 SSE4.2 instructions.
19763 AVX512F instructions.
19766 Here is an example:
19768 if (__builtin_cpu_supports ("popcnt"))
19770 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
19774 count = generic_countbits (n); //generic implementation.
19780 The following built-in functions are made available by @option{-mmmx}.
19781 All of them generate the machine instruction that is part of the name.
19784 v8qi __builtin_ia32_paddb (v8qi, v8qi)
19785 v4hi __builtin_ia32_paddw (v4hi, v4hi)
19786 v2si __builtin_ia32_paddd (v2si, v2si)
19787 v8qi __builtin_ia32_psubb (v8qi, v8qi)
19788 v4hi __builtin_ia32_psubw (v4hi, v4hi)
19789 v2si __builtin_ia32_psubd (v2si, v2si)
19790 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
19791 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
19792 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
19793 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
19794 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
19795 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
19796 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
19797 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
19798 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
19799 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
19800 di __builtin_ia32_pand (di, di)
19801 di __builtin_ia32_pandn (di,di)
19802 di __builtin_ia32_por (di, di)
19803 di __builtin_ia32_pxor (di, di)
19804 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
19805 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
19806 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
19807 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
19808 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
19809 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
19810 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
19811 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
19812 v2si __builtin_ia32_punpckhdq (v2si, v2si)
19813 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
19814 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
19815 v2si __builtin_ia32_punpckldq (v2si, v2si)
19816 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
19817 v4hi __builtin_ia32_packssdw (v2si, v2si)
19818 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
19820 v4hi __builtin_ia32_psllw (v4hi, v4hi)
19821 v2si __builtin_ia32_pslld (v2si, v2si)
19822 v1di __builtin_ia32_psllq (v1di, v1di)
19823 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
19824 v2si __builtin_ia32_psrld (v2si, v2si)
19825 v1di __builtin_ia32_psrlq (v1di, v1di)
19826 v4hi __builtin_ia32_psraw (v4hi, v4hi)
19827 v2si __builtin_ia32_psrad (v2si, v2si)
19828 v4hi __builtin_ia32_psllwi (v4hi, int)
19829 v2si __builtin_ia32_pslldi (v2si, int)
19830 v1di __builtin_ia32_psllqi (v1di, int)
19831 v4hi __builtin_ia32_psrlwi (v4hi, int)
19832 v2si __builtin_ia32_psrldi (v2si, int)
19833 v1di __builtin_ia32_psrlqi (v1di, int)
19834 v4hi __builtin_ia32_psrawi (v4hi, int)
19835 v2si __builtin_ia32_psradi (v2si, int)
19839 The following built-in functions are made available either with
19840 @option{-msse}, or with @option{-m3dnowa}. All of them generate
19841 the machine instruction that is part of the name.
19844 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
19845 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
19846 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
19847 v1di __builtin_ia32_psadbw (v8qi, v8qi)
19848 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
19849 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
19850 v8qi __builtin_ia32_pminub (v8qi, v8qi)
19851 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
19852 int __builtin_ia32_pmovmskb (v8qi)
19853 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
19854 void __builtin_ia32_movntq (di *, di)
19855 void __builtin_ia32_sfence (void)
19858 The following built-in functions are available when @option{-msse} is used.
19859 All of them generate the machine instruction that is part of the name.
19862 int __builtin_ia32_comieq (v4sf, v4sf)
19863 int __builtin_ia32_comineq (v4sf, v4sf)
19864 int __builtin_ia32_comilt (v4sf, v4sf)
19865 int __builtin_ia32_comile (v4sf, v4sf)
19866 int __builtin_ia32_comigt (v4sf, v4sf)
19867 int __builtin_ia32_comige (v4sf, v4sf)
19868 int __builtin_ia32_ucomieq (v4sf, v4sf)
19869 int __builtin_ia32_ucomineq (v4sf, v4sf)
19870 int __builtin_ia32_ucomilt (v4sf, v4sf)
19871 int __builtin_ia32_ucomile (v4sf, v4sf)
19872 int __builtin_ia32_ucomigt (v4sf, v4sf)
19873 int __builtin_ia32_ucomige (v4sf, v4sf)
19874 v4sf __builtin_ia32_addps (v4sf, v4sf)
19875 v4sf __builtin_ia32_subps (v4sf, v4sf)
19876 v4sf __builtin_ia32_mulps (v4sf, v4sf)
19877 v4sf __builtin_ia32_divps (v4sf, v4sf)
19878 v4sf __builtin_ia32_addss (v4sf, v4sf)
19879 v4sf __builtin_ia32_subss (v4sf, v4sf)
19880 v4sf __builtin_ia32_mulss (v4sf, v4sf)
19881 v4sf __builtin_ia32_divss (v4sf, v4sf)
19882 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
19883 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
19884 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
19885 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
19886 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
19887 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
19888 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
19889 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
19890 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
19891 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
19892 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
19893 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
19894 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
19895 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
19896 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
19897 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
19898 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
19899 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
19900 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
19901 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
19902 v4sf __builtin_ia32_maxps (v4sf, v4sf)
19903 v4sf __builtin_ia32_maxss (v4sf, v4sf)
19904 v4sf __builtin_ia32_minps (v4sf, v4sf)
19905 v4sf __builtin_ia32_minss (v4sf, v4sf)
19906 v4sf __builtin_ia32_andps (v4sf, v4sf)
19907 v4sf __builtin_ia32_andnps (v4sf, v4sf)
19908 v4sf __builtin_ia32_orps (v4sf, v4sf)
19909 v4sf __builtin_ia32_xorps (v4sf, v4sf)
19910 v4sf __builtin_ia32_movss (v4sf, v4sf)
19911 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
19912 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
19913 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
19914 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
19915 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
19916 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
19917 v2si __builtin_ia32_cvtps2pi (v4sf)
19918 int __builtin_ia32_cvtss2si (v4sf)
19919 v2si __builtin_ia32_cvttps2pi (v4sf)
19920 int __builtin_ia32_cvttss2si (v4sf)
19921 v4sf __builtin_ia32_rcpps (v4sf)
19922 v4sf __builtin_ia32_rsqrtps (v4sf)
19923 v4sf __builtin_ia32_sqrtps (v4sf)
19924 v4sf __builtin_ia32_rcpss (v4sf)
19925 v4sf __builtin_ia32_rsqrtss (v4sf)
19926 v4sf __builtin_ia32_sqrtss (v4sf)
19927 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
19928 void __builtin_ia32_movntps (float *, v4sf)
19929 int __builtin_ia32_movmskps (v4sf)
19932 The following built-in functions are available when @option{-msse} is used.
19935 @item v4sf __builtin_ia32_loadups (float *)
19936 Generates the @code{movups} machine instruction as a load from memory.
19937 @item void __builtin_ia32_storeups (float *, v4sf)
19938 Generates the @code{movups} machine instruction as a store to memory.
19939 @item v4sf __builtin_ia32_loadss (float *)
19940 Generates the @code{movss} machine instruction as a load from memory.
19941 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
19942 Generates the @code{movhps} machine instruction as a load from memory.
19943 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
19944 Generates the @code{movlps} machine instruction as a load from memory
19945 @item void __builtin_ia32_storehps (v2sf *, v4sf)
19946 Generates the @code{movhps} machine instruction as a store to memory.
19947 @item void __builtin_ia32_storelps (v2sf *, v4sf)
19948 Generates the @code{movlps} machine instruction as a store to memory.
19951 The following built-in functions are available when @option{-msse2} is used.
19952 All of them generate the machine instruction that is part of the name.
19955 int __builtin_ia32_comisdeq (v2df, v2df)
19956 int __builtin_ia32_comisdlt (v2df, v2df)
19957 int __builtin_ia32_comisdle (v2df, v2df)
19958 int __builtin_ia32_comisdgt (v2df, v2df)
19959 int __builtin_ia32_comisdge (v2df, v2df)
19960 int __builtin_ia32_comisdneq (v2df, v2df)
19961 int __builtin_ia32_ucomisdeq (v2df, v2df)
19962 int __builtin_ia32_ucomisdlt (v2df, v2df)
19963 int __builtin_ia32_ucomisdle (v2df, v2df)
19964 int __builtin_ia32_ucomisdgt (v2df, v2df)
19965 int __builtin_ia32_ucomisdge (v2df, v2df)
19966 int __builtin_ia32_ucomisdneq (v2df, v2df)
19967 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
19968 v2df __builtin_ia32_cmpltpd (v2df, v2df)
19969 v2df __builtin_ia32_cmplepd (v2df, v2df)
19970 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
19971 v2df __builtin_ia32_cmpgepd (v2df, v2df)
19972 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
19973 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
19974 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
19975 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
19976 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
19977 v2df __builtin_ia32_cmpngepd (v2df, v2df)
19978 v2df __builtin_ia32_cmpordpd (v2df, v2df)
19979 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
19980 v2df __builtin_ia32_cmpltsd (v2df, v2df)
19981 v2df __builtin_ia32_cmplesd (v2df, v2df)
19982 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
19983 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
19984 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
19985 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
19986 v2df __builtin_ia32_cmpordsd (v2df, v2df)
19987 v2di __builtin_ia32_paddq (v2di, v2di)
19988 v2di __builtin_ia32_psubq (v2di, v2di)
19989 v2df __builtin_ia32_addpd (v2df, v2df)
19990 v2df __builtin_ia32_subpd (v2df, v2df)
19991 v2df __builtin_ia32_mulpd (v2df, v2df)
19992 v2df __builtin_ia32_divpd (v2df, v2df)
19993 v2df __builtin_ia32_addsd (v2df, v2df)
19994 v2df __builtin_ia32_subsd (v2df, v2df)
19995 v2df __builtin_ia32_mulsd (v2df, v2df)
19996 v2df __builtin_ia32_divsd (v2df, v2df)
19997 v2df __builtin_ia32_minpd (v2df, v2df)
19998 v2df __builtin_ia32_maxpd (v2df, v2df)
19999 v2df __builtin_ia32_minsd (v2df, v2df)
20000 v2df __builtin_ia32_maxsd (v2df, v2df)
20001 v2df __builtin_ia32_andpd (v2df, v2df)
20002 v2df __builtin_ia32_andnpd (v2df, v2df)
20003 v2df __builtin_ia32_orpd (v2df, v2df)
20004 v2df __builtin_ia32_xorpd (v2df, v2df)
20005 v2df __builtin_ia32_movsd (v2df, v2df)
20006 v2df __builtin_ia32_unpckhpd (v2df, v2df)
20007 v2df __builtin_ia32_unpcklpd (v2df, v2df)
20008 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
20009 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
20010 v4si __builtin_ia32_paddd128 (v4si, v4si)
20011 v2di __builtin_ia32_paddq128 (v2di, v2di)
20012 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
20013 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
20014 v4si __builtin_ia32_psubd128 (v4si, v4si)
20015 v2di __builtin_ia32_psubq128 (v2di, v2di)
20016 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
20017 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
20018 v2di __builtin_ia32_pand128 (v2di, v2di)
20019 v2di __builtin_ia32_pandn128 (v2di, v2di)
20020 v2di __builtin_ia32_por128 (v2di, v2di)
20021 v2di __builtin_ia32_pxor128 (v2di, v2di)
20022 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
20023 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
20024 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
20025 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
20026 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
20027 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
20028 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
20029 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
20030 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
20031 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
20032 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
20033 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
20034 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
20035 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
20036 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
20037 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
20038 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
20039 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
20040 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
20041 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
20042 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
20043 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
20044 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
20045 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
20046 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
20047 v2df __builtin_ia32_loadupd (double *)
20048 void __builtin_ia32_storeupd (double *, v2df)
20049 v2df __builtin_ia32_loadhpd (v2df, double const *)
20050 v2df __builtin_ia32_loadlpd (v2df, double const *)
20051 int __builtin_ia32_movmskpd (v2df)
20052 int __builtin_ia32_pmovmskb128 (v16qi)
20053 void __builtin_ia32_movnti (int *, int)
20054 void __builtin_ia32_movnti64 (long long int *, long long int)
20055 void __builtin_ia32_movntpd (double *, v2df)
20056 void __builtin_ia32_movntdq (v2df *, v2df)
20057 v4si __builtin_ia32_pshufd (v4si, int)
20058 v8hi __builtin_ia32_pshuflw (v8hi, int)
20059 v8hi __builtin_ia32_pshufhw (v8hi, int)
20060 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
20061 v2df __builtin_ia32_sqrtpd (v2df)
20062 v2df __builtin_ia32_sqrtsd (v2df)
20063 v2df __builtin_ia32_shufpd (v2df, v2df, int)
20064 v2df __builtin_ia32_cvtdq2pd (v4si)
20065 v4sf __builtin_ia32_cvtdq2ps (v4si)
20066 v4si __builtin_ia32_cvtpd2dq (v2df)
20067 v2si __builtin_ia32_cvtpd2pi (v2df)
20068 v4sf __builtin_ia32_cvtpd2ps (v2df)
20069 v4si __builtin_ia32_cvttpd2dq (v2df)
20070 v2si __builtin_ia32_cvttpd2pi (v2df)
20071 v2df __builtin_ia32_cvtpi2pd (v2si)
20072 int __builtin_ia32_cvtsd2si (v2df)
20073 int __builtin_ia32_cvttsd2si (v2df)
20074 long long __builtin_ia32_cvtsd2si64 (v2df)
20075 long long __builtin_ia32_cvttsd2si64 (v2df)
20076 v4si __builtin_ia32_cvtps2dq (v4sf)
20077 v2df __builtin_ia32_cvtps2pd (v4sf)
20078 v4si __builtin_ia32_cvttps2dq (v4sf)
20079 v2df __builtin_ia32_cvtsi2sd (v2df, int)
20080 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
20081 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
20082 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
20083 void __builtin_ia32_clflush (const void *)
20084 void __builtin_ia32_lfence (void)
20085 void __builtin_ia32_mfence (void)
20086 v16qi __builtin_ia32_loaddqu (const char *)
20087 void __builtin_ia32_storedqu (char *, v16qi)
20088 v1di __builtin_ia32_pmuludq (v2si, v2si)
20089 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
20090 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
20091 v4si __builtin_ia32_pslld128 (v4si, v4si)
20092 v2di __builtin_ia32_psllq128 (v2di, v2di)
20093 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
20094 v4si __builtin_ia32_psrld128 (v4si, v4si)
20095 v2di __builtin_ia32_psrlq128 (v2di, v2di)
20096 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
20097 v4si __builtin_ia32_psrad128 (v4si, v4si)
20098 v2di __builtin_ia32_pslldqi128 (v2di, int)
20099 v8hi __builtin_ia32_psllwi128 (v8hi, int)
20100 v4si __builtin_ia32_pslldi128 (v4si, int)
20101 v2di __builtin_ia32_psllqi128 (v2di, int)
20102 v2di __builtin_ia32_psrldqi128 (v2di, int)
20103 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
20104 v4si __builtin_ia32_psrldi128 (v4si, int)
20105 v2di __builtin_ia32_psrlqi128 (v2di, int)
20106 v8hi __builtin_ia32_psrawi128 (v8hi, int)
20107 v4si __builtin_ia32_psradi128 (v4si, int)
20108 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
20109 v2di __builtin_ia32_movq128 (v2di)
20112 The following built-in functions are available when @option{-msse3} is used.
20113 All of them generate the machine instruction that is part of the name.
20116 v2df __builtin_ia32_addsubpd (v2df, v2df)
20117 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
20118 v2df __builtin_ia32_haddpd (v2df, v2df)
20119 v4sf __builtin_ia32_haddps (v4sf, v4sf)
20120 v2df __builtin_ia32_hsubpd (v2df, v2df)
20121 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
20122 v16qi __builtin_ia32_lddqu (char const *)
20123 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
20124 v4sf __builtin_ia32_movshdup (v4sf)
20125 v4sf __builtin_ia32_movsldup (v4sf)
20126 void __builtin_ia32_mwait (unsigned int, unsigned int)
20129 The following built-in functions are available when @option{-mssse3} is used.
20130 All of them generate the machine instruction that is part of the name.
20133 v2si __builtin_ia32_phaddd (v2si, v2si)
20134 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
20135 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
20136 v2si __builtin_ia32_phsubd (v2si, v2si)
20137 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
20138 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
20139 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
20140 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
20141 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
20142 v8qi __builtin_ia32_psignb (v8qi, v8qi)
20143 v2si __builtin_ia32_psignd (v2si, v2si)
20144 v4hi __builtin_ia32_psignw (v4hi, v4hi)
20145 v1di __builtin_ia32_palignr (v1di, v1di, int)
20146 v8qi __builtin_ia32_pabsb (v8qi)
20147 v2si __builtin_ia32_pabsd (v2si)
20148 v4hi __builtin_ia32_pabsw (v4hi)
20151 The following built-in functions are available when @option{-mssse3} is used.
20152 All of them generate the machine instruction that is part of the name.
20155 v4si __builtin_ia32_phaddd128 (v4si, v4si)
20156 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
20157 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
20158 v4si __builtin_ia32_phsubd128 (v4si, v4si)
20159 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
20160 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
20161 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
20162 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
20163 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
20164 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
20165 v4si __builtin_ia32_psignd128 (v4si, v4si)
20166 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
20167 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
20168 v16qi __builtin_ia32_pabsb128 (v16qi)
20169 v4si __builtin_ia32_pabsd128 (v4si)
20170 v8hi __builtin_ia32_pabsw128 (v8hi)
20173 The following built-in functions are available when @option{-msse4.1} is
20174 used. All of them generate the machine instruction that is part of the
20178 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
20179 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
20180 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
20181 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
20182 v2df __builtin_ia32_dppd (v2df, v2df, const int)
20183 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
20184 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
20185 v2di __builtin_ia32_movntdqa (v2di *);
20186 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
20187 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
20188 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
20189 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
20190 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
20191 v8hi __builtin_ia32_phminposuw128 (v8hi)
20192 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
20193 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
20194 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
20195 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
20196 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
20197 v4si __builtin_ia32_pminsd128 (v4si, v4si)
20198 v4si __builtin_ia32_pminud128 (v4si, v4si)
20199 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
20200 v4si __builtin_ia32_pmovsxbd128 (v16qi)
20201 v2di __builtin_ia32_pmovsxbq128 (v16qi)
20202 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
20203 v2di __builtin_ia32_pmovsxdq128 (v4si)
20204 v4si __builtin_ia32_pmovsxwd128 (v8hi)
20205 v2di __builtin_ia32_pmovsxwq128 (v8hi)
20206 v4si __builtin_ia32_pmovzxbd128 (v16qi)
20207 v2di __builtin_ia32_pmovzxbq128 (v16qi)
20208 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
20209 v2di __builtin_ia32_pmovzxdq128 (v4si)
20210 v4si __builtin_ia32_pmovzxwd128 (v8hi)
20211 v2di __builtin_ia32_pmovzxwq128 (v8hi)
20212 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
20213 v4si __builtin_ia32_pmulld128 (v4si, v4si)
20214 int __builtin_ia32_ptestc128 (v2di, v2di)
20215 int __builtin_ia32_ptestnzc128 (v2di, v2di)
20216 int __builtin_ia32_ptestz128 (v2di, v2di)
20217 v2df __builtin_ia32_roundpd (v2df, const int)
20218 v4sf __builtin_ia32_roundps (v4sf, const int)
20219 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
20220 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
20223 The following built-in functions are available when @option{-msse4.1} is
20227 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
20228 Generates the @code{insertps} machine instruction.
20229 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
20230 Generates the @code{pextrb} machine instruction.
20231 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
20232 Generates the @code{pinsrb} machine instruction.
20233 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
20234 Generates the @code{pinsrd} machine instruction.
20235 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
20236 Generates the @code{pinsrq} machine instruction in 64bit mode.
20239 The following built-in functions are changed to generate new SSE4.1
20240 instructions when @option{-msse4.1} is used.
20243 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
20244 Generates the @code{extractps} machine instruction.
20245 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
20246 Generates the @code{pextrd} machine instruction.
20247 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
20248 Generates the @code{pextrq} machine instruction in 64bit mode.
20251 The following built-in functions are available when @option{-msse4.2} is
20252 used. All of them generate the machine instruction that is part of the
20256 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
20257 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
20258 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
20259 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
20260 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
20261 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
20262 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
20263 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
20264 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
20265 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
20266 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
20267 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
20268 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
20269 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
20270 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
20273 The following built-in functions are available when @option{-msse4.2} is
20277 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
20278 Generates the @code{crc32b} machine instruction.
20279 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
20280 Generates the @code{crc32w} machine instruction.
20281 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
20282 Generates the @code{crc32l} machine instruction.
20283 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
20284 Generates the @code{crc32q} machine instruction.
20287 The following built-in functions are changed to generate new SSE4.2
20288 instructions when @option{-msse4.2} is used.
20291 @item int __builtin_popcount (unsigned int)
20292 Generates the @code{popcntl} machine instruction.
20293 @item int __builtin_popcountl (unsigned long)
20294 Generates the @code{popcntl} or @code{popcntq} machine instruction,
20295 depending on the size of @code{unsigned long}.
20296 @item int __builtin_popcountll (unsigned long long)
20297 Generates the @code{popcntq} machine instruction.
20300 The following built-in functions are available when @option{-mavx} is
20301 used. All of them generate the machine instruction that is part of the
20305 v4df __builtin_ia32_addpd256 (v4df,v4df)
20306 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
20307 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
20308 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
20309 v4df __builtin_ia32_andnpd256 (v4df,v4df)
20310 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
20311 v4df __builtin_ia32_andpd256 (v4df,v4df)
20312 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
20313 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
20314 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
20315 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
20316 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
20317 v2df __builtin_ia32_cmppd (v2df,v2df,int)
20318 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
20319 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
20320 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
20321 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
20322 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
20323 v4df __builtin_ia32_cvtdq2pd256 (v4si)
20324 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
20325 v4si __builtin_ia32_cvtpd2dq256 (v4df)
20326 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
20327 v8si __builtin_ia32_cvtps2dq256 (v8sf)
20328 v4df __builtin_ia32_cvtps2pd256 (v4sf)
20329 v4si __builtin_ia32_cvttpd2dq256 (v4df)
20330 v8si __builtin_ia32_cvttps2dq256 (v8sf)
20331 v4df __builtin_ia32_divpd256 (v4df,v4df)
20332 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
20333 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
20334 v4df __builtin_ia32_haddpd256 (v4df,v4df)
20335 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
20336 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
20337 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
20338 v32qi __builtin_ia32_lddqu256 (pcchar)
20339 v32qi __builtin_ia32_loaddqu256 (pcchar)
20340 v4df __builtin_ia32_loadupd256 (pcdouble)
20341 v8sf __builtin_ia32_loadups256 (pcfloat)
20342 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
20343 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
20344 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
20345 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
20346 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
20347 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
20348 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
20349 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
20350 v4df __builtin_ia32_maxpd256 (v4df,v4df)
20351 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
20352 v4df __builtin_ia32_minpd256 (v4df,v4df)
20353 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
20354 v4df __builtin_ia32_movddup256 (v4df)
20355 int __builtin_ia32_movmskpd256 (v4df)
20356 int __builtin_ia32_movmskps256 (v8sf)
20357 v8sf __builtin_ia32_movshdup256 (v8sf)
20358 v8sf __builtin_ia32_movsldup256 (v8sf)
20359 v4df __builtin_ia32_mulpd256 (v4df,v4df)
20360 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
20361 v4df __builtin_ia32_orpd256 (v4df,v4df)
20362 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
20363 v2df __builtin_ia32_pd_pd256 (v4df)
20364 v4df __builtin_ia32_pd256_pd (v2df)
20365 v4sf __builtin_ia32_ps_ps256 (v8sf)
20366 v8sf __builtin_ia32_ps256_ps (v4sf)
20367 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
20368 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
20369 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
20370 v8sf __builtin_ia32_rcpps256 (v8sf)
20371 v4df __builtin_ia32_roundpd256 (v4df,int)
20372 v8sf __builtin_ia32_roundps256 (v8sf,int)
20373 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
20374 v8sf __builtin_ia32_rsqrtps256 (v8sf)
20375 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
20376 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
20377 v4si __builtin_ia32_si_si256 (v8si)
20378 v8si __builtin_ia32_si256_si (v4si)
20379 v4df __builtin_ia32_sqrtpd256 (v4df)
20380 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
20381 v8sf __builtin_ia32_sqrtps256 (v8sf)
20382 void __builtin_ia32_storedqu256 (pchar,v32qi)
20383 void __builtin_ia32_storeupd256 (pdouble,v4df)
20384 void __builtin_ia32_storeups256 (pfloat,v8sf)
20385 v4df __builtin_ia32_subpd256 (v4df,v4df)
20386 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
20387 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
20388 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
20389 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
20390 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
20391 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
20392 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
20393 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
20394 v4sf __builtin_ia32_vbroadcastss (pcfloat)
20395 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
20396 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
20397 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
20398 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
20399 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
20400 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
20401 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
20402 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
20403 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
20404 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
20405 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
20406 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
20407 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
20408 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
20409 v2df __builtin_ia32_vpermilpd (v2df,int)
20410 v4df __builtin_ia32_vpermilpd256 (v4df,int)
20411 v4sf __builtin_ia32_vpermilps (v4sf,int)
20412 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
20413 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
20414 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
20415 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
20416 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
20417 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
20418 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
20419 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
20420 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
20421 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
20422 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
20423 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
20424 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
20425 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
20426 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
20427 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
20428 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
20429 void __builtin_ia32_vzeroall (void)
20430 void __builtin_ia32_vzeroupper (void)
20431 v4df __builtin_ia32_xorpd256 (v4df,v4df)
20432 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
20435 The following built-in functions are available when @option{-mavx2} is
20436 used. All of them generate the machine instruction that is part of the
20440 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
20441 v32qi __builtin_ia32_pabsb256 (v32qi)
20442 v16hi __builtin_ia32_pabsw256 (v16hi)
20443 v8si __builtin_ia32_pabsd256 (v8si)
20444 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
20445 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
20446 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
20447 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
20448 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
20449 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
20450 v8si __builtin_ia32_paddd256 (v8si,v8si)
20451 v4di __builtin_ia32_paddq256 (v4di,v4di)
20452 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
20453 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
20454 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
20455 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
20456 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
20457 v4di __builtin_ia32_andsi256 (v4di,v4di)
20458 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
20459 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
20460 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
20461 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
20462 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
20463 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
20464 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
20465 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
20466 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
20467 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
20468 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
20469 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
20470 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
20471 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
20472 v8si __builtin_ia32_phaddd256 (v8si,v8si)
20473 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
20474 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
20475 v8si __builtin_ia32_phsubd256 (v8si,v8si)
20476 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
20477 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
20478 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
20479 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
20480 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
20481 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
20482 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
20483 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
20484 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
20485 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
20486 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
20487 v8si __builtin_ia32_pminsd256 (v8si,v8si)
20488 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
20489 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
20490 v8si __builtin_ia32_pminud256 (v8si,v8si)
20491 int __builtin_ia32_pmovmskb256 (v32qi)
20492 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
20493 v8si __builtin_ia32_pmovsxbd256 (v16qi)
20494 v4di __builtin_ia32_pmovsxbq256 (v16qi)
20495 v8si __builtin_ia32_pmovsxwd256 (v8hi)
20496 v4di __builtin_ia32_pmovsxwq256 (v8hi)
20497 v4di __builtin_ia32_pmovsxdq256 (v4si)
20498 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
20499 v8si __builtin_ia32_pmovzxbd256 (v16qi)
20500 v4di __builtin_ia32_pmovzxbq256 (v16qi)
20501 v8si __builtin_ia32_pmovzxwd256 (v8hi)
20502 v4di __builtin_ia32_pmovzxwq256 (v8hi)
20503 v4di __builtin_ia32_pmovzxdq256 (v4si)
20504 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
20505 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
20506 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
20507 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
20508 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
20509 v8si __builtin_ia32_pmulld256 (v8si,v8si)
20510 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
20511 v4di __builtin_ia32_por256 (v4di,v4di)
20512 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
20513 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
20514 v8si __builtin_ia32_pshufd256 (v8si,int)
20515 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
20516 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
20517 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
20518 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
20519 v8si __builtin_ia32_psignd256 (v8si,v8si)
20520 v4di __builtin_ia32_pslldqi256 (v4di,int)
20521 v16hi __builtin_ia32_psllwi256 (16hi,int)
20522 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
20523 v8si __builtin_ia32_pslldi256 (v8si,int)
20524 v8si __builtin_ia32_pslld256(v8si,v4si)
20525 v4di __builtin_ia32_psllqi256 (v4di,int)
20526 v4di __builtin_ia32_psllq256(v4di,v2di)
20527 v16hi __builtin_ia32_psrawi256 (v16hi,int)
20528 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
20529 v8si __builtin_ia32_psradi256 (v8si,int)
20530 v8si __builtin_ia32_psrad256 (v8si,v4si)
20531 v4di __builtin_ia32_psrldqi256 (v4di, int)
20532 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
20533 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
20534 v8si __builtin_ia32_psrldi256 (v8si,int)
20535 v8si __builtin_ia32_psrld256 (v8si,v4si)
20536 v4di __builtin_ia32_psrlqi256 (v4di,int)
20537 v4di __builtin_ia32_psrlq256(v4di,v2di)
20538 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
20539 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
20540 v8si __builtin_ia32_psubd256 (v8si,v8si)
20541 v4di __builtin_ia32_psubq256 (v4di,v4di)
20542 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
20543 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
20544 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
20545 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
20546 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
20547 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
20548 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
20549 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
20550 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
20551 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
20552 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
20553 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
20554 v4di __builtin_ia32_pxor256 (v4di,v4di)
20555 v4di __builtin_ia32_movntdqa256 (pv4di)
20556 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
20557 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
20558 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
20559 v4di __builtin_ia32_vbroadcastsi256 (v2di)
20560 v4si __builtin_ia32_pblendd128 (v4si,v4si)
20561 v8si __builtin_ia32_pblendd256 (v8si,v8si)
20562 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
20563 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
20564 v8si __builtin_ia32_pbroadcastd256 (v4si)
20565 v4di __builtin_ia32_pbroadcastq256 (v2di)
20566 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
20567 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
20568 v4si __builtin_ia32_pbroadcastd128 (v4si)
20569 v2di __builtin_ia32_pbroadcastq128 (v2di)
20570 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
20571 v4df __builtin_ia32_permdf256 (v4df,int)
20572 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
20573 v4di __builtin_ia32_permdi256 (v4di,int)
20574 v4di __builtin_ia32_permti256 (v4di,v4di,int)
20575 v4di __builtin_ia32_extract128i256 (v4di,int)
20576 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
20577 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
20578 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
20579 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
20580 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
20581 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
20582 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
20583 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
20584 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
20585 v8si __builtin_ia32_psllv8si (v8si,v8si)
20586 v4si __builtin_ia32_psllv4si (v4si,v4si)
20587 v4di __builtin_ia32_psllv4di (v4di,v4di)
20588 v2di __builtin_ia32_psllv2di (v2di,v2di)
20589 v8si __builtin_ia32_psrav8si (v8si,v8si)
20590 v4si __builtin_ia32_psrav4si (v4si,v4si)
20591 v8si __builtin_ia32_psrlv8si (v8si,v8si)
20592 v4si __builtin_ia32_psrlv4si (v4si,v4si)
20593 v4di __builtin_ia32_psrlv4di (v4di,v4di)
20594 v2di __builtin_ia32_psrlv2di (v2di,v2di)
20595 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
20596 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
20597 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
20598 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
20599 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
20600 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
20601 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
20602 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
20603 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
20604 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
20605 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
20606 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
20607 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
20608 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
20609 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
20610 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
20613 The following built-in functions are available when @option{-maes} is
20614 used. All of them generate the machine instruction that is part of the
20618 v2di __builtin_ia32_aesenc128 (v2di, v2di)
20619 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
20620 v2di __builtin_ia32_aesdec128 (v2di, v2di)
20621 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
20622 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
20623 v2di __builtin_ia32_aesimc128 (v2di)
20626 The following built-in function is available when @option{-mpclmul} is
20630 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
20631 Generates the @code{pclmulqdq} machine instruction.
20634 The following built-in function is available when @option{-mfsgsbase} is
20635 used. All of them generate the machine instruction that is part of the
20639 unsigned int __builtin_ia32_rdfsbase32 (void)
20640 unsigned long long __builtin_ia32_rdfsbase64 (void)
20641 unsigned int __builtin_ia32_rdgsbase32 (void)
20642 unsigned long long __builtin_ia32_rdgsbase64 (void)
20643 void _writefsbase_u32 (unsigned int)
20644 void _writefsbase_u64 (unsigned long long)
20645 void _writegsbase_u32 (unsigned int)
20646 void _writegsbase_u64 (unsigned long long)
20649 The following built-in function is available when @option{-mrdrnd} is
20650 used. All of them generate the machine instruction that is part of the
20654 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
20655 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
20656 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
20659 The following built-in functions are available when @option{-msse4a} is used.
20660 All of them generate the machine instruction that is part of the name.
20663 void __builtin_ia32_movntsd (double *, v2df)
20664 void __builtin_ia32_movntss (float *, v4sf)
20665 v2di __builtin_ia32_extrq (v2di, v16qi)
20666 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
20667 v2di __builtin_ia32_insertq (v2di, v2di)
20668 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
20671 The following built-in functions are available when @option{-mxop} is used.
20673 v2df __builtin_ia32_vfrczpd (v2df)
20674 v4sf __builtin_ia32_vfrczps (v4sf)
20675 v2df __builtin_ia32_vfrczsd (v2df)
20676 v4sf __builtin_ia32_vfrczss (v4sf)
20677 v4df __builtin_ia32_vfrczpd256 (v4df)
20678 v8sf __builtin_ia32_vfrczps256 (v8sf)
20679 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
20680 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
20681 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
20682 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
20683 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
20684 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
20685 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
20686 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
20687 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
20688 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
20689 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
20690 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
20691 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
20692 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
20693 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
20694 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
20695 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
20696 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
20697 v4si __builtin_ia32_vpcomequd (v4si, v4si)
20698 v2di __builtin_ia32_vpcomequq (v2di, v2di)
20699 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
20700 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
20701 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
20702 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
20703 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
20704 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
20705 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
20706 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
20707 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
20708 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
20709 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
20710 v4si __builtin_ia32_vpcomged (v4si, v4si)
20711 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
20712 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
20713 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
20714 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
20715 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
20716 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
20717 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
20718 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
20719 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
20720 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
20721 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
20722 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
20723 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
20724 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
20725 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
20726 v4si __builtin_ia32_vpcomled (v4si, v4si)
20727 v2di __builtin_ia32_vpcomleq (v2di, v2di)
20728 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
20729 v4si __builtin_ia32_vpcomleud (v4si, v4si)
20730 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
20731 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
20732 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
20733 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
20734 v4si __builtin_ia32_vpcomltd (v4si, v4si)
20735 v2di __builtin_ia32_vpcomltq (v2di, v2di)
20736 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
20737 v4si __builtin_ia32_vpcomltud (v4si, v4si)
20738 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
20739 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
20740 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
20741 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
20742 v4si __builtin_ia32_vpcomned (v4si, v4si)
20743 v2di __builtin_ia32_vpcomneq (v2di, v2di)
20744 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
20745 v4si __builtin_ia32_vpcomneud (v4si, v4si)
20746 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
20747 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
20748 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
20749 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
20750 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
20751 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
20752 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
20753 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
20754 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
20755 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
20756 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
20757 v4si __builtin_ia32_vphaddbd (v16qi)
20758 v2di __builtin_ia32_vphaddbq (v16qi)
20759 v8hi __builtin_ia32_vphaddbw (v16qi)
20760 v2di __builtin_ia32_vphadddq (v4si)
20761 v4si __builtin_ia32_vphaddubd (v16qi)
20762 v2di __builtin_ia32_vphaddubq (v16qi)
20763 v8hi __builtin_ia32_vphaddubw (v16qi)
20764 v2di __builtin_ia32_vphaddudq (v4si)
20765 v4si __builtin_ia32_vphadduwd (v8hi)
20766 v2di __builtin_ia32_vphadduwq (v8hi)
20767 v4si __builtin_ia32_vphaddwd (v8hi)
20768 v2di __builtin_ia32_vphaddwq (v8hi)
20769 v8hi __builtin_ia32_vphsubbw (v16qi)
20770 v2di __builtin_ia32_vphsubdq (v4si)
20771 v4si __builtin_ia32_vphsubwd (v8hi)
20772 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
20773 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
20774 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
20775 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
20776 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
20777 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
20778 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
20779 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
20780 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
20781 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
20782 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
20783 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
20784 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
20785 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
20786 v4si __builtin_ia32_vprotd (v4si, v4si)
20787 v2di __builtin_ia32_vprotq (v2di, v2di)
20788 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
20789 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
20790 v4si __builtin_ia32_vpshad (v4si, v4si)
20791 v2di __builtin_ia32_vpshaq (v2di, v2di)
20792 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
20793 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
20794 v4si __builtin_ia32_vpshld (v4si, v4si)
20795 v2di __builtin_ia32_vpshlq (v2di, v2di)
20796 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
20799 The following built-in functions are available when @option{-mfma4} is used.
20800 All of them generate the machine instruction that is part of the name.
20803 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
20804 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
20805 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
20806 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
20807 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
20808 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
20809 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
20810 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
20811 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
20812 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
20813 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
20814 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
20815 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
20816 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
20817 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
20818 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
20819 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
20820 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
20821 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
20822 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
20823 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
20824 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
20825 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
20826 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
20827 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
20828 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
20829 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
20830 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
20831 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
20832 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
20833 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
20834 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
20838 The following built-in functions are available when @option{-mlwp} is used.
20841 void __builtin_ia32_llwpcb16 (void *);
20842 void __builtin_ia32_llwpcb32 (void *);
20843 void __builtin_ia32_llwpcb64 (void *);
20844 void * __builtin_ia32_llwpcb16 (void);
20845 void * __builtin_ia32_llwpcb32 (void);
20846 void * __builtin_ia32_llwpcb64 (void);
20847 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
20848 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
20849 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
20850 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
20851 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
20852 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
20855 The following built-in functions are available when @option{-mbmi} is used.
20856 All of them generate the machine instruction that is part of the name.
20858 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
20859 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
20862 The following built-in functions are available when @option{-mbmi2} is used.
20863 All of them generate the machine instruction that is part of the name.
20865 unsigned int _bzhi_u32 (unsigned int, unsigned int)
20866 unsigned int _pdep_u32 (unsigned int, unsigned int)
20867 unsigned int _pext_u32 (unsigned int, unsigned int)
20868 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
20869 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
20870 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
20873 The following built-in functions are available when @option{-mlzcnt} is used.
20874 All of them generate the machine instruction that is part of the name.
20876 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
20877 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
20878 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
20881 The following built-in functions are available when @option{-mfxsr} is used.
20882 All of them generate the machine instruction that is part of the name.
20884 void __builtin_ia32_fxsave (void *)
20885 void __builtin_ia32_fxrstor (void *)
20886 void __builtin_ia32_fxsave64 (void *)
20887 void __builtin_ia32_fxrstor64 (void *)
20890 The following built-in functions are available when @option{-mxsave} is used.
20891 All of them generate the machine instruction that is part of the name.
20893 void __builtin_ia32_xsave (void *, long long)
20894 void __builtin_ia32_xrstor (void *, long long)
20895 void __builtin_ia32_xsave64 (void *, long long)
20896 void __builtin_ia32_xrstor64 (void *, long long)
20899 The following built-in functions are available when @option{-mxsaveopt} is used.
20900 All of them generate the machine instruction that is part of the name.
20902 void __builtin_ia32_xsaveopt (void *, long long)
20903 void __builtin_ia32_xsaveopt64 (void *, long long)
20906 The following built-in functions are available when @option{-mtbm} is used.
20907 Both of them generate the immediate form of the bextr machine instruction.
20909 unsigned int __builtin_ia32_bextri_u32 (unsigned int,
20910 const unsigned int);
20911 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long,
20912 const unsigned long long);
20916 The following built-in functions are available when @option{-m3dnow} is used.
20917 All of them generate the machine instruction that is part of the name.
20920 void __builtin_ia32_femms (void)
20921 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
20922 v2si __builtin_ia32_pf2id (v2sf)
20923 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
20924 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
20925 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
20926 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
20927 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
20928 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
20929 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
20930 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
20931 v2sf __builtin_ia32_pfrcp (v2sf)
20932 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
20933 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
20934 v2sf __builtin_ia32_pfrsqrt (v2sf)
20935 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
20936 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
20937 v2sf __builtin_ia32_pi2fd (v2si)
20938 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
20941 The following built-in functions are available when @option{-m3dnowa} is used.
20942 All of them generate the machine instruction that is part of the name.
20945 v2si __builtin_ia32_pf2iw (v2sf)
20946 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
20947 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
20948 v2sf __builtin_ia32_pi2fw (v2si)
20949 v2sf __builtin_ia32_pswapdsf (v2sf)
20950 v2si __builtin_ia32_pswapdsi (v2si)
20953 The following built-in functions are available when @option{-mrtm} is used
20954 They are used for restricted transactional memory. These are the internal
20955 low level functions. Normally the functions in
20956 @ref{x86 transactional memory intrinsics} should be used instead.
20959 int __builtin_ia32_xbegin ()
20960 void __builtin_ia32_xend ()
20961 void __builtin_ia32_xabort (status)
20962 int __builtin_ia32_xtest ()
20965 The following built-in functions are available when @option{-mmwaitx} is used.
20966 All of them generate the machine instruction that is part of the name.
20968 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
20969 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
20972 The following built-in functions are available when @option{-mclzero} is used.
20973 All of them generate the machine instruction that is part of the name.
20975 void __builtin_i32_clzero (void *)
20978 The following built-in functions are available when @option{-mpku} is used.
20979 They generate reads and writes to PKRU.
20981 void __builtin_ia32_wrpkru (unsigned int)
20982 unsigned int __builtin_ia32_rdpkru ()
20985 @node x86 transactional memory intrinsics
20986 @subsection x86 Transactional Memory Intrinsics
20988 These hardware transactional memory intrinsics for x86 allow you to use
20989 memory transactions with RTM (Restricted Transactional Memory).
20990 This support is enabled with the @option{-mrtm} option.
20991 For using HLE (Hardware Lock Elision) see
20992 @ref{x86 specific memory model extensions for transactional memory} instead.
20994 A memory transaction commits all changes to memory in an atomic way,
20995 as visible to other threads. If the transaction fails it is rolled back
20996 and all side effects discarded.
20998 Generally there is no guarantee that a memory transaction ever succeeds
20999 and suitable fallback code always needs to be supplied.
21001 @deftypefn {RTM Function} {unsigned} _xbegin ()
21002 Start a RTM (Restricted Transactional Memory) transaction.
21003 Returns @code{_XBEGIN_STARTED} when the transaction
21004 started successfully (note this is not 0, so the constant has to be
21005 explicitly tested).
21007 If the transaction aborts, all side-effects
21008 are undone and an abort code encoded as a bit mask is returned.
21009 The following macros are defined:
21012 @item _XABORT_EXPLICIT
21013 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
21014 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
21015 @item _XABORT_RETRY
21016 Transaction retry is possible.
21017 @item _XABORT_CONFLICT
21018 Transaction abort due to a memory conflict with another thread.
21019 @item _XABORT_CAPACITY
21020 Transaction abort due to the transaction using too much memory.
21021 @item _XABORT_DEBUG
21022 Transaction abort due to a debug trap.
21023 @item _XABORT_NESTED
21024 Transaction abort in an inner nested transaction.
21027 There is no guarantee
21028 any transaction ever succeeds, so there always needs to be a valid
21032 @deftypefn {RTM Function} {void} _xend ()
21033 Commit the current transaction. When no transaction is active this faults.
21034 All memory side-effects of the transaction become visible
21035 to other threads in an atomic manner.
21038 @deftypefn {RTM Function} {int} _xtest ()
21039 Return a nonzero value if a transaction is currently active, otherwise 0.
21042 @deftypefn {RTM Function} {void} _xabort (status)
21043 Abort the current transaction. When no transaction is active this is a no-op.
21044 The @var{status} is an 8-bit constant; its value is encoded in the return
21045 value from @code{_xbegin}.
21048 Here is an example showing handling for @code{_XABORT_RETRY}
21049 and a fallback path for other failures:
21052 #include <immintrin.h>
21054 int n_tries, max_tries;
21055 unsigned status = _XABORT_EXPLICIT;
21058 for (n_tries = 0; n_tries < max_tries; n_tries++)
21060 status = _xbegin ();
21061 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
21064 if (status == _XBEGIN_STARTED)
21066 ... transaction code...
21071 ... non-transactional fallback path...
21076 Note that, in most cases, the transactional and non-transactional code
21077 must synchronize together to ensure consistency.
21079 @node Target Format Checks
21080 @section Format Checks Specific to Particular Target Machines
21082 For some target machines, GCC supports additional options to the
21084 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
21087 * Solaris Format Checks::
21088 * Darwin Format Checks::
21091 @node Solaris Format Checks
21092 @subsection Solaris Format Checks
21094 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
21095 check. @code{cmn_err} accepts a subset of the standard @code{printf}
21096 conversions, and the two-argument @code{%b} conversion for displaying
21097 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
21099 @node Darwin Format Checks
21100 @subsection Darwin Format Checks
21102 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
21103 attribute context. Declarations made with such attribution are parsed for correct syntax
21104 and format argument types. However, parsing of the format string itself is currently undefined
21105 and is not carried out by this version of the compiler.
21107 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
21108 also be used as format arguments. Note that the relevant headers are only likely to be
21109 available on Darwin (OSX) installations. On such installations, the XCode and system
21110 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
21111 associated functions.
21114 @section Pragmas Accepted by GCC
21116 @cindex @code{#pragma}
21118 GCC supports several types of pragmas, primarily in order to compile
21119 code originally written for other compilers. Note that in general
21120 we do not recommend the use of pragmas; @xref{Function Attributes},
21121 for further explanation.
21124 * AArch64 Pragmas::
21128 * RS/6000 and PowerPC Pragmas::
21131 * Solaris Pragmas::
21132 * Symbol-Renaming Pragmas::
21133 * Structure-Layout Pragmas::
21135 * Diagnostic Pragmas::
21136 * Visibility Pragmas::
21137 * Push/Pop Macro Pragmas::
21138 * Function Specific Option Pragmas::
21139 * Loop-Specific Pragmas::
21142 @node AArch64 Pragmas
21143 @subsection AArch64 Pragmas
21145 The pragmas defined by the AArch64 target correspond to the AArch64
21146 target function attributes. They can be specified as below:
21148 #pragma GCC target("string")
21151 where @code{@var{string}} can be any string accepted as an AArch64 target
21152 attribute. @xref{AArch64 Function Attributes}, for more details
21153 on the permissible values of @code{string}.
21156 @subsection ARM Pragmas
21158 The ARM target defines pragmas for controlling the default addition of
21159 @code{long_call} and @code{short_call} attributes to functions.
21160 @xref{Function Attributes}, for information about the effects of these
21165 @cindex pragma, long_calls
21166 Set all subsequent functions to have the @code{long_call} attribute.
21168 @item no_long_calls
21169 @cindex pragma, no_long_calls
21170 Set all subsequent functions to have the @code{short_call} attribute.
21172 @item long_calls_off
21173 @cindex pragma, long_calls_off
21174 Do not affect the @code{long_call} or @code{short_call} attributes of
21175 subsequent functions.
21179 @subsection M32C Pragmas
21182 @item GCC memregs @var{number}
21183 @cindex pragma, memregs
21184 Overrides the command-line option @code{-memregs=} for the current
21185 file. Use with care! This pragma must be before any function in the
21186 file, and mixing different memregs values in different objects may
21187 make them incompatible. This pragma is useful when a
21188 performance-critical function uses a memreg for temporary values,
21189 as it may allow you to reduce the number of memregs used.
21191 @item ADDRESS @var{name} @var{address}
21192 @cindex pragma, address
21193 For any declared symbols matching @var{name}, this does three things
21194 to that symbol: it forces the symbol to be located at the given
21195 address (a number), it forces the symbol to be volatile, and it
21196 changes the symbol's scope to be static. This pragma exists for
21197 compatibility with other compilers, but note that the common
21198 @code{1234H} numeric syntax is not supported (use @code{0x1234}
21202 #pragma ADDRESS port3 0x103
21209 @subsection MeP Pragmas
21213 @item custom io_volatile (on|off)
21214 @cindex pragma, custom io_volatile
21215 Overrides the command-line option @code{-mio-volatile} for the current
21216 file. Note that for compatibility with future GCC releases, this
21217 option should only be used once before any @code{io} variables in each
21220 @item GCC coprocessor available @var{registers}
21221 @cindex pragma, coprocessor available
21222 Specifies which coprocessor registers are available to the register
21223 allocator. @var{registers} may be a single register, register range
21224 separated by ellipses, or comma-separated list of those. Example:
21227 #pragma GCC coprocessor available $c0...$c10, $c28
21230 @item GCC coprocessor call_saved @var{registers}
21231 @cindex pragma, coprocessor call_saved
21232 Specifies which coprocessor registers are to be saved and restored by
21233 any function using them. @var{registers} may be a single register,
21234 register range separated by ellipses, or comma-separated list of
21238 #pragma GCC coprocessor call_saved $c4...$c6, $c31
21241 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
21242 @cindex pragma, coprocessor subclass
21243 Creates and defines a register class. These register classes can be
21244 used by inline @code{asm} constructs. @var{registers} may be a single
21245 register, register range separated by ellipses, or comma-separated
21246 list of those. Example:
21249 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
21251 asm ("cpfoo %0" : "=B" (x));
21254 @item GCC disinterrupt @var{name} , @var{name} @dots{}
21255 @cindex pragma, disinterrupt
21256 For the named functions, the compiler adds code to disable interrupts
21257 for the duration of those functions. If any functions so named
21258 are not encountered in the source, a warning is emitted that the pragma is
21259 not used. Examples:
21262 #pragma disinterrupt foo
21263 #pragma disinterrupt bar, grill
21264 int foo () @{ @dots{} @}
21267 @item GCC call @var{name} , @var{name} @dots{}
21268 @cindex pragma, call
21269 For the named functions, the compiler always uses a register-indirect
21270 call model when calling the named functions. Examples:
21279 @node RS/6000 and PowerPC Pragmas
21280 @subsection RS/6000 and PowerPC Pragmas
21282 The RS/6000 and PowerPC targets define one pragma for controlling
21283 whether or not the @code{longcall} attribute is added to function
21284 declarations by default. This pragma overrides the @option{-mlongcall}
21285 option, but not the @code{longcall} and @code{shortcall} attributes.
21286 @xref{RS/6000 and PowerPC Options}, for more information about when long
21287 calls are and are not necessary.
21291 @cindex pragma, longcall
21292 Apply the @code{longcall} attribute to all subsequent function
21296 Do not apply the @code{longcall} attribute to subsequent function
21300 @c Describe h8300 pragmas here.
21301 @c Describe sh pragmas here.
21302 @c Describe v850 pragmas here.
21304 @node S/390 Pragmas
21305 @subsection S/390 Pragmas
21307 The pragmas defined by the S/390 target correspond to the S/390
21308 target function attributes and some the additional options:
21315 Note that options of the pragma, unlike options of the target
21316 attribute, do change the value of preprocessor macros like
21317 @code{__VEC__}. They can be specified as below:
21320 #pragma GCC target("string[,string]...")
21321 #pragma GCC target("string"[,"string"]...)
21324 @node Darwin Pragmas
21325 @subsection Darwin Pragmas
21327 The following pragmas are available for all architectures running the
21328 Darwin operating system. These are useful for compatibility with other
21332 @item mark @var{tokens}@dots{}
21333 @cindex pragma, mark
21334 This pragma is accepted, but has no effect.
21336 @item options align=@var{alignment}
21337 @cindex pragma, options align
21338 This pragma sets the alignment of fields in structures. The values of
21339 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
21340 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
21341 properly; to restore the previous setting, use @code{reset} for the
21344 @item segment @var{tokens}@dots{}
21345 @cindex pragma, segment
21346 This pragma is accepted, but has no effect.
21348 @item unused (@var{var} [, @var{var}]@dots{})
21349 @cindex pragma, unused
21350 This pragma declares variables to be possibly unused. GCC does not
21351 produce warnings for the listed variables. The effect is similar to
21352 that of the @code{unused} attribute, except that this pragma may appear
21353 anywhere within the variables' scopes.
21356 @node Solaris Pragmas
21357 @subsection Solaris Pragmas
21359 The Solaris target supports @code{#pragma redefine_extname}
21360 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
21361 @code{#pragma} directives for compatibility with the system compiler.
21364 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
21365 @cindex pragma, align
21367 Increase the minimum alignment of each @var{variable} to @var{alignment}.
21368 This is the same as GCC's @code{aligned} attribute @pxref{Variable
21369 Attributes}). Macro expansion occurs on the arguments to this pragma
21370 when compiling C and Objective-C@. It does not currently occur when
21371 compiling C++, but this is a bug which may be fixed in a future
21374 @item fini (@var{function} [, @var{function}]...)
21375 @cindex pragma, fini
21377 This pragma causes each listed @var{function} to be called after
21378 main, or during shared module unloading, by adding a call to the
21379 @code{.fini} section.
21381 @item init (@var{function} [, @var{function}]...)
21382 @cindex pragma, init
21384 This pragma causes each listed @var{function} to be called during
21385 initialization (before @code{main}) or during shared module loading, by
21386 adding a call to the @code{.init} section.
21390 @node Symbol-Renaming Pragmas
21391 @subsection Symbol-Renaming Pragmas
21393 GCC supports a @code{#pragma} directive that changes the name used in
21394 assembly for a given declaration. While this pragma is supported on all
21395 platforms, it is intended primarily to provide compatibility with the
21396 Solaris system headers. This effect can also be achieved using the asm
21397 labels extension (@pxref{Asm Labels}).
21400 @item redefine_extname @var{oldname} @var{newname}
21401 @cindex pragma, redefine_extname
21403 This pragma gives the C function @var{oldname} the assembly symbol
21404 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
21405 is defined if this pragma is available (currently on all platforms).
21408 This pragma and the asm labels extension interact in a complicated
21409 manner. Here are some corner cases you may want to be aware of:
21412 @item This pragma silently applies only to declarations with external
21413 linkage. Asm labels do not have this restriction.
21415 @item In C++, this pragma silently applies only to declarations with
21416 ``C'' linkage. Again, asm labels do not have this restriction.
21418 @item If either of the ways of changing the assembly name of a
21419 declaration are applied to a declaration whose assembly name has
21420 already been determined (either by a previous use of one of these
21421 features, or because the compiler needed the assembly name in order to
21422 generate code), and the new name is different, a warning issues and
21423 the name does not change.
21425 @item The @var{oldname} used by @code{#pragma redefine_extname} is
21426 always the C-language name.
21429 @node Structure-Layout Pragmas
21430 @subsection Structure-Layout Pragmas
21432 For compatibility with Microsoft Windows compilers, GCC supports a
21433 set of @code{#pragma} directives that change the maximum alignment of
21434 members of structures (other than zero-width bit-fields), unions, and
21435 classes subsequently defined. The @var{n} value below always is required
21436 to be a small power of two and specifies the new alignment in bytes.
21439 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
21440 @item @code{#pragma pack()} sets the alignment to the one that was in
21441 effect when compilation started (see also command-line option
21442 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
21443 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
21444 setting on an internal stack and then optionally sets the new alignment.
21445 @item @code{#pragma pack(pop)} restores the alignment setting to the one
21446 saved at the top of the internal stack (and removes that stack entry).
21447 Note that @code{#pragma pack([@var{n}])} does not influence this internal
21448 stack; thus it is possible to have @code{#pragma pack(push)} followed by
21449 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
21450 @code{#pragma pack(pop)}.
21453 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
21454 directive which lays out structures and unions subsequently defined as the
21455 documented @code{__attribute__ ((ms_struct))}.
21458 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
21459 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
21460 @item @code{#pragma ms_struct reset} goes back to the default layout.
21463 Most targets also support the @code{#pragma scalar_storage_order} directive
21464 which lays out structures and unions subsequently defined as the documented
21465 @code{__attribute__ ((scalar_storage_order))}.
21468 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
21469 of the scalar fields to big-endian.
21470 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
21471 of the scalar fields to little-endian.
21472 @item @code{#pragma scalar_storage_order default} goes back to the endianness
21473 that was in effect when compilation started (see also command-line option
21474 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
21478 @subsection Weak Pragmas
21480 For compatibility with SVR4, GCC supports a set of @code{#pragma}
21481 directives for declaring symbols to be weak, and defining weak
21485 @item #pragma weak @var{symbol}
21486 @cindex pragma, weak
21487 This pragma declares @var{symbol} to be weak, as if the declaration
21488 had the attribute of the same name. The pragma may appear before
21489 or after the declaration of @var{symbol}. It is not an error for
21490 @var{symbol} to never be defined at all.
21492 @item #pragma weak @var{symbol1} = @var{symbol2}
21493 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
21494 It is an error if @var{symbol2} is not defined in the current
21498 @node Diagnostic Pragmas
21499 @subsection Diagnostic Pragmas
21501 GCC allows the user to selectively enable or disable certain types of
21502 diagnostics, and change the kind of the diagnostic. For example, a
21503 project's policy might require that all sources compile with
21504 @option{-Werror} but certain files might have exceptions allowing
21505 specific types of warnings. Or, a project might selectively enable
21506 diagnostics and treat them as errors depending on which preprocessor
21507 macros are defined.
21510 @item #pragma GCC diagnostic @var{kind} @var{option}
21511 @cindex pragma, diagnostic
21513 Modifies the disposition of a diagnostic. Note that not all
21514 diagnostics are modifiable; at the moment only warnings (normally
21515 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
21516 Use @option{-fdiagnostics-show-option} to determine which diagnostics
21517 are controllable and which option controls them.
21519 @var{kind} is @samp{error} to treat this diagnostic as an error,
21520 @samp{warning} to treat it like a warning (even if @option{-Werror} is
21521 in effect), or @samp{ignored} if the diagnostic is to be ignored.
21522 @var{option} is a double quoted string that matches the command-line
21526 #pragma GCC diagnostic warning "-Wformat"
21527 #pragma GCC diagnostic error "-Wformat"
21528 #pragma GCC diagnostic ignored "-Wformat"
21531 Note that these pragmas override any command-line options. GCC keeps
21532 track of the location of each pragma, and issues diagnostics according
21533 to the state as of that point in the source file. Thus, pragmas occurring
21534 after a line do not affect diagnostics caused by that line.
21536 @item #pragma GCC diagnostic push
21537 @itemx #pragma GCC diagnostic pop
21539 Causes GCC to remember the state of the diagnostics as of each
21540 @code{push}, and restore to that point at each @code{pop}. If a
21541 @code{pop} has no matching @code{push}, the command-line options are
21545 #pragma GCC diagnostic error "-Wuninitialized"
21546 foo(a); /* error is given for this one */
21547 #pragma GCC diagnostic push
21548 #pragma GCC diagnostic ignored "-Wuninitialized"
21549 foo(b); /* no diagnostic for this one */
21550 #pragma GCC diagnostic pop
21551 foo(c); /* error is given for this one */
21552 #pragma GCC diagnostic pop
21553 foo(d); /* depends on command-line options */
21558 GCC also offers a simple mechanism for printing messages during
21562 @item #pragma message @var{string}
21563 @cindex pragma, diagnostic
21565 Prints @var{string} as a compiler message on compilation. The message
21566 is informational only, and is neither a compilation warning nor an error.
21569 #pragma message "Compiling " __FILE__ "..."
21572 @var{string} may be parenthesized, and is printed with location
21573 information. For example,
21576 #define DO_PRAGMA(x) _Pragma (#x)
21577 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
21579 TODO(Remember to fix this)
21583 prints @samp{/tmp/file.c:4: note: #pragma message:
21584 TODO - Remember to fix this}.
21588 @node Visibility Pragmas
21589 @subsection Visibility Pragmas
21592 @item #pragma GCC visibility push(@var{visibility})
21593 @itemx #pragma GCC visibility pop
21594 @cindex pragma, visibility
21596 This pragma allows the user to set the visibility for multiple
21597 declarations without having to give each a visibility attribute
21598 (@pxref{Function Attributes}).
21600 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
21601 declarations. Class members and template specializations are not
21602 affected; if you want to override the visibility for a particular
21603 member or instantiation, you must use an attribute.
21608 @node Push/Pop Macro Pragmas
21609 @subsection Push/Pop Macro Pragmas
21611 For compatibility with Microsoft Windows compilers, GCC supports
21612 @samp{#pragma push_macro(@var{"macro_name"})}
21613 and @samp{#pragma pop_macro(@var{"macro_name"})}.
21616 @item #pragma push_macro(@var{"macro_name"})
21617 @cindex pragma, push_macro
21618 This pragma saves the value of the macro named as @var{macro_name} to
21619 the top of the stack for this macro.
21621 @item #pragma pop_macro(@var{"macro_name"})
21622 @cindex pragma, pop_macro
21623 This pragma sets the value of the macro named as @var{macro_name} to
21624 the value on top of the stack for this macro. If the stack for
21625 @var{macro_name} is empty, the value of the macro remains unchanged.
21632 #pragma push_macro("X")
21635 #pragma pop_macro("X")
21640 In this example, the definition of X as 1 is saved by @code{#pragma
21641 push_macro} and restored by @code{#pragma pop_macro}.
21643 @node Function Specific Option Pragmas
21644 @subsection Function Specific Option Pragmas
21647 @item #pragma GCC target (@var{"string"}...)
21648 @cindex pragma GCC target
21650 This pragma allows you to set target specific options for functions
21651 defined later in the source file. One or more strings can be
21652 specified. Each function that is defined after this point is as
21653 if @code{attribute((target("STRING")))} was specified for that
21654 function. The parenthesis around the options is optional.
21655 @xref{Function Attributes}, for more information about the
21656 @code{target} attribute and the attribute syntax.
21658 The @code{#pragma GCC target} pragma is presently implemented for
21659 x86, PowerPC, and Nios II targets only.
21663 @item #pragma GCC optimize (@var{"string"}...)
21664 @cindex pragma GCC optimize
21666 This pragma allows you to set global optimization options for functions
21667 defined later in the source file. One or more strings can be
21668 specified. Each function that is defined after this point is as
21669 if @code{attribute((optimize("STRING")))} was specified for that
21670 function. The parenthesis around the options is optional.
21671 @xref{Function Attributes}, for more information about the
21672 @code{optimize} attribute and the attribute syntax.
21676 @item #pragma GCC push_options
21677 @itemx #pragma GCC pop_options
21678 @cindex pragma GCC push_options
21679 @cindex pragma GCC pop_options
21681 These pragmas maintain a stack of the current target and optimization
21682 options. It is intended for include files where you temporarily want
21683 to switch to using a different @samp{#pragma GCC target} or
21684 @samp{#pragma GCC optimize} and then to pop back to the previous
21689 @item #pragma GCC reset_options
21690 @cindex pragma GCC reset_options
21692 This pragma clears the current @code{#pragma GCC target} and
21693 @code{#pragma GCC optimize} to use the default switches as specified
21694 on the command line.
21697 @node Loop-Specific Pragmas
21698 @subsection Loop-Specific Pragmas
21701 @item #pragma GCC ivdep
21702 @cindex pragma GCC ivdep
21705 With this pragma, the programmer asserts that there are no loop-carried
21706 dependencies which would prevent consecutive iterations of
21707 the following loop from executing concurrently with SIMD
21708 (single instruction multiple data) instructions.
21710 For example, the compiler can only unconditionally vectorize the following
21711 loop with the pragma:
21714 void foo (int n, int *a, int *b, int *c)
21718 for (i = 0; i < n; ++i)
21719 a[i] = b[i] + c[i];
21724 In this example, using the @code{restrict} qualifier had the same
21725 effect. In the following example, that would not be possible. Assume
21726 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
21727 that it can unconditionally vectorize the following loop:
21730 void ignore_vec_dep (int *a, int k, int c, int m)
21733 for (int i = 0; i < m; i++)
21734 a[i] = a[i + k] * c;
21739 @node Unnamed Fields
21740 @section Unnamed Structure and Union Fields
21741 @cindex @code{struct}
21742 @cindex @code{union}
21744 As permitted by ISO C11 and for compatibility with other compilers,
21745 GCC allows you to define
21746 a structure or union that contains, as fields, structures and unions
21747 without names. For example:
21761 In this example, you are able to access members of the unnamed
21762 union with code like @samp{foo.b}. Note that only unnamed structs and
21763 unions are allowed, you may not have, for example, an unnamed
21766 You must never create such structures that cause ambiguous field definitions.
21767 For example, in this structure:
21779 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
21780 The compiler gives errors for such constructs.
21782 @opindex fms-extensions
21783 Unless @option{-fms-extensions} is used, the unnamed field must be a
21784 structure or union definition without a tag (for example, @samp{struct
21785 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
21786 also be a definition with a tag such as @samp{struct foo @{ int a;
21787 @};}, a reference to a previously defined structure or union such as
21788 @samp{struct foo;}, or a reference to a @code{typedef} name for a
21789 previously defined structure or union type.
21791 @opindex fplan9-extensions
21792 The option @option{-fplan9-extensions} enables
21793 @option{-fms-extensions} as well as two other extensions. First, a
21794 pointer to a structure is automatically converted to a pointer to an
21795 anonymous field for assignments and function calls. For example:
21798 struct s1 @{ int a; @};
21799 struct s2 @{ struct s1; @};
21800 extern void f1 (struct s1 *);
21801 void f2 (struct s2 *p) @{ f1 (p); @}
21805 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
21806 converted into a pointer to the anonymous field.
21808 Second, when the type of an anonymous field is a @code{typedef} for a
21809 @code{struct} or @code{union}, code may refer to the field using the
21810 name of the @code{typedef}.
21813 typedef struct @{ int a; @} s1;
21814 struct s2 @{ s1; @};
21815 s1 f1 (struct s2 *p) @{ return p->s1; @}
21818 These usages are only permitted when they are not ambiguous.
21821 @section Thread-Local Storage
21822 @cindex Thread-Local Storage
21823 @cindex @acronym{TLS}
21824 @cindex @code{__thread}
21826 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
21827 are allocated such that there is one instance of the variable per extant
21828 thread. The runtime model GCC uses to implement this originates
21829 in the IA-64 processor-specific ABI, but has since been migrated
21830 to other processors as well. It requires significant support from
21831 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
21832 system libraries (@file{libc.so} and @file{libpthread.so}), so it
21833 is not available everywhere.
21835 At the user level, the extension is visible with a new storage
21836 class keyword: @code{__thread}. For example:
21840 extern __thread struct state s;
21841 static __thread char *p;
21844 The @code{__thread} specifier may be used alone, with the @code{extern}
21845 or @code{static} specifiers, but with no other storage class specifier.
21846 When used with @code{extern} or @code{static}, @code{__thread} must appear
21847 immediately after the other storage class specifier.
21849 The @code{__thread} specifier may be applied to any global, file-scoped
21850 static, function-scoped static, or static data member of a class. It may
21851 not be applied to block-scoped automatic or non-static data member.
21853 When the address-of operator is applied to a thread-local variable, it is
21854 evaluated at run time and returns the address of the current thread's
21855 instance of that variable. An address so obtained may be used by any
21856 thread. When a thread terminates, any pointers to thread-local variables
21857 in that thread become invalid.
21859 No static initialization may refer to the address of a thread-local variable.
21861 In C++, if an initializer is present for a thread-local variable, it must
21862 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
21865 See @uref{https://www.akkadia.org/drepper/tls.pdf,
21866 ELF Handling For Thread-Local Storage} for a detailed explanation of
21867 the four thread-local storage addressing models, and how the runtime
21868 is expected to function.
21871 * C99 Thread-Local Edits::
21872 * C++98 Thread-Local Edits::
21875 @node C99 Thread-Local Edits
21876 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
21878 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
21879 that document the exact semantics of the language extension.
21883 @cite{5.1.2 Execution environments}
21885 Add new text after paragraph 1
21888 Within either execution environment, a @dfn{thread} is a flow of
21889 control within a program. It is implementation defined whether
21890 or not there may be more than one thread associated with a program.
21891 It is implementation defined how threads beyond the first are
21892 created, the name and type of the function called at thread
21893 startup, and how threads may be terminated. However, objects
21894 with thread storage duration shall be initialized before thread
21899 @cite{6.2.4 Storage durations of objects}
21901 Add new text before paragraph 3
21904 An object whose identifier is declared with the storage-class
21905 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
21906 Its lifetime is the entire execution of the thread, and its
21907 stored value is initialized only once, prior to thread startup.
21911 @cite{6.4.1 Keywords}
21913 Add @code{__thread}.
21916 @cite{6.7.1 Storage-class specifiers}
21918 Add @code{__thread} to the list of storage class specifiers in
21921 Change paragraph 2 to
21924 With the exception of @code{__thread}, at most one storage-class
21925 specifier may be given [@dots{}]. The @code{__thread} specifier may
21926 be used alone, or immediately following @code{extern} or
21930 Add new text after paragraph 6
21933 The declaration of an identifier for a variable that has
21934 block scope that specifies @code{__thread} shall also
21935 specify either @code{extern} or @code{static}.
21937 The @code{__thread} specifier shall be used only with
21942 @node C++98 Thread-Local Edits
21943 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
21945 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
21946 that document the exact semantics of the language extension.
21950 @b{[intro.execution]}
21952 New text after paragraph 4
21955 A @dfn{thread} is a flow of control within the abstract machine.
21956 It is implementation defined whether or not there may be more than
21960 New text after paragraph 7
21963 It is unspecified whether additional action must be taken to
21964 ensure when and whether side effects are visible to other threads.
21970 Add @code{__thread}.
21973 @b{[basic.start.main]}
21975 Add after paragraph 5
21978 The thread that begins execution at the @code{main} function is called
21979 the @dfn{main thread}. It is implementation defined how functions
21980 beginning threads other than the main thread are designated or typed.
21981 A function so designated, as well as the @code{main} function, is called
21982 a @dfn{thread startup function}. It is implementation defined what
21983 happens if a thread startup function returns. It is implementation
21984 defined what happens to other threads when any thread calls @code{exit}.
21988 @b{[basic.start.init]}
21990 Add after paragraph 4
21993 The storage for an object of thread storage duration shall be
21994 statically initialized before the first statement of the thread startup
21995 function. An object of thread storage duration shall not require
21996 dynamic initialization.
22000 @b{[basic.start.term]}
22002 Add after paragraph 3
22005 The type of an object with thread storage duration shall not have a
22006 non-trivial destructor, nor shall it be an array type whose elements
22007 (directly or indirectly) have non-trivial destructors.
22013 Add ``thread storage duration'' to the list in paragraph 1.
22018 Thread, static, and automatic storage durations are associated with
22019 objects introduced by declarations [@dots{}].
22022 Add @code{__thread} to the list of specifiers in paragraph 3.
22025 @b{[basic.stc.thread]}
22027 New section before @b{[basic.stc.static]}
22030 The keyword @code{__thread} applied to a non-local object gives the
22031 object thread storage duration.
22033 A local variable or class data member declared both @code{static}
22034 and @code{__thread} gives the variable or member thread storage
22039 @b{[basic.stc.static]}
22044 All objects that have neither thread storage duration, dynamic
22045 storage duration nor are local [@dots{}].
22051 Add @code{__thread} to the list in paragraph 1.
22056 With the exception of @code{__thread}, at most one
22057 @var{storage-class-specifier} shall appear in a given
22058 @var{decl-specifier-seq}. The @code{__thread} specifier may
22059 be used alone, or immediately following the @code{extern} or
22060 @code{static} specifiers. [@dots{}]
22063 Add after paragraph 5
22066 The @code{__thread} specifier can be applied only to the names of objects
22067 and to anonymous unions.
22073 Add after paragraph 6
22076 Non-@code{static} members shall not be @code{__thread}.
22080 @node Binary constants
22081 @section Binary Constants using the @samp{0b} Prefix
22082 @cindex Binary constants using the @samp{0b} prefix
22084 Integer constants can be written as binary constants, consisting of a
22085 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
22086 @samp{0B}. This is particularly useful in environments that operate a
22087 lot on the bit level (like microcontrollers).
22089 The following statements are identical:
22098 The type of these constants follows the same rules as for octal or
22099 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
22102 @node C++ Extensions
22103 @chapter Extensions to the C++ Language
22104 @cindex extensions, C++ language
22105 @cindex C++ language extensions
22107 The GNU compiler provides these extensions to the C++ language (and you
22108 can also use most of the C language extensions in your C++ programs). If you
22109 want to write code that checks whether these features are available, you can
22110 test for the GNU compiler the same way as for C programs: check for a
22111 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
22112 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
22113 Predefined Macros,cpp,The GNU C Preprocessor}).
22116 * C++ Volatiles:: What constitutes an access to a volatile object.
22117 * Restricted Pointers:: C99 restricted pointers and references.
22118 * Vague Linkage:: Where G++ puts inlines, vtables and such.
22119 * C++ Interface:: You can use a single C++ header file for both
22120 declarations and definitions.
22121 * Template Instantiation:: Methods for ensuring that exactly one copy of
22122 each needed template instantiation is emitted.
22123 * Bound member functions:: You can extract a function pointer to the
22124 method denoted by a @samp{->*} or @samp{.*} expression.
22125 * C++ Attributes:: Variable, function, and type attributes for C++ only.
22126 * Function Multiversioning:: Declaring multiple function versions.
22127 * Type Traits:: Compiler support for type traits.
22128 * C++ Concepts:: Improved support for generic programming.
22129 * Deprecated Features:: Things will disappear from G++.
22130 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
22133 @node C++ Volatiles
22134 @section When is a Volatile C++ Object Accessed?
22135 @cindex accessing volatiles
22136 @cindex volatile read
22137 @cindex volatile write
22138 @cindex volatile access
22140 The C++ standard differs from the C standard in its treatment of
22141 volatile objects. It fails to specify what constitutes a volatile
22142 access, except to say that C++ should behave in a similar manner to C
22143 with respect to volatiles, where possible. However, the different
22144 lvalueness of expressions between C and C++ complicate the behavior.
22145 G++ behaves the same as GCC for volatile access, @xref{C
22146 Extensions,,Volatiles}, for a description of GCC's behavior.
22148 The C and C++ language specifications differ when an object is
22149 accessed in a void context:
22152 volatile int *src = @var{somevalue};
22156 The C++ standard specifies that such expressions do not undergo lvalue
22157 to rvalue conversion, and that the type of the dereferenced object may
22158 be incomplete. The C++ standard does not specify explicitly that it
22159 is lvalue to rvalue conversion that is responsible for causing an
22160 access. There is reason to believe that it is, because otherwise
22161 certain simple expressions become undefined. However, because it
22162 would surprise most programmers, G++ treats dereferencing a pointer to
22163 volatile object of complete type as GCC would do for an equivalent
22164 type in C@. When the object has incomplete type, G++ issues a
22165 warning; if you wish to force an error, you must force a conversion to
22166 rvalue with, for instance, a static cast.
22168 When using a reference to volatile, G++ does not treat equivalent
22169 expressions as accesses to volatiles, but instead issues a warning that
22170 no volatile is accessed. The rationale for this is that otherwise it
22171 becomes difficult to determine where volatile access occur, and not
22172 possible to ignore the return value from functions returning volatile
22173 references. Again, if you wish to force a read, cast the reference to
22176 G++ implements the same behavior as GCC does when assigning to a
22177 volatile object---there is no reread of the assigned-to object, the
22178 assigned rvalue is reused. Note that in C++ assignment expressions
22179 are lvalues, and if used as an lvalue, the volatile object is
22180 referred to. For instance, @var{vref} refers to @var{vobj}, as
22181 expected, in the following example:
22185 volatile int &vref = vobj = @var{something};
22188 @node Restricted Pointers
22189 @section Restricting Pointer Aliasing
22190 @cindex restricted pointers
22191 @cindex restricted references
22192 @cindex restricted this pointer
22194 As with the C front end, G++ understands the C99 feature of restricted pointers,
22195 specified with the @code{__restrict__}, or @code{__restrict} type
22196 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
22197 language flag, @code{restrict} is not a keyword in C++.
22199 In addition to allowing restricted pointers, you can specify restricted
22200 references, which indicate that the reference is not aliased in the local
22204 void fn (int *__restrict__ rptr, int &__restrict__ rref)
22211 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
22212 @var{rref} refers to a (different) unaliased integer.
22214 You may also specify whether a member function's @var{this} pointer is
22215 unaliased by using @code{__restrict__} as a member function qualifier.
22218 void T::fn () __restrict__
22225 Within the body of @code{T::fn}, @var{this} has the effective
22226 definition @code{T *__restrict__ const this}. Notice that the
22227 interpretation of a @code{__restrict__} member function qualifier is
22228 different to that of @code{const} or @code{volatile} qualifier, in that it
22229 is applied to the pointer rather than the object. This is consistent with
22230 other compilers that implement restricted pointers.
22232 As with all outermost parameter qualifiers, @code{__restrict__} is
22233 ignored in function definition matching. This means you only need to
22234 specify @code{__restrict__} in a function definition, rather than
22235 in a function prototype as well.
22237 @node Vague Linkage
22238 @section Vague Linkage
22239 @cindex vague linkage
22241 There are several constructs in C++ that require space in the object
22242 file but are not clearly tied to a single translation unit. We say that
22243 these constructs have ``vague linkage''. Typically such constructs are
22244 emitted wherever they are needed, though sometimes we can be more
22248 @item Inline Functions
22249 Inline functions are typically defined in a header file which can be
22250 included in many different compilations. Hopefully they can usually be
22251 inlined, but sometimes an out-of-line copy is necessary, if the address
22252 of the function is taken or if inlining fails. In general, we emit an
22253 out-of-line copy in all translation units where one is needed. As an
22254 exception, we only emit inline virtual functions with the vtable, since
22255 it always requires a copy.
22257 Local static variables and string constants used in an inline function
22258 are also considered to have vague linkage, since they must be shared
22259 between all inlined and out-of-line instances of the function.
22263 C++ virtual functions are implemented in most compilers using a lookup
22264 table, known as a vtable. The vtable contains pointers to the virtual
22265 functions provided by a class, and each object of the class contains a
22266 pointer to its vtable (or vtables, in some multiple-inheritance
22267 situations). If the class declares any non-inline, non-pure virtual
22268 functions, the first one is chosen as the ``key method'' for the class,
22269 and the vtable is only emitted in the translation unit where the key
22272 @emph{Note:} If the chosen key method is later defined as inline, the
22273 vtable is still emitted in every translation unit that defines it.
22274 Make sure that any inline virtuals are declared inline in the class
22275 body, even if they are not defined there.
22277 @item @code{type_info} objects
22278 @cindex @code{type_info}
22280 C++ requires information about types to be written out in order to
22281 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
22282 For polymorphic classes (classes with virtual functions), the @samp{type_info}
22283 object is written out along with the vtable so that @samp{dynamic_cast}
22284 can determine the dynamic type of a class object at run time. For all
22285 other types, we write out the @samp{type_info} object when it is used: when
22286 applying @samp{typeid} to an expression, throwing an object, or
22287 referring to a type in a catch clause or exception specification.
22289 @item Template Instantiations
22290 Most everything in this section also applies to template instantiations,
22291 but there are other options as well.
22292 @xref{Template Instantiation,,Where's the Template?}.
22296 When used with GNU ld version 2.8 or later on an ELF system such as
22297 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
22298 these constructs will be discarded at link time. This is known as
22301 On targets that don't support COMDAT, but do support weak symbols, GCC
22302 uses them. This way one copy overrides all the others, but
22303 the unused copies still take up space in the executable.
22305 For targets that do not support either COMDAT or weak symbols,
22306 most entities with vague linkage are emitted as local symbols to
22307 avoid duplicate definition errors from the linker. This does not happen
22308 for local statics in inlines, however, as having multiple copies
22309 almost certainly breaks things.
22311 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
22312 another way to control placement of these constructs.
22314 @node C++ Interface
22315 @section C++ Interface and Implementation Pragmas
22317 @cindex interface and implementation headers, C++
22318 @cindex C++ interface and implementation headers
22319 @cindex pragmas, interface and implementation
22321 @code{#pragma interface} and @code{#pragma implementation} provide the
22322 user with a way of explicitly directing the compiler to emit entities
22323 with vague linkage (and debugging information) in a particular
22326 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
22327 by COMDAT support and the ``key method'' heuristic
22328 mentioned in @ref{Vague Linkage}. Using them can actually cause your
22329 program to grow due to unnecessary out-of-line copies of inline
22333 @item #pragma interface
22334 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
22335 @kindex #pragma interface
22336 Use this directive in @emph{header files} that define object classes, to save
22337 space in most of the object files that use those classes. Normally,
22338 local copies of certain information (backup copies of inline member
22339 functions, debugging information, and the internal tables that implement
22340 virtual functions) must be kept in each object file that includes class
22341 definitions. You can use this pragma to avoid such duplication. When a
22342 header file containing @samp{#pragma interface} is included in a
22343 compilation, this auxiliary information is not generated (unless
22344 the main input source file itself uses @samp{#pragma implementation}).
22345 Instead, the object files contain references to be resolved at link
22348 The second form of this directive is useful for the case where you have
22349 multiple headers with the same name in different directories. If you
22350 use this form, you must specify the same string to @samp{#pragma
22353 @item #pragma implementation
22354 @itemx #pragma implementation "@var{objects}.h"
22355 @kindex #pragma implementation
22356 Use this pragma in a @emph{main input file}, when you want full output from
22357 included header files to be generated (and made globally visible). The
22358 included header file, in turn, should use @samp{#pragma interface}.
22359 Backup copies of inline member functions, debugging information, and the
22360 internal tables used to implement virtual functions are all generated in
22361 implementation files.
22363 @cindex implied @code{#pragma implementation}
22364 @cindex @code{#pragma implementation}, implied
22365 @cindex naming convention, implementation headers
22366 If you use @samp{#pragma implementation} with no argument, it applies to
22367 an include file with the same basename@footnote{A file's @dfn{basename}
22368 is the name stripped of all leading path information and of trailing
22369 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
22370 file. For example, in @file{allclass.cc}, giving just
22371 @samp{#pragma implementation}
22372 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
22374 Use the string argument if you want a single implementation file to
22375 include code from multiple header files. (You must also use
22376 @samp{#include} to include the header file; @samp{#pragma
22377 implementation} only specifies how to use the file---it doesn't actually
22380 There is no way to split up the contents of a single header file into
22381 multiple implementation files.
22384 @cindex inlining and C++ pragmas
22385 @cindex C++ pragmas, effect on inlining
22386 @cindex pragmas in C++, effect on inlining
22387 @samp{#pragma implementation} and @samp{#pragma interface} also have an
22388 effect on function inlining.
22390 If you define a class in a header file marked with @samp{#pragma
22391 interface}, the effect on an inline function defined in that class is
22392 similar to an explicit @code{extern} declaration---the compiler emits
22393 no code at all to define an independent version of the function. Its
22394 definition is used only for inlining with its callers.
22396 @opindex fno-implement-inlines
22397 Conversely, when you include the same header file in a main source file
22398 that declares it as @samp{#pragma implementation}, the compiler emits
22399 code for the function itself; this defines a version of the function
22400 that can be found via pointers (or by callers compiled without
22401 inlining). If all calls to the function can be inlined, you can avoid
22402 emitting the function by compiling with @option{-fno-implement-inlines}.
22403 If any calls are not inlined, you will get linker errors.
22405 @node Template Instantiation
22406 @section Where's the Template?
22407 @cindex template instantiation
22409 C++ templates were the first language feature to require more
22410 intelligence from the environment than was traditionally found on a UNIX
22411 system. Somehow the compiler and linker have to make sure that each
22412 template instance occurs exactly once in the executable if it is needed,
22413 and not at all otherwise. There are two basic approaches to this
22414 problem, which are referred to as the Borland model and the Cfront model.
22417 @item Borland model
22418 Borland C++ solved the template instantiation problem by adding the code
22419 equivalent of common blocks to their linker; the compiler emits template
22420 instances in each translation unit that uses them, and the linker
22421 collapses them together. The advantage of this model is that the linker
22422 only has to consider the object files themselves; there is no external
22423 complexity to worry about. The disadvantage is that compilation time
22424 is increased because the template code is being compiled repeatedly.
22425 Code written for this model tends to include definitions of all
22426 templates in the header file, since they must be seen to be
22430 The AT&T C++ translator, Cfront, solved the template instantiation
22431 problem by creating the notion of a template repository, an
22432 automatically maintained place where template instances are stored. A
22433 more modern version of the repository works as follows: As individual
22434 object files are built, the compiler places any template definitions and
22435 instantiations encountered in the repository. At link time, the link
22436 wrapper adds in the objects in the repository and compiles any needed
22437 instances that were not previously emitted. The advantages of this
22438 model are more optimal compilation speed and the ability to use the
22439 system linker; to implement the Borland model a compiler vendor also
22440 needs to replace the linker. The disadvantages are vastly increased
22441 complexity, and thus potential for error; for some code this can be
22442 just as transparent, but in practice it can been very difficult to build
22443 multiple programs in one directory and one program in multiple
22444 directories. Code written for this model tends to separate definitions
22445 of non-inline member templates into a separate file, which should be
22446 compiled separately.
22449 G++ implements the Borland model on targets where the linker supports it,
22450 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
22451 Otherwise G++ implements neither automatic model.
22453 You have the following options for dealing with template instantiations:
22457 Do nothing. Code written for the Borland model works fine, but
22458 each translation unit contains instances of each of the templates it
22459 uses. The duplicate instances will be discarded by the linker, but in
22460 a large program, this can lead to an unacceptable amount of code
22461 duplication in object files or shared libraries.
22463 Duplicate instances of a template can be avoided by defining an explicit
22464 instantiation in one object file, and preventing the compiler from doing
22465 implicit instantiations in any other object files by using an explicit
22466 instantiation declaration, using the @code{extern template} syntax:
22469 extern template int max (int, int);
22472 This syntax is defined in the C++ 2011 standard, but has been supported by
22473 G++ and other compilers since well before 2011.
22475 Explicit instantiations can be used for the largest or most frequently
22476 duplicated instances, without having to know exactly which other instances
22477 are used in the rest of the program. You can scatter the explicit
22478 instantiations throughout your program, perhaps putting them in the
22479 translation units where the instances are used or the translation units
22480 that define the templates themselves; you can put all of the explicit
22481 instantiations you need into one big file; or you can create small files
22488 template class Foo<int>;
22489 template ostream& operator <<
22490 (ostream&, const Foo<int>&);
22494 for each of the instances you need, and create a template instantiation
22495 library from those.
22497 This is the simplest option, but also offers flexibility and
22498 fine-grained control when necessary. It is also the most portable
22499 alternative and programs using this approach will work with most modern
22504 Compile your template-using code with @option{-frepo}. The compiler
22505 generates files with the extension @samp{.rpo} listing all of the
22506 template instantiations used in the corresponding object files that
22507 could be instantiated there; the link wrapper, @samp{collect2},
22508 then updates the @samp{.rpo} files to tell the compiler where to place
22509 those instantiations and rebuild any affected object files. The
22510 link-time overhead is negligible after the first pass, as the compiler
22511 continues to place the instantiations in the same files.
22513 This can be a suitable option for application code written for the Borland
22514 model, as it usually just works. Code written for the Cfront model
22515 needs to be modified so that the template definitions are available at
22516 one or more points of instantiation; usually this is as simple as adding
22517 @code{#include <tmethods.cc>} to the end of each template header.
22519 For library code, if you want the library to provide all of the template
22520 instantiations it needs, just try to link all of its object files
22521 together; the link will fail, but cause the instantiations to be
22522 generated as a side effect. Be warned, however, that this may cause
22523 conflicts if multiple libraries try to provide the same instantiations.
22524 For greater control, use explicit instantiation as described in the next
22528 @opindex fno-implicit-templates
22529 Compile your code with @option{-fno-implicit-templates} to disable the
22530 implicit generation of template instances, and explicitly instantiate
22531 all the ones you use. This approach requires more knowledge of exactly
22532 which instances you need than do the others, but it's less
22533 mysterious and allows greater control if you want to ensure that only
22534 the intended instances are used.
22536 If you are using Cfront-model code, you can probably get away with not
22537 using @option{-fno-implicit-templates} when compiling files that don't
22538 @samp{#include} the member template definitions.
22540 If you use one big file to do the instantiations, you may want to
22541 compile it without @option{-fno-implicit-templates} so you get all of the
22542 instances required by your explicit instantiations (but not by any
22543 other files) without having to specify them as well.
22545 In addition to forward declaration of explicit instantiations
22546 (with @code{extern}), G++ has extended the template instantiation
22547 syntax to support instantiation of the compiler support data for a
22548 template class (i.e.@: the vtable) without instantiating any of its
22549 members (with @code{inline}), and instantiation of only the static data
22550 members of a template class, without the support data or member
22551 functions (with @code{static}):
22554 inline template class Foo<int>;
22555 static template class Foo<int>;
22559 @node Bound member functions
22560 @section Extracting the Function Pointer from a Bound Pointer to Member Function
22562 @cindex pointer to member function
22563 @cindex bound pointer to member function
22565 In C++, pointer to member functions (PMFs) are implemented using a wide
22566 pointer of sorts to handle all the possible call mechanisms; the PMF
22567 needs to store information about how to adjust the @samp{this} pointer,
22568 and if the function pointed to is virtual, where to find the vtable, and
22569 where in the vtable to look for the member function. If you are using
22570 PMFs in an inner loop, you should really reconsider that decision. If
22571 that is not an option, you can extract the pointer to the function that
22572 would be called for a given object/PMF pair and call it directly inside
22573 the inner loop, to save a bit of time.
22575 Note that you still pay the penalty for the call through a
22576 function pointer; on most modern architectures, such a call defeats the
22577 branch prediction features of the CPU@. This is also true of normal
22578 virtual function calls.
22580 The syntax for this extension is
22584 extern int (A::*fp)();
22585 typedef int (*fptr)(A *);
22587 fptr p = (fptr)(a.*fp);
22590 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
22591 no object is needed to obtain the address of the function. They can be
22592 converted to function pointers directly:
22595 fptr p1 = (fptr)(&A::foo);
22598 @opindex Wno-pmf-conversions
22599 You must specify @option{-Wno-pmf-conversions} to use this extension.
22601 @node C++ Attributes
22602 @section C++-Specific Variable, Function, and Type Attributes
22604 Some attributes only make sense for C++ programs.
22607 @item abi_tag ("@var{tag}", ...)
22608 @cindex @code{abi_tag} function attribute
22609 @cindex @code{abi_tag} variable attribute
22610 @cindex @code{abi_tag} type attribute
22611 The @code{abi_tag} attribute can be applied to a function, variable, or class
22612 declaration. It modifies the mangled name of the entity to
22613 incorporate the tag name, in order to distinguish the function or
22614 class from an earlier version with a different ABI; perhaps the class
22615 has changed size, or the function has a different return type that is
22616 not encoded in the mangled name.
22618 The attribute can also be applied to an inline namespace, but does not
22619 affect the mangled name of the namespace; in this case it is only used
22620 for @option{-Wabi-tag} warnings and automatic tagging of functions and
22621 variables. Tagging inline namespaces is generally preferable to
22622 tagging individual declarations, but the latter is sometimes
22623 necessary, such as when only certain members of a class need to be
22626 The argument can be a list of strings of arbitrary length. The
22627 strings are sorted on output, so the order of the list is
22630 A redeclaration of an entity must not add new ABI tags,
22631 since doing so would change the mangled name.
22633 The ABI tags apply to a name, so all instantiations and
22634 specializations of a template have the same tags. The attribute will
22635 be ignored if applied to an explicit specialization or instantiation.
22637 The @option{-Wabi-tag} flag enables a warning about a class which does
22638 not have all the ABI tags used by its subobjects and virtual functions; for users with code
22639 that needs to coexist with an earlier ABI, using this option can help
22640 to find all affected types that need to be tagged.
22642 When a type involving an ABI tag is used as the type of a variable or
22643 return type of a function where that tag is not already present in the
22644 signature of the function, the tag is automatically applied to the
22645 variable or function. @option{-Wabi-tag} also warns about this
22646 situation; this warning can be avoided by explicitly tagging the
22647 variable or function or moving it into a tagged inline namespace.
22649 @item init_priority (@var{priority})
22650 @cindex @code{init_priority} variable attribute
22652 In Standard C++, objects defined at namespace scope are guaranteed to be
22653 initialized in an order in strict accordance with that of their definitions
22654 @emph{in a given translation unit}. No guarantee is made for initializations
22655 across translation units. However, GNU C++ allows users to control the
22656 order of initialization of objects defined at namespace scope with the
22657 @code{init_priority} attribute by specifying a relative @var{priority},
22658 a constant integral expression currently bounded between 101 and 65535
22659 inclusive. Lower numbers indicate a higher priority.
22661 In the following example, @code{A} would normally be created before
22662 @code{B}, but the @code{init_priority} attribute reverses that order:
22665 Some_Class A __attribute__ ((init_priority (2000)));
22666 Some_Class B __attribute__ ((init_priority (543)));
22670 Note that the particular values of @var{priority} do not matter; only their
22674 @cindex @code{warn_unused} type attribute
22676 For C++ types with non-trivial constructors and/or destructors it is
22677 impossible for the compiler to determine whether a variable of this
22678 type is truly unused if it is not referenced. This type attribute
22679 informs the compiler that variables of this type should be warned
22680 about if they appear to be unused, just like variables of fundamental
22683 This attribute is appropriate for types which just represent a value,
22684 such as @code{std::string}; it is not appropriate for types which
22685 control a resource, such as @code{std::lock_guard}.
22687 This attribute is also accepted in C, but it is unnecessary because C
22688 does not have constructors or destructors.
22692 @node Function Multiversioning
22693 @section Function Multiversioning
22694 @cindex function versions
22696 With the GNU C++ front end, for x86 targets, you may specify multiple
22697 versions of a function, where each function is specialized for a
22698 specific target feature. At runtime, the appropriate version of the
22699 function is automatically executed depending on the characteristics of
22700 the execution platform. Here is an example.
22703 __attribute__ ((target ("default")))
22706 // The default version of foo.
22710 __attribute__ ((target ("sse4.2")))
22713 // foo version for SSE4.2
22717 __attribute__ ((target ("arch=atom")))
22720 // foo version for the Intel ATOM processor
22724 __attribute__ ((target ("arch=amdfam10")))
22727 // foo version for the AMD Family 0x10 processors.
22734 assert ((*p) () == foo ());
22739 In the above example, four versions of function foo are created. The
22740 first version of foo with the target attribute "default" is the default
22741 version. This version gets executed when no other target specific
22742 version qualifies for execution on a particular platform. A new version
22743 of foo is created by using the same function signature but with a
22744 different target string. Function foo is called or a pointer to it is
22745 taken just like a regular function. GCC takes care of doing the
22746 dispatching to call the right version at runtime. Refer to the
22747 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
22748 Function Multiversioning} for more details.
22751 @section Type Traits
22753 The C++ front end implements syntactic extensions that allow
22754 compile-time determination of
22755 various characteristics of a type (or of a
22759 @item __has_nothrow_assign (type)
22760 If @code{type} is const qualified or is a reference type then the trait is
22761 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
22762 is true, else if @code{type} is a cv class or union type with copy assignment
22763 operators that are known not to throw an exception then the trait is true,
22764 else it is false. Requires: @code{type} shall be a complete type,
22765 (possibly cv-qualified) @code{void}, or an array of unknown bound.
22767 @item __has_nothrow_copy (type)
22768 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
22769 @code{type} is a cv class or union type with copy constructors that
22770 are known not to throw an exception then the trait is true, else it is false.
22771 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
22772 @code{void}, or an array of unknown bound.
22774 @item __has_nothrow_constructor (type)
22775 If @code{__has_trivial_constructor (type)} is true then the trait is
22776 true, else if @code{type} is a cv class or union type (or array
22777 thereof) with a default constructor that is known not to throw an
22778 exception then the trait is true, else it is false. Requires:
22779 @code{type} shall be a complete type, (possibly cv-qualified)
22780 @code{void}, or an array of unknown bound.
22782 @item __has_trivial_assign (type)
22783 If @code{type} is const qualified or is a reference type then the trait is
22784 false. Otherwise if @code{__is_pod (type)} is true then the trait is
22785 true, else if @code{type} is a cv class or union type with a trivial
22786 copy assignment ([class.copy]) then the trait is true, else it is
22787 false. Requires: @code{type} shall be a complete type, (possibly
22788 cv-qualified) @code{void}, or an array of unknown bound.
22790 @item __has_trivial_copy (type)
22791 If @code{__is_pod (type)} is true or @code{type} is a reference type
22792 then the trait is true, else if @code{type} is a cv class or union type
22793 with a trivial copy constructor ([class.copy]) then the trait
22794 is true, else it is false. Requires: @code{type} shall be a complete
22795 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22797 @item __has_trivial_constructor (type)
22798 If @code{__is_pod (type)} is true then the trait is true, else if
22799 @code{type} is a cv class or union type (or array thereof) with a
22800 trivial default constructor ([class.ctor]) then the trait is true,
22801 else it is false. Requires: @code{type} shall be a complete
22802 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22804 @item __has_trivial_destructor (type)
22805 If @code{__is_pod (type)} is true or @code{type} is a reference type then
22806 the trait is true, else if @code{type} is a cv class or union type (or
22807 array thereof) with a trivial destructor ([class.dtor]) then the trait
22808 is true, else it is false. Requires: @code{type} shall be a complete
22809 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22811 @item __has_virtual_destructor (type)
22812 If @code{type} is a class type with a virtual destructor
22813 ([class.dtor]) then the trait is true, else it is false. Requires:
22814 @code{type} shall be a complete type, (possibly cv-qualified)
22815 @code{void}, or an array of unknown bound.
22817 @item __is_abstract (type)
22818 If @code{type} is an abstract class ([class.abstract]) then the trait
22819 is true, else it is false. Requires: @code{type} shall be a complete
22820 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22822 @item __is_base_of (base_type, derived_type)
22823 If @code{base_type} is a base class of @code{derived_type}
22824 ([class.derived]) then the trait is true, otherwise it is false.
22825 Top-level cv qualifications of @code{base_type} and
22826 @code{derived_type} are ignored. For the purposes of this trait, a
22827 class type is considered is own base. Requires: if @code{__is_class
22828 (base_type)} and @code{__is_class (derived_type)} are true and
22829 @code{base_type} and @code{derived_type} are not the same type
22830 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
22831 type. A diagnostic is produced if this requirement is not met.
22833 @item __is_class (type)
22834 If @code{type} is a cv class type, and not a union type
22835 ([basic.compound]) the trait is true, else it is false.
22837 @item __is_empty (type)
22838 If @code{__is_class (type)} is false then the trait is false.
22839 Otherwise @code{type} is considered empty if and only if: @code{type}
22840 has no non-static data members, or all non-static data members, if
22841 any, are bit-fields of length 0, and @code{type} has no virtual
22842 members, and @code{type} has no virtual base classes, and @code{type}
22843 has no base classes @code{base_type} for which
22844 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
22845 be a complete type, (possibly cv-qualified) @code{void}, or an array
22848 @item __is_enum (type)
22849 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
22850 true, else it is false.
22852 @item __is_literal_type (type)
22853 If @code{type} is a literal type ([basic.types]) the trait is
22854 true, else it is false. Requires: @code{type} shall be a complete type,
22855 (possibly cv-qualified) @code{void}, or an array of unknown bound.
22857 @item __is_pod (type)
22858 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
22859 else it is false. Requires: @code{type} shall be a complete type,
22860 (possibly cv-qualified) @code{void}, or an array of unknown bound.
22862 @item __is_polymorphic (type)
22863 If @code{type} is a polymorphic class ([class.virtual]) then the trait
22864 is true, else it is false. Requires: @code{type} shall be a complete
22865 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22867 @item __is_standard_layout (type)
22868 If @code{type} is a standard-layout type ([basic.types]) the trait is
22869 true, else it is false. Requires: @code{type} shall be a complete
22870 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22872 @item __is_trivial (type)
22873 If @code{type} is a trivial type ([basic.types]) the trait is
22874 true, else it is false. Requires: @code{type} shall be a complete
22875 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22877 @item __is_union (type)
22878 If @code{type} is a cv union type ([basic.compound]) the trait is
22879 true, else it is false.
22881 @item __underlying_type (type)
22882 The underlying type of @code{type}. Requires: @code{type} shall be
22883 an enumeration type ([dcl.enum]).
22885 @item __integer_pack (length)
22886 When used as the pattern of a pack expansion within a template
22887 definition, expands to a template argument pack containing integers
22888 from @code{0} to @code{length-1}. This is provided for efficient
22889 implementation of @code{std::make_integer_sequence}.
22895 @section C++ Concepts
22897 C++ concepts provide much-improved support for generic programming. In
22898 particular, they allow the specification of constraints on template arguments.
22899 The constraints are used to extend the usual overloading and partial
22900 specialization capabilities of the language, allowing generic data structures
22901 and algorithms to be ``refined'' based on their properties rather than their
22904 The following keywords are reserved for concepts.
22908 States an expression as an assumption, and if possible, verifies that the
22909 assumption is valid. For example, @code{assume(n > 0)}.
22912 Introduces an axiom definition. Axioms introduce requirements on values.
22915 Introduces a universally quantified object in an axiom. For example,
22916 @code{forall (int n) n + 0 == n}).
22919 Introduces a concept definition. Concepts are sets of syntactic and semantic
22920 requirements on types and their values.
22923 Introduces constraints on template arguments or requirements for a member
22924 function of a class template.
22928 The front end also exposes a number of internal mechanism that can be used
22929 to simplify the writing of type traits. Note that some of these traits are
22930 likely to be removed in the future.
22933 @item __is_same (type1, type2)
22934 A binary type trait: true whenever the type arguments are the same.
22939 @node Deprecated Features
22940 @section Deprecated Features
22942 In the past, the GNU C++ compiler was extended to experiment with new
22943 features, at a time when the C++ language was still evolving. Now that
22944 the C++ standard is complete, some of those features are superseded by
22945 superior alternatives. Using the old features might cause a warning in
22946 some cases that the feature will be dropped in the future. In other
22947 cases, the feature might be gone already.
22949 While the list below is not exhaustive, it documents some of the options
22950 that are now deprecated:
22953 @item -fexternal-templates
22954 @itemx -falt-external-templates
22955 These are two of the many ways for G++ to implement template
22956 instantiation. @xref{Template Instantiation}. The C++ standard clearly
22957 defines how template definitions have to be organized across
22958 implementation units. G++ has an implicit instantiation mechanism that
22959 should work just fine for standard-conforming code.
22961 @item -fstrict-prototype
22962 @itemx -fno-strict-prototype
22963 Previously it was possible to use an empty prototype parameter list to
22964 indicate an unspecified number of parameters (like C), rather than no
22965 parameters, as C++ demands. This feature has been removed, except where
22966 it is required for backwards compatibility. @xref{Backwards Compatibility}.
22969 G++ allows a virtual function returning @samp{void *} to be overridden
22970 by one returning a different pointer type. This extension to the
22971 covariant return type rules is now deprecated and will be removed from a
22974 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
22975 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
22976 and are now removed from G++. Code using these operators should be
22977 modified to use @code{std::min} and @code{std::max} instead.
22979 The named return value extension has been deprecated, and is now
22982 The use of initializer lists with new expressions has been deprecated,
22983 and is now removed from G++.
22985 Floating and complex non-type template parameters have been deprecated,
22986 and are now removed from G++.
22988 The implicit typename extension has been deprecated and is now
22991 The use of default arguments in function pointers, function typedefs
22992 and other places where they are not permitted by the standard is
22993 deprecated and will be removed from a future version of G++.
22995 G++ allows floating-point literals to appear in integral constant expressions,
22996 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
22997 This extension is deprecated and will be removed from a future version.
22999 G++ allows static data members of const floating-point type to be declared
23000 with an initializer in a class definition. The standard only allows
23001 initializers for static members of const integral types and const
23002 enumeration types so this extension has been deprecated and will be removed
23003 from a future version.
23005 @node Backwards Compatibility
23006 @section Backwards Compatibility
23007 @cindex Backwards Compatibility
23008 @cindex ARM [Annotated C++ Reference Manual]
23010 Now that there is a definitive ISO standard C++, G++ has a specification
23011 to adhere to. The C++ language evolved over time, and features that
23012 used to be acceptable in previous drafts of the standard, such as the ARM
23013 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
23014 compilation of C++ written to such drafts, G++ contains some backwards
23015 compatibilities. @emph{All such backwards compatibility features are
23016 liable to disappear in future versions of G++.} They should be considered
23017 deprecated. @xref{Deprecated Features}.
23021 If a variable is declared at for scope, it used to remain in scope until
23022 the end of the scope that contained the for statement (rather than just
23023 within the for scope). G++ retains this, but issues a warning, if such a
23024 variable is accessed outside the for scope.
23026 @item Implicit C language
23027 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
23028 scope to set the language. On such systems, all header files are
23029 implicitly scoped inside a C language scope. Also, an empty prototype
23030 @code{()} is treated as an unspecified number of arguments, rather
23031 than no arguments, as C++ demands.
23034 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
23035 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr